1 2018-12-21 Alexandre Oliva <oliva@adacore.com>
3 * config.gcc (tmake_file): Add name to tmake_file for
4 each @name in --with-multilib-list on arm-*-* targets.
5 * doc/install.texi (with-multilib-list): Document it.
7 2018-12-21 Jakub Jelinek <jakub@redhat.com>
9 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Use
10 DR_TARGET_ALIGNMENT on dr_info rather than dr. Spelling fixes.
12 2018-12-20 Joern Rennecke <joern.rennecke@riscy-ip.com>
14 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Don't do
15 versioning for data accesses with misaligned step.
17 2018-12-20 H.J. Lu <hongjiu.lu@intel.com>
20 * doc/invoke.texi: Document -Wno-address-of-packed-member.
22 2018-12-20 Vladimir Makarov <vmakarov@redhat.com>
25 * ira-color.c (fast_allocation): Choose the best cost hard reg.
27 2018-12-20 Richard Sandiford <richard.sandiford@arm.com>
29 * config/aarch64/iterators.md (SVE_INT_UNARY, fp_int_op): Add abs.
32 2018-12-20 Richard Sandiford <richard.sandiford@arm.com>
34 * config/aarch64/aarch64-sve.md (*cond_<optab><mode>_4): Use
35 sve_fmla_op rather than sve_fmad_op for the movprfx alternative.
37 2018-12-20 Martin Jambor <mjambor@suse.cz>
40 * tree-ssa-alias.c (ao_ref_init_from_ptr_and_size): Assert that
43 2018-12-20 Richard Biener <rguenther@suse.de>
45 PR tree-optimization/84362
46 * tree-ssa-loop-im.c: Include alias.h, builtins.h and tree-dfa.h.
47 (struct im_mem_ref): add ref_canonical flag.
48 (struct mem_ref_hasher): Use ao_ref as compare_type.
49 (mem_ref_hasher::equal): Adjust and add variant comparing ao_ref
51 (mem_ref_alloc): Take ao_ref parameter, initialize ref_canonical
53 (gather_mem_refs_stmt): Set up ao_ref early and do the lookup
54 using it. If we have non-equal refs canonicalize the one
55 in the hashtable used for insertion.
56 (tree_ssa_lim_initialize): Adjust.
58 2018-12-20 Jakub Jelinek <jakub@redhat.com>
61 * config/i386/i386.c (ix86_expand_sse_movcc): For maskcmp, try to
62 emit vpmovm2? instruction perhaps after knot?. Reorganize code
63 so that it doesn't have to test !maskcmp in almost every conditional.
65 2018-12-20 Alan Modra <amodra@gmail.com>
67 * config/rs6000/sysv4.h (GNU_USER_DYNAMIC_LINKER): Define.
69 2018-12-19 Andreas Tobler <andreast@gcc.gnu.org>
71 * config.gcc: Enable TARGET_DEFAULT_ASYNC_UNWIND_TABLES on
74 2018-12-19 Uros Bizjak <ubizjak@gmail.com>
76 * config/i386/i386.md (SWI1248_AVX512BWDQ_64): Rename from
77 SWI1248_AVX512BWDQ2_64. Unconditionally enable HImode.
78 (*cmp<mode>_ccz_1): Emit kortest instead of ktest insn.
79 Use SWI1248_AVX512BWDQ_64 mode iterator and enable only for
82 2018-12-19 Thomas Preud'homme <thomas.preudhomme@linaro.org>
84 * config/arm/elf.h: Update comment about condition that need to
85 match with libgcc/config/arm/lib1funcs.S to also include
86 libgcc/config/arm/t-arm.
87 * doc/sourcebuild.texi (output-exists, output-exists-not): Rename
88 subsubsection these directives are in to "Check for output files".
89 Move scan-symbol to that section and add to it new scan-symbol-not
92 2018-12-19 Tom de Vries <tdevries@suse.de>
94 * config/nvptx/nvptx.c (PTX_CTA_SIZE): Define.
96 2018-12-19 Tom de Vries <tdevries@suse.de>
98 * omp-offload.c (oacc_fn_attrib_level): Remove static.
99 * omp-offload.h (oacc_fn_attrib_level): Declare.
101 2018-12-19 Tom de Vries <tdevries@suse.de>
103 * omp-offload.c (oacc_get_default_dim): New function.
104 * omp-offload.h (oacc_get_default_dim): Declare.
106 2018-12-19 David Malcolm <dmalcolm@redhat.com>
109 * gcc-rich-location.c
110 (maybe_range_label_for_tree_type_mismatch::get_text): Move here from
112 (binary_op_rich_location::binary_op_rich_location): New ctor.
113 (binary_op_rich_location::use_operator_loc_p): New function.
114 * gcc-rich-location.h
115 (class maybe_range_label_for_tree_type_mismatch)): Move here from
117 (struct op_location_t): New forward decl.
118 (class binary_op_rich_location): New class.
119 * tree.h (struct op_location_t): New struct.
121 2018-12-19 David Malcolm <dmalcolm@redhat.com>
125 * convert.c: Include "selftest.h".
126 (preserve_any_location_wrapper): New function.
127 (convert_to_pointer_maybe_fold): Update to handle location
129 (convert_to_real_maybe_fold): Likewise.
130 (convert_to_integer_1): Strip expr when using TREE_OVERFLOW.
131 Handle location wrappers when checking for INTEGER_CST.
132 (convert_to_integer_maybe_fold): Update to handle location
134 (convert_to_complex_maybe_fold): Likewise.
135 (selftest::test_convert_to_integer_maybe_fold): New functions.
136 (selftest::convert_c_tests): New function.
137 * convert.h (preserve_any_location_wrapper): New decl.
138 * fold-const.c (size_binop_loc): Strip location wrappers when
140 (operand_equal_p): Strip any location wrappers.
141 (integer_valued_real_p): Strip any location wrapper.
142 * selftest-run-tests.c (selftest::run_tests): Call
143 selftest::convert_c_tests.
144 * selftest.h (selftest::convert_c_tests): New decl.
145 * tree.c (build_complex): Assert that REAL and IMAG are constants.
146 (integer_zerop): Look through location wrappers.
147 (integer_onep): Likewise.
148 (integer_each_onep): Likewise.
149 (integer_all_onesp): Likewise.
150 (integer_minus_onep): Likewise.
151 (integer_pow2p): Likewise.
152 (integer_nonzerop): Likewise.
153 (integer_truep): Likewise.
154 (fixed_zerop): Likewise.
155 (real_zerop): Likewise.
156 (real_onep): Likewise.
157 (real_minus_onep): Likewise.
158 (tree_int_cst_equal): Likewise.
159 (simple_cst_equal): Treat location wrappers with non-equal source
160 locations as being unequal.
161 (uniform_integer_cst_p): Look through location wrappers.
162 (maybe_wrap_with_location): Don't create wrappers if any
163 auto_suppress_location_wrappers are active.
164 (suppress_location_wrappers): New variable.
165 (selftest::test_predicates): New test.
166 (selftest::tree_c_tests): Call it.
167 * tree.h (CONSTANT_CLASS_OR_WRAPPER_P): New macro.
168 (suppress_location_wrappers): New decl.
169 (class auto_suppress_location_wrappers): New class.
171 2018-12-19 Paul A. Clarke <pc@us.ibm.com>
173 * config/rs6000/tmmintrin.h (_mm_hadds_epi16): Vector lanes swapped.
174 (_mm_hsub_epi32): Likewise.
175 (_mm_shuffle_epi8): Fix reversed interpretation of parameters.
176 (_mm_shuffle_pi8): Likewise.
177 (_mm_addubs_pi16): Likewise.
179 2018-12-19 Thomas Preud'homme <thomas.preudhomme@linaro.org>
181 * config/arm/arm.h (TARGET_HARD_FLOAT): Restrict to TARGET_32BIT
183 * config/arm/arm.c (output_return_instruction): Only check
184 TARGET_HARD_FLOAT to decide whether FP instructions are available.
186 2018-12-19 Tom de Vries <tom@codesourcery.com>
188 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
189 dump files): Add offload-rtl.
191 2018-12-19 Segher Boessenkool <segher@kernel.crashing.org>
194 * config/rs6000/vsx.md (*vsx_extract_<P:mode>_<VSX_D:mode>_load):
195 Require TARGET_POWERPC64.
197 2018-12-19 Richard Biener <rguenther@suse.de>
199 PR tree-optimization/88533
201 2018-04-30 Richard Biener <rguenther@suse.de>
203 PR tree-optimization/28364
204 PR tree-optimization/85275
205 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
206 copying first exit test.
208 * tree-ssa-loop-ch.c: Include tree-phinodes.h and
210 (should_duplicate_loop_header_p): Track whether stmt compute
211 loop invariants or values based on IVs. Apart from the
212 original loop header only duplicate blocks with exit tests
213 that are based on IVs or invariants.
215 2018-12-19 Tom de Vries <tdevries@suse.de>
217 * config/nvptx/nvptx.c (nvptx_gen_shared_bcast, shared_prop_gen)
218 (nvptx_goacc_expand_accel_var): Use MAX and ROUND_UP.
219 (nvptx_assemble_value, nvptx_output_skip): Use MIN.
220 (nvptx_shared_propagate, nvptx_single, nvptx_expand_shared_addr): Use
223 2018-12-19 Tom de Vries <tdevries@suse.de>
225 * config/nvptx/nvptx.c (nvptx_gen_wcast): Rename as
226 nvptx_gen_warp_bcast.
227 (nvptx_gen_wcast): Rename to nvptx_gen_shared_bcast, add bool
228 vector argument, and update call to nvptx_gen_shared_bcast.
229 (propagator_fn): Add bool argument.
230 (nvptx_propagate): New bool argument, pass bool argument to fn.
231 (vprop_gen): Rename to warp_prop_gen, update call to
232 nvptx_gen_warp_bcast.
233 (nvptx_vpropagate): Rename to nvptx_warp_propagate, update call to
235 (wprop_gen): Rename to shared_prop_gen, update call to
236 nvptx_gen_shared_bcast.
237 (nvptx_wpropagate): Rename to nvptx_shared_propagate, update call
239 (nvptx_wsync): Rename to nvptx_cta_sync.
240 (nvptx_single): Update calls to nvptx_gen_warp_bcast,
241 nvptx_gen_shared_bcast and nvptx_cta_sync.
242 (nvptx_process_pars): Likewise.
243 (write_worker_buffer): Rename as write_shared_buffer.
244 (nvptx_file_end): Update calls to write_shared_buffer.
245 (nvptx_expand_worker_addr): Rename as nvptx_expand_shared_addr.
246 (nvptx_expand_builtin): Update call to nvptx_expand_shared_addr.
247 (nvptx_get_worker_red_addr): Rename as nvptx_get_shared_red_addr.
248 (nvptx_goacc_reduction_setup): Update call to
249 nvptx_get_shared_red_addr.
250 (nvptx_goacc_reduction_fini): Likewise.
251 (nvptx_goacc_reduction_teardown): Likewise.
253 2018-12-19 Tom de Vries <tdevries@suse.de>
255 * config/nvptx/nvptx.c (worker_bcast_size): Rename as
257 (worker_bcast_align): Rename as oacc_bcast_align.
258 (worker_bcast_sym): Rename as oacc_bcast_sym.
259 (nvptx_option_override): Update usage of oacc_bcast_*.
260 (struct wcast_data_t): Rename as broadcast_data_t.
261 (nvptx_gen_wcast): Update type of data argument and usage of
263 (wprop_gen): Update type of data_ and usage of oacc_bcast_align.
264 (nvptx_wpropagate): Update type of data and usage of
265 oacc_bcast_{sym,size}.
266 (nvptx_single): Update type of data and usage of oacc_bcast_size.
267 (nvptx_file_end): Update usage of oacc_bcast_{sym,align,size}.
269 2018-12-19 Tom de Vries <tdevries@suse.de>
271 * config/nvptx/nvptx.md (nvptx_barsync): Add and handle operand.
272 * config/nvptx/nvptx.c (nvptx_wsync): Update call to gen_nvptx_barsync.
274 2018-12-19 Tom de Vries <tdevries@suse.de>
276 * config/nvptx/nvptx.c (nvptx_single): Always pass false to
278 (nvptx_process_pars): Likewise.
280 2018-12-19 Tom de Vries <tdevries@suse.de>
282 * config/nvptx/nvptx.c (nvptx_previous_fndecl): Declare.
283 (nvptx_set_current_function): New function.
284 (TARGET_SET_CURRENT_FUNCTION): Define.
286 2018-12-19 Shaokun Zhang <zhangshaokun@hisilicon.com>
288 * config/aarch64/aarch64-cores.def (tsv110): Fix architecture. This
289 part is really Armv8.2 with some permitted Armv8.4 extensions.
291 2018-12-19 Jakub Jelinek <jakub@redhat.com>
294 * config/i386/vpclmulqdqintrin.h (_mm256_clmulepi64_epi128): Enable
295 for -mavx -mvpclmulqdq rather than just for -mavx512vl -mvpclmulqdq.
297 2018-12-19 Alexandre Oliva <aoliva@redhat.com>
302 (vr_values::vrp_evaluate_conditional_warnv_with_ops): Extend
303 simplification of overflow tests to cover cases in which we
304 can determine the result of the comparison.
306 2018-12-19 Bin Cheng <bin.cheng@linux.alibaba.com>
308 * auto-profile.c (afdo_indirect_call): Skip generating histogram
309 value if we can't find cgraph_node for then indirected callee. Save
310 profile_id of the cgraph_node in histogram value's first counter.
311 * value-prof.c (gimple_value_profile_transformations): Don't skip
312 for flag_auto_profile.
314 2018-12-18 Vladimir Makarov <vmakarov@redhat.com>
316 PR rtl-optimization/87759
317 * lra-assigns.c (lra_split_hard_reg_for): Recalculate
320 2018-12-18 Jakub Jelinek <jakub@redhat.com>
323 * config/i386/i386-builtin-types.def
324 (VOID_FTYPE_PDOUBLE_QI_V8SI_V4DF_INT,
325 VOID_FTYPE_PFLOAT_QI_V4DI_V8SF_INT,
326 VOID_FTYPE_PLONGLONG_QI_V8SI_V4DI_INT,
327 VOID_FTYPE_PINT_QI_V4DI_V8SI_INT,
328 VOID_FTYPE_PDOUBLE_QI_V4SI_V2DF_INT,
329 VOID_FTYPE_PFLOAT_QI_V2DI_V4SF_INT,
330 VOID_FTYPE_PLONGLONG_QI_V4SI_V2DI_INT,
331 VOID_FTYPE_PINT_QI_V2DI_V4SI_INT): New builtin types.
332 * config/i386/i386.c (enum ix86_builtins): Add
333 IX86_BUILTIN_SCATTERALTSIV4DF, IX86_BUILTIN_SCATTERALTDIV8SF,
334 IX86_BUILTIN_SCATTERALTSIV4DI, IX86_BUILTIN_SCATTERALTDIV8SI,
335 IX86_BUILTIN_SCATTERALTSIV2DF, IX86_BUILTIN_SCATTERALTDIV4SF,
336 IX86_BUILTIN_SCATTERALTSIV2DI and IX86_BUILTIN_SCATTERALTDIV4SI.
337 (ix86_init_mmx_sse_builtins): Fix up names of IX86_BUILTIN_GATHERALT*,
338 IX86_BUILTIN_GATHER3ALT* and IX86_BUILTIN_SCATTERALT* builtins to
339 match the IX86_BUILTIN codes. Build IX86_BUILTIN_SCATTERALTSIV4DF,
340 IX86_BUILTIN_SCATTERALTDIV8SF, IX86_BUILTIN_SCATTERALTSIV4DI,
341 IX86_BUILTIN_SCATTERALTDIV8SI, IX86_BUILTIN_SCATTERALTSIV2DF,
342 IX86_BUILTIN_SCATTERALTDIV4SF, IX86_BUILTIN_SCATTERALTSIV2DI and
343 IX86_BUILTIN_SCATTERALTDIV4SI decls.
344 (ix86_vectorize_builtin_scatter): Expand those new builtins.
346 2018-12-18 Bill Schmidt <wschmidt@linux.ibm.com>
348 * doc/extend.texi (PowerPC Altivec/VSX Built-in Functions):
349 Describe when a typedef name can be used as the type specifier for
350 a vector type, and when it cannot.
352 2018-12-18 Jozef Lawrynowicz <jozef.l@mittosystems.com>
354 * config/msp430/msp430.h: Define TARGET_VTABLE_ENTRY_ALIGN.
356 2018-12-18 Jakub Jelinek <jakub@redhat.com>
360 * optabs.def (vec_pack_sbool_trunc_optab, vec_unpacks_sbool_hi_optab,
361 vec_unpacks_sbool_lo_optab): New optabs.
362 * optabs.c (expand_widen_pattern_expr): Use vec_unpacks_sbool_*_optab
363 and pass additional argument if both input and target have the same
364 scalar mode of VECTOR_BOOLEAN_TYPE_P vectors.
365 * expr.c (expand_expr_real_2) <case VEC_PACK_TRUNC_EXPR>: Handle
366 VECTOR_BOOLEAN_TYPE_P pack where result has the same scalar mode
367 as the operands using vec_pack_sbool_trunc_optab.
368 * tree-vect-stmts.c (supportable_widening_operation): Use
369 vec_unpacks_sbool_{lo,hi}_optab for VECTOR_BOOLEAN_TYPE_P conversions
370 where both wider_vectype and vectype have the same scalar mode.
371 (supportable_narrowing_operation): Similarly use
372 vec_pack_sbool_trunc_optab if narrow_vectype and vectype have the same
374 * config/i386/i386.c (ix86_get_builtin)
375 <case IX86_BUILTIN_GATHER3ALTDIV8SF>: Check for VECTOR_MODE_P
376 rather than non-VOIDmode.
377 * config/i386/sse.md (vec_pack_trunc_qi, vec_pack_trunc_<mode>):
378 Remove useless ()s around "register_operand", formatting fixes.
379 (vec_pack_sbool_trunc_qi, vec_unpacks_sbool_lo_qi,
380 vec_unpacks_sbool_hi_qi): New expanders.
381 * doc/md.texi (vec_pack_sbool_trunc_M, vec_unpacks_sbool_hi_M,
382 vec_unpacks_sbool_lo_M): Document.
384 2018-12-18 Jozef Lawrynowicz <jozef.l@mittosystems.com>
386 * combine.c (update_rsp_from_reg_equal): Only look for the nonzero bits
387 of src in nonzero_bits_mode if the mode of src is MODE_INT and
389 (reg_nonzero_bits_for_combine): Add clarification to comment.
391 2018-12-18 Wei Xiao <wei3.xiao@intel.com>
393 * config/i386/driver-i386.c (host_detect_local_cpu): Detect cascadelake.
394 * config/i386/i386.c (fold_builtin_cpu): Handle cascadelake.
395 * doc/extend.texi: Add cascadelake.
396 2018-12-17 Peter Bergner <bergner@linux.ibm.com>
399 * config/rs6000/vsx.md (nW): New mode iterator.
400 (vsx_mov<mode>_64bit): Use it. Remove redundant GPR 0/-1 alternative.
401 Update length attribute for (<??r>, <nW>) alternative.
402 (vsx_mov<mode>_32bit): Likewise.
404 2018-12-17 Tom de Vries <tdevries@suse.de>
406 * config/nvptx/nvptx.c (PTX_VECTOR_LENGTH, PTX_WORKER_LENGTH,
407 PTX_DEFAULT_RUNTIME_DIM): Move to the top of the file.
409 2018-12-17 Tom de Vries <tdevries@suse.de>
411 * config/nvptx/nvptx.c (PTX_WARP_SIZE): Define.
412 (nvptx_simt_vf): Return PTX_WARP_SIZE instead of PTX_VECTOR_LENGTH.
414 2018-12-17 Tom de Vries <tdevries@suse.de>
416 * config/nvptx/nvptx.c (nvptx_single): Fix whitespace.
417 (nvptx_neuter_pars): Likewise.
419 2018-12-17 Tom de Vries <tdevries@suse.de>
421 * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Work around Fortran
422 bug PR72741 by overriding dims parameter for routines.
424 2018-12-17 Tom de Vries <tdevries@suse.de>
426 * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Rewrite using
429 2018-12-17 Steve Ellcey <sellcey@cavium.com>
431 * config/aarch64/aarch64-protos.h (aarch64_use_simple_return_insn_p):
433 (aarch64_epilogue_uses): Ditto.
434 * config/aarch64/aarch64.c (aarch64_attribute_table): New array.
435 (aarch64_simd_decl_p): New function.
436 (aarch64_reg_save_mode): New function.
437 (aarch64_function_ok_for_sibcall): Check for simd calls.
438 (aarch64_layout_frame): Check for simd function.
439 (aarch64_gen_storewb_pair): Handle E_TFmode.
440 (aarch64_push_regs): Use aarch64_reg_save_mode to get mode.
441 (aarch64_gen_loadwb_pair): Handle E_TFmode.
442 (aarch64_pop_regs): Use aarch64_reg_save_mode to get mode.
443 (aarch64_gen_store_pair): Handle E_TFmode.
444 (aarch64_gen_load_pair): Ditto.
445 (aarch64_save_callee_saves): Handle different mode sizes.
446 (aarch64_restore_callee_saves): Ditto.
447 (aarch64_components_for_bb): Check for simd function.
448 (aarch64_epilogue_uses): New function.
449 (aarch64_process_components): Check for simd function.
450 (aarch64_expand_prologue): Ditto.
451 (aarch64_expand_epilogue): Ditto.
452 (aarch64_expand_call): Ditto.
453 (aarch64_use_simple_return_insn_p): New function.
454 (TARGET_ATTRIBUTE_TABLE): New define.
455 * config/aarch64/aarch64.h (EPILOGUE_USES): Redefine.
456 (FP_SIMD_SAVED_REGNUM_P): New macro.
457 * config/aarch64/aarch64.md (simple_return): New define_expand.
458 (load_pair_dw_tftf): New instruction.
459 (store_pair_dw_tftf): Ditto.
460 (loadwb_pair<TX:mode>_<P:mode>): Ditto.
461 (storewb_pair<TX:mode>_<P:mode>): Ditto.
463 2018-12-17 Uros Bizjak <ubizjak@gmail.com>
466 * internal-fn.def (ACOSH): New.
469 * optabs.def (acosh_optab): New.
470 (asinh_optab): Ditto.
471 (atanh_optab): Ditto.
472 * config/i386/i386-protos.h (ix86_emit_i387_asinh): New prototype.
473 (ix86_emit_i387_acosh): Ditto.
474 (ix86_emit_i387_atanh): Ditto.
475 * config/i386/i386.c (ix86_emit_i387_asinh): New function.
476 (ix86_emit_i387_acosh): Ditto.
477 (ix86_emit_i387_atanh): Ditto.
478 * config/i386/i386.md (asinhxf2): New expander.
479 (asinh<mode>2): Ditto.
481 (acosh<mode>2): Ditto.
483 (atanh<mode>2): Ditto.
485 2018-12-17 David Edelsohn <dje.gcc@gmail.com>
487 * config.gcc (powerpc-ibm-aix6.*): Delete extra_headers.
488 (powerpc-ibm-aix7.1.*): Same.
489 (powerpc-ibm-aix[789].*): Same.
491 2018-12-17 H.J. Lu <hongjiu.lu@intel.com>
494 * dwarf2out.c (find_AT_string_in_table): Add insert argument
495 defaulting to INSERT and replace INSERT.
496 (find_AT_string): Likewise.
497 (output_macinfo_op): Pass NO_INSERT to find_AT_string.
499 2018-12-15 Jan Hubicka <hubicka@ucw.cz>
501 * coverage.c (struct conts_entry): Add n_counts.
502 (remap_counts_file): Record number of ocunts.
503 (get_coverage_counts): Verify that counts match.
504 * coverage.h (get_coverage_counts): Update prototype.
505 * profile.c (get_exec_counts. compute_value_histograms): Add
508 2018-12-17 Senthil Kumar Selvaraj <senthilkumar.selvaraj@microchip.com>
510 PR rtl-optimization/88253
511 * combine.c (combine_simplify_rtx): Test for side-effects before
512 substituting by zero.
514 2018-12-17 Richard Sandiford <richard.sandiford@arm.com>
516 * doc/invoke.texi (-fversion-loops-for-strides): Document
517 (loop-versioning-group-size, loop-versioning-max-inner-insns)
518 (loop-versioning-max-outer-insns): Document new --params.
519 * Makefile.in (OBJS): Add gimple-loop-versioning.o.
520 * common.opt (fversion-loops-for-strides): New option.
521 * opts.c (default_options_table): Enable fversion-loops-for-strides
523 * params.def (PARAM_LOOP_VERSIONING_GROUP_SIZE)
524 (PARAM_LOOP_VERSIONING_MAX_INNER_INSNS)
525 (PARAM_LOOP_VERSIONING_MAX_OUTER_INSNS): New parameters.
526 * passes.def: Add pass_loop_versioning.
527 * timevar.def (TV_LOOP_VERSIONING): New time variable.
528 * tree-ssa-propagate.h
529 (substitute_and_fold_engine::substitute_and_fold): Add an optional
531 * tree-ssa-propagate.c
532 (substitute_and_fold_engine::substitute_and_fold): Likewise.
533 When passed, only walk blocks dominated by that block.
534 * tree-vrp.h (range_includes_p): Declare.
535 (range_includes_zero_p): Turn into an inline wrapper around
537 * tree-vrp.c (range_includes_p): New function, generalizing...
538 (range_includes_zero_p): ...this.
539 * tree-pass.h (make_pass_loop_versioning): Declare.
540 * gimple-loop-versioning.cc: New file.
542 2018-12-15 Jan Hubicka <hubicka@ucw.cz>
544 * ipa-fnsummary.c (remap_edge_change_prob): Do not ICE when changes
547 2018-12-15 H.J. Lu <hongjiu.lu@intel.com>
550 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Revert
553 2018-12-15 Jan Hubicka <hubicka@ucw.cz>
555 * ipa-fnsummary.c (analyze_function_body): Do not loeak conds and
557 (ipa_fn_summary_generate): Add prevails parameter; do not allocate
558 data when symbol is not prevailing.
559 (inline_read_section): Likewise.
561 2018-12-15 Jan Hubicka <hubicka@ucw.cz>
563 * cgraph.h (cgraph_node): Add predicate prevailing_p.
564 (cgraph_edge): Add predicate possible_call_in_translation_unit_p.
565 * ipa-prop.c (ipa_write_jump_function): Optimize streaming of ADDR_EXPR.
566 (ipa_read_jump_function): Add prevails parameter; optimize streaming.
567 (ipa_read_edge_info): Break out from ...
568 (ipa_read_node_info): ... here; optimize streaming.
569 * cgraph.c (cgraph_edge::possibly_call_in_translation_unit_p): New
572 2018-12-15 Jan Hubicka <hubicka@ucw.cz>
574 * ipa-utils.c (ipa_merge_profiles): Do no merging when source function
577 2018-12-15 Jakub Jelinek <jakub@redhat.com>
579 PR tree-optimization/88464
581 * tree-vect-stmts.c (vect_build_gather_load_calls): For NARROWING
582 and mask with integral masktype, don't try to permute mask vectors,
583 instead emit VEC_UNPACK_{LO,HI}_EXPR. Fix up NOP_EXPR operand.
584 (vectorizable_store): Handle masked scatters with decl and integral
586 (permute_vec_elements): Allow scalar_dest to be NULL.
587 * config/i386/i386.c (ix86_get_builtin)
588 <case IX86_BUILTIN_GATHER3ALTDIV16SF>: Use lowpart_subreg for masks.
589 <case IX86_BUILTIN_GATHER3ALTDIV8SF>: Don't assume mask and src have
592 2018-12-15 Jan Hubicka <hubicka@ucw.cz>
594 * ipa.c (cgraph_build_static_cdtor_1): Add OPTIMIZATION and TARGET
596 (cgraph_build_static_cdtor): Update.
597 (build_cdtor): Use OPTIMIZATION and TARGET of the first real cdtor
600 2018-12-15 Alan Modra <amodra@gmail.com>
602 * config/rs6000/rs6000.c (generate_set_vrsave, rs6000_emit_savres_rtx),
603 (rs6000_emit_prologue, rs6000_call_aix, rs6000_call_sysv),
604 (rs6000_call_darwin_1): Use gen_hard_reg_clobber.
606 2018-12-15 Jakub Jelinek <jakub@redhat.com>
609 * config/i386/sse.md (UNSPEC_SFIXUPIMM): New unspec enumerator.
610 (avx512f_sfixupimm<mode><mask_name><round_saeonly_name>): Use it
611 instead of UNSPEC_FIXUPIMM.
613 PR rtl-optimization/88478
614 * cselib.c (cselib_record_sets): Move sets[i].src_elt tests
615 after REG_P (dest) test.
617 2018-12-14 Martin Sebor <msebor@redhat.com>
619 PR tree-optimization/88372
620 * calls.c (maybe_warn_alloc_args_overflow): Handle function pointers.
621 * tree-object-size.c (alloc_object_size): Same. Simplify.
622 * doc/extend.texi (Object Size Checking): Update.
623 (Other Builtins): Add __builtin_object_size.
624 (Common Type Attributes): Add alloc_size.
625 (Common Variable Attributes): Ditto.
627 2018-12-14 Martin Sebor <msebor@redhat.com>
629 PR rtl-optimization/87096
630 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Avoid
631 folding calls whose bound may exceed INT_MAX. Diagnose bound ranges
632 that exceed the limit.
634 2018-12-14 Martin Sebor <msebor@redhat.com>
637 * doc/extend.texi (attribute const, pure): Clarify.
639 2018-12-14 H.J. Lu <hongjiu.lu@intel.com>
641 * config/i386/i386.c (rest_of_insert_endbranch): Insert ENDBR
642 at the function entry only when -mmanual-endbr isn't used or
643 there is cf_check function attribute.
644 (ix86_attribute_table): Add cf_check.
645 * config/i386/i386.opt: Add -mmanual-endbr.
646 * doc/extend.texi: Document cf_check attribute.
647 * doc/invoke.texi: Document -mmanual-endbr.
649 2018-12-14 Thomas Schwinge <thomas@codesourcery.com>
650 Cesar Philippidis <cesar@codesourcery.com>
652 * omp-offload.c (inform_oacc_loop): New function.
653 (execute_oacc_device_lower): Use it to display loop parallelism.
655 2018-12-14 Jakub Jelinek <jakub@redhat.com>
659 * expr.h (categorize_ctor_elements): Add p_unique_nz_elts argument.
660 * expr.c (categorize_ctor_elements_1): Likewise. Compute it like
661 p_nz_elts, except don't multiply it by mult. Adjust recursive call.
662 Fix up COMPLEX_CST handling.
663 (categorize_ctor_elements): Add p_unique_nz_elts argument, initialize
664 it and pass it through to categorize_ctor_elements_1.
665 (mostly_zeros_p, all_zeros_p): Adjust categorize_ctor_elements callers.
666 * gimplify.c (gimplify_init_constructor): Likewise. Don't force
667 ctor into readonly data section if num_unique_nonzero_elements is
668 smaller or equal to 1/8 of num_nonzero_elements and size is >= 64
671 2018-12-14 Thomas Preud'homme <thomas.preudhomme@arm.com>
673 * config/arm/arm.md (arm_movdi): Split if -mslow-flash-data and
674 source is a constant that would be loaded by literal pool.
675 (movsf expander): Generate a no_literal_pool_sf_immediate insn if
676 -mslow-flash-data is present, targeting hardfloat ABI and source is a
677 float constant that cannot be loaded via vmov.
678 (movdf expander): Likewise but generate a no_literal_pool_df_immediate
680 (arm_movsf_soft_insn): Split if -mslow-flash-data and source is a
681 float constant that would be loaded by literal pool.
682 (softfloat constant movsf splitter): Splitter for the above case.
683 (movdf_soft_insn): Split if -mslow-flash-data and source is a float
684 constant that would be loaded by literal pool.
685 (softfloat constant movdf splitter): Splitter for the above case.
686 * config/arm/constraints.md (Pz): Document existing constraint.
687 (Ha): Define constraint.
689 * config/arm/predicates.md (hard_sf_operand): New predicate.
690 (hard_df_operand): Likewise.
691 * config/arm/thumb2.md (thumb2_movsi_insn): Split if
692 -mslow-flash-data and constant would be loaded by literal pool.
693 * constant/arm/vfp.md (thumb2_movsi_vfp): Likewise and disable constant
694 load in VFP register.
695 (movdi_vfp): Likewise.
696 (thumb2_movsf_vfp): Use hard_sf_operand as predicate for source to
697 prevent match for a constant load if -mslow-flash-data and constant
698 cannot be loaded via vmov. Adapt constraint accordingly by
699 using Ha instead of E for generic floating-point constant load.
700 (thumb2_movdf_vfp): Likewise using hard_df_operand predicate instead.
701 (no_literal_pool_df_immediate): Add a clobber to use as the
702 intermediate general purpose register and also enable it after reload
703 but disable it constant is a valid FP constant. Add constraints and
704 generate a DI intermediate load rather than 2 SI loads.
705 (no_literal_pool_sf_immediate): Add a clobber to use as the
706 intermediate general purpose register and also enable it after
709 2018-12-14 Uros Bizjak <ubizjak@gmail.com>
712 * internal-fn.def (HYPOT): New.
713 * optabs.def (hypot_optab): New.
714 * config/i386/i386.md (hypot<mode>3): New expander.
716 2018-12-14 Jeff Law <law@redhat.com>
718 * target.def (post_cfi_startproc): Fix text.
720 2018-12-14 Sam Tebbs <sam.tebbs@arm.com>
722 * doc/tm.texi (TARGET_ASM_POST_CFI_STARTPROC): Define.
723 * doc/tm.texi.in (TARGET_ASM_POST_CFI_STARTPROC): Define.
724 * dwarf2out.c (dwarf2out_do_cfi_startproc): Trigger the hook.
725 * hooks.c (hook_void_FILEptr_tree): Define.
726 * hooks.h (hook_void_FILEptr_tree): Define.
727 * target.def (post_cfi_startproc): Define.
729 2018-12-14 Tom de Vries <tdevries@suse.de>
731 * lto-cgraph.c (verify_node_partition): New function.
732 (input_overwrite_node, input_varpool_node): Use verify_node_partition.
734 2018-12-14 H.J. Lu <hongjiu.lu@intel.com>
737 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Don't
738 use get_frame_size ().
740 2018-12-14 Andrew Stubbs <ams@codesourcery.com>
742 * gcc/lra-int.h (lra_register_new_scratch_op): Add third parameter.
743 * gcc/lra-remat.c (update_scratch_ops): Pass icode to
744 lra_register_new_scratch_op.
745 * gcc/lra.c (struct sloc): Add icode field.
746 (lra_register_new_scratch_op): Add icode parameter.
747 Use icode to skip insns that have changed beyond recognition.
749 2018-12-14 Eric Botcazou <ebotcazou@adacore.com>
751 * dwarf2out.c (analyze_discr_in_predicate): Simplify.
752 (analyze_variants_discr): Deal with naked boolean discriminants.
754 2018-12-14 Claudiu Zissulescu <claziss@synopsys.com>
756 * config/arc/arc.c (arc_restore_callee_milli) Don't clobber off
759 2018-12-14 Claudiu Zissulescu <claziss@synopsys.com>
761 * config/arc/arc.h (reg_class): Add SIBCALL_REGS.
763 2018-12-14 Richard Biener <rguenther@suse.de>
765 * gimple-ssa-split-paths.c (is_feasible_trace): Remove
766 duplicated code block.
768 2018-12-14 Alan Modra <amodra@gmail.com>
770 PR rtl-optimization/88311
771 * config/rs6000/rs6000.c (rs6000_call_sysv): Do not mask cookie.
772 (rs6000_sibcall_sysv): Likewise.
774 2018-12-14 Segher Boessenkool <segher@kernel.crashing.org>
776 PR rtl-optimization/88001
777 * function.c (match_asm_constraints_1): Don't invalidly share RTL.
779 2018-12-14 Bin Cheng <bin.cheng@linux.alibaba.com>
781 * auto-profile.c (afdo_annotate_cfg): Call update_max_bb_count even
782 if autofdo counts are all zeros.
784 2018-12-14 Bin Cheng <bin.cheng@linux.alibaba.com>
786 * auto-profile.c (afdo_indirect_call): Skip generating histogram
787 value for internal call.
789 2018-12-14 Bin Cheng <bin.cheng@linux.alibaba.com>
791 * auto-profile.c (AFDO_EINFO): New macro.
792 (class edge_info): New class.
793 (is_edge_annotated, set_edge_annotated): Delete.
794 (afdo_propagate_edge, afdo_propagate_circuit, afdo_propagate): Remove
795 parameter. Adjust edge count computation and annotation using class
797 (afdo_calculate_branch_prob, afdo_annotate_cfg): Likewise.
799 2018-12-13 Michael Ploujnikov <michael.ploujnikov@oracle.com>
801 * ipa-cp.c (print_all_lattices): Skip cp clones.
803 * ipa-cp.c: Fix various comment typos.
805 2018-12-13 Jakub Jelinek <jakub@redhat.com>
807 PR tree-optimization/88444
808 * tree-inline.c (fold_marked_statements): Iterate up to
809 last_basic_block_for_fn rather than n_basic_blocks_for_fn.
811 PR rtl-optimization/88470
812 * cfgcleanup.c (outgoing_edges_match): If the function is
813 shrink-wrapped and bb1 ends with a JUMP_INSN with a single fake
814 edge to EXIT, return false.
816 2018-12-13 Martin Sebor <msebor@redhat.com>
818 * doc/extend.texi: Consistently use @code for const and volatile
819 qualifiers, the true and false constants, and asm statements.
821 2018-12-13 Vladimir Makarov <vmakarov@redhat.com>
823 PR rtl-optimization/88414
824 * lra-int.h (lra_asm_error_p): New.
825 * lra-assigns.c (lra_assign): Check lra_asm_error_p for checking
826 call crossed pseudo assignment correctness.
827 (lra_split_hard_reg_for): Set up lra_asm_error_p.
828 * lra-constraints.c (curr_insn_transform): Ditto.
829 * lra.c (lra_asm_error_p): New.
830 (lra): Reset lra_asm_error_p.
832 2018-12-13 Jakub Jelinek <jakub@redhat.com>
834 PR rtl-optimization/88416
835 * valtrack.c (cleanup_auto_inc_dec): Handle pre/post-inc/dec/modify
836 even if !AUTO_INC_DEC.
838 2018-12-13 David Edelsohn <dje.gcc@gmail.com>
840 * config/rs6000/rs6000.c (rs6000_function_arg): Ensure type is
842 (rs6000_arg_partial_bytes): Same.
844 2018-12-13 Jason Merrill <jason@redhat.com>
846 * gdbinit.in (pp): New macro.
849 2018-12-13 Jakub Jelinek <jakub@redhat.com>
851 PR tree-optimization/88464
852 * tree-vect-stmts.c (vect_build_gather_load_calls): Handle INTEGER_TYPE
853 masktype if mask is non-NULL.
854 (vectorizable_load): Don't reject masked gather loads if masktype
855 in the decl is INTEGER_TYPE.
857 2018-12-13 Andreas Krebbel <krebbel@linux.ibm.com>
859 * config/s390/s390-builtins.def (s390_vec_double_s64): Map to
860 s390_vec_double_s64 instead of s390_vcdgb.
861 (s390_vec_double_u64): Map to s390_vec_double_u64 instead of
864 2018-12-13 Andreas Krebbel <krebbel@linux.ibm.com>
866 * config/s390/vx-builtins.md ("vec_ctd_s64", "vec_ctd_u64")
867 ("vec_ctsl", "vec_ctul"): Replace 0 with VEC_NOINEXACT.
868 ("vec_double_s64", "vec_double_u64"): Replace 4 with VEC_INEXACT.
870 2018-12-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
872 * doc/invoke.texi (-msve-vector-bits): Clarify -msve-vector-bits=128
875 2018-12-13 Wei Xiao <wei3.xiao@intel.com>
877 * common/config/i386/i386-common.c (processor_names): Add cascadelake.
878 (processor_alias_table): Add cascadelake.
879 * config.gcc: Add -march=cascadelake.
880 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
882 * config/i386/i386.c (Add m_CASCADELAKE): New.
883 (processor_cost_table): Add cascadelake.
884 (get_builtin_code_for_version): Handle cascadelake.
885 * config/i386/i386.h (TARGET_CASCADELAKE, PROCESSOR_CASCADELAKE): New.
886 (PTA_CASCADELAKE): Ditto.
887 * doc/invoke.texi: Add -march=cascadelake.
889 2018-12-13 Jakub Jelinek <jakub@redhat.com>
892 * config/i386/i386.md (*movdi_internal, *movsi_internal,
893 *movhi_internal, *movqi_internal): Add alternative(s) to load
894 0 or -1 into k registers using kxor or kxnoq instructions.
897 * config/i386/sse.md (VI1248_AVX512VLBW, AVX512ZEXTMASK): New
899 (<avx512>_testm<mode>3<mask_scalar_merge_name>,
900 <avx512>_testnm<mode>3<mask_scalar_merge_name>): Merge patterns
901 with VI12_AVX512VL and VI48_AVX512VL iterators into ones with
902 VI1248_AVX512VLBW iterator.
903 (*<avx512>_testm<VI1248_AVX512VLBW:mode>3_zext,
904 *<avx512>_testm<VI1248_AVX512VLBW:mode>3_zext_mask,
905 *<avx512>_testnm<VI1248_AVX512VLBW:mode>3_zext,
906 *<avx512>_testnm<VI1248_AVX512VLBW:mode>3_zext_mask): New
910 * config/i386/i386.md (*zero_extendsidi2, zero_extend<mode>di2,
911 *zero_extend<mode>si2, *zero_extendqihi2): Add =*k, *km alternatives.
913 2018-12-12 Tom de Vries <tdevries@suse.de>
914 Jakub Jelinek <jakub@redhat.com>
916 * omp-builtins.def (BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
917 BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
918 BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
919 BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
920 BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
921 BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME): Fix up
922 function types - remove one argument.
924 2018-12-12 Martin Sebor <msebor@redhat.com>
926 * doc/extend.texi (attribute copy): Fix typos.
928 2018-12-12 Martin Sebor <msebor@redhat.com>
930 * doc/extend.texi (Function Attributes): Clarify C++ aspects.
931 (Variable Attributes): Same.
932 (Type Attributes): Same.
934 2018-12-12 Olivier Hainque <hainque@adacore.com>
936 * config/aarch64/aarch64.c (aarch64_override_options): Once arch,
937 cpu and tune were validated, insert SUBTARGET_OVERRIDE_OPTIONS if
940 2018-12-12 Olivier Hainque <hainque@adacore.com>
942 * config/aarch64/aarch64.c (PROBE_STACK_FIRST_REG) : Redefine as
943 R9_REGNUM instead of 9.
944 (PROBE_STACK_SECOND_REG): Redefine as R10_REGNUM instead of 10.
946 2018-12-12 Eric Botcazou <ebotcazou@adacore.com>
949 * config/sparc/sparc.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
950 (speculation_barrier): New instruction for V9.
952 2018-12-12 Segher Boessenkool <segher@kernel.crashing.org>
953 Iain Sandoe <iain@sandoe.co.uk>
956 * config/rs6000/rs6000.c (save_reg_p): Do not save the picbase reg
957 unless it has been used.
958 (first_reg_to_save): Remove dead code.
960 2018-12-12 Iain Sandoe <iain@sandoe.co.uk>
962 * config/rs6000/rs6000-protos.h (rs6000_call_darwin): New.
963 (rs6000_sibcall_darwin): New. (macho_call_template): Remove.
964 * config/rs6000/rs6000.c (get_prev_label): Forward declaration.
965 (rs6000_call_template_1): Handle Darwin.
966 (macho_call_template): Remove.
967 (rs6000_call_sysv): Remove handling for Darwin.
968 (rs6000_call_darwin_1, rs6000_call_darwin): New
969 (rs6000_sibcall_darwin): New.
970 * config/rs6000/rs6000.md (define_expand “call”): Handle Darwin
971 with its own expander. (define_expand “call_value”): Likewise.
972 (define_expand “sibcall”): Likewise.
973 (define_expand “sibcall_value”): Likewise.
974 (call_nonlocal_sysv): Remove Darwin special-casing.
975 (call_value_nonlocal_sysv): Likewise.
977 2018-12-12 Iain Sandoe <iain@sandoe.co.uk>
979 * config/rs6000/darwin.md (call_indirect_nonlocal_darwin64): Remove.
980 (call_nonlocal_darwin64): Remove.
981 (call_value_indirect_nonlocal_darwin64): Remove.
982 (call_value_nonlocal_darwin64): Remove.
983 * config/rs6000/rs6000.c (rs6000_call_template_1): Handle Darwin with
984 the same asm output as AIX/ELFv2. (rs6000_call_sysv): Preserve the
985 CALL_LONG flag when needed for Darwin.
986 * config/rs6000/rs6000.md (define expand “call”): Expand Darwin as
988 (define_expand “call_value”): Likewise.
989 (define_expand “sibcall”): Likewise.
990 (define_expand “sibcall_value”): Likewise.
991 (call_indirect_nonlocal_sysv): Mark the clobber mode P.
992 (call_nonlocal_sysv): Likewise.
993 (call_nonlocal_sysv_secure): Likewise.
994 (call_value_indirect_nonlocal_sysv): Likewise.
995 (call_value_nonlocal_sysv): Likewise.
996 (call_value_nonlocal_sysv_secure): Likewise.
997 (call_local64): Mark the clobber DI.
998 (call_value_local64): Likewise.
1000 2018-12-11 Jakub Jelinek <jakub@redhat.com>
1002 PR tree-optimization/88444
1003 * tree-vrp.c (register_edge_assert_for_2): Only register assertions
1004 for conversions if rhs1 is a SSA_NAME.
1006 2018-12-11 Dimitar Dimitrov <dimitar@dinux.eu>
1008 * cfgexpand.c (asm_clobber_reg_is_valid): Also produce
1009 error when stack pointer is clobbered.
1010 (expand_asm_stmt): Refactor clobber check in separate function.
1012 2018-12-11 Eric Botcazou <ebotcazou@adacore.com>
1014 * config/rs6000/vxworks.h (RS6000_STARTING_FRAME_OFFSET): Define,
1015 accounting for STACK_BOUNDARY 128.
1016 (STACK_DYNAMIC_OFFSET): Likewise.
1018 2018-12-11 Jakub Jelinek <jakub@redhat.com>
1021 * config/i386/i386.md (*x86_mov<SWI48:mode>cc_0_m1_neg_leu<SWI:mode>):
1022 New define_insn_and_split.
1024 2018-12-11 Richard Biener <rguenther@suse.de>
1028 * gimple.c (gimple_assign_set_rhs_with_ops): Revert previous
1030 * tree-complex.c (update_complex_assignment): Properly transfer
1031 or clean EH info around gimple_assign_set_rhs_with_ops.
1033 2018-12-11 Richard Earnshaw <rearnsha@arm.com>
1036 * config/aarch64/iterators.md (sizem1): Add sizes for SFmode and DFmode.
1037 (Vbtype): Add SFmode mapping.
1038 * config/aarch64/aarch64.md (copysigndf3, copysignsf3): Delete.
1039 (copysign<GPF:mode>3): New expand pattern.
1040 (copysign<GPF:mode>3_insn): New insn pattern.
1042 2018-12-11 Richard Biener <rguenther@suse.de>
1044 * ccmp.c (ccmp_candidate_p): Use GIMPLE API properly.
1046 2018-12-11 Jakub Jelinek <jakub@redhat.com>
1049 * doc/sourcebuild.texi (lto_incremental): Document new effective
1052 2018-12-11 Segher Boessenkool <segher@kernel.crashing.org>
1055 * config/rs6000/rs6000.c (rs6000_expand_zeroop_builtin): Use
1056 rs6000_isa_flags instead of rs6000_isa_flags_explicit to decide
1057 whether soft float is enabled.
1058 (rs6000_expand_mtfsb_builtin): Ditto.
1059 (rs6000_expand_set_fpscr_rn_builtin): Ditto.
1060 (rs6000_expand_set_fpscr_drn_builtin): Ditto.
1062 2018-12-10 Jeff Law <law@redhat.com>
1064 PR tree-optimization/80520
1065 * gimple-ssa-split-paths.c (is_feasible_trace): Recognize half
1066 diamonds that are likely if convertable.
1068 2018-12-10 Martin Sebor <msebor@redhat.com>
1070 PR tree-optimization/86196
1071 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
1072 base size only of arrays.
1074 2018-12-10 Segher Boessenkool <segher@kernel.crashing.org>
1076 * config.gcc (Obsolete configurations): Delete powerpc*-*-*spe*.
1077 (Unsupported targets): Add powerpc*-*-*spe*.
1078 (powerpc*-*-*spe*): Delete.
1079 (powerpc-*-eabispe*): Delete.
1080 (powerpc-*-rtems*spe*): Delete.
1081 (powerpc*-*-linux*spe*): Delete.
1082 (powerpc*-*-linux*): Do not handle the linux*spe* targets.
1083 (powerpc-wrs-vxworks*spe): Delete.
1084 (with_cpu setting code): Delete powerpc*-*-*spe* handling.
1085 * config.host (target powerpc*-*-*spe*): Delete.
1086 * doc/invoke.texi (PowerPC SPE Options): Delete.
1087 (PowerPC SPE Options): Delete.
1088 * config/powerpcspe: Delete.
1090 2018-12-10 Uros Bizjak <ubizjak@gmail.com>
1093 * config/i386/i386.c (ix86_expand_sse_cmp): For vector modes,
1094 check operand 1 with vector_operand predicate.
1095 (ix86_expand_sse_movcc): For vector modes, check op_true with
1096 vector_operand, not nonimmediate_operand.
1098 2018-12-10 Richard Biener <rguenther@suse.de>
1100 * tree-affine.c (tree_to_aff_combination): Remove unreachable
1102 (aff_combination_expand): Cache on SSA names, not possibly
1103 on conversion trees. Avoid expanding cases we do not handle.
1105 2018-12-10 Richard Biener <rguenther@suse.de>
1107 PR tree-optimization/88427
1108 * vr-values.c (vr_values::extract_range_from_phi_node):
1109 Handle symbolic ranges conservatively when trying to drop
1112 2018-12-10 Martin Jambor <mjambor@suse.cz>
1115 * ipa-prop.c (determine_locally_known_aggregate_parts): Make sure
1116 we check pointers against pointers.
1118 2018-12-10 Richard Biener <rguenther@suse.de>
1121 * gimple.c (gimple_assign_set_rhs_with_ops): Transfer EH
1122 info to a newly allocated stmt.
1124 2018-12-10 Jerome Lambourg <lambourg@adacore.com>
1126 * config/vxworksae.h (TARGET_VXWORKS_HAVE_CTORS_DTORS): Define.
1128 2018-12-10 Fredrik Nyström <fredrik@lysator.liu.se>
1131 * config/sol2.h: Only use libgcc-unwind.map if
1132 ENABLE_SHARED_LIBGCC.
1134 2018-12-10 Bin Cheng <bin.cheng@linux.alibaba.com>
1136 * auto-profile.c (afdo_calculate_branch_prob): Convert profile_count
1137 to afdo precision quality.
1138 (afdo_callsite_hot_enough_for_early_inline): Likewise.
1139 (afdo_annotate_cfg): Set profile_count for exit basic block.
1141 2018-12-10 Bin Cheng <bin.cheng@linux.alibaba.com>
1143 * auto-profile.c (string_table::get_index_by_decl)
1144 (function_instance::get_function_instance_by_decl): Avoid infinite
1145 recursion by using DECL_FROM_INLINE.
1147 2018-12-09 John David Anglin <danglin@gcc.gnu.org>
1149 * config.gcc (hppa*-*-linux*): Add pa/t-pa to tmake_file. Define
1151 * config/pa/pa-protos.h (pa_d_target_versions): New prototype.
1152 * config/pa/pa.h (TARGET_D_CPU_VERSIONS): Define.
1153 * config/pa/pa-d.c: New file.
1154 * config/pa/t-pa: New file.
1156 2018-12-08 Bin Cheng <bin.cheng@linux.alibaba.com>
1158 * ipa-cp.c (update_profiling_info): Call adjust_for_ipa_scaling for
1161 2018-12-08 Jakub Jelinek <jakub@redhat.com>
1164 * tree-nested.c (convert_local_reference_stmt): Handle clobbers where
1165 lhs is not a decl normally, don't call use_pointer_in_frame on that
1168 2018-12-08 Eric Botcazou <ebotcazou@adacore.com>
1170 PR rtl-optimization/88390
1171 * dwarf2cfi.c (struct dw_cfi_row): Add window_save field.
1172 (cfi_row_equal_p): Compare it.
1173 (dwarf2out_frame_debug_cfa_window_save): Add FAKE parameter.
1174 If FAKE is false, set window_save of the current row.
1175 (dwarf2out_frame_debug) <REG_CFA_TOGGLE_RA_MANGLE>: Call above
1176 function with FAKE parameter set to true.
1177 <REG_CFA_WINDOW_SAVE>: Likewise but with FAKE parameter set to false.
1178 (change_cfi_row): Emit a DW_CFA_GNU_window_save if necessary.
1180 2018-12-07 Peter Bergner <bergner@linux.ibm.com>
1183 * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
1184 -mabi=ieeelongdouble and -mabi=ibmlongdouble without -mlong-double-128.
1185 Do not error for -mabi=ibmlongdouble and no ISA 2.06 support.
1186 * doc/invoke.texi: Document -mabi=ibmlongdouble and -mabi=ieeelongdouble
1187 require -mlong-double-128.
1189 2018-12-07 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1191 * config/aarch64/aarch64.c (aarch64_tuning_override_functions): Add
1193 (aarch64_parse_sve_width_string): Define.
1195 2018-12-07 Jeff Law <law@redhat.com>
1198 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Add
1199 m_update_global_ranges member. Add corresponding argument to ctor.
1200 * gimple-ssa-evrp-analyze.c
1201 (evrp_range_analyzer::evrp_range_analyzer): Add new argument and
1202 initialize m_update_global_ranges.
1203 (evrp_range_analyzer::set_ssa_range_info): Assert that we are
1204 updating global ranges.
1205 (evrp_range_analyzer::record_ranges_from_incoming_edge): Only
1206 update global ranges if explicitly requested.
1207 (evrp_range_analyzer::record_ranges_from_phis): Similarly.
1208 (evrp_range_analyzer::record_ranges_from_stmt): Similarly.
1209 * gimple-ssa-evrp.c (evrp_dom_walker): Pass new argument to
1210 evrp_range_analyzer ctor.
1211 * gimple-ssa-sprintf.c (sprintf_dom_walker): Similarly.
1212 * tree-ssa-dom.c (dom_opt_dom_walker): Similarly.
1214 2018-12-07 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1216 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum):
1217 Add SVE_NOT_IMPLEMENTED value.
1218 * config/aarch64/aarch64-protos.h (struct tune_params): Add sve_width
1220 * config/aarch64/aarch64.c (generic_tunings,cortexa35_tunings,
1221 cortexa53_tunings, cortexa57_tunings, cortexa72_tunings,
1222 cortexa73_tunings, exynosm1_tunings, thunderx_tunings,
1223 thunderx_tunings, tsv110_tunings, xgene1_tunings, qdf24xx_tunings,
1224 saphira_tunings, thunderx2t99_tunings, emag_tunings):
1226 (aarch64_estimated_poly_value): Define.
1227 (TARGET_ESTIMATED_POLY_VALUE): Define.
1229 2018-12-07 Paul A. Clarke <pc@us.ibm.com>
1232 * config/rs6000/mmintrin.h (_mm_packs_pu16): Correctly use "__vector".
1234 2018-12-07 Vladimir Makarov <vmakarov@redhat.com>
1236 PR rtl-optimization/88349
1237 * ira-costs.c (record_operand_costs): Check bigger reg class on
1240 2018-12-07 Richard Sandiford <richard.sandiford@arm.com>
1242 * config/aarch64/aarch64-sve.md (*mul<mode>3, *v<optab><mode>3):
1243 Split the patterns after reload if we don't need the predicate
1245 (*post_ra_mul<mode>3, *post_ra_v<optab><mode>3): New patterns.
1247 2018-12-07 Richard Sandiford <richard.sandiford@arm.com>
1249 * config/aarch64/iterators.md (SVE_UNPRED_FP_BINARY): New code
1251 (sve_fp_op): Handle minus and mult.
1252 * config/aarch64/aarch64-sve.md (*add<mode>3, *sub<mode>3)
1253 (*mul<mode>3): Split the patterns after reload if we don't
1254 need the predicate operand.
1255 (*post_ra_<sve_fp_op><mode>3): New pattern.
1257 2018-12-07 Bin Cheng <bin.cheng@linux.alibaba.com>
1259 * profile-count.h (profile_count::oeprator>=): Fix typo by inverting
1260 return condition when *this is precise zero.
1262 2018-12-07 Jakub Jelinek <jakub@redhat.com>
1265 * final.c (rest_of_handle_final): Don't call collect_fn_hard_reg_usage
1266 for functions with naked attribute.
1269 * tree-vrp.c (extract_range_from_binary_expr): For POINTER_PLUS_EXPR
1270 with -fno-delete-null-pointer-checks, set_nonnull only if the pointer
1271 is non-NULL and offset is known to have most significant bit clear.
1272 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): For ADDR_EXPR
1273 of MEM_EXPR, return true if the MEM_EXPR has non-zero offset with
1274 most significant bit clear. If offset does have most significant bit
1275 set and -fno-delete-null-pointer-checks, don't return true even if
1276 the base pointer is non-NULL.
1278 2018-12-06 Alexandre Oliva <aoliva@redhat.com>
1280 * cselib.c (cselib_record_sets): Skip strict low part sets
1283 2018-12-06 Paul A. Clarke <pc@us.ibm.com>
1286 * config/rs6000/smmintrin.h: New file.
1287 * config.gcc: Add smmintrin.h to extra_headers for powerpc*-*-*.
1289 2018-12-06 Paul A. Clarke <pc@us.ibm.com>
1292 * config/rs6000/mmintrin.h (_mm_unpackhi_pi8): Fix for big-endian.
1293 (_mm_unpacklo_pi8): Likewise.
1294 (_mm_mulhi_pi16): Likewise.
1295 (_mm_packs_pi16): Fix for big-endian. Use preferred API.
1296 (_mm_packs_pi32): Likewise.
1297 (_mm_packs_pu16): Likewise.
1298 * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix for big-endian.
1299 (_mm_cvtss_si64): Likewise.
1300 (_mm_cvtpi32x2_ps): Likewise.
1301 (_mm_shuffle_ps): Likewise.
1302 (_mm_movemask_pi8): Likewise.
1303 (_mm_mulhi_pu16): Likewise.
1304 (_mm_sad_pu8): Likewise.
1305 (_mm_sad_pu8): Likewise.
1306 (_mm_cvtpu16_ps): Fix for big-endian. Use preferred API.
1307 (_mm_cvtpu8_ps): Likewise.
1308 (_mm_movemask_ps): Better #else case for big-endian (no functional
1310 (_mm_shuffle_pi16): Likewise.
1311 * config/rs6000/emmintrin.h (_mm_movemask_pd): Fix for big-endian.
1312 Better #else case for big-endian (no functional change).
1313 (_mm_movemask_epi8): Likewise.
1314 (_mm_shufflehi_epi16): Likewise.
1315 (_mm_shufflelo_epi16): Likewise.
1316 (_mm_shuffle_epi32): Likewise.
1317 (_mm_mul_epu32): Fix for big-endian.
1318 (_mm_bsrli_si128): Likewise.
1319 (_mm_cvtps_pd): Better #else case for big endian.
1320 (_mm_mulhi_epi16): Likewise.
1321 (_mm_mul_epu32): Likewise.
1322 (_mm_slli_si128): Likewise.
1323 (_mm_sll_epi16): Likewise.
1324 (_mm_sll_epi32): Likewise.
1325 (_mm_sra_epi16): Likewise.
1326 (_mm_sra_epi32): Likewise.
1327 (_mm_srl_epi16): Likewise.
1328 (_mm_srl_epi32): Likewise.
1329 (_mm_mulhi_epu16): Likewise.
1330 (_mm_sad_epu8): Likewise.
1331 * config/rs6000/pmmintrin.h (_mm_hadd_ps): Fix for big-endian.
1332 (_mm_sub_ps): Likewise.
1333 * config/rs6000/mmintrin.h (_mm_cmpeq_pi8): Fix for 32-bit mode.
1334 * gcc/config/rs6000/tmmintrin.h (_mm_alignr_epi8): Use ENDIAN
1335 macros consistently (no functional changes).
1336 (_mm_alignr_pi8): Likewise.
1338 2018-12-06 Iain Sandoe <iain@sandoe.co.uk>
1341 * config/darwin.h (TARGET_WEAK_NOT_IN_ARCHIVE_TOC) Remove, use the
1343 * config/rs6000/darwin7.h (TARGET_WEAK_NOT_IN_ARCHIVE_TOC): New.
1345 2018-12-06 Vladimir Makarov <vmakarov@redhat.com>
1348 * ira.c (ira_init_register_move_cost): Use info from
1349 hard_regno_mode_ok instead of contains_reg_of_mode.
1350 * ira-costs.c (contains_reg_of_mode): Don't use cost from bigger
1351 hard register class for some fixed hard registers.
1353 2018-12-06 Segher Boessenkool <segher@kernel.crashing.org>
1355 * doc/extend.texi (Using Assembly Language with C): Document asm inline.
1356 (Size of an asm): Fix typo. Document asm inline.
1357 * gimple-pretty-print.c (dump_gimple_asm): Handle asm inline.
1358 * gimple.h (enum gf_mask): Add GF_ASM_INLINE.
1359 (gimple_asm_set_volatile): Fix typo.
1360 (gimple_asm_inline_p): New.
1361 (gimple_asm_set_inline): New.
1362 * gimplify.c (gimplify_asm_expr): Propagate the asm inline flag from
1364 * ipa-icf-gimple.c (func_checker::compare_gimple_asm): Compare the
1365 gimple_asm_inline_p flag, too.
1366 * tree-core.h (tree_base): Document that protected_flag is ASM_INLINE_P
1368 * tree-inline.c (estimate_num_insns): If gimple_asm_inline_p return
1369 a minimum size for an asm.
1370 * tree.h (ASM_INLINE_P): New.
1372 2018-12-06 Segher Boessenkool <segher@kernel.crashing.org>
1375 * doc/extend.texi (Basic Asm): Update grammar.
1376 (Extended Asm): Update grammar.
1378 2018-12-06 Iain Sandoe <iain@sandoe.co.uk>
1381 * config/i386/darwin.h (STACK_BOUNDARY): Remove macro.
1382 * config/i386/i386.c (ix86_compute_frame_layout): Ensure at least 128b
1383 stack alignment in non-leaf functions.
1385 2018-12-06 Jakub Jelinek <jakub@redhat.com>
1388 * config/aarch64/aarch64.c (aarch64_print_address_internal): Don't
1389 call output_operand_lossage on VOIDmode CONST_INTs. After
1390 output_operand_lossage do return false.
1392 2018-12-06 Richard Biener <rguenther@suse.de>
1394 * df-problems.c (df_rd_local_compute): Use bitmap_release.
1395 (df_live_free): Likewise.
1396 (df_md_local_compute): Likewise.
1397 (df_md_free): Release df_md_scratch bitmap.
1398 * loop-invariant.c (calculate_loop_reg_pressure): Use
1400 * sched-deps.c (true_dependency_cache, output_dependency_cache,
1401 anti_dependency_cache, control_dependency_cache,
1402 spec_dependency_cache): Use bitmap instead of bitmap_head *.
1403 * sched-ebb.c (schedule_ebbs_init): Initialize non-GTY
1404 dont_calc_deps as bitmap allocated from obstack not GC.
1405 (schedule_ebbs_finish): Use bitmap_release.
1406 * sched-rgn.c (schedule_insns): Initialize non-GTY
1407 not_in_df as bitmap allocated from obstack not GC.
1409 * sel-sched.c (_forced_ebb_heads): Remove premature optimization.
1410 (sel_region_init): Allocate forced_ebb_heads.
1411 (sel_region_finish): Free forced_ebb_heads.
1413 2018-12-06 Richard Biener <rguenther@suse.de>
1415 * bitmap.c (bitmap_head::crashme): Define.
1416 * bitmap.h (bitmap_head): Add constexpr default constructor
1417 poisoning the obstack member.
1418 (bitmap_head::crashme): Declare.
1419 (bitmap_release): New function clearing a bitmap and poisoning
1421 * gengtype.c (main): Make it recognize CONSTEXPR.
1422 * lra-constraints.c (lra_inheritance): Use bitmap_release
1423 instead of bitmap_clear.
1424 * ira.c (ira): Work around class-memaccess warning.
1425 * regrename.c (create_new_chain): Likewise.
1427 2018-12-06 Jakub Jelinek <jakub@redhat.com>
1429 PR tree-optimization/85726
1430 * generic-match-head.c (optimize_successive_divisions_p): New function.
1431 * gimple-match-head.c (optimize_successive_divisions_p): Likewise.
1432 * match.pd: Don't combine successive divisions if they aren't exact
1433 and optimize_successive_divisions_p is false.
1436 * tree-nested.c (convert_nonlocal_reference_stmt): Remove clobbers
1437 for non-local automatic decls.
1439 2018-12-05 David Edelsohn <dje.gcc@gmail.com>
1441 * config/rs6000/aix72.h (ASM_DEFAULT_SPEC): Match Power7 processor
1444 2018-12-05 Uros Bizjak <ubizjak@gmail.com>
1446 * dwarf2out.c (field_byte_offset): Remove
1447 #ifdef PCC_BITFIELD_TYPE_MATTERS check.
1449 2018-12-05 Richard Biener <rguenther@suse.de>
1450 Martin Sebor <msebor@redhat.com>
1453 * calls.c (get_attr_nonstring_decl): Avoid setting *REF to
1455 * gcc/gimple-low.c (lower_stmt): Fold builtin calls here.
1456 * gimplify (maybe_fold_stmt): Avoid folding builtin calls.
1458 2018-12-05 Iain Sandoe <iain@sandoe.co.uk>
1460 * configure.ac (gcc_cv_otool): Set.
1461 * configure: Regenerate.
1463 2018-12-05 Uros Bizjak <ubizjak@gmail.com>
1465 * config/i386/cygming.h (PCC_BITFIELD_TYPE_MATTERS): Remove.
1466 (GROUP_BITFIELDS_BY_ALIGN): Ditto.
1468 2018-12-05 Richard Sandiford <richard.sandiford@arm.com>
1470 PR tree-optimization/88064
1471 * tree-vect-patterns.c (vect_convert_input): Convert the result of
1472 an existing cast if it has the right width but the wrong sign.
1473 Do not test the signedness of the required result when
1474 considering whether to split an existing cast; instead split to
1475 a type with the same signedness as the source of the cast, then
1476 convert it to the opposite signedness where necessary.
1477 (vect_recog_over_widening_pattern): Handle sign changes between
1478 the final PLUS_EXPR and the RSHIFT_EXPR.
1479 (vect_recog_average_pattern): Use an unsigned operation when
1480 truncating an addition, subtraction or multiplication. Cast the
1481 result back to the "real" signedness before promoting.
1483 2018-12-05 Richard Biener <rguenther@suse.de>
1485 PR tree-optimization/86637
1486 * tree-vectorizer.c (pass_slp_vectorize::execute): Reset
1487 vect_location at the end.
1489 2018-12-05 Jakub Jelinek <jakub@redhat.com>
1491 PR tree-optimization/87360
1492 * gimple-loop-jam.c (tree_loop_unroll_and_jam): On failure to analyze
1493 data dependencies, don't return false, just continue. Formatting
1495 (merge_loop_tree, bb_prevents_fusion_p, unroll_jam_possible_p,
1496 fuse_loops): Formatting fixes.
1499 * cfgexpand.c (expand_stack_vars): If asan_vec is empty, start with
1500 aligning frame offset to ASAN_RED_ZONE_SIZE bytes.
1502 2018-12-05 Xianmiao Qu <xianmiao_qu@c-sky.com>
1504 * config.gcc (csky-*-linux-gnu*): Force .init_array support.
1506 2018-12-05 Segher Boessenkool <segher@kernel.crashing.org>
1508 * config/rs6000/rs6000.c (rs6000_function_arg): Only do the special
1509 aggregate handling on actual AIX, not on somewhat similar systems.
1510 (rs6000_arg_partial_bytes): Ditto.
1512 2018-12-04 Vladimir Makarov <vmakarov@redhat.com>
1514 PR rtl-optimization/88317
1515 * lra-constraints.c (split_reg): Don't set up check_only_regs if
1516 we are outside the inheritance pass.
1518 2018-12-04 Jan Hubicka <hubicka@ucw.cz>
1520 * ipa-prop.c (jump_function_useful_p): New.
1521 (ipa_read_node_info): Do not allocated useless jump functions.
1523 2018-12-04 David Malcolm <dmalcolm@redhat.com>
1526 * spellcheck.c (selftest::test_suggestions): Add another assertion.
1528 2018-12-04 Martin Sebor <msebor@redhat.com>
1530 * doc/extend.texi (__builtin_has_attribute): Fix typo.
1532 2018-12-04 Jakub Jelinek <jakub@redhat.com>
1535 * config/rs6000/rs6000.c (ccr_bit): Return -1 instead of assertion
1537 (print_operand): Use REG_P instead of GET_CODE == REG.
1538 <case 'D'>: Also check CR_REGNO_P (REGNO (x)).
1539 <case 't'>: Likewise. Remove GET_MODE check.
1540 * config/rs6000/rs6000.md (scc patterns): Assert ccr_bit didn't
1543 2018-12-04 Martin Sebor <msebor@redhat.com>
1545 * doc/extend.texi (attribute aligned): Expand.
1547 2018-12-04 David Edelsohn <dje.gcc@gmail.com>
1550 * config/rs6000/rs6000.c (rs6000_function_arg): Don't pass aggregates
1552 (rs6000_arg_partial_bytes): Same.
1554 2018-12-04 Vladimir Makarov <vmakarov@redhat.com>
1557 * ira-costs.c (exec): Try bigger class to use smaller register
1560 2018-12-04 Michael Ploujnikov <michael.ploujnikov@oracle.com>
1563 * ipa-cp.c (create_specialized_node): Track clone counters by
1564 node assembler names.
1565 (ipcp_driver): Change type of clone_num_suffixes key to const char*.
1567 2018-12-04 Richard Biener <rguenther@suse.de>
1569 * tree-vect-loop.c (vect_transform_loop): Fix epilogue iterations
1570 for epilogue vectorization.
1572 2018-12-04 Richard Biener <rguenther@suse.de>
1574 PR tree-optimization/88301
1575 * tree-vrp.c (register_edge_assert_for_2): Fix sign-conversion
1576 issues in last commit.
1578 2018-12-04 Richard Biener <rguenther@suse.de>
1580 PR tree-optimization/88315
1581 * tree-vect-loop.c (get_initial_defs_for_reduction): Simplify
1582 and fix initialization vector for SAD and DOT_PROD SLP reductions.
1584 2018-12-03 Sandra Loosemore <sandra@codesourcery.com>
1587 * doc/extend.texi (Nonlocal gotos): New section.
1589 2018-12-03 Uros Bizjak <ubizjak@gmail.com>
1591 * config/i386/sse.md (vec_concatv2df): Change (v,xm,C)
1592 alternative to (v,vm,c).
1594 2018-12-03 Segher Boessenkool <segher@kernel.crashing.org>
1596 * gcc/config.gcc (Unsupported targets): Add powerpc*-*-linux*paired*.
1597 (powerpc*-*-linux*): Don't handle it here.
1599 2018-12-03 Jeff Law <law@redhat.com>
1601 * config/h8300/h8300.md (<code>qi3_1): Add gcc_unreachable to keep
1604 * config/h8300/h8300.md (low byte masking peepholes): Only accept
1605 nonimmediate_operand for the to-be-masked operand.
1607 2018-12-03 Richard Biener <rguenther@suse.de>
1609 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
1610 allow expected function-pointer cast re-instantiation.
1612 2018-12-03 Ilya Leoshkevich <iii@linux.ibm.com>
1614 * common/config/s390/s390-common.c (s390_option_init_struct):
1615 Use section anchors by default.
1616 * config/s390/s390.c (s390_check_symref_alignment): Handle
1618 (TARGET_MAX_ANCHOR_OFFSET): Use short displacement.
1619 * output.h (assemble_align): Pass `align' as unsigned int, so
1620 that the value 0x80000000, which corresponds to `aligned(1 <<
1621 28)', would pass the `align > BITS_PER_UNIT' check.
1622 * varasm.c (assemble_align): Likewise.
1624 2018-12-03 Julian Brown <julian@codesourcery.com>
1626 * tree-pretty-print.c (dump_omp_clause): Make default case
1629 2018-12-03 Richard Biener <rguenther@suse.de>
1631 PR tree-optimization/88301
1632 * tree-vrp.c (register_edge_assert_for_2): Handle conversions
1633 that do not change the value by registering the same assert
1636 2018-12-03 Richard Biener <rguenther@suse.de>
1638 * tree-vect-loop.c (vect_transform_loop): Properly compute
1639 upper bound for the epilogue when doing epilogue vectorization.
1641 2018-12-03 Ilya Leoshkevich <iii@linux.ibm.com>
1644 * cfgcleanup.c (class pass_postreload_jump): New pass.
1645 (pass_postreload_jump::execute): Likewise.
1646 (make_pass_postreload_jump): Likewise.
1647 * passes.def: Add pass_postreload_jump before
1648 pass_postreload_cse.
1649 * tree-pass.h (make_pass_postreload_jump): New pass.
1651 2018-12-02 Uros Bizjak <ubizjak@gmail.com>
1653 * config/i386/i386.c (inline_memory_move_cost): Correctly
1654 compute number of TFmode GPR moves.
1656 2018-12-02 Jakub Jelinek <jakub@redhat.com>
1659 * config/i386/sse.md (*vec_concatv4sf_0, *vec_concatv4si_0): New insns.
1661 2018-12-02 Jeff Law <law@redhat.com>
1663 * config/h8300/h8300.md (call, call_value): Drop mode from
1664 operand containing number of bytes pushed.
1665 (call_insn, call_value_insn): Use mode iterator for number
1666 of bytes pushed operand. Make pattern name vary appropriately.
1668 2018-12-02 Jan Hubicka <jh@suse.cz>
1670 * bb-reorer.c (find_rarely_executed_basic_blocks_and_crossin_edges):
1671 trust precise profiles.
1673 2018-12-02 Jakub Jelinek <jakub@redhat.com>
1675 * tree-nested.c (convert_nonlocal_omp_clauses,
1676 convert_local_omp_clauses): Handle OMP_CLAUSE_IN_REDUCTION,
1677 OMP_CLAUSE_TASK_REDUCTION and OMP_CLAUSE__SIMT_ clauses.
1678 (convert_nonlocal_reference_stmt, convert_local_reference_stmt):
1679 Convert clauses for GIMPLE_OMP_TASKGROUP.
1681 * omp-low.c (check_omp_nesting_restrictions): Allow cancel or
1682 cancellation point with taskgroup clause inside of taskloop. Consider
1683 a taskloop construct without nogroup clause as implicit taskgroup for
1684 diagnostics if cancel/cancellation point with taskgroup clause is
1685 closely nested inside of taskgroup region.
1688 * asan.c (asan_clear_shadow): Move assert that len is multiple of 4
1689 to the start of the function.
1690 (asan_emit_stack_protection): When emitting clearing sequence for
1691 epilogue, align offset down to ASAN_RED_ZONE_SIZE granularity,
1692 add last_size_aligned which is last_size padded to multiples of
1693 ASAN_RED_ZONE_SIZE and emit asan_clear_shadow always on 4 byte
1696 2018-12-01 Jakub Jelinek <jakub@redhat.com>
1699 * asan.c (asan_redzone_buffer::flush_redzone_payload): Fix up
1700 an off-by-one for BYTES_BIG_ENDIAN.
1703 * combine.c (find_split_point): For invalid memory address
1704 nonobj + obj + const, if reg + obj + const is valid addressing
1705 mode, split at nonobj. Use if rather than else if for the
1706 fallback. Comment fixes.
1708 2018-11-30 Indu Bhagat <indu.bhagat@oracle.com>
1710 * coverage.c (get_coverage_counts): Use from_function_decl for precise
1712 * profile-count.c (profile_count::dump): Add handling for precise
1714 * profile.c (compute_branch_probabilities): Rely on exec_counts instead
1715 of profile_info to set profile_status of function.
1716 (branch_prob): Do not set profile_status of function based on
1717 profile_info. Done above based on exec_counts.
1719 2018-11-30 Jakub Jelinek <jakub@redhat.com>
1721 PR tree-optimization/88274
1722 * tree-ssa-reassoc.c (optimize_range_tests_xor,
1723 optimize_range_tests_diff): If type has smaller precision than
1724 corresponding mode or if it has non-standard min/max, compute
1725 everything in a standard type for the precision.
1728 * params.def (PARAM_LOGICAL_OP_NON_SHORT_CIRCUIT): New param.
1729 * tree-ssa-ifcombine.c (ifcombine_ifandif): If
1730 --param logical-op-non-short-circuit is present, override
1731 LOGICAL_OP_NON_SHORT_CIRCUIT value from the param.
1732 * fold-const.c (fold_range_test, fold_truth_andor): Likewise.
1734 2018-11-30 Jeff Law <law@redhat.com>
1736 * optabs.c (expand_binop): Use "machine_mode" rather than
1737 "enum machine mode" in most recent change.
1739 2018-11-30 Wilco Dijkstra <wdijkstr@arm.com>
1742 * builtins.c (expand_builtin_longjmp): Use a temporary when restoring
1744 (expand_builtin_nonlocal_goto): Likewise.
1746 2018-11-30 David Malcolm <dmalcolm@redhat.com>
1748 * diagnostic-core.h (emit_diagnostic): New decl.
1749 * diagnostic.c (emit_diagnostic): New overload, taking a
1752 2018-11-30 David Malcolm <dmalcolm@redhat.com>
1754 * pretty-print.c (class selftest::test_pretty_printer): New
1755 subclass of pretty_printer.
1756 (selftest::test_prefixes_and_wrapping): New test.
1757 (selftest::pretty_print_c_tests): Call it.
1759 2018-11-30 Michael Ploujnikov <michael.ploujnikov@oracle.com>
1761 Minimize clone counter memory usage in create_virtual_clone.
1762 * cgraph.h (clone_function_name): Add a variant that takes a
1764 * cgraph.h (cgraph_node::create_virtual_clone): Add a new
1765 argument: num_suffix.
1766 * cgraphclones.c (cgraph_node::create_virtual_clone): Pass
1767 num_suffix to clone_function_name.
1768 (clone_function_name): Add a variant that takes a tree decl.
1769 * ipa-cp.c (create_specialized_node): Keep track of clone
1770 counters in clone_num_suffixes hash map.
1771 (ipcp_driver): Free the counter hash map.
1772 * ipa-hsa.c (process_hsa_functions): Creates at most one hsa
1775 2018-11-30 Michael Ploujnikov <michael.ploujnikov@oracle.com>
1776 Make function assembly more independent.
1778 This is achieved by having clone_function_name assign unique clone
1779 numbers for each function independently.
1781 * cgraphclones.c: Replaced clone_fn_id_num with clone_fn_ids;
1783 (clone_function_name_numbered): Use clone_fn_ids.
1785 2018-11-30 Vladimir Makarov <vmakarov@redhat.com>
1787 PR rtl-optimization/88179
1788 * lra-constraints.c (address_eliminator::address_eliminator):
1789 Don't eleminate regs in illegitimate address.
1791 2018-11-30 David Malcolm <dmalcolm@redhat.com>
1793 PR preprocessor/88257
1794 * input.c (get_substring_ranges_for_loc): Fix indentation.
1795 Bulletproof against getting a different files back from
1796 linemap_resolve_location and expand_location_to_spelling_point.
1798 2018-11-30 Alexander Monakov <amonakov@ispras.ru>
1800 PR gcov-profile/88279
1801 * tree-profile.c (init_ic_make_global_vars): Call
1802 decl_default_tls_model for variable and not it's type.
1804 2018-11-30 Sam Tebbs <sam.tebbs@arm.com>
1806 * config/arm/arm.h (TARGET_DOTPROD): Add TARGET_VFP5 constraint.
1808 2018-11-30 Martin Liska <mliska@suse.cz>
1811 * asan.c (asan_shadow_cst): Remove, partially transform
1812 into flush_redzone_payload.
1813 (RZ_BUFFER_SIZE): New.
1814 (struct asan_redzone_buffer): New.
1815 (asan_redzone_buffer::emit_redzone_byte): Likewise.
1816 (asan_redzone_buffer::flush_redzone_payload): Likewise.
1817 (asan_redzone_buffer::flush_if_full): Likewise.
1818 (asan_emit_stack_protection): Use asan_redzone_buffer class
1819 that is responsible for proper aligned stores and flushing
1820 of shadow memory payload.
1821 * asan.h (ASAN_MIN_RED_ZONE_SIZE): New.
1822 (asan_var_and_redzone_size): Likewise.
1823 * cfgexpand.c (expand_stack_vars): Use smaller alignment
1824 (ASAN_MIN_RED_ZONE_SIZE) in order to make shadow memory
1825 for automatic variables more compact.
1827 2018-11-30 Alan Modra <amodra@gmail.com>
1829 * config/rs6000/predicates.md (easy_fp_constant): Avoid long
1830 dependent insn sequences.
1831 * config/rs6000/rs6000.c (num_insns_constant): Support long
1833 * config/rs6000/rs6000.md (mov<mode>_softfloat <FMOVE128>): Adjust
1836 2018-11-30 Alan Modra <amodra@gmail.com>
1838 * config/rs6000/rs6000.c (num_insns_constant_gpr): Renamed from
1839 num_insns_constant_wide. Make static. Revise comment.
1840 (num_insns_constant_multi): New function.
1841 (num_insns_constant): Formatting. Correct CONST_WIDE_INT
1842 calculation. Simplify and extract code common to both
1843 CONST_INT and CONST_DOUBLE. Add gcc_unreachable for unhandled
1845 * config/rs6000/rs6000-protos.h (num_insns_const_wide): Delete.
1847 2018-11-30 Richard Biener <rguenther@suse.de>
1849 PR tree-optimization/88274
1850 * tree-vrp.c (ranges_from_anti_range): Fix handling of
1853 2018-11-29 Andi Kleen <ak@linux.intel.com>
1855 * config/i386/i386.c (current_fentry_section): Add.
1856 (x86_function_profiler): Handle fentry section.
1857 (ix86_attribute_table): Add fentry section.
1858 * config/i386/i386.opt: Add -mfentry-section.
1859 * doc/extend.texi: Document fentry_section attribute.
1860 * doc/invoke.texi: Document -mfentry-section.
1862 2018-11-29 Andi Kleen <ak@linux.intel.com>
1864 * config/i386/i386.c (x86_print_call_or_nop): Handle nop name.
1865 (current_fentry_name): Add.
1866 (x86_function_profiler): Handle fentry_name attribute.
1867 (ix86_handle_fentry_name): Add.
1868 (ix86_attribute_table): Add fentry_name.
1869 * config/i386/i386.opt: Add -mfentry-name
1870 * doc/extend.texi: Document fentry_name.
1871 * doc/invoke.texi: Document minstrument-return.
1873 2018-11-29 Andi Kleen <ak@linux.intel.com>
1875 * config/i386/i386-opts.h (enum instrument_return): Add.
1876 * config/i386/i386.c (output_return_instrumentation): Add.
1877 (ix86_output_function_return): Call output_return_instrumentation.
1878 (ix86_output_call_insn): Call output_return_instrumentation.
1879 * config/i386/i386.opt: Add -minstrument-return=.
1880 * doc/invoke.texi (-minstrument-return): Document.
1882 2018-11-29 Eric Botcazou <ebotcazou@adacore.com>
1885 * config/sparc/sparc-modes.def: Minor tweak.
1886 * config/sparc/sparc.c: Minor reordering.
1887 (sparc_pass_by_reference): Move around.
1888 (traverse_record_type): Change offset from HOST_WIDE_INT to int.
1889 (classify_registers): Likewise for bitpos.
1890 (function_arg_slotno): Remove dead test and tweak comments.
1891 <MODE_RANDOM>: Remove useless assertion and test whether the
1892 parameter is named in order to pass it in FP registers. Return
1893 the regno for floating-point vector types.
1894 (compute_int_layout): Change bitpos from HOST_WIDE_INT to int.
1895 (compute_fp_layout): Likewise.
1896 (count_registers): Likewise.
1897 (assign_int_registers): Likewise.
1898 (assign_fp_registers): Likewise.
1899 (assign_registers): Likewise.
1900 (function_arg_record_value): Change size from HOST_WIDE_INT to int
1901 and use CEIL_NWORDS to compute the number of registers.
1902 (function_arg_union_value): Minor tweaks.
1903 (function_arg_vector_value): Add slotno and named parameters, use
1904 CEIL_NWORDS to compute the number of registers.
1905 (sparc_function_arg_1): Rework handling of vector types. Change
1906 size from HOST_WIDE_INT to int.
1907 (sparc_arg_partial_bytes): Rework handling of 32-bit ABI and deal
1908 with vector types for the 64-bt ABI.
1909 (sparc_function_arg_advance): Likewise.
1910 (sparc_return_in_memory): Add reference to -fpcc-struct-return.
1911 (sparc_struct_value_rtx): Return NULL_RTX instead of 0.
1912 (sparc_function_value_1): Rework handling of vector types. Change
1913 size from HOST_WIDE_INT to int.
1915 2018-11-29 Jakub Jelinek <jakub@redhat.com>
1918 * config/i386/sse.md (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_lt,
1919 *<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt,
1920 *<sse>_movmsk<ssemodesuffix><avxsizesuffix>_shift,
1921 *<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift,
1922 *<sse2_avx2>_pmovmskb_lt, *<sse2_avx2>_pmovmskb_zext_lt): New
1923 define_insn_and_split patterns.
1926 * config/i386/sse.md
1927 (*<sse4_1>_blendv<ssemodesuffix><avxsizesuffix>_lt,
1928 *<sse4_1>_blendv<ssefltmodesuffix><avxsizesuffix>_ltint,
1929 *<sse4_1_avx2>_pblendvb_lt): Change define_insn into
1930 define_insn_and_split.
1932 2018-11-29 Peter Bergner <bergner@linux.ibm.com>
1935 * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
1936 -mabi=ieeelongdouble without both -mpopcntd and -mvsx.
1938 2018-11-29 Uros Bizjak <ubizjak@gmail.com>
1940 * config/i386/i386.c (inline_memory_move_cost):
1941 Check "in" for 2 in MMX_CLASS_P case.
1942 * config/i386/mmx.md (*mov<mode>_internal): Correct
1943 TARGET_INTER_UNIT_MOVES_FROM_VEC and TARGET_INTER_UNIT_MOVES_TO_VEC
1944 alternatives in preferred_for_speed attribute calculation.
1946 2018-11-29 Martin Sebor <msebor@redhat.com>
1950 * doc/extend.texi (attribute constructor): Clarify.
1952 2018-11-29 Martin Liska <mliska@suse.cz>
1955 * builtins.c (expand_movstr): Fix thinko introduced
1956 when switching to the new enum.
1958 2018-11-29 qing zhao <qing.zhao@oracle.com>
1960 * cif-code.def (EXTERN_LIVE_ONLY_STATIC): New CIF code.
1961 * common.opt: Add -flive-patching flag.
1962 * doc/invoke.texi: Document -flive-patching.
1963 * flag-types.h (enum live_patching_level): New enum.
1964 * ipa-inline.c (can_inline_edge_p): Disable external functions from
1965 inlining when flag_live_patching is LIVE_PATCHING_INLINE_ONLY_STATIC.
1966 * opts.c (control_options_for_live_patching): New function.
1967 (finish_options): Make flag_live_patching incompatible with flag_lto.
1968 Control IPA optimizations based on different levels of
1971 2018-11-29 Giuliano Belinassi <giuliano.belinassi@usp.br>
1973 * match.pd (sinh (atanh (x))): New simplification rules.
1974 (cosh (atanh (x))): Likewise.
1976 2018-11-29 Jakub Jelinek <jakub@redhat.com>
1979 * config/i386/sse.md (ssebytemode): Add V16SI, V8SI and V4SI entries.
1980 (ssefltmodesuffix, ssefltvecmode): New define_mode_attrs.
1981 (*<sse4_1>_blendv<ssemodesuffix><avxsizesuffix>_lt,
1982 *<sse4_1>_blendv<ssefltmodesuffix><avxsizesuffix>_ltint,
1983 *<sse4_1_avx2>_pblendvb_lt): New define_insns.
1986 * tree.h (build_uniform_cst, uniform_integer_cst_p): Declare.
1987 * tree.c (build_uniform_cst, uniform_integer_cst_p): New functions.
1988 * match.pd (define_predicates): Add uniform_integer_cst_p.
1989 (cmp @0 INTEGER_CST@1, cmp (convert?@2 @0) INTEGER_CST@1): Adjust
1990 so that it works also for vector comparisons with uniform constants
1991 with INTEGER_CST element.
1994 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): For
1995 vec_add and vec_sub builtins, perform PLUS_EXPR or MINUS_EXPR
1996 in unsigned_type_for instead of vector integral type where overflow
1999 2018-11-29 Michael Ploujnikov <michael.ploujnikov@oracle.com>
2001 There can be at most one .resolver clone per function
2002 * config/rs6000/rs6000.c (make_resolver_func): Generate
2003 resolver symbol with clone_function_name instead of
2004 clone_function_name_numbered.
2006 2018-11-29 Richard Biener <rguenther@suse.de>
2008 PR tree-optimization/88243
2009 * tree-vect-patterns.c (vect_mark_pattern_stmts): Set the def
2010 type of all pattern-sequence stmts to vect_internal_def.
2012 2018-11-29 Andre Vieira <andre.simoesdiasvieira@arm.com>
2015 * config/arm/arm-cpus.in (armv7-r): Add FP16conv configurations.
2016 (cortex-r7, cortex-r8): Update default and add new configuration.
2017 * doc/invoke.texi (armv7-r): Add two new vfp options.
2018 (nofp.dp): Add cortex-r7 and cortex-r8 to the list of targets that
2019 support this option.
2021 2018-11-29 Alan Modra <amodra@gmail.com>
2023 * config/rs6000/rs6000.c (rs6000_emit_move): Disable long
2024 double split for targets other than Darwin.
2026 2018-11-29 Alan Modra <amodra@gmail.com>
2028 * config/rs6000/rs6000.md (movdi_internal64): Correct lengths.
2030 2018-11-29 Alan Modra <amodra@gmail.com>
2032 * config/rs6000/predicates.md (easy_fp_constant): Remove code
2033 dealing with integers in const_double. Assert on unexpected
2034 modes. Delete superfluous ABI_V4 test.
2035 * config/rs6000/rs6000.c (rs6000_emit_move): Comment fixes.
2036 Don't call easy_fp_constant for Pmode.
2037 (rs6000_hash_constant): Remove code dealing with integers in
2039 (rs6000_legitimate_constant_p): Likewise.
2040 (output_toc): Formatting, use CONST_DOUBLE_P.
2042 2018-11-29 Alan Modra <amodra@gmail.com>
2044 * config/rs6000/constraints.md (G, H): Comment on purpose of
2045 constraint. Correct mode comments and "G" mode test.
2046 * config/rs6000/rs6000.md (movdi_internal32): Remove "GH" from
2047 alternative handling "F". Add length attr. Formatting.
2048 (mov<mode>_softfloat <FMOVE128>): Delete "GH" from
2049 alternative, and move "F" to separate alternative. Correct
2052 2018-11-29 Alan Modra <amodra@gmail.com>
2054 * config/rs6000/predicates.md (easy_vector_constant): Don't call
2057 2018-11-29 Alan Modra <amodra@gmail.com>
2059 * config.in (HAVE_AS_PLTSEQ): Add.
2060 * config/rs6000/predicates.md (indirect_call_operand): New.
2061 * config/rs6000/rs6000-protos.h (rs6000_pltseq_template),
2062 (rs6000_sibcall_sysv): Declare.
2063 * config/rs6000/rs6000.c (init_cumulative_args): Set cookie
2064 CALL_LONG for -fno-plt.
2065 (print_operand <T, z, 0>): Handle UNSPEC_PLTSEQ.
2066 (rs6000_indirect_call_template_1): Emit .reloc directives for
2067 UNSPEC_PLTSEQ calls.
2068 (rs6000_pltseq_template): New function.
2069 (rs6000_longcall_ref): Add arg parameter. Use PLT16 insns if
2070 relocs supported by assembler. Move SYMBOL_REF test to callers.
2071 (rs6000_call_aix): Adjust rs6000_longcall_ref call. Package
2072 insns in UNSPEC_PLTSEQ, preserving original func_desc.
2073 (rs6000_call_sysv): Likewise.
2074 (rs6000_sibcall_sysv): New function.
2075 * config/rs6000/rs6000.h (HAVE_AS_PLTSEQ): Provide default.
2076 * config/rs6000/rs6000.md (UNSPEC_PLTSEQ, UNSPEC_PLT16_HA,
2077 UNSPEC_PLT16_LO): New.
2078 (pltseq_tocsave, pltseq_plt16_ha, pltseq_plt16_lo, pltseq_mtctr): New.
2079 (call_indirect_nonlocal_sysv): Don't differentiate zero from non-zero
2080 cookie in constraints. Test explicitly for flags in length attr.
2081 Handle unspec operand 1.
2082 (call_value_indirect_nonlocal_sysv): Likewise.
2083 (call_indirect_aix, call_value_indirect_aix): Handle unspec operand 1.
2084 (call_indirect_elfv2, call_value_indirect_elfv2): Likewise.
2085 (sibcall, sibcall_value): Use rs6000_sibcall_sysv.
2086 (sibcall_indirect_nonlocal_sysv): New pattern.
2087 (sibcall_value_indirect_nonlocal_sysv): Likewise.
2088 (sibcall_nonlocal_sysv, sibcall_value_nonlocal_sysv): Remove indirect
2090 * configure.ac: Check for gas plt sequence marker support.
2091 * configure: Regenerate.
2093 2018-11-29 Alan Modra <amodra@gmail.com>
2095 * config/rs6000/predicates.md (unspec_tls): New.
2096 * config/rs6000/rs6000-protos.h (rs6000_call_template),
2097 (rs6000_sibcall_template): Update prototype.
2098 (rs6000_longcall_ref): Delete.
2099 (rs6000_call_sysv): Declare.
2100 * config/rs6000/rs6000.c (edit_tls_call_insn): New function.
2101 (global_tlsarg): New variable.
2102 (rs6000_legitimize_tls_address): Rewrite __tls_get_addr call
2104 (print_operand): Extract UNSPEC_TLSGD address operand.
2105 (rs6000_call_template, rs6000_sibcall_template): Remove arg
2106 parameter, extract from second call operand instead.
2107 (rs6000_longcall_ref): Make static, localize vars.
2108 (rs6000_call_aix): Rename parameter to reflect new usage. Take
2109 tlsarg from global_tlsarg. Don't create unused rtl or nop insns.
2110 (rs6000_sibcall_aix): Rename parameter to reflect new usage. Take
2111 tlsarg from global_tlsarg.
2112 (rs6000_call_sysv): New function.
2113 * config/rs6000/rs6000.md: Adjust rs6000_call_template and
2114 rs6000_sibcall_template throughout.
2115 (tls_gd_aix, tls_gd_sysv, tls_gd_call_aix, tls_gd_call_sysv): Delete.
2116 (tls_ld_aix, tls_ld_sysv, tls_ld_call_aix, tls_ld_call_sysv): Delete.
2117 (tls_gdld_nomark): New insn.
2118 (tls_gd): Swap operand order. Simplify mode selection.
2119 (tls_gd_high, tls_gd_low): Swap operand order.
2120 (tls_ld): Remove const_int 0 vector element from UNSPEC_TLSLD.
2121 Simplify mode selection.
2122 (tls_ld_high, tls_ld_low): Similarly adjust UNSPEC_TLSLD.
2123 (call, call_value): Don't assert for second call operand.
2124 Use rs6000_call_sysv.
2126 2018-11-29 Alan Modra <amodra@gmail.com>
2128 * config/rs6000/darwin.md (call_indirect_nonlocal_darwin64),
2129 (call_nonlocal_darwin64, call_value_indirect_nonlocal_darwin64),
2130 (call_value_nonlocal_darwin64): Remove constraints from second call
2131 arg, the rounded_stack_size_rtx arg.
2132 * config/rs6000/rs6000.md (tls_gd_aix, tls_gd_sysv, tls_gd_call_aix),
2133 (tls_gd_call_sysv, tls_ld_aix, tls_ld_sysv, tls_ld_call_aix),
2134 (tls_ld_call_sysv, call_local32, call_local64, call_value_local32),
2135 (call_value_local64, call_indirect_nonlocal_sysv),
2136 (call_nonlocal_sysv, call_nonlocal_sysv_secure),
2137 (call_value_indirect_nonlocal_sysv, call_value_nonlocal_sysv),
2138 (call_value_nonlocal_sysv_secure, call_local_aix),
2139 (call_value_local_aix, call_nonlocal_aix, call_value_nonlocal_aix),
2140 (call_indirect_aix, call_value_indirect_aix, call_indirect_elfv2),
2141 (call_value_indirect_elfv2, sibcall_local32, sibcall_local64),
2142 (sibcall_value_local32, sibcall_value_local64, sibcall_aix),
2143 (sibcall_value_aix): Likewise.
2145 2018-11-29 Alan Modra <amodra@gmail.com>
2147 * config/rs6000/rs6000.md (TLSmode): Delete mode iterator. Replace
2148 with P throughout except for call mems which should use SI.
2149 (tls_abi_suffix, tls_sysv_suffix, tls_insn_suffix): Delete mode
2150 attributes. Replace with bits, mode and ptrload respectively.
2152 2018-11-29 Alan Modra <amodra@gmail.com>
2154 * config/rs6000/rs6000-protos.h (rs6000_indirect_call_template),
2155 (rs6000_indirect_sibcall_template): Declare.
2156 * config/rs6000/rs6000.c (rs6000_indirect_call_template_1),
2157 (rs6000_indirect_call_template, rs6000_indirect_sibcall_template):
2159 * config/rs6000/rs6000.md (call_indirect_nonlocal_sysv),
2160 (call_value_indirect_nonlocal_sysv, sibcall_nonlocal_sysv),
2161 (call_indirect_aix, call_value_indirect_aix): Use
2162 rs6000_indirect_call_template and rs6000_indirect_sibcall_template.
2163 call_indirect_elfv2, call_value_indirect_elfv2): Likewise, and
2164 handle both speculation and non-speculation cases.
2165 (call_indirect_aix_nospec, call_value_indirect_aix_nospec): Delete.
2166 (call_indirect_elfv2_nospec, call_value_indirect_elfv2_nospec): Delete.
2168 2018-11-29 Alan Modra <amodra@gmail.com>
2170 * config/rs6000/rs6000-protos.h (rs6000_call_template): Declare.
2171 (rs6000_sibcall_template): Declare.
2172 (macho_call_template): Rename from output_call.
2173 * config/rs6000/rs6000.c (rs6000_call_template_1): New function.
2174 (rs6000_call_template, rs6000_sibcall_template): Likewise.
2175 (macho_call_template): Rename from output_call.
2176 * config/rs6000/rs6000.md (tls_gd_aix, tls_gd_sysv),
2177 (tls_gd_call_aix, tls_gd_call_sysv, tls_ld_aix, tls_ld_sysv),
2178 (tls_ld_call_aix, tls_ld_call_sysv, call_nonlocal_sysv),
2179 (call_nonlocal_sysv_secure, call_value_nonlocal_sysv),
2180 (call_value_nonlocal_sysv_secure, call_nonlocal_aix),
2181 (call_value_nonlocal_aix): Use rs6000_call_template and update
2182 occurrences of output_call to macho_call_template.
2183 (sibcall_nonlocal_sysv, sibcall_value_nonlocal_sysv, sibcall_aix),
2184 (sibcall_value_aix): Use rs6000_sibcall_template.
2186 2018-11-28 Aaron Sawdey <acsawdey@linux.ibm.com>
2188 * config/rs6000/rs6000-string.c (expand_block_clear): Change how
2189 we determine if unaligned vsx is ok.
2191 2018-11-28 Jan Hubicka <jh@suse.cz>
2193 * predict.c (determine_unlikely_bbs): Forward declare; also determine
2194 edges that are always known to be taken because it is only likely
2195 edge out of given BB.
2196 (tree_estimate_probability): Call before profile guessing to get bit
2199 2018-11-28 Jan Hubicka <jh@suse.cz>
2201 * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Handle
2202 profile_probability::always better.
2204 2018-11-28 Jan Hubicka <jh@suse.cz>
2206 * profile-count.h (profile_count::split): Give better result when
2207 splitting profile_probability::always.
2209 2018-11-28 Vladimir Makarov <vmakarov@redhat.com>
2212 * ira-costs.c (scan_one_insn): Process subregs when updating costs
2213 for pseudos and allocnos from insn.
2215 2018-11-28 David Edelsohn <dje.gcc@gmail.com>
2217 * config/rs6000/aix72.h: Update to match aix71.h changes.
2219 2018-11-28 Jeff Law <law@redhat.com>
2221 * config/h8300/constraints.md: Add "C" constraint for call insns.
2222 * config/h8300/h8300.md (call, call_value): Turn into a define_expand
2223 and define_insn pair. Move invalid call targets into a register in
2224 the expander and fix constraints in the matching pattern.
2225 * config/h8300/predicates.md (call_expander_operand): Renamed from
2226 call_insn_operand. Reject things we shouldn't be trying to handle.
2227 (call_insn_operand): New predicate for use by the call/call_value
2229 (small_call_insn_operand): Update appropriately.
2231 2018-11-28 Sam Tebbs <sam.tebbs@arm.com>
2233 * config/aarch64/aarch64.c (aarch64_process_target_attr): Replace
2234 calls to strtok with strtok_r.
2236 2018-11-28 Richard Biener <rguenther@suse.de>
2238 PR tree-optimization/88223
2239 * tree-ssa-sccvn.c (vn_reference_lookup_3): When skipping
2240 over a stored-same value may-alias store make sure to consider
2241 partial overlaps which are valid when TBAA reasonings do not
2242 apply and byte-granular overlaps are possible at all.
2244 2018-11-28 Richard Biener <rguenther@suse.de>
2246 PR tree-optimization/88217
2247 * vr-values.c (vr_values::extract_range_from_phi_node): Make
2248 sure to handle results > +INF and < -INF correctly when
2249 trying to drop down to +INF - 1 or -INF + 1.
2251 2018-11-28 Alan Modra <amodra@gmail.com>
2253 * xcoffout.c (do_block): Signed/unsigned warning fix.
2255 2018-11-28 Richard Biener <rguenther@suse.de>
2257 PR tree-optimization/79351
2258 * tree-ssa-sccvn.c (vn_reference_lookup_3): For assignments from
2259 empty CONSTRUCTORs ensure the store is at a constant position.
2261 2018-11-28 Richard Biener <rguenther@suse.de>
2263 PR tree-optimization/88229
2264 * tree-ssa.c (non_rewritable_mem_ref_base): Check DECL_SIZE_UNIT
2265 is an INTEGER_CST before accessing it so.
2267 2018-11-28 Sam Tebbs <sam.tebbs@arm.com>
2269 * doc/invoke.texi (-mtune=): Remove obsolete CPU names.
2271 2018-11-28 Jakub Jelinek <jakub@redhat.com>
2274 * config/i386/i386.c (ix86_expand_sse_movcc): Handle DFmode and
2275 SFmode using sse4_1_blendvs[sd] with TARGET_SSE4_1. Formatting fixes.
2276 * config/i386/sse.md (sse4_1_blendv<ssemodesuffix>): New pattern.
2278 2018-11-27 Martin Liska <mliska@suse.cz>
2280 * gcov.c (generate_results): Append current_working_directory
2283 2018-11-27 Mike Gulick <mgulick@mathworks.com>
2285 PR preprocessor/83173
2286 * input.c (dump_location_info): Dump reason and included_from
2287 fields from line_map_ordinary struct. Fix indentation when
2288 location > 5 digits.
2289 * diagnostic-show-locus.c (num_digits, num_digits): Move to
2290 diagnostic.c to allow it to be utilized by input.c.
2291 * diagnostic.c (num_digits, selftest::test_num_digits): Moved
2293 (selftest::diagnostic_c_tests): Run selftest::test_num_digits.
2294 * diagnostic.h (num_digits): Add extern definition.
2296 2018-11-27 Fredrik Noring <noring@nocrew.org>
2298 * config/mips/mips.c (mips_reorg_process_insns)
2299 (mips_option_override): Handle `-mfix-r5900'.
2300 * config/mips/mips.h (ASM_SPEC): Add `mfix-r5900' and
2302 * config/mips/mips.opt (mfix-r5900): New option.
2303 * doc/invoke.texi: Document the `r5900' processor name, and
2304 `-mfix-r5900' and `-mno-fix-r5900' options.
2306 2018-11-27 Jakub Jelinek <jakub@redhat.com>
2309 * config/rs6000/rs6000.c (print_operand) <case 'D'>: Use
2310 output_operand_lossage instead of gcc_assert.
2311 <case 't'>: Likewise.
2312 <case 'z'>: Likewise.
2313 <case 'V'>: Use output_operand_lossage instead of gcc_unreachable.
2315 2018-11-27 Jeff Law <law@redhat.com>
2317 * config/riscv/riscv (riscv_block_mvoe_straight): Use RETURN_BEGIN
2318 in call to move_by_pieces.
2320 * config/sh/sh-mem.c (expand_block_move): Use RETURN_BEGIN in call
2323 * config/lm32/lm32.c (lm32_block_move_inline): Use RETURN_BEGIN in
2324 call to move_by_pieces.
2326 * config/mips/mips.c (mips_block_move_straight): Use RETURN_BEGIN
2327 in call to move_by_pieces.
2329 * config/microblaze/microblaze.c (microblaze_block_move_straight): Use
2330 RETURN_BEGIN in call to move_by_pieces.
2331 (microblaze_expand_block_move): Likewise.
2333 2018-11-27 Tamar Christina <tamar.christina@arm.com>
2335 * doc/md.texi (xorsign): Document it.
2337 2018-11-27 Martin Liska <mliska@suse.cz>
2339 * config/gnu-user.h (TARGET_F951_OPTIONS): New.
2340 * gcc.c (find_fortran_preinclude_file): New function
2341 to handle Fortran pre-include.
2343 2018-11-27 Martin Liska <mliska@suse.cz>
2345 * asan.c (asan_emit_stack_protection): Use new enum values
2346 instead of int constants.
2347 * builtins.c (expand_builtin_memory_copy_args): Replace int
2348 type with memop_ret enum type.
2349 (expand_builtin_mempcpy_args): Likewise.
2350 (expand_builtin_memcpy): Use new enum values
2351 instead of int constants. Likewise.
2352 (expand_builtin_mempcpy): Likewise.
2353 (expand_movstr): Likewise.
2354 (expand_builtin_strcpy_args): Likewise.
2355 (expand_builtin_stpcpy_1): Likewise.
2356 (expand_builtin_strncpy): Likewise.
2357 (expand_builtin_memset_args): Likewise.
2358 * expr.c (move_by_pieces_d::finish_endp): Rename to ...
2359 (move_by_pieces_d::finish_retmode): ... this.
2360 (move_by_pieces): Change last argument type to memop_ret.
2361 (store_by_pieces): Use new enum values
2362 instead of int constants.
2363 (emit_block_move_hints): Likewise.
2364 (emit_push_insn): Likewise.
2365 (store_expr): Likewise.
2366 * expr.h (store_by_pieces): Change int to newly added enum
2368 * rtl.h (enum memop_ret): Define.
2369 (move_by_pieces): Use the enum type.
2371 2018-11-27 Alan Modra <amodra@gmail.com>
2373 * config/rs6000/aix71.h (ASM_SPEC): Don't select default -maix64
2375 (ASM_CPU_SPEC): Do so here. Rewrite using if .. else if .. specs
2376 form. Error on missing -mcpu case.
2377 * config/rs6000/driver-rs6000.c (asm_names <_AIX>): Update NULL case.
2378 (asm_names <!_AIX>): Add missing cpus. Update NULL case. Apply
2379 PR63177 fix for -mcpu=power8 and -mcpu=powerpc64le.
2380 * config/rs6000/rs6000.h (ASM_CPU_SPEC): Rewrite using if ..
2381 else if .. specs form. Error on missing -mcpu case. Don't output
2382 duplicate -maltivec. Apply PR63177 fix for -mcpu=powerpc64le.
2384 2018-11-26 David Malcolm <dmalcolm@redhat.com>
2386 * dump-context.h (dump_context::dump_loc): Convert 1st param from
2387 dump_flags_t to const dump_metadata_t &. Convert 2nd param from
2388 const dump_location_t & to const dump_user_location_t &.
2389 (dump_context::dump_loc_immediate): Convert 2nd param from
2390 const dump_location_t & to const dump_user_location_t &.
2391 (dump_context::dump_gimple_stmt): Convert 1st param from
2392 dump_flags_t to const dump_metadata_t &.
2393 (dump_context::void dump_gimple_stmt_loc): Likewise; convert
2394 2nd param from const dump_location_t & to
2395 const dump_user_location_t &.
2396 (dump_context::dump_gimple_expr): Convert 1st param from
2397 dump_flags_t to const dump_metadata_t &.
2398 (dump_context::dump_gimple_expr_loc): Likewise; convert
2399 2nd param from const dump_location_t & to
2400 const dump_user_location_t &.
2401 (dump_context::dump_generic_expr): Convert 1st param from
2402 dump_flags_t to const dump_metadata_t &.
2403 (dump_context::dump_generic_expr_loc): Likewise; convert
2404 2nd param from const dump_location_t & to
2405 const dump_user_location_t &.
2406 (dump_context::dump_printf_va): Convert 1st param from
2407 dump_flags_t to const dump_metadata_t &.
2408 (dump_context::dump_printf_loc_va): Likewise; convert
2409 2nd param from const dump_location_t & to
2410 const dump_user_location_t &.
2411 (dump_context::dump_dec): Convert 1st param from
2412 dump_flags_t to const dump_metadata_t &.
2413 (dump_context::dump_symtab_node): Likewise.
2414 (dump_context::begin_scope): Split out 2nd param into
2415 user and impl locations.
2416 (dump_context::ensure_pending_optinfo): Add metadata param.
2417 (dump_context::begin_next_optinfo): Replace dump_location_t param
2418 with metadata and user location.
2419 * dumpfile.c (dump_context::dump_loc): Convert 1st param from
2420 dump_flags_t to const dump_metadata_t &. Convert 2nd param from
2421 const dump_location_t & to const dump_user_location_t &.
2422 (dump_context::dump_loc_immediate): Convert 2nd param from
2423 const dump_location_t & to const dump_user_location_t &.
2424 (dump_context::dump_gimple_stmt): Convert 1st param from
2425 dump_flags_t to const dump_metadata_t &.
2426 (dump_context::void dump_gimple_stmt_loc): Likewise; convert
2427 2nd param from const dump_location_t & to
2428 const dump_user_location_t &.
2429 (dump_context::dump_gimple_expr): Convert 1st param from
2430 dump_flags_t to const dump_metadata_t &.
2431 (dump_context::dump_gimple_expr_loc): Likewise; convert
2432 2nd param from const dump_location_t & to
2433 const dump_user_location_t &.
2434 (dump_context::dump_generic_expr): Convert 1st param from
2435 dump_flags_t to const dump_metadata_t &.
2436 (dump_context::dump_generic_expr_loc): Likewise; convert
2437 2nd param from const dump_location_t & to
2438 const dump_user_location_t &.
2439 (dump_context::dump_printf_va): Convert 1st param from
2440 dump_flags_t to const dump_metadata_t &.
2441 (dump_context::dump_printf_loc_va): Likewise; convert
2442 2nd param from const dump_location_t & to
2443 const dump_user_location_t &.
2444 (dump_context::dump_dec): Convert 1st param from
2445 dump_flags_t to const dump_metadata_t &.
2446 (dump_context::dump_symtab_node): Likewise.
2447 (dump_context::begin_scope): Split out 2nd param into
2448 user and impl locations.
2449 (dump_context::ensure_pending_optinfo): Add metadata param.
2450 (dump_context::begin_next_optinfo): Replace dump_location_t param
2451 with metadata and user location.
2452 (dump_gimple_stmt): Convert 1st param from dump_flags_t to
2453 const dump_metadata_t &.
2454 (dump_gimple_stmt_loc): Likewise; convert 2nd param from
2455 const dump_location_t & to const dump_user_location_t &.
2456 (dump_gimple_expr): Convert 1st param from dump_flags_t to
2457 const dump_metadata_t &.
2458 (dump_gimple_expr_loc): Likewise; convert 2nd param from
2459 const dump_location_t & to const dump_user_location_t &.
2460 (dump_generic_expr): Convert 1st param from dump_flags_t to
2461 const dump_metadata_t &.
2462 (dump_generic_expr_loc): Likewise; convert 2nd param from
2463 const dump_location_t & to const dump_user_location_t &.
2464 (dump_printf): Convert 1st param from dump_flags_t to
2465 const dump_metadata_t &.
2466 (dump_printf_loc): Likewise; convert 2nd param from
2467 const dump_location_t & to const dump_user_location_t &.
2468 (dump_dec): Convert 1st param from dump_flags_t to
2469 const dump_metadata_t &.
2470 (dump_symtab_node): Likewise.
2471 (dump_begin_scope): Split out 2nd param into user and impl
2473 (selftest::assert_impl_location_eq): New function.
2474 (ASSERT_IMPL_LOCATION_EQ): New macro.
2475 (selftest::test_impl_location): Update to use
2476 ASSERT_IMPL_LOCATION_EQ.
2477 (selftest::test_capture_of_dump_calls): Convert "loc" to
2478 dump_user_location_t. Add ASSERT_IMPL_LOCATION_EQ throughout,
2479 verifying line numbers of dump emissions.
2480 * dumpfile.h (class dump_metadata_t): New class.
2481 (dump_printf): Convert 1st param from dump_flags_t to
2482 const dump_metadata_t &.
2483 (dump_printf_loc): Likewise; convert 2nd param from
2484 const dump_location_t & to const dump_user_location_t &.
2485 (dump_generic_expr_loc): Likewise.
2486 (dump_generic_expr): Convert 1st param from dump_flags_t to
2487 const dump_metadata_t &.
2488 (dump_gimple_stmt_loc): Likewise; convert 2nd param from
2489 const dump_location_t & to const dump_user_location_t &.
2490 (dump_gimple_stmt): Convert 1st param from dump_flags_t to
2491 const dump_metadata_t &.
2492 (dump_gimple_expr_loc): Likewise; convert 2nd param from
2493 const dump_location_t & to const dump_user_location_t &.
2494 (dump_gimple_expr): Convert 1st param from dump_flags_t to
2495 const dump_metadata_t &.
2496 (dump_symtab_node): Likewise.
2497 (dump_dec): Likewise.
2498 (dump_begin_scope): Split out 2nd param into user and impl
2500 (auto_dump_scope::auto_dump_scope): Split "loc" param into a user
2501 location and impl_location, and capture the impl_location.
2502 (AUTO_DUMP_SCOPE): Rename param from LOC to USER_LOC.
2503 * loop-unroll.c (report_unroll): Update for changes to
2504 dump_printf_loc and dump_printf.
2505 * opt-problem.cc (opt_problem::opt_problem): Update for change to
2507 * optinfo-emit-json.cc
2508 (selftest::test_building_json_from_dump_calls): Convert "loc" from
2509 dump_location_t to dump_user_location_t.
2510 * optinfo.cc (optinfo::emit_for_opt_problem): Update for change in
2512 * profile.c (compute_branch_probabilities): Update for change to
2514 * selftest.h (ASSERT_STR_CONTAINS_AT): New macro.
2515 * tree-vect-slp.c (vect_print_slp_tree): Update for change to
2518 2018-11-27 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2520 * doc/sourcebuild.texi: Document check_effective_target_int_eq_float
2521 and check_effective_target_ptr_eq_long.
2523 2018-11-27 Alan Modra <amodra@gmail.com>
2525 * config.gcc (powerpc*-*-freebsd*, powerpc-*-netbsd*),
2526 (powerpc-*-eabisimaltivec*, powerpc-*-eabisim*, powerpc-*-elf*),
2527 (powerpc-*-eabialtivec*, powerpc-*-eabi*, powerpc-*-rtems*),
2528 (powerpc-wrs-vxworks*, powerpc-*-lynxos*, powerpcle-*-elf*),
2529 (powerpcle-*-eabisim*, powerpcle-*-eabi*): Add gnu-user.h to tm_file.
2530 * config/rs6000/freebsd.h (CPLUSPLUS_CPP_SPEC),
2531 (LINK_GCC_C_SEQUENCE_SPEC): Undef.
2532 (ASM_APP_ON, ASM_APP_OFF): Don't define.
2533 * config/rs6000/freebsd64.h (ASM_APP_ON, ASM_APP_OFF): Don't define.
2534 * config/rs6000/lynx.h (ASM_APP_ON, ASM_APP_OFF): Don't define.
2535 * config/rs6000/linux64.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
2536 * config/rs6000/netbsd.h (CPLUSPLUS_CPP_SPEC),
2537 (LINK_GCC_C_SEQUENCE_SPEC): Undef.
2538 * config/rs6000/rtems.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
2539 * config/rs6000/sysv4.h (GNU_USER_TARGET_CRTI): Redefine.
2540 (GNU_USER_TARGET_CRTN): Redefine.
2541 (CC1_SPEC): Use GNU_USER_TARGET_CC1_SPEC.
2542 (LIB_LINUX_SPEC): Use GNU_USER_TARGET_LIB_SPEC.
2543 (CRTOFFLOADBEGIN, CRTOFFLOADEND): Don't define.
2544 (STARTFILE_LINUX_SPEC): Define as GNU_USER_TARGET_STARTFILE_SPEC.
2545 (ENDFILE_LINUX_SPEC): Define as GNU_USER_TARGET_ENDFILE_SPEC.
2546 (UCLIBC_DYNAMIC_LINKER, CHOOSE_DYNAMIC_LINKER): Don't define.
2547 (LINK_EH_SPEC): Don't define.
2549 2018-11-26 Martin Sebor <msebor@redhat.com>
2552 * expr.c (string_constant): Handle top-level decls of all character
2553 types and subobjects of narrow character type.
2555 2018-11-27 Alan Modra <amodra@gmail.com>
2557 * config.gcc (powerpc*-*-linux*): Add linux.h to tm_file.
2558 * config/rs6000/linux.h (TARGET_OS_CPP_BUILTINS): Use
2559 GNU_USER_TARGET_OS_CPP_BUILTINS.
2560 (RS6000_ABI_NAME): Define.
2561 * config/rs6000/linux64.h (TARGET_OS_CPP_BUILTINS): Use
2562 GNU_USER_TARGET_OS_CPP_BUILTINS.
2563 (MUSL_DYNAMIC_LINKER32): Undef before defining.
2564 (UCLIBC_DYNAMIC_LINKER32, UCLIBC_DYNAMIC_LINKER64): Don't define.
2565 (CHOOSE_DYNAMIC_LINKER): Don't define.
2566 (GNU_USER_DYNAMIC_LINKER32, GNU_USER_DYNAMIC_LINKER64): Don't define.
2567 * config/rs6000/sysv4.h (MUSL_DYNAMIC_LINKER): Undef before defining.
2568 (CHOOSE_DYNAMIC_LINKER, GNU_USER_DYNAMIC_LINKER): Only define when
2569 not already defined.
2570 (CPP_OS_LINUX_SPEC): Remove defines and asserts handled by
2571 TARGET_OS_CPP_BUILTINS.
2573 2018-11-26 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2575 * doc/sourcebuild.texi: Document check_effective_target_newlib_nano_io.
2577 2018-11-26 Jakub Jelinek <jakub@redhat.com>
2580 * config/i386/i386.c (def_builtin2): If tcode == VOID_FTYPE_UINT64
2581 and !TARGET_64BIT, return NULL_TREE.
2584 * dwarf2out.c (secname_for_decl): For functions with
2585 DECL_SECTION_NAME if in_cold_section_p, try to return
2586 current_function_section's name if it is a named section.
2588 2018-11-26 Maya Rashish <coypu@sdf.org>
2591 * config.host (*-*-netbsd*): Use x-netbsd and host-netbsd.o.
2592 * config/x-netbsd: New file.
2593 * config/host-netbsd: New file.
2595 2018-11-26 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2597 * doc/sourcebuild.texi: Document check_effective_target_size20plus.
2598 Clarify documentation for check_effective_target_size32plus.
2600 2018-11-26 Aaron Sawdey <acsawdey@linux.ibm.com>
2602 * config/rs6000/rs6000-string.c (expand_cmp_vec_sequence): Rename
2603 and modify expand_strncmp_vec_sequence.
2604 (emit_final_compare_vec): Rename and modify emit_final_str_compare_vec.
2605 (generate_6432_conversion): New function.
2606 (expand_block_compare): Add support for vsx.
2607 (expand_block_compare_gpr): New function.
2608 * config/rs6000/rs6000.opt (rs6000_block_compare_inline_limit): Increase
2609 default limit to 63 because of more compact vsx code.
2611 2018-11-26 Uros Bizjak <ubizjak@gmail.com>
2614 * config/i386/i386.c (dbx_register_map): Use IGNORED_DWARF_REGNUM
2615 for registers for which no debug information can be generated.
2616 Use INVALID_REGNUM for non-existent registers.
2617 (dbx64_register_map): Ditto.
2618 (svr4_dbx_register_map): Ditto.
2620 2018-11-26 Sandra Loosemore <sandra@codesourcery.com>
2623 * doc/invoke.texi (Optimize Options): Document that
2624 -freorder-functions also uses "hot" and "cold" attributes.
2626 2018-11-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2628 * config/aarch64/aarch64.c (aarch64_mangle_type): Fix typo in comment.
2630 2018-11-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2632 * config/arm/types.md (mul64): Rename to...
2633 (widen_mul64): ... This.
2634 * config/arm/arm-generic.md: Rename mul64 to widen_mul64.
2635 * config/arm/cortex-a15.md: Likewise.
2636 * config/arm/cortex-a5.md: Likewise.
2637 * config/arm/cortex-a53.md: Likewise.
2638 * config/arm/cortex-a57.md: Likewise.
2639 * config/arm/cortex-a7.md: Likewise.
2640 * config/arm/cortex-m4.md: Likewise.
2641 * config/arm/exynos-m1.md: Likewise.
2642 * config/arm/marvell-pj4.md: Likewise.
2643 * config/arm/xgene1.md: Likewise.
2645 2018-11-26 Richard Biener <rguenther@suse.de>
2647 PR tree-optimization/88182
2648 * tree-vect-loop.c (vectorizable_reduction): Pick up single
2649 correct reduc_def_info.
2650 * tree-vect-slp.c (vect_analyze_slp_instance): Set
2651 STMT_VINFO_REDUC_DEF of the first stmt.
2653 2018-11-26 Andreas Krebbel <krebbel@linux.ibm.com>
2655 * doc/invoke.texi: Document z14/arch12 -march option.
2657 2018-11-25 Sandra Loosemore <sandra@codesourcery.com>
2660 * doc/invoke.texi (Option Summary): Add -Wmissing-noreturn.
2661 (Warning Options): Likewise.
2663 2018-11-25 Sandra Loosemore <sandra@codesourcery.com>
2666 * doc/extend.texi (Common Function Attributes): Clarify that
2667 functions with "const" attribute can read const global variables.
2669 2018-11-26 Alan Modra <amodra@gmail.com>
2671 * config/rs6000/rs6000.h (OBJECT_PEF, TARGET_MACOS): Delete.
2673 2018-11-25 Sandra Loosemore <sandra@codesourcery.com>
2676 * doc/extend.texi (Common Variable Attributes): Use preferred
2677 placement of type attributes in examples, plus whitespace fixes.
2678 (Type Attributes): Clarify why placement of attributes
2679 immediately after struct/union/enum keyword is preferred.
2680 (Common Type Attributes): Use preferred placement of type
2681 attributes in examples, plus more whitespace fixes.
2683 2018-11-25 Paul Koning <ni1d@arrl.net>
2685 * config/pdp11/pdp11.h (TARGET_HAS_NO_HW_DIVIDE): Define.
2687 2018-11-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2689 * doc/sourcebuild.texi (Effective-Target Keywords, Environment
2690 attributes): Document d_runtime.
2692 2018-11-25 Vladimir Makarov <vmakarov@redhat.com>
2695 * ira-costs.c (record_operand_costs): Initiate register move cost
2698 2018-11-23 Jeff Law <law@redhat.com>
2700 PR rtl-optimization/87468
2701 * tree-ssa-threadupdate.c (create_block_for_threading): Clear
2702 EDGE_IGNORE on all outgoing edges of the duplicate block.
2704 2018-11-23 Vladimir Makarov <vmakarov@redhat.com>
2707 * ira-costs.c (record_operand_costs): Use bigger hard reg class if
2708 its mode does not fit to the original class.
2710 2018-11-23 Martin Sebor <msebor@redhat.com>
2712 PR tree-optimization/87756
2713 * expr.c (string_constant): Synthesize a string literal from
2714 the address of a constant character.
2715 * tree.c (build_string_literal): Add an argument.
2716 * tree.h (build_string_literal): Same.
2718 2018-11-23 Christoph Muellner <christoph.muellner@theobroma-systems.com>
2720 * config/aarch64/aarch64-cores.def: Define emag.
2721 * config/aarch64/aarch64-tune.md: Regenerated with emag.
2722 * config/aarch64/aarch64.c (emag_tunings): New struct.
2723 * doc/invoke.texi: Document mtune value.
2725 2018-11-23 Christoph Muellner <christoph.muellner@theobroma-systems.com>
2726 Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
2728 * config/aarch64/aarch64.c (xgene1_tunings): Optimize Xgene1 tunings
2730 * config/aarch64/aarch64.c (xgene1_tunings): Add Xgene1 specific
2732 * config/aarch64/aarch64.c (xgene1_addrcost_table): Correct the post-modify
2734 * config/arm/aarch-cost-tables.h (xgene1_extra_costs): Update the cost table
2737 2018-11-23 Richard Biener <rguenther@suse.de>
2739 PR tree-optimization/88149
2740 * tree-vect-slp.c (vect_slp_analyze_node_operations): Detect
2741 the case where there are two different def types for the
2742 same operand at different operand position in the same stmt.
2744 2018-23-11 Mihail Ionescu <mihail.ionescu@arm.com>
2746 * config/arm/arm.c (arm_expand_compare_and_swap): Simplify and call
2747 gen_atomic_compare_swap_1.
2748 (arm_evpc_neon_vuzp): Likewise gen_neon_vuzp_internal.
2749 (arm_evpc_neon_vtrn): Likewise gen_neon_vtrn_internal.
2750 (arm_evpc_neon_vext): Likewise gen_neon_vext_internal.
2751 (arm_evpc_neon_vzip): Likewise gen_neon_vzip_internal.
2752 (arm_evpc_neon_vrev): Replace the function pointer and simplify the mode
2754 * config/arm/arm.md (neon_vext<mode>),
2755 (neon_vrev64<mode>, neon_vrev32<mode>),
2756 (neon_vrev16<mode>, neon_vtrn<mode>_internal),
2757 (neon_vzip<mode>_internal, neon_vuzp<mode>_internal): Add an '@'character
2758 before the pattern name.
2759 * config/arm/sync.md:
2760 (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1),
2761 (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
2763 2018-11-23 Jakub Jelinek <jakub@redhat.com>
2765 PR tree-optimization/86614
2766 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return early
2767 if TREE_NO_WARNING is set on ref.ref.
2769 2018-11-22 Sandra Loosemore <sandra@codesourcery.com>
2770 Alan Coopersmith <alan.coopersmith@oracle.com>
2773 * doc/extend.texi (Designated Inits): Clarify handling of multiple
2774 initializers for unions.
2776 2018-11-22 Jan Hubicka <jh@suse.cz>
2779 * ipa-devirt.c (type_variants_equivalent_p): Drop warn and warned
2780 parameters; do not warn here.
2781 (odr_subtypes_equivalent_p): Likewise.
2782 (warn_odr): Fix typo.
2783 (warn_types_mismatch): Do not output confused warnings on integer types.
2784 (odr_types_equivalent_p): Update.
2786 2018-11-22 Jakub Jelinek <jakub@redhat.com>
2788 * config/i386/i386.c (ix86_option_override_internal,
2789 ix86_can_inline_p, classify_argument, construct_container,
2790 ix86_expand_prologue, ix86_expand_split_stack_prologue,
2791 ix86_expand_carry_flag_compare, expand_set_or_movmem_via_loop,
2792 expand_setmem_epilogue_via_loop, promote_duplicated_reg,
2793 ix86_expand_set_or_movmem, ix86_init_builtins_va_builtins_abi):
2796 * config/i386/i386.c (ix86_option_override_internal): For
2797 stack_protector_guard related options, use opts_set->x_ instead
2798 of global_options_set. and prefix options with opts->x_ . Move
2799 defaults for offset and reg into else block.
2801 2018-11-22 Uros Bizjak <ubizjak@gmail.com>
2803 * config/i386/i386.c (ix86_check_avx_upper_register):
2804 Return true for all SSE registers with mode bitsize > 128.
2806 2018-11-22 Vladimir Makarov <vmakarov@redhat.com>
2808 PR rtl-optimization/87718
2809 * ira-costs.c: Remove trailing white-spaces.
2810 (record_operand_costs): Add a special treatment for moves
2811 involving a hard register.
2813 2018-11-22 Uros Bizjak <ubizjak@gmail.com>
2815 * config/i386/i386.c (ix86_avx_emit_vzeroupper): Remove.
2816 (ix86_emit_mode_set) <case AVX_U128>: Emit vzeroupper here.
2818 2018-11-22 Martin Liska <mliska@suse.cz>
2820 * common/config/i386/i386-common.c (processor_names): Add
2821 static assert and add missing "znver2".
2822 (ix86_get_valid_option_values): Add checking assert for null
2823 values and add "native" value if feasible.
2824 * config/i386/i386.h: Do not declare size of processor_names.
2825 * common/config/i386/i386-common.c:
2826 * config/i386/i386.c: Add static assert for size
2827 of processor_cost_table.
2829 2018-11-22 Thomas Preud'homme <thomas.preudhomme@linaro.org>
2831 * target-insns.def (stack_protect_combined_set): Define new standard
2833 (stack_protect_combined_test): Likewise.
2834 * cfgexpand.c (stack_protect_prologue): Try new
2835 stack_protect_combined_set pattern first.
2836 * function.c (stack_protect_epilogue): Try new
2837 stack_protect_combined_test pattern first.
2838 * config/arm/arm.c (require_pic_register): Add pic_reg and compute_now
2839 parameters to control which register to use as PIC register and force
2840 reloading PIC register respectively. Insert in the stream of insns if
2842 (legitimize_pic_address): Expose above new parameters in prototype and
2843 adapt recursive calls accordingly. Use pic_reg if non null instead of
2845 (arm_load_pic_register): Add pic_reg parameter and use it if non null.
2846 (arm_legitimize_address): Adapt to new legitimize_pic_address
2848 (thumb_legitimize_address): Likewise.
2849 (arm_emit_call_insn): Adapt to require_pic_register prototype change.
2850 (arm_expand_prologue): Adapt to arm_load_pic_register prototype change.
2851 (thumb1_expand_prologue): Likewise.
2852 * config/arm/arm-protos.h (legitimize_pic_address): Adapt to prototype
2854 (arm_load_pic_register): Likewise.
2855 * config/arm/predicated.md (guard_addr_operand): New predicate.
2856 (guard_operand): New predicate.
2857 * config/arm/arm.md (movsi expander): Adapt to legitimize_pic_address
2859 (builtin_setjmp_receiver expander): Adapt to thumb1_expand_prologue
2861 (stack_protect_combined_set): New expander..
2862 (stack_protect_combined_set_insn): New insn_and_split pattern.
2863 (stack_protect_set_insn): New insn pattern.
2864 (stack_protect_combined_test): New expander.
2865 (stack_protect_combined_test_insn): New insn_and_split pattern.
2866 (arm_stack_protect_test_insn): New insn pattern.
2867 * config/arm/thumb1.md (thumb1_stack_protect_test_insn): New insn pattern.
2868 * config/arm/unspecs.md (UNSPEC_SP_SET): New unspec.
2869 (UNSPEC_SP_TEST): Likewise.
2870 * doc/md.texi (stack_protect_combined_set): Document new standard
2872 (stack_protect_set): Clarify that the operand for guard's address is
2874 (stack_protect_combined_test): Document new standard pattern name.
2875 (stack_protect_test): Clarify that the operand for guard's address is
2878 2018-11-22 Richard Biener <rguenther@suse.de>
2880 PR tree-optimization/88148
2881 * tree-ssa-loop-niter.c (simplify_replace_tree): Get optional
2882 valueization callback parameter and handle it.
2883 * tree-ssa-loop-niter.h (simplify_replace_tree): Export.
2884 * tree-ssa-sccvn.c (process_bb): Eliminate in loop niter trees.
2886 2018-11-22 Richard Biener <rguenther@suse.de>
2890 * lto-streamer-out.c (lto_is_streamable): Allow CALL_EXPRs
2891 which can appear in size expressions.
2892 * tree-streamer-in.c (unpack_ts_base_value_fields): Stream
2893 CALL_EXPR_BY_DESCRIPTOR.
2894 (streamer_read_tree_bitfields): Stream CALL_EXPR_IFN.
2895 * tree-streamer-out.c (pack_ts_base_value_fields): Stream
2896 CALL_EXPR_BY_DESCRIPTOR.
2897 (streamer_write_tree_bitfields): Stream CALL_EXPR_IFN.
2901 * tree.c (free_lang_data_in_one_sizepos): Free non-gimple-val
2904 2018-11-22 Richard Biener <rguenther@suse.de>
2906 PR tree-optimization/88069
2907 * tree-ssa-sccvn.c (visit_phi): Tweak previous fix to not
2908 apply to default defs.
2910 2018-11-22 Jakub Jelinek <jakub@redhat.com>
2914 * config/i386/i386.c (ix86_option_override_internal): Default
2915 ix86_stack_protector_guard to SSP_TLS only if TARGET_THREAD_SSP_OFFSET
2917 * config/i386/i386.md (stack_protect_set, stack_protect_set_<mode>,
2918 stack_protect_test, stack_protect_test_<mode>): Use empty condition
2919 instead of TARGET_SSP_TLS_GUARD.
2921 2018-11-22 Martin Liska <mliska@suse.cz>
2924 * config/i386/i386.c (ix86_option_override_internal): Error
2925 about usage -mabi=ms and -fsanitize=thread.
2927 2018-11-21 Uros Bizjak <ubizjak@gmail.com>
2930 2013-10-26 Vladimir Makarov <vmakarov@redhat.com>
2933 2013-10-25 Vladimir Makarov <vmakarov@redhat.com>
2935 * lra-spills.c (lra_final_code_change): Remove useless move insns.
2937 2018-11-21 Uros Bizjak <ubizjak@gmail.com>
2940 * function.c (expand_function_end): Do not emit extra blockage insn.
2942 2018-11-21 Lokesh Janghel <lokeshjanghel91@gmail.com>
2945 * config/i386/i386.c (function_value_ms_64): Return AX_REG instead
2946 of FIRST_SSE_REG for 4 or 8 byte modes.
2948 2018-11-21 Jan Hubicka <jh@suse.cz>
2951 * tree.c (fld_decl_context): Break out from ...
2952 (free_lang_data_in_decl): ... here; free TREE_PUBLIC, TREE_PRIVATE
2953 DECL_ARTIFICIAL of TYPE_DECL; do not free TREE_TYPE of TYPE_DECL.
2954 (fld_incomplete_type_of): Build copy of TYP_DECL.
2955 * ipa-devirt.c (free_enum_values): Rename to ...
2956 (free_odr_warning_data): ... this one; free also duplicated TYPE_DECLs
2957 and TREE_TYPEs of TYPE_DECLs.
2958 (get_odr_type): Initialize odr_vtable_hash if needed.
2960 2018-11-21 Alexandre Oliva <oliva@adacore.com>
2962 * final.c (compute_discriminator): Declare. Renamed from...
2963 (maybe_set_discriminator): ... this. Set and return a local.
2964 (override_discriminator): New.
2965 (final_scan_insn_1): Set it.
2966 (notice_source_line): Adjust. Always set discriminator.
2968 2018-11-21 Jakub Jelinek <jakub@redhat.com>
2971 * config/aarch64/atomics.md (@aarch64_compare_and_swap<mode>): Use
2972 rIJ constraint for aarch64_plus_operand rather than rn.
2974 2018-11-21 Renlin Li <renlin.li@arm.com>
2977 * explow.h (get_dynamic_stack_size): Declare it as external.
2978 * explow.c (record_new_stack_level): Remove function static attribute.
2979 * function.c (assign_stack_local_1): Dynamically align the stack slot
2980 addr for parameter copy on the stack.
2982 2018-11-21 Richard Biener <rguenther@suse.de>
2985 * bitmap.c (bitmap_last_set_bit): Refactor to avoid warning.
2986 * Makefile.in (bitmap.o-warn): Remove again.
2988 2018-11-20 Jeff Law <law@redhat.com>
2990 PR tree-optimization/88069
2991 * tree-ssa-dom.c (record_equivalences_from_phis): Propagate away
2992 degenerate virtual PHIs.
2994 2018-11-21 H.J. Lu <hongjiu.lu@intel.com>
2997 * config/i386/sse.md (sse4_1_<code>v8qiv8hi2<mask_name>): Replace
2998 nonimmediate_operand with register_operand.
2999 (avx2_<code>v8qiv8si2<mask_name>): Likewise.
3000 (sse4_1_<code>v4qiv4si2<mask_name>): Likewise.
3001 (sse4_1_<code>v4hiv4si2<mask_name>): Likewise.
3002 (sse4_1_<code>v2qiv2di2<mask_name>): Likewise.
3003 (avx512f_<code>v8qiv8di2<mask_name>): Likewise.
3004 (avx2_<code>v4qiv4di2<mask_name>): Likewise.
3005 (avx2_<code>v4hiv4di2<mask_name>): Likewise.
3006 (sse4_1_<code>v2hiv2di2<mask_name>): Likewise.
3007 (sse4_1_<code>v2siv2di2<mask_name>): Likewise.
3008 (*sse4_1_<code>v8qiv8hi2<mask_name>_1): New pattern.
3009 (*sse4_1_<code>v8qiv8hi2<mask_name>_2): Likewise.
3010 (*avx2_<code>v8qiv8si2<mask_name>_1): Likewise.
3011 (*avx2_<code>v8qiv8si2<mask_name>_2): Likewise.
3012 (*sse4_1_<code>v4qiv4si2<mask_name>_1): Likewise.
3013 (*sse4_1_<code>v4qiv4si2<mask_name>_2): Likewise.
3014 (*sse4_1_<code>v4hiv4si2<mask_name>_1): Likewise.
3015 (*sse4_1_<code>v4hiv4si2<mask_name>_2): Likewise.
3016 (*avx512f_<code>v8qiv8di2<mask_name>_1): Likewise.
3017 (*avx512f_<code>v8qiv8di2<mask_name>_2): Likewise.
3018 (*avx2_<code>v4qiv4di2<mask_name>_1): Likewise.
3019 (*avx2_<code>v4qiv4di2<mask_name>_2): Likewise.
3020 (*avx2_<code>v4hiv4di2<mask_name>_1): Likewise.
3021 (*avx2_<code>v4hiv4di2<mask_name>_2): Likewise.
3022 (*sse4_1_<code>v2hiv2di2<mask_name>_1): Likewise.
3023 (*sse4_1_<code>v2hiv2di2<mask_name>_2): Likewise.
3024 (*sse4_1_<code>v2siv2di2<mask_name>_1): Likewise.
3025 (*sse4_1_<code>v2siv2di2<mask_name>_2): Likewise.
3027 2018-11-21 H.J. Lu <hongjiu.lu@intel.com>
3029 * read-rtl.c (apply_subst_iterator): Handle define_split and
3030 define_insn_and_split.
3032 2018-11-21 Jakub Jelinek <jakub@redhat.com>
3034 PR rtl-optimization/87817
3035 * config/i386/i386.md (bmi2_bzhi_<mode>3, *bmi2_bzhi_<mode>3,
3036 *bmi2_bzhi_<mode>3_1, *bmi2_bzhi_<mode>3_1_ccz): Use IF_THEN_ELSE
3037 in the pattern to avoid triggering UB when operands[2] is zero.
3038 (tbm_bextri_<mode>): New expander. Renamed the old define_insn to ...
3039 (*tbm_bextri_<mode>): ... this.
3041 2018-11-21 Tom de Vries <tdevries@suse.de>
3044 * params.def (HOT_BB_COUNT_FRACTION): Terminate help message with
3047 2018-11-21 Andreas Krebbel <krebbel@linux.ibm.com>
3049 * configure.ac: Add check for Binutils to determine whether vector
3050 load/store alignments hints are being supported.
3051 * config.in: Regenerate.
3052 * configure: Regenerate.
3053 * config/s390/s390.c (print_operand): Support new output
3055 * config/s390/s390.md ("movti"): Append alignment hint output
3056 using the new output modifier 'A'.
3057 * config/s390/vector.md ("mov<mode>", "*vec_tf_to_v1tf")
3058 ("*vec_ti_to_v1ti"): Likewise.
3060 2018-11-20 Martin Sebor <msebor@redhat.com>
3062 * doc/extend.texi (Other Builtins): Add __builtin_has_attribute.
3064 2018-11-20 Jan Hubicka <hubicka@ucw.cz>
3067 * ipa-devirt.c (odr_types_equivalent_p): Use operand_equal_p to
3068 compare ENUM values.
3070 2018-11-20 Jakub Jelinek <jakub@redhat.com>
3072 PR tree-optimization/87895
3073 * omp-simd-clone.c (ipa_simd_modify_function_body): When removing
3074 or replacing GIMPLE_RETURN, set EDGE_FALLTHRU on the edge to EXIT.
3075 (simd_clone_adjust): Don't set EDGE_FALLTHRU here. In a loop that
3076 redirects edges to EXIT to edges to incr_bb, iterate while EXIT
3077 has any preds and always use EDGE_PRED (, 0).
3079 2018-11-20 Uros Bizjak <ubizjak@gmail.com>
3082 * mode-switching.c (create_pre_exit): After reload, always split the
3083 fallthrough edge to the exit block.
3085 2018-11-20 Jan Hubicka <hubicka@ucw.cz>
3087 * ipa-devirt.c (add_type_duplicate): Do not ICE on incomplete enums.
3088 * tree.c (build_array_type_1): Forward declare.
3089 (fld_type_variant_equal_p): Add INNER_TYPE parameter.
3090 (fld_type_variant): Likewise.
3091 (fld_simplified_types): New hash.
3092 (fld_process_array_type): New function.
3093 (fld_incomplete_type_of): Handle array and enumeration types.
3094 (fld_simplified_type): Handle simplification of arrays.
3095 (free_lang_data): Allocate and free simplified types hash.
3097 2018-11-20 Jan Hubicka <hubicka@ucw.cz>
3100 * ipa-devirt.c (odr_subtypes_equivalent_p): Report ODR violation
3101 when sybtype already violates ODR.
3102 (get_odr_type): Do not ICE when insert is false and type duplicate
3103 is not registered yet.
3104 (register_odr_type): Be sure to register subtypes first.
3106 2018-11-20 Andreas Krebbel <krebbel@linux.ibm.com>
3108 * config/s390/s390.md ("clztidi2"): Swap the RTX's written to the
3109 DImode parts of the target operand.
3111 2018-11-20 Nathan Sidwell <nathan@acm.org>
3114 * Makefile.in (bitmap.o-warn): Use -Wno-error=array-bounds.
3116 2018-11-20 Jan Hubicka <hubicka@ucw.cz>
3119 * tree.c (free_lang_data_in_cgraph): Add argument fld; break out
3121 (free_lang_data) ... here; update call of free_lang_data_in_cgraph.
3123 2018-11-20 Jan Hubicka <hubicka@ucw.cz>
3126 * ipa-fnsummary.c (pass_ipa_fnsummary): Do not remove functions
3127 * ipa.c (possible_inline_candidate_p): Break out from ..
3128 (process_references): ... here ; drop before_inlining_p;
3129 cleanup handling of alises.
3130 (walk_polymorphic_call_targets): Likewise.
3131 (symbol_table::remove_unreachable_nodes): Likewise.
3132 * passes.c (pass_data_ipa_remove_symbols): New structure.
3133 (pass_ipa_remove_symbols): New pass.
3134 (make_pass_ipa_remove_symbols): New function.
3135 * tree-pass.h (make_pass_ipa_remove_symbols): Declare.
3136 * passes.def (pass_ipa_remove_symbols): Schedule after early passes.
3138 2018-11-20 Richard Biener <rguenther@suse.de>
3140 * tree-vect-stmts.c (vectorizable_condition): Do not get
3141 at else_clause vect def for EXTRACT_LAST_REDUCTION. Remove
3142 pointless vect_is_simple_use calls.
3144 2018-11-20 Richard Biener <rguenther@suse.de>
3146 PR tree-optimization/88074
3147 * tree-ssa-dom.c (pass_dominator::execute): Do not walk
3150 2018-11-20 Richard Biener <rguenther@suse.de>
3152 PR tree-optimization/88069
3153 * tree-ssa-sccvn.c (visit_phi): Do not value-number to unvisited
3154 virtual PHI arguments.
3156 2018-11-20 Ilya Leoshkevich <iii@linux.ibm.com>
3159 * config/s390/s390.md: Skip LT(G) peephole when literal pool is
3161 * rtl.h (contains_constant_pool_address_p): New function.
3162 * rtlanal.c (contains_constant_pool_address_p): Likewise.
3164 2018-11-20 Richard Biener <rguenther@suse.de>
3167 * alias.c (component_uses_parent_alias_set_from): Remove
3168 alias-set zero and TYPE_TYPELESS_STORAGE case both already
3169 handled in other ways.
3171 2018-11-20 Richard Biener <rguenther@suse.de>
3173 PR tree-optimization/88087
3174 * tree-ssa-pre.c (create_expression_by_pieces): Re-materialize
3176 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Remember
3179 2018-11-20 Richard Biener <rguenther@suse.de>
3182 * tree-data-ref.c (lambda_matrix_right_hermite): Use abs_hwi.
3184 2018-11-19 Jan Hubicka <hubicka@ucw.cz>
3187 * ipa-devirt.c (free_enum_values): Do not ICE on ODR vilations.
3189 2018-11-19 Sandra Loosemore <sandra@codesourcery.com>
3192 * doc/invoke.texi (Link Options): Mention shared libraries
3193 in documentation for the -l option. Simplify discussion and
3194 point to the system linker documentation for details.
3196 2018-11-19 Peter Bergner <bergner@linux.ibm.com>
3198 PR rtl-optimization/88033
3199 * ira-lives.c (non_conflicting_reg_copy_p): Skip copies from a register
3200 to itself. Use HARD_REGISTER_NUM_P.
3202 2018-11-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3204 * tree-vect-loop.c (vect_transform_loop): Disable further unrolling
3205 of the loop if vf is non-constant.
3207 2018-11-19 David Malcolm <dmalcolm@redhat.com>
3209 PR tree-optimization/87025
3210 * dumpfile.c (dump_context::begin_scope): Call end_any_optinfo
3211 immediately after creating the scope optinfo.
3212 (selftest::test_pr87025): New function.
3213 (selftest::dumpfile_c_tests): Call it.
3214 * optinfo-emit-json.cc (optrecord_json_writer::pop_scope): Assert
3215 that we're not popping the top-level records array.
3216 * optinfo.cc (optinfo::handle_dump_file_kind): Assert that we're
3217 not changing the kind of a "scope" optinfo.
3219 2018-11-19 David Malcolm <dmalcolm@redhat.com>
3221 PR tree-optimization/87025
3222 * dump-context.h: Include "optinfo.h".
3223 (class optrecord_json_writer): New forward decl.
3224 (dump_context::forcibly_enable_optinfo_p): Delete.
3225 (dump_context::optinfo_enabled_p): New member function.
3226 (dump_context::optimization_records_enabled_p): New member
3228 (dump_context::set_json_writer): New member function.
3229 (dump_context::emit_optinfo): New member function.
3230 (dump_context::m_forcibly_enable_optinfo): Delete.
3231 (dump_context::m_json_writer): New member data.
3232 * dumpfile.c (dump_context::set_json_writer): New member function.
3233 (dump_context::finish_any_json_writer): New member function.
3234 (dump_context::end_scope): Replace call to
3235 optimization_records_maybe_pop_dump_scope with call to
3236 m_json_writer->pop_scope.
3237 (dump_context::optinfo_enabled_p): New member function.
3238 (dump_context::end_any_optinfo): Replace call to optinfo::emit with call
3239 to dump_context::emit_optinfo.
3240 (dump_context::emit_optinfo): New member function.
3241 (temp_dump_context::temp_dump_context): Replace
3242 m_forcibly_enable_optinfo with call to set_json_writer.
3243 (temp_dump_context::~temp_dump_context): Clean up any json writer.
3244 * optinfo-emit-json.cc (class optrecord_json_writer): Move to
3246 (the_json_writer): Delete.
3247 (optimization_records_start): Delete.
3248 (optimization_records_finish): Delete.
3249 (optimization_records_enabled_p): Delete, in favor of
3250 dump_context::optimization_records_enabled_p.
3251 (optimization_records_maybe_record_optinfo): Delete.
3252 (optimization_records_maybe_pop_dump_scope): Delete.
3253 * optinfo-emit-json.h: Include "json.h". Delete forward
3255 (optimization_records_start): Delete.
3256 (optimization_records_finish): Delete.
3257 (optimization_records_enabled_p): Delete.
3258 (optimization_records_maybe_record_optinfo): Delete.
3259 (optimization_records_maybe_pop_dump_scope): Delete.
3260 (class optrecord_json_writer): Move here from
3261 optinfo-emit-json.cc.
3262 * optinfo.cc (optinfo::emit_for_opt_problem): Replace call
3263 to optinfo::emit with call to dump_context::emit_optinfo.
3264 (optinfo::emit): Delete, in favor of dump_context::emit_optinfo.
3265 (optinfo_enabled_p): Delete, in favor of
3266 dump_context::optinfo_enabled_p.
3267 (optinfo_wants_inlining_info_p): Update for conversion o
3268 optimization_records_enabled_p to a member function of
3270 * optinfo.h (optinfo_enabled_p): Delete, in favor of
3271 dump_context::optinfo_enabled_p.
3272 (optinfo::emit): Delete, in favor of dump_context::emit_optinfo.
3273 * toplev.c: Include "dump-context.h".
3274 (compile_file): Replace call to optimization_records_finish with
3275 dump_context::finish_any_json_writer.
3276 (do_compile): Replace call to optimization_records_start with
3277 conditionally creating a optrecord_json_writer for the
3280 2018-11-19 Jakub Jelinek <jakub@redhat.com>
3282 PR tree-optimization/88071
3283 * tree-vect-loop.c (vectorize_fold_left_reduction): Pass true instead
3284 of false as last argument to gsi_remove.
3285 * tree-vect-stmts.c (vect_finish_replace_stmt): Pass true instead of
3286 false as last argument to gsi_replace.
3289 * omp-expand.c: Don't include debug.h.
3290 (adjust_context_and_scope): Add REGION argument. Find DECL_CONTEXT
3291 from innermost outer parallel, task, teams or target that has a
3292 child_fn set, or, if there is no such outer region, use
3293 current_function_decl. Do the DECL_CONTEXT adjustment regardless of
3294 whether a suitable BLOCK is found or not.
3295 (expand_parallel_call, expand_teams_call): Don't call
3296 adjust_context_and_scope here.
3297 (grid_expand_target_grid_body): Revert 2017-01-25 changes.
3298 (expand_omp_taskreg, expand_omp_target): Likewise. Call
3299 adjust_context_and_scope.
3300 * dwarf2out.c (dwarf2out_early_global_decl): For
3301 decl_function_context recurse instead of calling dwarf2out_decl.
