1 2018-05-16 Jim Wilson <jimw@sifive.com>
3 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
5 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
7 2018-05-16 Mark Wielaard <mark@klomp.org>
9 * dwarf2out.c (count_index_strings): New function.
10 (output_indirect_strings): Call count_index_strings and generate
11 header for dwarf_version >= 5.
13 2018-05-16 Mark Wielaard <mark@klomp.org>
15 * dwarf2out.c (dwarf_FORM): New function.
16 (set_indirect_string): Use dwarf_FORM.
17 (reset_indirect_string): Likewise.
18 (size_of_die): Likewise.
19 (value_format): Likewise.
20 (output_die): Likewise.
21 (add_skeleton_AT_string): Likewise.
22 (output_macinfo_op): Likewise.
23 (index_string): Likewise.
24 (output_index_string_offset): Likewise.
25 (output_index_string): Likewise.
26 (count_index_strings): Likewise.
28 2018-05-16 Carl Love <cel@us.ibm.com>
30 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
31 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
33 2018-05-16 Martin Jambor <mjambor@suse.cz>
35 * ipa-prop.c (ipa_free_all_edge_args): Remove.
36 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
38 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
40 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
41 (fnma<mode>4): Likewise.
42 (fms<mode>4): Likewise.
43 (fnms<mode>4): Likewise.
44 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
45 (aarch64_fnma<mode>4): Likewise.
46 (aarch64_fms<mode>4): Likewise.
47 (aarch64_fnms<mode>4): Likewise.
48 (aarch64_fnmadd<mode>4): Likewise.
50 2018-05-16 Jason Merrill <jason@redhat.com>
52 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
54 2018-05-16 Richard Biener <rguenther@suse.de>
56 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
57 (dump_stmt_cost): Declare.
58 (add_stmt_cost): Dump cost we add.
59 (add_stmt_costs): New function.
60 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
62 (vect_analyze_stmt): Adjust prototype.
63 (vectorizable_condition): Likewise.
64 (vectorizable_live_operation): Likewise.
65 (vectorizable_reduction): Likewise.
66 (vectorizable_induction): Likewise.
67 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
68 cost vector to pass to vectorizable_ and record afterwards.
69 (vect_model_reduction_cost): Take cost vector argument and adjust.
70 (vect_model_induction_cost): Likewise.
71 (vectorizable_reduction): Likewise.
72 (vectorizable_induction): Likewise.
73 (vectorizable_live_operation): Likewise.
74 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
75 SLP_TREE_NUMBER_OF_VEC_STMTS.
76 (vect_analyze_slp_cost_1): Remove.
77 (vect_analyze_slp_cost): Likewise.
78 (vect_slp_analyze_node_operations): Take visited args and
79 a target cost vector. Avoid processing already visited stmt sets.
80 (vect_slp_analyze_operations): Use a local cost vector to gather
81 costs and register those of non-discarded instances.
82 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
83 (vect_schedule_slp_instance): Remove copying of
84 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
86 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
87 adding cost. Record cost entry location.
88 (vect_prologue_cost_for_slp_op): Function to compute cost of
89 a constant or invariant generated for SLP vect in the prologue,
90 split out from vect_analyze_slp_cost_1.
91 (vect_model_simple_cost): Make static. Adjust for SLP costing.
92 (vect_model_promotion_demotion_cost): Likewise.
93 (vect_model_store_cost): Likewise, make static.
94 (vect_model_load_cost): Likewise.
95 (vectorizable_bswap): Add cost vector arg and adjust.
96 (vectorizable_call): Likewise.
97 (vectorizable_simd_clone_call): Likewise.
98 (vectorizable_conversion): Likewise.
99 (vectorizable_assignment): Likewise.
100 (vectorizable_shift): Likewise.
101 (vectorizable_operation): Likewise.
102 (vectorizable_store): Likewise.
103 (vectorizable_load): Likewise.
104 (vectorizable_condition): Likewise.
105 (vectorizable_comparison): Likewise.
106 (can_vectorize_live_stmts): Likewise.
107 (vect_analyze_stmt): Likewise.
108 (vect_transform_stmt): Adjust calls to vectorizable_*.
109 * tree-vectorizer.c: Include gimple-pretty-print.h.
110 (dump_stmt_cost): New function.
112 2018-05-16 Richard Biener <rguenther@suse.de>
114 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
115 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
116 * tree-ssa-dse.c: Include tree-ssa-loop.h.
117 (check_name): New callback.
118 (dse_classify_store): Track cycles via a visited bitmap of PHI
119 defs and simplify handling of in-loop and across loop dead stores
120 and properly fail for loop-variant refs. Handle byte-tracking with
121 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
124 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
126 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
127 (vect_get_mask_type_for_stmt): Likewise.
128 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
130 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
131 to determine the statement's vector type and the vector type that
132 should be used for calculating nunits. Deal with cases in which
133 the type has to be deferred.
134 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
135 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
136 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
137 (vect_determine_vf_for_stmt): New functions, split out from...
138 (vect_determine_vectorization_factor): ...here.
139 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
140 (vect_get_mask_type_for_stmt): New functions, split out from
141 vect_determine_vectorization_factor.
143 2018-05-16 Richard Biener <rguenther@suse.de>
145 * tree-cfg.c (verify_gimple_assign_ternary): Properly
146 verify the [VEC_]COND_EXPR embedded comparison.
148 2018-05-15 Martin Sebor <msebor@redhat.com>
150 PR tree-optimization/85753
151 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
152 RECORD_TYPE in addition to ARRAY_TYPE.
154 2018-05-15 Martin Sebor <msebor@redhat.com>
157 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
159 2018-05-15 Richard Biener <rguenther@suse.de>
161 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
162 add by_clobber_p one. Change algorithm to collect all defs
163 representing uses we need to walk and try reducing them to
164 a single one before failing.
165 (dse_dom_walker::dse_optimize_stmt): Adjust.
167 2018-05-13 Mark Wielaard <mark@klomp.org>
169 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
170 (size_of_loc_descr): Likewise.
171 (output_loc_operands): Likewise.
172 (output_loc_operands_raw): Likewise.
173 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
174 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
175 (hash_loc_operands): Likewise.
176 (compare_loc_operands): Likewise.
178 2018-05-14 Mark Wielaard <mark@klomp.org>
180 * dwarf2out.c (count_index_addrs): New function.
181 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
183 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
185 PR tree-optimization/83648
186 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
187 return value as malloc candidate.
189 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
192 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
193 as true in call to suggest_attribute.
195 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
197 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
200 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
202 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
204 (aarch64_crypto_aesd_fused): Likewise.
206 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
208 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
209 (movsi_aarch64): Likewise.
210 (load_pairsi): Likewise.
211 (load_pairdi): Likewise.
212 (store_pairsi): Likewise.
213 (store_pairdi): Likewise.
214 (load_pairsf): Likewise.
215 (load_pairdf): Likewise.
216 (store_pairsf): Likewise.
217 (store_pairdf): Likewise.
218 (zero_extend): Likewise.
219 (trunc): Swap alternatives.
220 (fcvt_target): Add '?' to prefer w over r.
222 2018-05-14 Jakub Jelinek <jakub@redhat.com>
225 * config/i386/i386.md: Disallow non-commutative arithmetics in
226 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
227 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
228 in the peephole2 before it.
230 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
232 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
233 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
234 (ix86_handle_option): Handle -mcldemote.
235 * config.gcc: New header.
236 * config/i386/cldemoteintrin.h: New file.
237 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
238 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
240 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
241 OPTION_MASK_ISA_CLDEMOTE.
242 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
243 (ix86_valid_target_attribute_inner_p): Ditto.
244 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
245 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
246 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
247 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
248 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
250 * config/i386/i386.opt: Add -mcldemote.
251 * config/i386/x86intrin.h: New header.
252 * doc/invoke.texi: Add -mcldemote.
254 2018-05-14 Richard Biener <rguenther@suse.de>
256 * doc/match-and-simplify.texi: Adjust :s documentation.
258 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
260 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
261 intended memcpy size.
262 (REORDER_45): Likewise.
264 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
267 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
268 * vec.c (qsort_chk): Use gcc_qsort.
269 * Makefile.in (OBJS-libcommon): Add sort.o.
270 (build/sort.o): New target. Use it...
271 (BUILD_RTL): ... here, and...
272 (build/gencfn-macros): ... here, and...
273 (build/genmatch): ... here.
275 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
276 Chung-Ju Wu <jasonwucj@gmail.com>
278 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
279 * config/nds32/nds32-graywolf.md: New file.
280 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
281 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
283 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
284 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
285 * config/nds32/nds32.md (pipeline_model): Add graywolf.
286 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
287 * config/nds32/pipelines.md: Include n15 settings.
289 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
290 Chung-Ju Wu <jasonwucj@gmail.com>
292 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
293 * config/nds32/nds32-n13.md: New file.
294 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
295 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
297 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
298 * config/nds32/nds32.md (pipeline_model): Add n13.
299 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
300 * config/nds32/pipelines.md: Include n13 settings.
302 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
303 Chung-Ju Wu <jasonwucj@gmail.com>
305 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
306 * config/nds32/nds32-n10.md: New file.
307 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
308 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
310 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
311 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
312 * config/nds32/nds32.md (pipeline_model): Add n10.
313 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
314 * config/nds32/pipelines.md: Include n10 settings.
316 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
317 Kito Cheng <kito.cheng@gmail.com>
318 Chung-Ju Wu <jasonwucj@gmail.com>
320 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
321 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
322 Add enum values for DSP extension instructions.
323 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
325 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
326 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
328 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
329 * config/nds32/nds32-dspext.md: New file for DSP implementation.
330 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
331 * config/nds32/nds32-intrinsic.md: Likewise.
332 * config/nds32/nds32_intrinsic.h: Likewise.
333 * config/nds32/nds32-md-auxiliary.c: Likewise.
334 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
335 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
336 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
337 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
338 * config/nds32/nds32-protos.h: New declarations for DSP extension.
339 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
340 TYPE_DMAC in switch statement.
341 * config/nds32/nds32.c: New checking and implementation for DSP
342 extension instructions.
343 * config/nds32/nds32.h: Likewise.
344 * config/nds32/nds32.md: Likewise.
345 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
346 * config/nds32/predicates.md: Implement new predicates for DSP
349 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
351 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
352 Reformat alternatives and attributes so it is easier to identify
353 which constraints/attributes go with which instruction.
354 (mov<mode>_hardfloat32, FMOVE64): Likewise.
355 (mov<mode>_softfloat32, FMOVE64): Likewise.