3303 2018-11-20 Eric Botcazou <ebotcazou@adacore.com>
3305 PR rtl-optimization/85925
3306 * rtl.h (word_register_operation_p): New predicate.
3307 * combine.c (record_dead_and_set_regs_1): Only apply specific handling
3308 for WORD_REGISTER_OPERATIONS targets to word_register_operation_p RTX.
3309 * rtlanal.c (nonzero_bits1): Likewise. Adjust couple of comments.
3310 (num_sign_bit_copies1): Likewise.
3312 2018-11-19 Richard Biener <rguenther@suse.de>
3315 * tree.c (free_lang_data_in_one_sizepos): Free non-gimple-val
3318 2018-11-19 Eric Botcazou <ebotcazou@adacore.com>
3320 * compare-elim.c (struct comparison): Add not_in_a field.
3321 (is_not): New static function.
3322 (strip_not): Likewise.
3323 (conforming_compare): Handle a NOT in the first operand.
3324 (can_eliminate_compare): Likewise.
3325 (find_comparison_dom_walker::before_dom_children): Likewise.
3326 (try_eliminate_compare): Likewise.
3327 * config/visium/visium.md (negsi2_insn_set_carry): Turn into...
3328 (neg<mode>2_insn_set_carry): ...this and add missing NEG operation.
3330 2018-11-19 Jonathan Wakely <jwakely@redhat.com>
3332 * doc/extend.texi (Common Type Attributes): Fix typos.
3334 2018-11-19 Martin Liska <mliska@suse.cz>
3336 PR gcov-profile/88045
3337 * coverage.c (coverage_begin_function): Add assert.
3339 2018-11-18 Sandra Loosemore <sandra@codesourcery.com>
3342 * doc/extend.texi (Common Function Attributes): Document that
3343 no_instrument_function applies to -p and -pg, too.
3344 * doc/invoke.texi (Instrumentation Options): Add cross-references
3345 to docs for -p, -pg, and -finstrument-functions.
3347 2018-11-18 Alan Modra <amodra@gmail.com>
3349 * config/gnu-user.h (GNU_USER_TARGET_CRTI): Define.
3350 (GNU_USER_TARGET_STARTFILE_SPEC): Use it here.
3351 (GNU_USER_TARGET_CRTN): Define.
3352 (GNU_USER_TARGET_ENDFILE_SPEC): Use it here.
3353 (GNU_USER_TARGET_LINK_GCC_C_SEQUENCE_SPEC): Define.
3355 2018-11-18 Alan Modra <amodra@gmail.com>
3357 * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Delete
3358 !HAVE_LD_PIE variant.
3359 (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
3361 2018-11-17 Nick Clifton <nickc@redhat.com>
3362 Sandra Loosemore <sandra@codesourcery.com>
3365 * doc/invoke.texi (Overall Options): Document that --help and
3366 --help=<value> options cannot be combined.
3368 2018-11-17 Jim Wilson <jimw@sifive.com>
3370 * config/riscv/riscv.c (epilogue_cfa_sp_offset): New.
3371 (riscv_restore_reg): If restoring HARD_FRAME_POINTER_REGNUM, and
3372 epilogue_cfa_sp_offset set, then add REG_CFA_DEF_CFA regnote.
3373 (riscv_expand_epilogue): Initialize epilogue_cfa_sp_offset. Set it
3374 to step2 if frame_pointer_needed and step1 is 0.
3376 2018-11-17 Sandra Loosemore <sandra@codesourcery.com>
3379 * doc/cpp.texi (System Headers): Add note about implicit
3380 extern "C" block on targets that define SYSTEM_IMPLICIT_EXTERN_C.
3382 2018-11-17 Sandra Loosemore <sandra@codesourcery.com>
3385 * doc/invoke.texi (C++ Dialect Options): Clarify usage of
3386 -fno-implicit-templates.
3388 2018-11-17 Jakub Jelinek <jakub@redhat.com>
3390 PR tree-optimization/87546
3391 * tree-vect-patterns.c (vect_look_through_possible_promotion): Add
3392 min_precision variable, initially set it to orig_precision, only does
3393 something if op_type's precision is <= min_precision and update
3394 min_precision whenever calling set_op.
3396 2018-11-16 Jan Hubicka <hubicka@ucw.cz>
3399 * ipa-devirt.c (warn_odr): Look for main variant to get TYPE_DECL.
3401 2018-11-16 Sandra Loosemore <sandra@codesourcery.com>
3403 * doc/invoke.texi (Option Summary): Fix whitespace and line
3404 breaks in @gccoptlist environments.
3405 (Warning Options): Likewise.
3406 (Optimize Options): Likewise.
3407 (PowerPC SPE Options): Likewise.
3408 (RS/6000 and PowerPC Options): Likewise.
3410 2018-11-16 Jeff Law <law@redhat.com>
3412 * config/mn10300/mn10300.md (adddi3_degenerate): Remove bogus
3415 2018-11-16 Sandra Loosemore <sandra@codesourcery.com>
3418 * doc/invoke.texi (Optimize Options): Update options enabled by
3419 fprofile-generate, -fprofile-use, and -fauto-profile.
3421 2018-11-16 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3424 * target-def.h: Initialize TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP.
3425 Add them to the TARGET_ASM_{,UN}ALIGNED_INT_OP structs.
3426 * target.def: Enumerate TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP in
3428 * target.h: Add psi, pdi, pti to struct asm_int_op definition.
3429 * targhooks.c (default_print_patchable_function_entry): Assert
3430 asm_int_op does not return a NULL string.
3431 * varasm.c (integer_asm_op): Return the op for a partial int type
3432 when the requested size does not correspond to an integer type.
3433 * config/msp430/msp430.c: Initialize TARGET_ASM_{,UN}ALIGNED_PSI_OP.
3434 * doc/tm.texi: Regenerate.
3436 2018-11-12 Jason Merrill <jason@redhat.com>
3438 * gimplify.c (gimplify_case_label_expr): Handle hot/cold attributes.
3440 2018-11-16 Michael Meissner <meissner@linux.ibm.com>
3442 * config/rs6000/constraints.md (wF constraint): Remove power9
3443 fusion documentation. Just document wF's use for power8 fusion.
3444 * config/rs6000/predicates.md (p9_fusion_reg_operand): Delete.
3445 (fusion_gpr_addis): Delete power9 fusion support. Change power8
3446 fusion support to require the upper 12 bits to be all 0's or all
3448 (fusion_gpr_mem_load): Add comment.
3449 (fusion_addis_mem_combo_load): Remove power9 fusion support.
3450 (fusion_addis_mem_combo_store): Delete.
3451 (fusion_offsettable_mem_operand): Delete.
3452 * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Do not set
3454 (ISA_3_0_MASKS_SERVER): Delete power9 fusion.
3455 (POWERPC_MASKS): Delete power9 fusion.
3456 * config/rs6000/rs6000-protos.h (emit_fusion_load_store): Delete.
3457 (fusion_p9_p): Delete.
3458 (expand_fusion_p9_load): Delete.
3459 (expand_fusion_p9_store): Delete.
3460 (emit_fusion_p9_load): Delete.
3461 (emit_fusion_p9_store): Delete.
3462 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Delete power9
3464 (rs6000_option_override_internal): Set power8 fusion based on
3465 whether we are tuning for power8. Delete power9 fusion support.
3466 (rs6000_opt_masks): Delete -mpower9-fusion switch.
3467 (emit_fusion_load): Rename emit_fusion_load_store to
3468 emit_fusion_load, and drop fusion store support. Update callers.
3469 (emit_fusion_load_store): Likewise.
3470 (emit_fusion_gpr_load): Likewise.
3471 (fusion_p9_p): Delete.
3472 (expand_fusion_p9_load): Delete.
3473 (expand_fusion_p9_store): Delete.
3474 (emit_fusion_p9_load): Delete.
3475 (emit_fusion_p9_store): Delete.
3476 * config/rs6000/rs6000.md (UNSPEC_FUSION_P9): Delete.
3477 (GPR_FUSION): Delete.
3478 (FPR_FUSION): Delete.
3479 (power9 fusion peephole2s): Delete.
3480 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Delete.
3481 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Delete.
3482 (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_load): Delete.
3483 (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_store): Delete.
3484 (fusion_p9_<mode>_constant): Delete.
3485 * config/rs6000/rs6000.opt (-mpower9-fusion): Delete undocumented
3486 power9 fusion switch.
3487 * doc/md.texi (PowerPC constraints): Update wF constraint
3488 documentation for power8 fusion only.
3490 2018-11-16 Jakub Jelinek <jakub@redhat.com>
3492 PR rtl-optimization/87475
3493 * cfgrtl.c (patch_jump_insn): Allow redirection failure for
3494 CROSSING_JUMP_P insns.
3495 (cfg_layout_redirect_edge_and_branch): Don't ICE if ret is NULL.
3497 2018-11-16 Uros Bizjak <ubizjak@gmail.com>
3500 * config/i386/i386.md (floatunsdidf2): Allow only 64bit AVX512F targets.
3501 * config/i386/sse.md (UNSPEC_MOVDI_TO_SSE): New UNSPEC.
3502 (movdi_to_sse): Rewrite using UNSPEC_MOVDI_TO_SSE unspec.
3504 2018-11-16 Jakub Jelinek <jakub@redhat.com>
3507 * optabs.c (expand_binop): For op0_mode use GET_MODE (op0), unless it
3508 is VOIDmode, in which case use int_mode. Similarly for op1_mode.
3510 2018-11-16 Eric Botcazou <ebotcazou@adacore.com>
3512 * tree-switch-conversion.h (switch_decision_tree::emit_case_nodes): Add
3513 location_t parameter.
3514 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
3515 (switch_decision_tree::do_jump_if_equal): Likewise.
3516 * tree-switch-conversion.c (switch_decision_tree::emit): Pass location
3517 of switch statement to emit_case_nodes.
3518 (switch_decision_tree::emit_cmp_and_jump_insns): Add LOC parameter and
3519 set it on the newly built GIMPLE comparison statement.
3520 (switch_decision_tree::do_jump_if_equal): Likewise.
3521 (switch_decision_tree::emit_case_nodes): Add LOC parameter and pass it
3522 in calls to do_jump_if_equal as well as recursive calls.
3524 2018-11-16 Andi Kleen <ak@linux.intel.com>
3526 * config/i386/i386.md: Allow memory operands to ptwrite.
3528 2018-11-16 Michael Matz <matz@suse.de>
3530 * system.h (PRsa): New macro.
3531 (SIZE_AMOUNT): Cast number to uint64_t.
3532 * alloc-pool.h (pool_usage::dump): Don't use %zu but PRsa.
3533 (pool_usage::dump_footer): Likewise and also use PRIu64.
3534 * bitmap.h (bitmap_usage::dump): Likewise.
3535 * ggc-common.c (ggc_usage::dump): Likewise.
3536 * ggc-page.c (ggc_print_statistics): Likewise.
3537 * input.c (dump_line_table_statistics): Likewise.
3538 * mem-stats.h (mem_usage::dump): Likewise.
3539 (mem_usage::dump_footer): Likewise.
3540 * rtl.c (dump_rtx_statistics): Likewise.
3541 * tree-cfg.c (dump_cfg_stats): Likewise.
3542 * tree-dfa.c (dump_dfa_stats): Likewise.
3543 * tree-phinodes.c (phinodes_print_statistics): Likewise.
3544 * tree-ssanames (ssanames_print_statistics): Likewise.
3545 * vec.c (vec_usage::dump): Likewise.
3546 (vec_usage::dump_footer): Likewise.
3548 2018-11-16 Richard Biener <rguenther@suse.de>
3550 PR tree-optimization/88011
3551 * tree-vrp.c (extract_range_from_binary_expr): Fix error in
3552 replacing set_value_range_to_undefined and
3553 set_value_range_to_varying with method calls.
3555 2018-11-16 Ilya Leoshkevich <iii@linux.ibm.com>
3557 * config/s390/s390.md
3558 (*r<noxa>sbg_<mode>_srl_bitmask): Do not delegate arithmetic to
3560 (*r<noxa>sbg_<mode>_sll): Likewise.
3561 (*r<noxa>sbg_<mode>_srl): Likewise.
3562 (*r<noxa>sbg_sidi_srl): New pattern.
3564 2018-11-16 Jerome Lambourg <lambourg@adacore.com>
3566 * config/vxworks.h (TARGET_VXWORKS_HAVE_CTORS_DTORS): New macro.
3567 Default to TARGET_VXWORKS_RTP.
3568 (SUPPORTS_INIT_PRIORITY): Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
3569 of TARGET_VXWORKS_RTP.
3570 * config/vxworks.c: Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
3571 of TARGET_VXWORKS_RTP to set targetm.have_ctors_dtors.
3573 2018-11-15 Sandra Loosemore <sandra@codesourcery.com>
3576 * doc/extend.texi (Common Type Attributes): Make it explicit
3577 that attribute "packed" can apply to C++ classes.
3579 2018-11-15 Martin Sebor <msebor@redhat.com>
3583 * tree.c (type_argument_type): New function.
3584 * tree.h (type_argument_type): Declare it.
3585 * gcc/doc/extend.texi (alloc_align): Update and clarify.
3586 (alloc_size, nonnull, sentinel): Same.
3588 2018-11-15 Andrew Stubbs <ams@codesourcery.com>
3589 Kwok Cheung Yeung <kcy@codesourcery.com>
3591 * tree-vect-stmts.c (vectorizable_store): Don't ICE when
3592 int_mode_for_size fails.
3593 (vectorizable_load): Likewise.
3595 2018-11-15 David Malcolm <dmalcolm@redhat.com>
3597 * doc/ux.texi (Group logically-related diagnostics): Move
3598 discussion of auto_diagnostic_group into this new subsection.
3599 Give an example of where this grouping is used.
3601 2018-11-15 David Malcolm <dmalcolm@redhat.com>
3604 * Makefile.in (OBJS): Move json.o to...
3605 (OBJS-libcommon): ...here and add diagnostic-format-json.o.
3606 * common.opt (fdiagnostics-format=): New option.
3607 (diagnostics_output_format): New enum.
3608 * diagnostic-format-json.cc: New file.
3609 * diagnostic.c (default_diagnostic_final_cb): New function, taken
3610 from start of diagnostic_finish.
3611 (diagnostic_initialize): Initialize final_cb to
3612 default_diagnostic_final_cb.
3613 (diagnostic_finish): Move "being treated as errors" messages to
3614 default_diagnostic_final_cb. Call any final_cb.
3615 (default_diagnostic_finalizer): Add diagnostic_t param.
3616 (diagnostic_report_diagnostic): Pass "orig_diag_kind" to
3617 diagnostic_finalizer callback.
3618 * diagnostic.h (enum diagnostics_output_format): New enum.
3619 (diagnostic_finalizer_fn): Reimplement, adding diagnostic_t param.
3620 (struct diagnostic_context): Add "final_cb".
3621 (default_diagnostic_finalizer): Add diagnostic_t param.
3622 (diagnostic_output_format_init): New decl.
3623 * doc/invoke.texi (-fdiagnostics-format): New option.
3624 * dwarf2out.c (gen_producer_string): Ignore
3625 OPT_fdiagnostics_format_.
3626 * gcc.c (driver_handle_option): Handle OPT_fdiagnostics_format_.
3627 * lto-wrapper.c (append_diag_options): Ignore it.
3628 * opts.c (common_handle_option): Handle it.
3630 2018-11-15 David Malcolm <dmalcolm@redhat.com>
3632 PR tree-optimization/88015
3633 * graphite-isl-ast-to-gimple.c
3634 (translate_isl_ast_to_gimple::scop_to_isl_ast): Add missing check
3636 * graphite-sese-to-poly.c (build_poly_scop): Likewise.
3638 2018-11-15 Richard Biener <rguenther@suse.de>
3641 * gimple.c (gimple_call_flags): Union flags from decl, type
3643 * trans-mem.c (is_tm_pure_call): Simplify.
3645 2018-11-15 Richard Biener <rguenther@suse.de>
3647 PR tree-optimization/88031
3648 * tree-vect-loop.c (vectorizable_reduction): Move check
3649 for multiple types earlier so we get the expected dump.
3650 Simplify calls to vectorizable_condition.
3651 * tree-vect-stmts.h (vectorizable_condition): Update prototype.
3652 * tree-vect-stmts.c (vectorizable_condition): Instead of
3653 reduc_def and reduc_index take just a flag. Simplify
3654 code-generation now that we can rely on the defs being set up.
3655 (vectorizable_comparison): Remove unused argument.
3657 2018-11-15 Richard Biener <rguenther@suse.de>
3659 PR tree-optimization/88030
3660 * tree-complex.c (need_eh_cleanup): New global.
3661 (update_complex_assignment): Mark blocks that need EH update.
3662 (expand_complex_comparison): Likewise.
3663 (tree_lower_complex): Allocate and deallocate need_eh_cleanup,
3664 perform EH cleanup and schedule CFG cleanup if that did anything.
3666 2018-11-15 Jakub Jelinek <jakub@redhat.com>
3668 PR rtl-optimization/88018
3669 * cfgrtl.c (fixup_abnormal_edges): Guard moving insns to fallthru edge
3670 on the presence of fallthru edge, rather than if it is a USE or not.
3672 2018-11-15 Richard Biener <rguenther@suse.de>
3675 * tree-data-ref.c (analyze_miv_subscript): Guard calls to
3676 analyze_subscript_affine_affine properly.
3678 2018-11-15 Xianmiao Qu <xianmiao_qu@c-sky.com>
3680 * config/csky/csky-linux-elf.h (LINUX_DYNAMIC_LINKER): Remove.
3681 (GLIBC_DYNAMIC_LINKER): Define.
3682 (LINUX_TARGET_LINK_SPEC): Update the dynamic linker's name.
3684 2018-11-15 Bin Cheng <bin.cheng@linux.alibaba.com>
3686 PR tree-optimization/84648
3687 * tree-ssa-loop-niter.c (adjust_cond_for_loop_until_wrap): New.
3688 (number_of_iterations_cond): Adjust exit cond for loop-until-wrap case
3689 by calling adjust_cond_for_loop_until_wrap.
3691 2018-11-15 Sandra Loosemore <sandra@codesourcery.com>
3694 * doc/extend.texi (Common Function Attributes): Clarify linker
3695 restrictions on "aligned" attribute.
3696 (Common Variable Attributes): Likewise. Mention that linker
3697 restrictions don't apply to stack-allocated variables.
3699 2018-11-15 Joern Rennecke <joern.rennecke@riscy-ip.com>
3701 * gensupport.c (add_predicate_code): Properly handle ZERO_EXTRACT
3704 2018-11-14 Sandra Loosemore <sandra@codesourcery.com>
3708 * doc/invoke.texi (Optimize Options): Remove bad example about
3709 interaction between -flto and -O. Replace it with a note that
3710 you need to compile with -O and not just link. Copy-edit -flto
3711 discussion to reduce verbiage and improve flow.
3713 2018-11-14 Sandra Loosemore <sandra@codesourcery.com>
3716 * doc/invoke.texi (Optimize Options): Clarify that -O0 and -Og
3717 also suppress many optimizations. Alphabetize option lists for
3718 -O1, -O2, and -Os. Add list of options disabled with -Og, and
3719 correct documentation for those options to say that.
3720 * opts.c (default_options_table): Sort table by level and option
3721 name, to make it easier to correlate to the manual.
3723 2018-11-14 Uros Bizjak <ubizjak@gmail.com>
3725 * config/i386/i386.c (ix86_print_operand_address_as): Simplify
3726 printing of the names of segment registers.
3728 2018-11-14 Nathan Sidwell <nathan@acm.org>
3732 * dwarf2out.c (dwarf2out_finish): Apply resolve_addr to comdat
3735 2018-11-14 David Malcolm <dmalcolm@redhat.com>
3737 * Makefile.in (CFLAGS-optinfo-emit-json.o): Add $(ZLIBINC).
3739 2018-11-14 Jakub Jelinek <jakub@redhat.com>
3741 P1236R1 - Signed integers are two's complement
3742 * doc/invoke.texi (Wshift-overflow): Adjust documentation for c++2a.
3745 * hash-map.h (hash_map::iterator::reference_pair): New class.
3746 (hash_map::iterator::operator*): Return it rather than std::pair.
3748 2018-11-14 Jeff Law <law@redhat.com>
3750 * optabs.c (expand_binop): Pass INT_MODE to operand_subword_force
3751 iff the operand is a constant.
3753 2018-11-14 Aldy Hernandez <aldyh@redhat.com>
3755 * gimple-ssa-evrp-analyze.c
3756 (evrp_range_analyzer::record_ranges_from_incoming_edge): Rename
3757 ignore_equivs_equal_p to equal_p.
3758 * ipa-cp.c (meet_with_1): Use equal_p instead of
3759 ignore_equivs_equal_p.
3760 * ipa-prop.c (ipa_vr_ggc_hash_traits::equal): Same.
3761 * tree-vrp.c (value_range::ignore_equivs_equal_p): Remove.
3762 (value_range::operator==): Remove.
3763 (value_range::operator!=): Remove.
3764 (vrp_prop::visit_stmt): Use equal_p.
3765 * tree-vrp.h (value_range): Remove operator==, operator!=,
3766 ignore_equivs_equal_p.
3767 * vr-values.c (update_value_range): Use equal_p.
3769 2018-11-14 Michael Matz <matz@suse.de>
3772 * gimplify.c (collect_fallthrough_labels): Add new argument,
3773 return location via that, don't modify statements.
3774 (warn_implicit_fallthrough_r): Adjust call, don't use
3775 statement location directly.
3777 2018-11-14 Richard Biener <rguenther@suse.de>
3780 * tree-data-ref.c (split_constant_offset): Add wrapper
3781 allocating a cache hash-map.
3782 (split_constant_offset_1): Cache results of expanding
3783 expressions from SSA def stmts.
3785 2018-11-14 Richard Biener <rguenther@suse.de>
3788 * tree-data-ref.c (lambda_matrix_row_add): Change const1 argument
3790 (lambda_vector_mult_const): Likewise.
3791 (lambda_matrix_right_hermite): Use lambda_int temporaries.
3793 2018-11-14 Wilco Dijkstra <wdijkstr@arm.com>
3794 Jackson Woodruff <jackson.woodruff@arm.com>
3796 PR 71026/tree-optimization
3797 * match.pd: Simplify floating point comparisons.
3799 2018-11-14 Jakub Jelinek <jakub@redhat.com>
3801 PR rtl-optimization/87817
3802 * config/i386/i386.c (ix86_fold_builtin): For _bzhi_u{32,64} if
3803 last argument has low 8 bits clear, fold to 0.
3805 2018-11-14 Iain Sandoe <iain@sandoe.co.uk>
3807 * tree-vect-data-refs.c (vect_can_force_dr_alignment_p): Cast
3808 MAX_OFILE_ALIGNMENT as needed.
3810 2018-11-14 Richard Biener <rguenther@suse.de>
3812 PR tree-optimization/88019
3813 * graphite-isl-ast-to-gimple.c (ternary_op_to_tree): Rewrite
3814 COND_EXPR arguments to non-trapping overflow.
3816 2018-11-14 Claudiu Zissulescu <claziss@synopsys.com>
3818 * config/arc/arc.h (reg_class): Reorder registers classes, remove
3819 unused register classes.
3820 (REG_CLASS_NAMES): Likewise.
3821 (REG_CLASS_CONTENTS): Likewise.
3822 (FIXED_REGISTERS): Make lp_count fixed.
3823 (BASE_REG_CLASS): Remove ACC16_BASE_REGS reference.
3824 (PROGRAM_COUNTER_REGNO): Remove.
3825 * config/arc/arc.c (arc_conditional_register_usage): Remove unused
3826 register classes, use constants for register numbers, remove
3827 reg_class_contents references.
3828 (arc_process_double_reg_moves): Add asserts.
3829 (arc_secondary_reload): Remove LPCOUNT_REG reference, use
3830 lra_in_progress predicate.
3831 (arc_init_reg_tables): Remove unused register classes.
3832 (arc_register_move_cost): Likewise.
3833 (arc_preferred_reload_class): Likewise.
3834 (hwloop_optimize): Update rtx patterns involving lp_count register.
3835 (arc_return_address_register): Rename ILINK1, INLINK2 regnums macros.
3836 * config/arc/constraints.md ("c"): Choose between GENERAL_REGS and
3837 CHEAP_CORE_REGS. Former one will be used for LRA.
3838 ("Rac"): Choose between GENERAL_REGS and ALL_CORE_REGS. Former
3839 one will be used for LRA.
3840 ("w"): Choose between GENERAL_REGS and WRITABLE_CORE_REGS. Former
3841 one will be used for LRA.
3842 ("W"): Choose between GENERAL_REGS and MPY_WRITABLE_CORE_REGS.
3843 Former one will be used for LRA.
3844 ("f"): Delete constraint.
3847 ("l"): Change it from register constraint to constraint.
3848 * config/arc/arc.md (movqi_insn): Remove unsed lp_count constraints.
3849 (movhi_insn): Likewise.
3850 (movsi_insn): Update pattern.
3853 (stack_tie): Remove 'b' constraint letter.
3855 (R9_REG, R15_REG, R16_REG, R25_REG): Likewise.
3856 (R32_REG, R40_REG, R41_REG, R42_REG, R43_REG, R44_REG): Likewise.
3857 (R57_REG, R59_REG, PCL_REG): Likewise.
3858 (ILINK1_REGNUM): Renamed to ILINK1_REG.
3859 (ILINK2_REGNUM): Renamed to ILINK2_REG.
3861 (SP_REGS): Likewise.
3862 (Rcw): Remove unused reg classes.
3863 * config/arc/predicates.md (dest_reg_operand): Just default on
3864 register_operand predicate.
3865 (mpy_dest_reg_operand): Likewise.
3866 (move_dest_operand): Use macros instead of constants.
3867 * config/arc/arc.opt (mlra): Switch to lra as default.
3869 2018-11-14 Richard Biener <rguenther@suse.de>
3871 PR tree-optimization/87974
3872 * tree-vect-loop.c (vectorizable_reduction): When computing
3873 the vectorized reduction PHI vector type ignore constant
3876 2018-11-14 Jakub Jelinek <jakub@redhat.com>
3878 PR tree-optimization/87977
3879 * tree-ssa-math-opts.c (optimize_recip_sqrt): Don't reuse division
3880 stmt, build a new one and replace the old one with it. Formatting fix.
3881 Call release_ssa_name (x) if !has_other_use and !delete_div.
3882 (pass_cse_reciprocals::execute): Before calling optimize_recip_sqrt
3883 verify lhs of stmt is still def.
3885 2018-11-13 Peter Bergner <bergner@linux.ibm.com>
3887 PR rtl-optimization/87507
3888 * lower-subreg.c (operand_for_swap_move_operator): New function.
3889 (simple_move): Strip simple operators.
3890 (find_pseudo_copy): Likewise.
3891 (resolve_operand_for_swap_move_operator): New function.
3892 (resolve_simple_move): Strip simple operators and swap operands.
3894 2018-11-13 Jakub Jelinek <jakub@redhat.com>
3896 PR tree-optimization/87898
3897 * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Formatting fix.
3898 (ipa_simd_modify_function_body): Remove debug stmts where the first
3899 argument was changed into a non-decl.
3901 2018-11-13 Peter Bergner <bergner@linux.ibm.com>
3903 PR rtl-optimization/87899
3904 * lra-lives.c (start_living): Update white space in comment.
3905 (enum point_type): New.
3906 (sparseset_contains_pseudos_p): New function.
3907 (update_pseudo_point): Likewise.
3908 (make_hard_regno_live): Use HARD_REGISTER_NUM_P macro.
3909 (make_hard_regno_dead): Likewise. Remove ignore_reg_for_conflicts
3910 handling. Move early exit after adding conflicts.
3911 (mark_pseudo_live): Use HARD_REGISTER_NUM_P macro. Add early exit
3912 if regno is already live. Remove all handling of program points.
3913 (mark_pseudo_dead): Use HARD_REGISTER_NUM_P macro. Add early exit
3914 after adding conflicts. Remove all handling of program points and
3915 ignore_reg_for_conflicts.
3916 (mark_regno_live): Use HARD_REGISTER_NUM_P macro. Remove return value
3917 and do not guard call to mark_pseudo_live.
3918 (mark_regno_dead): Use HARD_REGISTER_NUM_P macro. Remove return value
3919 and do not guard call to mark_pseudo_dead.
3920 (check_pseudos_live_through_calls): Use HARD_REGISTER_NUM_P macro.
3921 (process_bb_lives): Use HARD_REGISTER_NUM_P and HARD_REGISTER_P macros.
3922 Use new function update_pseudo_point. Handle register copies by
3923 removing the source register from the live set. Handle INOUT operands.
3924 Update to the next program point using the unused_set, dead_set and
3926 (lra_create_live_ranges_1): Use HARD_REGISTER_NUM_P macro.
3928 2018-11-13 David Malcolm <dmalcolm@redhat.com>
3930 * builtins.c: Replace "source_location" with "location_t".
3931 * diagnostic-show-locus.c: Likewise.
3932 * diagnostic.c: Likewise.
3933 * dumpfile.c: Likewise.
3934 * gcc-rich-location.h: Likewise.
3935 * genmatch.c: Likewise.
3936 * gimple.h: Likewise.
3937 * gimplify.c: Likewise.
3938 * input.c: Likewise.
3939 * input.h: Likewise. Eliminate the typedef.
3940 * omp-expand.c: Likewise.
3941 * selftest.h: Likewise.
3942 * substring-locations.h (get_source_location_for_substring):
3944 (get_location_within_string): ...this.
3945 * tree-cfg.c: Replace "source_location" with "location_t".
3946 * tree-cfgcleanup.c: Likewise.
3947 * tree-diagnostic.c: Likewise.
3948 * tree-into-ssa.c: Likewise.
3949 * tree-outof-ssa.c: Likewise.
3950 * tree-parloops.c: Likewise.
3951 * tree-phinodes.c: Likewise.
3952 * tree-phinodes.h: Likewise.
3953 * tree-ssa-loop-ivopts.c: Likewise.
3954 * tree-ssa-loop-manip.c: Likewise.
3955 * tree-ssa-phiopt.c: Likewise.
3956 * tree-ssa-phiprop.c: Likewise.
3957 * tree-ssa-threadupdate.c: Likewise.
3958 * tree-ssa.c: Likewise.
3959 * tree-ssa.h: Likewise.
3960 * tree-vect-loop-manip.c: Likewise.
3962 2018-11-13 Michael Ploujnikov <michael.ploujnikov@oracle.com>
3964 * doc/extend.texi: Fix typo in the weakref description.
3966 2018-11-13 Richard Biener <rguenther@suse.de>
3968 PR tree-optimization/86991
3969 * tree-vect-loop.c (vect_is_slp_reduction): Delay reduction
3970 group building until we have successfully detected the SLP
3972 (vect_is_simple_reduction): Remove fixup code here.
3974 2018-11-13 David Malcolm <dmalcolm@redhat.com>
3976 * dumpfile.c (VERIFY_DUMP_ENABLED_P): New macro.
3977 (dump_gimple_stmt): Use it.
3978 (dump_gimple_stmt_loc): Likewise.
3979 (dump_gimple_expr): Likewise.
3980 (dump_gimple_expr_loc): Likewise.
3981 (dump_generic_expr): Likewise.
3982 (dump_generic_expr_loc): Likewise.
3983 (dump_printf): Likewise.
3984 (dump_printf_loc): Likewise.
3985 (dump_dec): Likewise.
3986 (dump_dec): Likewise.
3987 (dump_hex): Likewise.
3988 (dump_symtab_node): Likewise.
3989 * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
3990 Guard dump call with dump_enabled_p.
3991 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
3992 * graphite-optimize-isl.c (optimize_isl): Likewise.
3993 * graphite.c (graphite_transform_loops): Likewise.
3994 * tree-loop-distribution.c (pass_loop_distribution::execute): Likewise.
3995 * tree-parloops.c (parallelize_loops): Likewise.
3996 * tree-ssa-loop-niter.c (number_of_iterations_exit): Likewise.
3997 * tree-vect-data-refs.c (vect_analyze_group_access_1): Likewise.
3998 (vect_prune_runtime_alias_test_list): Likewise.
3999 * tree-vect-loop.c (vect_update_vf_for_slp): Likewise.
4000 (vect_estimate_min_profitable_iters): Likewise.
4001 * tree-vect-slp.c (vect_record_max_nunits): Likewise.
4002 (vect_build_slp_tree_2): Likewise.
4003 (vect_supported_load_permutation_p): Likewise.
4004 (vect_slp_analyze_operations): Likewise.
4005 (vect_slp_analyze_bb_1): Likewise.
4006 (vect_slp_bb): Likewise.
4007 * tree-vect-stmts.c (vect_analyze_stmt): Likewise.
4008 * tree-vectorizer.c (try_vectorize_loop_1): Likewise.
4009 (pass_slp_vectorize::execute): Likewise.
4010 (increase_alignment): Likewise.
4012 2018-11-13 David Malcolm <dmalcolm@redhat.com>
4015 * ipa-inline.c (report_inline_failed_reason): Guard calls to
4016 cl_target_option_print_diff and cl_optimization_print_diff with
4019 2018-11-13 David Malcolm <dmalcolm@redhat.com>
4021 * doc/invoke.texi (-fsave-optimization-record): Note that the
4022 output is compressed.
4023 * optinfo-emit-json.cc: Include <zlib.h>.
4024 (optrecord_json_writer::write): Compress the output.
4026 2018-11-13 Aldy Hernandez <aldyh@redhat.com>
4028 * tree-vrp.c (value_range_base::dump): Dump type.
4029 Do not use INF nomenclature for 1-bit types.
4030 (dump_value_range): Group all variants to common dumping code.
4031 (debug): New overloaded functions for value_ranges.
4032 (value_range_base::dump): Remove no argument version.
4033 (value_range::dump): Same.
4035 2018-11-13 Richard Biener <rguenther@suse.de>
4037 PR tree-optimization/87931
4038 * tree-vect-loop.c (vect_is_simple_reduction): Restrict
4039 nested cycles we support to latch computations vectorizable_reduction
4042 2018-11-13 Martin Liska <mliska@suse.cz>
4044 PR tree-optimization/87885
4045 * cfghooks.c (account_profile_record): Rename
4047 (profile_record_check_consistency): ... this.
4048 Calculate missing num_mismatched_freq_in.
4049 (profile_record_account_profile): New function
4050 that calculates time and size of a function.
4051 * cfghooks.h (struct profile_record): Remove
4053 (struct cfg_hooks): Remove after_pass flag.
4054 (account_profile_record): Rename to ...
4055 (profile_record_check_consistency): ... this.
4056 (profile_record_account_profile): New.
4057 * cfgrtl.c (rtl_account_profile_record): Remove
4059 * passes.c (check_profile_consistency): Do only
4061 (account_profile): Calculate size and time of
4063 (pass_manager::dump_profile_report): Reformat
4065 (execute_one_ipa_transform_pass): Call
4066 consistency check before clean upand call account_profile
4067 after a clean up is done.
4068 (execute_one_pass): Call check_profile_consistency and
4069 account_profile instead of using after_pass flag..
4070 * tree-cfg.c (gimple_account_profile_record): Likewise.
4072 2018-11-13 Martin Liska <mliska@suse.cz>
4075 * config/i386/i386.c (ix86_option_override_internal): Error
4076 about usage -mabi=ms and -fsanitize={,kernel-}address.
4078 2018-11-13 Andre Vieira <andre.simoesdiasvieira@arm.com>
4080 * config/aarch64/aarch64.c
4081 (aarch64_vectorize_preferred_vector_alignment): Change return type to
4083 (aarch64_simd_vector_alignment_reachable): Adapt to preferred vector
4084 alignment being a poly int.
4085 * doc/tm.texi (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Change
4086 return type to poly_uint64.
4087 * target.def (default_preferred_vector_alignment): Likewise.
4088 * targhooks.c (default_preferred_vector_alignment): Likewise.
4089 * targhooks.h (default_preferred_vector_alignment): Likewise.
4090 * tree-vect-data-refs.c (vect_calculate_target_alignment): Likewise.
4091 (vect_compute_data_ref_alignment): Adapt to vector alignment being a
4093 (vect_update_misalignment_for_peel): Likewise.
4094 (vect_enhance_data_refs_alignment): Likewise.
4095 (vect_find_same_alignment_drs): Likewise.
4096 (vect_duplicate_ssa_name_ptr_info): Likewise.
4097 (vect_setup_realignment): Likewise.
4098 (vect_can_force_dr_alignment_p): Change alignment parameter type to
4100 * tree-vect-loop-manip.c (get_misalign_in_elems): Learn to construct a
4101 mask with a compile time variable vector alignment.
4102 (vect_gen_prolog_loop_niters): Adapt to vector alignment being a poly
4104 (vect_do_peeling): Exit early if vector alignment is not constant.
4105 * tree-vect-stmts.c (ensure_base_align): Adapt to vector alignment being
4107 (vectorizable_store): Likewise.
4108 (vectorizable_load): Likweise.
4109 * tree-vectorizer.h (struct dr_vec_info): Make target_alignment field a
4111 (vect_known_alignment_in_bytes): Adapt to vector alignment being a
4113 (vect_can_force_dr_alignment_p): Change alignment parameter type to
4116 2018-11-13 Richard Biener <rguenther@suse.de>
4118 PR tree-optimization/87962
4119 * tree-vect-loop.c (vect_is_simple_reduction): More reliably
4120 detect outer reduction for disqualifying in-loop uses.
4122 2018-11-13 Richard Biener <rguenther@suse.de>
4124 PR tree-optimization/87967
4125 * tree-vect-loop.c (vect_transform_loop): Also copy PHIs
4126 for constants for the scalar loop.
4128 2018-11-13 Alan Modra <amodra@gmail.com>
4130 * config/rs6000/linux64.h (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Exclude
4131 integer constants when -mcmodel=medium.
4133 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
4135 * config/arc/arc.h (FUNCTION_PROFILER): Redefine to empty.
4136 * config/arc/elf.h (PROFILE_HOOK): Define.
4137 * config/arc/linux.h (PROFILE_HOOK): Likewise.
4139 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
4141 * config/arc/arc.c (hwloop_optimize): Bailout when detecting a
4142 jump table data in the text section.
4144 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
4146 * config/arc/arc.c (arc_eh_return_address_location): Repurpose it
4147 to fit the eh_return pattern.
4148 * config/arc/arc.md (eh_return): Define.
4149 (VUNSPEC_ARC_EH_RETURN): Likewise.
4150 * config/arc/arc-protos.h (arc_eh_return_address_location): Match
4152 * config/arc/arc.h (EH_RETURN_HANDLER_RTX): Remove it.
4155 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
4157 * common/config/arc/arc-common.c (arc_option_optimization_table):
4158 Millicode optimization is default on for size optimizations.
4159 * config/arc/arc-protos.h (arc_check_multi): New function.
4160 * config/arc/arc.c (RTX_OK_FOR_OFFSET_P): Rearange.
4161 (ENTER_LEAVE_START_REG): Define.
4162 (ENTER_LEAVE_END_REG): Likewise.
4163 (arc_override_options): Disable millicode when long calls option
4165 (arc_frame_info): Change it from int to bool.
4166 (arc_compute_frame_size): Clean up.
4167 (arc_save_restore): Remove.
4168 (frame_save_reg): New function.
4169 (frame_restore_reg): Likewise.
4170 (arc_enter_leave_p): Likewise.
4171 (arc_save_callee_saves): Likewise.
4172 (arc_restore_callee_saves): Likewise.
4173 (arc_save_callee_enter): Likewise.
4174 (arc_restore_callee_leave): Likewise.
4175 (arc_save_callee_milli): Likewise.
4176 (arc_restore_callee_milli): Likewise.
4177 (arc_expand_prologue): Reimplement to emit enter/leave
4179 (arc_expand_epilogue): Likewise.
4180 (arc_check_multi): New function.
4181 * config/arc/arc.md (push_multi_fp): New pattern.
4182 (push_multi_fp_blink): Likewise.
4183 (pop_multi_fp): Likewise.
4184 (pop_multi_fp_blink): Likewise.
4185 (pop_multi_fp_ret): Likewise.
4186 (pop_multi_fp_blink_ret): Likewise.
4187 * config/arc/arc.opt (mmillicode): Update option.
4188 (mcode-density-frame): New option.
4189 * config/arc/predicates.md (push_multi_operand): New predicate.
4190 (pop_multi_operand): Likewise.
4191 * doc/invoke.texi (ARC): Update ARC options information.
4194 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
4196 * config/arc/arc-protos.h (gen_operands_ldd_std): Add.
4197 * config/arc/arc.c (operands_ok_ldd_std): New function.
4198 (mem_ok_for_ldd_std): Likewise.
4199 (gen_operands_ldd_std): Likewise.
4200 * config/arc/arc.md: Add peephole2 rules for std/ldd.
4202 2018-11-13 Eric Botcazou <ebotcazou@adacore.com>
4204 * toplev.c (output_stack_usage): Turn test on flag_stack_usage into
4205 test on stack_usage_file.
4206 (lang_dependent_init): Do not open the .su file if generating LTO.
4208 2018-11-13 Jakub Jelinek <jakub@redhat.com>
4210 PR rtl-optimization/87918
4211 * simplify-rtx.c (simplify_merge_mask): For COMPARISON_P, use
4212 simplify_gen_relational rather than simplify_gen_binary.
4214 2018-11-13 Richard Biener <rguenther@suse.de>
4216 * tree-ssanames.h (set_range_info): Use value_range_base.
4217 (get_range_info): Likewise.
4218 * tree-ssanames.c (set_range_info): Likewise.
4219 (get_range_info): Likewise.
4220 * tree-vrp.c (value_range_base::union_helper): Split
4221 out common parts of value_range[_base]::union_.
4222 (value_range_base::union_): Update.
4223 (value_range::union_): Likewise.
4224 (determine_value_range_1): Use value_range_base.
4225 (determine_value_range): Likewise.
4226 * tree-vrp.h (value_range_base::union_helper): Move ...
4227 (value_range::union_helper): ... from here.
4229 2018-11-13 Alan Modra <amodra@gmail.com>
4231 * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Negate
4233 (rs6000_secondary_reload_gpr): Don't call find_replacement.
4235 2018-11-13 Sandra Loosemore <sandra@codesourcery.com>
4238 * doc/invoke.texi (Optimize Options): Clarify that the
4239 l1-cache-line-size, l1-cache-size, and l2-cache-size parameters
4240 apply to data cache size.
4242 2018-11-13 Alan Modra <amodra@gmail.com>
4244 * config/rs6000/predicates.md (logical_const_operand),
4245 (logical_operand): Correct comment.
4246 * config/rs6000/rs6000.c (rs6000_emit_prologue): Comment fix.
4248 2018-11-13 Alan Modra <amodra@gmail.com>
4250 * config/rs6000/rs6000.md (addsi3_high): Prefix with '*'.
4252 2018-11-13 Alan Modra <amodra@gmail.com>
4254 * gcc/doc/md.texi (Machine Constraints): Correct rs6000 h constraint
4256 * config/rs6000/rs6000.md (movsi_internal1): Delete MT%0 case
4257 covered by alternative.
4258 (movcc_internal1): Ignore h for register preference.
4259 (mov<mode>_hardfloat64): Likewise.
4260 (mov<mode>_softfloat): Ignore c, l, h for register preference.
4262 2018-11-12 Sandra Loosemore <sandra@codesourcery.com>
4264 PR preprocessor/47823
4265 * doc/cpp.texi (Alternatives to Wrapper #ifndef): Move #pragma once
4268 * doc/extend.texi (Pragmas): Note additional pragmas documented
4271 2018-11-13 Kugan Vivekanandarajah <kuganv@linaro.org>
4275 * tree-scalar-evolution.c (expression_expensive_p): Make
4276 BUILTIN POPCOUNT as expensive when backend does not define it.
4278 2018-11-12 Fredrik Noring <noring@nocrew.org>
4280 * config.gcc: Update with-llsc defaults for MIPS r5900.
4282 2018-11-12 Martin Liska <mliska@suse.cz>
4284 PR gcov-profile/87442
4285 * common.opt: Add -fprofile-filter-files and -fprofile-exclude-files
4287 * doc/invoke.texi: Document them.
4288 * tree-profile.c (parse_profile_filter): New.
4289 (parse_profile_file_filtering): Likewise.
4290 (release_profile_file_filtering): Likewise.
4291 (include_source_file_for_profile): Likewise.
4292 (tree_profiling): Filter source files based on the
4293 newly added options.
4295 2018-11-12 Martin Liska <mliska@suse.cz>
4298 * doc/extend.texi: Add missing values for __builtin_cpu_is and
4299 __builtin_cpu_supports for x86 target.
4301 2018-11-12 Sandra Loosemore <sandra@codesourcery.com>
4304 * doc/rtl.texi (Constants): Clarify that mode of "high" doesn't
4306 (Arithmetic): Likewise for "lo_sum".
4308 2018-11-12 Renlin Li <renlin.li@arm.com>
4311 * dse.c (get_stored_val): Add check for compile-time constantness
4314 2018-11-12 Sudakshina Das <sudi.das@arm.com>
4316 * config/arm/arm-cpus.in (armv8_5, sb, predres): New features.
4317 (ARMv8_5a): New fgroup.
4318 (armv8.5-a): New arch.
4319 (armv8-a, armv8.1-a, armv8.2-a, armv8.3-a, armv8.4-a): New options
4321 * config/arm/arm-tables.opt: Regenerate.
4322 * config/arm/t-aprofile: Add matching rules for -march=armv8.5-a
4323 * config/arm/t-arm-elf (all_v8_archs): Add armv8.5-a.
4324 * config/arm/t-multilib (v8_5_a_simd_variants): New variable.
4325 Add matching rules for -march=armv8.5-a and extensions.
4326 * doc/invoke.texi (ARM options): Document -march=armv8.5-a.
4327 Add sb and predres to all armv8-a except armv8.5-a.
4329 2018-11-12 Richard Biener <rguenther@suse.de>
4331 * tree-vrp.h (value_range[_base]::set): Make public. Provide
4332 overload for single value.
4333 (value_range[_base]::set_nonnull): New.
4334 (value_range[_base]::set_null): Likewise.
4335 (value_range): Document bitmap copying behavior, mark
4336 copy constructor and assignment operator deleted.
4337 (value_range::move): New.
4338 (value_range::set_and_canonicalize): Default bitmap to zero.
4339 (set_value_range_to_nonnull): Remove.
4340 (set_value_range_to_null): Likewise.
4341 (set_value_range): Likewise.
4342 (set_value_range_to_value): Likewise.
4343 (extract_range_from_unary_expr): Work on value_range_base.
4344 (extract_range_from_binary_expr_1): Likewise. Rename to...
4345 (extract_range_from_binary_expr): ... this.
4346 * tree-vrp.c (value_range::update): Clear equiv bitmap
4348 (value_range::move): New, move equiv bitmap.
4349 (value_range_base::set_undefined): Avoid assignment.
4350 (value_range::set_undefined): Likewise.
4351 (value_range_base::set_varying): Likewise.
4352 (value_range::set_varying): Likewise.
4353 (set_value_range): Remove.
4354 (value_range_base::set): New overload for value.
4355 (value_range::set): Likewise.
4356 (set_value_range_to_nonnull): Remove.
4357 (value_range_base::set_nonnull): New.
4358 (value_range::set_nonnull): Likewise.
4359 (set_value_range_to_null): Remove.
4360 (value_range_base::set_null): New.
4361 (value_range::set_null): Likewise.
4362 (range_is_null): Work on value_range_base.
4363 (range_is_nonnull): Likewise.
4364 (ranges_from_anti_range): Likewise.
4365 (extract_range_into_wide_ints): Likewise.
4366 (extract_range_from_multiplicative_op): Likewise.
4367 (extract_range_from_binary_expr): Likewise. Update for API changes.
4368 (extract_range_from_unary_expr): Likewise. Remove OBJ_TYPE_REF
4370 (value_range::intersect_helper): Avoid copy and assignment.
4371 (value_range::union_helper): Likewise.
4372 (determine_value_range_1): Adjust.
4373 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
4374 Avoid assignment by using move.
4375 (evrp_range_analyzer::record_ranges_from_stmt): Avoid assignment.
4376 * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
4378 * tree-ssanames.c (get_range_info): Likewise.
4379 * vr-values.h (vr_values::get_vr_for_comparison): Adjust API.
4380 * vr-values.c (vr_values::get_value_range): Adjust.
4381 (vr_values::update_value_range): Likewise.
4382 (symbolic_range_based_on_p): Work on value_range_base.
4383 (vr_values::extract_range_from_binary_expr): Use value_range_base.
4384 (vr_values::extract_range_from_unary_expr): Likewise.
4385 (vr_values::extract_range_from_cond_expr): Avoid assignment.
4386 (vr_values::extract_range_from_comparison): Adjust.
4387 (vr_values::check_for_binary_op_overflow): Use value_range_base.
4388 (vr_values::extract_range_basic): Adjust.
4389 (vr_values::adjust_range_with_scev): Likewise.
4390 (vr_values::vrp_visit_assignment_or_call): Likewise.
4391 (vr_values::get_vr_for_comparison): Change API to avoid
4392 assignment and copy construction.
4393 (vr_values::compare_name_with_value): Adjust accordingly.
4394 (vr_values::compare_names): Likewise.
4395 (vr_values::extract_range_from_phi_node): Avoid assignment and
4396 bogus in-place modify of equiv bitmap.
4397 (vr_values::simplify_bit_ops_using_ranges): Use value_range_base.
4398 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Adjust
4399 for extract_range_from_unary_expr API change.
4400 * ipa-cp.c (ipa_vr_operation_and_type_effects): Likewise.
4402 2018-11-12 Eric Botcazou <ebotcazou@adacore.com>
4404 * config/mcore/mcore.h (WORD_REGISTER_OPERATIONS): Remove duplicate.
4406 2018-11-12 Richard Biener <rguenther@suse.de>
4408 * tree-vrp.h (value_range_base::symbolic_p,
4409 value_range_base::constant_p, value_range_base::zero_p,
4410 value_range_base::singleton_p): Move from value_range.
4411 (value_range::dump): Add.
4412 * gimple-ssa-evrp-analyze.c
4413 (evrp_range_analyzer::record_ranges_from_phis): Use set_varying.
4414 * ipa-cp.c (ipcp_vr_lattice::print): Use dump_value_range.
4415 * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
4417 * tree-vrp.c (value_range::symbolic_p): Move to value_range_base.
4418 (value_range::constant_p): Likewise.
4419 (value_range::singleton_p): Likewise.
4420 (value_range_base::dump): Add.
4421 (set_value_range_to_undefined): Remove.
4422 (set_value_range_to_varying): Likewise.
4423 (range_int_cst_p): Take value_range_base argument.
4424 (range_int_cst_singleton_p): Likewise.
4425 (value_range_constant_singleton): Likewise.
4426 (vrp_set_zero_nonzero_bits): Likewise.
4427 (extract_range_from_multiplicative_op): Use set_varying.
4428 (extract_range_from_binary_expr_1): Likewise. Use set_undefined.
4429 (extract_range_from_unary_expr): Likewise.
4430 (dump_value_range_base): Change to overload of dump_value_range.
4431 (vrp_prop::vrp_initialize): Use set_varying and set_undefined.
4432 (vrp_prop::visit_stmt): Likewise.
4433 (value_range::intersect_helper): Likewise.
4434 (value_range::union_helper): Likewise.
4435 (determine_value_range_1): Likewise.
4437 2018-11-12 Richard Biener <rguenther@suse.de>
4439 * tree-vrp.c (set_value_range_to_nonnull): Clear equiv.
4440 (set_value_range_to_null): Likewise.
4441 * vr-values.c (vr_values::extract_range_from_comparison):
4442 Clear equiv for constant singleton ranges.
4444 2018-11-12 Wei Xiao <wei3.xiao@intel.com>
4446 * config/i386/sse.md: Combine VFIXUPIMM* patterns
4447 (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
4448 (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
4449 (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Remove.
4450 (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
4451 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
4452 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Remove.
4454 2018-11-11 Sandra Loosemore <sandra@codesourcery.com>
4457 * doc/extend.texi (Common Type Attributes): For the align type
4458 attribute, copy language about decreasing alignment from the
4459 corresponding variable attribute.
4461 2018-11-11 Xianmiao Qu <xianmiao_qu@c-sky.com>
4463 * config/csky/csky.md (*fpuv2_nmulsf3_1, *fpuv2_nmuldf3_1): Handle
4466 2018-11-11 Sandra Loosemore <sandra@codesourcery.com>
4469 * doc/invoke.texi (C++ Dialect Options): Add warning about mixing
4470 -frtti and -fno-rtti code.
4472 2018-11-11 Sandra Loosemore <sandra@codesourcery.com>
4475 * doc/extend.texi (Other Builtins): Document probability associated
4476 with __builtin_expect.
4478 2018-11-11 Uros Bizjak <ubizjak@gmail.com>
4481 * config/i386/i386.h (STACK_BOUNDARY): Use TARGET_64BIT_MS_ABI
4482 instead of (TARGET_64BIT && ix86_abi == MS_ABI).
4483 * config/i386/darwin.h (STACK_BOUNDARY): Ditto.
4484 * config/i386/cygming.h (STACK_BOUNDARY): Remove.
4486 2018-11-11 Xianmiao Qu <xianmiao_qu@c-sky.com>
4488 * config/csky/csky-linux-elf.h (CC1_SPEC): Support -profile.
4490 2018-11-11 Xianmiao Qu <xianmiao_qu@c-sky.com>
4492 * config/csky/csky.h (ASM_PREFERRED_EH_DATA_FORMAT): Define.
4494 2018-11-11 Richard Biener <rguenther@suse.de>
4496 * tree-vrp.h (class value_range_base): New base class for
4497 value_range containing all but the m_equiv member.
4498 (dump_value_range_base): Add.
4499 (range_includes_zero_p): Work on value_range_base.
4500 * tree-vrp.c (value_range_base::set): Split out base handling
4502 (value_range::set): this.
4503 (value_range::set_equiv): New.
4504 (value_range_base::value_range_base): New constructors.
4505 (value_range_base::check): Split out base handling from...
4506 (value_range::check): this.
4507 (value_range::equal_p): Refactor in terms of
4508 ignore_equivs_equal_p which is now member of the base.
4509 (value_range_base::set_undefined): New.
4510 (value_range_base::set_varying): Likewise.
4511 (value_range_base::dump):Split out base handling from...
4512 (value_range::dump): this.
4513 (value_range_base::set_and_canonicalize): Split out base handling
4515 (value_range::set_and_canonicalize): this.
4516 (value_range_base::union_): New.
4517 * ipa-prop.h (struct ipa_jump_func): Use value_range_base *
4519 * ipa-cp.c (class ipcp_vr_lattice): Use value_range_base
4520 instead of value_range everywhere.
4521 (ipcp_vr_lattice::print): Use dump_value_range_base.
4522 (ipcp_vr_lattice::meet_with): Adjust.
4523 (ipcp_vr_lattice::meet_with_1): Likewise.
4524 (ipa_vr_operation_and_type_effects): Likewise.
4525 (propagate_vr_across_jump_function): Likewise.
4526 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Likewise.
4527 (ipa_get_value_range): Likewise.
4528 (ipa_set_jfunc_vr): Likewise.
4529 (ipa_compute_jump_functions_for_edge): Likewise.
4531 2018-11-10 Sandra Loosemore <sandra@codesourcery.com>
4534 * doc/invoke.texi (Optimize Options): Add @opindex entries
4535 for the positive forms of -fno-xxx and -mno-xxx options
4536 that were lacking them.
4538 2018-11-10 Segher Boessenkool <segher@kernel.crashing.org>
4540 * combine.c (make_more_copies): Only make an intermediate copy if the
4541 dest of a move is a pseudo.
4543 2018-11-09 Maya Rashish <coypu@sdf.org>
4546 * config/netbsd-elf.h (NETBSD_STARTFILE_SPEC): Use crtbeginS.o for PIE.
4547 (NETBSD_ENDFILE_SPEC): Use crtendS.o for PIE.
4549 2018-11-09 Sandra Loosemore <sandra@codesourcery.com>
4553 * doc/invoke.texi (Optimize Options): Clarify default behavior
4554 for -fno-toplevel-reorder, -fno-defer-pop, and -fno-branch-count-reg.
4556 2018-11-09 Bernd Edlinger <bernd.edlinger@hotmail.de>
4558 PR tree-optimization/87940
4559 * expr.c (string_constant): Don't strip NOPS in subexpressions.
4560 Fold PLUS_EXPR correctly.
4562 2018-11-09 Ilya Leoshkevich <iii@linux.ibm.com>
4565 * config/s390/s390.c (s390_safe_relative_long_p): New function.
4566 (annotate_constant_pool_refs): Skip insns which support
4567 relative addressing.
4568 (annotate_constant_pool_refs_1): New helper function.
4569 (find_constant_pool_ref): Skip insns which support relative
4571 (find_constant_pool_ref_1): New helper function.
4572 (replace_constant_pool_ref): Skip insns which support
4573 relative addressing.
4574 (replace_constant_pool_ref_1): New helper function.
4575 (s390_mainpool_start): Adapt to the new signature.
4576 (s390_mainpool_finish): Likewise.
4577 (s390_chunkify_start): Likewise.
4578 (s390_chunkify_finish): Likewise.
4579 (pass_s390_early_mach::execute): Likewise.
4580 (s390_prologue_plus_offset): Likewise.
4581 (s390_emit_prologue): Likewise.
4582 (s390_emit_epilogue): Likewise.
4584 2018-11-09 Jakub Jelinek <jakub@redhat.com>
4586 * gimplify.c (gimplify_scan_omp_clauses): Call sorry_at for valid
4587 but unsupported lastprivate with conditional modifier.
4589 2018-11-09 Jeff Law <law@redhat.com>
4591 * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): Handle
4592 unused argument better. Add gcc_unreachable to silence warning.
4594 2018-11-09 Martin Sebor <msebor@redhat.com>
4597 * attribs.c (has_attribute): New helper function.
4598 (decls_mismatched_attributes, maybe_diag_alias_attributes): Same.
4599 * attribs.h (decls_mismatched_attributes): Declare.
4600 * cgraphunit.c (handle_alias_pairs): Call maybe_diag_alias_attributes.
4601 (maybe_diag_incompatible_alias): Use OPT_Wattribute_alias_.
4602 * common.opt (-Wattribute-alias): Take an argument.
4603 (-Wno-attribute-alias): New option.
4604 * doc/extend.texi (Common Function Attributes): Document copy.
4605 (Common Variable Attributes): Same.
4606 * doc/invoke.texi (-Wmissing-attributes): Document enhancement.
4607 (-Wattribute-alias): Document new option argument.
4609 2018-11-09 Richard Earnshaw <rearnsha@arm.com>
4611 * config/arm/parsecpu.awk (/alias/): Tighten invisible alias
4612 matching criteria. Remove unused array initializer.
4614 2018-11-09 Bill Schmidt <wschmidt@linux.ibm.com>
4615 Jinsong Ji <jji@us.ibm.com>
4617 * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix incorrect
4618 constraints by introducing a new temporary.
4619 (_mm_cvtss_si64): Likewise.
4621 2018-11-09 Martin Liska <mliska@suse.cz>
4623 * common.opt: Add -fipa-stack-alignment flag.
4624 * doc/invoke.texi: Document it.
4625 * final.c (rest_of_clean_state): Guard stack
4626 shrinking with flag.
4628 2018-11-09 Martin Liska <mliska@suse.cz>
4630 * cgraph.h (ipa_discover_readonly_nonaddressable_vars): Rename
4632 (ipa_discover_variable_flags): ... this.
4633 * common.opt: Come up with new flag -fipa-reference-addressable.
4634 * doc/invoke.texi: Document it.
4635 * ipa-reference.c (propagate): Call the renamed fn.
4636 * ipa-visibility.c (whole_program_function_and_variable_visibility):
4638 * ipa.c (ipa_discover_readonly_nonaddressable_vars): Renamed to
4640 (ipa_discover_variable_flags): ... this. Discover
4641 non-addressable variables only with the newly added flag.
4642 * opts.c: Enable the newly added flag with -O1 and higher
4645 2018-11-09 David Malcolm <dmalcolm@redhat.com>
4647 * json.cc (selftest::test_writing_literals): Fix comment.
4649 2018-11-09 Martin Liska <mliska@suse.cz>
4651 * config/aarch64/aarch64.c (aarch64_parse_arch): Do not copy
4652 string to a stack buffer.
4653 (aarch64_parse_cpu): Likewise.
4654 (aarch64_parse_tune): Likewise.
4656 2018-11-09 Richard Biener <rguenther@suse.de>
4658 PR tree-optimization/87953
4659 * tree-vect-loop.c (vectorizable_reduction): For analysis
4660 always pass ops[0] to vectorizable_condition.
4662 2018-11-09 Stafford Horne <shorne@gmail.com>
4663 Richard Henderson <rth@twiddle.net>
4664 Joel Sherrill <joel@rtems.org>
4666 * common/config/or1k/or1k-common.c: New file.
4667 * config/or1k/*: New.
4668 * config.gcc (or1k*-*-*): New.
4669 * configure.ac (or1k*-*-*): New test for openrisc tls.
4670 * configure: Regenerated.
4671 * doc/install.texi: Document OpenRISC triplets.
4672 * doc/invoke.texi: Document OpenRISC arguments.
4673 * doc/md.texi: Document OpenRISC.
4675 2018-11-09 Richard Earnshaw <rearnsha@arm.com>
4677 * config/arm/arm-cpus.in (arm7tdmi): Add an alias for arm7tdmi-s.
4678 (arm7tdmi-s): Delete CPU.
4679 (arm710t): Add aliases for arm720t and arm740t.
4680 (arm720t, arm740t): Delete CPUs.
4681 (arm920t): Add aliases for arm920, arm922t and arm940t.
4682 (arm920, arm922t, arm940t): Delete CPUs.
4683 (arm10tdmi): Add alias for arm1020t.
4684 (arm1020t): Delete CPU.
4685 (arm9e): Add aliases for arm946e-s, arm966e-s and arm968e-s.
4686 (arm946e-s, arm966e-s, arm968e-s): Delete CPUs.
4687 (arm10e): Add aliases for arm1020e and arm1022e.
4688 (arm1020e, arm1022e): Delete CPU.
4689 * config/arm/arm.md (generic_sched): Remove entries that are now
4691 (generic_vfp): Likewise.
4692 * config/arm/arm1020e.md: Simplify tuning selection based on alias
4694 * config/arm/arm-tune.md: Regenerated.
4695 * config/arm/arm-tables.opt: Regenerated.
4697 2018-11-09 Richard Biener <rguenther@suse.de>
4699 PR tree-optimization/87621
4700 * tree-vect-loop.c (vectorizable_reduction): Handle reduction
4701 op with only phi inputs.
4702 * tree-ssa-loop-ch.c: Include tree-ssa-sccvn.h.
4703 (ch_base::copy_headers): Run CSE on copied loop headers.
4704 (pass_ch_vect::process_loop_p): Simplify.
4706 2018-11-09 Alexandre Oliva <oliva@adacore.com>
4708 * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Adjust
4709 for non-w64 x86_64 biarch.
4711 2018-11-09 Alexandre Oliva <aoliva@redhat.com>
4713 PR rtl-optimization/86438
4714 * compare-elim.c (try_eliminate_compare): Use SET_SRC instead
4715 of in_b for the compare if in_b is SET_DEST.
4718 * config/i386/i386.c (ix86_const_not_ok_for_debug_p): Reject
4719 non-toplevel UNSPEC.
4721 2018-11-08 Aldy Hernandez <aldyh@redhat.com>
4723 * tree-vrp.c (value_range::check): Do not access internals
4725 (value_range::singleton_p): Same.
4726 (value_range::type): Same.
4727 (vrp_finalize): Use value_range API.
4729 2018-11-08 Aldy Hernandez <aldyh@redhat.com>
4731 * tree-vrp.c (may_contain_p): Do not access m_min/m_max directly.
4733 2018-11-08 Aldy Hernandez <aldyh@redhat.com>
4735 * vr-values.c (vr_values::get_value_range): Use value_range API
4736 instead of piecing together ranges.
4737 (vr_values::update_value_range): Same.
4739 2018-11-08 Aldy Hernandez <aldyh@redhat.com>
4741 * gimple-fold.c (size_must_be_zero_p): Use value_range API instead
4742 of performing ad-hoc calculations.
4743 * tree-ssanames.c (set_range_info): New overloaded function
4744 accepting value_range &.
4745 (get_range_info): Same.
4746 * tree-ssanames.h (set_range_info_raw): Remove.
4747 (set_range_info): New prototype.
4748 (get_range_info): Same.
4749 * tree-vrp.h (value_range::null_p): Rename to zero_p.
4750 * tree-vrp.c (value_range::null_p): Same.
4752 2018-11-09 Jan Hubicka <jh@suse.cz>
4754 * tree.c (fld_type_variant_equal_p): Test user align flag.
4755 (flt_type_variant): Copy user align flag.
4756 (fld_incomplete_type_of): Clear it.
4758 2018-11-09 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4760 * config/arm/neon.md (div<mode>3): New pattern.
4762 2018-11-08 Andi Kleen <ak@linux.intel.com>
4764 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PTWRITE_SET): New.
4765 (OPTION_MASK_ISA_PTWRITE_UNSET): New.
4766 (ix86_handle_option): Handle OPT_mptwrite.
4767 * config/i386/cpuid.h (bit_PTWRITE): Add.
4768 * config/i386/driver-i386.c (host_detect_local_cpu): Detect ptwrite.
4769 * config/i386/i386-builtin.def (BDESC): Add ptwrite32/64.
4770 * config/i386/i386-c.c (ix86_target_macros_internal): Define __PTWRITE__.
4771 * config/i386/i386.c (ix86_target_string): Handle ptwrite.
4772 (ix86_option_override_internal): Handle PTA_PTWRITE.
4773 (ix86_valid_target_attribute_inner_p): Define ptwrite.
4774 (def_builtin2): Force UINT64 to be 64bit only.
4775 * config/i386/i386.h (TARGET_PTWRITE): Add.
4776 (TARGET_PTWRITE_P): Add.
4778 * config/i386/i386.md: Define ptwrite.
4779 * config/i386/i386.opt: Add -mptwrite.
4780 * config/i386/immintrin.h (_ptwrite64): Add.
4782 * doc/extend.texi: Document __builtin_ia32_ptwrite*.
4783 * doc/invoke.texi: Document -mptwrite.
4785 2018-11-08 Peter Bergner <bergner@linux.ibm.com>
4787 PR rtl-optimization/87600
4788 * cfgexpand.c (expand_asm_stmt): Catch illegal asm constraint usage.
4789 * lra-constraints.c (process_alt_operands): Skip illegal hard
4790 register usage. Prefer reloading non hard register operands.
4792 2018-11-08 Sandra Loosemore <sandra@codesourcery.com>
4795 * doc/invoke.texi (Optimize Options): Clarify default behavior
4796 for -fno-sched-interblock and -fno-sched-spec.
4798 2018-11-08 Roman Geissler <roman.geissler@amadeus.com>
4800 * collect2.c (linker_select): Add USE_LLD_LD.
4801 (ld_suffixes): Add ld.lld.
4802 (main): Handle -fuse-ld=lld.
4803 * common.opt (-fuse-ld=lld): New option.
4804 * doc/invoke.texi (-fuse-ld=lld): Document.
4805 * opts.c (common_handle_option): Handle OPT_fuse_ld_lld.
4807 2018-11-08 Paul Koning <ni1d@arrl.net>
4809 * config/pdp11/constraints.md: Add "Z" series constraints for use
4810 with pre-dec and post-inc addressing.
4811 * config/pdp11/pdp11-protos.m (expand_block_move): Delete.
4812 (pdp11_expand_operands): Add int argument (word count).
4813 (pdp11_sp_frame_offset): Delete.
4814 (pdp11_cmp_length): New function.
4815 (pushpop_regeq): New function.
4816 * config/pdp11/pdp11.c (TARGET_STACK_PROTECT_RUNTIME_ENABLED_P):
4818 (pdp11_expand_prologue, pdp11_expand_epilogue): Rewrite for new
4820 (pdp11_initial_elimination_offset): Ditto.
4821 (pdp11_expand_operands): Add word count argument. Bugfixes.
4822 (output_move_multiple): Change how pointer adjustment is done.
4823 (pdp11_gen_int_label): Correct format.
4824 (output_ascii): Ditto.
4825 (pdp11_asm_output_var): Add code for DEC assembler case.
4826 (pdp11_asm_print_operand): Bugfix for CONST_DOUBLE holding integer
4828 (legitimate_const_double_p): Ditto.
4829 (pdp11_register_move_cost): Adjust for new register classes.
4830 (pdp11_regno_reg_class): Ditto.
4831 (expand_block_move): Delete.
4832 (pushpop_regeq): New function.
4833 (pdp11_legitimate_address_p): Bugfix in check for constant
4835 (pdp11_sp_frame_offset): Delete.
4836 (pdp11_reg_save_size): New helper function for new frame layout.
4837 (output_addr_const_pdp11): Remove CONST_DOUBLE case.
4838 (pdp11_expand_shift): Bugfix in check for constant shift count.
4839 (pdp11_shift_length): Ditto.
4840 (pdp11_assemble_shift): Copy input to pdp11_expand_operands.
4841 (pdp11_cmp_length): New function.
4842 * config/pdp11/pdp11.h (TARGET_CPU_CPP_BUILTINS): Add macros for
4843 some compile options.
4844 (FIXED_REGISTERS): Remove HARD_FRAME_POINTER_REGNUM.
4845 (CALL_USED_REGISTERS): Ditto.
4846 (ELIMINABLE_REGS): Ditto.
4847 (REGISTER_NAMES): Ditto.
4848 (reg_class): Add classes NOTR0_REG through NOTSP_REG for use by Z
4850 (REG_CLASS_NAMES): Ditto.
4851 (REG_CLASS_CONTENTS): Ditto. Also remove
4852 HARD_FRAME_POINTER_REGNUM.
4853 (CPU_REG_CLASS): New macro.
4854 (CLASS_MAX_NREGS): Adjust for new register classes.
4855 (FUNCTION_PROFILER): Make no-op.
4856 (may_call_alloca): Remove unused declaration.
4857 (ASM_OUTPUT_ALIGN): Add workaround for PR87795.
4858 (ASM_OUTPUT_SKIP): Fix format.
4859 * config/pdp11/pdp11.md (unspecv): Add UNSPECV_MOVMEM.
4860 (HARD_FRAME_POINTER_REGNUM): Remove.
4862 (*rts): Rename. Remove epilogue related checks.
4863 (cmpsi, cmpdi): New insn.
4864 (cbranch<mode>4): Change to apply to SI and DI modes as well.
4865 (mov<mode>): Change constraints to enforce that push/pop
4866 destination cannot use the same register as source.
4867 (*mov<mode><cc_cc>): Ditto.
4868 (movmemhi, movmemhi1, movmemhi_nocc): Change to expand block move
4869 at assembly output rather than as RTL expander.
4870 (zero_extendqihi2): Bugfix in check for same registers.
4871 (adddi3_nocc): Bugfix in check for constant operand.
4872 (addsi3_nocc): Ditto.
4873 (subdi3_nocc): Ditto.
4874 (subsi3_nocc): Ditto.
4875 (negdi2_nocc): Copy input to pdp11_expand_operands.
4876 (negsi2_nocc): Ditto.
4877 (bswap2_nocc): Ditto.
4878 * config/pdp11/pdp11.opt (mlra): Fix documentation.
4879 * config/pdp11/t-pdp11: Use -Os.
4881 2018-11-08 Richard Earnshaw <rearnsha@arm.com>
4883 * config/arm/parsecpu.awk (/alias/): New parsing rule.
4884 (/begin cpu/): Check that the cpu name hasn't been previously defined.
4885 (gen_comm_data): Print out CPU alias tables.
4886 (check_cpu): Match aliases when checking the CPU name.
4887 * config/arm/arm-protos.h (cpu_alias): New structure.
4888 (cpu_option): Add entry for aliases.
4889 * config/arm/arm-cpus.in (strongarm): Add aliases for strongarm110
4890 strongarm1100 and strongarm1110.
4891 (strongarm110, strongarm1100, strongarm1110): Delete CPU entries.
4892 (config/arm/arm-generic.md): Remove redundant references to
4893 strongarm110, strongarm1100 and strongarm1110.
4894 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4895 Scan aliases for additional hints.
4896 (arm_parse_cpu_option_name): Also match a cpu name against the list
4898 * config/arm/arm-tables.opt: Regenerated.
4899 * config/arm/arm-tune.md: Regenerated.
4901 2018-11-08 Jakub Jelinek <jakub@redhat.com>
4903 * builtin-types.def (BT_FN_VOID_BOOL, BT_FN_VOID_SIZE_SIZE_PTR,
4904 BT_FN_UINT_UINT_PTR_PTR, BT_FN_UINT_OMPFN_PTR_UINT_UINT,
4905 BT_FN_BOOL_UINT_LONGPTR_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
4906 BT_FN_BOOL_UINT_ULLPTR_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR,
4907 BT_FN_BOOL_LONG_LONG_LONG_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
4908 BT_FN_BOOL_BOOL_ULL_ULL_ULL_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR): New.
4909 * gengtype.c (open_base_files): Add omp-general.h.
4910 * gimple.c (gimple_build_omp_critical):
4911 (gimple_build_omp_taskgroup): Add CLAUSES argument. Call
4912 gimple_omp_taskgroup_set_clauses.
4913 (gimple_build_omp_atomic_load): Add mo argument, call
4914 gimple_omp_atomic_set_memory_order.
4915 (gimple_build_omp_atomic_store): Likewise.
4916 (gimple_copy): Adjust handling of GIMPLE_OMP_TASKGROUP.
4917 * gimple.def (GIMPLE_OMP_TASKGROUP): Use GSS_OMP_SINGLE_LAYOUT
4919 (GIMPLE_OMP_TEAMS): Use GSS_OMP_PARALLEL_LAYOUT instead
4920 of GSS_OMP_SINGLE_LAYOUT, adjust comments.
4921 * gimple.h (enum gf_mask): Add GF_OMP_TEAMS_HOST, GF_OMP_TASK_TASKWAIT
4922 and GF_OMP_ATOMIC_MEMORY_ORDER. Remove GF_OMP_ATOMIC_SEQ_CST, use
4923 different value for GF_OMP_ATOMIC_NEED_VALUE.
4924 (struct gimple_statement_omp_taskreg): Add GIMPLE_OMP_TEAMS to
4926 (struct gimple_statement_omp_single_layout): And remove here.
4927 (struct gomp_teams): Inherit from gimple_statement_omp_taskreg rather
4928 than gimple_statement_omp_single_layout.
4929 (is_a_helper <gimple_statement_omp_taskreg *>::test): Allow
4931 (is_a_helper <const gimple_statement_omp_taskreg *>::test): Likewise.
4932 (gimple_omp_subcode): Formatting fix.
4933 (gimple_omp_teams_child_fn, gimple_omp_teams_child_fn_ptr,
4934 gimple_omp_teams_set_child_fn, gimple_omp_teams_data_arg,
4935 gimple_omp_teams_data_arg_ptr, gimple_omp_teams_set_data_arg,
4936 gimple_omp_teams_host, gimple_omp_teams_set_host,
4937 gimple_omp_task_taskwait_p, gimple_omp_task_set_taskwait_p,
4938 gimple_omp_taskgroup_clauses, gimple_omp_taskgroup_clauses_ptr,
4939 gimple_omp_taskgroup_set_clauses): New inline functions.
4940 (gimple_build_omp_atomic_load): Add enum omp_memory_order argument.
4941 (gimple_build_omp_atomic_store): Likewise.
4942 (gimple_omp_atomic_seq_cst_p): Remove.
4943 (gimple_omp_atomic_memory_order): New function.
4944 (gimple_omp_atomic_set_seq_cst): Remove.
4945 (gimple_omp_atomic_set_memory_order): New function.
4946 (gimple_build_omp_taskgroup): Add clauses argument.
4947 * gimple-pretty-print.c (dump_gimple_omp_taskgroup): New function.
4948 (dump_gimple_omp_task): Print taskwait with depend clauses.
4949 (dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Use
4950 dump_omp_atomic_memory_order.
4951 (pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP.
4952 * gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP_ALLOC_ONLY,
4953 GOVD_MAP_FROM_ONLY and GOVD_NONTEMPORAL.
4954 (enum omp_region_type): Reserve bits 1 and 2 for auxiliary flags,
4955 renumber values of most of ORT_* enumerators, add ORT_HOST_TEAMS,
4956 ORT_COMBINED_HOST_TEAMS, ORT_TASKGROUP, ORT_TASKLOOP and
4957 ORT_UNTIED_TASKLOOP enumerators.
4958 (enum gimplify_defaultmap_kind): New.
4959 (struct gimplify_omp_ctx): Remove target_map_scalars_firstprivate and
4960 target_map_pointers_as_0len_arrays members, add defaultmap.
4961 (new_omp_context): Initialize defaultmap member.
4962 (gimple_add_tmp_var): Handle ORT_TASKGROUP like ORT_WORKSHARE.
4963 (maybe_fold_stmt): Don't fold even in host teams regions.
4964 (omp_firstprivatize_variable): Handle ORT_TASKGROUP like
4965 ORT_WORKSHARE. Test ctx->defaultmap[GDMK_SCALAR] instead of
4966 ctx->omp_firstprivatize_variable.
4967 (omp_add_variable): Don't add private/firstprivate for VLAs in
4969 (omp_default_clause): Print "taskloop" rather than "task" if
4971 (omp_notice_variable): Handle ORT_TASKGROUP like ORT_WORKSHARE.
4972 Handle new defaultmap clause kinds.
4973 (omp_is_private): Handle ORT_TASKGROUP like ORT_WORKSHARE. Allow simd
4974 iterator to be lastprivate or private. Fix up diagnostics if linear
4975 is used on collapse>1 simd iterator.
4976 (omp_check_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.
4977 (gimplify_omp_depend): New function.
4978 (gimplify_scan_omp_clauses): Add shared clause on parallel for
4979 combined parallel master taskloop{, simd} if taskloop has
4980 firstprivate, lastprivate or reduction clause. Handle
4981 OMP_CLAUSE_REDUCTION_TASK diagnostics. Adjust tests for
4982 ORT_COMBINED_TEAMS. Gimplify depend clauses with iterators. Handle
4983 cancel and simd OMP_CLAUSE_IF_MODIFIERs. Handle
4984 OMP_CLAUSE_NONTEMPORAL. Handle new defaultmap clause kinds. Handle
4985 OMP_CLAUSE_{TASK,IN}_REDUCTION. Diagnose invalid conditional
4987 (gimplify_adjust_omp_clauses_1): Ignore GOVD_NONTEMPORAL. Handle
4988 GOVD_MAP_ALLOC_ONLY and GOVD_MAP_FROM_ONLY.
4989 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_NONTEMPORAL. Handle
4990 OMP_CLAUSE_{TASK,IN}_REDUCTION.
4991 (gimplify_omp_task): Handle taskwait with depend clauses.
4992 (gimplify_omp_for): Add shared clause on parallel for combined
4993 parallel master taskloop{, simd} if taskloop has firstprivate,
4994 lastprivate or reduction clause. Use ORT_TASKLOOP or
4995 ORT_UNTIED_TASKLOOP instead of ORT_TASK or ORT_UNTIED_TASK. Adjust
4996 tests for ORT_COMBINED_TEAMS. Handle C++ range for loops with
4997 NULL TREE_PURPOSE in OMP_FOR_ORIG_DECLS. Firstprivatize
4998 __for_end and __for_range temporaries on OMP_PARALLEL for
4999 distribute parallel for{, simd}. Move OMP_CLAUSE_REDUCTION
5000 and OMP_CLAUSE_IN_REDUCTION from taskloop to the task construct
5001 sandwiched in between two taskloops.
5002 (computable_teams_clause): Test ctx->defaultmap[GDMK_SCALAR]
5003 instead of ctx->omp_firstprivatize_variable.
5004 (gimplify_omp_workshare): Set ort to ORT_HOST_TEAMS or
5005 ORT_COMBINED_HOST_TEAMS if not inside of target construct. If
5006 host teams, use gimplify_and_return_first etc. for body like
5007 for target or target data constructs, and at the end call
5008 gimple_omp_teams_set_host on the GIMPLE_OMP_TEAMS object.
5009 (gimplify_omp_atomic): Use OMP_ATOMIC_MEMORY_ORDER instead
5010 of OMP_ATOMIC_SEQ_CST, pass it as new argument to
5011 gimple_build_omp_atomic_load and gimple_build_omp_atomic_store, remove
5012 gimple_omp_atomic_set_seq_cst calls.
5013 (gimplify_expr) <case OMP_TASKGROUP>: Move handling into a separate
5014 case, handle taskgroup clauses.
5015 * lto-streamer-out.c (hash_tree): Handle
5016 OMP_CLAUSE_{TASK,IN}_REDUCTION.
5017 * Makefile.in (GTFILES): Add omp-general.h.
5018 * omp-builtins.def (BUILT_IN_GOMP_TASKWAIT_DEPEND,
5019 BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
5020 BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
5021 BUILT_IN_GOMP_LOOP_START, BUILT_IN_GOMP_LOOP_ORDERED_START,
5022 BUILT_IN_GOMP_LOOP_DOACROSS_START,
5023 BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_NEXT,
5024 BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
5025 BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
5026 BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
5027 BUILT_IN_GOMP_LOOP_ULL_START, BUILT_IN_GOMP_LOOP_ULL_ORDERED_START,
5028 BUILT_IN_GOMP_LOOP_ULL_DOACROSS_START,
5029 BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_NEXT,
5030 BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
5031 BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
5032 BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME,
5033 BUILT_IN_GOMP_PARALLEL_REDUCTIONS, BUILT_IN_GOMP_SECTIONS2_START,
5034 BUILT_IN_GOMP_TEAMS_REG, BUILT_IN_GOMP_TASKGROUP_REDUCTION_REGISTER,
5035 BUILT_IN_GOMP_TASKGROUP_REDUCTION_UNREGISTER,
5036 BUILT_IN_GOMP_TASK_REDUCTION_REMAP,
5037 BUILT_IN_GOMP_WORKSHARE_TASK_REDUCTION_UNREGISTER): New builtins.
5038 * omp-expand.c (workshare_safe_to_combine_p): Return false for
5039 non-worksharing loops.
5040 (omp_adjust_chunk_size): Don't adjust anything if chunk_size is zero.
5041 (determine_parallel_type): Don't combine parallel with worksharing
5042 which has _reductemp_ clause.
5043 (expand_parallel_call): Emit the GOMP_*nonmonotonic_runtime* or
5044 GOMP_*maybe_nonmonotonic_runtime* builtins instead of GOMP_*runtime*
5045 if there is nonmonotonic modifier or if there is no modifier and no
5046 ordered clause. For dynamic and guided schedule without monotonic
5047 and nonmonotonic modifier, default to nonmonotonic.
5048 (expand_omp_for): Likewise. Adjust expand_omp_for_generic caller, use
5049 GOMP_loop{,_ull}{,_ordered,_doacross}_start builtins if there are
5051 (expand_task_call): Add GOMP_TASK_FLAG_REDUCTION flag to flags if
5052 there are any reduction clauses.
5053 (expand_taskwait_call): New function.
5054 (expand_teams_call): New function.
5055 (expand_omp_taskreg): Allow GIMPLE_OMP_TEAMS and call
5056 expand_teams_call for it. Formatting fix. Handle taskwait with
5058 (expand_omp_for_generic): Add SCHED_ARG argument. Handle expansion
5059 of worksharing loops with task reductions.
5060 (expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Handle
5061 expansion of worksharing loops with task reductions.
5062 (expand_omp_sections): Handle expansion of sections with task
5064 (expand_omp_synch): For host teams call expand_omp_taskreg.
5065 (omp_memory_order_to_memmodel): New function.
5066 (expand_omp_atomic_load, expand_omp_atomic_store,
5067 expand_omp_atomic_fetch_op): Use it and gimple_omp_atomic_memory_order
5068 instead of gimple_omp_atomic_seq_cst_p.
5069 (build_omp_regions_1, omp_make_gimple_edges): Treat taskwait with
5070 depend clauses as a standalone directive.
5071 * omp-general.c (enum omp_requires): New variable.
5072 (omp_extract_for_data): Initialize have_reductemp member. Allow
5073 NE_EXPR even in OpenMP loops, transform them into LT_EXPR or
5074 GT_EXPR loops depending on incr sign. Formatting fixes.
5075 * omp-general.h (struct omp_for_data): Add have_reductemp member.
5076 (enum omp_requires): New enum.
5077 (omp_requires_mask): Declare.
5078 * omp-grid.c (grid_eliminate_combined_simd_part): Formatting fix.
5080 * omp-low.c (struct omp_context): Add task_reductions and
5081 task_reduction_map fields.
5082 (is_host_teams_ctx): New function.
5083 (is_taskreg_ctx): Return true also if is_host_teams_ctx.
5084 (use_pointer_for_field): Use is_global_var instead of
5085 TREE_STATIC || DECL_EXTERNAL, and apply only if not privatized
5087 (build_outer_var_ref): Ignore taskgroup outer contexts.
5088 (delete_omp_context): Release task_reductions and task_reduction_map.
5089 (scan_sharing_clauses): Don't add any fields for reduction clause on
5090 taskloop. Handle OMP_CLAUSE__REDUCTEMP_. Handle
5091 OMP_CLAUSE_{IN,TASK}_REDUCTION and OMP_CLAUSE_REDUCTION with task
5092 modifier. Don't ignore shared clauses in is_host_teams_ctx contexts.
5093 Handle OMP_CLAUSE_NONTEMPORAL.
5094 (add_taskreg_looptemp_clauses): Add OMP_CLAUSE__REDUCTEMP_ clause if
5096 (scan_omp_parallel): Add _reductemp_ clause if there are any reduction
5097 clauses with task modifier.
5098 (scan_omp_task): Handle taskwait with depend clauses.
5099 (finish_taskreg_scan): Move field corresponding to _reductemp_ clause
5100 first. Move also OMP_CLAUSE__REDUCTEMP_ clause in front if present.
5101 Handle GIMPLE_OMP_TEAMS like GIMPLE_OMP_PARALLEL.
5102 (scan_omp_for): Fix comment formatting.
5103 (scan_omp_teams): Handle host teams constructs.
5104 (check_omp_nesting_restrictions): Allow teams with no outer
5105 OpenMP context. Adjust diagnostics for teams strictly nested into
5106 some explicit OpenMP construct other than target. Allow OpenMP atomics
5107 inside of simd regions.
5108 (scan_omp_1_stmt): Call scan_sharing_clauses for taskgroups.
5109 (scan_omp_1_stmt) <case GIMPLE_OMP_TEAMS>: Temporarily bump
5110 taskreg_nesting_level while scanning host teams construct.
5111 (task_reduction_read): New function.
5112 (lower_rec_input_clauses): Handle OMP_CLAUSE_REDUCTION on taskloop
5113 construct. Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE__REDUCTEMP_
5114 clauses. Handle OMP_CLAUSE_REDUCTION with task modifier. Remove
5115 second argument create_tmp_var if it is NULL. Don't ignore shared
5116 clauses in is_host_teams_ctx contexts. Handle
5117 OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE on OMP_CLAUSE_FIRSTPRIVATE
5119 (lower_reduction_clauses): Ignore reduction clauses with task
5120 modifier. Remove second argument create_tmp_var if it is NULL.
5121 Initialize OMP_ATOMIC_MEMORY_ORDER to relaxed.
5122 (lower_send_clauses): Ignore reduction clauses with task modifier.
5123 Handle OMP_CLAUSE__REDUCTEMP_. Don't send anything for
5124 OMP_CLAUSE_REDUCTION on taskloop. Handle OMP_CLAUSE_IN_REDUCTION.
5125 (maybe_add_implicit_barrier_cancel): Add OMP_RETURN argument, don't
5126 rely that it is the last stmt in body so far. Ignore outer taskgroup
5128 (omp_task_reductions_find_first, omp_task_reduction_iterate,
5129 lower_omp_task_reductions): New functions.
5130 (lower_omp_sections): Handle reduction clauses with taskgroup
5131 modifiers. Adjust maybe_add_implicit_barrier_cancel caller.
5132 (lower_omp_single): Adjust maybe_add_implicit_barrier_cancel caller.
5133 (lower_omp_for): Likewise. Handle reduction clauses with taskgroup
5135 (lower_omp_taskgroup): Handle taskgroup reductions.
5136 (create_task_copyfn): Copy over OMP_CLAUSE__REDUCTEMP_ pointer.
5137 Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE_REDUCTION clauses.
5138 (lower_depend_clauses): If there are any
5139 OMP_CLAUSE_DEPEND_DEPOBJ or OMP_CLAUSE_DEPEND_MUTEXINOUTSET
5140 depend clauses, use a new array format. If OMP_CLAUSE_DEPEND_LAST is
5141 seen, assume lowering is done already and return early. Set kind
5142 on artificial depend clause to OMP_CLAUSE_DEPEND_LAST.
5143 (lower_omp_taskreg): Handle reduction clauses with task modifier on
5144 parallel construct. Handle reduction clause on taskloop construct.
5145 Handle taskwait with depend clauses.
5146 (lower_omp_1): Use lower_omp_taskreg instead of lower_omp_teams
5147 for host teams constructs.
5148 * tree.c (omp_clause_num_ops): Add in_reduction, task_reduction,
5149 nontemporal and _reductemp_ clause entries.
5150 (omp_clause_code_name): Likewise.
5151 (walk_tree_1): Handle OMP_CLAUSE_{IN,TASK}_REDUCTION,
5152 OMP_CLAUSE_NONTEMPORAL and OMP_CLAUSE__REDUCTEMP_.
5153 * tree-core.h (enum omp_clause_code): Add
5154 OMP_CLAUSE_{{IN,TASK}_REDUCTION,NONTEMPORAL,_REDUCTEMP_}.
5155 (enum omp_clause_defaultmap_kind, enum omp_memory_order): New.
5156 (struct tree_base): Add omp_atomic_memory_order field into union.
5157 Remove OMP_ATOMIC_SEQ_CST comment.
5158 (enum omp_clause_depend_kind): Add OMP_CLAUSE_DEPEND_MUTEXINOUTSET
5159 and OMP_CLAUSE_DEPEND_DEPOBJ.
5160 (struct tree_omp_clause): Add subcode.defaultmap_kind.
5161 * tree.def (OMP_TASKGROUP): Add another operand, move next to other
5162 OpenMP constructs with body and clauses operands.
5163 * tree.h (OMP_BODY): Use OMP_MASTER instead of OMP_TASKGROUP.
5164 (OMP_CLAUSES): Use OMP_TASKGROUP instead of OMP_SINGLE.
5165 (OMP_TASKGROUP_CLAUSES): Define.
5166 (OMP_CLAUSE_DECL): Use OMP_CLAUSE__REDUCTEMP_ instead of
5167 OMP_CLAUSE__LOOPTEMP_.
5168 (OMP_ATOMIC_SEQ_CST): Remove.
5169 (OMP_ATOMIC_MEMORY_ORDER, OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE,
5170 OMP_CLAUSE_LASTPRIVATE_CONDITIONAL): Define.
5171 (OMP_CLAUSE_REDUCTION_CODE, OMP_CLAUSE_REDUCTION_INIT,
5172 OMP_CLAUSE_REDUCTION_MERGE, OMP_CLAUSE_REDUCTION_PLACEHOLDER,
5173 OMP_CLAUSE_REDUCTION_DECL_PLACEHOLDER,
5174 OMP_CLAUSE_REDUCTION_OMP_ORIG_REF): Handle
5175 OMP_CLAUSE_{,IN_,TASK_}REDUCTION.
5176 (OMP_CLAUSE_REDUCTION_TASK, OMP_CLAUSE_REDUCTION_INSCAN,
5177 OMP_CLAUSE_DEFAULTMAP_KIND, OMP_CLAUSE_DEFAULTMAP_CATEGORY,
5178 OMP_CLAUSE_DEFAULTMAP_BEHAVIOR, OMP_CLAUSE_DEFAULTMAP_SET_KIND):
5180 * tree-inline.c (remap_gimple_stmt): Remap taskgroup clauses.
5181 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
5182 OMP_CLAUSE__REDUCTEMP_, OMP_CLAUSE_NONTEMPORAL.
5183 (convert_local_omp_clauses): Likewise. Remove useless test.
5184 * tree-parloops.c (create_call_for_reduction_1): Pass
5185 OMP_MEMORY_ORDER_RELAXED as new argument to
5186 dump_gimple_omp_atomic_load and dump_gimple_omp_atomic_store.
5187 * tree-pretty-print.c (dump_omp_iterators): New function.
5188 (dump_omp_clause): Handle OMP_CLAUSE__REDUCTEMP_,
5189 OMP_CLAUSE_NONTEMPORAL, OMP_CLAUSE_{TASK,IN}_REDUCTION. Print
5190 reduction modifiers. Handle OMP_CLAUSE_DEPEND_DEPOBJ and
5191 OMP_CLAUSE_DEPEND_MUTEXINOUTSET. Print iterators in depend clauses.
5192 Print __internal__ for OMP_CLAUSE_DEPEND_LAST. Handle cancel and
5193 simd OMP_CLAUSE_IF_MODIFIERs. Handle new kinds of
5194 OMP_CLAUSE_DEFAULTMAP. Print conditional: for
5195 OMP_CLAUSE_LASTPRIVATE_CONDITIONAL.
5196 (dump_omp_atomic_memory_order): New function.
5197 (dump_generic_node): Use it. Print taskgroup clauses. Print
5198 taskwait with depend clauses.
5199 * tree-pretty-print.h (dump_omp_atomic_memory_order): Declare.
5200 * tree-streamer-in.c (unpack_ts_omp_clause_value_fields):
5201 Handle OMP_CLAUSE_{TASK,IN}_REDUCTION.
5202 * tree-streamer-out.c (pack_ts_omp_clause_value_fields,
5203 write_ts_omp_clause_tree_pointers): Likewise.
5205 2018-11-08 David Malcolm <dmalcolm@redhat.com>
5208 * doc/invoke.texi (-fdump-ipa-): Document the "-optimized",
5209 "-missed", "-note", and "-all" sub-options.
5210 * ipa-inline.c (caller_growth_limits): Port from fprintf to dump
5212 (can_early_inline_edge_p): Likewise.
5213 (want_early_inline_function_p): Likewise.
5214 (want_inline_self_recursive_call_p): Likewise.
5215 (recursive_inlining): Likewise.
5216 (inline_small_functions): Likewise.
5217 (flatten_function): Likewise.
5218 (ipa_inline): Likewise.
5219 (inline_always_inline_functions): Likewise.
5220 (early_inline_small_functions): Likewise.
5221 (early_inliner): Likewise.
5222 * tree-inline.c (expand_call_inline): Likewise.
5224 2018-11-08 David Malcolm <dmalcolm@redhat.com>
5226 * pretty-print.c (pp_format): Handle %f.
5227 (selftest::test_pp_format): Add test of %f.
5228 * pretty-print.h (pp_double): New macro.
5230 2018-11-08 David Malcolm <dmalcolm@redhat.com>
5232 * dump-context.h (ASSERT_IS_CGRAPH_NODE): New macro.
5233 * dumpfile.c (make_item_for_dump_cgraph_node): Move to before...
5234 (dump_pretty_printer::decode_format): Implement "%C" for
5236 (selftest::test_capture_of_dump_calls): Rename "where" to
5237 "stmt_loc". Convert test_decl to a function decl and set its
5238 location. Add a symbol_table_test RAII instance and a
5239 cgraph_node, using it to test "%C" and dump_symtab_node.
5241 2018-11-08 Eric Botcazou <ebotcazou@adacore.com>
5244 * cgraphclones.c (duplicate_thunk_for_node): Also set DECL_IGNORED_P.
5246 2018-11-08 David Malcolm <dmalcolm@redhat.com>
5248 * cgraph.c: Include "selftest.h".
5249 (saved_symtab): New variable.
5250 (selftest::symbol_table_test::symbol_table_test): New ctor.
5251 (selftest::symbol_table_test::~symbol_table_test): New dtor.
5252 (selftest::test_symbol_table_test): New test.
5253 (selftest::cgraph_c_tests): New.
5254 * cgraph.h (saved_symtab): New decl.
5255 (selftest::symbol_table_test): New class.
5256 * selftest-run-tests.c (selftest::run_tests): Call
5257 selftest::cgraph_c_tests.
5258 * selftest.h (selftest::cgraph_c_tests): New decl.
5260 2018-11-08 Richard Biener <rguenther@suse.de>
5262 * tree-data-ref.h (lambda_int): New typedef.
5263 (lambda_vector_gcd): Adjust.
5264 (lambda_vector_new): Likewise.
5265 (lambda_matrix_new): Likewise.
5266 * tree-data-ref.c (print_lambda_vector): Adjust.
5268 2018-11-08 Richard Biener <rguenther@suse.de>
5270 PR tree-optimization/87929
5271 * tree-complex.c (expand_complex_comparison): Clean EH.
5273 2018-11-08 Martin Liska <mliska@suse.cz>
5275 * doc/extend.texi: Reword.
5276 * predict.c (expr_expected_value_1): Likewise.
5278 2018-11-08 Richard Biener <rguenther@suse.de>
5280 PR tree-optimization/87913
5281 * tree-ssa-phiopt.c (minmax_replacement): Turn EQ/NE compares
5282 of extreme values to ordered comparisons.
5284 2018-11-07 Sandra Loosemore <sandra@codesourcery.com>
5287 * doc/invoke.texi (Code Gen Options): Clarify -fno-common behavior.
5289 2018-11-07 Sandra Loosemore <sandra@codesourcery.com>
5291 * doc/invoke.texi: Remove leading dash from @opindex entries
5292 throughout the file.
5294 2018-11-07 Sandra Loosemore <sandra@codesourcery.com>
5297 * doc/invoke.texi (Option Summary): Add -e and --entry.
5298 (Link Options): Likewise.
5300 2018-11-07 Nathan Sidwell <nathan@acm.org>
5303 * Makefile.in (bitmap.o-warn): Add -Wno-error to unbreak
5304 --disable-checking bootstrap.
5306 2018-11-07 Hafiz Abid Qadeer <abidh@codesourcery.com>
5308 * configure: Regenerated.
5310 2018-11-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5313 * stor-layout.c (compute_record_mode): Set TYPE_MODE of UNION_TYPE
5314 to the mode of the widest field iff the widest field has mode class
5315 MODE_INT, or MODE_PARTIAL_INT and the union would be passed by
5318 2018-11-07 Nikolai Merinov <n.merinov@inango-systems.com>
5320 * common.opt: Add -Wattribute-warning.
5321 * doc/invoke.texi: Add documentation for -Wno-attribute-warning.
5322 * expr.c (expand_expr_real_1): Add new attribute to warning_at
5323 call to allow user configure behavior of "warning" attribute.
5325 2018-11-07 Segher Boessenkool <segher@kernel.crashing.org>
5327 * target.def: Put @: after every vs., e.g., and i.e. where it is
5328 followed by whitespace.
5329 * doc/extend.texi: Ditto.
5330 * doc/fragments.texi: Ditto.
5331 * doc/gimple.texi: Ditto.
5332 * doc/implement-c.texi: Ditto.
5333 * doc/install.texi: Ditto.
5334 * doc/invoke.texi: Ditto.
5335 * doc/md.texi: Ditto.
5336 * doc/plugins.texi: Ditto.
5337 * doc/rtl.texi: Ditto.
5338 * doc/sourcebuild.texi: Ditto.
5339 * doc/tm.texi.in: Ditto.
5340 * doc/ux.texi: Ditto.
5341 * doc/tm.texi: Regenerate.
5343 2018-11-07 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5345 * config/arm/arm-cpus.in (ares): New entry.
5346 * config/arm/arm-tables.opt: Regenerate.
5347 * config/arm/arm-tune.md: Likewise.
5348 * doc/invoke.texi (ARM Options): Document ares.
5350 2018-11-07 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5352 * config/aarch64/aarch64-cores.def (ares): Define.
5353 * config/aarch64/aarch64-tune.md: Regenerate.
5354 * doc/invoke.texi (AArch64 Options): Document ares value for mtune.
5356 2018-11-07 Jan Hubicka <jh@suse.cz>
5358 * tree.c (fld_incomplete_type_of): Clear TREE_ADDRESSABLE.
5359 (free_lang_data_in_decl): Set TREE_ADDRESSABLE for public vars and
5360 functions; clear TYPE_DECL_SUPPRESS_DEBUG and DECL_MODE for
5363 2018-11-07 Richard Biener <rguenther@suse.de>
5365 PR tree-optimization/87914
5366 * tree-vect-loop.c (vect_is_simple_reduction): Improve detection
5368 (vectorizable_reduction): Handle shifts and rotates by dispatching
5369 to vectorizable_shift.
5370 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Handle
5371 in-loop uses of vect_nested_cycle defs. Merge cycle and internal
5373 (vectorizable_shift): Export and handle being called as
5375 (vect_analyze_stmt): Call vectorizable_shift after
5376 vectorizable_reduction.
5377 * tree-vectorizer.h (vectorizable_shift): Declare.
5379 2018-11-07 Jan Hubicka <jh@suse.cz>
5381 * ipa-devirt.c (odr_types_equivalent_p): Expect constants
5382 than const decls in TREE_VALUE of enum.
5383 (dump_type_inheritance_graph): Improve duplicate dumping.
5384 (free_enum_values): New.
5385 (build_type_inheritance_graph): Use it.
5386 * tree.c (free_lang_data_in_type): Free TYPE_VALUES of enums
5387 which are not main variants or not ODR types.
5388 (verify_type_variant): Expect variants to have no TYPE_VALUES.
5390 2018-11-07 Richard Biener <rguenther@suse.de>
5392 * ipa-inline.c (want_inline_small_function_p): Compute
5393 big_speedup_p lazily and last.
5395 2018-11-07 Jan Hubicka <jh@suse.cz>
5397 * tree.c (fld_type_variant_equal_p): Skip TYPE_ALIGN check when
5398 building incomplete variant of complete type.
5399 (fld_type_variant): Do not copy TYPE_ALIGN when building incomplete
5400 variant of complete type.
5402 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
5404 * config/mips/mips.c: Fix typo in documentation of
5405 mips_loongson_ext2_prefetch_cookie.
5406 (mips_option_override): fix brain twister logical.
5407 * config/mips/mips.h: Fix typo in documentation of
5408 ISA_HAS_CTZ_CTO and define pattern.
5409 * config/mips/mips.md (prefetch): Hoist EXT2 above
5411 (prefetch_indexed): Hoist EXT2 above the EXT block.
5413 2018-11-07 Jan Hubicka <jh@suse.cz>
5415 * tree.c (free_lang_data_in_type): Add fld parameter; simplify
5416 return and parameter types of function and method types.
5417 (free_lang_data_in_cgraph): Update.
5419 2018-11-07 Martin Liska <mliska@suse.cz>
5421 PR rtl-optimization/87868
5422 * postreload-gcse.c (eliminate_partially_redundant_load): Set
5423 threshold to max_count if we would overflow.
5424 * profile-count.h: Make max_count a public constant.
5426 2018-11-07 Martin Liska <mliska@suse.cz>
5428 * mem-stats.h: Fix GNU coding style.
5430 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
5432 * config/mips/gs264e.md: New.
5433 * config/mips/mips-cpus.def: Define gs264e.
5434 * config/mips/mips-tables.opt: Regenerate.
5435 * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
5437 (mips_issue_rate): Add support for gs264e.
5438 (mips_multipass_dfa_lookahead): Likewise.
5439 * config/mips/mips.h: Define TARGET_GS264E and TUNE_GS264E.
5440 (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs264e.
5441 (MIPS_ASE_MSA_SPEC): New.
5442 (BASE_DRIVER_SELF_SPECS): march=gs264e implies -mmsa.
5443 (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS264E.
5444 (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS264E.
5445 * config/mips/mips.md: Include gs264e.md.
5446 (processor): Add gs264e.
5447 * config/mips/mips.opt (MSA): Use Mask instead of Var.
5448 * doc/invoke.texi: Add gs264e to supported architectures.
5450 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
5452 * config/mips/gs464e.md: New.
5453 * config/mips/mips-cpus.def: Define gs464e.
5454 * config/mips/mips-tables.opt: Regenerate.
5455 * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
5457 (mips_issue_rate): Add support for gs464e.
5458 (mips_multipass_dfa_lookahead): Likewise.
5459 (mips_option_override): Enable MMI, EXT and EXT2 for gs464e.
5460 * config/mips/mips.h: Define TARGET_GS464E and TUNE_GS464E.
5461 (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464e.
5462 (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS464E.
5463 (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS464E.
5464 * config/mips/mips.md: Include gs464e.md.
5465 (processor): Add gs464e.
5466 * doc/invoke.texi: Add gs464e to supported architectures.
5468 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
5470 * config/mips/loongson3a.md: Rename to ...
5471 * config/mips/gs464.md: ... here.
5472 * config/mips/mips-cpus.def: Define gs464; Add loongson3a
5473 as an alias of gs464 processor.
5474 * config/mips/mips-tables.opt: Regenerate.
5475 * config/mips/mips.c (mips_issue_rate): Use PROCESSOR_GS464
5476 instead of PROCESSOR_LOONGSON_3A.
5477 (mips_multipass_dfa_lookahead): Use TUNE_GS464 instead of
5479 (mips_option_override): Enable MMI and EXT for gs464.
5480 * config/mips/mips.h: Rename TARGET_LOONGSON_3A to TARGET_GS464;
5481 Rename TUNE_LOONGSON_3A to TUNE_GS464.
5482 (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464.
5483 (ISA_HAS_ODD_SPREG, ISA_AVOID_DIV_HILO, ISA_HAS_FUSED_MADD4,
5484 ISA_HAS_UNFUSED_MADD4): Use TARGET_GS464 instead of
5486 * config/mips/mips.md: Include gs464.md instead of loongson3a.md.
5487 (processor): Add gs464;
5488 * doc/invoke.texi: Add gs464 to supported architectures.
5490 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
5492 * config/mips/mips-protos.h
5493 (mips_loongson_ext2_prefetch_cookie): New prototype.
5494 * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): New.
5495 (mips_option_override): Enable TARGET_LOONGSON_EXT when
5496 TARGET_LOONGSON_EXT2 is true.
5497 * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define
5498 __mips_loongson_ext2, __mips_loongson_ext_rev=2.
5499 (ISA_HAS_CTZ_CTO): New, true if TARGET_LOONGSON_EXT2.
5500 (ISA_HAS_PREFETCH): Include TARGET_LOONGSON_EXT and
5501 TARGET_LOONGSON_EXT2.
5502 (ASM_SPEC): Add mloongson-ext2 and mno-loongson-ext2.
5503 (define_insn "ctz<mode>2"): New insn pattern.
5504 (define_insn "prefetch"): Include TARGET_LOONGSON_EXT2.
5505 (define_insn "prefetch_indexed_<mode>"): Include
5506 TARGET_LOONGSON_EXT and TARGET_LOONGSON_EXT2.
5507 * config/mips/mips.opt (-mloongson-ext2): Add option.
5508 * gcc/doc/invoke.texi (-mloongson-ext2): Document.
5510 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
5512 * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Add
5513 __mips_loongson_ext.
5514 (MIPS_ASE_LOONGSON_EXT_SPEC): New.
5515 (BASE_DRIVER_SELF_SPECS): march=loongson3a implies
5517 (ASM_SPEC): Add mloongson-ext and mno-loongson-ext.
5518 * config/mips/mips.md (mul<mode>3, mul<mode>3_mul3_nohilo,
5519 <u>div<mode>3, <u>mod<mode>3, prefetch): Use TARGET_LOONGSON_EXT
5520 instead of TARGET_LOONGSON_3A.
5521 * config/mips/mips.opt (-mloongson-ext): Add option.
5522 * gcc/doc/invoke.texi (-mloongson-ext): Document.
5524 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
5526 * config.gcc (extra_headers): Add loongson-mmiintrin.h.
5527 * config/mips/loongson.md: Move to ...
5528 * config/mips/loongson-mmi.md: here; Adjustment.
5529 * config/mips/loongson.h: Move to ...
5530 State as deprecated. Include loongson-mmiintrin.h for back
5531 compatibility and warning.
5532 * config/mips/loongson-mmiintrin.h: ... here.
5533 * config/mips/mips.c (mips_hard_regno_mode_ok_uncached,
5534 mips_vector_mode_supported_p, AVAIL_NON_MIPS16): Use
5535 TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
5536 (mips_option_override): Make sure MMI use hard float;
5537 (mips_shift_truncation_mask, mips_expand_vpc_loongson_even_odd,
5538 mips_expand_vpc_loongson_pshufh, mips_expand_vpc_loongson_bcast,
5539 mips_expand_vector_init): Use TARGET_LOONGSON_MMI instead of
5540 TARGET_LOONGSON_VECTORS.
5541 * gcc/config/mips/mips.h (TARGET_LOONGSON_VECTORS): Delete.
5542 (TARGET_CPU_CPP_BUILTINS): Add __mips_loongson_mmi.
5543 (MIPS_ASE_DSP_SPEC, MIPS_ASE_LOONGSON_MMI_SPEC): New.
5544 (BASE_DRIVER_SELF_SPECS): march=loongson2e/2f/3a implies
5546 (SHIFT_COUNT_TRUNCATED): Use TARGET_LOONGSON_MMI instead of
5547 TARGET_LOONGSON_VECTORS.
5548 * gcc/config/mips/mips.md (MOVE64, MOVE128): Use
5549 TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
5550 (Loongson MMI patterns): Include loongson-mmi.md instead of
5552 * gcc/config/mips/mips.opt (-mloongson-mmi): New option.
5553 * gcc/doc/invoke.texi (-mloongson-mmi): Document.
5555 2018-11-07 Richard Biener <rguenther@suse.de>
5558 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup
5559 BLOCK_ABSTRACT_ORIGIN to be the ultimate origin.
5561 2018-11-07 Alexandre Oliva <aoliva@redhat.com>
5563 PR rtl-optimization/87874
5564 * lra.c (lra_substitute_pseudo): Do not create a subreg for
5567 2018-11-06 Aaron Sawdey <acsawdey@linux.ibm.com>
5569 * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
5570 if not in indexed or indirect form.
5571 (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
5572 (bswap<mode>2_store): Ditto.
5574 2018-11-06 Richard Earnshaw <rearnsha@arm.com>
5576 * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
5579 2018-11-06 Richard Biener <rguenther@suse.de>
5581 PR tree-optimization/86850
5582 * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
5583 instead of src.m_vec.
5585 2018-11-06 Jan Hubicka <jh@suse.cz>
5587 * tree.c (fld_simplified_type_name): Break out form ...
5588 (free_lang_data_in_type): ... here.
5589 (fld_type_variant_equal_p): Use it.
5591 2018-11-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5593 * config/default-d.c: Include memmodel.h.
5595 * config/sol2-d.c: New file.
5596 * config/t-sol2 (sol2-d.o): New rule.
5597 * config.gcc <*-*-solaris2*>: Set d_target_objs,
5598 target_has_targetdm.
5600 2018-11-06 Jan Hubicka <jh@suse.cz>
5602 * tree.c (fld_type_variant): Also copy alignment; be sure that
5603 new variant is equal.
5605 2018-11-06 Ilya Leoshkevich <iii@linux.ibm.com>
5608 * config/s390/s390.md: Add relative_long attribute.
5610 2018-11-06 Jan Hubicka <jh@suse.cz>
5612 * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
5613 * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
5614 * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
5615 stream TYPE_NEEDS_CONSTRUCTING.
5616 * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
5617 * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
5619 2018-11-06 Richard Biener <rguenther@suse.de>
5621 * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
5623 (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
5625 2018-11-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5627 * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
5628 REG_CLASS_CONTENTS[GEN_REGS].
5629 (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
5631 2018-11-06 Jan Hubicka <jh@suse.cz>
5633 * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
5635 2018-11-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5638 * config/sol2.h (ASAN_CC1_SPEC): Define.
5639 (LD_WHOLE_ARCHIVE_OPTION): Define.
5640 (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
5641 (ASAN_REJECT_SPEC): Provide default.
5642 (LIBASAN_EARLY_SPEC): Define.
5643 (LIBTSAN_EARLY_SPEC): Define.
5644 (LIBLSAN_EARLY_SPEC): Define.
5645 * config/i386/sol2.h (CC1_SPEC): Redefine.
5646 (ASAN_REJECT_SPEC): Define.
5648 * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
5649 (TARGET_ASAN_SHADOW_OFFSET): Define.
5650 (sparc_asan_shadow_offset): New function.
5651 * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
5652 (ASAN_REJECT_SPEC): Define.
5654 2018-11-06 Jan Hubicka <jh@suse.cz>
5656 * tree.c (fld_type_variant): Copy canonical type.
5657 (fld_incomplete_type_of): Check that canonical types looks sane;
5658 copy canonical type.
5659 (verify_type): Accept when incomplete type has complete canonical type.
5661 2018-11-06 Jan Hubicka <jh@suse.cz>
5663 * tree.c (free_lang_data): Reset overwite_assembler_name,
5664 print_xnode, print_decl, print_type and print_identifier of
5667 2018-11-06 Richard Biener <rguenther@suse.de>
5669 PR tree-optimization/87889
5670 * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
5671 Do nothing if old and new arg are the same
5673 2018-11-06 Andreas Krebbel <krebbel@linux.ibm.com>
5676 * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
5677 attributes for operands 3 and 4.
5679 2018-11-06 Richard Biener <rguenther@suse.de>
5682 * simplify-rtx.c (simplify_binary_operation_1): Add pattern
5683 matching bitfield insertion.
5685 2018-11-06 Alexandre Oliva <aoliva@redhat.com>
5687 * auto-inc-dec.c: Include valtrack.h. Improve comments.
5688 (reg_next_debug_use): New.
5689 (attempt_change): Propagate adjusted expression into affected
5691 (merge_in_block): Track uses in debug insns.
5692 (pass_inc_dec::execute): Allocate and release
5695 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
5697 * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
5698 (_mm512_fixupimm_round_pd): Update parameters and builtin.
5699 (_mm512_maskz_fixupimm_round_pd): Ditto.
5700 (_mm512_fixupimm_round_ps): Ditto.
5701 (_mm512_maskz_fixupimm_round_ps): Ditto.
5702 (_mm_fixupimm_round_sd): Ditto.
5703 (_mm_maskz_fixupimm_round_sd): Ditto.
5704 (_mm_fixupimm_round_ss): Ditto.
5705 (_mm_maskz_fixupimm_round_ss): Ditto.
5706 (_mm512_fixupimm_pd): Ditto.
5707 (_mm512_maskz_fixupimm_pd): Ditto.
5708 (_mm512_fixupimm_ps): Ditto.
5709 (_mm512_maskz_fixupimm_ps): Ditto.
5710 (_mm_fixupimm_sd): Ditto.
5711 (_mm_maskz_fixupimm_sd): Ditto.
5712 (_mm_fixupimm_ss): Ditto.
5713 (_mm_maskz_fixupimm_ss): Ditto.
5714 (_mm512_mask_fixupimm_round_pd): Update builtin.
5715 (_mm512_mask_fixupimm_round_ps): Ditto.
5716 (_mm_mask_fixupimm_round_sd): Ditto.
5717 (_mm_mask_fixupimm_round_ss): Ditto.
5718 (_mm512_mask_fixupimm_pd): Ditto.
5719 (_mm512_mask_fixupimm_ps): Ditto.
5720 (_mm_mask_fixupimm_sd): Ditto.
5721 (_mm_mask_fixupimm_ss): Ditto.
5722 * config/i386/avx512vlintrin.h:
5723 (_mm256_fixupimm_pd): Update parameters and builtin.
5724 (_mm256_maskz_fixupimm_pd): Ditto.
5725 (_mm256_fixupimm_ps): Ditto.
5726 (_mm256_maskz_fixupimm_ps): Ditto.
5727 (_mm_fixupimm_pd): Ditto.
5728 (_mm_maskz_fixupimm_pd): Ditto.
5729 (_mm_fixupimm_ps): Ditto.
5730 (_mm_maskz_fixupimm_ps): Ditto.
5731 (_mm256_mask_fixupimm_pd): Update builtin.
5732 (_mm256_mask_fixupimm_ps): Ditto.
5733 (_mm_mask_fixupimm_pd): Ditto.
5734 (_mm_mask_fixupimm_ps): Ditto.
5735 * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
5736 * config/i386/i386-builtin.def: Update builtin definitions.
5737 * config/i386/i386.c: Handle new builtin types and remove useless ones.
5738 * config/i386/sse.md: Update VFIXUPIMM* patterns.
5739 (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
5740 (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
5741 (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
5742 (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
5743 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
5744 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
5745 * config/i386/subst.md:
5746 (round_saeonly_sd_mask_operand4): Add new subst_attr.
5747 (round_saeonly_sd_mask_op4): Ditto.
5748 (round_saeonly_expand_operand5): Ditto.
5749 (round_saeonly_expand): Update.
5751 2018-11-05 Max Filippov <jcmvbkbc@gmail.com>
5753 * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
5755 2018-11-05 Segher Boessenkool <segher@kernel.crashing.org>
5757 PR rtl-optimization/87871
5758 * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
5760 2018-11-05 Paul Koning <ni1d@arrl.net>
5762 * doc/sourcebuild.texi (target attributes): Document new "inf"
5763 effective target keyword.
5765 2018-11-05 Robin Dapp <rdapp@linux.ibm.com>
5767 * config/s390/s390.c (s390_register_move_cost): Increase costs for
5768 moves involving the CC reg.
5770 2018-11-05 Richard Biener <rguenther@suse.de>
5772 PR tree-optimization/87873
5773 * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
5775 * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
5776 * tree-vect-loop.c (vect_transform_loop): When splitting the
5777 loop exit also create forwarder PHIs for constants.
5778 * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
5779 Handle constant to_arg, add extra checking we match up the correct
5782 2018-11-05 Robin Dapp <rdapp@linux.ibm.com>
5784 * config/s390/s390.md: QImode and HImode for load on condition.
5786 2018-11-05 Robin Dapp <rdapp@linux.ibm.com>
5788 * config/s390/predicates.md: Fix typo.
5789 * config/s390/s390.md: Allow immediates for load on condition.
5791 2018-11-05 Martin Liska <mliska@suse.cz>
5793 * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
5794 * bitmap.h (struct bitmap_usage): Likewise.
5795 * ggc-common.c (SCALE): Remove.
5797 (struct ggc_usage): Use SIZE_AMOUNT. And update
5799 * ggc-page.c (SCALE): Remove.
5800 (STAT_LABEL): Remove.
5801 (ggc_print_statistics): Use SIZE_AMOUNT.
5802 * gimple.h (SCALE): Remove.
5804 * input.c (ONE_K): Remove.
5807 (STAT_LABEL): Likewise.
5808 (FORMAT_AMOUNT): Likewise.
5809 (dump_line_table_statistics): Use SIZE_AMOUNT.
5810 * mem-stats.h (struct mem_usage): Likewise.
5811 * rtl.c (dump_rtx_statistics): Likewise.
5812 (rtx_alloc_counts): Change type to size_t.
5813 (rtx_alloc_sizes): Likewise.
5814 (rtx_count_cmp): New.
5815 (dump_rtx_statistics): Sort first based on counts.
5816 * tree.c (tree_nodes_cmp): New.
5817 (tree_codes_cmp): New.
5818 (dump_tree_statistics): Sort first based on counts.
5819 * system.h (ONE_K): New.
5821 (SIZE_SCALE): Likewise.
5822 (SIZE_LABEL): Likewise.
5823 (SIZE_AMOUNT): Likewise.
5824 * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
5825 * tree-dfa.c (dump_dfa_stats): Likewise.
5826 * tree-phinodes.c (phinodes_print_statistics): Likewise.
5827 * tree-ssanames.c (ssanames_print_statistics): Likewise.
5828 * tree.c (dump_tree_statistics): Likewise.
5829 * vec.c (struct vec_usage): Likewise.
5830 * trans-mem.c (tm_mangle): Enlarge buffer in order to not
5831 trigger a -Werror=format-overflow with
5832 --enable-gather-detailed-stats.
5834 2018-11-05 Martin Liska <mliska@suse.cz>
5836 * mem-stats.h (mem_alloc_description::release_instance_overhead):
5838 * vec.c (struct vec_usage): Register m_element_size.
5839 (vec_prefix::register_overhead): New arguments: elements and
5841 (vec_prefix::release_overhead): Subtract elements.
5842 * vec.h (struct vec_prefix): Change signature.
5843 (va_heap::reserve): Pass proper arguments.
5844 (va_heap::release): Likewise.
5846 2018-11-05 Martin Liska <mliska@suse.cz>
5848 * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
5852 2018-11-05 Richard Biener <rguenther@suse.de>
5854 * tree-scalar-evolution.h (final_value_replacement_loop): Update
5856 * tree-scalar-evolution.c (final_value_replacement_loop): Return
5857 whether anything was done.
5858 (scev_const_prop): Remove constant propagation part, fold
5860 * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
5861 (pass_data_scev_cprop): TODO_cleanup_cfg is now done
5864 2018-11-05 Jakub Jelinek <jakub@redhat.com>
5866 PR tree-optimization/87859
5867 * gimple-ssa-store-merging.c (struct merged_store_group): Add
5868 only_constants and first_nonmergeable_order members.
5869 (merged_store_group::merged_store_group): Initialize them.
5870 (merged_store_group::do_merge): Clear only_constants member if
5871 adding something other than INTEGER_CST store.
5872 (imm_store_chain_info::coalesce_immediate_stores): Don't merge
5873 stores with order >= first_nonmergeable_order. Use
5874 merged_store->only_constants instead of always recomputing it.
5875 Set merged_store->first_nonmergeable_order if we've skipped any
5876 stores. Attempt to merge overlapping INTEGER_CST stores that
5877 we would otherwise skip.
5880 * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
5882 2018-11-05 Xuepeng Guo <xuepeng.guo@intel.com>
5885 * config/i386/emmintrin.h (__v16qs): New to cope with option
5887 (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
5888 (_mm_cmplt_epi8): Likewise.
5889 (_mm_cmpgt_epi8): Likewise.
5891 2018-11-05 Richard Biener <rguenther@suse.de>
5893 PR rtl-optimization/87852
5894 * fwprop.c (use_killed_between): Only consider single-defs of the
5895 use whose definition statement dominates the use.
5897 2018-11-05 Martin Liska <mliska@suse.cz>
5900 * doc/invoke.texi: Remove options that are
5901 not disabled with -Os.
5903 2018-11-05 Martin Liska <mliska@suse.cz>
5906 * doc/extend.texi: Update constrain about the last argument
5907 of __builtin_expect_with_probability.
5909 2018-11-05 Martin Liska <mliska@suse.cz>
5912 * predict.c (expr_expected_value_1): Verify
5913 that last argument is a real constants and emit
5916 2018-11-05 Martin Liska <mliska@suse.cz>
5918 PR gcov-profile/77698
5919 * ipa-profile.c (ipa_profile): Adjust hotness threshold
5922 2018-11-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
5924 PR tree-optimization/86572
5925 * builtins.c (c_strlen): Handle negative offsets in a safe way.
5927 2018-11-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
5929 PR tree-optimization/87672
5930 * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
5931 * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
5933 2018-11-04 Uros Bizjak <ubizjak@gmail.com>
5936 * cfgexpand.c (pass_expand::execute): Move the call to
5937 finish_eh_generation in front of the call to expand_stack_alignment.
5939 2018-11-04 Venkataramanan Kumar <venkataramanan.kumar@amd.com>
5941 * common/config/i386/i386-common.c (processor_alias_table): Add
5943 * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
5944 (case ${target}): Add znver2.
5945 * config/i386/driver-i386.c: (host_detect_local_cpu): Let
5946 -march=native recognize znver2 processors.
5947 * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
5948 * config/i386/i386.c (m_znver2): New definition.
5949 (m_ZNVER): New definition.
5950 (m_AMD_MULTIPLE): Includes m_znver2.
5951 (processor_cost_table): Add znver2 entry.
5952 (processor_target_table): Add znver2 entry.
5953 (get_builtin_code_for_version): Set priority for
5955 (processor_model): Add M_AMDFAM17H_ZNVER2.
5956 (arch_names_table): Ditto.
5957 (ix86_reassociation_width): Include znver2.
5958 * config/i386/i386.h (TARGET_znver2): New definition.
5959 (struct ix86_size_cost): Add TARGET_ZNVER2.
5960 (enum processor_type): Add PROCESSOR_ZNVER2.
5961 * config/i386/i386.md (define_attr "cpu"): Add znver2.
5962 * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
5963 * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
5964 (ix86_adjust_cost): Add znver2.
5965 * config/i386/x86-tune.def: Replace m_ZNVER1 by m_ZNVER.
5966 * gcc/doc/extend.texi: Add details about znver2.
5967 * gcc/doc/invoke.texi: Add details about znver2.
5969 2018-11-03 Sandra Loosemore <sandra@codesourcery.com>
5973 * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
5976 2018-11-02 Aaron Sawdey <acsawdey@linux.ibm.com>
5978 * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
5979 attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
5981 2018-11-02 Richard Earnshaw <rearnsha@arm.com>
5983 * config/aarch64/aarch64.c ((aarch64_override_options): Disable
5984 shrink-wrapping when -mtrack-speculation.
5986 2018-11-02 Richard Biener <rguenther@suse.de>
5988 * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
5989 (pop_cost_one_pair): Do not free pair.
5990 (pop_best_coalesce): Likewise.
5991 (create_coalesce_list): Initialize obstack.
5992 (delete_coalesce_list): Free obstack.
5993 (find_coalesce_pair): Obstack-allocate coalesce pairs.
5994 (add_cost_one_coalesce): Likewise.
5995 (struct live_track): Remove bitmap pointer indirections.
5996 (new_live_track): Adjust.
5997 (delete_live_track): Likewise.
5998 (live_track_remove_partition): Likewise.
5999 (live_track_add_partition): Likewise.
6000 (live_track_live_p): Likewise.
6001 (live_track_process_def): Likewise.
6002 (live_track_clear_base_vars): Likewise.
6004 2018-11-02 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6006 * configure.ac (gcc_cv_as_sparc_register_op): Remove.
6007 * configure: Regenerate.
6008 * config.in: Regenerate.
6009 * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
6010 HAVE_AS_REGISTER_PSEUDO_OP guard.
6011 * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
6012 !HAVE_AS_REGISTER_PSEUDO_OP support.
6013 (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
6016 2018-11-02 Richard Biener <rguenther@suse.de>
6018 * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
6021 2018-11-02 Richard Biener <rguenther@suse.de>
6023 PR tree-optimization/87776
6024 * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
6025 executable when iterating but running into rpo-vn-max-loop-depth
6026 and not eliding the iteration.
6028 2018-11-30 Jan Hubicka <jh@suse.cz>
6030 * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
6033 2018-11-01 Aaron Sawdey <acsawdey@linux.ibm.com>
6035 * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
6037 * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
6038 Combine with rs6000_address_for_fpconvert.
6039 (rs6000_address_for_fpconvert): Combine with
6040 rs6000_force_indexed_or_indirect_mem.
6041 (rs6000_expand_vector_init): Change function call from
6042 rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
6043 * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
6044 rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
6045 (floatsi<mode>2_lfiwax_mem): Ditto.
6046 (floatunssi<mode>2_lfiwzx): Ditto.
6047 (floatunssi<mode>2_lfiwzx_mem): Ditto.
6048 (float<QHI:mode><FP_ISA3:mode>2): Ditto.
6049 (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
6050 (fix_trunc<mode>si2_stfiwx): Ditto.
6051 (fixuns_trunc<mode>si2_stfiwx): Ditto.
6052 (float_<mode>si2_hw): Ditto.
6053 (floatuns_<mode>si2_hw): Ditto.
6054 * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
6055 (vsx_splat_<mode>): Ditto.
6057 2018-11-01 Joseph Myers <joseph@codesourcery.com>
6059 * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
6060 at end of assembler input text.
6061 * configure: Regenerate.
6063 2018-11-01 Jakub Jelinek <jakub@redhat.com>
6065 PR tree-optimization/87826
6066 * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
6067 negative or larger or equal to type's precision.
6069 2018-10-31 Alexandre Oliva <aoliva@redhat.com>
6071 * opts.c (default_options_table): Do not enable
6072 OPT_fdelayed_branch at -Og.
6073 * doc/invoke.texi (-fdelayed-branch): Document it.
6075 2018-10-31 Richard Henderson <richard.henderson@linaro.org>
6077 * optabs-libfuncs.c (build_libfunc_function_visibility):
6078 New, split out from...
6079 (build_libfunc_function): ... here.
6080 (init_one_libfunc_visibility): New, split out from ...
6081 (init_one_libfunc): ... here.
6083 * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
6084 scratch register need not be early-clobber. Document the reason
6085 why we cannot use ST<OP>.
6087 2018-10-31 Joseph Myers <joseph@codesourcery.com>
6090 * configure.ac: Remove AC_PREREQ. Use AC_LANG_SOURCE. Use single
6091 line for second argument of AC_DEFINE_UNQUOTED.
6092 * doc/install.texi (Tools/packages necessary for modifying GCC):
6093 Update to autoconf 2.69 and automake 1.15.1.
6094 * aclocal.m4, config.in, configure: Regenerate.
6096 2018-10-31 Pat Haugen <pthaugen@us.ibm.com>
6098 * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
6101 2018-10-31 Martin Liska <mliska@suse.cz>
6104 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
6105 Add new argument invalid_extension.
6106 (aarch64_get_all_extension_candidates): New function.
6107 (aarch64_rewrite_selected_cpu): Add NULL to function call.
6108 * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
6110 (aarch64_get_all_extension_candidates): New function.
6111 * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
6112 argument invalid_extension.
6113 (aarch64_parse_cpu): Likewise.
6114 (aarch64_print_hint_for_extensions): New function.
6115 (aarch64_validate_mcpu): Provide hint about invalid extension.
6116 (aarch64_validate_march): Likewise.
6117 (aarch64_handle_attr_arch): Pass new argument.
6118 (aarch64_handle_attr_cpu): Provide hint about invalid extension.
6119 (aarch64_handle_attr_isa_flags): Likewise.
6121 2018-10-31 Richard Biener <rguenther@suse.de>
6125 * tree-outof-ssa.c (insert_backedge_copies): Restrict
6126 copy generation to useful cases. Place the copy before
6127 the definition of the backedge value when possible.
6129 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
6131 * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
6132 * config/arc/arc.c (arc_active_insn): New function.
6133 (check_store_cacheline_hazard): Likewise.
6134 (workaround_arc_anomaly): Use check_store_cacheline_hazard.
6135 (arc_override_options): Disable delay slot scheduler for older
6137 (arc_store_addr_hazard_p): New implementation, old one renamed to
6139 (arc_store_addr_hazard_internal_p): Renamed.
6140 (arc_reorg): Don't combine into brcc instructions which are part
6141 of hardware hazard solution.
6142 * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
6143 (tune_arc700): Likewise.
6144 * config/arc/arc.opt (arc7xx): New tune value.
6145 * config/arc/arc700.md: Improve A7 scheduler.
6147 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
6149 * config/arc/arc.c (arc_override_options): Remove
6150 TARGET_COMPACT_CASESI.
6151 * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
6152 (CASE_VECTOR_MODE): Likewise.
6153 (CASE_VECTOR_PC_RELATIVE): Likewise.
6154 (CASE_VECTOR_SHORTEN_MODE): Likewise.
6155 (CASE_VECTOR_SHORTEN_MODE1): Delete.
6156 (ADDR_VEC_ALIGN): Update.
6157 (ASM_OUTPUT_CASE_LABEL): Undefine.
6158 (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
6159 (TARGET_BI_BIH): Define.
6160 (DEFAULT_BRANCH_INDEX): Likewise.
6161 * config/arc/arc.md (casesi): Rework to accept BI/BIH
6162 instructions, remove compact_casesi use case.
6163 (casesi_compact_jump): Remove.
6164 (casesi_dispatch): New pattern.
6165 * config/arc/arc.opt: Add mbranch-index option. Deprecate
6166 compact_casesi option.
6167 * doc/invoke.texi: Document mbranch-index option.
6169 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
6171 * config/arc/arc.c (arc_get_tp): Remove function.
6172 (arc_emit_call_tls_get_addr): Likewise.
6173 (arc_call_tls_get_addr): New function.
6174 (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
6175 * config/arc/arc.md (tls_load_tp_soft): Remove.
6176 (tls_gd_get_addr): Likewise.
6178 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
6180 * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
6181 (mulsi3_600_lib): Remove pattern.
6182 (umulsi3_highpart_600_lib_le): Likewise.
6183 (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
6184 (umulsidi3): Remove call to umulsidi3_600_lib.
6185 (umulsidi3_600_lib): Remove pattern.
6186 (peephole2): Remove peephole using the above deprecated patterns.
6188 2018-10-31 Thomas Preud'homme <thomas.preudhomme@linaro.org>
6191 * config/arm/arm.c (arm_option_check_internal): Disable the combined
6192 use of -mslow-flash-data and -mword-relocations.
6193 (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
6194 * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
6196 * doc/invoke.texi (-mword-relocations): Mention conflict with
6198 (-mslow-flash-data): Reciprocally.
6200 2018-10-31 Richard Henderson <richard.henderson@linaro.org>
6202 * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
6203 16-byte modes held in GP registers to use an even regno.
6205 * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
6206 (aarch64_atomic_ldop_supported_p): Remove.
6207 (aarch64_gen_atomic_ldop): Remove.
6208 * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
6209 Fully expand LSE operations here.
6210 (atomic_fetch_<atomic_optab><ALLI>): Likewise.
6211 (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
6212 (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
6213 and use ATOMIC_LDOP instead; use register_operand for the input;
6214 drop the split and emit insns directly.
6215 (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
6216 (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
6217 (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
6219 * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
6220 (aarch64_gen_atomic_ldop): Don't call it.
6221 * config/aarch64/atomics.md (atomic_exchange<ALLI>):
6222 Use aarch64_reg_or_zero.
6223 (aarch64_atomic_exchange<ALLI>): Likewise.
6224 (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
6225 operand 0; use aarch64_reg_or_zero for input; merge ...
6226 (@aarch64_atomic_swp<ALLI>): ... this and remove.
6228 * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
6229 (aarch64_split_compare_and_swap): Use it.
6230 (aarch64_expand_compare_and_swap): Likewise. Remove convert_modes;
6231 test oldval against the proper predicate.
6232 * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
6233 Use nonmemory_operand for expected.
6234 (cas_short_expected_pred): New.
6235 (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
6236 (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
6237 * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
6238 (aarch64_plushi_operand): New.
6240 * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
6241 Force oldval into the rval register for TARGET_LSE; emit the compare
6242 during initial expansion so that it may be deleted if unused.
6243 (aarch64_gen_atomic_cas): Remove.
6244 * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
6245 Change =&r to +r for operand 0; use match_dup for operand 2;
6246 remove is_weak and mod_f operands as unused. Drop the split
6248 (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
6249 (@aarch64_compare_and_swap<GPI>_lse): Similarly.
6250 (@aarch64_atomic_cas<GPI>): Similarly.
6252 2018-10-31 Richard Biener <rguenther@suse.de>
6254 * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
6257 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
6259 * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
6260 * config/aarch64/aarch64.md: Include saphira.md
6261 * config/aarch64/saphira.md: New file for pipeline description.
6263 2018-10-30 Martin Sebor <msebor@redhat.com>
6266 * gimple-ssa-sprintf.c (format_directive): Use %G to include
6268 (sprintf_dom_walker::compute_format_length):
6269 Avoid setting POSUNDER4K here.
6270 (get_destination_size): Handle null argument values.
6271 (get_user_idx_format): New function.
6272 (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
6273 functions, including user-defined with attribute format printf.
6274 Use %G to include inlining context.
6275 Set POSUNDER4K here.
6277 2018-10-30 Jan Hubicka <jh@suse.cz>
6279 * params.def (lto-partitions): Bump from 32 to 128.
6281 2018-10-30 Jan Hubicka <jh@suse.cz>
6284 (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
6286 (free_lang_data_in_type): Forward declare.
6287 (fld_type_variant_equal_p): New function.
6288 (fld_type_variant): New function
6289 (fld_incomplete_types): New hash.
6290 (fld_incomplete_type_of): New function
6291 (fld_simplfied-type): New function.
6292 (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
6293 (free_lang_data): Allocate and free fld_incomplete_type; update call
6294 of free_lang_data_in_decl.
6296 2018-10-30 Eric Botcazou <ebotcazou@adacore.com>
6298 * gcov.c (output_lines): Remove duplicate line.
6300 2018-10-30 Aaron Sawdey <acsawdey@linux.ibm.com>
6302 * config/rs6000/rs6000.md (bswapdi2): Force address into register
6303 if not in indexed or indirect form.
6304 (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
6305 (bswapdi2_store): Ditto.
6306 * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
6308 * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
6309 Prototype for helper function.
6311 2018-10-30 Martin Sebor <msebor@redhat.com>
6313 * doc/extend.texi (optimize): Clarify/expand attribute documentation.
6314 (target, pragma GCC optimize, pragma GCC target): Ditto.
6316 2018-10-30 Jonathan Wakely <jwakely@redhat.com>
6318 * doc/extend.texi: Fix prototype and description of
6319 __builtin_expect_with_probability.
6321 2018-10-30 Michael Ploujnikov <michael.ploujnikov@oracle.com>
6323 * cgraph.h (clone_function_name_1): Replaced by new
6324 clone_function_name_numbered that takes name as string; for
6325 privatize_symbol_name_1 use only.
6326 (clone_function_name): Renamed to
6327 clone_function_name_numbered to be explicit about numbering.
6328 (clone_function_name): New two-argument function that does
6329 not number its output.
6330 (clone_function_name): New three-argument function that
6331 takes a number to append to its output.
6332 * cgraphclones.c (duplicate_thunk_for_node):
6333 (clone_function_name_1): Renamed.
6334 (clone_function_name_numbered): Two new functions.
6335 (clone_function_name): Improved documentation.
6336 (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
6337 * config/rs6000/rs6000.c (make_resolver_func): Ditto.
6338 * final.c (final_scan_insn_1): Use the new clone_function_name
6340 * multiple_target.c (create_dispatcher_calls): Ditto.
6341 (create_target_clone): Ditto.
6342 * omp-expand.c (grid_expand_target_grid_body): Ditto.
6343 * omp-low.c (create_omp_child_function_name): Ditto.
6344 * omp-simd-clone.c (simd_clone_create): Ditto.
6345 * symtab.c (simd_symtab_node::noninterposable_alias): Use the
6346 new clone_function_name without numbering.
6348 2018-10-30 Richard Earnshaw <rearnsha@arm.com>
6350 * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
6351 Assert that the allocation size is not zero.
6353 2018-10-30 Richard Biener <rguenther@suse.de>
6355 PR tree-optimization/87800
6356 * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
6357 non-induction or reduction PHIs.
6359 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org>
6361 * config/aarch64/falkor-tag-collision-avoidance.c
6362 (execute_tag_collision_avoidance): Call df_note_add_problem.
6364 2018-10-30 Martin Liska <mliska@suse.cz>
6366 * doc/extend.texi: Fix typo in documentation
6367 of __builtin_expect_with_probability.
6369 2018-10-29 David Malcolm <dmalcolm@redhat.com>
6372 * input.c (get_substring_ranges_for_loc): Detect if
6373 linemap_resolve_location gives us a NULL map, and reject
6376 2018-10-29 Iain Buclaw <ibuclaw@gdcproject.org>
6378 * config.gcc (xstormy16-*-elf): Set tm_d_file.
6380 2018-10-29 David Malcolm <dmalcolm@redhat.com>
6381 Martin Sebor <msebor@redhat.com>
6382 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
6384 * doc/ux.texi (Quoting): New subsection, adapted from material at
6385 https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
6386 MartinSebor and ManuelLopezIbanez.
6387 (Fix-it hints): Note that fix-it hints shouldn't be marked for
6390 2018-10-29 Kugan Vivekanandarajah <kuganv@linaro.org>
6393 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
6396 2018-10-29 Paul A. Clarke <pc@us.ibm.com>
6398 * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
6400 2018-10-29 Paul Koning <ni1d@arrl.net>
6402 * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
6404 2018-10-29 Paul A. Clarke <pc@us.ibm.com>
6406 * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
6407 _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
6408 _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
6409 _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
6410 _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
6411 _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
6412 _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
6413 _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
6414 Change 'vector' to '__vector'.
6415 * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
6416 _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
6417 _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
6418 _mm_avg_pu16): Likewise. And, whitespace corrections.
6420 2018-10-29 Richard Biener <rguenther@suse.de>
6422 PR tree-optimization/87785
6423 * tree-vect-slp.c (vect_gather_slp_loads): Only gather
6426 2018-10-29 Olivier Hainque <hainque@adacore.com>
6428 * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
6430 2018-10-29 Olivier Hainque <hainque@adacore.com>
6432 * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
6433 ports configuration macro, defaults to "gnu".
6434 (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
6435 TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
6437 2018-10-29 Olivier Hainque <hainque@adacore.com>
6439 * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
6441 2018-10-29 Richard Biener <rguenther@suse.de>
6443 PR tree-optimization/87790
6444 * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
6445 (vect_make_slp_decision): Adjust.
6446 (vect_slp_analyze_bb_1): Likewise.
6447 (vect_detect_hybrid_slp_stmts): Properly union SLP type over
6450 2018-10-29 Richard Biener <rguenther@suse.de>
6452 PR tree-optimization/87785
6453 * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
6455 (vect_build_slp_tree): Likewise.
6456 (vect_gather_slp_loads): New function.
6457 (vect_analyze_slp_instance): Gather loads separately from the
6460 2018-10-29 Martin Liska <mliska@suse.cz>
6462 * Makefile.in: Make dependency to json.o.
6463 * doc/gcov.texi: Document new JSON format, remove
6464 old intermediate format documentation.
6465 * gcov.c (struct function_info): Come up with m_name and
6467 (function_info::function_info): Initialize it.
6468 (function_info::~function_info): Release it.
6469 (main): Rename flag_intermediate_format to flag_json_format.
6470 (print_usage): Describe --json-format.
6471 (process_args): Set flag_json_format.
6472 (output_intermediate_line): Remove.
6473 (output_intermediate_json_line): Likewise.
6474 (get_gcov_intermediate_filename): Return new extension
6476 (output_intermediate_file): Implement JSON emission.
6477 (output_json_intermediate_file): Implement JSON emission.
6478 (generate_results): Use ::get_name for function name.
6479 Handle JSON output file.
6480 (read_graph_file): Use ::get_name instead of cplus_demangle.
6481 (read_count_file): Likewise.
6482 (solve_flow_graph): Likewise.
6483 (add_line_counts): Likewise.
6484 (accumulate_line_counts): Use new flag_json_format.
6485 (output_function_details): Use ::get_name instead of cplus_demangle.
6486 (output_lines): Likewise.
6487 * json.cc (test_writing_literals): Add new tests.
6488 * json.h (class literal): Add new boolean constructor.
6490 2018-10-29 Segher Boessenkool <segher@kernel.crashing.org>
6492 PR rtl-optimization/87701
6493 PR rtl-optimization/87780
6494 * combine.c (make_more_copies): Rewrite.
6496 2018-10-28 Kugan Vivekanandarajah <kuganv@linaro.org>
6498 * doc/generic.texi (ABSU_EXPR): Document.
6499 * match.pd (absu(x)*absu(x) -> x*x): Handle.
6500 (absu(absu(X)) -> absu(X)): Likewise.
6501 (absu(-X) -> absu(X)): Likewise.
6502 (absu(X) where X is nonnegative -> X): Likewise.
6504 2018-10-28 Iain Buclaw <ibuclaw@gdcproject.org>
6506 * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
6507 (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
6508 (tm_d.h, cs-tm_d.h, default-d.o): New rules.
6509 (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
6510 (s-tm-texi): Also check timestamp on d-target.def.
6511 (generated_files): Add TM_D_H and d-target-hooks-def.h.
6512 (build/genhooks.o): Also depend on D_TARGET_DEF.
6513 * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
6515 * config/aarch64/aarch64-d.c: New file.
6516 * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
6518 * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
6520 * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
6521 * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
6522 * config/arm/arm-d.c: New file.
6523 * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
6524 * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
6525 * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
6526 * config/arm/t-arm (arm-d.o): New rule.
6527 * config/default-d.c: New file.
6528 * config/glibc-d.c: New file.
6529 * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6530 * config/i386/i386-d.c: New file.
6531 * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
6532 * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
6533 * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
6534 (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
6535 * config/i386/t-i386 (i386-d.o): New rule.
6536 * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6537 * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6538 * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
6539 * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6540 * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
6541 * config/mips/mips-d.c: New file.
6542 * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
6543 * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
6544 * config/mips/t-mips (mips-d.o): New rule.
6545 * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6546 * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6547 * config/powerpcspe/powerpcspe-d.c: New file.
6548 * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
6550 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
6551 Support GNU D by using 0 as the language type.
6552 * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
6553 * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
6554 * config/riscv/riscv-d.c: New file.
6555 * config/riscv/riscv-protos.h (riscv_d_target_versions): New
6557 * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
6558 * config/riscv/t-riscv (riscv-d.o): New rule.
6559 * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6560 * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
6561 * config/rs6000/rs6000-d.c: New file.
6562 * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
6564 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
6565 Support GNU D by using 0 as the language type.
6566 * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
6567 * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
6568 * config/s390/s390-d.c: New file.
6569 * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
6570 * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
6571 * config/s390/t-s390 (s390-d.o): New rule.
6572 * config/sparc/sparc-d.c: New file.
6573 * config/sparc/sparc-protos.h (sparc_d_target_versions): New
6575 * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
6576 * config/sparc/t-sparc (sparc-d.o): New rule.
6577 * config/t-glibc (glibc-d.o): New rule.
6578 * configure: Regenerated.
6579 * configure.ac (tm_d_file): New variable.
6580 (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
6581 * doc/contrib.texi (Contributors): Add self for the D frontend.
6582 * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
6583 * doc/install.texi (Configuration): Mention libphobos as an option for
6584 --enable-shared. Mention d as an option for --enable-languages.
6585 (Testing): Mention check-d as a target.
6586 * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
6587 name suffixes. Mention d as a -x option.
6588 * doc/sourcebuild.texi (Top Level): Mention libphobos.
6589 * doc/standards.texi (Standards): Add section on D language.
6590 * doc/tm.texi: Regenerated.
6591 * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
6592 TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
6593 * dwarf2out.c (is_dlang): New function.
6594 (gen_compile_unit_die): Use DW_LANG_D for D.
6595 (declare_in_namespace): Return module die for D, instead of adding
6596 extra declarations into the namespace.
6597 (gen_namespace_die): Generate DW_TAG_module for D.
6598 (gen_decl_die): Handle CONST_DECLSs for D.
6599 (dwarf2out_decl): Likewise.
6600 (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
6601 (prune_unused_types_walk): Handle DW_tag_interface_type same as other
6602 kinds of aggregates.
6603 * gcc.c (default_compilers): Add entries for .d, .dd and .di.
6604 * genhooks.c: Include d/d-target.def.
6606 2018-10-28 Iain Sandoe <iain@sandoe.co.uk>
6609 * config/rs6000/darwin.h (STACK_BOUNDARY): New.
6610 (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
6611 (STACK_DYNAMIC_OFFSET): Likewise.
6613 2018-10-27 Sandra Loosemore <sandra@codesourcery.com>
6616 * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
6619 2018-10-26 Tulio Magno Quites Machado Filho <tuliom@linux.ibm.com>
6621 * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
6622 return type and other typos.
6624 2018-10-26 Aaron Sawdey <acsawdey@linux.ibm.com>
6626 * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
6627 a shorter sequence with fewer branches.
6628 (emit_final_str_compare_gpr): Ditto.
6630 2018-10-26 Paul A. Clarke <pc@us.ibm.com>
6632 * config/rs6000/tmmintrin.h: New file.
6633 * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
6635 2018-10-26 Paul A. Clarke <pc@us.ibm.com>
6637 * config/rs6000/mmintrin.h: Enable 32bit compilation.
6638 * config/rs6000/xmmintrin.h: Likewise.
6640 2018-10-26 Paul A. Clarke <pc@us.ibm.com>
6642 * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
6644 2018-10-26 Richard Biener <rguenther@suse.de>
6646 * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
6648 (vect_mark_slp_stmts_relevant): Likewise.
6649 (vect_detect_hybrid_slp_stmts): Likewise.
6650 (vect_bb_slp_scalar_cost): Likewise.
6651 (vect_remove_slp_scalar_calls): Likewise.
6653 2018-10-26 Jan Hubicka <jh@suse.cz>
6655 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
6656 (warn_types_mismatch): Fix walk of DECL_NAME.
6657 (odr_types_equivalent_p): Fix overactive assert.
6659 2018-10-26 Richard Biener <rguenther@suse.de>
6661 PR tree-optimization/87105
6662 * tree-vectorizer.h (_slp_tree::refcnt): New member.
6663 * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
6665 (vect_create_new_slp_node): Initialize refcnt to one.
6667 (scalar_stmts_set_t, bst_fail): Remove.
6668 (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
6669 (vect_build_slp_tree): Add bst_map argument and lookup
6670 already created SLP nodes.
6671 (vect_print_slp_tree): Handle a SLP graph, print SLP node
6673 (vect_slp_rearrange_stmts): Handle a SLP graph.
6674 (vect_analyze_slp_instance): Adjust and free SLP nodes from
6675 the CSE map. Fix indenting.
6676 (vect_schedule_slp_instance): Add short-cut.
6678 2018-10-26 Martin Liska <mliska@suse.cz>
6681 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
6682 addr_expr and not with pointers.
6684 2018-10-26 Jan Hubicka <jh@suse.cz>
6686 * tree.c (free_lang_data_in_type): Only check main variants.
6687 * ipa-devirt.c (warn_odr): Make static.
6688 (types_same_for_odr): Drop strict variant.
6689 (types_odr_comparable): Likewise.
6690 (odr_or_derived_type_p): Look for main variants.
6691 (odr_name_hasher::equal): Cleanup comment.
6692 (odr_subtypes_equivalent): Add warn and warned arguments; check main
6694 (type_variants_equivalent_p): break out from ...
6695 (odr_types_equivalent): ... here; go for main variants where needed.
6696 (warn_odr): ... here; turn static.
6697 (warn_types_mismatch): Compare mangled names of main variants.
6698 * ipa-utils.h (types_odr_comparable): Drop strict parameter.
6699 (type_with_linkage_p): Sanity check that we look at main variant.
6700 * lto.c (lto_read_decls): Only consider main variant to be ODR type.
6701 * tree.h (types_same_for_odr): Drop strict argument.
6703 2018-10-26 Richard Biener <rguenther@suse.de>
6705 PR tree-optimization/87746
6706 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
6707 Simplify and fix WRT strided store groups with size not
6708 equal to step in element count.
6709 (vect_analyze_group_access_1): Dump the whole group.
6711 2018-10-25 Carl Love <cel@us.ibm.com>
6713 * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
6714 P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
6715 P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name. Add quad
6716 precicion entry for each overloaded builtin.
6717 * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
6718 VSCEDPUO): Rename overloaded name.
6719 (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
6720 VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
6721 * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
6722 define_expand for xscmpexqp instruction.
6723 (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
6725 2018-10-25 Bill Schmidt <wschmidt@linux.ibm.com>
6726 Jinsong Ji <jji@us.ibm.com>
6728 * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
6729 function with vec_sl.
6730 (_mm_slli_epi32): Likewise.
6731 (_mm_slli_epi64): Likewise.
6732 (_mm_srai_epi16): Replace deprecated function with vec_sra.
6733 (_mm_srai_epi32): Likewise.
6734 (_mm_srli_epi16): Replace deprecated function with vec_sr.
6735 (_mm_srli_epi32): Likewise.
6736 (_mm_srli_epi64): Likewise.
6737 (_mm_sll_epi16): Replace deprecated function with vec_sl.
6738 (_mm_sll_epi32): Likewise.
6739 (_mm_sll_epi64): Likewise.
6740 (_mm_sra_epi16): Replace deprecated function with vec_sra.
6741 (_mm_sra_epi32): Likewise.
6742 (_mm_srl_epi16): Replace deprecated function with vec_sr.
6743 (_mm_srl_epi32): Likewise.
6744 (_mm_srl_epi64): Likewise.
6746 2018-10-25 Bill Schmidt <wschmidt@linux.ibm.com>
6747 Jinsong Ji <jji@us.ibm.com>
6749 * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
6750 comparison operators with vec_cmp* for compatibility due to
6751 unfortunate history; clean up formatting and use types more
6753 (_mm_sll_epi32): Likewise.
6754 (_mm_sll_epi64): Likewise.
6755 (_mm_srl_epi16): Likewise.
6756 (_mm_srl_epi32): Likewise.
6757 (_mm_srl_epi64): Likewise.
6759 2018-10-25 Bill Schmidt <wschmidt@linux.ibm.com>
6760 Jinsong Ji <jji@us.ibm.com>
6762 * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
6763 * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
6764 __vector __bool int. Use vec_cmpgt in preference to deprecated
6765 function vec_vcmpgtfp.
6766 (_mm_max_ps): Likewise.
6768 2018-10-25 Jeff Law <law@redhat.com>
6770 * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
6773 2018-10-25 Martin Sebor <msebor@redhat.com>
6775 * doc/extend.texi (aligned): Expand attribute description.
6776 (Alignment): Rename section. Discuss function arguments.
6778 2018-10-25 Jan Hubicka <jh@suse.cz>
6780 * ipa-devirt.c (main_odr_variant): Remove.
6781 (hash_odr_name, types_same_for_odr, types_odr_comparable,
6782 odr_name_hasher::equal, odr_subtypes_equivalent_p):
6783 Drop use of main_odr_variant.
6784 (add_type_duplicate): Silence confused warnings on integer types.
6785 (get_odr_type): Always look for main variant.
6786 (register_odr_type): Simplify.
6788 2018-10-25 Richard Biener <rguenther@suse.de>
6790 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
6791 Initialize ng to silence error with release checking bootstrap.
6793 2018-10-25 Richard Biener <rguenther@suse.de>
6795 * tree-if-conv.c: Include tree-ssa-sccvn.h.
6796 (tree_if_conversion): Run CSE on the if-converted loop body.
6798 2018-10-25 Ilya Leoshkevich <iii@linux.ibm.com>
6800 * config/s390/constraints.md (ZL): New constraint.
6801 * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
6803 * config/s390/s390.md (movdi_larl): Remove.
6804 (movdi_64): Add the LARL alternative.
6806 2018-10-25 Ilya Leoshkevich <iii@linux.ibm.com>
6809 * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
6810 (RTX_CODE_HWINT_P): New macro.
6811 (rtx_code_size): Use RTX_CODE_HWINT_P ().
6813 2018-10-25 Jan Hubicka <jh@suse.cz>
6815 * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
6818 2018-10-25 Richard Biener <rguenther@suse.de>
6820 PR tree-optimization/87665
6821 PR tree-optimization/87745
6822 * tree-vectorizer.h (get_earlier_stmt): Remove.
6823 (get_later_stmt): Pick up UID from the original non-pattern stmt.
6825 2018-10-25 Sam Tebbs <sam.tebbs@arm.com>
6827 * options.texi (Deprecated): Move list to Var section.
6829 2018-10-24 Bill Schmidt <wschmidt@linux.ibm.com>
6830 Jinsong Ji <jji@us.ibm.com>
6832 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
6833 __vector long to __vector long long.
6834 (_mm_cvtpd_ps): Likewise.
6835 (_mm_cvttpd_epi32): Likewise.
6836 (_mm_cvtpi32_pd): Likewise.
6837 (_mm_unpackhi_epi64): Likewise.
6838 (_mm_unpacklo_epi64): Likewise.
6840 2018-10-24 Segher Boessenkool <segher@kernel.crashing.org>
6842 PR rtl-optimization/87720
6843 * combine.c (make_more_copies): Skip if the dest is pc_rtx.
6845 2018-10-24 Alexandre Oliva <aoliva@redhat.com>
6847 * gimple-ssa-isolate-paths.c
6848 (find_implicit_erroneous_behavior): Do not change code if the
6849 pass is running for warnings only.
6850 (find_explicit_erroneous_behavior): Likewise.
6852 2018-10-24 Michael Meissner <meissner@linux.ibm.com>
6854 * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
6855 Define as rs6000_mangle_decl_assembler_name.
6856 (rs6000_mangle_decl_assembler_name): If the user switched from IBM
6857 long double to IEEE long double, switch the names of the long
6858 double built-in functions to be <func>f128 instead of <func>l.
6860 2018-10-24 Martin Sebor <msebor@redhat.com>
6862 * doc/extend.texi (nonnull): List no-argument form. Reference
6863 -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
6865 2018-10-24 Richard Biener <rguenther@suse.de>
6867 * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
6869 2018-10-24 Martin Liska <mliska@suse.cz>
6871 PR tree-optimization/84436
6872 * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
6874 (switch_conversion::contains_linear_function_p): New.
6875 (switch_conversion::build_one_array): Support linear
6876 transformation on input.
6877 * tree-switch-conversion.h (struct switch_conversion): Add
6878 contains_linear_function_p declaration.
6880 2018-10-24 Richard Biener <rguenther@suse.de>
6882 * varasm.c (const_hash_1): Return hash of ADDR_EXPR
6883 if its argument is CONSTANT_CLASS_P.
6885 2018-10-11 Jan Hubicka <hubicka@ucw.cz>
6887 * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
6888 it is wrong for forward declarations.
6890 2018-10-24 Ilya Leoshkevich <iii@linux.ibm.com>
6892 * config/s390/s390.c (s390_check_qrst_address): Add the missing
6893 SYMBOL_REF_P () check.
6895 2018-10-24 Richard Biener <rguenther@suse.de>
6897 PR tree-optimization/87105
6898 * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
6899 dump classification.
6900 (vect_analyze_data_ref_accesses): Handle duplicate loads and
6901 stores by splitting the affected group after the fact.
6902 * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
6903 fail the SLP build because of size constraints.
6905 2018-10-24 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6907 * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
6908 * configure: Regenerate.
6909 * config.in: Regenerate.
6910 * varasm.c (mergeable_string_section): Use readonly_data_section
6911 if linker doesn't support SHF_MERGE with alignment > 8.
6912 (mergeable_constant_section): Likewise.
6914 2018-10-24 Richard Biener <rguenther@suse.de>
6916 PR tree-optimization/84013
6917 * tree-ssa-structalias.c (struct msdi_data): New struct for
6918 marshalling data to walk_stmt_load_store_ops.
6919 (maybe_set_dependence_info): Refactor as callback for
6920 walk_stmt_load_store_ops.
6921 (compute_dependence_clique): Set restrict info on all stmt kinds.
6923 2018-10-24 Martin Liska <mliska@suse.cz>
6925 * cgraph.c (cgraph_node::dump):
6926 Remove reduntant dumps and make tp_first_run dump more compact.
6928 2018-10-24 Richard Biener <rguenther@suse.de>
6930 PR tree-optimization/87665
6931 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
6934 2018-10-12 Jeff Law <law@redhat.com>
6936 * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
6939 2018-10-23 Richard Biener <rguenther@suse.de>
6941 * tree-vrp.c (add_assert_info): Guard dump_printf with
6943 * gimple-ssa-evrp-analyze.c
6944 (evrp_range_analyzer::record_ranges_from_incoming_edge):
6945 Use value_range::ignore_equivs_equal_p.
6947 2018-10-23 Richard Biener <rguenther@suse.de>
6949 PR tree-optimization/87105
6950 PR tree-optimization/87608
6951 * passes.def (pass_all_early_optimizations): Add early phi-opt
6953 * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
6954 addition to debug stmts.
6955 (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
6956 and abs replacement early.
6957 * tree-cfg.c (gimple_empty_block_p): Likewise.
6959 2018-10-23 Richard Earnshaw <rearnsha@arm.com>
6962 * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
6963 specified to configure.
6964 (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
6966 2018-10-23 Richard Biener <rguenther@suse.de>
6968 PR tree-optimization/87700
6969 * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
6971 2018-10-23 Jakub Jelinek <jakub@redhat.com>
6974 * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
6975 second argument from __mmask16 to __mmask8.
6976 * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
6977 _mm_mask_packs_epi32): Likewise.
6978 * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
6980 (_mm512_mask_prefetch_i64scatter_pd): Likewise. Formatting fix.
6982 2018-10-23 Richard Biener <rguenther@suse.de>
6984 * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
6986 2018-10-23 Richard Biener <rguenther@suse.de>
6988 PR tree-optimization/86144
6989 * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
6990 over simd attribute.
6992 2018-10-23 Richard Biener <rguenther@suse.de>
6994 PR tree-optimization/87693
6995 * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
6996 the case we do not find the taken edge.
6998 2018-10-22 Bill Schmidt <wschmidt@linux.ibm.com>
6999 Jinsong Ji <jji@us.ibm.com>
7001 * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
7002 (_mm_store_pd): Use unaligned vector type for pointer cast.
7003 (_mm_maskmoveu_si128): Likewise.
7004 * config/rs6000/xmmintrin.h (__m128_u): New typedef.
7005 (_mm_store_ps): Use unaligned vector type for pointer cast.
7007 2018-10-22 Paul Koning <ni1d@arrl.net>
7009 * symtab.c (symtab_node::increase_alignment): Correct max
7012 2018-10-22 Yury Gribov <tetra2005@gmail.com>
7014 PR tree-optimization/87633
7015 * match.pd: Do not generate unordered integer comparisons.
7017 2018-10-22 Segher Boessenkool <segher@kernel.crashing.org>
7019 PR rtl-optimization/87600
7020 * combine.c: Add include of expr.h.
7021 (cant_combine_insn_p): Do not combine moves from any hard non-fixed
7022 register to a pseudo.
7023 (make_more_copies): New function, add a copy to a new pseudo after
7024 the moves from hard registers into pseudos.
7025 (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
7026 later. Call make_more_copies.
7028 2018-10-22 Andrew Stubbs <ams@codesourcery.com>
7030 * lra-constraints.c (process_alt_operands): New local array,
7031 matching_early_clobber. Check matching_early_clobber before
7032 decrementing reject, and set matching_early_clobber after.
7034 2018-10-22 Segher Boessenkool <segher@kernel.crashing.org>
7037 * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
7038 call output_addr_const and hope for the best.
7040 2018-10-22 Richard Biener <rguenther@suse.de>
7042 * gimple-ssa-evrp-analyze.c
7043 (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
7044 smarter about what ranges to use.
7045 * tree-vrp.c (add_assert_info): Dump here.
7046 (register_edge_assert_for_2): Instead of here at multiple but
7049 * gcc.dg/tree-ssa/evrp12.c: New testcase.
7050 * gcc.dg/predict-6.c: Adjust.
7051 * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
7052 * gcc.dg/tree-ssa/vrp02.c: Likewise.
7053 * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
7055 2018-10-22 Steven Bosscher <steven@gcc.gnu.org>
7056 Richard Biener <rguenther@suse.de>
7058 * bitmap.h: Update data structure documentation, including a
7059 description of bitmap views as either linked-lists or splay trees.
7060 (struct bitmap_element_def): Update comments for splay tree bitmaps.
7061 (struct bitmap_head_def): Likewise.
7062 (bitmap_list_view, bitmap_tree_view): New prototypes.
7063 (bitmap_initialize_stat): Initialize a bitmap_head's indx and
7065 (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
7066 (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
7067 * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
7068 released bitmap element here.
7069 (bitmap_element_free): Remove.
7070 (bitmap_elt_clear_from): Work on splay tree bitmaps.
7071 (bitmap_list_link_element): Renamed from bitmap_element_link. Move
7072 this function similar ones such that linked-list bitmap implementation
7073 functions are grouped.
7074 (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
7075 and moved for grouping.
7076 (bitmap_list_insert_element_after): Renamed from
7077 bitmap_elt_insert_after, and moved for grouping.
7078 (bitmap_list_find_element): New function spliced from bitmap_find_bit.
7079 (bitmap_tree_link_left, bitmap_tree_link_right,
7080 bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
7081 bitmap_tree_link_element, bitmap_tree_unlink_element,
7082 bitmap_tree_find_element): New functions for splay-tree bitmap
7084 (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
7085 Renamed and moved, see above entries.
7086 (bitmap_tree_listify_from): New function to convert part of a splay
7087 tree bitmap to a linked-list bitmap.
7088 (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
7089 (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
7090 (bitmap_find_bit): Remove.
7091 (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
7092 bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
7093 Handle splay tree bitmaps.
7094 (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
7095 bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
7096 bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
7097 bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
7098 bitmap_intersect_compl_p, bitmap_ior_and_compl,
7099 bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
7100 bitmap_hash): Reject trying to act on splay tree bitmaps. Make
7101 corresponding changes to use linked-list specific bitmap_element
7102 manipulation functions as applicable for efficiency.
7103 (bitmap_tree_to_vec): New function.
7104 (debug_bitmap_elt_file): New function split out from ...
7105 (debug_bitmap_file): ... here. Handle splay tree bitmaps.
7106 (bitmap_print): Likewise.
7108 PR tree-optimization/63155
7109 * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
7111 * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
7114 2018-10-22 Martin Liska <mliska@suse.cz>
7116 PR tree-optimization/87686
7118 2018-08-29 Martin Liska <mliska@suse.cz>
7120 * tree-switch-conversion.c (switch_conversion::expand):
7121 Strenghten assumption about gswitch statements.
7123 2018-10-22 Martin Liska <mliska@suse.cz>
7125 * ipa-icf.c (sem_item::compare_attributes): Remove.
7126 (sem_item::compare_referenced_symbol_properties): Use
7127 attribute_list_equal instead.
7128 (sem_function::equals_wpa): Likewise.
7129 * ipa-icf.h: Remove compare_attributes.
7131 2018-10-22 Richard Biener <rguenther@suse.de>
7134 * mem-stats.h (mem_usage::operator==): Fix pasto.
7136 2018-10-22 Richard Biener <rguenther@suse.de>
7138 PR tree-optimization/87640
7139 * tree-vrp.c (set_value_range_with_overflow): Decompose
7141 (extract_range_from_binary_expr_1): Adjust.
7143 2018-10-22 Martin Jambor <mjambor@suse.cz>
7145 * tree-eh.h (stmt_could_throw_p): Add function parameter.
7146 (stmt_can_throw_external): Likewise.
7147 (stmt_can_throw_internal): Likewise.
7148 * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
7149 (lower_eh_constructs_2): Likewise.
7150 (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
7151 (stmt_can_throw_external): Likewise.
7152 (stmt_can_throw_internal): Likewise.
7153 (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
7154 (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
7155 (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
7156 (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
7157 (pass_lower_eh_dispatch::execute): Pass cfun to
7158 stmt_can_throw_external.
7159 (cleanup_empty_eh): Likewise.
7160 (verify_eh_edges): Pass cfun to stmt_could_throw_p.
7161 * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
7162 stmt_can_throw_external instead of pushing it to cfun.
7163 (symbol_table::create_edge): Likewise.
7164 * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
7165 stmt_can_throw_internal.
7166 * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
7167 to stmt_could_throw_p.
7168 * gimple-ssa-store-merging.c (handled_load): Pass cfun to
7169 stmt_can_throw_internal.
7170 (pass_store_merging::execute): Likewise.
7171 * gimple-ssa-strength-reduction.c
7172 (find_candidates_dom_walker::before_dom_children): Pass cfun to
7174 * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
7175 stmt_can_throw_internal.
7176 * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
7177 to stmt_can_throw_external.
7178 (check_stmt): Pass cfun to stmt_could_throw_p.
7179 (check_stmt): Pass cfun to stmt_can_throw_external.
7180 (pass_nothrow::execute): Likewise.
7181 * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
7182 * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
7183 stmt_can_throw_internal.
7184 (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
7185 (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
7186 (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
7187 * tree-complex.c (expand_complex_libcall): Pass cfun to
7188 stmt_could_throw_p and to stmt_can_throw_internal.
7189 (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
7190 * tree-inline.c (copy_edges_for_bb): Likewise.
7191 (maybe_move_debug_stmts_to_successors): Likewise.
7192 * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
7194 * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
7195 * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
7196 * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
7197 stmt_can_throw_internal.
7198 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
7199 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
7201 (mark_aliased_reaching_defs_necessary_1): Pass cfun to
7202 stmt_can_throw_internal.
7203 * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
7204 * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
7206 * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
7207 (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
7208 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
7209 (convert_mult_to_fma_1): Likewise.
7210 (convert_to_divmod): Likewise.
7211 * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
7212 * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
7213 * tree-ssa-propagate.c
7214 (substitute_and_fold_dom_walker::before_dom_children): Likewise.
7215 * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
7216 (maybe_optimize_range_tests): Likewise.
7217 (linearize_expr_tree): Likewise.
7218 (reassociate_bb): Likewise.
7219 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
7220 * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
7221 * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
7222 (handle_char_store): Likewise.
7223 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
7224 stmt_can_throw_internal.
7225 * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
7227 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
7228 (vectorizable_call): Pass cfun to stmt_can_throw_internal.
7229 (vectorizable_simd_clone_call): Likewise.
7230 * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
7231 (gimple_stringop_fixed_value): Likewise.
7233 2018-10-22 Ilya Leoshkevich <iii@linux.ibm.com>
7235 * config/s390/s390.c (s390_loadrelative_operand_p): Accept
7236 literal pool references.
7237 (s390_check_qrst_address): Adapt to the new behavior of
7238 s390_loadrelative_operand_p ().
7240 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
7243 * config/i386/sse.md (*andnot<mode>3_bcst): New.
7245 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
7248 * config/i386/sse.md (*<code><mode>3_bcst): New.
7250 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
7253 * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
7254 V4DI, V16SI and V8DI.
7255 (*sub<mode>3<mask_name>_bcst): New.
7256 (*add<mode>3<mask_name>_bcst): Likewise.
7258 2018-10-21 Bill Schmidt <wschmidt@linux.ibm.com>
7259 Jinsong Ji <jji@us.ibm.com>
7261 * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
7262 __m64 with __vector unsigned long long for compatibility.
7263 (_mm_movemask_epi8): Likewise.
7264 * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
7265 (_mm_cvttps_pi32): Likewise.
7266 (_mm_cvtpi32_ps): Likewise.
7267 (_mm_cvtps_pi16): Likewise.
7268 (_mm_loadh_pi): Likewise.
7269 (_mm_storeh_pi): Likewise.
7270 (_mm_movehl_ps): Likewise.
7271 (_mm_movelh_ps): Likewise.
7272 (_mm_loadl_pi): Likewise.
7273 (_mm_storel_pi): Likewise.
7274 (_mm_movemask_ps): Likewise.
7275 (_mm_shuffle_pi16): Likewise.
7277 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
7280 * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
7281 __builtin_ia32_vfnmsubpd512_mask.
7282 (_mm512_mask_fnmsub_round_pd): Likewise.
7283 (_mm512_fnmsub_pd): Likewise.
7284 (_mm512_mask_fnmsub_pd): Likewise.
7285 (_mm512_maskz_fnmsub_round_pd): Use
7286 __builtin_ia32_vfnmsubpd512_maskz.
7287 (_mm512_maskz_fnmsub_pd): Likewise.
7288 (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
7289 (_mm512_mask_fnmsub_round_ps): Likewise.
7290 (_mm512_fnmsub_ps): Likewise.
7291 (_mm512_mask_fnmsub_ps): Likewise.
7292 (_mm512_maskz_fnmsub_round_ps): Use
7293 __builtin_ia32_vfnmsubps512_maskz.
7294 (_mm512_maskz_fnmsub_ps): Likewise.
7295 * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
7296 __builtin_ia32_vfnmsubpd256_mask.
7297 (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
7298 (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
7299 (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
7300 (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
7301 (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
7302 (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
7303 (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
7304 (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
7305 * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
7306 __builtin_ia32_vfnmsubpd.
7307 (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
7308 (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
7309 (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
7310 (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
7311 (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
7312 * config/i386/i386-builtin.def: Add
7313 __builtin_ia32_vfnmsubpd256_mask,
7314 __builtin_ia32_vfnmsubpd256_maskz,
7315 __builtin_ia32_vfnmsubpd128_mask,
7316 __builtin_ia32_vfnmsubpd128_maskz,
7317 __builtin_ia32_vfnmsubps256_mask,
7318 __builtin_ia32_vfnmsubps256_maskz,
7319 __builtin_ia32_vfnmsubps128_mask,
7320 __builtin_ia32_vfnmsubps128_maskz,
7321 __builtin_ia32_vfnmsubpd512_mask,
7322 __builtin_ia32_vfnmsubpd512_maskz,
7323 __builtin_ia32_vfnmsubps512_mask,
7324 __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
7325 __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
7326 __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
7327 __builtin_ia32_vfnmsubpd256.
7328 * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
7329 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
7330 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
7332 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
7334 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
7336 (fmai_vmfnmsub_<mode><round_name>): Likewise.
7338 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
7341 * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
7342 __builtin_ia32_vfnmaddpd512_mask.
7343 (_mm512_mask_fnmadd_round_pd): Likewise.
7344 (_mm512_fnmadd_pd): Likewise.
7345 (_mm512_mask_fnmadd_pd): Likewise.
7346 (_mm512_maskz_fnmadd_round_pd): Use
7347 __builtin_ia32_vfnmaddpd512_maskz.
7348 (_mm512_maskz_fnmadd_pd): Likewise.
7349 (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
7350 (_mm512_mask_fnmadd_round_ps): Likewise.
7351 (_mm512_fnmadd_ps): Likewise.
7352 (_mm512_mask_fnmadd_ps): Likewise.
7353 (_mm512_maskz_fnmadd_round_ps): Use
7354 __builtin_ia32_vfnmaddps512_maskz.
7355 (_mm512_maskz_fnmadd_ps): Likewise.
7356 * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
7357 __builtin_ia32_vfnmaddpd256_mask.
7358 (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
7359 (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
7360 (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
7361 (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
7362 (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
7363 (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
7364 (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
7365 (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
7366 * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
7367 __builtin_ia32_vfnmaddpd.
7368 (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
7369 (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
7370 (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
7371 (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
7372 (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
7373 * config/i386/i386-builtin.def: Add
7374 __builtin_ia32_vfnmaddpd256_mask,
7375 __builtin_ia32_vfnmaddpd256_maskz,
7376 __builtin_ia32_vfnmaddpd128_mask,
7377 __builtin_ia32_vfnmaddpd128_maskz,
7378 __builtin_ia32_vfnmaddps256_mask,
7379 __builtin_ia32_vfnmaddps256_maskz,
7380 __builtin_ia32_vfnmaddps128_mask,
7381 __builtin_ia32_vfnmaddps128_maskz,
7382 __builtin_ia32_vfnmaddpd512_mask,
7383 __builtin_ia32_vfnmaddpd512_maskz,
7384 __builtin_ia32_vfnmaddps512_mask,
7385 __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
7386 __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
7387 __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
7388 __builtin_ia32_vfnmaddpd256.
7389 * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
7390 (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
7391 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
7393 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
7395 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
7397 (fmai_vmfnmadd_<mode><round_name>): Likewise.
7399 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
7402 * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
7403 __builtin_ia32_vfmsubpd512_mask.
7404 (_mm512_mask_fmsub_round_pd): Likewise.
7405 (_mm512_fmsub_pd): Likewise.
7406 (_mm512_mask_fmsub_pd): Likewise.
7407 (_mm512_maskz_fmsub_round_pd): Use
7408 __builtin_ia32_vfmsubpd512_maskz.
7409 (_mm512_maskz_fmsub_pd): Likewise.
7410 (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
7411 (_mm512_mask_fmsub_round_ps): Likewise.
7412 (_mm512_fmsub_ps): Likewise.
7413 (_mm512_mask_fmsub_ps): Likewise.
7414 (_mm512_maskz_fmsub_round_ps): Use
7415 __builtin_ia32_vfmsubps512_maskz.
7416 (_mm512_maskz_fmsub_ps): Likewise.
7417 * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
7418 __builtin_ia32_vfmsubpd256_mask.
7419 (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
7420 (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
7421 (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
7422 (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
7423 (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
7424 (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
7425 (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
7426 (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
7427 * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
7428 __builtin_ia32_vfmsubpd.
7429 (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
7430 (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
7431 (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
7432 (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
7433 (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
7434 * config/i386/i386-builtin.def: Add
7435 __builtin_ia32_vfmsubpd256_mask,
7436 __builtin_ia32_vfmsubpd256_maskz,
7437 __builtin_ia32_vfmsubpd128_mask,
7438 __builtin_ia32_vfmsubpd128_maskz,
7439 __builtin_ia32_vfmsubps256_mask,
7440 __builtin_ia32_vfmsubps256_maskz,
7441 __builtin_ia32_vfmsubps128_mask,
7442 __builtin_ia32_vfmsubps128_maskz,
7443 __builtin_ia32_vfmsubpd512_mask,
7444 __builtin_ia32_vfmsubpd512_maskz,
7445 __builtin_ia32_vfmsubps512_mask,
7446 __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
7447 __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
7448 __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
7449 __builtin_ia32_vfmsubpd256.
7450 * config/i386/sse.md (fma4i_fmsub_<mode>): New.
7451 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
7452 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
7454 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
7456 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
7458 (fmai_vmfmsub_<mode><round_name>): Likewise.
7460 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
7462 * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
7463 Remove plus. Renamed to ...
7464 (*sub<mode>3<mask_name>_bcst): This.
7465 (*add<mode>3<mask_name>_bcst_2): Renamede to ...
7466 (*add<mode>3<mask_name>_bcst): This.
7468 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
7471 * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
7473 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
7476 * i386/avx512vlintrin.h (_mm256_or_epi32): New.
7477 (_mm_or_epi32): Likewise.
7478 (_mm256_xor_epi32): Likewise.
7479 (_mm_xor_epi32): Likewise.
7480 (_mm256_or_epi64): Likewise.
7481 (_mm_or_epi64): Likewise.
7482 (_mm256_xor_epi64): Likewise.
7483 (_mm_xor_epi64): Likewise.
7485 2018-10-20 H.J. Lu <hongjiu.lu@intel.com>
7488 * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
7490 2018-10-20 Jakub Jelinek <jakub@redhat.com>
7493 * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
7495 2018-10-20 Andreas Schwab <schwab@linux-m68k.org>
7497 * doc/ux.texi: Move @section directly after @node.
7499 2018-10-19 Jakub Jelinek <jakub@redhat.com>
7503 * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
7504 depend closely nested inside of loop with ordered clause with
7507 2018-10-19 David Malcolm <dmalcolm@redhat.com>
7509 * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
7510 * doc/gccint.texi: Include ux.texi and use it in top-level menu.
7511 * doc/ux.texi: New file.
7513 2018-10-19 Segher Boessenkool <segher@kernel.crashing.org>
7515 * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
7516 be the first CR field allocated.
7518 2018-10-19 Richard Biener <rguenther@suse.de>
7521 * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
7522 TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
7524 2018-10-19 H.J. Lu <hongjiu.lu@intel.com>
7527 * config/i386/sse.md
7528 (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
7529 (*add<mode>3<mask_name>_bcst_2): Likewise.
7531 2018-10-19 H.J. Lu <hongjiu.lu@intel.com>
7533 * config/i386/sse.md
7534 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
7535 Replace nonimmediate_operand with register_operand.
7536 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
7538 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
7541 2018-10-19 Ilya Leoshkevich <iii@linux.ibm.com>
7543 PR rtl-optimization/87596
7544 * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
7545 lra_get_insn_recog_data () instead of lra_insn_recog_data[]
7546 for instructions in FROM..TO range.
7548 2018-10-19 Eric Botcazou <ebotcazou@adacore.com>
7550 * cfgexpand.c (expand_one_var): Use specific wording in error message
7551 for non-local frame variables.
7552 * stor-layout.c (layout_decl): Do not issue a warning for them.
7554 2018-10-19 Robin Dapp <rdapp@linux.ibm.com>
7556 * haifa-sched.c (priority): Add force_recompute parameter.
7557 (apply_replacement): Call priority () with force_recompute = true.
7558 (restore_pattern): Likewise.
7560 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
7562 * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
7563 HOST_BITS_PER_WIDE_INT.
7564 (test_vector_ops_duplicate): Likewise.
7566 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
7569 * config/i386/sse.md (VF_AVX512): New.
7570 (avx512bcst): Likewise.
7571 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
7573 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
7575 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
7578 2018-10-18 Jonathan Wakely <jwakely@redhat.com>
7580 * doc/invoke.texi (-dumpversion): Improve grammar.
7581 (-dumpfullversion): Make more consistent with -dumpversion.
7583 2018-10-18 Uros Bizjak <ubizjak@gmail.com>
7585 * config/i386/i386.c (ix86_emit_fp_unordered_jump):
7586 Set JUMP_LABEL to the jump insn.
7587 (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
7588 Predict emitted jump and add label to jump insn.
7590 2018-10-18 David Malcolm <dmalcolm@redhat.com>
7592 PR tree-optimization/87562
7593 * input.c (get_substring_ranges_for_loc): Use
7594 LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
7595 getting the linemap for the endpoint. Verify that it's either
7596 in the same linemap as the start point's spelling location, or
7597 at least in the same file.
7599 2018-10-18 Richard Biener <rguenther@suse.de>
7601 * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
7602 feed width-specific load/store costs through ix86_vec_cost.
7603 * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
7604 (k8_cost): Likewise.
7605 (bdver_cost): Likewise.
7606 (znver1_cost): Likewise.
7607 (btver1_cost): Likewise.
7608 (btver2_cost): Likewise.
7610 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
7612 * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
7613 to simplify subreg of vec_merge.
7615 2018-10-18 Richard Biener <rguenther@suse.de>
7617 * config/i386/i386.c: Fix costing of vector FMA.
7619 2018-10-18 Richard Biener <rguenther@suse.de>
7621 * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
7623 (ix86_builtin_vectorization_cost): For vec_construct properly
7624 cost insertion into SSE regs.
7625 (...): Adjust calls to ix86_vec_cost.
7627 2018-10-18 Richard Biener <rguenther@suse.de>
7631 2018-02-07 Richard Biener <rguenther@suse.de>
7633 PR tree-optimization/84204
7634 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
7637 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
7640 * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
7642 (test_vector_ops_duplicate): Add test for a scalar subreg of a
7643 VEC_MERGE of a VEC_DUPLICATE.
7645 2018-10-17 Joseph Myers <joseph@codesourcery.com>
7647 * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
7648 * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
7649 * doc/standards.texi (C Language): Document C2X.
7650 * dwarf2out.c (highest_c_language), config/rl78/rl78.c
7651 (rl78_option_override): Handle "GNU C2X" language name.
7653 2018-10-17 Joseph Myers <joseph@codesourcery.com>
7655 * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
7656 Document C17 as published in 2018.
7658 2018-10-17 Eric Botcazou <ebotcazou@adacore.com>
7661 * fold-const.c (fold_truth_andor_1): If the right side is not constant,
7662 bail out if both sides do not have the same storage order.
7664 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
7666 * bitmap.c (bitmap_head::dump): New.
7667 * bitmap.h (bitmap_head): Add dump().
7668 * gimple-ssa-evrp-analyze.c
7669 (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
7670 (evrp_range_analyzer::set_ssa_range_info): Same.
7671 (evrp_range_analyzer::record_ranges_from_phis): Same.
7672 (evrp_range_analyzer::record_ranges_from_stmt): Same.
7673 * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
7674 * gimple-ssa-sprintf.c (get_int_range): Same.
7675 (format_integer): Same.
7676 (sprintf_dom_walker::handle_gimple_call): Same.
7677 * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
7678 (ipcp_vr_lattice::top_p): Same.
7679 (ipcp_vr_lattice::bottom_p): Same.
7680 (ipcp_vr_lattice::set_to_bottom): Same.
7681 (ipa_vr_operation_and_type_effects): Same.
7682 (propagate_vr_across_jump_function): Same.
7683 (ipcp_store_vr_results): Same.
7684 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
7685 (ipa_print_node_jump_functions_for_edge): Same.
7686 (ipa_get_value_range): Same.
7687 (ipa_compute_jump_functions_for_edge): Same.
7688 (ipa_write_jump_function): Same.
7689 * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
7690 * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
7692 * vr-values.c (set_value_range_to_nonnegative): Same.
7693 (set_value_range_to_truthvalue): Same.
7694 (vr_values::get_value_range): Same.
7695 (vr_values::set_defs_to_varying): Same.
7696 (vr_values::update_value_range): Same.
7697 (symbolic_range_based_on_p): Same.
7698 (vr_values::op_with_boolean_value_range_p): Same.
7699 (vr_values::extract_range_for_var_from_comparison_expr): Same.
7700 (vr_values::extract_range_from_ssa_name): Same.
7701 (vr_values::extract_range_from_binary_expr): Same.
7702 (vr_values::extract_range_from_unary_expr): Same.
7703 (vr_values::extract_range_from_cond_expr): Same.
7704 (vr_values::extract_range_from_comparison): Same.
7705 (vr_values::check_for_binary_op_overflow): Same.
7706 (vr_values::extract_range_basic): Same.
7707 (vr_values::extract_range_from_assignment): Same.
7708 (compare_ranges): Same.
7709 (compare_range_with_value): Same.
7710 (vr_values::adjust_range_with_scev): Same.
7711 (vrp_valueize): Same.
7712 (vrp_valueize_1): Same.
7713 (vr_values::get_vr_for_comparison): Same.
7714 (vr_values::compare_name_with_value): Same.
7715 (vr_values::compare_names): Same.
7716 (vr_values::vrp_evaluate_conditional): Same.
7717 (find_case_label_ranges): Same.
7718 (vr_values::vrp_visit_switch_stmt): Same.
7719 (vr_values::extract_range_from_phi_node): Same.
7720 (vr_values::simplify_div_or_mod_using_ranges): Same.
7721 (vr_values::simplify_bit_ops_using_ranges): Same.
7722 (test_for_singularity): Same.
7723 (range_fits_type_p): Same.
7724 (vr_values::simplify_cond_using_ranges_1): Same.
7725 (vr_values::simplify_switch_using_ranges): Same.
7726 (vr_values::simplify_float_conversion_using_ranges): Same.
7727 (vr_values::two_valued_val_range_p): Same.
7728 (vr_values::add_equivalence): Move to value_range::equiv_add.
7729 * vr-values.h (vr_values::add_equivalence): Remove.
7730 (VR_INITIALIZER): Remove.
7731 * tree-vrp.c (value_range::set): New.
7732 (value_range::equiv_add): New.
7733 (value_range::value_range): New.
7734 (value_range::deep_copy): New.
7735 (value_range::check): New.
7736 (value_range::equal_p): New.
7737 (value_range::ignore_equivs_equal_p): New.
7738 (value_range::operator==): New.
7739 (value_range::operator!=): New.
7740 (value_range::symbolic_p): New.
7741 (value_range::numeric_p): New.
7742 (value_range::set_undefined): New.
7743 (value_range::set_varying): New.
7744 (value_range::may_contain_p): New.
7745 (value_range::equiv_clear): New.
7746 (value_range::singleton_p): New.
7747 (value_range::intersect): New.
7748 (value_range::dump): New.
7749 (value_range::set_and_canonicalize): New.
7750 (set_value_range): Adjust for value_range API.
7751 (set_value_range_to_undefined): Same.
7752 (set_value_range_to_varying): Same.
7753 (set_and_canonicalize_value_range): Same.
7754 (set_value_range_to_nonnull): Same.
7755 (set_value_range_to_null): Same.
7756 (range_is_null): Same.
7757 (range_is_nonnull): Same.
7758 (range_int_cst_p): Same.
7759 (range_int_cst_singleton_p): Same.
7760 (symbolic_range_p): Same.
7761 (range_includes_zero_p): Same.
7762 (value_range_constant_singleton): Same.
7763 (vrp_set_zero_nonzero_bits): Same.
7764 (ranges_from_anti_range): Same.
7765 (extract_range_into_wide_ints): Same.
7766 (extract_range_from_multiplicative_op): Same.
7767 (set_value_range_with_overflow): Same.
7768 (extract_range_from_binary_expr_1): Same.
7769 (extract_range_from_unary_expr): Same.
7770 (dump_value_range): Same.
7771 (debug_value_range): Same.
7772 (vrp_prop::check_array_ref): Same.
7773 (vrp_prop::check_mem_ref): Same.
7774 (vrp_prop::vrp_initialize): Same.
7775 (vrp_prop::visit_stmt): Same.
7776 (intersect_ranges): Same.
7777 (vrp_prop::visit_phi): Same.
7778 (vrp_prop::vrp_finalize): Same.
7779 (determine_value_range_1): Same.
7780 (determine_value_range): Same.
7781 (vrp_intersect_ranges_1): Rename to...
7782 (vrp_intersect_1): this.
7783 (vrp_intersect_ranges): Rename to...
7784 (value_range::intersect_helper): ...this.
7785 (vrp_meet_1): Rename to...
7786 (value_range::union_helper): ...this.
7787 (vrp_meet): Rename to...
7788 (value_range::union_): ...this.
7789 (copy_value_range): Remove.
7790 * tree-vrp.h (struct value_range): Rewrite into a proper class.
7791 (value_range::vrtype): New.
7792 (value_range::type): New.
7793 (value_range::equiv): New.
7794 (value_range::min): New.
7795 (value_range::max): New.
7796 (value_range::varying_p): New.
7797 (value_range::undefined_p): New.
7798 (value_range::null_p): New.
7799 (value_range::equiv_add): New.
7800 (copy_value_range): Remove.
7802 2018-10-17 David Malcolm <dmalcolm@redhat.com>
7804 * Makefile.in (SELFTEST_TARGETS): New.
7805 (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
7806 (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
7807 (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
7809 (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
7810 (selftest-c++-gdb, selftest-c++-valgrind): Move to
7812 * configure: Regenerate.
7813 * configure.ac (selftest_languages): New.
7815 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
7817 * tree-vrp.c (extract_range_from_multiplicative_op): Remove
7818 overflow wraps argument.
7819 (extract_range_from_binary_expr_1): Do not pass overflow wraps to
7820 wide_int_range_multiplicative_op.
7821 * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
7822 overflow wraps argument.
7823 (wide_int_range_multiplicative_op): Same.
7824 (wide_int_range_lshift): Same.
7825 (wide_int_range_div): Same.
7826 * wide-int-range.h (wide_int_range_multiplicative_op): Same.
7827 (wide_int_range_lshift): Same.
7828 (wide_int_range_div): Same.
7830 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
7832 * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
7833 use sign as argument.
7834 * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
7835 wide_int_range_shift_undefined_p.
7837 2018-10-16 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7839 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
7841 (@despeculate_copy<ALLI_TI:mode>): ... This.
7842 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
7845 2018-10-16 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7847 * config.gcc: Obsolete *-*-solaris2.10*.
7848 * doc/install.texi (Specific, *-*-solaris2*): Document it.
7850 2018-10-12 Jeff Law <law@redhat.com>
7852 * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
7853 reg + sym +- const_int addressing modes.
7855 2018-10-15 David Malcolm <dmalcolm@redhat.com>
7857 * common.opt (fdiagnostics-minimum-margin-width=): New option.
7858 * diagnostic-show-locus.c (layout::layout): Apply the minimum
7860 (layout::start_annotation_line): Only print up to 3 of the
7861 margin character, to avoid touching the left-hand side.
7862 (selftest::test_diagnostic_show_locus_fixit_lines): Update for
7863 minimum margin width, as set by test_diagnostic_context's ctor.
7864 (selftest::test_fixit_insert_containing_newline): Likewise.
7865 (selftest::test_fixit_insert_containing_newline_2): Likewise.
7866 (selftest::test_line_numbers_multiline_range): Clear
7867 dc.min_margin_width.
7868 * diagnostic.c (diagnostic_initialize): Initialize
7870 * diagnostic.h (struct diagnostic_context): Add field
7872 * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
7873 * opts.c (common_handle_option): Handle
7874 OPT_fdiagnostics_minimum_margin_width_.
7875 * selftest-diagnostic.c
7876 (selftest::test_diagnostic_context::test_diagnostic_context):
7877 Initialize min_margin_width to 6.
7878 * toplev.c (general_init): Initialize global_dc->min_margin_width.
7880 2018-10-15 David Malcolm <dmalcolm@redhat.com>
7882 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
7883 Fix usage of "error_at_rich_loc" in the comment.
7885 2018-10-15 Renlin Li <renlin.li@arm.com>
7888 * tree-vectorizer.c (try_vectorize_loop_1): Don't use
7889 if-conversioned loop when it contains ifn with types not
7890 supported by backend.
7891 * internal-fn.c (expand_direct_optab_fn): Add an assert.
7892 (direct_internal_fn_supported_p): New helper function.
7893 * internal-fn.h (direct_internal_fn_supported_p): Declare.
7895 2018-10-15 Jakub Jelinek <jakub@redhat.com>
7898 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
7899 Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
7900 OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
7902 2018-10-15 Bin Cheng <bin.cheng@linux.alibaba.com>
7904 PR tree-optimization/87022
7905 * tree-loop-distribution.c (pg_add_dependence_edges): Check all
7906 bits in dist vector rather than the first one.
7908 2018-10-15 Richard Biener <rguenther@suse.de>
7911 * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
7912 (visit_loadstore): When a used restrict tag escaped verify that
7913 the points-to solution of "other" pointers do not include
7915 (compute_dependence_clique): If a used restrict tag escaped
7916 communicated that down to visit_loadstore.
7918 2018-10-15 Andreas Krebbel <krebbel@linux.ibm.com>
7920 * config/s390/s390.c (s390_expand_vec_init): Force vector element
7921 into reg if it isn't a general operand.
7923 2018-10-14 H.J. Lu <hongjiu.lu@intel.com>
7926 * config/i386/sse.md (*vec_dupv2di): Add register source to
7929 2018-10-14 H.J. Lu <hongjiu.lu@intel.com>
7932 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
7933 Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
7934 OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
7935 OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
7937 2018-10-13 Eric Botcazou <ebotcazou@adacore.com>
7939 * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
7940 (notice_args_size): Set it in the current trace if no insn that can
7941 throw internally has been seen yet.
7942 (connect_traces): When connecting args_size between traces, allow the
7943 incoming values not to match if there is an insn setting it before the
7944 first insn that can throw internally; in that case, force the creation
7945 of a CFI note on this latter insn.
7947 2018-10-13 Jonathan Wakely <jwakely@redhat.com>
7949 * opt-problem.h (opt_wrapper): Use template-argument-list when naming
7950 the base class, because using the injected-class-name was not clearly
7951 specified until DR 176.
7953 2018-10-12 Paul Koning <ni1d@arrl.net>
7955 * config/pdp11/pdp11.md (doloop_end): New expander.
7956 (doloop_end_insn): renamed from "doloop_end".
7957 (addqi3): New pattern.
7958 (subqi3): New pattern.
7959 * config/pdp11/predicates.md (incdec_operand): New predicate.
7961 2018-10-12 Yury Gribov <tetra2005@gmail.com>
7964 * real.c (format_helper::can_represent_integral_type_p): New function
7965 * real.h (format_helper::can_represent_integral_type_p): Ditto.
7966 * match.pd: New pattern.
7968 2018-10-12 Alexandre Oliva <oliva@adacore.com>
7970 * configure.ac: Introduce --enable-large-address-aware
7971 to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
7972 * doc/install.texi: Document it.
7973 * configure, config.in: Rebuilt.
7974 * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
7975 based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
7976 (LINK_SPEC): Insert it.
7977 * config/i386/mingw-w64.h: Likewise.
7979 * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
7981 2018-10-12 Peter Bergner <bergner@linux.ibm.com>
7983 PR rtl-optimization/87600
7984 * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
7986 2018-10-12 Paul Koning <ni1d@arrl.net>
7988 * doc/md.texi (doloop_end): Document that the pattern code may
7989 need to check operand mode.
7991 2018-10-12 Wilco Dijkstra <wdijkstr@arm.com>
7993 * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
7994 to zero-extend between int and floating-point registers.
7995 (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
7996 ldp into floating-point registers. Add type and arch attributes.
7997 (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
7998 Use f_loads for type attribute.
8000 2018-10-11 Martin Sebor <msebor@redhat.com>
8002 * doc/extend.texi (attribute packed): Correct typos.
8004 2018-10-11 Martin Sebor <msebor@redhat.com>
8006 * doc/extend.texi (attribute flatten): Mention interaction with
8009 2018-10-11 Jan Hubicka <hubicka@ucw.cz>
8012 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8013 Set new_decl virtual flag to zero.
8015 2018-10-11 Martin Sebor <msebor@redhat.com>
8018 * doc/extend.texi (attribute format_arg): Discuss using multiple
8019 attributes on a single function.
8021 2018-10-11 Giuliano Belinassi <giuliano.belinassi@usp.br>
8023 PR tree-optimization/86829
8024 * match.pd (sin (atan (x))): New simplification rules.
8025 (cos (atan (x))): Likewise.
8026 * real.c (build_sinatan_real): New function.
8027 * real.h (build_sinatan_real): Prototype.
8029 2018-10-11 Will Schmidt <will_schmidt@vnet.ibm.com>
8031 * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
8033 (fold_mergeeo_helper): New helper function.
8034 (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
8035 intrinsics. Correct some whitespace indentation issues.
8037 2018-10-11 Wilco Dijkstra <wdijkstr@arm.com>
8040 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
8041 Use HOST_WIDE_INT_1U for shift.
8043 2018-10-11 Doug Rupp <rupp@adacore.com>
8044 Olivier Hainque <hainque@adacore.com>
8046 * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
8047 Pass --relax to the linker for RTPs.
8048 (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
8050 2018-10-11 Andrew Stubbs <ams@codesourcery.com>
8051 Jan Hubicka <jh@suse.cz>
8052 Martin Jambor <mjambor@suse.cz>
8054 * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
8055 the same elements are repeated rather than printing all of them.
8056 * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
8057 "repeated" elements.
8058 * read-rtl-function.c (test_loading_repeat): New function.
8059 (read_rtl_function_c_tests): Call test_loading_repeat.
8060 * rtl-tests.c (test_dumping_repeat): New function.
8061 (rtl_tests_c_tests): Call test_dumping_repeat.
8063 2018-10-11 Richard Biener <rguenther@suse.de>
8065 * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
8066 bdver?_cost): Unify to ...
8067 (bdver_memcpy, bdver_memset, bdver_cost): ... this.
8068 * config/i386/i386.c (processor_cost_table): Adjust.
8070 2018-10-10 Eric Botcazou <ebotcazou@adacore.com>
8073 * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
8074 the thunk when expanding to GIMPLE.
8076 2018-10-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
8079 * varasm.c (mergeable_string_section): Don't try to move zero-length
8080 strings to the merge section.
8082 2018-10-10 Uros Bizjak <ubizjak@gmail.com>
8085 * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
8087 2018-10-10 Jakub Jelinek <jakub@redhat.com>
8090 * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
8091 to special_args set.
8093 2018-10-10 Richard Biener <rguenther@suse.de>
8095 * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
8096 reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
8097 reduc_plus_scal_v4sf): Merge into pattern reducing to half width
8098 and recursing and pattern terminating the recursion on SSE
8099 vector width using ix86_expand_reduc.
8100 (reduc_sminmax_scal_<mode>): Split into part reducing to half
8101 width and recursing and SSE2 vector variant doing the final
8102 reduction with ix86_expand_reduc.
8103 (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
8104 with terminating the recursion at AVX level, splitting that
8107 2018-10-09 David Malcolm <dmalcolm@redhat.com>
8109 * genmatch.c (error_cb): Rename to...
8110 (diagnostic_cb): ...this, converting int params to enums.
8111 (fatal_at): Update for renaming.
8112 (warning_at): Likewise.
8114 * input.c (selftest::ebcdic_execution_charset::apply):
8115 Update for renaming of...
8116 (selftest::ebcdic_execution_charset::on_error): ...this, renaming
8118 (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
8119 converting level and reason to enums.
8120 (class selftest::lexer_error_sink): Rename to...
8121 (class selftest::lexer_test_options): ...this, renaming field
8122 "m_errors" to "m_diagnostics".
8123 (selftest::lexer_test_options::apply): Update for renaming of...
8124 (selftest::lexer_test_options::on_error): ...this, renaming to...
8125 (selftest::lexer_test_options::on_diagnostic): ...this
8126 converting level and reason to enums.
8127 (selftest::test_lexer_string_locations_raw_string_unterminated):
8128 Update for renamings.
8129 * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
8132 2018-10-09 Paul A. Clarke <pc@us.ibm.com>
8134 * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
8135 * config/rs6000/pmmintrin.h: New file.
8137 2018-10-09 Eric Botcazou <ebotcazou@adacore.com>
8139 PR tree-optimization/86659
8140 * gimple-match.h (gimple_match_op constructors): Initialize reverse.
8142 2018-10-09 Richard Biener <rguenther@suse.de>
8144 PR tree-optimization/63155
8145 * tree-ssa-structalias.c: Include tree-ssa.h.
8146 (get_constraint_for_ssa_var): For undefs return nothing_id.
8147 (find_func_aliases): Cleanup PHI handling.
8149 2018-10-09 Richard Biener <rguenther@suse.de>
8151 * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
8154 2018-10-09 Martin Liska <mliska@suse.cz>
8156 * asan.c (asan_emit_stack_protection): If a stack variable
8157 is located in a same file as current function, then emit
8158 line info into variable definition string.
8160 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
8162 * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
8165 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
8167 * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
8170 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
8172 PR tree-optimization/86659
8173 * gimple-match.h (struct gimple_match_op): Add reverse field.
8174 (gimple_match_op::set_op): New overloaded method.
8175 * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
8176 the REF_REVERSE_STORAGE_ORDER flag on the value.
8177 (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
8178 REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
8180 2018-10-08 Richard Sandiford <richard.sandiford@arm.com>
8183 * gimple-ssa-backprop.c (backprop::intersect_uses): Use
8184 FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
8186 2018-10-08 H.J. Lu <hongjiu.lu@intel.com>
8189 * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
8190 Defined with __builtin_ia32_vfmaddsubpd512_mask.
8192 2018-10-08 Richard Biener <rguenther@suse.de>
8194 * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
8195 cost the same as AVX128 ones.
8197 2018-10-08 Paul Koning <ni1d@arrl.net>
8199 * config/pdp11/pdp11-protos.h (output_block_move): Remove.
8200 (expand_block_move): New function.
8201 * config/pdp11/pdp11.c (output_block_move): Remove.
8202 (expand_block_move): New function.
8203 * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
8204 * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
8205 (*movmemhi1): Remove.
8207 2018-10-08 Robin Dapp <rdapp@linux.ibm.com>
8209 * config/s390/2827.md: Increase latencies for some FP instructions.
8211 2018-10-08 Richard Biener <rguenther@suse.de>
8213 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
8215 * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
8216 * tree-vectorizer.h (dump_stmt_cost): Adjust.
8217 (add_stmt_cost): Dump return value of the hook.
8219 2018-10-08 Richard Biener <rguenther@suse.de>
8221 PR tree-optimization/63155
8222 * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
8223 (ssa_propagation_engine::ssa_propagate): Remove redundant
8224 bitmap bit clearing.
8226 2018-10-05 Peter Bergner <bergner@linux.ibm.com>
8228 PR rtl-optimization/86939
8229 PR rtl-optimization/87479
8230 * ira.h (non_conflicting_reg_copy_p): New prototype.
8231 * ira-lives.c (ignore_reg_for_conflicts): New static variable.
8232 (make_hard_regno_dead): Don't add conflicts for register
8233 ignore_reg_for_conflicts.
8234 (make_object_dead): Likewise.
8235 (non_conflicting_reg_copy_p): New function.
8236 (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
8237 Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
8238 * lra-lives.c (ignore_reg_for_conflicts): New static variable.
8239 (make_hard_regno_dead): Don't add conflicts for register
8240 ignore_reg_for_conflicts. Remove special conflict handling of
8241 REAL_PIC_OFFSET_TABLE_REGNUM. Remove now unused argument
8242 check_pic_pseudo_p and update callers.
8243 (mark_pseudo_dead): Don't add conflicts for register
8244 ignore_reg_for_conflicts.
8245 (process_bb_lives): Set ignore_reg_for_conflicts for copies.
8247 2018-10-05 Andrew Waterman <andrew@sifive.com>
8248 Jim Wilson <jimw@sifive.com>
8250 * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
8251 Add define_expand. Add ! HONOR_SNANS check to current pattern. Add
8252 new pattern using HONOR_SNANS that emits one extra instruction.
8254 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
8256 * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
8257 patterns): Merge SI and DI patterns to a GPR pattern.
8258 (unnamed define_insn and define_split for record form of that): Merge
8259 to a single define_insn_and_split pattern.
8261 2018-10-05 David Malcolm <dmalcolm@redhat.com>
8264 * input.c (expand_location_to_spelling_point): Add param "aspect"
8265 and use rather than hardcoding LOCATION_ASPECT_CARET.
8266 (get_substring_ranges_for_loc): Handle the case of a single token
8267 within a macro expansion.
8268 * input.h (expand_location_to_spelling_point): Add "aspect" param,
8269 defaulting to LOCATION_ASPECT_CARET.
8271 2018-10-05 Paul Koning <ni1d@arrl.net>
8273 * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
8274 (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
8275 (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
8276 (pdp11_guard_type): New function.
8278 2018-10-05 Paul Koning <ni1d@arrl.net>
8280 * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
8281 * config/pdp11/pdp11.opt (mfloat32): Remove.
8283 * doc/invoke.texi (pdp11 -mfloat32): Remove:
8284 (pdp11 -mfloat64): Remove.
8286 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
8288 * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
8289 (*cmp<mode>_cc_i387): Ditto.
8290 (*cmpu<mode>_cc_i387): Ditto.
8291 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
8292 * config/i386/i386.c (ix86_expand_fp_compare): Remove
8294 <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
8295 Emit x86_sahf_1 pattern.
8296 (ix86_expand_compare): Update call to ix86_expand_fp_compare.
8297 (ix86_expand_carry_flag_compare): Ditto.
8299 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
8301 * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
8302 to reg_or_0_operand. Add "C" constraint.
8303 (*cmpxf_cc_i387): Ditto.
8304 (*cmp<mode>_i387): Change operand 2 predicate
8305 to nonimm_or_0_operand. Add "C" constraint.
8306 (*cmp<mode>_cc_i387): Ditto.
8307 (*cmp<mode>_0_i387): Remove insn pattern.
8308 (*cmp<mode>_0_cc_i387): Ditto.
8310 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
8312 * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
8313 * config/i386/predicates.md (nonimm_or_0_operand): Rename
8314 from vector_move_operand. Update all uses.
8316 2018-10-05 Martin Sebor <msebor@redhat.com>
8318 PR tree-optimization/87490
8319 * builtins.c (expand_builtin_strnlen): Handle a null data.decl
8322 2018-10-05 Richard Biener <rguenther@suse.de>
8324 PR tree-optimization/63155
8325 * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
8326 vertical space in dumpfiles.
8327 * tree-ssa-propagate.h
8328 (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
8329 * tree-ssa-propagate.c (cfg_blocks_back): New global.
8330 (ssa_edge_worklist_back): Likewise.
8331 (curr_order): Likewise.
8332 (cfg_blocks_get): Remove abstraction.
8333 (cfg_blocks_add): Likewise.
8334 (cfg_blocks_empty_p): Likewise.
8335 (add_ssa_edge): Add to current or next worklist based on
8337 (add_control_edge): Likewise.
8338 (ssa_propagation_engine::process_ssa_edge_worklist): Fold
8340 (ssa_propagation_engine::ssa_propagate): ... here. Unify
8341 iteration from CFG and SSA edge worklist so we process
8342 everything in RPO order, prioritizing forward progress
8344 (ssa_prop_init): Allocate new worklists, do not dump
8346 (ssa_prop_fini): Free new worklists.
8348 2018-10-05 Richard Biener <rguenther@suse.de>
8350 * tree-core.h (tree_block::abstract_flag): Remove.
8351 (tree_block::block_num): Make full 32bits.
8352 * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
8353 * tree.h (BLOCK_ABSTRACT): Remove.
8354 * dwarf2out.c (gen_lexical_block_die): Remove dead code
8355 resulting from BLOCK_ABSTRACT being always false.
8356 (gen_inlined_subroutine_die): Likewise.
8357 (gen_block_die): Likewise.
8358 * tree.c (block_ultimate_origin): Likewise.
8359 * tree-pretty-print.c (dump_block_node): Remove code dealing
8360 with BLOCK_ABSTRACT.
8361 * tree-ssa-live.c (dump_scope_block): Likewise.
8362 * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
8363 * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
8365 2018-10-05 Richard Biener <rguenther@suse.de>
8367 * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
8368 is asked for initialize mode to the component mode of the
8371 2018-10-05 H.J. Lu <hongjiu.lu@intel.com>
8374 * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
8375 assembler for -mavx.
8376 * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
8378 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
8381 * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
8383 * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
8384 to be DImode. When using mffscrn, force the operand to a register.
8386 2018-10-04 Uros Bizjak <ubizjak@gmail.com>
8388 * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
8389 from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
8390 X87MODEF mode iterator.
8391 (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
8392 *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
8393 X87MODEF mode iterator.
8395 2018-10-04 Vinay Kumar <vinay.kumar@blackfigtech.com>
8397 * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
8398 -Wno-prio-ctor-dtor.
8400 2018-10-04 David Malcolm <dmalcolm@redhat.com>
8402 * Makefile.in (OBJS): Add opt-problem.o.
8403 * dump-context.h: Include "selftest.h.
8404 (selftest::temp_dump_context): New forward decl.
8405 (class dump_context): Make friend of class
8406 selftest::temp_dump_context.
8407 (dump_context::dump_loc_immediate): New decl.
8408 (class dump_pretty_printer): Move here from dumpfile.c.
8409 (class temp_dump_context): Move to namespace selftest.
8410 (temp_dump_context::temp_dump_context): Add param
8411 "forcibly_enable_dumping".
8412 (selftest::verify_dumped_text):
8413 (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
8414 (selftest::verify_item):
8415 (ASSERT_IS_TEXT): Move here from dumpfile.c.
8416 (ASSERT_IS_TREE): Likewise.
8417 (ASSERT_IS_GIMPLE): Likewise.
8418 * dumpfile.c (dump_context::dump_loc): Move immediate dumping
8420 (dump_context::dump_loc_immediate): ...this new function.
8421 (class dump_pretty_printer): Move to dump-context.h.
8422 (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
8423 (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
8424 (temp_dump_context::temp_dump_context): Move to "selftest"
8425 namespace. Add param "forcibly_enable_dumping", and use it to
8426 conditionalize the use of m_pp;
8427 (selftest::verify_dumped_text): Make non-static.
8428 (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
8429 (selftest::verify_item): Make non-static.
8430 (ASSERT_IS_TEXT): Move to dump-context.h.
8431 (ASSERT_IS_TREE): Likewise.
8432 (ASSERT_IS_GIMPLE): Likewise.
8433 (selftest::test_capture_of_dump_calls): Pass "true" for new
8434 param of temp_dump_context.
8435 * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
8436 it to MSG_ALL_PRIORITIES. Update values of TDF_COMPARE_DEBUG and
8438 * opt-problem.cc: New file.
8439 * opt-problem.h: New file.
8440 * optinfo-emit-json.cc
8441 (selftest::test_building_json_from_dump_calls): Pass "true" for
8442 new param of temp_dump_context.
8443 * optinfo.cc (optinfo_kind_to_dump_flag): New function.
8444 (optinfo::emit_for_opt_problem): New function.
8445 (optinfo::emit): Clarity which emit_item is used.
8446 * optinfo.h (optinfo::get_dump_location): New accessor.
8447 (optinfo::emit_for_opt_problem): New decl.
8448 (optinfo::emit): Make const.
8449 * selftest-run-tests.c (selftest::run_tests): Call
8450 selftest::opt_problem_cc_tests.
8451 * selftest.h (selftest::opt_problem_cc_tests): New decl.
8452 * tree-data-ref.c (dr_analyze_innermost): Convert return type from
8453 bool to opt_result, converting fprintf messages to
8454 opt_result::failure_at calls. Add "stmt" param for use by the
8456 (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
8457 (runtime_alias_check_p): Convert return type from bool to
8458 opt_result, converting dump_printf calls to
8459 opt_result::failure_at, using the statement DDR_A for their
8461 (find_data_references_in_stmt): Convert return type from bool to
8462 opt_result, converting "return false" to opt_result::failure_at
8464 * tree-data-ref.h: Include "opt-problem.h".
8465 (dr_analyze_innermost): Convert return type from bool to opt_result,
8466 and add a const gimple * param.
8467 (find_data_references_in_stmt): Convert return type from bool to
8469 (runtime_alias_check_p): Likewise.
8470 * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
8471 dr_analyze_innermost.
8472 * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
8473 Convert return type from bool to opt_result, adding a message for
8474 the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
8475 (vect_analyze_data_ref_dependence): Convert return type from bool
8476 to opt_result. Change sense of return type from "false"
8477 effectively meaning "no problems" to "false" meaning a problem,
8478 so that "return false" becomes "return opt_result::success".
8479 Convert "return true" calls to opt_result::failure_at, using
8480 the location of statement A rather than vect_location.
8481 (vect_analyze_data_ref_dependences): Convert return type from bool
8483 (verify_data_ref_alignment): Likewise, converting dump_printf_loc
8484 calls to opt_result::failure_at, using the stmt location rather
8486 (vect_verify_datarefs_alignment): Convert return type from bool
8488 (vect_enhance_data_refs_alignment): Likewise. Split local "stat"
8489 into multiple more-tightly-scoped copies.
8490 (vect_analyze_data_refs_alignment): Convert return type from bool
8492 (vect_analyze_data_ref_accesses): Likewise, converting a
8493 "return false" to a "return opt_result::failure_at", adding a
8495 (vect_prune_runtime_alias_test_list): Convert return type from
8496 bool to opt_result, converting dump_printf_loc to
8497 opt_result::failure_at. Add a %G to show the pertinent statement,
8498 and use the stmt's location rather than vect_location.
8499 (vect_find_stmt_data_reference): Convert return type from
8500 bool to opt_result, converting dump_printf_loc to
8501 opt_result::failure_at, using stmt's location.
8502 (vect_analyze_data_refs): Convert return type from bool to
8503 opt_result. Convert "return false" to "return
8504 opt_result::failure_at", adding messages as needed.
8505 * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
8506 type from bool to opt_result.
8507 (vect_determine_vf_for_stmt): Likewise.
8508 (vect_determine_vectorization_factor): Likewise, converting
8509 dump_printf_loc to opt_result::failure_at, using location of phi
8510 rather than vect_location.
8511 (vect_analyze_loop_form_1): Convert return type from bool to
8512 opt_result, converting dump_printf_loc calls, retaining the use of
8514 (vect_analyze_loop_form): Convert return type from loop_vec_info
8515 to opt_loop_vec_info.
8516 (vect_analyze_loop_operations): Convert return type from bool to
8517 opt_result, converting dump_printf_loc calls, using the location
8518 of phi/stmt rather than vect_location where available. Convert
8519 various "return false" to "return opt_result::failure_at" with
8520 "unsupported phi" messages.
8521 (vect_get_datarefs_in_loop): Convert return type from bool to
8522 opt_result. Add a message for the
8523 PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
8524 (vect_analyze_loop_2): Convert return type from bool to
8525 opt_result. Ensure "ok" is set to a opt_result::failure_at before
8526 each "goto again;", adding new messages where needed.
8527 Add "unsupported grouped {store|load}" messages.
8528 (vect_analyze_loop): Convert return type from loop_vec_info to
8530 * tree-vect-slp.c (vect_analyze_slp): Convert return type from
8532 * tree-vect-stmts.c (process_use): Likewise, converting
8533 dump_printf_loc call and using stmt location, rather than
8535 (vect_mark_stmts_to_be_vectorized): Likeise.
8536 (vect_analyze_stmt): Likewise, adding a %G.
8537 (vect_get_vector_types_for_stmt): Convert return type from bool to
8538 opt_result, converting dump_printf_loc calls and using stmt
8539 location, rather than vect_location.
8540 (vect_get_mask_type_for_stmt): Convert return type from tree to
8541 opt_tree, converting dump_printf_loc calls and using stmt location.
8542 * tree-vectorizer.c: Include "opt-problem.h.
8543 (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
8544 MSG_PRIORITY_INTERNALS. Convert local "loop_vinfo" from
8545 loop_vec_info to opt_loop_vec_info. If if fails, and dumping is
8546 enabled, use it to report at the top level "couldn't vectorize
8547 loop" followed by the problem.
8548 * tree-vectorizer.h (opt_loop_vec_info): New typedef.
8549 (vect_mark_stmts_to_be_vectorized): Convert return type from bool
8551 (vect_analyze_stmt): Likewise.
8552 (vect_get_vector_types_for_stmt): Likewise.
8553 (tree vect_get_mask_type_for_stmt): Likewise.
8554 (vect_analyze_data_ref_dependences): Likewise.
8555 (vect_enhance_data_refs_alignment): Likewise.
8556 (vect_analyze_data_refs_alignment): Likewise.
8557 (vect_verify_datarefs_alignment): Likewise.
8558 (vect_analyze_data_ref_accesses): Likewise.
8559 (vect_prune_runtime_alias_test_list): Likewise.
8560 (vect_find_stmt_data_reference): Likewise.
8561 (vect_analyze_data_refs): Likewise.
8562 (vect_analyze_loop): Convert return type from loop_vec_info to
8564 (vect_analyze_loop_form): Likewise.
8565 (vect_analyze_slp): Convert return type from bool to opt_result.
8567 2018-10-04 David Malcolm <dmalcolm@redhat.com>
8569 * doc/invoke.texi (-fopt-info): Document new "internals"
8571 * dump-context.h (dump_context::apply_dump_filter_p): New decl.
8572 * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
8574 (optinfo_verbosity_options): Add "internals".
8575 (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
8576 (dump_context::apply_dump_filter_p): New member function.
8577 (dump_context::dump_loc): Use apply_dump_filter_p rather than
8578 explicitly masking the dump_kind.
8579 (dump_context::begin_scope): Increment the scope depth first. Use
8580 apply_dump_filter_p rather than explicitly masking the dump_kind.
8581 (dump_context::emit_item): Use apply_dump_filter_p rather than
8582 explicitly masking the dump_kind.
8583 (dump_dec): Likewise.
8584 (dump_hex): Likewise.
8585 (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
8586 (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
8587 (opt_info_switch_p): Update handling of default
8588 MSG_OPTIMIZED_LOCATIONS to cope with default of
8589 MSG_PRIORITY_USER_FACING.
8590 (dump_basic_block): Use apply_dump_filter_p rather than explicitly
8591 masking the dump_kind.
8592 (selftest::test_capture_of_dump_calls): Update test_dump_context
8593 instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
8594 than MSG_ALL. Generalize scope test to be run at all four
8595 combinations of with/without MSG_PRIORITY_USER_FACING and
8596 MSG_PRIORITY_INTERNALS, adding examples of explicit priority
8597 for each of the two values.
8598 * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
8599 Rename MSG_ALL to MSG_ALL_KINDS. Add MSG_PRIORITY_USER_FACING,
8600 MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
8601 values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
8602 (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
8603 with MSG_PRIORITY_*.
8604 * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
8605 dump messages as MSG_PRIORITY_USER_FACING.
8606 * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
8607 about the interaction with MSG_PRIORITY_*.
8609 2018-10-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
8611 * varasm.c (output_constant): Add new parameter merge_strings.
8612 Make strings properly zero terminated in merge string sections.
8613 (mergeable_string_section): Don't fail if the last char is non-zero.
8614 (assemble_variable_contents): Handle merge string sections.
8615 (assemble_variable): Likewise.
8616 (assemble_constant_contents): Likewise.
8617 (output_constant_def_contents): Likewise.
8618 (output_constructor_array_range,
8619 output_constructor_regular_field): Adjust call to output_constant.
8620 (output_object_block): Adjust call to assemble_constant_contents
8621 and assemble_variable_contents.
8623 2018-10-04 Martin Liska <mliska@suse.cz>
8626 * cgraphunit.c (process_function_and_variable_attributes):
8627 Warn about a function with alias attribute and a body.
8629 2018-10-04 Martin Liska <mliska@suse.cz>
8632 * multiple_target.c (redirect_to_specific_clone): New function.
8633 (ipa_target_clone): Use it.
8634 * tree-inline.c: Fix comment.
8636 2018-10-04 David Malcolm <dmalcolm@redhat.com>
8638 * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
8640 (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
8641 (gcc::dump_manager::register_pass): New member function, adapted
8642 from loop body in gcc::pass_manager::register_pass, adding a
8643 call to update_dfi_for_opt_info.
8644 (gcc::dump_manager::opt_info_enable_passes): Store the
8645 -fopt-info options into the new fields. Move the loop
8647 (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
8649 * dumpfile.h (struct opt_pass): New forward decl.
8650 (gcc::dump_manager::register_pass): New decl.
8651 (gcc::dump_manager::update_dfi_for_opt_info): New decl.
8652 (class gcc::dump_manager): Add fields "m_optgroup_flags",
8653 "m_optinfo_flags", and "m_optinfo_filename".
8654 * passes.c (gcc::pass_manager::register_pass): Move all of the
8655 dump-handling code to gcc::dump_manager::register_pass.
8657 2018-10-04 Peter Bergner <bergner@linux.ibm.com>
8659 PR rtl-optimization/87466
8660 * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
8661 * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
8662 * doc/tm.texi: Regenerate.
8663 * ira-lives.c (process_bb_node_lives): Use the new target hook.
8664 * lra-lives.c (process_bb_lives): Likewise.
8665 * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
8668 2018-10-04 Tamar Christina <tamar.christina@arm.com>
8670 * params.c (add_params): Fix initialization.
8672 2018-10-04 Martin Liska <mliska@suse.cz>
8674 PR gcov-profile/84107
8675 * tree-profile.c (init_ic_make_global_vars):
8676 Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
8677 Come up with new ic_tuple* variables. Emit
8678 __gcov_indirect_call{,_topn} variables.
8679 (gimple_gen_ic_profiler): Access the variable
8681 (gimple_gen_ic_func_profiler): Access
8682 __gcov_indirect_call.callee field.
8683 (gimple_init_gcov_profiler): Use ptr_type_node.
8684 * value-prof.c (gimple_ic): Use ptr_type_node.
8686 2018-10-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8688 PR tree-optimization/85787
8689 * ipa-pure-const.c (malloc_candidate_p_1): Move most of
8690 malloc_candidate_p into this function and add support for
8691 detecting multiple phis.
8692 (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
8694 2018-10-04 Martin Liska <mliska@suse.cz>
8697 * ipa-inline.c (inline_to_all_callers_1):
8698 Call ultimate_alias_target for node being inlined.
8700 2018-10-03 Jeff Law <law@redhat.com>
8702 * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
8704 * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
8705 * tree.h (get_typenode_from_name): Prototype.
8707 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
8709 * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
8710 Change operand 2 predicate to nonimmediate_operand.
8711 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
8713 2018-10-03 Martin Sebor <msebor@redhat.com>
8714 Jeff Law <law@redhat.com>
8716 * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
8718 (get_string_length): Detect unterminated arrays.
8719 (format_string): Same.
8720 (format_directive): Warn about unterminated arrays.
8721 (handle_gimple_call): Mark statements with no_warning as needed.
8723 2018-10-03 Jim Wilson <jimw@sifive.com>
8725 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
8726 also define __riscv_abi_rve. Delete trailing white space.
8728 2018-10-03 Paul Koning <ni1d@arrl.net>
8730 Enable LRA register allocator for PDP11.
8731 * config/pdp11/constraints.md (Q): Use define_memory_constraint.
8734 * config/pdp11/pdp11.c (pdp11_lra_p): New function.
8735 * config/pdp11/pdp11.opt (-mlra): New option.
8736 * doc/invoke.texi (PDP-11 Options): Document -mlra.
8738 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
8740 * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
8741 (*<absneg:code>extend<mode>xf2): Ditto.
8743 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
8745 PR tree-optimization/87415
8746 * tree-vrp.c (set_value_range_with_overflow): Special case one bit
8749 2018-10-02 Jeff Law <law@redhat.com>
8751 * gimple-fold.c (get_range_strlen): Only set *nonstr when
8752 an unterminated string is discovered. Bubble up range
8753 even for unterminated strings.
8754 (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
8755 indicates the string was not terminated via NONSTR.
8757 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
8759 * tree-vrp.c (extract_range_from_unary_expr): Special case all
8760 pointer conversions.
8761 Do not do anything special for anti-ranges.
8763 2018-10-03 Jérôme Lambourg <lambourg@adacore.com>
8765 * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
8766 DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
8768 2018-10-03 Martin Liska <mliska@suse.cz>
8770 PR gcov-profile/86109
8771 * coverage.c (coverage_begin_function): Do not
8772 mark lambdas as artificial.
8773 * tree-core.h (struct GTY): Remove tm_clone_flag
8774 and introduce new lambda_function.
8775 * tree.h (DECL_LAMBDA_FUNCTION): New macro.
8777 2018-10-02 Aaron Sawdey <acsawdey@linux.ibm.com>
8780 * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
8781 P8_VECTOR and VSX are enabled.
8783 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
8785 * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
8786 0x3907 as CPU model number.
8788 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
8790 * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
8791 * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
8792 PF_Z14. Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
8793 TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
8794 TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
8795 * config/s390/s390.md: Likewise. Rename also the cpu attribute
8796 value from arch12 to z14.
8798 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
8800 * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
8802 (isinf<mode>2): Ditto.
8804 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
8806 * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
8807 before emitting fxam. Perform calculations in XFmode.
8809 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
8811 * match.pd (((X /[ex] A) +- B) * A): New transformation.
8813 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
8816 * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
8817 * tree.c (signed_or_unsigned_type_for): Handle complex.
8819 2018-10-02 Jeff Law <law@redhat.com>
8821 * gimple-fold.c (get_range_strlen): Remove dead code.
8823 2018-10-02 Martin Sebor <msebor@redhat.com>
8824 Jeff Law <law@redhat.com>
8826 * builtins.c (unterminated_array): Add new arguments.
8827 If argument is not terminated, bubble up size and exact
8829 (expand_builtin_strnlen): Detect, avoid expanding
8830 and diagnose unterminated arrays.
8831 (c_strlen): Fill in offset of start of unterminated strings.
8832 * builtins.h (unterminated_array): Update prototype.
8834 2018-10-02 Richard Biener <rguenther@suse.de>
8836 * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
8837 of haddv4df, first reduce to SSE width and exploit the fact
8838 that we only need element zero with the reduction result.
8839 (reduc_plus_scal_v2df): Likewise.
8841 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
8843 * dojump.h (do_jump): Delete.
8844 (do_jump_1): Likewise.
8845 (split_comparison): Move around.
8846 * dojump.c (do_jump): Make static.
8847 (do_jump_1): Likewise.
8848 (jumpifnot): Move around.
8849 (jumpifnot_1): Likewise.
8851 (jumpif_1): Likewise.
8852 * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
8854 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
8856 * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
8857 insns in the delay slot and add_insn_after for the jump insn.
8859 2018-10-02 Richard Biener <rguenther@suse.de>
8861 * tree-inline.c (expand_call_inline): Use the location of
8862 the callee declaration for the inline-entry marker.
8863 * final.c (notice_source_line): Remove special-casing of
8864 NOTE_INSN_INLINE_ENTRY.
8866 2018-10-01 Carl Love <cel@us.ibm.com>
8869 * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
8870 (__builtin_mtfsb0): New.
8871 (__builtin_mtfsb1): New.
8872 ( __builtin_set_fpscr_rn): New.
8873 (__builtin_set_fpscr_drn): New.
8874 * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
8875 (rs6000_expand_set_fpscr_rn_builtin): Add.
8876 (rs6000_expand_set_fpscr_drn_builtin): Add.
8877 (rs6000_expand_builtin): Add case statement entries for
8878 RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
8879 RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
8880 RS6000_BUILTIN_MFFSL.
8881 (rs6000_init_builtins): Add ftype initialization and def_builtin
8882 calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
8883 __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
8884 * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
8885 rs6000_mffscdrn): Add define_insn.
8886 (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
8887 * doc/extend.texi: Add documentation for the builtins.
8889 2018-10-01 Richard Biener <rguenther@suse.de>
8891 PR tree-optimization/87465
8892 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
8893 causing branch miscounts.
8895 2018-10-01 Tamar Christina <tamar.christina@arm.com>
8897 * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
8898 aarch64_option_default_param): New.
8899 (params.h): Include.
8900 (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
8901 * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
8902 stack-clash protection validation code.
8904 2018-10-01 Tamar Christina <tamar.christina@arm.com>
8906 * params.c (validate_param): New.
8907 (add_params): Use it.
8908 (set_param_value): Refactor param validation into validate_param.
8909 (diagnostic.h): Include.
8910 * diagnostic.h (diagnostic_ready_p): New.
8912 2018-10-01 Tamar Christina <tamar.christina@arm.com>
8914 * params.c (set_param_value):
8915 Add index of parameter being validated.
8916 * common/common-target.def (option_validate_param): New.
8917 * common/common-targhooks.h (default_option_validate_param): New.
8918 * common/common-targhooks.c (default_option_validate_param): New.
8919 * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
8920 * doc/tm.texi: Regenerate.
8922 2018-10-01 Tamar Christina <tamar.christina@arm.com>
8925 * config/aarch64/aarch64.c (aarch64_override_options_internal):
8926 Add validation for stack-clash parameters and set defaults.
8928 2018-10-01 Tamar Christina <tamar.christina@arm.com>
8931 * configure.ac: Add stack-clash-protection-guard-size.
8932 * doc/install.texi: Document it.
8933 * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
8934 * params.def: Update comment for guard-size.
8935 (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
8936 PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
8937 * configure: Regenerate.
8939 2018-10-01 Tamar Christina <tamar.christina@arm.com>
8942 * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
8943 STACK_DYNAMIC_OFFSET): New.
8944 * config/aarch64/aarch64.c (aarch64_layout_frame):
8945 Update outgoing args size.
8946 (aarch64_stack_clash_protection_alloca_probe_range,
8947 TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
8949 2018-10-01 Tamar Christina <tamar.christina@arm.com>
8952 * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
8954 * target.def (stack_clash_protection_alloca_probe_range): New.
8955 (stack_clash_protection_final_dynamic_probe): Remove.
8956 * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
8957 (default_stack_clash_protection_final_dynamic_probe): Remove.
8958 * targhooks.c: Likewise.
8959 * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
8960 (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
8961 * doc/tm.texi: Regenerate.
8963 2018-10-01 Tamar Christina <tamar.christina@arm.com>
8966 * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
8967 * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
8968 aarch64_clamp_to_uimm12_shift): New.
8969 (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
8970 * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
8972 2018-10-01 Tamar Christina <tamar.christina@arm.com>
8975 * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
8977 2018-10-01 Jeff Law <law@redhat.com>
8978 Richard Sandiford <richard.sandiford@linaro.org>
8979 Tamar Christina <tamar.christina@arm.com>
8982 * config/aarch64/aarch64.md
8983 (probe_stack_range): Add k (SP) constraint.
8984 * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
8985 STACK_CLASH_MAX_UNROLL_PAGES): New.
8986 * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
8987 stack probes for stack clash.
8988 (aarch64_allocate_and_probe_stack_space): New.
8989 (aarch64_expand_prologue): Use it.
8990 (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
8991 (aarch64_sub_sp): Add emit_move_imm optional param.
8993 2018-10-01 MCC CS <deswurstes@users.noreply.github.com>
8995 PR tree-optimization/87261
8996 * match.pd: Remove trailing whitespace.
8997 Add (x & y) | ~(x | y) -> ~(x ^ y),
8998 (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
9000 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
9002 * config/arc/arc.md (*add_n): Clean up pattern, update instruction
9004 (ashlsi3_insn): Update instruction constraints.
9005 (ashrsi3_insn): Likewise.
9006 (rotrsi3): Likewise.
9007 (add_shift): Likewise.
9008 * config/arc/constraints.md (Csz): New 32 bit constraint. It
9009 avoids placing in the limm field small constants which, otherwise,
9010 could end into a small instruction.
9012 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
9014 * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
9015 destination register is not odd-even.
9016 (umaddsidi4_split): Likewise.
9018 2018-10-01 Richard Biener <rguenther@suse.de>
9020 * tree-inline.c (expand_call_inline): Store origin of fn
9021 in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
9022 * tree.c (block_ultimate_origin): Simplify and do some
9025 2018-09-30 Uros Bizjak <ubizjak@gmail.com>
9027 * config/i386/mmx.md (EMMS): New int iterator.
9028 (emms): New int attribute.
9029 (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
9030 EMMS int iterator. Explicitly declare clobbers.
9031 (mmx_emms): Remove expander.
9033 * config/i386/predicates.md (emms_operation): Remove predicate.
9034 (vzeroall_pattern): New predicate.
9035 (vzeroupper_pattern): Rename from vzeroupper_operation.
9036 * config/i386/i386.c (ix86_avx_u128_mode_after): Use
9037 vzeroupper_pattern and vzeroall_pattern predicates.
9039 2018-09-30 Peter Bergner <bergner@linux.ibm.com>
9041 PR rtl-optimization/86939
9042 * ira-lives.c (make_hard_regno_born): Rename from this...
9043 (make_hard_regno_live): ... to this. Remove update to conflict
9044 information. Update function comment.
9045 (make_hard_regno_dead): Add conflict information update. Update
9047 (make_object_born): Rename from this...
9048 (make_object_live): ... to this. Remove update to conflict information.
9049 Update function comment.
9050 (make_object_dead): Add conflict information update. Update function
9052 (mark_pseudo_regno_live): Call make_object_live.
9053 (mark_pseudo_regno_subword_live): Likewise.
9054 (mark_hard_reg_dead): Update function comment.
9055 (mark_hard_reg_live): Call make_hard_regno_live.
9056 (process_bb_node_lives): Likewise.
9057 * lra-lives.c (make_hard_regno_born): Rename from this...
9058 (make_hard_regno_live): ... to this. Remove update to conflict
9059 information. Remove now uneeded check_pic_pseudo_p argument.
9060 Update function comment.
9061 (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
9062 to conflict information. Update function comment.
9063 (mark_pseudo_live): Remove update to conflict information. Update
9065 (mark_pseudo_dead): Add conflict information update.
9066 (mark_regno_live): Call make_hard_regno_live.
9067 (mark_regno_dead): Call make_hard_regno_dead with new arguement.
9068 (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
9070 2018-09-29 H.J. Lu <hongjiu.lu@intel.com>
9073 * config/i386/i386.c (construct_container): Use TImode for
9074 BLKmode values in 2 integer registers.
9076 2018-09-29 Jeff Law <law@redhat.com>
9078 * builtins.c (unterminated_array): Pass in c_strlen_data * to
9079 c_strlen rather than just a tree *.
9080 (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
9081 Update recursive calls appropriately. If caller did not provide a
9082 suitable data pointer, create a local one. When a non-terminated
9083 string is discovered, bubble up information about the string via the
9084 c_strlen_data object.
9085 * builtins.h (c_strlen): Update prototype.
9086 (c_strlen_data): New structure.
9087 * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
9088 For a type 2 call, if c_strlen indicates a non-terminated string
9089 use the length of the non-terminated string.
9090 (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
9092 2018-09-29 Jakub Jelinek <jakub@redhat.com>
9095 * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
9096 __m512d type for __A argument rather than __m512.
9098 2018-09-28 John David Anglin <danglin@gcc.gnu.org>
9100 * match.pd (simple_comparison): Don't optimize if either operand is
9101 a function pointer when target needs function pointer canonicalization.
9103 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
9105 * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
9106 power5 .. power9 to remove indirection.
9107 * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
9108 ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
9109 ASM_CPU_476_SPEC): Delete.
9110 (ASM_CPU_SPEC): Adjust.
9111 (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
9112 asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
9114 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
9116 * config.in: Delete HAVE_AS_DCI.
9117 * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
9118 * config/rs6000/rs6000.h: Ditto.
9119 * configure.ac: Delete HAVE_AS_DCI.
9120 * configure: Regenerate.
9122 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
9124 * config.in (HAVE_AS_LWSYNC): Delete.
9125 * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
9126 * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
9128 * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
9129 * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
9131 * configure.ac: Delete HAVE_AS_LWSYNC.
9132 * configure: Regenerate.
9134 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
9135 Pierre-Marie de Rodat <derodat@adacore.com>
9137 * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
9138 * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
9139 (cgraph_node::create_thunk): Add indirect_offset parameter.
9140 (thunk_adjust): Likewise.
9141 * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
9142 and initialize the corresponding field with it.
9143 (cgraph_node::dump): Dump indirect_offset field.
9144 * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
9145 * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
9146 (thunk_adjust): Add indirect_offset parameter and deal with it.
9147 (cgraph_node::expand_thunk): Deal with the indirect_offset field and
9148 pass it to thunk_adjust. Do not call the target hook if it's non-zero
9149 or if the thunk is external or local. Fix formatting. Do not chain
9150 the RESULT_DECL to BLOCK_VARS. Pass the static chain to the target,
9151 if any, in the GIMPLE representation.
9152 * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
9153 * lto-cgraph.c (lto_output_node): Write indirect_offset field.
9154 (input_node): Read indirect_offset field.
9155 * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
9156 call to thunk_adjust.
9157 * tree-nested.c (struct nesting_info): Add thunk_p field.
9158 (create_nesting_tree): Set it.
9159 (convert_all_function_calls): Copy static chain from targets to thunks.
9160 (finalize_nesting_tree_1): Return early for thunks.
9161 (unnest_nesting_tree_1): Do not finalize thunks.
9162 (gimplify_all_functions): Do not gimplify thunks.
9164 2018-09-28 David Malcolm <dmalcolm@redhat.com>
9166 * opt-suggestions.c (option_proposer::build_option_suggestions):
9167 Release "option_values".
9169 2018-09-28 David Malcolm <dmalcolm@redhat.com>
9171 * coverage.c (get_coverage_counts): Convert problem-reporting dump
9172 messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
9173 * dumpfile.c (kind_as_string): New function.
9174 (dump_loc): Rather than a hardcoded prefix of "note: ", use
9175 kind_as_string to vary the prefix based on dump_kind.
9176 (selftest::test_capture_of_dump_calls): Update for above.
9178 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
9180 * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
9181 (GET_SSE_REGNO): Rename from SSE_REGNO. Update all uses for rename.
9183 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
9185 * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
9186 * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
9187 INVALID_REGNUM instead of FPSR_REG.
9188 (ix86_md_asm_adjust): Do not clobber FPSR_REG.
9189 * config/i386/i386.md: Update comment of FP compares.
9190 (fldenv): Do not clobber FPSR_REG.
9192 2018-09-28 Richard Biener <rguenther@suse.de>
9194 * tree.h (BLOCK_ORIGIN): New.
9195 * omp-expand.c (grid_expand_target_grid_body): Assign
9196 BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
9197 * tree-inline.c (remap_block): Likewise.
9198 * auto-profile.c (get_function_decl_from_block): Simplify
9199 by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
9200 * langhooks.c (lhd_print_error_function): Likewise.
9201 * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
9203 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
9204 * tree.c (block_nonartificial_location): Likewise.
9205 (block_ultimate_origin): Likewise.
9206 * tree-pretty-print.c (percent_K_format): Likewise. Remove
9207 no longer needed LTO case.
9209 2018-09-28 Andrew Stubbs <ams@codesourcery.com>
9210 Jan Hubicka <jh@suse.cz>
9211 Martin Jambor <mjambor@suse.cz>
9213 * simplify-rtx.c (simplify_merge_mask): New function.
9214 (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
9215 same masks are used in op1 or op2.
9216 (test_vec_merge): New function.
9217 (test_vector_ops): Call test_vec_merge.
9219 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
9221 * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
9222 * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
9223 * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
9224 (cypress_costs): Set it.
9225 (supersparc_costs): Likewise.
9226 (hypersparc_costs): Likewise.
9227 (leon_cost): Likewise.
9228 (leon3_costs): Likewise.
9229 (sparclet_costs): Likewise.
9230 (ultrasparc_costs): Likewise.
9231 (ultrasparc_costs): Likewise.
9232 (niagara_costs): Likewise.
9233 (niagara2_costs): Likewise.
9234 (niagara3_costs): Likewise.
9235 (niagara4_costs): Likewise.
9236 (niagara7_costs): Likewise.
9237 (m8_costs): Likewise.
9238 (TARGET_CAN_FOLLOW_JUMP): Define.
9239 (pass_work_around_errata::gate): Minor tweak.
9240 (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
9241 Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
9242 Automaitcally clear MASK_FSMULD mask for V7 processors.
9243 (sparc_can_follow_jump): New static function.
9244 (output_ubranch): Deal with CROSSING_JUMP_P.
9245 (sparc_use_sched_lookahead): Rewrite using switch statement.
9246 (sparc_issue_rate): Reorder.
9247 (sparc_branch_cost): New function.
9249 2018-09-27 Martin Sebor <msebor@redhat.com>
9251 * tree.h (tree_to_shwi): Add attribute nonnull and pure.
9252 (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
9253 (int_fits_type_p): Same.
9255 2018-09-27 Uros Bizjak <ubizjak@gmail.com>
9257 * config/i386/i386.md (FPCR_REG): Remove.
9258 (UNSPEC_FLDCW): Remove.
9259 (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
9260 (x86_fldcw_1): Remove insn pattern.
9261 (fnstenv): Do not clobber FPCR_REG.
9263 * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
9264 (CALL_USED_REGISTERS): Ditto.
9265 (REG_ALLOC_ORDER): Ditto.
9266 (REG_CLASS_CONTENTS): Ditto.
9267 (HI_REGISTER_NAMES): Ditto.
9268 (ADDITIONAL_REGISTER_NAMES): Use defines instead
9269 of numerical constants.
9270 * config/i386/i386.c (regclass_map): Remove fpsr register.
9271 (dbx_register_map): Ditto.
9272 (dbx64_register_map): Ditto.
9273 (svr4_dbx_register_map): Ditto.
9274 (print_reg): Do not handle FPCR_REG.
9276 2018-09-27 Segher Boessenkool <segher@kernel.crashing.org>
9279 * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
9280 HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
9281 Delete, always treat as true.
9282 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
9283 Ditto. Simplify remaining code.
9284 * config/powerpcspe/powerpcspe.h: Ditto.
9285 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
9286 Simplify remaining code.
9287 (rs6000_expand_builtin): Ditto.
9288 * config/rs6000/rs6000.h: Ditto.
9289 * configure.ac: Ditto.
9290 * configure: Regenerate.
9292 2018-09-27 Martin Liska <mliska@suse.cz>
9294 * coverage.c (get_coverage_counts): Revert the formatting
9295 of missing profile opt info.
9297 2018-09-27 Richard Biener <rguenther@suse.de>
9301 * dwarf2out.c (set_block_origin_self): Do not mark outermost
9302 block as we do not output that.
9303 (gen_inlined_subroutine_die): Elide the originally outermost
9304 block, matching what we do for concrete instances.
9305 (decls_for_scope): Add parameter specifying whether to recurse
9308 2018-09-27 Andrew Stubbs <ams@codesourcery.com>
9309 Tom de Vries <tom@codesourcery.com>
9313 * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
9314 STORE_FLAG_VALUE == 1.
9316 2018-09-27 Andreas Krebbel <krebbel@linux.ibm.com>
9318 * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
9319 constant definitions.
9320 ("tx_assist"): Replace magic number with PPA_TX_ABORT.
9321 ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
9322 ("speculation_barrier"): New expander definition.
9324 2018-09-26 Indu Bhagat <indu.bhagat@oracle.com>
9326 PR gcov-profile/86957
9327 * common.opt: New warning option -Wmissing-profile.
9328 * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
9329 * doc/invoke.texi: Document -Wmissing-profile.
9331 2018-09-26 Jim Wilson <jimw@sifive.com>
9333 * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
9334 (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
9336 2018-09-26 Martin Sebor <msebor@redhat.com>
9338 * tree.c (zerop): Change return type to bool.
9339 (integer_zerop, integer_onep, integer_each_onep): Same.
9340 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
9341 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
9342 (real_onep, real_minus_onep, chain_index): Same.
9343 (print_type_hash_statistics, type_list_equal): Same.
9344 * tree.h (zerop): Same.
9345 (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
9346 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
9347 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
9348 (real_onep, real_minus_onep, chain_index): Same.
9349 (print_type_hash_statistics, type_list_equal): Same.
9351 2018-09-26 Jim Wilson <jimw@sifive.com>
9353 * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
9355 2018-09-26 Jakub Jelinek <jakub@redhat.com>
9358 * config/i386/i386.c: Include debug.h and dwarf2out.h.
9359 (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
9362 2018-09-25 Andrew Stubbs <ams@codesourcery.com>
9364 * builtins.c (get_builtin_sync_mem): Force address mode conversion.
9366 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
9368 * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
9369 and FP_SECOND_SSE_REGS.
9370 (REG_CLASS_NAMES): Ditto.
9371 (REG_CLASS_CONTENTS): Ditto.
9372 * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
9373 FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
9374 (ix86_preferred_output_reload_class): Ditto.
9375 * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
9376 clobber constraint to "=&f".
9377 (fix_truncdi_i387): Ditto.
9378 (lrintxfdi2): Ditto.
9379 (fistdi2_<rounding>): Ditto.
9380 (fpremxf4_i387): Change "=u" constraint to "=f".
9381 (fprem1xf4_i387): Ditto.
9383 (fptanxf4_i387): Ditto.
9384 (fxtractxf3_i387): Ditto.
9385 (fscalexf4_i387): Ditto.
9386 (atan2xf3): Change "u" constraint to "f".
9387 (fyl2xxf3_i387): Ditto.
9388 (fyl2xp1xf3_i387): Ditto.
9390 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
9393 * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
9394 for removed I387_MASK_PM entity.
9397 2018-09-26 Jeff Law <law@redhat.com>
9399 2018-09-26 Alexey Neyman <stilor@att.net>
9401 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
9402 headers are no longer pulled in by <isl/val.h>.
9404 2018-09-26 Richard Biener <rguenther@suse.de>
9407 * dwarf2out.c (gen_lexical_block_die): Do not equate inline
9408 or concrete instance DIE to the tree. Create abstract origin
9409 attributes also for concrete instances.
9411 2018-09-26 Alexey Neyman <stilor@att.net>
9413 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
9414 headers are no longer pulled in by <isl/val.h>.
9416 2018-09-26 Matthew Malcomson <matthew.malcomson@arm.com>
9418 * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
9419 Use new helper functions.
9420 * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
9421 Use new helper functions.
9422 * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
9423 aarch_mm_needs_release): New declarations.
9424 * config/arm/aarch-common.c (aarch_mm_needs_acquire,
9425 aarch_mm_needs_release): New.
9427 2018-09-26 Eric Botcazou <ebotcazou@adacore.com>
9429 * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
9430 (arm32_output_mi_thunk): Deal with long calls.
9432 2018-09-26 Richard Biener <rguenther@suse.de>
9436 * tree-inline.c (expand_call_inline): When the location
9437 of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
9438 or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
9439 the inserted BLOCK to make inlined_function_outer_scope_p
9441 * dwarf2out.c (add_call_src_coords_attributes): Do not add
9442 coords for reserved locations.
9444 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
9446 * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
9447 (*call_indirect_nonlocal_sysv<mode>): Ditto.
9448 (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
9449 (*sibcall_nonlocal_sysv<mode>): Ditto.
9450 (*sibcall_value_nonlocal_sysv<mode>): Ditto.
9451 (<bd>_<mode>): Ditto.
9452 (<bd>tf_<mode>): Ditto.
9454 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
9456 * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
9457 control string as a list of templates instead of as C code.
9458 (*altivec_movti): Ditto.
9459 * config/rs6000/darwin.md (movdf_low_di): Ditto.
9461 2018-09-25 Jim Wilson <jimw@sifive.com>
9463 * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
9464 when target symbol is weak.
9466 2018-09-25 Bernd Edlinger <bernd.edlinger@hotmail.de>
9469 * builtins.c (unterminated_array): Simplify.
9470 * expr.c (string_constant): Handle SSA_NAME. Add more exceptions
9471 where pointer arithmetic is safe.
9473 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
9476 * config/rs6000/altivec.md (altivec_vspltb): Use
9477 const_0_to_15_operand instead of u5bit_cint_operand.
9478 (*altivec_vspltb_internal): Ditto.
9479 (altivec_vspltb_direct): Ditto.
9480 (altivec_vsplth): Use const_0_to_7_operand instead of
9482 (*altivec_vsplth_internal): Ditto.
9483 (altivec_vsplth_direct): Ditto.
9484 (altivec_vspltw): Use const_0_to_3_operand instead of
9486 (*altivec_vspltw_internal): Ditto.
9487 (altivec_vspltw_direct): Ditto.
9488 (altivec_vspltsf): Ditto.
9489 (*altivec_vspltsf_internal): Ditto.
9490 * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
9491 various splats with the proper size immediate. Reorder the various
9492 cases by ascending size of immediate, and put all such together.
9494 2018-09-25 Richard Biener <rguenther@suse.de>
9497 * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
9499 (maybe_create_die_with_external_ref): Do not create
9500 DW_TAG_imported_unit here.
9501 (add_abstract_origin_attribute): Handle external BLOCK refs.
9502 (dwarf2out_abstract_function): Simplify LTO case.
9503 (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
9504 rather than using maybe_create_die_with_external_ref.
9506 2018-09-25 Uros Bizjak <ubizjak@gmail.com>
9509 * config/i386/i386.md (frndintxf2_mask_pm): Remove.
9510 (frndintxf2_mask_pm_i387): Ditto.
9511 (nearbyintxf2): Rewrite expander pattern to match rintxf2.
9512 Enable for !flag_trapping_math.
9513 (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
9514 Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
9515 Change operand 1 predicate to nonimmediate_operand.
9516 (attr "i387_cw"): Remove mask_pm.
9517 * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
9518 (enum ix86_entity): Remove I387_MASK_PM.
9519 * config/i386/i386.c (ix86_i387_mode_needed): Do not
9520 handle I387_MASK_PM.
9521 (ix86_mode_needed): Ditto.
9522 (ix86_mode_after): Ditto.
9523 (ix86_mode_entry): Ditto.
9524 (ix86_mode_exit): Ditto.
9525 (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
9527 2018-09-25 Jakub Jelinek <jakub@redhat.com>
9529 * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
9530 to_update_switch_stmts to vNULL instead of calling create on them
9533 2018-09-25 Richard Biener <rguenther@suse.de>
9535 PR tree-optimization/87402
9536 * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
9537 (visit_phi): Re-instantiate handling of supposed to be VARYING
9538 but non-VARYING backedge value.
9540 2018-09-25 Richard Biener <rguenther@suse.de>
9543 * dwarf2out.c (struct sym_off_pair): New.
9544 (external_die_map): New global.
9545 (lookup_decl_die): When in LTO create DIEs lazily from the
9547 (lookup_block_die): New function, create DIEs lazily in LTO.
9548 (equate_block_to_die): New function.
9549 (dwarf2out_die_ref_for_decl): During WPA get the association
9550 from the external DIE map.
9551 (dwarf2out_register_external_die): Record mapping into the
9553 (maybe_create_die_with_external_ref): New function split out from
9554 DIE generation part of old dwarf2out_register_external_die.
9555 (add_abstract_origin_attribute): Do not return the DIE. When
9556 in LTO reference externals directly.
9557 (dwarf2out_abstract_function): When in LTO ignore calls for
9558 decls with external DIEs (already present abstract instances).
9559 (gen_call_site_die): Adjust.
9560 (add_high_low_attributes): Likewise.
9561 (gen_lexical_block_die): Likewise.
9562 (gen_inlined_subroutine_die): Likewie.
9563 (gen_block_die): Likewise.
9564 (dwarf2out_inline_entry): Likewise.
9565 (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
9568 2018-09-25 Martin Liska <mliska@suse.cz>
9570 * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
9571 integers and not by a float value.
9573 2018-09-25 Martin Liska <mliska@suse.cz>
9576 * dbgcnt.c (dbg_cnt_process_single_pair): Return false
9578 * dumpfile.c (dump_enable_all): Remove extra parenthesis.
9579 * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
9580 * godump.c (go_format_type): Remove extra parenthesis.
9582 2018-09-25 Martin Liska <mliska@suse.cz>
9584 * alias.c (set_dest_equal_p): Remove unused function.
9585 * config/i386/i386.c (def_builtin_pure2): Likewise.
9586 * diagnostic-show-locus.c (class layout): Remove
9588 (layout::layout): Likewise here.
9589 * dump-context.h (class temp_dump_context): Likewise.
9590 * dwarf2out.c (add_AT_fde_ref): Remove unused function.
9591 (add_AT_loclistsptr): Likewise.
9592 (add_AT_offset): Likewise.
9593 (get_AT_hi_pc): Likewise.
9594 (is_comdat_die): Likewise.
9595 (type_is_enum): Likewise.
9596 (ceiling): Likewise.
9597 (add_AT_vms_delta): Likewise.
9598 (is_class_die): Likewise.
9599 * edit-context.c (class line_event): Remove unused field.
9600 * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
9602 * ipa-cp.c (ipa_get_vr_lat): Likewise.
9603 * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
9604 (ok_for_base_p_nonstrict): Likewise.
9605 * tree-chrec.c (is_not_constant_evolution): Likewise.
9606 (chrec_fold_poly_cst): Likewise.
9607 * tree-if-conv.c (has_pred_critical_p): Likewise.
9608 * tree-ssa-coalesce.c (print_exprs): Likewise.
9609 * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
9610 * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
9611 * tree-vrp.c (value_ranges_intersect_p): Likewise.
9612 (value_range_nonnegative_p): Likewise.
9614 2018-09-25 Martin Liska <mliska@suse.cz>
9616 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
9617 Do not handle "GNU Pascal".
9618 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
9620 * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
9621 from documentation. Likewise.
9622 * dbxout.c (dbxout_range_type): Likewise.
9623 * doc/cpp.texi: Likewise.
9624 * doc/extend.texi: Likewise.
9625 * doc/frontends.texi: Likewise.
9626 * doc/invoke.texi: Remove Pascal entry.
9627 * tree.def (CLEANUP_POINT_EXPR): Likewise.
9628 * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
9630 2018-09-25 Martin Liska <mliska@suse.cz>
9633 * doc/invoke.texi: Document all parameters and remove default
9636 2018-09-25 Ilya Leoshkevich <iii@linux.ibm.com>
9639 * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
9640 contains HOST_WIDE_INTs when computing its size.
9642 2018-09-24 Jim Wilson <jimw@sifive.com>
9645 * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
9647 (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
9649 2018-09-24 Andrew Pinski <apinski@marvell.com>
9651 * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
9652 access prev before checking it for NULLness in the
9653 AARCH64_FUSE_CMP_BRANCH case.
9655 2018-09-24 H.J. Lu <hongjiu.lu@intel.com>
9658 * config/i386/i386.c (rest_of_insert_endbranch): Set
9659 endbr_queued_at_entrance to true and don't insert ENDBR if
9660 x86_function_profiler will be called.
9661 (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
9663 * config/i386/i386.h (machine_function): Add
9664 endbr_queued_at_entrance.
9666 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
9668 * genattrtab.c (mk_attr_alt): Use alternative_mask.
9669 (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
9671 (check_attr_test): Use alternative_mask.
9672 (get_attr_value): Likewise.
9673 (compute_alternative_mask): Use alternative_mask and XWINT.
9674 (make_alternative_compare): Use alternative_mask.
9675 (attr_alt_subset_p): Use XWINT.
9676 (attr_alt_subset_of_compl_p): Likewise.
9677 (attr_alt_intersection): Use alternative_mask and XWINT.
9678 (attr_alt_union): Likewise.
9679 (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
9680 (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
9681 (simplify_test_exp): Use alternative_mask and XWINT.
9682 (write_test_expr): Use alternative_mask and XWINT, adjust bit
9683 number calculation to support 64 bits. Generate code that
9684 checks 64-bit masks.
9685 (main): Use alternative_mask.
9686 * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
9688 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
9691 * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
9692 RETURN+USE when returning via %r14.
9694 2018-09-24 Martin Liska <mliska@suse.cz>
9696 * gcov.c (output_lines): Print colorization legend
9697 for both flag_use_colors and flag_use_hotness_colors.
9700 2018-09-24 Martin Liska <mliska@suse.cz>
9702 * coverage.c (get_coverage_counts): Use warning_at
9703 with current_function_decl location. Use %qD in warning
9706 2018-09-24 Martin Liska <mliska@suse.cz>
9708 * memory-block.h (memory_block_pool::release): Annotate with
9709 valgrind that the memory is not accessible.
9711 2018-09-24 Martin Liska <mliska@suse.cz>
9714 * asan.c: Make asan_handled_variables extern.
9716 * cfgexpand.c (expand_stack_vars): Make sure
9717 a representative is unpoison if another
9718 variable in the partition is handled by
9719 use-after-scope sanitization.
9721 2018-09-24 Richard Biener <rguenther@suse.de>
9723 PR tree-optimization/63155
9724 * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
9725 the worklist when the edge of the respective argument isn't
9728 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
9730 * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
9731 ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
9732 (MASK_CLASS_P): Update for rename.
9733 (MAYBE_MASK_CLASS_P): Ditto.
9734 (REG_CLASS_NAMES): Update.
9735 (REG_CLASS_CONTENT): Update.
9736 * config/i386/i386.c (regclass_map): Update for MASK_REG
9737 and ALL_MASK_REGS rename.
9738 * config/i386/constraints.md (Yk): Update for rename.
9741 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
9743 * config/i386/i386.h (enum reg_class): Remove
9744 EVEX_SSE_REGS and MOD4_SSE_REGS.
9745 (REG_CLASS_NAMES): Update.
9746 (REG_CLASS_CONTENT): Update.
9747 * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
9748 registers as ALL_SSE_REGS.
9749 (ix86_additional_allocno_class_p): Remove.
9750 (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
9751 (ix86_register_priority): Lower priority of EVEX SSE registers.
9752 Use IN_RANGE macro where appropriate.
9753 (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
9754 AVX-5124VNNIW checks.
9755 (ix86_modes_tieable_p): Tie 512-bit SSE modes.
9756 * config/i386/sse.md (avx5124fmaddps_4fmaddps)
9757 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
9758 (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
9759 (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
9760 (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
9761 (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
9762 (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
9763 (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
9764 (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
9765 (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
9766 * config/i386/constraints.md (Yh): Remove.
9768 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
9770 * config/i386/i386.c (regclass_map): Declare integer REX registers
9773 2018-09-23 Gerald Pfeifer <gerald@pfeifer.com>
9775 * doc/service.texi (Service): Switch the fsf.org link to https.
9777 2018-09-22 Chung-Ju Wu <jasonwucj@gmail.com>
9780 * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9781 Define to speculation_safe_value_not_needed.
9783 2018-09-21 Florian Weimer <fweimer@redhat.com>
9786 * doc/extend.texi (Common Function Attributes): Mention that
9787 noreturn suppresses tail call optimization.
9789 2018-09-21 Jeff Law <law@redhat.com>
9791 * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
9792 vr_values::cleanup_edges_and_switches.
9793 * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
9795 (identify_jump_threads): Remove EDGE_IGNORE handling.
9796 (execute_vrp): Move handling of to_remove_edges and
9797 to_update_switch_stmts into vr_values class member functions.
9798 * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
9799 (to_update_switch_stmts): Likewise.
9800 * vr-values.c: Include cfghooks.h.
9801 (vr_values::vr_values): Initialize to_remove_edges and
9802 to_update_switch_stmts.
9803 (vr_values::~vr_values): Verify to_remove_edges and
9804 to_update_switch_stmts are empty.
9805 (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
9806 (vr_values::cleanup_edges_and_switches): New member function.
9807 * vr-values.h (vr_values): Add cleanup_edges_and_switches member
9808 function. Add new data members.
9810 2018-09-21 David Malcolm <dmalcolm@redhat.com>
9812 PR tree-optimization/87309
9813 * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
9814 calls with pflags and alt_flags.
9815 (selftest::test_capture_of_dump_calls): Add test of interaction of
9816 MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
9818 2018-09-21 Olivier Hainque <hainque@adacore.com>
9820 * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
9822 2018-09-21 Olivier Hainque <hainque@adacore.com>
9824 * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
9826 2018-09-21 Olivier Hainque <hainque@adacore.com>
9828 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
9829 Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
9830 (VXWORKS_LIBS_RTP): Minor reordering.
9832 2018-09-21 Olivier Hainque <hainque@adacore.com>
9834 * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
9835 (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
9837 2018-09-21 Olivier Hainque <hainque@adacore.com>
9839 * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
9840 (PTRDIFF_TYPE): Likewise.
9842 2018-09-21 Olivier Hainque <hainque@adacore.com>
9844 * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
9845 triplet, similar to support for VxWorks7.
9846 * config/vxworks-dummy.h: Provide a default definition
9847 of TARGET_VXWORKS64 to 0.
9849 2018-09-21 Olivier Hainque <hainque@adacore.com>
9851 * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
9852 * config/vxworks-dummy.h: here.
9854 2018-09-21 Olivier Hainque <hainque@adacore.com>
9856 * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
9858 2018-09-21 Shaokun Zhang <zhangshaokun@hisilicon.com>
9859 Bo Zhou <zbo.zhou@hisilicon.com>
9861 * config/aarch64/aarch64-cores.def (tsv110): New CPU.
9862 * config/aarch64/aarch64-tune.md: Regenerated.
9863 * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
9864 * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
9865 * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
9867 2018-09-21 Andrew Stubbs <ams@codesourcery.com>
9868 Julian Brown <julian@codesourcery.com>
9870 * builtins.c (get_builtin_sync_mem): Handle address spaces.
9872 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
9874 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
9875 if the call takes a static chain.
9877 2018-09-21 Martin Liska <mliska@suse.cz>
9879 * auto-profile.c (autofdo_source_profile::read): Do not
9881 (read_profile): Do not add working sets.
9882 (read_autofdo_file): Remove sum_all.
9883 (afdo_callsite_hot_enough_for_early_inline): Remove const
9885 * coverage.c (struct counts_entry): Remove gcov_summary.
9886 (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
9887 do not support GCOV_TAG_PROGRAM_SUMMARY.
9888 (get_coverage_counts): Remove summary and expected
9890 * coverage.h (get_coverage_counts): Likewise.
9891 * doc/gcov-dump.texi: Remove -w option.
9892 * gcov-dump.c (dump_working_sets): Remove.
9893 (main): Do not support '-w' option.
9894 (print_usage): Likewise.
9895 (tag_summary): Likewise.
9896 * gcov-io.c (gcov_write_summary): Do not dump
9898 (gcov_read_summary): Likewise.
9899 (gcov_histo_index): Remove.
9900 (gcov_histogram_merge): Likewise.
9901 (compute_working_sets): Likewise.
9902 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
9904 (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
9905 (GCOV_TAG_SUMMARY_LENGTH): Adjust.
9906 (GCOV_HISTOGRAM_SIZE): Remove.
9907 (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
9908 (struct gcov_summary): Simplify rapidly just
9909 to runs and sum_max fields.
9910 (gcov_histo_index): Remove.
9911 (NUM_GCOV_WORKING_SETS): Likewise.
9912 (compute_working_sets): Likewise.
9913 * gcov-tool.c (print_overlap_usage_message): Remove
9914 trailing empty line.
9915 * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
9916 (output_lines): Remove program related line.
9917 * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
9918 * lto-cgraph.c (output_profile_summary): Do not stream GCOV
9920 (input_profile_summary): Do not read it.
9921 (merge_profile_summaries): And do not merge it.
9922 (input_symtab): Do not call removed function.
9923 * modulo-sched.c (sms_schedule): Do not print sum_max.
9924 * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
9925 removed when histogram method was invented.
9926 (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
9928 * postreload-gcse.c (eliminate_partially_redundant_load): Fix
9930 * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
9931 and dump selected value.
9932 * profile.c (add_working_set): Remove.
9933 (get_working_sets): Likewise.
9934 (find_working_set): Likewise.
9935 (get_exec_counts): Do not work with working sets.
9936 (read_profile_edge_counts): Do not inform as sum_max is removed.
9937 (compute_branch_probabilities): Likewise.
9938 (compute_value_histograms): Remove argument for call of
9939 get_coverage_counts.
9940 * profile.h: Do not make gcov_summary const.
9942 2018-09-21 Monk Chiang <sh.chiang04@gmail.com>
9944 * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
9946 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
9948 PR tree-optimization/86990
9949 * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
9950 Check that the entire merged store group is made of constants only for
9953 2018-09-20 Allan Sandfeld Jensen <allan.jensen@qt.io>
9955 * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
9956 (VTABLE_VERIFICATION_SPEC): Likewise.
9957 (SANITIZER_EARLY_SPEC): Likewise.
9958 (SANITIZER_SPEC): Likewise.
9959 * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
9960 * doc/invoke.texi (Link Options): Document -r.
9962 2018-09-20 Richard Biener <rguenther@suse.de>
9965 * gimplify.c (gimplify_expr): Retain alignment of
9966 addressable lvalue in dereference.
9968 2018-09-20 Alexandre Oliva <aoliva@redhat.com>
9971 * configure.ac: Check for .loc is_stmt support.
9972 * configure, config.in: Rebuilt.
9973 * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
9976 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
9978 * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
9979 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
9982 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
9984 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
9986 * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
9989 2018-09-20 Marek Polacek <polacek@redhat.com>
9991 * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
9993 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
9995 PR tree-optimization/87288
9996 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
9997 into account when determining PEELING_FOR_NITERS.
9999 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
10001 PR tree-optimization/86877
10002 * tree-vect-loop.c (vect_analyze_loop_2): Call
10003 vect_verify_datarefs_alignment.
10005 2018-09-19 Marek Polacek <polacek@redhat.com>
10007 * doc/invoke.texi: Document -Wclass-conversion.
10009 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
10011 * config/pa/pa.c (pa_adjust_priority): Delete.
10012 (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
10014 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
10015 (atomic_storehi): Likewise.
10016 (atomic_storesi): Likewise.
10017 (atomic_loaddi): Restore compare and swap exchange loop code.
10019 2018-09-19 Segher Boessenkool <segher@kernel.crashing.org>
10021 PR rtl-optimization/86902
10022 * combine.c (try_combine): When changing the CC mode used, don't change
10023 an unrelated mode in other_insn to that new CC mode.
10025 2018-09-19 David Malcolm <dmalcolm@redhat.com>
10027 * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
10028 with %T in place of calls to dump_generic_expr.
10029 (prune_runtime_alias_test_list): Likewise.
10030 (create_runtime_alias_checks): Likewise.
10031 * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
10032 (vect_analyze_data_ref_dependence): Likewise.
10033 (vect_slp_analyze_data_ref_dependence): Likewise.
10034 (vect_record_base_alignment): Likewise. Use %G in place of call
10035 to dump_gimple_stmt.
10036 (vect_compute_data_ref_alignment): Likewise.
10037 (verify_data_ref_alignment): Likewise.
10038 (vect_find_same_alignment_drs): Likewise.
10039 (vect_analyze_group_access_1): Likewise.
10040 (vect_analyze_data_ref_accesses): Likewise.
10041 (dependence_distance_ge_vf): Likewise.
10042 (dump_lower_bound): Likewise.
10043 (vect_prune_runtime_alias_test_list): Likewise.
10044 (vect_find_stmt_data_reference): Likewise.
10045 (vect_analyze_data_refs): Likewise.
10046 (vect_create_addr_base_for_vector_ref): Likewise.
10047 (vect_create_data_ref_ptr): Likewise.
10048 * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
10049 (vect_can_advance_ivs_p): Likewise.
10050 (vect_update_ivs_after_vectorizer): Likewise.
10051 (vect_gen_prolog_loop_niters): Likewise.
10052 (vect_prepare_for_masked_peels): Likewise.
10053 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
10054 (vect_determine_vectorization_factor): Likewise.
10055 (vect_is_simple_iv_evolution): Likewise.
10056 (vect_analyze_scalar_cycles_1): Likewise.
10057 (vect_analyze_loop_operations): Likewise.
10058 (report_vect_op): Likewise.
10059 (vect_is_slp_reduction): Likewise.
10060 (check_reduction_path): Likewise.
10061 (vect_is_simple_reduction): Likewise.
10062 (vect_create_epilog_for_reduction): Likewise.
10063 (vect_finalize_reduction:): Likewise.
10064 (vectorizable_induction): Likewise.
10065 (vect_transform_loop_stmt): Likewise.
10066 (vect_transform_loop): Likewise.
10067 (optimize_mask_stores): Likewise.
10068 * tree-vect-patterns.c (vect_pattern_detected): Likewise.
10069 (vect_split_statement): Likewise.
10070 (vect_recog_over_widening_pattern): Likewise.
10071 (vect_recog_average_pattern): Likewise.
10072 (vect_determine_min_output_precision_1): Likewise.
10073 (vect_determine_precisions_from_range): Likewise.
10074 (vect_determine_precisions_from_users): Likewise.
10075 (vect_mark_pattern_stmts): Likewise.
10076 (vect_pattern_recog_1): Likewise.
10077 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
10078 (vect_record_max_nunits): Likewise.
10079 (vect_build_slp_tree_1): Likewise.
10080 (vect_build_slp_tree_2): Likewise.
10081 (vect_print_slp_tree): Likewise.
10082 (vect_analyze_slp_instance): Likewise.
10083 (vect_detect_hybrid_slp_stmts): Likewise.
10084 (vect_detect_hybrid_slp_1): Likewise.
10085 (vect_slp_analyze_operations): Likewise.
10086 (vect_slp_analyze_bb_1): Likewise.
10087 (vect_transform_slp_perm_load): Likewise.
10088 (vect_schedule_slp_instance): Likewise.
10089 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
10090 (vect_mark_stmts_to_be_vectorized): Likewise.
10091 (vect_init_vector_1): Likewise.
10092 (vect_get_vec_def_for_operand): Likewise.
10093 (vect_finish_stmt_generation_1): Likewise.
10094 (vect_check_load_store_mask): Likewise.
10095 (vectorizable_call): Likewise.
10096 (vectorizable_conversion): Likewise.
10097 (vectorizable_operation): Likewise.
10098 (vectorizable_load): Likewise.
10099 (vect_analyze_stmt): Likewise.
10100 (vect_is_simple_use): Likewise.
10101 (vect_get_vector_types_for_stmt): Likewise.
10102 (vect_get_mask_type_for_stmt): Likewise.
10103 * tree-vectorizer.c (increase_alignment): Likewise.
10105 2018-09-19 Andrew Stubbs <ams@codesourcery.com>
10107 * doc/rtl.texi: Adjust vec_select description.
10108 * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
10109 non-constant selectors.
10111 2018-09-19 Matthew Malcomson <matthew.malcomson@arm.com>
10113 * config/aarch64/aarch64-protos.h
10114 (aarch64_offset_9bit_signed_unscaled_p): New declaration.
10115 * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
10116 (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
10117 * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
10118 (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
10119 (AARCH64_FL_PROFILE): Move index so flags are ordered.
10120 (AARCH64_ISA_RCPC8_4): New flag.
10121 * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
10122 to aarch64_offset_9bit_signed_unscaled_p.
10123 * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
10125 * config/aarch64/constraints.md (Ust): New constraint.
10126 * config/aarch64/predicates.md.
10127 (aarch64_9bit_offset_memory_operand): New predicate.
10128 (aarch64_rcpc_memory_operand): New predicate.
10130 2018-09-19 Eric Botcazou <ebotcazou@adacore.com>
10132 PR rtl-optimization/87361
10133 * rtlanal.c (nonzero_bits1): Revert accidental change.
10135 2018-09-19 Richard Biener <rguenther@suse.de>
10137 PR tree-optimization/87349
10138 PR tree-optimization/87342
10139 * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
10141 2018-09-18 Marek Polacek <polacek@redhat.com>
10143 P1064R0 - Allowing Virtual Function Calls in Constant Expressions
10144 * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
10146 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
10148 * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
10150 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
10152 PR rtl-optimization/86882
10153 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
10155 2018-09-18 Uros Bizjak <ubizjak@gmail.com>
10157 * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
10158 *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
10160 2018-09-18 Jonathan Wakely <jwakely@redhat.com>
10163 * doc/invoke.texi (Link Options): Fix formatting and grammar.
10165 2018-09-18 Richard Biener <rguenther@suse.de>
10167 PR middle-end/63155
10168 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
10169 (compute_samebase_partition_bases): Likewise.
10170 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
10171 (gimple_can_coalesce_p): Simplify.
10173 2018-09-18 Hans-Peter Nilsson <hp@bitrange.com>
10175 Handle a library implementation of ffs calling __builtin_ffs.
10176 * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
10177 (mmix_init_libfuncs): New function: make __builtin_ffs expand
10180 2018-09-17 David Malcolm <dmalcolm@redhat.com>
10182 * diagnostic-show-locus.c (class layout_range): Add field
10184 (layout_range::layout_range): Add "original_idx" param and use it
10185 to initialize new field.
10186 (make_range): Use 0 for original_idx.
10187 (layout::layout): Pass in index to calls to
10188 maybe_add_location_range.
10189 (layout::maybe_add_location_range): Add param "original_idx" and
10190 pass it on to layout_range.
10191 (layout::print_any_labels): Pass on range->m_original_idx to
10193 (gcc_rich_location::add_location_if_nearby): Use 0 for
10195 * gcc-rich-location.h (text_range_label::get_text): Update for new
10197 (range_label_for_type_mismatch::get_text): Likewise.
10199 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
10201 * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
10203 2018-09-17 David Malcolm <dmalcolm@redhat.com>
10205 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
10206 format_string_diagnostic_t.
10207 (fmtwarn_n): Likewise.
10208 * substring-locations.c
10209 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
10210 (format_warning_n_va): Convert to...
10211 (format_string_diagnostic_t::emit_warning_n_va): ...this.
10212 (format_warning_va): Convert to...
10213 (format_string_diagnostic_t::emit_warning_va): ...this.
10214 (format_warning_at_substring): Convert to...
10215 (format_string_diagnostic_t::emit_warning): ...this.
10216 (format_warning_at_substring_n): Convert to...
10217 (format_string_diagnostic_t::emit_warning_n): ...this.
10218 * substring-locations.h (class format_string_diagnostic_t): New
10220 (format_warning_va): Convert to
10221 format_string_diagnostic_t::emit_warning_va.
10222 (format_warning_n_va): Convert to
10223 format_string_diagnostic_t::emit_warning_n_va.
10224 (format_warning_at_substring): Convert to
10225 format_string_diagnostic_t::emit_warning.
10226 (format_warning_at_substring_n): Convert to
10227 format_string_diagnostic_t::emit_warning_n.
10229 2018-09-17 Cesar Philippidis <cesar@codesourcery.com>
10230 Bernd Schmidt <bernds_cb1@t-online.de>
10232 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
10235 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
10237 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
10238 operand 0 predicate to nonimmediate operand.
10239 (rint<mode>2_frndint): Remove insn pattern.
10240 (rint<mode>2): Change operand 1 predicate to general_operand.
10241 Extend operand 1 to XFmode and generate rintxf2 insn.
10242 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
10243 Do not use X87MODEF mode macro.
10244 (frndintxf2_<rounding>_i387): Rename from
10245 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro.
10246 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
10247 to XFmode and generate significandxf3 insn.
10249 2018-09-17 Richard Biener <rguenther@suse.de>
10251 PR tree-optimization/87328
10252 * tree-ssa-sccvn.c (process_bb): Remove assertion about not
10253 visiting unexecutable backedges when not iterating.
10254 (do_rpo_vn): Mark all edges not executable even when not
10257 2018-09-17 Martin Jambor <mjambor@suse.cz>
10260 * doc/invoke.texi (Warning Options): Likewise.
10262 2018-09-17 Richard Biener <rguenther@suse.de>
10264 PR tree-optimization/87301
10265 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
10266 clean EH info from leftover copy assignments.
10268 2018-09-17 Martin Liska <mliska@suse.cz>
10270 PR gcov-profile/85871
10271 * gcov.c (output_intermediate_file): Fix out of bounds
10274 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
10276 * config/arc/arc.c: Object attributes for core4 not reflected
10278 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
10281 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
10283 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
10285 2018-09-17 Martin Liska <mliska@suse.cz>
10287 * doc/gcov.texi: Document new option --use-hotness-colors.
10288 * gcov.c (struct source_info): Declare new field.
10289 (source_info::source_info): Set default for maximum_count.
10290 (print_usage): Add new -q option.
10291 (process_args): Process it.
10292 (accumulate_line_info): Save src->maximum_count.
10293 (output_line_beginning): Make color line number if
10294 flag_use_hotness_colors is set.
10295 (output_line_details): Pass default argument value.
10296 (output_lines): Pass src->maximum_count.
10298 2018-09-17 Martin Liska <mliska@suse.cz>
10300 * common/config/i386/i386-common.c (ix86_get_valid_option_values):
10301 Use processor_names table.
10302 * config/i386/i386.c (ix86_default_align): Use
10303 processor_cost_table for alignment values.
10304 (ix86_option_override_internal): Use processor_names.
10305 (ix86_function_specific_print): Likewise.
10306 * config/i386/i386.h (struct processor_costs):
10307 Add alignment values.
10308 (struct ptt): Remove and replace with const char *.
10309 * config/i386/x86-tune-costs.h (struct processor_costs):
10310 Declare default alignments for all costs.
10312 2018-09-17 Aldy Hernandez <aldyh@redhat.com>
10314 * tree-vrp.c (extract_range_from_unary_expr): Do not special case
10315 symbolics or VR_VARYING ranges for ABS_EXPR.
10316 * wide-int-range.cc (wide_int_range_abs): Return positive numbers
10317 when range will wrap.
10319 2018-09-15 Eric Botcazou <ebotcazou@adacore.com>
10321 PR middle-end/86864
10322 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
10323 before and after a JUMP_TABLE_DATA.
10325 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
10327 PR middle-end/87188
10328 * dojump.c (do_compare_and_jump): Canonicalize function pointers
10329 when one operand is a function pointer. Use POINTER_TYPE_P and
10330 FUNC_OR_METHOD_TYPE_P.
10331 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
10332 * fold-const.c (build_range_check): Likewise.
10333 * match.pd (simple_comparison): Likewise.
10335 2018-09-14 David Malcolm <dmalcolm@redhat.com>
10338 * spellcheck.c (get_edit_distance_cutoff): New function.
10339 (selftest::test_edit_distance_unit_test_oneway): Rename to...
10340 (selftest::test_get_edit_distance_one_way): ...this.
10341 (selftest::test_get_edit_distance_unit): Rename to...
10342 (selftest::test_get_edit_distance_both_ways): ...this.
10343 (selftest::test_edit_distances): Move tests to this new function,
10344 and test some more pairs of strings. Update for above renaming.
10345 (selftest::get_old_cutoff): New function.
10346 (selftest::test_get_edit_distance_cutoff): New function.
10347 (selftest::assert_suggested_for): New function.
10348 (ASSERT_SUGGESTED_FOR): New macro.
10349 (selftest::assert_not_suggested_for): New function.
10350 (ASSERT_NOT_SUGGESTED_FOR): New macro.
10351 (selftest::test_suggestions): New function.
10352 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
10353 tests to selftest::test_edit_distances and call it. Add calls to
10354 selftest::test_get_edit_distance_cutoff and
10355 selftest::test_suggestions.
10356 * spellcheck.h (get_edit_distance_cutoff): New function declaration.
10357 (best_match::consider): Replace hard-coded cutoff calculation with
10359 (best_match::get_cutoff): New declaration.
10360 (best_match::get_best_meaningful_candidate): Likewise.
10362 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
10364 * builtins.c (fold_builtin_strlen): Remove TODO comment.
10366 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
10369 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
10371 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
10372 terminated string literal.
10374 2018-09-14 Martin Sebor <msebor@redhat.com>
10376 * builtins.c (unterminated_array): Handle ARRAY_REF.
10377 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
10378 * builtins.h (unterminated_array): Declare extern.
10379 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
10381 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
10384 2018-09-14 Martin Sebor <msebor@redhat.com>
10385 Jeff Law <law@redhat.com>
10387 * builtins.c (unterminated_array): New.
10388 (expand_builtin_strcpy): Adjust.
10389 (expand_builtin_strcpy_args): Detect unterminated arrays.
10390 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
10391 unterminated arrays.
10392 * gimple-fold.h (get_maxval_strlen): Add argument.
10393 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
10395 * gimple-fold.c (get_range_strlen): Add argument.
10396 (get_maxval_strlen): Adjust.
10397 * gimple-fold.h (get_range_strlen): Add argument.
10399 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
10401 * config/i386/movdirintrin.h: Fix copyright year.
10403 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
10405 * reg-stack.c: Include regs.h.
10406 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
10407 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
10408 FIRST_STACK_REG, not DFmode.
10409 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
10410 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
10411 (change stack): Default register mode to the reg_raw_mode of
10412 FIRST_STACK_REG, not DFmode.
10413 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
10414 (*swapxf): Rename from swapxf.
10416 2018-09-14 Carl Love <cel@us.ibm.com>
10418 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
10419 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
10421 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
10424 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
10427 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
10430 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
10432 2018-09-14 Jason Merrill <jason@redhat.com>
10434 Fix --enable-gather-detailed-mem-stats.
10435 * hash-table.c (hash_table_usage): Change from variable to function.
10436 * hash-table.h: Adjust.
10437 * Makefile.in: Add missing dependencies on hash-table.h.
10439 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10441 PR tree-optimization/87259
10443 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
10444 execute_cse_reciprocals_1 has tried transforming.
10446 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
10448 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
10449 VR_VARYING for PLUS/MINUS_EXPR.
10451 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
10453 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
10456 2018-09-14 Richard Biener <rguenther@suse.de>
10458 PR middle-end/63155
10459 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
10460 bits for the merged partition.
10462 2018-09-13 Martin Sebor <msebor@redhat.com>
10463 Bernd Edlinger <bernd.edlinger@hotmail.de>
10465 * builtins.h (c_srlen): Add argument.
10466 * builtins.c (warn_string_no_nul): New function.
10467 (c_strlen): Add argument and use it. Update recursive calls.
10468 Pass DECL argument to string_constant to get info on non
10469 terminated strings. Update *NONSTR as needed.
10470 (fold_builtin_strlen): Add argument to calls to c_strlen.
10471 Warn for unterminated arrays.
10472 (warn_string_no_null): Add prototype.
10473 * expr.c (string_constant): Update arguments. Update recursive
10474 calls appropriately. Detect missing NUL terminator and outermost
10475 declaration its missing in.
10476 Improve checks for arrays with nonzero lower bound or elements
10477 that are not a single byte. Simplify offset computation.
10478 Simplify checks for non-NUL terminated strings.
10479 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
10480 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
10482 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
10484 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
10486 * fold-const.c (c_getstr): Fix function comment. Remove unused third
10487 argument. Fix range checks.
10488 * fold-const.h (c_getstr): Adjust protoype.
10489 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
10490 string is constant but contains no NUL byte.
10492 * expr.c (string_constant): Adjust function comment.
10493 Remove bogus check for zero termination.
10495 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
10497 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
10498 (get_constant_size): Don't make STRING_CSTs larger than they are.
10499 (check_string_literal): New check function for STRING_CSTs.
10500 (output_constant): Use it.
10502 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
10505 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
10507 2018-09-13 Richard Biener <rguenther@suse.de>
10509 PR tree-optimization/87263
10510 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
10511 (struct unwind_state): Add max_rpo field.
10512 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
10513 Compute max_rpo, the max RPO number a block can be backwards reached
10514 from. Re-write non-iterating mode to a RPO ordered worklist approach,
10515 separating it from the iterating mode.
10517 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
10519 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
10520 (rfs_decision): New scheduling decision.
10522 2018-09-13 Richard Biener <rguenther@suse.de>
10525 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
10526 (vn_nary_op_insert_pieces_predicated): Do not write useless
10527 valid_dominated_by_p entry outside of the allocated storage.
10529 2018-09-13 Omar Sandoval <osandov@osandov.com>
10530 Tom de Vries <tdevries@suse.de>
10533 * dwarf2out.c (is_c): New function.
10534 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
10536 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
10539 * config/aarch64/aarch64.md (*aarch64_bfxil):
10541 * config/aarch64/constraints.md (Ulc): Define.
10542 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
10544 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
10547 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
10549 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
10550 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
10551 aarch64_layout_frame call.
10552 (aarch64_expand_epilogue): Likewise.
10553 (aarch64_initial_elimination_offset): Likewise.
10554 (aarch64_get_separate_components): Likewise.
10555 (aarch64_use_return_insn_p): Likewise.
10556 (aarch64_layout_frame): Remove unneeded check.
10558 2018-09-13 Jakub Jelinek <jakub@redhat.com>
10560 * configure.ac: Only append
10561 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
10562 gcc_config_arguments if it was never reconfigured or last reconfigure
10563 was with different arguments.
10564 * configure: Regenerated.
10566 2018-09-13 Jakub Jelinek <jakub@redhat.com>
10567 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10569 PR middle-end/87290
10570 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
10571 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
10573 2018-09-13 Jakub Jelinek <jakub@redhat.com>
10575 PR tree-optimization/87287
10576 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
10577 X % C == 0 to X % (unsigned) C == 0 optimization to ...
10578 * match.pd (X % C == 0): ... here. New optimization.
10580 2018-09-12 Jakub Jelinek <jakub@redhat.com>
10582 PR middle-end/82853
10583 * expr.h (maybe_optimize_mod_cmp): Declare.
10584 * expr.c (mod_inv): New function.
10585 (maybe_optimize_mod_cmp): New function.
10586 (do_store_flag): Use it.
10587 * cfgexpand.c (expand_gimple_cond): Likewise.
10589 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
10590 Julian Brown <julian@codesourcery.com>
10592 PR middle-end/86336
10593 * gimplify.c (gimplify_scan_omp_clauses): Set
10594 target_firstprivatize_array_bases in OpenACC parallel and kernels
10595 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
10596 OpenACC data regions.
10598 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
10600 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
10601 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
10603 2018-09-12 Richard Biener <rguenther@suse.de>
10605 PR tree-optimization/87280
10606 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
10607 edge but unreachable target.
10608 (do_rpo_vn): For conservatively handling a PHI only mark
10609 the backedge executable but not the block reachable.
10611 2018-09-12 Richard Biener <rguenther@suse.de>
10613 PR tree-optimization/87266
10614 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
10617 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
10619 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
10621 ("trunc<BFP:mode><DFP_ALL:mode>2")
10622 ("trunc<DFP_ALL:mode><BFP:mode>2")
10623 ("extend<BFP:mode><DFP_ALL:mode>2")
10624 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
10625 according to the target operand type.
10627 2018-09-12 Jakub Jelinek <jakub@redhat.com>
10628 Andreas Krebbel <krebbel@linux.ibm.com>
10630 PR tree-optimization/86844
10631 * gimple-ssa-store-merging.c
10632 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
10633 there are any overlapping stores in between them, make sure they are
10634 also coalesced or we give up completely.
10636 2018-09-12 Jakub Jelinek <jakub@redhat.com>
10638 PR middle-end/87248
10639 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
10640 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
10642 2018-09-12 Tom de Vries <tdevries@suse.de>
10644 * common.opt (gdescribe-dies): Add option.
10645 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
10646 attribute for artifical and nameless decls.
10647 (dwarf2out_register_external_die): Add description attribute to
10648 external reference die.
10649 (add_desc_attribute): New functions.
10650 (gen_subprogram_die): Add description attribute to
10651 DW_TAG_call_site_parameter.
10652 * tree-pretty-print.c (print_generic_expr_to_str): New function.
10653 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
10654 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
10655 -gno-describe-dies.
10656 (@item -gdescribe-dies): Add.
10658 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
10660 * tree-vrp.c (vrp_shift_undefined_p): Remove.
10661 (extract_range_from_binary_expr_1: Call
10662 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
10663 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
10666 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
10668 * gimple-ssa-warn-alloca.c
10669 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
10670 field for ALLOCA_BOUND_*_LARGE.
10672 2018-09-11 Nathan Sidwell <nathan@acm.org>
10674 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
10676 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
10678 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
10679 for clobbers. Remove obsolete comment.
10681 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
10683 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
10684 mpxchk, mpxld and mpxst types.
10685 (define_attr length_immediate): Remove all processing of mpx types.
10686 (define_attr prefix_0f): Ditto.
10687 (define_attr memory): Ditto.
10689 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
10691 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
10692 (log<mode>2): Change operand 1 predicate to general_operand.
10693 Extend operand 1 to XFmode and generate logxf3 insn.
10694 (log10<mode>2): Change operand 1 predicate to general_operand.
10695 Extend operand 1 to XFmode and generate log10xf3 insn.
10696 (log2<mode>2): Change operand 1 predicate to general_operand.
10697 Extend operand 1 to XFmode and generate log2xf3 insn.
10698 (fyl2xp1_extend<mode>xf3_i387): Remove.
10699 (log1p<mode>2): Change operand 1 predicate to general_operand.
10700 Extend operand 1 to XFmode and generate log1pxf3 insn.
10701 (fxtract_extend<mode>xf3_i387): Remove.
10702 (logb<mode>2): Change operand 1 predicate to general_operand.
10703 Extend operand 1 to XFmode and generate logbxf3 insn.
10704 (ilogb<mode>2): Change operand 1 predicate to general_operand.
10705 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
10706 (significand<mode>2): Change operand 1 predicate to general_operand.
10707 Extend operand 1 to XFmode and generate significandxf3 insn.
10709 2018-09-11 Nathan Sidwell <nathan@acm.org>
10711 * gcc.c (perror_with_name, pfatal_with_name): Delete.
10712 (load_specs): Use fatal_error.
10713 (DELETE_IF_ORDINARY, process_command): Use error.
10714 (execute, run_attempt): Use fatal_error.
10716 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
10718 * diagnostic-core.h (sorry_at): New prototype.
10719 * diagnostic.c (sorry_at): New function.
10721 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
10723 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
10724 by zero as VR_UNDEFINED.
10726 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
10728 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
10729 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
10730 (<sincos>mode2): New expander.
10731 (sincos_extend<mode>xf3_i387): Remove insn pattern.
10732 (sincos -> sin, cos splitters): Remove splitter patterns.
10733 (sincos<mode>3): Change operand 2 predicate to general_operand.
10734 Extend operand 2 to XFmode and generate sincosxf3 insn.
10735 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
10736 Change operand 3 predicate to const1_operand.
10737 (fptan_extend<mode>xf4_i387): Remove insn pattern.
10738 (tanxf2): Update operands in the call to fptanxf4_i387.
10739 (tan<mode>2): Change operand 1 predicate to general_operand.
10740 Extend operand 1 to XFmode and generate tanxf3 insn.
10741 (atan2xf3): Rename from *fpatanxf3_i387.
10742 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
10743 (atan2xf3): Remove expander.
10744 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
10745 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
10746 (atan<mode>2): Change operand 1 predicate to general_operand.
10747 Extend operand 1 to XFmode and generate atanxf3 insn.
10749 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
10751 * config/i386/i386.md (x87/SSE constant load splitter): Use
10752 memory_operand instead of nonimmediate_operand for input operand
10755 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
10757 * config/i386/i386.md (float partial SSE register stall splitter): Move
10758 splitter near its instruction pattern.
10759 (float_extend partial SSE register stall splitter): Ditto.
10760 (float_truncate partial SSE register stall splitter): Ditto.
10762 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
10765 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
10766 to speculation_safe_value_not_needed.
10769 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
10770 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
10771 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
10772 leaf_function_p, instead use has_hard_reg_initial_val.
10774 2018-09-09 Nathan Sidwell <nathan@acm.org>
10776 * gcc.h (pfatal_with_name): Don't declare here.
10777 * gcc.c (pfatal_with_name): Make static.
10779 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
10781 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
10784 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
10786 PR rtl-optimization/85458
10787 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
10788 priority hook to reduce the priority of EXPR.
10790 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
10792 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
10793 DImode for x87 on 32bit targets. Conditionally disable x87 modes
10794 with X87_ENABLE_FLOAT. Remove preparation code.
10795 (*float<SWI48:mode><MODEF:mode>2): Rename from
10796 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
10797 math using "enabled" attribute.
10798 (*floatdi<MODEF:mode>2_i387): Rename from
10799 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
10800 enable for 32bit targets only.
10801 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
10803 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
10804 as operand 1 predicate. Rewrite as define_insn_and_split.
10805 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
10807 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
10809 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
10810 to fallthru to FLOAT case.
10812 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
10815 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
10816 around folding of vec_sl to handle out of range shift values.
10818 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
10820 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
10821 Update callers to gen_fix_trunc<mode>_i387_fisttp
10822 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
10823 nonimmediate_operand.
10824 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
10825 and corresponding splitters.
10826 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
10827 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
10828 (fix_truncdi_i387_with_temp): Remove insn pattern
10829 and corresponding splitters.
10830 (fix_trunc<mode>_i387): Change operand 0 predicate to
10831 nonimmediate_operand.
10832 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
10833 and corresponding splitters.
10834 (*fistdi2_1): Remove.
10836 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
10837 (lrintxfdi2): Remove expander. Reimplement as define_insn.
10838 (*fist<mode>2_1): Remove.
10839 (fist<mode>2): Ditto.
10840 (fist<mode>2_with_temp): Remove insn pattern and corresponding
10842 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
10843 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
10844 (fistdi2_<rounding>): Change operand 0 predicate to
10845 nonimmediate_operand.
10846 (fistdi2_<rounding>_with_temp): Remove insn pattern
10847 and corresponding splitters.
10848 (fist<mode>2_<rounding>): Change operand 0 predicate to
10849 nonimmediate_operand.
10850 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
10851 and corresponding splitters.
10853 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
10855 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
10857 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
10860 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
10862 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
10863 early gimple folding of vec_splat().
10864 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
10865 * gimple-fold.h: Add an extern define for tree_vec_extract().
10867 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
10869 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
10870 wrappers around TREE_TYPE comparisons.
10872 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
10875 * config/s390/predicates.md: Add nonsym_memory_operand.
10876 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
10877 contains a SYMBOL_REF, load it into an intermediate pseudo.
10878 (s390_emit_compare_and_swap): Legitimize operand.
10879 * config/s390/s390.md: Use the new nonsym_memory_operand
10880 with UNSPECV_CAS patterns.
10882 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
10885 * config/s390/s390-passes.def: New file.
10886 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
10888 (make_pass_s390_early_mach): Add declaration.
10889 * config/s390/s390.c (make_pass_s390_early_mach):
10890 (s390_option_override): Remove dynamic registration.
10891 * config/s390/t-s390: Add s390-passes.def.
10893 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
10895 * config/s390/s390.c (s390_decompose_constant_pool_ref):
10896 Remove UNSPEC_LTREL_BASE check.
10897 (annotate_constant_pool_refs): Likewise.
10898 (find_constant_pool_ref): Likewise.
10899 (find_ltrel_base): Removed.
10900 (replace_ltrel_base): Removed.
10901 (s390_mainpool_finish): Remove replace_ltrel_base call.
10902 (s390_chunkify_start): Remove pending LTREL_BASE logic.
10903 (s390_chunkify_finish): Remove replace_ltrel_base call.
10904 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
10906 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
10909 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
10910 to speculation_safe_value_not_needed.
10912 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
10913 Bernd Schmidt <bernds_cb1@t-online.de>
10915 * config/nvptx/nvptx-opts.h: New file.
10916 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
10917 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
10918 (ASM_SPEC): Define.
10919 (TARGET_SM35): New macro.
10920 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
10922 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
10924 (misa=): New option.
10925 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
10927 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
10929 * config/i386/i386.md (truncdfsf2): Remove expander.
10930 (truncdfsf2_with_temp): Ditto.
10931 (truncxf<mode>2): Ditto.
10932 (*truncdfsf_fast_mixed): Remove insn pattern.
10933 (*truncdfsf_fast_i387): Ditto.
10934 (*truncdfsf_mixed): Ditto.
10935 (*truncdfsf_i387): Ditto.
10936 (*truncdfsf2_i387_1): Ditto.
10937 (*truncxfsf2_mixed): Ditto.
10938 (*truncxfdf2_mixed): Ditto.
10939 (*truncxf<mode>2_i387_noop): Ditto. Update callers
10940 to call gen_truncxf<mode>2 instead.
10941 (*truncxf<mode>2_i387): Remove.
10942 (reg->reg splitters): Remove splitter pattern.
10943 (reg->mem splitters): Ditto.
10945 (truncdfsf2): New insn pattern.
10946 (truncxf<mode>2): Ditto.
10948 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10950 * tree-ssa-math-opts.c (is_mult_by): New function.
10951 (is_square_of): Use the above.
10952 (optimize_recip_sqrt): New function.
10953 (pass_cse_reciprocals::execute): Use the above.
10955 2018-09-05 Richard Biener <rguenther@suse.de>
10958 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
10959 to zero-init the emplaced vec.
10961 2018-09-05 Martin Liska <mliska@suse.cz>
10963 PR tree-optimization/87205
10964 * tree-switch-conversion.c (pass_lower_switch::execute):
10965 Group cases for switch statements.
10967 2018-09-05 Richard Biener <rguenther@suse.de>
10969 PR tree-optimization/87217
10970 * tree-ssa-sccvn.c (vuse_valueize): New.
10971 (vn_reference_lookup_pieces): Use it.
10972 (vn_reference_lookup): Likewise.
10974 2018-09-05 Nathan Sidwell <nathan@acm.org>
10977 * stor-layout.c (place_field): Scan forwards to check last
10978 bitfield when ms_bitfield_placement is in effect.
10980 2018-09-05 Richard Biener <rguenther@suse.de>
10983 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
10986 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
10987 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
10989 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
10990 * config.gcc (extra_objs): Build it.
10991 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
10993 * config/aarch64/aarch64-passes.def
10994 (pass_tag_collision_avoidance): New pass.
10995 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
10996 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
10997 (aarch64_classify_address): Remove static qualifier.
10998 (aarch64_address_info, aarch64_address_type): Move to...
10999 * config/aarch64/aarch64-protos.h: ... here.
11000 (make_pass_tag_collision_avoidance): New function.
11001 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
11004 2018-09-05 Martin Liska <mliska@suse.cz>
11006 * doc/gcov.texi: Update documentation of humar
11008 * gcov.c (format_count): Print one decimal place, it provides
11009 more fine number of situations like '1G' vs. '1.4G'.
11011 2018-09-05 Martin Liska <mliska@suse.cz>
11014 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
11015 * optc-gen.awk: Allow 'Var' for Deprecated options in order
11016 to generate a MASK value.
11018 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
11021 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
11022 if frame pointer isn't used.
11023 (compute_frame_pointer_to_fb_displacement): Likewise.
11025 2018-09-04 Jakub Jelinek <jakub@redhat.com>
11028 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
11029 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
11030 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
11031 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
11032 and OPTION_MASK_ISA_XSAVEC_UNSET.
11034 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
11036 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
11037 XOR operations in NAND case.
11039 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
11041 * wide-int-range.cc (wide_int_range_convert): New.
11042 * wide-int-range.h (wide_int_range_convert): New.
11043 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
11044 code into wide_int_range_convert.
11045 (extract_range_into_wide_ints): Do not munge anti range constants
11046 into the entire domain. Just return the range back.
11048 2018-09-04 Martin Liska <mliska@suse.cz>
11050 * genmatch.c (output_line_directive): Add new argument
11052 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
11054 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
11056 * doc/invoke.texi (Option Summary): Add whitespace.
11058 * doc/invoke.texi (Option Summary): Add -Waligned-new.
11060 2018-09-04 Richard Biener <rguenther@suse.de>
11062 PR tree-optimization/87211
11063 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
11064 backedge value we're supposed to treat as VARYING also number
11065 the PHI to VARYING in case it got a different value-number already.
11067 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
11069 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
11070 (extract_range_from_binary_expr_1): Do not call
11071 vrp_can_optimize_bit_op.
11072 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
11074 (wide_int_range_get_mask_and_bounds): New.
11075 (wide_int_range_optimize_bit_op): New.
11076 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
11077 (wide_int_range_bit_and): Same.
11078 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
11079 (wide_int_range_optimize_bit_op): New.
11080 (wide_int_range_get_mask_and_bounds): New.
11082 2018-09-04 Richard Biener <rguenther@suse.de>
11084 PR tree-optimization/87176
11085 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
11086 variable. When value-numbering a virtual PHI node make sure
11087 to not value-number to the backedge value.
11089 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
11091 * doc/extend.texi (Long Long, Hex Floats): Document support for
11092 long long and hex floats in more recent versions of ISO C++.
11094 2018-09-03 Richard Biener <rguenther@suse.de>
11096 PR tree-optimization/87177
11097 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
11100 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
11102 * bb-reorder.c (edge_order): Convert to C-qsort-style
11103 tri-state comparator.
11104 (reorder_basic_blocks_simple): Change std::stable_sort to
11107 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
11109 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
11110 tri-state comparator.
11111 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
11113 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
11115 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
11116 (mergesort): ... here as maximum count for using netsort.
11117 (gcc_qsort): Set nlim to 3 if stable sort is requested.
11118 (gcc_stablesort): New.
11119 * system.h (gcc_stablesort): Declare.
11121 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
11123 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
11124 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
11125 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
11127 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
11129 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
11130 lxsdx and stxsdx alternatives.
11131 (*mov<mode>_hardfloat64): Ditto.
11132 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
11134 2018-09-03 Richard Biener <rguenther@suse.de>
11136 PR tree-optimization/87200
11137 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
11140 2018-09-03 Martin Liska <mliska@suse.cz>
11142 PR tree-optimization/87201
11143 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
11144 Fix parenthesis in an expression.
11146 2018-09-03 Richard Biener <rguenther@suse.de>
11148 PR tree-optimization/87197
11149 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
11150 visited. CSE the VN_INFO hashtable lookup.
11152 PR tree-optimization/87169
11153 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
11154 iterating make sure there's no extra backedges from irreducible
11155 regions feeding the header. Mark the destination block
11158 2018-09-03 Martin Liska <mliska@suse.cz>
11161 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
11162 * common/common-targhooks.c (default_get_valid_option_values):
11164 * common/common-targhooks.h (default_get_valid_option_values):
11166 * common/config/i386/i386-common.c: Move processor_target_table
11168 (ix86_get_valid_option_values): New function.
11169 (TARGET_GET_VALID_OPTION_VALUES): New macro.
11170 * config/i386/i386.c (struct ptt): Move to i386-common.c.
11171 (PTA_*): Move all defined masks into i386-common.c.
11172 (ix86_function_specific_restore): Use new processor_cost_table.
11173 * config/i386/i386.h (struct ptt): Moved from i386.c.
11174 (struct pta): Likewise.
11175 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
11176 * doc/tm.texi.in: Likewise.
11177 * opt-suggestions.c (option_proposer::suggest_option):
11178 Pass prefix to build_option_suggestions.
11179 (option_proposer::get_completions): Likewise.
11180 (option_proposer::build_option_suggestions): Use the new target
11182 * opts.c (struct option_help_tuple): New struct.
11183 (print_filtered_help): Use the new target hook.
11185 2018-09-03 Martin Liska <mliska@suse.cz>
11187 PR middle-end/59521
11188 * predict.c (set_even_probabilities): Add likely_edges
11189 argument and handle cases where we have precisely one
11191 (combine_predictions_for_bb): Catch also likely_edges.
11192 (tree_predict_by_opcode): Handle gswitch statements.
11193 * tree-cfg.h (find_case_label_for_value): New declaration.
11194 (find_taken_edge_switch_expr): Likewise.
11195 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
11196 Find pivot in decision tree based on probabily, not by number of
11199 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
11201 * doc/standards.texi (Standards): Update Objective-C reference.
11203 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
11205 * doc/install.texi (Prerequisites): Update link for MPC.
11207 2018-09-01 Michael Matz <matz@suse.de>
11209 PR tree-optimization/87074
11210 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
11211 PHIs for outer-loop uses.
11213 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
11215 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
11216 * doc/invoke.texi (C Dialect Options): Ditto.
11218 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
11220 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
11222 2018-08-31 Richard Biener <rguenther@suse.de>
11224 PR tree-optimization/87168
11225 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
11226 (rpo_elim::eliminate_avail): When OP was not visited it must
11229 2018-08-31 David Malcolm <dmalcolm@redhat.com>
11231 * tree-vrp.c (copy_value_range): Convert param "from" from
11232 "value_range *" to "const value_range *".
11233 (range_is_null): Likewise for param "vr".
11234 (range_int_cst_p): Likewise.
11235 (range_int_cst_singleton_p): Likewise.
11236 (symbolic_range_p): Likewise.
11237 (value_ranges_intersect_p): Likewise for both params.
11238 (value_range_nonnegative_p): Likewise for param "vr".
11239 (value_range_constant_singleton): Likewise.
11240 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
11241 (extract_range_into_wide_ints): Likewise for param "vr".
11242 (extract_range_from_multiplicative_op): Likewise for params "vr0"
11244 (vrp_can_optimize_bit_op): Likewise.