356 (mov<mode>_hardfloat64, FMOVE64): Likewise.
357 (mov<mode>_softfloat64, FMOVE64): Likewise.
359 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
361 * doc/extend.texi (PowerPC Built-in Functions): Rename this
363 (Basic PowerPC Built-in Functions): The new name of the
364 subsection previously known as "PowerPC Built-in Functions".
365 (Basic PowerPC Built-in Functions Available on all Configurations):
367 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
368 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
369 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
370 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
372 2018-05-11 Martin Jambor <mjambor@suse.cz>
375 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
378 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
381 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
383 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
385 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
386 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
387 (ix86_handle_option): Handle -mwaitpkg.
388 * config.gcc: New header.
389 * config/i386/cpuid.h (bit_WAITPKG): New bit.
390 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
391 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
393 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
394 OPTION_MASK_ISA_WAITPKG.
395 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
396 (ix86_option_override_internal): Add PTA_WAITPKG.
397 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
398 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
400 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
401 __builtin_ia32_umwait and __builtin_ia32_tpause.
402 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
403 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
404 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
405 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
406 UNSPECV_TPAUSE): New.
407 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
408 * config/i386/i386.opt: Add -mwaitpkg.
409 * config/i386/waitpkgintrin.h: New file.
410 * config/i386/x86intrin.h: New header.
411 * doc/invoke.texi: Add -mwaitpkg.
413 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
416 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
418 (cortex-m0): Use armv6s-m isa.
419 (cortex-m0plus): Likewise.
420 (cortex-m1): Likewise.
421 (cortex-m0.small-multiply): Likewise.
422 (cortex-m0plus.small-multiply): Likewise.
423 (cortex-m1.small-multiply): Likewise.
425 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
426 Jakub Jelinek <jakub@redhat.com>
428 PR tree-optimization/85692
429 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
430 source permute as well.
432 2018-05-11 Martin Liska <mliska@suse.cz>
435 * doc/extend.texi: Document LLVM style format for no_sanitize
438 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
440 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
441 mode_supports_vsx_dform_quad to mode_supports_dq_form.
442 (mode_supports_vsx_dform_quad): Likewise.
443 (mode_supports_vmx_dform): Move these functions to be next to the
444 other mode_supports functions.
445 (mode_supports_dq_form): Likewise.
446 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
447 mode_supports_dq_form.
448 (reg_offset_addressing_ok_p): Likewise.
449 (offsettable_ok_by_alignment): Likewise.
450 (rs6000_legitimate_offset_address_p): Likewise.
451 (legitimate_lo_sum_address_p): Likewise.
452 (rs6000_legitimize_address): Likewise.
453 (rs6000_legitimize_reload_address): Likewise.
454 (rs6000_secondary_reload_inner): Likewise.
455 (rs6000_preferred_reload_class): Likewise.
456 (rs6000_output_move_128bit): Likewise.
458 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
460 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
461 Generate SImode target register for null target.
462 <case IX86_BUILTIN_XGETBV>: Ditto.
463 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
464 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
466 2018-05-10 Carl Love <cel@us.ibm.com>
468 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
469 dcbtt and dcbtstt if operands[2] is 0.
471 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
474 * config/i386/sse.md (usadv64qi): New expander.
476 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
478 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
479 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
480 -maltivec=be support.
481 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
482 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
483 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
484 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
485 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
486 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
487 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
488 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
489 altivec_vsumsws): Adjust.
490 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
491 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
492 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
494 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
495 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
496 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
497 (altivec_lve<VI_char>x): Delete expand.
498 (*altivec_lve<VI_char>x_internal): Rename to...
499 (altivec_lve<VI_char>x): ... this.
500 (altivec_lvxl_<mode>): Delete expand.
501 (*altivec_lvxl_<mode>_internal): Rename to ...
502 (altivec_lvxl_<mode>): ... this.
503 (altivec_stvxl_<mode>): Delete expand.
504 (*altivec_stvxl_<mode>_internal): Rename to ...
505 (altivec_stvxl_<mode>): ... this.
506 (altivec_stve<VI_char>x): Delete expand.
507 (*altivec_stve<VI_char>x_internal): Rename to ...
508 (altivec_stve<VI_char>x): ... this.
509 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
510 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
511 reduc_plus_scal_<mode>): Adjust.
512 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
514 (rs6000_cpu_cpp_builtins): Adjust.
515 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
516 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
517 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
518 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
519 -maltivec=be support.
520 (rs6000_split_vec_extract_var): Adjust.
521 (rs6000_split_v4si_init): Adjust.
522 (swap_selector_for_mode): Delete.
523 (altivec_expand_lvx_be, altivec_expand_stvx_be,
524 altivec_expand_stvex_be): Delete.
525 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
526 -maltivec=be support.
527 (rs6000_gimple_fold_builtin): Ditto.
528 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
530 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
531 (TARGET_DIRECT_MOVE_64BIT): Adjust.
532 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
533 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
534 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
535 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
536 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
537 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
538 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
539 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
540 anonymous split): Adjust.
541 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
542 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
544 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
546 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
547 when --with-gxx-include-dir is also specified.
548 * configure: Regenerate.
550 2018-05-09 Jim Wilson <jimw@sifive.com>
553 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
554 * config/riscv/t-withmultilib: New.
555 * config/riscv/withmultilib.h: New.
556 * doc/install.texi: Document RISC-V --with-multilib-list support.
558 2018-05-09 Richard Biener <rguenther@suse.de>
560 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
562 (vect_bb_vectorization_profitable_p): Adjust. Compute
563 actual scalar cost using the cost vector and the add_stmt_cost
566 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
568 PR rtl-optimization/85645
569 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
570 in the REG_CFA_REGISTER note for LR, don't leave it empty.
572 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
574 PR rtl-optimization/85645
575 * shrink-wrap.c (spread_components): Return a boolean saying if
576 anything was changed.
577 (try_shrink_wrapping_separate): Iterate spread_components until
578 nothing changes anymore.
580 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
582 PR rtl-optimization/85645
583 * regrename.c (build_def_use): Also kill the chains that include the
584 destination of a REG_CFA_REGISTER note.
586 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
588 PR rtl-optimization/85645
589 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
590 insn that has a REG_CFA_REGISTER note.
592 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
594 * cfgexpand.c (expand_clobber): New function.
595 (expand_gimple_stmt_1): Use it.
596 * tree-vect-stmts.c (vect_clobber_variable): New function,
598 (vectorizable_simd_clone_call): ...here.
599 (vectorizable_store): Emit a clobber either side of an
600 IFN_STORE_LANES sequence.
601 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
603 2018-05-09 Tom de Vries <tom@codesourcery.com>
606 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
607 (define_insn "trap_if_false"): Add exit after trap.
609 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
611 PR rtl-optimization/85638
612 * bb-reorder.c: Include common/common-target.h.
613 (create_forwarder_block): New function extracted from...
614 (fix_up_crossing_landing_pad): ...here. Rename into...
615 (dw2_fix_up_crossing_landing_pad): ...this.
616 (sjlj_fix_up_crossing_landing_pad): New function.
617 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
618 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
619 from both partitions and exit the loop after one iteration.
621 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
624 * doc/extend.texi (PowerPC Built-in Functions): Rename this
626 (Basic PowerPC Built-in Functions): The new name of the
627 subsection previously known as "PowerPC Built-in Functions".
628 (Basic PowerPC Built-in Functions Available on all Configurations):
630 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
632 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
634 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
636 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
639 2018-05-08 Jim Wilson <jimw@sifive.com>
641 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
642 (LD_EMUL_SUFFIX): New.
645 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
647 * doc/extend.texi (PowerPC Built-in Functions): Rename this
649 (Basic PowerPC Built-in Functions): The new name of the
650 subsection previously known as "PowerPC Built-in Functions".
651 (Basic PowerPC Built-in Functions Available on all Configurations):
653 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
655 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
657 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
659 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
662 2018-05-08 Jakub Jelinek <jakub@redhat.com>
665 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
666 after cmpelim optimization.
668 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
670 * config.gcc: Support "goldmont".
671 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
672 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
674 * config/i386/i386.c (m_GOLDMONT): Define.
675 (processor_target_table): Add "goldmont".
676 (PTA_GOLDMONT): Define.
677 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
678 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
679 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
680 (fold_builtin_cpu): Add "goldmont".
681 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
682 (ix86_option_override_internal): Add "goldmont".
683 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
684 (processor_type): Add PROCESSOR_GOLDMONT.
685 * config/i386/i386.md: Add CPU "glm".
686 * config/i386/glm.md: New file.
687 * config/i386/x86-tune.def: Add m_GOLDMONT.
688 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
690 2018-05-08 Jakub Jelinek <jakub@redhat.com>
693 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
695 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
696 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
697 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
700 * config/i386/i386.c (ix86_fold_builtin): Handle
701 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
704 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
705 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
707 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
710 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
711 (check_arch): Likewise.
712 (check_fpu): Return the result rather than printing it.
713 (end arch): Fix operator precedence.
715 (END): Print the result from check_fpu.
717 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
718 Alan Hayward <alan.hayward@arm.com>
719 David Sherwood <david.sherwood@arm.com>
721 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
722 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
723 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
724 (*fcmuo<mode>_and): New patterns.
726 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
728 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
729 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
730 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
731 (cmp_op, sve_imm_con): New code attributes.
732 (SVE_COND_INT_CMP, imm_con): Delete.
733 (cmp_op): Remove above unspecs from int attribute.
734 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
736 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
737 comparison-specific unspecs.
738 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
739 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
740 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
741 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
742 (*vec_fcm<cmp_op><mode>): Rename to...
743 (*fcm<cmp_op><mode>): ...this and adjust likewise.
744 (*vec_fcmuo<mode>): Rename to...
745 (*fcmuo<mode>): ...this and adjust likewise.
746 (*pred_fcm<cmp_op><mode>): New pattern.
747 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
748 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
750 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
752 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
753 (aarch64_emit_sve_predicated_cond): New function.
754 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
755 (aarch64_emit_unspec_cond_or): Replace with...
756 (aarch64_emit_sve_or_conds): ...this new function. Use
757 aarch64_emit_sve_ptrue_op for the individual comparisons and
758 aarch64_emit_binop to OR them together.
759 (aarch64_emit_inverted_unspec_cond): Replace with...
760 (aarch64_emit_sve_inverted_cond): ...this new function. Use
761 aarch64_emit_sve_ptrue_op for the comparison and
762 aarch64_emit_unop to invert the result.
763 (aarch64_expand_sve_vec_cmp_float): Update after the above
764 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
766 2018-05-07 Nathan Sidwell <nathan@acm.org>
768 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
769 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
770 (Backwards Compatibility): Likewise.
772 2018-05-07 Luis Machado <luis.machado@linaro.org>
776 2018-05-07 Luis Machado <luis.machado@linaro.org>
778 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
779 <prefetch_dynamic_strides>: New const bool field.
780 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
781 prefetch_dynamic_strides.
782 (exynosm1_prefetch_tune): Likewise.
783 (thunderxt88_prefetch_tune): Likewise.
784 (thunderx_prefetch_tune): Likewise.
785 (thunderx2t99_prefetch_tune): Likewise.
786 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
787 (aarch64_override_options_internal): Update to set
788 PARAM_PREFETCH_DYNAMIC_STRIDES.
789 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
790 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
791 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
792 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
793 prefetch-dynamic-strides setting.
795 2018-05-07 Luis Machado <luis.machado@linaro.org>
797 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
798 <minimum_stride>: New const int field.
799 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
800 minimum_stride field.
801 (exynosm1_prefetch_tune): Likewise.
802 (thunderxt88_prefetch_tune): Likewise.
803 (thunderx_prefetch_tune): Likewise.
804 (thunderx2t99_prefetch_tune): Likewise.
805 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
806 (aarch64_override_options_internal): Update to set
807 PARAM_PREFETCH_MINIMUM_STRIDE.
808 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
809 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
810 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
811 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
812 stride is constant and is below the minimum stride threshold.
814 2018-05-07 Luis Machado <luis.machado@linaro.org>
816 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
819 2018-05-07 Luis Machado <luis.machado@linaro.org>
821 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
822 <prefetch_dynamic_strides>: New const bool field.
823 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
824 prefetch_dynamic_strides.
825 (exynosm1_prefetch_tune): Likewise.
826 (thunderxt88_prefetch_tune): Likewise.
827 (thunderx_prefetch_tune): Likewise.
828 (thunderx2t99_prefetch_tune): Likewise.
829 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
830 (aarch64_override_options_internal): Update to set
831 PARAM_PREFETCH_DYNAMIC_STRIDES.
832 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
833 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
834 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
835 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
836 prefetch-dynamic-strides setting.
838 2018-05-07 Luis Machado <luis.machado@linaro.org>
840 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
841 <minimum_stride>: New const int field.
842 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
843 minimum_stride field.
844 (exynosm1_prefetch_tune): Likewise.
845 (thunderxt88_prefetch_tune): Likewise.
846 (thunderx_prefetch_tune): Likewise.
847 (thunderx2t99_prefetch_tune): Likewise.
848 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
849 (aarch64_override_options_internal): Update to set
850 PARAM_PREFETCH_MINIMUM_STRIDE.
851 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
852 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
853 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
854 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
855 stride is constant and is below the minimum stride threshold.
857 2018-05-06 Jakub Jelinek <jakub@redhat.com>
860 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
861 the type is addressable. Don't force op into register if it has
864 2018-05-05 Roland McGrath <mcgrathr@google.com>
867 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
868 any section for which we don't know a specific type it should have,
869 regardless of name. Previously this was done only for the exact
870 names ".init_array", ".fini_array", and ".preinit_array".
871 (default_elf_asm_named_section): Add comment about
872 relationship with default_section_type_flags and SECTION_NOTYPE.
873 (get_section): Don't consider it a type conflict if one side has
874 SECTION_NOTYPE and the other doesn't, as long as neither has the
875 SECTION_BSS et al used in the default_section_type_flags logic.
877 2018-05-05 Tom de Vries <tom@codesourcery.com>
880 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
881 (workaround_barsyncs): New function.
882 (nvptx_reorg): Use workaround_barsyncs.
883 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
884 (define_expand "nvptx_membar_cta"): New define_expand.
885 (define_insn "*nvptx_membar_cta"): New insn.
887 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
889 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
890 To improve optimization opportunities.
891 * builtin-types.def: The new needed builtin types for the above.
893 2018-05-04 Richard Biener <rguenther@suse.de>
895 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
896 * gimple-ssa-store-merging.c
897 (imm_store_chain_info::output_merged_store): Remove redundant create,
898 release split_store vector contents on failure.
899 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
900 scalar stmt vector on cache hit.
902 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
904 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
906 * config.gcc (powerpc-xilinx-eabi*): Remove.
907 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
909 (fusion_addis_mem_combo_load): Ditto.
910 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
912 (rs6000_cpu_cpp_builtins): Ditto.
913 * config/rs6000/rs6000-linux.c
914 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
915 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
916 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
918 (rs6000_setup_reg_addr_masks): Ditto.
919 (rs6000_init_hard_regno_mode_ok): Ditto.
920 (rs6000_option_override_internal): Ditto.
921 (legitimate_lo_sum_address_p): Ditto.
922 (rs6000_legitimize_address): Ditto.
923 (rs6000_legitimize_reload_address): Ditto.
924 (rs6000_legitimate_address_p): Ditto.
925 (abi_v4_pass_in_fpr): Ditto.
926 (setup_incoming_varargs): Ditto.
927 (rs6000_gimplify_va_arg): Ditto.
928 (rs6000_split_multireg_move): Ditto.
929 (rs6000_savres_strategy): Ditto.
930 (rs6000_emit_prologue_components): Ditto.
931 (rs6000_emit_epilogue_components): Ditto.
932 (rs6000_emit_prologue): Ditto.
933 (rs6000_emit_epilogue): Ditto.
934 (rs6000_elf_file_end): Ditto.
935 (rs6000_function_value): Ditto.
936 (rs6000_libcall_value): Ditto.
937 * config/rs6000/rs6000.h: Ditto.
938 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
939 (TARGET_MINMAX): ... this. New.
940 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
941 * config/rs6000/rs6000.md: Remove Xilinx FP support.
942 (*movsi_internal1_single): Delete.
943 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
944 mfpu=, mxilinx-fpu): Delete.
945 * config/rs6000/singlefp.h: Delete.
946 * config/rs6000/sysv4.h: Remove Xilinx FP support.
947 * config/rs6000/t-rs6000: Ditto.
948 * config/rs6000/t-xilinx: Delete.
949 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
950 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
951 (VStype_simple): Delete.
952 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
953 * config/rs6000/xfpu.h: Delete.
954 * config/rs6000/xfpu.md: Delete.
955 * config/rs6000/xilinx.h: Delete.
956 * config/rs6000/xilinx.opt: Delete.
957 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
958 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
960 2018-05-04 Tom de Vries <tom@codesourcery.com>
963 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
966 2018-05-04 Richard Biener <rguenther@suse.de>
969 * tree-complex.c (update_complex_assignment): We are always in SSA form.
970 (expand_complex_div_wide): Likewise.
971 (expand_complex_operations_1): Likewise.
972 (expand_complex_libcall): Preserve EH info of the original stmt.
973 (tree_lower_complex): Handle removed blocks.
974 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
975 on complex multiplication and division libcall builtins.
977 2018-05-04 Richard Biener <rguenther@suse.de>
980 * fold-const.c (negate_expr_p): Restrict negation of operand
981 zero of a division to when we know that can happen without
983 (fold_negate_expr_1): Likewise.
985 2018-05-04 Jakub Jelinek <jakub@redhat.com>
988 * real.h (real_nextafter): Declare.
989 * real.c (real_nextafter): New function.
990 * fold-const-call.c (fold_const_nextafter): New function.
991 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
993 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
994 even when arg1_mode is different from arg0_mode.
996 2018-05-03 Nathan Sidwell <nathan@acm.org>
998 * doc/extend.texi (Deprecated Features): Remove
1000 (Backwards Compatibility): Likewise.
1001 * doc/invoke.texi (C++ Language Options): Likewise.
1002 (C++ Dialect Options): Likewise.
1004 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1007 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1008 _mm512_mask_mullox_epi64): New intrinsics.
1010 2018-05-03 Tom de Vries <tom@codesourcery.com>
1013 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1014 dump files): Add offload-tree.
1016 2018-05-03 Richard Biener <rguenther@suse.de>
1018 PR tree-optimization/85615
1019 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1020 to loops not nested in BBs loop father to avoid creating multi-entry
1023 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1025 PR tree-optimization/70291
1026 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1027 arguments. Change return type to tree. Emit libcall as a new
1028 statement rather than replacing existing one when inplace_p is true.
1029 (expand_complex_multiplication_components): New function.
1030 (expand_complex_multiplication): Expand floating-point complex
1031 multiplication using the above.
1032 (expand_complex_division): Rename inner_type parameter to type.
1033 Update expand_complex_libcall call-site.
1034 (expand_complex_operations_1): Update expand_complex_multiplication
1035 and expand_complex_division call-sites.
1037 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1040 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1041 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1042 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1043 the highest significant bit of the shift count mask is clear. In
1044 check whether and[sq]i3 is needed verify that all significant bits
1045 of the shift count other than the highest are set.
1047 2018-05-02 Tom de Vries <tom@codesourcery.com>
1050 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1051 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1052 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1053 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1054 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1055 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1056 __builtin_goacc_parlevel_size.
1058 2018-05-02 Richard Biener <rguenther@suse.de>
1060 PR tree-optimization/85597
1061 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1062 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1065 2018-05-02 Tom de Vries <tom@codesourcery.com>
1068 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1069 dump files): Add ltrans-tree.
1071 2018-05-02 Tom de Vries <tom@codesourcery.com>
1074 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1075 dump files): Add wpa-ipa.
1077 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1079 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1080 powerpc*-*-linux*paired* target.
1081 * config/rs6000/750cl.h: Delete.
1082 * config/rs6000/paired.h: Delete.
1083 * config/rs6000/paired.md: Delete.
1084 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1086 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1087 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1088 comment. Remove paired float support.
1089 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1090 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1092 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1093 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1095 * config/rs6000/rs6000.c: Remove paired float support.
1096 (paired_expand_vector_init, paired_expand_vector_move,
1097 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1098 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1099 paired_expand_builtin, paired_expand_predicate_builtin,
1100 paired_init_builtins): Delete.
1101 * config/rs6000/rs6000.h: Remove paired float support.
1102 * config/rs6000/rs6000.md: Remove paired float support.
1103 (move_from_CR_ov_bit): Delete.
1104 * config/rs6000/rs6000.opt (mpaired): Delete.
1105 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1106 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1108 2018-05-02 Richard Biener <rguenther@suse.de>
1111 * gimplify.c (gimplify_save_expr): When in SSA form allow
1112 SAVE_EXPRs to compute to SSA vars.
1114 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1117 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1118 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1119 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1120 clobber operands[2], instead use a new pseudo. Formatting fixes.