11245 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
11247 (extract_range_from_unary_expr): Likewise.
11248 (debug_value_range): Likewise for param "vr".
11249 (value_range::dump): Add "const" qualifier.
11250 (vrp_prop::check_array_ref): Convert local "vr" from
11251 "value_range *" to "const value_range *".
11252 (vrp_prop::check_mem_ref): Likewise.
11253 (vrp_prop::visit_stmt): Likewise for local "old_vr".
11254 (vrp_intersect_ranges_1): Likewise for param "vr_1".
11255 (vrp_intersect_ranges): Likewise.
11256 (simplify_stmt_for_jump_threading): Likewise for local "vr".
11257 (vrp_prop::vrp_finalize): Likewise.
11258 * tree-vrp.h (value_range::dump): Add "const" qualifier.
11259 (vrp_intersect_ranges): Add "const" qualifier to params as above.
11260 (extract_range_from_unary_expr): Likewise.
11261 (value_range_constant_singleton): Likewise.
11262 (symbolic_range_p): Likewise.
11263 (copy_value_range): Likewise.
11264 (extract_range_from_binary_expr_1): Likewise.
11265 (range_int_cst_p): Likewise.
11266 (vrp_set_zero_nonzero_bits): Likewise.
11267 (range_int_cst_singleton_p): Likewise.
11269 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
11271 * config/aarch64/arm_neon.h (vabsd_s64): New.
11272 (vnegd_s64): Likewise.
11274 2018-08-31 Martin Jambor <mjambor@suse.cz>
11276 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
11278 2018-08-31 Martin Liska <mliska@suse.cz>
11280 * ipa-icf.c (sem_item::add_type): Use
11281 sem_item::m_type_hash_cache.
11282 * ipa-icf.h: Move the cache from sem_item_optimizer
11285 2018-08-31 Nathan Sidwell <nathan@acm.org>
11287 * doc/extend.texi (Backwards Compatibility): Remove implicit
11288 extern C leeway of () being (...).
11290 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11292 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
11294 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
11298 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
11300 2018-08-31 Jakub Jelinek <jakub@redhat.com>
11302 PR middle-end/87138
11303 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
11304 gen_int_mode. Formatting fixes.
11306 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
11308 * target.def (custom_function_descriptors): Improve documentation.
11309 * doc/tm.texi.in (Trampolines): Expand discussion of function
11310 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
11311 beginning of the section.
11312 * doc/tm.texi: Regenerated.
11314 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
11316 * cfg.h (class auto_edge_flag): Spell out the template-id of the
11317 base class in the initializer list. This is a workaround for
11318 building with older GCC.
11319 (class auto_bb_flag): Likewise.
11321 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
11323 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
11324 (altivec_vcmpequ<VI_char>_p): Remove star.
11325 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
11327 (expand_strncmp_vec_sequence): New function.
11328 (emit_final_str_compare_vec): New function.
11329 (expand_strn_compare): Add support for vector strncmp.
11330 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
11331 length specification to bytes.
11332 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
11333 (vcmpnezb_p): New pattern.
11334 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
11335 for option -mstring-compare-inline-limit.
11337 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
11339 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
11340 (PTA_SKYLAKE): Add PTA_AES.
11341 (PTA_GOLDMONT): Likewise.
11343 2018-08-29 Jan Hubicka <jh@suse.cz>
11346 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
11347 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
11349 2018-08-29 Jan Hubicka <jh@suse.cz>
11351 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
11353 (hash_tree): Do not visit TYPE_STUB_DECL.
11354 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
11355 stream TYPE_STUB_DECL.
11356 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
11357 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
11358 after free_lang_data.
11359 (type_in_anonymous_namespace_p): Likewise.
11361 2018-08-29 Jan Hubicka <jh@suse.cz>
11363 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
11364 comment that it has to be even number.
11365 (class sreal): Change m_sig type to int32_t.
11366 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
11367 int64_t for temporary calculations.
11368 (sreal_verify_basics): Drop one bit from minimum and maximum.
11370 2018-08-30 Richard Biener <rguenther@suse.de>
11372 PR tree-optimization/87147
11373 * tree-ssa-sccvn.c (SSA_VISITED): New function.
11374 (visit_phi): When the degenerate result is from the backedge and
11375 we didn't visit its definition yet drop to VARYING.
11376 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
11378 2018-08-29 Jan Hubicka <jh@suse.cz>
11380 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
11382 (hash_tree): Likewise.
11384 2018-08-29 Jan Hubicka <jh@suse.cz>
11386 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
11387 and TYPE_NEXT_REF_TO.
11389 2018-08-29 Jan Hubicka <jh@suse.cz>
11391 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
11392 comment that it has to be even number.
11393 (class sreal): Change m_sig type to int32_t.
11394 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
11395 int64_t for temporary calculations.
11396 (sreal_verify_basics): Drop one bit from minimum and maximum.
11398 2018-08-30 Tamar Christina <tamar.christina@arm.com>
11400 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
11402 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
11404 PR middle-end/86995
11405 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
11406 if to_add is negative.
11408 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
11410 PR middle-end/87053
11411 * builtins.c (c_strlen): Improve range checks.
11413 2018-08-29 Martin Sebor <msebor@redhat.com>
11414 Jeff Law <law@redhat.com>
11416 PR tree-optimization/86714
11417 PR tree-optimization/86711
11418 * builtins.c (c_strlen): Add arguments to call to string_constant.
11419 * expr.c (string_constant): Add argument. Detect missing nul
11420 terminator and outermost declaration it's missing in.
11421 * expr.h (string_constant): Add argument.
11422 * fold-const.c (read_from_constant_string): Add arguments to call to
11424 (c_getstr): Likewise.
11425 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
11426 to string_constant.
11427 * tree-ssa-strlen.c (get_stridx): Likewise.
11429 2018-08-29 Jan Hubicka <jh@suse.cz>
11431 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
11432 Do not stream DECL_VINDEX.
11433 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
11434 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
11435 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
11437 2018-08-29 Richard Biener <rguenther@suse.de>
11439 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
11440 virtual operands that are not default defs to honor region
11442 (rpo_vn_valueize): Remove ineffective code here.
11444 2018-08-29 Richard Biener <rguenther@suse.de>
11446 PR tree-optimization/87132
11447 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
11448 when skipping defs reachable over backedges.
11450 2018-08-29 Richard Biener <rguenther@suse.de>
11452 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
11453 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
11454 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
11455 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
11456 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
11457 (vn_reference_lookup_3): Remove use of const_parms.
11458 (free_rpo_vn): Do not free const_parms.
11459 (do_rpo_vn): Do not call init_const_parms.
11460 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
11461 SSA_NAME_POINTS_TO_READONLY_MEMORY.
11462 (call_may_clobber_ref_p_1): Likewise.
11464 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
11467 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
11469 (-ftree-scev-cprop): Document.
11471 2018-08-29 Jan Hubicka <jh@suse.cz>
11473 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
11475 (sreal constructor): Update.
11476 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
11477 sreal:operator/): Update.
11479 2018-08-29 Martin Liska <mliska@suse.cz>
11481 * tree-switch-conversion.c (switch_conversion::expand):
11482 Strenghten assumption about gswitch statements.
11484 2018-08-29 Richard Biener <rguenther@suse.de>
11486 PR tree-optimization/87117
11487 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
11488 re-value-number released SSA VDEFs.
11490 2018-08-29 Richard Biener <rguenther@suse.de>
11492 PR tree-optimization/87126
11493 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
11495 2018-08-28 Jim Wilson <jimw@sifive.com>
11497 * config/riscv/pic.md: Rewrite.
11498 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
11500 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
11501 (SOFTF, default_load, softload, softstore): New.
11503 2018-08-28 Jeff Law <law@redhat.com>
11505 * fold-const.c (fold_binary_loc): Remove recently added assert.
11507 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
11509 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
11510 to OP parmeter of generated function.
11512 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
11514 PR tree-optimization/87009
11515 * match.pd: Add boolean optimizations.
11517 2018-08-28 Martin Sebor <msebor@redhat.com>
11519 PR middle-end/86631
11520 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
11521 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
11522 (pass_walloca::gate): Use it.
11523 (alloca_call_type): Same.
11524 (pass_walloca::execute): Same.
11525 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
11527 2018-08-28 David Malcolm <dmalcolm@redhat.com>
11529 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
11530 GCC_VERSION for usage of "__gcc_dump_printf__" format from
11531 >= 3005 to >= 9000.
11533 2018-08-28 Richard Biener <rguenther@suse.de>
11535 PR tree-optimization/87124
11536 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
11537 constants before looking up avail.
11539 2018-08-28 Jakub Jelinek <jakub@redhat.com>
11541 PR middle-end/87099
11542 * calls.c (maybe_warn_nonstring_arg): Punt early if
11543 warn_stringop_overflow is zero. Don't call get_range_strlen
11544 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
11545 Swap comparison operands to have constants on rhs. Only use
11546 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
11547 increment lenrng[0].
11549 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
11551 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
11552 use of tree_to_shwi. Remove duplicated test for the size being
11553 a whole number of bytes.
11555 2018-08-28 Richard Biener <rguenther@suse.de>
11557 PR tree-optimization/87117
11558 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
11559 Handle removed stmt without LHS (GIMPLE_NOP).
11561 2018-08-28 Richard Biener <rguenther@suse.de>
11563 PR tree-optimization/87117
11564 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
11565 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
11567 2018-08-28 Richard Biener <rguenther@suse.de>
11569 PR tree-optimization/87117
11570 * tree-ssa-pre.c (compute_avail): Do not make expressions
11571 with predicated values available.
11572 (get_expr_value_id): Assert we do not run into predicated value
11575 2018-08-28 Richard Biener <rguenther@suse.de>
11577 PR tree-optimization/87117
11578 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
11579 get virtual operands.
11580 (get_expr_operands): Handle STRING_CST like other decls.
11582 2018-08-28 Martin Liska <mliska@suse.cz>
11584 * tree.h: Update documentation of fndecl_built_in_p
11588 2018-08-27 Jeff Law <law@redhat.com>
11589 PR tree-optimization/87110
11590 * tree-ssa-dse.c (compute_trims): Handle non-constant
11593 2018-08-27 Martin Sebor <msebor@redhat.com>
11595 PR tree-optimization/86914
11596 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
11598 2018-08-27 Martin Sebor <msebor@redhat.com>
11600 PR tree-optimization/87112
11601 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
11602 the type of the bound argument.
11604 2018-08-27 Jeff Law <law@redhat.com>
11606 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
11607 type does not have a TYPE_SIZE_UNIT.
11609 2018-08-27 Steve Ellcey <sellcey@cavium.com>
11611 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
11612 with include of backend.h.
11614 2018-08-27 Richard Biener <rguenther@suse.de>
11616 PR tree-optimization/86927
11617 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
11618 use const cond reduction code.
11620 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
11622 PR tree-optimization/85758
11623 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
11625 2018-08-27 David Malcolm <dmalcolm@redhat.com>
11628 * diagnostic-show-locus.c (class layout_range): Update for
11629 conversion of show_caret_p to a tri-state.
11630 (layout_range::layout_range): Likewise.
11631 (make_range): Likewise.
11632 (layout::maybe_add_location_range): Likewise.
11633 (layout::should_print_annotation_line_p): Don't show annotation
11634 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
11635 (layout::get_state_at_point): Update for conversion of
11636 show_caret_p to a tri-state. Bail out early for
11637 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
11638 underlining or source colorization.
11639 (gcc_rich_location::add_location_if_nearby): Update for conversion
11640 of show_caret_p to a tri-state.
11641 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
11642 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
11644 (selftest::test_one_liner_labels): Likewise.
11645 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
11646 conversion of show_caret_p to a tri-state.
11647 * pretty-print.c (text_info::set_location): Likewise.
11648 * pretty-print.h (text_info::set_location): Likewise.
11649 * substring-locations.c (format_warning_n_va): Likewise.
11650 * tree-diagnostic.c (default_tree_printer): Likewise.
11651 * tree-pretty-print.c (newline_and_indent): Likewise.
11653 2018-08-27 David Malcolm <dmalcolm@redhat.com>
11656 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
11657 line above for line-insertion fix-it hints.
11658 (selftest::test_fixit_insert_containing_newline): Update the
11659 expected results, and add a test with line-numbering enabled.
11661 2018-08-27 Martin Liska <mliska@suse.cz>
11664 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
11665 params with DECL_HAS_VALUE_EXPR_P.
11667 2018-08-27 Martin Liska <mliska@suse.cz>
11669 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
11670 selected expansion strategy.
11672 2018-08-27 Martin Liska <mliska@suse.cz>
11674 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
11675 * builtins.c (is_builtin_fn): Likewise.
11676 * attribs.c (diag_attr_exclusions): Use new function
11677 fndecl_built_in_p and remove check for FUNCTION_DECL if
11679 (builtin_mathfn_code): Likewise.
11680 (fold_builtin_expect): Likewise.
11681 (fold_call_expr): Likewise.
11682 (fold_builtin_call_array): Likewise.
11683 (fold_call_stmt): Likewise.
11684 (set_builtin_user_assembler_name): Likewise.
11685 (is_simple_builtin): Likewise.
11686 * calls.c (gimple_alloca_call_p): Likewise.
11687 (maybe_warn_nonstring_arg): Likewise.
11688 * cfgexpand.c (expand_call_stmt): Likewise.
11689 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
11690 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
11691 (cgraph_node::verify_node): Likewise.
11692 * cgraphclones.c (build_function_decl_skip_args): Likewise.
11693 (cgraph_node::create_clone): Likewise.
11694 * config/arm/arm.c (arm_insert_attributes): Likewise.
11695 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
11696 * dse.c (scan_insn): Likewise.
11697 * expr.c (expand_expr_real_1): Likewise.
11698 * fold-const.c (operand_equal_p): Likewise.
11699 (fold_binary_loc): Likewise.
11700 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
11701 * gimple-low.c (lower_stmt): Likewise.
11702 * gimple-pretty-print.c (dump_gimple_call): Likewise.
11703 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11705 * gimple.c (gimple_build_call_from_tree): Likewise.
11706 (gimple_call_builtin_p): Likewise.
11707 (gimple_call_combined_fn): Likewise.
11708 * gimplify.c (gimplify_call_expr): Likewise.
11709 (gimple_boolify): Likewise.
11710 (gimplify_modify_expr): Likewise.
11711 (gimplify_addr_expr): Likewise.
11712 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
11713 * ipa-cp.c (determine_versionability): Likewise.
11714 * ipa-fnsummary.c (compute_fn_summary): Likewise.
11715 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
11716 * ipa-split.c (visit_bb): Likewise.
11717 (split_function): Likewise.
11718 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
11719 * lto-cgraph.c (input_node): Likewise.
11720 * lto-streamer-out.c (write_symbol): Likewise.
11721 * omp-low.c (setjmp_or_longjmp_p): Likewise.
11722 (lower_omp_1): Likewise.
11723 * predict.c (strip_predict_hints): Likewise.
11724 * print-tree.c (print_node): Likewise.
11725 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
11726 * trans-mem.c (is_tm_irrevocable): Likewise.
11727 (is_tm_load): Likewise.
11728 (is_tm_simple_load): Likewise.
11729 (is_tm_store): Likewise.
11730 (is_tm_simple_store): Likewise.
11731 (is_tm_abort): Likewise.
11732 (tm_region_init_1): Likewise.
11733 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
11734 * tree-cfg.c (verify_gimple_call): Likewise.
11735 (move_stmt_r): Likewise.
11736 (stmt_can_terminate_bb_p): Likewise.
11737 * tree-eh.c (lower_eh_constructs_2): Likewise.
11738 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
11739 * tree-inline.c (remap_gimple_stmt): Likewise.
11740 (copy_bb): Likewise.
11741 (estimate_num_insns): Likewise.
11742 (fold_marked_statements): Likewise.
11743 * tree-sra.c (scan_function): Likewise.
11744 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
11745 (optimize_stack_restore): Likewise.
11746 (pass_fold_builtins::execute): Likewise.
11747 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
11748 (mark_all_reaching_defs_necessary_1): Likewise.
11749 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
11750 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
11751 (pass_forwprop::execute): Likewise.
11752 * tree-ssa-loop-im.c (stmt_cost): Likewise.
11753 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
11754 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
11755 * tree-ssa-strlen.c (get_string_length): Likewise.
11756 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
11757 (find_func_aliases_for_call): Likewise.
11758 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
11759 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
11760 * tree-tailcall.c (find_tail_calls): Likewise.
11761 * tree.c (need_assembler_name_p): Likewise.
11762 (free_lang_data_in_decl): Likewise.
11763 (get_call_combined_fn): Likewise.
11764 * ubsan.c (is_ubsan_builtin_p): Likewise.
11765 * varasm.c (incorporeal_function_p): Likewise.
11766 * tree.h (DECL_BUILT_IN): Remove and replace with
11768 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
11769 (fndecl_built_in_p): New.
11771 2018-08-27 Martin Liska <mliska@suse.cz>
11773 PR tree-optimization/86847
11774 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
11775 Dump also subtree probability.
11776 (switch_decision_tree::do_jump_if_equal): New function.
11777 (switch_decision_tree::emit_case_nodes): Handle special
11778 situations in balanced tree that can be emitted much simpler.
11779 Fix calculation of probabilities that happen in tree expansion.
11780 * tree-switch-conversion.h (struct cluster): Add
11782 (struct simple_cluster): Likewise.
11783 (struct case_tree_node): Add new function has_child.
11784 (do_jump_if_equal): New.
11786 2018-08-27 Martin Liska <mliska@suse.cz>
11788 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
11789 Add new argument to bit_test_cluster constructor.
11790 (bit_test_cluster::emit): Set bits really number of values
11792 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
11793 probability argument.
11794 * tree-switch-conversion.h (struct bit_test_cluster):
11795 Add m_handles_entire_switch.
11797 2018-08-27 Martin Liska <mliska@suse.cz>
11799 PR tree-optimization/86702
11800 * tree-switch-conversion.c (jump_table_cluster::emit):
11801 Make probabilities even for values in jump table
11802 according to number of cases handled.
11803 (switch_decision_tree::compute_cases_per_edge): Pass
11804 argument to reset_out_edges_aux function.
11805 (switch_decision_tree::analyze_switch_statement): Likewise.
11806 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
11809 2018-08-27 Martin Liska <mliska@suse.cz>
11811 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
11812 cfun argument explicitly.
11813 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
11814 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
11815 function gimple_switch_default_bb.
11816 (convert_switch_statements):
11818 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
11819 * stmt.c (label_to_block_fn): Use label_to_block and pass
11820 cfun argument explicitly and use gimple_switch_label_bb.
11821 (expand_case): Likewise.
11822 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
11823 cfun argument explicitly. Likewise.
11824 (make_edges_bb): Likewise.
11825 (make_cond_expr_edges): Likewise.
11826 (get_cases_for_edge): Likewise.
11827 (make_gimple_switch_edges): Likewise.
11828 (label_to_block_fn): Likewise.
11829 (label_to_block): Likewise.
11830 (make_goto_expr_edges): Likewise.
11831 (make_gimple_asm_edges): Likewise.
11832 (main_block_label): Likewise.
11833 (group_case_labels_stmt): Likewise.
11834 (find_taken_edge_computed_goto): Likewise.
11835 (find_taken_edge_switch_expr): Likewise.
11836 (gimple_verify_flow_info): Likewise.
11837 (gimple_redirect_edge_and_branch): Likewise.
11838 (gimple_switch_label_bb): New function.
11839 (gimple_switch_default_bb): Likewise.
11840 (gimple_switch_edge): Likewise.
11841 (gimple_switch_default_edge): Likewise.
11842 * tree-cfg.h (label_to_block_fn): Remove and replace ...
11843 (label_to_block): ... with this.
11844 (gimple_switch_label_bb): New.
11845 (gimple_switch_default_bb): Likewise.
11846 (gimple_switch_edge): Likewise.
11847 (gimple_switch_default_edge): Likewise.
11848 * tree-cfgcleanup.c (convert_single_case_switch): Use
11849 new gimple functions and pass new argument to label_to_block.
11850 (cleanup_control_flow_bb):
11851 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
11852 cfun argument explicitly.
11853 (make_eh_edges): Likewise.
11854 (redirect_eh_dispatch_edge): Likewise.
11855 (lower_resx): Likewise.
11856 (lower_eh_dispatch): Likewise.
11857 (maybe_remove_unreachable_handlers): Likewise.
11858 (unsplit_eh): Likewise.
11859 (cleanup_empty_eh): Likewise.
11860 (verify_eh_edges): Likewise.
11861 (verify_eh_dispatch_edge): Likewise.
11862 * tree-ssa-dom.c (record_edge_info): Likewise.
11863 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
11864 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
11865 (thread_through_normal_block): Likewise.
11866 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
11867 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
11868 * tree-switch-conversion.c (switch_conversion::collect): Use new
11870 (switch_conversion::check_final_bb): Likewise.
11871 (switch_conversion::gather_default_values): Pass new argument
11873 (switch_conversion::build_constructors): Likewise.
11874 (switch_decision_tree::compute_cases_per_edge): Use new
11875 gimple_switch_edge function.
11876 (switch_decision_tree::analyze_switch_statement): Pass new argument
11878 (switch_decision_tree::try_switch_expansion): Use
11879 gimple_switch_default_edge.
11880 * tree-vrp.c (find_switch_asserts): Pass new argument
11882 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
11883 (vr_values::simplify_switch_using_ranges): Likewise.
11885 2018-08-27 Richard Biener <rguenther@suse.de>
11887 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
11888 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
11890 * tree-ssa-sccvn.h (struct vn_pval): New structure.
11891 (struct vn_nary_op_s): Add unwind_to member. Add
11892 predicated_values flag and put result into a union together
11893 with a linked list of vn_pval.
11894 (struct vn_ssa_aux): Add name member to make maintaining
11895 a map of SSA name to vn_ssa_aux possible. Remove no longer
11896 needed info, dfsnum, low, visited, on_sccstack, use_processed
11897 and range_info_anti_range_p members.
11898 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
11899 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
11901 (vn_valueize): New global.
11902 (vn_context_bb): Likewise.
11903 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
11904 VN_INFO_PTR_INFO): Remove.
11905 * tree-ssa-sccvn.c: ... (rewrite)
11906 (pass_fre::execute): For -O2+ initialize loops and run
11907 RPO VN in optimistic mode (iterating). For -O1 and -Og
11908 run RPO VN in non-optimistic mode.
11909 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
11910 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
11911 * doc/invoke.texi (sccvn-max-scc-size): Remove.
11912 (rpo-vn-max-loop-depth): Document.
11913 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
11914 when valuezing the VUSE signals we walked out of the region.
11915 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
11916 (phi_translate): Set VN context block to use for availability
11918 (compute_avail): Likewise.
11919 (pre_valueize): New function.
11920 (pass_pre::execute): Adjust to the RPO VN API.
11922 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
11923 (propagate_constants_for_unrolling): Remove.
11924 (tree_unroll_loops_completely): Perform value-numbering
11925 on the unrolled bodies loop parent.
11927 2018-08-27 Richard Biener <rguenther@suse.de>
11929 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
11930 for partial antic compute.
11932 2018-08-27 Jakub Jelinek <jakub@redhat.com>
11934 PR rtl-optimization/87065
11935 * combine.c (simplify_if_then_else): Formatting fix.
11936 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
11938 (known_cond): Don't return const_true_rtx for vector modes. Use
11939 CONST0_RTX instead of const0_rtx. Formatting fixes.
11941 2018-08-27 Martin Liska <mliska@suse.cz>
11943 PR gcov-profile/87069
11944 * gcov.c (process_file): Record files already processed
11945 and warn about a file being processed multiple times.
11947 2018-08-27 Martin Liska <mliska@suse.cz>
11950 * config/aarch64/aarch64.c (aarch64_override_options_internal):
11951 Set default values for x_aarch64_*_string strings.
11952 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
11953 prefix. For -mabi do not print '=ABI' in help and use
11954 <option_value> format for -msve-vector-bits and -moverride
11957 2018-08-26 Jeff Law <law@redhat.com>
11959 * config/mips/frame-header-opt.c: Include "backend.h" rather than
11962 2018-08-26 Marek Polacek <polacek@redhat.com>
11964 PR c++/87029, Implement -Wredundant-move.
11965 * doc/invoke.texi: Document -Wredundant-move.
11967 2018-08-25 Martin Sebor <msebor@redhat.com>
11969 PR tree-optimization/87059
11970 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
11971 to the same type as the other.
11972 * fold-const.c (fold_binary_loc): Assert expectation.
11974 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
11976 * config/darwin.c (machopic_legitimize_pic_address): Clean up
11977 extraneous parentheses, dead code section and formatting.
11979 2018-08-24 David Malcolm <dmalcolm@redhat.com>
11982 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
11983 wide enough for jumps in the line-numbering to be visible.
11984 (layout::print_gap_in_line_numbering): New member function.
11985 (layout::calculate_line_spans): When using line numbering, merge
11986 line spans that are only 1 line apart.
11987 (diagnostic_show_locus): When printing line numbers, show gaps in
11988 line numbering directly, rather than printing headers.
11989 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
11990 line-numbering with multiple line spans.
11991 (selftest::test_fixit_insert_containing_newline_2): Add test of
11992 line-numbering, in which the spans are close enough to be merged.
11994 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
11996 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
11997 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
11998 * tree-vrp.c (range_is_nonnull): Remove.
11999 (range_includes_zero_p): Accept value_range instead of min/max.
12000 (extract_range_from_binary_expr_1): Do not early bail on
12002 Use range_includes_zero_p instead of range_is_nonnull.
12003 (extract_range_from_unary_expr): Use range_includes_zero_p instead
12004 of range_is_nonnull.
12005 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
12006 special case VR_ANTI_RANGE.
12007 (vrp_finalize): Same.
12008 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
12009 instead of min/max.
12010 (range_is_nonnull): Remove.
12011 * vr-values.c (vrp_stmt_computes_nonzero): Use
12012 range_includes_zero_p instead of range_is_nonnull.
12013 (extract_range_basic): Pass value_range to range_includes_zero_p
12014 instead of range_is_nonnull.
12016 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
12018 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
12019 * emit-rtl.h (rtl_data): Remove return_bnd.
12020 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
12021 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
12022 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
12023 (POINTER_BOUNDS_MODE): Remove definition.
12024 (make_pointer_bounds_mode): Remove.
12025 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
12026 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
12027 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
12028 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
12029 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
12030 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
12031 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
12033 * config/i386/i386-modes.def (BND32, BND64): Remove.
12034 * config/i386/i386.c (dbx_register_map): Remove bound registers.
12035 (dbx64_register_map): Ditto.
12036 (svr4_dbx_register_map): Ditto.
12037 (indirect_thunk_bnd_needed): Remove.
12038 (indirect_thunks_bnd_used): Ditto.
12039 (indirect_return_bnd_needed): Ditto.
12040 (indirect_return_via_cx_bnd): Ditto.
12041 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
12042 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
12043 (output_indirect_thunk): Ditto. Remove need_prefix argument.
12044 (output_indirect_thunk_function): Remove handling of
12045 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
12046 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
12047 (ix86_save_reg): Remove handling of crtl->return_bnd.
12048 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
12049 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
12050 and UNSPEC_BNDLX_ADDR.
12051 (ix86_output_indirect_branch_via_reg): Remove handling of
12052 indirect_thunk_prefix_bnd.
12053 (ix86_output_indirect_branch_via_push): Ditto.
12054 (ix86_output_function_return): Ditto.
12055 (ix86_output_indirect_function_return): Ditto.
12056 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
12057 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
12058 (CALL_USED_REGISTERS): Ditto.
12059 (REG_ALLOC_ORDER): Update for removal of bound registers.
12060 (HI_REGISTER_NAMES): Ditto.
12061 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
12062 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
12063 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
12064 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
12065 (FIRST_PSEUDO_REG): Update.
12066 (BND): Remove mode iterator.
12067 * config/i386/predicates.md (bnd_mem_operator): Remove.
12069 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
12071 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
12074 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
12076 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
12077 the case in which the permute needs only a single element and
12078 repeats for every vector of the result. Extend that case to
12079 handle variable-length vectors.
12080 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
12082 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
12085 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
12086 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
12088 2018-08-24 Richard Biener <rguenther@suse.de>
12090 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
12091 bb_flags_allocated members.
12092 (auto_flag): New RAII class for allocating flags.
12093 (auto_edge_flag): New RAII class for allocating edge flags.
12094 (auto_bb_flag): New RAII class for allocating bb flags.
12095 * cfgloop.c (verify_loop_structure): Allocate temporary edge
12097 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
12098 in favor of temporarily allocated BB flag.
12099 * hsa-brig.c: Re-order includes.
12100 * hsa-dump.c: Likewise.
12101 * hsa-regalloc.c: Likewise.
12102 * print-rtl.c: Likewise.
12103 * profile-count.c: Likewise.
12105 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
12108 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
12111 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
12114 * wide-int-range.cc (wide_int_range_div): Do not ignore result
12115 from wide_int_range_multiplicative_op.
12117 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12119 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
12122 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
12124 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
12127 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
12129 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
12130 full GENERIC expression used for replacement.
12132 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
12134 * tree-vrp.c (abs_extent_range): Remove.
12135 (extract_range_into_wide_ints): Pass wide ints by reference.
12136 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
12137 Pass wide ints by reference in all calls to
12138 extract_range_into_wide_ints.
12139 * wide-int-range.cc (wide_int_range_div): New.
12140 * wide-int-range.h (wide_int_range_div): New.
12141 (wide_int_range_includes_zero_p): New.
12142 (wide_int_range_zero_p): New.
12144 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
12146 * config/aarch64/aarch64.md (arches): New enum.
12147 (arch): New enum attr.
12148 (arch_enabled): New attr.
12149 (enabled): Now uses arch_enabled only.
12150 (simd, sve, fp16): Removed attribute.
12151 (fp): Attr now defined in terms of 'arch'.
12152 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
12153 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
12154 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
12155 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
12156 attributes into 'arch'.
12157 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
12158 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
12159 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
12160 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
12161 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
12162 'simd' attribute into 'arch'.
12163 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
12164 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
12165 Convert use of 'fp' attribute to 'arch'.
12166 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
12167 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
12169 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
12170 (different modes) Merge 'fp' and 'simd' into 'arch'.
12171 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
12172 'simd' into 'arch'.
12174 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
12176 PR rtl-optimization/87026
12177 * expmed.c (canonicalize_comparison): If we can no longer create
12178 pseudoregisters, don't.
12180 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
12183 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
12185 * config/arm/arm.c (speculation_barrier_libfunc): New static
12187 (arm_init_libfuncs): Initialize it.
12188 (arm_emit_speculation_barrier): New function.
12189 * config/arm/arm.md (speculation_barrier): Call
12190 arm_emit_speculation_barrier for architectures that do not have
12192 (speculation_barrier_insn): Only match on Armv7 or later.
12194 2018-08-23 Richard Biener <rguenther@suse.de>
12196 PR middle-end/87024
12197 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
12200 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
12202 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
12203 of single-vector TBLs.
12204 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
12205 one input is given.
12207 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
12210 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
12211 aarch64_evpc_tbl guard.
12213 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
12215 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
12218 2018-08-22 Martin Sebor <msebor@redhat.com>
12220 PR middle-end/87052
12221 * tree-pretty-print.c (pretty_print_string): Add argument.
12222 (dump_generic_node): Call to pretty_print_string with string size.
12224 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
12226 PR rtl-optimization/86771
12227 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
12228 of two SETs into those two SETs, one to be placed at i2, if that SETs
12229 destination is modified between i2 and i3.
12231 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
12233 PR tree-optimization/86725
12234 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
12236 (vect_analyze_scalar_cycles_1): Check it.
12238 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
12240 PR tree-optimization/86725
12241 * tree-vect-loop.c (vect_is_simple_reduction): When treating
12242 an outer loop phi as a double reduction, make sure that the
12243 single user of the phi result is an inner loop phi.
12245 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
12247 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
12248 grouped stores with gaps to a strided group.
12250 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
12252 * tree-vect-stmts.c (get_group_load_store_type)
12253 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
12254 first statement in a group.
12256 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
12258 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
12259 the sequence used in gcc/gcc.c.
12261 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
12264 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
12267 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
12269 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
12270 Darwin10-specific unwinder-shim.
12271 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
12272 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
12273 New to cater for Darwin10 Rosetta.
12275 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
12277 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
12278 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
12280 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
12285 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
12286 (dwarf2out_switch_text_section): Generate a local label for the
12287 second function sub-section and apply it as the second FDE start
12289 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
12290 second sub-section start.
12292 2018-08-22 Richard Biener <rguenther@suse.de>
12294 PR tree-optimization/86988
12295 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
12297 2018-08-22 Richard Biener <rguenther@suse.de>
12299 PR tree-optimization/86945
12300 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
12302 2018-08-22 Alexandre Oliva <oliva@adacore.com>
12304 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
12305 a comment about how uses of r2 for .sdata2 come about.
12307 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
12309 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
12311 2018-08-21 Marek Polacek <polacek@redhat.com>
12313 PR c++/86981, Implement -Wpessimizing-move.
12314 * doc/invoke.texi: Document -Wpessimizing-move.
12316 2018-08-21 Jan Hubicka <jh@suse.cz>
12318 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
12319 * tree.h (is_redundant_typedef): Remove.
12320 * dwarf2out.c (is_redundant_typedef): Turn into static function.
12322 2018-08-21 Jan Hubicka <jh@suse.cz>
12324 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
12327 2018-08-21 Tamar Christina <tamar.christina@arm.com>
12329 * expmed.c (extract_low_bits): Reject invalid subregs early.
12331 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
12333 PR middle-end/86121
12334 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
12337 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
12339 * config/vxworks.h: Guard vxworks_asm_out_constructor and
12340 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
12341 * config/vxworks.c: Likewise.
12343 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
12345 * config/vxworks.c: Set targetm.have_ctors_dtors
12346 if HAVE_INITFINI_ARRAY_SUPPORT.
12347 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
12348 if HAVE_INITFINI_ARRAY_SUPPORT.
12350 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
12352 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
12353 default search path for VxWorks < 7.
12355 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
12357 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
12358 (get_format_string): Refer to c_getstr.
12360 2018-08-21 Tom de Vries <tdevries@suse.de>
12362 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
12363 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
12364 (debuginfo_early_stop): Declare.
12365 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
12366 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
12367 (debuginfo_early_stop): New function.
12368 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
12369 and debuginfo_early_stop.
12370 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
12371 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
12372 (general_init): Call debuginfo_early_init.
12373 (finalize): Call debuginfo_fini.
12374 (do_compile): Call debuginfo_init.
12375 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
12376 -fdump-early-debug.
12377 (@item -fdump-debug, @item -fdump-earlydebug): Add.
12379 2018-08-21 Tom de Vries <tdevries@suse.de>
12381 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
12382 flag_dump_noaddr and flag_dump_unnumbered.
12384 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
12386 * wide-int-range.cc (wide_int_range_abs): New.
12387 (wide_int_range_order_set): Rename from wide_int_range_min_max.
12388 * wide-int-range.h (wide_int_range_abs): New.
12389 (wide_int_range_min_max): New.
12390 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
12391 case to call wide_int_range_abs.
12392 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
12393 (extract_range_from_abs_expr): Delete.
12395 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
12398 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
12399 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
12402 2018-08-20 Nathan Sidwell <nathan@acm.org>
12403 Jeff Law <law@redhat.com>
12405 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
12406 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
12408 2018-08-20 David Malcolm <dmalcolm@redhat.com>
12411 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
12412 (decl_attributes): Likewise.
12413 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
12415 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
12416 * diagnostic-core.h (class auto_diagnostic_group): New class.
12417 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
12418 (diagnostic_report_diagnostic): Handle the first diagnostics within
12420 (emit_diagnostic): Add auto_diagnostic_group instance.
12421 (inform): Likewise.
12422 (inform_n): Likewise.
12423 (warning): Likewise.
12424 (warning_at): Likewise.
12425 (warning_n): Likewise.
12426 (pedwarn): Likewise.
12427 (permerror): Likewise.
12429 (error_n): Likewise.
12430 (error_at): Likewise.
12432 (fatal_error): Likewise.
12433 (internal_error): Likewise.
12434 (internal_error_no_backtrace): Likewise.
12435 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
12436 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
12437 * diagnostic.h (struct diagnostic_context): Add fields
12438 "diagnostic_group_nesting_depth",
12439 "diagnostic_group_emission_count", "begin_group_cb",
12441 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
12442 Add auto_diagnostic_group instance(s).
12443 (find_explicit_erroneous_behavior): Likewise.
12444 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
12445 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
12446 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
12447 (gimplify_va_arg_expr): Likewise.
12448 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
12449 (HSA_SORRY_AT): Likewise.
12450 * ipa-devirt.c (compare_virtual_tables): Likewise.
12451 (warn_odr): Likewise.
12452 * multiple_target.c (expand_target_clones): Likewise.
12453 * opts-common.c (cmdline_handle_error): Likewise.
12454 * reginfo.c (globalize_reg): Likewise.
12455 * substring-locations.c (format_warning_n_va): Likewise.
12456 * tree-inline.c (expand_call_inline): Likewise.
12457 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
12458 * tree-ssa-loop-niter.c
12459 (do_warn_aggressive_loop_optimizations): Likewise.
12460 * tree-ssa-uninit.c (warn_uninit): Likewise.
12461 * tree.c (warn_deprecated_use): Likewise.
12463 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
12466 * config/i386/i386.md (eh_return): Always update EH return
12467 address in word_mode.
12469 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
12471 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
12472 TARGET_SPLIT_COMPLEX_ARG is defined.
12474 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
12476 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
12478 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
12481 * expr.c (expand_assignment): Assert that bitpos is positive.
12482 (store_field): Likewise
12483 (expand_expr_real_1): Make sure that bitpos is positive.
12484 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
12487 2018-08-20 Nathan Sidwell <nathan@acm.org>
12489 * Makefile.in (CPP_ID_DATA_H): Delete.
12490 (CPP_INTERNAL_H): Don't add it.
12491 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
12492 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
12494 2018-08-20 Richard Biener <rguenther@suse.de>
12496 PR tree-optimization/78655
12497 * tree-vrp.c (extract_range_from_binary_expr_1): Make
12498 pointer + offset nonnull if either operand is nonnull work.
12500 2018-08-20 Tom de Vries <tdevries@suse.de>
12502 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
12503 unless the referenced die describes the added property using
12504 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
12505 Otherwise, add a DW_AT_location to the referenced die.
12507 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
12510 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
12511 register_operand when calling ix86_set_reg_reg_cost.
12512 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
12513 Set *total to 0 for operands that satisfy x86_64_immediate_operand
12514 predicate and to 1 otherwise.
12516 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
12518 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
12519 emit a diagnostic that it is not supported and reset the option.
12520 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
12521 supported and consume the option. (ASM_FINAL_SPEC): New.
12523 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
12525 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
12528 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
12530 C-SKY port: Documentation
12532 * doc/extend.texi (C-SKY Function Attributes): New section.
12533 * doc/invoke.texi (Option Summary): Add C-SKY options.
12534 (C-SKY Options): New section.
12535 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
12537 2018-08-17 Jojo <jijie_rong@c-sky.com>
12538 Huibin Wang <huibin_wang@c-sky.com>
12539 Sandra Loosemore <sandra@codesourcery.com>
12540 Chung-Lin Tang <cltang@codesourcery.com>
12542 C-SKY port: Backend implementation
12544 * config/csky/*: New.
12545 * common/config/csky/*: New.
12547 2018-08-17 Jojo <jijie_rong@c-sky.com>
12548 Huibin Wang <huibin_wang@c-sky.com>
12549 Sandra Loosemore <sandra@codesourcery.com>
12550 Chung-Lin Tang <cltang@codesourcery.com>
12551 Andrew Jenner <andrew@codesourcery.com>
12553 C-SKY port: Configury
12555 * config.gcc (csky-*-*): New.
12556 * configure.ac: Add csky to targets for dwarf2 debug_line support.
12557 * configure: Regenerated.
12559 2018-08-17 David Malcolm <dmalcolm@redhat.com>
12561 * dump-context.h: Include "dumpfile.h".
12562 (dump_context::dump_printf_va): Convert final param from va_list
12563 to va_list *. Convert from ATTRIBUTE_PRINTF to
12564 ATTRIBUTE_GCC_DUMP_PRINTF.
12565 (dump_context::dump_printf_loc_va): Likewise.
12566 * dumpfile.c: Include "stringpool.h".
12567 (make_item_for_dump_printf_va): Delete.
12568 (make_item_for_dump_printf): Delete.
12569 (class dump_pretty_printer): New class.
12570 (dump_pretty_printer::dump_pretty_printer): New ctor.
12571 (dump_pretty_printer::emit_items): New member function.
12572 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
12574 (dump_pretty_printer::emit_item): New member function.
12575 (dump_pretty_printer::stash_item): New member function.
12576 (dump_pretty_printer::format_decoder_cb): New member function.
12577 (dump_pretty_printer::decode_format): New member function.
12578 (dump_context::dump_printf_va): Reimplement in terms of
12579 dump_pretty_printer.
12580 (dump_context::dump_printf_loc_va): Convert final param from va_list
12582 (dump_context::begin_scope): Reimplement call to
12583 make_item_for_dump_printf.
12584 (dump_printf): Update for change to dump_printf_va.
12585 (dump_printf_loc): Likewise.
12586 (selftest::test_capture_of_dump_calls): Convert "stmt" from
12587 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
12588 with %T, %E, and %G.
12589 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
12590 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
12591 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
12592 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
12593 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
12594 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
12595 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
12596 within a dump_printf_loc call to "%wu".
12597 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
12598 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
12599 missing space after "=".
12600 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
12601 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
12602 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
12603 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
12604 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
12605 duplicate "vectorized" from message.
12607 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
12609 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
12610 polyNxK_t element's TYPE_STRING_FLAG.
12612 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
12614 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
12615 (they were unnamed before). Fix comments.
12617 2018-08-17 Nathan Sidwell <nathan@acm.org>
12619 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
12621 2018-08-17 Richard Biener <rguenther@suse.de>
12623 PR tree-optimization/86841
12624 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
12626 2018-08-17 Martin Liska <mliska@suse.cz>
12628 * common.opt: Remove Warn, Init and Report for options with
12629 Ignore/Deprecated flag. Warning is done automatically for
12631 * config/i386/i386.opt: Likewise.
12632 * config/ia64/ia64.opt: Likewise.
12633 * config/rs6000/rs6000.opt: Likewise.
12634 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
12635 Remove usage of flag_check_pointer_bounds.
12636 * lto-wrapper.c (merge_and_complain): Do not handle
12637 OPT_fcheck_pointer_bounds.
12638 (append_compiler_options): Likewise.
12639 * opt-functions.awk: Do not handle Deprecated.
12640 * optc-gen.awk: Check that Var, Report and Init are not
12641 used for an option with Ignore/Deprecated flag.
12642 * opts-common.c (decode_cmdline_option): Do not report
12644 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
12646 * opts.h (struct cl_option): Remove cl_deprecated flag.
12647 (CL_ERR_DEPRECATED): Remove error enum value.
12649 2018-08-17 Richard Biener <rguenther@suse.de>
12651 PR middle-end/86505
12652 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
12653 across a va-arg-pack using call adjust its return value accordingly.
12655 2018-08-16 Martin Sebor <msebor@redhat.com>
12657 PR tree-optimization/86853
12658 * gimple-ssa-sprintf.c (struct format_result): Rename member.
12659 (struct fmtresult): Add member and initialize it in ctors.
12660 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
12661 (format_string): Handle %S the same as %ls. Set MAYFAIL.
12662 (format_directive): Set POSUNDER4K when MAYFAIL is set.
12663 (parse_directive): Handle %C same as %c.
12664 (sprintf_dom_walker::compute_format_length): Adjust.
12665 (is_call_safe): Adjust.
12667 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
12669 * builtins.c (c_strlen): Add new parameter eltsize. Use it
12670 for determining how to count the elements.
12671 * builtins.h (c_strlen): Adjust prototype.
12672 * expr.c (string_constant): Add new parameter mem_size.
12673 Set *mem_size appropriately.
12674 * expr.h (string_constant): Adjust protoype.
12675 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
12676 * gimple-fold.h (get_range_strlen): Adjust prototype.
12677 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
12678 (format_string): Call get_string_length with eltsize.
12680 2018-08-16 David Malcolm <dmalcolm@redhat.com>
12682 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
12683 to emit the span, rather than setting it as the prefix.
12685 2018-08-16 David Malcolm <dmalcolm@redhat.com>
12687 * diagnostic-show-locus.c (layout::start_annotation_line): Add
12688 "margin_char" parameter, defaulting to space. Use it in place
12689 of pp_space for the initial part of the margin.
12690 (layout::print_leading_fixits): Use '+' when filling the margin
12691 of line-insertion fix-it hints.
12693 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
12695 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
12698 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
12700 * config/rs6000/altivec.md: Don't set length attribute to the default
12702 * config/rs6000/darwin.md: Ditto.
12703 * config/rs6000/dfp.md: Ditto.
12704 * config/rs6000/htm.md: Ditto.
12705 * config/rs6000/rs6000.md: Ditto.
12706 * config/rs6000/sync.md: Ditto.
12707 * config/rs6000/vsx.md: Ditto.
12709 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
12711 * config/rs6000/altivec.md: Don't set length attribute to the default
12712 value, for branch instructions.
12713 * config/rs6000/darwin.md: Ditto.
12714 * config/rs6000/rs6000.md: Ditto.
12716 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
12718 * config/rs6000/rs6000.md (length): Always define as const_int 4.
12719 (unnamed conditional branch define_insn): Set length to 4 or 8
12720 depending on offset.
12721 (<bd>_<mode>): Similar, for alternative 0.
12722 (<bd>tf_<mode>): Ditto.
12724 2018-08-16 Tamar Christina <tamar.christina@arm.com>
12726 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
12728 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
12730 * doc/rtl.texi: Replace old RTX class names with new names.
12733 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
12735 * expmed.h (canonicalize_comparison): New declaration.
12736 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
12737 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
12738 * optabs.c (prepare_cmp_insn): Likewise.
12739 * rtl.h (unsigned_condition_p): New function which checks if a
12740 comparison operator is unsigned.
12742 2018-08-16 Nathan Sidwell <nathan@acm.org>
12744 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
12745 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
12747 2018-08-16 Tamar Christina <tamar.christina@arm.com>
12750 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
12751 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
12752 (mov<mov>): ..this and enable unconditionally.
12754 2018-08-16 Tamar Christina <tamar.christina@arm.com>
12756 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
12758 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
12760 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
12761 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
12762 "Common" with "Target".
12764 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
12766 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
12767 * doc/invoke.texi (mmitigate-rop): Remove.
12768 * config/i386/i386.c: Do not include "regrename.h".
12769 (ix86_rop_should_change_byte_p, reg_encoded_number)
12770 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
12772 (ix86_reorg): Remove call to ix86_mitigate_rop.
12773 * config/i386/i386.md (attr "modrm_class"): Remove.
12774 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
12775 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
12776 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
12778 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
12780 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
12781 allow folding of mergeh() and mergel() for the float and double types.
12782 (fold_mergehl_helper): Rework to handle building a permute tree
12785 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
12787 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
12790 2018-08-15 David Malcolm <dmalcolm@redhat.com>
12792 * common.opt (fdiagnostics-show-labels): New option.
12793 * diagnostic-show-locus.c (class layout_range): Add field
12795 (class layout): Add field "m_show_labels_p".
12796 (layout_range::layout_range): Add param "label" and use it to
12797 initialize m_label.
12798 (make_range): Pass in NULL for new "label" param of layout_range's
12800 (layout::layout): Initialize m_show_labels_p.
12801 (layout::maybe_add_location_range): Pass in loc_range->m_label
12802 when constructing layout_range instances.
12803 (struct line_label): New struct.
12804 (layout::print_any_labels): New member function.
12805 (layout::print_line): Call it if label-printing is enabled.
12806 (selftest::test_one_liner_labels): New test.
12807 (selftest::test_diagnostic_show_locus_one_liner): Call it.
12808 * diagnostic.c (diagnostic_initialize): Initialize
12809 context->show_labels_p.
12810 * diagnostic.h (struct diagnostic_context): Add field
12812 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
12813 -fno-diagnostics-show-labels.
12814 * dwarf2out.c (gen_producer_string): Add
12815 OPT_fdiagnostics_show_labels to the ignored options.
12816 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
12818 (gcc_rich_location::maybe_add_expr): Likewise.
12819 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
12820 label" param, defaulting to NULL.
12821 (gcc_rich_location::add_expr): Add "label" param.
12822 (gcc_rich_location::maybe_add_expr): Likewise.
12823 (class text_range_label): New class.
12824 (class range_label_for_type_mismatch): New class.
12825 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
12826 of format_warning_va.
12827 (fmtwarn_n): Likewise for new params of format_warning_n_va.
12828 * lto-wrapper.c (merge_and_complain): Add
12829 OPT_fdiagnostics_show_labels to the "pick one setting" options.
12830 (append_compiler_options): Likewise to the dropped options.
12831 (append_diag_options): Likewise to the passed-on options.
12832 * opts.c (common_handle_option): Handle the new option.
12833 * selftest-diagnostic.c
12834 (test_diagnostic_context::test_diagnostic_context): Enable
12836 * substring-locations.c: Include "gcc-rich-location.h".
12837 (format_warning_n_va): Add "fmt_label" and "param_label" params
12838 and use them as appropriate.
12839 (format_warning_va): Add "fmt_label" and "param_label" params,
12840 passing them on to format_warning_n_va.
12841 (format_warning_at_substring): Likewise.
12842 (format_warning_at_substring_n): Likewise.
12843 * substring-locations.h (format_warning_va): Add "fmt_label" and
12844 "param_label" params.
12845 (format_warning_n_va): Likewise.
12846 (format_warning_at_substring): Likewise.
12847 (format_warning_at_substring_n): Likewise.
12848 * toplev.c (general_init): Initialize global_dc->show_labels_p.
12850 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
12853 * builtins.c (expand_builtin_memcmp): Do not expand the call
12854 when overflow is detected.
12856 2018-08-15 Martin Sebor <msebor@redhat.com>
12858 PR tree-optimization/71625
12859 * config/aarch64/aarch64-builtins.c
12860 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
12862 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
12864 * config/s390/s390.c (s390_reorg): Remove loop.
12866 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
12869 (darwin_function_switched_text_sections): Delete.
12870 * gcc/config/darwin.h
12871 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
12873 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
12876 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
12877 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
12878 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
12880 2018-08-15 Martin Liska <mliska@suse.cz>
12882 PR tree-optimization/86925
12883 * predict.c (expr_expected_value_1): When taking
12884 later predictor, assign also probability.
12885 Use fold_build2_initializer_loc in order to fold
12886 the expression in -frounding-math.
12888 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
12890 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
12892 (expand_vec_perm_1): Try the new method.
12894 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
12897 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
12898 Check whether lra_live_max_point is 0 before dividing.
12900 2018-08-14 Martin Sebor <msebor@redhat.com>
12902 PR tree-optimization/86650
12903 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
12904 (vrp_prop::check_mem_ref): Same.
12906 2018-08-13 Liu Hao <lh_mouse@126.com>
12908 * pretty-print.c (eat_esc_sequence): Swap the foreground and
12909 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
12910 and clear it thereafter, as it only works for DBCS.
12912 2018-08-13 Liu Hao <lh_mouse@126.com>
12914 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
12915 handle returned by _get_osf_handle().
12917 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
12919 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
12920 for folding vec_perm.
12922 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
12924 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
12925 Add support for gimple-folding of vec_pack() and vec_unpack()
12928 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
12930 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
12931 vec_xst variants to the list.
12932 (rs6000_gimple_fold_builtin): Add support for folding unaligned
12933 vector loads and stores.
12935 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
12937 * config.gcc (rs6000-ibm-aix4.x): Delete.
12938 (rs6000-ibm-aix5.1): Delete.
12939 (rs6000-ibm-aix5.2): Delete.
12940 (rs6000-ibm-aix5.3): Delete.
12941 * config/rs6000/aix43.h: Delete.
12942 * config/rs6000/aix51.h: Delete.
12943 * config/rs6000/aix52.h: Delete.
12944 * config/rs6000/t-aix43: Delete.
12946 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
12948 * config/s390/s390.c (s390_decompose_constant_pool_ref):
12950 (s390_decompose_address): Factor out constant pool ref
12953 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
12955 * config/nds32/nds32-predicates.c
12956 (nds32_can_use_bclr_p): Change return type as bool.
12957 (nds32_can_use_bset_p): Ditto.
12958 (nds32_can_use_btgl_p): Ditto.
12959 (nds32_can_use_bitci_p): Ditto.
12960 * config/nds32/nds32-protos.h
12961 (nds32_can_use_bclr_p): Change declaration.
12962 (nds32_can_use_bset_p): Ditto.
12963 (nds32_can_use_btgl_p): Ditto.
12964 (nds32_can_use_bitci_p): Ditto.
12966 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
12968 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
12969 Support -msched-prolog-epilog option.
12970 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
12972 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
12974 * common/config/nds32/nds32-common.c
12975 (nds32_option_optimization_table): Enalbe -malways-align.
12977 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
12979 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
12981 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
12982 OPT_misr_secure_ case.
12983 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
12984 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
12985 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
12987 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
12988 (nds32_isr_info): New field security_level.
12989 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
12990 * config/nds32/nds32.md (return_internal): Consider critical attribute.
12991 * config/nds32/nds32.opt (misr-secure): New option.
12992 * config/nds32/nds32_init.inc: New file.
12993 * config/nds32/nds32_isr.h: New file.
12995 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
12997 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
12998 Update comment for atomic instructions.
12999 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
13000 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
13002 (atomic_loaddi): Revise fence expansion to only emit fence prior to
13003 load for __ATOMIC_SEQ_CST model.
13004 (atomic_loaddi_1): Remove float register target.
13005 (atomic_storedi): Handle CONST_INT values.
13006 (atomic_storedi_1): Remove float register source. Add special case
13008 (memory_barrier): New expander and insn.
13010 2018-08-11 Jakub Jelinek <jakub@redhat.com>
13012 PR tree-optimization/86835
13013 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
13014 new_stmt after def_gsi, make sure to insert new_square_stmt after
13015 that stmt, not 2 stmts before it.
13017 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
13020 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
13023 2018-08-10 Martin Liska <mliska@suse.cz>
13026 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
13028 * builtins.c (expand_builtin_expect_with_probability):
13030 (expand_builtin_expect_with_probability): New function.
13031 (build_builtin_expect_predicate): Add new argumnet probability
13032 for BUILT_IN_EXPECT_WITH_PROBABILITY.
13033 (fold_builtin_expect):
13036 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
13037 * builtins.h (fold_builtin_expect): Set new argument.
13038 * doc/extend.texi: Document __builtin_expect_with_probability.
13039 * doc/invoke.texi: Likewise.
13040 * gimple-fold.c (gimple_fold_call): Pass new argument.
13041 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
13042 also BUILT_IN_EXPECT_WITH_PROBABILITY.
13043 * predict.c (get_predictor_value): New function.
13044 (expr_expected_value): Add new argument probability. Assume
13045 that predictor and probability are always non-null.
13046 (expr_expected_value_1): Likewise. For __builtin_expect and
13047 __builtin_expect_with_probability set probability. Handle
13048 combination in binary expressions.
13049 (tree_predict_by_opcode): Simplify code by simply calling
13050 get_predictor_value.
13051 (pass_strip_predict_hints::execute): Add handling of
13052 BUILT_IN_EXPECT_WITH_PROBABILITY.
13053 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
13055 * tree.h (DECL_BUILT_IN_P): New function.
13057 2018-08-10 Martin Liska <mliska@suse.cz>
13059 PR tree-optimization/85799
13060 * passes.def: Add argument for pass_strip_predict_hints.
13061 * predict.c (class pass_strip_predict_hints): Add new argument
13063 (strip_predictor_early): New function.
13064 (pass_strip_predict_hints::execute): Call the function to
13066 (strip_predict_hints): New function.
13067 * predict.def: Fix comment.
13069 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
13071 * Makefile.in: Clarify which tm.texi to copy over to assert the
13072 right to grant a GFDL license for all.
13074 2018-08-09 Jeff Law <law@redhat.com>
13076 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
13079 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
13081 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
13084 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
13086 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
13087 reductions for variable-length vectors.
13089 2018-08-09 David Malcolm <dmalcolm@redhat.com>
13092 * common.opt (fdiagnostics-show-line-numbers): New option.
13093 * diagnostic-show-locus.c (class layout): Add fields
13094 "m_show_line_numbers_p" and "m_linenum_width";
13095 (num_digits): New function.
13096 (test_num_digits): New function.
13097 (layout::layout): Initialize new fields. Update m_x_offset
13098 logic to handle any left margin.
13099 (layout::print_source_line): Print line number when requested.
13100 (layout::start_annotation_line): New member function.
13101 (layout::print_annotation_line): Call it.
13102 (layout::print_leading_fixits): Likewise.
13103 (layout::print_trailing_fixits): Likewise. Update calls to
13104 move_to_column for new parameter.
13105 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
13106 it to potentially call start_annotation_line.
13107 (layout::show_ruler): Call start_annotation_line.
13108 (selftest::test_line_numbers_multiline_range): New selftest.
13109 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
13110 and selftest::test_line_numbers_multiline_range.
13111 * diagnostic.c (diagnostic_initialize): Initialize
13112 show_line_numbers_p.
13113 * diagnostic.h (struct diagnostic_context): Add field
13114 "show_line_numbers_p".
13115 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
13116 -fno-diagnostics-show-line-numbers.
13117 * dwarf2out.c (gen_producer_string): Add
13118 OPT_fdiagnostics_show_line_numbers to the ignored options.
13119 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
13120 one setting" options.
13121 (append_compiler_options): Likewise to the dropped options.
13122 (append_diag_options): Likewise to the passed-on options.
13123 * opts.c (common_handle_option): Handle the new option.
13124 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
13126 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
13128 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
13129 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
13130 third argument of type "const signed char" to descriptions of
13131 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
13132 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
13133 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
13134 __builtin_bcdsub_ov functions.
13136 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
13138 PR tree-optimization/86858
13139 * tree-vect-loop.c (vect_is_simple_reduction): Restore
13140 flow_bb_inside_loop_p calls.
13142 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
13144 PR tree-optimization/86871
13145 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
13146 instead of gimple_assign_lhs.
13148 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
13151 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
13152 register constraint to operand 0.
13153 (add<mode>3_carryinC): Likewise.
13154 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
13156 2018-08-09 Martin Liska <mliska@suse.cz>
13159 * common.opt: Remove extra line.
13161 2018-08-09 Martin Liska <mliska@suse.cz>
13163 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
13164 at the end of a line, make first letter capital and end up
13165 a sentence with a dot.
13166 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
13167 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
13168 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
13169 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
13170 (PARAM_MAX_ISL_OPERATIONS): Likewise.
13171 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
13172 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
13173 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
13174 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
13175 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
13176 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
13177 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
13178 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
13179 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
13180 (PARAM_TREE_REASSOC_WIDTH): Likewise.
13181 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
13182 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
13183 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
13185 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
13188 * config/s390/s390.c (s390_option_override_internal): Reduce the
13189 stack-clash-protection-probe-interval param if it would be too big
13192 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
13195 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
13196 * config/m68k/m68k.c (handle_move_double): Don't call
13197 m68k_final_prescan_insn.
13198 (m68k_adjust_decorated_operand): Renamed from
13199 m68k_final_prescan_insn, remove first and third operand and
13201 (print_operand): Call it.
13202 (print_operand_address): Call it.
13204 2018-08-08 Nathan Sidwell <nathan@acm.org>
13206 * diagnostic.c (diagnostic_report_current_module): Use
13207 linemap_included_from & linemap_included_from_linemap.
13209 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
13211 * config/aarch64/aarch64-cores.def: Add phecda core.
13212 * config/aarch64/aarch64-tune.md: Regenerate.
13213 * doc/invoke.texi: Add phecda core.
13215 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
13218 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
13220 * config/s390/s390.md ("movti"): Add more alternatives for
13221 constant to GPR copies.
13223 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
13225 * config/s390/s390.c: Fix whitespace damage throughout the file.
13226 * config/s390/s390.h: Likewise.
13227 * config/s390/tpf.h: Likewise.
13229 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
13231 * config/s390/s390.c (s390_loadrelative_operand_p):
13232 Remove TARGET_CPU_ZARCH usages.
13233 (s390_rtx_costs): Likewise.
13234 (s390_legitimate_constant_p): Likewise.
13235 (s390_cannot_force_const_mem): Likewise.
13236 (legitimate_reload_constant_p): Likewise.
13237 (s390_preferred_reload_class): Likewise.
13238 (legitimize_pic_address): Likewise.
13239 (legitimize_tls_address): Likewise.
13240 (s390_split_branches): Removed.
13241 (s390_add_execute): Removed.
13242 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
13243 (s390_mainpool_start): Likewise.
13244 (s390_mainpool_finish): Likewise.
13245 (s390_mainpool_cancel): Removed.
13246 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
13247 (s390_chunkify_cancel): Likewise.
13248 (s390_return_addr_rtx): Likewise.
13249 (s390_register_info): Remove split_branches_pending_p uages.
13250 (s390_optimize_register_info): Likewise.
13251 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
13252 split_branches_pending_p usages.
13253 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
13254 (s390_load_got): Likewise.
13255 (s390_expand_split_stack_prologue): Likewise.
13256 (output_asm_nops): Likewise.
13257 (s390_function_profiler): Likewise.
13258 (s390_emit_call): Likewise.
13259 (s390_conditional_register_usage): Likewise.
13260 (s390_optimize_prologue): Likewise.
13261 (s390_reorg): Remove TARGET_CPU_ZARCH and
13262 split_branches_pending_p usages.
13263 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
13265 (s390_output_indirect_thunk_function): Likewise.
13266 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
13267 (TARGET_CPU_ZARCH_P): Removed.
13268 (struct machine_function): Remove split_branches_pending_p.
13269 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
13271 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
13273 * common/config/s390/s390-common.c (processor_flags_table):
13275 * config.gcc: Remove with_arch/with_tune support.
13276 * config/s390/2064.md: Remove cpu attribute comparisons.
13277 * config/s390/driver-native.c (s390_host_detect_local_cpu):
13279 * config/s390/linux.h (ASM_SPEC):
13280 Remove -march support.
13281 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
13282 Use a table to get an arch level.
13283 * config/s390/s390-opts.h (enum processor_type):
13284 Remove enum values.
13285 * config/s390/s390.c
13286 (processor_table): Remove entries, add arch_level values.
13287 (s390_issue_rate): Remove cases.
13288 (s390_option_override): Adjust
13289 s390_option_override_internal() call.
13290 (s390_option_override_internal): Remove deprecation warning.
13291 (s390_valid_target_attribute_tree): Adjust
13292 s390_option_override_internal() call.
13293 * config/s390/s390.h (struct s390_processor):
13294 Share with s390-c.c, add arch_level field.
13295 * config/s390/s390.md:
13296 Remove occurrences in cpu attribute.
13297 * config/s390/s390.opt: Remove -march/-mtune support.
13298 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
13299 * doc/invoke.texi: Remove deprecation warning.
13301 2018-08-08 Luis Machado <luis.machado@linaro.org>
13303 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
13305 (qdf24xx_tunings): Set vector cost structure to
13306 qdf24xx_vector_cost.
13308 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
13309 <register_sextend>: Set to 3.
13311 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
13314 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
13315 * config/aarch64/aarch64-simd.md
13316 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
13317 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
13318 (aarch64_frecpx<mode>): ...this new pattern.
13319 * config/aarch64/aarch64-simd-builtins.def: Remove comment
13320 about aarch64_frecp<FRECP:frecp_suffix><mode>.
13322 2018-08-07 Martin Liska <mliska@suse.cz>
13324 PR middle-end/83023
13325 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
13326 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
13327 * predict.def (PRED_MALLOC_NONNULL): New predictor.
13328 * doc/extend.texi: Document that malloc attribute adds
13331 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
13334 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13335 Define to speculation_safe_value_not_needed.
13337 2018-08-06 Jeff Law <law@redhat.com>
13339 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
13340 the vr_values instance to cprop_into_stmt.
13341 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
13342 (cprop_operand): Also query EVRP to determine if OP is a constant.
13344 2018-08-06 Nathan Sidwell <nathan@acm.org>
13346 * diagnostic.c (diagnostic_report_current_module): Reroll
13347 included-at loop. Translate text.
13349 2018-08-06 David Malcolm <dmalcolm@redhat.com>
13351 * function-tests.c (selftest::test_expansion_to_rtl): Call
13352 free_after_compilation.
13354 2018-08-06 Alan Hayward <alan.hayward@arm.com>
13356 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
13358 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
13360 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
13361 loops with memory block operations from getting unrolled.
13363 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
13366 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13367 Define to speculation_safe_value_not_needed.
13369 2018-08-06 Jeff Law <law@redhat.com>
13371 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
13374 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
13377 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
13378 with all enabled __intN types.
13380 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
13382 2018-08-06 Alan Hayward <alan.hayward@arm.com>
13384 * alias.c (record_set): Check for clobber high.
13385 * cfgexpand.c (expand_gimple_stmt): Likewise.
13386 * combine-stack-adj.c (single_set_for_csa): Likewise.
13387 * combine.c (find_single_use_1): Likewise.
13388 (set_nonzero_bits_and_sign_copies): Likewise.
13389 (get_combine_src_dest): Likewise.
13390 (is_parallel_of_n_reg_sets): Likewise.
13391 (try_combine): Likewise.
13392 (record_dead_and_set_regs_1): Likewise.
13393 (reg_dead_at_p_1): Likewise.
13394 (reg_dead_at_p): Likewise.
13395 * dce.c (deletable_insn_p): Likewise.
13396 (mark_nonreg_stores_1): Likewise.
13397 (mark_nonreg_stores_2): Likewise.
13398 * df-scan.c (df_find_hard_reg_defs): Likewise.
13399 (df_uses_record): Likewise.
13400 (df_get_call_refs): Likewise.
13401 * dwarf2out.c (mem_loc_descriptor): Likewise.
13402 * haifa-sched.c (haifa_classify_rtx): Likewise.
13403 * ira-build.c (create_insn_allocnos): Likewise.
13404 * ira-costs.c (scan_one_insn): Likewise.
13405 * ira.c (equiv_init_movable_p): Likewise.
13406 (rtx_moveable_p): Likewise.
13407 (interesting_dest_for_shprep): Likewise.
13408 * jump.c (mark_jump_label_1): Likewise.
13409 * postreload-gcse.c (record_opr_changes): Likewise.
13410 * postreload.c (reload_cse_simplify): Likewise.
13411 (struct reg_use): Add source expr.
13412 (reload_combine): Check for clobber high.
13413 (reload_combine_note_use): Likewise.
13414 (reload_cse_move2add): Likewise.
13415 (move2add_note_store): Likewise.
13416 * print-rtl.c (print_pattern): Likewise.
13417 * recog.c (decode_asm_operands): Likewise.
13418 (store_data_bypass_p): Likewise.
13419 (if_test_bypass_p): Likewise.
13420 * regcprop.c (kill_clobbered_value): Likewise.
13421 (kill_set_value): Likewise.
13422 * reginfo.c (reg_scan_mark_refs): Likewise.
13423 * reload1.c (maybe_fix_stack_asms): Likewise.
13424 (eliminate_regs_1): Likewise.
13425 (elimination_effects): Likewise.
13426 (mark_not_eliminable): Likewise.
13427 (scan_paradoxical_subregs): Likewise.
13428 (forget_old_reloads_1): Likewise.
13429 * reorg.c (find_end_label): Likewise.
13430 (try_merge_delay_insns): Likewise.
13431 (redundant_insn): Likewise.
13432 (own_thread_p): Likewise.
13433 (fill_simple_delay_slots): Likewise.
13434 (fill_slots_from_thread): Likewise.
13435 (dbr_schedule): Likewise.
13436 * resource.c (update_live_status): Likewise.
13437 (mark_referenced_resources): Likewise.
13438 (mark_set_resources): Likewise.
13439 * rtl.c (copy_rtx): Likewise.
13440 * rtlanal.c (reg_referenced_p): Likewise.
13441 (single_set_2): Likewise.
13442 (noop_move_p): Likewise.
13443 (note_stores): Likewise.
13444 * sched-deps.c (sched_analyze_reg): Likewise.
13445 (sched_analyze_insn): Likewise.
13447 2018-08-06 Alan Hayward <alan.hayward@arm.com>
13449 * cse.c (invalidate_reg): New function extracted from...
13450 (invalidate): ...here.
13451 (canonicalize_insn): Check for clobber high.
13452 (invalidate_from_clobbers): invalidate clobber highs.
13453 (invalidate_from_sets_and_clobbers): Likewise.
13454 (count_reg_usage): Check for clobber high.
13455 (insn_live_p): Likewise.
13456 * cselib.c (cselib_expand_value_rtx_1):Likewise.
13457 (cselib_invalidate_regno): Check for clobber in setter.
13458 (cselib_invalidate_rtx): Pass through setter.
13459 (cselib_invalidate_rtx_note_stores):
13460 (cselib_process_insn): Check for clobber high.
13461 * cselib.h (cselib_invalidate_rtx): Add operand.
13463 2018-08-06 Alan Hayward <alan.hayward@arm.com>
13465 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
13466 (mark_not_eliminable): Likewise.
13467 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
13468 * lra-lives.c (process_bb_lives): Check for clobber high.
13469 * lra.c (new_insn_reg): Remember clobber highs.
13470 (collect_non_operand_hard_regs): Check for clobber high.
13471 (lra_set_insn_recog_data): Likewise.
13472 (add_regs_to_insn_regno_info): Likewise.
13473 (lra_update_insn_regno_info): Likewise.
13475 2018-08-06 Alan Hayward <alan.hayward@arm.com>
13477 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
13478 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
13480 2018-08-06 Alan Hayward <alan.hayward@arm.com>
13482 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
13483 (copy_insn_1): Likewise.
13484 (gen_hard_reg_clobber_high): New gen function.
13485 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
13486 * genemit.c (gen_exp): Likewise.
13487 (gen_emit_seq): Pass through info.
13488 (gen_insn): Check for CLOBBER_HIGH.
13489 (gen_expand): Pass through info.
13490 (gen_split): Likewise.
13491 (output_add_clobbers): Likewise.
13492 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
13493 (remove_clobbers): Likewise.
13494 * rtl.h (gen_hard_reg_clobber_high): New declaration.
13496 2018-08-06 Alan Hayward <alan.hayward@arm.com>
13498 * doc/rtl.texi (clobber_high): Add.
13499 (parallel): Add in clobber high
13500 * rtl.c (rtl_check_failed_code3): Add function.
13501 * rtl.def (CLOBBER_HIGH): Add expression.
13502 * rtl.h (RTL_CHECKC3): Add macro.
13503 (rtl_check_failed_code3): Add declaration.
13504 (XC3EXP): Add macro.
13506 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
13509 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
13510 cfun->machine->max_used_stack_alignment if needed.
13512 2018-08-04 Martin Sebor <msebor@redhat.com>
13514 PR tree-optimization/86571
13515 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
13518 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
13521 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13524 2018-08-03 Jeff Law <law@redhat.com>
13527 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13528 Define to speculation_safe_value_not_needed.
13530 2018-08-03 David Malcolm <dmalcolm@redhat.com>
13532 * doc/gcov.texi (-x): Remove duplicate "to".
13533 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
13534 (-Wif-not-aligned): Remove duplicate "is".
13535 (-flto): Remove duplicate "the".
13536 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
13537 duplicate "v5.00.b".
13538 (MSP430 Options): Remove duplicate "and" from the description
13539 of "-mgprel-sec=regexp".
13540 (x86 Options): Remove duplicate copies of "vmldLog102" and
13541 vmlsLog104 from description of "-mveclibabi=type".
13543 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
13545 * internal-fn.h (first_commutative_argument): Declare.
13546 * internal-fn.c (first_commutative_argument): New function.
13547 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
13548 restrictions for pattern statements. Use first_commutative_argument
13549 to look for commutative operands in calls to internal functions.
13551 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
13553 * Makefile.in (wide-int-range.o): New.
13554 * tree-vrp.c: Move all the wide_int_* functions to...
13555 * wide-int-range.cc: ...here.
13556 * tree-vrp.h: Move all the wide_int_* prototypes to...
13557 * wide-int-range.h: ...here.
13559 2018-08-03 Tom de Vries <tdevries@suse.de>
13561 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
13563 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
13564 * except.c (output_function_exception_table): Do early exit if
13565 targetm_common.except_unwind_info (&global_options) == UI_NONE.
13567 2018-08-03 Martin Liska <mliska@suse.cz>
13569 * predict.c (dump_prediction): Change to 2 digits
13572 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
13574 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
13576 (falkor_am_1_gtov_gtov): ... a new insn reservation.
13578 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
13580 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
13581 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
13582 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
13583 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
13584 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
13586 2018-08-02 David Malcolm <dmalcolm@redhat.com>
13588 * diagnostic-show-locus.c (diagnostic_show_locus): Use
13589 pp_take_prefix when saving the existing prefix.
13590 * diagnostic.c (diagnostic_append_note): Likewise.
13591 * langhooks.c (lhd_print_error_function): Likewise.
13592 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
13593 param's type. Free the existing prefix.
13594 (pp_take_prefix): New function.
13595 (pretty_printer::pretty_printer): Drop the prefix parameter.
13596 Rename the length parameter to match the comment.
13597 (pretty_printer::~pretty_printer): Free the prefix.
13598 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
13600 (struct pretty_printer): Drop the "const" from "prefix" field's
13601 type and clarify memory management.
13602 (pp_set_prefix): Drop the "const" from the 2nd param.
13603 (pp_take_prefix): New decl.
13605 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
13607 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
13608 for word_mode_ok here instead of passing as argument.
13609 (expand_block_compare): Change select_block_compare_mode() call.
13610 (expand_strncmp_gpr_sequence): New function.
13611 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
13613 2018-08-02 Jeff Law <law@redhat.com>
13616 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13617 Define to speculation_safe_value_not_needed.
13620 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13621 Define to speculation_safe_value_not_needed.
13623 2018-08-02 Tom de Vries <tdevries@suse.de>
13626 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
13627 function. Return UI_TARGET unconditionally.
13628 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
13629 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
13631 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
13633 * genemit.c (print_overload_test): Fix typo.
13635 2018-08-02 Richard Biener <rguenther@suse.de>
13637 PR tree-optimization/86816
13638 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
13639 which checks for value availability before querying it.
13640 (gvn_uses_equal): Use it.
13641 (same_succ_hash): Likewise.
13642 (gimple_equal_p): Likewise.
13644 2018-08-02 Nick Clifton <nickc@redhat.com>
13647 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13648 Define to speculation_safe_value_not_needed.
13651 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13652 Define to speculation_safe_value_not_needed.
13655 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13656 Define to speculation_safe_value_not_needed.
13659 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13660 Define to speculation_safe_value_not_needed.
13663 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13664 Define to speculation_safe_value_not_needed.
13667 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13668 Define to speculation_safe_value_not_needed.
13671 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13672 Define to speculation_safe_value_not_needed.
13675 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
13676 Define to speculation_safe_value_not_needed.
13679 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
13680 speculation_safe_value_not_needed.
13683 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
13684 to speculation_safe_value_not_needed.
13686 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
13688 * doc/md.texi: Expand the documentation of instruction names
13689 to mention port-local uses. Document '@' in pattern names.
13690 * read-md.h (overloaded_instance, overloaded_name): New structs.
13691 (mapping): Declare.
13692 (md_reader::handle_overloaded_name): New member function.
13693 (md_reader::get_overloads): Likewise.
13694 (md_reader::m_first_overload): New member variable.
13695 (md_reader::m_next_overload_ptr): Likewise.
13696 (md_reader::m_overloads_htab): Likewise.
13697 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
13698 m_next_overload_ptr and m_overloads_htab.
13699 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
13700 (get_mode_token, get_code_token, get_int_token): New functions.
13701 (map_attr_string): Add an optional argument that passes back
13702 the associated iterator.
13703 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
13704 (md_reader::handle_overloaded_name, add_overload_instance): New
13706 (apply_iterators): Handle '@' names. Report an error if '@'
13707 is used without iterators.
13708 (initialize_iterators): Initialize the new iterator_group fields.
13709 * genopinit.c (handle_overloaded_code_for)
13710 (handle_overloaded_gen): New functions.
13711 (main): Use them to print declarations of maybe_code_for_* and
13712 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
13713 * genemit.c (print_overload_arguments, print_overload_test)
13714 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
13715 (main): Use it to print definitions of maybe_code_for_* and
13716 maybe_gen_* functions.
13717 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
13718 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
13719 instead of explicit mode checks.
13720 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
13721 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
13722 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
13723 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
13724 (aarch64_expand_compare_and_swap): Likewise
13725 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
13726 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
13727 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
13728 (aarch64_constant_pool_reload_icode): Delete.
13729 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
13730 instead of aarch64_constant_pool_reload_icode. Use
13731 code_for_aarch64_reload_mov instead of explicit mode checks.
13732 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
13733 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
13734 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
13735 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
13736 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
13737 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
13738 (aarch64_atomic_load_op_code): Delete.
13739 (aarch64_emit_atomic_load_op): Likewise.
13740 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
13741 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
13742 instead of aarch64_emit_atomic_load_op.
13743 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
13744 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
13745 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
13746 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
13747 character before the pattern name.
13748 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
13749 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
13750 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
13751 (aarch64_frecps<mode>): Likewise.
13752 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
13753 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
13754 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
13755 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
13756 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
13758 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
13760 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
13761 Allow HFmode constants if TARGET_FP_F16INST.
13763 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
13766 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
13767 No longer check last store for clobber of address register.
13769 2018-08-02 Martin Liska <mliska@suse.cz>
13771 PR gcov-profile/86817
13772 * gcov.c (process_all_functions): New function.
13774 (process_file): Move functions processing to
13775 process_all_functions.
13777 2018-08-02 David Malcolm <dmalcolm@redhat.com>
13779 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
13780 "const" to the "gimple *" and "rtx_insn *" parameters.
13781 * dumpfile.h (dump_user_location_t::dump_user_location_t):
13783 (dump_location_t::dump_location_t): Likewise.
13785 2018-08-01 Martin Sebor <msebor@redhat.com>
13787 PR tree-optimization/86650
13788 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
13789 rather than a "gcall *". Directly pass the data of interest
13790 to percent_K_format, rather than building a temporary CALL_EXPR
13792 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
13793 (gimple_fold_builtin_strncat): Adjust.
13794 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
13795 gcall* argument with gimple*.
13796 * gimple-ssa-warn-restrict.c (check_call): Same.
13797 (wrestrict_dom_walker::before_dom_children): Same.
13798 (builtin_access::builtin_access): Same.
13799 (check_bounds_or_overlap): Same
13800 (maybe_diag_overlap): Same.
13801 (maybe_diag_offset_bounds): Same.
13802 * tree-diagnostic.c (default_tree_printer): Move usage of
13803 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
13805 * tree-pretty-print.c (percent_K_format): Add argument.
13806 * tree-pretty-print.h: Add argument.
13807 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
13808 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
13809 (maybe_diag_stxncpy_trunc): Same.
13810 (handle_builtin_stxncpy): Same.
13811 (handle_builtin_strcat): Same.
13813 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
13815 * match.pd: Optimise pointer range checks.
13817 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
13819 PR tree-optimization/86758
13820 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
13821 to remove pattern statements.
13823 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
13825 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
13826 result of dfs_enumerate_from when constructing stmt_vec_infos,
13827 instead of additionally calling get_loop_body.
13829 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
13831 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
13833 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
13834 When creating an iv, assert that the step is not known to be zero.
13835 (vect_setup_realignment): Update call accordingly.
13836 * tree-vect-stmts.c (vectorizable_store): Likewise.
13837 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
13839 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
13841 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
13842 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
13843 (vectorizable_reduction): Likewise.
13844 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13845 (vect_detect_hybrid_slp_stmts): Likewise.
13846 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
13848 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
13850 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
13851 (wide_int_set_zero_nonzero_bits): ...this.
13852 (zero_nonzero_bits_from_vr): Rename to...
13853 (vrp_set_zero_nonzero_bits): ...this.
13854 (extract_range_from_multiplicative_op_1): Abstract wide int
13856 (wide_int_range_multiplicative_op): ...here.
13857 (extract_range_from_binary_expr_1): Extract wide int binary
13858 operations into their own functions.
13859 (wide_int_range_lshift): New.
13860 (wide_int_range_can_optimize_bit_op): New.
13861 (wide_int_range_shift_undefined_p): New.
13862 (wide_int_range_bit_xor): New.
13863 (wide_int_range_bit_ior): New.
13864 (wide_int_range_bit_and): New.
13865 (wide_int_range_trunc_mod): New.
13866 (extract_range_into_wide_ints): New.
13867 (vrp_shift_undefined_p): New.
13868 (extract_range_from_multiplicative_op): New.
13869 (vrp_can_optimize_bit_op): New.
13870 * tree-vrp.h (value_range::dump): New.
13871 (wide_int_range_multiplicative_op): New.
13872 (wide_int_range_lshift):New.
13873 (wide_int_range_shift_undefined_p): New.
13874 (wide_int_range_bit_xor): New.
13875 (wide_int_range_bit_ior): New.
13876 (wide_int_range_bit_and): New.
13877 (wide_int_range_trunc_mod): New.
13878 (zero_nonzero_bits_from_bounds): Rename to...
13879 (wide_int_set_zero_nonzero_bits): ...this.
13880 (zero_nonzero_bits_from_vr): Rename to...
13881 (vrp_set_zero_nonzero_bits): ...this.
13882 (range_easy_mask_min_max): Rename to...
13883 (wide_int_range_can_optimize_bit_op): this.
13884 * vr-values.c (simplify_bit_ops_using_ranges): Rename
13885 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
13887 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
13889 * tree-vectorizer.h (vect_orig_stmt): New function.
13890 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
13891 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
13892 (vect_create_epilog_for_reduction): Likewise.
13893 (vectorizable_live_operation): Likewise.
13894 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
13895 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
13896 * tree-vect-stmts.c (vectorizable_call): Likewise.
13897 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
13899 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
13901 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
13903 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
13904 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
13905 (vect_transform_loop): Likewise.
13906 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
13908 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
13910 * tree-vectorizer.h (vect_schedule_slp): Return void.
13911 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
13912 (vect_schedule_slp): Likewise.
13914 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
13916 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
13918 (vect_transform_loop): Update calls accordingly. Schedule SLP
13919 instances before the main loop, if any exist.
13921 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
13923 PR tree-optimization/86749
13924 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
13925 If the lhs is used in a COND_EXPR, check that it is being used
13926 as the "then" or "else" value.
13928 2018-08-01 Tom de Vries <tdevries@suse.de>
13931 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
13932 speculation_safe_value_not_needed.
13934 2018-08-01 Richard Biener <rguenther@suse.de>
13936 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
13937 as base and offset.
13939 2018-08-01 Martin Liska <mliska@suse.cz>
13941 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
13942 format how successful transformation is dumped.
13943 (gimple_mod_pow2_value_transform): Likewise.
13944 (gimple_mod_subtract_transform): Likewise.
13945 (gimple_stringops_transform): Likewise.
13947 2018-08-01 Martin Liska <mliska@suse.cz>
13949 PR value-prof/35543
13950 * value-prof.c (interesting_stringop_to_profile_p):
13951 Simplify the code and add BUILT_IN_MEMMOVE.
13952 (gimple_stringops_transform): Likewise.
13954 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
13956 * config/aarch64/aarch64-simd.md
13957 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
13958 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
13959 use GPI iterator instead of SI mode.
13961 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
13963 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
13964 rs6000_speculation_barrier.
13965 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
13966 new barrier pattern name.
13968 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
13970 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
13971 (speculation_barrier): New insn.
13973 2018-08-01 Richard Biener <rguenther@suse.de>
13976 * graphite.h: Include isl/id.h and isl/space.h to allow build
13979 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
13982 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
13983 mode for COFF targets.
13984 * defaults.h (TARGET_COFF): Define.
13985 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
13986 TARGET_COFF): Define.
13987 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
13988 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
13989 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
13991 2018-07-31 Alexandre Oliva <oliva@adacore.com>
13993 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
13994 * gimple-streamer-out.c (output_bb): Save it.
13995 * lto-streamer-in.c (input_struct_function_base): Restore
13996 instance discriminator if available. Create map on demand.
13997 * lto-streamer-out.c (output_struct_function_base): Save it if
13999 * final.c (decl_to_instance_map): Document LTO strategy.
14001 2018-07-31 Alexandre Oliva <oliva@adacore.com>
14002 Olivier Hainque <hainque@adacore.com>
14004 * debug.h (decl_to_instance_map_t): New type.
14005 (decl_to_instance_map): Declare.
14006 (maybe_create_decl_to_instance_map): New inline function.
14007 * final.c (bb_discriminator, last_bb_discriminator): New statics,
14008 to track basic block discriminators.
14009 (final_start_function_1): Initialize them.
14010 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
14012 (decl_to_instance_map): New variable.
14013 (map_decl_to_instance, maybe_set_discriminator): New functions.
14014 (notice_source_line): Set discriminator.
14016 2018-07-31 Ian Lance Taylor <iant@golang.org>
14018 * targhooks.c (default_have_speculation_safe_value): Add
14021 2018-07-31 David Malcolm <dmalcolm@redhat.com>
14023 * dump-context.h: Include "pretty-print.h".
14024 (dump_context::refresh_dumps_are_enabled): New decl.
14025 (dump_context::emit_item): New decl.
14026 (class dump_context): Add fields "m_test_pp" and
14028 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
14029 (temp_dump_context::get_dumped_text): New decl.
14030 (class temp_dump_context): Add field "m_pp".
14031 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
14032 (dump_context::refresh_dumps_are_enabled): ...and add a test for
14034 (set_dump_file): Update for above change.
14035 (set_alt_dump_file): Likewise.
14036 (dump_loc): New overload, taking a pretty_printer *.
14037 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
14038 to any test pretty-printer.
14039 (make_item_for_dump_gimple_stmt): New function, adapted from
14040 optinfo::add_gimple_stmt.
14041 (dump_context::dump_gimple_stmt): Call it, and use the result,
14042 eliminating the direct usage of dump_file and alt_dump_file in
14043 favor of indirectly using them via emit_item.
14044 (make_item_for_dump_gimple_expr): New function, adapted from
14045 optinfo::add_gimple_expr.
14046 (dump_context::dump_gimple_expr): Call it, and use the result,
14047 eliminating the direct usage of dump_file and alt_dump_file in
14048 favor of indirectly using them via emit_item.
14049 (make_item_for_dump_generic_expr): New function, adapted from
14051 (dump_context::dump_generic_expr): Call it, and use the result,
14052 eliminating the direct usage of dump_file and alt_dump_file in
14053 favor of indirectly using them via emit_item.
14054 (make_item_for_dump_printf_va): New function, adapted from
14055 optinfo::add_printf_va.
14056 (make_item_for_dump_printf): New function.
14057 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
14058 and use the result, eliminating the direct usage of dump_file and
14059 alt_dump_file in favor of indirectly using them via emit_item.
14060 (make_item_for_dump_dec): New function.
14061 (dump_context::dump_dec): Call it, and use the result,
14062 eliminating the direct usage of dump_file and alt_dump_file in
14063 favor of indirectly using them via emit_item.
14064 (make_item_for_dump_symtab_node): New function, adapted from
14065 optinfo::add_symtab_node.
14066 (dump_context::dump_symtab_node): Call it, and use the result,
14067 eliminating the direct usage of dump_file and alt_dump_file in
14068 favor of indirectly using them via emit_item.
14069 (dump_context::begin_scope): Reimplement, avoiding direct usage
14070 of dump_file and alt_dump_file in favor of indirectly using them
14072 (dump_context::emit_item): New member function.
14073 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
14074 Set up test pretty-printer on the underlying context. Call
14075 refresh_dumps_are_enabled.
14076 (temp_dump_context::~temp_dump_context): Call
14077 refresh_dumps_are_enabled.
14078 (temp_dump_context::get_dumped_text): New member function.
14079 (selftest::verify_dumped_text): New function.
14080 (ASSERT_DUMPED_TEXT_EQ): New macro.
14081 (selftest::test_capture_of_dump_calls): Run all tests twice, with
14082 and then without optinfo enabled. Add uses of
14083 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
14084 * dumpfile.h: Update comment for the dump_* API.
14085 * optinfo-emit-json.cc
14086 (selftest::test_building_json_from_dump_calls): Update for new
14087 param for temp_dump_context ctor.
14088 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
14089 and "m_owned" field.
14090 (optinfo_item::~optinfo_item): Likewise.
14091 (optinfo::add_item): New member function.
14092 (optinfo::emit): Update comment.
14093 (optinfo::add_string): Delete.
14094 (optinfo::add_printf): Delete.
14095 (optinfo::add_printf_va): Delete.
14096 (optinfo::add_gimple_stmt): Delete.
14097 (optinfo::add_gimple_expr): Delete.
14098 (optinfo::add_tree): Delete.
14099 (optinfo::add_symtab_node): Delete.
14100 (optinfo::add_dec): Delete.
14101 * optinfo.h (class dump_context): New forward decl.
14102 (optinfo::add_item): New decl.
14103 (optinfo::add_string): Delete.
14104 (optinfo::add_printf): Delete.
14105 (optinfo::add_printf_va): Delete.
14106 (optinfo::add_gimple_stmt): Delete.
14107 (optinfo::add_gimple_expr): Delete.
14108 (optinfo::add_tree): Delete.
14109 (optinfo::add_symtab_node): Delete.
14110 (optinfo::add_dec): Delete.
14111 (optinfo::add_poly_int): Delete.
14112 (optinfo_item::optinfo_item): Remove "owned" param.
14113 (class optinfo_item): Remove field "m_owned".
14115 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
14117 PR middle-end/86705
14118 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
14119 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
14120 requested variable alignment.
14121 (expand_one_ssa_partition): Likewise.
14122 (expand_one_var): Likewise.
14124 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
14126 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
14127 to speculation_safe_value_not_needed.
14129 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
14131 * targhooks.h (speculation_safe_value_not_needed): New prototype.
14132 * targhooks.c (speculation_safe_value_not_needed): New function.
14133 * target.def (have_speculation_safe_value): Update documentation.
14134 * doc/tm.texi: Regenerated.
14136 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
14138 * config/aarch64/iterators.md (ALLI_TI): New iterator.
14139 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
14141 (despeculate_copy<ALLI:mode>_insn): New insn.
14142 (despeculate_copyti_insn): New insn.
14143 (despeculate_simple<ALLI:mode>): New insn
14144 (despeculate_simpleti): New insn.
14145 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
14147 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
14148 aarch64_speculation_safe_value.
14149 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
14151 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
14153 * config/aarch64/aarch64-speculation.cc: New file.
14154 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
14155 before pass_reorder_blocks.
14156 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
14158 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
14159 X14 and X15 when tracking speculation.
14160 * config/aarch64/aarch64.md (register name constants): Add
14161 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
14162 (unspec): Add UNSPEC_SPECULATION_TRACKER.
14163 (speculation_barrier): New insn attribute.
14164 (cmp<mode>): Allow SP in comparisons.
14165 (speculation_tracker): New insn.
14166 (speculation_barrier): Add speculation_barrier attribute.
14167 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
14168 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
14169 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
14171 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
14173 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
14174 aarch64_track_speculation is true.
14175 (tb<optab><mode>1): Likewise.
14176 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
14177 generate CB[N]Z when tracking speculation.
14178 (aarch64_split_compare_and_swap): Likewise.
14179 (aarch64_split_atomic_op): Likewise.
14181 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
14183 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
14185 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
14187 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
14188 (speculation_barrier): New insn.
14190 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
14192 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
14193 * config/arm/arm.md (speculation_barrier): New expand.
14194 (speculation_barrier_insn): New pattern.
14196 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
14198 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
14199 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
14200 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
14201 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
14203 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
14204 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
14205 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
14206 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
14207 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
14208 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
14209 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
14210 * builtins.c (expand_speculation_safe_value): New function.
14211 (expand_builtin): Call it.
14212 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
14213 * doc/extend.texi: Document __builtin_speculation_safe_value.
14214 * doc/md.texi: Document "speculation_barrier" pattern.
14215 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
14216 TARGET_HAVE_SPECULATION_SAFE_VALUE.
14217 * doc/tm.texi: Regenerated.
14218 * target.def (have_speculation_safe_value, speculation_safe_value): New
14220 * targhooks.c (default_have_speculation_safe_value): New function.
14221 (default_speculation_safe_value): New function.
14222 * targhooks.h (default_have_speculation_safe_value): Add prototype.
14223 (default_speculation_safe_value): Add prototype.
14225 2018-07-31 David Malcolm <dmalcolm@redhat.com>
14227 * dump-context.h (dump_context::dump_loc): New decl.
14228 * dumpfile.c (dump_context::dump_loc): New member function.
14229 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
14230 and dump_gimple_stmt.
14231 (dump_context::dump_gimple_expr_loc): Likewise, using
14233 (dump_context::dump_generic_expr_loc): Likewise, using
14235 (dump_context::dump_printf_loc_va): Likewise, using
14237 (dump_context::begin_scope): Explicitly using the global function
14238 "dump_loc", rather than the member function.
14240 2018-07-31 Martin Sebor <msebor@redhat.com>
14242 PR tree-optimization/86741
14243 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
14245 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
14247 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
14248 depend on whether prefetch instructions will be emitted or not.
14249 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
14250 will be emitted or not.
14251 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
14252 (TARGET_SETMEM_PFD): New macros.
14254 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14256 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
14257 (NULL_STMT_VEC_INFO): Delete.
14258 (stmt_vec_info::operator*): Likewise.
14259 (stmt_vec_info::operator gimple *): Likewise.
14260 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
14261 of NULL_STMT_VEC_INFO.
14262 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
14263 (vect_reassociating_reduction_p): Likewise.
14264 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
14265 (vectorizable_store): Likewise.
14266 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
14267 (vec_info::free_stmt_vec_infos): Likewise.
14269 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14271 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
14272 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
14274 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14276 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
14277 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
14278 (vec_info::free_stmt_vec_info): New private member functions.
14279 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
14280 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
14281 * tree-parloops.c (gather_scalar_reductions): Remove calls to
14282 set_stmt_vec_info_vec and free_stmt_vec_infos.
14283 * tree-vect-loop.c (_loop_vec_info): Remove call to
14284 set_stmt_vec_info_vec.
14285 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
14286 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
14287 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
14288 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
14289 (vec_info::free_stmt_vec_info): ...these new functions. Remove
14290 assignments in {vec_info::,}new_stmt_vec_info that are redundant
14291 with the clearing in the xcalloc.
14292 (stmt_vec_info_vec): Delete.
14293 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
14294 (vectorize_loops): Likewise.
14295 (vec_info::~vec_info): Remove argument from call to
14296 free_stmt_vec_infos.
14297 (vec_info::add_stmt): Remove vinfo argument from call to
14300 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14302 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
14303 rather than a gimple stmt.
14304 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
14305 information for pattern statements when passed the original
14306 statement; instead wait to be passed the pattern statement itself.
14307 Don't call set_vinfo_for_stmt here.
14308 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
14309 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
14310 stmt_vec_infos here.
14311 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
14312 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
14313 stmt_vec_infos entry.
14315 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14317 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
14318 * tree-vectorizer.c (vec_info::replace_stmt): New function.
14319 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
14320 * tree-vect-stmts.c (vectorizable_call): Likewise.
14321 (vectorizable_simd_clone_call): Likewise.
14323 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14325 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
14326 * tree-vectorizer.c (vec_info::remove_stmt): New function.
14327 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
14328 * tree-vect-loop.c (vect_transform_loop): Likewise.
14329 * tree-vect-slp.c (vect_schedule_slp): Likewise.
14330 * tree-vect-stmts.c (vect_remove_stores): Likewise.
14332 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14334 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
14335 (vect_dr_stmt): Delete.
14336 * tree-vectorizer.c (vec_info::lookup_dr): New function.
14337 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
14339 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14340 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
14341 (vect_verify_datarefs_alignment, vect_peeling_supportable)
14342 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
14343 (vect_analyze_data_refs): Likewise.
14344 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
14346 (vect_find_same_alignment_drs): Likewise.
14347 (vect_slp_analyze_node_dependences): Update calls accordingly.
14348 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
14349 instead of DR_VECT_AUX.
14350 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
14351 of a vector data references. Use vec_info::lookup_dr instead of
14353 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
14354 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
14355 instead of DR_VECT_AUX.
14357 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14359 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
14361 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
14363 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
14364 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
14365 (vect_gen_prolog_loop_niters): Likewise.
14367 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14369 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
14370 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
14371 (vect_known_alignment_in_bytes, vect_dr_behavior)
14372 (vect_get_scalar_dr_size): Take references as dr_vec_infos
14373 instead of data_references. Update calls to other routines for
14374 which the same change has been made.
14375 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
14376 dr_vec_infos instead of stmt_vec_infos.
14377 (vect_analyze_data_ref_dependence): Update call accordingly.
14378 (vect_slp_analyze_data_ref_dependence)
14379 (vect_record_base_alignments): Use DR_VECT_AUX.
14380 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
14381 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
14382 (vector_alignment_reachable_p, vect_get_data_access_cost)
14383 (vect_peeling_supportable, vect_analyze_group_access_1)
14384 (vect_analyze_group_access, vect_analyze_data_ref_access)
14385 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
14386 (vect_compile_time_alias, vect_small_gap_p)
14387 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
14388 (vect_supportable_dr_alignment): Take references as dr_vec_infos
14389 instead of data_references. Update calls to other routines for
14390 which the same change has been made.
14391 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
14392 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
14393 (vect_slp_analyze_and_verify_node_alignment)
14394 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
14395 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
14396 (vect_setup_realignment): Use dr_vec_infos. Update calls after
14398 (_vect_peel_info::dr): Replace with...
14399 (_vect_peel_info::dr_info): ...this new field.
14400 (vect_peeling_hash_get_most_frequent)
14401 (vect_peeling_hash_choose_best_peeling): Update accordingly.
14402 (vect_peeling_hash_get_lowest_cost):
14403 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
14404 routines for which the same change has been made.
14405 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
14407 * tree-vect-loop-manip.c (get_misalign_in_elems)
14408 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
14410 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
14411 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
14412 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
14413 (get_group_load_store_type, get_negative_load_store_type)
14414 (vect_get_data_ptr_increment, vectorizable_store)
14415 (vectorizable_load): Likewise.
14416 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
14417 Update calls to other routines for which the same change has been made.
14419 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14421 * tree-vectorizer.h (vec_info::move_dr): New member function.
14422 (dataref_aux): Rename to...
14423 (dr_vec_info): ...this and add "dr" and "stmt" fields.
14424 (_stmt_vec_info::dr_aux): Update accordingly.
14425 (_stmt_vec_info::data_ref_info): Delete.
14426 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
14427 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
14428 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
14429 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
14431 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
14432 (STMT_VINFO_DR_INFO): New macro.
14433 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
14434 (set_dr_misalignment): Update after rename of dataref_aux.
14435 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
14436 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
14437 initialization of STMT_VINFO_DATA_REF.
14438 * tree-vectorizer.c (vec_info::move_dr): New function.
14439 * tree-vect-patterns.c (vect_recog_bool_pattern)
14440 (vect_recog_mask_conversion_pattern)
14441 (vect_recog_gather_scatter_pattern): Use it.
14442 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
14443 the "dr" and "stmt" fields of dr_vec_info instead of
14444 STMT_VINFO_DATA_REF.
14446 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14448 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
14449 (is_pattern_stmt_p): Use it.
14450 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
14451 on pattern statements.
14453 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14455 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
14456 original stmt as a stmt_vec_info rather than a gimple stmt.
14457 (vect_pattern_recog_1): Take the statement directly as a
14458 stmt_vec_info, rather than via a gimple_stmt_iterator.
14459 Update call to vect_mark_pattern_stmts.
14460 (vect_pattern_recog): Update calls accordingly.
14462 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14464 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
14465 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
14466 a vect_def_type for the first argument.
14467 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
14468 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
14469 operand if it isn't defined by a vectorized statement.
14470 (vect_build_gather_load_calls): Remove the mask_dt argument and
14471 update calls to vect_get_vec_def_for_stmt_copy.
14472 (vectorizable_bswap): Likewise the dt argument.
14473 (vectorizable_call): Update calls to vectorizable_bswap and
14474 vect_get_vec_def_for_stmt_copy.
14475 (vectorizable_simd_clone_call, vectorizable_assignment)
14476 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
14477 (vectorizable_comparison): Update calls to
14478 vect_get_vec_def_for_stmt_copy.
14479 (vectorizable_store): Likewise. Remove now-unnecessary calls to
14480 vect_is_simple_use.
14481 (vect_get_loop_based_defs): Remove dt argument and update call
14482 to vect_get_vec_def_for_stmt_copy.
14483 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
14484 and vect_get_vec_def_for_stmt_copy.
14485 (vectorizable_load): Update calls to vect_build_gather_load_calls
14486 and vect_get_vec_def_for_stmt_copy.
14487 * tree-vect-loop.c (vect_create_epilog_for_reduction)
14488 (vectorizable_reduction, vectorizable_live_operation): Update calls
14489 to vect_get_vec_def_for_stmt_copy.
14491 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14493 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
14494 and gimple stmt arguments with a stmt_vec_info.
14495 (vect_record_base_alignments): Update calls accordingly.
14496 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
14497 and gimple stmt arguments with a stmt_vec_info.
14498 (vect_build_slp_tree_1): Remove vinfo argument and update call
14499 to vect_record_max_nunits.
14500 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
14501 and vect_record_max_nunits.
14503 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14505 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
14506 file and take a stmt_vec_info instead of a gimple stmt.
14507 (supportable_widening_operation, vect_finish_replace_stmt)
14508 (vect_finish_stmt_generation, vect_get_store_rhs)
14509 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
14510 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
14511 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
14512 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
14513 (vect_create_data_ref_ptr, bump_vector_ptr)
14514 (vect_permute_store_chain, vect_setup_realignment)
14515 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
14516 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
14517 (vectorizable_reduction, vectorizable_induction)
14518 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
14519 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
14520 than gimple stmts as arguments.
14521 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
14522 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
14523 (can_group_stmts_p, vect_check_gather_scatter)
14524 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
14525 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
14526 (vect_permute_load_chain, vect_shift_permute_load_chain)
14527 (vect_transform_grouped_load)
14528 (vect_record_grouped_load_vectors): Likewise.
14529 * tree-vect-loop.c (vect_fixup_reduc_chain)
14530 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
14531 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
14532 (vectorizable_reduction, vectorizable_induction)
14533 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
14534 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
14535 (vect_get_load_store_mask): Likewise.
14536 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
14537 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
14538 * tree-vect-stmts.c (vect_mark_relevant)
14539 (is_simple_and_all_uses_invariant)
14540 (exist_non_indexing_operands_for_use_p, process_use)
14541 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
14542 (vect_get_vec_def_for_operand, vect_get_vec_defs)
14543 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
14544 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
14545 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
14546 (get_negative_load_store_type, get_load_store_type)
14547 (vect_check_load_store_mask, vect_check_store_rhs)
14548 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
14549 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
14550 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
14551 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
14552 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
14553 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
14554 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
14555 (supportable_widening_operation): Likewise.
14557 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14559 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
14560 a stmt_vec_info instead of a gcall.
14561 (vect_check_gather_scatter): Update call accordingly.
14562 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
14564 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
14565 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
14566 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
14567 instead of a gimple stmt.
14568 (vect_transform_loop): Update calls accordingly.
14569 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
14570 stmt_vec_infos instead of gimple stmts.
14571 (vect_analyze_slp_instance): Update use accordingly.
14572 * tree-vect-stmts.c (read_vector_array, write_vector_array)
14573 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
14574 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
14575 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
14576 (vect_gen_widened_results_half, vect_get_loop_based_defs)
14577 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
14578 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
14579 down to subroutines.
14581 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14583 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
14584 of the worklist from a vector of gimple stmts to a vector of
14586 * tree-vect-stmts.c (vect_mark_relevant, process_use)
14587 (vect_mark_stmts_to_be_vectorized): Likewise
14589 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14591 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
14592 statement before passing it to vect_analyze_stmt.
14593 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
14594 the chain of phi vector definitions. Track the exit phi via its
14596 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
14597 STMT_VINFO_REDUC_DEF.
14598 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
14599 stmt_vec_infos to handle the statement chains.
14600 (vect_get_slp_defs): Record the first statement in the node
14601 using a stmt_vec_info.
14602 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
14603 statements here and pass their stmt_vec_info down to subroutines.
14604 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
14605 down to vect_finish_stmt_generation.
14606 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
14607 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
14608 stmt_vec_infos to subroutines.
14609 (vect_remove_stores): Use stmt_vec_infos to handle the statement
14612 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14614 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
14615 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
14616 (vect_permute_store_chain, vect_setup_realignment)
14617 (vect_permute_load_chain, vect_shift_permute_load_chain)
14618 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
14619 stmts internally, and when passing values to other vectorizer routines.
14620 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
14621 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
14622 (vect_analyze_loop_operations, get_initial_def_for_reduction)
14623 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
14624 (vectorizable_reduction, vectorizable_induction)
14625 (vectorizable_live_operation, vect_transform_loop_stmt)
14626 (vect_transform_loop): Likewise.
14627 * tree-vect-patterns.c (vect_reassociating_reduction_p)
14628 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
14629 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
14630 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14631 (vect_slp_analyze_node_operations_1): Likewise.
14632 * tree-vect-stmts.c (vect_mark_relevant, process_use)
14633 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
14634 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
14635 (vect_finish_stmt_generation_1, get_group_load_store_type)
14636 (get_load_store_type, vect_build_gather_load_calls)
14637 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
14638 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
14639 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
14640 (vectorizable_store, vectorizable_load, vectorizable_condition)
14641 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
14642 (supportable_widening_operation): Likewise.
14643 (vect_get_vector_types_for_stmt): Likewise.
14644 * tree-vectorizer.h (vect_dr_behavior): Likewise.
14646 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14648 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
14649 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
14650 (vect_permute_store_chain, vect_permute_load_chain)
14651 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
14652 repeated stmt_vec_info lookups.
14653 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
14654 (vect_update_ivs_after_vectorizer): Likewise.
14655 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
14656 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
14657 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
14658 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14659 (vect_bb_slp_scalar_cost): Likewise.
14660 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
14662 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14664 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
14665 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
14666 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
14667 to get gassigns and gcalls, rather than operating on generc gimple
14669 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
14670 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
14671 (vectorizable_load, vect_analyze_stmt): Likewise.
14672 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
14674 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14676 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
14677 return stmt_vec_infos rather than gimple stmts. Do not accept
14679 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
14681 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
14682 Update use of get_later_stmt.
14683 (vect_get_constant_vectors): Update call accordingly.
14684 (vect_schedule_slp_instance): Likewise
14685 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
14686 (vect_slp_analyze_instance_dependence): Likewise.
14687 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
14689 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14691 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
14692 (stmt_info_for_cost::stmt_info): ...this new field.
14693 (add_stmt_costs): Update accordingly.
14694 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
14695 (vect_get_known_peeling_cost): Likewise.
14696 (vect_estimate_min_profitable_iters): Likewise.
14697 * tree-vect-stmts.c (record_stmt_cost): Likewise.
14699 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14701 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
14702 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
14703 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
14705 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
14707 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14709 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
14710 a gimple stmt to a stmt_vec_info.
14711 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
14713 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14715 * tree-vectorizer.h (vec_info::grouped_stores): Change from
14716 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
14717 (_loop_vec_info::reduction_chains): Likewise.
14718 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
14720 * tree-vect-slp.c (vect_analyze_slp): Likewise.
14722 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14724 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
14725 a gimple stmt to a stmt_vec_info.
14726 (_stmt_vec_info::next_element): Likewise.
14727 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
14728 (vect_slp_analyze_and_verify_node_alignment)
14729 (vect_analyze_group_access_1, vect_analyze_group_access)
14730 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
14731 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
14732 (vect_supportable_dr_alignment): Update accordingly.
14733 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
14734 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
14735 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
14736 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
14737 * tree-vect-slp.c (vect_build_slp_tree_1)
14738 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
14739 (vect_split_slp_store_group, vect_analyze_slp_instance)
14740 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
14741 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
14742 (get_group_load_store_type, get_load_store_type)
14743 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
14744 (vect_transform_stmt, vect_remove_stores): Likewise.
14746 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14748 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
14749 than a gimple stmt.
14750 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
14751 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
14752 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
14753 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
14754 (vector_alignment_reachable_p, vect_get_data_access_cost)
14755 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
14756 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
14757 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
14758 (vect_analyze_group_access_1, vect_analyze_group_access)
14759 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
14760 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
14761 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
14762 result of vect_dr_stmt and use the stmt_vec_info instead of
14763 the associated gimple stmt.
14764 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
14765 (vect_gen_prolog_loop_niters): Likewise.
14766 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
14768 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14770 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
14771 to a vec<stmt_vec_info>.
14772 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
14773 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
14774 vec<stmt_vec_info>.
14775 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
14776 to a vec<stmt_vec_info>.
14777 (bst_traits::value_type, bst_traits::value_type): Likewise.
14778 (bst_traits::hash): Update accordingly.
14779 (vect_get_and_check_slp_defs): Change the stmts parameter from
14780 a vec<gimple *> to a vec<stmt_vec_info>.
14781 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
14782 (vect_build_slp_tree): Likewise.
14783 (vect_build_slp_tree_2): Likewise. Update uses of
14784 SLP_TREE_SCALAR_STMTS.
14785 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
14786 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
14787 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
14788 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
14789 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
14790 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
14791 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
14792 (vect_get_constant_vectors, vect_get_slp_defs)
14793 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
14794 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
14795 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
14796 instead of gimple stmts.
14797 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
14798 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
14799 (vect_slp_analyze_instance_dependence): Update uses of
14800 SLP_TREE_SCALAR_STMTS.
14801 (vect_slp_analyze_and_verify_node_alignment): Likewise.
14802 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
14803 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
14804 (get_initial_defs_for_reduction): Likewise.
14805 (vect_create_epilog_for_reduction): Likewise.
14806 (vectorize_fold_left_reduction): Likewise.
14807 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
14808 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
14809 (can_vectorize_live_stmts): Likewise.
14811 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14813 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
14814 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
14815 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
14817 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
14819 (gather_scalar_reductions): Update after above interface changes.
14820 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
14821 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
14823 (vect_force_simple_reduction): Likewise.
14824 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
14825 LOOP_VINFO_REDUCTIONS.
14826 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
14828 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14830 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
14831 a gimple stmt to a stmt_vec_info.
14832 * tree-vect-loop.c (vect_active_double_reduction_p)
14833 (vect_force_simple_reduction, vectorizable_reduction): Update
14836 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14838 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
14839 vec<gimple *> to a vec<stmt_vec_info>.
14840 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
14841 the reduction_phis argument from a vec<gimple *> to a
14842 vec<stmt_vec_info>.
14843 (vectorizable_reduction): Likewise the phis local variable that
14844 is passed to vect_create_epilog_for_reduction. Update for new type
14845 of SLP_TREE_VEC_STMTS.
14846 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
14847 (vectorizable_live_operation): Likewise.
14848 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
14849 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
14851 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14853 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
14854 a gimple stmt to a stmt_vec_info.
14855 (vectorizable_condition, vectorizable_live_operation)
14856 (vectorizable_reduction, vectorizable_induction): Pass back the
14857 vectorized statement as a stmt_vec_info.
14858 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
14859 use of STMT_VINFO_VEC_STMT.
14860 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
14861 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
14862 as stmt_vec_infos rather than gimple stmts.
14863 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
14864 to a stmt_vec_info.
14865 (vectorizable_live_operation): Likewise.
14866 (vectorizable_reduction, vectorizable_induction): Likewise,
14867 updating use of STMT_VINFO_VEC_STMT.
14868 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
14869 of STMT_VINFO_VEC_STMT.
14870 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
14871 (vectorizable_simd_clone_call, vectorizable_conversion)
14872 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
14873 (vectorizable_store, vectorizable_load, vectorizable_condition)
14874 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
14875 from a gimple stmt to a stmt_vec_info.
14876 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
14877 pointer to a stmt_vec_info to the vectorizable_* routines.
14879 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14881 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
14882 a gimple stmt to a stmt_vec_info.
14883 (is_pattern_stmt_p): Update accordingly.
14884 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
14885 (vect_record_grouped_load_vectors): Likewise.
14886 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
14887 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
14888 (vect_model_reduction_cost): Likewise.
14889 (vect_create_epilog_for_reduction): Likewise.
14890 (vectorizable_reduction, vectorizable_induction): Likewise.
14891 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
14892 Return the stmt_vec_info for the pattern statement.
14893 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
14894 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
14895 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
14896 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
14897 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
14898 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
14899 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
14900 (free_stmt_vec_info, vect_is_simple_use): Likewise.
14902 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14904 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
14905 (vect_finish_stmt_generation): Likewise.
14906 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
14907 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
14908 (vect_build_gather_load_calls): Use the return value of the above
14909 functions instead of a separate call to vinfo_for_stmt. Use narrow
14910 scopes for the input gimple stmt and wider scopes for the associated
14911 stmt_vec_info. Use vec_info::lookup_def when setting these
14912 stmt_vec_infos from an SSA_NAME definition.
14913 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
14914 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
14915 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
14916 (vectorizable_store, vectorizable_load, vectorizable_condition)
14917 (vectorizable_comparison): Likewise.
14918 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
14919 (vectorizable_reduction): Likewise.
14921 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14923 * tree-vectorizer.h (vect_is_simple_use): Add an optional
14924 stmt_vec_info * parameter before the optional gimple **.
14925 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
14926 (process_use, vect_get_vec_def_for_operand_1): Update callers.
14927 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
14928 * tree-vect-loop.c (vectorizable_reduction): Likewise.
14929 (vectorizable_live_operation): Likewise.
14930 * tree-vect-patterns.c (type_conversion_p): Likewise.
14931 (vect_look_through_possible_promotion): Likewise.
14932 (vect_recog_rotate_pattern): Likewise.
14933 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
14935 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14937 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
14938 a typedef to a wrapper class.
14939 (NULL_STMT_VEC_INFO): New macro.
14940 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
14941 (stmt_vec_info::operator*): New function.
14942 (stmt_vec_info::operator gimple *): Likewise.
14943 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
14944 (add_stmt_costs): Likewise.
14945 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
14946 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
14947 (vect_get_known_peeling_cost): Likewise.
14948 (vect_estimate_min_profitable_iters): Likewise.
14949 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
14950 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
14951 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
14952 (vectorizable_store, free_stmt_vec_infos): Likewise.
14953 (new_stmt_vec_info): Change return type of xcalloc to
14956 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14958 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
14959 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
14960 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
14961 a single_imm_use-based sequence.
14962 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
14964 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14966 * tree-vectorizer.h (vec_info::lookup_def): Declare.
14967 * tree-vectorizer.c (vec_info::lookup_def): New function.
14968 * tree-vect-patterns.c (vect_get_internal_def): Use it.
14969 (vect_widened_op_tree): Likewise.
14970 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
14971 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
14972 (vectorizable_reduction): Likewise.
14973 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
14975 (vect_is_slp_reduction): Update calls accordingly. Use
14976 vec_info::lookup_def.
14977 (vect_is_simple_reduction): Likewise
14978 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
14980 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
14982 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
14983 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
14984 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
14986 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
14987 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
14988 (vect_update_vf_for_slp, vect_analyze_loop_operations)
14989 (vect_is_slp_reduction, vectorizable_induction)
14990 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
14991 * tree-vect-patterns.c (vect_init_pattern_stmt):
14992 (vect_determine_min_output_precision_1, vect_determine_precisions)
14993 (vect_pattern_recog): Likewise.
14994 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
14995 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
14996 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
14997 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
14998 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
14999 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
15000 info field from a loop to a loop_vec_info.
15002 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
15004 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
15005 (vec_info::add_stmt): Declare.
15006 * tree-vectorizer.c (vec_info::add_stmt): New function.
15007 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
15008 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
15009 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
15010 (vectorizable_induction): Likewise.
15011 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
15012 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
15013 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
15014 (vectorizable_load): Likewise.
15015 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
15016 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
15017 (vect_recog_gather_scatter_pattern): Likewise.
15018 (append_pattern_def_seq): Likewise. Remove a check that is
15019 performed by add_stmt itself.
15021 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
15023 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
15024 which make_ssa_name was called with new_stmt before new_stmt
15027 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
15029 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
15031 (vect_is_slp_reduction): ...here...
15032 (vect_is_simple_reduction): ...and here. Remove repetition of tests
15033 that are already known to be false.
15035 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
15037 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
15038 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
15039 STMT_VINFO_NUM_SLP_USES when it's true.
15040 (vect_free_slp_instance): Add a final_p parameter and pass it to
15041 vect_free_slp_tree.
15042 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
15043 (vect_analyze_slp_instance): Likewise.
15044 (vect_slp_analyze_operations): Likewise.
15045 (vect_slp_analyze_bb_1): Likewise.
15046 * tree-vectorizer.c (vec_info): Likewise.
15047 * tree-vect-loop.c (vect_transform_loop): Likewise.
15049 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
15051 * tree-vect-loop.c (vectorizable_reduction): Assert that the
15052 function is not called for second and subsequent members of
15055 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
15057 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
15058 cases for nested loops from here to ...
15059 (vect_create_epilog_for_reduction): ...here. Only call
15060 vect_is_simple_use for inner-loop reductions.
15062 2018-07-31 Martin Liska <mliska@suse.cz>
15064 PR gcov-profile/85338
15065 PR gcov-profile/85350
15066 PR gcov-profile/85372
15067 * profile.c (struct location_triplet): New.
15068 (struct location_triplet_hash): Likewise.
15069 (output_location): Do not output a BB that
15070 is already recorded for a line.
15071 (branch_prob): Use streamed_locations.
15073 2018-07-31 Martin Liska <mliska@suse.cz>
15075 PR gcov-profile/85370
15076 * coverage.c (coverage_begin_function): Do not mark target
15077 clones as artificial functions.
15079 2018-07-31 Martin Liska <mliska@suse.cz>
15081 PR gcov-profile/83813
15082 PR gcov-profile/84758
15083 PR gcov-profile/85217
15084 PR gcov-profile/85332
15085 * profile.c (branch_prob): Do not record GOTO expressions
15086 for GIMPLE statements which locations are already streamed.
15088 2018-07-31 Olivier Hainque <hainque@adacore.com>
15090 * gcc.c (handle_spec_function): Accept a soft_matched_part
15091 argument, as do_spec_1. Pass it down to ...
15092 (eval_spec_function): Accept a soft_matched_part argument,
15093 and pass it down to ...
15094 (do_spec_2): Accept a soft_matched_part argument, and pass
15095 it down to do_spec_1.
15096 (do_spec_1): Pass soft_matched_part to handle_spec_function.
15097 (handle_braces): Update call to handle_spec_function.
15098 (driver::set_up_specs): Update calls to do_spec_2.
15099 (compare_debug_dump_opt_spec_function): Likewise.
15100 (compare_debug_self_opt_spec_function): Likewise.
15102 2018-07-31 Olivier Hainque <hainque@adacore.com>
15104 * common.opt (nolibc): New option.
15105 * doc/invoke.texi (Link Options): Document it.
15106 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
15107 * config/alpha/linux.h: Likewise.
15108 * config/arc/elf.h: Likewise.
15109 * config/arm/uclinux-elf.h: Likewise.
15110 * config/arm/unknown-elf.h: Likewise.
15111 * config/avr/avrlibc.h: Likewise.
15112 * config/bfin/bfin.h: Likewise.
15113 * config/bfin/linux.h: Likewise.
15114 * config/bfin/uclinux.h: Likewise.
15115 * config/darwin.h: Likewise.
15116 * config/darwin10.h: Likewise.
15117 * config/darwin12.h: Likewise.
15118 * config/gnu-user.h: Likewise.
15119 * config/lm32/uclinux-elf.h: Likewise.
15120 * config/pa/pa-hpux11.h: Likewise.
15121 * config/pa/pa64-hpux.h: Likewise.
15122 * config/sparc/sparc.h: Likewise.
15124 2018-07-31 Olivier Hainque <hainque@adacore.com>
15126 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
15127 undefined variables.
15129 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
15132 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
15133 instead of GEN_INT.
15135 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
15137 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
15138 terminated string literal.
15140 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
15142 PR rtl-optimization/85160
15143 * combine.c (is_just_move): New function.
15144 (try_combine): Allow combining two instructions into two if neither of
15145 the original instructions was a move.
15147 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
15150 * doc/extend.texi (Global Register Variables): Discourage use of type
15152 (Local Register Variables): Likewise.
15154 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
15156 PR tree-optimization/86506
15157 * hwint.h (ceil_log2): Resync with hwint.c implementation.
15159 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
15162 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
15163 hard_regno, make sure no insn between `from` and `to` clobbers it.
15165 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
15166 Tom de Vries <tdevries@suse.de>
15168 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
15169 (PTX_DEFAULT_RUNTIME_DIM): ... this.
15170 (nvptx_goacc_validate_dims): Set default worker and gang dims to
15171 PTX_DEFAULT_RUNTIME_DIM.
15172 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
15174 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
15176 * config/pa/pa.c (pa_output_addr_vec): Align address table.
15177 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
15178 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
15180 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
15182 * config/rs6000/constraints.md (wG constraint): Delete, no longer
15184 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
15185 predicate to reflect toc fusion has been deleted.
15186 (toc_fusion_mem_raw): Delete, no longer used.
15187 (toc_fusion_mem_wrapped): Likewise.
15188 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
15190 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
15191 Delete, no longer used.
15192 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
15193 meant to be used for toc fusion.
15194 (rs6000_debug_print_mode): Delete toc fusion debugging.
15195 (rs6000_debug_reg_global): Likewise.
15196 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
15197 fusion and secondary reload support that were never used.
15198 (rs6000_option_override_internal): Delete TOC fusion, that was only
15199 partially defined, and it did not work unless you also used the
15201 (rs6000_legitimate_address_p): Delete TOC fusion support.
15202 (rs6000_opt_masks): Likewise.
15203 (fusion_wrap_memory_address): Delete function, no longer used.
15204 (fusion_split_address); Delete TOC fusion support.
15205 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
15206 longer used with toc fusion being deleted.
15207 (TARGET_TOC_FUSION_FP): Likewise.
15208 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
15210 (toc fusion spliter): Delete TOC fusion support.
15211 (toc_fusionload_<mode>): Likewise.
15212 (toc_fusionload_di): Likewise.
15213 (fusion_gpr_load_<mode>): Delete generator function, this insn no
15214 longer needs to be named. Rename predicate to delete TOC fusion.
15215 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
15216 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
15217 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
15218 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
15219 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
15221 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
15223 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
15224 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
15225 __int128 in built-in function prototypes.
15226 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
15227 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
15229 2018-07-27 Martin Sebor <msebor@redhat.com>
15231 PR tree-optimization/86696
15232 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
15233 types, including enums.
15234 (handle_char_store): Be prepared for the above function to fail.
15236 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
15238 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
15239 when optimization level is lower than 2 or optimize for size.
15241 2018-07-26 Martin Sebor <msebor@redhat.com>
15243 PR tree-optimization/86043
15244 PR tree-optimization/86042
15245 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
15246 (get_string_cst_length): Rename...
15247 (get_min_string_length): ...to this. Add argument.
15248 (handle_char_store): Extend to handle multi-character stores by
15250 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
15251 * tree.h (initializer_zerop): Add argument.
15253 2018-07-26 Jakub Jelinek <jakub@redhat.com>
15255 PR middle-end/86660
15256 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
15257 declare target to variables if they have always,{to,from,tofrom} map
15260 2018-07-26 Martin Liska <mliska@suse.cz>
15263 * lto-wrapper.c: Add linker_output as prefix
15264 for ltrans_output_file.
15266 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
15268 PR rtl-optimization/85805
15269 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
15270 value for hard registers if that was written in the same mode.
15272 2018-07-26 Martin Liska <mliska@suse.cz>
15274 PR gcov-profile/86536
15275 * gcov.c (format_gcov): Use printf format %.*f directly
15276 and do not handle special values.
15278 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
15280 * common/config/arc/arc-common.c (arc_option_optimization_table):
15281 Update default optimizations for size.
15283 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
15285 * config/arc/arc.md (movsf_insn): Add short instruction selection.
15286 * config/arc/constraints.md (CfZ): New constraint.
15287 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
15288 (subsf3_fpu): Likewise.
15289 (cmpsf_fpu): Likewise.
15290 (cmpsf_fpu_uneq): Likewise.
15292 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
15294 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
15296 (arc_is_uncached_mem_p): uncached applies to both the variable and
15299 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
15301 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
15304 2018-07-25 David Malcolm <dmalcolm@redhat.com>
15306 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
15307 field "m_scopes" from vec to auto_vec.
15309 2018-07-25 Martin Liska <mliska@suse.cz>
15311 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
15314 2018-07-25 Richard Biener <rguenther@suse.de>
15317 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
15318 special wrt context_die late.
15319 (gen_subprogram_die): Re-use DIEs in local scope.
15321 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
15323 PR tree-optimization/86644
15324 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
15326 2018-07-25 Martin Liska <mliska@suse.cz>
15328 PR middle-end/86645
15329 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
15330 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
15332 2018-07-25 Martin Liska <mliska@suse.cz>
15335 * params.def: Explain ASan abbreviation and provide
15336 a documentation link.
15338 2018-07-24 Martin Sebor <msebor@redhat.com>
15340 PR tree-optimization/86622
15341 PR tree-optimization/86532
15342 * builtins.h (string_length): Declare.
15343 * builtins.c (c_strlen): Correct handling of non-constant offsets.
15344 (check_access): Be prepared for non-constant length ranges.
15345 (string_length): Make extern.
15346 * expr.c (string_constant): Only handle the minor non-constant
15347 array index. Use string_constant to compute the length of
15348 a generic string constant.
15350 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
15352 PR tree-optimization/86618
15353 * tree-vect-stmts.c (vectorizable_call): Don't take the address
15354 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
15356 2018-07-24 David Malcolm <dmalcolm@redhat.com>
15358 PR tree-optimization/86636
15359 * json.cc (json::object::set): Fix comment. Add assertions.
15360 (json::array::append): Move here from json.h. Add comment and an
15362 (json::string::string): Likewise.
15363 * json.h (json::array::append): Move to json.cc.
15364 (json::string::string): Likewise.
15365 * optinfo-emit-json.cc
15366 (optrecord_json_writer::impl_location_to_json): Assert that we
15367 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
15368 wrapper around it. Expand the location once, rather than three
15370 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
15371 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
15373 (optrecord_json_writer::optinfo_to_json): Likewise, in four
15374 places. Fix some overlong lines.
15376 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
15378 * config/aarch64/aarch64-simd.md
15379 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
15380 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
15381 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
15382 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
15383 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
15384 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
15385 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
15386 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
15387 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
15389 2018-07-24 Jakub Jelinek <jakub@redhat.com>
15391 PR middle-end/86627
15392 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
15393 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
15394 and abs_d == d, do the power of two handling if profitable.
15396 2018-07-24 Richard Biener <rguenther@suse.de>
15398 * match.pd: Add BIT_FIELD_REF canonicalizations.
15400 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
15403 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
15405 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
15407 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
15408 terminated STRING_CST object.
15410 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
15412 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
15414 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
15416 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
15417 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
15418 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
15419 the elements into a register.
15420 (rs6000_split_v4si_init_di_reg): Delete.
15421 (rs6000_split_v4si_init): Delete.
15422 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
15423 (vsx_init_v4si): Rewrite as a define_expand.
15425 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
15427 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
15428 zero_extend argument from memory): New.
15430 2018-07-22 Martin Sebor <msebor@redhat.com>
15433 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
15434 diagnosing calls with unknown arguments unless -Walloca-larger-than
15435 is restricted to less than PTRDIFF_MAX bytes.
15437 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
15439 * doc/gcov.texi (Invoking Gcov): Editorial changes.
15441 2018-07-20 David Malcolm <dmalcolm@redhat.com>
15443 * pretty-print.c (text_info::set_location): Remove redundant
15444 "line_table" parameter from call to rich_location::set_range.
15446 2018-07-20 Martin Sebor <msebor@redhat.com>
15448 PR middle-end/82063
15449 * builtins.c (expand_builtin_alloca): Adjust.
15450 * calls.c (alloc_max_size): Simplify.
15451 * cgraphunit.c (cgraph_node::expand): Adjust.
15452 * common.opt (larger_than_size, warn_frame_larger_than): Remove
15454 (frame_larger_than_size): Same.
15455 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
15456 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
15458 * doc/invoke.texi (GCC Command Options): Document option arguments.
15459 Explain byte-size arguments and suffixes.
15460 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
15461 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
15462 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
15463 * doc/options.texi (UInteger): Expand.
15464 (Host_Wide_Int, ByteSize): Document new properties.
15465 * final.c (final_start_function_1): Include sizes in an error message.
15466 * function.c (frame_offset_overflow): Same.
15467 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
15468 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
15469 Diagnose unbounded alloca calls only for limits of less than
15471 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
15472 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
15474 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
15475 only for limits of less than PTRDIFF_MAX.
15476 * langhooks-def.h (lhd_handle_option): Change function argument
15478 * langhooks.c (lhd_handle_option): Same.
15479 * langhooks.h (handle_option): Same.
15480 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
15482 (var_type, var_type_struct): Same.
15483 (var_set): Handle ByteSize flag.
15484 * optc-gen.awk: Add comments to output to ease debugging. Make
15485 use of HOST_WIDE_INT where appropriate.
15486 * opts-gen-save.awk: Use %lx to format unsigned long.
15487 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
15488 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
15489 arguments. Parse bytes-size suffixes.
15490 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
15491 (enum_value_to_arg): Same.
15492 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
15493 (handle_option): Adjust.
15494 (generate_option): Change function argument to HOST_WIDE_INT.
15495 (cmdline_handle_error): Adjust.
15496 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
15497 (set_option): Change function argument to HOST_WIDE_INT.
15498 (option_enabled): Handle cl_host_wide_int.
15499 (get_option_state): Handle CLVC_SIZE.
15500 (control_warning_option): Same.
15501 * opts.c (common_handle_option): Change function argument to
15502 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
15503 OPT_Wvla_larger_than_.
15504 * opts.h (enum cl_var_type): Add an enumerator.
15505 * stor-layout.c (layout_decl): Print a more meaningful warning.
15506 * toplev.c (output_stack_usage): Adjust.
15508 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
15510 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
15511 call to inline_expand_builtin_string_cmp.
15512 (expand_builtin_strcmp): Likewise.
15513 (expand_builtin_strncmp): Likewise.
15514 (inline_string_cmp): Delete the last parameter, change char_type_node
15515 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
15517 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
15518 the inlining expansion on target where the type of the call has same or
15519 narrower precision than unsigned char.
15521 2018-07-20 David Malcolm <dmalcolm@redhat.com>
15523 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
15524 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
15525 * common.opt (fsave-optimization-record): New option.
15526 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
15527 * doc/invoke.texi (-fsave-optimization-record): New option.
15528 * dumpfile.c: Include "optinfo-emit-json.h".
15529 (struct kv_pair): Move to coretypes.h.
15530 (optgroup_options): Make non-static.
15531 (dump_context::end_scope): Call
15532 optimization_records_maybe_pop_dump_scope.
15533 * dumpfile.h (optgroup_options): New decl.
15534 * json.cc: New file.
15535 * json.h: New file.
15536 * optinfo-emit-json.cc: New file.
15537 * optinfo-emit-json.h: New file.
15538 * optinfo.cc: Include "optinfo-emit-json.h".
15539 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
15540 (optinfo_enabled_p): Check optimization_records_enabled_p.
15541 (optinfo_wants_inlining_info_p): Likewise.
15542 * optinfo.h: Update comment.
15543 * profile-count.c (profile_quality_as_string): New function.
15544 * profile-count.h (profile_quality_as_string): New decl.
15545 (profile_count::quality): New accessor.
15546 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
15547 and optinfo_emit_json_cc_tests.
15548 * selftest.h (selftest::json_cc_tests): New decl.
15549 (selftest::optinfo_emit_json_cc_tests): New decl.
15550 * toplev.c: Include "optinfo-emit-json.h".
15551 (compile_file): Call optimization_records_finish.
15552 (do_compile): Call optimization_records_start.
15553 * tree-ssa-live.c: Include optinfo.h.
15554 (remove_unused_scope_block_p): Retain inlining information if
15555 optinfo_wants_inlining_info_p returns true.
15557 2018-07-20 Richard Biener <rguenther@suse.de>
15560 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
15561 to cover -flto-partition=none.
15563 2018-07-20 Martin Liska <mliska@suse.cz>
15565 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
15566 (get_decl_source_range): Remove unused function.
15568 2018-07-20 Richard Biener <rguenther@suse.de>
15570 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
15571 (struct vn_phi_s): Likewise.
15572 (struct vn_reference_s): Likewise.
15573 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
15574 for searching the slot of an entry known to be in the hash itself.
15575 (vn_phi_hasher::equal): Likewise.
15576 (vn_reference_hasher::equal): Likewise.
15577 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
15579 (optimistic_info, current_info): Remove, keeping only valid_info.
15580 (vn_reference_lookup_1): Remove fallback lookup.
15581 (vn_reference_lookup_2): Likewise.
15582 (vn_nary_op_lookup_1): Likewise.
15583 (vn_phi_lookup): Likewise.
15584 (vn_nary_build_or_lookup_1): Make sure to not chain the built
15586 (vn_reference_insert): Adjust, chain the inserted hash element
15587 at last_inserted_ref.
15588 (vn_reference_insert_pieces): Likewise.
15589 (visit_reference_op_call): Likewise.
15590 (vn_nary_op_insert_into): Chain the inserted hash element at
15591 last_inserted_nary.
15592 (vn_nary_op_insert_pieces): Adjust.
15593 (vn_nary_op_insert): Likewise.
15594 (vn_nary_op_insert_stmt): Likewise.
15595 (vn_phi_insert): Adjust, chain the inserted hash element at
15597 (process_scc): Remove clearing and copying the optimistic
15598 table. Instead remove elements inserted during an optimistic
15599 iteration from the single table we maintain.
15600 (init_scc_vn): Adjust.
15601 (free_scc_vn): Likewise.
15602 (sccvn_dom_walker::record_cond): Likewise.
15603 (sccvn_dom_walker::after_dom_children): Likewise.
15605 2018-07-19 Martin Sebor <msebor@redhat.com>
15607 PR tree-optimization/84047
15608 PR tree-optimization/83776
15609 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
15610 (check_array_bounds): Call it.
15612 2018-07-19 Martin Sebor <msebor@redhat.com>
15614 * align.h (align_flags): Use member initialization.
15616 2018-07-19 David Malcolm <dmalcolm@redhat.com>
15618 * Makefile.in (OBJS): Add optinfo.o.
15619 * coretypes.h (class symtab_node): New forward decl.
15620 (struct cgraph_node): New forward decl.
15621 (class varpool_node): New forward decl.
15622 * dump-context.h: New file.
15623 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
15625 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
15626 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
15627 (set_alt_dump_file): Likewise.
15628 (dump_context::~dump_context): New dtor.
15629 (dump_gimple_stmt): Move implementation to...
15630 (dump_context::dump_gimple_stmt): ...this new member function.
15631 Add the stmt to any pending optinfo, creating one if need be.
15632 (dump_gimple_stmt_loc): Move implementation to...
15633 (dump_context::dump_gimple_stmt_loc): ...this new member function.
15634 Start a new optinfo and add the stmt to it.
15635 (dump_gimple_expr): Move implementation to...
15636 (dump_context::dump_gimple_expr): ...this new member function.
15637 Add the stmt to any pending optinfo, creating one if need be.
15638 (dump_gimple_expr_loc): Move implementation to...
15639 (dump_context::dump_gimple_expr_loc): ...this new member function.
15640 Start a new optinfo and add the stmt to it.
15641 (dump_generic_expr): Move implementation to...
15642 (dump_context::dump_generic_expr): ...this new member function.
15643 Add the tree to any pending optinfo, creating one if need be.
15644 (dump_generic_expr_loc): Move implementation to...
15645 (dump_context::dump_generic_expr_loc): ...this new member
15646 function. Add the tree to any pending optinfo, creating one if
15648 (dump_printf): Move implementation to...
15649 (dump_context::dump_printf_va): ...this new member function. Add
15650 the text to any pending optinfo, creating one if need be.
15651 (dump_printf_loc): Move implementation to...
15652 (dump_context::dump_printf_loc_va): ...this new member function.
15653 Start a new optinfo and add the stmt to it.
15654 (dump_dec): Move implementation to...
15655 (dump_context::dump_dec): ...this new member function. Add the
15656 value to any pending optinfo, creating one if need be.
15657 (dump_context::dump_symtab_node): New member function.
15658 (dump_context::get_scope_depth): New member function.
15659 (dump_context::begin_scope): New member function.
15660 (dump_context::end_scope): New member function.
15661 (dump_context::ensure_pending_optinfo): New member function.
15662 (dump_context::begin_next_optinfo): New member function.
15663 (dump_context::end_any_optinfo): New member function.
15664 (dump_context::s_current): New global.
15665 (dump_context::s_default): New global.
15666 (dump_scope_depth): Delete global.
15667 (dumpfile_ensure_any_optinfo_are_flushed): New function.
15668 (dump_symtab_node): New function.
15669 (get_dump_scope_depth): Reimplement in terms of dump_context.
15670 (dump_begin_scope): Likewise.
15671 (dump_end_scope): Likewise.
15672 (selftest::temp_dump_context::temp_dump_context): New ctor.
15673 (selftest::temp_dump_context::~temp_dump_context): New dtor.
15674 (selftest::verify_item): New function.
15675 (ASSERT_IS_TEXT): New macro.
15676 (ASSERT_IS_TREE): New macro.
15677 (ASSERT_IS_GIMPLE): New macro.
15678 (selftest::test_capture_of_dump_calls): New test.
15679 (selftest::dumpfile_c_tests): Call it.
15680 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
15681 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
15682 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
15683 descriptive comment.
15684 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
15685 (dump_node, dump_bb): Move these unrelated decls.
15686 (class dump_manager): Add leading comment.
15687 * optinfo.cc: New file.
15688 * optinfo.h: New file.
15690 2018-07-19 Michael Collison <michael.collison@arm.com>
15691 Richard Henderson <rth@redhat.com>
15693 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
15694 (subti): Handle op1 zero.
15695 (subvti4, usub4ti4): New.
15696 (*sub<GPI>3_compare1_imm): New.
15697 (sub<GPI>3_carryinCV): New.
15698 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
15699 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
15701 2018-07-19 Michael Collison <michael.collison@arm.com>
15702 Richard Henderson <rth@redhat.com>
15704 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
15705 (addti3): Create simpler code if low part is already known to be 0.
15706 (addvti4, uaddvti4): New.
15707 (*add<GPI>3_compareC_cconly_imm): New.
15708 (*add<GPI>3_compareC_cconly): New.
15709 (*add<GPI>3_compareC_imm): New.
15710 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
15711 handle constants within this pattern..
15712 (*add<GPI>3_compareV_cconly_imm): New.
15713 (*add<GPI>3_compareV_cconly): New.
15714 (*add<GPI>3_compareV_imm): New.
15715 (add<GPI>3_compareV): New.
15716 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
15717 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
15718 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
15719 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
15720 with 'comparison' operator.
15721 (*add<GPI>3_compareV_cconly_imm): Ditto.
15722 (*add<GPI>3_compareV_cconly): Ditto.
15723 (*add<GPI>3_compareV_imm): Ditto.
15724 (add<GPI>3_compareV): Ditto.
15725 (add<mode>3_carryinC): Ditto.
15726 (*add<mode>3_carryinC_zero): Ditto.
15727 (*add<mode>3_carryinC): Ditto.
15728 (add<mode>3_carryinV): Ditto.
15729 (*add<mode>3_carryinV_zero): Ditto.
15730 (*add<mode>3_carryinV): Ditto.
15732 2018-07-19 Michael Collison <michael.collison@arm.com>
15733 Richard Henderson <rth@redhat.com>
15735 * config/aarch64/aarch64-modes.def (CC_V): New.
15736 * config/aarch64/aarch64-protos.h
15737 (aarch64_addti_scratch_regs): Declare
15738 (aarch64_subvti_scratch_regs): Declare.
15739 (aarch64_expand_subvti): Declare.
15740 (aarch64_gen_unlikely_cbranch): Declare
15741 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
15742 for signed overflow using CC_Vmode.
15743 (aarch64_get_condition_code_1): Handle CC_Vmode.
15744 (aarch64_gen_unlikely_cbranch): New function.
15745 (aarch64_addti_scratch_regs): New function.
15746 (aarch64_subvti_scratch_regs): New function.
15747 (aarch64_expand_subvti): New function.
15749 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
15751 * config/aarch64/aarch64-option-extensions.def: New entry for profile
15753 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
15754 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
15757 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
15760 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
15761 address check not strict.
15763 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
15765 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
15767 (store_pair_lanes<mode>): Likewise.
15768 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
15769 enum value 'ADDR_QUERY_LDP_STP_N'.
15770 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
15771 (aarch64_print_address_internal): Add declaration.
15772 (aarch64_print_ldpstp_address): Remove.
15773 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
15774 (aarch64_print_operand): Change printing of 'y'.
15775 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
15776 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
15777 'true' rather than '1'.
15778 * config/aarch64/constraints.md (Uml): Likewise.
15779 (Uml): Rename to Umn.
15782 2018-07-19 Richard Biener <rguenther@suse.de>
15784 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
15786 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
15787 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
15788 (vn_reference_hasher): Likewise.
15789 (struct vn_tables_s): Remove obstack and alloc-pool members.
15790 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
15791 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
15792 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
15793 (vn_reference_insert_pieces): Likewise.
15794 (alloc_vn_nary_op_noinit): Adjust.
15795 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
15796 (vn_phi_eq): Adjust.
15797 (shared_lookup_phiargs): Remove.
15798 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
15799 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
15800 (visit_reference_op_call): Likewise.
15801 (copy_nary, copy_phi, copy_reference): Remove.
15802 (process_scc): Rewind the obstack when iterating. Do not
15803 copy the elements to valid_info but just move them from one
15804 hashtable to the other.
15805 (allocate_vn_table): Adjust.
15806 (free_vn_table): Likewise.
15807 (init_scc_vn): Likewise.
15808 (free_scc_vn): Likewise.
15810 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
15813 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
15814 indirect_return as function type attribute.
15815 (ix86_attribute_table): Change indirect_return to function
15817 * doc/extend.texi: Update indirect_return attribute.
15819 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
15821 * wide-int.h (widest2_int): New.
15822 * gimple-fold.c (arith_overflowed_p): Use it.
15823 * tree.h (widest2_int_cst): New.
15824 * tree-vrp.c (wide_int_binop_overflow): Rename from
15825 vrp_int_const_binop.
15826 Rewrite to work on trees.
15827 (extract_range_from_multiplicative_op_1): Abstract code to...
15828 (wide_int_range_min_max): ...here.
15829 (wide_int_range_cross_product): ...and here.
15830 (extract_range_from_binary_expr_1): Abstract overflow code to...
15831 (wide_int_range_mult_wrapping): ...here.
15832 * tree-vrp.h (wide_int_range_cross_product): New.
15833 (wide_int_range_mult_wrapping): New.
15835 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
15836 Julia Koval <julia.koval@intel.com>
15838 * config/i386/x86-tune-costs.h (skylake_memcpy,
15839 skylake_memset): Replace rep_prefix with unrolling for size 512.
15841 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
15843 PR middle-end/86544
15844 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
15845 comparision with EQ_EXPR in last stmt.
15847 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
15849 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
15850 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
15851 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
15852 previously known as "PowerPC AltiVec Built-in Functions". Move
15853 some material to new subsubsections "PowerPC AltiVec Built-in
15854 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
15856 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
15857 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
15858 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
15859 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
15861 2018-07-18 Richard Biener <rguenther@suse.de>
15863 PR tree-optimization/86557
15864 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
15867 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
15869 * config/s390/s390.c (s390_function_profiler): Generate CFI.
15871 2018-07-17 Jeff Law <law@redhat.com>
15873 * config/arm/arm.c (get_label_padding): Update for recent
15874 changes to label_to_alignment.
15876 PR tree-optimization/86010
15877 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
15879 * config/mips/mips.c (vr4130_align_insns): Update for recent
15880 changes to label_to_alignment.
15882 * config/frv/frv.c (frv_label_align): Update for recent changes
15883 to label_to_alignment.
15885 * config/nios2/nios2.c (nios2_label_align): Update for recent
15886 changes which dropped ALIGN_LABELS_LOG.
15888 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
15890 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
15891 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
15893 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
15895 * config/arc/arc.c (arc_label_align): Use align_labels instead of
15896 deprecated align_labels_log.
15898 2018-07-17 Richard Biener <rguenther@suse.de>
15901 * dwarf2out.c (init_sections_and_labels): Always generate
15902 a debug_line_str_section for early LTO debug.
15903 (dwarf2out_finish): Reset debug_line_str_hash output early.
15904 Bump counter for extra dwarf5 .debug_loc labels to not conflict
15906 (dwarf2out_early_finish): Output debug_line_str.
15908 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
15910 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
15911 index register on z196 or later.
15913 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
15915 * config/s390/s390.c (s390_default_align): Set default function
15917 (s390_override_options_after_change): Call s390_default align.
15918 (s390_option_override_internal): Call s390_default_align.
15919 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
15921 2018-07-17 Jakub Jelinek <jakub@redhat.com>
15923 PR middle-end/86542
15924 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
15925 to _looptemp_ clauses, other than the first two.
15927 2018-07-17 Martin Liska <mliska@suse.cz>
15929 * opts.c: Do not enable OPT_falign_* for -Os.
15931 2018-07-17 Martin Liska <mliska@suse.cz>
15933 * align.h (MAX_CODE_ALIGN): New.
15934 (MAX_CODE_ALIGN_VALUE): New.
15935 * common/config/i386/i386-common.c (ix86_handle_option):
15936 (MAX_CODE_ALIGN): Moved to align.h.
15937 * final.c (MAX_CODE_ALIGN): Likewise.
15938 * opts.c (parse_and_check_align_values):
15939 (MAX_CODE_ALIGN): Likewise.
15940 (MAX_CODE_ALIGN_VALUE): Likewise.
15942 2018-07-17 Martin Liska <mliska@suse.cz>
15944 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
15945 in order to fulfil coding style.
15946 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
15947 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15948 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
15949 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
15950 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
15951 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
15952 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
15953 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
15954 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
15956 2018-07-17 Martin Liska <mliska@suse.cz>
15958 * align.h: New file.
15959 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
15961 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
15962 align_flags of label_to_alignment.
15963 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
15965 * config/m68k/m68k.c: Do not use removed align_labels_value and
15967 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
15968 (LOOP_ALIGN): Likewise.
15969 (LABEL_ALIGN): Likewise.
15970 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
15971 Remove not used macro.
15972 (rs6000_loop_align): Change return type to align_flags.
15973 (rs6000_loop_align_max_skip): Remove.
15974 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
15975 Change return type to align_flags.
15976 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
15977 Remove not used macro.
15978 (rs6000_loop_align): Change return type to align_flags.
15979 (rs6000_loop_align_max_skip): Remove.
15980 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
15981 * config/rx/rx-protos.h (rx_align_for_label): Make it
15983 * config/rx/rx.c (rx_align_for_label): Change return type
15985 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
15987 into align_flags class.
15988 (LABEL_ALIGN): Likewise.
15989 (LOOP_ALIGN): Likewise.
15990 * config/s390/s390.c (s390_label_align): Use align_flags
15992 (s390_asm_output_function_label): Likewise.
15993 * config/sh/sh.c (sh_override_options_after_change):
15994 Use align_flags class directly without macros.
15995 (find_barrier): Likewise.
15996 (barrier_align): Likewise.
15997 (sh_loop_align): Likewise.
15998 * config/spu/spu.c (spu_option_override):
15999 Use align_flags_tuple::get_value instead of removed macros.
16000 (spu_sched_init): Likewise.
16001 * config/spu/spu.h (GTY): Likewise.
16002 * config/visium/visium.c (visium_option_override):
16003 Set "8" as default secondary alignment.
16004 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
16005 in order to guarantee secondary alignment of 8.
16006 * coretypes.h: Include align.h header file.
16007 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
16008 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
16009 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
16010 * doc/tm.texi.in: Likewise.
16011 * final.c (struct label_alignment): Remove not used structure.
16012 (LABEL_ALIGN): Change type to align_flags.
16013 (LOOP_ALIGN): Likewise.
16014 (JUMP_ALIGN): Likewise.
16015 (default_loop_align_max_skip): Remove.
16016 (default_label_align_max_skip): Likewise.
16017 (default_jump_align_max_skip): Likewise.
16018 (default_label_align_after_barrier_max_skip):
16019 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
16020 (LABEL_TO_MAX_SKIP): Remove.
16021 (label_to_alignment): Return align_flags type instead of integer.
16022 (label_to_max_skip): Remove.
16023 (align_fuzz): Use align_flags type.
16024 (compute_alignments): Use align_flags type and use align_flags::max
16025 to combine multiple alignments.
16026 (grow_label_align): Grow vec instead of C array.
16027 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
16028 (shorten_branches): Use align_flags type and use align_flags::max
16029 to combine multiple alignments.
16030 (final_scan_insn_1): Remove usage of secondary alignment that comes
16031 from label alignment, but instead use proper secondary alignment
16032 which is computed in grow_label_align.
16033 * flags.h (struct align_flags_tuple): Move to align.h.
16034 (struct align_flags): Likewise.
16035 (state_align_loops): Rename to align_loops.
16036 (state_align_jumps): Rename to align_jumps.
16037 (state_align_labels): Rename to align_labels.
16038 (state_align_functions): Rename to align_functions.
16039 (align_loops_log): Remove.
16040 (align_jumps_log): Remove.
16041 (align_labels_log): Remove.
16042 (align_functions_log): Remove.
16043 (align_loops_max_skip): Remove.
16044 (align_jumps_max_skip): Remove.
16045 (align_labels_max_skip): Remove.
16046 (align_functions_max_skip): Remove.
16047 (align_loops_value): Remove.
16048 (align_jumps_value): Remove.
16049 (align_labels_value): Remove.
16050 (align_functions_value): Remove.
16051 * output.h (label_to_alignment): Change return type to align_flags.
16052 (label_to_max_skip): Remove.
16053 * target.def: Remove loop_align_max_skip, label_align_max_skip,
16054 jump_align_max_skip macros.
16055 * targhooks.h (default_loop_align_max_skip): Remove.
16056 (default_label_align_max_skip): Likewise.
16057 (default_jump_align_max_skip): Likewise.
16058 (default_label_align_after_barrier_max_skip): Remove.
16059 * toplev.c (read_log_maxskip): Use ::normalize function.
16060 (parse_N_M): Remove not used argument and also call ::normalize.
16061 (parse_alignment_opts): Do not pass unused arguments.
16062 * varasm.c (assemble_start_function): Use directly align_functions
16063 instead of removed macros.
16064 * system.h: Do not poison removed macros.
16066 2018-07-17 Jakub Jelinek <jakub@redhat.com>
16068 PR middle-end/86539
16069 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
16070 and cond temporaries don't have reference type if iterator has
16071 pointer type. For init use &for_pre_body instead of pre_p if
16072 for_pre_body is non-empty.
16074 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
16076 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
16077 double-double modes to SFmode directly directly.
16078 (trunc<mode>sf2_fprs): Delete.
16080 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
16082 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
16083 for conversions between IFmode and the decimal floating point modes.
16084 (init_float128_ieee): Use the correct names for conversions between
16085 KFmode and the decimal floating point modes.
16087 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
16089 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
16090 for the conversions between TDmode and IFmode.
16091 (init_float128_ieee): Use more correct names for the conversions
16092 between TDmode and KFmode.
16094 2018-07-16 Jakub Jelinek <jakub@redhat.com>
16096 PR tree-optimization/86526
16097 * builtins.c (expand_builtin_memcmp): Formatting fixes.
16098 (inline_expand_builtin_string_cmp): Likewise.
16099 (inline_string_cmp): Likewise. Use c_readstr instead of
16100 builtin_memcpy_read_str. Add unit_mode temporary.
16102 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
16104 PR middle-end/86528
16105 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
16106 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
16108 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
16110 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
16111 Alphabetize prototypes of built-in functions, separating out
16112 built-in functions that are listed in this section but should be
16113 described elsewhere.
16115 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
16118 * expmed.c (emit_store_flag): Do not emit setcc followed by a
16119 conditional move when trapping comparison was split to a
16120 non-trapping one (and vice versa).
16122 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
16124 * config/s390/s390.c (s390_function_profiler): Generate nops
16125 instead of profiler call sequences.
16126 * config/s390/s390.opt: Add the new option.
16128 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
16130 * config/s390/s390.c (s390_function_profiler): Generate
16131 __mcount_loc section.
16132 * config/s390/s390.opt: Add the new option.
16134 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
16136 * common.opt: Add the new warning.
16137 * config/s390/s390.c (s390_function_profiler): Emit "brasl
16138 %r0,__fentry__" when -mfentry is specified.
16139 (s390_option_override_internal): Disallow -mfentry for 31-bit
16141 * config/s390/s390.opt: Add the new option.
16143 2018-07-16 Richard Biener <rguenther@suse.de>
16146 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
16147 for function-local FUNCTION_DECL and RESULT_DECL immediately.
16149 2018-07-16 Martin Liska <mliska@suse.cz>
16152 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
16155 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
16157 * config/arc/arcHS.md: Update ARCHS scheduling rules.
16159 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
16161 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
16163 * config/arc/arc-cpus.def (hs4x): New cpu.
16165 * config/arc/arc-tables.opt: Regenerate.
16166 * config/arc/arc.c (arc_sched_issue_rate): New function.
16167 (TARGET_SCHED_ISSUE_RATE): Define.
16168 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
16169 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
16171 (attr tune): Add ARCHS4x tune values.
16172 (attr tune_dspmpy): Define.
16173 (*tst): Correct instruction type.
16174 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
16175 * config/arc/arcHS4x.md: New file.
16176 * config/arc/fpu.md: Update instruction type attributes.
16177 * config/arc/t-multilib: Regenerate.
16179 2018-07-16 Tom de Vries <tdevries@suse.de>
16182 * var-tracking.c (vt_initialize): Fix pre_dec handling.
16184 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
16186 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
16189 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
16191 PR tree-optimization/86514
16192 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
16193 conversion to a boolean type from a type with greater precision.
16195 2018-07-16 Tom de Vries <tdevries@suse.de>
16197 * var-tracking.c (vt_initialize): Print adjusted insn slim if
16198 dump_flags request TDF_SLIM.
16200 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
16202 * fold-const.c (int_const_binop_1): Abstract...
16203 (wide_int_binop): ...wide int code here.
16204 (poly_int_binop): ...poly int code here.
16205 Abstract the rest of int_const_binop_1 into int_const_binop.
16206 * fold-const.h (wide_int_binop): New.
16207 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
16208 Remove useless PLUS/MINUS_EXPR case.
16209 (zero_nonzero_bits_from_vr): Move wide int code...
16210 (zero_nonzero_bits_from_bounds): ...here.
16211 (extract_range_from_binary_expr_1): Move mask optimization code...
16212 (range_easy_mask_min_max): ...here.
16213 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
16214 (range_easy_mask_min_max): New.
16216 2018-07-15 Jeff Law <law@redhat.com>
16219 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
16222 2018-07-14 Jim Wilson <jimw@sifive.com>
16224 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
16226 2018-07-14 Paul Koning <ni1d@arrl.net>
16228 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
16230 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
16232 * lto-streamer-out.c (copy_function_or_variable): Dump info about
16235 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
16236 Steve Munroe <munroesj52@gmail.com>
16238 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
16239 (_mm_andnot_si128): Likewise.
16240 (_mm_or_si128): Likewise.
16241 (_mm_xor_si128): Likewise.
16243 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
16245 PR middle-end/78809
16246 * builtins.c (expand_builtin_memcmp): Inline the calls first
16247 when result_eq is false.
16248 (expand_builtin_strcmp): Inline the calls first.
16249 (expand_builtin_strncmp): Likewise.
16250 (inline_string_cmp): New routine. Expand a string compare
16251 call by using a sequence of char comparison.
16252 (inline_expand_builtin_string_cmp): New routine. Inline expansion
16253 a call to str(n)cmp/memcmp.
16254 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
16256 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
16258 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
16260 * config/arm/driver-arm.c: Include arm-native.h.
16261 (host_detect_local_cpu): Use auto-generated data tables.
16262 (vendors, arm_cpu_table): Delete. Move part information to ...
16263 * config/arm/arm-cpus.in: ... here.
16264 * config/arm/parsecpu.awk (gen_native): New function.
16265 (vendor, part): New CPU fields.
16266 (END): Add support for building the native CPU detection tables.
16267 * config/arm/t-arm (arm-native.h): Add build rule.
16268 (driver-arm.o): Add dependency on arm-native.h.
16270 2018-07-13 Richard Biener <rguenther@suse.de>
16272 PR middle-end/85974
16273 * match.pd (addr1 - addr2): Allow either of the operand to
16276 2018-07-13 Tom de Vries <tdevries@suse.de>
16278 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
16281 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
16283 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
16284 arrays instead of numbered variables.
16286 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
16288 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
16289 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
16291 2018-07-13 Richard Biener <rguenther@suse.de>
16294 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
16295 instead of get_context_die.
16297 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
16298 Richard Biener <rguenther@suse.de>
16300 PR middle-end/86489
16301 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
16302 that the loop latch destination where phi is defined.
16304 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
16306 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
16307 (riscv_expand_epilogue): Add assertion to check interrupt mode.
16308 (riscv_set_current_function): Extract getting interrupt type to new
16310 (riscv_get_interrupt_type): New function.
16311 (riscv_merge_decl_attributes): New function, checking interrupt type is
16313 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
16315 2018-07-12 Paul Koning <ni1d@arrl.net>
16317 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
16320 2018-07-12 Paul Koning <ni1d@arrl.net>
16322 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
16323 zero reference, add doloop_end instead.
16324 * doc/md.texi (decrement_and_branch_until_zero): Remove.
16325 (Looping patterns): Remove decrement_and_branch_until_zero. Add
16326 detail for doloop_end.
16328 2018-07-12 Martin Sebor <msebor@redhat.com>
16331 * attribs.c (decl_attributes): Reject conflicting attributes before
16332 calling attribute handlers.
16334 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
16336 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
16338 (gcc::dump_manager::get_dump_file_name): likewise.
16339 (dump_begin): Likewise.
16340 * dumpfile.h (dump_begin): Update prototype.
16341 (gcc::dump_manager::get_dump_file_name,
16342 gcc::dump_manager::get_dump_file_name): Update prototype.
16344 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
16346 * internal-fn.h (vectorizable_internal_fn_p): New function.
16347 * tree-vect-slp.c (compatible_calls_p): Likewise.
16348 (vect_build_slp_tree_1): Remove nops argument. Handle calls
16349 to internal functions.
16350 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
16352 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
16354 * fold-const.h (inverse_conditions_p): Declare.
16355 * fold-const.c (inverse_conditions_p): New function.
16356 * match.pd: Use inverse_conditions_p. Add folds of view_converts
16357 that test the inverse condition of a conditional internal function.
16358 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
16359 * internal-fn.c (internal_fn_mask_index): Handle conditional
16360 internal functions.
16361 (vectorized_internal_fn_supported_p): New function.
16362 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
16363 (any_pred_load_store): Replace with...
16364 (need_to_predicate): ...this new variable.
16365 (redundant_ssa_names): New variable.
16366 (ifcvt_can_use_mask_load_store): Move initial checks to...
16367 (ifcvt_can_predicate): ...this new function. Handle tree codes
16368 for which a conditional internal function exists.
16369 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
16370 instead of ifcvt_can_use_mask_load_store. Update after variable
16372 (predicate_load_or_store): New function, split out from
16373 predicate_mem_writes.
16374 (check_redundant_cond_expr): New function.
16375 (value_available_p): Likewise.
16376 (predicate_rhs_code): Likewise.
16377 (predicate_mem_writes): Rename to...
16378 (predicate_statements): ...this. Use predicate_load_or_store
16379 and predicate_rhs_code.
16380 (combine_blocks, tree_if_conversion): Update after above name changes.
16381 (ifcvt_local_dce): Handle redundant_ssa_names.
16382 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
16383 general conditional functions.
16384 * tree-vect-stmts.c (vectorizable_call): Likewise.
16386 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
16387 Alan Hayward <alan.hayward@arm.com>
16388 David Sherwood <david.sherwood@arm.com>
16390 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
16391 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
16392 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
16393 plus and minus and convert them into IFN_COND_FMA-based sequences.
16394 (convert_mult_to_fma): Handle conditional plus and minus.
16396 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
16398 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
16399 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
16400 (cond_fnms_optab): New optabs.
16401 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
16402 internal functions.
16403 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
16404 * internal-fn.h (get_conditional_internal_fn): Declare.
16405 (get_unconditional_internal_fn): Likewise.
16406 * internal-fn.c (cond_ternary_direct): New macro.
16407 (expand_cond_ternary_optab_fn): Likewise.
16408 (direct_cond_ternary_optab_supported_p): Likewise.
16409 (FOR_EACH_COND_FN_PAIR): Likewise.
16410 (get_conditional_internal_fn): New function.
16411 (get_unconditional_internal_fn): Likewise.
16412 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
16413 (gimple_match_op::gimple_match_op): Add a new overload for 5
16415 (gimple_match_op::set_op): Likewise.
16416 (gimple_resimplify5): Declare.
16417 * genmatch.c (decision_tree::gen): Generate simplifications for
16419 * gimple-match-head.c (gimple_simplify): Define an overload for
16420 5 operands. Handle calls with 5 arguments in the top-level overload.
16421 (convert_conditional_op): Handle conversions from unconditional
16422 internal functions to conditional ones.
16423 (gimple_resimplify5): New function.
16424 (build_call_internal): Pass a fifth operand.
16425 (maybe_push_res_to_seq): Likewise.
16426 (try_conditional_simplification): Try converting conditional
16427 internal functions to unconditional internal functions.
16428 Handle 3-operand unconditional forms.
16429 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
16430 Define ternary equivalents of the current rules for binary conditional
16431 internal functions.
16432 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
16433 ternary operations.
16434 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
16435 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
16436 (optab): Handle them.
16437 (SVE_COND_FP_TERNARY): New int iterator.
16438 (sve_fmla_op, sve_fmad_op): New int attributes.
16439 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
16440 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
16441 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
16443 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
16445 * target.def (preferred_else_value): New target hook.
16446 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
16447 * doc/tm.texi: Regenerate.
16448 * targhooks.h (default_preferred_else_value): Declare.
16449 * targhooks.c (default_preferred_else_value): New function.
16450 * internal-fn.h (conditional_internal_fn_code): Declare.
16451 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
16452 (get_conditional_internal_fn): Use it.
16453 (conditional_internal_fn_code): New function.
16454 * gimple-match.h (gimple_match_cond): New struct.
16455 (gimple_match_op): Add a cond member function.
16456 (gimple_match_op::gimple_match_op): Update all forms to take a
16458 * genmatch.c (expr::gen_transform): Use the same condition as res_op
16459 for the suboperation, but don't specify a particular else_value.
16460 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
16461 (visit_nary_op, visit_reference_op_load): Pass
16462 gimple_match_cond::UNCOND to the gimple_match_op constructor.
16463 * gimple-match-head.c: Include tree-eh.h
16464 (convert_conditional_op): New function.
16465 (maybe_resimplify_conditional_op): Likewise.
16466 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
16467 (gimple_resimplify2): Likewise.
16468 (gimple_resimplify3): Likewise.
16469 (gimple_resimplify4): Likewise.
16470 (maybe_push_res_to_seq): Return null for conditional operations.
16471 (try_conditional_simplification): New function.
16472 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
16474 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
16476 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
16478 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
16480 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
16482 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
16484 (hash_tree): Do not hash DECL_FCONTEXT
16485 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
16486 Do not stream DECL_FCONTEXT.
16487 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
16488 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
16490 2018-07-12 Richard Biener <rguenther@suse.de>
16493 * dwarf2out.c (gen_block_die): Only output blocks when they have
16494 at least one !DECL_IGNORED_P variable.
16496 2018-07-12 Richard Biener <rguenther@suse.de>
16499 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
16500 Remove -mieee-fp handling.
16502 2018-07-12 Richard Biener <rguenther@suse.de>
16504 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
16505 left-over from last patch.
16507 2018-07-12 Jakub Jelinek <jakub@redhat.com>
16509 PR tree-optimization/86492
16510 * gimple-ssa-store-merging.c
16511 (imm_store_chain_info::coalesce_immediate_stores): Call
16512 check_no_overlap even for the merge_overlapping case. Formatting fix.
16514 2018-07-12 Richard Biener <rguenther@suse.de>
16516 PR middle-end/86479
16517 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
16518 move possibly trapping operations into the conditional.
16520 2018-07-12 Richard Biener <rguenther@suse.de>
16522 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
16523 (vn_lookup_simplify_result): Remove recursion limit applied
16525 (vn_nary_build_or_lookup_1): Adjust.
16526 (try_to_simplify): Likewise.
16527 * gimple-match-head.c (gimple_resimplify1): Instead apply one
16529 (gimple_resimplify2): Likewise.
16530 (gimple_resimplify3): Likewise.
16531 (gimple_resimplify4): Likewise.
16533 2018-07-11 Jakub Jelinek <jakub@redhat.com>
16535 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
16536 Use __mmask64 type instead of __mmask8 for __M argument.
16537 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
16538 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
16540 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
16541 __mmask16 for __M argument.
16542 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
16543 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
16544 to __mmask16 instead of __mmask8.
16545 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
16546 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
16547 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
16548 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
16549 instead of __mmask16 for __U argument.
16550 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
16551 __mmask16 instead of __mmask8 for __U argument.
16552 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
16554 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
16556 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
16558 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
16560 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
16562 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
16564 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
16566 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
16568 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
16570 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
16572 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
16573 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
16574 return type as well as __M argument type and all casts from __mmask8
16576 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
16577 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
16578 return type as well as __M argument type and all casts from __mmask8
16580 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
16581 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
16582 return type as well as __M argument type and all casts from __mmask8
16584 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
16585 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
16586 return type as well as __M argument type and all casts from __mmask8
16588 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
16589 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
16592 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
16594 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
16595 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
16598 2018-07-11 Paul Koning <ni1d@arrl.net>
16600 * doc/md.texi (define_subst): Document how multiple occurrences of
16601 the same argument in the replacement pattern are handled.
16603 2018-07-11 Paul Koning <ni1d@arrl.net>
16605 * doc/extend.texi (Common Variable Attributes): Move "mode" into
16606 alphabetical order.
16607 (Common Type Attributes): Add "mode" attribute.
16609 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
16611 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
16612 stream DECL_ORIGINAL_TYPE.
16613 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
16614 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
16615 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
16616 Do not walk original type.
16617 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
16619 (write_ts_decl_non_common_tree_pointers): Do not stream
16621 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
16622 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
16624 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
16626 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
16627 thread twice from the same starting edge.
16629 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
16631 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
16632 * gimple.c (gimple_call_nonnull_result_p): ...here...
16633 (gimple_call_nonnull_arg): ...and here.
16634 * gimple.h (gimple_call_nonnull_result_p): New.
16635 (gimple_call_nonnull_arg): New.
16637 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
16639 * config/arm/arm-cpus.in: Move information from fpu field of each
16640 cpu definition to the isa field.
16641 * config/arm/parsecpu.awk (fpu): Delete match rule.
16642 (gen_comm_data): Don't add bits from the CPU's FPU entry.
16644 2018-07-11 Richard Biener <rguenther@suse.de>
16647 * dwarf2out.c (init_sections_and_labels): Use
16648 output_asm_line_debug_info consistently.
16649 (dwarf2out_early_finish): Likewise.
16650 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
16653 2018-07-11 Richard Biener <rguenther@suse.de>
16655 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
16656 Rework father_bb setting in a way to avoid propagating constants
16657 multiple times on a loop body.
16659 2018-07-10 Mark Wielaard <mark@klomp.org>
16662 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
16664 2018-07-10 Richard Biener <rguenther@suse.de>
16666 * hash-map.h (hash_map::iterator::operator*): Return
16667 references to key and value.
16669 2018-07-10 Jakub Jelinek <jakub@redhat.com>
16672 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
16673 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
16674 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
16675 (gimplify_omp_for): For composite loops, move outer
16676 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
16677 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
16678 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
16679 TREE_LIST for both the original class iterator and the "last" helper
16680 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
16681 loop, remember has_decl_expr from outer composite loops for the
16682 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
16684 2018-07-09 Martin Sebor <msebor@redhat.com>
16686 PR middle-end/77357
16687 PR middle-end/86428
16688 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
16689 accessing implicitly initialized array elements.
16690 * expr.c (string_constant): Handle string initializers of
16691 character arrays within aggregates.
16692 * gimple-fold.c (fold_array_ctor_reference): Add argument.
16693 Store element offset. As a special case, handle zero size.
16694 (fold_nonarray_ctor_reference): Same.
16695 (fold_ctor_reference): Add argument. Store subobject offset.
16696 * gimple-fold.h (fold_ctor_reference): Add argument.
16698 2018-07-09 Paul Koning <ni1d@arrl.net>
16700 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
16701 (pdp11_insn_cost): New function.
16702 (pdp11_md_asm_adjust): New function.
16703 (TARGET_INVALID_WITHIN_DOLOOP): Define.
16704 (pdp11_rtx_costs): Update to match machine better.
16705 (output_addr_const_pdp11): Correct format mismatch warnings.
16706 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
16707 * config/pdp11/pdp11.md: General change to add base_cost and/or
16708 length attributes for use by new pdp11_insn_cost function.
16709 (MIN_BRANCH): Correct definition.
16711 (doloop_end): Use standard pattern name for looping pattern.
16712 (doloop_end_nocc): New.
16713 (movsf): Add another constraint alternative.
16714 (zero_extendqihi2): Add constraint alternatives for not in place
16716 (zero_extendhisi2): Remove.
16717 (shift patterns): Add CC handling variants.
16721 (define_peephole2): New peephole to recognize mov that sets CC for
16724 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
16726 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
16727 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
16730 2018-07-09 Jakub Jelinek <jakub@redhat.com>
16733 * real.c (real_nextafter): Return true if result is denormal.
16735 2018-07-09 Martin Liska <mliska@suse.cz>
16737 * common.opt: Add back wrongly removed attribute.
16739 2018-07-09 Richard Biener <rguenther@suse.de>
16742 * dwarf2out.c (gen_block_die): For an early generated DIE
16743 always output high/low PC attributes.
16745 2018-07-09 Tom de Vries <tdevries@suse.de>
16747 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
16748 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
16749 onto VAR_DECL with abstract origin.
16751 2018-07-07 Jim Wilson <jimw@sifive.com>
16753 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
16755 2018-07-07 Tom de Vries <tdevries@suse.de>
16757 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
16759 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
16761 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
16763 (combine_bound): Use wide-int overflow calculation instead of
16765 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
16767 * fold-const.c (int_const_binop_2): Same.
16768 (extract_muldiv_1): Same.
16769 (fold_div_compare): Same.
16770 (fold_abs_const): Same.
16772 * poly-int.h (add): Same.
16776 * predict.c (predict_iv_comparison): Same.
16777 * profile-count.c (slow_safe_scale_64bit): Same.
16778 * simplify-rtx.c (simplify_const_binary_operation): Same.
16779 * tree-chrec.c (tree_fold_binomial): Same.
16780 * tree-data-ref.c (split_constant_offset_1): Same.
16781 * tree-if-conv.c (idx_within_array_bound): Same.
16782 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
16783 * tree-ssa-phiopt.c (minmax_replacement): Same.
16784 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
16785 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
16786 * vr-values.c (vr_values::adjust_range_with_scev): Same.
16787 * wide-int.cc (wi::add_large): Same.
16788 (wi::mul_internal): Same.
16789 (wi::sub_large): Same.
16790 (wi::divmod_internal): Same.
16791 * wide-int.h: Change overflow type to overflow_type for neg, add,
16792 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
16793 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
16794 mul_internal, divmod_internal.
16795 (overflow_type): New enum.
16796 (accumulate_overflow): New.
16798 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
16800 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
16801 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
16803 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
16805 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
16806 argument is checked for zero before entering loop, avoid checking again.
16808 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
16810 * gimplify.h (generic_expr_could_trap_p): Set as global function.
16811 * gimplify.h (generic_expr_could_trap_p): Likwise.
16812 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
16814 2018-07-06 Jakub Jelinek <jakub@redhat.com>
16816 PR tree-optimization/86401
16817 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
16818 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
16819 (fold_bit_and_mask): ... here. New helper function for match.pd.
16820 * fold-const.h (fold_bit_and_mask): Declare.
16821 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
16823 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
16826 * target.def (translate_mode_attribute): New hook.
16827 * targhooks.h (default_translate_mode_attribute): Declare.
16828 * targhooks.c (default_translate_mode_attribute): New function.
16829 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
16830 * doc/tm.texi: Regenerate.
16831 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
16832 (rs6000_translate_mode_attribute): New function.
16834 2018-07-06 Paul Koning <ni1d@arrl.net>
16836 * doc/md.texi (define_split): Document DONE and FAIL.
16837 (define_peephole2): Ditto.
16839 2018-07-05 Jeff Law <law@redhat.com>
16841 PR tree-optimization/86010
16842 * tree-ssa-dse.c (compute_trims): More aggressively trim at
16843 both the head and tail of mem* and str* calls.
16845 2018-07-05 Jim Wilson <jimw@sifive.com>
16847 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
16849 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
16851 * config/aarch64/aarch64-simd.md: correct flags text for
16852 MIN_EXPR replacement.
16854 2018-07-05 James Clarke <jrtc27@jrtc27.com>
16856 * configure: Regenerated.
16858 2018-07-05 Carl Love <cel@us.ibm.com>
16860 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
16861 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
16862 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
16863 VSX_BUILTIN_DOUBLEL_V4SF.
16865 2018-07-05 Martin Sebor <msebor@redhat.com>
16868 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
16869 than its domain to compute its the upper bound of a char array.
16871 2018-07-05 Nathan Sidwell <nathan@acm.org>
16873 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
16874 * doc/cpp.texi: Update comment.
16875 * doc/tm.texi: Rebuilt.
16876 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
16877 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
16878 * doc/extend.texi (Backwards Compatibility): Clarify it is system
16879 headers affected by extern "C".
16880 * system.h: Poison NO_IMPLICIT_EXTERN_C.
16881 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
16882 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
16883 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
16884 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
16885 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
16886 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
16887 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
16888 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
16889 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
16890 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
16891 config/sparc/sp64-elf.h, config/spu/spu.h,
16892 config/stormy16/stormy16.h, config/v850/v850.h,
16893 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
16894 define NO_IMPLICIT_EXTERN_C.
16895 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
16897 2018-07-05 Tamar Christina <tamar.christina@arm.com>
16900 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
16901 instead of GET_MODE_SIZE when comparing Units.
16903 2018-07-05 Tamar Christina <tamar.christina@arm.com>
16906 * rtlanal.c (set_noop_p): Constrain on mode change,
16907 include hard-reg-set.h
16909 2018-07-05 Tamar Christina <tamar.christina@arm.com>
16911 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
16913 2018-07-05 Jakub Jelinek <jakub@redhat.com>
16916 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
16919 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
16921 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
16923 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
16926 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
16928 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
16930 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
16932 PR middle-end/86380
16933 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
16935 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
16937 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
16938 neg_*_op* variables.
16940 2018-07-04 Martin Liska <mliska@suse.cz>
16942 * tree-switch-conversion.c: Define
16943 max_ratio_for_speed and max_ratio_for_size constants.
16945 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
16946 Martin Liska <mliska@suse.cz>
16948 PR middle-end/66240
16951 * common.opt: Rename align options with 'str_' prefix.
16952 * common/config/i386/i386-common.c (set_malign_value): New
16954 (ix86_handle_option): Use it to set -falign-* options/
16955 * config/aarch64/aarch64-protos.h (struct tune_params): Change
16956 type from int to string.
16957 * config/aarch64/aarch64.c: Update default values from int
16959 * config/alpha/alpha.c (alpha_override_options_after_change):
16961 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
16962 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
16963 max skip conditionally.
16964 * config/i386/freebsd.h (SUBALIGN_LOG): New.
16965 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
16966 max skip conditionally.
16967 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
16968 max skip conditionally.
16969 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
16970 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
16971 max skip conditionally.
16972 * config/i386/i386.c (struct ptt): Change type from int to
16974 (ix86_default_align): Set default values.
16975 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
16976 max skip conditionally.
16977 * config/i386/iamcu.h (SUBALIGN_LOG): New.
16978 (ASM_OUTPUT_MAX_SKIP_ALIGN):
16979 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
16980 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
16981 max skip conditionally.
16982 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
16983 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
16984 * config/i386/x86-64.h (SUBALIGN_LOG): New.
16985 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
16986 max skip conditionally.
16987 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
16988 * config/ia64/ia64.c (ia64_option_override): Set default values
16989 for alignment options.
16990 * config/m68k/m68k.c: Handle new str_align_* options.
16991 * config/mips/mips.c (mips_set_compression_mode): Change
16993 (mips_option_override): Set default values for options.
16994 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
16996 * config/rs6000/rs6000.c (rs6000_option_override_internal):
16998 * config/rx/rx.c (rx_option_override): Likewise.
16999 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
17000 (LABEL_ALIGN): Use align_labels_log.
17001 (LOOP_ALIGN): Use align_loops_align.
17002 * config/s390/s390.c (s390_asm_output_function_label): Use new
17004 * config/sh/sh.c (sh_override_options_after_change):
17005 Change type of constants.
17006 * config/spu/spu.c (spu_sched_init): Likewise.
17007 * config/sparc/sparc.c (sparc_option_override): Set default
17008 values for options.
17009 * config/visium/visium.c (visium_option_override): Likewise.
17010 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
17011 emit p2align format with last argument if it's not needed.
17012 * doc/invoke.texi: Document extended format of -falign-*.
17013 * final.c: Use align_labels alignment.
17014 * flags.h (struct target_flag_state): Change type to use
17016 (struct align_flags_tuple): New.
17017 (struct align_flags): Likewise.
17018 (align_loops_log): Redefine macro to use new types.
17019 (align_loops_max_skip): Redefine macro to use new types.
17020 (align_jumps_log): Redefine macro to use new types.
17021 (align_jumps_max_skip): Redefine macro to use new types.
17022 (align_labels_log): Redefine macro to use new types.
17023 (align_labels_max_skip): Redefine macro to use new types.
17024 (align_functions_log): Redefine macro to use new types.
17025 (align_loops): Redefine macro to use new types.
17026 (align_jumps): Redefine macro to use new types.
17027 (align_labels): Redefine macro to use new types.
17028 (align_functions): Redefine macro to use new types.
17029 (align_functions_max_skip): Redefine macro to use new types.
17030 (align_loops_value): New macro.
17031 (align_jumps_value): New macro.
17032 (align_labels_value): New macro.
17033 (align_functions_value): New macro.
17034 * function.c (invoke_set_current_function_hook): Propagate
17035 alignment values from flags to global variables default in
17037 * ipa-icf.c (sem_function::equals_wpa): Use
17038 cl_optimization_option_eq instead of memcmp.
17039 * lto-streamer.h (cl_optimization_stream_out): Support streaming
17041 (cl_optimization_stream_in): Likewise.
17042 * optc-save-gen.awk: Support strings in cl_optimization.
17043 * opth-gen.awk: Likewise.
17044 * opts.c (finish_options): Remove error checking of invalid
17046 (MAX_CODE_ALIGN): Remove.
17047 (MAX_CODE_ALIGN_VALUE): Likewise.
17048 (parse_and_check_align_values): New function.
17049 (check_alignment_argument): Likewise.
17050 (common_handle_option): Use check_alignment_argument.
17051 * opts.h (parse_and_check_align_values): Declare.
17052 * toplev.c (init_alignments): Remove.
17053 (read_log_maxskip): New.
17054 (parse_N_M): Likewise.
17055 (parse_alignment_opts): Likewise.
17056 (backend_init_target): Remove usage of init_alignments.
17057 * toplev.h (parse_alignment_opts): Declare.
17058 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
17060 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
17061 * tree.c (cl_option_hasher::equal): New.
17062 * varasm.c: Use new global macros.
17064 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
17066 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
17067 Use a simpler align directive also if MAXSKIP = ALIGN-1.
17068 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17069 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17070 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17071 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
17072 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
17073 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
17074 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
17075 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17076 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17077 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17078 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17080 2018-07-04 Martin Liska <mliska@suse.cz>
17081 Jonathan Wakely <jwakely@redhat.com>
17083 * coverage.c: Use correct type.
17084 * doc/invoke.texi: Language correction.
17086 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
17089 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
17090 ENDBRANCH for non-tail call which may return via indirect branch.
17091 * doc/extend.texi: Document indirect_return attribute.
17093 2018-07-03 Martin Sebor <msebor@redhat.com>
17095 PR tree-optimization/86274
17096 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
17098 (format_floating): Correct handling of infinities and NaNs.
17100 2018-07-03 Martin Sebor <msebor@redhat.com>
17102 * print-tree.c (print_real_cst): New function.
17103 (print_node_brief): Call it.
17104 (print_node): Ditto.
17106 2018-07-03 Jeff Law <law@redhat.com>
17108 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
17109 into a single pattern.
17111 * config/h8300/h8300.md (ors code_iterator): New.
17112 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
17113 a single pattern and single splitter.
17114 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
17115 (iorqi3_1, xorqi3_1): Likewise.
17116 (iorqi3, xorqi3 expanders): Similarly.
17118 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
17119 (movmd_internal) into a single pattern using the P mode iterator.
17120 (movmd splitters): Similarly.
17121 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
17122 (movsd splitters): Similarly.
17124 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
17125 ADDB, ADDW and ADDL into a single ADD attribute which selects the
17126 right table based on the size of the operand.
17127 * config/h8300/h8300.md (length_table): Corresponding changes. All
17128 references to "addb", "addw" and "addl" changed to "add".
17129 (btst patterns): Merge two variants into a single pattern.
17130 (tstqi, tsthi): Likewise.
17131 (addhi3_incdec, addsi3_incdec): Likewise.
17132 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
17133 (mulhi3, mulsi3): Likewise.
17134 (udivhi3, udivsi3): Likewise.
17135 (divhi3, divsi3): Likewise.
17136 (andorqi3, andorhi3, andorsi3): Likewise.
17138 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
17141 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
17142 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
17144 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
17146 PR tree-optimization/85694
17147 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
17148 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
17150 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
17151 (<u>avg<mode>3_ceil): New patterns.
17153 2018-07-03 David Malcolm <dmalcolm@redhat.com>
17155 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
17156 scan-tree-dump directive.
17157 * gcc.dg/vect/slp-perm-2.c: Likewise.
17158 * gcc.dg/vect/slp-perm-3.c: Likewise.
17159 * gcc.dg/vect/slp-perm-5.c: Likewise.
17160 * gcc.dg/vect/slp-perm-6.c: Likewise.
17161 * gcc.dg/vect/slp-perm-7.c: Likewise.
17162 * gcc.dg/vect/slp-perm-8.c: Likewise.
17164 2018-07-03 Marek Polacek <polacek@redhat.com>
17166 PR middle-end/86202
17167 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
17169 2018-07-03 Richard Biener <rguenther@suse.de>
17172 * tree-ssa-structalias.c (find_func_clobbers): Properly
17173 handle indirect calls.
17175 2018-07-03 Jeff Law <law@redhat.com>
17177 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
17178 (shifts): New code iterator.
17179 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
17180 expander. Fix HImode handling on H8/SX.
17181 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
17182 (subqi3, subhi3, subsi3 expanders): Likewise.
17183 (andqi3, andhi3, andsi3 expanders): Likewise.
17184 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
17185 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
17186 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
17187 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
17188 (zero_extendqihi2, zero_extendqisi2): Likewise.
17189 (extendqihi2, extendqisi2): Likewise.
17190 (rotlqi3, rotlhi3, rotlsi3): Likewise.
17191 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
17192 (rotlqi3_1, rotlhi3_1): Likewise.
17193 (logicalhi3_sn, logicalsi3_sn): Likewise.
17194 (logicalhi3, logicalsi3): Likewise.
17196 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
17198 * tree-vect-patterns.c (vect_recog_rotate_pattern)
17199 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
17200 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
17201 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
17202 type to append_pattern_def_seq instead of creating a stmt_vec_info
17204 (build_mask_conversion): Likewise. Remove vinfo argument.
17205 (vect_add_conversion_to_patterm): Likewise, renaming to...
17206 (vect_add_conversion_to_pattern): ...this.
17207 (vect_recog_mask_conversion_pattern): Update call to
17208 build_mask_conversion. Pass the vector type to
17209 append_pattern_def_seq here too.
17210 (vect_recog_gather_scatter_pattern): Update call to
17211 vect_add_conversion_to_pattern.
17213 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
17215 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
17216 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
17217 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
17218 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
17219 STMT_VINFO_PATTERN_DEF_SEQ to null here.
17220 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
17221 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
17222 append_pattern_def_seq instead of new_pattern_def_seq.
17223 (vect_recog_divmod_pattern): Do both of the above.
17224 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
17227 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
17229 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
17230 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
17231 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
17232 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
17233 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
17234 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
17235 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
17236 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
17237 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
17238 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
17239 parameter with a single stmt_vec_info.
17240 (vect_recog_func_ptr): Likewise.
17241 (vect_recog_gather_scatter_pattern): Likewise, folding in...
17242 (vect_try_gather_scatter_pattern): ...this.
17243 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
17244 the stmt_vec_info of the statement to be matched. Don't clear
17245 STMT_VINFO_RELATED_STMT.
17246 (vect_pattern_recog): Update call accordingly.
17248 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
17250 PR tree-optimization/85694
17251 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
17252 (uavgM3_ceil): Document new optabs.
17253 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
17254 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
17256 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
17257 (savg_ceil_optab): New optabs.
17258 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
17259 (vect_vect_recog_func_ptrs): Add it.
17260 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
17261 constant directly from the associated lhs.
17263 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
17265 * tree-vect-patterns.c (vect_split_statement): New function.
17266 (vect_convert_input): Use it to try to split an existing cast.
17268 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
17270 * poly-int.h (print_hex): New function.
17271 * dumpfile.h (dump_dec, dump_hex): Declare.
17272 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
17273 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
17274 min_input_precision, operation_precision and operation_sign.
17275 * tree-vect-patterns.c (vect_get_range_info): New function.
17276 (vect_same_loop_or_bb_p, vect_single_imm_use)
17277 (vect_operation_fits_smaller_type): Delete.
17278 (vect_look_through_possible_promotion): Add an optional
17279 single_use_p parameter.
17280 (vect_recog_over_widening_pattern): Rewrite to use new
17281 stmt_vec_info infomration. Handle one operation at a time.
17282 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
17283 (vect_truncatable_operation_p, vect_set_operation_type)
17284 (vect_set_min_input_precision): New functions.
17285 (vect_determine_min_output_precision_1): Likewise.
17286 (vect_determine_min_output_precision): Likewise.
17287 (vect_determine_precisions_from_range): Likewise.
17288 (vect_determine_precisions_from_users): Likewise.
17289 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
17290 (vect_vect_recog_func_ptrs): Put over_widening first.
17292 (vect_pattern_recog): Call vect_determine_precisions.
17294 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
17296 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
17297 statements that have been replaced by further pattern statements.
17298 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
17300 2018-07-03 Richard Biener <rguenther@suse.de>
17302 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
17303 always set *dt. Dump vectype in vectype overload.
17304 * dumpfile.h (dump_gimple_expr): New function.
17305 (dump_gimple_expr_loc): Likewise.
17306 * dumpfile.c (dump_gimple_expr): New function.
17307 (dump_gimple_expr_loc): Likewise.
17309 2018-07-02 Jeff Law <law@redhat.com>
17311 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
17312 the H8/300, H8/300H and H8/S variants into a single pattern.
17313 (movhi_h8300, movqi_h8300hs): Similarly.
17314 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
17315 (QHI mode iterator): New.
17317 * config/h8300/h8300.md: Remove trailing whitespace.
17319 2018-07-02 Jim Wilson <jimw@sifive.com>
17321 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
17322 instead of emit_insn for interrupt returns.
17323 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
17324 (riscv_sret, riscv_uret): Likewise.
17326 2018-07-02 David Malcolm <dmalcolm@redhat.com>
17328 * pretty-print.c (selftest::test_pp_format): Move save and restore
17329 of quotes to class auto_fix_quotes, and add an instance.
17330 * selftest.c: Include "intl.h".
17331 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
17332 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
17333 * selftest.h (selftest::auto_fix_quotes): New class.
17335 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
17337 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
17338 (aarch64_sve_prepare_conditional_op): Remove.
17339 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
17340 Allow aarch64_simd_reg_or_zero as select operand; remove
17341 the aarch64_sve_prepare_conditional_op call.
17342 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
17343 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
17344 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
17345 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
17346 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
17347 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
17348 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
17349 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
17350 and a splitters to match all of the *_any patterns.
17351 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
17353 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
17354 (SVE_COND_FP_BINARY_REV): Remove.
17355 (sve_int_op_rev, sve_fp_op_rev): New.
17356 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
17357 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
17358 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
17359 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
17360 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
17361 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
17362 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
17363 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
17364 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
17366 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
17367 Remove match_dup 1 from the inner unspec.
17368 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
17370 * config/aarch64/aarch64.md (movprfx): New attr.
17371 (length): Default movprfx to 8.
17372 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
17373 (*madd<SVE_I>, *msub<SVE_I): Likewise.
17374 (*<su>mul<SVE_I>3_highpart): Likewise.
17375 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
17376 (*v<ASHIFT><SVE_I>3): Likewise.
17377 (*<su><MAXMIN><SVE_I>3): Likewise.
17378 (*<su><MAXMIN><SVE_F>3): Likewise.
17379 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
17380 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
17381 (*div<SVE_F>4): Likewise.
17383 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
17385 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
17388 2018-07-02 Richard Biener <rguenther@suse.de>
17390 PR tree-optimization/86363
17391 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
17392 memset argument refers to a non-variable address.
17394 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
17396 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
17397 {PLUS,MINUS}_EXPR code to...
17398 (adjust_symbolic_bound): ...here,
17399 (combine_bound): ...here,
17400 (set_value_range_with_overflow): ...and here.
17402 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
17404 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
17406 (extract_range_from_abs_expr): ...here.
17408 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
17410 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
17411 -fno-omit-frame-pointer when not optimizing.
17413 2018-07-02 Martin Liska <mliska@suse.cz>
17416 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
17417 (propagate_nothrow): Likewise.
17419 2018-07-02 Martin Liska <mliska@suse.cz>
17422 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
17424 2018-07-02 David Malcolm <dmalcolm@redhat.com>
17426 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
17427 function in r262149, changing "loc" param from source_location to
17428 const dump_location_t &.
17429 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
17430 declaration, as above.
17432 2018-07-01 Paul Koning <ni1d@arrl.net>
17434 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
17435 -munit-asm, -mgnu-asm, -mdec-asm.
17436 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
17437 (pdp11_output_labelref): New.
17438 (pdp11_output_def): New.
17439 (pdp11_output_addr_vec_elt): New.
17440 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
17441 %# and %@ format codes.
17442 (pdp11_option_override): New.
17443 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
17444 (pdp11_output_ident): New.
17445 (pdp11_asm_named_section): New.
17446 (pdp11_asm_init_sections): New.
17447 (pdp11_file_start): New.
17448 (pdp11_file_end): New.
17449 (output_ascii): Use .ascii/.asciz for -mdec-asm.
17450 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
17451 %o, like %c but octal.
17452 (pdp11_option_override): New.
17453 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
17455 (DATA_SECTION_ASM_OP): Ditto.
17456 (READONLY_DATA_SECTION_ASM_OP): New.
17457 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
17458 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
17459 (ASM_OUTPUT_LABELREF): Ditto.
17460 (ASM_OUTPUT_DEF): Ditto.
17461 (ASM_OUTPUT_EXTERNAL): New.
17462 (ASM_OUTPUT_SOURCE_FILENAME): New.
17463 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
17464 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
17465 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
17466 %# and %@ format codes.
17467 * config/pdp11/pdp11.opt (mgnu-asm): New.
17468 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
17469 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
17470 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
17472 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
17474 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
17475 dereferencing path[] beyond its length.
17477 (debug_all_paths): New.
17478 (rewire_first_differing_edge): New.
17479 (adjust_paths_after_duplication): New.
17480 (duplicate_thread_path): Call adjust_paths_after_duplication.
17482 (thread_through_all_blocks): Add new argument to
17483 duplicate_thread_path.
17485 2018-06-30 Jim Wilson <jimw@sifive.com>
17487 * config/riscv/predicates.md (p2m1_shift_operand): New.
17488 (high_mask_shift_operand): New.
17489 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
17490 pattern using p2m1_shift_operand.
17491 (lshsi3_zero_extend_3+2): New combiner pattern using
17492 high_mask_shift_operand.
17494 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
17496 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
17498 (vect_recog_rotate_pattern): ...here.
17499 (vect_convert_input): Try to insert casts of invariants in the
17501 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
17502 preheader to be empty.
17504 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
17506 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
17507 vector type. If given, install it in the new statement's
17508 STMT_VINFO_VECTYPE.
17509 (vect_element_precision): New function.
17510 (vect_unpromoted_value): New struct.
17511 (vect_unpromoted_value::vect_unpromoted_value): New function.
17512 (vect_unpromoted_value::set_op): Likewise.
17513 (vect_look_through_possible_promotion): Likewise.
17514 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
17515 (vect_widened_op_tree, vect_convert_input): Likewise.
17516 (vect_convert_inputs, vect_convert_output): Likewise.
17517 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
17518 to handle the optional cast of the multiplication result and
17519 vect_widened_op_tree to detect the widened multiplication itself.
17520 Do not require the input and output of promotion casts to have
17521 the same sign, but base the signedness of the operation on the
17522 input rather than the result. If the pattern includes two
17523 promotions, check that those promotions have the same sign.
17524 Do not restrict the MULT_EXPR handling to a double-width result;
17525 handle quadruple-width results and wider. Use vect_convert_inputs
17526 to convert the inputs to the common type.
17527 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
17528 to handle the optional cast of the ABS result. Also allow a sign
17529 change or a sign extension between the ABS and MINUS.
17530 Use vect_widened_op_tree to detect the widened subtraction and use
17531 vect_convert_inputs to convert the inputs to the common type.
17532 (vect_handle_widen_op_by_const): Delete.
17533 (vect_recog_widen_op_pattern): New function.
17534 (vect_recog_widen_mult_pattern): Use it.
17535 (vect_recog_widen_shift_pattern): Likewise.
17536 (vect_recog_widen_sum_pattern): Use
17537 vect_look_through_possible_promotion to handle the promoted
17540 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
17542 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
17543 the containing gimple_seq *.
17544 * gimple-iterator.h (gsi_for_stmt): Declare it.
17545 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
17546 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
17547 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
17548 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
17549 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
17551 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
17553 (vect_mark_pattern_stmts): ...here. Handle cases in which the
17554 statement being replaced is part of an existing pattern
17555 definition sequence, inserting the new pattern statements before
17557 (vect_pattern_recog_1): Don't return a bool. If the statement
17558 is already part of a pattern, instead apply pattern matching
17559 to the pattern definition statements. Don't clear the
17560 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
17561 (vect_pattern_recog): Don't break after the first match;
17562 continue processing the pattern definition statements instead.
17563 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
17565 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
17567 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
17568 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
17569 (vect_recog_widen_sum_pattern): Use it.
17571 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
17573 * tree-vect-loop.c (vectorizable_reduction): Assert that the
17574 phi is not a pattern statement and has not been replaced by
17575 a pattern statement.
17576 * tree-vect-patterns.c (type_conversion_p): Don't check
17577 STMT_VINFO_IN_PATTERN_P.
17578 (vect_recog_vector_vector_shift_pattern): Likewise.
17579 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
17580 the pattern statement rather than the original statement; check
17581 directly for a WIDEN_MULT_EXPR here.
17582 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
17583 vect_is_simple_use to return the pattern statement rather
17584 than the original statement; use is_pattern_stmt_p to check
17585 for such a pattern statement.
17586 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
17587 to return the pattern statement rather than the original statement;
17588 don't do the same transformation here.
17589 (vect_is_simple_use): If the defining statement has been replaced
17590 by a pattern statement, return the pattern statement instead.
17591 Remove the corresponding (local) transformation from the vectype
17594 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
17596 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
17597 end and default to null.
17598 * tree-vect-loop.c (vect_create_epilog_for_reduction)
17599 (vectorizable_reduction): Update calls accordingly, dropping the
17600 gimple ** argument if the passed-back statement isn't needed.
17601 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
17602 (vect_recog_rotate_pattern): Likewise.
17603 (vect_recog_mask_conversion_pattern): Likewise.
17604 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
17605 (vect_mask_constant_operand_p): Likewise.
17606 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
17607 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
17608 (get_group_load_store_type, get_load_store_type): Likewise.
17609 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
17610 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
17611 (vectorizable_conversion, vectorizable_assignment): Likewise.
17612 (vectorizable_shift, vectorizable_operation): Likewise.
17613 (vectorizable_store, vect_is_simple_cond): Likewise.
17614 (vectorizable_condition, vectorizable_comparison): Likewise.
17615 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
17616 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
17617 and move it to the end. Cope with null def_stmt_outs.
17619 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
17621 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
17623 2018-06-29 Jeff Law <law@redhat.com>
17625 * config/v850/v850.c (v850_legitimate_address_p): Handle large
17626 displacements for TARGET_V850E2V3 and newer.
17627 (TARGET_LRA_P): Remove. Defaults to LRA now.
17628 * config/v850/v850.md (sign23byte_load): Remove.
17629 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
17630 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
17632 2018-06-29 Martin Liska <mliska@suse.cz>
17635 * coverage.c (coverage_init): Mangle full path name.
17636 * doc/invoke.texi: Document the change.
17637 * gcov-io.c (mangle_path): New.
17638 * gcov-io.h (mangle_path): Likewise.
17639 * gcov.c (mangle_name): Use mangle_path for path mangling.
17641 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17643 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
17644 if starting source register is not even.
17646 2018-06-29 Martin Liska <mliska@suse.cz>
17648 PR tree-optimization/86263
17649 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
17650 Make edge redirection.
17652 2018-06-29 David Malcolm <dmalcolm@redhat.com>
17654 * dumpfile.c (dump_loc): Add indentation based on scope depth.
17655 (dump_scope_depth): New variable.
17656 (get_dump_scope_depth): New function.
17657 (dump_begin_scope): New function.
17658 (dump_end_scope): New function.
17659 * dumpfile.h (get_dump_scope_depth): New declaration.
17660 (dump_begin_scope): New declaration.
17661 (dump_end_scope): New declaration.
17662 (class auto_dump_scope): New class.
17663 (AUTO_DUMP_SCOPE): New macro.
17664 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
17667 2018-06-29 Richard Biener <rguenther@suse.de>
17669 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
17670 compute_all_dependences succeeds.
17671 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
17672 exceed --param loop-max-datarefs-for-datadeps.
17674 2018-06-29 Jakub Jelinek <jakub@redhat.com>
17676 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
17678 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
17681 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
17682 alternative 0 in preferred_for_speed attribute.
17684 2018-06-28 Paul Koning <ni1d@arrl.net>
17686 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
17687 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
17688 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
17689 * config/pdp11/pdp11.md: Correct "length" attribute calculation
17690 for shift insn patterns.
17692 2018-06-28 David Malcolm <dmalcolm@redhat.com>
17694 * cgraph.c (cgraph_node::get_body): Replace assignments to
17695 "dump_file" with calls to set_dump_file.
17696 * dumpfile.c (alt_dump_file): Make static, and group with...
17697 (alt_flags): ...this definition.
17698 (dumps_are_enabled): New variable.
17699 (refresh_dumps_are_enabled): New function.
17700 (set_dump_file): New function.
17701 (set_alt_dump_file): New function.
17702 (gcc::dump_manager::dump_start): Replace assignments to
17703 "dump_file" and "alt_dump_file" with calls to set_dump_file and
17705 (gcc::dump_manager::dump_finish): Likewise.
17706 * dumpfile.h (alt_dump_file): Delete decl.
17707 (dumps_are_enabled): New variable decl.
17708 (set_dump_file): New function decl.
17709 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
17711 * tree-nested.c (lower_nested_functions): Replace assignments to
17712 "dump_file" with calls to set_dump_file.
17714 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
17716 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
17717 goto_locus of each outgoing edge of each basic block.
17719 2018-06-28 Richard Biener <rguenther@suse.de>
17721 * dwarf2out.c (decl_scope_table): Remove.
17722 (push_decl_scope): Likewise.
17723 (pop_decl_scope): Likewise.
17724 (gen_type_die_for_member): Do not call push/pop_decl_scope.
17725 (gen_struct_or_union_type_die): Likewise.
17726 (gen_tagged_type_die): Likewise.
17727 (dwarf2out_init): Do not initialize decl_scope_table.
17728 (dwarf2out_c_finalize): Do not free it.
17730 2018-06-28 Richard Biener <rguenther@suse.de>
17732 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
17733 deciding whether to not re-use a DIE.
17735 2018-06-28 Richard Biener <rguenther@suse.de>
17737 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
17738 DW_AT_abstract_origin attribute.
17740 2018-06-28 Martin Liska <mliska@suse.cz>
17742 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
17743 Use newly introduced constants.
17744 * tree-switch-conversion.h (struct jump_table_cluster):
17745 Define max_ratio_for_size and max_ratio_for_speed.
17747 2018-06-28 Martin Liska <mliska@suse.cz>
17749 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
17750 Add new checking assert to catch invalid state.
17751 (jump_table_cluster::can_be_handled): Handle single case
17753 (jump_table_cluster::is_beneficial): Bail out for such case.
17754 (bit_test_cluster::find_bit_tests):
17755 Add new checking assert to catch invalid state.
17756 (bit_test_cluster::can_be_handled): Handle single case
17758 (bit_test_cluster::is_beneficial): Bail out for such case.
17759 (switch_decision_tree::analyze_switch_statement):
17762 2018-06-28 Martin Liska <mliska@suse.cz>
17764 * common.opt: Introduce -completion option.
17765 * gcc.c (driver_handle_option): Handle it.
17766 (driver::main): Print completions if completion
17768 * opt-suggestions.c (option_proposer::get_completions):
17770 (option_proposer::suggest_completion): Likewise.
17771 (option_proposer::find_param_completions): Likewise.
17772 (verify_autocompletions): Likewise.
17773 (test_completion_valid_options): Likewise.
17774 (test_completion_valid_params): Likewise.
17775 (in_completion_p): Likewise.
17776 (empty_completion_p): Likewise.
17777 (test_completion_partial_match): Likewise.
17778 (test_completion_garbage): Likewise.
17779 (opt_proposer_c_tests): Likewise.
17780 * opt-suggestions.h: Declare new functions.
17781 * opts.c (common_handle_option): Handle OPT__completion_.
17782 * selftest-run-tests.c (selftest::run_tests): Add
17783 opt_proposer_c_tests.
17784 * selftest.c (assert_str_startswith): New.
17785 * selftest.h (assert_str_startswith): Likewise.
17786 (opt_proposer_c_tests): New.
17787 (ASSERT_STR_STARTSWITH): Likewise.
17789 2018-06-28 Martin Liska <mliska@suse.cz>
17791 * Makefile.in: Add opt-suggestions.o.
17792 * gcc-main.c: Include opt-suggestions.h.
17793 * gcc.c (driver::driver): Likewise.
17794 (driver::~driver): Remove m_option_suggestions.
17795 (driver::build_option_suggestions): Moved to option_proposer.
17796 (driver::suggest_option): Likewise.
17797 (driver::handle_unrecognized_options): Use option_proposer.
17798 * gcc.h (class driver): Add new memver m_option_proposer.
17799 * opt-suggestions.c: New file.
17800 * opt-suggestions.h: New file.
17802 2018-06-28 Martin Liska <mliska@suse.cz>
17804 * vec.h (class auto_string_vec): New (moved from auto_argvec).
17805 (auto_string_vec::~auto_string_vec): Likewise.
17807 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
17809 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
17810 prevent_decl_creation_for_types fields up and add reset_location field.
17811 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
17812 statement if id->reset_location is true.
17813 (copy_edges_for_bb): Do not set goto_locus on the new edges if
17814 id->reset_location is true.
17815 (copy_phis_for_bb): Force input_location on the arguments if
17816 id->reset_location is true.
17817 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
17818 is set on the function to be inlined.
17820 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
17822 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
17824 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
17826 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
17827 registers for Pmode.
17828 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
17829 hard registers for the clobbered pseudo.
17831 2018-06-27 Paul Koning <ni1d@arrl.net>
17833 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
17834 mutually exclusive options.
17835 * config/pdp11/constraints.md (h): New constraint.
17836 (O): Update definition to match shift code generation.
17837 (D): New constraint.
17838 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
17840 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
17842 (output_jump): Change arguments.
17843 (pdp11_fixed_cc_regs): New function.
17844 (pdp11_cc_mode): Ditto.
17845 (pdp11_expand_shift): Ditto.
17846 (pdp11_assemble_shift): Ditto.
17847 (pdp11_small_shift): Ditto.
17848 (pdp11_branch_cost): Remove.
17849 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
17851 (pdp11_register_move_cost): Update for CC registers.
17852 (pdp11_rtx_costs): Add case for LSHIFTRT.
17853 (pdp11_output_jump): Add CCNZ mode conditional branches.
17854 (notice_update_cc_on_set): Remove.
17855 (pdp11_cc_mode): New function.
17856 (simple_memory_operand): Correct pre/post decrement case.
17857 (no_side_effect_operand): New function.
17858 (pdp11_regno_reg_class): Add CC_REGS class.
17859 (pdp11_fixed_cc_regs): New function.
17860 (pdp11_small_shift): New function.
17861 (pdp11_expand_shift): New function to expand shift insns.
17862 (pdp11_assemble_shift): New function to output shifts.
17863 (pdp11_branch_cost): Remove.
17864 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
17865 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
17866 (WCHAR_TYPE): Ditto.
17867 (PTRDIFF_TYPE): Ditto.
17868 (ADJUST_INSN_LENGTH): New macro.
17869 (FIXED_REGISTERS): Add CC registers.
17870 (CALL_USED_REGISTERS): Ditto.
17871 (reg_class): Ditto.
17872 (REG_CLASS_NAMES): Ditto.
17873 (REG_CLASS_CONTENTS): Ditto.
17874 (SELECT_CC_MODE): Use new function.
17875 (TARGET_FLAGS_REGNUM): New macro.
17876 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
17877 (cc0_reg_rtx): Remove.
17878 (CC_STATUS_MDEP): Remove.
17879 (CC_STATUS_MDEFP_INIT): Remove.
17880 (CC_IN_FPU): Remove.
17881 (NOTICE_UPDATE_CC): Remove.
17882 (REGISTER_NAMES): Add CC registers.
17883 (BRANCH_COST): Change to constant 1.
17884 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
17886 * config/pdp11/pdp11.opt (mbcopy): Remove.
17887 (mbcopy-builtin): Remove.
17888 (mbranch-cheap): Remove.
17889 (mbranch-expensive): Remove.
17890 * config/pdp11/predicates.md (expand_shift_operand): Update to
17891 match shift code generation.
17892 (ccnz_operator): New predicate.
17893 * doc/invoke.texi (PDP-11 Options): Remove deleted options
17894 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
17895 Remove non-existent option -mabshi, -mno-abshi. Document mutually
17897 * doc/md.texi (PDP-11): Document new D and h constraints. Update
17898 description of O constraint.
17900 2018-06-27 Jeff Law <law@redhat.com>
17901 Austin Law <austinklaw@gmail.com>
17903 * config/v850/v850.md (addsi3_set_flags): New pattern.
17904 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
17905 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
17906 (zero_extendhisi2_v850_set_flags): Likewise.
17907 (zero_extendqisi2_v850_set_flags): Likewise.
17908 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
17909 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
17910 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
17912 * config/v850/v850-protos.h (notice_update_cc): Remove.
17913 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
17914 (v850_print_operand): Handle 'D' and "d".
17915 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
17916 Add handling of arithmetic/logical operations compared against zero.
17917 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
17918 Do not look at v850_compare_op, instead get mode from last argument.
17919 (v850_gen_compare): Remove
17920 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
17921 after reload for prologue insns.
17922 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
17924 (construct_save_jarl): Likewise.
17925 (TARGET_FLAGS_REGNUM): Define.
17926 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
17927 (NOTICE_UPDATE_CC): Remove.
17928 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
17929 than cc0. Conditionalize on reload_completed.
17930 (cmpsi_insn, setfcc_insn): Likewise.
17931 (tst1 splitter): Turn into define_and_split which sets the flags
17933 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
17934 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
17935 (cstoresf4, cstoredf4): Clobber the flags.
17936 (cmpsi, cmpsf, cmpdf): Remove expanders.
17937 (setf_insn): Remove pattern.
17938 (addsi3): Turn into define_and_split which clobbers the flags after
17939 reload and a suitable pattern (addsi3_clobber_flags) for use after
17941 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
17942 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
17943 (ashrsi3, ashrsi3_v850e2): Likewise.
17944 (bins): Clobber the flags.
17945 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
17946 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
17947 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
17948 (call_value_internal_short, call_value_internal_long): Likewise.
17949 (callt_save_interrupt, callt_return_interrupt): Likewise.
17950 (save_interrupt, return_interrupt): Likewise.
17951 (callt_save_all_interrupt, save_all_interrupt): Likewise.
17952 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
17953 (restore_all_interrupt, _restore_all_interrupt): Likewise.
17954 (All FP comparisons): Only allow after reload has completed.
17956 (divh, divhu): Tweak output template.
17957 (branch_z_normal, branch_z_invert): Remove
17958 (branch_nz_normal, branch_nz_invert): Likewise.
17959 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
17961 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
17962 * config/v850/v850.c (notice_update_cc): Remove.
17963 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
17964 (CC_NO_CARRY): Likewise.
17965 (NOTICE_UPDATE_CC): Define to nothing.
17966 * config/v850/v850.md: Remove block comment on cc0 handling
17967 Remove "cc" attribute from all patterns. Remove cc_status handling
17968 from all patterns. Minor formatting fixes.
17970 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17972 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
17973 (cortex-a76.cortex-a55): Likewise.
17974 * config/aarch64/aarch64-tune.md: Regenerate.
17975 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
17976 cortex-a76.cortex-a55.
17978 2018-06-27 Jeff Law <law@redhat.com>
17980 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
17981 (MULTILIB_DIRNAMES): Similarly.
17983 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
17985 * gimple.h (gimple_return_retbnd): Delete.
17986 (gimple_return_set_retbnd): Likewise.
17987 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
17988 gimple_return_set_retbnd.
17989 * gimple-pretty-print.c (dump_gimple_return): Remove call to
17990 gimple_return_retbnd and adjust.
17991 * tree-inline.h (struct copy_body_data): Remove retbnd field.
17992 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
17993 Explicitly return NULL in a couple more cases. Move assertion
17994 on debug statements and remove unreachable code.
17995 (reset_debug_binding): Do not test id->retbnd.
17996 (expand_call_inline): Do not set it.
17998 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
18000 * configure.ac: Add --disable-gcov option.
18001 * configure: Regenerate.
18002 * Makefile.in: Honour @enable_gcov@.
18003 * doc/install.texi: Document --disable-gcov.
18005 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18007 * config/arm/arm-cpus.in (cortex-a76): New entry.
18008 (cortex-a76.cortex-a55): Likewise.
18009 * config/arm/arm-tables.opt: Regenerate.
18010 * config/arm/arm-tune.md: Likewise.
18011 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
18012 * doc/invoke.texi (ARM Options): Document cortex-a76 and
18013 cortex-a76.cortex-a55.
18015 2018-06-27 Tamar Christina <tamar.christina@arm.com>
18018 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
18020 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
18022 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
18024 (EPILOGUE_USES): Likewise.
18026 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
18028 * tree-inline.c (remap_location): New function extracted from...
18029 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
18030 (copy_phis_for_bb): ...here. Call remap_location.
18031 (copy_cfg_body): Adjust call to copy_edges_for_bb.
18033 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
18035 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
18036 unaligned vsx for 16B memset.
18038 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
18041 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
18042 ieee128_float_type_node to long_double_type_node unless
18043 TARGET_LONG_DOUBLE_128 is set.
18045 2018-06-26 David Malcolm <dmalcolm@redhat.com>
18047 * cfgloop.c (get_loop_location): Convert return type from
18048 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
18049 by implicit construction from rtx_insn *, and using
18050 dump_user_location_t::from_function_decl for the fallback case.
18051 * cfgloop.h (get_loop_location): Convert return type from
18052 location_t to dump_user_location_t.
18053 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
18054 dump_printf_loc to pass in a dump_location_t rather than a
18055 location_t, via the gimple stmt.
18056 * coverage.c (get_coverage_counts): Update calls to
18057 dump_printf_loc to pass in dump_location_t rather than a
18059 * doc/optinfo.texi (Dump types): Convert example of
18060 dump_printf_loc from taking "locus" to taking "insn". Update
18061 description of the "_loc" calls to cover dump_location_t.
18062 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
18064 (dump_user_location_t::dump_user_location_t): New constructors,
18065 from gimple *stmt and rtx_insn *.
18066 (dump_user_location_t::from_function_decl): New function.
18067 (dump_loc): Make static.
18068 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
18069 const dump_location_t &.
18070 (dump_generic_expr_loc): Delete.
18071 (dump_printf_loc): Convert param "loc" from location_t to
18072 const dump_location_t &.
18073 (selftest::test_impl_location): New function.
18074 (selftest::dumpfile_c_tests): New function.
18075 * dumpfile.h: Include "profile-count.h".
18076 (class dump_user_location_t): New class.
18077 (struct dump_impl_location_t): New struct.
18078 (class dump_location_t): New class.
18079 (dump_printf_loc): Convert 2nd param from source_location to
18080 const dump_location_t &.
18081 (dump_generic_expr_loc): Delete.
18082 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
18083 const dump_location_t &.
18084 * gimple-fold.c (fold_gimple_assign): Update call to
18085 dump_printf_loc to pass in a dump_location_t rather than a
18086 location_t, via the gimple stmt.
18087 (gimple_fold_call): Likewise.
18088 * gimple-loop-interchange.cc
18089 (loop_cand::analyze_iloop_reduction_var): Update for change to
18090 check_reduction_path.
18091 (tree_loop_interchange::interchange): Update for change to
18092 find_loop_location.
18093 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
18094 change in return-type of find_loop_location.
18095 (graphite_regenerate_ast_isl): Likewise.
18096 * graphite-optimize-isl.c (optimize_isl): Likewise.
18097 * graphite.c (graphite_transform_loops): Likewise.
18098 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
18099 pass in a dump_location_t rather than a location_t, via the
18101 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
18102 * ipa.c (walk_polymorphic_call_targets): Likewise.
18103 * loop-unroll.c (report_unroll): Convert "locus" param from
18104 location_t to dump_location_t.
18105 (decide_unrolling): Update for change to get_loop_location's
18107 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
18108 location_t to dump_user_location_t.
18109 (grid_find_single_omp_among_assignments_1): Updates calls to
18110 dump_printf_loc to pass in a dump_location_t rather than a
18111 location_t, via the gimple stmt.
18112 (grid_parallel_clauses_gridifiable): Convert "tloc" from
18113 location_t to dump_location_t. Updates calls to dump_printf_loc
18114 to pass in a dump_location_t rather than a location_t, via the
18116 (grid_inner_loop_gridifiable_p): Likewise.
18117 (grid_dist_follows_simple_pattern): Likewise.
18118 (grid_gfor_follows_tiling_pattern): Likewise.
18119 (grid_target_follows_gridifiable_pattern): Likewise.
18120 (grid_attempt_target_gridification): Convert initialization
18121 of local "grid" from memset to zero-initialization; FIXME: does
18122 this require C++11? Update call to dump_printf_loc to pass in a
18123 optinfo_location rather than a location_t, via the gimple stmt.
18124 * profile.c (read_profile_edge_counts): Updates call to
18125 dump_printf_loc to pass in a dump_location_t rather than a
18127 (compute_branch_probabilities): Likewise.
18128 * selftest-run-tests.c (selftest::run_tests): Call
18130 * selftest.h (dumpfile_c_tests): New decl.
18131 * tree-loop-distribution.c (pass_loop_distribution::execute):
18132 Update for change in return type of find_loop_location.
18133 * tree-parloops.c (parallelize_loops): Likewise.
18134 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
18135 "locus" from location_t to dump_user_location_t.
18136 (canonicalize_loop_induction_variables): Likewise.
18137 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
18138 for change in return type of find_loop_location.
18139 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
18140 to dump_printf_loc to pass in a dump_location_t rather than a
18141 location_t, via the stmt.
18142 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
18144 * tree-vect-loop-manip.c (find_loop_location): Convert return
18145 type from source_location to dump_user_location_t.
18146 (vect_do_peeling): Update for above change.
18147 (vect_loop_versioning): Update for change in type of
18149 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
18150 from location_t to dump_user_location_t.
18151 (vect_estimate_min_profitable_iters): Update for change in type
18153 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
18154 location_t to dump_location_t.
18155 (vect_slp_bb): Update for change in type of vect_location.
18156 * tree-vectorizer.c (vect_location): Convert from source_location
18157 to dump_user_location_t.
18158 (try_vectorize_loop_1): Update for change in vect_location's type.
18159 (vectorize_loops): Likewise.
18160 (increase_alignment): Likewise.
18161 * tree-vectorizer.h (vect_location): Convert from source_location
18162 to dump_user_location_t.
18163 (find_loop_location): Convert return type from source_location to
18164 dump_user_location_t.
18165 (check_reduction_path): Convert 1st param from location_t to
18166 dump_user_location_t.
18167 * value-prof.c (check_counter): Update call to dump_printf_loc to
18168 pass in a dump_user_location_t rather than a location_t; update
18169 call to error_at for change in type of "locus".
18170 (check_ic_target): Update call to dump_printf_loc to
18171 pass in a dump_user_location_t rather than a location_t, via the
18174 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
18176 * config/s390/s390.h (enum processor_flags): Do not use
18177 default tune parameter when -march was specified.
18179 2018-06-26 Jakub Jelinek <jakub@redhat.com>
18182 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
18183 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
18186 2018-06-26 Richard Biener <rguenther@suse.de>
18188 PR tree-optimization/86287
18190 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
18191 (vect_analyze_loop): Initialize n_stmts.
18193 2018-06-26 Richard Biener <rguenther@suse.de>
18195 PR middle-end/86271
18196 * fold-const.c (fold_convertible_p): Pointer extension
18199 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
18202 * dwarf2out.c (loc_list_has_views): Adjust comments.
18203 (dw_loc_list): Split single cross-partition range with
18206 2018-06-25 Jeff Law <law@redhat.com>
18208 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
18209 on -mbig-switch by default.
18211 * config/v850/predicates.md (const_float_1_operand): Fix match_code
18213 (const_float_0_operand): Remove unused predicate.
18214 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
18215 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
18216 (recipsf2): New expander. Original pattern now called
18218 (recipdf2, recipdf2_insn): Similarly.
18219 (rsqrtsf2, rsqrtsf2_insn): Similarly
18220 (rsqrtdf2, rsqrtdf2_insn): Similarly
18222 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
18224 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
18225 Simplify logic for FreeBSD (twice).
18227 2018-06-25 Martin Sebor <msebor@redhat.com>
18229 PR tree-optimization/86204
18230 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
18231 a strnlen result if it's less than the length of the string.
18233 2018-06-25 Martin Sebor <msebor@redhat.com>
18235 PR tree-optimization/85700
18236 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
18237 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
18238 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
18240 2018-06-25 Martin Sebor <msebor@redhat.com>
18242 * doc/extend.texi (Zero-length arrays): Update and clarify.
18244 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
18246 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
18247 added IEEE/IBM long double multilib support on PowerPC little
18248 endian Linux systems.
18249 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
18250 (MULTILIB_DEFAULTS): Likewise.
18251 * config/rs6000/rs6000.c (rs6000_option_override_internal):
18253 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
18254 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
18255 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
18257 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
18259 PR middle-end/86311
18260 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
18261 (REORDER_45): Likewise.
18263 2018-06-25 Jeff Law <law@redhat.com>
18265 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
18266 dividend to 32 bits. Adjust length.
18267 (udivmodhi4): Cleanup output template. Fix length.
18269 2018-06-25 Carl Love <cel@us.ibm.com>
18271 * config/rs6000/vsx.md: Change word selector to prefered location.
18273 2018-06-25 Richard Biener <rguenther@suse.de>
18275 PR tree-optimization/86304
18276 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
18277 epilogue-if-converted loops as well.
18279 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
18281 * lto-section-out.c (lto_begin_section): Do not print section
18282 name for noaddr and unnumbered dumps.
18284 2018-06-25 Richard Biener <rguenther@suse.de>
18286 * tree-vectorizer.h (struct vec_info_shared): New structure
18287 with parts split out from struct vec_info and loop_nest from
18288 struct _loop_vec_info.
18289 (struct vec_info): Adjust accordingly.
18290 (struct _loop_vec_info): Likewise.
18291 (LOOP_VINFO_LOOP_NEST): Adjust.
18292 (LOOP_VINFO_DATAREFS): Likewise.
18293 (LOOP_VINFO_DDRS): Likewise.
18294 (struct _bb_vec_info): Likewise.
18295 (BB_VINFO_DATAREFS): Likewise.
18296 (BB_VINFO_DDRS): Likewise.
18297 (struct _stmt_vec_info): Add dr_aux member.
18298 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
18299 (DR_MISALIGNMENT_UNINITIALIZED): New.
18300 (set_dr_misalignment): Adjust.
18301 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
18302 (vect_analyze_loop): Adjust prototype.
18303 (vect_analyze_loop_form): Likewise.
18304 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
18305 Compute dependences lazily.
18306 (vect_record_base_alignments): Use shared datarefs/ddrs.
18307 (vect_verify_datarefs_alignment): Likewise.
18308 (vect_analyze_data_refs_alignment): Likewise.
18309 (vect_analyze_data_ref_accesses): Likewise.
18310 (vect_analyze_data_refs): Likewise.
18311 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
18312 constructor parameter for shared part.
18313 (vect_analyze_loop_form): Pass in shared part and adjust.
18314 (vect_analyze_loop_2): Pass in storage for the number of
18315 stmts. Move loop nest finding to the caller. Compute
18317 (vect_analyze_loop): Pass in shared part.
18318 (vect_transform_loop): Verify shared datarefs are unchanged.
18319 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
18320 constructor parameter for shared part.
18321 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
18322 (vect_slp_bb): Verify shared datarefs are unchanged before
18324 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
18326 (new_stmt_vec_info): Initialize DR_AUX misalignment to
18327 DR_MISALIGNMENT_UNINITIALIZED.
18328 * tree-vectorizer.c (vec_info::vec_info): Add constructor
18329 parameter for shared part.
18330 (vec_info::~vec_info): Adjust.
18331 (vec_info_shared::vec_info_shared): New.
18332 (vec_info_shared::~vec_info_shared): Likewise.
18333 (vec_info_shared::save_datarefs): Likewise.
18334 (vec_info_shared::check_datarefs): Likewise.
18335 (try_vectorize_loop_1): Construct shared part live for analyses
18336 of a single loop for multiple vector sizes.
18337 * tree-parloops.c (gather_scalar_reductions): Adjust.
18339 2018-06-25 Richard Biener <rguenther@suse.de>
18341 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
18342 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
18343 (vect_analyze_data_refs): Remove similar code from here and
18344 simplify accordingly.
18346 2018-06-25 Richard Biener <rguenther@suse.de>
18348 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
18349 for reverse storage order accesses rather than asserting
18350 they cannot happen here.
18352 2018-06-25 Tom de Vries <tdevries@suse.de>
18355 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
18356 Use data16 instead of .byte for insn prefix.
18358 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
18361 * parser.c (make_char_string_pack): Pass this literal chars
18362 through cpp_interpret_string.
18363 (cp_parser_userdef_numeric_literal): Check the result of
18364 make_char_string_pack.
18366 2018-06-24 Maya Rashish <coypu@sdf.org>
18368 * ginclude/stddef.h: Simplify conditions around avoiding
18369 re-definition of __size_t.
18371 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
18373 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
18374 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
18376 2018-06-22 Maya Rashish <coypu@sdf.org>
18378 * doc/invoke.texi (mno-fancy-math-387): Update for changes
18379 made to OpenBSD and NetBSD through the years.
18381 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
18383 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
18384 behavior of vec_pack (vector double, vector double) to match
18385 behavior of vec_float2 (vector double, vector double).
18387 2018-06-22 Olivier Hainque <hainque@adacore.com>
18389 * gimplify.c (gimplify_function_tree): Prevent creation
18390 of a trampoline for the address of the current function
18391 passed to entry/exit instrumentation hooks.
18393 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
18396 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
18399 2018-06-22 Martin Liska <mliska@suse.cz>
18401 PR tree-optimization/86263
18402 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
18403 Bail out if is_enabled is false.
18404 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
18406 (jump_table_cluster::is_enabled): New function.
18408 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
18410 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
18411 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
18412 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
18413 (lto_input_ts_binfo_tree_pointers): Likewise.
18414 * tree-streamer-out.c (streamer_write_tree_bitfields,
18415 write_ts_binfo_tree_pointers): Likewise.
18416 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
18418 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
18420 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
18422 2018-06-22 Martin Liska <mliska@suse.cz>
18424 * symbol-summary.h (get): Make it pure and inline move
18425 functionality from ::get function.
18426 (get): Remove and inline into ::get and ::get_create.
18427 (get_create): Move code from ::get function.
18429 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18432 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
18433 -x assembler-with-cpp.
18435 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18437 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
18438 _FILE_OFFSET_BITS=64 for C++.
18440 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
18442 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
18443 conversion insn that shows up when pr85657-3.c is compiled using
18444 IEEE 128-bit long double.
18445 (neg<mode>2_internal): Use the correct mode to check whether the
18446 mode is IBM extended.
18447 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
18448 multiply and divide external functions from being created more
18451 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
18453 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
18455 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
18456 the edge can be forwarded.
18457 (cfg_layout_merge_blocks): Likewise.
18459 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
18461 * except.c (finish_eh_generation): Commit edge insertions only after
18462 the EH edges have been redirected from post-landing to landing pads.
18464 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
18466 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
18467 create_tmp_var_for to create the FRAME decl.
18468 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
18470 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
18472 * tree-inline.c (copy_edges_for_bb): Minor tweak.
18473 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
18474 debug statement when resetting its value.
18475 (expand_call_inline): Copy the locus of the call onto the assignment
18476 of the return value, if any. Use local variable in more cases.
18478 2018-06-21 Martin Liska <mliska@suse.cz>
18480 * ipa-pure-const.c (propagate_nothrow): Use
18481 funct_state_summaries->get.
18482 (dump_malloc_lattice): Likewise.
18483 (propagate_malloc): Likewise.
18485 2018-06-21 Richard Biener <rguenther@suse.de>
18487 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
18488 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
18489 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
18491 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
18492 BLOCK_ABSTRACT_ORIGIN unconditionally.
18494 2018-06-21 David Malcolm <dmalcolm@redhat.com>
18496 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
18498 * ipa-reference.c (ipa_reference_c_finalize): Delete
18499 ipa_ref_opt_sum_summaries and set it to NULL.
18501 2018-06-21 Tom de Vries <tdevries@suse.de>
18503 PR tree-optimization/85859
18504 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
18505 test with comment from bb_no_side_effects_p.
18507 2018-06-21 Richard Biener <rguenther@suse.de>
18509 PR tree-optimization/86232
18510 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
18511 max for constant niter.
18513 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
18515 * config/aarch64/aarch64-simd.md
18516 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
18518 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
18520 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
18521 Make opernads of the unspec commutative.
18523 2018-06-21 Richard Biener <rguenther@suse.de>
18525 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
18526 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
18527 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
18528 (vect_analyze_data_ref_dependence): Re-order checks to deal with
18530 (vect_record_base_alignments): Do not record base alignment
18531 for gathers or scatters.
18532 (vect_compute_data_ref_alignment): Drop return value that is always
18533 true. Bail out early for gathers or scatters.
18534 (vect_enhance_data_refs_alignment): Bail out early for gathers
18536 (vect_find_same_alignment_drs): Likewise.
18537 (vect_analyze_data_refs_alignment): Remove dead code.
18538 (vect_slp_analyze_and_verify_node_alignment): Likewise.
18539 (vect_analyze_data_refs): For possible gathers or scatters do
18540 not create an alternate DR, just check their possible validity
18541 and mark them. Adjust DECL_NONALIASED handling to not rely
18542 on DR_BASE_ADDRESS.
18543 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
18544 update inits of gathers or scatters.
18545 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
18546 Also copy gather/scatter flag to pattern vinfo.
18548 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
18550 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
18551 behavior of vec_packsu (vector unsigned long long, vector unsigned
18552 long long) to match behavior of vec_packs with same signature.
18554 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
18555 Thomas Schwinge <thomas@codesourcery.com>
18556 Cesar Philippidis <cesar@codesourcery.com>
18558 * gimplify.c (gimplify_scan_omp_clauses): Add support for
18559 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
18560 (gimplify_adjust_omp_clauses): Likewise.
18561 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
18562 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
18563 (gimplify_omp_target_update): Update handling of acc update and
18565 * omp-low.c (install_var_field): Remove unused parameter
18566 base_pointers_restrict.
18567 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
18568 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
18570 (omp_target_base_pointers_restrict_p): Delete.
18571 (scan_omp_target): Update call to scan_sharing_clauses.
18572 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
18574 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
18575 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
18576 (convert_local_omp_clauses): Likewise.
18577 * tree-pretty-print.c (dump_omp_clause): Likewise.
18578 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
18580 (omp_clause_code_name): Likewise.
18582 2018-06-20 Jakub Jelinek <jakub@redhat.com>
18585 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
18588 PR tree-optimization/86231
18589 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
18590 anti-range don't overwrite *vr0min before using it to compute *vr0max.
18592 2018-06-20 Tom de Vries <tdevries@suse.de>
18594 PR tree-optimization/86097
18595 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
18596 iv type if signedness of iv type is not the same as that of *nit.
18598 2018-06-20 Jakub Jelinek <jakub@redhat.com>
18600 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
18601 EDGE_EH edges, verify they are all EDGE_EH.
18603 2018-06-20 Maya Rashish <coypu@sdf.org>
18605 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
18607 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18609 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
18610 * config/aarch64/aarch64.c (xgene1_tunings): Add
18611 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
18612 (aarch64_mode_valid_for_sched_fusion_p):
18613 Allow 16-byte modes.
18614 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
18615 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
18617 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
18619 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
18620 * config/aarch64/iterators.md (VQ2): New mode iterator.
18622 2018-06-20 Martin Liska <mliska@suse.cz>
18624 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
18625 Change default ratio from 10 to 8.
18627 2018-06-20 Martin Liska <mliska@suse.cz>
18629 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
18631 (bit_test_cluster::find_bit_tests): Likewise.
18632 (switch_decision_tree::analyze_switch_statement): Find clusters.
18633 * tree-switch-conversion.h (struct jump_table_cluster): Document
18636 2018-06-20 Martin Liska <mliska@suse.cz>
18638 * tree-switch-conversion.c (switch_conversion::collect):
18639 Record m_uniq property.
18640 (switch_conversion::expand): Bail out for special conditions.
18641 (group_cluster::~group_cluster): New.
18642 (group_cluster::group_cluster): Likewise.
18643 (group_cluster::dump): Likewise.
18644 (jump_table_cluster::emit): New.
18645 (switch_decision_tree::fix_phi_operands_for_edges): New.
18646 (struct case_node): Remove struct.
18647 (jump_table_cluster::can_be_handled): New.
18648 (case_values_threshold): Moved to header.
18649 (reset_out_edges_aux): Likewise.
18650 (jump_table_cluster::is_beneficial): New.
18651 (bit_test_cluster::can_be_handled): Likewise.
18652 (add_case_node): Remove.
18653 (bit_test_cluster::is_beneficial): New.
18654 (case_bit_test::cmp): New.
18655 (bit_test_cluster::emit): New.
18656 (expand_switch_as_decision_tree_p): Remove.
18657 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
18658 (fix_phi_operands_for_edge): Likewise.
18659 (switch_decision_tree::analyze_switch_statement): New.
18660 (compute_cases_per_edge): Move ...
18661 (switch_decision_tree::compute_cases_per_edge): ... here.
18662 (try_switch_expansion): Likewise.
18663 (switch_decision_tree::try_switch_expansion): Likewise.
18664 (record_phi_operand_mapping): Likewise.
18665 (switch_decision_tree::record_phi_operand_mapping): Likewise.
18666 (emit_case_decision_tree): Likewise.
18667 (switch_decision_tree::emit): Likewise.
18668 (balance_case_nodes): Likewise.
18669 (switch_decision_tree::balance_case_nodes): Likewise.
18670 (dump_case_nodes): Likewise.
18671 (switch_decision_tree::dump_case_nodes): Likewise.
18672 (emit_jump): Likewise.
18673 (switch_decision_tree::emit_jump): Likewise.
18674 (emit_cmp_and_jump_insns): Likewise.
18675 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
18676 (emit_case_nodes): Likewise.
18677 (switch_decision_tree::emit_case_nodes): Likewise.
18678 (conditional_probability): Remove.
18679 * tree-switch-conversion.h (enum cluster_type): New.
18681 (struct cluster): Likewise.
18682 (cluster::cluster): Likewise.
18683 (struct simple_cluster): Likewise.
18684 (simple_cluster::simple_cluster): Likewise.
18685 (struct group_cluster): Likewise.
18686 (struct jump_table_cluster): Likewise.
18687 (struct bit_test_cluster): Likewise.
18688 (struct min_cluster_item): Likewise.
18689 (struct case_tree_node): Likewise.
18690 (case_tree_node::case_tree_node): Likewise.
18691 (jump_table_cluster::case_values_threshold): Likewise.
18692 (struct case_bit_test): Likewise.
18693 (struct switch_decision_tree): Likewise.
18694 (struct switch_conversion): Likewise.
18695 (switch_decision_tree::reset_out_edges_aux): Likewise.
18697 2018-06-20 Martin Liska <mliska@suse.cz>
18699 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
18700 (hoist_edge_and_branch_if_true): Likewise.
18701 (expand_switch_using_bit_tests_p): Likewise.
18702 (struct case_bit_test): Likewise.
18703 (case_bit_test_cmp): Likewise.
18704 (emit_case_bit_tests): Likewise.
18705 (switch_conversion::switch_conversion): New class.
18706 (struct switch_conv_info): Remove old struct.
18707 (collect_switch_conv_info): More to ...
18708 (switch_conversion::collect): ... this.
18709 (check_range): Likewise.
18710 (switch_conversion::check_range): Likewise.
18711 (check_all_empty_except_final): Likewise.
18712 (switch_conversion::check_all_empty_except_final): Likewise.
18713 (check_final_bb): Likewise.
18714 (switch_conversion::check_final_bb): Likewise.
18715 (create_temp_arrays): Likewise.
18716 (switch_conversion::create_temp_arrays): Likewise.
18717 (free_temp_arrays): Likewise.
18718 (gather_default_values): Likewise.
18719 (switch_conversion::gather_default_values): Likewise.
18720 (build_constructors): Likewise.
18721 (switch_conversion::build_constructors): Likewise.
18722 (constructor_contains_same_values_p): Likewise.
18723 (switch_conversion::contains_same_values_p): Likewise.
18724 (array_value_type): Likewise.
18725 (switch_conversion::array_value_type): Likewise.
18726 (build_one_array): Likewise.
18727 (switch_conversion::build_one_array): Likewise.
18728 (build_arrays): Likewise.
18729 (switch_conversion::build_arrays): Likewise.
18730 (gen_def_assigns): Likewise.
18731 (switch_conversion::gen_def_assigns): Likewise.
18732 (prune_bbs): Likewise.
18733 (switch_conversion::prune_bbs): Likewise.
18734 (fix_phi_nodes): Likewise.
18735 (switch_conversion::fix_phi_nodes): Likewise.
18736 (gen_inbound_check): Likewise.
18737 (switch_conversion::gen_inbound_check): Likewise.
18738 (process_switch): Use the newly created class.
18739 (switch_conversion::expand): New.
18740 (switch_conversion::~switch_conversion): New.
18741 * tree-switch-conversion.h: New file.
18743 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
18745 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
18746 tree-vect-patterns.c.
18747 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
18748 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
18749 (vect_recog_sad_pattern): Likewise.
18750 (vect_recog_widen_sum_pattern): Likewise.
18751 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
18752 (vect_recog_widen_shift_pattern): Remove the type_in argument.
18753 (vect_recog_rotate_pattern): Likewise.
18754 (vect_recog_mult_pattern): Likewise.
18755 (vect_recog_vector_vector_shift_pattern): Likewise.
18756 (vect_recog_divmod_pattern): Likewise.
18757 (vect_recog_mixed_size_cond_pattern): Likewise.
18758 (vect_recog_bool_pattern): Likewise.
18759 (vect_recog_mask_conversion_pattern): Likewise.
18760 (vect_try_gather_scatter_pattern): Likewise.
18761 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
18762 (vect_recog_over_widening_pattern): Likewise.
18763 (vect_recog_gather_scatter_pattern): Likewise.
18764 (vect_recog_func_ptr): Move from tree-vectorizer.h
18765 (vect_vect_recog_func_ptrs): Move further down the file.
18766 (vect_recog_func): Likewise. Remove the third argument.
18767 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
18768 (vect_pattern_recog_1): Expect the pattern function to do any
18769 necessary target tests. Also expect it to provide a vector type.
18770 Remove the type_in handling.
18772 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
18774 * tree-vect-patterns.c (vect_pattern_detected): New function.
18775 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
18776 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
18777 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
18778 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
18779 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
18780 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
18781 (vect_recog_mask_conversion_pattern)
18782 (vect_try_gather_scatter_pattern): Likewise.
18784 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
18786 * tree-vect-patterns.c (vect_get_internal_def): New function.
18787 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
18788 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
18789 (search_type_for_mask_1): Use it.
18791 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
18793 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
18794 redundant WIDEN_SUM_EXPR handling.
18795 (vect_recog_sad_pattern): Likewise.
18797 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
18799 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
18800 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
18801 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
18802 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
18803 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
18805 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
18807 * tree-vect-stmts.c (vectorizable_call): Make sure that we
18808 use the stmt_vec_info of the original bb statement for the
18809 new zero assignment, even if the call is part of a pattern.
18811 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
18813 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
18814 that the sequence is attached to the original statement rather
18815 than the pattern statement.
18816 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
18817 PATTERN_DEF_SEQ from the original statement rather than
18818 the main pattern statement.
18819 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
18820 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
18821 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
18823 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
18825 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
18826 definition statements before the early exit for statements that aren't
18828 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
18830 (vect_transform_loop): ...here. Process pattern definition
18831 statements without first checking whether the main pattern
18832 statement is live or relevant.
18834 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
18836 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
18837 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
18839 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
18841 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
18842 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
18843 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
18844 (expand_block_compare): Change select_block_compare_mode call.
18845 (expand_strncmp_align_check): Use new functions, fix comment.
18846 (emit_final_str_compare_gpr): New function.
18847 (expand_strn_compare): Refactor and clean up code.
18848 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
18850 2018-06-19 Tony Reix <tony.reix@atos.com>
18851 Damien Bergamini <damien.bergamini@atos.com>
18852 David Edelsohn <dje.gcc@gmail.com>
18854 * collect2.c (static_obj): New variable.
18855 (static_libs): New variable.
18856 (is_in_list): Uncomment declaration.
18857 (main): Track AIX libraries linked statically.
18858 (is_in_list): Uncomment definition.
18859 (scan_prog_file): Don't add AIX shared libraries initializer
18860 to constructor list if linking statically.
18862 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
18864 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
18866 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
18868 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
18870 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
18873 2018-06-19 Martin Liska <mliska@suse.cz>
18875 * config/i386/i386.c (ix86_can_inline_p): Do not use
18876 ipa_fn_summaries::get_create.
18877 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
18879 (devirtualization_time_bonus): Likewise.
18880 (ipcp_propagate_stage): Likewise.
18881 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
18882 (edge_set_predicate): Likewise.
18883 (evaluate_conditions_for_known_args): Likewise.
18884 (evaluate_properties_for_edge): Likewise.
18885 (ipa_call_summary::reset): Tranform to ...
18886 (ipa_call_summary::~ipa_call_summary): ... this.
18887 (ipa_fn_summary::reset): Transform to ...
18888 (ipa_fn_summary::~ipa_fn_summary): ... this.
18889 (ipa_fn_summary_t::remove): Rename to ...
18890 (ipa_fn_summary_t::remove_callees): ... this.
18891 (ipa_fn_summary_t::duplicate): Use placement new
18892 instead of memory copy.
18893 (ipa_call_summary_t::duplicate): Likewise.
18894 (ipa_call_summary_t::remove): Remove.
18895 (dump_ipa_call_summary): Change get_create to get.
18896 (ipa_dump_fn_summary): Dump only when summary exists.
18897 (analyze_function_body): Use symbol_summary::get instead
18899 (compute_fn_summary): Likewise.
18900 (estimate_edge_devirt_benefit): Likewise.
18901 (estimate_edge_size_and_time): Likewise.
18902 (inline_update_callee_summaries): Likewise.
18903 (remap_edge_change_prob): Likewise.
18904 (remap_edge_summaries): Likewise.
18905 (ipa_merge_fn_summary_after_inlining): Likewise.
18906 (write_ipa_call_summary): Likewise.
18907 (ipa_fn_summary_write): Likewise.
18908 (ipa_free_fn_summary): Likewise.
18909 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
18910 (struct ipa_call_summary): Likewise.
18911 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
18913 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
18914 (estimate_size_after_inlining): Likewise.
18915 (estimate_growth): Likewise.
18916 (growth_likely_positive): Likewise.
18917 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
18918 (inline_call): Likewise.
18919 * ipa-inline.c (caller_growth_limits): Likewise.
18920 (can_inline_edge_p): Likewise.
18921 (can_inline_edge_by_limits_p): Likewise.
18922 (compute_uninlined_call_time): Likewise.
18923 (compute_inlined_call_time): Likewise.
18924 (want_inline_small_function_p): Likewise.
18925 (edge_badness): Likewise.
18926 (update_caller_keys): Likewise.
18927 (update_callee_keys): Likewise.
18928 (inline_small_functions): Likewise.
18929 (inline_to_all_callers_1): Likewise.
18930 (dump_overall_stats): Likewise.
18931 (early_inline_small_functions): Likewise.
18932 (early_inliner): Likewise.
18933 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
18934 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
18935 * ipa-pure-const.c (malloc_candidate_p): Likewise.
18936 * ipa-split.c (execute_split_functions): Likewise.
18937 * symbol-summary.h: Likewise.
18938 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
18940 2018-06-19 Richard Biener <rguenther@suse.de>
18942 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
18943 (vectorize_loops): ... here. Fix dbgcnt handling.
18944 (try_vectorize_loop): Wrap try_vectorize_loop_1.
18946 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
18949 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
18950 ieee128 argument takes up only one (vector) register, not two (floating
18953 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
18955 * gimplify.c (gimplify_init_constructor): Really never clear for an
18956 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
18958 2018-06-19 Richard Biener <rguenther@suse.de>
18960 PR tree-optimization/86179
18961 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
18962 after failed recognition.
18964 2018-06-18 Martin Sebor <msebor@redhat.com>
18966 PR middle-end/85602
18967 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
18968 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
18969 Handle integer subtraction.
18970 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
18971 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
18973 2018-06-18 David Malcolm <dmalcolm@redhat.com>
18975 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
18976 param from rtx to rtx_insn *.
18977 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
18979 (frv_ifcvt_modify_insn): Likwise.
18980 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
18981 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
18982 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
18983 as_a <rtx_insn *> cast to local "unprotected_region" once
18984 it's been established that it's not NULL or pc_rtx.
18985 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
18986 param "sethi" from rtx to rtx_insn *.
18987 (nds32_group_float_insns): Likewise for param "insn".
18988 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
18990 (vax_output_int_subtract): Likewise.
18991 * config/vax/vax.c (vax_output_int_add): Likewise for param
18993 (vax_output_int_subtract): Likewise.
18994 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
18995 (emit_pattern_after): Likewise for param "after".
18996 (emit_insn_after): Likewise.
18997 (emit_jump_insn_after): Likewise.
18998 (emit_call_insn_after): Likewise.
18999 (emit_debug_insn_after): Likewise.
19000 (emit_pattern_before): Likewise for param "before".
19001 (emit_insn_before): Likewise.
19002 (emit_jump_insn_before): Likewise.
19003 * final.c (get_insn_template): Likewise for param "insn", removing
19005 * output.h (get_insn_template): Likewise for 2nd param.
19006 * rtl.h (emit_insn_before): Likewise.
19007 (emit_jump_insn_before): Likewise.
19008 (emit_debug_insn_before_noloc): Likewise.
19009 (emit_insn_after): Likewise.
19010 (emit_jump_insn_after): Likewise.
19011 (emit_call_insn_after): Likewise.
19012 (emit_debug_insn_after): Likewise.
19013 (set_insn_deleted): Likewise for param.
19015 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
19018 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
19019 floating point modes, so that IFmode is numerically greater than
19020 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
19021 to declare the ordering. This prevents IFmode from being
19022 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
19023 machine. Include rs6000-modes.h to share the fractional values
19024 between genmodes* and the rest of the compiler.
19025 (IFmode): Likewise.
19026 (KFmode): Likewise.
19027 (TFmode): Likewise.
19028 * config/rs6000/rs6000-modes.h: New file.
19029 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
19030 meaning of rs6000_long_double_size so that 126..128 selects an
19031 appropriate 128-bit floating point type.
19032 (rs6000_option_override_internal): Likewise.
19033 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
19034 (TARGET_LONG_DOUBLE_128): Change the meaning of
19035 rs6000_long_double_size so that 126..128 selects an appropriate
19036 128-bit floating point type.
19037 (LONG_DOUBLE_TYPE_SIZE): Update comment.
19038 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
19039 source and destination to match the standard usage.
19040 (truncifkf2): Likewise.
19041 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
19042 ISA 2.07 to use an explicit clobber, instead of passing in a
19044 (copysign<mode>3_soft): Likewise.
19046 2018-06-18 David Malcolm <dmalcolm@redhat.com>
19048 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
19049 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
19050 (vect_slp_analyze_instance_dependence): Likewise.
19051 (vect_enhance_data_refs_alignment): Likewise.
19052 (vect_analyze_data_refs_alignment): Likewise.
19053 (vect_slp_analyze_and_verify_instance_alignment
19054 (vect_analyze_data_ref_accesses): Likewise.
19055 (vect_prune_runtime_alias_test_list): Likewise.
19056 (vect_analyze_data_refs): Likewise.
19057 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
19058 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
19059 (vect_analyze_scalar_cycles_1): Likewise.
19060 (vect_get_loop_niters): Likewise.
19061 (vect_analyze_loop_form_1): Likewise.
19062 (vect_update_vf_for_slp): Likewise.
19063 (vect_analyze_loop_operations): Likewise.
19064 (vect_analyze_loop): Likewise.
19065 (vectorizable_induction): Likewise.
19066 (vect_transform_loop): Likewise.
19067 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
19068 * tree-vect-slp.c (vect_analyze_slp): Likewise.
19069 (vect_make_slp_decision): Likewise.
19070 (vect_detect_hybrid_slp): Likewise.
19071 (vect_slp_analyze_operations): Likewise.
19072 (vect_slp_bb): Likewise.
19073 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
19074 (vectorizable_bswap): Likewise.
19075 (vectorizable_call): Likewise.
19076 (vectorizable_simd_clone_call): Likewise.
19077 (vectorizable_conversion): Likewise.
19078 (vectorizable_assignment): Likewise.
19079 (vectorizable_shift): Likewise.
19080 (vectorizable_operation): Likewise.
19081 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
19083 2018-06-18 Martin Sebor <msebor@redhat.com>
19085 PR tree-optimization/81384
19086 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
19087 * builtins.c (expand_builtin_strnlen): New function.
19088 (expand_builtin): Call it.
19089 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
19090 * builtins.def (BUILT_IN_STRNLEN): New.
19091 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
19092 Warn for bounds in excess of maximum object size.
19093 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
19094 single-value ranges. Handle strnlen.
19095 (handle_builtin_strlen): Handle strnlen.
19096 (strlen_check_and_optimize_stmt): Same.
19097 * doc/extend.texi (Other Builtins): Document strnlen.
19099 2018-06-18 Maya Rashish <coypu@sdf.org>
19101 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
19102 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
19103 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
19105 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
19107 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
19109 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19111 * tree.c (escaped_string::escape): Replace cast to char * by
19112 const_cast<char *> (unescaped).
19114 2018-06-18 Nick Clifton <nickc@redhat.com>
19117 * tree.c (escaped_string): New class. Converts an unescaped
19118 string into its escaped equivalent.
19119 (warn_deprecated_use): Use the new class to convert the
19120 deprecation message, if present.
19121 (test_escaped_strings): New self test.
19122 (test_c_tests): Add test_escaped_strings.
19123 * doc/extend.texi (deprecated): Add a note that the
19124 deprecation message is affected by the -fmessage-length
19125 option, and that control characters will be escaped.
19126 (#pragma GCC error): Document this pragma.
19127 (#pragma GCC warning): Likewise.
19128 * doc/invoke.texi (-fmessage-length): Document this option's
19129 effect on the #warning and #error preprocessor directives and
19130 the deprecated attribute.
19132 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
19134 * tree.c (decl_value_expr_lookup): Revert latest change.
19135 (decl_value_expr_insert): Likewise.
19137 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
19139 * gimplify.c (nonlocal_vlas): Delete.
19140 (nonlocal_vla_vars): Likewise.
19141 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
19143 (gimplify_body): Do not create and destroy nonlocal_vlas.
19144 * tree-nested.c: Include diagnostic.h.
19145 (use_pointer_in_frame): Tweak.
19146 (lookup_field_for_decl): Add assertion and declare the transformation.
19147 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
19148 internal error when the reference is in a wrong context. Do not
19149 create a debug decl by default.
19150 (note_nonlocal_block_vlas): Delete.
19151 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
19152 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
19153 create a debug decl by default.
19154 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
19155 call is in a wrong context.
19156 (fixup_vla_decls): New function.
19157 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
19158 debug variables were created.
19159 * tree.c (decl_value_expr_lookup): Add checking assertion.
19160 (decl_value_expr_insert): Likewise.
19162 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
19164 PR middle-end/82479
19165 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
19166 * tree-scalar-evolution.c (interpret_expr): Likewise.
19167 (expression_expensive_p): Likewise.
19168 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
19169 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
19170 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
19171 (ssa_defined_by_minus_one_stmt_p): New.
19173 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
19175 PR middle-end/64946
19176 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
19177 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
19178 * dojump.c (do_jump): Likewise.
19179 * expr.c (expand_expr_real_2): Check operand type's sign.
19180 * fold-const.c (const_unop): Handle ABSU_EXPR.
19181 (fold_abs_const): Likewise.
19182 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
19183 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
19184 (strip_sign_op_1): Likesise.
19185 * match.pd: Add new pattern to generate ABSU_EXPR.
19186 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
19187 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
19188 * tree-eh.c (operation_could_trap_helper_p): Likewise.
19189 * tree-inline.c (estimate_operator_cost): Likewise.
19190 * tree-pretty-print.c (dump_generic_node): Likewise.
19191 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
19192 * tree.def (ABSU_EXPR): New.
19194 2018-06-16 Jakub Jelinek <jakub@redhat.com>
19196 PR middle-end/86095
19197 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
19198 documented as preserved for backward compatibility only.
19199 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
19201 PR rtl-optimization/86108
19202 * bb-reorder.c (create_forwarder_block): Renamed to ...
19203 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
19204 jump from new landing pad to the second part.
19205 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
19208 2018-06-15 Jakub Jelinek <jakub@redhat.com>
19210 PR middle-end/85878
19211 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
19212 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
19213 Only call store_expr for halves if the mode is the same.
19215 PR middle-end/86123
19216 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
19217 Fix up comment formatting.
19219 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
19221 * typed-splay-tree.h (typed_splay_tree::remove): New function.
19222 (typed_splay_tree::closure,
19223 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
19224 (typed_splay_tree::typed_splay_tree,
19225 typed_splay_tree::operator =): Declared private.
19226 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
19227 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
19228 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
19229 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
19230 typed_splay_tree::splay_tree_splay,
19231 typed_splay_tree::splay_tree_foreach_helper,
19232 typed_splay_tree::splay_tree_insert,
19233 typed_splay_tree::splay_tree_remove,
19234 typed_splay_tree::splay_tree_lookup,
19235 typed_splay_tree::splay_tree_predecessor,
19236 typed_splay_tree::splay_tree_successor,
19237 typed_splay_tree::splay_tree_min,
19238 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
19239 (typed_splay_tree::root, typed_splay_tree::comp,
19240 typed_splay_tree::delete_key,
19241 typed_splay_tree::delete_value): New data members.
19242 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
19243 typed_splay_tree::remove.
19245 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
19247 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
19248 -mginv and -mno-ginv to the assembler.
19249 * config/mips/mips.opt (-mcrc): New option.
19250 (-mginv): Likewise.
19251 * doc/invoke.text (-mcrc): Document.
19252 (-mginv): Likewise.
19254 2018-06-15 Nick Clifton <nickc@redhat.com>
19257 * tree.c (escaped_string): New class. Converts an unescaped
19258 string into its escaped equivalent.
19259 (warn_deprecated_use): Use the new class to convert the
19260 deprecation message, if present.
19261 (test_escaped_strings): New self test.
19262 (test_c_tests): Add test_escaped_strings.
19263 * doc/extend.texi (deprecated): Add a note that the
19264 deprecation message is affected by the -fmessage-length
19265 option, and that control characters will be escaped.
19266 (#pragma GCC error): Document this pragma.
19267 (#pragma GCC warning): Likewise.
19268 * doc/invoke.texi (-fmessage-length): Document this option's
19269 effect on the #warning and #error preprocessor directives and
19270 the deprecated attribute.
19272 2018-06-15 Richard Biener <rguenther@suse.de>
19274 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
19275 here, also noting vector size used.
19276 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
19277 size used in MSG_OPTIMIZED_LOCATIONS dump.
19278 (pass_slp_vectorize::execute): Adjust.
19280 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
19283 * config/arc/arc.c (arc_return_address_register): Fix
19286 2018-06-15 Richard Biener <rguenther@suse.de>
19288 PR middle-end/86159
19289 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
19290 leave useless conversion stripping to force_gimple_operand_gsi.
19291 (gimplify_build2): Likewise.
19292 (gimplify_build1): Likewise.
19294 2018-06-15 Richard Biener <rguenther@suse.de>
19296 PR middle-end/86076
19297 * tree-cfg.c (move_stmt_op): unshare invariant addresses
19298 before adjusting their block.
19300 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
19302 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
19303 multilibs for *-*-rtems*.
19304 * config/riscv/t-rtems: New file.
19306 2018-06-14 Jakub Jelinek <jakub@redhat.com>
19308 PR middle-end/86122
19309 * match.pd ((A +- CST1) +- CST2): Punt if last resort
19310 unsigned_type_for returns NULL.
19313 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
19314 subregs of multi-word pseudos unless the float mode has word size.
19316 2018-06-14 Richard Biener <rguenther@suse.de>
19318 PR middle-end/86139
19319 * tree-vect-generic.c (build_word_mode_vector_type): Remove
19320 duplicate and harmful type_hash_canon.
19321 * tree.c (type_hash_canon): Assert we didn't find ourselves.
19323 2018-06-14 Richard Biener <rguenther@suse.de>
19326 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
19329 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
19331 * config/rtems.h (STDINT_LONG32): Define.
19333 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
19334 Prachi Godbole <prachi.godbole@imgtec.com>
19336 * config/mips/mips-cpus.def: Define P6600.
19337 * config/mips/mips-tables.opt: Regenerate.
19338 * config/mips/mips.c (mips_ucbranch_type): New enum.
19339 (mips_rtx_cost_data): Add support for P6600.
19340 (mips_issue_rate): Likewise.
19341 (mips_multipass_dfa_lookahead): Likewise.
19342 (mips_avoid_hazard): Likewise.
19343 (mips_reorg_process_insns): Likewise.
19344 (mips_classify_branch_p6600): New function.
19345 * config/mips/mips.h (TUNE_P6600): New define.
19346 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
19347 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
19348 * config/mips/mips.md: Include p6600.md.
19349 (processor): Add p6600.
19350 * config/mips/p6600.md: New file.
19351 * doc/invoke.texi: Add p6600 to supported architectures.
19353 2018-06-13 Martin Sebor <msebor@redhat.com>
19355 PR tree-optimization/86114
19356 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
19358 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
19360 2018-06-13 Richard Biener <rguenther@suse.de>
19362 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
19363 Properly set vector type of the intermediate stmt.
19364 * tree-vect-stmts.c (vectorizable_operation): The destination
19365 var always has vectype_out type.
19367 2018-06-13 Jeff Law <law@redhat.com>
19369 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
19370 integer 0 for argument to print_rtl_with_bb.
19371 (rl78_reorg): Likewise.
19373 2018-06-13 David Malcolm <dmalcolm@redhat.com>
19375 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
19376 from rtx to rtx_insn *.
19377 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
19379 (add_sched_insns_for_speculation): Likewise for local "target",
19380 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
19381 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
19382 from rtx_insn ** to rtx_code_label **.
19383 (reorg_emit_nops): Likewise.
19384 (c6x_reorg): Likewise for local "call_labels".
19385 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
19387 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
19388 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
19389 the loops over LABEL_REFS.
19390 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
19392 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
19393 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
19394 (split_branches): Strengthen local "olabel" from rtx to
19395 rtx_insn *, adding a safe_as_a cast.
19396 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
19398 (add_insn_after): Likewise for first two params.
19399 (add_insn_before): Likewise.
19400 (remove_insn): Likewise for param.
19401 (emit_pattern_before_noloc): Likewise for second and third params.
19402 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
19403 (emit_call_insn_before_noloc): Likewise.
19404 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
19406 (emit_barrier_before): Likewise.
19407 (emit_label_before): Strengthen "label" param from "rtx" to
19408 "rtx_code_label *". Strengthen "before" param from "rtx" to
19410 (emit_insn_after_1): Strengthen "after" param from "rtx" to
19412 (emit_pattern_after_noloc): Likewise.
19413 (emit_insn_after_noloc): Likewise.
19414 (emit_jump_insn_after_noloc): Likewise.
19415 (emit_call_insn_after_noloc): Likewise.
19416 (emit_debug_insn_after_noloc): Likewise.
19417 (emit_barrier_after): Likewise.
19418 (emit_label_after): Likewise for both params.
19419 (emit_pattern_after_setloc): Likewise for "after" param. Convert
19420 "loc" param from "int" to "location_t".
19421 (emit_insn_after_setloc): Likewise.
19422 (emit_jump_insn_after_setloc): Likewise.
19423 (emit_call_insn_after_setloc): Likewise.
19424 (emit_debug_insn_after_setloc): Likewise.
19425 (emit_pattern_before_setloc): Likewise for "before" param. Convert
19426 "loc" param from "int" to "location_t".
19427 (emit_pattern_before): Convert NULL_RTX to NULL.
19428 (emit_insn_before_setloc): Convert "loc" param from "int" to
19430 (emit_jump_insn_before_setloc): Likewise.
19431 (emit_call_insn_before_setloc): Likewise.
19432 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
19433 rtx_insn *. Convert "loc" param from "int" to "location_t".
19434 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
19435 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
19436 Convert 3rd param from "int" to "location_t".
19437 (emit_barrier_before, emit_barrier_after, next_real_insn):
19438 Strengthen param from rtx to rtx_insn *.
19439 (emit_label_before): Strengthen 1st param from "rtx" to
19440 "rtx_code_label *". Strengthen 2nd param from "rtx" to
19442 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
19443 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
19444 Strengthen 2nd param from "rtx" to "rtx_insn *".
19445 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
19446 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
19447 Likewise. Convert 3rd param from "int" to "location_t".
19448 (emit_label_after): Strengthen 1st param from "rtx" to
19449 "rtx_code_label *".
19450 (next_real_insn, remove_insn): Strengthen param from "rtx" to
19452 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
19453 from "rtx" to "rtx_insn *".
19455 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
19457 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
19458 bodies streamed in with -Q.
19459 * dumpfile.c (dump_files): Add lto-stream-out dump file.
19460 * dumpfile.h (tree_dump_index): Add lto_stream_out.
19461 * gimple-streamer-out.c: Include gimple-pretty-print.h
19462 (output_bb): Dump stmts streamed.
19463 * lto-section-out.c: Include print-tree.h
19464 (lto_begin_section): Dump sections created.
19465 (lto_output_decl_index): Dump decl encoded.
19466 * lto-streamer-out.c: Include print-tree.h
19467 (create_output_block): Dump output block created.
19468 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
19469 (output_function): Dump function output.
19470 (output_constructor): Dump constructor streamed.
19471 (write_global_stream): Output indexes encoded.
19472 (produce_asm_for_decls): Dump streams encoded.
19473 * lto-streamer.c (streamer_dump_file): New global var.
19474 * lto-streamer.h (streamer_dump_file): Declare.
19475 * passes.c (ipa_write_summaries): Initialize streamer dump.
19476 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
19479 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
19482 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
19483 offsets for register save directives. Emit a second batch of save
19484 directives, if need be, when the function accesses prior frames.
19486 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
19488 * config/arc/fpu.md (fmasf4): Force operand to register.
19489 (fnmasf4): Likewise.
19491 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
19493 * config/arc/arc-protos.h (arc_pad_return): Remove.
19494 * config/arc/arc.c (machine_function): Remove force_short_suffix
19496 (arc_print_operand): Adjust printing of '&'.
19497 (arc_verify_short): Remove conditional printing of short suffix.
19498 (arc_final_prescan_insn): Remove reference to size_reason.
19499 (pad_return): New function.
19500 (arc_reorg): Call pad_return.
19501 (arc_pad_return): Remove.
19502 (arc_init_machine_status): Remove reference to force_short_suffix.
19503 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
19504 (attr length): When attribute iscompact is true force to 2
19505 regardless; in the case of maybe check if we want to force the
19506 instruction to have 4 bytes length.
19507 (nopv): Change it to generate 4 byte long nop as well.
19508 (blockage): New pattern.
19509 (simple_return): Remove call to arc_pad_return.
19510 (p_return_i): Likewise.
19512 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
19514 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
19516 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
19518 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
19521 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
19523 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
19524 for ARC700 and ARCv2.
19526 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
19529 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
19530 operands[2] instead of operands[1].
19533 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
19535 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
19536 case, check whether the outer register overlaps an unallocatable
19537 register, not just whether it fits the required class.
19539 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
19541 * poly-int.h (can_div_trunc_p): Add new overload in which all values
19543 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
19544 (memrefs_conflict_p): Likewise.
19545 (init_alias_analysis): Likewise.
19546 * cfgexpand.c (expand_debug_expr): Likewise.
19547 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
19548 * cse.c (fold_rtx): Likewise.
19549 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
19550 * expr.c (emit_block_move_hints): Likewise.
19551 (clear_storage_hints, push_block, emit_push_insn): Likewise.
19552 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
19553 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
19554 (emit_group_store): Likewise.
19555 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
19556 to read the PRE/POST_MODIFY increment.
19557 * calls.c (store_one_arg): Use strip_offset.
19558 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
19560 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
19562 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
19563 (simplify_binary_operation_1): Extend CONST_INT handling to
19565 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
19566 than a HOST_WIDE_INT.
19567 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
19569 (adjust_mems, add_stores): Update accodingly.
19570 (vt_canonicalize_addr): Track polynomial offsets.
19571 (emit_note_insn_var_location): Likewise.
19572 (vt_add_function_parameter): Likewise.
19573 (vt_initialize): Likewise.
19575 2018-06-12 Jeff Law <law@redhat.com>
19577 * config.gcc (alpha*-*-freebsd*): Remove.
19578 * config/alpha/freebsd.h: Remove.
19580 2018-06-12 David Malcolm <dmalcolm@redhat.com>
19583 * spellcheck-tree.c (levenshtein_distance): Rename to...
19584 (get_edit_distance): ...this, and update for underlying renaming.
19585 * spellcheck-tree.h (levenshtein_distance): Rename to...
19586 (get_edit_distance): ...this.
19587 * spellcheck.c (levenshtein_distance): Rename to...
19588 (get_edit_distance): ...this. Convert from Levenshtein distance
19589 to Damerau-Levenshtein distance by supporting transpositions of
19590 adjacent characters. Rename "v1" to "v_next" and "v0" to
19592 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
19593 (selftest::test_edit_distance_unit_test_oneway): ...this, and
19594 update for underlying renaming.
19595 (selftest::levenshtein_distance_unit_test): Rename to...
19596 (selftest::test_get_edit_distance_unit): ...this, and update for
19597 underlying renaming.
19598 (selftest::test_find_closest_string): Add example from PR 69968
19599 where transposition helps
19600 (selftest::test_metric_conditions): Update for renaming.
19601 (selftest::test_metric_conditions): Likewise.
19602 (selftest::spellcheck_c_tests): Likewise.
19603 * spellcheck.h (levenshtein_distance): Rename both overloads to...
19604 (get_edit_distance): ...this.
19605 (best_match::consider): Update for renaming.
19607 2018-06-12 Martin Sebor <msebor@redhat.com>
19609 PR tree-optimization/85259
19610 * builtins.c (compute_objsize): Handle constant offsets.
19611 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
19612 true iff a warning has been issued.
19613 * gimple.h (gimple_nonartificial_location): New function.
19614 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
19615 gimple_nonartificial_location and handle -Wno-system-headers.
19616 (handle_builtin_stxncpy): Same.
19618 2018-06-12 Martin Sebor <msebor@redhat.com>
19621 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
19623 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
19625 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
19626 BUILTIN_VEC_XST entries for pointer to double and long long.
19628 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
19631 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
19633 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
19636 2018-06-12 Martin Liska <mliska@suse.cz>
19638 * doc/options.texi: Document IntegerRange.
19640 2018-06-12 Martin Liska <mliska@suse.cz>
19642 * config/i386/i386.opt: Make MPX-related options as Deprecated.
19643 * opt-functions.awk: Handle Deprecated flag.
19644 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
19646 (read_cmdline_option): Report warning for a deprecated option.
19647 * opts.h (struct cl_option): Add new field cl_deprecated.
19648 (CL_ERR_DEPRECATED): New.
19650 2018-06-12 Martin Liska <mliska@suse.cz>
19652 * doc/options.texi: Document Deprecated option flag.
19654 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
19656 * config/arc/arc-arch.h (arc_extras): New enum.
19657 (arc_cpu_t):Add field extra.
19658 (arc_cpu_types): Consider the extras.
19659 * config/arc/arc-cpus.def: Add extras info.
19660 * config/arc/arc-opts.h (processor_type): Consider extra field.
19661 * config/arc/arc.c (arc_override_options): Handle extra field.
19663 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
19665 * config/arc/arc-arch.h: Update ARC_OPTX macro.
19666 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
19668 * config/arc/arc.c (arc_init): Update pic warning.
19669 (irq_range): Update irq range parsing warnings.
19670 (arc_override_options): Update various warning messages.
19671 (arc_handle_aux_attribute): Likewise.
19673 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
19675 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
19677 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
19679 * doc/sourcebuild.texi: Document usage of line number 0 in verify
19680 compiler messages directives.
19682 2018-06-12 Matthew Fortune <mfortune@gmail.com>
19684 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
19685 * config/mips/mips-tables.opt: Regenerate.
19686 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
19688 * doc/invoke.texi: Document -march=i6500.
19690 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
19692 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
19693 (i6400_gpmul): Add cpu_unit.
19694 (i6400_gpdiv): Likewise.
19695 (i6400_msa_add_d): Update reservations.
19696 (i6400_msa_int_add) Likewise.
19697 (i6400_msa_short_logic3) Likewise.
19698 (i6400_msa_short_logic2) Likewise.
19699 (i6400_msa_short_logic) Likewise.
19700 (i6400_msa_move) Likewise.
19701 (i6400_msa_cmp) Likewise.
19702 (i6400_msa_short_float2) Likewise.
19703 (i6400_msa_div_d) Likewise.
19704 (i6400_msa_long_logic1) Likewise.
19705 (i6400_msa_long_logic2) Likewise.
19706 (i6400_msa_mult) Likewise.
19707 (i6400_msa_long_float2) Likewise.
19708 (i6400_msa_long_float4) Likewise.
19709 (i6400_msa_long_float5) Likewise.
19710 (i6400_msa_long_float8) Likewise.
19711 (i6400_fpu_fadd): Include frint type.
19712 (i6400_fpu_store): New define_insn_reservation.
19713 (i6400_fpu_load): Likewise.
19714 (i6400_fpu_move): Likewise.
19715 (i6400_fpu_fcmp): Likewise.
19716 (i6400_fpu_fmadd): Likewise.
19717 (i6400_int_mult): Include imul3nc type and update reservation.
19718 (i6400_int_div): Include idiv3 type and update reservation.
19719 (i6400_int_load): Update to check type not move_type.
19720 (i6400_int_store): Likewise.
19721 (i6400_int_prefetch): Set zero latency.
19723 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
19725 * gcc.c: Document new %@{...} sequence.
19726 (LINK_COMMAND_SPEC): Use it for the -L switches.
19727 (cpp_unique_options): Use it for the -I switches.
19728 (at_file_argbuf): New global variable.
19729 (in_at_file): Likewise.
19730 (alloc_args): Create at_file_argbuf.
19731 (clear_args): Truncate at_file_argbuf.
19732 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
19733 (open_at_file): New function.
19734 (close_at_file): Likewise.
19735 (create_at_file): Delete.
19736 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
19739 (validate_switches_from_spec): Deal with %@{...} sequence.
19740 (validate_switches): Likewise.
19741 (driver::finalize): Call clear_args.
19743 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
19745 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
19747 2018-06-11 Martin Sebor <msebor@redhat.com>
19749 * doc/invoke.texi (-Wall): List -Wc++17-compat.
19750 (Wno-class-memaccess): Add @opindex.
19751 (Wno-templates, Wno-multiple-inheritance): Same.
19752 (Wno-virtual-inheritance, Wno-namespaces): Same.
19753 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
19754 (Wno-format-overflow, Wno-format-truncation): Same.
19755 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
19756 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
19757 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
19758 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
19759 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
19760 (Wno-misspelled-isr): Same.
19762 2018-06-11 Martin Sebor <msebor@redhat.com>
19764 * PR tree-optimization/86083
19765 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
19767 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
19769 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
19771 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
19774 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
19775 on the correct operand.
19776 (*movdi_internal64): Ditto.
19778 2018-06-11 Martin Liska <mliska@suse.cz>
19780 PR tree-optimization/86089
19781 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
19783 2018-06-11 Julia Koval <julia.koval@intel.com>
19785 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
19786 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
19787 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
19789 2018-06-11 Olivier Hainque <hainque@adacore.com>
19791 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
19792 for Ada with strict dwarf2.
19794 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
19797 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
19800 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
19802 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
19804 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
19806 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
19809 2018-06-08 Martin Liska <mliska@suse.cz>
19811 * tree-cfg.h (debug_function): Fix argument type to match
19814 2018-06-08 Martin Liska <mliska@suse.cz>
19816 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
19817 Remove usage of MPX-related (and removed) fields.
19818 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
19820 2018-06-08 David Malcolm <dmalcolm@redhat.com>
19822 * cfg.c (debug): Use TDF_NONE rather than 0.
19823 * cfghooks.c (debug): Likewise.
19824 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
19825 (struct dump_option_value_info): Convert to...
19826 (struct kv_pair): ...this template type.
19827 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
19829 (optinfo_verbosity_options): Likewise.
19830 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
19832 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
19833 than int for "optgroup_flags" param.
19834 (dump_generic_expr_loc): Use dump_flags_t rather than int for
19836 (dump_dec): Likewise.
19837 (dump_finish): Use TDF_NONE rather than 0.
19838 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
19839 rather than int for "optgroup_flags" param. Use TDF_NONE rather
19840 than 0. Update for change to option_ptr.
19841 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
19842 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
19843 0. Update for changes to optinfo_verbosity_options and
19845 (opt_info_switch_p): Convert optgroup_flags from int to
19847 (dump_basic_block): Use dump_flags_t rather than int
19848 for "dump_kind" param.
19849 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
19850 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
19851 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
19852 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
19853 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
19854 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
19855 TDF_NONE): Convert from macros to...
19856 (enum dump_flag): ...this new enum.
19857 (dump_flags_t): Update to use enum.
19858 (operator|, operator&, operator~, operator|=, operator&=):
19859 Implement for dump_flags_t.
19860 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
19861 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
19862 Convert from macros to...
19863 (enum optgroup_flag): ...this new enum.
19864 (optgroup_flags_t): New typedef.
19865 (operator|, operator|=): Implement for optgroup_flags_t.
19866 (struct dump_file_info): Convert field "alt_flags" to
19867 dump_flags_t. Convert field "optgroup_flags" to
19869 (dump_basic_block): Use dump_flags_t rather than int for param.
19870 (dump_generic_expr_loc): Likewise.
19871 (dump_dec): Likewise.
19872 (dump_register): Convert param "optgroup_flags" to
19874 (opt_info_enable_passes): Likewise.
19875 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
19877 * gimple-pretty-print.c (debug): Likewise.
19878 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
19879 (merged_store_group::apply_stores): Likewise.
19880 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
19881 * gimple.c (verify_gimple_pp): Likewise.
19882 * graphite-poly.c (print_pbb_body): Likewise.
19883 * passes.c (pass_manager::register_one_dump_file): Convert
19884 local "optgroup_flags" to optgroup_flags_t.
19885 * print-tree.c (print_node): Use TDF_NONE rather than 0.
19887 (debug_body): Likewise.
19888 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
19889 to optgroup_flags_t.
19890 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
19892 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
19893 (convert_mult_to_fma): Likewise.
19894 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
19895 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
19896 * tree-vect-data-refs.c (dump_lower_bound): Convert param
19897 "dump_kind" to dump_flags_t.
19899 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
19901 * config/rs6000/rs6000.c (min, max): Delete.
19903 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
19905 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
19906 -mabi=spe and -mabi=no-spe.
19908 2018-06-08 Martin Liska <mliska@suse.cz>
19910 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
19911 where we expect an existing summary.
19913 2018-06-08 Martin Liska <mliska@suse.cz>
19915 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
19916 * ipa-inline.h (estimate_edge_growth): Likewise.
19918 2018-06-08 Martin Liska <mliska@suse.cz>
19920 * cgraph.c (function_version_hasher::hash): Use
19921 cgraph_node::get_uid ().
19922 (function_version_hasher::equal):
19923 * cgraph.h (cgraph_node::get_uid): New method.
19924 * ipa-inline.c (update_caller_keys): Use
19925 cgraph_node::get_uid ().
19926 (update_callee_keys): Likewise.
19927 * ipa-utils.c (searchc): Likewise.
19928 (ipa_reduced_postorder): Likewise.
19929 * lto-cgraph.c (input_node): Likewise.
19930 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
19931 * symbol-summary.h (symtab_insertion): Likewise.
19932 (symtab_removal): Likewise.
19933 (symtab_duplication): Likewise.
19934 * tree-pretty-print.c (dump_function_header): Likewise.
19935 * tree-sra.c (convert_callers_for_node): Likewise.
19937 2018-06-08 Martin Liska <mliska@suse.cz>
19939 * cgraph.c (symbol_table::create_edge): Always assign a new
19941 (symbol_table::free_edge): Do not recycle numbers.
19942 * cgraph.h (cgraph_edge::get): New method.
19943 * symbol-summary.h (symtab_removal): Use it.
19944 (symtab_duplication): Likewise.
19945 (call_summary::hashable_uid): Remove.
19947 2018-06-08 Martin Liska <mliska@suse.cz>
19949 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
19950 (initialize_growth_caches): Remove.
19951 (free_growth_caches): Likewise.
19952 (do_estimate_edge_time): Use edge_growth_cache.
19953 (do_estimate_edge_size): Likewise.
19954 (do_estimate_edge_hints): Likewise.
19955 * ipa-inline.c (reset_edge_caches): Likewise.
19956 (recursive_inlining): Likewise.
19957 (inline_small_functions): Likewise.
19958 * ipa-inline.h (initialize_growth_caches): Remove.
19959 (estimate_edge_size): Likewise.
19960 (estimate_edge_time): Likewise.
19961 (estimate_edge_hints): Likewise.
19962 (reset_edge_growth_cache): Likewise.
19963 * symbol-summary.h (call_summary::remove): New method.
19965 2018-06-08 Martin Liska <mliska@suse.cz>
19967 * ipa-cp.c (class edge_clone_summary): New summary.
19968 (grow_edge_clone_vectors): Remove.
19969 (ipcp_edge_duplication_hook): Remove.
19970 (class edge_clone_summary_t): New call_summary class.
19971 (ipcp_edge_removal_hook): Remove.
19972 (edge_clone_summary_t::duplicate): New function.
19973 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
19974 (create_specialized_node): Likewise.
19975 (ipcp_driver): Initialize edge_clone_summaries and do not
19978 2018-06-08 Martin Liska <mliska@suse.cz>
19980 * symbol-summary.h (get): New function.
19981 (call_summary::m_initialize_when_cloning): New class member.
19983 2018-06-08 Martin Liska <mliska@suse.cz>
19985 * cgraph.c (cgraph_node::remove): Do not recycle uid.
19986 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
19987 (symbol_table::allocate_cgraph_symbol): Do not set uid.
19988 * passes.c (uid_hash_t): Record removed_nodes by their uids.
19989 (remove_cgraph_node_from_order): Use the removed_nodes set.
19990 (do_per_function_toporder): Likwise.
19991 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
19992 instead of summary_uid.
19993 (symtab_removal): Likewise.
19994 (symtab_duplication): Likewise.
19996 2018-06-08 Martin Liska <mliska@suse.cz>
19998 * ipa-cp.c (ipcp_store_bits_results): Use
19999 ipcp_transformation_sum.
20000 (ipcp_store_vr_results): Likewise.
20001 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
20003 (ipcp_transformation_initialize): ... this.
20004 (ipa_set_node_agg_value_chain):
20005 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
20006 (write_ipcp_transformation_info): Likewise.
20007 (read_ipcp_transformation_info): Likewise.
20008 (ipcp_update_bits): Likewise.
20009 (ipcp_update_vr): Likewise.
20010 (ipcp_transform_function): Likewise.
20011 * ipa-prop.h: Rename ipcp_transformation_summary to
20012 ipcp_transformation.
20013 (class ipcp_transformation_t): New function summary.
20014 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
20015 (ipa_get_agg_replacements_for_node): Likewise.
20017 2018-06-08 Martin Liska <mliska@suse.cz>
20019 * ipa-pure-const.c (struct funct_state_d): Do it class instead
20021 (class funct_state_summary_t): New function_summary class.
20022 (has_function_state): Remove.
20023 (get_function_state): Likewise.
20024 (set_function_state): Likewise.
20025 (add_new_function): Likewise.
20026 (funct_state_summary_t::insert): New function.
20027 (duplicate_node_data): Remove.
20028 (remove_node_data): Remove.
20029 (funct_state_summary_t::duplicate): New function.
20030 (register_hooks): Create new funct_state_summaries.
20031 (pure_const_generate_summary): Use it.
20032 (pure_const_write_summary): Likewise.
20033 (pure_const_read_summary): Likewise.
20034 (propagate_pure_const): Likewise.
20035 (propagate_nothrow): Likewise.
20036 (dump_malloc_lattice): Likewise.
20037 (propagate_malloc): Likewise.
20038 (execute): Do not register hooks, just remove summary
20040 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
20043 2018-06-08 Martin Liska <mliska@suse.cz>
20045 * ipa-reference.c (remove_node_data): Remove.
20046 (duplicate_node_data): Likewise.
20047 (class ipa_ref_var_info_summary_t): New class.
20048 (class ipa_ref_opt_summary_t): Likewise.
20049 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
20050 (get_reference_optimization_summary): Use
20051 ipa_ref_opt_sum_summaries.
20052 (set_reference_vars_info): Remove.
20053 (set_reference_optimization_summary): Likewise.
20054 (ipa_init): Create summaries.
20055 (init_function_info): Use function summary.
20056 (ipa_ref_opt_summary_t::duplicate): New function.
20057 (ipa_ref_opt_summary_t::remove): New function.
20058 (get_read_write_all_from_node): Fix GNU coding style.
20059 (propagate): Use function summary.
20060 (write_node_summary_p): Fix GNU coding style.
20061 (stream_out_bitmap): Likewise.
20062 (ipa_reference_read_optimization_summary): Use function summary.
20063 (ipa_reference_c_finalize): Do not release hooks.
20065 2018-06-08 Martin Liska <mliska@suse.cz>
20067 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
20068 (analyze_function_body): Extract multiple calls of get_create.
20069 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
20070 * ipa-inline.c (recursive_inlining): Use ::get method.
20071 * ipa-inline.h (estimate_edge_growth): Likewise.
20073 2018-06-08 Martin Liska <mliska@suse.cz>
20075 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
20077 (hsa_function_summary::hsa_function_summary): Use the new enum
20079 (hsa_gpu_implementation_p): Use hsa_summaries::get.
20080 * hsa-gen.c (hsa_get_host_function): Likewise.
20081 (get_brig_function_name): Likewise.
20082 * ipa-hsa.c (process_hsa_functions): Likewise.
20083 (ipa_hsa_write_summary): Likewise.
20084 * symbol-summary.h (symtab_duplication): Use ::get function/
20085 (get): New function.
20087 2018-06-08 Martin Liska <mliska@suse.cz>
20089 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
20091 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
20092 (hsa_register_kernel): Likewise.
20093 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
20094 * hsa-gen.c (hsa_get_host_function): Likewise.
20095 (get_brig_function_name): Likewise.
20096 (generate_hsa): Likewise.
20097 (pass_gen_hsail::execute): Likewise.
20098 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
20099 (devirtualization_time_bonus): Likewise.
20100 (ipcp_propagate_stage): Likewise.
20101 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
20102 (edge_set_predicate): Likewise.
20103 (evaluate_conditions_for_known_args): Likewise.
20104 (evaluate_properties_for_edge): Likewise.
20105 (ipa_fn_summary::reset): Likewise.
20106 (ipa_fn_summary_t::duplicate): Likewise.
20107 (dump_ipa_call_summary): Likewise.
20108 (ipa_dump_fn_summary): Likewise.
20109 (analyze_function_body): Likewise.
20110 (compute_fn_summary): Likewise.
20111 (estimate_edge_devirt_benefit): Likewise.
20112 (estimate_edge_size_and_time): Likewise.
20113 (estimate_calls_size_and_time): Likewise.
20114 (estimate_node_size_and_time): Likewise.
20115 (inline_update_callee_summaries): Likewise.
20116 (remap_edge_change_prob): Likewise.
20117 (remap_edge_summaries): Likewise.
20118 (ipa_merge_fn_summary_after_inlining): Likewise.
20119 (ipa_update_overall_fn_summary): Likewise.
20120 (read_ipa_call_summary): Likewise.
20121 (inline_read_section): Likewise.
20122 (write_ipa_call_summary): Likewise.
20123 (ipa_fn_summary_write): Likewise.
20124 (ipa_free_fn_summary): Likewise.
20125 * ipa-hsa.c (process_hsa_functions): Likewise.
20126 (ipa_hsa_write_summary): Likewise.
20127 (ipa_hsa_read_section): Likewise.
20128 * ipa-icf.c (sem_function::merge): Likewise.
20129 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
20130 (do_estimate_edge_time): Likewise.
20131 (estimate_size_after_inlining): Likewise.
20132 (estimate_growth): Likewise.
20133 (growth_likely_positive): Likewise.
20134 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
20135 (inline_call): Likewise.
20136 * ipa-inline.c (caller_growth_limits): Likewise.
20137 (can_inline_edge_p): Likewise.
20138 (can_inline_edge_by_limits_p): Likewise.
20139 (compute_uninlined_call_time): Likewise.
20140 (compute_inlined_call_time): Likewise.
20141 (want_inline_small_function_p): Likewise.
20142 (edge_badness): Likewise.
20143 (update_caller_keys): Likewise.
20144 (update_callee_keys): Likewise.
20145 (recursive_inlining): Likewise.
20146 (inline_small_functions): Likewise.
20147 (inline_to_all_callers_1): Likewise.
20148 (dump_overall_stats): Likewise.
20149 (early_inline_small_functions): Likewise.
20150 (early_inliner): Likewise.
20151 * ipa-inline.h (estimate_edge_growth): Likewise.
20152 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
20153 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
20154 * ipa-prop.h (IPA_NODE_REF): Likewise.
20155 (IPA_EDGE_REF): Likewise.
20156 * ipa-pure-const.c (malloc_candidate_p): Likewise.
20157 (propagate_malloc): Likewise.
20158 * ipa-split.c (execute_split_functions): Likewise.
20159 * symbol-summary.h: Rename get to get_create.
20161 (get_create): Likewise.
20162 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
20164 2018-06-08 Martin Liska <mliska@suse.cz>
20166 * symbol-summary.h (release): Move definition out of class
20168 (symtab_removal): Likewise.
20169 (symtab_duplication): Likewise.
20171 2018-06-08 Martin Liska <mliska@suse.cz>
20173 * symbol-summary.h (function_summary): Move constructor
20174 implementation out of class declaration.
20175 (release): Likewise.
20176 (symtab_insertion): Likewise.
20177 (symtab_removal): Likewise.
20178 (symtab_duplication): Likewise.
20181 2018-06-08 Martin Liska <mliska@suse.cz>
20183 * Makefile.in: Remove support for MPX (macros, related functions,
20184 fields in cgraph_node, ...).
20185 * builtin-types.def (BT_BND): Likewise.
20186 (BT_FN_BND_CONST_PTR): Likewise.
20187 (BT_FN_CONST_PTR_BND): Likewise.
20188 (BT_FN_VOID_PTR_BND): Likewise.
20189 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
20190 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
20191 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
20192 (expand_builtin_mempcpy_with_bounds): Likewise.
20193 (expand_builtin_memset_with_bounds): Likewise.
20194 (expand_builtin_memset_args): Likewise.
20195 (std_expand_builtin_va_start): Likewise.
20196 (expand_builtin): Likewise.
20197 (expand_builtin_with_bounds): Likewise.
20198 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
20199 (DEF_LIB_BUILTIN_CHKP): Likewise.
20200 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
20201 (DEF_CHKP_BUILTIN): Likewise.
20202 (BUILT_IN_MEMCPY): Likewise.
20203 (BUILT_IN_MEMMOVE): Likewise.
20204 (BUILT_IN_MEMPCPY): Likewise.
20205 (BUILT_IN_MEMSET): Likewise.
20206 (BUILT_IN_STPCPY): Likewise.
20207 (BUILT_IN_STRCAT): Likewise.
20208 (BUILT_IN_STRCHR): Likewise.
20209 (BUILT_IN_STRCPY): Likewise.
20210 (BUILT_IN_STRLEN): Likewise.
20211 (BUILT_IN_MEMCPY_CHK): Likewise.
20212 (BUILT_IN_MEMMOVE_CHK): Likewise.
20213 (BUILT_IN_MEMPCPY_CHK): Likewise.
20214 (BUILT_IN_MEMSET_CHK): Likewise.
20215 (BUILT_IN_STPCPY_CHK): Likewise.
20216 (BUILT_IN_STRCAT_CHK): Likewise.
20217 (BUILT_IN_STRCPY_CHK): Likewise.
20218 * calls.c (store_bounds): Likewise.
20219 (emit_call_1): Likewise.
20220 (special_function_p): Likewise.
20221 (maybe_warn_nonstring_arg): Likewise.
20222 (initialize_argument_information): Likewise.
20223 (finalize_must_preallocate): Likewise.
20224 (compute_argument_addresses): Likewise.
20225 (expand_call): Likewise.
20226 * cfgexpand.c (expand_call_stmt): Likewise.
20227 (expand_return): Likewise.
20228 (expand_gimple_stmt_1): Likewise.
20229 (pass_expand::execute): Likewise.
20230 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
20231 (cgraph_node::remove): Likewise.
20232 (cgraph_node::dump): Likewise.
20233 (cgraph_node::verify_node): Likewise.
20234 * cgraph.h (chkp_function_instrumented_p): Likewise.
20235 (symtab_node::get_alias_target): Likewise.
20236 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
20237 (cgraph_local_p): Likewise.
20238 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
20239 (cgraph_edge::rebuild_references): Likewise.
20240 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
20241 (walk_polymorphic_call_targets): Likewise.
20242 (cgraph_node::expand_thunk): Likewise.
20243 (symbol_table::output_weakrefs): Likewise.
20244 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
20245 (ix86_handle_option): Likewise.
20246 * config/i386/constraints.md: Likewise.
20247 * config/i386/i386-builtin-types.def (BND): Likewise.
20251 * config/i386/i386-builtin.def (BDESC_END): Likewise.
20252 (BDESC_FIRST): Likewise.
20254 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
20255 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
20256 * config/i386/i386.c (enum reg_class): Likewise.
20257 (ix86_target_string): Likewise.
20258 (ix86_option_override_internal): Likewise.
20259 (ix86_conditional_register_usage): Likewise.
20260 (ix86_valid_target_attribute_inner_p): Likewise.
20261 (ix86_set_indirect_branch_type): Likewise.
20262 (ix86_set_current_function): Likewise.
20263 (ix86_function_arg_regno_p): Likewise.
20264 (init_cumulative_args): Likewise.
20265 (ix86_function_arg_advance): Likewise.
20266 (ix86_function_arg): Likewise.
20267 (ix86_pass_by_reference): Likewise.
20268 (ix86_function_value_regno_p): Likewise.
20269 (ix86_function_value_1): Likewise.
20270 (ix86_function_value_bounds): Likewise.
20271 (ix86_return_in_memory): Likewise.
20272 (ix86_setup_incoming_vararg_bounds): Likewise.
20273 (ix86_va_start): Likewise.
20274 (indirect_thunk_need_prefix): Likewise.
20275 (print_reg): Likewise.
20276 (ix86_print_operand): Likewise.
20277 (ix86_expand_call): Likewise.
20278 (ix86_output_function_return): Likewise.
20279 (reg_encoded_number): Likewise.
20280 (BDESC_VERIFYS): Likewise.
20281 (ix86_init_mpx_builtins): Likewise.
20282 (ix86_init_builtins): Likewise.
20283 (ix86_emit_cmove): Likewise.
20284 (ix86_emit_move_max): Likewise.
20285 (ix86_expand_builtin): Likewise.
20286 (ix86_builtin_mpx_function): Likewise.
20287 (ix86_get_arg_address_for_bt): Likewise.
20288 (ix86_load_bounds): Likewise.
20289 (ix86_store_bounds): Likewise.
20290 (ix86_load_returned_bounds): Likewise.
20291 (ix86_store_returned_bounds): Likewise.
20292 (ix86_class_likely_spilled_p): Likewise.
20293 (ix86_hard_regno_mode_ok): Likewise.
20294 (x86_order_regs_for_local_alloc): Likewise.
20295 (ix86_mitigate_rop): Likewise.
20296 (ix86_bnd_prefixed_insn_p): Likewise.
20297 (ix86_mpx_bound_mode): Likewise.
20298 (ix86_make_bounds_constant): Likewise.
20299 (ix86_initialize_bounds): Likewise.
20300 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
20301 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
20302 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
20303 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
20304 (TARGET_CHKP_BOUND_MODE): Likewise.
20305 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
20306 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
20307 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
20308 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
20309 * config/i386/i386.h (TARGET_MPX): Likewise.
20310 (TARGET_MPX_P): Likewise.
20311 (VALID_BND_REG_MODE): Likewise.
20312 (FIRST_BND_REG): Likewise.
20313 (LAST_BND_REG): Likewise.
20314 (enum reg_class): Likewise.
20315 (BND_REG_P): Likewise.
20316 (BND_REGNO_P): Likewise.
20317 (BNDmode): Likewise.
20318 (ADJUST_INSN_LENGTH): Likewise.
20319 * config/i386/i386.md: Likewise.
20320 * config/i386/i386.opt: Likewise.
20321 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
20322 (defined): Likewise.
20323 (LINK_MPX): Likewise.
20324 (MPX_SPEC): Likewise.
20325 (LIBMPX_SPEC): Likewise.
20326 (LIBMPXWRAPPERS_SPEC): Likewise.
20327 (CHKP_SPEC): Likewise.
20328 * config/i386/predicates.md: Likewise.
20329 * dbxout.c (dbxout_type): Likewise.
20330 * doc/extend.texi: Likewise.
20331 * doc/invoke.texi: Likewise.
20332 * doc/md.texi: Likewise.
20333 * doc/tm.texi: Likewise.
20334 * doc/tm.texi.in: Likewise.
20335 * dwarf2out.c (is_base_type): Likewise.
20336 (gen_formal_types_die): Likewise.
20337 (gen_subprogram_die): Likewise.
20338 (gen_type_die_with_usage): Likewise.
20339 (gen_decl_die): Likewise.
20340 (dwarf2out_late_global_decl): Likewise.
20341 * expr.c (expand_assignment): Likewise.
20342 (emit_storent_insn): Likewise.
20343 (store_expr_with_bounds): Likewise.
20344 (store_expr): Likewise.
20345 (expand_expr_real_1): Likewise.
20346 * expr.h (store_expr_with_bounds): Likewise.
20347 * function.c (use_register_for_decl): Likewise.
20348 (struct bounds_parm_data): Likewise.
20349 (assign_parms_augmented_arg_list): Likewise.
20350 (assign_parm_find_entry_rtl): Likewise.
20351 (assign_parm_is_stack_parm): Likewise.
20352 (assign_parm_load_bounds): Likewise.
20353 (assign_bounds): Likewise.
20354 (assign_parms): Likewise.
20355 (expand_function_start): Likewise.
20356 * gcc.c (CHKP_SPEC): Likewise.
20357 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
20358 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
20359 (wrestrict_dom_walker::check_call): Likewise.
20360 * gimple.c (gimple_build_call_from_tree): Likewise.
20361 * gimple.h (enum gf_mask): Likewise.
20362 (gimple_call_with_bounds_p): Likewise.
20363 (gimple_call_set_with_bounds): Likewise.
20364 * gimplify.c (gimplify_init_constructor): Likewise.
20365 * ipa-cp.c (initialize_node_lattices): Likewise.
20366 (propagate_constants_across_call): Likewise.
20367 (find_more_scalar_values_for_callers_subset): Likewise.
20368 * ipa-hsa.c (process_hsa_functions): Likewise.
20369 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
20370 * ipa-icf.c (sem_function::merge): Likewise.
20371 * ipa-inline.c (early_inliner): Likewise.
20372 * ipa-pure-const.c (warn_function_noreturn): Likewise.
20373 (warn_function_cold): Likewise.
20374 (propagate_pure_const): Likewise.
20375 * ipa-ref.h (enum GTY): Likewise.
20376 * ipa-split.c (find_retbnd): Likewise.
20377 (consider_split): Likewise.
20378 (split_function): Likewise.
20379 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
20380 * ipa.c (walk_polymorphic_call_targets): Likewise.
20381 (symbol_table::remove_unreachable_nodes): Likewise.
20382 (process_references): Likewise.
20383 (cgraph_build_static_cdtor_1): Likewise.
20384 * lto-cgraph.c (lto_output_node): Likewise.
20385 (output_refs): Likewise.
20386 (compute_ltrans_boundary): Likewise.
20387 (input_overwrite_node): Likewise.
20388 (input_node): Likewise.
20389 (input_cgraph_1): Likewise.
20390 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
20391 * passes.c (pass_manager::execute_early_local_passes): Likewise.
20392 (class pass_chkp_instrumentation_passes): Likewise.
20393 (make_pass_chkp_instrumentation_passes): Likewise.
20394 * passes.def: Likewise.
20395 * rtl.h (struct GTY): Likewise.
20396 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
20397 * stor-layout.c (layout_type): Likewise.
20398 * symtab.c: Likewise.
20399 * target.def: Likewise.
20400 * targhooks.c (default_chkp_bound_type): Likewise.
20401 (default_chkp_bound_mode): Likewise.
20402 (default_builtin_chkp_function): Likewise.
20403 (default_chkp_function_value_bounds): Likewise.
20404 (default_chkp_make_bounds_constant): Likewise.
20405 (default_chkp_initialize_bounds): Likewise.
20406 * targhooks.h (default_chkp_bound_type): Likewise.
20407 (default_chkp_bound_mode): Likewise.
20408 (default_builtin_chkp_function): Likewise.
20409 (default_chkp_function_value_bounds): Likewise.
20410 (default_chkp_make_bounds_constant): Likewise.
20411 (default_chkp_initialize_bounds): Likewise.
20412 * toplev.c (compile_file): Likewise.
20413 (process_options): Likewise.
20414 * tree-core.h (DEF_BUILTIN): Likewise.
20415 (DEF_BUILTIN_CHKP): Likewise.
20416 * tree-inline.c (declare_return_variable): Likewise.
20417 (remap_gimple_stmt): Likewise.
20418 (copy_bb): Likewise.
20419 (initialize_inlined_parameters): Likewise.
20420 (expand_call_inline): Likewise.
20421 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
20422 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
20423 (make_pass_ipa_chkp_produce_thunks): Likewise.
20424 (make_pass_chkp): Likewise.
20425 (make_pass_chkp_opt): Likewise.
20426 (make_pass_chkp_instrumentation_passes): Likewise.
20427 * tree-pretty-print.c (dump_generic_node): Likewise.
20428 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
20429 * tree-ssa-dce.c (propagate_necessity): Likewise.
20430 (eliminate_unnecessary_stmts): Likewise.
20431 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
20432 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
20433 * tree-ssa-sccvn.h: Likewise.
20434 * tree-ssa-strlen.c (get_string_length): Likewise.
20435 (valid_builtin_call): Likewise.
20436 (adjust_last_stmt): Likewise.
20437 (handle_builtin_strchr): Likewise.
20438 (handle_builtin_strcpy): Likewise.
20439 (handle_builtin_stxncpy): Likewise.
20440 (handle_builtin_memcpy): Likewise.
20441 (handle_builtin_strcat): Likewise.
20442 (strlen_check_and_optimize_stmt): Likewise.
20443 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
20444 * tree-streamer-in.c: Likewise.
20445 * tree-streamer.c (record_common_node): Likewise.
20446 * tree.c (tree_code_size): Likewise.
20447 (wide_int_to_tree_1): Likewise.
20448 (type_contains_placeholder_1): Likewise.
20449 (build_common_tree_nodes): Likewise.
20450 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
20451 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
20452 (POINTER_BOUNDS_P): Likewise.
20453 (BOUNDED_TYPE_P): Likewise.
20454 (BOUNDED_P): Likewise.
20455 (CALL_WITH_BOUNDS_P): Likewise.
20456 (pointer_bounds_type_node): Likewise.
20457 * value-prof.c (gimple_ic): Likewise.
20458 * var-tracking.c (vt_add_function_parameters): Likewise.
20459 * varasm.c (make_decl_rtl): Likewise.
20460 (assemble_start_function): Likewise.
20461 (output_constant): Likewise.
20462 (maybe_assemble_visibility): Likewise.
20463 * varpool.c (ctor_for_folding): Likewise.
20464 * chkp-builtins.def: Remove.
20465 * ipa-chkp.c: Remove.
20466 * ipa-chkp.h: Remove.
20467 * rtl-chkp.c: Remove.
20468 * rtl-chkp.h: Remove.
20469 * tree-chkp-opt.c: Remove.
20470 * tree-chkp.c: Remove.
20471 * tree-chkp.h: Remove.
20473 2018-06-07 Carl Love <cel@us.ibm.com>
20475 * config/rs6000/vsx.md (vextract_fp_from_shorth,
20476 vextract_fp_from_shortl): Add BE support.
20478 2018-06-07 Paul Koning <ni1d@arrl.net>
20480 * compare-elim.c (try_merge_compare): Don't merge compare if
20481 address contains a side effect.
20482 (try_eliminate_compare): Likewise.
20484 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
20486 * config.gcc: Support "tremont".
20487 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
20488 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20490 * config/i386/i386.c (m_TREMONT): Define.
20491 (processor_target_table): Add "tremont".
20492 (PTA_TREMONT): Define.
20493 (ix86_lea_outperforms): Add TARGET_TREMONT.
20494 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
20495 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
20496 and M_INTEL_GOLDMONT_PLUS.
20497 (fold_builtin_cpu): Add "tremont".
20498 (ix86_add_stmt_cost): Add TARGET_TREMONT.
20499 (ix86_option_override_internal): Add "tremont".
20500 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
20501 (processor_type): Add PROCESSOR_TREMONT.
20502 * config/i386/x86-tune.def: Add m_TREMONT.
20503 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
20505 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
20507 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
20508 symbol defined for msp430i* devices to be lower case.
20510 2018-06-07 Richard Biener <rguenther@suse.de>
20512 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
20513 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
20514 Properly wrap signed arithmetic if overflow wraps.
20516 2018-06-07 Jakub Jelinek <jakub@redhat.com>
20518 PR tree-optimization/69615
20519 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
20520 of a cast from a same precision integral SSA_NAME in a bb dominated
20521 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
20522 cast to utype if rhs2 has already a compatible type.
20524 2018-06-07 Richard Biener <rguenther@suse.de>
20526 PR tree-optimization/85935
20527 * graphite-scop-detection.c (find_params_in_bb): Analyze
20528 condition operands with respect to the correct loop. Assert
20529 the analysis doesn't fail.
20531 2018-06-04 Carl Love <cel@us.ibm.com>
20533 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
20534 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
20535 as it is slightly cheaper.
20536 (first_match_or_eos_index_<mode>):
20537 Calculate index using natural element order.
20538 (first_match_index_<mode>):
20539 Calculate index using natural element order.
20540 (first_match_or_eos_index_<mode>):
20541 Calculate index using natural order.
20542 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
20543 for BE and LE modes.
20544 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
20545 P9V_BUILTIN_VCLZLSBB_V16QI.
20546 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
20549 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
20551 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
20552 indentation and line wrap for many prototypes. Add missing
20553 @smallexample directives around block of prototypes for vec_xl and
20556 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
20558 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
20559 track if we pass or return IEEE 128-bit floating point.
20560 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
20561 C++ mangling that is compatible with GCC 8.1.
20562 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
20563 (init_cumulative_args): Note if we pass or return IEEE 128-bit
20564 floating point types.
20565 (rs6000_function_arg_advance_1): Likewise.
20566 (rs6000_mangle_type): Optionally generate mangled names that match
20567 what GCC 8.1 generated for IEEE 128-bit floating point types.
20568 (rs6000_globalize_decl_name): If we have an external function that
20569 passes or returns IEEE 128-bit types, generate a weak reference
20570 from the mangled name used in GCC 8.1 to the current mangled
20572 (rs6000_init_builtins): Make __ibm128 use the long double type if
20573 long double is IBM extended double. Make __float128 use the long
20574 double type if long double is IEEE 128-bit.
20577 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
20578 macro for __ibm128 built-in functions.
20579 (PACK_IF): Add __ibm128 pack/unpack functions.
20580 (UNPACK_IF): Likewise.
20581 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
20582 enable long double built-in functions if long double is IEEE
20583 128-bit floating point.
20584 (rs6000_invalid_builtin): Update long double built-in function
20586 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
20587 functions, adjust the built-in function to use the long double
20588 built-in function if __ibm128 and long double are the same type.
20589 * doc/extend.texi (PowerPC builtins): Update documention for
20590 __builtin_{,un}pack_longdouble. Add documentation for
20591 __builtin_{,un}pack_ibm128.
20593 2018-06-06 Jim Wilson <jimw@sifive.com>
20595 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
20596 (struct machine_function): New field interrupt_mode.
20597 (riscv_handle_type_attribute): New function. Add forward declaration.
20598 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
20599 (riscv_expand_epilogue): Check interrupt_mode field.
20600 (riscv_set_current_function): Check interrupt attribute args and
20601 set interrupt_mode field.
20602 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
20603 (riscv_sret, riscv_uret): New.
20604 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
20605 new arguments to interrupt attribute.
20607 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
20610 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
20611 Don't handle -mcpu=power8 if -mpower9-vector is also used.
20613 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
20615 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
20616 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
20617 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
20618 several redundant entries.