1122 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1124 PR tree-optimization/85586
1125 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1126 exit early for statements in the same group if the accesses are
1129 2018-05-02 Tom de Vries <tom@codesourcery.com>
1132 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1135 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1137 PR tree-optimization/85143
1138 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1140 2018-05-01 Tom de Vries <tom@codesourcery.com>
1143 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1144 not found" error message.
1146 2018-05-01 Tom de Vries <tom@codesourcery.com>
1149 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1150 * vec.c (test_ordered_remove_if): New function.
1151 (vec_c_tests): Call test_ordered_remove_if.
1152 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1153 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1154 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1155 VEC_ORDERED_REMOVE_IF.
1157 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1159 PR tree-optimization/82665
1160 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1161 pointer subtraction where arguments come from a memchr call.
1163 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1165 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1166 --push-state --as-needed and --pop-state instead of --as-needed and
1167 --no-as-needed if ld supports it.
1168 * configure: Regenerated.
1171 * doc/install.texi2html: Replace _002d with - and _002a with * in
1172 generated html files using sed.
1174 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1177 * gcc-rich-location.c (blank_line_before_p): New function.
1178 (use_new_line): New function.
1179 (gcc_rich_location::add_fixit_insert_formatted): New function.
1180 * gcc-rich-location.h
1181 (gcc_rich_location::add_fixit_insert_formatted): New function.
1183 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1185 * selftest.c (assert_streq): Rename "expected" and "actual" to
1186 "val1" and "val2". Extend NULL-handling to cover both inputs
1187 symmetrically, while still requiring both to be non-NULL for a pass.
1188 * selftest.h (assert_streq): Rename "expected" and "actual" to
1190 (ASSERT_EQ): Likewise.
1191 (ASSERT_EQ_AT): Likewise.
1192 (ASSERT_KNOWN_EQ): Likewise.
1193 (ASSERT_KNOWN_EQ_AT): Likewise.
1194 (ASSERT_NE): Likewise.
1195 (ASSERT_MAYBE_NE): Likewise.
1196 (ASSERT_MAYBE_NE_AT): Likewise.
1197 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1198 the assertion to pass.
1199 (ASSERT_STREQ_AT): Likewise.
1201 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1203 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1204 interaction with -pie.
1206 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1208 * selftest.h: Fix alphabetization of per-source-file selftest
1211 2018-04-30 Jason Merrill <jason@redhat.com>
1213 PR c++/61982 - dead stores to destroyed objects.
1214 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1217 2018-04-30 Jason Merrill <jason@redhat.com>
1219 * tree.c (build_clobber): New.
1220 * tree.h: Declare it.
1221 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1223 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1225 * diagnostic-show-locus.c (layout::layout): Update for
1226 location_get_source_line returning a char_span.
1227 (struct char_span): Move to input.h.
1228 (struct correction): Update for fields in char_span becoming
1230 (struct source_line): Update for location_get_source_line
1231 returning a char_span.
1232 (layout::print_line): Likewise.
1233 * edit-context.c (edited_file::print_content): Likewise.
1234 (edited_file::print_diff_hunk): Likewise.
1235 (edited_file::print_run_of_changed_lines): Likewise.
1236 (edited_file::get_num_lines): Likewise.
1237 (edited_line::edited_line): Likewise.
1238 * final.c (asm_show_source): Likewise.
1239 * input.c (location_get_source_line): Convert return type
1240 from const char * to char_span, losing the final "line_len"
1242 (dump_location_info): Update for the above.
1243 (get_substring_ranges_for_loc): Likewise. Use a char_span
1244 when handling the literal within the line.
1245 (test_reading_source_line): Update for location_get_source_line
1246 returning a char_span.
1247 * input.h (class char_span): Move here from
1248 diagnostic-show-locus.c, converting from a struct to a class.
1249 Make data members private.
1250 (char_span::operator bool): New.
1251 (char_span::length): New.
1252 (char_span::get_buffer): New.
1253 (char_span::operator[]): New.
1254 (char_span::subspan): Make const.
1255 (char_span::xstrdup): New.
1256 (location_get_source_line): Convert return type from const char *
1257 to char_span, losing the final "line_size" param.
1259 2018-04-30 Jan Hubicka <jh@suse.cz>
1261 * lto-wrapper.c (ltrans_priorities): New static var.
1262 (cmp_priority): New.
1263 (run_gcc): Read priorities and if doing parallel build order
1264 the Makefile by them.
1266 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1268 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1270 2018-04-30 Richard Biener <rguenther@suse.de>
1272 * tree-cfg.c (verify_address): Remove base argument, add
1273 flag whether to check TREE_ADDRESSABLE and do that.
1274 (verify_expr): Remove.
1275 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1276 (verify_gimple_assign_single): Likewise.
1277 (verify_gimple_switch): Likewise.
1278 (verify_expr_location_1): Dereference tp once. Add (disabled)
1279 piece from verify_expr.
1280 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1282 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1284 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1286 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1288 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1289 (small_data_pattern): Likewise.
1290 (arc_rewrite_small_data): Likewise.
1291 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1292 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1293 (get_symbol_alignment): New function.
1294 (legitimate_small_data_address_p): Likewise.
1295 (legitimate_scaled_address): Update, call
1296 legitimate_small_data_address_p.
1297 (output_sdata): New static variable.
1298 (arc_print_operand): Update how we handle small data operands.
1299 (arc_print_operand_address): Likewise.
1300 (arc_legitimate_address_p): Update, use
1301 legitimate_small_data_address_p.
1302 (arc_rewrite_small_data_p): Remove.
1303 (arc_rewrite_small_data_1): Likewise.
1304 (arc_rewrite_small_data): Likewise.
1305 (small_data_pattern): Likewise.
1306 (compact_sda_memory_operand): Update to use
1307 legitimate_small_data_address_p and get_symbol_alignment.
1308 (prepare_move_operands): Don't rewite sdata pattern.
1309 (prepare_extend_operands): Remove.
1310 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1312 (zero_extendqisi2): Likewise.
1313 (zero_extendhisi2): Likewise.
1314 (extendqihi2): Likewise.
1315 (extendqisi2): Likewise.
1316 (extendhisi2): Likewise.
1320 * config/arc/constraints.md (Usd): Change it to memory constraint.
1322 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1324 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1325 as source of std instructions.
1326 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1327 allow 6-bit constants as source for store instructions.
1328 (movdi_insn): Update instruction pattern to allow 6-bit constants
1329 as source for store instructions.
1331 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1333 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1335 2018-04-30 Nathan Sidwell <nathan@acm.org>
1336 Sandra Loosemore <sandra@codesourcery.com>
1338 * dumpfile.c (dump_open): Allow '-' for stdout.
1339 * doc/invoke.texi (Developer Options): Document dump filename
1340 determination early. Document stdin/stdout selection.
1342 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1344 Microblaze Target: PIC data text relative
1346 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1347 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1349 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1350 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1351 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1352 New addressing mode for data-text relative position indepenedent code.
1353 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1354 'ADDRESS_SYMBOLIC_TXT_REL'.
1355 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1356 (microblaze_legitimate_pic_operand): Exclude function calls from
1357 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1358 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1360 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1361 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1362 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1363 for 'address + offset'.
1364 (microblaze_expand_prologue): Add new function prologue call for
1366 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1367 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1368 table in case of TARGET_PIC_DATA_TEXT_REL.
1369 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1370 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1371 Add new macros 'UNSPEC_TEXT',
1372 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1373 + exclude function calls from 'UNSPEC_PLT' in case of data text
1375 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1376 new target hook for generating address diff vector tables in case of
1378 * doc/tm.texi : Regenerate.
1379 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1380 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1381 of addr diff vector generation.
1382 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1383 target hook definition.
1384 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1385 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1386 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1387 option for data text relative.
1389 2018-04-30 Richard Biener <rguenther@suse.de>
1391 * tree-chrec.h (evolution_function_is_constant_p): Remove
1393 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1395 2018-04-30 Richard Biener <rguenther@suse.de>
1398 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1400 2018-04-30 Richard Biener <rguenther@suse.de>
1402 PR tree-optimization/28364
1403 PR tree-optimization/85275
1404 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1405 copying first exit test.
1407 2018-04-28 Mark Wielaard <mark@klomp.org>
1409 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1411 (dwarf_AT): Handle DW_AT_addr_base.
1412 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1414 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1417 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1418 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1419 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1420 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1422 2018-04-28 Richard Biener <rguenther@suse.de>
1424 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1425 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1426 to reflect use. Only add interesting stmts.
1428 2018-04-27 Martin Jambor <mjambor@suse.cz>
1431 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1432 the jump function allows for passing through aggregate values.
1434 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1436 * input.h (in_system_header_at): Convert from macro to inline
1438 (from_macro_expansion_at): Likewise.
1439 (from_macro_definition_at): Likewise.
1441 2018-04-27 Jeff Law <law@redhat.com>
1443 * config.gcc: Mark tile* targets as deprecated/obsolete.
1445 2018-04-27 Richard Biener <rguenther@suse.de>
1447 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1450 2018-04-27 Richard Biener <rguenther@suse.de>
1452 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1454 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1456 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1457 with Yd constraint. Set "preferred_for_speed" attribute from
1458 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1460 (*movdi_internal): Ditto.
1461 (movti_interunit splitters): Remove
1462 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1463 (movdi_interunit splitters): Ditto.
1464 * config/i386/constraints.md (Ye): Remove.
1465 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1467 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1470 * config/aarch64/constraints.md (Usg): Limit to 31.
1473 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1475 PR tree-optimization/85529
1476 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1477 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1478 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1479 zero extension or masking of the MSB bit.
1480 (optimize_range_tests): Add FIRST_BB argument, pass it through
1481 to optimize_range_tests_var_bound.
1482 (maybe_optimize_range_tests, reassociate_bb): Adjust
1483 optimize_range_tests callers.
1485 2018-04-26 Richard Biener <rguenther@suse.de>
1486 Jakub Jelinek <jakub@redhat.com>
1488 * cgraph.h (symbol_table): Just declare debug method here.
1489 * symtab.c (symbol_table::debug): Define.
1491 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1493 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1495 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1497 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1498 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1499 (*movdi_internal): Substitute Yi and Yj constraint with x
1500 and Ym and Yn constraint with y constraint. Update "isa"
1501 attribute and set "preferred_for_speed" attribute from
1502 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1503 (*movsi_internal): Ditto.
1504 (*movdf_internal): Ditto.
1505 (*movsf_internal): Ditto.