20620 2018-06-06 David Malcolm <dmalcolm@redhat.com>
20622 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
20623 type from "rtx" to "rtx_insn *".
20624 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
20625 for local "call_insn", removing cast.
20626 (ix86_expand_call): Likewise, introducing a "call_insn" local.
20628 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
20630 PR tree-optimization/86066
20631 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
20632 for BIT_INSERT_EXPR stores.
20634 2018-06-06 Richard Biener <rguenther@suse.de>
20636 PR tree-optimization/86062
20637 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
20638 component refs ontop
20639 of to be offsetted base.
20641 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
20643 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
20644 to be static and remove check on interrupt attribute name.
20646 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
20648 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
20649 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
20651 2018-06-05 Steve Ellcey <sellcey@cavium.com>
20654 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
20656 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
20657 Remove second argument, change how error is called.
20658 (aarch64_layout_arg): Remove second argument from
20659 aarch64_err_no_fpadvsimd call.
20660 (aarch64_init_cumulative_args): Ditto.
20661 (aarch64_gimplify_va_arg_expr): Ditto.
20662 * config/aarch64/aarch64.md (mov<mode>): Ditto.
20664 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
20666 * config/i386/i386.md (simple_return_indirect_internal): New expander.
20667 (*simple_return_indirect_internal<mode>): Rename from
20668 simple_return_indirect_internal. Use W mode iterator.
20669 (rstorssp): New expander.
20670 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
20671 (clrssbsy): New expander.
20672 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
20674 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
20676 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
20678 (cmse_check_pointed_object): Likewise.
20680 2018-06-05 Martin Liska <mliska@suse.cz>
20682 PR gcov-profile/47618
20683 * doc/invoke.texi: Document how -fprofile-dir format
20686 2018-06-05 Richard Biener <rguenther@suse.de>
20688 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
20689 removal pretend DOM info isn't available so we do not update
20690 it and only remove edges, not dominated blocks. Actually free
20691 DOM info in case we removed something. Remove unreachable blocks.
20692 (mfb_keep_latches): Work with either DOM info or marked backedges.
20693 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
20694 first. Mark backedges if DOM info isn't available.
20695 (Re-)compute DOM info after cleanup_control_flow_pre.
20697 2018-06-05 Richard Biener <rguenther@suse.de>
20699 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
20700 (locus_discrim_hasher::hash): Adjust.
20701 (locus_discrim_hasher::equal): Likewise.
20702 (next_discriminator_for_locus): Work on line directly.
20703 (same_line_p): Pass in expanded locus1 as well.
20704 (assign_discriminators): Avoid redundant location expansions.
20706 2018-06-05 Richard Biener <rguenther@suse.de>
20708 PR tree-optimization/86046
20709 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
20710 if required after clearing TREE_ADDRESSABLE.
20712 2018-06-05 Richard Biener <rguenther@suse.de>
20714 PR tree-optimization/86047
20715 * tree-ssa-loop.c (for_each_index): Glob handling of all
20716 decls and constants and really handle all of them.
20718 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20721 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
20722 qualifier_void_pointer and qualifier_const_void_pointer.
20723 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
20724 (arm_init_builtins): Handle the above.
20725 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
20726 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
20727 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
20730 2018-06-05 Martin Liska <mliska@suse.cz>
20732 * auto-profile.c (read_autofdo_file): Do not use
20733 gcov_ctr_summary struct.
20734 (afdo_callsite_hot_enough_for_early_inline): Likewise.
20735 * coverage.c (struct counts_entry): Likewise.
20736 (read_counts_file): Read just single summary entry.
20737 (get_coverage_counts): Use gcov_summary struct.
20738 * coverage.h (get_coverage_counts): Likewise.
20739 * gcov-dump.c (dump_working_sets): Likewise.
20740 (tag_summary): Dump just single summary.
20741 * gcov-io.c (gcov_write_summary): Write just histogram
20743 (gcov_read_summary): Read just single summary.
20744 (compute_working_sets): Use gcov_summary struct.
20745 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
20746 of GCOV_COUNTERS_SUMMABLE.
20747 (GCOV_COUNTERS_SUMMABLE): Remove.
20748 (GCOV_FIRST_VALUE_COUNTER): Replace with
20749 GCOV_COUNTER_V_INTERVAL.
20750 (struct gcov_ctr_summary): Remove.
20751 (struct gcov_summary): Directly use fields of former
20753 (compute_working_sets): Use gcov_summary struct.
20754 * gcov.c (read_count_file): Do not use ctrs fields.
20755 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
20757 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
20759 * profile.c: Likewise.
20760 * profile.h: Likewise.
20762 2018-06-05 Martin Liska <mliska@suse.cz>
20764 PR gcov-profile/84846
20765 * gcov.c (output_lines): Print working directory only
20766 in intermediate format.
20768 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
20770 * config/s390/s390-builtin-types.def: Add void function type.
20771 * config/s390/s390-builtins.def: Use the function type for the
20774 2018-06-04 Jim Wilson <jimw@sifive.com>
20776 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
20778 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
20779 and maybe_eh_return. Change regno to unsigned int. Use new args to
20780 handle EH_RETURN_DATA_REGNO registers properly.
20781 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
20782 (riscv_expand_epilogue): Update comment. Change argument name and
20783 type. Update code to use new name and type. Pass new args to
20784 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
20786 * config/riscv/riscv.md (NORMAL_RETURN): New.
20787 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
20788 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
20789 (eh_return): Call gen_eh_return_internal and emit barrier.
20790 (eh_return_internal): Call riscv_expand_epilogue.
20792 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
20794 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
20795 bit_insertion field and declare can_be_merged_into method.
20796 (merged_store_group::can_be_merged_into): New method.
20797 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
20798 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
20799 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
20801 2018-06-04 Richard Biener <rguenther@suse.de>
20803 PR tree-optimization/85955
20804 * builtins.c (fold_builtin_sincos): Convert pointers to
20805 destination to appropriate type before dereferencing.
20807 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
20809 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
20811 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
20813 * expr.c (expand_expr_real_1): Force the operand into memory if
20814 its TYPE_MODE is BLKmode and if there is no integer mode for
20815 the number of bits being extracted.
20817 2018-06-04 Jakub Jelinek <jakub@redhat.com>
20821 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
20822 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
20824 2018-06-04 Richard Biener <rguenther@suse.de>
20826 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
20827 (cleanup_tree_cfg_noloop): ... single caller. Do
20828 start_recording_case_labels later.
20830 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
20832 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
20833 to _IMMINTRIN_H_INCLUDED.
20834 * config/i386/pconfigintrin.h: Ditto.
20835 * config/i386/waitpkgintrin.h: Ditto.
20836 * config/i386/immintrin.h: Add includes for sgxintrin.h,
20837 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
20838 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
20839 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
20840 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
20841 waitpkgintrin.h and cldemoteintrin.h.
20843 2018-06-04 Richard Biener <rguenther@suse.de>
20845 PR tree-optimization/86038
20846 * tracer.c (find_best_successor): Check probability for
20847 being initialized, bail out if not.
20849 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
20852 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
20853 of bits to ignore when comparing architectures.
20855 2018-06-04 Jakub Jelinek <jakub@redhat.com>
20857 PR tree-optimization/69615
20858 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
20859 maximum or minimum of the type, try to merge it also as if
20860 range1 is + [-, x - 1] or + [x + 1, -].
20863 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
20865 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
20867 PR tree-optimization/86034
20868 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
20869 the unsigned bitfield type in a bit insertion sequence if it does not
20870 have a larger precision than the bitfield size.
20871 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
20873 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
20875 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
20877 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
20879 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
20880 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
20881 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
20882 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
20884 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
20886 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
20887 Disable -fdelete-null-pointer-checks for ELF toolchain.
20889 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
20890 Kito Cheng <kito.cheng@gmail.com>
20892 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
20893 (nds32le-*-*, nds32be-*-*): Integrate checking process.
20894 (nds32*-*-*): Add glibc and uclibc conditions.
20895 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
20896 (TARGET_EXCEPT_UNWIND_INFO): Define.
20897 * config/nds32/elf.h: New file.
20898 * config/nds32/linux.h: New file.
20899 * config/nds32/nds32-elf.opt: New file.
20900 * config/nds32/nds32-linux.opt: New file.
20901 * config/nds32/nds32-fp-as-gp.c
20902 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
20903 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
20905 (nds32_asm_file_end): Ditto.
20906 (nds32_print_operand): Ditto.
20907 (nds32_insert_attributes): Ditto.
20908 (nds32_init_libfuncs): New function.
20909 (TARGET_HAVE_TLS): Define.
20910 (TARGET_INIT_LIBFUNCS): Define.
20911 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
20913 (TARGET_ELF): Apply different mcmodel setting.
20914 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
20915 been migrated into elf.h and linux.h files.
20916 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
20917 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
20918 (mcmodel): The content has been migrated into nds32-elf.opt and
20919 nds32-linux.opt files.
20920 * config/nds32/t-elf: New file.
20921 * config/nds32/t-linux: New file.
20923 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
20924 Shiva Chen <shiva0217@gmail.com>
20926 * config/nds32/constants.md (unspec_volatile_element): Add
20927 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
20928 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
20930 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
20931 (make_pass_nds32_fp_as_gp): Declare.
20932 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
20934 (nds32_asm_function_end_prologue): Remove unused asm output.
20935 (nds32_asm_function_begin_epilogue): Remove unused asm output.
20936 (nds32_asm_file_start): Output necessary fp_as_gp information.
20937 (nds32_option_override): Adjust register usage.
20938 (nds32_expand_prologue): Consider fp_as_gp situation.
20939 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
20940 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
20943 (simple_return): Ditto.
20944 (omit_fp_begin): Output special directive for fp_as_gp.
20945 (omit_fp_end): Output special directive for fp_as_gp.
20946 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
20947 mforbid-fp-as-gp): New options.
20949 2018-06-01 Mark Wielaard <mark@klomp.org>
20951 * dwarf2out.c (dwarf2out_finish): Remove generation of
20952 DW_AT_loclists_base.
20954 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
20956 * gimple-ssa-store-merging.c: Include gimple-fold.h.
20957 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
20958 (struct merged_store_group): Add bit_insertion field.
20959 (dump_char_array): Use standard hexadecimal format.
20960 (merged_store_group::merged_store_group): Set bit_insertion to false.
20961 (merged_store_group::apply_stores): Use optimal buffer size. Deal
20962 with BIT_INSERT_EXPR stores. Move up code updating the mask and
20963 also print the mask in the dump file.
20964 (pass_store_merging::gate): Minor tweak.
20965 (imm_store_chain_info::coalesce_immediate): Fix wrong association
20966 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
20967 stores with INTEGER_CST stores.
20968 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
20969 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
20970 and use it throughout. Generate bit insertion sequences if need be.
20971 (pass_store_merging::process_store): Remove redundant condition.
20972 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
20974 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
20976 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
20977 the 128-bit floating point types. Fix function comment.
20979 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20981 * config/aarch64/aarch64-simd.md
20982 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
20984 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
20987 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
20989 PR tree-optimization/85989
20990 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
20992 (backprop::intersect_uses): Check it when deciding whether this
20993 is a backedge reference.
20994 (backprop::process_block): Add each phi to m_visited_phis
20995 after visiting it, then clear it at the end.
20997 2018-06-01 Richard Biener <rguenther@suse.de>
20999 * tree-vectorizer.h (vect_dr_stmt): New function.
21000 (vect_get_load_cost): Adjust.
21001 (vect_get_store_cost): Likewise.
21002 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
21003 Use vect_dr_stmt instead of DR_SMTT.
21004 (vect_record_base_alignments): Likewise.
21005 (vect_calculate_target_alignment): Likewise.
21006 (vect_compute_data_ref_alignment): Likewise and make static.
21007 (vect_update_misalignment_for_peel): Likewise.
21008 (vect_verify_datarefs_alignment): Likewise.
21009 (vector_alignment_reachable_p): Likewise.
21010 (vect_get_data_access_cost): Likewise. Pass down
21011 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
21012 (vect_get_peeling_costs_all_drs): Likewise.
21013 (vect_peeling_hash_get_lowest_cost): Likewise.
21014 (vect_enhance_data_refs_alignment): Likewise.
21015 (vect_find_same_alignment_drs): Likewise.
21016 (vect_analyze_data_refs_alignment): Likewise.
21017 (vect_analyze_group_access_1): Likewise.
21018 (vect_analyze_group_access): Likewise.
21019 (vect_analyze_data_ref_access): Likewise.
21020 (vect_analyze_data_ref_accesses): Likewise.
21021 (vect_vfa_segment_size): Likewise.
21022 (vect_small_gap_p): Likewise.
21023 (vectorizable_with_step_bound_p): Likewise.
21024 (vect_prune_runtime_alias_test_list): Likewise.
21025 (vect_analyze_data_refs): Likewise.
21026 (vect_supportable_dr_alignment): Likewise.
21027 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
21028 (vect_gen_prolog_loop_niters): Likewise.
21029 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
21030 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
21032 (vect_recog_mask_conversion_pattern): Likewise.
21033 (vect_try_gather_scatter_pattern): Likewise.
21034 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
21035 to vect_get_store_cost.
21036 (vect_get_store_cost): Get stmt_info instead of DR.
21037 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
21038 (vect_get_load_cost): Get stmt_info instead of DR.
21040 2018-06-01 Richard Biener <rguenther@suse.de>
21042 PR middle-end/86017
21043 * gimple-fold.c (var_decl_component_p): Also allow offsetted
21044 vars wrapped in MEM_REFs.
21046 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
21048 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
21049 Fix subreg tests so that we only return a choice between
21050 GENERAL_REGS and FP_REGS if the original classes included both.
21052 2018-06-01 Richard Biener <rguenther@suse.de>
21055 * tree-ssa-structalias.c (get_function_part_constraint):
21056 Handle NULL fi->decl.
21057 (find_func_aliases_for_call): Properly handle indirect
21058 fi from direct call.
21059 (find_func_clobbers): Likewise.
21060 (ipa_pta_execute): Likewise.
21061 (create_variable_info_for): For functions that are ifunc_resolver
21062 resolve to a varinfo that contains the result of the resolver call.
21063 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
21066 2018-05-31 Michael Collison <michael.collison@arm.com>
21068 * config/aarch64/aarch64.md:
21069 (*fix_to_zero_extenddfdi2): New pattern.
21070 * gcc.target/aarch64/fix_extend1.c: New testcase.
21072 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
21074 PR middle-end/78809
21075 PR middle-end/83026
21076 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
21077 and BUILT_IN_STRNCMP_EQ.
21078 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
21079 BUILT_IN_STRNCMP_EQ.
21080 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
21081 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
21082 (gimple_fold_builtin): Likewise.
21083 * tree-ssa-strlen.c (compute_string_length): New function.
21084 (determine_min_obsize): New function.
21085 (handle_builtin_string_cmp): New function to handle calls to
21086 string compare functions.
21087 (strlen_optimize_stmt): Add handling to builtin string compare
21089 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
21090 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
21091 * tree.c (build_common_builtin_nodes): Add new defines of
21092 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
21094 2018-05-31 Jakub Jelinek <jakub@redhat.com>
21097 * bb-reorder.c (pass_partition_blocks::gate): Return false for
21098 functions with naked attribute.
21100 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
21102 * config/i386/sse.md (avx_vec_concat<mode>):
21103 Substitute concat_tg_mode mode attribute with xtg_mode.
21104 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
21105 (concat_tg_mode): Remove mode attribute.
21107 2018-05-31 Martin Sebor <msebor@redhat.com>
21110 * calls.c (alloc_max_size): Correct a logic error/typo.
21111 Treat excessive arguments as infinite. Warn for invalid arguments.
21112 * doc/invoke.texi (-Walloc-size-larger-than): Update.
21114 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
21117 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
21118 and movx for Haswell.
21120 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
21121 Cesar Philippidis <cesar@codesourcery.com>
21123 PR middle-end/85879
21124 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
21125 when emitting error on private/firstprivate reductions.
21126 * omp-low.c (lower_omp_target): Avoid reference-type processing
21127 on pointers for firstprivate clause.
21129 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
21131 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
21134 * config/aarch64/aarch64-simd.md
21135 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
21136 (aarch64_ld1_x3_<mode>): Likewise
21137 (aarch64_st1x2<VALLDIF:mode>): Likewise
21138 (aarch64_st1_x2_<mode>): Likewise
21139 (aarch64_st1x3<VALLDIF:mode>): Likewise
21140 (aarch64_st1_x3_<mode>): Likewise
21141 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
21142 (vld1_s8_x3): Likewise.
21143 (vld1_u16_x3): Likewise.
21144 (vld1_s16_x3): Likewise.
21145 (vld1_u32_x3): Likewise.
21146 (vld1_s32_x3): Likewise.
21147 (vld1_u64_x3): Likewise.
21148 (vld1_s64_x3): Likewise.
21149 (vld1_f16_x3): Likewise.
21150 (vld1_f32_x3): Likewise.
21151 (vld1_f64_x3): Likewise.
21152 (vld1_p8_x3): Likewise.
21153 (vld1_p16_x3): Likewise.
21154 (vld1_p64_x3): Likewise.
21155 (vld1q_u8_x3): Likewise.
21156 (vld1q_s8_x3): Likewise.
21157 (vld1q_u16_x3): Likewise.
21158 (vld1q_s16_x3): Likewise.
21159 (vld1q_u32_x3): Likewise.
21160 (vld1q_s32_x3): Likewise.
21161 (vld1q_u64_x3): Likewise.
21162 (vld1q_s64_x3): Likewise.
21163 (vld1q_f16_x3): Likewise.
21164 (vld1q_f32_x3): Likewise.
21165 (vld1q_f64_x3): Likewise.
21166 (vld1q_p8_x3): Likewise.
21167 (vld1q_p16_x3): Likewise.
21168 (vld1q_p64_x3): Likewise.
21169 (vst1_s64_x2): Likewise.
21170 (vst1_u64_x2): Likewise.
21171 (vst1_f64_x2): Likewise.
21172 (vst1_s8_x2): Likewise.
21173 (vst1_p8_x2): Likewise.
21174 (vst1_s16_x2): Likewise.
21175 (vst1_p16_x2): Likewise.
21176 (vst1_s32_x2): Likewise.
21177 (vst1_u8_x2): Likewise.
21178 (vst1_u16_x2): Likewise.
21179 (vst1_u32_x2): Likewise.
21180 (vst1_f16_x2): Likewise.
21181 (vst1_f32_x2): Likewise.
21182 (vst1_p64_x2): Likewise.
21183 (vst1q_s8_x2): Likewise.
21184 (vst1q_p8_x2): Likewise.
21185 (vst1q_s16_x2): Likewise.
21186 (vst1q_p16_x2): Likewise.
21187 (vst1q_s32_x2): Likewise.
21188 (vst1q_s64_x2): Likewise.
21189 (vst1q_u8_x2): Likewise.
21190 (vst1q_u16_x2): Likewise.
21191 (vst1q_u32_x2): Likewise.
21192 (vst1q_u64_x2): Likewise.
21193 (vst1q_f16_x2): Likewise.
21194 (vst1q_f32_x2): Likewise.
21195 (vst1q_f64_x2): Likewise.
21196 (vst1q_p64_x2): Likewise.
21197 (vst1_s64_x3): Likewise.
21198 (vst1_u64_x3): Likewise.
21199 (vst1_f64_x3): Likewise.
21200 (vst1_s8_x3): Likewise.
21201 (vst1_p8_x3): Likewise.
21202 (vst1_s16_x3): Likewise.
21203 (vst1_p16_x3): Likewise.
21204 (vst1_s32_x3): Likewise.
21205 (vst1_u8_x3): Likewise.
21206 (vst1_u16_x3): Likewise.
21207 (vst1_u32_x3): Likewise.
21208 (vst1_f16_x3): Likewise.
21209 (vst1_f32_x3): Likewise.
21210 (vst1_p64_x3): Likewise.
21211 (vst1q_s8_x3): Likewise.
21212 (vst1q_p8_x3): Likewise.
21213 (vst1q_s16_x3): Likewise.
21214 (vst1q_p16_x3): Likewise.
21215 (vst1q_s32_x3): Likewise.
21216 (vst1q_s64_x3): Likewise.
21217 (vst1q_u8_x3): Likewise.
21218 (vst1q_u16_x3): Likewise.
21219 (vst1q_u32_x3): Likewise.
21220 (vst1q_u64_x3): Likewise.
21221 (vst1q_f16_x3): Likewise.
21222 (vst1q_f32_x3): Likewise.
21223 (vst1q_f64_x3): Likewise.
21224 (vst1q_p64_x3): Likewise.
21226 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
21228 * config/msp430/msp430.c (msp430_output_labelref): Prepend
21229 user_label_prefix to name.
21231 * tree-core.h: Update comment about the format of NAME string
21232 passed to handler in attribute_spec.
21234 * config/msp430/msp430.md: Remove erroneous subreg expression from
21235 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
21236 zero_extend{q,h}isi2.
21238 2018-05-30 Borislav Petkov <bp@suse.de>
21240 * doc/extend.texi: Document some architecture specific
21241 constraints and sort entries.
21243 2018-05-30 Martin Sebor <msebor@redhat.com>
21245 PR middle-end/85369
21246 * builtins.c (expand_builtin_stpcpy_1): New function.
21247 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
21248 only if the former succeeds.
21250 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
21252 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
21255 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
21257 * doc/invoke.texi (-flinker-output): Document
21259 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
21261 * passes.c (ipa_write_summaries): Only modify statements if body
21263 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
21264 incrementally linking.
21265 (ipa_passes): Likewise.
21266 * lto-cgraph.c (lto_output_node): When incrementally linking do not
21267 pass down resolution info.
21268 * common.opt (flag_incremental_link): Update info.
21269 * gcc.c (plugin specs): Turn flinker-output=* to
21270 -plugin-opt=-linker-output-known
21271 * toplev.c (compile_file): Also cut compilation when doing incremental
21273 * flag-types. (enum lto_partition_model): Add
21274 LTO_LINKER_OUTPUT_NOLTOREL.
21275 (invoke.texi): Add -flinker-output docs.
21276 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
21277 link same way as WPA; do not stream in dead initializers.
21279 * dwarf2out.c (dwarf2out_die_ref_for_decl,
21280 darf2out_register_external_decl): Support incremental link.
21282 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
21284 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
21286 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
21288 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
21289 it down to simple_object_copy_lto_debug_sections.
21290 (run_gcc): Determine incremental LTO link time and configure
21291 lto1 into non-wpa mode, disable renaming of debug sections.
21293 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
21295 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
21296 descriptions of various incorrectly documented functions.
21298 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
21301 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
21302 address check not strict.
21304 2018-05-30 Richard Biener <rguenther@suse.de>
21306 PR tree-optimization/85964
21307 * tracer.c (better_p): Drop initialized count check, we only
21308 call the function with initialized counts now.
21309 (find_best_successor): Do find a best edge if one
21310 has uninitialized count.
21311 (find_best_predecessor): Likewise. Do BB frequency check only
21312 if count is initialized.
21314 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
21316 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
21317 (aarch64_ldrstr_offset_compare): New.
21318 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
21319 load/store orderings.
21320 (aarch64_gen_adjusted_ldpstp): Likewise.
21322 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
21324 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
21325 Check for subset of GENERAL_REGS and FP_REGS.
21326 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
21329 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
21331 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
21332 and wi::to_poly_offset. Add the current offset and then check
21333 whether the sum fits, rather than using an unchecked addition of
21334 a checked term. Check for a shwi rather than a uhwi.
21335 * expr.c (get_bit_range): Use tree_to_poly_uint64.
21336 (store_constructor): Use poly_int_tree_p.
21337 (expand_expr_real_1): Likewise.
21338 * function.c (assign_temp): Likewise.
21339 * fold-const.c (const_binop): Use poly_int_tree_p and
21340 wi::to_poly_offset.
21341 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
21343 * ipa-icf-gimple.c (func_checker::compare_operand): Use
21344 to_poly_offset for MEM offsets.
21345 * ipa-icf.c (sem_variable::equals): Likewise.
21346 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
21347 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
21348 wi::to_poly_offset for BIT_FIELD_REF offsets.
21349 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
21350 wi::to_poly_offset.
21351 * var-tracking.c (emit_note_insn_var_location): Use
21352 tree_to_poly_uint64.
21354 2018-05-29 Jim Wilson <jimw@sifive.com>
21356 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
21358 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
21361 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
21362 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
21364 (sse4_1_round<mode>2): Use nonimmediate_operand
21365 for operand 1 predicate.
21367 2018-05-29 Martin Sebor <msebor@redhat.com>
21368 Richard Biener <rguenther@suse.de>
21371 * calls.c (get_size_range): Call determine_value_range instead
21372 of get_value_range..
21373 * tree-vrp.h (determine_value_range): Declared new function.
21374 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
21376 2018-05-29 Richard Biener <rguenther@suse.de>
21378 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
21379 sure to use non-pattern stmts for get_earlier_stmt arguments.
21380 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
21381 called on pattern stmts.
21382 (get_later_stmt): Likewise.
21384 2018-05-29 Martin Liska <mliska@suse.cz>
21386 PR gcov-profile/85759
21387 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
21390 2018-05-29 Jakub Jelinek <jakub@redhat.com>
21392 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
21394 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
21398 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
21399 VEC_PACK_FLOAT_EXPR): New tree codes.
21400 * tree-pretty-print.c (op_code_prio): Handle
21401 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
21402 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
21403 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
21404 * tree-inline.c (estimate_operator_cost): Likewise.
21405 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
21406 * fold-const.c (const_binop): Likewise.
21407 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
21408 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
21409 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
21410 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
21411 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
21412 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
21413 * expr.c (expand_expr_real_2): Likewise.
21414 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
21415 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
21416 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
21418 * optabs.c (expand_widen_pattern_expr): For
21419 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
21420 sign from result type rather than operand's type.
21421 (expand_binop_directly): For vec_packu_float_optab and
21422 vec_packs_float_optab allow result type to be different from operand's
21424 * optabs-tree.c (optab_for_tree_code): Handle
21425 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
21426 VEC_PACK_FLOAT_EXPR. Formatting fixes.
21427 * tree-vect-generic.c (expand_vector_operations_1): Handle
21428 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
21429 VEC_PACK_FLOAT_EXPR.
21430 * tree-vect-stmts.c (supportable_widening_operation): Handle
21432 (supportable_narrowing_operation): Handle FLOAT_EXPR.
21433 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
21434 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
21435 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
21436 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
21438 (vec_pack<floatprefix>_float_<mode>): New expander.
21439 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
21441 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
21442 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
21443 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
21444 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
21445 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
21447 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
21448 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
21449 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
21450 VEC_PACK_FLOAT_EXPR): Document.
21452 2018-05-29 Richard Biener <rguenther@suse.de>
21454 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
21456 (stmt_vec_info_vec): Make pointer.
21457 (init_stmt_vec_info_vec): Remove.
21458 (free_stmt_vec_info_vec): Likewise.
21459 (set_stmt_vec_info_vec): New function.
21460 (free_stmt_vec_infos): Likewise.
21461 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
21462 (set_vinfo_for_stmt): Likewise.
21463 (get_earlier_stmt): Likewise.
21464 (get_later_stmt): Likewise.
21465 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
21466 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
21467 (vec_info::~vec_info): Free stmt_vec_infos.
21468 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
21469 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
21470 (pass_slp_vectorize::execute): Likewise.
21471 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
21472 (free_stmt_vec_info_vec): Likewise.
21473 (set_stmt_vec_info_vec): New function.
21474 (free_stmt_vec_infos): Likewise.
21475 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
21476 the global stmt_vec_info_vec.
21477 * tree-parloops.c (gather_scalar_reductions): Use
21478 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
21481 2018-05-29 Richard Biener <rguenther@suse.de>
21483 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
21485 2018-05-29 Martin Liska <mliska@suse.cz>
21486 David Malcolm <dmalcolm@redhat.com>
21488 * vec.c (test_reverse): New.
21489 (vec_c_tests): Add new test.
21490 * vec.h (vl_ptr>::reverse): New function.
21492 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
21494 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
21496 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
21499 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
21501 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
21503 2018-05-28 Richard Biener <rguenther@suse.de>
21505 PR tree-optimization/85933
21506 * tree-vect-data-refs.c (vect_record_base_alignments): Only
21507 look at stmts marked as vectorizable.
21509 2018-05-28 Richard Biener <rguenther@suse.de>
21511 PR tree-optimization/85934
21512 * tree-vect-generic.c (expand_vector_operations_1): Hoist
21513 vector boolean check before scalar optimization.
21515 2018-05-28 Jakub Jelinek <jakub@redhat.com>
21517 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
21520 2018-05-28 Mark Wielaard <mark@klomp.org>
21522 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
21523 if it is an expression containing a minus sign.
21525 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
21527 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
21529 2018-05-27 Paul Koning <ni1d@arrl.net>
21531 * config/pdp11/pdp11.md (truncsihi2): Remove.
21533 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
21534 Chung-Ju Wu <jasonwucj@gmail.com>
21536 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
21538 (unaligned_store_dw): Ditto.
21539 * config/nds32/nds32-memory-manipulation.c
21540 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
21541 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
21542 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
21543 (emit_setmem_word_loop): Rename to ...
21544 (emit_setmem_doubleword_loop): ... this.
21545 (nds32_gen_dup_4_byte_to_word_value): New function.
21546 (nds32_gen_dup_8_byte_to_double_word_value): New function.
21547 (nds32_expand_setmem_loop): Refine implementation.
21548 (nds32_expand_setmem_loop_v3m): Ditto.
21549 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
21552 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
21554 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
21556 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
21558 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
21559 (nds32_init_machine_status): Initialize machine->attr_naked_p and
21560 machine->attr_no_prologue_p.
21561 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
21562 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
21563 (nds32_expand_epilogue): Consider attr_naked_p.
21564 (nds32_expand_epilogue_v3pop): Likewise.
21565 (nds32_can_use_return_insn): Likewise.
21566 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
21567 attr_no_prologue_p fields.
21568 * config/nds32/nds32.opt (mret-in-naked-func): New option.
21570 2018-05-27 Jakub Jelinek <jakub@redhat.com>
21573 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
21575 * config/i386/sse.md
21576 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
21578 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
21580 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
21582 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
21584 (*<floatsuffix>floatv2div2sf2): Rename to ...
21585 (*float<floatunssuffix>v2div2sf2): ... this.
21586 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
21587 (float<floatunssuffix>v2div2sf2_mask): ... this.
21588 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
21589 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
21590 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
21592 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
21594 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
21596 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
21598 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
21600 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
21602 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
21603 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
21604 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
21605 gen_ufix_truncv8dfv8si2.
21606 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
21607 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
21608 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
21609 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
21610 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
21611 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
21612 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
21613 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
21615 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
21619 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
21621 2018-05-25 Jim Wilson <jimw@sifive.com>
21623 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
21624 * config/riscv/riscv.c (struct machine_function): Add
21625 interrupt_handler_p and attribute_checked_p fields.
21626 (riscv_attribute_table): Add interrupt.
21627 (riscv_interrupt_type_p): New.
21628 (riscv_save_reg_p): Save extra regs for interrupt handler.
21629 (riscv_use_save_libcall): Return false for interrupt handler.
21630 (riscv_first_stack_step): Add forward declaration.
21631 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
21632 for interrupt handler with large frame. Use it for saved reg list.
21633 (riscv_expand_prologue): Move flag_stack_usage_info support to
21634 eliminate duplication.
21635 (riscv_expand_epilogue): Generate mret for interrupt handler.
21636 (riscv_epilogue_uses): New.
21637 (riscv_can_use_return_insn): Return false for interrupt handler.
21638 (riscv_function_ok_for_sibcall): Likewise.
21639 (riscv_set_current_function): Add interrupt handler support.
21640 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
21641 * config/riscv/riscv.md (UNSPECV_MRET): New.
21643 (riscv_frflags, riscv_fsflags): Use tab after opcode.
21645 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
21647 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
21649 PR tree-optimization/85712
21650 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
21651 this candidate has already been replaced in-situ by a copy.
21653 2018-05-25 Jason Merrill <jason@redhat.com>
21655 PR c++/80485 - inline function non-zero address.
21656 * symtab.c (nonzero_address): Check DECL_COMDAT.
21658 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
21661 * config/alpha/alpha.md (ashlsi3): New insn pattern.
21662 (*ashlsi_se): Rename from *ashldi_se. Define as sign
21663 extension of SImode operation. Use const123_operand predicate.
21664 (*saddsi_1): Remove.
21665 (*saddl_se_1): Ditto.
21666 (*ssubsi_1): Ditto.
21667 (*ssubl_se_1): Ditto.
21668 * config/alpha/predicates.md (const123_operand): New predicate.
21669 * config/alpha/constraints.md (P): Use IN_RANGE.
21671 2018-05-25 Richard Biener <rguenther@suse.de>
21673 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
21675 (ref_maybe_used_by_stmt_p): Likewise.
21676 (stmt_may_clobber_ref_p): Likewise.
21677 (stmt_may_clobber_ref_p_1): Likewise.
21678 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
21680 (ref_maybe_used_by_stmt_p): Likewise.
21681 (stmt_may_clobber_ref_p): Likewise.
21682 (stmt_may_clobber_ref_p_1): Likewise.
21683 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
21684 the alias oracle to disambiguate DRs with stmts DR analysis
21686 (vect_analyze_data_refs): Do not give up on not analyzable
21687 DRs for BB vectorization. Remove code truncating the dataref
21690 2018-05-25 Jakub Jelinek <jakub@redhat.com>
21693 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
21694 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
21695 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
21697 2018-05-25 Richard Biener <rguenther@suse.de>
21699 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
21700 function, combining stmt data ref gathering and fatal analysis
21702 (vect_analyze_data_refs): Remove now redudnant code and simplify.
21703 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
21704 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
21705 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
21706 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
21708 2018-05-25 Bin Cheng <bin.cheng@arm.com>
21710 PR tree-optimization/85720
21711 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
21712 SCC if all partitions are builtins.
21713 (version_loop_by_alias_check): New parameter. Generate cancelable
21714 runtime alias check if all partitions are builtins.
21715 (distribute_loop): Update call to above function.
21717 2018-05-25 Bin Cheng <bin.cheng@arm.com>
21719 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
21720 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
21721 (parm_default_def_partition_arg): Ditto.
21722 (set_parm_default_def_partition): Ditto.
21723 (get_parm_default_def_partitions): Ditto and make it static.
21724 (get_undefined_value_partitions): Ditto and make it static.
21725 (remove_ssa_form): Refactor call to init_var_map here.
21726 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
21727 computation for loop region.
21728 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
21729 (register_default_def): Delete.
21730 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
21731 (parm_default_def_partition_arg): Ditto.
21732 (set_parm_default_def_partition): Ditto.
21733 (get_parm_default_def_partitions): Ditto and make it static.
21734 (get_undefined_value_partitions): Ditto and make it static.
21735 (coalesce_with_default, coalesce_with_default): Update comment.
21736 (create_coalesce_list_for_region): New func factored out from
21737 create_outofssa_var_map.
21738 (populate_coalesce_list_for_outofssa): New func factored out from
21739 create_outofssa_var_map and coalesce_ssa_name.
21740 (create_outofssa_var_map): Delete.
21741 (coalesce_ssa_name): Refactor to support live range computation.
21742 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
21743 (get_parm_default_def_partitions): Delete.
21744 (get_undefined_value_partitions): Ditto.
21745 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
21746 computation for loop region.
21747 (new_tree_live_info, loe_visit_block): Ditto.
21748 (live_worklist, set_var_live_on_entry): Ditto.
21749 (calculate_live_on_exit, verify_live_on_entry): Ditto.
21750 * tree-ssa-live.h (struct _var_map): New fields.
21751 (init_var_map): Change decl.
21752 (region_contains_p): New.
21754 2018-05-25 Bin Cheng <bin.cheng@arm.com>
21756 * tree-ssa-live.h (live_merge_and_clear): Delete.
21758 2018-05-25 Richard Biener <rguenther@suse.de>
21761 * tree-dump.c (dequeue_and_dump): Remove access to removed
21762 operand 2 of a SWITCH_EXPR.
21764 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
21766 * doc/sourcebuild.texi (vect_double_cond_arith): Include
21767 multiplication and division.
21768 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
21769 (cond_udiv@var{m}, cond_umod@var{m}): Document.
21770 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
21771 (cond_udiv_optab, cond_umod_optab): New optabs.
21772 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
21773 (IFN_COND_RDIV): New internal functions.
21774 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
21775 TRUNC_MOD_EXPR and RDIV_EXPR.
21776 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
21777 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
21779 (SVE_INT_BINARY): Include mult.
21780 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
21781 (optab, sve_int_op): Handle mult.
21782 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
21784 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
21785 for SVE_INT_BINARY_SD.
21787 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
21789 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
21790 (optab, sve_int_op): Handle div and udiv.
21791 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
21792 for SVE_INT_BINARY_SD.
21793 (*<optab><mode>3): New insn for the same.
21795 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
21797 * tree-vect-patterns.c: Include predict.h.
21798 (vect_recog_divmod_pattern): Restrict check for division support
21799 to when optimizing for size.
21801 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
21803 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
21804 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
21805 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
21806 (gimple_match_op::set_op): Likewise.
21807 (gimple_resimplify4): Declare.
21808 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
21809 (expr::gen_transform): Likewise.
21810 (decision_tree::gen): Generate a simplification routine for 4 operands.
21811 * gimple-match-head.c (gimple_simplify): Add an overload for
21812 4 operands. In the top-level function, handle up to 4 call
21813 arguments and call gimple_resimplify4.
21814 (gimple_resimplify4): New function.
21815 (build_call_internal): Pass a fourth operand.
21816 (maybe_push_to_seq): Likewise.
21817 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
21818 Fold VEC_COND_EXPRs of an operation and a default value into
21819 an IFN_COND_* function if possible.
21820 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
21822 (SVE_COND_FP_BINARY): Include them.
21823 (optab, sve_fp_op): Handle them.
21824 (SVE_INT_BINARY_REV): New code iterator.
21825 (SVE_COND_FP_BINARY_REV): New int iterator.
21826 (commutative): New int attribute.
21827 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
21829 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
21831 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
21832 (*cond_<optab><mode>): New patterns for reversed operands.
21834 2018-05-25 Richard Biener <rguenther@suse.de>
21836 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
21837 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
21838 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
21839 (STMT_VINFO_GROUPED_ACCESS): Adjust.
21840 * tree-vect-data-refs.c (everywhere): Adjust users.
21841 * tree-vect-loop.c (everywhere): Likewise.
21842 * tree-vect-slp.c (everywhere): Likewise.
21843 * tree-vect-stmts.c (everywhere): Likewise.
21844 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
21846 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21848 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
21850 (gcc_cv_as_section_exclude): ... this.
21851 Try Solaris as #exclude syntax.
21852 * configure: Regenerate.
21853 * config.in: Regenerate.
21854 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
21856 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
21857 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
21859 * varasm.c (default_elf_asm_named_section): Don't check if
21860 HAVE_GAS_SECTION_EXCLUDE is defined.
21862 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
21864 * doc/md.texi: Update the documentation of the cond_* optabs
21865 to mention the new final operand. Fix GET_MODE_NUNITS call.
21866 Describe the scalar case too.
21867 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
21868 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
21870 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
21871 (get_conditional_internal_fn): Update comment.
21872 * tree-vect-loop.c (vectorizable_reduction): Pass the original
21873 accumulator value as a final argument to conditional functions.
21874 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
21875 a define_expand and add an "else" operand. Assert for now that
21876 the else operand is equal to operand 2. Use SVE_INT_BINARY and
21877 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
21878 (*cond_<optab><mode>): New patterns.
21879 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
21880 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
21881 (UNSPEC_COND_EOR): Delete.
21882 (optab): Remove associated mappings.
21883 (SVE_INT_BINARY): New code iterator.
21884 (sve_int_op): Remove int attribute and add "minus" to the code
21886 (SVE_COND_INT_OP): Delete.
21887 (SVE_COND_FP_OP): Rename to...
21888 (SVE_COND_FP_BINARY): ...this.
21890 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
21892 * optabs.c (can_reuse_operands_p): New function.
21893 (maybe_legitimize_operands): Try to reuse the results for
21896 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
21898 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
21899 Add {q} suffix to insn mnemonic.
21901 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
21903 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
21904 (msp430_warn_func_return): New.
21906 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
21908 * fold-const.c (tree_nonzero_bits): New function.
21909 * fold-const.h (tree_nonzero_bits): Likewise.
21910 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
21911 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
21913 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
21917 * varasm.c (assemble_alias): Check ifunc_resolver only on
21920 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
21923 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
21924 when memory input operand is handled.
21926 2018-05-24 Luis Machado <luis.machado@linaro.org>
21928 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
21930 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
21932 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
21934 * match.pd: Delay FMA folds until after vectorization.
21936 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
21939 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
21940 address check not strict.
21942 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
21944 * gimple-match.h (gimple_match_op): New class.
21945 (mprts_hook): Replace parameters with a gimple_match_op *.
21946 (maybe_build_generic_op): Likewise.
21947 (gimple_simplified_result_is_gimple_val): Replace parameters with
21948 a const gimple_match_op *.
21949 (gimple_simplify): Replace code_helper * and tree * parameters with
21950 a gimple_match_op * parameter.
21951 (gimple_resimplify1): Replace code_helper *, tree and tree *
21952 parameters with a gimple_match_op * parameter.
21953 (gimple_resimplify2): Likewise.
21954 (gimple_resimplify3): Likewise.
21955 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
21956 parameters with a gimple_match_op * parameter.
21957 * gimple-match-head.c (gimple_simplify): Change prototypes of
21958 auto-generated functions to take a gimple_match_op * instead of
21959 separate code_helper * and tree * parameters. Make the same
21960 change in the top-level overload and update calls to the
21961 gimple_resimplify routines. Update calls to the auto-generated
21962 functions and to maybe_push_res_to_seq in the publicly-facing
21963 operation-specific gimple_simplify overloads.
21964 (gimple_match_op::MAX_NUM_OPS): Define.
21965 (gimple_resimplify1): Replace rcode and ops with a single res_op
21966 parameter. Update call to gimple_simplify.
21967 (gimple_resimplify2): Likewise.
21968 (gimple_resimplify3): Likewise.
21969 (mprts_hook): Replace parameters with a gimple_match_op *.
21970 (maybe_build_generic_op): Likewise.
21971 (build_call_internal): Replace type, nargs and ops with
21972 a gimple_match_op *.
21973 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
21974 with a single gimple_match_op *. Update calls to mprts_hook,
21975 build_call_internal and gimple_simplified_result_is_gimple_val.
21976 Factor out code that is common to the tree_code and combined_fn cases.
21977 * genmatch.c (expr::gen_transform): Replace tem_code and
21978 tem_ops with a gimple_match_op called tem_op. Update calls
21979 to the gimple_resimplify functions and maybe_push_res_to_seq.
21980 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
21981 res_ops. Update call to the gimple_resimplify functions.
21982 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
21983 (decision_tree::gen): Make the functions take a gimple_match_op *
21984 called res_op instead of separate res_code and res_ops parameters.
21985 Update call accordingly.
21986 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
21987 and ops with a single res_op parameter. Update calls to
21988 maybe_build_generic_op and maybe_push_res_to_seq.
21989 (fold_stmt_1): Update calls to gimple_simplify and
21990 replace_stmt_with_simplification.
21991 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
21992 and gimple_simplified_result_is_gimple_val.
21993 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
21995 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
21996 with a gimple_match_op *.
21997 (vn_nary_build_or_lookup): Likewise. Update call to
21998 vn_nary_build_or_lookup_1.
21999 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
22000 gimple_match_op *. Update calls to the gimple_resimplify routines
22001 and to gimple_simplified_result_is_gimple_val.
22002 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
22003 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
22004 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
22005 (visit_nary_op): Likewise.
22006 (visit_reference_op_load): Likewise.
22008 2018-05-23 Luis Machado <luis.machado@linaro.org>
22010 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
22011 modifier for printing the step amount.
22013 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
22016 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
22019 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
22021 * doc/sourcebuild.texi (Endianness): New subsubsection.
22023 2018-05-23 Luis Machado <luis.machado@linaro.org>
22025 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
22026 <prefetch_dynamic_strides>: New const bool field.
22027 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
22028 prefetch_dynamic_strides.
22029 (exynosm1_prefetch_tune): Likewise.
22030 (thunderxt88_prefetch_tune): Likewise.
22031 (thunderx_prefetch_tune): Likewise.
22032 (thunderx2t99_prefetch_tune): Likewise.
22033 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
22035 (aarch64_override_options_internal): Update to set
22036 PARAM_PREFETCH_DYNAMIC_STRIDES.
22037 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
22038 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
22039 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
22040 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
22041 prefetch-dynamic-strides setting.
22043 2018-05-23 Luis Machado <luis.machado@linaro.org>
22045 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
22046 <minimum_stride>: New const int field.
22047 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
22048 minimum_stride field defaulting to -1.
22049 (exynosm1_prefetch_tune): Likewise.
22050 (thunderxt88_prefetch_tune): Likewise.
22051 (thunderx_prefetch_tune): Likewise.
22052 (thunderx2t99_prefetch_tune): Likewise.
22053 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
22054 <default_opt_level>: Set to 3.
22055 (aarch64_override_options_internal): Update to set
22056 PARAM_PREFETCH_MINIMUM_STRIDE.
22057 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
22058 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
22059 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
22060 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
22061 stride is constant and is below the minimum stride threshold.
22063 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22065 * config/arm/arm-cpus.in (mode26): Delete.
22066 (armv4): Delete mode26 reference.
22067 * config/arm/arm.c (arm_configure_build_target): Delete use of
22070 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
22072 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
22074 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
22075 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
22077 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
22078 (floatunsdidf2): Ditto.
22080 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
22082 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
22083 (fixuns_trunc<mode>si2_avx512f): Ditto.
22084 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
22085 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
22086 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
22088 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
22090 PR rtl-optimization/79985
22091 * df-scan.c (df_insn_refs_collect): Remove special case for
22092 global registers and asm statements.
22094 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
22096 * extend.texi (Global Register Variables): Rewrite the bullet list.
22097 Note that the register is available for allocation. Note that access
22098 via inline asm must use constraints. Add note about async-signal
22099 handlers. Remove paragraph about automagic register selection.
22101 2018-05-23 Richard Biener <rguenther@suse.de>
22103 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
22104 of fixed offset from memset VN.
22106 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
22108 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
22109 first_interp field.
22110 (alloc_cand_and_find_basis): Initialize first_interp field.
22111 (slsr_process_mul): Modify first_interp field.
22112 (slsr_process_add): Likewise.
22113 (slsr_process_cast): Modify first_interp field for each new
22115 (slsr_process_copy): Likewise.
22116 (dump_candidate): Dump first_interp field.
22117 (replace_mult_candidate): Process all interpretations, not just
22119 (replace_rhs_if_not_dup): Likewise.
22120 (replace_one_candidate): Likewise.
22122 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
22124 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
22126 (aarch64_needs_frame_chain): New function.
22127 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
22129 2018-05-23 Sudakshina Das <sudi.das@arm.com>
22132 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
22133 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
22134 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
22135 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
22136 as true for strict-align.
22137 (aarch64_can_inline_p): Perform checks even when callee has no
22138 attributes to check for strict alignment.
22139 * doc/extend.texi (AArch64 Function Attributes): Document
22141 * doc/invoke.texi: (AArch64 Options): Likewise.
22143 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
22145 PR tree-optimization/85853
22146 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
22147 the handling of the root of the node to...
22148 (vect_slp_analyze_node_operations_1): ...this new function,
22149 and run the whole thing with the child nodes' def types
22150 set according to their SLP node's def type.
22152 2018-05-23 Richard Biener <rguenther@suse.de>
22154 PR middle-end/85874
22155 * tree-data-ref.c (create_runtime_alias_checks): Defer
22156 and ignore overflow warnings.
22158 2018-05-23 Yury Gribov <tetra2005@gmail.com>
22160 PR tree-optimization/85822
22161 * tree-vrp.c (is_masked_range_test): Fix handling of negative
22164 2018-05-23 Richard Biener <rguenther@suse.de>
22166 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
22167 memset constants via native_interpret_expr.
22169 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
22172 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
22174 (cgraph_node::create_alias): Likewise.
22175 (cgraph_node::get_availability): Check ifunc_resolver instead
22176 of looking up ifunc attribute.
22177 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
22178 * varasm.c (do_assemble_alias): Likewise.
22179 (assemble_alias): Likewise.
22180 (default_binds_local_p_3): Likewise.
22181 * cgraph.h (cgraph_node): Add ifunc_resolver.
22182 (cgraph_node::only_called_directly_or_aliased_p): Return false
22183 for IFUNC resolver.
22184 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
22186 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
22187 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
22188 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
22189 instead of looking up ifunc attribute.
22191 2018-05-22 Luis Machado <luis.machado@linaro.org>
22193 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
22195 2018-05-22 Martin Sebor <msebor@redhat.com>
22197 PR middle-end/85359
22198 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
22199 only when expasion succeeds.
22200 (expand_builtin_strcmp): Same.
22201 (expand_builtin_strncmp): Same.
22203 2018-05-22 Martin Sebor <msebor@redhat.com>
22205 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
22207 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
22208 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22210 * config/aarch64/aarch64-ldpstp.md: Replace uses of
22211 aarch64_mem_pair_operand with memory_operand and delete operand swapping
22213 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
22214 Add check for legitimate_address.
22215 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
22216 (aarch64_swap_ldrstr_operands): New.
22217 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
22220 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
22221 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22223 * config/aarch64/aarch64.md: New patterns to generate stp
22225 (store_pair_sw, store_pair_dw): New patterns to generate stp for
22226 single words and double words.
22227 (load_pair_sw, load_pair_dw): Likewise.
22228 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
22230 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
22232 * config/aarch64/aarch64-ldpstp.md: Modify peephole
22233 for different mode ldpstp and add peephole for merged zero stores.
22234 Likewise for loads.
22235 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
22237 (aarch64_gen_store_pair): Rename calls to match new patterns.
22238 (aarch64_gen_load_pair): Rename calls to match new patterns.
22239 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
22240 (load_pair<DREG:mode><DREG2:mode>): ... This.
22241 (store_pair<mode>): Rename to...
22242 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
22243 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
22244 New mode iterators.
22245 (V_INT_EQUIV): Handle SImode.
22246 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
22249 2018-05-22 Martin Sebor <msebor@redhat.com>
22252 * calls.c (maybe_warn_nonstring_arg): Use string length to set
22253 or ajust the presumed bound on an operation to avoid unnecessary
22256 2018-05-22 Martin Sebor <msebor@redhat.com>
22258 PR tree-optimization/85826
22259 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
22260 assuming that a DECL necesarily has a constant size.
22262 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
22264 PR middle-end/85862
22265 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
22267 2018-05-22 Richard Biener <rguenther@suse.de>
22269 PR tree-optimization/85834
22270 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
22271 non-constant and non-zero memset arguments.
22273 2018-05-22 Martin Liska <mliska@suse.cz>
22276 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
22278 2018-05-22 Richard Biener <rguenther@suse.de>
22280 PR tree-optimization/85863
22281 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
22282 comparisons when vectype is specified.
22283 (vectorizable_condition): Do not specify vectype for
22284 vect_is_simple_cond when SLP vectorizing.
22286 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
22289 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
22290 define __ibm128 as long double.
22291 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
22292 as a distinct type when IEEE 128-bit support is enabled.
22293 (init_float128_ieee): Fix up conversions between IFmode and IEEE
22294 128-bit types to use the correct functions.
22295 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
22296 convert between 128-bit floating point types that have different
22297 modes but the same representation, instead of using gen_lowpart to
22299 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
22301 (IFKF_reg): New attributes to give the register constraints for
22303 (extend<mode>tf2_internal): New insns to mark an explicit
22304 conversion between 128-bit floating point types that have a
22305 different mode but share the same representation.
22307 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
22309 PR tree-optimization/85814
22310 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
22311 a null return from get_strinfo when unsharing the next
22312 strinfo in the chain.
22314 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
22317 * varasm.c (weak_finish): Clean up weak_decls.
22319 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22321 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
22322 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
22323 UNSPEC_UADALP values.
22324 * config/aarch64/iterators.md (ABAL): New int iterator.
22327 (sur): Add mappings for the above.
22328 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
22330 (aarch64_<sur>abal<mode>_4): Likewise.
22331 (aarch64_<sur>adalp<mode>_3): Likewise.
22332 (<sur>sadv16qi): New define_expand.
22334 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
22336 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
22337 (*movdf_internal): Ditto.
22338 (*rcpsf2_sse): Ditto.
22339 (*rsqrtsf2_sse): Ditto.
22340 (*sqrt<mode>2_sse): Ditto.
22342 2018-05-21 Tamar Christina <tamar.christina@arm.com>
22344 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
22346 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
22347 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
22348 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
22349 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
22351 * config/aarch64/arm_neon.h: Likewise.
22352 * config/aarch64/iterators.md (VQ_I): New.
22354 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
22356 * config.gcc: Add arc/t-multilib-linux to tmake_file for
22358 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
22361 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
22363 * config/nds32/constraints.md (S): New constraint.
22364 * config/nds32/nds32.md (call_internal): Use constraint S.
22365 (call_value_internal): Likewise.
22366 (sibcall_internal): Likewise.
22367 (sibcall_value_internal): Likewise.
22369 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
22370 Chung-Ju Wu <jasonwucj@gmail.com>
22372 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
22373 into consideration.
22375 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
22376 Chung-Ju Wu <jasonwucj@gmail.com>
22378 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
22379 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
22380 (nds32_rtx_costs_impl): Simplify.
22381 (nds32_address_cost_impl): Simplify.
22382 (nds32_init_rtx_costs): New function.
22383 (nds32_rtx_costs_speed_prefer): Likewise.
22384 (nds32_rtx_costs_size_prefer): Likewise.
22385 (nds32_address_cost_speed_prefer): Likewise.
22386 (nds32_address_cost_speed_fwprop): Likewise.
22387 (nds32_address_cost_size_prefer): Likewise.
22388 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
22389 * config/nds32/nds32.c (nds32_option_override): Use
22390 nds32_init_rtx_costs function.
22392 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
22394 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
22395 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
22396 (TARGET_PIPELINE_N8): Likewise.
22397 (TARGET_PIPELINE_N10): Likewise.
22398 (TARGET_PIPELINE_N13): Likewise.
22399 (TARGET_PIPELINE_GRAYWOLF): Likewise.
22401 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
22403 * config/nds32/nds32-fpu.md: Update copyright year.
22405 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
22407 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
22409 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
22411 * config/nds32/nds32.c
22412 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
22413 * config/nds32/nds32.opt (minline-asm-r15): New option.
22415 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
22417 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
22419 * config/nds32/nds32.md (abssi2): New pattern.
22421 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
22423 * config/i386/i386.md (rex64namesuffix): New mode attribute.
22424 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
22425 Merge insn pattern from sse_cvtsi2ss<round_name> and
22426 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
22427 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
22428 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
22429 using SWI48 mode iterator.
22430 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
22431 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
22432 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
22433 pattern from sse_cvttss2si<round_saeonly_name>
22434 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
22435 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
22436 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
22437 using SWI48 mode iterator.
22438 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
22439 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
22440 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
22441 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
22442 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
22443 using SWI48 mode iterator.
22444 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
22445 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
22446 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
22447 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
22448 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
22449 SWI48 mode iterator.
22450 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
22451 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
22452 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
22453 pattern from sse_cvttsd2si<round_saeonly_name>
22454 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
22456 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
22458 * config/nds32/nds32-md-auxiliary.c
22459 (nds32_valid_smw_lwm_base_p): Refine.
22460 (nds32_output_smw_single_word): Refine.
22461 (nds32_output_smw_double_word): New.
22462 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
22464 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
22466 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
22467 (nds32_output_stack_pop): Refine.
22468 (nds32_expand_unaligned_load): Refine.
22469 (nds32_expand_unaligned_store): Refine.
22471 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
22472 Chung-Ju Wu <jasonwucj@gmail.com>
22474 * config/nds32/constants.md: Add TP_REGNUM constant.
22475 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
22476 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
22478 * config/nds32/nds32-doubleword.md: Consider flag_pic.
22479 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
22480 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
22481 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
22482 and PIC code generation.
22483 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
22485 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
22487 * config/nds32/nds32.md: Support TLS and PIC.
22488 * config/nds32/nds32.c: Support TLS and PIC.
22489 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
22490 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
22493 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
22495 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
22496 mode with E_ prefix.
22498 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
22499 Chung-Ju Wu <jasonwucj@gmail.com>
22501 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
22502 * config/nds32/nds32-md-auxiliary.c
22503 (symbolic_reference_mentioned_p): New.
22504 (nds32_legitimize_ict_address): New.
22505 (nds32_expand_ict_move): New.
22506 (nds32_indirect_call_referenced_p): New.
22507 (nds32_symbol_binds_local_p): Delete.
22508 (nds32_long_call_p): Modify.
22509 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
22510 * config/nds32/nds32-protos.h
22511 (symbolic_reference_mentioned_p): Declare.
22512 (nds32_legitimize_ict_address): Declare.
22513 (nds32_expand_ict_move): Declare.
22514 (nds32_indirect_call_referenced_p): Declare.
22515 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
22516 (nds32_relax_group): Use nds32_ict_const_p as condition.
22517 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
22518 (nds32_asm_file_start): Output ict_model directive in asm code.
22519 (nds32_legitimate_address_p): Consider indirect call.
22520 (nds32_print_operand): Consider indirect call.
22521 (nds32_print_operand_address): Consider indirect call.
22522 (nds32_insert_attributes): Handle "indirect_call" attribute.
22523 (TARGET_LEGITIMATE_ADDRESS_P): Define.
22524 (TARGET_LEGITIMATE_CONSTANT_P): Define.
22525 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
22526 (TARGET_DELEGITIMIZE_ADDRESS): Define.
22527 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
22528 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
22529 (TARGET_ICT_MODEL_SMALL): Define.
22530 (TARGET_ICT_MODEL_LARGE): Define.
22531 * config/nds32/nds32.md (movsi): Consider ict model.
22532 (call, call_value): Consider ict model.
22533 (sibcall, sibcall_value): Consider ict model.
22534 * config/nds32/nds32.opt (mict-model): New option.
22535 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
22538 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
22539 Monk Chiang <sh.chiang04@gmail.com>
22540 Jim Wilson <jimw@sifive.com>
22542 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
22543 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
22544 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
22545 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
22546 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
22547 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
22548 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
22549 compute save_libcall_adjustment properly.
22550 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
22551 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
22552 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
22553 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
22554 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
22555 (ABI_SPEC): Handle mabi=ilp32e.
22556 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
22557 (RVE): Add RVE mask.
22558 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
22559 <-march>: Add rv32e as an example.
22561 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
22564 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
22565 (intra_create_variable_infos): Handle C++ constructors.
22567 2018-05-18 Martin Liska <mliska@suse.cz>
22569 * passes.def: Remove a redundant pass.
22571 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
22574 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
22576 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22578 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
22580 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
22582 (armv2, armv2a, armv3, armv3m): Delete architectures.
22583 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
22584 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
22585 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
22587 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
22588 (*mulsidi3adddi): Likewise.
22589 (mulsidi3): Likewise.
22590 (*mulsidi3_nov6): Likewise.
22591 (umulsidi3): Likewise.
22592 (umulsidi3_nov6): Likewise.
22593 (umaddsidi4): Likewise.
22594 (*umulsidi3adddi): Likewise.
22595 (smulsi3_highpart): Likewise.
22596 (*smulsi3_highpart_nov6): Likewise.
22597 (umulsi3_highpart): Likewise.
22598 (*umulsi3_highpart_nov6): Likewise.
22599 * config/arm/arm.h (arm_arch3m): Delete.
22600 * config/arm/arm.c (arm_arch3m): Delete.
22601 (arm_option_override_internal): Update armv3-related comment.
22602 (arm_configure_build_target): Delete use of isa_bit_mode32.
22603 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
22604 (arm_rtx_costs_internal): Delete check of arm_arch3m.
22605 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
22606 (mulsa3): Likewise.
22607 (mulusa3): Likewise.
22608 * config/arm/arm-protos.h (arm_arch3m): Delete.
22609 * config/arm/arm-tables.opt: Regenerate.
22610 * config/arm/arm-tune.md: Likewise.
22611 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
22612 deleted architectures.
22614 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22616 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
22617 (armv5t, armv5te): New features.
22618 (ARMv5, ARMv5e): Delete fgroups.
22619 (ARMv5t, ARMv5te): Adjust for above changes.
22620 (ARMv6m): Likewise.
22621 (armv5, armv5e): Delete arches.
22622 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
22624 (*call_reg_arm): Likewise.
22625 (*call_value_reg_armv5): Likewise.
22626 (*call_value_reg_arm): Likewise.
22627 (*call_symbol): Likewise.
22628 (*call_value_symbol): Likewise.
22629 (*sibcall_insn): Likewise.
22630 (*sibcall_value_insn): Likewise.
22631 (clzsi2): Likewise.
22632 (prefetch): Likewise.
22633 (define_split and define_peephole2 dependent on arm_arch5):
22635 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
22637 (TARGET_ARM_QBIT): Likewise.
22638 (TARGET_DSP_MULTIPLY): Likewise.
22639 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
22640 (arm_arch5, arm_arch5e): Delete.
22641 (arm_arch5t, arm_arch5te): Declare.
22642 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
22643 (arm_arch5t): Declare.
22644 (arm_option_reconfigure_globals): Update for the above.
22645 (arm_options_perform_arch_sanity_checks): Update comment, replace
22646 use of arm_arch5 with arm_arch5t.
22647 (use_return_insn): Likewise.
22648 (arm_emit_call_insn): Likewise.
22649 (output_return_instruction): Likewise.
22650 (arm_final_prescan_insn): Likewise.
22651 (arm_coproc_builtin_available): Likewise.
22652 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
22653 arm_arch5e with arm_arch5t and arm_arch5te.
22654 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
22655 (arm_arch5t, arm_arch5te): Declare.
22656 * config/arm/arm-tables.opt: Regenerate.
22657 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
22658 * config/arm/t-multilib: Likewise.
22659 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
22660 instead of arm_arch5.
22661 (*call_reg_thumb1): Likewise.
22662 (*call_value_reg_thumb1_v5): Likewise.
22663 (*call_value_reg_thumb1): Likewise.
22664 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
22666 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
22668 2018-05-18 Martin Liska <mliska@suse.cz>
22670 PR gcov-profile/84846
22671 * doc/gcov.texi: Document -t option of gcov tool.
22673 2018-05-18 Martin Liska <mliska@suse.cz>
22675 PR gcov-profile/84846
22676 * gcov.c (print_usage): Add new -t option.
22677 (process_args): Handle the option.
22678 (generate_results): Use stdout as output when requested by
22681 2018-05-18 Martin Liska <mliska@suse.cz>
22683 PR gcov-profile/84846
22684 * coverage.c (coverage_init): Write PWD to .gcno file.
22685 * doc/gcov.texi: Document how working directory is printed.
22686 * gcov-dump.c (dump_gcov_file): Print PWD.
22687 * gcov.c (output_intermediate_file): Likewise.
22688 (read_graph_file): Read PWD string.
22689 (output_lines): Print PWD.
22691 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22693 PR middle-end/85817
22694 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
22695 for retval and return false if all args to phi are zero.
22697 2018-05-18 Richard Biener <rguenther@suse.de>
22699 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
22701 (evrp_dom_walker::before_dom_children): Call it.
22703 2018-05-18 Richard Biener <rguenther@suse.de>
22705 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
22706 results when processing array refs with variable index.
22708 2018-05-18 Toon Moene <toon@moene.org>
22710 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
22711 directly after that of -floop-interchange. Indicate that both
22712 options are enabled by default when specifying -O3.
22714 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22716 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
22717 iterator. Delete separate integer-mode vec_set<mode> expander.
22718 (aarch64_simd_vec_setv2di): Delete.
22719 (vec_setv2di): Delete.
22720 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
22721 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
22722 the "w, r" alternative.
22724 2018-05-18 Martin Liska <mliska@suse.cz>
22726 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
22727 * tree-pass.h (make_pass_lower_switch_O0): New function.
22728 * tree-switch-conversion.c (node_has_low_bound): Remove.
22729 (node_has_high_bound): Likewise.
22730 (node_is_bounded): Likewise.
22731 (class pass_lower_switch): Make it a template type and create
22733 (pass_lower_switch::execute): Add template argument.
22734 (make_pass_lower_switch): New function.
22735 (make_pass_lower_switch_O0): New function.
22736 (do_jump_if_equal): Remove.
22737 (emit_case_nodes): Simplify to just handle all 3 cases and leave
22738 all the hard work to tree optimization passes.
22740 2018-05-18 Martin Liska <mliska@suse.cz>
22742 * dbgcnt.c (limit_low): Renamed from limit.
22743 (limit_high): New variable.
22744 (dbg_cnt_is_enabled): Check for upper limit.
22745 (dbg_cnt): Adjust dumping.
22746 (dbg_cnt_set_limit_by_index): Add new argument for high
22748 (dbg_cnt_set_limit_by_name): Likewise.
22749 (dbg_cnt_process_single_pair): Parse new format.
22750 (dbg_cnt_process_opt): Use strtok.
22751 (dbg_cnt_list_all_counters): Remove 'value' and add
22753 * doc/invoke.texi: Document changes.
22755 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
22757 * doc/sourcebuild.texi (scalar_all_fma): Document.
22758 * tree.def (FMA_EXPR): Delete.
22759 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
22760 * internal-fn.c (ternary_direct): New macro.
22761 (expand_ternary_optab_fn): Likewise.
22762 (direct_ternary_optab_supported_p): Likewise.
22763 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
22764 * builtins.c (fold_builtin_fma): Delete.
22765 (fold_builtin_3): Don't call it.
22766 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
22767 * expr.c (expand_expr_real_2): Likewise.
22768 * fold-const.c (operand_equal_p): Likewise.
22769 (fold_ternary_loc): Likewise.
22770 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
22771 * gimple.c (DEFTREECODE): Likewise.
22772 * gimplify.c (gimplify_expr): Likewise.
22773 * optabs-tree.c (optab_for_tree_code): Likewise.
22774 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
22775 * tree-eh.c (operation_could_trap_p): Likewise.
22776 (stmt_could_throw_1_p): Likewise.
22777 * tree-inline.c (estimate_operator_cost): Likewise.
22778 * tree-pretty-print.c (dump_generic_node): Likewise.
22779 (op_code_prio): Likewise.
22780 * tree-ssa-loop-im.c (stmt_cost): Likewise.
22781 * tree-ssa-operands.c (get_expr_operands): Likewise.
22782 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
22783 * fold-const-call.h (fold_fma): Delete.
22784 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
22785 CFN_FNMA and CFN_FNMS.
22786 (fold_fma): Delete.
22787 * genmatch.c (combined_fn): New enum.
22788 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
22789 (commutative_op): New function.
22790 (commutate): Use it. Handle more than 2 operands.
22791 (dt_operand::gen_gimple_expr): Use commutative_op.
22792 (parser::parse_expr): Allow :c to be used with non-binary
22793 operators if the commutative operand is known.
22794 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
22795 CFN_FMS, CFN_FNMA and CFN_FNMS.
22796 (backprop::process_assign_use): Remove FMA_EXPR handling.
22797 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
22798 (gen_hsa_fma): New function.
22799 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
22800 IFN_FNMA and IFN_FNMS.
22801 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
22802 * gimple-fold.h (follow_all_ssa_edges): Declare.
22803 * gimple-fold.c (follow_all_ssa_edges): New function.
22804 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
22805 gimple_build interface and use follow_all_ssa_edges to fold the result.
22806 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
22807 instead of checking for optabs directly.
22808 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
22809 rather than FMA_EXPRs.
22810 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
22811 call to IFN_FMA instead of an FMA_EXPR.
22813 2018-05-17 Jim Wilson <jimw@sifive.com>
22815 * expr.c (do_tablejump): When converting index to Pmode, if we have a
22816 sign extended promoted subreg, and the range does not have the sign bit
22817 set, then do a sign extend.
22819 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
22820 test, check for sign extended subreg and/or constant operands, and
22821 do a sign extend in that case.
22823 2018-05-17 Steve Ellcey <sellcey@cavium.com>
22825 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
22826 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
22828 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
22829 Change logics_shift_reg to logics_shift_imm.
22830 (thunderx2t99_fp_loadpair_basic): Delete.
22831 (thunderx2t99_fp_storepair_basic): Delete.
22832 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
22833 (thunderx2t99_asimd_polynomial): Delete.
22834 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
22835 and neon_fp_mul_d_scalar_q.
22836 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
22837 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
22838 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
22839 (thunderx2t99_asimd_lut): Add missing tbl types.
22840 (thunderx2t99_asimd_ext): Delete.
22841 (thunderx2t99_asimd_load1_1_mult): Delete.
22842 (thunderx2t99_asimd_load1_2_mult): Delete.
22843 (thunderx2t99_asimd_load1_ldp): New.
22844 (thunderx2t99_asimd_load1): New.
22845 (thunderx2t99_asimd_load2): Add missing *load2* types.
22846 (thunderx2t99_asimd_load3): New.
22847 (thunderx2t99_asimd_load4): New.
22848 (thunderx2t99_asimd_store1_1_mult): Delete.
22849 (thunderx2t99_asimd_store1_2_mult): Delete.
22850 (thunderx2t99_asimd_store2_mult): Delete.
22851 (thunderx2t99_asimd_store2_onelane): Delete.
22852 (thunderx2t99_asimd_store_stp): New.
22853 (thunderx2t99_asimd_store1): New.
22854 (thunderx2t99_asimd_store2): New.
22855 (thunderx2t99_asimd_store3): New.
22856 (thunderx2t99_asimd_store4): New.
22858 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
22860 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
22861 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
22863 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
22864 Segher Boessenkool <segher@kernel.crashing.org>
22867 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
22870 2018-05-17 Richard Biener <rguenther@suse.de>
22872 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
22873 for pruning loop and prune defs feeding only already visited PHIs.
22875 2018-05-17 Richard Biener <rguenther@suse.de>
22877 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
22879 2018-05-17 Bin Cheng <bin.cheng@arm.com>
22880 Richard Biener <rguenther@suse.de>
22882 PR tree-optimization/85793
22883 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
22884 for VMAT_ELEMENTWISE.
22886 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
22888 * internal-fn.h (lookup_internal_fn): Declare
22889 * internal-fn.c (lookup_internal_fn): New function.
22890 * gimple.c (gimple_build_call_from_tree): Handle calls to
22891 internal functions.
22892 * gimple-pretty-print.c (dump_gimple_call): Print "." before
22893 internal function names.
22894 * tree-pretty-print.c (dump_generic_node): Likewise.
22895 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
22897 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
22899 * gimple-fold.h (gimple_build): Make the function forms take
22900 combined_fn rather than built_in_function.
22901 (gimple_simplify): Likewise.
22902 * gimple-match-head.c (gimple_simplify): Likewise.
22903 * gimple-fold.c (gimple_build): Likewise.
22904 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
22905 rather than gimple_build_call_internal.
22906 (get_initial_defs_for_reduction): Likewise.
22907 (vect_create_epilog_for_reduction): Likewise.
22908 (vectorizable_live_operation): Likewise.
22910 2018-05-17 Martin Liska <mliska@suse.cz>
22912 * gimple-ssa-sprintf.c (format_directive): Do not use
22913 space in between 'G_' and '('.
22915 2018-05-17 Jakub Jelinek <jakub@redhat.com>
22918 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
22919 even if the mask is not all ones.
22922 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
22924 (ix86_gimple_fold_builtin): Likewise.
22927 * config/i386/i386.c: Include tree-vector-builder.h.
22928 (ix86_vector_shift_count): New function.
22929 (ix86_fold_builtin): Fold shift builtins by scalar count.
22930 (ix86_gimple_fold_builtin): Likewise.
22932 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
22933 _mm512_setzero): New intrinsics.
22935 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
22936 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22938 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
22939 code generation for cases where splatting a value is not useful.
22940 * simplify-rtx.c (simplify_ternary_operation): Simplify
22941 vec_merge across a vec_duplicate and a paradoxical subreg forming
22942 a vector mode to a vec_concat.
22944 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
22946 * config.gcc: Support "goldmont-plus".
22947 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
22949 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
22950 PROCESSOR_GOLDMONT_PLUS.
22951 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
22952 (processor_target_table): Add "goldmont-plus".
22953 (PTA_GOLDMONT_PLUS): Define.
22954 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
22955 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
22956 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
22957 (fold_builtin_cpu): Add "goldmont-plus".
22958 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
22959 (ix86_option_override_internal): Add "goldmont-plus".
22960 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
22961 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
22962 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
22963 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
22965 2018-05-17 Richard Biener <rguenther@suse.de>
22967 PR tree-optimization/85757
22968 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
22969 remove defs that only feed that PHI from further processing.
22971 2018-05-16 Jim Wilson <jimw@sifive.com>
22973 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
22975 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
22977 2018-05-16 Mark Wielaard <mark@klomp.org>
22979 * dwarf2out.c (count_index_strings): New function.
22980 (output_indirect_strings): Call count_index_strings and generate
22981 header for dwarf_version >= 5.
22983 2018-05-16 Mark Wielaard <mark@klomp.org>
22985 * dwarf2out.c (dwarf_FORM): New function.
22986 (set_indirect_string): Use dwarf_FORM.
22987 (reset_indirect_string): Likewise.
22988 (size_of_die): Likewise.
22989 (value_format): Likewise.
22990 (output_die): Likewise.
22991 (add_skeleton_AT_string): Likewise.
22992 (output_macinfo_op): Likewise.
22993 (index_string): Likewise.
22994 (output_index_string_offset): Likewise.
22995 (output_index_string): Likewise.
22996 (count_index_strings): Likewise.
22998 2018-05-16 Carl Love <cel@us.ibm.com>
23000 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
23001 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
23003 2018-05-16 Martin Jambor <mjambor@suse.cz>
23005 * ipa-prop.c (ipa_free_all_edge_args): Remove.
23006 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
23008 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
23010 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
23011 (fnma<mode>4): Likewise.
23012 (fms<mode>4): Likewise.
23013 (fnms<mode>4): Likewise.
23014 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
23015 (aarch64_fnma<mode>4): Likewise.
23016 (aarch64_fms<mode>4): Likewise.
23017 (aarch64_fnms<mode>4): Likewise.
23018 (aarch64_fnmadd<mode>4): Likewise.
23020 2018-05-16 Jason Merrill <jason@redhat.com>
23022 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
23024 2018-05-16 Richard Biener <rguenther@suse.de>
23026 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
23027 (dump_stmt_cost): Declare.
23028 (add_stmt_cost): Dump cost we add.
23029 (add_stmt_costs): New function.
23030 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
23031 No longer exported.
23032 (vect_analyze_stmt): Adjust prototype.
23033 (vectorizable_condition): Likewise.
23034 (vectorizable_live_operation): Likewise.
23035 (vectorizable_reduction): Likewise.
23036 (vectorizable_induction): Likewise.
23037 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
23038 cost vector to pass to vectorizable_ and record afterwards.
23039 (vect_model_reduction_cost): Take cost vector argument and adjust.
23040 (vect_model_induction_cost): Likewise.
23041 (vectorizable_reduction): Likewise.
23042 (vectorizable_induction): Likewise.
23043 (vectorizable_live_operation): Likewise.
23044 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
23045 SLP_TREE_NUMBER_OF_VEC_STMTS.
23046 (vect_analyze_slp_cost_1): Remove.
23047 (vect_analyze_slp_cost): Likewise.
23048 (vect_slp_analyze_node_operations): Take visited args and
23049 a target cost vector. Avoid processing already visited stmt sets.
23050 (vect_slp_analyze_operations): Use a local cost vector to gather
23051 costs and register those of non-discarded instances.
23052 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
23053 (vect_schedule_slp_instance): Remove copying of
23054 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
23056 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
23057 adding cost. Record cost entry location.
23058 (vect_prologue_cost_for_slp_op): Function to compute cost of
23059 a constant or invariant generated for SLP vect in the prologue,
23060 split out from vect_analyze_slp_cost_1.
23061 (vect_model_simple_cost): Make static. Adjust for SLP costing.
23062 (vect_model_promotion_demotion_cost): Likewise.
23063 (vect_model_store_cost): Likewise, make static.
23064 (vect_model_load_cost): Likewise.
23065 (vectorizable_bswap): Add cost vector arg and adjust.
23066 (vectorizable_call): Likewise.
23067 (vectorizable_simd_clone_call): Likewise.
23068 (vectorizable_conversion): Likewise.
23069 (vectorizable_assignment): Likewise.
23070 (vectorizable_shift): Likewise.
23071 (vectorizable_operation): Likewise.
23072 (vectorizable_store): Likewise.
23073 (vectorizable_load): Likewise.
23074 (vectorizable_condition): Likewise.
23075 (vectorizable_comparison): Likewise.
23076 (can_vectorize_live_stmts): Likewise.
23077 (vect_analyze_stmt): Likewise.
23078 (vect_transform_stmt): Adjust calls to vectorizable_*.
23079 * tree-vectorizer.c: Include gimple-pretty-print.h.
23080 (dump_stmt_cost): New function.
23082 2018-05-16 Richard Biener <rguenther@suse.de>
23084 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
23085 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
23086 * tree-ssa-dse.c: Include tree-ssa-loop.h.
23087 (check_name): New callback.
23088 (dse_classify_store): Track cycles via a visited bitmap of PHI
23089 defs and simplify handling of in-loop and across loop dead stores
23090 and properly fail for loop-variant refs. Handle byte-tracking with
23091 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
23094 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
23096 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
23097 (vect_get_mask_type_for_stmt): Likewise.
23098 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
23100 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
23101 to determine the statement's vector type and the vector type that
23102 should be used for calculating nunits. Deal with cases in which
23103 the type has to be deferred.
23104 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
23105 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
23106 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
23107 (vect_determine_vf_for_stmt): New functions, split out from...
23108 (vect_determine_vectorization_factor): ...here.
23109 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
23110 (vect_get_mask_type_for_stmt): New functions, split out from
23111 vect_determine_vectorization_factor.
23113 2018-05-16 Richard Biener <rguenther@suse.de>
23115 * tree-cfg.c (verify_gimple_assign_ternary): Properly
23116 verify the [VEC_]COND_EXPR embedded comparison.
23118 2018-05-15 Martin Sebor <msebor@redhat.com>
23120 PR tree-optimization/85753
23121 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
23122 RECORD_TYPE in addition to ARRAY_TYPE.
23124 2018-05-15 Martin Sebor <msebor@redhat.com>
23126 PR middle-end/85643
23127 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
23129 2018-05-15 Richard Biener <rguenther@suse.de>
23131 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
23132 add by_clobber_p one. Change algorithm to collect all defs
23133 representing uses we need to walk and try reducing them to
23134 a single one before failing.
23135 (dse_dom_walker::dse_optimize_stmt): Adjust.
23137 2018-05-13 Mark Wielaard <mark@klomp.org>
23139 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
23140 (size_of_loc_descr): Likewise.
23141 (output_loc_operands): Likewise.
23142 (output_loc_operands_raw): Likewise.
23143 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
23144 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
23145 (hash_loc_operands): Likewise.
23146 (compare_loc_operands): Likewise.
23148 2018-05-14 Mark Wielaard <mark@klomp.org>
23150 * dwarf2out.c (count_index_addrs): New function.
23151 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
23153 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23155 PR tree-optimization/83648
23156 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
23157 return value as malloc candidate.
23159 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23162 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
23163 param as true in call to suggest_attribute.
23165 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
23167 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
23168 -mreadonly-in-sdata.
23170 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23172 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
23174 (aarch64_crypto_aesd_fused): Likewise.
23176 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
23178 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
23179 (movsi_aarch64): Likewise.
23180 (load_pairsi): Likewise.
23181 (load_pairdi): Likewise.
23182 (store_pairsi): Likewise.
23183 (store_pairdi): Likewise.
23184 (load_pairsf): Likewise.
23185 (load_pairdf): Likewise.
23186 (store_pairsf): Likewise.
23187 (store_pairdf): Likewise.
23188 (zero_extend): Likewise.
23189 (trunc): Swap alternatives.
23190 (fcvt_target): Add '?' to prefer w over r.
23192 2018-05-14 Jakub Jelinek <jakub@redhat.com>
23195 * config/i386/i386.md: Disallow non-commutative arithmetics in
23196 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
23197 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
23198 in the peephole2 before it.
23200 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
23202 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
23203 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
23204 (ix86_handle_option): Handle -mcldemote.
23205 * config.gcc: New header.
23206 * config/i386/cldemoteintrin.h: New file.
23207 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
23208 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
23210 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
23211 OPTION_MASK_ISA_CLDEMOTE.
23212 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
23213 (ix86_valid_target_attribute_inner_p): Ditto.
23214 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
23215 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
23216 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
23217 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
23218 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
23220 * config/i386/i386.opt: Add -mcldemote.
23221 * config/i386/x86intrin.h: New header.
23222 * doc/invoke.texi: Add -mcldemote.
23224 2018-05-14 Richard Biener <rguenther@suse.de>
23226 * doc/match-and-simplify.texi: Adjust :s documentation.
23228 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
23230 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
23231 intended memcpy size.
23232 (REORDER_45): Likewise.
23234 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
23236 * sort.cc: New file.
23237 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
23238 * vec.c (qsort_chk): Use gcc_qsort.
23239 * Makefile.in (OBJS-libcommon): Add sort.o.
23240 (build/sort.o): New target. Use it...
23241 (BUILD_RTL): ... here, and...
23242 (build/gencfn-macros): ... here, and...
23243 (build/genmatch): ... here.
23245 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
23246 Chung-Ju Wu <jasonwucj@gmail.com>
23248 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
23249 * config/nds32/nds32-graywolf.md: New file.
23250 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
23251 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
23253 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
23254 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
23255 * config/nds32/nds32.md (pipeline_model): Add graywolf.
23256 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
23257 * config/nds32/pipelines.md: Include n15 settings.
23259 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
23260 Chung-Ju Wu <jasonwucj@gmail.com>
23262 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
23263 * config/nds32/nds32-n13.md: New file.
23264 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
23265 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
23267 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
23268 * config/nds32/nds32.md (pipeline_model): Add n13.
23269 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
23270 * config/nds32/pipelines.md: Include n13 settings.
23272 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
23273 Chung-Ju Wu <jasonwucj@gmail.com>
23275 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
23276 * config/nds32/nds32-n10.md: New file.
23277 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
23278 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
23280 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
23281 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
23282 * config/nds32/nds32.md (pipeline_model): Add n10.
23283 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
23284 * config/nds32/pipelines.md: Include n10 settings.
23286 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
23287 Kito Cheng <kito.cheng@gmail.com>
23288 Chung-Ju Wu <jasonwucj@gmail.com>
23290 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
23291 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
23292 Add enum values for DSP extension instructions.
23293 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
23295 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
23296 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
23297 New code iterators.
23298 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
23299 * config/nds32/nds32-dspext.md: New file for DSP implementation.
23300 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
23301 * config/nds32/nds32-intrinsic.md: Likewise.
23302 * config/nds32/nds32_intrinsic.h: Likewise.
23303 * config/nds32/nds32-md-auxiliary.c: Likewise.
23304 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
23305 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
23306 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
23307 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
23308 * config/nds32/nds32-protos.h: New declarations for DSP extension.
23309 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
23310 TYPE_DMAC in switch statement.
23311 * config/nds32/nds32.c: New checking and implementation for DSP
23312 extension instructions.
23313 * config/nds32/nds32.h: Likewise.
23314 * config/nds32/nds32.md: Likewise.
23315 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
23316 * config/nds32/predicates.md: Implement new predicates for DSP
23319 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
23321 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
23322 Reformat alternatives and attributes so it is easier to identify
23323 which constraints/attributes go with which instruction.
23324 (mov<mode>_hardfloat32, FMOVE64): Likewise.
23325 (mov<mode>_softfloat32, FMOVE64): Likewise.
23326 (mov<mode>_hardfloat64, FMOVE64): Likewise.
23327 (mov<mode>_softfloat64, FMOVE64): Likewise.
23329 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
23331 * doc/extend.texi (PowerPC Built-in Functions): Rename this
23333 (Basic PowerPC Built-in Functions): The new name of the
23334 subsection previously known as "PowerPC Built-in Functions".
23335 (Basic PowerPC Built-in Functions Available on all Configurations):
23337 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
23338 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
23339 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
23340 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
23342 2018-05-11 Martin Jambor <mjambor@suse.cz>
23345 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
23348 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
23351 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
23353 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
23355 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
23356 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
23357 (ix86_handle_option): Handle -mwaitpkg.
23358 * config.gcc: New header.
23359 * config/i386/cpuid.h (bit_WAITPKG): New bit.
23360 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
23361 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
23363 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
23364 OPTION_MASK_ISA_WAITPKG.
23365 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
23366 (ix86_option_override_internal): Add PTA_WAITPKG.
23367 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
23368 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
23369 IX86_BUILTIN_TPAUSE.
23370 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
23371 __builtin_ia32_umwait and __builtin_ia32_tpause.
23372 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
23373 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
23374 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
23375 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
23376 UNSPECV_TPAUSE): New.
23377 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
23378 * config/i386/i386.opt: Add -mwaitpkg.
23379 * config/i386/waitpkgintrin.h: New file.
23380 * config/i386/x86intrin.h: New header.
23381 * doc/invoke.texi: Add -mwaitpkg.
23383 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
23386 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
23388 (cortex-m0): Use armv6s-m isa.
23389 (cortex-m0plus): Likewise.
23390 (cortex-m1): Likewise.
23391 (cortex-m0.small-multiply): Likewise.
23392 (cortex-m0plus.small-multiply): Likewise.
23393 (cortex-m1.small-multiply): Likewise.
23395 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
23396 Jakub Jelinek <jakub@redhat.com>
23398 PR tree-optimization/85692
23399 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
23400 source permute as well.
23402 2018-05-11 Martin Liska <mliska@suse.cz>
23405 * doc/extend.texi: Document LLVM style format for no_sanitize
23408 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
23410 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
23411 mode_supports_vsx_dform_quad to mode_supports_dq_form.
23412 (mode_supports_vsx_dform_quad): Likewise.
23413 (mode_supports_vmx_dform): Move these functions to be next to the
23414 other mode_supports functions.
23415 (mode_supports_dq_form): Likewise.
23416 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
23417 mode_supports_dq_form.
23418 (reg_offset_addressing_ok_p): Likewise.
23419 (offsettable_ok_by_alignment): Likewise.
23420 (rs6000_legitimate_offset_address_p): Likewise.
23421 (legitimate_lo_sum_address_p): Likewise.
23422 (rs6000_legitimize_address): Likewise.
23423 (rs6000_legitimize_reload_address): Likewise.
23424 (rs6000_secondary_reload_inner): Likewise.
23425 (rs6000_preferred_reload_class): Likewise.
23426 (rs6000_output_move_128bit): Likewise.
23428 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
23430 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
23431 Generate SImode target register for null target.
23432 <case IX86_BUILTIN_XGETBV>: Ditto.
23433 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
23434 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
23436 2018-05-10 Carl Love <cel@us.ibm.com>
23438 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
23439 dcbtt and dcbtstt if operands[2] is 0.
23441 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
23444 * config/i386/sse.md (usadv64qi): New expander.
23446 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
23448 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
23449 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
23450 -maltivec=be support.
23451 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
23452 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
23453 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
23454 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
23455 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
23456 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
23457 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
23458 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
23459 altivec_vsumsws): Adjust.
23460 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
23461 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
23462 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
23464 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
23465 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
23466 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
23467 (altivec_lve<VI_char>x): Delete expand.
23468 (*altivec_lve<VI_char>x_internal): Rename to...
23469 (altivec_lve<VI_char>x): ... this.
23470 (altivec_lvxl_<mode>): Delete expand.
23471 (*altivec_lvxl_<mode>_internal): Rename to ...
23472 (altivec_lvxl_<mode>): ... this.
23473 (altivec_stvxl_<mode>): Delete expand.
23474 (*altivec_stvxl_<mode>_internal): Rename to ...
23475 (altivec_stvxl_<mode>): ... this.
23476 (altivec_stve<VI_char>x): Delete expand.
23477 (*altivec_stve<VI_char>x_internal): Rename to ...
23478 (altivec_stve<VI_char>x): ... this.
23479 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
23480 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
23481 reduc_plus_scal_<mode>): Adjust.
23482 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
23484 (rs6000_cpu_cpp_builtins): Adjust.
23485 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
23486 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
23487 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
23488 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
23489 -maltivec=be support.
23490 (rs6000_split_vec_extract_var): Adjust.
23491 (rs6000_split_v4si_init): Adjust.
23492 (swap_selector_for_mode): Delete.
23493 (altivec_expand_lvx_be, altivec_expand_stvx_be,
23494 altivec_expand_stvex_be): Delete.
23495 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
23496 -maltivec=be support.
23497 (rs6000_gimple_fold_builtin): Ditto.
23498 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
23500 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
23501 (TARGET_DIRECT_MOVE_64BIT): Adjust.
23502 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
23503 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
23504 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
23505 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
23506 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
23507 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
23508 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
23509 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
23510 anonymous split): Adjust.
23511 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
23512 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
23514 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
23516 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
23517 when --with-gxx-include-dir is also specified.
23518 * configure: Regenerate.
23520 2018-05-09 Jim Wilson <jimw@sifive.com>
23523 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
23524 * config/riscv/t-withmultilib: New.
23525 * config/riscv/withmultilib.h: New.
23526 * doc/install.texi: Document RISC-V --with-multilib-list support.
23528 2018-05-09 Richard Biener <rguenther@suse.de>
23530 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
23532 (vect_bb_vectorization_profitable_p): Adjust. Compute
23533 actual scalar cost using the cost vector and the add_stmt_cost
23536 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
23538 PR rtl-optimization/85645
23539 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
23540 in the REG_CFA_REGISTER note for LR, don't leave it empty.
23542 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
23544 PR rtl-optimization/85645
23545 * shrink-wrap.c (spread_components): Return a boolean saying if
23546 anything was changed.
23547 (try_shrink_wrapping_separate): Iterate spread_components until
23548 nothing changes anymore.
23550 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
23552 PR rtl-optimization/85645
23553 * regrename.c (build_def_use): Also kill the chains that include the
23554 destination of a REG_CFA_REGISTER note.
23556 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
23558 PR rtl-optimization/85645
23559 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
23560 insn that has a REG_CFA_REGISTER note.
23562 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
23564 * cfgexpand.c (expand_clobber): New function.
23565 (expand_gimple_stmt_1): Use it.
23566 * tree-vect-stmts.c (vect_clobber_variable): New function,
23568 (vectorizable_simd_clone_call): ...here.
23569 (vectorizable_store): Emit a clobber either side of an
23570 IFN_STORE_LANES sequence.
23571 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
23573 2018-05-09 Tom de Vries <tom@codesourcery.com>
23576 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
23577 (define_insn "trap_if_false"): Add exit after trap.
23579 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
23581 PR rtl-optimization/85638
23582 * bb-reorder.c: Include common/common-target.h.
23583 (create_forwarder_block): New function extracted from...
23584 (fix_up_crossing_landing_pad): ...here. Rename into...
23585 (dw2_fix_up_crossing_landing_pad): ...this.
23586 (sjlj_fix_up_crossing_landing_pad): New function.
23587 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
23588 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
23589 from both partitions and exit the loop after one iteration.
23591 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
23594 * doc/extend.texi (PowerPC Built-in Functions): Rename this
23596 (Basic PowerPC Built-in Functions): The new name of the
23597 subsection previously known as "PowerPC Built-in Functions".
23598 (Basic PowerPC Built-in Functions Available on all Configurations):
23600 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
23602 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
23604 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
23606 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
23609 2018-05-08 Jim Wilson <jimw@sifive.com>
23611 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
23612 (LD_EMUL_SUFFIX): New.
23613 (LINK_SPEC): Use it.
23615 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
23617 * doc/extend.texi (PowerPC Built-in Functions): Rename this
23619 (Basic PowerPC Built-in Functions): The new name of the
23620 subsection previously known as "PowerPC Built-in Functions".
23621 (Basic PowerPC Built-in Functions Available on all Configurations):
23623 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
23625 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
23627 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
23629 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
23632 2018-05-08 Jakub Jelinek <jakub@redhat.com>
23635 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
23636 after cmpelim optimization.
23638 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
23640 * config.gcc: Support "goldmont".
23641 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
23642 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
23643 PROCESSOR_GOLDMONT.
23644 * config/i386/i386.c (m_GOLDMONT): Define.
23645 (processor_target_table): Add "goldmont".
23646 (PTA_GOLDMONT): Define.
23647 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
23648 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
23649 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
23650 (fold_builtin_cpu): Add "goldmont".
23651 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
23652 (ix86_option_override_internal): Add "goldmont".
23653 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
23654 (processor_type): Add PROCESSOR_GOLDMONT.
23655 * config/i386/i386.md: Add CPU "glm".
23656 * config/i386/glm.md: New file.
23657 * config/i386/x86-tune.def: Add m_GOLDMONT.
23658 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
23660 2018-05-08 Jakub Jelinek <jakub@redhat.com>
23663 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
23665 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
23666 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
23667 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
23670 * config/i386/i386.c (ix86_fold_builtin): Handle
23671 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
23674 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
23675 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
23677 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
23680 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
23681 (check_arch): Likewise.
23682 (check_fpu): Return the result rather than printing it.
23683 (end arch): Fix operator precedence.
23684 (end cpu): Likewise.
23685 (END): Print the result from check_fpu.
23687 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
23688 Alan Hayward <alan.hayward@arm.com>
23689 David Sherwood <david.sherwood@arm.com>
23691 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
23692 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
23693 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
23694 (*fcmuo<mode>_and): New patterns.
23696 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
23698 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
23699 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
23700 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
23701 (cmp_op, sve_imm_con): New code attributes.
23702 (SVE_COND_INT_CMP, imm_con): Delete.
23703 (cmp_op): Remove above unspecs from int attribute.
23704 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
23706 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
23707 comparison-specific unspecs.
23708 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
23709 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
23710 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
23711 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
23712 (*vec_fcm<cmp_op><mode>): Rename to...
23713 (*fcm<cmp_op><mode>): ...this and adjust likewise.
23714 (*vec_fcmuo<mode>): Rename to...
23715 (*fcmuo<mode>): ...this and adjust likewise.
23716 (*pred_fcm<cmp_op><mode>): New pattern.
23717 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
23718 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
23720 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
23722 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
23723 (aarch64_emit_sve_predicated_cond): New function.
23724 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
23725 (aarch64_emit_unspec_cond_or): Replace with...
23726 (aarch64_emit_sve_or_conds): ...this new function. Use
23727 aarch64_emit_sve_ptrue_op for the individual comparisons and
23728 aarch64_emit_binop to OR them together.
23729 (aarch64_emit_inverted_unspec_cond): Replace with...
23730 (aarch64_emit_sve_inverted_cond): ...this new function. Use
23731 aarch64_emit_sve_ptrue_op for the comparison and
23732 aarch64_emit_unop to invert the result.
23733 (aarch64_expand_sve_vec_cmp_float): Update after the above
23734 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
23736 2018-05-07 Nathan Sidwell <nathan@acm.org>
23738 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
23739 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
23740 (Backwards Compatibility): Likewise.
23742 2018-05-07 Luis Machado <luis.machado@linaro.org>
23746 2018-05-07 Luis Machado <luis.machado@linaro.org>
23748 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
23749 <prefetch_dynamic_strides>: New const bool field.
23750 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
23751 prefetch_dynamic_strides.
23752 (exynosm1_prefetch_tune): Likewise.
23753 (thunderxt88_prefetch_tune): Likewise.
23754 (thunderx_prefetch_tune): Likewise.
23755 (thunderx2t99_prefetch_tune): Likewise.
23756 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
23758 (aarch64_override_options_internal): Update to set
23759 PARAM_PREFETCH_DYNAMIC_STRIDES.
23760 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
23761 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
23762 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
23763 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
23764 prefetch-dynamic-strides setting.
23766 2018-05-07 Luis Machado <luis.machado@linaro.org>
23768 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
23769 <minimum_stride>: New const int field.
23770 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
23771 minimum_stride field.
23772 (exynosm1_prefetch_tune): Likewise.
23773 (thunderxt88_prefetch_tune): Likewise.
23774 (thunderx_prefetch_tune): Likewise.
23775 (thunderx2t99_prefetch_tune): Likewise.
23776 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
23777 (aarch64_override_options_internal): Update to set
23778 PARAM_PREFETCH_MINIMUM_STRIDE.
23779 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
23780 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
23781 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
23782 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
23783 stride is constant and is below the minimum stride threshold.
23785 2018-05-07 Luis Machado <luis.machado@linaro.org>
23787 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
23790 2018-05-07 Luis Machado <luis.machado@linaro.org>
23792 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
23793 <prefetch_dynamic_strides>: New const bool field.
23794 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
23795 prefetch_dynamic_strides.
23796 (exynosm1_prefetch_tune): Likewise.
23797 (thunderxt88_prefetch_tune): Likewise.
23798 (thunderx_prefetch_tune): Likewise.
23799 (thunderx2t99_prefetch_tune): Likewise.
23800 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
23802 (aarch64_override_options_internal): Update to set
23803 PARAM_PREFETCH_DYNAMIC_STRIDES.
23804 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
23805 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
23806 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
23807 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
23808 prefetch-dynamic-strides setting.
23810 2018-05-07 Luis Machado <luis.machado@linaro.org>
23812 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
23813 <minimum_stride>: New const int field.
23814 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
23815 minimum_stride field.
23816 (exynosm1_prefetch_tune): Likewise.
23817 (thunderxt88_prefetch_tune): Likewise.
23818 (thunderx_prefetch_tune): Likewise.
23819 (thunderx2t99_prefetch_tune): Likewise.
23820 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
23821 (aarch64_override_options_internal): Update to set
23822 PARAM_PREFETCH_MINIMUM_STRIDE.
23823 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
23824 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
23825 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
23826 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
23827 stride is constant and is below the minimum stride threshold.
23829 2018-05-06 Jakub Jelinek <jakub@redhat.com>
23832 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
23833 the type is addressable. Don't force op into register if it has
23836 2018-05-05 Roland McGrath <mcgrathr@google.com>
23839 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
23840 any section for which we don't know a specific type it should have,
23841 regardless of name. Previously this was done only for the exact
23842 names ".init_array", ".fini_array", and ".preinit_array".
23843 (default_elf_asm_named_section): Add comment about
23844 relationship with default_section_type_flags and SECTION_NOTYPE.
23845 (get_section): Don't consider it a type conflict if one side has
23846 SECTION_NOTYPE and the other doesn't, as long as neither has the
23847 SECTION_BSS et al used in the default_section_type_flags logic.
23849 2018-05-05 Tom de Vries <tom@codesourcery.com>
23852 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
23853 (workaround_barsyncs): New function.
23854 (nvptx_reorg): Use workaround_barsyncs.
23855 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
23856 (define_expand "nvptx_membar_cta"): New define_expand.
23857 (define_insn "*nvptx_membar_cta"): New insn.
23859 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
23861 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
23862 To improve optimization opportunities.
23863 * builtin-types.def: The new needed builtin types for the above.
23865 2018-05-04 Richard Biener <rguenther@suse.de>
23867 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
23868 * gimple-ssa-store-merging.c
23869 (imm_store_chain_info::output_merged_store): Remove redundant create,
23870 release split_store vector contents on failure.
23871 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
23872 scalar stmt vector on cache hit.
23874 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
23876 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
23878 * config.gcc (powerpc-xilinx-eabi*): Remove.
23879 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
23881 (fusion_addis_mem_combo_load): Ditto.
23882 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
23884 (rs6000_cpu_cpp_builtins): Ditto.
23885 * config/rs6000/rs6000-linux.c
23886 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
23887 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
23888 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
23890 (rs6000_setup_reg_addr_masks): Ditto.
23891 (rs6000_init_hard_regno_mode_ok): Ditto.
23892 (rs6000_option_override_internal): Ditto.
23893 (legitimate_lo_sum_address_p): Ditto.
23894 (rs6000_legitimize_address): Ditto.
23895 (rs6000_legitimize_reload_address): Ditto.
23896 (rs6000_legitimate_address_p): Ditto.
23897 (abi_v4_pass_in_fpr): Ditto.
23898 (setup_incoming_varargs): Ditto.
23899 (rs6000_gimplify_va_arg): Ditto.
23900 (rs6000_split_multireg_move): Ditto.
23901 (rs6000_savres_strategy): Ditto.
23902 (rs6000_emit_prologue_components): Ditto.
23903 (rs6000_emit_epilogue_components): Ditto.
23904 (rs6000_emit_prologue): Ditto.
23905 (rs6000_emit_epilogue): Ditto.
23906 (rs6000_elf_file_end): Ditto.
23907 (rs6000_function_value): Ditto.
23908 (rs6000_libcall_value): Ditto.
23909 * config/rs6000/rs6000.h: Ditto.
23910 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
23911 (TARGET_MINMAX): ... this. New.
23912 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
23913 * config/rs6000/rs6000.md: Remove Xilinx FP support.
23914 (*movsi_internal1_single): Delete.
23915 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
23916 mfpu=, mxilinx-fpu): Delete.
23917 * config/rs6000/singlefp.h: Delete.
23918 * config/rs6000/sysv4.h: Remove Xilinx FP support.
23919 * config/rs6000/t-rs6000: Ditto.
23920 * config/rs6000/t-xilinx: Delete.
23921 * config/rs6000/titan.md: Adjust for fp_type removal.
23922 * config/rs6000/vsx.md: Remove Xilinx FP support.
23923 (VStype_simple): Delete.
23924 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
23925 * config/rs6000/xfpu.h: Delete.
23926 * config/rs6000/xfpu.md: Delete.
23927 * config/rs6000/xilinx.h: Delete.
23928 * config/rs6000/xilinx.opt: Delete.
23929 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
23930 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
23932 2018-05-04 Tom de Vries <tom@codesourcery.com>
23935 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
23938 2018-05-04 Richard Biener <rguenther@suse.de>
23940 PR middle-end/85627
23941 * tree-complex.c (update_complex_assignment): We are always in SSA form.
23942 (expand_complex_div_wide): Likewise.
23943 (expand_complex_operations_1): Likewise.
23944 (expand_complex_libcall): Preserve EH info of the original stmt.
23945 (tree_lower_complex): Handle removed blocks.
23946 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
23947 on complex multiplication and division libcall builtins.
23949 2018-05-04 Richard Biener <rguenther@suse.de>
23951 PR middle-end/85574
23952 * fold-const.c (negate_expr_p): Restrict negation of operand
23953 zero of a division to when we know that can happen without
23955 (fold_negate_expr_1): Likewise.
23957 2018-05-04 Jakub Jelinek <jakub@redhat.com>
23960 * real.h (real_nextafter): Declare.
23961 * real.c (real_nextafter): New function.
23962 * fold-const-call.c (fold_const_nextafter): New function.
23963 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
23964 CASE_CFN_NEXTTOWARD.
23965 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
23966 even when arg1_mode is different from arg0_mode.
23968 2018-05-03 Nathan Sidwell <nathan@acm.org>
23970 * doc/extend.texi (Deprecated Features): Remove
23971 -ffriend-injection.
23972 (Backwards Compatibility): Likewise.
23973 * doc/invoke.texi (C++ Language Options): Likewise.
23974 (C++ Dialect Options): Likewise.
23976 2018-05-03 Jakub Jelinek <jakub@redhat.com>
23979 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
23980 _mm512_mask_mullox_epi64): New intrinsics.
23982 2018-05-03 Tom de Vries <tom@codesourcery.com>
23985 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
23986 dump files): Add offload-tree.
23988 2018-05-03 Richard Biener <rguenther@suse.de>
23990 PR tree-optimization/85615
23991 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
23992 to loops not nested in BBs loop father to avoid creating multi-entry
23995 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23997 PR tree-optimization/70291
23998 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
23999 arguments. Change return type to tree. Emit libcall as a new
24000 statement rather than replacing existing one when inplace_p is true.
24001 (expand_complex_multiplication_components): New function.
24002 (expand_complex_multiplication): Expand floating-point complex
24003 multiplication using the above.
24004 (expand_complex_division): Rename inner_type parameter to type.
24005 Update expand_complex_libcall call-site.
24006 (expand_complex_operations_1): Update expand_complex_multiplication
24007 and expand_complex_division call-sites.
24009 2018-05-02 Jakub Jelinek <jakub@redhat.com>
24012 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
24013 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
24014 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
24015 the highest significant bit of the shift count mask is clear. In
24016 check whether and[sq]i3 is needed verify that all significant bits
24017 of the shift count other than the highest are set.
24019 2018-05-02 Tom de Vries <tom@codesourcery.com>
24022 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
24023 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
24024 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
24025 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
24026 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
24027 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
24028 __builtin_goacc_parlevel_size.
24030 2018-05-02 Richard Biener <rguenther@suse.de>
24032 PR tree-optimization/85597
24033 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
24034 do not use split vect_get_vec_defs call but call vect_get_slp_defs
24037 2018-05-02 Tom de Vries <tom@codesourcery.com>
24040 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
24041 dump files): Add ltrans-tree.
24043 2018-05-02 Tom de Vries <tom@codesourcery.com>
24046 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
24047 dump files): Add wpa-ipa.
24049 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
24051 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
24052 powerpc*-*-linux*paired* target.
24053 * config/rs6000/750cl.h: Delete.
24054 * config/rs6000/paired.h: Delete.
24055 * config/rs6000/paired.md: Delete.
24056 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
24058 * config/rs6000/rs6000-builtin.def: Remove paired float support.
24059 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
24060 comment. Remove paired float support.
24061 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
24062 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
24064 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
24065 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
24067 * config/rs6000/rs6000.c: Remove paired float support.
24068 (paired_expand_vector_init, paired_expand_vector_move,
24069 paired_emit_vector_compare, paired_emit_vector_cond_expr,
24070 (paired_expand_lv_builtin, paired_expand_stv_builtin,
24071 paired_expand_builtin, paired_expand_predicate_builtin,
24072 paired_init_builtins): Delete.
24073 * config/rs6000/rs6000.h: Remove paired float support.
24074 * config/rs6000/rs6000.md: Remove paired float support.
24075 (move_from_CR_ov_bit): Delete.
24076 * config/rs6000/rs6000.opt (mpaired): Delete.
24077 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
24078 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
24080 2018-05-02 Richard Biener <rguenther@suse.de>
24082 PR middle-end/85567
24083 * gimplify.c (gimplify_save_expr): When in SSA form allow
24084 SAVE_EXPRs to compute to SSA vars.
24086 2018-05-02 Jakub Jelinek <jakub@redhat.com>
24089 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
24090 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
24091 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
24092 clobber operands[2], instead use a new pseudo. Formatting fixes.
24094 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
24096 PR tree-optimization/85586
24097 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
24098 exit early for statements in the same group if the accesses are
24101 2018-05-02 Tom de Vries <tom@codesourcery.com>
24104 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
24107 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
24109 PR tree-optimization/85143
24110 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
24112 2018-05-01 Tom de Vries <tom@codesourcery.com>
24115 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
24116 not found" error message.
24118 2018-05-01 Tom de Vries <tom@codesourcery.com>
24121 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
24122 * vec.c (test_ordered_remove_if): New function.
24123 (vec_c_tests): Call test_ordered_remove_if.
24124 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
24125 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
24126 * tree-vect-patterns.c (vect_pattern_recog_1): Use
24127 VEC_ORDERED_REMOVE_IF.
24129 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24131 PR tree-optimization/82665
24132 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
24133 pointer subtraction where arguments come from a memchr call.
24135 2018-05-01 Jakub Jelinek <jakub@redhat.com>
24137 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
24138 --push-state --as-needed and --pop-state instead of --as-needed and
24139 --no-as-needed if ld supports it.
24140 * configure: Regenerated.
24143 * doc/install.texi2html: Replace _002d with - and _002a with * in
24144 generated html files using sed.
24146 2018-04-30 David Malcolm <dmalcolm@redhat.com>
24149 * gcc-rich-location.c (blank_line_before_p): New function.
24150 (use_new_line): New function.
24151 (gcc_rich_location::add_fixit_insert_formatted): New function.
24152 * gcc-rich-location.h
24153 (gcc_rich_location::add_fixit_insert_formatted): New function.
24155 2018-04-30 David Malcolm <dmalcolm@redhat.com>
24157 * selftest.c (assert_streq): Rename "expected" and "actual" to
24158 "val1" and "val2". Extend NULL-handling to cover both inputs
24159 symmetrically, while still requiring both to be non-NULL for a pass.
24160 * selftest.h (assert_streq): Rename "expected" and "actual" to
24162 (ASSERT_EQ): Likewise.
24163 (ASSERT_EQ_AT): Likewise.
24164 (ASSERT_KNOWN_EQ): Likewise.
24165 (ASSERT_KNOWN_EQ_AT): Likewise.
24166 (ASSERT_NE): Likewise.
24167 (ASSERT_MAYBE_NE): Likewise.
24168 (ASSERT_MAYBE_NE_AT): Likewise.
24169 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
24170 the assertion to pass.
24171 (ASSERT_STREQ_AT): Likewise.
24173 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
24175 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
24176 interaction with -pie.
24178 2018-04-30 David Malcolm <dmalcolm@redhat.com>
24180 * selftest.h: Fix alphabetization of per-source-file selftest
24183 2018-04-30 Jason Merrill <jason@redhat.com>
24185 PR c++/61982 - dead stores to destroyed objects.
24186 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
24189 2018-04-30 Jason Merrill <jason@redhat.com>
24191 * tree.c (build_clobber): New.
24192 * tree.h: Declare it.
24193 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
24195 2018-04-30 David Malcolm <dmalcolm@redhat.com>
24197 * diagnostic-show-locus.c (layout::layout): Update for
24198 location_get_source_line returning a char_span.
24199 (struct char_span): Move to input.h.
24200 (struct correction): Update for fields in char_span becoming
24202 (struct source_line): Update for location_get_source_line
24203 returning a char_span.
24204 (layout::print_line): Likewise.
24205 * edit-context.c (edited_file::print_content): Likewise.
24206 (edited_file::print_diff_hunk): Likewise.
24207 (edited_file::print_run_of_changed_lines): Likewise.
24208 (edited_file::get_num_lines): Likewise.
24209 (edited_line::edited_line): Likewise.
24210 * final.c (asm_show_source): Likewise.
24211 * input.c (location_get_source_line): Convert return type
24212 from const char * to char_span, losing the final "line_len"
24214 (dump_location_info): Update for the above.
24215 (get_substring_ranges_for_loc): Likewise. Use a char_span
24216 when handling the literal within the line.
24217 (test_reading_source_line): Update for location_get_source_line
24218 returning a char_span.
24219 * input.h (class char_span): Move here from
24220 diagnostic-show-locus.c, converting from a struct to a class.
24221 Make data members private.
24222 (char_span::operator bool): New.
24223 (char_span::length): New.
24224 (char_span::get_buffer): New.
24225 (char_span::operator[]): New.
24226 (char_span::subspan): Make const.
24227 (char_span::xstrdup): New.
24228 (location_get_source_line): Convert return type from const char *
24229 to char_span, losing the final "line_size" param.
24231 2018-04-30 Jan Hubicka <jh@suse.cz>
24233 * lto-wrapper.c (ltrans_priorities): New static var.
24234 (cmp_priority): New.
24235 (run_gcc): Read priorities and if doing parallel build order
24236 the Makefile by them.
24238 2018-04-30 David Malcolm <dmalcolm@redhat.com>
24240 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
24242 2018-04-30 Richard Biener <rguenther@suse.de>
24244 * tree-cfg.c (verify_address): Remove base argument, add
24245 flag whether to check TREE_ADDRESSABLE and do that.
24246 (verify_expr): Remove.
24247 (verify_types_in_gimple_reference): Add pieces from verify_expr.
24248 (verify_gimple_assign_single): Likewise.
24249 (verify_gimple_switch): Likewise.
24250 (verify_expr_location_1): Dereference tp once. Add (disabled)
24251 piece from verify_expr.
24252 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
24254 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
24256 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
24258 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
24260 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
24261 (small_data_pattern): Likewise.
24262 (arc_rewrite_small_data): Likewise.
24263 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
24264 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
24265 (get_symbol_alignment): New function.
24266 (legitimate_small_data_address_p): Likewise.
24267 (legitimate_scaled_address): Update, call
24268 legitimate_small_data_address_p.
24269 (output_sdata): New static variable.
24270 (arc_print_operand): Update how we handle small data operands.
24271 (arc_print_operand_address): Likewise.
24272 (arc_legitimate_address_p): Update, use
24273 legitimate_small_data_address_p.
24274 (arc_rewrite_small_data_p): Remove.
24275 (arc_rewrite_small_data_1): Likewise.
24276 (arc_rewrite_small_data): Likewise.
24277 (small_data_pattern): Likewise.
24278 (compact_sda_memory_operand): Update to use
24279 legitimate_small_data_address_p and get_symbol_alignment.
24280 (prepare_move_operands): Don't rewite sdata pattern.
24281 (prepare_extend_operands): Remove.
24282 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
24284 (zero_extendqisi2): Likewise.
24285 (zero_extendhisi2): Likewise.
24286 (extendqihi2): Likewise.
24287 (extendqisi2): Likewise.
24288 (extendhisi2): Likewise.
24289 (addsi3): Likewise.
24290 (subsi3): Likewise.
24291 (andsi3): Likewise.
24292 * config/arc/constraints.md (Usd): Change it to memory constraint.
24294 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
24296 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
24297 as source of std instructions.
24298 * config/arc/arc.md (movsi_insn): Update pattern predicate to
24299 allow 6-bit constants as source for store instructions.
24300 (movdi_insn): Update instruction pattern to allow 6-bit constants
24301 as source for store instructions.
24303 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
24305 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
24307 2018-04-30 Nathan Sidwell <nathan@acm.org>
24308 Sandra Loosemore <sandra@codesourcery.com>
24310 * dumpfile.c (dump_open): Allow '-' for stdout.
24311 * doc/invoke.texi (Developer Options): Document dump filename
24312 determination early. Document stdin/stdout selection.
24314 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
24316 Microblaze Target: PIC data text relative
24318 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
24319 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
24321 * config/microblaze/microblaze.h (microblaze_constant_address_p):
24322 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
24323 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
24324 New addressing mode for data-text relative position indepenedent code.
24325 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
24326 'ADDRESS_SYMBOLIC_TXT_REL'.
24327 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
24328 (microblaze_legitimate_pic_operand): Exclude function calls from
24329 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
24330 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
24332 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
24333 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
24334 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
24335 for 'address + offset'.
24336 (microblaze_expand_prologue): Add new function prologue call for
24338 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
24339 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
24340 table in case of TARGET_PIC_DATA_TEXT_REL.
24341 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
24342 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
24343 Add new macros 'UNSPEC_TEXT',
24344 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
24345 + exclude function calls from 'UNSPEC_PLT' in case of data text
24347 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
24348 new target hook for generating address diff vector tables in case of
24350 * doc/tm.texi : Regenerate.
24351 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
24352 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
24353 of addr diff vector generation.
24354 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
24355 target hook definition.
24356 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
24357 Add default function for generate_pic_addr_diff_vec -> flag_pic.
24358 * doc/invoke.texi (Add new pic option): Add new microblaze pic
24359 option for data text relative.
24361 2018-04-30 Richard Biener <rguenther@suse.de>
24363 * tree-chrec.h (evolution_function_is_constant_p): Remove
24365 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
24367 2018-04-30 Richard Biener <rguenther@suse.de>
24370 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
24372 2018-04-30 Richard Biener <rguenther@suse.de>
24374 PR tree-optimization/28364
24375 PR tree-optimization/85275
24376 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
24377 copying first exit test.
24379 2018-04-28 Mark Wielaard <mark@klomp.org>
24381 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
24382 dwarf_version >= 5.
24383 (dwarf_AT): Handle DW_AT_addr_base.
24384 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
24386 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
24389 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
24390 (*ashl<dwi>3_doubleword_mask_1): Ditto.
24391 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
24392 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
24394 2018-04-28 Richard Biener <rguenther@suse.de>
24396 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
24397 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
24398 to reflect use. Only add interesting stmts.
24400 2018-04-27 Martin Jambor <mjambor@suse.cz>
24403 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
24404 the jump function allows for passing through aggregate values.
24406 2018-04-27 David Malcolm <dmalcolm@redhat.com>
24408 * input.h (in_system_header_at): Convert from macro to inline
24410 (from_macro_expansion_at): Likewise.
24411 (from_macro_definition_at): Likewise.
24413 2018-04-27 Jeff Law <law@redhat.com>
24415 * config.gcc: Mark tile* targets as deprecated/obsolete.
24417 2018-04-27 Richard Biener <rguenther@suse.de>
24419 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
24422 2018-04-27 Richard Biener <rguenther@suse.de>
24424 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
24426 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
24428 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
24429 with Yd constraint. Set "preferred_for_speed" attribute from
24430 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
24431 with Yd constraint.
24432 (*movdi_internal): Ditto.
24433 (movti_interunit splitters): Remove
24434 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
24435 (movdi_interunit splitters): Ditto.
24436 * config/i386/constraints.md (Ye): Remove.
24437 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
24439 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24442 * config/aarch64/constraints.md (Usg): Limit to 31.
24443 (Usj): Limit to 63.
24445 2018-04-27 Jakub Jelinek <jakub@redhat.com>
24447 PR tree-optimization/85529
24448 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
24449 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
24450 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
24451 zero extension or masking of the MSB bit.
24452 (optimize_range_tests): Add FIRST_BB argument, pass it through
24453 to optimize_range_tests_var_bound.
24454 (maybe_optimize_range_tests, reassociate_bb): Adjust
24455 optimize_range_tests callers.
24457 2018-04-26 Richard Biener <rguenther@suse.de>
24458 Jakub Jelinek <jakub@redhat.com>
24460 * cgraph.h (symbol_table): Just declare debug method here.
24461 * symtab.c (symbol_table::debug): Define.
24463 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
24465 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
24467 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
24469 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
24470 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
24471 (*movdi_internal): Substitute Yi and Yj constraint with x
24472 and Ym and Yn constraint with y constraint. Update "isa"
24473 attribute and set "preferred_for_speed" attribute from
24474 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
24475 (*movsi_internal): Ditto.
24476 (*movdf_internal): Ditto.
24477 (*movsf_internal): Ditto.
24478 (*zero_extendsidi2): Ditto.
24479 * config/i386/sse.md (vec_set<mode>_0): Ditto.
24480 (sse2_loadld): Ditto.
24481 (*vec_extract<ssevecmodelower>_0): Ditto.
24482 (*vec_extractv4si_0_zext_sse4): Ditto.
24483 (vec_concatv2di): Ditto.
24484 (*vec_dup<mode>): Ditto.
24485 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
24486 * config/i386/constraints.md (Yi): Remove.
24491 2018-04-26 Nathan Sidwell <nathan@acm.org>
24493 * dumpfile.c (dump_open): New.
24494 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
24495 (dump_finish): Detect stdio/stderr by value not name.
24497 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
24499 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
24501 2018-04-26 Tom de Vries <tom@codesourcery.com>
24504 * config/nvptx/nvptx.c (verify_neutering_jumps)
24505 (verify_neutering_labels): New function
24506 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
24508 2018-04-26 Tom de Vries <tom@codesourcery.com>
24511 * config/nvptx/nvptx.c (needs_neutering_p): New function.
24512 (nvptx_single): Use needs_neutering_p to skip over insns that do not
24515 2018-04-26 Richard Biener <rguenther@suse.de>
24516 Tom de Vries <tom@codesourcery.com>
24519 * lto-streamer-out.c (output_function): Fixup loops if required to match
24520 discovery done in the reader.
24522 2018-04-26 Richard Biener <rguenther@suse.de>
24524 PR tree-optimization/85116
24525 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
24526 have a loop exit from the single latch predecessor. Remove
24527 case of header with just condition.
24528 (ch_base::copy_headers): Exclude infinite loops from any
24530 (pass_ch::execute): Record exits.
24532 2018-04-26 Richard Biener <rguenther@suse.de>
24534 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
24535 prologue cost vector and pass it to vect_get_load_cost.
24536 (vect_get_peeling_costs_all_drs): Likewise.
24537 (vect_peeling_hash_get_lowest_cost): Likewise.
24538 (vect_enhance_data_refs_alignment): Likewise.
24540 2018-04-26 Richard Biener <rguenther@suse.de>
24542 PR middle-end/85450
24543 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
24544 checking of integer<->pointer conversions.
24545 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
24546 sign-/zero-extending pointer types.
24547 (expand_omp_for_static_chunk): Likewise.
24549 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
24550 Jean Lee <xiaoyur347@gmail.com>
24552 * config/mips/mips.c (mips_asan_shadow_offset): New function.
24553 (TARGET_ASAN_SHADOW_OFFSET): Define.
24554 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
24555 true for -fsanitize=address.
24557 2018-04-25 Mark Wielaard <mark@klomp.org>
24559 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
24562 2018-04-25 Jakub Jelinek <jakub@redhat.com>
24564 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
24565 than "alu", remove explicit "memory" and "imm_disp" attributes.
24566 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
24568 PR middle-end/85414
24569 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
24570 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
24571 gen_lowpart_no_emit.
24573 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
24576 * config/i386/i386.c (ix86_expand_builtin): Change memory
24577 operand to XI, extend p0 to Pmode.
24578 * config/i386/i386.md: Change unspec volatile and operand
24579 1 mode to XI, change operand 0 mode to P.
24581 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
24583 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
24584 GET_MODE_MASK before any checking.
24585 (nds32_can_use_bset_p): Likewise.
24586 (nds32_can_use_btgl_p): Likewise.
24588 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
24590 * config/nds32/nds32-doubleword.md: New define_split pattern for
24591 illegal register number.
24593 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
24595 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
24597 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
24599 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
24601 2018-04-25 Richard Biener <rguenther@suse.de>
24603 * lto-streamer.h (LTO_major_version): Bump to 8.
24605 2018-04-25 Jakub Jelinek <jakub@redhat.com>
24607 * BASE-VER: Set to 9.0.0.
24609 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
24611 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
24612 in __abskf2 and __powikf2.
24614 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24617 * config/aarch64/constraints.md (Usg, Usj): New constraints.
24618 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
24619 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
24620 Use the above on operand 2. Reindent.
24621 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
24623 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
24626 * common/config/i386/i386-common.c (ix86_handle_option): Don't
24628 * config/i386/i386.opt (mcet): Removed.
24629 * doc/install.texi: Remove -mcet documentation.
24630 * doc/invoke.texi: Likewise.
24632 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
24635 * doc/install.texi: Remove -mcet from bootstrap-cet.
24637 2018-04-24 Jakub Jelinek <jakub@redhat.com>
24640 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
24641 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
24645 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
24646 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
24647 containing a CONST_VECTOR.
24649 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
24651 * doc/install.texi: Update newlib dependency for nvptx.
24653 2018-04-24 Jakub Jelinek <jakub@redhat.com>
24656 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
24657 instead of INTVAL when shifting x left.
24659 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
24661 PR tree-optimization/85478
24662 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
24663 vect_grouped_store_supported for single element vectors.
24665 2018-04-24 Richard Biener <rguenther@suse.de>
24668 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
24669 load cost increase to the case of non-constant step.
24671 2018-04-24 Jakub Jelinek <jakub@redhat.com>
24674 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
24675 destination if any_malformed_asm.
24677 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
24679 PR middle-end/85496
24680 * expr.c (store_field): In the bitfield case, if the value comes from
24681 a function call and is returned in registers by means of a PARALLEL,
24682 do not change the mode of the temporary unless BLKmode and VOIDmode.
24684 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
24686 PR rtl-optimization/85423
24687 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
24688 dependencies to debug insns when the previous insn is non-debug.
24690 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
24692 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
24693 enums into a single definition.
24694 (fls): Fix predicates and printing.
24697 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
24699 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
24700 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
24701 and short u6 immediate.
24702 (check_if_valid_sleep_operand): Remove.
24703 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
24705 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
24707 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
24708 flag_always_save_lp condition.
24709 * config/nds32/nds32.opt (malways-save-lp): New option.
24711 2018-04-22 Shiva Chen <shiva0217@gmail.com>
24713 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
24714 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
24715 * config/nds32/nds32.h
24716 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
24717 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
24719 2018-04-22 Shiva Chen <shiva0217@gmail.com>
24721 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
24722 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
24724 2018-04-22 Shiva Chen <shiva0217@gmail.com>
24725 Chung-Ju Wu <jasonwucj@gmail.com>
24727 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
24729 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
24730 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
24732 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
24734 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
24736 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
24738 * config/nds32/nds32-protos.h (nds32_data_alignment,
24739 nds32_local_alignment): Declare.
24740 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
24741 nds32_local_alignment): New functions.
24742 (TARGET_CONSTANT_ALIGNMENT): Define.
24743 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
24745 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
24747 * config/nds32/nds32.c
24748 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
24749 (TARGET_MODES_TIEABLE_P): Likewise.
24751 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
24753 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
24754 level Ofast and Og.
24756 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
24757 Chung-Ju Wu <jasonwucj@gmail.com>
24759 * config/nds32/constants.md (unspec_volatile_element): Add enum values
24760 for unaligned access.
24761 * config/nds32/nds32-intrinsic.c: Implementation of expanding
24763 * config/nds32/nds32-intrinsic.md: Likewise.
24764 * config/nds32/nds32_intrinsic.h: Likewise.
24765 * config/nds32/nds32.h (nds32_builtins): Likewise.
24766 * config/nds32/nds32.opt (munaligned-access): New option.
24767 * config/nds32/nds32.c (nds32_asm_file_start): Display
24768 flag_unaligned_access status.
24770 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
24772 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
24773 -mno-relax is present.
24774 * config/riscv/linux.h (LINK_SPEC): Ditto.
24776 2018-04-20 Martin Sebor <msebor@redhat.com>
24779 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
24781 (gimple_fold_builtin_stxcpy_chk): Same.
24782 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
24784 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
24787 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
24788 __powikf2 when long double is IEEE 128-bit.
24790 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
24792 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
24793 step to make sure stack always aligned.
24795 2018-04-20 Carl Love <cel@us.ibm.com>
24798 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
24799 size check for arg0.
24801 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
24802 Tom de Vries <tom@codesourcery.com>
24805 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
24806 Emit insns for calls too.
24807 (nvptx_find_par): Always look for worker-level predecessor insn.
24808 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
24810 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
24811 (nvptx_process_pars): Propagate frames for calls.
24813 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
24816 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
24818 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
24819 (ix86_handle_option): Don't handle OPT_mibt.
24820 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
24822 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
24824 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
24825 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
24826 (ix86_target_macros): Define __CET__ with flag_cf_protection
24827 for -fcf-protection.
24828 * config/i386/i386.c (isa2_opts): Remove -mibt.
24829 * config/i386/i386.h (TARGET_IBT): Removed.
24830 (TARGET_IBT_P): Likewise.
24831 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
24832 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
24833 * config/i386/i386.opt (mcet): Update help message.
24834 (mshstk): Likewise.
24836 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
24837 -mcet as an alias for -mshstk.
24839 2018-04-20 Richard Biener <rguenther@suse.de>
24841 PR middle-end/85475
24842 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
24843 complexity by forcing a single use of the multiply operand.
24845 2018-04-20 Martin Jambor <mjambor@suse.cz>
24848 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
24849 recursion dependency to only apply to non-clones.
24851 2018-04-20 Martin Jambor <mjambor@suse.cz>
24854 * ipa-cp.c (create_specialized_node): Check that clones of
24855 self-recursive edges exist during IPA-CP.
24857 2018-04-19 Toon Moene <toon@moene.org>
24859 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
24862 2018-04-19 Jakub Jelinek <jakub@redhat.com>
24864 PR tree-optimization/85467
24865 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
24866 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
24867 VECTOR_CST element to type.
24869 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
24872 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
24873 * config/i386/i386.md (builtin_setjmp_setup): Removed.
24874 (builtin_longjmp): Likewise.
24875 (save_stack_nonlocal): New pattern.
24876 (restore_stack_nonlocal): Likewise.
24878 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
24881 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
24882 Replace ASM_OUTPUT_LABEL with fprintf.
24884 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
24887 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
24888 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
24889 * config/i386/i386-c.c (ix86_target_macros_internal): Also
24890 define __IBT__ and __SHSTK__ for -fcf-protection.
24891 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
24893 (ix86_trampoline_init): Likewise.
24894 (x86_output_mi_thunk): Likewise.
24895 (ix86_notrack_prefixed_insn_p): Likewise.
24896 (ix86_option_override_internal): Don't disallow -fcf-protection.
24897 * config/i386/i386.md (rdssp<mode>): Also enable for
24899 (incssp<mode>): Likewise.
24900 (nop_endbr): Likewise.
24901 * config/i386/i386.opt (mcet): Change help message to built-in
24904 (mshstk): Likewise.
24905 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
24906 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
24907 enable CET built-in functions.
24909 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
24911 * common/config/i386/i386-common.c
24912 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
24913 OPTION_MASK_ISA_MOVDIRI_UNSET,
24914 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
24915 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
24916 * config.gcc (movdirintrin.h): New header.
24917 * config/i386/cpuid.h (bit_MOVDIRI,
24918 bit_MOVDIR64B): New bits.
24919 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
24921 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
24922 (VOID, PVOID, PCVOID)): New function types.
24923 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
24924 __builtin_ia32_directstoreu_u64,
24925 __builtin_ia32_movdir64b): New builtins.
24926 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
24927 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
24929 (ix86_valid_target_attribute_inner_p): Ditto.
24930 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
24931 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
24932 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
24933 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
24934 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
24935 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
24936 (movdiri<mode>, movdir64b_<mode>): New.
24937 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
24938 * config/i386/immintrin.h: Include movdirintrin.h.
24939 * config/i386/movdirintrin.h: New file.
24940 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
24942 2018-04-19 Richard Biener <rguenther@suse.de>
24944 PR middle-end/85455
24945 * cfg.c (clear_bb_flags): When loop state says we have
24946 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
24948 2018-04-19 Richard Biener <rguenther@suse.de>
24950 PR tree-optimization/84737
24951 * tree-vect-data-refs.c (vect_copy_ref_info): New function
24952 copying restrict info.
24953 (vect_setup_realignment): Use it.
24954 * tree-vectorizer.h (vect_copy_ref_info): Declare.
24955 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
24956 the first DR to all generated stores.
24957 (vectorizable_load): Likewise for loads.
24959 2018-04-19 Jakub Jelinek <jakub@redhat.com>
24961 PR tree-optimization/85446
24962 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
24963 the integral and pointer types to have the same precision.
24965 * doc/install.texi: Document --disable-cet being the default and
24968 2018-04-18 Martin Liska <mliska@suse.cz>
24970 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
24973 2018-04-18 Martin Liska <mliska@suse.cz>
24976 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
24979 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
24980 arguments if they are comparable.
24982 2018-04-18 Martin Liska <mliska@suse.cz>
24985 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
24988 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
24991 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
24994 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
24995 ENDBR after calling __morestack.
24997 2018-04-18 David Malcolm <dmalcolm@redhat.com>
25000 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
25001 by using gcc_base_ver to generate a gcc_driver_version, and use
25002 it when generating GCC_DRIVER_NAME.
25003 * configure: Regenerate.
25005 2018-04-18 Jakub Jelinek <jakub@redhat.com>
25008 * config.gcc: Obsolete powerpc*-*-*spe*.
25010 2018-04-17 Jakub Jelinek <jakub@redhat.com>
25013 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
25014 (stabstr_D): Change type of unum from unsigned int to
25015 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
25018 2018-04-17 Jim Wilson <jimw@sifive.com>
25021 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
25022 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
25023 Set arg_pointer_offset after using pretend_args_size.
25025 2018-04-17 Jakub Jelinek <jakub@redhat.com>
25027 PR rtl-optimization/85431
25028 * dse.c (record_store): Ignore zero width stores.
25031 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
25032 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
25033 __builtin_stack_restore rather than after it.
25034 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
25035 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
25036 argument instead of virtual_dynamic_stack_rtx.
25038 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
25040 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
25042 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
25043 Add note to error message to explain internal mapping of overloaded
25044 built-in function name to non-overloaded built-in function name.
25045 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
25048 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
25051 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
25052 where the inputs overlap with the output.
25054 2018-04-17 Jakub Jelinek <jakub@redhat.com>
25056 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
25057 (=v, v) alternative and explicit "memory" attribute.
25058 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
25059 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
25061 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
25062 "sselog1" type instead of "sselog".
25063 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
25064 "sselog". Remove explicit "memory" attribute.
25065 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
25066 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
25068 (vec_extract_hi_v32hi): Merge all alternatives into one, use
25069 "sselog1" type instead of "sselog". Remove explicit "memory"
25071 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
25072 use "sselog1" type instead of "sselog". Remove explicit "memory"
25074 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
25075 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
25077 (vec_extract_hi_v64qi): Merge all alternatives into one, use
25078 "sselog1" type instead of "sselog". Remove explicit "memory"
25080 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
25081 use "sselog1" type instead of "sselog". Remove explicit "memory"
25085 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
25087 PR middle-end/85414
25088 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
25091 2018-04-17 Martin Jambor <mjambor@suse.cz>
25094 * ipa-cp.c (create_specialized_node): Call
25095 expand_all_artificial_thunks if necessary.
25097 2018-04-17 Martin Liska <mliska@suse.cz>
25100 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
25101 in message, remote space in between '_G' and '('.
25103 2018-04-17 Jakub Jelinek <jakub@redhat.com>
25106 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
25107 avx512f_vmcmp<mode>3<round_saeonly_name>,
25108 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
25109 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
25110 avx512f_rndscale<mode><round_saeonly_name>,
25111 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
25112 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
25113 Use %<iptr>2 instead of %2 for -masm=intel.
25114 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
25115 avx512f_vcvttss2usi<round_saeonly_name>,
25116 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
25118 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
25119 avx512f_vcvttsd2usi<round_saeonly_name>,
25120 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
25121 Use %q1 instead of %1 for -masm=intel.
25122 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
25123 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
25124 of %3 for -masm=intel.
25125 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
25127 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
25129 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
25131 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
25132 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
25134 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
25136 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
25137 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
25138 %g1 and one with %0 and %1.
25139 (avx512er_vmrcp28<mode><round_saeonly_name>,
25140 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
25141 %1 for -masm=intel.
25142 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
25143 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
25144 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
25145 of %0 and %{%4%} for -masm=intel.
25146 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
25147 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
25148 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
25149 order of %0 and %{%5%}%{z%} for -masm=intel.
25151 2018-04-17 Jan Hubicka <jh@suse.cz>
25154 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
25156 2018-04-17 Martin Liska <mliska@suse.cz>
25159 * multiple_target.c (create_dispatcher_calls): Set apostrophes
25160 for target_clone error message. Make default implementation
25161 clone to be a local declaration.
25162 (separate_attrs): Add new argument and check for an empty
25164 (expand_target_clones): Handle it.
25165 (ipa_target_clone): Make redirection just for target_clones
25168 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
25169 Tom de Vries <tom@codesourcery.com>
25171 PR middle-end/84955
25172 * omp-expand.c (expand_oacc_for): Add dummy false branch for
25173 tiled basic blocks without omp continue statements.
25175 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
25178 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
25179 vec_extract expression as having side effects to make sure it gets
25182 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
25185 * config/i386/i386.c (get_builtin_code_for_version): Check
25188 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
25191 * config.gcc: Support "skylake".
25192 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
25194 * config/i386/i386.c (m_SKYLAKE): Define.
25195 (processor_target_table): Add "skylake".
25196 (ix86_option_override_internal): Add "skylake".
25197 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
25198 PROCESSOR_CANNONLAKE.
25199 (get_builtin_code_for_version): Fix priority for
25200 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
25201 PROCESSOR_SKYLAKE-AVX512.
25202 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
25203 (processor_type): Add PROCESSOR_SKYLAKE.
25205 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
25206 Jason Merrill <jason@redhat.com>
25209 * convert.c (convert_to_integer_1): Use direct recursion for
25210 enumeral types and types with a precision less than the number
25211 of bits in their mode.
25213 2018-04-16 Julia Koval <julia.koval@intel.com>
25216 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
25217 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
25219 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
25222 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
25223 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
25224 and -mno-direct-move.
25226 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
25229 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
25230 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
25231 Ensure negative shifts result in {0}.
25233 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
25235 PR rtl-optimization/79916
25236 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
25237 regs (if any) to define how to gnerate SD moves when LRA is in
25240 2018-04-13 Jakub Jelinek <jakub@redhat.com>
25242 PR rtl-optimization/85393
25243 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
25244 * except.c (expand_dw2_landing_pad_for_region): Make static.
25245 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
25246 a label and unconditional jump to old_bb, rather than
25247 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
25250 PR rtl-optimization/85376
25251 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
25252 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
25253 instead of a specific value.
25255 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
25256 Bin Cheng <bin.cheng@arm.com>
25258 PR tree-optimization/82965
25259 PR tree-optimization/83991
25260 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
25261 by_profile_only parameter.
25262 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
25263 information if the loop was predicted to iterate too many times.
25264 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
25266 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
25269 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
25272 2018-04-13 Martin Liska <mliska@suse.cz>
25273 Jakub Jelinek <jakub@redhat.com>
25275 PR middle-end/81657
25276 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
25277 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
25278 * builtins.c (expand_builtin_memory_copy_args): Use
25279 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
25280 handle dest_addr == pc_rtx.
25282 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
25285 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
25286 asked to not generate direct moves.
25287 (fix_trunc<mode>si2_stfiwx): Similar.
25288 (fix_trunc<mode>si2_internal): Similar.
25290 2018-04-12 Jakub Jelinek <jakub@redhat.com>
25293 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
25294 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
25295 lookup if dest in some wider mode is known to be const0_rtx and
25296 if so, record permanent equivalence for it to be ZERO_EXTEND of
25297 the narrower mode destination.
25299 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
25301 * lto-streamer-out.c (output_function): Revert 259346.
25302 * omp-expand.c (expand_oacc_for): Likewise.
25304 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
25306 PR rtl-optimization/85354
25307 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
25308 * sel-sched.c (sel_global_init): ... here.
25310 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
25313 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
25314 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
25315 mode for PE-COFF targets.
25316 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
25317 (i386_pe_asm_lto_end): Likewise.
25318 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
25319 (TARGET_ASM_LTO_END): Likewise.
25320 * config/i386/winnt.c (saved_debug_info_level): New static variable.
25321 (i386_pe_asm_lto_start): New function.
25322 (i386_pe_asm_lto_end): Likewise.
25324 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
25325 Richard Biener <rguenther@suse.de>
25327 PR middle-end/84955
25328 * lto-streamer-out.c (output_function): Fix CFG loop state before
25330 * omp-expand.c (expand_oacc_for): Handle calls to internal
25331 functions like regular functions.
25333 2018-04-12 Richard Biener <rguenther@suse.de>
25336 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
25337 for the early LTO debug to properly generate references to it
25338 during DIE emission. Do not re-use that for the skeleton for
25340 (dwarf2out_early_finish): Likewise.
25342 2018-04-12 Jakub Jelinek <jakub@redhat.com>
25345 * config/i386/sse.md
25346 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
25347 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
25348 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
25349 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
25350 and output is a reg, avoid creating invalid lowpart subreg, but
25351 instead split into a 512-bit move. Don't split if not AVX512VL,
25352 input is xmm16+ reg and output is a mem.
25353 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
25354 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
25355 xmm16+ reg and output is a mem.
25357 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
25359 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
25360 also for flag_dwarf2_cfi_asm.
25362 2018-04-12 Jakub Jelinek <jakub@redhat.com>
25364 PR rtl-optimization/85342
25365 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
25366 a bool scalar var inside of the loop instead. Don't try to update
25367 recog_data.operand after failed apply_change_group.
25369 2018-04-12 Tom de Vries <tom@codesourcery.com>
25372 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
25373 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
25374 array with flexible array member as array without given dimension.
25375 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
25376 argument for undefined param to true.
25378 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
25381 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
25382 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
25383 from PowerPC section.
25384 * config/rs6000/sysv4.opt (mcall-): Improve help text.
25385 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
25386 help text that is too long.
25387 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
25388 help text that is too long.
25389 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
25390 help text that is too long.
25392 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
25394 * config/alpha/alpha.md (stack_probe_internal): Rename
25395 from "probe_stack". Update all callers.
25397 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
25399 PR rtl-optimization/84566
25400 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
25401 sched_macro_fuse_insns.
25403 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
25406 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
25407 (compute_block_dependences): ... from here.
25409 2018-04-11 Jakub Jelinek <jakub@redhat.com>
25411 PR tree-optimization/85331
25412 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
25413 from int to HOST_WIDE_INT.
25415 2018-04-11 Martin Jambor <mjambor@suse.cz>
25418 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
25419 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
25420 not the same as the source val.
25421 (cgraph_edge_brings_value_p): New parameter.
25422 (gather_edges_for_value): Pass destination value to
25423 cgraph_edge_brings_value_p.
25424 (perhaps_add_new_callers): Likewise.
25425 (get_info_about_necessary_edges): Likewise and exclude values brought
25426 only by self-recursive edges.
25427 (create_specialized_node): Redirect only clones of self-calling edges.
25428 (+self_recursive_pass_through_p): New function.
25429 (find_more_scalar_values_for_callers_subset): Use it.
25430 (find_aggregate_values_for_callers_subset): Likewise.
25431 (known_aggs_to_agg_replacement_list): Removed.
25432 (decide_whether_version_node): Re-calculate known constants for all
25433 remaining context clones.
25435 2018-04-11 Richard Biener <rguenther@suse.de>
25438 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
25439 from early DWARF output.
25440 (dwarf2out_early_finish): Output line info unconditionally into
25441 early DWARF and add reference to it.
25443 2018-04-11 Jakub Jelinek <jakub@redhat.com>
25446 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
25447 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
25448 other than V2DFmode using iptr mode attribute.
25449 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
25451 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
25453 PR rtl-optimization/84659
25454 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
25456 2018-04-11 Jakub Jelinek <jakub@redhat.com>
25459 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
25461 (output_loc_list): Pass address of a dummy size variable even in the
25462 locview handling loop.
25463 (index_location_lists): Add comment on why skip_loc_list_entry can't
25466 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
25469 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
25472 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
25475 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
25476 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
25477 and -mstring-compare-inline-limit.
25479 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
25482 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
25483 for stack clash protection in a register whenever we need it to be in
25486 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
25488 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
25489 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
25491 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
25494 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
25496 (mlong-double-): Ditto.
25497 * config/rs6000/sysv4.opt (msdata=): Ditto.
25498 (mtls-size=): Ditto.
25500 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
25502 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
25503 erroneous entries for
25504 "vector int vec_ldl (int, long int *)", and
25505 "vector unsigned int vec_ldl (int, unsigned long int *)".
25506 Add comments and entries for
25507 "vector bool char vec_ldl (int, bool char *)",
25508 "vector bool short vec_ldl (int, bool short *)",
25509 "vector bool int vec_ldl (int, bool int *)",
25510 "vector bool long long vec_ldl (int, bool long long *)",
25511 "vector pixel vec_ldl (int, pixel *)",
25512 "vector long long vec_ldl (int, long long *)",
25513 "vector unsigned long long vec_ldl (int, unsigned long long *)".
25514 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
25515 type tree bool_long_long_type_node and correct definition of
25516 bool_V2DI_type_node to make reference to this new type tree.
25517 (rs6000_mangle_type): Replace erroneous reference to
25518 bool_long_type_node with bool_long_long_type_node.
25519 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
25520 comments to emphasize sign distinctions for char and int types and
25521 replace RS6000_BTI_bool_long constant with
25522 RS6000_BTI_bool_long_long constant. Also add comment to restrict
25523 use of RS6000_BTI_pixel.
25524 (bool_long_type_node): Remove this macro definition.
25525 (bool_long_long_type_node): New macro definition
25527 2018-04-10 Jakub Jelinek <jakub@redhat.com>
25529 PR rtl-optimization/85300
25530 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
25531 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
25532 simplify_unary_operation fails.
25534 2018-04-10 Martin Liska <mliska@suse.cz>
25536 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
25537 cgraph_edge and ipa_ref.
25539 2018-04-10 Jakub Jelinek <jakub@redhat.com>
25543 * config/i386/sse.md
25544 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
25545 computation of the VEC_MERGE selector from mask.
25546 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
25547 Fix decoding of the VEC_MERGE selector into mask.
25549 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
25551 PR tree-optimization/85286
25552 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
25554 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
25556 * final.c (final_1): Set insn_last_address as well as
25557 insn_current_address.
25559 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25562 * explow.c (emit_stack_probe): Call validize_mem on memory location
25563 before passing it to gen_probe_stack. Create address operand and
25564 legitimize it for the probe_stack_address case.
25566 2018-04-09 Jan Hubicka <jh@suse.cz>
25569 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
25570 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
25571 * tree.c (free_lang_data_in_type): Fix handling of binfos;
25573 (free_lang_data): Rebuild type inheritance graph.
25575 2018-04-09 Martin Sebor <msebor@redhat.com>
25577 * invoke.texi (-finline-small-functions): Mention other optimization
25579 (-findirect-inlining, -fpartial-inlining): Same.
25580 (-finline-functions-called-once): Same.
25581 (-freorder-blocks-and-partition): Same.
25583 2018-04-09 Jan Hubicka <jh@suse.cz>
25586 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
25587 jumps; choose last target that matches the criteria (i.e.
25588 no partition changes for non-crossing jumps).
25589 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
25590 support for redirecting crossing jumps to non-crossing.
25592 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
25594 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
25595 also for naked functions.
25597 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
25599 * config/arc/arc.md (add_shift): New pattern.
25600 (add_shift2): Likewise.
25601 (sub_shift): Likewise.
25602 (sub_shift_cmp0_noout): Likewise.
25603 (compare_si_ashiftsi): Likewise.
25604 (xbfu_cmp0_noout): New combine pattern.
25605 (xbfu_cmp0"): Likewise.
25606 (movsi_set_cc_insn): Place the predicable variant first.
25607 (commutative_binary_cmp0_noout): Remove clobber.
25608 (commutative_binary_cmp0): New pattern.
25609 (noncommutative_binary_cmp0): Likewise.
25610 (noncommutative_binary_cmp0_noout): Likewise.
25611 (noncommutative_binary_comparison_result_used): Removed.
25612 (rsub_cmp0): New pattern.
25613 (rsub_cmp0_noout): Likewise.
25614 (extzvsi): Changed, keep only meaningful variants.
25615 (SQH, SEZ): New iterators.
25616 (SQH_postfix): New mode attribute.
25617 (SEZ_prefix): New code attribute.
25618 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
25619 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
25620 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
25621 of numerical value.
25622 (noncommutative_operator): Check the availability of barrel
25625 2018-04-09 Richard Biener <rguenther@suse.de>
25627 PR tree-optimization/85284
25628 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
25629 Only use the niter constraining form of simple_iv when the exit
25630 is always executed.
25632 2018-04-09 Tom de Vries <tom@codesourcery.com>
25635 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
25636 (define_expand "*memory_barrier"): New define_expand.
25637 (define_insn "memory_barrier"): New insn.
25639 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
25641 PR rtl-optimization/80463
25642 PR rtl-optimization/83972
25643 PR rtl-optimization/83480
25645 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
25646 correct producer for the insn.
25647 (tidy_control_flow): Fixup seqnos in case of debug insns.
25649 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
25651 PR rtl-optimization/83913
25653 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
25654 different sched-times when merging exprs.
25656 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
25658 PR rtl-optimization/83962
25660 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
25661 tidy_fallthru_edge and tidy_control_flow.
25663 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
25665 PR rtl-optimization/83530
25667 * sel-sched.c (force_next_insn): New global variable.
25668 (remove_insn_for_debug): When force_next_insn is true, also leave only
25669 next insn in the ready list.
25670 (sel_sched_region): When the region wasn't scheduled, make another pass
25671 over it with force_next_insn set to 1.
25673 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
25675 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
25677 * config/nds32/constants.md (unspec_volatile_element): Add enum values
25678 for interrupt control.
25679 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
25680 functions for interrupt control.
25681 * config/nds32/nds32-intrinsic.md: Likewise.
25682 * config/nds32/nds32_intrinsic.h: Likewise.
25683 * config/nds32/nds32.h (nds32_builtins): Likewise.
25685 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
25687 * config/nds32/nds32.c (nds32_init_machine_status,
25688 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
25689 strict_aligned_p field.
25690 (nds32_expand_to_rtl_hook): New function.
25691 (TARGET_EXPAND_TO_RTL_HOOK): Define.
25692 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
25694 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
25695 Chung-Ju Wu <jasonwucj@gmail.com>
25697 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
25698 * config/nds32/nds32-n7.md: New file.
25699 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
25700 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
25702 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
25703 * config/nds32/nds32.md (pipeline_model): Add n7.
25704 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
25705 * config/nds32/pipelines.md: Include n7 settings.
25707 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
25708 Chung-Ju Wu <jasonwucj@gmail.com>
25710 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
25711 * config/nds32/nds32-e8.md: New file.
25712 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
25713 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
25715 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
25716 * config/nds32/nds32.md (pipeline_model): Add e8.
25717 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
25718 * config/nds32/pipelines.md: Include e8 settings.
25720 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
25721 Chung-Ju Wu <jasonwucj@gmail.com>
25723 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
25724 * config/nds32/nds32-n8.md: New file.
25725 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
25726 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
25728 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
25729 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
25730 * config/nds32/nds32.md (pipeline_model): Add n8.
25731 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
25732 * config/nds32/pipelines.md: Include n8 settings.
25734 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
25735 Chung-Ju Wu <jasonwucj@gmail.com>
25737 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
25738 * config/nds32/nds32-n9-2r1w.md: New file.
25739 * config/nds32/nds32-n9-3r2w.md: New file.
25740 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
25741 nds32_register_ports): New or modify for cpu n9.
25742 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
25744 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
25745 * config/nds32/nds32-utils.c: New file.
25746 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
25747 TARGET_MUL_SLOW): Define.
25748 * config/nds32/nds32.md (pipeline_model): New attribute.
25749 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
25750 New options that support cpu n9.
25751 * config/nds32/pipelines.md: Include n9 settings.
25752 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
25754 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
25756 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
25757 information if necessary.
25758 (output_cond_branch_compare_zero): Likewise.
25759 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
25760 (nds32_target_alignment): Refine for alignment.
25761 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
25762 (FUNCTION_BOUNDARY): Modify.
25763 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
25765 * config/nds32/nds32.opt (malways-align, malign-functions): New.
25767 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
25769 * config/nds32/constants.md (unspec_volatile_element): Add values for
25770 TLB operation and data prefetch.
25771 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
25772 functions for TLB operation and data prefetch.
25773 * config/nds32/nds32-intrinsic.md: Likewise.
25774 * config/nds32/nds32_intrinsic.h: Likewise.
25775 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
25776 (nds32_print_operand): Likewise.
25777 * config/nds32/nds32.h (nds32_builtins): Likewise.
25779 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
25780 Andrew Pinski <pinsika@gcc.gnu.org>
25782 PR middle-end/82976
25783 * match.pd: Use constant_boolean_node of correct type instead of
25784 boolean_true_node or boolean_false_node for simplifying
25785 pointer comparisons to zero.
25787 2018-04-07 Jakub Jelinek <jakub@redhat.com>
25789 PR tree-optimization/80021
25790 * tree.c (verify_type_variant): Make error call in verify_variant_match
25791 translatable and remove final full stop.
25793 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
25795 * config/nds32/constants.md (unspec_volatile_element): Add
25796 UNSPEC_VOLATILE_EH_RETURN.
25797 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
25798 nds32_output_stack_pop): Support dwarf exception handling process.
25799 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
25800 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
25801 exception handling process.
25802 (nds32_compute_stack_frame): Likewise.
25803 (nds32_return_addr_rtx): Likewise.
25804 (nds32_initial_elimination_offset): Likewise.
25805 (nds32_expand_prologue): Likewise.
25806 (nds32_expand_epilogue): Likewise.
25807 (nds32_dynamic_chain_address): New function.
25808 * config/nds32/nds32.h (machine_function): Add fields for dwarf
25809 exception handling.
25810 (DYNAMIC_CHAIN_ADDRESS): Define.
25811 (EH_RETURN_DATA_REGNO): Define.
25812 (EH_RETURN_STACKADJ_RTX): Define.
25813 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
25814 patterns for dwarf exception handling.