1506 (*zero_extendsidi2): Ditto.
1507 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1508 (sse2_loadld): Ditto.
1509 (*vec_extract<ssevecmodelower>_0): Ditto.
1510 (*vec_extractv4si_0_zext_sse4): Ditto.
1511 (vec_concatv2di): Ditto.
1512 (*vec_dup<mode>): Ditto.
1513 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1514 * config/i386/constraints.md (Yi): Remove.
1519 2018-04-26 Nathan Sidwell <nathan@acm.org>
1521 * dumpfile.c (dump_open): New.
1522 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1523 (dump_finish): Detect stdio/stderr by value not name.
1525 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1527 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1529 2018-04-26 Tom de Vries <tom@codesourcery.com>
1532 * config/nvptx/nvptx.c (verify_neutering_jumps)
1533 (verify_neutering_labels): New function
1534 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1536 2018-04-26 Tom de Vries <tom@codesourcery.com>
1539 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1540 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1543 2018-04-26 Richard Biener <rguenther@suse.de>
1544 Tom de Vries <tom@codesourcery.com>
1547 * lto-streamer-out.c (output_function): Fixup loops if required to match
1548 discovery done in the reader.
1550 2018-04-26 Richard Biener <rguenther@suse.de>
1552 PR tree-optimization/85116
1553 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1554 have a loop exit from the single latch predecessor. Remove
1555 case of header with just condition.
1556 (ch_base::copy_headers): Exclude infinite loops from any
1558 (pass_ch::execute): Record exits.
1560 2018-04-26 Richard Biener <rguenther@suse.de>
1562 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1563 prologue cost vector and pass it to vect_get_load_cost.
1564 (vect_get_peeling_costs_all_drs): Likewise.
1565 (vect_peeling_hash_get_lowest_cost): Likewise.
1566 (vect_enhance_data_refs_alignment): Likewise.
1568 2018-04-26 Richard Biener <rguenther@suse.de>
1571 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1572 checking of integer<->pointer conversions.
1573 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1574 sign-/zero-extending pointer types.
1575 (expand_omp_for_static_chunk): Likewise.
1577 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1578 Jean Lee <xiaoyur347@gmail.com>
1580 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1581 (TARGET_ASAN_SHADOW_OFFSET): Define.
1582 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1583 true for -fsanitize=address.
1585 2018-04-25 Mark Wielaard <mark@klomp.org>
1587 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1590 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1592 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1593 than "alu", remove explicit "memory" and "imm_disp" attributes.
1594 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1597 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1598 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1599 gen_lowpart_no_emit.
1601 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1604 * config/i386/i386.c (ix86_expand_builtin): Change memory
1605 operand to XI, extend p0 to Pmode.
1606 * config/i386/i386.md: Change unspec volatile and operand
1607 1 mode to XI, change operand 0 mode to P.
1609 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1611 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1612 GET_MODE_MASK before any checking.
1613 (nds32_can_use_bset_p): Likewise.
1614 (nds32_can_use_btgl_p): Likewise.
1616 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1618 * config/nds32/nds32-doubleword.md: New define_split pattern for
1619 illegal register number.
1621 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1623 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1625 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1627 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1629 2018-04-25 Richard Biener <rguenther@suse.de>
1631 * lto-streamer.h (LTO_major_version): Bump to 8.
1633 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1635 * BASE-VER: Set to 9.0.0.
1637 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1639 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1640 in __abskf2 and __powikf2.
1642 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1645 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1646 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1647 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1648 Use the above on operand 2. Reindent.
1649 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1651 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1654 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1656 * config/i386/i386.opt (mcet): Removed.
1657 * doc/install.texi: Remove -mcet documentation.
1658 * doc/invoke.texi: Likewise.
1660 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1663 * doc/install.texi: Remove -mcet from bootstrap-cet.
1665 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1668 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1669 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1673 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1674 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1675 containing a CONST_VECTOR.
1677 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1679 * doc/install.texi: Update newlib dependency for nvptx.
1681 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1684 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1685 instead of INTVAL when shifting x left.
1687 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1689 PR tree-optimization/85478
1690 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1691 vect_grouped_store_supported for single element vectors.
1693 2018-04-24 Richard Biener <rguenther@suse.de>
1696 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1697 load cost increase to the case of non-constant step.
1699 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1702 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1703 destination if any_malformed_asm.
1705 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1708 * expr.c (store_field): In the bitfield case, if the value comes from
1709 a function call and is returned in registers by means of a PARALLEL,
1710 do not change the mode of the temporary unless BLKmode and VOIDmode.
1712 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1714 PR rtl-optimization/85423
1715 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1716 dependencies to debug insns when the previous insn is non-debug.
1718 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1720 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1721 enums into a single definition.
1722 (fls): Fix predicates and printing.
1725 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1727 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1728 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1729 and short u6 immediate.
1730 (check_if_valid_sleep_operand): Remove.
1731 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1733 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1735 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1736 flag_always_save_lp condition.
1737 * config/nds32/nds32.opt (malways-save-lp): New option.
1739 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1741 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1742 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1743 * config/nds32/nds32.h
1744 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1745 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1747 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1749 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1750 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1752 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1753 Chung-Ju Wu <jasonwucj@gmail.com>
1755 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1757 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1758 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1760 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1762 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1764 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1766 * config/nds32/nds32-protos.h (nds32_data_alignment,
1767 nds32_local_alignment): Declare.
1768 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1769 nds32_local_alignment): New functions.
1770 (TARGET_CONSTANT_ALIGNMENT): Define.
1771 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1773 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1775 * config/nds32/nds32.c
1776 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1777 (TARGET_MODES_TIEABLE_P): Likewise.
1779 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1781 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1784 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1785 Chung-Ju Wu <jasonwucj@gmail.com>
1787 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1788 for unaligned access.
1789 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1791 * config/nds32/nds32-intrinsic.md: Likewise.
1792 * config/nds32/nds32_intrinsic.h: Likewise.
1793 * config/nds32/nds32.h (nds32_builtins): Likewise.
1794 * config/nds32/nds32.opt (munaligned-access): New option.
1795 * config/nds32/nds32.c (nds32_asm_file_start): Display
1796 flag_unaligned_access status.
1798 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1800 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1801 -mno-relax is present.
1802 * config/riscv/linux.h (LINK_SPEC): Ditto.
1804 2018-04-20 Martin Sebor <msebor@redhat.com>
1807 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1809 (gimple_fold_builtin_stxcpy_chk): Same.
1810 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1812 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1815 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1816 __powikf2 when long double is IEEE 128-bit.
1818 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1820 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1821 step to make sure stack always aligned.
1823 2018-04-20 Carl Love <cel@us.ibm.com>
1826 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1827 size check for arg0.
1829 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1830 Tom de Vries <tom@codesourcery.com>
1833 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1834 Emit insns for calls too.
1835 (nvptx_find_par): Always look for worker-level predecessor insn.
1836 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1838 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1839 (nvptx_process_pars): Propagate frames for calls.
1841 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1844 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1846 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1847 (ix86_handle_option): Don't handle OPT_mibt.
1848 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1850 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1852 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1853 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1854 (ix86_target_macros): Define __CET__ with flag_cf_protection
1855 for -fcf-protection.
1856 * config/i386/i386.c (isa2_opts): Remove -mibt.
1857 * config/i386/i386.h (TARGET_IBT): Removed.
1858 (TARGET_IBT_P): Likewise.
1859 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1860 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1861 * config/i386/i386.opt (mcet): Update help message.
1864 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1865 -mcet as an alias for -mshstk.
1867 2018-04-20 Richard Biener <rguenther@suse.de>
1870 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1871 complexity by forcing a single use of the multiply operand.
1873 2018-04-20 Martin Jambor <mjambor@suse.cz>
1876 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1877 recursion dependency to only apply to non-clones.
1879 2018-04-20 Martin Jambor <mjambor@suse.cz>
1882 * ipa-cp.c (create_specialized_node): Check that clones of
1883 self-recursive edges exist during IPA-CP.
1885 2018-04-19 Toon Moene <toon@moene.org>
1887 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1890 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1892 PR tree-optimization/85467
1893 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1894 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1895 VECTOR_CST element to type.
1897 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1900 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
1901 * config/i386/i386.md (builtin_setjmp_setup): Removed.
1902 (builtin_longjmp): Likewise.
1903 (save_stack_nonlocal): New pattern.
1904 (restore_stack_nonlocal): Likewise.
1906 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1909 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1910 Replace ASM_OUTPUT_LABEL with fprintf.
1912 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
1915 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
1916 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
1917 * config/i386/i386-c.c (ix86_target_macros_internal): Also
1918 define __IBT__ and __SHSTK__ for -fcf-protection.
1919 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
1921 (ix86_trampoline_init): Likewise.
1922 (x86_output_mi_thunk): Likewise.
1923 (ix86_notrack_prefixed_insn_p): Likewise.
1924 (ix86_option_override_internal): Don't disallow -fcf-protection.
1925 * config/i386/i386.md (rdssp<mode>): Also enable for
1927 (incssp<mode>): Likewise.
1928 (nop_endbr): Likewise.
1929 * config/i386/i386.opt (mcet): Change help message to built-in
1933 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
1934 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
1935 enable CET built-in functions.
1937 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
1939 * common/config/i386/i386-common.c
1940 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
1941 OPTION_MASK_ISA_MOVDIRI_UNSET,
1942 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
1943 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
1944 * config.gcc (movdirintrin.h): New header.
1945 * config/i386/cpuid.h (bit_MOVDIRI,
1946 bit_MOVDIR64B): New bits.
1947 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
1949 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
1950 (VOID, PVOID, PCVOID)): New function types.
1951 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
1952 __builtin_ia32_directstoreu_u64,
1953 __builtin_ia32_movdir64b): New builtins.
1954 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
1955 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
1957 (ix86_valid_target_attribute_inner_p): Ditto.
1958 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
1959 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
1960 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
1961 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
1962 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
1963 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
1964 (movdiri<mode>, movdir64b_<mode>): New.
1965 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
1966 * config/i386/immintrin.h: Include movdirintrin.h.
1967 * config/i386/movdirintrin.h: New file.
1968 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
1970 2018-04-19 Richard Biener <rguenther@suse.de>
1973 * cfg.c (clear_bb_flags): When loop state says we have
1974 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
1976 2018-04-19 Richard Biener <rguenther@suse.de>
1978 PR tree-optimization/84737
1979 * tree-vect-data-refs.c (vect_copy_ref_info): New function
1980 copying restrict info.
1981 (vect_setup_realignment): Use it.