25816 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
25818 * config/nds32/nds32.h: Clean up obsolete macros.
25820 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
25822 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
25823 Add enum values for particular instructions.
25824 * config/nds32/nds32-intrinsic.c: Implementation of expanding
25825 particular intrinsic functions.
25826 * config/nds32/nds32-intrinsic.md: Likewise.
25827 * config/nds32/nds32_intrinsic.h: Likewise.
25828 * config/nds32/nds32.h (nds32_builtins): Likewise.
25829 * config/nds32/nds32.md (type): Add pbsad and pbsada.
25830 (btst, ave): New patterns for particular instructions.
25832 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
25834 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
25835 Add enum values for atomic load/store and memory sync.
25836 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
25838 * config/nds32/nds32-intrinsic.md: Likewise.
25839 * config/nds32/nds32_intrinsic.h: Likewise.
25840 * config/nds32/nds32.h (nds32_builtins): Likewise.
25842 2018-04-07 Jakub Jelinek <jakub@redhat.com>
25844 PR tree-optimization/85257
25845 * fold-const.c (native_encode_vector): If not all elts could fit
25846 and off is -1, return 0 rather than offset.
25847 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
25848 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
25849 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
25850 adjust buffer in native_interpret_expr call.
25852 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
25854 * config/nds32/constants.md (unspec_volatile_element): Add cache
25855 control enum values.
25856 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
25857 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
25858 * config/nds32/nds32.c (nds32_cctl_names): New.
25859 (nds32_print_operand): Handle cache control register names.
25860 * config/nds32/nds32.h (nds32_builtins): New enum values.
25861 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
25863 * config/nds32/nds32.md (type): Add mmu.
25864 * config/nds32/pipelines.md (simple_insn): Add mmu.
25866 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
25868 * config/nds32/nds32.md (type): Remove call.
25869 * config/nds32/pipelines.md (simple_insn): Likewise.
25871 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
25873 * config/nds32/constants.md (unspec_volatile_element): Add
25874 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
25875 UNSPEC_VOLATILE_FMFCFG.
25876 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
25877 description for fmfcfg and fmfcsr.
25878 (bdesc_1arg): Add fmtcsr.
25879 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
25880 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
25881 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
25882 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
25883 unspec_fmfcfg): New patterns.
25884 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
25885 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
25886 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
25887 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
25888 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
25889 __nds32__fmfcfg): Define.
25891 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
25893 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
25894 intrinsic register names.
25895 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
25896 intrinsic register enum values and macros.
25898 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
25900 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
25901 for load/store addressing form.
25902 (nds32_print_operand_address): Likewise.
25904 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
25907 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
25908 based on LABEL_REF. Remove useless assertion.
25909 (pic_address_needs_scratch): Fix formatting.
25910 (sparc_legitimize_pic_address): Minor tweaks.
25911 (sparc_delegitimize_address): Adjust assertion accordingly.
25912 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
25913 into symbolic_operand.
25914 (movsi_high_pic_label_ref): Likewise.
25915 (movsi_lo_sum_pic_label_ref): Likewise.
25916 (movdi_pic_label_ref): Likewise.
25917 (movdi_high_pic_label_ref): Likewise.
25918 (movdi_lo_sum_pic_label_ref): Likewise.
25920 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
25922 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
25923 custom LIB_SPEC setup.
25925 2018-04-06 Ruslan Bukin <br@bsdpad.com>
25926 Kito Cheng <kito.cheng@gmail.com>
25928 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
25929 * config/riscv/freebsd.h: New.
25931 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
25933 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
25934 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
25937 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
25938 Kito Cheng <kito.cheng@gmail.com>
25940 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
25941 nds32_output_call, nds32_symbol_binds_local_p): New functions.
25942 * config/nds32/nds32-protos.h (nds32_output_call,
25943 nds32_output_return): Declare.
25944 * config/nds32/nds32.md: Refine all the call and return patterns.
25946 2018-04-06 Jakub Jelinek <jakub@redhat.com>
25949 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
25950 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
25952 PR rtl-optimization/84872
25953 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
25954 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
25955 EDGE_CROSSING edge.
25957 2018-04-06 Tamar Christina <tamar.christina@arm.com>
25959 * expr.c (copy_blkmode_to_reg): Revert 254862.
25960 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
25962 2018-04-06 Richard Biener <rguenther@suse.de>
25964 PR middle-end/85244
25965 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
25966 after seeing a component reference with an adjacent field. Treat
25967 refs to arrays at struct end of external decls similar to
25968 refs to unconstrained commons.
25970 2018-04-06 Jakub Jelinek <jakub@redhat.com>
25973 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
25974 look through SAVE_EXPRs with non-side-effects argument. Adjust
25976 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
25979 2018-04-06 Richard Biener <rguenther@suse.de>
25981 PR middle-end/85180
25982 * alias.c (find_base_term): New wrapper around find_base_term
25983 unwinding CSELIB_VAL_PTR changes.
25984 (find_base_term): Do not restore CSELIB_VAL_PTR during the
25987 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
25989 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
25991 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
25992 constant definitions.
25993 ("nop"): lr 0,0 -> nopr r0
25994 ("nop_lr0", "nop_lr1"): New insn definitions.
25996 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
25998 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
25999 NDS32_V3PUSH_AVAILABLE_P macro.
26001 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
26002 Chung-Ju Wu <jasonwucj@gmail.com>
26004 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
26005 (nds32*-*-*): Add float and fpu_config into supported_defaults.
26006 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
26007 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
26008 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
26009 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
26010 * config/nds32/constraints.md: New constraints and checking for hard
26011 float configuration.
26012 * config/nds32/iterators.md: New mode iterator and attribute for hard
26013 float configuration.
26014 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
26016 * config/nds32/nds32-fpu.md: New file.
26017 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
26018 deal with hard float code generation.
26019 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
26021 (abi_type, float_reg_number): New enum type.
26022 * config/nds32/nds32-predicates.c: New predicates for hard float.
26023 * config/nds32/nds32-protos.h: Declare functions for hard float.
26024 * config/nds32/nds32.c: Implementation for hard float configuration.
26025 * config/nds32/nds32.h: Definitions for hard float configuration.
26026 * config/nds32/nds32.md: Include hard float machine description and
26027 modify patterns for hard float configuration.
26028 * config/nds32/nds32.opt: New options for hard float configuration.
26029 * config/nds32/predicates.md: New predicates for hard float
26032 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
26034 * common/config/nds32/nds32-common.c
26035 (nds32_option_optimization_table): Enable -mreleax-hint by default.
26037 2018-04-05 Jakub Jelinek <jakub@redhat.com>
26039 PR middle-end/85195
26040 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
26041 CONSTRUCTOR_ELT (ctor, ...)->value.
26043 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
26046 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
26048 2018-04-05 Tom de Vries <tom@codesourcery.com>
26051 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
26054 2018-04-05 Shiva Chen <shiva0217@gmail.com>
26055 Kito Cheng <kito.cheng@gmail.com>
26057 * config/nds32/constraints.md (U33): Fine-tune checking condition.
26058 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
26059 * config/nds32/nds32.h (nds32_16bit_address_type): Add
26060 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
26062 2018-04-05 Shiva Chen <shiva0217@gmail.com>
26063 Kito Cheng <kito.cheng@gmail.com>
26065 * config/nds32/constraints.md (Ufe): New memory constraint.
26066 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
26067 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
26068 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
26070 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
26071 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
26073 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
26075 * config/nds32/nds32.md: Use optimize_size in the condition for
26076 alu-shift instructions.
26078 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
26080 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
26082 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
26084 * config/nds32/nds32.md (negsi2): Refine pattern.
26086 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
26087 Chung-Ju Wu <jasonwucj@gmail.com>
26089 * config/nds32/iterators.md (shift_rotate): New code iterator.
26090 (shift): New code attribute.
26091 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
26092 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
26093 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
26094 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
26095 bit-wise operations.
26096 (andsi3, *andsi3): Ditto.
26097 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
26098 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
26099 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
26100 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
26101 nds32_ior_operand, nds32_xor_operand): New predicates.
26103 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
26105 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
26106 (addsi3, subsi3): ... this.
26108 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
26110 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
26112 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
26114 * config/nds32/nds32.md: Adjust indention.
26116 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
26118 * config/nds32/nds32.md (feature): New attribute.
26120 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
26122 * config/nds32/nds32.md (subtype): New attribute.
26124 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
26127 * config/arm/arm-builtins.c (arm_expand_builtin): Change
26128 expansion to perform a bitwise AND of the argument followed by a
26129 boolean negation of the result.
26131 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
26133 PR rtl-optimization/84878
26134 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
26135 the basic block. Assert the use reference is not artificial and that
26136 it has an associated insn.
26138 2018-04-04 Michael Matz <matz@suse.de>
26140 * builtins.c (compute_objsize): Pass correct operand
26141 to array_at_struct_end_p.
26143 2018-04-04 Richard Biener <rguenther@suse.de>
26146 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
26147 from contexts for DINFO_LEVEL_TERSE and below.
26149 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
26151 * config/nds32/nds32-doubleword.md (move_<mode>): Require
26152 resiter_operand condition.
26153 * config/nds32/nds32.md (*move<mode>): Ditto.
26155 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
26156 Monk Chiang <sh.chiang04@gmail.com>
26158 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
26160 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
26162 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
26164 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
26165 Kito Cheng <kito.cheng@gmail.com>
26167 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
26168 nds32_cond_code_str, output_cond_branch,
26169 output_cond_branch_compare_zero, nds32_expand_cbranch,
26170 nds32_expand_cstore, nds32_expand_movcc,
26171 nds32_output_cbranchsi4_equality_zero,
26172 nds32_output_cbranchsi4_equality_reg,
26173 nds32_output_cbranchsi4_equality_reg_or_const_int,
26174 nds32_output_cbranchsi4_greater_less_zero: New functions.
26175 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
26176 nds32_expand_cstore, nds32_expand_movcc,
26177 nds32_output_cbranchsi4_equality_zero,
26178 nds32_output_cbranchsi4_equality_reg,
26179 nds32_output_cbranchsi4_equality_reg_or_const_int,
26180 nds32_output_cbranchsi4_greater_less_zero): Declare.
26181 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
26182 nds32_rimm11s_operand): New predicates.
26183 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
26184 * config/nds32/nds32.md: Rewrite all the branch and conditional move
26187 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
26189 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
26190 * config/nds32/nds32.md: Ditto.
26191 * config/nds32/pipelines.md: Ditto.
26193 2018-04-04 Richard Biener <rguenther@suse.de>
26195 PR tree-optimization/85168
26196 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
26197 propagating abnormals.
26199 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
26201 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
26203 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
26204 Kito Cheng <kito.cheng@gmail.com>
26206 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
26207 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
26208 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
26209 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
26210 * config/nds32/nds32.md (sibcall_internal): New.
26211 (sibcall_register): Remove.
26212 (sibcall_immediate): Remove.
26213 (sibcall_value_internal): New.
26214 (sibcall_value_register): Remove.
26215 (sibcall_value_immediate): Remove.
26216 * config/nds32/predicates.md (nds32_general_register_operand): New.
26217 (nds32_call_address_operand): New.
26219 2018-04-03 Jakub Jelinek <jakub@redhat.com>
26221 PR rtl-optimization/85167
26222 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
26223 bb_defs if *split_p, instead preinitialize it to NULL.
26225 PR tree-optimization/85156
26226 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
26227 evaluating the argument multiple times.
26229 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
26231 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
26233 (_mm_cvtpd_ps): Likewise.
26234 (_mm_cvttpd_epi32): Likewise.
26235 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
26236 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
26237 vector, pixel, and bool following altivec.h include.
26239 2018-04-03 Martin Sebor <msebor@redhat.com>
26241 * doc/extend.texi (Common Function Attributes): Clarify.
26242 (const attribute): Likewise.
26243 (pure attribute): Likewise.
26245 2018-04-03 Jakub Jelinek <jakub@redhat.com>
26248 * config/i386/i386.c (ix86_expand_vector_set): Use
26249 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
26251 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
26253 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
26254 instructions when changing rounding bits to preserve precision bits
26255 in the x87 control word.
26257 2018-04-03 Martin Liska <mliska@suse.cz>
26259 PR tree-optimization/82491
26260 * rtl.h (strip_offset_and_add): Replace += suboffset with
26261 poly_uint64 () + suboffset.
26263 2018-03-29 Martin Liska <mliska@suse.cz>
26264 Martin Jambor <mjambor@suse.cz>
26267 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
26268 param_type is not an integral or pointer type.
26270 2018-04-03 Richard Biener <rguenther@suse.de>
26272 * sese.h (recompute_all_dominators): Remove.
26274 2018-04-02 Martin Sebor <msebor@redhat.com>
26276 * doc/invoke.texi (-Wrestrict): Fix typos.
26278 2018-04-02 Jim Wilson <jimw@sifive.com>
26280 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
26281 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
26282 (<optab>di3, <optab>si3_extend): Likewise.
26283 (<optab>si3_mask, <optab>si3_mask_1): New.
26284 (<optab>di3_mask, <optab>di3_mask_1): New.
26285 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
26286 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
26287 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
26289 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
26291 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
26294 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
26296 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
26297 (nds32_canonicalize_comparison): New function.
26299 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
26300 Kito Cheng <kito.cheng@gmail.com>
26301 Kuan-Lin Chen <kuanlinchentw@gmail.com>
26303 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
26304 * config/nds32/constants.md (unspec_volatile_element): Add
26305 UNSPEC_VOLATILE_RELAX_GROUP.
26306 * config/nds32/nds32-relax-opt.c: New file.
26307 * config/nds32/nds32-predicates.c
26308 (nds32_symbol_load_store_p): New function.
26309 * config/nds32/nds32-protos.h
26310 (nds32_symbol_load_store_p): Declare function.
26311 (make_pass_nds32_relax_opt): Declare new rtl pass function.
26312 * config/nds32/nds32.c
26313 (nds32_register_pass): New function to register pass.
26314 (nds32_register_passes): New function to register passes.
26315 * config/nds32/nds32.md (relax_group): New pattern.
26316 * config/nds32/nds32.opt (mrelax-hint): New option.
26317 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
26319 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
26321 * config/nds32/t-nds32: Modify files dependency.
26323 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
26325 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
26326 (PROFILE_HOOK): Define its implementation.
26328 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
26330 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
26331 type and 32-bit size.
26333 2018-04-01 Jakub Jelinek <jakub@redhat.com>
26335 PR middle-end/85090
26336 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
26337 (V_128_256): New mode iterator.
26338 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
26339 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
26340 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
26342 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
26343 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
26345 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
26348 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
26349 NaN inputs correctly.
26351 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
26354 * config/rs6000/vsx.md (??r): New mode attribute.
26355 (*vsx_mov<mode>_64bit): Use it.
26356 (*vsx_mov<mode>_32bit): Likewise.
26358 2018-03-30 Martin Sebor <msebor@redhat.com>
26360 PR tree-optimization/84818
26361 * builtins.c (check_access): Use warning_n.
26363 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
26366 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
26368 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
26371 2018-03-30 Julia Koval <julia.koval@intel.com>
26374 * x86-tune.def (movx, partial_reg_dependency): Enable for
26377 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
26379 PR inline-asm/84985
26380 * lra-constraints.c (process_alt_operands): Move setting
26381 this_alternative_matches below.
26383 2018-03-29 Martin Liska <mliska@suse.cz>
26386 * doc/invoke.texi: Document how LTO works with debug info.
26387 Describe auto-load support of binutils. Mention 'x86-64'
26388 as valid option value of -march option.
26390 2018-03-29 Jakub Jelinek <jakub@redhat.com>
26392 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
26395 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
26396 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
26397 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
26400 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
26403 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
26404 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
26405 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
26406 for RS6000_BTM_POWERPC64.
26407 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
26408 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
26409 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
26412 (DIVDEU): Likewise.
26414 2018-03-28 Carl Love <cel@us.ibm.com>
26417 2017-09-27 Carl Love <cel@us.ibm.com>
26419 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
26420 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
26421 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
26424 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
26426 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
26427 instead of __vector bool.
26428 (_mm_max_pu8): Likewise.
26429 (_mm_min_pi16): Likewise.
26431 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
26434 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
26435 (DIVWEUO): Likewise.
26436 (DIVDEO): Likewise.
26437 (DIVDEUO): Likewise.
26438 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
26439 DIVWEUO and DIVDEUO.
26440 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
26441 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
26442 (div_extend): Likewise.
26443 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
26445 (__builtin_divweuo): Likewise.
26446 (__builtin_divdeo): Likewise.
26447 (__builtin_divdeuo): Likewise.
26449 2018-03-28 Jakub Jelinek <jakub@redhat.com>
26452 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
26453 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
26455 PR tree-optimization/82004
26456 * gimple-match-head.c (optimize_pow_to_exp): New function.
26457 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
26458 Don't fold to exp if optimize_pow_to_exp is false.
26460 2018-03-28 Martin Liska <mliska@suse.cz>
26463 * calls.c (initialize_argument_information): Fix trailing space.
26464 * common.opt: Fix typo and provide better explanation for
26465 -fsanitize-coverage option.
26466 * config/i386/i386.opt: Fix typo.
26468 2018-03-28 Jakub Jelinek <jakub@redhat.com>
26469 Martin Liska <mliska@suse.cz>
26472 * gimplify.c (asan_poison_variable): Don't do the check for
26473 gimplify_omp_ctxp here.
26474 (gimplify_decl_expr): Do it here.
26475 (gimplify_target_expr): Likewise.
26477 2018-03-28 Martin Liska <mliska@suse.cz>
26480 * config/i386/i386.c (ix86_function_arg_advance): Do not call
26481 chkp_type_bounds_count if MPX is not enabled.
26483 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
26485 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
26487 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
26490 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
26491 function to create the function decl for complex long double
26492 multiply and divide for -mabi=ieeelongdouble.
26493 (init_float128_ieee): Call it.
26495 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
26498 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
26499 -fcf-protection=branch -mibt.
26500 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
26502 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
26505 * config/arm/arm.c (arm_valid_symbolic_address): Handle
26506 arm_word_relocations.
26508 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
26511 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
26512 extern array declarations.
26514 2018-03-27 Richard Biener <rguenther@suse.de>
26516 PR middle-end/84067
26517 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
26518 explicit single_use checks.
26520 2018-03-27 Richard Biener <rguenther@suse.de>
26522 PR tree-optimization/85082
26523 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
26526 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
26528 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
26529 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
26530 Turn on fasynchronous-unwind-tables and funwind-tables.
26532 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
26535 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
26536 (*bmi_blsr_<mode>_ccz): Ditto.
26538 2018-03-26 Tom de Vries <tom@codesourcery.com>
26540 PR tree-optimization/85063
26541 * omp-general.c (offloading_function_p): New function. Factor out
26543 * omp-offload.c (pass_omp_target_link::gate): ... here.
26544 * omp-general.h (offloading_function_p): Declare.
26545 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
26546 with attribute omp declare target for offloading functions.
26548 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
26550 PR tree-optimization/84005
26551 * tree-data-ref.h (get_base_for_alignment): Declare.
26552 * tree-data-ref.c (get_base_for_alignment_1): New function.
26553 (get_base_for_alignment): Likewise.
26554 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
26555 get_base_for_alignment to find a suitable base object, instead
26556 of always using drb->base_address.
26558 2018-03-23 Jakub Jelinek <jakub@redhat.com>
26560 PR inline-asm/85022
26561 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
26562 known size by default.
26564 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
26566 PR inline-asm/85030
26567 * lra-constraints.c (process_alt_operands): Don't match BLKmode
26568 and non BLKmode operands.
26570 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26573 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
26574 Clean up attributes.
26576 2018-03-23 Richard Biener <rguenther@suse.de>
26579 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
26580 we are going to emit early debug for LTO.
26582 2018-03-23 Jakub Jelinek <jakub@redhat.com>
26584 PR inline-asm/85034
26585 * function.c (match_asm_constraints_1): Don't optimize if input
26586 doesn't satisfy general_operand predicate for output's mode.
26588 PR inline-asm/85022
26589 * alias.c (write_dependence_p): Don't require for x_canonicalized
26590 non-VOIDmode if x has VOIDmode.
26593 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
26594 just don't try to optimize it rather than assert it never happens.
26596 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
26598 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
26599 macro expansions for definition of ST_INTERNAL_<mode> and
26600 LD_INTERNAL_<mode> builtins.
26601 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
26603 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
26605 (altivec_expand_st_builtin): Likewise.
26606 (altivec_expand_builtin): Remove calls to deleted functions.
26607 (rs6000_address_for_altivec): Delete this function.
26608 * config/rs6000/vector.md: Remove expands for
26609 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
26611 2018-03-22 Sudakshina Das <sudi.das@arm.com>
26614 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
26615 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
26616 re-computing once computed.
26617 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
26618 (arm_init_machine_status): Initialize
26619 machine->static_chain_stack_bytes.
26621 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
26624 * doc/extend.texi: Add four new prototypes for vec_ld.
26625 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
26626 definitions for more logical presentation.
26627 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
26628 entries for V1TI variants of __builtin_altivec_ld builtin.
26629 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
26630 handling of V1TI variant of LVX icode pattern.
26631 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
26632 (rs6000_gimple_fold_builtin): Likewise.
26633 (altivec_init_builtins): Add code to define
26634 __builtin_altivec_lvx_v1ti function.
26636 2018-03-22 Jakub Jelinek <jakub@redhat.com>
26638 PR inline-asm/84941
26639 * function.c (match_asm_constraints_1): Don't do the optimization
26640 if input isn't a REG, SUBREG, MEM or constant.
26642 2018-03-22 Tom de Vries <tom@codesourcery.com>
26644 PR tree-optimization/84956
26645 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
26646 bb_has_abnormal_pred.
26648 2018-03-22 Jakub Jelinek <jakub@redhat.com>
26651 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
26652 DECL_INITIAL (decl) to decl at the end.
26653 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
26654 adjust the comment.
26656 2018-03-21 Joseph Myers <joseph@codesourcery.com>
26658 * doc/extend.texi (__builtin_tgmath): Document when complex
26659 integer types are treated as _Complex _Float64.
26661 2018-03-21 Tom de Vries <tom@codesourcery.com>
26663 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
26665 2018-03-21 Jakub Jelinek <jakub@redhat.com>
26667 PR tree-optimization/84960
26668 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
26669 if it is ENTRY block, move them into single succ of ENTRY in that case.
26671 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
26673 PR tree-optimization/84811
26674 * poly-int.h (poly_span_traits): Remove the T3 parameter and
26675 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
26676 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
26677 (known_subrange_p): Update accordingly. Cast each value involved
26678 in the size comparison, rather than casting the result of the
26681 2018-03-21 Jakub Jelinek <jakub@redhat.com>
26683 PR tree-optimization/84982
26684 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
26685 by flipping the least significant bit rather than all bits from
26686 bitpos to bitpos + bitsize - 1.
26688 2018-03-21 Nathan Sidwell <nathan@acm.org>
26690 * doc/extend.texi (Deprecated Features): Remove mention of
26691 long-deleted deprecations.
26693 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
26696 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
26697 * configure: Regenerate.
26699 2018-03-21 Tom de Vries <tom@codesourcery.com>
26701 PR tree-optimization/83126
26702 * tree-parloops.c (num_phis): New function.
26703 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
26705 2018-03-21 Nathan Sidwell <nathan@acm.org>
26707 * doc/extend.texi (Deprecated Features): Update deprecated flags,
26708 mention anon-struct/union members and trailing attributes.
26710 2018-03-21 Bin Cheng <bin.cheng@arm.com>
26712 PR tree-optimization/84969
26713 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
26714 builtin memset partitions if they set different rhs values.
26716 2018-03-21 Jakub Jelinek <jakub@redhat.com>
26718 PR rtl-optimization/84989
26719 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
26720 VEC_DUPLICATE with scalar result mode.
26722 2018-03-21 Martin Liska <mliska@suse.cz>
26725 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
26726 not intended return statement.
26728 2018-03-21 Martin Liska <mliska@suse.cz>
26731 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
26732 (chkp_find_bound_slots_1): Limit number of iterations.
26734 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
26737 * Minor grammar fixes for x86 options.
26739 2018-03-20 Jakub Jelinek <jakub@redhat.com>
26742 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
26743 holding REG_CFA_RESTORE notes, instead turn them into a USE.
26745 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
26748 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
26749 (altivec_lvx_<mode>_1op): Likewise.
26750 (altivec_stvx_<mode>_2op): Likewise.
26751 (altivec_stvx_<mode>_1op): Likewise.
26752 (altivec_lvx_<VM2:mode>): New define_expand.
26753 (altivec_stvx_<VM2:mode>): Likewise.
26754 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
26755 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
26756 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
26757 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
26758 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
26759 (rs6000_gen_lvx): Likewise.
26760 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
26761 (altivec_expand_stv_builtin): Likewise.
26762 (altivec_expand_builtin): Likewise.
26763 * config/rs6000/vector.md: Likewise.
26765 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26768 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
26771 2018-03-20 Richard Biener <rguenther@suse.de>
26774 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
26775 sign-conversions as zero, fall back to standard scalar_stmt
26778 2018-03-20 Martin Liska <mliska@suse.cz>
26781 * predict.c (rebuild_frequencies): Handle case when we have
26782 PROFILE_ABSENT, but flag_guess_branch_prob is false.
26784 2018-03-20 Jakub Jelinek <jakub@redhat.com>
26787 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
26788 flag_section_anchors.
26789 * varasm.c (use_blocks_for_decl_p): Remove hack for
26790 dw2_force_const_mem.
26793 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
26795 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
26796 be created, use lowpart_subreg of operands[0] rather than operands[0]
26798 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
26799 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
26800 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
26801 and n constraint instead of aarch64_shift_imm_di and Usd.
26802 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
26803 (*aarch64_<optab>_reg_minus<mode>3): ... this.
26805 2018-03-20 Sudakshina Das <sudi.das@arm.com>
26808 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
26809 to favor GPR over NEON registers.
26810 (<shift>di3_neon): Likewise.
26812 2018-03-20 Tom de Vries <tom@codesourcery.com>
26815 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
26816 (nvptx_process_pars): Emit bar.sync asap and alap.
26818 2018-03-20 Tom de Vries <tom@codesourcery.com>
26821 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
26822 seen_label if seen_label is already set.
26824 2018-03-20 Jakub Jelinek <jakub@redhat.com>
26827 * config/i386/i386.c (fold_builtin_cpu): For features above 31
26828 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
26829 Use 1U instead of 1. Formatting fixes.
26832 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
26833 instead of TREE_TYPE (s1) for the return value.
26835 2018-03-19 Jakub Jelinek <jakub@redhat.com>
26837 PR tree-optimization/84946
26838 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
26839 bitsize + bitsize in poly_uint64 rather than poly_int64.
26842 * dwarf2asm.c: Include fold-const.c.
26843 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
26844 of decl rather than decl itself.
26846 PR rtl-optimization/84643
26847 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
26849 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
26852 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
26853 calling assemble_variable.
26855 2018-03-19 Sudakshina Das <sudi.das@arm.com>
26858 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
26859 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
26861 2018-03-19 Jim Wilson <jimw@sifive.com>
26864 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
26865 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
26866 (riscv_first_stack_step): Likewise.
26867 (riscv_option_override): Use STACK_BOUNDARY instead of
26868 MIN_STACK_BOUNDARY.
26869 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
26870 MIN_STACK_BOUNDARY.
26871 (BIGGEST_ALIGNMENT): Set to 128.
26872 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
26873 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
26876 2018-03-19 Richard Biener <rguenther@suse.de>
26878 PR tree-optimization/84933
26879 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
26880 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
26882 2018-03-19 Richard Biener <rguenther@suse.de>
26884 PR tree-optimization/84859
26885 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
26886 (cond_if_else_store_replacement): Perform sinking operation on
26887 single-store BBs regardless of MAX_STORES_TO_SINK setting.
26888 Generalize what a BB with a single eligible store is.
26890 2018-03-19 Richard Biener <rguenther@suse.de>
26892 PR tree-optimization/84929
26893 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
26894 chrec_is_positive against non-chrec arg.
26896 2018-03-19 Tamar Christina <tamar.christina@arm.com>
26899 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
26901 2018-03-18 Martin Liska <mliska@suse.cz>
26903 PR rtl-optimization/84635
26904 * regrename.c (build_def_use): Use matches_mode only when
26907 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
26909 PR tree-optimization/84913
26910 * tree-vect-loop.c (vectorizable_reduction): Don't try to
26911 vectorize chains of COND_EXPRs.
26913 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
26915 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
26917 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
26919 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
26921 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
26923 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
26925 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
26926 Kito Cheng <kito.cheng@gmail.com>
26928 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
26929 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
26930 (nds32_adjust_reg_alloc_order): New function.
26931 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
26933 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
26935 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
26936 nds32_print_operand, nds32_print_operand_address): Use
26937 HOST_WIDE_INT_PRINT_DEC instead.
26939 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
26941 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
26943 2018-03-17 Jakub Jelinek <jakub@redhat.com>
26946 * config/i386/i386.c (initial_ix86_tune_features,
26947 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
26948 unsigned long long.
26949 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
26950 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
26951 rather than 1u << ix86_tune. Formatting fix.
26952 (ix86_option_override_internal): Change ix86_arch_mask from
26953 unsigned int to unsigned HOST_WIDE_INT, initialize to
26954 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
26955 (ix86_function_specific_restore): Likewise.
26957 2018-03-16 Jakub Jelinek <jakub@redhat.com>
26960 * postreload.c (reload_combine_recognize_pattern): Perform
26961 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
26962 truncate_int_for_mode the result for the destination's mode.
26965 * hsa-gen.c (mem_type_for_type): Fix comment typo.
26966 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
26968 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
26971 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
26974 * lra-assigns.c (lra_split_hard_reg_for): Don't use
26975 regno_allocno_class_array and sorted_pseudos.
26976 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
26977 insns where regno is used.
26979 2018-03-16 Martin Liska <mliska@suse.cz>
26982 * multiple_target.c (create_dispatcher_calls): Redirect
26983 reference in the symbol table.
26985 2018-03-16 Martin Liska <mliska@suse.cz>
26988 * multiple_target.c (create_dispatcher_calls): Redirect also
26991 2018-03-16 Jakub Jelinek <jakub@redhat.com>
26995 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
26996 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
26997 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
26999 2018-03-16 Julia Koval <julia.koval@intel.com>
27001 * doc/invoke.texi (Skylake Server): Add CLWB.
27002 Cannonlake): Remove CLWB.
27004 2018-03-16 Jakub Jelinek <jakub@redhat.com>
27006 PR tree-optimization/84841
27007 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
27009 (FLOAT_ONE_CONST_TYPE): Define.
27010 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
27011 (sort_by_operand_rank): Put entries with higher constant_type last
27012 rather than first to match comments.
27014 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
27016 * config/nios2/nios2.md (movsi_internal): Fix thinko in
27019 2018-03-15 Jakub Jelinek <jakub@redhat.com>
27022 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
27023 check and use address of target always.
27025 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
27028 * config/i386/i386.c (indirect_thunk_needed): Update comments.
27029 (indirect_thunk_bnd_needed): Likewise.
27030 (indirect_thunks_used): Likewise.
27031 (indirect_thunks_bnd_used): Likewise.
27032 (indirect_return_needed): New.
27033 (indirect_return_bnd_needed): Likewise.
27034 (output_indirect_thunk_function): Add a bool argument for
27036 (output_indirect_thunk_function): Don't generate alias for
27037 function return thunk.
27038 (ix86_code_end): Call output_indirect_thunk_function to generate
27039 function return thunks.
27040 (ix86_output_function_return): Set indirect_return_bnd_needed
27041 and indirect_return_needed instead of indirect_thunk_bnd_needed
27042 and indirect_thunk_needed.
27044 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
27046 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
27047 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
27048 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
27050 2018-03-15 David Malcolm <dmalcolm@redhat.com>
27051 Paul Hua <paul.hua.gm@gmail.com>
27054 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
27056 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
27058 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
27059 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
27060 resp. SFmode cases.
27062 2018-03-15 Tamar Christina <tamar.christina@arm.com>
27065 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
27066 instead of GET_MODE_SIZE when comparing Units.
27068 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
27071 * varasm.c (hash_section): Return an unchangeble hash value
27072 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
27073 Return !aarch64_can_use_per_function_literal_pools_p ().
27075 2018-03-15 Jakub Jelinek <jakub@redhat.com>
27078 * optabs.c (emit_conditional_move): Pass address of cmode's copy
27079 rather than address of cmode as last argument to prepare_cmp_insn.
27081 2018-03-15 Julia Koval <julia.koval@intel.com>
27083 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
27084 F_AVX512VNNI, F_AVX512BITALG): New.
27086 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
27089 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
27090 insn for floating-point loads and stores.
27092 2018-03-14 Carl Love <cel@us.ibm.com>
27094 * config/rs6000/rs6000-c.c: Add macro definitions for
27095 ALTIVEC_BUILTIN_VEC_PERMXOR.
27096 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
27097 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
27098 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
27099 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
27101 * config/doc/extend.texi: Add prototypes for vec_permxor.
27103 2018-03-14 David Malcolm <dmalcolm@redhat.com>
27106 * diagnostic-show-locus.c (class layout_point): Convert m_line
27107 from int to linenum_type.
27108 (line_span::comparator): Use linenum "compare" function when
27109 comparing line numbers.
27110 (test_line_span): New function.
27111 (layout_range::contains_point): Convert param "row" from int to
27113 (layout_range::intersects_line_p): Likewise.
27114 (layout::will_show_line_p): Likewise.
27115 (layout::print_source_line): Likewise.
27116 (layout::should_print_annotation_line_p): Likewise.
27117 (layout::print_annotation_line): Likewise.
27118 (layout::print_leading_fixits): Likewise.
27119 (layout::annotation_line_showed_range_p): Likewise.
27120 (struct line_corrections): Likewise for field m_row.
27121 (line_corrections::line_corrections): Likewise for param "row".
27122 (layout::print_trailing_fixits): Likewise.
27123 (layout::get_state_at_point): Likewise.
27124 (layout::get_x_bound_for_row): Likewise.
27125 (layout::print_line): Likewise.
27126 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
27127 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
27128 * input.c (selftest::test_linenum_comparisons): New function.
27129 (selftest::input_c_tests): Call it.
27130 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
27131 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
27132 * selftest.h (ASSERT_GT): New macro.
27133 (ASSERT_GT_AT): New macro.
27134 (ASSERT_LT): New macro.
27135 (ASSERT_LT_AT): New macro.
27137 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
27139 PR rtl-optimization/84780
27140 * combine.c (distribute_links): Don't make a link based on pc_rtx.
27142 2018-03-14 Martin Liska <mliska@suse.cz>
27144 * tree.c (record_node_allocation_statistics): Use
27145 get_stats_node_kind.
27146 (get_stats_node_kind): New function extracted from
27147 record_node_allocation_statistics.
27148 (free_node): Use get_stats_node_kind.
27150 2018-03-14 Richard Biener <rguenther@suse.de>
27152 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
27153 that the value-set of ANTIC_IN doesn't grow.
27156 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
27158 (BB_VISITED_WITH_VISITED_SUCCS): New define.
27159 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
27161 2018-03-14 Julia Koval <julia.koval@intel.com>
27163 * config.gcc (icelake-client, icelake-server): New.
27165 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
27166 (initial_ix86_arch_features): Ditto.
27167 (PTA_SKYLAKE): Add SGX.
27168 (PTA_ICELAKE): Remove.
27169 (PTA_ICELAKE_CLIENT): New.
27170 (PTA_ICELAKE_SERVER): New.
27171 (ix86_option_override_internal): Split up icelake on icelake client and
27173 (get_builtin_code_for_version): Ditto.
27174 (fold_builtin_cpu): Ditto.
27175 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
27176 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
27177 * config/i386/i386.h (processor_type): Ditto.
27178 * doc/invoke.texi: Ditto.
27180 2018-03-14 Jakub Jelinek <jakub@redhat.com>
27183 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
27184 INTEGER_CST offset, add it together with bitpos / 8 and
27185 sign extend based on POINTER_SIZE.
27189 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
27192 * config/i386/constraints.md (Yc): New register constraint.
27193 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
27194 Use Yc constraint for alternative 2 of operand 0. Remove
27195 preferred_for_speed attribute.
27197 2018-03-14 Richard Biener <rguenther@suse.de>
27199 PR tree-optimization/84830
27200 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
27201 with the old one to avoid oscillations.
27203 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
27206 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
27208 (assign_by_spills): Return a flag of reload assignment failure.
27209 Do not process the reload assignment failures. Do not spill other
27210 reload pseudos if they has the same reg class. Update n if
27212 (lra_assign): Add a return arg. Set up from the result of
27213 assign_by_spills call.
27214 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
27215 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
27216 usage_insns if it is not NULL.
27217 (spill_hard_reg_in_range): New function.
27218 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
27219 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
27220 function prototypes.
27221 (lra_assign): Change prototype.
27222 * lra.c (lra): Add code to deal with fails by splitting hard reg
27225 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
27227 * config/riscv/riscv.opt (mrelax): New option.
27228 * config/riscv/riscv.c (riscv_file_start): Emit ".option
27229 "norelax" when riscv_mrelax is disabled.
27230 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
27232 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
27235 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
27236 reassociation for int modes.
27238 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
27240 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
27241 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
27243 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
27244 * config/aarch64/aarch64-sve.md
27245 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
27246 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
27247 (*extend<mode><Vwide>2): Rename to...
27248 (aarch64_sve_extend<mode><Vwide>2): ...this.
27249 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
27250 renaming the old pattern to...
27251 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
27253 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
27254 define_expand, renaming the old pattern to...
27255 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
27256 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
27257 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
27258 account when deciding which SVE instruction the optab should use.
27259 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
27261 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
27263 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
27264 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
27265 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
27266 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
27267 (tlsdesc_small_<mode>): Turn a define_expand and use
27268 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
27269 (tlsdesc_small_advsimd_<mode>): ...this.
27270 (tlsdesc_small_sve_<mode>): New pattern.
27272 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
27274 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
27275 (UNSPEC_UMUL_HIGHPART): New constants.
27276 (MUL_HIGHPART): New int iteraor.
27277 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
27278 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
27280 (*<su>mul<mode>3_highpart): New define_insn.
27282 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
27285 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
27288 2018-03-13 Martin Liska <mliska@suse.cz>
27291 * (sem_item_optimizer::sem_item_optimizer): Initialize new
27293 (sem_item_optimizer::~sem_item_optimizer): Release it.
27294 (sem_item_optimizer::merge_classes): Register variable aliases.
27295 (sem_item_optimizer::fixup_pt_set): New function.
27296 (sem_item_optimizer::fixup_points_to_sets): Likewise.
27297 * ipa-icf.h: Declare new variables and functions.
27299 2018-03-13 Jakub Jelinek <jakub@redhat.com>
27301 PR middle-end/84834
27302 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
27303 integer_pow2p@2 and test integer_pow2p in condition.
27304 (A < 0 ? C : 0): Similarly for @1.
27306 PR middle-end/84831
27307 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
27308 characters starting at p contain '\0' character, don't look beyond
27312 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
27313 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
27316 * reg-stack.c (change_stack): Change update_end var from int to
27317 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
27318 also call set_block_for_insn on the newly added insns and rescan.
27321 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
27322 on the last operand.
27325 * tree.c (stabilize_reference_1): Return save_expr (e) for
27326 STATEMENT_LIST even if it doesn't have side-effects.
27328 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
27330 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
27332 2018-03-12 Renlin Li <renlin.li@arm.com>
27334 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
27335 aarch64_output_scalar_simd_mov_immediate.
27337 2018-03-12 Martin Sebor <msebor@redhat.com>
27339 PR tree-optimization/83456
27340 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
27341 for perfectly overlapping calls to memcpy.
27342 (gimple_fold_builtin_memory_chk): Same.
27343 (gimple_fold_builtin_strcpy): Handle no-warning.
27344 (gimple_fold_builtin_stxcpy_chk): Same.
27345 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
27347 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
27349 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
27350 parameter. Use it for SFmode.
27351 (rs6000_function_arg_advance_1): Adjust.
27352 (rs6000_function_arg): Adjust.
27353 (rs6000_gimplify_va_arg): Pass false for that new parameter.
27355 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
27357 PR rtl-optimization/84169
27358 PR rtl-optimization/84780
27359 * combine.c (can_combine_p): Check for a 2-insn combination whether
27360 the destination register is used between the two insns, too.
27362 2018-03-12 Richard Biener <rguenther@suse.de>
27364 PR tree-optimization/84803
27365 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
27366 for refs DR analysis didn't process.
27368 2018-03-12 Richard Biener <rguenther@suse.de>
27370 PR tree-optimization/84777
27371 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
27372 force-vectorize loops ignore whether we are optimizing for size.
27374 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
27376 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
27377 (TARGET_MD_ASM_ADJUST): Define.
27379 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
27380 Kito Cheng <kito.cheng@gmail.com>
27381 Chung-Ju Wu <jasonwucj@gmail.com>
27383 * config/nds32/nds32.c (nds32_compute_stack_frame,
27384 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
27385 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
27386 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
27387 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
27388 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
27389 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
27390 * config/nds32/nds32.md (prologue, epilogue): Use macro
27391 NDS32_V3PUSH_AVAILABLE_P to do checking.
27393 2018-03-11 Jakub Jelinek <jakub@redhat.com>
27396 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
27397 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
27398 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
27399 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
27400 addition of most attributes on !orig_type_die or the attribute not
27401 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
27403 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
27404 Chung-Ju Wu <jasonwucj@gmail.com>
27406 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
27407 __NDS32_VH__ macro.
27408 * config/nds32/nds32.opt (mvh): New option.
27410 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
27411 Chung-Ju Wu <jasonwucj@gmail.com>
27413 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
27415 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
27416 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
27419 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
27420 Chung-Ju Wu <jasonwucj@gmail.com>
27422 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
27424 * config/nds32/nds32-multiple.md (strlensi): New pattern.
27425 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
27427 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
27428 Kito Cheng <kito.cheng@gmail.com>
27429 Chung-Ju Wu <jasonwucj@gmail.com>
27431 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
27432 UNSPEC_FFMISM and UNSPEC_FLMISM.
27433 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
27434 for ffb, ffmism and flmism.
27435 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
27436 (unspec_ffmism): Ditto.
27437 (unspec_flmism): Ditto.
27438 (nds32_expand_builtin_impl): Check if string extension is available.
27439 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
27440 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
27442 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
27445 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
27448 * lra-assigns.c (assign_by_spills): Return a flag of reload
27449 assignment failure. Do not process the reload assignment
27450 failures. Do not spill other reload pseudos if they has the same
27452 (lra_assign): Add a return arg. Set up from the result of
27453 assign_by_spills call.
27454 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
27455 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
27456 usage_insns if it is not NULL.
27457 (spill_hard_reg_in_range): New function.
27458 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
27459 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
27460 function prototypes.
27461 (lra_assign): Change prototype.
27462 * lra.c (lra): Add code to deal with fails by splitting hard reg
27465 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
27468 * config/i386/i386.opt: Replace Enforcment with Enforcement.
27470 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
27473 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
27474 (dw_val_node): Add val_symbolic_view.
27475 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
27476 (symview_upper_bound): New.
27477 (new_line_info_table): Initialize symviews_since_reset.
27478 (dwarf2out_source_line): Count symviews_since_reset and set
27479 symview_upper_bound.
27480 (dw_val_equal_p): Handle symview.
27481 (add_AT_symview): New.
27482 (print_dw_val): Handle symview.
27483 (attr_checksum, attr_checksum_ordered): Likewise.
27484 (same_dw_val_p, size_of_die): Likewise.
27485 (value_format, output_die): Likewise.
27486 (add_high_low_attributes): Use add_AT_symview for entry_view.
27487 (dwarf2out_finish): Reset symview_upper_bound, clear
27490 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
27493 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
27494 Add strict argument and use it.
27495 (rs6000_split_multireg_move): Update for new strict argument.
27496 (mem_operand_gpr): Disallow all non-offsettable addresses.
27497 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
27499 2018-03-09 Jakub Jelinek <jakub@redhat.com>
27502 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
27503 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
27504 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
27507 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
27508 decl, use remap_type if we want to use the type.
27510 2018-03-09 Martin Sebor <msebor@redhat.com>
27512 PR tree-optimization/84526
27513 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
27515 (builtin_access::generic_overlap): Be prepared to handle non-array
27518 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
27520 PR rtl-optimization/84682
27521 * lra-constraints.c (process_address_1): Check is_address flag
27522 for address constraints.
27523 (process_alt_operands): Likewise.
27524 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
27525 preprocess_constraints.
27526 * recog.h (preprocess_constraints): Add oploc parameter.
27528 * recog.c (preprocess_constraints): Test address_operand for
27529 CT_ADDRESS constraints.
27531 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
27534 * lra-assigns.c (assign_by_spills): Return a flag of reload
27535 assignment failure. Do not process the reload assignment
27536 failures. Do not spill other reload pseudos if they has the same
27538 (lra_assign): Add a return arg. Set up from the result of
27539 assign_by_spills call.
27540 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
27541 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
27542 usage_insns if it is not NULL.
27543 (spill_hard_reg_in_range): New function.
27544 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
27545 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
27546 function prototypes.
27547 (lra_assign): Change prototype.
27548 * lra.c (lra): Add code to deal with fails by splitting hard reg
27551 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27554 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
27555 Accept complain bool parameter. Only emit errors if it is true.
27556 (arm_parse_cpu_option_name): Likewise.
27557 (arm_target_thumb_only): Adjust callers of the above.
27558 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
27559 prototype to take a default true bool parameter.
27560 (arm_parse_arch_option_name): Likewise.
27562 2018-03-09 David Malcolm <dmalcolm@redhat.com>
27563 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
27567 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
27568 * configure: Regenerate.
27569 * configure.ac ("linker --version-script option"): New.
27570 ("linker soname option"): New.
27572 2018-03-09 Richard Biener <rguenther@suse.de>
27574 PR tree-optimization/84775
27575 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
27576 immediate uses of predicate stmts and mark them modified.
27579 PR tree-optimization/84178
27580 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
27582 (version_loop_for_if_conversion): Delay update_ssa call.
27583 (tree_if_conversion): Delay update_ssa until after predicate
27586 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
27589 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
27590 when the function accesses prior frames.
27592 2018-03-08 Jakub Jelinek <jakub@redhat.com>
27595 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
27596 gen_llsym, otherwise call maybe_gen_llsym.
27598 PR inline-asm/84742
27599 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
27600 has ',' character inside of it.
27602 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27605 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
27606 as clobbering CC_REGNUM.
27608 2018-03-08 Richard Biener <rguenther@suse.de>
27610 PR middle-end/84552
27611 * tree-scalar-evolution.c: Include tree-into-ssa.h.
27612 (follow_copies_to_constant): Do not follow SSA names registered
27615 2018-03-08 Richard Biener <rguenther@suse.de>
27617 PR tree-optimization/84178
27618 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
27620 (version_loop_for_if_conversion): Delay update_ssa call.
27621 (tree_if_conversion): Delay update_ssa until after predicate
27624 2018-03-08 David Malcolm <dmalcolm@redhat.com>
27626 PR tree-optimization/84178
27627 * tree-if-conv.c (release_bb_predicate): Remove the
27628 the assertion that the stmts have NULL use_ops.
27629 Discard the statements, asserting that they haven't
27630 yet been added to a BB.
27632 2018-03-08 Richard Biener <rguenther@suse.de>
27634 PR tree-optimization/84746
27635 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
27636 (phi_translate): Pass in destination ANTIC_OUT set.
27637 (phi_translate_1): Likewise. For a simplified result lookup
27638 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
27639 (phi_translate_set): Adjust.
27640 (do_pre_regular_insertion): Likewise.
27641 (do_pre_partial_partial_insertion): Likewise.
27643 2018-03-08 Martin Liska <mliska@suse.cz>
27645 PR gcov-profile/84735
27646 * doc/gcov.texi: Document usage of profile files.
27647 * gcov-io.h: Document changes in the format.
27649 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
27653 * dwarf2out.c (struct dw_line_info_table): Update comments for
27655 (FORCE_RESET_NEXT_VIEW): New.
27656 (FORCE_RESETTING_VIEW_P): New.
27657 (RESETTING_VIEW_P): Check for -1 too.
27658 (ZERO_VIEW_P): Likewise.
27659 (new_line_info_table): Force-reset next view.
27660 (dwarf2out_begin_function): Likewise.
27661 (dwarf2out_source_line): Simplify zero_view_p initialization.
27662 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
27663 view directly. Omit view when omitting .loc at line 0.
27665 2018-03-08 Jakub Jelinek <jakub@redhat.com>
27667 PR tree-optimization/84740
27668 * tree-switch-conversion.c (process_switch): Call build_constructors
27669 only if info.phi_count is non-zero.
27671 PR tree-optimization/84739
27672 * tree-tailcall.c (find_tail_calls): Check call arguments against
27673 DECL_ARGUMENTS (current_function_decl) rather than
27674 DECL_ARGUMENTS (func) when checking for tail recursion.
27676 2018-03-07 Jakub Jelinek <jakub@redhat.com>
27678 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
27679 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
27680 Volker Reichelt's entry and add entries for people that perform
27681 GCC fuzzy testing and report numerous bugs.
27683 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
27686 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
27687 readonly data in sdata, if that is disabled.
27688 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
27689 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
27690 -mreadonly-in-sdata option.
27692 2018-03-07 Martin Sebor <msebor@redhat.com>
27694 PR tree-optimization/84468
27695 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
27696 basic block when looking for nul assignment.
27698 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
27701 * except.h (output_function_exception_table): Adjust prototype.
27702 * except.c (output_function_exception_table): Remove FNNAME parameter
27703 and add SECTION parameter. Ouput one part of the table at a time.
27704 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
27705 the first part of the exception table and emit unwind directives.
27706 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
27707 (i386_pe_seh_cold_init): Likewise.
27708 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
27709 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
27710 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
27711 (ix86_output_call_insn): Emit a nop in one more case for SEH.
27712 * config/i386/winnt.c: Include except.h.
27713 (struct seh_frame_state): Add reg_offset, after_prologue and
27714 in_cold_section fields.
27715 (i386_pe_seh_end_prologue): Set seh->after_prologue.
27716 (i386_pe_seh_cold_init): New function.
27717 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
27718 to seh->in_cold_section.
27719 (seh_emit_push): Record the offset of the push.
27720 (seh_emit_save): Record the offet of the save.
27721 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
27722 Test seh->after_prologue to disregard the epilogue.
27723 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
27724 (i386_pe_end_cold_function): New function.
27726 2018-03-07 Jakub Jelinek <jakub@redhat.com>
27729 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
27730 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
27733 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
27735 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
27736 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
27738 PR middle-end/84723
27739 * multiple_target.c: Include tree-inline.h and intl.h.
27740 (expand_target_clones): Diagnose and fail if node->definition and
27741 !tree_versionable_function_p (node->decl).
27743 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
27745 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
27747 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
27748 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
27749 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
27751 2018-03-06 Jakub Jelinek <jakub@redhat.com>
27754 * combine.c (try_combine): Use reg_or_subregno instead of handling
27755 just paradoxical SUBREGs and REGs.
27757 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
27759 * config/arc/arc.c (arc_finalize_pic): Remove function.
27760 (arc_must_save_register): We use single base PIC register, remove
27761 checks to save/restore the PIC register.
27762 (arc_expand_prologue): Likewise.
27763 * config/arc/arc-protos.h (arc_set_default_type_attributes):
27765 (arc_verify_short): Likewise.
27766 (arc_attr_type): Likewise.
27767 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
27768 (walk_stores): Likewise.
27769 (arc_address_cost): Make it static.
27770 (arc_verify_short): Likewise.
27771 (branch_dest): Likewise.
27772 (arc_attr_type): Likewise.
27773 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
27774 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
27775 (arc_final_prescan_insn): Remove inserting the nops due to
27776 hardware hazards. It is done in reorg step.
27777 (insn_length_variant_t): Remove.
27778 (insn_length_parameters_t): Likewise.
27779 (arc_insn_length_parameters): Likewise.
27780 (arc_get_insn_variants): Likewise.
27781 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
27783 2018-03-06 Jakub Jelinek <jakub@redhat.com>
27785 PR inline-asm/84683
27786 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
27789 PR tree-optimization/84687
27790 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
27792 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
27794 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
27796 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
27797 Rename to ppc_speculation_barrier.
27798 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
27799 __builtin_ppc_speculation_barrier.
27801 2018-03-05 Jakub Jelinek <jakub@redhat.com>
27804 * combine.c (combine_simplify_rtx): Don't try to simplify if
27805 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
27808 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
27810 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
27811 to 32 bytes when compiling for POWER9.
27813 2018-03-05 Jakub Jelinek <jakub@redhat.com>
27816 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
27817 regparm >= 3 with no arg reg available also for calls with
27818 flag_force_indirect_call. Pass decl to ix86_function_regparm.
27821 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
27823 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
27825 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
27828 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
27830 2018-03-05 Richard Biener <rguenther@suse.de>
27832 PR tree-optimization/84486
27833 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
27834 When inserting a __builtin_assume_aligned call set the LHS
27835 SSA name alignment info accordingly.
27837 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
27839 PR tree-optimization/84114
27840 * config/aarch64/aarch64.c (aarch64_reassociation_width)
27841 Avoid reassociation of FLOAT_MODE addition.
27843 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
27845 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
27846 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
27847 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
27848 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
27849 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
27850 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
27851 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
27853 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
27854 __builtin_ia32_wbinvd): New builtins.
27855 (SPECIAL_ARGS2): New.
27856 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
27857 (SPECIAL_ARGS2): New.
27858 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
27859 (ix86_valid_target_attribute_inner_p): Ditto.
27860 (ix86_init_mmx_sse_builtins): Add special_args2.
27861 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
27862 TARGET_WBNOINVD_P): New.
27863 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
27864 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
27865 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
27866 * config/i386/immintrin.h (_wbinvd): New intrinsic.
27867 * config/i386/pconfigintrin.h: New file.
27868 * config/i386/wbnoinvdintrin.h: Ditto.
27869 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
27871 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
27873 2018-03-05 Richard Biener <rguenther@suse.de>
27875 PR tree-optimization/84670
27876 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
27878 (BB_VISITED_WITH_VISITED_SUCCS): New define.
27879 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
27880 (compute_antic_aux): Only assert the number of values in ANTIC_IN
27881 doesn't grow if all successors (recursively) were visited at least
27884 2018-03-05 Richard Biener <rguenther@suse.de>
27886 PR tree-optimization/84650
27887 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
27888 if executed in the loop pipeline.
27890 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
27892 * doc/configfiles.texi (Configuration Files): Move info about
27893 conditionalizing $target-protos.h to...
27894 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
27895 differs from $target-protos.h.
27897 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
27898 Chung-Ju Wu <jasonwucj@gmail.com>
27900 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
27901 * config/nds32/nds32-multiple.md (setmemsi): Define.
27902 * config/nds32/nds32-memory-manipulation.c
27903 (nds32_gen_dup_4_byte_to_word_value): New.
27904 (emit_setmem_word_loop): New.
27905 (emit_setmem_byte_loop): New.
27906 (nds32_expand_setmem_loop): New.
27907 (nds32_expand_setmem_loop_v3m): New.
27908 (nds32_expand_setmem_unroll): New.
27909 (nds32_expand_setmem): New.
27911 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
27912 Chung-Ju Wu <jasonwucj@gmail.com>
27914 * config/nds32/nds32-memory-manipulation.c
27915 (nds32_emit_load_store): New.
27916 (nds32_emit_post_inc_load_store): New.
27917 (nds32_emit_mem_move): New.
27918 (nds32_emit_mem_move_block): New.
27919 (nds32_expand_movmemsi_loop_unknown_size): New.
27920 (nds32_expand_movmemsi_loop_known_size): New.
27921 (nds32_expand_movmemsi_loop): New.
27922 (nds32_expand_movmemsi_unroll): New.
27923 (nds32_expand_movmemqi): Rename ...
27924 (nds32_expand_movmemsi): ... to this.
27925 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
27926 (movmemsi): ... to this.
27927 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
27928 (nds32_expand_movmemsi): ... to this.
27930 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
27931 Monk Chiang <sh.chiang04@gmail.com>
27932 Chung-Ju Wu <jasonwucj@gmail.com>
27934 * config/nds32/nds32-protos.h
27935 (nds32_expand_load_multiple): New arguments.
27936 (nds32_expand_store_multiple): Ditto.
27937 (nds32_valid_multiple_load_store): Rename ...
27938 (nds32_valid_multiple_load_store_p): ... to this.
27939 * config/nds32/nds32-memory-manipulation.c
27940 (nds32_expand_load_multiple): Refine implementation.
27941 (nds32_expand_store_multiple): Ditto.
27942 * config/nds32/nds32-multiple.md
27943 (load_multiple): Update nds32_expand_load_multiple interface.
27944 (store_multiple): Update nds32_expand_store_multiple interface.
27945 * config/nds32/nds32-predicates.c
27946 (nds32_valid_multiple_load_store): Rename ...
27947 (nds32_valid_multiple_load_store_p): ... to this and refine
27949 * config/nds32/predicates.md
27950 (nds32_load_multiple_and_update_address_operation): New predicate.
27951 (nds32_store_multiple_and_update_address_operation): New predicate.
27953 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
27954 Chung-Ju Wu <jasonwucj@gmail.com>
27956 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
27957 (combo): New attribute.
27958 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
27960 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
27962 * config/nds32/nds32.opt: Change -mcmodel= default value.
27964 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
27965 Monk Chiang <sh.chiang04@gmail.com>
27966 Chung-Ju Wu <jasonwucj@gmail.com>
27968 * config/nds32/constants.md (unspec_element): New enum.
27969 * config/nds32/constraints.md (Umw): New constraint.
27970 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
27971 * config/nds32/nds32-intrinsic.md: Likewise.
27972 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
27973 (nds32_valid_smw_lwm_base_p): New.
27974 (nds32_output_smw_single_word): New.
27975 (nds32_output_lmw_single_word): New.
27976 (nds32_expand_unaligned_load): New.
27977 (nds32_expand_unaligned_store): New.
27978 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
27979 (nds32_output_smw_single_word): Declare.
27980 (nds32_output_lmw_single_word): Declare.
27981 (nds32_expand_unaligned_load): Declare.
27982 (nds32_expand_unaligned_store): Declare.
27983 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
27984 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
27985 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
27986 NDS32_BUILTIN_UASTORE_DW.
27987 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
27990 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
27991 Kito Cheng <kito.cheng@gmail.com>
27992 Chung-Ju Wu <jasonwucj@gmail.com>
27994 * config/nds32/nds32-intrinsic.c
27995 (nds32_expand_builtin_null_ftype_reg): Delete.
27996 (nds32_expand_builtin_reg_ftype_imm): Ditto.
27997 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
27998 (nds32_read_argument): New.
27999 (nds32_legitimize_target): Ditto.
28000 (nds32_legitimize_argument): Ditto.
28001 (nds32_check_constant_argument): Ditto.
28002 (nds32_expand_unop_builtin): Ditto.
28003 (nds32_expand_unopimm_builtin): Ditto.
28004 (nds32_expand_binop_builtin): Ditto.
28005 (nds32_builtin_decl_impl): Ditto.
28006 (builtin_description): Ditto.
28007 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
28008 (nds32_init_builtins_impl): Ditto.
28009 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
28010 (nds32_builtin_decl): New.
28011 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
28012 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
28014 2018-03-02 Jeff Law <law@redhat.com>
28016 * reorg.c (stop_search_p): Handle DEBUG_INSN.
28017 (redundant_insn, fill_simple_delay_slots): Likewise.
28018 (fill_slots_from_thread): Likewise.
28019 * resource.c (mark_referenced_resources): Likewise.
28020 (mark_set_resources, find_dead_or_set_registers): Likewise.
28022 2018-03-02 Jakub Jelinek <jakub@redhat.com>
28024 * substring-locations.h (format_warning_va): Formatting fix for
28025 ATTRIBUTE_GCC_DIAG.
28026 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
28028 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
28029 * substring-locations.c: Include intl.h.
28030 (format_warning_va): Turned into small wrapper around
28031 format_warning_n_va, renamed to ...
28032 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
28033 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
28035 (format_warning_at_substring_n): New function.
28036 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
28037 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
28038 format_warning_at_substring with just a shorter name instead of
28039 const function pointer.
28040 (fmtwarn_n): New function.
28041 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
28042 appropriate, get rid of all the fmtstr temporaries, move conditionals
28043 with G_() wrapped string literals directly into fmtwarn arguments,
28044 cast dir.len to (int), formatting fixes.
28046 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
28048 * doc/invoke.texi: Remove "Cilk Plus" references.
28050 2018-03-02 Jakub Jelinek <jakub@redhat.com>
28051 Richard Biener <rguenther@suse.de>
28054 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
28055 for error or warning attributes if CALL_FROM_THUNK_P is set.
28058 2018-03-02 Jakub Jelinek <jakub@redhat.com>
28061 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
28062 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
28065 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
28066 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
28068 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
28069 instead of -1U in last predictors element's probability member.
28071 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
28074 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
28075 arguments if they are comparable.
28077 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
28079 PR tree-optimization/84634
28080 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
28081 masks and masked_loop_p with a single loop_masks, making sure it's
28082 null for bb vectorization.
28084 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
28086 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
28087 (vect_analyze_data_ref_access): Use loop->safe_len rather than
28088 loop->force_vectorize to check whether there is no alias.
28090 2018-03-02 Jakub Jelinek <jakub@redhat.com>
28093 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
28095 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
28097 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
28098 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
28099 instead of a loop around prev_real_insn.
28100 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
28103 PR inline-asm/84625
28104 * config/i386/i386.c (ix86_print_operand): Use conditional
28105 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
28108 2018-03-02 Richard Biener <rguenther@suse.de>
28110 PR tree-optimization/84427
28111 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
28112 (bitmap_set_subtract_values): Rewrite to handle multiple
28115 (prune_clobbered_mems): Likewise.
28116 (phi_translate): Take edge instead of pred/phiblock.
28117 (phi_translate_1): Likewise.
28118 (phi_translate_set): Likewise. Insert all translated
28119 exprs for a value into the set, keeping possibly multiple
28120 expressions per value.
28121 (compute_antic_aux): Adjust for phi_translate changes.
28122 When intersecting union the expressions and prune those
28123 not in the final value set, keeping possibly multiple
28124 expressions per value. Do not use value-insertion
28125 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
28126 all expressions. Add verification that the value-sets
28127 only shrink during iteration.
28128 (compute_partial_antic_aux): Adjust for the phi_translate changes.
28129 (do_pre_regular_insertion): Likewise.
28130 (do_pre_partial_partial_insertion): Likewise.
28132 2018-03-02 Richard Biener <rguenther@suse.de>
28135 * config/darwin.c (saved_debug_info_level): New static global.
28136 (darwin_asm_lto_start): Disable debug info generation for LTO out.
28137 (darwin_asm_lto_end): Restore debug info generation settings.
28139 2018-03-01 Martin Liska <mliska@suse.cz>
28142 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
28143 volatile arguments.
28145 2018-03-01 Richard Biener <rguenther@suse.de>
28148 * dwarf2out.c (gen_variable_die): Properly handle late VLA
28149 type annotation with LTO when debug was disabled at compile-time.
28151 2018-03-01 Matthew Fortune <mfortune@gmail.com>
28153 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
28155 (mips_final_postscan_insn): Likewise.
28157 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
28159 PR rtl-optimization/84528
28160 * alias.c (init_alias_target): Add commentary.
28161 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
28162 a unique base value if the frame pointer is not eliminated
28163 to the stack pointer.
28165 2018-03-01 Tom de Vries <tom@codesourcery.com>
28167 PR rtl-optimization/83327
28168 * lra-int.h (hard_regs_spilled_into): Declare.
28169 * lra.c (hard_regs_spilled_into): Define.
28170 (init_reg_info): Init hard_regs_spilled_into.
28171 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
28172 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
28173 (process_bb_lives): Handle hard_regs_spilled_into.
28174 (lra_create_live_ranges_1): Before doing liveness propagation, clear
28175 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
28177 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
28179 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
28180 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
28181 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
28182 * config/rs6000/aix72.h: New file.
28184 2018-02-28 Jakub Jelinek <jakub@redhat.com>
28186 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
28187 instead of warning_at with conditional singular and plural messages
28191 * stor-layout.c (update_alignment_for_field): For
28192 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
28193 && !DECL_PACKED (field), do the alignment update, just use
28194 only desired_align instead of MAX (type_align, desired_align)
28196 (place_field): Don't do known_align < desired_align handling
28197 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
28198 is non-NULL, instead do it after rli->prev_field handling and
28199 only if not within a bitfield word. For DECL_PACKED (field)
28200 use type_align of BITS_PER_UNIT.
28202 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
28204 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
28205 superfluous parentheses and trailing spaces.
28207 2018-02-28 Richard Biener <rguenther@suse.de>
28209 PR tree-optimization/84584
28210 * graphite-scop-detection.c (scop_detection::add_scop): Discard
28211 SCoPs with fake exit edge.
28213 2018-02-28 Martin Liska <mliska@suse.cz>
28216 * timevar.c (timer::print): Fix format to properly print 100%
28219 2018-02-28 Richard Biener <rguenther@suse.de>
28221 PR middle-end/84607
28222 * genmatch.c (capture_info::walk_match): Do not mark
28223 captured expressions without operands as expr_p given
28224 they act more like predicates and should be subject to
28225 "lost tail" side-effect preserving.
28227 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
28229 PR rtl-optimization/81611
28230 * auto-inc-dec.c (attempt_change): Move dead note from
28231 mem_insn if it's the next use of regno
28232 (find_address): Take address use of reg holding
28233 non-incremented value. Add parm to limit search to the named
28235 (merge_in_block): Attempt to use a mem insn that is the next
28236 use of the original regno.
28238 2018-02-27 Martin Sebor <msebor@redhat.com>
28241 * doc/invoke.texi (-Wmissing-attributes): New option.
28242 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
28244 2018-02-27 Martin Sebor <msebor@redhat.com>
28246 PR translation/84207
28247 * diagnostic-core.h (warning_n, error_n, inform_n): Change
28248 n argument to unsigned HOST_WIDE_INT.
28249 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
28250 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
28251 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
28252 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
28254 2018-02-27 Richard Biener <rguenther@suse.de>
28256 PR tree-optimization/84512
28257 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
28258 Do not use the estimate returned from record_stmt_cost for
28259 the scalar iteration cost but sum properly using add_stmt_cost.
28261 2018-02-27 Richard Biener <rguenther@suse.de>
28263 PR tree-optimization/84466
28264 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
28265 Adjust last change to less strictly validate use operands.
28267 2018-02-27 Martin Liska <mliska@suse.cz>
28269 PR gcov-profile/84548
28270 * gcov.c (process_file): Allow partial overlap and consider it
28271 also as group functions.
28272 (output_lines): Properly calculate range of lines for a group.
28274 2018-02-27 Martin Liska <mliska@suse.cz>
28276 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
28277 'ggc' suffixes. Change first column width.
28278 (timer::print): Fix formatting of the column.
28280 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
28282 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
28283 preserve inline entry blocks for the sake of debug inline
28284 entry point markers alone.
28285 (remove_unused_locals): Suggest in comments a better place to
28286 force the preservation of inline entry blocks that are
28287 otherwise unused, but do not preserve them.
28289 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
28291 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
28293 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
28296 * config/i386/constraints.md (Bs): Replace
28297 ix86_indirect_branch_register with
28298 TARGET_INDIRECT_BRANCH_REGISTER.
28300 * config/i386/i386.md (indirect_jump): Likewise.
28301 (tablejump): Likewise.
28302 (*sibcall_memory): Likewise.
28303 (*sibcall_value_memory): Likewise.
28304 Peepholes of indirect call and jump via memory: Likewise.
28305 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
28306 (*sibcall_value_GOT_32): Likewise.
28307 * config/i386/predicates.md (indirect_branch_operand): Likewise.
28308 (GOT_memory_operand): Likewise.
28309 (call_insn_operand): Likewise.
28310 (sibcall_insn_operand): Likewise.
28311 (GOT32_symbol_operand): Likewise.
28312 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
28314 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
28316 PR rtl-optimization/83496
28317 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
28318 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
28319 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
28320 redundant insn, if any.
28321 (relax_delay_slots): Likewise.
28322 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
28324 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
28326 PR tree-optimization/83965
28327 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
28328 that grouped statements are part of a reduction chain. Return
28329 true if the statement is not marked as a reduction itself but
28330 is part of a group.
28331 (vect_recog_dot_prod_pattern): Don't check whether the statement
28332 is part of a group here.
28333 (vect_recog_sad_pattern): Likewise.
28334 (vect_recog_widen_sum_pattern): Likewise.
28336 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
28339 * final.c (rest_of_clean_state): Also look for calls inside sequences.
28341 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
28344 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
28346 (ix86_output_indirect_function_return): New prototype.
28347 (ix86_split_simple_return_pop_internal): Likewise.
28348 * config/i386/i386.c (indirect_return_via_cx): New.
28349 (indirect_return_via_cx_bnd): Likewise.
28350 (indirect_thunk_name): Handle return va CX_REG.
28351 (output_indirect_thunk_function): Create alias for
28352 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
28353 (ix86_output_indirect_jmp): Remove the bool argument.
28354 (ix86_output_indirect_function_return): New function.
28355 (ix86_split_simple_return_pop_internal): Likewise.
28356 * config/i386/i386.md (*indirect_jump): Don't pass false
28357 to ix86_output_indirect_jmp.
28358 (*tablejump_1): Likewise.
28359 (simple_return_pop_internal): Change it to define_insn_and_split.
28360 Call ix86_split_simple_return_pop_internal to split it for
28361 -mfunction-return=.
28362 (simple_return_indirect_internal): Call
28363 ix86_output_indirect_function_return instead of
28364 ix86_output_indirect_jmp.
28366 2018-02-26 Jakub Jelinek <jakub@redhat.com>
28369 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
28370 memset and value initialization afterwards.
28372 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
28374 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
28376 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
28379 * common/config/aarch64/aarch64-common.c
28380 (aarch_option_optimization_table[]): Switch
28381 off fomit-frame-pointer
28383 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
28384 Chung-Ju Wu <jasonwucj@gmail.com>
28386 * config/nds32/nds32-multiple.md (load_multiple): Disallow
28388 (store_multiple): Ditto.
28390 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
28392 * config.gcc: Add --with-cpu support for nds32 target.
28393 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
28394 * config/nds32/nds32.opt: Add -mcpu= option.
28396 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
28398 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
28399 isel=yes): Warn for these deprecated options.
28401 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
28403 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
28404 ISA_2_5_MASKS_EMBEDDED.
28406 2018-02-23 Jakub Jelinek <jakub@redhat.com>
28408 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
28409 p->max as pointers rather than using iterative_hash_expr.
28411 2018-02-23 Carl Love <cel@us.ibm.com>
28413 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
28414 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
28416 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
28417 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
28418 P8V_BUILTIN_VEC_VUNSIGNED2.
28420 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
28423 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
28424 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
28426 * lra-constraints.c (curr_insn_transform): Set up
28427 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
28429 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
28430 * lra-eliminations.c (spill_pseudos): Ditto.
28431 (process_insn_for_elimination): Ditto.
28432 * lra-lives.c (reg_early_clobber_p): Use the new macros.
28433 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
28434 LRA_NON_CLOBBERED_ALT.
28436 2018-02-22 Martin Sebor <msebor@redhat.com>
28438 PR tree-optimization/84480
28439 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
28440 to maybe_diag_stxncpy_trunc. Call it.
28441 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
28442 from gimple_fold_builtin_strcpy. Print inlining stack.
28443 (handle_builtin_stxncpy): Print inlining stack.
28444 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
28446 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
28449 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
28450 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
28451 and -fcheck-pointer-bounds are used together.
28452 (indirect_thunk_prefix): New enum.
28453 (indirect_thunk_need_prefix): New function.
28454 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
28455 "_nt" instead of "_bnd" for NOTRACK prefix.
28456 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
28457 (output_indirect_thunk_function): Likewise.
28459 (ix86_code_end): Update output_indirect_thunk_function calls.
28460 (ix86_output_indirect_branch_via_reg): Replace
28461 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
28462 (ix86_output_indirect_branch_via_push): Likewise.
28463 (ix86_output_function_return): Likewise.
28464 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
28465 incompatible with -fcf-protection=branch and
28466 -fcheck-pointer-bounds.
28468 2018-02-22 Steve Ellcey <sellcey@cavium.com>
28471 * config/aarch64/aarch64.c (aarch64_print_address_internal):
28472 Change gcc_assert call to output_operand_lossage.
28474 2018-02-22 Steve Ellcey <sellcey@cavium.com>
28476 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
28478 2018-02-22 DJ Delorie <dj@redhat.com>
28479 Sebastian Perta <sebastian.perta@renesas.com>
28480 Oleg Endo <olegendo@gcc.gnu.org>
28482 * config/rx/rx.c (rx_rtx_costs): New function.
28483 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
28485 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
28487 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
28489 2018-02-22 Martin Liska <mliska@suse.cz>
28492 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
28493 Add "native" as a possible value.
28495 2018-02-22 Martin Liska <mliska@suse.cz>
28498 * config/i386/i386.c (ix86_option_override_internal):
28499 Add "native" as a possible value for -march and -mtune.
28501 2018-02-22 Jakub Jelinek <jakub@redhat.com>
28504 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
28505 to all type variants.
28507 PR tree-optimization/84503
28508 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
28509 width as info->bitpos + info->bitsize - start.
28510 (merged_store_group::merge_overlapping): Simplify width computation.
28511 (check_no_overlap): New function.
28512 (imm_store_chain_info::try_coalesce_bswap): Compute expected
28513 start + width and last_order of the group, fail if check_no_overlap
28515 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
28516 to group if check_no_overlap fails.
28518 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
28520 * config/rs6000/altivec.md: Delete contraint arguments to
28521 define_expand, define_split, and define_peephole2, and in
28522 define_insn_and_split if always unused.
28523 * config/rs6000/darwin.md: Ditto.
28524 * config/rs6000/dfp.md: Ditto.
28525 * config/rs6000/rs6000.md: Ditto.
28526 * config/rs6000/sync.md: Ditto.
28527 * config/rs6000/vector.md: Ditto.
28528 * config/rs6000/vsx.md: Ditto.
28530 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
28532 * config/rs6000/altivec.md: Write output control strings as braced
28533 blocks instead of double-quoted strings.
28534 * config/rs6000/darwin.md: Ditto.
28535 * config/rs6000/rs6000.md: Ditto.
28536 * config/rs6000/vector.md: Ditto.
28537 * config/rs6000/vsx.md: Ditto.
28539 2018-02-21 Jason Merrill <jason@redhat.com>
28541 PR c++/84314 - ICE with templates and fastcall attribute.
28542 * attribs.c (build_type_attribute_qual_variant): Remove assert.
28544 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
28546 * ipa-cp.c (determine_versionability): Fix comment typos.
28548 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
28551 * ipa-cp.c (determine_versionability): Do not version functions caling
28554 2018-02-21 Martin Liska <mliska@suse.cz>
28557 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
28558 Add "native" as a possible value.
28559 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
28560 the macro when native cpu detection is available.
28562 2018-02-21 Martin Liska <mliska@suse.cz>
28565 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
28566 Add "native" as a possible value.
28567 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
28568 when native cpu detection is available.
28570 2018-02-21 Jakub Jelinek <jakub@redhat.com>
28571 Martin Sebor <msebor@redhat.com>
28573 PR tree-optimization/84478
28574 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
28576 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
28577 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
28578 support which is conservatively correct, for 2 only stay conservative
28579 for maxlen. Formatting and comment capitalization fixes. Add STRICT
28580 argument to the 2 argument get_range_strlen, adjust 6 arg
28581 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
28583 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
28584 (gimple_fold_builtin_strlen): Pass true as last argument to
28587 2018-02-20 Martin Sebor <msebor@redhat.com>
28589 PR middle-end/84095
28590 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
28591 (builtin_memref::set_base_and_offset): Same. Handle inner references.
28592 (builtin_memref::builtin_memref): Factor out parts into
28593 set_base_and_offset and call it.
28595 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
28597 PR middle-end/84406
28598 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
28599 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
28600 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
28601 search at the associated MODE_INT.
28603 2018-02-20 Jeff Law <law@redhat.com>
28605 PR middle-end/82123
28606 PR tree-optimization/81592
28607 PR middle-end/79257
28608 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
28609 for range data rather than using global data.
28610 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
28611 range data rather than using global data.
28612 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
28613 pass it to children as needed.
28614 (struct directive::fmtresult): Similarly.
28615 (struct directive::set_width): Similarly.
28616 (struct directive::set_precision): Similarly.
28617 (format_integer, format_directive, parse_directive): Similarly.
28618 (format_none): Accept unnamed vr_values parameter.
28619 (format_percent, format_floating, format_character): Similarly.
28620 (format_string, format_plain): Similarly.
28621 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
28622 the EVRP range analyzer for range data rather than using global data.
28623 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
28624 gimple-ssa-evrp-analyze.h
28625 (class sprintf_dom_walker): Add after_dom_children member function.
28626 Add evrp_range_analyzer member.
28627 (sprintf_dom_walker::before_dom_children): Call into the EVRP
28628 range analyzer as needed.
28629 (sprintf_dom_walker::after_dom_children): New member function.
28630 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
28632 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
28633 (evrp_range_analyzer::pop_to_marker): Likewise.
28635 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
28637 PR tree-optimization/84419
28638 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
28639 with the required type if its current type is compatible but
28642 2018-02-20 Jakub Jelinek <jakub@redhat.com>
28644 PR middle-end/82004
28645 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
28646 after vectorization.
28648 2018-02-20 Martin Liska <mliska@suse.cz>
28651 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
28652 possible values if we don't have a hint.
28654 2018-02-20 Martin Liska <mliska@suse.cz>
28658 * final.c (shorten_branches): Build align_tab array with one
28660 * opts.c (finish_options): Add alignment option limit check.
28661 (MAX_CODE_ALIGN): Likewise.
28662 (MAX_CODE_ALIGN_VALUE): Likewise.
28663 * doc/invoke.texi: Document maximum allowed option value for
28664 all -falign-* options.
28666 2018-02-19 Jakub Jelinek <jakub@redhat.com>
28669 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
28670 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
28671 * var-tracking.c (emit_note_insn_var_location): Remove all references
28672 to NOTE_INSN_CALL_ARG_LOCATION.
28673 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
28674 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
28675 Use copy_rtx_if_shared.
28676 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
28677 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
28678 (dwarf2out_var_location): Remove handling of
28679 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
28681 * final.c (final_scan_insn): Remove all references to
28682 NOTE_INSN_CALL_ARG_LOCATION.
28683 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
28684 before dumping final insns.
28685 * except.c (emit_note_eh_region_end): Remove all references to
28686 NOTE_INSN_CALL_ARG_LOCATION.
28687 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
28688 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
28689 * config/arc/arc.c (hwloop_optimize): Likewise.
28690 * config/arm/arm.c (create_fix_barrier): Likewise.
28691 * config/s390/s390.c (s390_chunkify_start): Likewise.
28692 * config/sh/sh.c (find_barrier): Likewise.
28693 * config/i386/i386.c (rest_of_insert_endbranch,
28694 ix86_seh_fixup_eh_fallthru): Likewise.
28695 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
28696 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
28697 * config/frv/frv.c (frv_function_prologue): Likewise.
28698 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
28700 (note_outside_basic_block_p): Remove all references to
28701 NOTE_INSN_CALL_ARG_LOCATION.
28702 * gengtype.c (adjust_field_rtx_def): Likewise.
28703 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
28705 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
28706 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
28709 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
28712 PR tree-optimization/84452
28713 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
28714 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
28717 2018-02-19 Martin Liska <mliska@suse.cz>
28720 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
28722 2018-02-19 Martin Liska <mliska@suse.cz>
28723 Richard Sandiford <richard.sandiford@linaro.org>
28725 PR tree-optimization/82491
28726 * gimple-fold.c (get_base_constructor): Make earlier bail out
28729 2018-02-19 Carl Love <cel@us.ibm.com>
28731 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
28732 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
28734 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
28735 P8V_BUILTIN_VEC_NEG.
28737 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
28739 * config/rl78/rl78.md (movdf): New define expand.
28741 2018-02-19 Martin Liska <mliska@suse.cz>
28744 * doc/invoke.texi: Fix typo.
28745 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
28747 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
28749 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
28750 handle rs6000_single_float and rs6000_double_float specially for
28753 2018-02-16 Jeff Law <law@redhat.com>
28755 * config/rx/rx.c (add_pop_cfi_notes): New function.;
28756 (pop_regs): Use it.
28758 2018-02-16 Jakub Jelinek <jakub@redhat.com>
28761 * ipa-inline.c (inline_small_functions): Fix a typo.
28763 2018-02-16 Nathan Sidwell <nathan@acm.org>
28765 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
28767 2018-02-16 Carl Love <cel@us.ibm.com>
28769 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
28770 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
28771 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
28772 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
28773 expansion to P8V_BUILTIN_VEC_FLOAT2.
28775 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
28777 PR rtl-optimization/70023
28778 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
28779 src_regno into account.
28781 2018-02-16 Carl Love <cel@us.ibm.com>
28783 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
28784 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
28785 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
28786 * config/rs6000/rs6000.c: Remove case statements for
28787 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
28788 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
28789 and P9V_BUILTIN_VEC_VINSERT4B.
28790 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
28791 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
28792 * config/rs6000/vsx.md:
28793 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
28796 2018-02-16 Carl Love <cel@us.ibm.com>
28798 * config/rs6000/altivec.h: Add builtin names vec_extract4b
28800 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
28802 * config/rs6000/rs6000-c.c: Add the definitions for
28803 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
28804 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
28805 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
28806 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
28807 definition for insert4b and define insn *insert3b_internal.
28808 * doc/extend.texi: Add documentation for vec_extract4b.
28810 2018-02-16 Nathan Sidwell <nathan@acm.org>
28812 * doc/extend.texi (Backwards Compatibility): Mention friend
28813 injection. Note for-scope is deprecated.
28814 * doc/invoke.texi (-ffriend-injection): Deprecate.
28816 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
28818 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
28819 that moved to I2, also allow destinations that are a paradoxical
28820 subreg (instead of a normal reg).
28822 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
28825 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
28828 2018-02-16 Richard Biener <rguenther@suse.de>
28830 PR tree-optimization/84037
28831 PR tree-optimization/84016
28833 * config/i386/i386.c (ix86_builtin_vectorization_cost):
28834 Adjust vec_construct for the fact we need additional higher latency
28835 128bit inserts for AVX256 and AVX512 vector builds.
28836 (ix86_add_stmt_cost): Scale vector construction cost for
28839 2018-02-16 Richard Biener <rguenther@suse.de>
28841 PR tree-optimization/84417
28842 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
28843 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
28844 (non_rewritable_lvalue_p): Likewise, use poly-ints.
28846 2018-02-16 Martin Liska <mliska@suse.cz>
28849 * internal-fn.def (ASAN_CHECK): Set proper flags.
28850 (ASAN_MARK): Likewise.
28852 2018-02-16 Julia Koval <julia.koval@intel.com>
28854 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
28855 from PTA_CANNONLAKE.
28857 2018-02-16 Jakub Jelinek <jakub@redhat.com>
28860 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
28861 Use ++iter rather than iter++ for std::list iterators.
28862 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
28863 defer deleting them until all nodes in the forest are processed. Do
28864 free even leaf nodes. Change to_process into auto_vec.
28867 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
28868 * vec.h (vec_default_construct): Use memset instead of placement new
28869 if BROKEN_VALUE_INITIALIZATION is defined.
28870 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
28871 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
28874 PR rtl-optimization/83723
28875 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
28876 * lra.c (lra_substitute_pseudo): Likewise. If true, use
28877 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
28879 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
28881 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
28883 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
28885 PR rtl-optimization/81443
28886 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
28887 from inner REGs to paradoxical SUBREGs.
28889 2018-02-16 Richard Biener <rguenther@suse.de>
28891 PR tree-optimization/84399
28892 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
28893 For operands we can analyze at their definition make sure we can
28894 analyze them at each use as well.
28896 2018-02-16 Richard Biener <rguenther@suse.de>
28898 PR tree-optimization/84190
28899 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
28900 volatile accesses if the decl isn't volatile.
28902 2018-02-15 Jason Merrill <jason@redhat.com>
28904 PR c++/84314 - ICE with templates and fastcall attribute.
28905 * attribs.c (build_type_attribute_qual_variant): Don't clobber
28906 TYPE_CANONICAL on an existing type.
28908 2018-02-15 Jakub Jelinek <jakub@redhat.com>
28910 PR tree-optimization/84383
28911 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
28912 dstoff nor call operand_equal_p if dstbase is NULL.
28914 PR tree-optimization/84334
28915 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
28916 also a CONSTANT_CLASS_P, punt.
28918 2018-02-14 Jim Wilson <jimw@sifive.com>
28920 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
28921 first SMALL_OPERAND check. New local min_second_step. Move assert
28922 to where locals are set. Add TARGET_RVC support.
28923 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
28925 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
28927 * doc/invoke.texi: Correct -Wformat-overflow code sample.
28929 2018-02-14 Martin Sebor <msebor@redhat.com>
28931 PR tree-optimization/83698
28932 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
28933 arrays constrain the offset range to their bounds.
28934 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
28935 (builtin_access::overlap): Avoid setting the size of overlap if it's
28937 (maybe_diag_overlap): Also consider arrays when deciding what values
28938 of offsets to include in diagnostics.
28940 2018-02-14 Martin Sebor <msebor@redhat.com>
28943 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
28944 that correspond to the kind of a declaration.
28946 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
28949 * config/pa/pa.md: Load address of PIC label using the linkage table
28950 if the label is nonlocal.
28952 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
28954 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
28955 warning message if user requests -maltivec=be.
28956 * doc/invoke.texi: Document deprecation of -maltivec=be.
28958 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
28961 * config/rs6000/rs6000-c.c: Update definitions for
28962 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
28963 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
28965 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
28968 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
28969 add _get_ssp intrinsics. Remove argument from
28970 __builtin_ia32_rdssp[d|q].
28971 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
28972 * config/i386/i386-builtin.def: Remove argument from
28973 __builtin_ia32_rdssp[d|q].
28974 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
28975 ix86_expand_special_args_builtin for _rdssp[d|q].
28976 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
28977 Clear register before usage.
28978 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
28979 Add documentation for new _get_ssp and _inc_ssp intrinsics.
28981 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
28983 PR tree-optimization/84357
28984 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
28985 operand 1 of an ARRAY_REF too.
28987 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
28990 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
28991 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
28993 (set_of_reg): New struct.
28994 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
28995 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
28996 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
28998 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
28999 Split into bitclr, bitset, bitinvert patterns if appropriate.
29000 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
29001 use rx_fuse_in_memory_bitop.
29002 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
29003 to named insn, correct maximum insn length.
29005 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
29008 * machmode.def: Define a complex mode for PARTIAL_INT.
29009 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
29011 * doc/rtl.texi: Document CSPImode.
29012 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
29014 (msp430_hard_regno_nregs_with_padding): Likewise.
29016 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
29019 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
29021 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
29023 PR rtl-optimization/84169
29024 * combine.c (try_combine): New variable split_i2i3. Set it to true if
29025 we generated a parallel as new i3 and we split that to new i2 and i3
29026 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
29027 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
29028 those to i2, not i1. Partially rewrite this scan code.
29030 2018-02-13 Jakub Jelinek <jakub@redhat.com>
29033 * stor-layout.c (place_field): For variable length fields, adjust
29034 offset_align afterwards not just based on the field's alignment,
29035 but also on the size.
29037 PR middle-end/84309
29038 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
29039 of exps and logs in the use_exp2 case.
29041 2018-02-13 Jeff Law <law@redhat.com>
29043 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
29044 entry for "vector".
29046 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
29049 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
29053 * common.opt (gas-loc-support, gas-locview-support): New.
29054 (ginline-points, ginternal-reset-location-views): New.
29055 * doc/invoke.texi: Document them. Use @itemx where intended.
29056 (gvariable-location-views): Adjust.
29057 * target.def (reset_location_view): New.
29058 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
29059 (TARGET_RESET_LOCATION_VIEW): New.
29060 * doc/tm.texi: Rebuilt.
29061 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
29062 (dwarf2out_default_as_locview_support): New.
29063 (output_asm_line_debug_info): Use option variables.
29064 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
29065 (output_loc_list): Likewise.
29066 (add_high_low_attributes): Check option variables.
29067 Don't output entry view attribute in strict mode.
29068 (gen_inlined_subroutine_die): Check option variables.
29069 (dwarf2out_inline_entry): Likewise.
29070 (init_sections_and_labels): Likewise.
29071 (dwarf2out_early_finish): Likewise.
29072 (maybe_reset_location_view): New, from...
29073 (dwarf2out_var_location): ... here. Call it.
29074 * debug.h (dwarf2out_default_as_loc_support): Declare.
29075 (dwarf2out_default_as_locview_support): Declare.
29076 * hooks.c (hook_int_rtx_insn_0): New.
29077 * hooks.h (hook_int_rtx_insn_0): Declare.
29078 * toplev.c (process_options): Take -gas-loc-support and
29079 -gas-locview-support from dwarf2out. Enable
29080 -gvariable-location-views by default only with locview
29081 assembler support. Enable -ginternal-reset-location-views by
29082 default only if the target defines the corresponding hook.
29083 Enable -ginline-points by default if location views are
29084 enabled; force it disabled if statement frontiers are
29086 * tree-inline.c (expand_call_inline): Check option variables.
29087 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
29089 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
29091 PR tree-optimization/84321
29092 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
29093 handling. Also check whether the anti-range contains any values
29094 that satisfy the mask; switch to a VR_RANGE if not.
29096 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
29099 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
29101 2018-02-13 Martin Jambor <mjambor@suse.cz>
29104 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
29105 of call statements, also set location of a load to a temporary.
29107 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
29109 * config/rl78/rl78.c (add_vector_labels): New function.
29110 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
29111 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
29112 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
29113 which checks that no arguments are passed.
29114 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
29115 * doc/extend.texi: Documentation for the new attribute.
29117 2018-02-13 Andreas Schwab <schwab@suse.de>
29119 * config/riscv/linux.h (CPP_SPEC): Define.
29121 2018-02-13 Jakub Jelinek <jakub@redhat.com>
29124 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
29125 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
29126 OPTION_MASK_ISA_AES as first argument to def_builtin_const
29127 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
29128 instead of OPTION_MASK_ISA_PCLMUL as first argument to
29129 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
29130 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
29131 temporarily for AES and PCLMUL builtins.
29133 PR tree-optimization/84339
29134 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
29135 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
29138 PR middle-end/84309
29139 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
29140 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
29141 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
29143 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
29145 * omp-simd-clone.h: New file.
29146 * omp-simd-clone.c: Include omp-simd-clone.h.
29147 (expand_simd_clones): No longer static.
29148 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
29149 cgraph.h and omp-simd-clone.h.
29150 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
29151 (vect_recog_widen_shift_pattern): Formatting fix.
29152 (vect_pattern_recog_1): Don't check optab for calls.
29155 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
29156 operands[2] into a REG before using gen_lowpart on it.
29158 2018-02-12 Jeff Law <law@redhat.com>
29161 * config/sh/sh.c (find_barrier): Consider a sibling call
29164 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
29165 successfully back substituting a reg.
29167 2018-02-12 Richard Biener <rguenther@suse.de>
29169 PR tree-optimization/84037
29170 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
29171 parameter, move visited init to caller.
29172 (vect_slp_analyze_operations): Separate cost from validity
29173 check, initialize visited once for all instances.
29174 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
29176 * tree-vect-stmts.c (vect_model_simple_cost): Make early
29178 (vect_model_promotion_demotion_cost): Likewise.
29179 (vectorizable_bswap): Guard cost modeling with !slp_node
29180 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
29182 (vectorizable_call): Likewise.
29183 (vectorizable_conversion): Likewise.
29184 (vectorizable_assignment): Likewise.
29185 (vectorizable_shift): Likewise.
29186 (vectorizable_operation): Likewise.
29187 (vectorizable_store): Likewise.
29188 (vectorizable_load): Likewise.
29189 (vectorizable_condition): Likewise.
29190 (vectorizable_comparison): Likewise.
29192 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
29195 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
29196 (ASAN_MARK): Fix fnspec to account for return value, change pointer
29197 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
29199 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
29201 PR middle-end/83665
29202 * params.def (inline-min-speedup): Increase from 8 to 15.
29203 (max-inline-insns-auto): Decrease from 40 to 30.
29204 * ipa-split.c (consider_split): Add some buffer for function to
29205 be considered inlining candidate.
29206 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
29209 2018-02-12 Richard Biener <rguenther@suse.de>
29211 PR tree-optimization/84037
29212 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
29213 matched stmts if we cannot swap the non-matched ones.
29215 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
29217 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
29218 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
29219 _mm_maskz_scalef_round_ss): New intrinsics.
29220 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
29221 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
29222 __builtin_ia32_scalefss_round): Remove.
29223 (__builtin_ia32_scalefsd_mask_round,
29224 __builtin_ia32_scalefss_mask_round): New intrinsics.
29225 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
29226 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
29227 ((match_operand:VF_128 2 "<round_nimm_predicate>"
29228 "<round_constraint>")): Changed to ...
29229 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
29230 "<round_scalar_constraint>")): ... this.
29231 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
29232 %0, %1, %2<round_op3>}"): Changed to ...
29233 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
29234 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
29235 %2<round_scalar_mask_op3>}"): ... this.
29236 * config/i386/subst.md (round_scalar_nimm_predicate): New.
29238 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
29240 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
29241 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
29242 (_mm_maskz_sqrt_round_ss): New intrinsics.
29243 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
29244 (__builtin_ia32_sqrtsd_mask_round)
29245 (__builtin_ia32_sqrtss_mask_round): New builtins.
29246 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
29247 (__builtin_ia32_sqrtss_round): Remove.
29248 (__builtin_ia32_sqrtsd_mask_round)
29249 (__builtin_ia32_sqrtss_mask_round): New builtins.
29250 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
29251 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
29252 ((match_operand:VF_128 1 "vector_operand"
29253 "xBm,<round_constraint>")): Changed to ...
29254 ((match_operand:VF_128 1 "vector_operand"
29255 "xBm,<round_scalar_constraint>")): ... this.
29256 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
29257 %0, %2, %<iptr>1<round_op3>}): Changed to ...
29258 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
29259 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
29260 %<iptr>1<round_scalar_mask_op3>}): ... this.
29261 ((set_attr "prefix" "<round_prefix>")): Changed to ...
29262 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
29264 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
29267 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
29268 Cast vec_cmpeq result to correct type.
29269 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
29270 Cast vec_cmpgt result to correct type.
29272 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
29274 * final.c (final_scan_insn_1): Renamed from...
29275 (final_scan_insn): ... this. New wrapper, to recover
29276 seen from the outermost call in recursive ones.
29277 * config/sparc/sparc.c (output_return): Drop seen from call.
29278 (output_sibcall): Likewise.
29279 * config/visium/visium.c (output_branch): Likewise.
29281 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
29283 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
29286 2018-02-10 Alan Modra <amodra@gmail.com>
29289 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
29290 Specify LR as an input.
29292 2018-02-10 Jakub Jelinek <jakub@redhat.com>
29295 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
29296 remove_member_access_dummy_vars): New functions.
29297 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
29298 lower_omp_1, execute_lower_omp): Use them.
29300 PR rtl-optimization/84308
29301 * shrink-wrap.c (spread_components): Release todo vector.
29303 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
29305 PR rtl-optimization/57193
29306 * ira-color.c (struct allocno_color_data): Add member
29307 conflict_allocno_hard_prefs.
29308 (update_conflict_allocno_hard_prefs): New.
29309 (bucket_allocno_compare_func): Add a preference based on
29310 conflict_allocno_hard_prefs.
29311 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
29312 (color_allocnos): Remove a dead code. Initiate
29313 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
29315 2018-02-09 Jakub Jelinek <jakub@redhat.com>
29318 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
29319 constraint from =wa to wa. Avoid a subreg on the output operand,
29320 instead use a pseudo and subreg it in a move.
29321 (p9_xxbrd_<mode>): Changed to ...
29322 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
29323 (p9_xxbrd_v2df): New expander.
29324 (p9_xxbrw_<mode>): Changed to ...
29325 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
29326 (p9_xxbrw_v4sf): New expander.
29328 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
29330 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
29332 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
29335 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
29336 multiply in 32-bit mode.
29337 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
29338 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
29341 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
29343 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
29344 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
29345 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
29346 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
29348 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
29351 * dwarf2out.c (is_trivial_indirect_ref): New function.
29352 (dwarf2out_late_global_decl): Do not generate a location
29353 attribute for variables that have a non-trivial DECL_VALUE_EXPR
29354 and that are not defined in the current unit.
29356 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
29358 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
29359 instead of a libcall for UNORDERED.
29361 2018-02-09 Tamar Christina <tamar.christina@arm.com>
29364 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
29365 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
29367 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
29370 * config/s390/s390.c (s390_set_current_function): Invoke
29371 s390_indirect_branch_settings also if fndecl didn't change.
29373 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
29375 * config/rs6000/rs6000.md (blockage): Set length to zero.
29377 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
29379 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
29381 2018-02-09 Jakub Jelinek <jakub@redhat.com>
29384 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
29385 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
29389 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
29390 PARALLEL incoming that failed vt_get_decl_and_offset check.
29392 PR middle-end/84237
29393 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
29394 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
29396 (get_variable_section): For decls in named .bss* sections pass true as
29397 second argument to bss_initializer_p.
29399 2018-02-09 Marek Polacek <polacek@redhat.com>
29400 Jakub Jelinek <jakub@redhat.com>
29403 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
29404 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
29405 Sync some changes from cxx_fold_indirect_ref.
29407 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
29409 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
29411 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
29412 (BLOCK_INLINE_ENTRY_LABEL): New.
29413 (dwarf2out_var_location): Disregard inline entry markers.
29414 (inline_entry_data): New struct.
29415 (inline_entry_data_hasher): New hashtable type.
29416 (inline_entry_data_hasher::hash): New.
29417 (inline_entry_data_hasher::equal): New.
29418 (inline_entry_data_table): New variable.
29419 (add_high_low_attributes): Add DW_AT_entry_pc and
29420 DW_AT_GNU_entry_view attributes if a pending entry is found
29421 in inline_entry_data_table. Add old entry_pc attribute only
29422 if debug nonbinding markers are disabled.
29423 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
29424 markers are enabled.
29425 (block_within_block_p, dwarf2out_inline_entry): New.
29426 (dwarf2out_finish): Check that no entries remained in
29427 inline_entry_data_table.
29428 * final.c (reemit_insn_block_notes): Handle inline entry notes.
29429 (final_scan_insn, notice_source_line): Likewise.
29430 (rest_of_clean_state): Skip inline entry markers.
29431 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
29433 * gimple.c (gimple_build_debug_inline_entry): New.
29434 * gimple.h (enum gimple_debug_subcode): Add
29435 GIMPLE_DEBUG_INLINE_ENTRY.
29436 (gimple_build_debug_inline_entry): Declare.
29437 (gimple_debug_inline_entry_p): New.
29438 (gimple_debug_nonbind_marker_p): Adjust.
29439 * insn-notes.def (INLINE_ENTRY): New.
29440 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
29441 inline entry marker notes.
29442 (print_insn): Likewise.
29443 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
29444 (INSN_DEBUG_MARKER_KIND): Likewise.
29445 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
29446 * tree-inline.c (expand_call_inline): Build and insert
29447 debug_inline_entry stmt.
29448 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
29449 inline entry blocks early, if nonbind markers are enabled.
29450 (dump_scope_block): Dump fragment info.
29451 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
29452 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
29453 (gimple_build_debug_inline_entry): New.
29454 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
29455 Enable/disable inline entry points too.
29456 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
29457 (DEBUG_INSN): Describe inline entry markers.
29459 * common.opt (gvariable-location-views): New.
29460 (gvariable-location-views=incompat5): New.
29461 * config.in: Rebuilt.
29462 * configure: Rebuilt.
29463 * configure.ac: Test assembler for view support.
29464 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
29465 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
29466 * dwarf2out.c (var_loc_view): New typedef.
29467 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
29468 (dwarf2out_locviews_in_attribute): New.
29469 (dwarf2out_locviews_in_loclist): New.
29470 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
29471 (enum dw_line_info_opcode): Add LI_adv_address.
29472 (struct dw_line_info_table): Add view.
29473 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
29474 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
29475 (zero_view_p): New variable.
29476 (ZERO_VIEW_P): New macro.
29477 (output_asm_line_debug_info): New.
29478 (struct var_loc_node): Add view.
29479 (add_AT_view_list, AT_loc_list): New.
29480 (add_var_loc_to_decl): Add view param. Test it against last.
29481 (new_loc_list): Add view params. Record them.
29482 (AT_loc_list_ptr): Handle loc and view lists.
29483 (view_list_to_loc_list_val_node): New.
29484 (print_dw_val): Handle dw_val_class_view_list.
29485 (size_of_die): Likewise.
29486 (value_format): Likewise.
29487 (loc_list_has_views): New.
29488 (gen_llsym): Set vl_symbol too.
29489 (maybe_gen_llsym, skip_loc_list_entry): New.
29490 (dwarf2out_maybe_output_loclist_view_pair): New.
29491 (output_loc_list): Output view list or entries too.
29492 (output_view_list_offset): New.
29493 (output_die): Handle dw_val_class_view_list.
29494 (output_dwarf_version): New.
29495 (output_compilation_unit_header): Use it.
29496 (output_skeleton_debug_sections): Likewise.
29497 (output_rnglists, output_line_info): Likewise.
29498 (output_pubnames, output_aranges): Update version comments.
29499 (output_one_line_info_table): Output view numbers in asm comments.
29500 (dw_loc_list): Determine current endview, pass it to new_loc_list.
29501 Call maybe_gen_llsym.
29502 (loc_list_from_tree_1): Adjust.
29503 (add_AT_location_description): Create view list attribute if
29504 needed, check it's absent otherwise.
29505 (convert_cfa_to_fb_loc_list): Adjust.
29506 (maybe_emit_file): Call output_asm_line_debug_info for test.
29507 (dwarf2out_var_location): Reset views as needed. Precompute
29508 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
29509 attribute. Set view.
29510 (new_line_info_table): Reset next view.
29511 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
29512 (dwarf2out_source_line): Likewise. Output view resets and labels to
29513 the assembler, or select appropriate line info opcodes.
29514 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
29515 (optimize_string_length): Catch it. Adjust.
29516 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
29517 dw_val_class_view_list, and remove it if no longer needed.
29518 (hash_loc_list): Hash view numbers.
29519 (loc_list_hasher::equal): Compare them.
29520 (optimize_location_lists): Check whether a view list symbol is
29521 needed, and whether the locview attribute is present, and
29522 whether they match. Remove the locview attribute if no longer
29524 (index_location_lists): Call skip_loc_list_entry for test.
29525 (dwarf2out_finish): Call output_asm_line_debug_info for test.
29526 Use output_dwarf_version.
29527 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
29528 (struct dw_val_node): Add val_view_list.
29529 * final.c (SEEN_NEXT_VIEW): New.
29530 (set_next_view_needed): New.
29531 (clear_next_view_needed): New.
29532 (maybe_output_next_view): New.
29533 (final_start_function): Rename to...
29534 (final_start_function_1): ... this. Take pointer to FIRST,
29535 add SEEN parameter. Emit param bindings in the initial view.
29536 (final_start_function): Reintroduce SEEN-less interface.
29537 (final): Rename to...
29538 (final_1): ... this. Take SEEN parameter. Output final pending
29539 next view at the end.
29540 (final): Reintroduce seen-less interface.
29541 (final_scan_insn): Output pending next view before switching
29542 sections or ending a block. Mark the next view as needed when
29543 outputting variable locations. Notify debug backend of section
29544 changes, and of location view changes.
29545 (rest_of_handle_final): Adjust.
29546 * toplev.c (process_options): Autodetect value for debug variable
29547 location views option. Warn on incompat5 without -gdwarf-5.
29548 * doc/invoke.texi (gvariable-location-views): New.
29549 (gvariable-location-views=incompat5): New.
29550 (gno-variable-location-views): New.
29552 2018-02-08 David Malcolm <dmalcolm@redhat.com>
29554 PR tree-optimization/84136
29555 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
29556 that the result of find_edge is non-NULL.
29558 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
29561 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
29562 storing integer register in SImode. Fix cost of 256 and 512
29563 byte aligned SSE register store.
29565 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
29567 * config/i386/i386.c (ix86_multiplication_cost): Fix
29568 multiplication cost for TARGET_AVX512DQ.
29570 2018-02-08 Marek Polacek <polacek@redhat.com>
29572 PR tree-optimization/84238
29573 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
29576 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
29578 PR tree-optimization/84265
29579 * tree-vect-stmts.c (vectorizable_store): Don't treat
29580 VMAT_CONTIGUOUS accesses as grouped.
29581 (vectorizable_load): Likewise.
29583 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
29585 PR tree-optimization/81635
29586 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
29587 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
29588 (test_round_for_mask): New functions.
29589 (wide_int_cc_tests): Call test_round_for_mask.
29590 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
29591 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
29592 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
29593 range returned by get_range_info.
29595 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
29598 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
29599 * symtab.c: Include builtins.h
29600 (symtab_node::output_to_lto_symbol_table_p): Move here
29601 from lto-streamer-out.c:output_symbol_p.
29602 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
29603 (output_symbol_p): Move all logic to symtab.c
29604 (produce_symtab): Update.
29606 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
29608 * config/s390/s390-opts.h (enum indirect_branch): Define.
29609 * config/s390/s390-protos.h (s390_return_addr_from_memory)
29610 (s390_indirect_branch_via_thunk)
29611 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
29612 (enum s390_indirect_branch_type): Define.
29613 * config/s390/s390.c (struct s390_frame_layout, struct
29614 machine_function): Remove.
29615 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
29616 (indirect_branch_table_label_no, indirect_branch_table_name):
29618 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
29619 (enum s390_indirect_branch_option): Define.
29620 (s390_return_addr_from_memory): New function.
29621 (s390_handle_string_attribute): New function.
29622 (s390_attribute_table): Add new attribute handler.
29623 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
29624 (s390_indirect_branch_via_thunk): New function.
29625 (s390_indirect_branch_via_inline_thunk): New function.
29626 (s390_function_ok_for_sibcall): When jumping via thunk disallow
29627 sibling call optimization for non z10 compiles.
29628 (s390_emit_call): Force indirect branch target to be a single
29629 register. Add r1 clobber for non-z10 compiles.
29630 (s390_emit_epilogue): Emit return jump via return_use expander.
29631 (s390_reorg): Handle JUMP_INSNs as execute targets.
29632 (s390_option_override_internal): Perform validity checks for the
29633 new command line options.
29634 (s390_indirect_branch_attrvalue): New function.
29635 (s390_indirect_branch_settings): New function.
29636 (s390_set_current_function): Invoke s390_indirect_branch_settings.
29637 (s390_output_indirect_thunk_function): New function.
29638 (s390_code_end): Implement target hook.
29639 (s390_case_values_threshold): Implement target hook.
29640 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
29642 * config/s390/s390.h (struct s390_frame_layout)
29643 (struct machine_function): Move here from s390.c.
29644 (TARGET_INDIRECT_BRANCH_NOBP_RET)
29645 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
29646 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
29647 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
29648 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
29649 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
29650 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
29651 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
29652 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
29653 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
29654 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
29655 (mnemonic attribute): Add values which aren't recognized
29657 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
29658 pattern for branch conversion. Fix mnemonic attribute.
29659 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
29660 indirect branch via thunk if requested.
29661 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
29662 ("*indirect_jump"): Disable for branch conversion using out of
29664 ("indirect_jump_via_thunk<mode>_z10")
29665 ("indirect_jump_via_thunk<mode>")
29666 ("indirect_jump_via_inlinethunk<mode>_z10")
29667 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
29668 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
29669 ("casesi_jump_via_inlinethunk<mode>_z10")
29670 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
29671 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
29672 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
29673 ("*indirect2_jump"): Disable for branch conversion.
29674 ("casesi_jump"): Turn into expander and expand patterns for branch
29676 ("return_use"): New expander.
29677 ("*return"): Emit return via thunk and rename it to ...
29678 ("*return<mode>"): ... this one.
29679 * config/s390/s390.opt: Add new options and and enum for the
29682 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
29684 * lra-constraints.c (match_reload): Unconditionally use
29685 gen_lowpart_SUBREG, rather than selecting between that
29686 and equivalent gen_rtx_SUBREG code.
29688 2018-02-08 Richard Biener <rguenther@suse.de>
29690 PR tree-optimization/84233
29691 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
29692 changed flag instead of boguously re-using phi_inserted.
29694 2018-02-08 Martin Jambor <mjambor@suse.cz>
29696 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
29697 static local variables.
29699 2018-02-08 Richard Biener <rguenther@suse.de>
29701 PR tree-optimization/84278
29702 * tree-vect-stmts.c (vectorizable_store): When looking for
29703 smaller vector types to perform grouped strided loads/stores
29704 make sure the mode is supported by the target.
29705 (vectorizable_load): Likewise.
29707 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
29709 * config/aarch64/aarch64.c (aarch64_components_for_bb):
29710 Increase LDP/STP opportunities by adding adjacent callee-saves.
29712 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
29714 PR rtl-optimization/84068
29715 PR rtl-optimization/83459
29716 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
29718 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
29720 PR tree-optimization/84224
29721 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
29722 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
29723 non-zero arguments.
29725 2018-02-07 Iain Sandoe <iain@codesourcery.com>
29728 * config/rs6000/altivec.md (*restore_world): Remove LR use.
29729 * config/rs6000/predicates.md (restore_world_operation): Adjust op
29730 count, remove one USE.
29732 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
29734 * doc/install.texi (Configuration): Document the
29735 --with-long-double-format={ibm,ieee} PowerPC configuration
29739 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
29740 Convert from define_expand to be define_insn_and_split. Rework
29741 float/double/_Float128 conversions to QI/HI/SImode to work with
29742 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
29743 conversions to QI/HImode types did a store and then a load to
29744 truncate the value. For conversions to VSX registers, don't split
29745 the insn, instead emit the code directly. Use the code iterator
29746 any_fix to combine signed and unsigned conversions.
29747 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
29748 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
29749 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
29750 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
29751 (fix_<mode>di2_hw): Likewise.
29752 (fixuns_<mode>di2_hw): Likewise.
29753 (fix_<mode>si2_hw): Likewise.
29754 (fixuns_<mode>si2_hw): Likewise.
29755 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
29756 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
29757 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
29758 fix<uns>_trunc<SFDF:mode>si2_p8.
29759 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
29760 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
29761 (fix<uns>_<mode>_mem): Likewise.
29762 (fctiw<u>z_<mode>_mem): Likewise.
29763 (fix<uns>_<mode>_mem): Likewise.
29764 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
29765 the register allocator from doing a direct move to the GPRs to do
29766 a store, and instead use the ISA 3.0 store byte/half-word from
29767 vector register instruction. For IEEE 128-bit floating point,
29768 also optimize stores of 32-bit ints.
29769 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
29771 2018-02-07 Alan Hayward <alan.hayward@arm.com>
29773 * genextract.c (push_pathstr_operand): New function to support
29775 (walk_rtx): Call push_pathstr_operand.
29776 (print_path): Support [a-zA-Z].
29778 2018-02-07 Richard Biener <rguenther@suse.de>
29780 PR tree-optimization/84037
29781 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
29782 (cse_and_gimplify_to_preheader): Declare.
29783 (vect_get_place_in_interleaving_chain): Likewise.
29784 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
29786 (_loop_vec_info::~_loop_vec_info): Delete it.
29787 (cse_and_gimplify_to_preheader): New function.
29788 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
29789 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
29790 (vectorizable_load): Likewise. For grouped stores always base
29791 the IV on the first element.
29792 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
29793 condition before gimplifying.
29795 2018-02-07 Jakub Jelinek <jakub@redhat.com>
29797 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
29798 *DIV_EXPR and *MOD_EXPR.
29800 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
29803 * config/i386/i386.c (ix86_option_override_internal): Mask out
29804 the CF_SET bit when checking -fcf-protection.
29806 2018-02-07 Tom de Vries <tom@codesourcery.com>
29809 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
29812 2018-02-07 Richard Biener <rguenther@suse.de>
29814 PR tree-optimization/84204
29815 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
29818 PR tree-optimization/84205
29819 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
29820 special-case isl_ast_op_zdiv_r.
29822 PR tree-optimization/84223
29823 * graphite-scop-detection.c (gather_bbs::before_dom_children):
29824 Only add conditions from within the region.
29825 (gather_bbs::after_dom_children): Adjust.
29827 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
29830 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
29831 * config/avr/avr.md: Only post-reload split REG-REG moves if
29832 either register is GENERAL_REG_P.
29834 2018-02-07 Jakub Jelinek <jakub@redhat.com>
29836 PR tree-optimization/84235
29837 * tree-ssa-scopedtables.c
29838 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
29839 if the subtraction is performed in floating point type where NaNs are
29840 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
29841 build 1. Formatting fix.
29843 2018-02-06 Jakub Jelinek <jakub@redhat.com>
29846 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
29847 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
29848 and skip it regardless of bb boundaries. Use CALL_P macro,
29849 don't test INSN_P (insn) together with CALL_P or JUMP_P check
29850 unnecessarily, formatting fix.
29852 2018-02-06 Michael Collison <michael.collison@arm.com>
29854 * config/arm/thumb2.md:
29855 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
29856 (*thumb_mov_notscc): Ditto.
29858 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
29861 * config/rs6000/rs6000.md (su code attribute): Use "u" for
29862 unsigned_fix, not "s".
29864 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
29866 * configure.ac (gcc_fn_eh_frame_ro): New function.
29867 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
29868 correct .eh_frame permissions.
29869 * configure: Regenerate.
29871 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
29873 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
29874 irrelevant options.
29876 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
29878 * config/rs6000/rs6000.c (rs6000_option_override_internal):
29879 Display warning message for -mno-speculate-indirect-jumps.
29881 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
29883 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
29885 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
29887 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
29889 PR tree-optimization/84225
29890 * tree-eh.c (find_trapping_overflow): Only call
29891 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
29893 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
29896 * config/i386/i386.c: Reimplement the check of possible options
29897 -mibt/-mshstk conbination. Change error messages.
29898 * doc/invoke.texi: Fix a typo: remove extra '='.
29900 2018-02-06 Marek Polacek <polacek@redhat.com>
29902 PR tree-optimization/84228
29903 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
29905 2018-02-06 Tamar Christina <tamar.christina@arm.com>
29908 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
29909 emitted arch directives.
29910 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
29911 __ARM_FEATURE_COPROC before changing architectures.
29913 2018-02-06 Richard Biener <rguenther@suse.de>
29915 * config/i386/i386.c (print_reg): Fix typo.
29916 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
29918 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
29920 * configure: Regenerate.
29922 2018-02-05 Martin Sebor <msebor@redhat.com>
29924 PR tree-optimization/83369
29925 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
29928 2018-02-05 Martin Liska <mliska@suse.cz>
29930 * doc/invoke.texi: Cherry-pick upstream r323995.
29932 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
29934 * ira.c (ira_init_register_move_cost): Adjust comment.
29936 2018-02-05 Martin Liska <mliska@suse.cz>
29938 PR gcov-profile/84137
29939 * doc/gcov.texi: Fix typo in documentation.
29941 2018-02-05 Martin Liska <mliska@suse.cz>
29943 PR gcov-profile/83879
29944 * doc/gcov.texi: Document necessity of --dynamic-list-data when
29945 using dlopen functionality.
29947 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
29949 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
29950 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
29951 _mm_maskz_range_ss, _mm_mask_range_round_ss,
29952 _mm_maskz_range_round_ss): New intrinsics.
29953 (__builtin_ia32_rangesd128_round)
29954 (__builtin_ia32_rangess128_round): Remove.
29955 (__builtin_ia32_rangesd128_mask_round,
29956 __builtin_ia32_rangess128_mask_round): New builtins.
29957 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
29958 __builtin_ia32_rangess128_round): Remove.
29959 (__builtin_ia32_rangesd128_mask_round,
29960 __builtin_ia32_rangess128_mask_round): New builtins.
29961 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
29962 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
29963 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
29964 "<round_saeonly_constraint>")): Changed to ...
29965 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
29966 "<round_saeonly_scalar_constraint>")): ... this.
29967 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
29968 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
29969 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
29970 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
29971 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
29973 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
29975 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
29977 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
29978 Remove all values except native, 8540 and 8548.
29980 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
29982 * config/i386/i386.c (ix86_output_function_return): Pass
29983 INVALID_REGNUM, instead of -1, as invalid register number to
29984 indirect_thunk_name and output_indirect_thunk.
29986 2018-02-02 Julia Koval <julia.koval@intel.com>
29988 * config.gcc: Add -march=icelake.
29989 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
29990 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
29991 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
29992 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
29993 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
29994 (processor_target_table): Add icelake.
29995 (ix86_option_override_internal): Handle new PTAs.
29996 (get_builtin_code_for_version): Handle icelake.
29997 (M_INTEL_COREI7_ICELAKE): New.
29998 (fold_builtin_cpu): Handle icelake.
29999 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
30000 * doc/invoke.texi: Add -march=icelake.
30002 2018-02-02 Julia Koval <julia.koval@intel.com>
30004 * config/i386/i386.c (ix86_option_override_internal): Change flags type
30005 to wide_int_bitmask.
30006 * wide-int-bitmask.h: New.
30008 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
30011 * config/i386/i386.md: Replace Pmode with word_mode in
30012 builtin_setjmp_setup and builtin_longjmp to support x32.
30014 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
30018 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
30020 (rs6000_supported_cpu_names): New static variable.
30021 (linux_cpu_translation_table): Likewise.
30022 (elf_platform) <cpu>: Define new static variable and use it.
30023 Translate kernel AT_PLATFORM name to canonical name if needed.
30024 Error if platform name is unknown.
30026 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
30029 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
30031 2018-02-01 Jeff Law <law@redhat.com>
30034 * config/i386/i386.c (release_scratch_register_on_entry): Add new
30035 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
30036 the scratch if RELEASE_VIA_POP is false.
30037 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
30038 If we have to save a temporary register, decrement SIZE appropriately.
30039 Pass new arguments to release_scratch_register_on_entry.
30040 (ix86_adjust_stack_and_probe): Likewise.
30041 (ix86_emit_probe_stack_range): Pass new arguments to
30042 release_scratch_register_on_entry.
30044 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
30046 PR rtl-optimization/84157
30047 * combine.c (change_zero_ext): Use REG_P predicate in
30048 front of HARD_REGISTER_P predicate.
30050 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
30052 * config/avr/avr.c (avr_option_override): Move disabling of
30053 -fdelete-null-pointer-checks to...
30054 * common/config/avr/avr-common.c (avr_option_optimization_table):
30057 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
30059 PR tree-optimization/81635
30060 * tree-data-ref.c (split_constant_offset_1): For types that
30061 wrap on overflow, try to use range info to prove that wrapping
30064 2018-02-01 Renlin Li <renlin.li@arm.com>
30067 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
30068 TAILCALL_ADDR_REGS.
30069 (aarch64_register_move_cost): Likewise.
30070 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
30071 TAILCALL_ADDR_REGS.
30072 (REG_CLASS_NAMES): Likewise.
30073 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
30074 TAILCALL_ADDR_REGS. Remove IP registers.
30075 * config/aarch64/aarch64.md (Ucs): Update register constraint.
30077 2018-02-01 Richard Biener <rguenther@suse.de>
30079 * domwalk.h (dom_walker::dom_walker): Add additional constructor
30080 for specifying RPO order and allow NULL for that.
30081 * domwalk.c (dom_walker::dom_walker): Likewise.
30082 (dom_walker::walk): Handle NULL RPO order.
30083 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
30085 (rewrite_update_dom_walker): Likewise.
30086 (mark_def_dom_walker): Likewise.
30088 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
30090 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
30091 (aarch64_maybe_expand_sve_subreg_move): Declare.
30092 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
30093 * config/aarch64/predicates.md (aarch64_any_register_operand): New
30095 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
30096 that are semantically a reverse operation.
30097 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
30098 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
30099 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
30101 (aarch64_can_change_mode_class): For big-endian, forbid changes
30102 between two SVE modes if they have different element sizes.
30104 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
30106 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
30107 the TImode handling for big-endian targets.
30109 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
30111 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
30112 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
30114 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
30115 Remove BSWAP handing for big-endian targets and use the form of
30116 LD1RQ appropariate for the mode.
30118 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
30120 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
30121 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
30122 duplicated element.
30124 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
30127 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
30128 check for operands that need to go through aarch64_sve_reload_be.
30130 2018-02-01 Jakub Jelinek <jakub@redhat.com>
30132 PR tree-optimization/81661
30133 PR tree-optimization/84117
30134 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
30135 * tree-eh.c: Include gimplify.h.
30136 (find_trapping_overflow, replace_trapping_overflow,
30137 rewrite_to_non_trapping_overflow): New functions.
30138 * tree-vect-loop.c: Include tree-eh.h.
30139 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
30140 * tree-data-ref.c: Include tree-eh.h.
30141 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
30143 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
30145 PR rtl-optimization/84123
30146 * combine.c (change_zero_ext): Check if hard register satisfies
30147 can_change_dest_mode before calling gen_lowpart_SUBREG.
30149 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
30152 * ira.c (ira_init_register_move_cost): Remove assert.
30154 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
30156 PR rtl-optimization/84071
30157 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
30158 * doc/tm.texi: Regenerate.
30160 2018-01-31 Richard Biener <rguenther@suse.de>
30162 PR tree-optimization/84132
30163 * tree-data-ref.c (analyze_miv_subscript): Properly
30164 check whether evolution_function_is_affine_multivariate_p
30165 before calling gcd_of_steps_may_divide_p.
30167 2018-01-31 Julia Koval <julia.koval@intel.com>
30170 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
30171 * config/i386/i386.md (rdpid_rex64) New.
30172 (rdpid): Make 32bit only.
30174 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
30177 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
30178 an IDENTIFIER_NODE for FUNCTION_TYPE's.
30180 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
30183 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
30185 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
30187 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
30189 PR rtl-optimization/84071
30190 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
30191 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
30193 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
30195 * config/arc/arc.c (arc_handle_aux_attribute): New function.
30196 (arc_attribute_table): Add 'aux' attribute.
30197 (arc_in_small_data_p): Consider aux like variables.
30198 (arc_is_aux_reg_p): New function.
30199 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
30200 (arc_get_aux_arg): New function.
30201 (prepare_move_operands): Handle aux-register access.
30202 (arc_handle_aux_attribute): New function.
30203 * doc/extend.texi (ARC Variable attributes): Add subsection.
30205 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
30207 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
30208 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
30209 (arc_attribute_table): Add 'uncached' attribute.
30210 (arc_print_operand): Print '.di' flag for uncached memory
30212 (arc_in_small_data_p): Do not consider for small data the uncached
30214 (arc_is_uncached_mem_p): New function.
30215 * config/arc/predicates.md (compact_store_memory_operand): Check
30216 for uncached memory accesses.
30217 (nonvol_nonimm_operand): Likewise.
30218 * doc/extend.texi (ARC Type Attribute): New subsection.
30220 2018-01-31 Jakub Jelinek <jakub@redhat.com>
30223 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
30224 falign-loops=): Add Optimization flag.
30226 2018-01-30 Jeff Law <law@redhat.com>
30229 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
30230 INT_REGISTERS_SAVED. Check it prior to calling
30231 get_scratch_register_on_entry.
30232 (ix86_adjust_stack_and_probe): Similarly.
30233 (ix86_emit_probe_stack_range): Similarly.
30234 (ix86_expand_prologue): Corresponding changes.
30236 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
30239 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
30240 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
30242 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
30245 * lra-constraints.c (curr_insn_transform): Process AND in the
30248 2018-01-30 Jakub Jelinek <jakub@redhat.com>
30250 PR rtl-optimization/83986
30251 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
30252 dependence against last_pending_memory_flush in addition to
30253 pending_jump_insns.
30255 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
30257 PR tree-optimization/81611
30258 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
30261 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
30264 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
30267 2018-01-30 Richard Biener <rguenther@suse.de>
30268 Jakub Jelinek <jakub@redhat.com>
30270 PR tree-optimization/84111
30271 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
30272 inner loops added during recursion, as they don't have up-to-date
30275 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
30278 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
30279 (can_inline_edge_by_limits_p): ... here.
30280 (can_early_inline_edge_p, check_callers,
30281 update_caller_keys, update_callee_keys, recursive_inlining,
30282 add_new_edges_to_heap, speculation_useful_p,
30283 inline_small_functions,
30284 inline_small_functions, flatten_function,
30285 inline_to_all_callers_1): Update.
30287 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
30289 * profile-count.c (profile_count::combine_with_ipa_count): Handle
30292 2018-01-30 Richard Biener <rguenther@suse.de>
30294 PR tree-optimization/83008
30295 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
30296 invariant and constant vector uses in stmts when they need
30297 more than one stmt.
30299 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
30302 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
30303 * configure: Regenerate.
30305 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
30307 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
30309 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
30310 Use gen_rtx_REG rather than gen_lowpart.
30312 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
30314 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
30315 rather than 0 when creating partial subregs.
30317 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
30319 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
30322 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
30325 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
30326 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
30327 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
30328 flags. This restores the settings used before the 2017-07-24.
30329 Turning off pre increment/decrement/modify allows IVOPTS to
30330 optimize DF/SF loops where the index is an int.
30332 2018-01-29 Richard Biener <rguenther@suse.de>
30333 Kelvin Nilsen <kelvin@gcc.gnu.org>
30336 * tree-vect-stmts.c (vectorizable_call): Don't call
30337 targetm.vectorize_builtin_md_vectorized_function if callee is
30340 2018-01-22 Carl Love <cel@us.ibm.com>
30342 * doc/extend.tex: Fix typo in second arg in
30343 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
30345 2018-01-29 Richard Biener <rguenther@suse.de>
30347 PR tree-optimization/84086
30348 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
30349 (flush_ssaname_freelist): When SSA names were released reset
30350 the SCEV hash table.
30352 2018-01-29 Richard Biener <rguenther@suse.de>
30354 PR tree-optimization/84057
30355 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
30356 removed paths when removing edges.
30358 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
30360 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
30361 -mfunction-return=@var{choice}.
30363 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
30365 PR diagnostic/84034
30366 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
30367 Handle CR like TAB.
30368 (layout::print_source_line): Likewise.
30369 (test_get_line_width_without_trailing_whitespace): Add test cases.
30371 2018-01-27 Jakub Jelinek <jakub@redhat.com>
30373 PR middle-end/84040
30374 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
30377 2018-01-26 Jim Wilson <jimw@sifive.com>
30379 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
30381 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
30384 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30386 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
30387 and CMP + SUB-immediate -> SUBS.
30389 2018-01-26 Martin Sebor <msebor@redhat.com>
30391 PR tree-optimization/83896
30392 * tree-ssa-strlen.c (get_string_len): Rename...
30393 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
30394 Avoid assuming length is constant.
30395 (handle_char_store): Use HOST_WIDE_INT for string length.
30397 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
30400 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
30401 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
30403 2018-01-26 Richard Biener <rguenther@suse.de>
30405 PR rtl-optimization/84003
30406 * dse.c (record_store): Only record redundant stores when
30407 the earlier store aliases at least all accesses the later one does.
30409 2018-01-26 Jakub Jelinek <jakub@redhat.com>
30411 PR rtl-optimization/83985
30412 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
30413 REG_CFA_RESTORE insns.
30414 (delete_unmarked_insns): Don't ignore separate shrink wrapping
30415 REG_CFA_RESTORE insns here.
30418 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
30419 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
30421 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
30423 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
30424 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
30425 (arc_init): Likewise.
30426 (arc_override_options): Likewise.
30427 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
30429 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
30431 * config/arc/arc.h (TARGET_DBNZ): Define.
30432 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
30433 properly set the tune attribute.
30434 (dbnz): Use TARGET_DBNZ guard.
30435 * config/arc/arc.opt (mtune): Add core3 option.
30437 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
30439 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
30440 recognize new pic like addresses.
30441 (arc_delegitimize_address): Clean up.
30443 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
30445 * config/arc/arc-arches.def: Option mrf16 valid for all
30447 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
30448 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
30449 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
30450 * config/arc/arc-tables.opt: Regenerate.
30451 * config/arc/arc.c (arc_conditional_register_usage): Handle
30452 reduced register file case.
30453 (arc_file_start): Set must have build attributes.
30454 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
30455 mrf16 option value.
30456 * config/arc/arc.opt (mrf16): Add new option.
30457 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
30458 * config/arc/genmultilib.awk: Handle new mrf16 option.
30459 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
30460 * config/arc/t-multilib: Regenerate.
30461 * doc/invoke.texi (ARC Options): Document mrf16 option.
30463 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
30465 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
30466 * config/arc/arc.c (arc_handle_secure_attribute): New function.
30467 (arc_attribute_table): Add 'secure_call' attribute.
30468 (arc_print_operand): Print secure call operand.
30469 (arc_function_ok_for_sibcall): Don't optimize tail calls when
30471 (arc_is_secure_call_p): New function. * config/arc/arc.md
30472 (call_i): Add support for sjli instruction.
30473 (call_value_i): Likewise.
30474 * config/arc/constraints.md (Csc): New constraint.
30476 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
30477 John Eric Martin <John.Martin@emmicro-us.com>
30479 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
30480 * config/arc/arc.c (_arc_jli_section): New struct.
30481 (arc_jli_section): New type.
30482 (rc_jli_sections): New static variable.
30483 (arc_handle_jli_attribute): New function.
30484 (arc_attribute_table): Add jli_always and jli_fixed attribute.
30485 (arc_file_end): New function.
30486 (TARGET_ASM_FILE_END): Define.
30487 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
30488 (arc_add_jli_section): New function.
30489 (jli_call_scan): Likewise.
30490 (arc_reorg): Call jli_call_scan.
30491 (arc_output_addsi): Remove 'S' from printing asm operand.
30492 (arc_is_jli_call_p): New function.
30493 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
30495 (movhi_insn): Likewise.
30496 (movsi_insn): Likewise.
30497 (movsi_set_cc_insn): Likewise.
30498 (loadqi_update): Likewise.
30499 (load_zeroextendqisi_update): Likewise.
30500 (load_signextendqisi_update): Likewise.
30501 (loadhi_update): Likewise.
30502 (load_zeroextendhisi_update): Likewise.
30503 (load_signextendhisi_update): Likewise.
30504 (loadsi_update): Likewise.
30505 (loadsf_update): Likewise.
30506 (movsicc_insn): Likewise.
30507 (bset_insn): Likewise.
30508 (bxor_insn): Likewise.
30509 (bclr_insn): Likewise.
30510 (bmsk_insn): Likewise.
30511 (bicsi3_insn): Likewise.
30512 (cmpsi_cc_c_insn): Likewise.
30513 (movsi_ne): Likewise.
30514 (movsi_cond_exec): Likewise.
30515 (clrsbsi2): Likewise.
30516 (norm_f): Likewise.
30526 (call_i): Remove 'S' asm letter, add jli instruction.
30527 (call_value_i): Likewise.
30528 * config/arc/arc.op (mjli-always): New option.
30529 * config/arc/constraints.md (Cji): New constraint.
30530 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
30532 (subsf3_fpx): Likewise.
30533 (mulsf3_fpx): Likewise.
30534 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
30536 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
30537 function attrbutes.
30538 * doc/invoke.texi (ARC): Document mjli-always option.
30540 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
30542 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
30543 avoid addition with 0 and use incw and decw where possible.
30545 2018-01-26 Richard Biener <rguenther@suse.de>
30547 PR tree-optimization/81082
30548 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
30549 association if it requires casting to unsigned.
30550 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
30551 from fold_plusminus_mult_expr to catch important cases late when
30552 range info is available.
30554 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
30556 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
30557 * configure.ac (hidden_linkonce): New test.
30558 * configure: Regenerate.
30559 * config.in: Regenerate.
30561 2018-01-26 Julia Koval <julia.koval@intel.com>
30563 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
30564 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
30565 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
30566 _mm_mask_bitshuffle_epi64_mask): Fix type.
30567 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
30568 USI_FTYPE_V4DI_V4DI_USI): Remove.
30569 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
30570 __builtin_ia32_vpshufbitqmb256_mask,
30571 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
30572 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
30573 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
30575 2018-01-26 Alan Modra <amodra@gmail.com>
30578 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
30579 UNSPEC_VBPERMQ. Sort other unspecs.
30581 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
30583 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
30585 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
30587 PR middle-end/83055
30588 * predict.c (drop_profile): Do not push/pop cfun; update also
30590 (handle_missing_profiles): Fix logic looking for zero profiles.
30592 2018-01-25 Jakub Jelinek <jakub@redhat.com>
30594 PR middle-end/83977
30595 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
30596 on functions with #pragma omp declare simd or functions with simd
30598 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
30599 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
30600 Remove trailing \n from warning_at calls.
30602 2018-01-25 Tom de Vries <tom@codesourcery.com>
30605 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
30606 for neutered workers.
30608 2018-01-24 Joseph Myers <joseph@codesourcery.com>
30611 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
30612 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
30614 2018-01-24 Jeff Law <law@redhat.com>
30617 * i386.c (get_probe_interval): Move to earlier point.
30618 (ix86_compute_frame_layout): If -fstack-clash-protection and
30619 the frame is larger than the probe interval, then use pushes
30620 to save registers rather than reg->mem moves.
30621 (ix86_expand_prologue): Remove conditional for int_registers_saved
30624 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
30627 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
30628 min/max for never referenced object.
30630 2018-01-24 Jakub Jelinek <jakub@redhat.com>
30632 PR middle-end/83977
30633 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
30635 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
30636 attributes from DECL_ATTRIBUTES (decl) without affecting
30637 DECL_ATTRIBUTES (current_function_decl).
30638 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
30639 functions with non-NULL DECL_ABSTRACT_ORIGIN.
30641 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
30643 PR tree-optimization/83979
30644 * fold-const.c (fold_comparison): Use constant_boolean_node
30645 instead of boolean_{true,false}_node.
30647 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
30649 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
30652 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
30654 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
30655 Simplify the clause that sets the length attribute.
30656 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
30657 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
30658 clause that sets the length attribute.
30659 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
30661 2018-01-24 Tom de Vries <tom@codesourcery.com>
30664 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
30665 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
30666 Add strict parameter.
30667 (prevent_branch_around_nothing): Insert dummy insn between branch to
30668 label and label with no ptx insn inbetween.
30669 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
30671 2018-01-24 Tom de Vries <tom@codesourcery.com>
30674 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
30675 for neutered threads in warp.
30676 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
30678 2018-01-24 Richard Biener <rguenther@suse.de>
30680 PR tree-optimization/83176
30681 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
30684 2018-01-24 Richard Biener <rguenther@suse.de>
30686 PR tree-optimization/82819
30687 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
30688 code generating pluses that are no-ops in the target precision.
30690 2018-01-24 Richard Biener <rguenther@suse.de>
30692 PR middle-end/84000
30693 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
30695 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
30697 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
30698 to merge probabilities.
30699 * predict.c (probably_never_executed): Also mark as cold functions
30700 with global 0 profile and guessed local profile.
30701 * profile-count.c (profile_probability::combine_with_count): New
30703 * profile-count.h (profile_probability::operator*,
30704 profile_probability::operator*=, profile_probability::operator/,
30705 profile_probability::operator/=): Reduce precision to adjusted
30706 and set value to guessed on contradictory divisions.
30707 (profile_probability::combine_with_freq): Remove.
30708 (profile_probability::combine_wiht_count): Declare.
30709 (profile_count::force_nonzero):: Set to adjusted.
30710 (profile_count::probability_in):: Set quality to adjusted.
30711 * tree-ssa-tail-merge.c (replace_block_by): Use
30712 combine_with_count.
30714 2018-01-23 Andrew Waterman <andrew@sifive.com>
30715 Jim Wilson <jimw@sifive.com>
30717 * config/riscv/riscv.c (riscv_stack_boundary): New.
30718 (riscv_option_override): Set riscv_stack_boundary. Handle
30719 riscv_preferred_stack_boundary_arg.
30720 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
30721 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
30722 (STACK_BOUNDARY): Set to riscv_stack_boundary.
30723 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
30724 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
30725 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
30727 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
30730 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
30731 of struct ix86_frame.
30732 (ix86_expand_epilogue): Likewise. Add a local variable for
30733 the reg_save_offset field in struct ix86_frame.
30735 2018-01-23 Bin Cheng <bin.cheng@arm.com>
30737 PR tree-optimization/82604
30738 * tree-loop-distribution.c (enum partition_kind): New enum item
30739 PKIND_PARTIAL_MEMSET.
30740 (partition_builtin_p): Support above new enum item.
30741 (generate_code_for_partition): Ditto.
30742 (compute_access_range): Differentiate cases that equality can be
30743 proven at all loops, the innermost loops or no loops.
30744 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
30745 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
30746 (finalize_partitions, distribute_loop): Don't fuse partition of
30747 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
30748 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
30749 parloop is enabled.
30751 2018-01-23 Martin Liska <mliska@suse.cz>
30753 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
30754 order to ignore the predictor.
30755 (PRED_POLYMORPHIC_CALL): Likewise.
30756 (PRED_RECURSIVE_CALL): Likewise.
30758 2018-01-23 Martin Liska <mliska@suse.cz>
30760 * tree-profile.c (tree_profiling): Print function header to
30761 aware reader which function we are working on.
30762 * value-prof.c (gimple_find_values_to_profile): Do not print
30763 not interesting value histograms.
30765 2018-01-23 Martin Liska <mliska@suse.cz>
30767 * profile-count.h (enum profile_quality): Add
30768 profile_uninitialized as the first value. Do not number values
30769 as they are zero based.
30770 (profile_count::verify): Update sanity check.
30771 (profile_probability::verify): Likewise.
30773 2018-01-23 Nathan Sidwell <nathan@acm.org>
30775 * doc/invoke.texi (ffor-scope): Deprecate.
30777 2018-01-23 David Malcolm <dmalcolm@redhat.com>
30779 PR tree-optimization/83510
30780 * domwalk.c (set_all_edges_as_executable): New function.
30781 (dom_walker::dom_walker): Convert bool param
30782 "skip_unreachable_blocks" to enum reachability. Move setup of
30783 edge flags to set_all_edges_as_executable and only do it when
30784 reachability is REACHABLE_BLOCKS.
30785 * domwalk.h (enum dom_walker::reachability): New enum.
30786 (dom_walker::dom_walker): Convert bool param
30787 "skip_unreachable_blocks" to enum reachability.
30788 (set_all_edges_as_executable): New decl.
30789 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
30790 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
30792 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
30793 but converting true to REACHABLE_BLOCKS.
30794 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
30796 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
30797 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
30798 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
30800 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
30801 if check_all_array_refs will be called.
30803 2018-01-23 David Malcolm <dmalcolm@redhat.com>
30805 * tree.c (selftest::test_location_wrappers): Add more test
30808 2018-01-23 David Malcolm <dmalcolm@redhat.com>
30810 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
30811 (selftest::test_bit_in_range): Likewise.
30813 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
30816 * doc/sourcebuild.texi (vect_float): Say that the selector
30817 only describes the situation when -funsafe-math-optimizations is on.
30818 (vect_float_strict): Document.
30820 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
30822 PR tree-optimization/83965
30823 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
30824 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
30825 instead of checking only for a reduction.
30826 (vect_recog_widen_sum_pattern): Likewise.
30828 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
30830 * predict.c (probably_never_executed): Only use precise profile info.
30831 (compute_function_frequency): Skip after inlining hack since we now
30832 have quality checking.
30834 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
30836 * profile-count.h (profile_probability::very_unlikely,
30837 profile_probability::unlikely, profile_probability::even): Set
30838 precision to guessed.
30840 2018-01-23 Richard Biener <rguenther@suse.de>
30842 PR tree-optimization/83963
30843 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
30844 Properly terminate dominator walk when crossing the exit edge not
30845 when visiting its source block.
30847 2018-01-23 Jakub Jelinek <jakub@redhat.com>
30850 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
30851 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
30853 2018-01-22 Jakub Jelinek <jakub@redhat.com>
30855 PR tree-optimization/83957
30856 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
30857 semicolon after for body surrounded by braces.
30859 PR tree-optimization/83081
30860 * profile-count.h (profile_probability::split): New method.
30861 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
30862 Use profile_probability::split.
30863 (do_compare_rtx_and_jump): Fix adjustment of probabilities
30864 when splitting a single conditional jump into 2.
30866 2018-01-22 David Malcolm <dmalcolm@redhat.com>
30868 PR tree-optimization/69452
30869 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
30872 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
30874 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
30875 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
30876 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
30878 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
30880 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
30882 * config/rl78/rl78.md (movdi): New define_expand.
30883 * config/rl78/rl78.c (rl78_split_movdi): New function.
30885 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
30888 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
30890 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
30891 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
30892 128-bit to produce an UNSPEC move to get the double word with the
30893 signbit and then a shift directly to do signbit.
30894 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
30895 implementation with a new version that just does either a direct
30896 move or a regular move. Move memory interface to separate insns.
30897 Move insns so they are next to the expander.
30898 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
30899 with signbit move. Split big and little endian case.
30900 (signbit<mode>2_dm_mem_le): Likewise.
30901 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
30902 (signbit<mode>2_dm2): Likewise.
30904 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
30906 * config/rl78/rl78.md (anddi3): New define_expand.
30908 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
30910 * config/rl78/rl78.md (umindi3): New define_expand.
30912 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
30914 * config/rl78/rl78.md (smindi3): New define_expand.
30916 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
30918 * config/rl78/rl78.md (smaxdi3): New define_expand.
30920 2018-01-22 Carl Love <cel@us.ibm.com>
30922 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
30923 LVX_V1TI): Add macro expansion.
30924 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
30925 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
30926 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
30927 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
30928 Change check to determine if the instruction is a byte reversing
30929 entry. Fix typo in comment.
30930 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
30931 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
30932 Add def_builtin calls for new builtins.
30933 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
30934 Add define_insn expansion.
30936 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
30938 * config/rl78/rl78.md (umaxdi3): New define_expand.
30940 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
30942 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
30943 for non-QImode registers.
30945 2018-01-22 Richard Biener <rguenther@suse.de>
30947 PR tree-optimization/83963
30948 * graphite-scop-detection.c (scop_detection::get_sese): Delay
30949 including the loop exit block.
30950 (scop_detection::merge_sese): Likewise.
30951 (scop_detection::add_scop): Do it here instead.
30953 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30955 * doc/sourcebuild.texi (arm_softfloat): Document.
30957 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
30960 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
30961 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
30962 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
30964 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
30965 David Edelsohn <dje.gcc@gmail.com>
30968 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
30969 Change "crset eq" to "crset 2".
30970 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
30971 (*call_indirect_aix<mode>_nospec): Likewise.
30972 (*call_value_indirect_aix<mode>_nospec): Likewise.
30973 (*call_indirect_elfv2<mode>_nospec): Likewise.
30974 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
30975 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
30976 change assembly output from . to $.
30977 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
30978 (indirect_jump<mode>_nospec): Change assembly output from . to $.
30979 (*tablejump<mode>_internal1_nospec): Likewise.
30981 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
30984 * config/sh/sh_optimize_sett_clrt.cc:
30985 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
30987 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
30989 PR tree-optimization/83940
30990 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
30991 offset_dt to vect_constant_def rather than vect_unknown_def_type.
30992 (vect_check_load_store_mask): Add a mask_dt_out parameter and
30993 use it to pass back the definition type.
30994 (vect_check_store_rhs): Likewise rhs_dt_out.
30995 (vect_build_gather_load_calls): Add a mask_dt argument and use
30996 it instead of a call to vect_is_simple_use.
30997 (vectorizable_store): Update calls to vect_check_load_store_mask
30998 and vect_check_store_rhs. Use the dt returned by the latter instead
30999 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
31000 instead of calls to vect_is_simple_use. Pass the scalar rather
31001 than the vector operand to vect_is_simple_use when handling
31002 second and subsequent copies of an rhs value.
31003 (vectorizable_load): Update calls to vect_check_load_store_mask
31004 and vect_build_gather_load_calls. Use the cached mask_dt and
31005 gs_info.offset_dt instead of calls to vect_is_simple_use.
31007 2018-01-20 Jakub Jelinek <jakub@redhat.com>
31009 PR middle-end/83945
31010 * tree-emutls.c: Include gimplify.h.
31011 (lower_emutls_2): New function.
31012 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
31013 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
31014 it before further processing.
31017 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
31018 UINTVAL (trueop1) instead of INTVAL (op1).
31020 2018-01-19 Jakub Jelinek <jakub@redhat.com>
31024 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
31025 INCOMING_FRAME_SP_OFFSET if not defined.
31026 (scan_trace): Add ENTRY argument. If true and
31027 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
31028 emit a note to adjust the CFA offset.
31029 (create_cfi_notes): Adjust scan_trace callers.
31030 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
31031 INCOMING_FRAME_SP_OFFSET in the CIE.
31032 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
31033 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
31035 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
31036 * doc/tm.texi: Regenerated.
31038 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
31040 PR rtl-optimization/83147
31041 * lra-constraints.c (remove_inheritance_pseudos): Use
31042 lra_substitute_pseudo_within_insn.
31044 2018-01-19 Tom de Vries <tom@codesourcery.com>
31045 Cesar Philippidis <cesar@codesourcery.com>
31048 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
31050 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
31053 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
31054 spaces for function labels.
31056 2018-01-19 Martin Liska <mliska@suse.cz>
31058 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
31059 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
31060 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
31061 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
31062 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
31063 (PRED_CONST_RETURN): Change from 69 to 65.
31064 (PRED_NULL_RETURN): Change from 91 to 71.
31065 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
31066 (PRED_LOOP_GUARD): Change from 66 to 73.
31068 2018-01-19 Martin Liska <mliska@suse.cz>
31070 * predict.c (predict_insn_def): Add new assert.
31071 (struct branch_predictor): Change type to signed integer.
31072 (test_prediction_value_range): Amend test to cover
31073 PROB_UNINITIALIZED.
31074 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
31075 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
31076 (PRED_LOOP_ITERATIONS_MAX): Likewise.
31077 (PRED_LOOP_IV_COMPARE): Likewise.
31078 * predict.h (PROB_UNINITIALIZED): Define new constant.
31080 2018-01-19 Martin Liska <mliska@suse.cz>
31082 * predict.c (dump_prediction): Add new format for
31083 analyze_brprob.py script which is enabled with -details
31085 * profile-count.h (precise_p): New function.
31087 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
31089 PR tree-optimization/83922
31090 * tree-vect-loop.c (vect_verify_full_masking): Return false if
31091 there are no statements that need masking.
31092 (vect_active_double_reduction_p): New function.
31093 (vect_analyze_loop_operations): Use it when handling phis that
31094 are not in the loop header.
31096 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
31098 PR tree-optimization/83914
31099 * tree-vect-loop.c (vectorizable_induction): Don't convert
31100 init_expr or apply the peeling adjustment for inductions
31101 that are nested within the vectorized loop.
31103 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31105 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
31108 2018-01-18 Jakub Jelinek <jakub@redhat.com>
31112 * function.h (gimplify_parameters): Add gimple_seq * argument.
31113 * function.c: Include gimple.h and options.h.
31114 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
31115 for the added local temporaries if needed.
31116 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
31117 if there are any parameter cleanups, wrap whole body into a
31118 try/finally with the cleanups.
31120 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
31123 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
31124 Use GET_MODE_CLASS for scalar floating point.
31126 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
31130 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
31131 Fix call of call_cgraph_insertion_hooks.
31133 2018-01-18 Martin Sebor <msebor@redhat.com>
31135 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
31137 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
31140 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
31143 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
31146 * common.opt: (-ffile-prefix-map): New option.
31147 * opts.c (common_handle_option): Defer it.
31148 * opts-global.c (handle_common_deferred_options): Handle it.
31149 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
31150 * file-prefix-map.h: New file.
31151 (remap_debug_filename, add_debug_prefix_map): ...here.
31152 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
31153 * final.c (debug_prefix_map, add_debug_prefix_map
31154 remap_debug_filename): Move to...
31155 * file-prefix-map.c: New file.
31156 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
31157 generalize, get rid of alloca(), use strrchr() instead of strchr().
31158 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
31159 Implement in terms of add_prefix_map().
31160 (remap_macro_filename, remap_debug_filename): Implement in term of
31162 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
31163 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
31164 * dbxout.c: Include file-prefix-map.h.
31165 * varasm.c: Likewise.
31166 * vmsdbgout.c: Likewise.
31167 * xcoffout.c: Likewise.
31168 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
31169 * doc/cppopts.texi (-fmacro-prefix-map): Document.
31170 * doc/invoke.texi (-ffile-prefix-map): Document.
31171 (-fdebug-prefix-map): Update description.
31173 2018-01-18 Martin Liska <mliska@suse.cz>
31175 * config/i386/i386.c (indirect_thunk_name): Document that also
31177 (output_indirect_thunk): Document why both instructions
31178 (pause and lfence) are generated.
31180 2018-01-18 Richard Biener <rguenther@suse.de>
31182 PR tree-optimization/83887
31183 * graphite-scop-detection.c
31184 (scop_detection::get_nearest_dom_with_single_entry): Remove.
31185 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
31186 (scop_detection::merge_sese): Re-implement with a flood-fill
31187 algorithm that properly finds a SESE region if it exists.
31189 2018-01-18 Jakub Jelinek <jakub@redhat.com>
31192 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
31193 pointer_diff optimizations use view_convert instead of convert.
31195 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
31197 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
31198 Generate different code for -mno-speculate-indirect-jumps.
31199 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
31200 (*call_indirect_aix<mode>): Disable for
31201 -mno-speculate-indirect-jumps.
31202 (*call_indirect_aix<mode>_nospec): New define_insn.
31203 (*call_value_indirect_aix<mode>): Disable for
31204 -mno-speculate-indirect-jumps.
31205 (*call_value_indirect_aix<mode>_nospec): New define_insn.
31206 (*sibcall_nonlocal_sysv<mode>): Generate different code for
31207 -mno-speculate-indirect-jumps.
31208 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
31210 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
31212 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
31213 long double type, set the flags for noting the default long double
31214 type, even if we don't pass or return a long double type.
31216 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
31219 * ipa-inline.c (flatten_function): Do not overwrite final inlining
31222 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
31224 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
31225 support for merge[hl].
31226 (fold_mergehl_helper): New helper function.
31227 (tree-vector-builder.h): New #include for tree_vector_builder usage.
31228 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
31229 (altivec_vmrglw_direct): Add xxmrglw insn.
31231 2018-01-17 Andrew Waterman <andrew@sifive.com>
31233 * config/riscv/riscv.c (riscv_conditional_register_usage): If
31234 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
31236 2018-01-17 David Malcolm <dmalcolm@redhat.com>
31239 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
31240 call the lto_location_cache before reading the
31241 DECL_SOURCE_LOCATION of the types.
31243 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
31244 Richard Sandiford <richard.sandiford@linaro.org>
31246 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
31247 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
31248 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
31249 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
31250 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
31252 * config/aarch64/constraints.md (aarch64_movti_operand):
31254 * config/aarch64/predicates.md (Uti): Add new constraint.
31256 2018-01-17 Carl Love <cel@us.ibm.com>
31258 * config/rs6000/vsx.md (define_expand xl_len_r,
31259 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
31260 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
31262 (define_expand, define_insn): Move the shift left from the
31263 define_insn to the define_expand for lxvl and stxvl instructions.
31264 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
31265 and XL_LEN_R definitions to PURE.
31267 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
31269 * config/i386/i386.c (indirect_thunk_name): Declare regno
31270 as unsigned int. Compare regno with INVALID_REGNUM.
31271 (output_indirect_thunk): Ditto.
31272 (output_indirect_thunk_function): Ditto.
31273 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
31274 in the call to output_indirect_thunk_function.
31276 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
31278 PR middle-end/83884
31279 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
31280 rather than the size of inner_type to determine the stack slot size
31281 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
31283 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
31286 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
31289 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
31291 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
31292 endian Linux systems to optionally enable multilibs for selecting
31293 the long double type if the user configured an explicit type.
31294 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
31295 have no long double multilibs if not defined.
31296 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
31297 warn if the user used -mabi={ieee,ibm}longdouble and we built
31298 multilibs for long double.
31299 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
31300 appropriate multilib option.