1982 * tree-vectorizer.h (vect_copy_ref_info): Declare.
1983 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
1984 the first DR to all generated stores.
1985 (vectorizable_load): Likewise for loads.
1987 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1989 PR tree-optimization/85446
1990 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
1991 the integral and pointer types to have the same precision.
1993 * doc/install.texi: Document --disable-cet being the default and
1996 2018-04-18 Martin Liska <mliska@suse.cz>
1998 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2001 2018-04-18 Martin Liska <mliska@suse.cz>
2004 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2007 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2008 arguments if they are comparable.
2010 2018-04-18 Martin Liska <mliska@suse.cz>
2013 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2016 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2019 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2022 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2023 ENDBR after calling __morestack.
2025 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2028 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2029 by using gcc_base_ver to generate a gcc_driver_version, and use
2030 it when generating GCC_DRIVER_NAME.
2031 * configure: Regenerate.
2033 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2036 * config.gcc: Obsolete powerpc*-*-*spe*.
2038 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2041 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2042 (stabstr_D): Change type of unum from unsigned int to
2043 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2046 2018-04-17 Jim Wilson <jimw@sifive.com>
2049 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2050 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2051 Set arg_pointer_offset after using pretend_args_size.
2053 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2055 PR rtl-optimization/85431
2056 * dse.c (record_store): Ignore zero width stores.
2059 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2060 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2061 __builtin_stack_restore rather than after it.
2062 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2063 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2064 argument instead of virtual_dynamic_stack_rtx.
2066 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2068 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2070 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2071 Add note to error message to explain internal mapping of overloaded
2072 built-in function name to non-overloaded built-in function name.
2073 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2076 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2079 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2080 where the inputs overlap with the output.
2082 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2084 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2085 (=v, v) alternative and explicit "memory" attribute.
2086 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2087 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2089 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2090 "sselog1" type instead of "sselog".
2091 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2092 "sselog". Remove explicit "memory" attribute.
2093 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2094 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2096 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2097 "sselog1" type instead of "sselog". Remove explicit "memory"
2099 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2100 use "sselog1" type instead of "sselog". Remove explicit "memory"
2102 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2103 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2105 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2106 "sselog1" type instead of "sselog". Remove explicit "memory"
2108 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2109 use "sselog1" type instead of "sselog". Remove explicit "memory"
2113 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2116 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2119 2018-04-17 Martin Jambor <mjambor@suse.cz>
2122 * ipa-cp.c (create_specialized_node): Call
2123 expand_all_artificial_thunks if necessary.
2125 2018-04-17 Martin Liska <mliska@suse.cz>
2128 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2129 in message, remote space in between '_G' and '('.
2131 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2134 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2135 avx512f_vmcmp<mode>3<round_saeonly_name>,
2136 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2137 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2138 avx512f_rndscale<mode><round_saeonly_name>,
2139 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2140 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2141 Use %<iptr>2 instead of %2 for -masm=intel.
2142 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2143 avx512f_vcvttss2usi<round_saeonly_name>,
2144 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2146 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2147 avx512f_vcvttsd2usi<round_saeonly_name>,
2148 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2149 Use %q1 instead of %1 for -masm=intel.
2150 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2151 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2152 of %3 for -masm=intel.
2153 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2155 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2157 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2159 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2160 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2162 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2164 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2165 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2166 %g1 and one with %0 and %1.
2167 (avx512er_vmrcp28<mode><round_saeonly_name>,
2168 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2170 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2171 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2172 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2173 of %0 and %{%4%} for -masm=intel.
2174 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2175 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2176 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2177 order of %0 and %{%5%}%{z%} for -masm=intel.
2179 2018-04-17 Jan Hubicka <jh@suse.cz>
2182 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2184 2018-04-17 Martin Liska <mliska@suse.cz>
2187 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2188 for target_clone error message. Make default implementation
2189 clone to be a local declaration.
2190 (separate_attrs): Add new argument and check for an empty
2192 (expand_target_clones): Handle it.
2193 (ipa_target_clone): Make redirection just for target_clones
2196 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2197 Tom de Vries <tom@codesourcery.com>
2200 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2201 tiled basic blocks without omp continue statements.
2203 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2206 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2207 vec_extract expression as having side effects to make sure it gets
2210 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2213 * config/i386/i386.c (get_builtin_code_for_version): Check
2216 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2219 * gcc/config.gcc: Support "skylake".
2220 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2222 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2223 (processor_target_table): Add "skylake".
2224 (ix86_option_override_internal): Add "skylake".
2225 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2226 PROCESSOR_CANNONLAKE.
2227 (get_builtin_code_for_version): Fix priority for
2228 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2229 PROCESSOR_SKYLAKE-AVX512.
2230 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2231 (processor_type): Add PROCESSOR_SKYLAKE.
2233 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2234 Jason Merrill <jason@redhat.com>
2237 * convert.c (convert_to_integer_1): Use direct recursion for
2238 enumeral types and types with a precision less than the number
2239 of bits in their mode.
2241 2018-04-16 Julia Koval <julia.koval@intel.com>
2244 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2245 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2247 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2250 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2251 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2252 and -mno-direct-move.
2254 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2257 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2258 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2259 Ensure negative shifts result in {0}.
2261 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2263 PR rtl-optimization/79916
2264 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2265 regs (if any) to define how to gnerate SD moves when LRA is in
2268 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2270 PR rtl-optimization/85393
2271 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2272 * except.c (expand_dw2_landing_pad_for_region): Make static.
2273 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2274 a label and unconditional jump to old_bb, rather than
2275 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2278 PR rtl-optimization/85376
2279 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2280 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2281 instead of a specific value.
2283 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2284 Bin Cheng <bin.cheng@arm.com>
2286 PR tree-optimization/82965
2287 PR tree-optimization/83991
2288 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2289 by_profile_only parameter.
2290 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2291 information if the loop was predicted to iterate too many times.
2292 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2294 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2297 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2300 2018-04-13 Martin Liska <mliska@suse.cz>
2301 Jakub Jelinek <jakub@redhat.com>
2304 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2305 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2306 * builtins.c (expand_builtin_memory_copy_args): Use
2307 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2308 handle dest_addr == pc_rtx.
2310 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2313 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2314 asked to not generate direct moves.
2315 (fix_trunc<mode>si2_stfiwx): Similar.
2316 (fix_trunc<mode>si2_internal): Similar.
2318 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2321 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2322 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2323 lookup if dest in some wider mode is known to be const0_rtx and
2324 if so, record permanent equivalence for it to be ZERO_EXTEND of
2325 the narrower mode destination.
2327 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2329 * lto-streamer-out.c (output_function): Revert 259346.
2330 * omp-expand.c (expand_oacc_for): Likewise.
2332 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2334 PR rtl-optimization/85354
2335 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2336 * sel-sched.c (sel_global_init): ... here.
2338 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2341 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2342 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2343 mode for PE-COFF targets.
2344 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2345 (i386_pe_asm_lto_end): Likewise.
2346 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2347 (TARGET_ASM_LTO_END): Likewise.
2348 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2349 (i386_pe_asm_lto_start): New function.
2350 (i386_pe_asm_lto_end): Likewise.
2352 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2353 Richard Biener <rguenther@suse.de>
2356 * lto-streamer-out.c (output_function): Fix CFG loop state before
2358 * omp-expand.c (expand_oacc_for): Handle calls to internal
2359 functions like regular functions.
2361 2018-04-12 Richard Biener <rguenther@suse.de>
2364 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2365 for the early LTO debug to properly generate references to it
2366 during DIE emission. Do not re-use that for the skeleton for
2368 (dwarf2out_early_finish): Likewise.
2370 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2373 * config/i386/sse.md
2374 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2375 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2376 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2377 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2378 and output is a reg, avoid creating invalid lowpart subreg, but
2379 instead split into a 512-bit move. Don't split if not AVX512VL,
2380 input is xmm16+ reg and output is a mem.
2381 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2382 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2383 xmm16+ reg and output is a mem.
2385 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2387 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2388 also for flag_dwarf2_cfi_asm.
2390 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2392 PR rtl-optimization/85342
2393 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2394 a bool scalar var inside of the loop instead. Don't try to update
2395 recog_data.operand after failed apply_change_group.
2397 2018-04-12 Tom de Vries <tom@codesourcery.com>
2400 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2401 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2402 array with flexible array member as array without given dimension.
2403 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2404 argument for undefined param to true.
2406 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2409 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2410 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2411 from PowerPC section.
2412 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2413 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2414 help text that is too long.
2415 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2416 help text that is too long.
2417 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2418 help text that is too long.
2420 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2422 * config/alpha/alpha.md (stack_probe_internal): Rename
2423 from "probe_stack". Update all callers.
2425 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2427 PR rtl-optimization/84566
2428 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2429 sched_macro_fuse_insns.
2431 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2434 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2435 (compute_block_dependences): ... from here.
2437 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2439 PR tree-optimization/85331
2440 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2441 from int to HOST_WIDE_INT.
2443 2018-04-11 Martin Jambor <mjambor@suse.cz>
2446 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2447 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2448 not the same as the source val.
2449 (cgraph_edge_brings_value_p): New parameter.
2450 (gather_edges_for_value): Pass destination value to
2451 cgraph_edge_brings_value_p.
2452 (perhaps_add_new_callers): Likewise.
2453 (get_info_about_necessary_edges): Likewise and exclude values brought
2454 only by self-recursive edges.
2455 (create_specialized_node): Redirect only clones of self-calling edges.
2456 (+self_recursive_pass_through_p): New function.
2457 (find_more_scalar_values_for_callers_subset): Use it.
2458 (find_aggregate_values_for_callers_subset): Likewise.
2459 (known_aggs_to_agg_replacement_list): Removed.
2460 (decide_whether_version_node): Re-calculate known constants for all
2461 remaining context clones.
2463 2018-04-11 Richard Biener <rguenther@suse.de>
2466 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2467 from early DWARF output.
2468 (dwarf2out_early_finish): Output line info unconditionally into
2469 early DWARF and add reference to it.
2471 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2474 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2475 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2476 other than V2DFmode using iptr mode attribute.
2477 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2479 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2481 PR rtl-optimization/84659
2482 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2484 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2487 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2489 (output_loc_list): Pass address of a dummy size variable even in the
2490 locview handling loop.
2491 (index_location_lists): Add comment on why skip_loc_list_entry can't
2494 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2497 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2500 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2503 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2504 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2505 and -mstring-compare-inline-limit.
2507 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2510 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2511 for stack clash protection in a register whenever we need it to be in
2514 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2516 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2517 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2519 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2522 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2524 (mlong-double-): Ditto.