31301 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
31303 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
31304 for building long double multilibs.
31305 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
31307 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
31309 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
31312 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
31314 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
31317 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
31320 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
31323 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
31325 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
31326 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
31328 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
31330 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
31331 different rtl trees depending on TARGET_64BIT.
31332 (rs6000_gen_lvx): Likewise.
31334 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
31336 * config/visium/visium.md (nop): Tweak comment.
31337 (hazard_nop): Likewise.
31339 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
31341 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
31342 -mspeculate-indirect-jumps.
31343 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
31344 for -mno-speculate-indirect-jumps.
31345 (*call_indirect_elfv2<mode>_nospec): New define_insn.
31346 (*call_value_indirect_elfv2<mode>): Disable for
31347 -mno-speculate-indirect-jumps.
31348 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
31349 (indirect_jump): Emit different RTL for
31350 -mno-speculate-indirect-jumps.
31351 (*indirect_jump<mode>): Disable for
31352 -mno-speculate-indirect-jumps.
31353 (*indirect_jump<mode>_nospec): New define_insn.
31354 (tablejump): Emit different RTL for
31355 -mno-speculate-indirect-jumps.
31356 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
31357 (tablejumpsi_nospec): New define_expand.
31358 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
31359 (tablejumpdi_nospec): New define_expand.
31360 (*tablejump<mode>_internal1): Disable for
31361 -mno-speculate-indirect-jumps.
31362 (*tablejump<mode>_internal1_nospec): New define_insn.
31363 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
31366 2018-01-16 Artyom Skrobov tyomitch@gmail.com
31368 * caller-save.c (insert_save): Drop unnecessary parameter. All
31371 2018-01-16 Jakub Jelinek <jakub@redhat.com>
31372 Richard Biener <rguenth@suse.de>
31375 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
31376 return early, inline manually is_gimple_sizepos. Make sure if we
31377 call gimplify_expr we don't end up with a gimple constant.
31378 * tree.c (variably_modified_type_p): Don't return true for
31379 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
31380 * gimplify.h (is_gimple_sizepos): Remove.
31382 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
31384 PR tree-optimization/83857
31385 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
31386 vectorizable_live_operation for pure SLP statements.
31387 (vectorizable_live_operation): Handle PHIs.
31389 2018-01-16 Richard Biener <rguenther@suse.de>
31391 PR tree-optimization/83867
31392 * tree-vect-stmts.c (vect_transform_stmt): Precompute
31393 nested_in_vect_loop_p since the scalar stmt may get invalidated.
31395 2018-01-16 Jakub Jelinek <jakub@redhat.com>
31398 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
31399 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
31400 If off is not INTEGER_CST, issue a may not be aligned warning
31401 rather than isn't aligned. Use isn%'t rather than isn't.
31402 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
31404 <case MULT_EXPR>: Improve the case when bottom and one of the
31405 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
31406 operand, in that case check if the other operand is multiple of
31407 bottom divided by the INTEGER_CST operand.
31409 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
31412 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
31413 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
31414 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
31415 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
31416 * config/pa/pa.c (pa_function_arg_advance): Likewise.
31417 (pa_function_arg, pa_arg_partial_bytes): Likewise.
31418 (pa_function_arg_size): New function.
31420 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
31422 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
31423 in a separate statement.
31425 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
31427 PR tree-optimization/83847
31428 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
31429 group gathers and scatters.
31431 2018-01-16 Jakub Jelinek <jakub@redhat.com>
31433 PR rtl-optimization/86620
31434 * params.def (max-sched-ready-insns): Bump minimum value to 1.
31436 PR rtl-optimization/83213
31437 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
31438 to last if both are JUMP_INSNs.
31440 PR tree-optimization/83843
31441 * gimple-ssa-store-merging.c
31442 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
31443 store_immediate_info for bswap/nop orig_stores.
31445 2018-01-15 Andrew Waterman <andrew@sifive.com>
31447 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
31449 <UDIV>: Increase cost if !TARGET_DIV.
31451 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
31453 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
31454 (define_attr "cr_logical_3op"): New.
31455 (cceq_ior_compare): Adjust.
31456 (cceq_ior_compare_complement): Adjust.
31457 (*cceq_rev_compare): Adjust.
31458 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
31459 (is_cracked_insn): Adjust.
31460 (insn_must_be_first_in_group): Adjust.
31461 * config/rs6000/40x.md: Adjust.
31462 * config/rs6000/440.md: Adjust.
31463 * config/rs6000/476.md: Adjust.
31464 * config/rs6000/601.md: Adjust.
31465 * config/rs6000/603.md: Adjust.
31466 * config/rs6000/6xx.md: Adjust.
31467 * config/rs6000/7450.md: Adjust.
31468 * config/rs6000/7xx.md: Adjust.
31469 * config/rs6000/8540.md: Adjust.
31470 * config/rs6000/cell.md: Adjust.
31471 * config/rs6000/e300c2c3.md: Adjust.
31472 * config/rs6000/e500mc.md: Adjust.
31473 * config/rs6000/e500mc64.md: Adjust.
31474 * config/rs6000/e5500.md: Adjust.
31475 * config/rs6000/e6500.md: Adjust.
31476 * config/rs6000/mpc.md: Adjust.
31477 * config/rs6000/power4.md: Adjust.
31478 * config/rs6000/power5.md: Adjust.
31479 * config/rs6000/power6.md: Adjust.
31480 * config/rs6000/power7.md: Adjust.
31481 * config/rs6000/power8.md: Adjust.
31482 * config/rs6000/power9.md: Adjust.
31483 * config/rs6000/rs64.md: Adjust.
31484 * config/rs6000/titan.md: Adjust.
31486 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
31488 * config/i386/predicates.md (indirect_branch_operand): Rewrite
31489 ix86_indirect_branch_register logic.
31491 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
31493 * config/i386/constraints.md (Bs): Update
31494 ix86_indirect_branch_register check. Don't check
31495 ix86_indirect_branch_register with GOT_memory_operand.
31497 * config/i386/predicates.md (GOT_memory_operand): Don't check
31498 ix86_indirect_branch_register here.
31499 (GOT32_symbol_operand): Likewise.
31501 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
31503 * config/i386/predicates.md (constant_call_address_operand):
31504 Rewrite ix86_indirect_branch_register logic.
31505 (sibcall_insn_operand): Likewise.
31507 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
31509 * config/i386/constraints.md (Bs): Replace
31510 ix86_indirect_branch_thunk_register with
31511 ix86_indirect_branch_register.
31513 * config/i386/i386.md (indirect_jump): Likewise.
31514 (tablejump): Likewise.
31515 (*sibcall_memory): Likewise.
31516 (*sibcall_value_memory): Likewise.
31517 Peepholes of indirect call and jump via memory: Likewise.
31518 * config/i386/i386.opt: Likewise.
31519 * config/i386/predicates.md (indirect_branch_operand): Likewise.
31520 (GOT_memory_operand): Likewise.
31521 (call_insn_operand): Likewise.
31522 (sibcall_insn_operand): Likewise.
31523 (GOT32_symbol_operand): Likewise.
31525 2018-01-15 Jakub Jelinek <jakub@redhat.com>
31527 PR middle-end/83837
31528 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
31529 type rather than type addr's type points to.
31530 (expand_omp_atomic_mutex): Likewise.
31531 (expand_omp_atomic): Likewise.
31533 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
31536 * config/i386/i386.c (output_indirect_thunk_function): Use
31537 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
31538 for __x86_return_thunk.
31540 2018-01-15 Richard Biener <rguenther@suse.de>
31542 PR middle-end/83850
31543 * expmed.c (extract_bit_field_1): Fix typo.
31545 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31548 * config/arm/iterators.md (VF): New mode iterator.
31549 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
31550 Remove integer-related logic from pattern.
31551 (neon_vabd<mode>_3): Likewise.
31553 2018-01-15 Jakub Jelinek <jakub@redhat.com>
31555 PR middle-end/82694
31556 * common.opt (fstrict-overflow): No longer an alias.
31557 (fwrapv-pointer): New option.
31558 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
31559 also for pointer types based on flag_wrapv_pointer.
31560 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
31561 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
31562 opts->x_flag_wrapv got set.
31563 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
31564 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
31565 POINTER_TYPE_OVERFLOW_UNDEFINED.
31566 * match.pd: Likewise in address comparison pattern.
31567 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
31569 2018-01-15 Richard Biener <rguenther@suse.de>
31572 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
31573 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
31574 Reset type names to their identifier if their TYPE_DECL doesn't
31575 have linkage (and thus is used for ODR and devirt).
31576 (save_debug_info_for_decl): Remove.
31577 (save_debug_info_for_type): Likewise.
31578 (add_tree_to_fld_list): Adjust.
31579 * tree-pretty-print.c (dump_generic_node): Make dumping of
31580 type names more robust.
31582 2018-01-15 Richard Biener <rguenther@suse.de>
31584 * BASE-VER: Bump to 8.0.1.
31586 2018-01-14 Martin Sebor <msebor@redhat.com>
31589 * builtins.c (check_access): Avoid warning when the no-warning bit
31592 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
31594 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
31595 * ira-color (allocno_hard_regs_compare): Likewise.
31597 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
31600 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
31601 Use .pushsection/.popsection.
31603 2018-01-14 Martin Sebor <msebor@redhat.com>
31606 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
31608 2018-01-14 Jakub Jelinek <jakub@redhat.com>
31610 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
31611 entry from extra_headers.
31612 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
31613 extra_headers, make the list bitwise identical to the i?86-*-* one.
31615 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
31617 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
31618 -mcmodel=large with -mindirect-branch=thunk,
31619 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
31620 -mfunction-return=thunk-extern.
31621 * doc/invoke.texi: Document -mcmodel=large is incompatible with
31622 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
31623 -mfunction-return=thunk and -mfunction-return=thunk-extern.
31625 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
31627 * config/i386/i386.c (print_reg): Print the name of the full
31628 integer register without '%'.
31629 (ix86_print_operand): Handle 'V'.
31630 * doc/extend.texi: Document 'V' modifier.
31632 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
31634 * config/i386/constraints.md (Bs): Disallow memory operand for
31635 -mindirect-branch-register.
31637 * config/i386/predicates.md (indirect_branch_operand): Likewise.
31638 (GOT_memory_operand): Likewise.
31639 (call_insn_operand): Likewise.
31640 (sibcall_insn_operand): Likewise.
31641 (GOT32_symbol_operand): Likewise.
31642 * config/i386/i386.md (indirect_jump): Call convert_memory_address
31643 for -mindirect-branch-register.
31644 (tablejump): Likewise.
31645 (*sibcall_memory): Likewise.
31646 (*sibcall_value_memory): Likewise.
31647 Disallow peepholes of indirect call and jump via memory for
31648 -mindirect-branch-register.
31649 (*call_pop): Replace m with Bw.
31650 (*call_value_pop): Likewise.
31651 (*sibcall_pop_memory): Replace m with Bs.
31652 * config/i386/i386.opt (mindirect-branch-register): New option.
31653 * doc/invoke.texi: Document -mindirect-branch-register option.
31655 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
31657 * config/i386/i386-protos.h (ix86_output_function_return): New.
31658 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
31659 set function_return_type.
31660 (indirect_thunk_name): Add ret_p to indicate thunk for function
31662 (output_indirect_thunk_function): Pass false to
31663 indirect_thunk_name.
31664 (ix86_output_indirect_branch_via_reg): Likewise.
31665 (ix86_output_indirect_branch_via_push): Likewise.
31666 (output_indirect_thunk_function): Create alias for function
31667 return thunk if regno < 0.
31668 (ix86_output_function_return): New function.
31669 (ix86_handle_fndecl_attribute): Handle function_return.
31670 (ix86_attribute_table): Add function_return.
31671 * config/i386/i386.h (machine_function): Add
31672 function_return_type.
31673 * config/i386/i386.md (simple_return_internal): Use
31674 ix86_output_function_return.
31675 (simple_return_internal_long): Likewise.
31676 * config/i386/i386.opt (mfunction-return=): New option.
31677 (indirect_branch): Mention -mfunction-return=.
31678 * doc/extend.texi: Document function_return function attribute.
31679 * doc/invoke.texi: Document -mfunction-return= option.
31681 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
31683 * config/i386/i386-opts.h (indirect_branch): New.
31684 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
31685 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
31686 with local indirect jump when converting indirect call and jump.
31687 (ix86_set_indirect_branch_type): New.
31688 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
31689 (indirectlabelno): New.
31690 (indirect_thunk_needed): Likewise.
31691 (indirect_thunk_bnd_needed): Likewise.
31692 (indirect_thunks_used): Likewise.
31693 (indirect_thunks_bnd_used): Likewise.
31694 (INDIRECT_LABEL): Likewise.
31695 (indirect_thunk_name): Likewise.
31696 (output_indirect_thunk): Likewise.
31697 (output_indirect_thunk_function): Likewise.
31698 (ix86_output_indirect_branch_via_reg): Likewise.
31699 (ix86_output_indirect_branch_via_push): Likewise.
31700 (ix86_output_indirect_branch): Likewise.
31701 (ix86_output_indirect_jmp): Likewise.
31702 (ix86_code_end): Call output_indirect_thunk_function if needed.
31703 (ix86_output_call_insn): Call ix86_output_indirect_branch if
31705 (ix86_handle_fndecl_attribute): Handle indirect_branch.
31706 (ix86_attribute_table): Add indirect_branch.
31707 * config/i386/i386.h (machine_function): Add indirect_branch_type
31708 and has_local_indirect_jump.
31709 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
31711 (tablejump): Likewise.
31712 (*indirect_jump): Use ix86_output_indirect_jmp.
31713 (*tablejump_1): Likewise.
31714 (simple_return_indirect_internal): Likewise.
31715 * config/i386/i386.opt (mindirect-branch=): New option.
31716 (indirect_branch): New.
31719 (thunk-inline): Likewise.
31720 (thunk-extern): Likewise.
31721 * doc/extend.texi: Document indirect_branch function attribute.
31722 * doc/invoke.texi: Document -mindirect-branch= option.
31724 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
31727 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
31729 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
31731 * ipa-inline.c (want_inline_small_function_p): Return false if
31732 inlining has already failed with CIF_FINAL_ERROR.
31733 (update_caller_keys): Call want_inline_small_function_p before
31735 (update_callee_keys): Likewise.
31737 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
31739 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
31741 (rs6000_quadword_masked_address_p): Likewise.
31742 (quad_aligned_load_p): Likewise.
31743 (quad_aligned_store_p): Likewise.
31744 (const_load_sequence_p): Add comment to describe the outer-most loop.
31745 (mimic_memory_attributes_and_flags): New function.
31746 (rs6000_gen_stvx): Likewise.
31747 (replace_swapped_aligned_store): Likewise.
31748 (rs6000_gen_lvx): Likewise.
31749 (replace_swapped_aligned_load): Likewise.
31750 (replace_swapped_load_constant): Capitalize argument name in
31751 comment describing this function.
31752 (rs6000_analyze_swaps): Add a third pass to search for vector loads
31753 and stores that access quad-word aligned addresses and replace
31754 with stvx or lvx instructions when appropriate.
31755 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
31756 New function prototype.
31757 (rs6000_quadword_masked_address_p): Likewise.
31758 (rs6000_gen_lvx): Likewise.
31759 (rs6000_gen_stvx): Likewise.
31760 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
31761 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
31762 when memory address is aligned.
31763 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
31764 this split to select lvx instruction when memory address is aligned.
31765 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
31766 instruction when memory address is aligned.
31767 (*vsx_le_perm_load_v16qi): Likewise.
31768 (four unnamed splitters): Modify to select the stvx instruction
31769 when memory is aligned.
31771 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
31773 * predict.c (determine_unlikely_bbs): Handle correctly BBs
31774 which appears in the queue multiple times.
31776 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
31777 Alan Hayward <alan.hayward@arm.com>
31778 David Sherwood <david.sherwood@arm.com>
31780 * tree-vectorizer.h (vec_lower_bound): New structure.
31781 (_loop_vec_info): Add check_nonzero and lower_bounds.
31782 (LOOP_VINFO_CHECK_NONZERO): New macro.
31783 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
31784 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
31785 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
31786 fields. Make seg_len the distance travelled, not including the
31788 (dr_direction_indicator): Declare.
31789 (dr_zero_step_indicator): Likewise.
31790 (dr_known_forward_stride_p): Likewise.
31791 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
31793 (runtime_alias_check_p): Allow runtime alias checks with
31795 (operator ==): Compare access_size and align.
31796 (prune_runtime_alias_test_list): Rework for new distinction between
31797 the access_size and seg_len.
31798 (create_intersect_range_checks_index): Likewise. Cope with polynomial
31800 (get_segment_min_max): New function.
31801 (create_intersect_range_checks): Use it.
31802 (dr_step_indicator): New function.
31803 (dr_direction_indicator): Likewise.
31804 (dr_zero_step_indicator): Likewise.
31805 (dr_known_forward_stride_p): Likewise.
31806 * tree-loop-distribution.c (data_ref_segment_size): Return
31807 DR_STEP * (niters - 1).
31808 (compute_alias_check_pairs): Update call to the dr_with_seg_len
31810 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
31811 (vect_preserves_scalar_order_p): New function, split out from...
31812 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
31813 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
31814 (vect_vfa_access_size): New function.
31815 (vect_vfa_align): Likewise.
31816 (vect_compile_time_alias): Take access_size_a and access_b arguments.
31817 (dump_lower_bound): New function.
31818 (vect_check_lower_bound): Likewise.
31819 (vect_small_gap_p): Likewise.
31820 (vectorizable_with_step_bound_p): Likewise.
31821 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
31822 depencies if the vectorization factor is 1. Convert the checks
31823 for nonzero steps into checks on the bounds of DR_STEP. Try using
31824 a bunds check for variable steps if the minimum required step is
31825 relatively small. Update calls to the dr_with_seg_len
31826 constructor and to vect_compile_time_alias.
31827 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
31829 (vect_loop_versioning): Call it.
31830 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
31832 (vect_estimate_min_profitable_iters): Account for any bounds checks.
31834 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
31835 Alan Hayward <alan.hayward@arm.com>
31836 David Sherwood <david.sherwood@arm.com>
31838 * doc/sourcebuild.texi (vect_scatter_store): Document.
31839 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
31841 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
31843 * genopinit.c (main): Add supports_vec_scatter_store and
31844 supports_vec_scatter_store_cached to target_optabs.
31845 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
31846 IFN_MASK_SCATTER_STORE.
31847 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
31849 * internal-fn.h (internal_store_fn_p): Declare.
31850 (internal_fn_stored_value_index): Likewise.
31851 * internal-fn.c (scatter_store_direct): New macro.
31852 (expand_scatter_store_optab_fn): New function.
31853 (direct_scatter_store_optab_supported_p): New macro.
31854 (internal_store_fn_p): New function.
31855 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
31856 IFN_MASK_SCATTER_STORE.
31857 (internal_fn_mask_index): Likewise.
31858 (internal_fn_stored_value_index): New function.
31859 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
31860 for scatter stores.
31861 * optabs-query.h (supports_vec_scatter_store_p): Declare.
31862 * optabs-query.c (supports_vec_scatter_store_p): New function.
31863 * tree-vectorizer.h (vect_get_store_rhs): Declare.
31864 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
31865 true for scatter stores.
31866 (vect_gather_scatter_fn_p): Handle scatter stores too.
31867 (vect_check_gather_scatter): Consider using scatter stores if
31868 supports_vec_scatter_store_p.
31869 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
31870 scatter stores too.
31871 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
31872 internal_fn_stored_value_index.
31873 (check_load_store_masking): Handle scatter stores too.
31874 (vect_get_store_rhs): Make public.
31875 (vectorizable_call): Use internal_store_fn_p.
31876 (vectorizable_store): Handle scatter store internal functions.
31877 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
31878 when deciding whether the end of the group has been reached.
31879 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
31880 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
31881 (mask_scatter_store<mode>): New insns.
31883 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
31884 Alan Hayward <alan.hayward@arm.com>
31885 David Sherwood <david.sherwood@arm.com>
31887 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
31888 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
31889 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
31891 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
31892 Use vect_truncate_gather_scatter_offset if we can't treat the
31893 operation as a normal gather load or scatter store.
31894 (get_group_load_store_type): Take the gather_scatter_info
31895 as argument. Try using a gather load or scatter store for
31896 single-element groups.
31897 (get_load_store_type): Update calls to get_group_load_store_type
31898 and vect_use_strided_gather_scatters_p.
31900 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
31901 Alan Hayward <alan.hayward@arm.com>
31902 David Sherwood <david.sherwood@arm.com>
31904 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
31905 optional tree argument.
31906 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
31908 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
31909 but continue to use the current value as a fallback.
31910 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
31911 to compare the updates.
31912 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
31913 (get_load_store_type): Use it when handling a strided access.
31914 (vect_get_strided_load_store_ops): New function.
31915 (vect_get_data_ptr_increment): Likewise.
31916 (vectorizable_load): Handle strided gather loads. Always pass
31917 a step to vect_create_data_ref_ptr and bump_vector_ptr.
31919 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
31920 Alan Hayward <alan.hayward@arm.com>
31921 David Sherwood <david.sherwood@arm.com>
31923 * doc/md.texi (gather_load@var{m}): Document.
31924 (mask_gather_load@var{m}): Likewise.
31925 * genopinit.c (main): Add supports_vec_gather_load and
31926 supports_vec_gather_load_cached to target_optabs.
31927 * optabs-tree.c (init_tree_optimization_optabs): Use
31928 ggc_cleared_alloc to allocate target_optabs.
31929 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
31930 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
31932 * internal-fn.h (internal_load_fn_p): Declare.
31933 (internal_gather_scatter_fn_p): Likewise.
31934 (internal_fn_mask_index): Likewise.
31935 (internal_gather_scatter_fn_supported_p): Likewise.
31936 * internal-fn.c (gather_load_direct): New macro.
31937 (expand_gather_load_optab_fn): New function.
31938 (direct_gather_load_optab_supported_p): New macro.
31939 (direct_internal_fn_optab): New function.
31940 (internal_load_fn_p): Likewise.
31941 (internal_gather_scatter_fn_p): Likewise.
31942 (internal_fn_mask_index): Likewise.
31943 (internal_gather_scatter_fn_supported_p): Likewise.
31944 * optabs-query.c (supports_at_least_one_mode_p): New function.
31945 (supports_vec_gather_load_p): Likewise.
31946 * optabs-query.h (supports_vec_gather_load_p): Declare.
31947 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
31948 and memory_type field.
31949 (NUM_PATTERNS): Bump to 15.
31950 * tree-vect-data-refs.c: Include internal-fn.h.
31951 (vect_gather_scatter_fn_p): New function.
31952 (vect_describe_gather_scatter_call): Likewise.
31953 (vect_check_gather_scatter): Try using internal functions for
31954 gather loads. Recognize existing calls to a gather load function.
31955 (vect_analyze_data_refs): Consider using gather loads if
31956 supports_vec_gather_load_p.
31957 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
31958 (vect_get_gather_scatter_offset_type): Likewise.
31959 (vect_convert_mask_for_vectype): Likewise.
31960 (vect_add_conversion_to_patterm): Likewise.
31961 (vect_try_gather_scatter_pattern): Likewise.
31962 (vect_recog_gather_scatter_pattern): New pattern recognizer.
31963 (vect_vect_recog_func_ptrs): Add it.
31964 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
31965 internal_fn_mask_index and internal_gather_scatter_fn_p.
31966 (check_load_store_masking): Take the gather_scatter_info as an
31967 argument and handle gather loads.
31968 (vect_get_gather_scatter_ops): New function.
31969 (vectorizable_call): Check internal_load_fn_p.
31970 (vectorizable_load): Likewise. Handle gather load internal
31972 (vectorizable_store): Update call to check_load_store_masking.
31973 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
31974 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
31975 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
31976 (aarch64_gather_scale_operand_d): New predicates.
31977 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
31978 (mask_gather_load<mode>): New insns.
31980 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
31981 Alan Hayward <alan.hayward@arm.com>
31982 David Sherwood <david.sherwood@arm.com>
31984 * optabs.def (fold_left_plus_optab): New optab.
31985 * doc/md.texi (fold_left_plus_@var{m}): Document.
31986 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
31987 * internal-fn.c (fold_left_direct): Define.
31988 (expand_fold_left_optab_fn): Likewise.
31989 (direct_fold_left_optab_supported_p): Likewise.
31990 * fold-const-call.c (fold_const_fold_left): New function.
31991 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
31992 * tree-parloops.c (valid_reduction_p): New function.
31993 (gather_scalar_reductions): Use it.
31994 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
31995 (vect_finish_replace_stmt): Declare.
31996 * tree-vect-loop.c (fold_left_reduction_fn): New function.
31997 (needs_fold_left_reduction_p): New function, split out from...
31998 (vect_is_simple_reduction): ...here. Accept reductions that
31999 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
32000 (vect_force_simple_reduction): Also store the reduction type in
32001 the assignment's STMT_VINFO_REDUC_TYPE.
32002 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
32003 (merge_with_identity): New function.
32004 (vect_expand_fold_left): Likewise.
32005 (vectorize_fold_left_reduction): Likewise.
32006 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
32007 scalar phi in place for it. Check for target support and reject
32008 cases that would reassociate the operation. Defer the transform
32009 phase to vectorize_fold_left_reduction.
32010 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
32011 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
32012 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
32014 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32016 * tree-if-conv.c (predicate_mem_writes): Remove redundant
32017 call to ifc_temp_var.
32019 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32020 Alan Hayward <alan.hayward@arm.com>
32021 David Sherwood <david.sherwood@arm.com>
32023 * target.def (legitimize_address_displacement): Take the original
32024 offset as a poly_int.
32025 * targhooks.h (default_legitimize_address_displacement): Update
32027 * targhooks.c (default_legitimize_address_displacement): Likewise.
32028 * doc/tm.texi: Regenerate.
32029 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
32030 as an argument, moving assert of ad->disp == ad->disp_term to...
32031 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
32032 Try calling targetm.legitimize_address_displacement before expanding
32033 the address rather than afterwards, and adjust for the new interface.
32034 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
32035 Match the new hook interface. Handle SVE addresses.
32036 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
32037 new hook interface.
32039 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32041 * Makefile.in (OBJS): Add early-remat.o.
32042 * target.def (select_early_remat_modes): New hook.
32043 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
32044 * doc/tm.texi: Regenerate.
32045 * targhooks.h (default_select_early_remat_modes): Declare.
32046 * targhooks.c (default_select_early_remat_modes): New function.
32047 * timevar.def (TV_EARLY_REMAT): New timevar.
32048 * passes.def (pass_early_remat): New pass.
32049 * tree-pass.h (make_pass_early_remat): Declare.
32050 * early-remat.c: New file.
32051 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
32053 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
32055 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32056 Alan Hayward <alan.hayward@arm.com>
32057 David Sherwood <david.sherwood@arm.com>
32059 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
32060 vfm1 with a bound_epilog parameter.
32061 (vect_do_peeling): Update calls accordingly, and move the prologue
32062 call earlier in the function. Treat the base bound_epilog as 0 for
32063 fully-masked loops and retain vf - 1 for other loops. Add 1 to
32064 this base when peeling for gaps.
32065 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
32066 with fully-masked loops.
32067 (vect_estimate_min_profitable_iters): Handle the single peeled
32068 iteration in that case.
32070 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32071 Alan Hayward <alan.hayward@arm.com>
32072 David Sherwood <david.sherwood@arm.com>
32074 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
32075 single-element interleaving even if the size is not a power of 2.
32076 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
32077 accesses for single-element interleaving if the group size is
32080 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32081 Alan Hayward <alan.hayward@arm.com>
32082 David Sherwood <david.sherwood@arm.com>
32084 * doc/md.texi (fold_extract_last_@var{m}): Document.
32085 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
32086 * optabs.def (fold_extract_last_optab): New optab.
32087 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
32088 * internal-fn.c (fold_extract_direct): New macro.
32089 (expand_fold_extract_optab_fn): Likewise.
32090 (direct_fold_extract_optab_supported_p): Likewise.
32091 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
32092 * tree-vect-loop.c (vect_model_reduction_cost): Handle
32093 EXTRACT_LAST_REDUCTION.
32094 (get_initial_def_for_reduction): Do not create an initial vector
32095 for EXTRACT_LAST_REDUCTION reductions.
32096 (vectorizable_reduction): Leave the scalar phi in place for
32097 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
32098 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
32099 epilogue code for EXTRACT_LAST_REDUCTION and defer the
32100 transform phase to vectorizable_condition.
32101 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
32103 (vect_finish_stmt_generation): ...here.
32104 (vect_finish_replace_stmt): New function.
32105 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
32106 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
32108 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
32110 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32111 Alan Hayward <alan.hayward@arm.com>
32112 David Sherwood <david.sherwood@arm.com>
32114 * doc/md.texi (extract_last_@var{m}): Document.
32115 * optabs.def (extract_last_optab): New optab.
32116 * internal-fn.def (EXTRACT_LAST): New internal function.
32117 * internal-fn.c (cond_unary_direct): New macro.
32118 (expand_cond_unary_optab_fn): Likewise.
32119 (direct_cond_unary_optab_supported_p): Likewise.
32120 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
32121 loops using EXTRACT_LAST.
32122 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
32123 (extract_last_<mode>): ...this optab.
32124 (vec_extract<mode><Vel>): Update accordingly.
32126 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32127 Alan Hayward <alan.hayward@arm.com>
32128 David Sherwood <david.sherwood@arm.com>
32130 * target.def (empty_mask_is_expensive): New hook.
32131 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
32132 * doc/tm.texi: Regenerate.
32133 * targhooks.h (default_empty_mask_is_expensive): Declare.
32134 * targhooks.c (default_empty_mask_is_expensive): New function.
32135 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
32136 if the target says that empty masks are expensive.
32137 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
32139 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
32141 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32142 Alan Hayward <alan.hayward@arm.com>
32143 David Sherwood <david.sherwood@arm.com>
32145 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
32146 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
32147 (vect_use_loop_mask_for_alignment_p): New function.
32148 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
32149 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
32150 niters_skip argument. Make sure that the first niters_skip elements
32151 of the first iteration are inactive.
32152 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
32153 Update call to vect_set_loop_masks_directly.
32154 (get_misalign_in_elems): New function, split out from...
32155 (vect_gen_prolog_loop_niters): ...here.
32156 (vect_update_init_of_dr): Take a code argument that specifies whether
32157 the adjustment should be added or subtracted.
32158 (vect_update_init_of_drs): Likewise.
32159 (vect_prepare_for_masked_peels): New function.
32160 (vect_do_peeling): Skip prologue peeling if we're using a mask
32161 instead. Update call to vect_update_inits_of_drs.
32162 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
32164 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
32165 alignment. Do not include the number of peeled iterations in
32166 the minimum threshold in that case.
32167 (vectorizable_induction): Adjust the start value down by
32168 LOOP_VINFO_MASK_SKIP_NITERS iterations.
32169 (vect_transform_loop): Call vect_prepare_for_masked_peels.
32170 Take the number of skipped iterations into account when calculating
32172 * tree-vect-stmts.c (vect_gen_while_not): New function.
32174 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32175 Alan Hayward <alan.hayward@arm.com>
32176 David Sherwood <david.sherwood@arm.com>
32178 * doc/sourcebuild.texi (vect_fully_masked): Document.
32179 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
32180 default value to 0.
32181 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
32183 (vect_analyze_loop_2): ...here. Don't check the vectorization
32184 factor against the number of loop iterations if the loop is
32187 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32188 Alan Hayward <alan.hayward@arm.com>
32189 David Sherwood <david.sherwood@arm.com>
32191 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
32192 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
32193 (dump_groups): Update accordingly.
32194 (iv_use::mem_type): New member variable.
32195 (address_p): New function.
32196 (record_use): Add a mem_type argument and initialize the new
32198 (record_group_use): Add a mem_type argument. Use address_p.
32199 Remove obsolete null checks of base_object. Update call to record_use.
32200 (find_interesting_uses_op): Update call to record_group_use.
32201 (find_interesting_uses_cond): Likewise.
32202 (find_interesting_uses_address): Likewise.
32203 (get_mem_type_for_internal_fn): New function.
32204 (find_address_like_use): Likewise.
32205 (find_interesting_uses_stmt): Try find_address_like_use before
32206 calling find_interesting_uses_op.
32207 (addr_offset_valid_p): Use the iv mem_type field as the type
32208 of the addressed memory.
32209 (add_autoinc_candidates): Likewise.
32210 (get_address_cost): Likewise.
32211 (split_small_address_groups_p): Use address_p.
32212 (split_address_groups): Likewise.
32213 (add_iv_candidate_for_use): Likewise.
32214 (autoinc_possible_for_pair): Likewise.
32215 (rewrite_groups): Likewise.
32216 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
32217 (determine_group_iv_cost): Update after split of USE_ADDRESS.
32218 (get_alias_ptr_type_for_ptr_address): New function.
32219 (rewrite_use_address): Rewrite address uses in calls that were
32220 identified by find_address_like_use.
32222 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32223 Alan Hayward <alan.hayward@arm.com>
32224 David Sherwood <david.sherwood@arm.com>
32226 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
32228 * gimple-expr.h (is_gimple_addressable: Likewise.
32229 * gimple-expr.c (is_gimple_address): Likewise.
32230 * internal-fn.c (expand_call_mem_ref): New function.
32231 (expand_mask_load_optab_fn): Use it.
32232 (expand_mask_store_optab_fn): Likewise.
32234 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32235 Alan Hayward <alan.hayward@arm.com>
32236 David Sherwood <david.sherwood@arm.com>
32238 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
32239 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
32240 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
32241 (cond_umax@var{mode}): Document.
32242 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
32243 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
32244 (cond_umin_optab, cond_umax_optab): New optabs.
32245 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
32246 (COND_IOR, COND_XOR): New internal functions.
32247 * internal-fn.h (get_conditional_internal_fn): Declare.
32248 * internal-fn.c (cond_binary_direct): New macro.
32249 (expand_cond_binary_optab_fn): Likewise.
32250 (direct_cond_binary_optab_supported_p): Likewise.
32251 (get_conditional_internal_fn): New function.
32252 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
32253 Cope with reduction statements that are vectorized as calls rather
32255 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
32256 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
32257 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
32258 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
32259 (UNSPEC_COND_EOR): New unspecs.
32260 (optab): Add mappings for them.
32261 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
32262 (sve_int_op, sve_fp_op): New int attributes.
32264 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32265 Alan Hayward <alan.hayward@arm.com>
32266 David Sherwood <david.sherwood@arm.com>
32268 * optabs.def (while_ult_optab): New optab.
32269 * doc/md.texi (while_ult@var{m}@var{n}): Document.
32270 * internal-fn.def (WHILE_ULT): New internal function.
32271 * internal-fn.h (direct_internal_fn_supported_p): New override
32272 that takes two types as argument.
32273 * internal-fn.c (while_direct): New macro.
32274 (expand_while_optab_fn): New function.
32275 (convert_optab_supported_p): Likewise.
32276 (direct_while_optab_supported_p): New macro.
32277 * wide-int.h (wi::udiv_ceil): New function.
32278 * tree-vectorizer.h (rgroup_masks): New structure.
32279 (vec_loop_masks): New typedef.
32280 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
32281 and fully_masked_p.
32282 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
32283 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
32284 (vect_max_vf): New function.
32285 (slpeel_make_loop_iterate_ntimes): Delete.
32286 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
32287 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
32288 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
32289 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
32290 internal-fn.h, stor-layout.h and optabs-query.h.
32291 (vect_set_loop_mask): New function.
32292 (add_preheader_seq): Likewise.
32293 (add_header_seq): Likewise.
32294 (interleave_supported_p): Likewise.
32295 (vect_maybe_permute_loop_masks): Likewise.
32296 (vect_set_loop_masks_directly): Likewise.
32297 (vect_set_loop_condition_masked): Likewise.
32298 (vect_set_loop_condition_unmasked): New function, split out from
32299 slpeel_make_loop_iterate_ntimes.
32300 (slpeel_make_loop_iterate_ntimes): Rename to..
32301 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
32302 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
32303 (vect_do_peeling): Update call accordingly.
32304 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
32306 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
32307 mask_compare_type, can_fully_mask_p and fully_masked_p.
32308 (release_vec_loop_masks): New function.
32309 (_loop_vec_info): Use it to free the loop masks.
32310 (can_produce_all_loop_masks_p): New function.
32311 (vect_get_max_nscalars_per_iter): Likewise.
32312 (vect_verify_full_masking): Likewise.
32313 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
32314 retries, and free the mask rgroups before retrying. Check loop-wide
32315 reasons for disallowing fully-masked loops. Make the final decision
32316 about whether use a fully-masked loop or not.
32317 (vect_estimate_min_profitable_iters): Do not assume that peeling
32318 for the number of iterations will be needed for fully-masked loops.
32319 (vectorizable_reduction): Disable fully-masked loops.
32320 (vectorizable_live_operation): Likewise.
32321 (vect_halve_mask_nunits): New function.
32322 (vect_double_mask_nunits): Likewise.
32323 (vect_record_loop_mask): Likewise.
32324 (vect_get_loop_mask): Likewise.
32325 (vect_transform_loop): Handle the case in which the final loop
32326 iteration might handle a partial vector. Call vect_set_loop_condition
32327 instead of slpeel_make_loop_iterate_ntimes.
32328 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
32329 (check_load_store_masking): New function.
32330 (prepare_load_store_mask): Likewise.
32331 (vectorizable_store): Handle fully-masked loops.
32332 (vectorizable_load): Likewise.
32333 (supportable_widening_operation): Use vect_halve_mask_nunits for
32335 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
32336 (vect_gen_while): New function.
32337 * config/aarch64/aarch64.md (umax<mode>3): New expander.
32338 (aarch64_uqdec<mode>): New insn.
32340 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32341 Alan Hayward <alan.hayward@arm.com>
32342 David Sherwood <david.sherwood@arm.com>
32344 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
32345 (reduc_xor_scal_optab): New optabs.
32346 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
32347 (reduc_xor_scal_@var{m}): Document.
32348 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
32349 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
32350 internal functions.
32351 * fold-const-call.c (fold_const_call): Handle them.
32352 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
32353 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
32354 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
32355 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
32356 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
32357 (UNSPEC_XORV): New unspecs.
32358 (optab): Add entries for them.
32359 (BITWISEV): New int iterator.
32360 (bit_reduc_op): New int attributes.
32362 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32363 Alan Hayward <alan.hayward@arm.com>
32364 David Sherwood <david.sherwood@arm.com>
32366 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
32367 * internal-fn.def (VEC_SHL_INSERT): New internal function.
32368 * optabs.def (vec_shl_insert_optab): New optab.
32369 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
32370 (duplicate_and_interleave): Likewise.
32371 * tree-vect-loop.c: Include internal-fn.h.
32372 (neutral_op_for_slp_reduction): New function, split out from
32373 get_initial_defs_for_reduction.
32374 (get_initial_def_for_reduction): Handle option 2 for variable-length
32375 vectors by loading the neutral value into a vector and then shifting
32376 the initial value into element 0.
32377 (get_initial_defs_for_reduction): Replace the code argument with
32378 the neutral value calculated by neutral_op_for_slp_reduction.
32379 Use gimple_build_vector for constant-length vectors.
32380 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
32381 but the first group_size elements have a neutral value.
32382 Use duplicate_and_interleave otherwise.
32383 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
32384 Update call to get_initial_defs_for_reduction. Handle SLP
32385 reductions for variable-length vectors by creating one vector
32386 result for each scalar result, with the elements associated
32387 with other scalar results stubbed out with the neutral value.
32388 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
32389 Require IFN_VEC_SHL_INSERT for double reductions on
32390 variable-length vectors, or SLP reductions that have
32391 a neutral value. Require can_duplicate_and_interleave_p
32392 support for variable-length unchained SLP reductions if there
32393 is no neutral value, such as for MIN/MAX reductions. Also require
32394 the number of vector elements to be a multiple of the number of
32395 SLP statements when doing variable-length unchained SLP reductions.
32396 Update call to vect_create_epilog_for_reduction.
32397 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
32398 and remove initial values.
32399 (duplicate_and_interleave): Make public.
32400 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
32401 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
32403 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32404 Alan Hayward <alan.hayward@arm.com>
32405 David Sherwood <david.sherwood@arm.com>
32407 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
32408 (can_duplicate_and_interleave_p): New function.
32409 (vect_get_and_check_slp_defs): Take the vector of statements
32410 rather than just the current one. Remove excess parentheses.
32411 Restriction rejectinon of vect_constant_def and vect_external_def
32412 for variable-length vectors to boolean types, or types for which
32413 can_duplicate_and_interleave_p is false.
32414 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
32415 (duplicate_and_interleave): New function.
32416 (vect_get_constant_vectors): Use gimple_build_vector for
32417 constant-length vectors and suitable variable-length constant
32418 vectors. Use duplicate_and_interleave for other variable-length
32419 vectors. Don't defer the update when inserting new statements.
32421 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32422 Alan Hayward <alan.hayward@arm.com>
32423 David Sherwood <david.sherwood@arm.com>
32425 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
32426 min_profitable_iters doesn't go negative.
32428 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32429 Alan Hayward <alan.hayward@arm.com>
32430 David Sherwood <david.sherwood@arm.com>
32432 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
32433 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
32434 * optabs.def (vec_mask_load_lanes_optab): New optab.
32435 (vec_mask_store_lanes_optab): Likewise.
32436 * internal-fn.def (MASK_LOAD_LANES): New internal function.
32437 (MASK_STORE_LANES): Likewise.
32438 * internal-fn.c (mask_load_lanes_direct): New macro.
32439 (mask_store_lanes_direct): Likewise.
32440 (expand_mask_load_optab_fn): Handle masked operations.
32441 (expand_mask_load_lanes_optab_fn): New macro.
32442 (expand_mask_store_optab_fn): Handle masked operations.
32443 (expand_mask_store_lanes_optab_fn): New macro.
32444 (direct_mask_load_lanes_optab_supported_p): Likewise.
32445 (direct_mask_store_lanes_optab_supported_p): Likewise.
32446 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
32448 (vect_load_lanes_supported): Likewise.
32449 * tree-vect-data-refs.c (strip_conversion): New function.
32450 (can_group_stmts_p): Likewise.
32451 (vect_analyze_data_ref_accesses): Use it instead of checking
32452 for a pair of assignments.
32453 (vect_store_lanes_supported): Take a masked_p parameter.
32454 (vect_load_lanes_supported): Likewise.
32455 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
32456 vect_store_lanes_supported and vect_load_lanes_supported.
32457 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
32458 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
32459 parameter. Don't allow gaps for masked accesses.
32460 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
32461 and vect_load_lanes_supported.
32462 (get_load_store_type): Take a masked_p parameter and update
32463 call to get_group_load_store_type.
32464 (vectorizable_store): Update call to get_load_store_type.
32465 Handle IFN_MASK_STORE_LANES.
32466 (vectorizable_load): Update call to get_load_store_type.
32467 Handle IFN_MASK_LOAD_LANES.
32469 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32470 Alan Hayward <alan.hayward@arm.com>
32471 David Sherwood <david.sherwood@arm.com>
32473 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
32475 * config/aarch64/aarch64-protos.h
32476 (aarch64_sve_struct_memory_operand_p): Declare.
32477 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
32478 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
32479 (VPRED, vpred): Handle SVE structure modes.
32480 * config/aarch64/constraints.md (Utx): New constraint.
32481 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
32482 (aarch64_sve_struct_nonimmediate_operand): New predicates.
32483 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
32484 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
32485 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
32486 structure modes. Split into pieces after RA.
32487 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
32488 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
32490 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
32491 SVE structure modes.
32492 (aarch64_classify_address): Likewise.
32493 (sizetochar): Move earlier in file.
32494 (aarch64_print_operand): Handle SVE register lists.
32495 (aarch64_array_mode): New function.
32496 (aarch64_sve_struct_memory_operand_p): Likewise.
32497 (TARGET_ARRAY_MODE): Redefine.
32499 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32500 Alan Hayward <alan.hayward@arm.com>
32501 David Sherwood <david.sherwood@arm.com>
32503 * target.def (array_mode): New target hook.
32504 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
32505 * doc/tm.texi: Regenerate.
32506 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
32507 * hooks.c (hook_optmode_mode_uhwi_none): New function.
32508 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
32509 targetm.array_mode.
32510 * stor-layout.c (mode_for_array): Likewise. Support polynomial
32513 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32514 Alan Hayward <alan.hayward@arm.com>
32515 David Sherwood <david.sherwood@arm.com>
32517 * fold-const.c (fold_binary_loc): Check the argument types
32518 rather than the result type when testing for a vector operation.
32520 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32522 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
32523 * doc/tm.texi: Regenerate.
32525 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32526 Alan Hayward <alan.hayward@arm.com>
32527 David Sherwood <david.sherwood@arm.com>
32529 * doc/invoke.texi (-msve-vector-bits=): Document new option.
32530 (sve): Document new AArch64 extension.
32531 * doc/md.texi (w): Extend the description of the AArch64
32532 constraint to include SVE vectors.
32533 (Upl, Upa): Document new AArch64 predicate constraints.
32534 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
32536 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
32537 (msve-vector-bits=): New option.
32538 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
32539 SVE when these are disabled.
32540 (sve): New extension.
32541 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
32542 modes. Adjust their number of units based on aarch64_sve_vg.
32543 (MAX_BITSIZE_MODE_ANY_MODE): Define.
32544 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
32545 aarch64_addr_query_type.
32546 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
32547 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
32548 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
32549 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
32550 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
32551 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
32552 (aarch64_simd_imm_zero_p): Delete.
32553 (aarch64_check_zero_based_sve_index_immediate): Declare.
32554 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
32555 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
32556 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
32557 (aarch64_sve_float_mul_immediate_p): Likewise.
32558 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
32559 rather than an rtx.
32560 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
32561 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
32562 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
32563 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
32564 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
32565 (aarch64_regmode_natural_size): Likewise.
32566 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
32567 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
32569 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
32570 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
32571 for VG and the SVE predicate registers.
32572 (V_ALIASES): Add a "z"-prefixed alias.
32573 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
32574 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
32575 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
32576 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
32577 (REG_CLASS_NAMES): Add entries for them.
32578 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
32579 and the predicate registers.
32580 (aarch64_sve_vg): Declare.
32581 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
32582 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
32583 (REGMODE_NATURAL_SIZE): Define.
32584 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
32586 * config/aarch64/aarch64.c: Include cfgrtl.h.
32587 (simd_immediate_info): Add a constructor for series vectors,
32588 and an associated step field.
32589 (aarch64_sve_vg): New variable.
32590 (aarch64_dbx_register_number): Handle VG and the predicate registers.
32591 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
32592 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
32593 (VEC_ANY_DATA, VEC_STRUCT): New constants.
32594 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
32595 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
32596 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
32597 (aarch64_get_mask_mode): New functions.
32598 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
32599 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
32600 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
32601 predicate modes and predicate registers. Explicitly restrict
32602 GPRs to modes of 16 bytes or smaller. Only allow FP registers
32603 to store a vector mode if it is recognized by
32604 aarch64_classify_vector_mode.
32605 (aarch64_regmode_natural_size): New function.
32606 (aarch64_hard_regno_caller_save_mode): Return the original mode
32608 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
32609 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
32610 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
32611 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
32613 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
32614 does not overlap dest if the function is frame-related. Handle
32616 (aarch64_split_add_offset): New function.
32617 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
32618 them aarch64_add_offset.
32619 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
32620 and update call to aarch64_sub_sp.
32621 (aarch64_add_cfa_expression): New function.
32622 (aarch64_expand_prologue): Pass extra temporary registers to the
32623 functions above. Handle the case in which we need to emit new
32624 DW_CFA_expressions for registers that were originally saved
32625 relative to the stack pointer, but now have to be expressed
32626 relative to the frame pointer.
32627 (aarch64_output_mi_thunk): Pass extra temporary registers to the
32629 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
32630 IP0 and IP1 values for SVE frames.
32631 (aarch64_expand_vec_series): New function.
32632 (aarch64_expand_sve_widened_duplicate): Likewise.
32633 (aarch64_expand_sve_const_vector): Likewise.
32634 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
32635 Handle SVE constants. Use emit_move_insn to move a force_const_mem
32636 into the register, rather than emitting a SET directly.
32637 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
32638 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
32639 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
32640 (offset_9bit_signed_scaled_p): New functions.
32641 (aarch64_replicate_bitmask_imm): New function.
32642 (aarch64_bitmask_imm): Use it.
32643 (aarch64_cannot_force_const_mem): Reject expressions involving
32644 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
32645 (aarch64_classify_index): Handle SVE indices, by requiring
32646 a plain register index with a scale that matches the element size.
32647 (aarch64_classify_address): Handle SVE addresses. Assert that
32648 the mode of the address is VOIDmode or an integer mode.
32649 Update call to aarch64_classify_symbol.
32650 (aarch64_classify_symbolic_expression): Update call to
32651 aarch64_classify_symbol.
32652 (aarch64_const_vec_all_in_range_p): New function.
32653 (aarch64_print_vector_float_operand): Likewise.
32654 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
32655 "vN" for FP registers with SVE modes. Handle (const ...) vectors
32656 and the FP immediates 1.0 and 0.5.
32657 (aarch64_print_address_internal): Handle SVE addresses.
32658 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
32659 (aarch64_regno_regclass): Handle predicate registers.
32660 (aarch64_secondary_reload): Handle big-endian reloads of SVE
32662 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
32663 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
32664 (aarch64_convert_sve_vector_bits): New function.
32665 (aarch64_override_options): Use it to handle -msve-vector-bits=.
32666 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
32667 rather than an rtx.
32668 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
32669 Handle SVE vector and predicate modes. Accept VL-based constants
32670 that need only one temporary register, and VL offsets that require
32671 no temporary registers.
32672 (aarch64_conditional_register_usage): Mark the predicate registers
32673 as fixed if SVE isn't available.
32674 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
32675 Return true for SVE vector and predicate modes.
32676 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
32677 rather than an unsigned int. Handle SVE modes.
32678 (aarch64_preferred_simd_mode): Update call accordingly. Handle
32680 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
32682 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
32683 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
32684 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
32685 (aarch64_sve_float_mul_immediate_p): New functions.
32686 (aarch64_sve_valid_immediate): New function.
32687 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
32688 Explicitly reject structure modes. Check for INDEX constants.
32689 Handle PTRUE and PFALSE constants.
32690 (aarch64_check_zero_based_sve_index_immediate): New function.
32691 (aarch64_simd_imm_zero_p): Delete.
32692 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
32693 vector modes. Accept constants in the range of CNT[BHWD].
32694 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
32695 ask for an Advanced SIMD mode.
32696 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
32697 (aarch64_simd_vector_alignment): Handle SVE predicates.
32698 (aarch64_vectorize_preferred_vector_alignment): New function.
32699 (aarch64_simd_vector_alignment_reachable): Use it instead of
32701 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
32702 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
32704 (MAX_VECT_LEN): Delete.
32705 (expand_vec_perm_d): Add a vec_flags field.
32706 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
32707 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
32708 (aarch64_evpc_ext): Don't apply a big-endian lane correction
32710 (aarch64_evpc_rev): Rename to...
32711 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
32712 (aarch64_evpc_rev_global): New function.
32713 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
32714 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
32716 (aarch64_evpc_sve_tbl): New function.
32717 (aarch64_expand_vec_perm_const_1): Update after rename of
32718 aarch64_evpc_rev. Handle SVE permutes too, trying
32719 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
32720 than aarch64_evpc_tbl.
32721 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
32722 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
32723 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
32724 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
32725 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
32726 (aarch64_expand_sve_vcond): New functions.
32727 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
32728 of aarch64_vector_mode_p.
32729 (aarch64_dwarf_poly_indeterminate_value): New function.
32730 (aarch64_compute_pressure_classes): Likewise.
32731 (aarch64_can_change_mode_class): Likewise.
32732 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
32733 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
32734 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
32735 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
32736 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
32737 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
32738 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
32739 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
32741 (Dn, Dl, Dr): Accept const as well as const_vector.
32742 (Dz): Likewise. Compare against CONST0_RTX.
32743 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
32744 of "vector" where appropriate.
32745 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
32746 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
32747 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
32748 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
32749 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
32750 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
32751 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
32752 (v_int_equiv): Extend to SVE modes.
32753 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
32755 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
32756 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
32757 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
32758 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
32759 (SVE_COND_FP_CMP): New int iterators.
32760 (perm_hilo): Handle the new unpack unspecs.
32761 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
32763 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
32764 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
32765 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
32766 (aarch64_equality_operator, aarch64_constant_vector_operand)
32767 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
32768 (aarch64_sve_nonimmediate_operand): Likewise.
32769 (aarch64_sve_general_operand): Likewise.
32770 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
32771 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
32772 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
32773 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
32774 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
32775 (aarch64_sve_float_arith_immediate): Likewise.
32776 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
32777 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
32778 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
32779 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
32780 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
32781 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
32782 (aarch64_sve_float_arith_operand): Likewise.
32783 (aarch64_sve_float_arith_with_sub_operand): Likewise.
32784 (aarch64_sve_float_mul_operand): Likewise.
32785 (aarch64_sve_vec_perm_operand): Likewise.
32786 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
32787 (aarch64_mov_operand): Accept const_poly_int and const_vector.
32788 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
32789 as well as const_vector.
32790 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
32791 in file. Use CONST0_RTX and CONSTM1_RTX.
32792 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
32793 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
32794 Use aarch64_simd_imm_zero.
32795 * config/aarch64/aarch64-sve.md: New file.
32796 * config/aarch64/aarch64.md: Include it.
32797 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
32798 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
32799 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
32800 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
32801 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
32802 (sve): New attribute.
32803 (enabled): Disable instructions with the sve attribute unless
32805 (movqi, movhi): Pass CONST_POLY_INT operaneds through
32806 aarch64_expand_mov_immediate.
32807 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
32808 CNT[BHSD] immediates.
32809 (movti): Split CONST_POLY_INT moves into two halves.
32810 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
32811 Split additions that need a temporary here if the destination
32812 is the stack pointer.
32813 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
32814 (*add<mode>3_poly_1): New instruction.
32815 (set_clobber_cc): New expander.
32817 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32819 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
32820 parameter and use it instead of GET_MODE_SIZE (innermode). Use
32821 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
32822 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
32823 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
32824 Change innermode from fixed_mode_size to machine_mode.
32825 (simplify_subreg): Update call accordingly. Handle a constant-sized
32826 subreg of a variable-length CONST_VECTOR.
32828 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
32829 Alan Hayward <alan.hayward@arm.com>
32830 David Sherwood <david.sherwood@arm.com>
32832 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
32833 (add_offset_to_base): New function, split out from...
32834 (create_mem_ref): ...here. When handling a scale other than 1,
32835 check first whether the address is valid without the offset.
32836 Add it into the base if so, leaving the index and scale as-is.
32838 2018-01-12 Jakub Jelinek <jakub@redhat.com>
32841 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
32842 fold_for_warn before checking if arg2 is INTEGER_CST.
32844 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
32846 * config/rs6000/predicates.md (load_multiple_operation): Delete.
32847 (store_multiple_operation): Delete.
32848 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
32849 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
32850 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
32851 guarded by TARGET_STRING.
32852 (rs6000_output_load_multiple): Delete.
32853 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
32854 OPTION_MASK_STRING / TARGET_STRING handling.
32855 (print_operand) <'N', 'O'>: Add comment that these are unused now.
32856 (const rs6000_opt_masks) <"string">: Change mask to 0.
32857 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
32858 (MASK_STRING): Delete.
32859 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
32861 (load_multiple): Delete.
32868 (store_multiple): Delete.
32875 (movmemsi_8reg): Delete.
32876 (corresponding unnamed define_insn): Delete.
32877 (movmemsi_6reg): Delete.
32878 (corresponding unnamed define_insn): Delete.
32879 (movmemsi_4reg): Delete.
32880 (corresponding unnamed define_insn): Delete.
32881 (movmemsi_2reg): Delete.
32882 (corresponding unnamed define_insn): Delete.
32883 (movmemsi_1reg): Delete.
32884 (corresponding unnamed define_insn): Delete.
32885 * config/rs6000/rs6000.opt (mno-string): New.
32886 (mstring): Replace by deprecation warning stub.
32887 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
32889 2018-01-12 Jakub Jelinek <jakub@redhat.com>
32891 * regrename.c (regrename_do_replace): If replacing the same
32892 reg multiple times, try to reuse last created gen_raw_REG.
32895 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
32896 main to workaround a bug in GDB.
32898 2018-01-12 Tom de Vries <tom@codesourcery.com>
32901 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
32903 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
32905 PR rtl-optimization/80481
32906 * ira-color.c (get_cap_member): New function.
32907 (allocnos_conflict_by_live_ranges_p): Use it.
32908 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
32909 (setup_slot_coalesced_allocno_live_ranges): Ditto.
32911 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
32914 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
32915 (*saddl_se_1): Ditto.
32916 (*ssubsi_1): Ditto.
32917 (*ssubl_se_1): Ditto.
32919 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
32921 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
32922 rather than wi::to_widest for DR_INITs.
32923 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
32924 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
32925 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
32927 (vect_analyze_group_access_1): Note that here.
32929 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
32931 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
32932 polynomial type sizes.
32934 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
32936 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
32937 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
32938 (gimple_add_tmp_var): Likewise.
32940 2018-01-12 Martin Liska <mliska@suse.cz>
32942 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
32943 (gimple_alloc_sizes): Likewise.
32944 (dump_gimple_statistics): Use PRIu64 in printf format.
32945 * gimple.h: Change uint64_t to int.
32947 2018-01-12 Martin Liska <mliska@suse.cz>
32949 * tree-core.h: Use uint64_t instead of int.
32950 * tree.c (tree_node_counts): Likewise.
32951 (tree_node_sizes): Likewise.
32952 (dump_tree_statistics): Use PRIu64 in printf format.
32954 2018-01-12 Martin Liska <mliska@suse.cz>
32956 * Makefile.in: As qsort_chk is implemented in vec.c, add
32957 vec.o to linkage of gencfn-macros.
32958 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
32959 passing the info to record_node_allocation_statistics.
32960 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
32962 * ggc-common.c (struct ggc_usage): Add operator== and use
32963 it in operator< and compare function.
32964 * mem-stats.h (struct mem_usage): Likewise.
32965 * vec.c (struct vec_usage): Remove operator< and compare
32966 function. Can be simply inherited.
32968 2018-01-12 Martin Jambor <mjambor@suse.cz>
32971 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
32972 * tree-ssa-math-opts.c: Include domwalk.h.
32973 (convert_mult_to_fma_1): New function.
32974 (fma_transformation_info): New type.
32975 (fma_deferring_state): Likewise.
32976 (cancel_fma_deferring): New function.
32977 (result_of_phi): Likewise.
32978 (last_fma_candidate_feeds_initial_phi): Likewise.
32979 (convert_mult_to_fma): Added deferring logic, split actual
32980 transformation to convert_mult_to_fma_1.
32981 (math_opts_dom_walker): New type.
32982 (math_opts_dom_walker::after_dom_children): New method, body moved
32983 here from pass_optimize_widening_mul::execute, added deferring logic
32985 (pass_optimize_widening_mul::execute): Moved most of code to
32986 math_opts_dom_walker::after_dom_children.
32987 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
32988 * config/i386/i386.c (ix86_option_override_internal): Added
32989 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
32991 2018-01-12 Richard Biener <rguenther@suse.de>
32994 * dwarf2out.c (gen_variable_die): Do not reset old_die for
32995 inline instance vars.
32997 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
33000 * config/rx/rx.c (rx_is_restricted_memory_address):
33001 Handle SUBREG case.
33003 2018-01-12 Richard Biener <rguenther@suse.de>
33005 PR tree-optimization/80846
33006 * target.def (split_reduction): New target hook.
33007 * targhooks.c (default_split_reduction): New function.
33008 * targhooks.h (default_split_reduction): Declare.
33009 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
33010 target requests first reduce vectors by combining low and high
33012 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
33013 (get_vectype_for_scalar_type_and_size): Export.
33014 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
33015 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
33016 * doc/tm.texi: Regenerate.
33017 * config/i386/i386.c (ix86_split_reduction): Implement
33018 TARGET_VECTORIZE_SPLIT_REDUCTION.
33020 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
33023 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
33024 in PIC mode except for TARGET_VXWORKS_RTP.
33025 * config/sparc/sparc.c: Include cfgrtl.h.
33026 (TARGET_INIT_PIC_REG): Define.
33027 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
33028 (sparc_pic_register_p): New predicate.
33029 (sparc_legitimate_address_p): Use it.
33030 (sparc_legitimize_pic_address): Likewise.
33031 (sparc_delegitimize_address): Likewise.
33032 (sparc_mode_dependent_address_p): Likewise.
33033 (gen_load_pcrel_sym): Remove 4th parameter.
33034 (load_got_register): Adjust call to above. Remove obsolete stuff.
33035 (sparc_expand_prologue): Do not call load_got_register here.
33036 (sparc_flat_expand_prologue): Likewise.
33037 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
33038 (sparc_use_pseudo_pic_reg): New function.
33039 (sparc_init_pic_reg): Likewise.
33040 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
33041 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
33043 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
33045 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
33046 Add item for branch_cost.
33048 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
33050 PR rtl-optimization/83565
33051 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
33052 not extend the result to a larger mode for rotate operations.
33053 (num_sign_bit_copies1): Likewise.
33055 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
33058 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
33060 Use values-Xc.o for -pedantic.
33061 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
33063 2018-01-12 Martin Liska <mliska@suse.cz>
33066 * ipa-devirt.c (final_warning_record::grow_type_warnings):
33068 (possible_polymorphic_call_targets): Use it.
33069 (ipa_devirt): Likewise.
33071 2018-01-12 Martin Liska <mliska@suse.cz>
33073 * profile-count.h (enum profile_quality): Use 0 as invalid
33074 enum value of profile_quality.
33076 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
33078 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
33079 -mext-string options.
33081 2018-01-12 Richard Biener <rguenther@suse.de>
33083 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
33084 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
33085 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
33087 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
33089 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
33091 * configure.ac (--with-long-double-format): Add support for the
33092 configuration option to change the default long double format on
33094 * config.gcc (powerpc*-linux*-*): Likewise.
33095 * configure: Regenerate.
33096 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
33097 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
33098 used without modification.
33100 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
33102 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
33103 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
33104 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
33105 MISC_BUILTIN_SPEC_BARRIER.
33106 (rs6000_init_builtins): Likewise.
33107 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
33109 (speculation_barrier): New define_insn.
33110 * doc/extend.texi: Document __builtin_speculation_barrier.
33112 2018-01-11 Jakub Jelinek <jakub@redhat.com>
33115 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
33116 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
33117 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
33119 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
33120 integral modes instead of "ss" and "sd".
33121 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
33122 vectors with 32-bit and 64-bit elements.
33123 (vecdupssescalarmodesuffix): New mode attribute.
33124 (vec_dup<mode>): Use it.
33126 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
33129 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
33130 frame if argument is passed on stack.
33132 2018-01-11 Jakub Jelinek <jakub@redhat.com>
33135 * ree.c (combine_reaching_defs): Optimize also
33136 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
33137 reg2=any_extend(exp); reg1=reg2;, formatting fix.
33139 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
33141 PR middle-end/83189
33142 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
33144 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
33146 PR middle-end/83718
33147 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
33148 after they are computed.
33150 2018-01-11 Bin Cheng <bin.cheng@arm.com>
33152 PR tree-optimization/83695
33153 * gimple-loop-linterchange.cc
33154 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
33155 reset cached scev information after interchange.
33156 (pass_linterchange::execute): Remove call to scev_reset_htab.
33158 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33160 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
33161 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
33162 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
33163 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
33164 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
33165 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
33166 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
33167 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
33168 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
33169 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
33170 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
33171 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
33172 (V_lane_reg): Likewise.
33173 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
33175 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
33176 (vfmal_lane_low<mode>_intrinsic,
33177 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
33178 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
33179 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
33180 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
33181 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
33182 vfmsl_lane_high<mode>_intrinsic): New define_insns.
33184 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33186 * config/arm/arm-cpus.in (fp16fml): New feature.
33187 (ALL_SIMD): Add fp16fml.
33188 (armv8.2-a): Add fp16fml as an option.
33189 (armv8.3-a): Likewise.
33190 (armv8.4-a): Add fp16fml as part of fp16.
33191 * config/arm/arm.h (TARGET_FP16FML): Define.
33192 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
33194 * config/arm/arm-modes.def (V2HF): Define.
33195 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
33196 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
33197 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
33198 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
33199 vfmsl_low, vfmsl_high): New set of builtins.
33200 * config/arm/iterators.md (PLUSMINUS): New code iterator.
33201 (vfml_op): New code attribute.
33202 (VFMLHALVES): New int iterator.
33203 (VFML, VFMLSEL): New mode attributes.
33204 (V_reg): Define mapping for V2HF.
33205 (V_hi, V_lo): New mode attributes.
33206 (VF_constraint): Likewise.
33207 (vfml_half, vfml_half_selector): New int attributes.
33208 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
33210 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
33211 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
33213 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
33214 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
33215 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
33216 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
33218 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
33219 Document new effective target and option set.
33221 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
33223 * config/arm/arm-cpus.in (armv8_4): New feature.
33224 (ARMv8_4a): New fgroup.
33225 (armv8.4-a): New arch.
33226 * config/arm/arm-tables.opt: Regenerate.
33227 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
33228 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
33229 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
33230 Add matching rules for -march=armv8.4-a and extensions.
33231 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
33233 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
33236 * config/rx/rx.md (BW): New mode attribute.
33237 (sync_lock_test_and_setsi): Add mode suffix to insn output.
33239 2018-01-11 Richard Biener <rguenther@suse.de>
33241 PR tree-optimization/83435
33242 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
33243 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
33244 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
33246 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
33247 Alan Hayward <alan.hayward@arm.com>
33248 David Sherwood <david.sherwood@arm.com>
33250 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
33252 (aarch64_classify_address): Initialize it. Track polynomial offsets.
33253 (aarch64_print_address_internal): Use it to check for a zero offset.
33255 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
33256 Alan Hayward <alan.hayward@arm.com>
33257 David Sherwood <david.sherwood@arm.com>
33259 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
33260 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
33261 Return a poly_int64 rather than a HOST_WIDE_INT.
33262 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
33263 rather than a HOST_WIDE_INT.
33264 * config/aarch64/aarch64.h (aarch64_frame): Protect with
33265 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
33266 hard_fp_offset, frame_size, initial_adjust, callee_offset and
33267 final_offset from HOST_WIDE_INT to poly_int64.
33268 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
33269 to_constant when getting the number of units in an Advanced SIMD
33271 (aarch64_builtin_vectorized_function): Check for a constant number
33273 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
33275 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
33276 attribute instead of GET_MODE_NUNITS.
33277 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
33278 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
33279 GET_MODE_SIZE for fixed-size registers.
33280 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
33281 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
33282 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
33283 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
33284 (aarch64_print_operand, aarch64_print_address_internal)
33285 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
33286 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
33287 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
33288 Handle polynomial GET_MODE_SIZE.
33289 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
33290 wider than SImode without modification.
33291 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
33292 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
33293 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
33294 passing and returning SVE modes.
33295 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
33296 rather than GEN_INT.
33297 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
33298 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
33299 (aarch64_allocate_and_probe_stack_space): Likewise.
33300 (aarch64_layout_frame): Cope with polynomial offsets.
33301 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
33302 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
33303 polynomial offsets.
33304 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
33305 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
33306 poly_int64 rather than a HOST_WIDE_INT.
33307 (aarch64_get_separate_components, aarch64_process_components)
33308 (aarch64_expand_prologue, aarch64_expand_epilogue)
33309 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
33310 (aarch64_anchor_offset): New function, split out from...
33311 (aarch64_legitimize_address): ...here.
33312 (aarch64_builtin_vectorization_cost): Handle polynomial
33313 TYPE_VECTOR_SUBPARTS.
33314 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
33316 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
33317 number of elements from the PARALLEL rather than the mode.
33318 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
33319 rather than GET_MODE_BITSIZE.
33320 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
33321 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
33322 (aarch64_expand_vec_perm_const_1): Handle polynomial
33323 d->perm.length () and d->perm elements.
33324 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
33325 Apply to_constant to d->perm elements.
33326 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
33327 polynomial CONST_VECTOR_NUNITS.
33328 (aarch64_move_pointer): Take amount as a poly_int64 rather
33330 (aarch64_progress_pointer): Avoid temporary variable.
33331 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
33332 the mode attribute instead of GET_MODE.
33334 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
33335 Alan Hayward <alan.hayward@arm.com>
33336 David Sherwood <david.sherwood@arm.com>
33338 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
33339 x exists before using it.
33340 (aarch64_add_constant_internal): Rename to...
33341 (aarch64_add_offset_1): ...this. Replace regnum with separate
33342 src and dest rtxes. Handle the case in which they're different,
33343 including when the offset is zero. Replace scratchreg with an rtx.
33344 Use 2 additions if there is no spare register into which we can
33345 move a 16-bit constant.
33346 (aarch64_add_constant): Delete.
33347 (aarch64_add_offset): Replace reg with separate src and dest
33348 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
33349 Use aarch64_add_offset_1.
33350 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
33351 an rtx rather than an int. Take the delta as a poly_int64
33352 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
33353 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
33354 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
33355 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
33356 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
33357 and aarch64_add_sp.
33358 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
33359 aarch64_add_constant.
33361 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
33363 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
33364 Use scalar_float_mode.
33366 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
33368 * config/aarch64/aarch64-simd.md
33369 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
33370 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
33371 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
33372 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
33373 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
33374 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
33375 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
33376 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
33377 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
33378 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
33380 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
33383 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
33384 targ_options->x_arm_arch_string is non NULL.
33386 2018-01-11 Tamar Christina <tamar.christina@arm.com>
33388 * config/aarch64/aarch64.h
33389 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
33391 2018-01-11 Sudakshina Das <sudi.das@arm.com>
33394 * expmed.c (emit_store_flag_force): Swap if const op0
33395 and change VOIDmode to mode of op0.
33397 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
33399 PR rtl-optimization/83761
33400 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
33401 than bytes to mode_for_size.
33403 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
33405 PR middle-end/83189
33406 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
33407 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
33410 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
33412 PR middle-end/83575
33413 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
33414 when in layout mode.
33415 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
33416 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
33419 2018-01-10 Michael Collison <michael.collison@arm.com>
33421 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
33422 * config/aarch64/aarch64-option-extension.def: Add
33423 AARCH64_OPT_EXTENSION of 'fp16fml'.
33424 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
33425 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
33426 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
33427 * config/aarch64/constraints.md (Ui7): New constraint.
33428 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
33429 (VFMLA_SEL_W): Ditto.
33432 (VFMLA16_LOW): New int iterator.
33433 (VFMLA16_HIGH): Ditto.
33434 (UNSPEC_FMLAL): New unspec.
33435 (UNSPEC_FMLSL): Ditto.
33436 (UNSPEC_FMLAL2): Ditto.
33437 (UNSPEC_FMLSL2): Ditto.
33438 (f16mac): New code attribute.
33439 * config/aarch64/aarch64-simd-builtins.def
33440 (aarch64_fmlal_lowv2sf): Ditto.
33441 (aarch64_fmlsl_lowv2sf): Ditto.
33442 (aarch64_fmlalq_lowv4sf): Ditto.
33443 (aarch64_fmlslq_lowv4sf): Ditto.
33444 (aarch64_fmlal_highv2sf): Ditto.
33445 (aarch64_fmlsl_highv2sf): Ditto.
33446 (aarch64_fmlalq_highv4sf): Ditto.
33447 (aarch64_fmlslq_highv4sf): Ditto.
33448 (aarch64_fmlal_lane_lowv2sf): Ditto.
33449 (aarch64_fmlsl_lane_lowv2sf): Ditto.
33450 (aarch64_fmlal_laneq_lowv2sf): Ditto.
33451 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
33452 (aarch64_fmlalq_lane_lowv4sf): Ditto.
33453 (aarch64_fmlsl_lane_lowv4sf): Ditto.
33454 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
33455 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
33456 (aarch64_fmlal_lane_highv2sf): Ditto.
33457 (aarch64_fmlsl_lane_highv2sf): Ditto.
33458 (aarch64_fmlal_laneq_highv2sf): Ditto.
33459 (aarch64_fmlsl_laneq_highv2sf): Ditto.
33460 (aarch64_fmlalq_lane_highv4sf): Ditto.
33461 (aarch64_fmlsl_lane_highv4sf): Ditto.
33462 (aarch64_fmlalq_laneq_highv4sf): Ditto.
33463 (aarch64_fmlsl_laneq_highv4sf): Ditto.
33464 * config/aarch64/aarch64-simd.md:
33465 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
33466 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
33467 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
33468 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
33469 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
33470 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
33471 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
33472 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
33473 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
33474 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
33475 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
33476 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
33477 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
33478 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
33479 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
33480 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
33481 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
33482 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
33483 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
33484 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
33485 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
33486 (vfmlsl_low_u32): Ditto.
33487 (vfmlalq_low_u32): Ditto.
33488 (vfmlslq_low_u32): Ditto.
33489 (vfmlal_high_u32): Ditto.
33490 (vfmlsl_high_u32): Ditto.
33491 (vfmlalq_high_u32): Ditto.
33492 (vfmlslq_high_u32): Ditto.
33493 (vfmlal_lane_low_u32): Ditto.
33494 (vfmlsl_lane_low_u32): Ditto.
33495 (vfmlal_laneq_low_u32): Ditto.
33496 (vfmlsl_laneq_low_u32): Ditto.
33497 (vfmlalq_lane_low_u32): Ditto.
33498 (vfmlslq_lane_low_u32): Ditto.
33499 (vfmlalq_laneq_low_u32): Ditto.
33500 (vfmlslq_laneq_low_u32): Ditto.
33501 (vfmlal_lane_high_u32): Ditto.
33502 (vfmlsl_lane_high_u32): Ditto.
33503 (vfmlal_laneq_high_u32): Ditto.
33504 (vfmlsl_laneq_high_u32): Ditto.
33505 (vfmlalq_lane_high_u32): Ditto.
33506 (vfmlslq_lane_high_u32): Ditto.
33507 (vfmlalq_laneq_high_u32): Ditto.
33508 (vfmlslq_laneq_high_u32): Ditto.
33509 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
33510 (AARCH64_FL_FOR_ARCH8_4): New.
33511 (AARCH64_ISA_F16FML): New ISA flag.
33512 (TARGET_F16FML): New feature flag for fp16fml.
33513 (doc/invoke.texi): Document new fp16fml option.
33515 2018-01-10 Michael Collison <michael.collison@arm.com>
33517 * config/aarch64/aarch64-builtins.c:
33518 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
33519 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
33520 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
33521 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
33522 (AARCH64_ISA_SHA3): New ISA flag.
33523 (TARGET_SHA3): New feature flag for sha3.
33524 * config/aarch64/iterators.md (sha512_op): New int attribute.
33525 (CRYPTO_SHA512): New int iterator.
33526 (UNSPEC_SHA512H): New unspec.
33527 (UNSPEC_SHA512H2): Ditto.
33528 (UNSPEC_SHA512SU0): Ditto.
33529 (UNSPEC_SHA512SU1): Ditto.
33530 * config/aarch64/aarch64-simd-builtins.def
33531 (aarch64_crypto_sha512hqv2di): New builtin.
33532 (aarch64_crypto_sha512h2qv2di): Ditto.
33533 (aarch64_crypto_sha512su0qv2di): Ditto.
33534 (aarch64_crypto_sha512su1qv2di): Ditto.
33535 (aarch64_eor3qv8hi): Ditto.
33536 (aarch64_rax1qv2di): Ditto.
33537 (aarch64_xarqv2di): Ditto.
33538 (aarch64_bcaxqv8hi): Ditto.
33539 * config/aarch64/aarch64-simd.md:
33540 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
33541 (aarch64_crypto_sha512su0qv2di): Ditto.
33542 (aarch64_crypto_sha512su1qv2di): Ditto.
33543 (aarch64_eor3qv8hi): Ditto.
33544 (aarch64_rax1qv2di): Ditto.
33545 (aarch64_xarqv2di): Ditto.
33546 (aarch64_bcaxqv8hi): Ditto.
33547 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
33548 (vsha512h2q_u64): Ditto.
33549 (vsha512su0q_u64): Ditto.
33550 (vsha512su1q_u64): Ditto.
33551 (veor3q_u16): Ditto.
33552 (vrax1q_u64): Ditto.
33553 (vxarq_u64): Ditto.
33554 (vbcaxq_u16): Ditto.
33555 * config/arm/types.md (crypto_sha512): New type attribute.
33556 (crypto_sha3): Ditto.
33557 (doc/invoke.texi): Document new sha3 option.
33559 2018-01-10 Michael Collison <michael.collison@arm.com>
33561 * config/aarch64/aarch64-builtins.c:
33562 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
33563 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
33564 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
33565 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
33566 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
33567 (AARCH64_ISA_SM4): New ISA flag.
33568 (TARGET_SM4): New feature flag for sm4.
33569 * config/aarch64/aarch64-simd-builtins.def
33570 (aarch64_sm3ss1qv4si): Ditto.
33571 (aarch64_sm3tt1aq4si): Ditto.
33572 (aarch64_sm3tt1bq4si): Ditto.
33573 (aarch64_sm3tt2aq4si): Ditto.
33574 (aarch64_sm3tt2bq4si): Ditto.
33575 (aarch64_sm3partw1qv4si): Ditto.
33576 (aarch64_sm3partw2qv4si): Ditto.
33577 (aarch64_sm4eqv4si): Ditto.
33578 (aarch64_sm4ekeyqv4si): Ditto.
33579 * config/aarch64/aarch64-simd.md:
33580 (aarch64_sm3ss1qv4si): Ditto.
33581 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
33582 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
33583 (aarch64_sm4eqv4si): Ditto.
33584 (aarch64_sm4ekeyqv4si): Ditto.
33585 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
33586 (sm3part_op): Ditto.
33587 (CRYPTO_SM3TT): Ditto.
33588 (CRYPTO_SM3PART): Ditto.
33589 (UNSPEC_SM3SS1): New unspec.
33590 (UNSPEC_SM3TT1A): Ditto.
33591 (UNSPEC_SM3TT1B): Ditto.
33592 (UNSPEC_SM3TT2A): Ditto.
33593 (UNSPEC_SM3TT2B): Ditto.
33594 (UNSPEC_SM3PARTW1): Ditto.
33595 (UNSPEC_SM3PARTW2): Ditto.
33596 (UNSPEC_SM4E): Ditto.
33597 (UNSPEC_SM4EKEY): Ditto.
33598 * config/aarch64/constraints.md (Ui2): New constraint.
33599 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
33600 * config/arm/types.md (crypto_sm3): New type attribute.
33601 (crypto_sm4): Ditto.
33602 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
33603 (vsm3tt1aq_u32): Ditto.
33604 (vsm3tt1bq_u32): Ditto.
33605 (vsm3tt2aq_u32): Ditto.
33606 (vsm3tt2bq_u32): Ditto.
33607 (vsm3partw1q_u32): Ditto.
33608 (vsm3partw2q_u32): Ditto.
33609 (vsm4eq_u32): Ditto.
33610 (vsm4ekeyq_u32): Ditto.
33611 (doc/invoke.texi): Document new sm4 option.
33613 2018-01-10 Michael Collison <michael.collison@arm.com>
33615 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
33616 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
33617 (AARCH64_FL_FOR_ARCH8_4): New.
33618 (AARCH64_FL_V8_4): New flag.
33619 (doc/invoke.texi): Document new armv8.4-a option.
33621 2018-01-10 Michael Collison <michael.collison@arm.com>
33623 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
33624 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
33625 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
33626 * config/aarch64/aarch64-option-extension.def: Add
33627 AARCH64_OPT_EXTENSION of 'sha2'.
33628 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
33629 (crypto): Disable sha2 and aes if crypto disabled.
33630 (crypto): Enable aes and sha2 if enabled.
33631 (simd): Disable sha2 and aes if simd disabled.
33632 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
33634 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
33635 (TARGET_SHA2): New feature flag for sha2.
33636 (TARGET_AES): New feature flag for aes.
33637 * config/aarch64/aarch64-simd.md:
33638 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
33639 conditional on TARGET_AES.
33640 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
33641 (aarch64_crypto_sha1hsi): Make pattern conditional
33643 (aarch64_crypto_sha1hv4si): Ditto.
33644 (aarch64_be_crypto_sha1hv4si): Ditto.
33645 (aarch64_crypto_sha1su1v4si): Ditto.
33646 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
33647 (aarch64_crypto_sha1su0v4si): Ditto.
33648 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
33649 (aarch64_crypto_sha256su0v4si): Ditto.
33650 (aarch64_crypto_sha256su1v4si): Ditto.
33651 (doc/invoke.texi): Document new aes and sha2 options.
33653 2018-01-10 Martin Sebor <msebor@redhat.com>
33655 PR tree-optimization/83781
33656 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
33659 2018-01-11 Martin Sebor <msebor@gmail.com>
33660 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
33662 PR tree-optimization/83501
33663 PR tree-optimization/81703
33665 * tree-ssa-strlen.c (get_string_cst): Rename...
33666 (get_string_len): ...to this. Handle global constants.
33667 (handle_char_store): Adjust.
33669 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
33670 Jim Wilson <jimw@sifive.com>
33672 * config/riscv/riscv-protos.h (riscv_output_return): New.
33673 * config/riscv/riscv.c (struct machine_function): New naked_p field.
33674 (riscv_attribute_table, riscv_output_return),
33675 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
33676 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
33677 (riscv_compute_frame_info): Only compute frame->mask if not a naked
33679 (riscv_expand_prologue): Add early return for naked function.
33680 (riscv_expand_epilogue): Likewise.
33681 (riscv_function_ok_for_sibcall): Return false for naked function.
33682 (riscv_set_current_function): New.
33683 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
33684 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
33685 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
33686 * doc/extend.texi (RISC-V Function Attributes): New.
33688 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
33690 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
33691 check for 128-bit long double before checking TCmode.
33692 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
33693 128-bit long doubles before checking TFmode or TCmode.
33694 (FLOAT128_IBM_P): Likewise.
33696 2018-01-10 Martin Sebor <msebor@redhat.com>
33698 PR tree-optimization/83671
33699 * builtins.c (c_strlen): Unconditionally return zero for the empty
33701 Use -Warray-bounds for warnings.
33702 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
33703 for non-constant array indices with COMPONENT_REF, arrays of
33704 arrays, and pointers to arrays.
33705 (gimple_fold_builtin_strlen): Determine and set length range for
33706 non-constant character arrays.
33708 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
33710 PR middle-end/81897
33711 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
33714 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
33716 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
33718 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
33721 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
33722 VECTOR_MEM_ALTIVEC_OR_VSX_P.
33723 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
33724 indexed_or_indirect_operand predicate.
33725 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
33726 (*vsx_le_perm_load_v8hi): Likewise.
33727 (*vsx_le_perm_load_v16qi): Likewise.
33728 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
33729 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
33730 (*vsx_le_perm_store_v8hi): Likewise.
33731 (*vsx_le_perm_store_v16qi): Likewise.
33732 (eight unnamed splitters): Likewise.
33734 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
33736 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
33737 * config/rs6000/emmintrin.h: Likewise.
33738 * config/rs6000/mmintrin.h: Likewise.
33739 * config/rs6000/xmmintrin.h: Likewise.
33741 2018-01-10 David Malcolm <dmalcolm@redhat.com>
33744 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
33746 * tree.c (tree_nop_conversion): Return true for location wrapper
33748 (maybe_wrap_with_location): New function.
33749 (selftest::check_strip_nops): New function.
33750 (selftest::test_location_wrappers): New function.
33751 (selftest::tree_c_tests): Call it.
33752 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
33753 (maybe_wrap_with_location): New decl.
33754 (EXPR_LOCATION_WRAPPER_P): New macro.
33755 (location_wrapper_p): New inline function.
33756 (tree_strip_any_location_wrapper): New inline function.
33758 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
33761 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
33762 stack_realign_offset for the largest alignment of stack slot
33764 (ix86_find_max_used_stack_alignment): New function.
33765 (ix86_finalize_stack_frame_flags): Use it. Set
33766 max_used_stack_alignment if we don't realign stack.
33767 * config/i386/i386.h (machine_function): Add
33768 max_used_stack_alignment.
33770 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
33772 * config/arm/arm.opt (-mbranch-cost): New option.
33773 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
33776 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
33779 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
33780 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
33782 2018-01-10 Richard Biener <rguenther@suse.de>
33785 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
33786 early out so it also covers the case where we have a non-NULL
33789 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
33791 PR tree-optimization/83753
33792 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
33793 for non-strided grouped accesses if the number of elements is 1.
33795 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
33798 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
33799 * i386.h (TARGET_USE_GATHER): Define.
33800 * x86-tune.def (X86_TUNE_USE_GATHER): New.
33802 2018-01-10 Martin Liska <mliska@suse.cz>
33805 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
33806 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
33808 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
33809 CLEANUP_NO_PARTITIONING is not set.
33811 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
33813 * doc/rtl.texi: Remove documentation of (const ...) wrappers
33814 for vectors, as a partial revert of r254296.
33815 * rtl.h (const_vec_p): Delete.
33816 (const_vec_duplicate_p): Don't test for vector CONSTs.
33817 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
33818 * expmed.c (make_tree): Likewise.
33821 * common.md (E, F): Use CONSTANT_P instead of checking for
33823 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
33824 checking for CONST_VECTOR.
33826 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
33828 PR middle-end/83575
33829 * predict.c (force_edge_cold): Handle in more sane way edges
33830 with no prediction.
33832 2018-01-09 Carl Love <cel@us.ibm.com>
33834 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
33836 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
33837 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
33838 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
33839 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
33840 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
33841 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
33842 * config/rs6000/rs6000-protos.h: Add extern defition for
33843 rs6000_generate_float2_double_code.
33844 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
33846 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
33847 (float2_v2df): Add define_expand.
33849 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
33852 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
33853 op_mode in the force_to_mode call.
33855 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
33857 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
33858 instead of checking each element individually.
33859 (aarch64_evpc_uzp): Likewise.
33860 (aarch64_evpc_zip): Likewise.
33861 (aarch64_evpc_ext): Likewise.
33862 (aarch64_evpc_rev): Likewise.
33863 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
33864 instead of checking each element individually. Return true without
33866 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
33867 whether all selected elements come from the same input, instead of
33868 checking each element individually. Remove calls to gen_rtx_REG,
33869 start_sequence and end_sequence and instead assert that no rtl is
33872 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
33874 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
33875 order of HIGH and CONST checks.
33877 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
33879 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
33880 if the destination isn't an SSA_NAME.
33882 2018-01-09 Richard Biener <rguenther@suse.de>
33884 PR tree-optimization/83668
33885 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
33887 (canonicalize_loop_form): ... here, renamed from ...
33888 (canonicalize_loop_closed_ssa_form): ... this and amended to
33889 swap successor edges for loop exit blocks to make us use
33890 the RPO order we need for initial schedule generation.
33892 2018-01-09 Joseph Myers <joseph@codesourcery.com>
33894 PR tree-optimization/64811
33895 * match.pd: When optimizing comparisons with Inf, avoid
33896 introducing or losing exceptions from comparisons with NaN.
33898 2018-01-09 Martin Liska <mliska@suse.cz>
33901 * asan.c (shadow_mem_size): Add gcc_assert.
33903 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
33905 Don't save registers in main().
33908 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
33909 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
33910 * config/avr/avr.c (avr_set_current_function): Don't error if
33911 naked, OS_task or OS_main are specified at the same time.
33912 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
33914 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
33916 * common/config/avr/avr-common.c (avr_option_optimization_table):
33917 Switch on -mmain-is-OS_task for optimizing compilations.
33919 2018-01-09 Richard Biener <rguenther@suse.de>
33921 PR tree-optimization/83572
33922 * graphite.c: Include cfganal.h.
33923 (graphite_transform_loops): Connect infinite loops to exit
33924 and remove fake edges at the end.
33926 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
33928 * ipa-inline.c (edge_badness): Revert accidental checkin.
33930 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
33933 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
33934 symbols; not inline clones.
33936 2018-01-09 Jakub Jelinek <jakub@redhat.com>
33939 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
33940 hard registers. Formatting fixes.
33942 PR preprocessor/83722
33943 * gcc.c (try_generate_repro): Pass
33944 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
33945 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
33948 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
33949 Kito Cheng <kito.cheng@gmail.com>
33951 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
33952 (riscv_leaf_function_p): Delete.
33953 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
33955 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
33957 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
33959 (do_ifelse): New function.
33960 (do_isel): New function.
33961 (do_sub3): New function.
33962 (do_add3): New function.
33963 (do_load_mask_compare): New function.
33964 (do_overlap_load_compare): New function.
33965 (expand_compare_loop): New function.
33966 (expand_block_compare): Call expand_compare_loop() when appropriate.
33967 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
33968 option description.
33969 (-mblock-compare-inline-loop-limit): New option.
33971 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
33974 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
33975 Reverse order of second and third operands in first alternative.
33976 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
33977 of first and second elements in UNSPEC_VPERMR vector.
33978 (altivec_expand_vec_perm_le): Likewise.
33980 2018-01-08 Jeff Law <law@redhat.com>
33982 PR rtl-optimizatin/81308
33983 * tree-switch-conversion.c (cfg_altered): New file scoped static.
33984 (process_switch): If group_case_labels makes a change, then set
33986 (pass_convert_switch::execute): If a switch is converted, then
33987 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
33989 PR rtl-optimization/81308
33990 * recog.c (split_all_insns): Conditionally cleanup the CFG after
33993 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
33995 PR target/83663 - Revert r255946
33996 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
33997 generation for cases where splatting a value is not useful.
33998 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
33999 across a vec_duplicate and a paradoxical subreg forming a vector
34000 mode to a vec_concat.
34002 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
34004 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
34005 -march=armv8.3-a variants.
34006 * config/arm/t-multilib: Likewise.
34007 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
34009 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
34011 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
34013 (cceq_ior_compare_complement): Give it a name so I can use it, and
34014 change boolean_or_operator predicate to boolean_operator so it can
34015 be used to generate a crand.
34016 (eqne): New code iterator.
34017 (bd/bd_neg): New code_attrs.
34018 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
34019 a single define_insn.
34020 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
34021 decrement (bdnzt/bdnzf/bdzt/bdzf).
34022 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
34023 with the new names of the branch decrement patterns, and added the
34024 names of the branch decrement conditional patterns.
34026 2018-01-08 Richard Biener <rguenther@suse.de>
34028 PR tree-optimization/83563
34029 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
34032 2018-01-08 Richard Biener <rguenther@suse.de>
34034 PR middle-end/83713
34035 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
34037 2018-01-08 Richard Biener <rguenther@suse.de>
34039 PR tree-optimization/83685
34040 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
34041 references to abnormals.
34043 2018-01-08 Richard Biener <rguenther@suse.de>
34046 * dwarf2out.c (output_indirect_strings): Handle empty
34047 skeleton_debug_str_hash.
34048 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
34050 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
34052 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
34053 (emit_store_direct): Likewise.
34054 (arc_trampoline_adjust_address): Likewise.
34055 (arc_asm_trampoline_template): New function.
34056 (arc_initialize_trampoline): Use asm_trampoline_template.
34057 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
34058 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
34059 * config/arc/arc.md (flush_icache): Delete pattern.
34061 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
34063 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
34064 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
34067 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
34070 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
34071 by not USED_FOR_TARGET.
34072 (make_pass_resolve_sw_modes): Likewise.
34074 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
34076 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
34079 2018-01-08 Richard Biener <rguenther@suse.de>
34081 PR middle-end/83580
34082 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
34084 2018-01-08 Richard Biener <rguenther@suse.de>
34086 PR middle-end/83517
34087 * match.pd ((t * 2) / 2) -> t): Add missing :c.
34089 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
34091 PR middle-end/81897
34092 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
34093 basic blocks with a small number of successors.
34094 (convert_control_dep_chain_into_preds): Improve handling of
34096 (dump_predicates): Split apart into...
34097 (dump_pred_chain): ...here...
34098 (dump_pred_info): ...and here.
34099 (can_one_predicate_be_invalidated_p): Add debugging printfs.
34100 (can_chain_union_be_invalidated_p): Improve check for invalidation
34102 (uninit_uses_cannot_happen): Avoid unnecessary if
34103 convert_control_dep_chain_into_preds yielded nothing.
34105 2018-01-06 Martin Sebor <msebor@redhat.com>
34107 PR tree-optimization/83640
34108 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
34109 subtracting negative offset from size.
34110 (builtin_access::overlap): Adjust offset bounds of the access to fall
34111 within the size of the object if possible.
34113 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
34115 PR rtl-optimization/83699
34116 * expmed.c (extract_bit_field_1): Restrict the vector usage of
34117 extract_bit_field_as_subreg to cases in which the extracted
34118 value is also a vector.
34120 * lra-constraints.c (process_alt_operands): Test for the equivalence
34121 substitutions when detecting a possible reload cycle.
34123 2018-01-06 Jakub Jelinek <jakub@redhat.com>
34126 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
34127 by default if flag_selective_schedling{,2}. Formatting fixes.
34129 PR rtl-optimization/83682
34130 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
34131 if it has non-VECTOR_MODE element mode.
34132 (vec_duplicate_p): Likewise.
34134 PR middle-end/83694
34135 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
34136 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
34138 2018-01-05 Jakub Jelinek <jakub@redhat.com>
34141 * config/i386/i386-builtin.def
34142 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
34143 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
34144 Require also OPTION_MASK_ISA_AVX512F in addition to
34145 OPTION_MASK_ISA_GFNI.
34146 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
34147 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
34148 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
34149 to OPTION_MASK_ISA_GFNI.
34150 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
34151 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
34152 OPTION_MASK_ISA_AVX512BW.
34153 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
34154 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
34155 addition to OPTION_MASK_ISA_GFNI.
34156 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
34157 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
34158 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
34159 to OPTION_MASK_ISA_GFNI.
34160 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
34161 a requirement for all ISAs rather than any of them with a few
34163 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
34165 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
34166 bitmasks to be enabled with 3 exceptions, instead of requiring any
34167 enabled ISA with lots of exceptions.
34168 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
34169 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
34170 Change avx512bw in isa attribute to avx512f.
34171 * config/i386/sgxintrin.h: Add license boilerplate.
34172 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
34173 to __AVX512F__ and __AVX512VL to __AVX512VL__.
34174 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
34175 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
34177 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
34178 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
34179 temporarily sse2 rather than sse if not enabled already.
34182 * config/i386/sse.md (VI248_VLBW): Rename to ...
34183 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
34184 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
34185 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
34186 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
34187 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
34188 mode iterator instead of VI248_VLBW.
34190 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
34192 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
34193 (record_modified): Skip clobbers; add debug output.
34194 (param_change_prob): Use sreal frequencies.
34196 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
34198 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
34199 punt for user-aligned variables.
34201 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
34203 * tree-chrec.c (chrec_contains_symbols): Return true for
34206 2018-01-05 Sudakshina Das <sudi.das@arm.com>
34209 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
34210 of (x|y) == x for BICS pattern.
34212 2018-01-05 Jakub Jelinek <jakub@redhat.com>
34214 PR tree-optimization/83605
34215 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
34216 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
34219 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
34221 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
34222 * config/epiphany/rtems.h: New file.
34224 2018-01-04 Jakub Jelinek <jakub@redhat.com>
34225 Uros Bizjak <ubizjak@gmail.com>
34228 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
34229 QIreg_operand instead of register_operand predicate.
34230 * config/i386/i386.c (ix86_rop_should_change_byte_p,
34231 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
34232 comments instead of -fmitigate[-_]rop.
34234 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
34237 * cgraphunit.c (symbol_table::compile): Switch to text_section
34238 before calling assembly_start debug hook.
34239 * run-rtl-passes.c (run_rtl_passes): Likewise.
34242 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
34244 * tree-vrp.c (extract_range_from_binary_expr_1): Check
34245 range_int_cst_p rather than !symbolic_range_p before calling
34246 extract_range_from_multiplicative_op_1.
34248 2018-01-04 Jeff Law <law@redhat.com>
34250 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
34251 redundant test in assertion.
34253 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
34255 * doc/rtl.texi: Document machine_mode wrapper classes.
34257 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
34259 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
34260 using tree_to_uhwi.
34262 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
34264 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
34265 the VEC_PERM_EXPR fold to fail.
34267 2018-01-04 Jakub Jelinek <jakub@redhat.com>
34270 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
34271 to switched_sections.
34273 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
34276 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
34277 test for d.testing.
34279 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
34282 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
34283 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
34285 2018-01-04 Jakub Jelinek <jakub@redhat.com>
34288 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
34289 is BLKmode and bitpos not zero or mode change is needed.
34291 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
34294 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
34297 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
34300 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
34301 instead of MULT rtx. Update all corresponding splitters.
34302 (*saddl_se): Ditto.
34303 (*ssub<modesuffix>): Ditto.
34304 (*ssubl_se): Ditto.
34305 (*cmp_sadd_di): Update split patterns.
34306 (*cmp_sadd_si): Ditto.
34307 (*cmp_sadd_sidi): Ditto.
34308 (*cmp_ssub_di): Ditto.
34309 (*cmp_ssub_si): Ditto.
34310 (*cmp_ssub_sidi): Ditto.
34311 * config/alpha/predicates.md (const23_operand): New predicate.
34312 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
34313 Look for ASHIFT, not MULT inner operand.
34314 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
34316 2018-01-04 Martin Liska <mliska@suse.cz>
34318 PR gcov-profile/83669
34319 * gcov.c (output_intermediate_file): Add version to intermediate
34321 * doc/gcov.texi: Document new field 'version' in intermediate
34322 file format. Fix location of '-k' option of gcov command.
34324 2018-01-04 Martin Liska <mliska@suse.cz>
34327 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
34329 2018-01-04 Jakub Jelinek <jakub@redhat.com>
34331 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
34333 2018-01-03 Martin Sebor <msebor@redhat.com>
34335 PR tree-optimization/83655
34336 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
34337 checking calls with invalid arguments.
34339 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34341 * tree-vect-stmts.c (vect_get_store_rhs): New function.
34342 (vectorizable_mask_load_store): Delete.
34343 (vectorizable_call): Return false for masked loads and stores.
34344 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
34345 instead of gimple_assign_rhs1.
34346 (vectorizable_load): Handle IFN_MASK_LOAD.
34347 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
34349 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34351 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
34353 (vectorizable_mask_load_store): ...here.
34354 (vectorizable_load): ...and here.
34356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34358 * tree-vect-stmts.c (vect_build_all_ones_mask)
34359 (vect_build_zero_merge_argument): New functions, split out from...
34360 (vectorizable_load): ...here.
34362 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34364 * tree-vect-stmts.c (vect_check_store_rhs): New function,
34366 (vectorizable_mask_load_store): ...here.
34367 (vectorizable_store): ...and here.
34369 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34371 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
34373 (vectorizable_mask_load_store): ...here.
34375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34377 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
34378 (vect_model_store_cost): Take a vec_load_store_type instead of a
34380 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
34381 (vect_model_store_cost): Take a vec_load_store_type instead of a
34383 (vectorizable_mask_load_store): Update accordingly.
34384 (vectorizable_store): Likewise.
34385 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
34387 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34389 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
34390 IFN_MASK_LOAD calls here rather than...
34391 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
34393 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34394 Alan Hayward <alan.hayward@arm.com>
34395 David Sherwood <david.sherwood@arm.com>
34397 * expmed.c (extract_bit_field_1): For vector extracts,
34398 fall back to extract_bit_field_as_subreg if vec_extract
34401 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34402 Alan Hayward <alan.hayward@arm.com>
34403 David Sherwood <david.sherwood@arm.com>
34405 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
34406 they are variable or constant sized.
34407 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
34408 slots for constant-sized data.
34410 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34411 Alan Hayward <alan.hayward@arm.com>
34412 David Sherwood <david.sherwood@arm.com>
34414 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
34415 handling COND_EXPRs with boolean comparisons, try to find a better
34416 basis for the mask type than the boolean itself.
34418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34420 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
34421 is calculated and how it can be overridden.
34422 * genmodes.c (max_bitsize_mode_any_mode): New variable.
34423 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
34425 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
34428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34429 Alan Hayward <alan.hayward@arm.com>
34430 David Sherwood <david.sherwood@arm.com>
34432 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
34433 Remove the mode argument.
34434 (aarch64_simd_valid_immediate): Remove the mode and inverse
34436 * config/aarch64/iterators.md (bitsize): New iterator.
34437 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
34438 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
34439 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
34440 aarch64_simd_valid_immediate.
34441 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
34442 (aarch64_reg_or_bic_imm): Likewise.
34443 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
34444 with an insn_type enum and msl with a modifier_type enum.
34445 Replace element_width with a scalar_mode. Change the shift
34446 to unsigned int. Add constructors for scalar_float_mode and
34447 scalar_int_mode elements.
34448 (aarch64_vect_float_const_representable_p): Delete.
34449 (aarch64_can_const_movi_rtx_p)
34450 (aarch64_simd_scalar_immediate_valid_for_move)
34451 (aarch64_simd_make_constant): Update call to
34452 aarch64_simd_valid_immediate.
34453 (aarch64_advsimd_valid_immediate_hs): New function.
34454 (aarch64_advsimd_valid_immediate): Likewise.
34455 (aarch64_simd_valid_immediate): Remove mode and inverse
34456 arguments. Rewrite to use the above. Use const_vec_duplicate_p
34457 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
34458 and aarch64_float_const_representable_p on the result.
34459 (aarch64_output_simd_mov_immediate): Remove mode argument.
34460 Update call to aarch64_simd_valid_immediate and use of
34461 simd_immediate_info.
34462 (aarch64_output_scalar_simd_mov_immediate): Update call
34465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34466 Alan Hayward <alan.hayward@arm.com>
34467 David Sherwood <david.sherwood@arm.com>
34469 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
34470 (mode_nunits): Likewise CONST_MODE_NUNITS.
34471 * machmode.def (ADJUST_NUNITS): Document.
34472 * genmodes.c (mode_data::need_nunits_adj): New field.
34473 (blank_mode): Update accordingly.
34474 (adj_nunits): New variable.
34475 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
34477 (emit_mode_size_inline): Set need_bytesize_adj for all modes
34478 listed in adj_nunits.
34479 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
34480 listed in adj_nunits. Don't emit case statements for such modes.
34481 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
34482 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
34483 nothing if adj_nunits is nonnull.
34484 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
34485 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
34486 (emit_mode_fbit): Update use of print_maybe_const_decl.
34487 (emit_move_size): Likewise. Treat the array as non-const
34489 (emit_mode_adjustments): Handle adj_nunits.
34491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34493 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
34494 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
34495 (VECTOR_MODES): Use it.
34496 (make_vector_modes): Take the prefix as an argument.
34498 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34499 Alan Hayward <alan.hayward@arm.com>
34500 David Sherwood <david.sherwood@arm.com>
34502 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
34503 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
34504 for MODE_VECTOR_BOOL.
34505 * machmode.def (VECTOR_BOOL_MODE): Document.
34506 * genmodes.c (VECTOR_BOOL_MODE): New macro.
34507 (make_vector_bool_mode): New function.
34508 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
34510 * lto-streamer-in.c (lto_input_mode_table): Likewise.
34511 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
34513 * stor-layout.c (int_mode_for_mode): Likewise.
34514 * tree.c (build_vector_type_for_mode): Likewise.
34515 * varasm.c (output_constant_pool_2): Likewise.
34516 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
34517 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
34518 for MODE_VECTOR_BOOL.
34519 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
34520 of mode class checks.
34521 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
34522 instead of a list of mode class checks.
34523 (expand_vector_scalar_condition): Likewise.
34524 (type_for_widest_vector_mode): Handle BImode as an inner mode.
34526 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34527 Alan Hayward <alan.hayward@arm.com>
34528 David Sherwood <david.sherwood@arm.com>
34530 * machmode.h (mode_size): Change from unsigned short to
34532 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
34533 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
34534 or if measurement_type is not polynomial.
34535 (fixed_size_mode::includes_p): Check for constant-sized modes.
34536 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
34537 return a poly_uint16 rather than an unsigned short.
34538 (emit_mode_size): Change the type of mode_size from unsigned short
34539 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
34540 (emit_mode_adjustments): Cope with polynomial vector sizes.
34541 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
34543 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
34545 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
34546 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
34547 * caller-save.c (setup_save_areas): Likewise.
34548 (replace_reg_with_saved_mem): Likewise.
34549 * calls.c (emit_library_call_value_1): Likewise.
34550 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
34551 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
34552 (gen_lowpart_for_combine): Likewise.
34553 * convert.c (convert_to_integer_1): Likewise.
34554 * cse.c (equiv_constant, cse_insn): Likewise.
34555 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
34556 (cselib_subst_to_values): Likewise.
34557 * dce.c (word_dce_process_block): Likewise.
34558 * df-problems.c (df_word_lr_mark_ref): Likewise.
34559 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
34560 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
34561 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
34562 (rtl_for_decl_location): Likewise.
34563 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
34564 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
34565 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
34566 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
34567 (expand_expr_real_1): Likewise.
34568 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
34569 (pad_below): Likewise.
34570 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
34571 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
34572 * ira.c (get_subreg_tracking_sizes): Likewise.
34573 * ira-build.c (ira_create_allocno_objects): Likewise.
34574 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
34575 (ira_sort_regnos_for_alter_reg): Likewise.
34576 * ira-costs.c (record_operand_costs): Likewise.
34577 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
34578 (resolve_simple_move): Likewise.
34579 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
34580 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
34581 (lra_constraints): Likewise.
34582 (CONST_POOL_OK_P): Reject variable-sized modes.
34583 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
34584 (add_pseudo_to_slot, lra_spill): Likewise.
34585 * omp-low.c (omp_clause_aligned_alignment): Likewise.
34586 * optabs-query.c (get_best_extraction_insn): Likewise.
34587 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
34588 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
34589 (expand_mult_highpart, valid_multiword_target_p): Likewise.
34590 * recog.c (offsettable_address_addr_space_p): Likewise.
34591 * regcprop.c (maybe_mode_change): Likewise.
34592 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
34593 * regrename.c (build_def_use): Likewise.
34594 * regstat.c (dump_reg_info): Likewise.
34595 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
34596 (find_reloads, find_reloads_subreg_address): Likewise.
34597 * reload1.c (eliminate_regs_1): Likewise.
34598 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
34599 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
34600 (simplify_binary_operation_1, simplify_subreg): Likewise.
34601 * targhooks.c (default_function_arg_padding): Likewise.
34602 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
34603 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
34604 (verify_gimple_assign_ternary): Likewise.
34605 * tree-inline.c (estimate_move_cost): Likewise.
34606 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
34607 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
34608 (get_address_cost_ainc): Likewise.
34609 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
34610 (vect_supportable_dr_alignment): Likewise.
34611 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
34612 (vectorizable_reduction): Likewise.
34613 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
34614 (vectorizable_operation, vectorizable_load): Likewise.
34615 * tree.c (build_same_sized_truth_vector_type): Likewise.
34616 * valtrack.c (cleanup_auto_inc_dec): Likewise.
34617 * var-tracking.c (emit_note_insn_var_location): Likewise.
34618 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
34619 (ADDR_VEC_ALIGN): Likewise.
34621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34622 Alan Hayward <alan.hayward@arm.com>
34623 David Sherwood <david.sherwood@arm.com>
34625 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
34627 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
34628 or if measurement_type is polynomial.
34629 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
34630 * combine.c (make_extraction): Likewise.
34631 * dse.c (find_shift_sequence): Likewise.
34632 * dwarf2out.c (mem_loc_descriptor): Likewise.
34633 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
34634 (extract_bit_field, extract_low_bits): Likewise.
34635 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
34636 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
34637 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
34638 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
34639 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
34640 * reload.c (find_reloads): Likewise.
34641 * reload1.c (alter_reg): Likewise.
34642 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
34643 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
34644 * tree-if-conv.c (predicate_mem_writes): Likewise.
34645 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
34646 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
34647 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
34648 * valtrack.c (dead_debug_insert_temp): Likewise.
34649 * varasm.c (mergeable_constant_section): Likewise.
34650 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
34652 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34653 Alan Hayward <alan.hayward@arm.com>
34654 David Sherwood <david.sherwood@arm.com>
34656 * expr.c (expand_assignment): Cope with polynomial mode sizes
34657 when assigning to a CONCAT.
34659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34660 Alan Hayward <alan.hayward@arm.com>
34661 David Sherwood <david.sherwood@arm.com>
34663 * machmode.h (mode_precision): Change from unsigned short to
34665 (mode_to_precision): Return a poly_uint16 rather than an unsigned
34667 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
34668 or if measurement_type is not polynomial.
34669 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
34670 in which the mode is already known to be a scalar_int_mode.
34671 * genmodes.c (emit_mode_precision): Change the type of mode_precision
34672 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
34674 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
34675 for GET_MODE_PRECISION.
34676 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
34677 for GET_MODE_PRECISION.
34678 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
34680 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
34681 (expand_field_assignment, make_extraction): Likewise.
34682 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
34683 (get_last_value): Likewise.
34684 * convert.c (convert_to_integer_1): Likewise.
34685 * cse.c (cse_insn): Likewise.
34686 * expr.c (expand_expr_real_1): Likewise.
34687 * lra-constraints.c (simplify_operand_subreg): Likewise.
34688 * optabs-query.c (can_atomic_load_p): Likewise.
34689 * optabs.c (expand_atomic_load): Likewise.
34690 (expand_atomic_store): Likewise.
34691 * ree.c (combine_reaching_defs): Likewise.
34692 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
34693 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
34694 * tree.h (type_has_mode_precision_p): Likewise.
34695 * ubsan.c (instrument_si_overflow): Likewise.
34697 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34698 Alan Hayward <alan.hayward@arm.com>
34699 David Sherwood <david.sherwood@arm.com>
34701 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
34702 polynomial numbers of units.
34703 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
34704 (valid_vector_subparts_p): New function.
34705 (build_vector_type): Remove temporary shim and take the number
34706 of units as a poly_uint64 rather than an int.
34707 (build_opaque_vector_type): Take the number of units as a
34708 poly_uint64 rather than an int.
34709 * tree.c (build_vector_from_ctor): Handle polynomial
34710 TYPE_VECTOR_SUBPARTS.
34711 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
34712 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
34713 (build_vector_from_val): If the number of units is variable,
34714 use build_vec_duplicate_cst for constant operands and
34715 VEC_DUPLICATE_EXPR otherwise.
34716 (make_vector_type): Remove temporary is_constant ().
34717 (build_vector_type, build_opaque_vector_type): Take the number of
34718 units as a poly_uint64 rather than an int.
34719 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
34721 * cfgexpand.c (expand_debug_expr): Likewise.
34722 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
34723 (store_constructor, expand_expr_real_1): Likewise.
34724 (const_scalar_mask_from_tree): Likewise.
34725 * fold-const-call.c (fold_const_reduction): Likewise.
34726 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
34727 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
34728 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
34729 (fold_relational_const): Likewise.
34730 (native_interpret_vector): Likewise. Change the size from an
34731 int to an unsigned int.
34732 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
34733 TYPE_VECTOR_SUBPARTS.
34734 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
34735 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
34736 duplicating a non-constant operand into a variable-length vector.
34737 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
34738 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
34739 * ipa-icf.c (sem_variable::equals): Likewise.
34740 * match.pd: Likewise.
34741 * omp-simd-clone.c (simd_clone_subparts): Likewise.
34742 * print-tree.c (print_node): Likewise.
34743 * stor-layout.c (layout_type): Likewise.
34744 * targhooks.c (default_builtin_vectorization_cost): Likewise.
34745 * tree-cfg.c (verify_gimple_comparison): Likewise.
34746 (verify_gimple_assign_binary): Likewise.
34747 (verify_gimple_assign_ternary): Likewise.
34748 (verify_gimple_assign_single): Likewise.
34749 * tree-pretty-print.c (dump_generic_node): Likewise.
34750 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
34751 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
34752 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
34753 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
34754 (vect_shift_permute_load_chain): Likewise.
34755 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
34756 (expand_vector_condition, optimize_vector_constructor): Likewise.
34757 (lower_vec_perm, get_compute_type): Likewise.
34758 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
34759 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
34760 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
34761 (vect_recog_mask_conversion_pattern): Likewise.
34762 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
34763 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
34764 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
34765 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
34766 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
34767 (vectorizable_shift, vectorizable_operation, vectorizable_store)
34768 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
34769 (supportable_widening_operation): Likewise.
34770 (supportable_narrowing_operation): Likewise.
34771 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
34773 * varasm.c (output_constant): Likewise.
34775 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34776 Alan Hayward <alan.hayward@arm.com>
34777 David Sherwood <david.sherwood@arm.com>
34779 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
34780 so that both the length == 3 and length != 3 cases set up their
34781 own permute vectors. Add comments explaining why we know the
34782 number of elements is constant.
34783 (vect_permute_load_chain): Likewise.
34785 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34786 Alan Hayward <alan.hayward@arm.com>
34787 David Sherwood <david.sherwood@arm.com>
34789 * machmode.h (mode_nunits): Change from unsigned char to
34791 (ONLY_FIXED_SIZE_MODES): New macro.
34792 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
34793 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
34794 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
34796 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
34797 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
34798 or if measurement_type is not polynomial.
34799 * genmodes.c (ZERO_COEFFS): New macro.
34800 (emit_mode_nunits_inline): Make mode_nunits_inline return a
34802 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
34803 Use ZERO_COEFFS when emitting initializers.
34804 * data-streamer.h (bp_pack_poly_value): New function.
34805 (bp_unpack_poly_value): Likewise.
34806 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
34807 for GET_MODE_NUNITS.
34808 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
34809 for GET_MODE_NUNITS.
34810 * tree.c (make_vector_type): Remove temporary shim and make
34811 the real function take the number of units as a poly_uint64
34812 rather than an int.
34813 (build_vector_type_for_mode): Handle polynomial nunits.
34814 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
34815 * emit-rtl.c (const_vec_series_p_1): Likewise.
34816 (gen_rtx_CONST_VECTOR): Likewise.
34817 * fold-const.c (test_vec_duplicate_folding): Likewise.
34818 * genrecog.c (validate_pattern): Likewise.
34819 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
34820 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
34821 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
34822 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
34823 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
34824 * rtlanal.c (subreg_get_info): Likewise.
34825 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
34826 (vect_grouped_load_supported): Likewise.
34827 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
34828 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
34829 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
34830 (simplify_const_unary_operation, simplify_binary_operation_1)
34831 (simplify_const_binary_operation, simplify_ternary_operation)
34832 (test_vector_ops_duplicate, test_vector_ops): Likewise.
34833 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
34834 instead of CONST_VECTOR_NUNITS.
34835 * varasm.c (output_constant_pool_2): Likewise.
34836 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
34837 explicit-encoded elements in the XVEC for variable-length vectors.
34839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34841 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
34843 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34844 Alan Hayward <alan.hayward@arm.com>
34845 David Sherwood <david.sherwood@arm.com>
34847 * coretypes.h (fixed_size_mode): Declare.
34848 (fixed_size_mode_pod): New typedef.
34849 * builtins.h (target_builtins::x_apply_args_mode)
34850 (target_builtins::x_apply_result_mode): Change type to
34851 fixed_size_mode_pod.
34852 * builtins.c (apply_args_size, apply_result_size, result_vector)
34853 (expand_builtin_apply_args_1, expand_builtin_apply)
34854 (expand_builtin_return): Update accordingly.
34856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34858 * cse.c (hash_rtx_cb): Hash only the encoded elements.
34859 * cselib.c (cselib_hash_rtx): Likewise.
34860 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
34861 CONST_VECTOR encoding.
34863 2018-01-03 Jakub Jelinek <jakub@redhat.com>
34864 Jeff Law <law@redhat.com>
34867 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
34868 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
34869 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
34870 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
34873 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
34874 explicitly probe *sp in a noreturn function if there were any callee
34875 register saves or frame pointer is needed.
34877 2018-01-03 Jakub Jelinek <jakub@redhat.com>
34880 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
34881 BLKmode for ternary, binary or unary expressions.
34884 * var-tracking.c (delete_vta_debug_insn): New inline function.
34885 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
34886 insns from get_insns () to NULL instead of each bb separately.
34887 Use delete_vta_debug_insn. No longer static.
34888 (vt_debug_insns_local, variable_tracking_main_1): Adjust
34889 delete_vta_debug_insns callers.
34890 * rtl.h (delete_vta_debug_insns): Declare.
34891 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
34892 instead of variable_tracking_main.
34894 2018-01-03 Martin Sebor <msebor@redhat.com>
34896 PR tree-optimization/83603
34897 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
34898 arguments past the endof the argument list in functions declared
34899 without a prototype.
34900 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
34901 Avoid checking when arguments are null.
34903 2018-01-03 Martin Sebor <msebor@redhat.com>
34906 * doc/extend.texi (attribute const): Fix a typo.
34907 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
34908 issuing -Wsuggest-attribute for void functions.
34910 2018-01-03 Martin Sebor <msebor@redhat.com>
34912 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
34913 offset_int::from instead of wide_int::to_shwi.
34914 (maybe_diag_overlap): Remove assertion.
34915 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
34916 * gimple-ssa-sprintf.c (format_directive): Same.
34917 (parse_directive): Same.
34918 (sprintf_dom_walker::compute_format_length): Same.
34919 (try_substitute_return_value): Same.
34921 2018-01-03 Jeff Law <law@redhat.com>
34923 PR middle-end/83654
34924 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
34925 non-constant residual for zero at runtime and avoid probing in
34926 that case. Reorganize code for trailing problem to mirror handling
34929 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
34931 PR tree-optimization/83501
34932 * tree-ssa-strlen.c (get_string_cst): New.
34933 (handle_char_store): Call get_string_cst.
34935 2018-01-03 Martin Liska <mliska@suse.cz>
34937 PR tree-optimization/83593
34938 * tree-ssa-strlen.c: Include tree-cfg.h.
34939 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
34940 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
34941 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
34943 (strlen_dom_walker::before_dom_children): Call
34944 gimple_purge_dead_eh_edges. Dump tranformation with details
34946 (strlen_dom_walker::before_dom_children): Update call by adding
34947 new argument cleanup_eh.
34948 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
34950 2018-01-03 Martin Liska <mliska@suse.cz>
34953 * cif-code.def (VARIADIC_THUNK): New enum value.
34954 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
34957 2018-01-03 Jan Beulich <jbeulich@suse.com>
34959 * sse.md (mov<mode>_internal): Tighten condition for when to use
34960 vmovdqu<ssescalarsize> for TI and OI modes.
34962 2018-01-03 Jakub Jelinek <jakub@redhat.com>
34964 Update copyright years.
34966 2018-01-03 Martin Liska <mliska@suse.cz>
34969 * ipa-visibility.c (function_and_variable_visibility): Skip
34970 functions with noipa attribure.
34972 2018-01-03 Jakub Jelinek <jakub@redhat.com>
34974 * gcc.c (process_command): Update copyright notice dates.
34975 * gcov-dump.c (print_version): Ditto.
34976 * gcov.c (print_version): Ditto.
34977 * gcov-tool.c (print_version): Ditto.
34978 * gengtype.c (create_file): Ditto.
34979 * doc/cpp.texi: Bump @copying's copyright year.
34980 * doc/cppinternals.texi: Ditto.
34981 * doc/gcc.texi: Ditto.
34982 * doc/gccint.texi: Ditto.
34983 * doc/gcov.texi: Ditto.
34984 * doc/install.texi: Ditto.
34985 * doc/invoke.texi: Ditto.
34987 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
34989 * vector-builder.h (vector_builder::m_full_nelts): Change from
34990 unsigned int to poly_uint64.
34991 (vector_builder::full_nelts): Update prototype accordingly.
34992 (vector_builder::new_vector): Likewise.
34993 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
34994 (vector_builder::operator ==): Likewise.
34995 (vector_builder::finalize): Likewise.
34996 * int-vector-builder.h (int_vector_builder::int_vector_builder):
34997 Take the number of elements as a poly_uint64 rather than an
34999 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
35000 from unsigned int to poly_uint64.
35001 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
35002 (vec_perm_indices::new_vector): Likewise.
35003 (vec_perm_indices::length): Likewise.
35004 (vec_perm_indices::nelts_per_input): Likewise.
35005 (vec_perm_indices::input_nelts): Likewise.
35006 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
35007 number of elements per input as a poly_uint64 rather than an
35008 unsigned int. Use the original encoding for variable-length
35009 vectors, rather than clamping each individual element.
35010 For the second and subsequent elements in each pattern,
35011 clamp the step and base before clamping their sum.
35012 (vec_perm_indices::series_p): Handle polynomial element counts.
35013 (vec_perm_indices::all_in_range_p): Likewise.
35014 (vec_perm_indices_to_tree): Likewise.
35015 (vec_perm_indices_to_rtx): Likewise.
35016 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
35017 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
35018 (tree_vector_builder::new_binary_operation): Handle polynomial
35019 element counts. Return false if we need to know the number
35020 of elements at compile time.
35021 * fold-const.c (fold_vec_perm): Punt if the number of elements
35022 isn't known at compile time.
35024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35026 * vec-perm-indices.h (vec_perm_builder): Change element type
35027 from HOST_WIDE_INT to poly_int64.
35028 (vec_perm_indices::element_type): Update accordingly.
35029 (vec_perm_indices::clamp): Handle polynomial element_types.
35030 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
35031 (vec_perm_indices::all_in_range_p): Likewise.
35032 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
35034 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
35035 polynomial vec_perm_indices element types.
35036 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
35037 * fold-const.c (fold_vec_perm): Likewise.
35038 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
35039 * tree-vect-generic.c (lower_vec_perm): Likewise.
35040 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
35041 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
35042 element type to HOST_WIDE_INT.
35044 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35045 Alan Hayward <alan.hayward@arm.com>
35046 David Sherwood <david.sherwood@arm.com>
35048 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
35049 rather than an int. Use plus_constant.
35050 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
35051 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
35053 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35054 Alan Hayward <alan.hayward@arm.com>
35055 David Sherwood <david.sherwood@arm.com>
35057 * calls.c (emit_call_1, expand_call): Change struct_value_size from
35058 a HOST_WIDE_INT to a poly_int64.
35060 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35061 Alan Hayward <alan.hayward@arm.com>
35062 David Sherwood <david.sherwood@arm.com>
35064 * calls.c (load_register_parameters): Cope with polynomial
35065 mode sizes. Require a constant size for BLKmode parameters
35066 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
35067 forces a parameter to be padded at the lsb end in order to
35068 fill a complete number of words, require the parameter size
35069 to be ordered wrt UNITS_PER_WORD.
35071 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35072 Alan Hayward <alan.hayward@arm.com>
35073 David Sherwood <david.sherwood@arm.com>
35075 * reload1.c (spill_stack_slot_width): Change element type
35076 from unsigned int to poly_uint64_pod.
35077 (alter_reg): Treat mode sizes as polynomial.
35079 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35080 Alan Hayward <alan.hayward@arm.com>
35081 David Sherwood <david.sherwood@arm.com>
35083 * reload.c (complex_word_subreg_p): New function.
35084 (reload_inner_reg_of_subreg, push_reload): Use it.
35086 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35087 Alan Hayward <alan.hayward@arm.com>
35088 David Sherwood <david.sherwood@arm.com>
35090 * lra-constraints.c (process_alt_operands): Reject matched
35091 operands whose sizes aren't ordered.
35092 (match_reload): Refer to this check here.
35094 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35095 Alan Hayward <alan.hayward@arm.com>
35096 David Sherwood <david.sherwood@arm.com>
35098 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
35099 that the mode size is in the set {1, 2, 4, 8, 16}.
35101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35102 Alan Hayward <alan.hayward@arm.com>
35103 David Sherwood <david.sherwood@arm.com>
35105 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
35106 Use plus_constant instead of gen_rtx_PLUS.
35108 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35109 Alan Hayward <alan.hayward@arm.com>
35110 David Sherwood <david.sherwood@arm.com>
35112 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
35113 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
35114 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
35115 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
35116 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
35117 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
35118 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
35119 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
35120 * config/i386/i386.c (ix86_push_rounding): ...this new function.
35121 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
35123 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
35124 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
35125 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
35126 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
35127 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
35128 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
35129 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
35130 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
35131 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
35132 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
35134 * expr.c (emit_move_resolve_push): Treat the input and result
35135 of PUSH_ROUNDING as a poly_int64.
35136 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
35137 (emit_push_insn): Likewise.
35138 * lra-eliminations.c (mark_not_eliminable): Likewise.
35139 * recog.c (push_operand): Likewise.
35140 * reload1.c (elimination_effects): Likewise.
35141 * rtlanal.c (nonzero_bits1): Likewise.
35142 * calls.c (store_one_arg): Likewise. Require the padding to be
35143 known at compile time.
35145 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35146 Alan Hayward <alan.hayward@arm.com>
35147 David Sherwood <david.sherwood@arm.com>
35149 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
35150 Use plus_constant instead of gen_rtx_PLUS.
35152 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35153 Alan Hayward <alan.hayward@arm.com>
35154 David Sherwood <david.sherwood@arm.com>
35156 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
35157 rather than an int.
35159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35160 Alan Hayward <alan.hayward@arm.com>
35161 David Sherwood <david.sherwood@arm.com>
35163 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
35164 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
35165 via stack temporaries. Treat the mode size as polynomial too.
35167 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35168 Alan Hayward <alan.hayward@arm.com>
35169 David Sherwood <david.sherwood@arm.com>
35171 * expr.c (expand_expr_real_2): When handling conversions involving
35172 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
35173 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
35174 as a poly_uint64 too.
35176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35177 Alan Hayward <alan.hayward@arm.com>
35178 David Sherwood <david.sherwood@arm.com>
35180 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
35182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35183 Alan Hayward <alan.hayward@arm.com>
35184 David Sherwood <david.sherwood@arm.com>
35186 * combine.c (can_change_dest_mode): Handle polynomial
35187 REGMODE_NATURAL_SIZE.
35188 * expmed.c (store_bit_field_1): Likewise.
35189 * expr.c (store_constructor): Likewise.
35190 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
35191 and polynomial REGMODE_NATURAL_SIZE.
35192 (gen_lowpart_common): Likewise.
35193 * reginfo.c (record_subregs_of_mode): Likewise.
35194 * rtlanal.c (read_modify_subreg_p): Likewise.
35196 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35197 Alan Hayward <alan.hayward@arm.com>
35198 David Sherwood <david.sherwood@arm.com>
35200 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
35201 numbers of elements.
35203 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35204 Alan Hayward <alan.hayward@arm.com>
35205 David Sherwood <david.sherwood@arm.com>
35207 * match.pd: Cope with polynomial numbers of vector elements.
35209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35210 Alan Hayward <alan.hayward@arm.com>
35211 David Sherwood <david.sherwood@arm.com>
35213 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
35214 in a POINTER_PLUS_EXPR.
35216 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35217 Alan Hayward <alan.hayward@arm.com>
35218 David Sherwood <david.sherwood@arm.com>
35220 * omp-simd-clone.c (simd_clone_subparts): New function.
35221 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
35222 (ipa_simd_modify_function_body): Likewise.
35224 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35225 Alan Hayward <alan.hayward@arm.com>
35226 David Sherwood <david.sherwood@arm.com>
35228 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
35229 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
35230 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
35231 (expand_vector_condition, vector_element): Likewise.
35232 (subparts_gt): New function.
35233 (get_compute_type): Use subparts_gt.
35234 (count_type_subparts): Delete.
35235 (expand_vector_operations_1): Use subparts_gt instead of
35236 count_type_subparts.
35238 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35239 Alan Hayward <alan.hayward@arm.com>
35240 David Sherwood <david.sherwood@arm.com>
35242 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
35243 (vect_compile_time_alias): ...this new function. Do the calculation
35244 on poly_ints rather than trees.
35245 (vect_prune_runtime_alias_test_list): Update call accordingly.
35247 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35248 Alan Hayward <alan.hayward@arm.com>
35249 David Sherwood <david.sherwood@arm.com>
35251 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
35253 (vect_schedule_slp_instance): Likewise.
35255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35256 Alan Hayward <alan.hayward@arm.com>
35257 David Sherwood <david.sherwood@arm.com>
35259 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
35260 constant and extern definitions for variable-length vectors.
35261 (vect_get_constant_vectors): Note that the number of units
35262 is known to be constant.
35264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35265 Alan Hayward <alan.hayward@arm.com>
35266 David Sherwood <david.sherwood@arm.com>
35268 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
35269 of units as polynomial. Choose between WIDE and NARROW based
35272 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35273 Alan Hayward <alan.hayward@arm.com>
35274 David Sherwood <david.sherwood@arm.com>
35276 * tree-vect-stmts.c (simd_clone_subparts): New function.
35277 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
35279 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35280 Alan Hayward <alan.hayward@arm.com>
35281 David Sherwood <david.sherwood@arm.com>
35283 * tree-vect-stmts.c (vectorizable_call): Treat the number of
35284 vectors as polynomial. Use build_index_vector for
35285 IFN_GOMP_SIMD_LANE.
35287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35288 Alan Hayward <alan.hayward@arm.com>
35289 David Sherwood <david.sherwood@arm.com>
35291 * tree-vect-stmts.c (get_load_store_type): Treat the number of
35292 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
35293 for variable-length vectors.
35294 (vectorizable_mask_load_store): Treat the number of units as
35295 polynomial, asserting that it is constant if the condition has
35296 already been enforced.
35297 (vectorizable_store, vectorizable_load): Likewise.
35299 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35300 Alan Hayward <alan.hayward@arm.com>
35301 David Sherwood <david.sherwood@arm.com>
35303 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
35304 of units as polynomial. Punt if we can't tell at compile time
35305 which vector contains the final result.
35307 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35308 Alan Hayward <alan.hayward@arm.com>
35309 David Sherwood <david.sherwood@arm.com>
35311 * tree-vect-loop.c (vectorizable_induction): Treat the number
35312 of units as polynomial. Punt on SLP inductions. Use an integer
35313 VEC_SERIES_EXPR for variable-length integer reductions. Use a
35314 cast of such a series for variable-length floating-point
35317 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35318 Alan Hayward <alan.hayward@arm.com>
35319 David Sherwood <david.sherwood@arm.com>
35321 * tree.h (build_index_vector): Declare.
35322 * tree.c (build_index_vector): New function.
35323 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
35324 of units as polynomial, forcibly converting it to a constant if
35325 vectorizable_reduction has already enforced the condition.
35326 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
35327 to create a {1,2,3,...} vector.
35328 (vectorizable_reduction): Treat the number of units as polynomial.
35329 Choose vectype_in based on the largest scalar element size rather
35330 than the smallest number of units. Enforce the restrictions
35333 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35334 Alan Hayward <alan.hayward@arm.com>
35335 David Sherwood <david.sherwood@arm.com>
35337 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
35338 number of units as polynomial.
35340 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35341 Alan Hayward <alan.hayward@arm.com>
35342 David Sherwood <david.sherwood@arm.com>
35344 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
35345 * target.def (autovectorize_vector_sizes): Return the vector sizes
35346 by pointer, using vector_sizes rather than a bitmask.
35347 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
35348 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
35349 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
35351 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
35352 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
35353 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
35354 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
35355 * omp-general.c (omp_max_vf): Likewise.
35356 * omp-low.c (omp_clause_aligned_alignment): Likewise.
35357 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
35358 * tree-vect-loop.c (vect_analyze_loop): Likewise.
35359 * tree-vect-slp.c (vect_slp_bb): Likewise.
35360 * doc/tm.texi: Regenerate.
35361 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
35363 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
35364 the vector size as a poly_uint64 rather than an unsigned int.
35365 (current_vector_size): Change from an unsigned int to a poly_uint64.
35366 (get_vectype_for_scalar_type): Update accordingly.
35367 * tree.h (build_truth_vector_type): Take the size and number of
35368 units as a poly_uint64 rather than an unsigned int.
35369 (build_vector_type): Add a temporary overload that takes
35370 the number of units as a poly_uint64 rather than an unsigned int.
35371 * tree.c (make_vector_type): Likewise.
35372 (build_truth_vector_type): Take the number of units as a poly_uint64
35373 rather than an unsigned int.
35375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35376 Alan Hayward <alan.hayward@arm.com>
35377 David Sherwood <david.sherwood@arm.com>
35379 * target.def (get_mask_mode): Take the number of units and length
35380 as poly_uint64s rather than unsigned ints.
35381 * targhooks.h (default_get_mask_mode): Update accordingly.
35382 * targhooks.c (default_get_mask_mode): Likewise.
35383 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
35384 * doc/tm.texi: Regenerate.
35386 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35387 Alan Hayward <alan.hayward@arm.com>
35388 David Sherwood <david.sherwood@arm.com>
35390 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
35391 * omp-general.c (omp_max_vf): Likewise.
35392 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
35393 (expand_omp_simd): Handle polynomial safelen.
35394 * omp-low.c (omplow_simd_context): Add a default constructor.
35395 (omplow_simd_context::max_vf): Change from int to poly_uint64.
35396 (lower_rec_simd_input_clauses): Update accordingly.
35397 (lower_rec_input_clauses): Likewise.
35399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35400 Alan Hayward <alan.hayward@arm.com>
35401 David Sherwood <david.sherwood@arm.com>
35403 * tree-vectorizer.h (vect_nunits_for_cost): New function.
35404 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
35405 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
35406 (vect_analyze_slp_cost): Likewise.
35407 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
35408 (vect_model_load_cost): Likewise.
35410 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35411 Alan Hayward <alan.hayward@arm.com>
35412 David Sherwood <david.sherwood@arm.com>
35414 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
35415 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
35416 from an unsigned int * to a poly_uint64_pod *.
35417 (calculate_unrolling_factor): New function.
35418 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
35420 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35421 Alan Hayward <alan.hayward@arm.com>
35422 David Sherwood <david.sherwood@arm.com>
35424 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
35425 from an unsigned int to a poly_uint64.
35426 (_loop_vec_info::slp_unrolling_factor): Likewise.
35427 (_loop_vec_info::vectorization_factor): Change from an int
35429 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
35430 (vect_get_num_vectors): New function.
35431 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
35432 (vect_get_num_copies): Use vect_get_num_vectors.
35433 (vect_analyze_data_ref_dependences): Change max_vf from an int *
35434 to an unsigned int *.
35435 (vect_analyze_data_refs): Change min_vf from an int * to a
35437 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
35438 than an unsigned HOST_WIDE_INT.
35439 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
35440 (vect_analyze_data_ref_dependence): Change max_vf from an int *
35441 to an unsigned int *.
35442 (vect_analyze_data_ref_dependences): Likewise.
35443 (vect_compute_data_ref_alignment): Handle polynomial vf.
35444 (vect_enhance_data_refs_alignment): Likewise.
35445 (vect_prune_runtime_alias_test_list): Likewise.
35446 (vect_shift_permute_load_chain): Likewise.
35447 (vect_supportable_dr_alignment): Likewise.
35448 (dependence_distance_ge_vf): Take the vectorization factor as a
35449 poly_uint64 rather than an unsigned HOST_WIDE_INT.
35450 (vect_analyze_data_refs): Change min_vf from an int * to a
35452 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
35453 vfm1 as a poly_uint64 rather than an int. Make the same change
35454 for the returned bound_scalar.
35455 (vect_gen_vector_loop_niters): Handle polynomial vf.
35456 (vect_do_peeling): Likewise. Update call to
35457 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
35458 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
35460 * tree-vect-loop.c (vect_determine_vectorization_factor)
35461 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
35462 (vect_get_known_peeling_cost): Likewise.
35463 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
35464 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
35465 (vect_transform_loop): Likewise. Use the lowest possible VF when
35466 updating the upper bounds of the loop.
35467 (vect_min_worthwhile_factor): Make static. Return an unsigned int
35468 rather than an int.
35469 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
35470 polynomial unroll factors.
35471 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
35472 (vect_make_slp_decision): Likewise.
35473 (vect_supported_load_permutation_p): Likewise, and polynomial
35475 (vect_analyze_slp_cost): Handle polynomial vf.
35476 (vect_slp_analyze_node_operations): Likewise.
35477 (vect_slp_analyze_bb_1): Likewise.
35478 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
35479 than an unsigned HOST_WIDE_INT.
35480 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
35481 (vectorizable_load): Handle polynomial vf.
35482 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
35484 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
35486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35487 Alan Hayward <alan.hayward@arm.com>
35488 David Sherwood <david.sherwood@arm.com>
35490 * match.pd: Handle bit operations involving three constants
35491 and try to fold one pair.
35493 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
35495 * tree-vect-loop-manip.c: Include gimple-fold.h.
35496 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
35497 niters_maybe_zero parameters. Handle other cases besides a step of 1.
35498 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
35499 Add a path that uses a step of VF instead of 1, but disable it
35501 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
35502 and niters_no_overflow parameters. Update calls to
35503 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
35504 Create a new SSA name if the latter choses to use a ste other
35505 than zero, and return it via niters_vector_mult_vf_var.
35506 * tree-vect-loop.c (vect_transform_loop): Update calls to
35507 vect_do_peeling, vect_gen_vector_loop_niters and
35508 slpeel_make_loop_iterate_ntimes.
35509 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
35510 (vect_gen_vector_loop_niters): Update declarations after above changes.
35512 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
35514 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
35515 128-bit round to integer instructions.
35516 (ceil<mode>2): Likewise.
35517 (btrunc<mode>2): Likewise.
35518 (round<mode>2): Likewise.
35520 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
35522 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
35523 unaligned VSX load/store on P8/P9.
35524 (expand_block_clear): Allow the use of unaligned VSX
35525 load/store on P8/P9.
35527 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
35529 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
35531 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
35532 swap associated with both a load and a store.
35534 2018-01-02 Andrew Waterman <andrew@sifive.com>
35536 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
35537 * config/riscv/riscv.md (clear_cache): Use it.
35539 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
35541 * web.c: Remove out-of-date comment.
35543 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35545 * expr.c (fixup_args_size_notes): Check that any existing
35546 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
35547 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
35548 (emit_single_push_insn): ...here.
35550 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35552 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
35553 (const_vector_encoded_nelts): New function.
35554 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
35555 (const_vector_int_elt, const_vector_elt): Declare.
35556 * emit-rtl.c (const_vector_int_elt_1): New function.
35557 (const_vector_elt): Likewise.
35558 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
35559 of CONST_VECTOR_ELT.
35561 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35563 * expr.c: Include rtx-vector-builder.h.
35564 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
35565 directly on the tree encoding.
35566 (const_vector_from_tree): Likewise.
35567 * optabs.c: Include rtx-vector-builder.h.
35568 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
35569 sequence of "u" values.
35570 * vec-perm-indices.c: Include rtx-vector-builder.h.
35571 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
35572 directly on the vec_perm_indices encoding.
35574 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35576 * doc/rtl.texi (const_vector): Describe new encoding scheme.
35577 * Makefile.in (OBJS): Add rtx-vector-builder.o.
35578 * rtx-vector-builder.h: New file.
35579 * rtx-vector-builder.c: Likewise.
35580 * rtl.h (rtx_def::u2): Add a const_vector field.
35581 (CONST_VECTOR_NPATTERNS): New macro.
35582 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
35583 (CONST_VECTOR_DUPLICATE_P): Likewise.
35584 (CONST_VECTOR_STEPPED_P): Likewise.
35585 (CONST_VECTOR_ENCODED_ELT): Likewise.
35586 (const_vec_duplicate_p): Check for a duplicated vector encoding.
35587 (unwrap_const_vec_duplicate): Likewise.
35588 (const_vec_series_p): Check for a non-duplicated vector encoding.
35589 Say that the function only returns true for integer vectors.
35590 * emit-rtl.c: Include rtx-vector-builder.h.
35591 (gen_const_vec_duplicate_1): Delete.
35592 (gen_const_vector): Call gen_const_vec_duplicate instead of
35593 gen_const_vec_duplicate_1.
35594 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
35595 (gen_const_vec_duplicate): Use rtx_vector_builder.
35596 (gen_const_vec_series): Likewise.
35597 (gen_rtx_CONST_VECTOR): Likewise.
35598 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
35599 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
35600 Build a new vector rather than modifying a CONST_VECTOR in-place.
35601 (handle_special_swappables): Update call accordingly.
35602 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
35603 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
35604 Build a new vector rather than modifying a CONST_VECTOR in-place.
35605 (handle_special_swappables): Update call accordingly.
35607 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35609 * simplify-rtx.c (simplify_const_binary_operation): Use
35610 CONST_VECTOR_ELT instead of XVECEXP.
35612 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35614 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
35615 the selector elements to be different from the data elements
35616 if the selector is a VECTOR_CST.
35617 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
35618 ssizetype for the selector.
35620 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35622 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
35623 before testing each element individually.
35624 * tree-vect-generic.c (lower_vec_perm): Likewise.
35626 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35628 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
35629 * selftest-run-tests.c (selftest::run_tests): Call it.
35630 * vector-builder.h (vector_builder::operator ==): New function.
35631 (vector_builder::operator !=): Likewise.
35632 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
35633 (vec_perm_indices::all_from_input_p): New function.
35634 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
35635 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
35636 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
35637 instead of reading the VECTOR_CST directly. Detect whether both
35638 vector inputs are the same before constructing the vec_perm_indices,
35639 and update the number of inputs argument accordingly. Use the
35640 utility functions added above. Only construct sel2 if we need to.
35642 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35644 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
35645 the broadcast of the low byte.
35646 (expand_mult_highpart): Use an explicit encoding for the permutes.
35647 * optabs-query.c (can_mult_highpart_p): Likewise.
35648 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
35649 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
35650 (vectorizable_bswap): Likewise.
35651 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
35652 explicit encoding for the power-of-2 permutes.
35653 (vect_permute_store_chain): Likewise.
35654 (vect_grouped_load_supported): Likewise.
35655 (vect_permute_load_chain): Likewise.
35657 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35659 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
35660 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
35661 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
35662 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
35663 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
35664 (vect_gen_perm_mask_any): Likewise.
35666 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35668 * int-vector-builder.h: New file.
35669 * vec-perm-indices.h: Include int-vector-builder.h.
35670 (vec_perm_indices): Redefine as an int_vector_builder.
35671 (auto_vec_perm_indices): Delete.
35672 (vec_perm_builder): Redefine as a stand-alone class.
35673 (vec_perm_indices::vec_perm_indices): New function.
35674 (vec_perm_indices::clamp): Likewise.
35675 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
35676 (vec_perm_indices::new_vector): New function.
35677 (vec_perm_indices::new_expanded_vector): Update for new
35678 vec_perm_indices class.
35679 (vec_perm_indices::rotate_inputs): New function.
35680 (vec_perm_indices::all_in_range_p): Operate directly on the
35681 encoded form, without computing elided elements.
35682 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
35683 encoding. Update for new vec_perm_indices class.
35684 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
35685 the given vec_perm_builder.
35686 (expand_vec_perm_var): Update vec_perm_builder constructor.
35687 (expand_mult_highpart): Use vec_perm_builder instead of
35688 auto_vec_perm_indices.
35689 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
35690 vec_perm_indices instead of auto_vec_perm_indices. Use a single
35691 or double series encoding as appropriate.
35692 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
35693 vec_perm_indices instead of auto_vec_perm_indices.
35694 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
35695 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
35696 (vect_permute_store_chain): Likewise.
35697 (vect_grouped_load_supported): Likewise.
35698 (vect_permute_load_chain): Likewise.
35699 (vect_shift_permute_load_chain): Likewise.
35700 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
35701 (vect_transform_slp_perm_load): Likewise.
35702 (vect_schedule_slp_instance): Likewise.
35703 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
35704 (vectorizable_mask_load_store): Likewise.
35705 (vectorizable_bswap): Likewise.
35706 (vectorizable_store): Likewise.
35707 (vectorizable_load): Likewise.
35708 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
35709 vec_perm_indices instead of auto_vec_perm_indices. Use
35710 tree_to_vec_perm_builder to read the vector from a tree.
35711 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
35712 vec_perm_builder instead of a vec_perm_indices.
35713 (have_whole_vector_shift): Use vec_perm_builder and
35714 vec_perm_indices instead of auto_vec_perm_indices. Leave the
35715 truncation to calc_vec_perm_mask_for_shift.
35716 (vect_create_epilog_for_reduction): Likewise.
35717 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
35718 from auto_vec_perm_indices to vec_perm_indices.
35719 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
35720 instead of changing individual elements.
35721 (aarch64_vectorize_vec_perm_const): Use new_vector to install
35722 the vector in d.perm.
35723 * config/arm/arm.c (expand_vec_perm_d::perm): Change
35724 from auto_vec_perm_indices to vec_perm_indices.
35725 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
35726 instead of changing individual elements.
35727 (arm_vectorize_vec_perm_const): Use new_vector to install
35728 the vector in d.perm.
35729 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
35730 Update vec_perm_builder constructor.
35731 (rs6000_expand_interleave): Likewise.
35732 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
35733 (rs6000_expand_interleave): Likewise.
35735 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35737 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
35738 to qimode could truncate the indices.
35739 * optabs.c (expand_vec_perm_var): Likewise.
35741 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35743 * Makefile.in (OBJS): Add vec-perm-indices.o.
35744 * vec-perm-indices.h: New file.
35745 * vec-perm-indices.c: Likewise.
35746 * target.h (vec_perm_indices): Replace with a forward class
35748 (auto_vec_perm_indices): Move to vec-perm-indices.h.
35749 * optabs.h: Include vec-perm-indices.h.
35750 (expand_vec_perm): Delete.
35751 (selector_fits_mode_p, expand_vec_perm_var): Declare.
35752 (expand_vec_perm_const): Declare.
35753 * target.def (vec_perm_const_ok): Replace with...
35754 (vec_perm_const): ...this new hook.
35755 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
35756 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
35757 * doc/tm.texi: Regenerate.
35758 * optabs.def (vec_perm_const): Delete.
35759 * doc/md.texi (vec_perm_const): Likewise.
35760 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
35761 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
35762 expand_vec_perm for constant permutation vectors. Assert that
35763 the mode of variable permutation vectors is the integer equivalent
35764 of the mode that is being permuted.
35765 * optabs-query.h (selector_fits_mode_p): Declare.
35766 * optabs-query.c: Include vec-perm-indices.h.
35767 (selector_fits_mode_p): New function.
35768 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
35769 is defined, instead of checking whether the vec_perm_const_optab
35770 exists. Use targetm.vectorize.vec_perm_const instead of
35771 targetm.vectorize.vec_perm_const_ok. Check whether the indices
35772 fit in the vector mode before using a variable permute.
35773 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
35774 vec_perm_indices instead of an rtx.
35775 (expand_vec_perm): Replace with...
35776 (expand_vec_perm_const): ...this new function. Take the selector
35777 as a vec_perm_indices rather than an rtx. Also take the mode of
35778 the selector. Update call to shift_amt_for_vec_perm_mask.
35779 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
35780 Use vec_perm_indices::new_expanded_vector to expand the original
35781 selector into bytes. Check whether the indices fit in the vector
35782 mode before using a variable permute.
35783 (expand_vec_perm_var): Make global.
35784 (expand_mult_highpart): Use expand_vec_perm_const.
35785 * fold-const.c: Includes vec-perm-indices.h.
35786 * tree-ssa-forwprop.c: Likewise.
35787 * tree-vect-data-refs.c: Likewise.
35788 * tree-vect-generic.c: Likewise.
35789 * tree-vect-loop.c: Likewise.
35790 * tree-vect-slp.c: Likewise.
35791 * tree-vect-stmts.c: Likewise.
35792 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
35794 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
35795 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
35796 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
35797 (aarch64_vectorize_vec_perm_const): ...this new function.
35798 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
35799 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
35800 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
35801 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
35802 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
35803 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
35804 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
35806 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
35807 check for NEON modes.
35808 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
35809 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
35810 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
35811 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
35813 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
35814 the old VEC_PERM_CONST conditions.
35815 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
35816 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
35817 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
35818 (ia64_vectorize_vec_perm_const_ok): Merge into...
35819 (ia64_vectorize_vec_perm_const): ...this new function.
35820 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
35821 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
35822 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
35823 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
35824 * config/mips/mips.c (mips_expand_vec_perm_const)
35825 (mips_vectorize_vec_perm_const_ok): Merge into...
35826 (mips_vectorize_vec_perm_const): ...this new function.
35827 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
35828 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
35829 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
35830 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
35831 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
35832 (rs6000_expand_vec_perm_const): Delete.
35833 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
35835 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
35836 (altivec_expand_vec_perm_const_le): Take each operand individually.
35837 Operate on constant selectors rather than rtxes.
35838 (altivec_expand_vec_perm_const): Likewise. Update call to
35839 altivec_expand_vec_perm_const_le.
35840 (rs6000_expand_vec_perm_const): Delete.
35841 (rs6000_vectorize_vec_perm_const_ok): Delete.
35842 (rs6000_vectorize_vec_perm_const): New function.
35843 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
35844 an element count and rtx array.
35845 (rs6000_expand_extract_even): Update call accordingly.
35846 (rs6000_expand_interleave): Likewise.
35847 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
35848 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
35849 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
35850 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
35851 (rs6000_expand_vec_perm_const): Delete.
35852 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
35853 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
35854 (altivec_expand_vec_perm_const_le): Take each operand individually.
35855 Operate on constant selectors rather than rtxes.
35856 (altivec_expand_vec_perm_const): Likewise. Update call to
35857 altivec_expand_vec_perm_const_le.
35858 (rs6000_expand_vec_perm_const): Delete.
35859 (rs6000_vectorize_vec_perm_const_ok): Delete.
35860 (rs6000_vectorize_vec_perm_const): New function. Remove stray
35861 reference to the SPE evmerge intructions.
35862 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
35863 an element count and rtx array.
35864 (rs6000_expand_extract_even): Update call accordingly.
35865 (rs6000_expand_interleave): Likewise.
35866 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
35867 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
35869 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
35871 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35873 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
35874 vector mode and that that mode matches the mode of the data
35876 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
35877 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
35878 directly using expand_vec_perm_1 when forcing selectors into
35880 (expand_vec_perm_var): New function, split out from expand_vec_perm.
35882 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35884 * optabs-query.h (can_vec_perm_p): Delete.
35885 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
35886 * optabs-query.c (can_vec_perm_p): Split into...
35887 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
35888 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
35889 particular selector is valid.
35890 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
35891 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
35892 (vect_grouped_load_supported): Likewise.
35893 (vect_shift_permute_load_chain): Likewise.
35894 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
35895 (vect_transform_slp_perm_load): Likewise.
35896 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
35897 (vectorizable_bswap): Likewise.
35898 (vect_gen_perm_mask_checked): Likewise.
35899 * fold-const.c (fold_ternary_loc): Likewise. Don't take
35900 implementations of variable permutation vectors into account
35901 when deciding which selector to use.
35902 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
35903 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
35904 with a false third argument.
35905 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
35906 to test whether the constant selector is valid and can_vec_perm_var_p
35907 to test whether a variable selector is valid.
35909 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35911 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
35912 * optabs-query.c (can_vec_perm_p): Likewise.
35913 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
35914 instead of vec_perm_indices.
35915 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
35916 (vect_gen_perm_mask_checked): Likewise,
35917 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
35918 (vect_gen_perm_mask_checked): Likewise,
35920 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
35922 * optabs-query.h (qimode_for_vec_perm): Declare.
35923 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
35924 (qimode_for_vec_perm): ...this new function.
35925 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
35927 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
35929 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
35930 does not have a conditional at the top.
35932 2018-01-02 Richard Biener <rguenther@suse.de>
35934 * ipa-inline.c (big_speedup_p): Fix expression.
35936 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
35939 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
35942 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
35946 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
35947 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
35948 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
35949 cond_taken_branch_cost 3->4.
35951 2018-01-01 Jakub Jelinek <jakub@redhat.com>
35953 PR tree-optimization/83581
35954 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
35955 TODO_cleanup_cfg if any changes have been made.
35957 PR middle-end/83608
35958 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
35959 convert_modes if target mode has the right side, but different mode
35962 PR middle-end/83609
35963 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
35964 last argument when extracting from CONCAT. If either from_real or
35965 from_imag is NULL, use expansion through memory. If result is not
35966 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
35967 the parts directly to inner mode, if even that fails, use expansion
35970 PR middle-end/83623
35971 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
35972 check for bswap in mode rather than HImode and use that in expand_unop
35975 Copyright (C) 2018 Free Software Foundation, Inc.
35977 Copying and distribution of this file, with or without modification,
35978 are permitted in any medium without royalty provided the copyright
35979 notice and this notice are preserved.