2525 * config/rs6000/sysv4.opt (msdata=): Ditto.
2526 (mtls-size=): Ditto.
2528 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2530 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2531 erroneous entries for
2532 "vector int vec_ldl (int, long int *)", and
2533 "vector unsigned int vec_ldl (int, unsigned long int *)".
2534 Add comments and entries for
2535 "vector bool char vec_ldl (int, bool char *)",
2536 "vector bool short vec_ldl (int, bool short *)",
2537 "vector bool int vec_ldl (int, bool int *)",
2538 "vector bool long long vec_ldl (int, bool long long *)",
2539 "vector pixel vec_ldl (int, pixel *)",
2540 "vector long long vec_ldl (int, long long *)",
2541 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2542 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2543 type tree bool_long_long_type_node and correct definition of
2544 bool_V2DI_type_node to make reference to this new type tree.
2545 (rs6000_mangle_type): Replace erroneous reference to
2546 bool_long_type_node with bool_long_long_type_node.
2547 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2548 comments to emphasize sign distinctions for char and int types and
2549 replace RS6000_BTI_bool_long constant with
2550 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2551 use of RS6000_BTI_pixel.
2552 (bool_long_type_node): Remove this macro definition.
2553 (bool_long_long_type_node): New macro definition
2555 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2557 PR rtl-optimization/85300
2558 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2559 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2560 simplify_unary_operation fails.
2562 2018-04-10 Martin Liska <mliska@suse.cz>
2564 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2565 cgraph_edge and ipa_ref.
2567 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2571 * config/i386/sse.md
2572 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2573 computation of the VEC_MERGE selector from mask.
2574 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2575 Fix decoding of the VEC_MERGE selector into mask.
2577 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2579 PR tree-optimization/85286
2580 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2582 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2584 * final.c (final_1): Set insn_last_address as well as
2585 insn_current_address.
2587 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2590 * explow.c (emit_stack_probe): Call validize_mem on memory location
2591 before passing it to gen_probe_stack. Create address operand and
2592 legitimize it for the probe_stack_address case.
2594 2018-04-09 Jan Hubicka <jh@suse.cz>
2597 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2598 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2599 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2601 (free_lang_data): Rebuild type inheritance graph.
2603 2018-04-09 Martin Sebor <msebor@redhat.com>
2605 * invoke.texi (-finline-small-functions): Mention other optimization
2607 (-findirect-inlining, -fpartial-inlining): Same.
2608 (-finline-functions-called-once): Same.
2609 (-freorder-blocks-and-partition): Same.
2611 2018-04-09 Jan Hubicka <jh@suse.cz>
2614 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2615 jumps; choose last target that matches the criteria (i.e.
2616 no partition changes for non-crossing jumps).
2617 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2618 support for redirecting crossing jumps to non-crossing.
2620 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2622 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2623 also for naked functions.
2625 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2627 * config/arc/arc.md (add_shift): New pattern.
2628 (add_shift2): Likewise.
2629 (sub_shift): Likewise.
2630 (sub_shift_cmp0_noout): Likewise.
2631 (compare_si_ashiftsi): Likewise.
2632 (xbfu_cmp0_noout): New combine pattern.
2633 (xbfu_cmp0"): Likewise.
2634 (movsi_set_cc_insn): Place the predicable variant first.
2635 (commutative_binary_cmp0_noout): Remove clobber.
2636 (commutative_binary_cmp0): New pattern.
2637 (noncommutative_binary_cmp0): Likewise.
2638 (noncommutative_binary_cmp0_noout): Likewise.
2639 (noncommutative_binary_comparison_result_used): Removed.
2640 (rsub_cmp0): New pattern.
2641 (rsub_cmp0_noout): Likewise.
2642 (extzvsi): Changed, keep only meaningful variants.
2643 (SQH, SEZ): New iterators.
2644 (SQH_postfix): New mode attribute.
2645 (SEZ_prefix): New code attribute.
2646 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2647 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2648 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2650 (noncommutative_operator): Check the availability of barrel
2653 2018-04-09 Richard Biener <rguenther@suse.de>
2655 PR tree-optimization/85284
2656 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2657 Only use the niter constraining form of simple_iv when the exit
2660 2018-04-09 Tom de Vries <tom@codesourcery.com>
2663 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2664 (define_expand "*memory_barrier"): New define_expand.
2665 (define_insn "memory_barrier"): New insn.
2667 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2669 PR rtl-optimization/80463
2670 PR rtl-optimization/83972
2671 PR rtl-optimization/83480
2673 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2674 correct producer for the insn.
2675 (tidy_control_flow): Fixup seqnos in case of debug insns.
2677 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2679 PR rtl-optimization/83913
2681 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2682 different sched-times when merging exprs.
2684 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2686 PR rtl-optimization/83962
2688 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2689 tidy_fallthru_edge and tidy_control_flow.
2691 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2693 PR rtl-optimization/83530
2695 * sel-sched.c (force_next_insn): New global variable.
2696 (remove_insn_for_debug): When force_next_insn is true, also leave only
2697 next insn in the ready list.
2698 (sel_sched_region): When the region wasn't scheduled, make another pass
2699 over it with force_next_insn set to 1.
2701 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2703 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2705 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2706 for interrupt control.
2707 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2708 functions for interrupt control.
2709 * config/nds32/nds32-intrinsic.md: Likewise.
2710 * config/nds32/nds32_intrinsic.h: Likewise.
2711 * config/nds32/nds32.h (nds32_builtins): Likewise.
2713 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2715 * config/nds32/nds32.c (nds32_init_machine_status,
2716 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2717 strict_aligned_p field.
2718 (nds32_expand_to_rtl_hook): New function.
2719 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2720 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2722 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2723 Chung-Ju Wu <jasonwucj@gmail.com>
2725 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2726 * config/nds32/nds32-n7.md: New file.
2727 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2728 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2730 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2731 * config/nds32/nds32.md (pipeline_model): Add n7.
2732 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2733 * config/nds32/pipelines.md: Include n7 settings.
2735 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2736 Chung-Ju Wu <jasonwucj@gmail.com>
2738 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2739 * config/nds32/nds32-e8.md: New file.
2740 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2741 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2743 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2744 * config/nds32/nds32.md (pipeline_model): Add e8.
2745 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2746 * config/nds32/pipelines.md: Include e8 settings.
2748 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2749 Chung-Ju Wu <jasonwucj@gmail.com>
2751 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2752 * config/nds32/nds32-n8.md: New file.
2753 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2754 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2756 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2757 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2758 * config/nds32/nds32.md (pipeline_model): Add n8.
2759 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2760 * config/nds32/pipelines.md: Include n8 settings.
2762 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2763 Chung-Ju Wu <jasonwucj@gmail.com>
2765 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2766 * config/nds32/nds32-n9-2r1w.md: New file.
2767 * config/nds32/nds32-n9-3r2w.md: New file.
2768 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2769 nds32_register_ports): New or modify for cpu n9.
2770 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2772 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2773 * config/nds32/nds32-utils.c: New file.
2774 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2775 TARGET_MUL_SLOW): Define.
2776 * config/nds32/nds32.md (pipeline_model): New attribute.
2777 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2778 New options that support cpu n9.
2779 * config/nds32/pipelines.md: Include n9 settings.
2780 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2782 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2784 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2785 information if necessary.
2786 (output_cond_branch_compare_zero): Likewise.
2787 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2788 (nds32_target_alignment): Refine for alignment.
2789 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2790 (FUNCTION_BOUNDARY): Modify.
2791 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2793 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2795 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2797 * config/nds32/constants.md (unspec_volatile_element): Add values for
2798 TLB operation and data prefetch.
2799 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2800 functions for TLB operation and data prefetch.
2801 * config/nds32/nds32-intrinsic.md: Likewise.
2802 * config/nds32/nds32_intrinsic.h: Likewise.
2803 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2804 (nds32_print_operand): Likewise.
2805 * config/nds32/nds32.h (nds32_builtins): Likewise.
2807 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2808 Andrew Pinski <pinsika@gcc.gnu.org>
2811 * match.pd: Use constant_boolean_node of correct type instead of
2812 boolean_true_node or boolean_false_node for simplifying
2813 pointer comparisons to zero.
2815 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2817 PR tree-optimization/80021
2818 * tree.c (verify_type_variant): Make error call in verify_variant_match
2819 translatable and remove final full stop.
2821 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2823 * config/nds32/constants.md (unspec_volatile_element): Add
2824 UNSPEC_VOLATILE_EH_RETURN.
2825 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2826 nds32_output_stack_pop): Support dwarf exception handling process.
2827 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2828 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2829 exception handling process.
2830 (nds32_compute_stack_frame): Likewise.
2831 (nds32_return_addr_rtx): Likewise.
2832 (nds32_initial_elimination_offset): Likewise.
2833 (nds32_expand_prologue): Likewise.
2834 (nds32_expand_epilogue): Likewise.
2835 (nds32_dynamic_chain_address): New function.
2836 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2838 (DYNAMIC_CHAIN_ADDRESS): Define.
2839 (EH_RETURN_DATA_REGNO): Define.
2840 (EH_RETURN_STACKADJ_RTX): Define.
2841 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2842 patterns for dwarf exception handling.
2844 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2846 * config/nds32/nds32.h: Clean up obsolete macros.
2848 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2850 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2851 Add enum values for particular instructions.
2852 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2853 particular intrinsic functions.
2854 * config/nds32/nds32-intrinsic.md: Likewise.
2855 * config/nds32/nds32_intrinsic.h: Likewise.
2856 * config/nds32/nds32.h (nds32_builtins): Likewise.
2857 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2858 (btst, ave): New patterns for particular instructions.
2860 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2862 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2863 Add enum values for atomic load/store and memory sync.
2864 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2866 * config/nds32/nds32-intrinsic.md: Likewise.
2867 * config/nds32/nds32_intrinsic.h: Likewise.
2868 * config/nds32/nds32.h (nds32_builtins): Likewise.
2870 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2872 PR tree-optimization/85257
2873 * fold-const.c (native_encode_vector): If not all elts could fit
2874 and off is -1, return 0 rather than offset.
2875 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2876 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2877 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2878 adjust buffer in native_interpret_expr call.
2880 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2882 * config/nds32/constants.md (unspec_volatile_element): Add cache
2883 control enum values.
2884 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2885 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2886 * config/nds32/nds32.c (nds32_cctl_names): New.
2887 (nds32_print_operand): Handle cache control register names.
2888 * config/nds32/nds32.h (nds32_builtins): New enum values.
2889 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2891 * config/nds32/nds32.md (type): Add mmu.
2892 * config/nds32/pipelines.md (simple_insn): Add mmu.
2894 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2896 * config/nds32/nds32.md (type): Remove call.
2897 * config/nds32/pipelines.md (simple_insn): Likewise.
2899 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2901 * config/nds32/constants.md (unspec_volatile_element): Add
2902 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
2903 UNSPEC_VOLATILE_FMFCFG.
2904 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
2905 description for fmfcfg and fmfcsr.
2906 (bdesc_1arg): Add fmtcsr.
2907 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
2908 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
2909 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
2910 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
2911 unspec_fmfcfg): New patterns.
2912 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
2913 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
2914 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
2915 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
2916 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
2917 __nds32__fmfcfg): Define.
2919 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2921 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
2922 intrinsic register names.
2923 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
2924 intrinsic register enum values and macros.
2926 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2928 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
2929 for load/store addressing form.
2930 (nds32_print_operand_address): Likewise.
2932 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
2935 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
2936 based on LABEL_REF. Remove useless assertion.
2937 (pic_address_needs_scratch): Fix formatting.
2938 (sparc_legitimize_pic_address): Minor tweaks.
2939 (sparc_delegitimize_address): Adjust assertion accordingly.
2940 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
2941 into symbolic_operand.
2942 (movsi_high_pic_label_ref): Likewise.
2943 (movsi_lo_sum_pic_label_ref): Likewise.
2944 (movdi_pic_label_ref): Likewise.
2945 (movdi_high_pic_label_ref): Likewise.
2946 (movdi_lo_sum_pic_label_ref): Likewise.
2948 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
2950 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
2951 custom LIB_SPEC setup.
2953 2018-04-06 Ruslan Bukin <br@bsdpad.com>
2954 Kito Cheng <kito.cheng@gmail.com>
2956 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
2957 * config/riscv/freebsd.h: New.
2959 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2961 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
2962 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
2965 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2966 Kito Cheng <kito.cheng@gmail.com>
2968 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
2969 nds32_output_call, nds32_symbol_binds_local_p): New functions.
2970 * config/nds32/nds32-protos.h (nds32_output_call,
2971 nds32_output_return): Declare.
2972 * config/nds32/nds32.md: Refine all the call and return patterns.
2974 2018-04-06 Jakub Jelinek <jakub@redhat.com>
2977 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
2978 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
2980 PR rtl-optimization/84872
2981 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
2982 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
2985 2018-04-06 Tamar Christina <tamar.christina@arm.com>
2987 * expr.c (copy_blkmode_to_reg): Revert 254862.
2988 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
2990 2018-04-06 Richard Biener <rguenther@suse.de>
2993 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
2994 after seeing a component reference with an adjacent field. Treat
2995 refs to arrays at struct end of external decls similar to
2996 refs to unconstrained commons.
2998 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3001 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3002 look through SAVE_EXPRs with non-side-effects argument. Adjust
3004 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3007 2018-04-06 Richard Biener <rguenther@suse.de>
3010 * alias.c (find_base_term): New wrapper around find_base_term
3011 unwinding CSELIB_VAL_PTR changes.
3012 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3015 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3017 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3019 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3020 constant definitions.
3021 ("nop"): lr 0,0 -> nopr r0
3022 ("nop_lr0", "nop_lr1"): New insn definitions.
3024 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3026 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3027 NDS32_V3PUSH_AVAILABLE_P macro.
3029 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3030 Chung-Ju Wu <jasonwucj@gmail.com>
3032 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3033 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3034 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3035 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3036 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3037 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3038 * config/nds32/constraints.md: New constraints and checking for hard
3039 float configuration.
3040 * config/nds32/iterators.md: New mode iterator and attribute for hard
3041 float configuration.
3042 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3044 * config/nds32/nds32-fpu.md: New file.
3045 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3046 deal with hard float code generation.
3047 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3049 (abi_type, float_reg_number): New enum type.
3050 * config/nds32/nds32-predicates.c: New predicates for hard float.
3051 * config/nds32/nds32-protos.h: Declare functions for hard float.
3052 * config/nds32/nds32.c: Implementation for hard float configuration.
3053 * config/nds32/nds32.h: Definitions for hard float configuration.
3054 * config/nds32/nds32.md: Include hard float machine description and
3055 modify patterns for hard float configuration.
3056 * config/nds32/nds32.opt: New options for hard float configuration.
3057 * config/nds32/predicates.md: New predicates for hard float
3060 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3062 * common/config/nds32/nds32-common.c
3063 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3065 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3068 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3069 CONSTRUCTOR_ELT (ctor, ...)->value.
3071 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3074 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3076 2018-04-05 Tom de Vries <tom@codesourcery.com>
3079 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3082 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3083 Kito Cheng <kito.cheng@gmail.com>
3085 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3086 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3087 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3088 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3090 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3091 Kito Cheng <kito.cheng@gmail.com>
3093 * config/nds32/constraints.md (Ufe): New memory constraint.
3094 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3095 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3096 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3098 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3099 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3101 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3103 * config/nds32/nds32.md: Use optimize_size in the condition for
3104 alu-shift instructions.
3106 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3108 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3110 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3112 * config/nds32/nds32.md (negsi2): Refine pattern.
3114 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3115 Chung-Ju Wu <jasonwucj@gmail.com>
3117 * config/nds32/iterators.md (shift_rotate): New code iterator.
3118 (shift): New code attribute.
3119 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3120 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3121 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3122 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3123 bit-wise operations.
3124 (andsi3, *andsi3): Ditto.
3125 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3126 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3127 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3128 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3129 nds32_ior_operand, nds32_xor_operand): New predicates.
3131 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3133 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3134 (addsi3, subsi3): ... this.
3136 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3138 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3140 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3142 * config/nds32/nds32.md: Adjust indention.
3144 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3146 * config/nds32/nds32.md (feature): New attribute.
3148 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3150 * config/nds32/nds32.md (subtype): New attribute.
3152 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3155 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3156 expansion to perform a bitwise AND of the argument followed by a
3157 boolean negation of the result.
3159 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3161 PR rtl-optimization/84878
3162 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3163 the basic block. Assert the use reference is not artificial and that
3164 it has an associated insn.
3166 2018-04-04 Michael Matz <matz@suse.de>
3168 * builtins.c (compute_objsize): Pass correct operand
3169 to array_at_struct_end_p.
3171 2018-04-04 Richard Biener <rguenther@suse.de>
3174 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3175 from contexts for DINFO_LEVEL_TERSE and below.
3177 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3179 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3180 resiter_operand condition.
3181 * config/nds32/nds32.md (*move<mode>): Ditto.
3183 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3184 Monk Chiang <sh.chiang04@gmail.com>
3186 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3188 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3190 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3192 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3193 Kito Cheng <kito.cheng@gmail.com>
3195 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3196 nds32_cond_code_str, output_cond_branch,
3197 output_cond_branch_compare_zero, nds32_expand_cbranch,
3198 nds32_expand_cstore, nds32_expand_movcc,
3199 nds32_output_cbranchsi4_equality_zero,
3200 nds32_output_cbranchsi4_equality_reg,
3201 nds32_output_cbranchsi4_equality_reg_or_const_int,
3202 nds32_output_cbranchsi4_greater_less_zero: New functions.
3203 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3204 nds32_expand_cstore, nds32_expand_movcc,
3205 nds32_output_cbranchsi4_equality_zero,
3206 nds32_output_cbranchsi4_equality_reg,
3207 nds32_output_cbranchsi4_equality_reg_or_const_int,
3208 nds32_output_cbranchsi4_greater_less_zero): Declare.
3209 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3210 nds32_rimm11s_operand): New predicates.
3211 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3212 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3215 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3217 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3218 * config/nds32/nds32.md: Ditto.
3219 * config/nds32/pipelines.md: Ditto.
3221 2018-04-04 Richard Biener <rguenther@suse.de>
3223 PR tree-optimization/85168
3224 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3225 propagating abnormals.
3227 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3229 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3231 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3232 Kito Cheng <kito.cheng@gmail.com>
3234 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3235 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3236 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3237 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3238 * config/nds32/nds32.md (sibcall_internal): New.
3239 (sibcall_register): Remove.
3240 (sibcall_immediate): Remove.
3241 (sibcall_value_internal): New.
3242 (sibcall_value_register): Remove.
3243 (sibcall_value_immediate): Remove.
3244 * config/nds32/predicates.md (nds32_general_register_operand): New.
3245 (nds32_call_address_operand): New.
3247 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3249 PR rtl-optimization/85167
3250 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3251 bb_defs if *split_p, instead preinitialize it to NULL.
3253 PR tree-optimization/85156
3254 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3255 evaluating the argument multiple times.
3257 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3259 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3261 (_mm_cvtpd_ps): Likewise.
3262 (_mm_cvttpd_epi32): Likewise.
3263 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3264 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3265 vector, pixel, and bool following altivec.h include.
3267 2018-04-03 Martin Sebor <msebor@redhat.com>
3269 * doc/extend.texi (Common Function Attributes): Clarify.
3270 (const attribute): Likewise.
3271 (pure attribute): Likewise.
3273 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3276 * config/i386/i386.c (ix86_expand_vector_set): Use
3277 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3279 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3281 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3282 instructions when changing rounding bits to preserve precision bits
3283 in the x87 control word.
3285 2018-04-03 Martin Liska <mliska@suse.cz>
3287 PR tree-optimization/82491
3288 * rtl.h (strip_offset_and_add): Replace += suboffset with
3289 poly_uint64 () + suboffset.
3291 2018-03-29 Martin Liska <mliska@suse.cz>
3292 Martin Jambor <mjambor@suse.cz>
3295 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3296 param_type is not an integral or pointer type.
3298 2018-04-03 Richard Biener <rguenther@suse.de>
3300 * sese.h (recompute_all_dominators): Remove.
3302 2018-04-02 Martin Sebor <msebor@redhat.com>
3304 * doc/invoke.texi (-Wrestrict): Fix typos.
3306 2018-04-02 Jim Wilson <jimw@sifive.com>
3308 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3309 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3310 (<optab>di3, <optab>si3_extend): Likewise.
3311 (<optab>si3_mask, <optab>si3_mask_1): New.
3312 (<optab>di3_mask, <optab>di3_mask_1): New.
3313 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3314 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3315 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3317 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3319 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3322 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3324 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3325 (nds32_canonicalize_comparison): New function.
3327 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3328 Kito Cheng <kito.cheng@gmail.com>
3329 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3331 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3332 * config/nds32/constants.md (unspec_volatile_element): Add
3333 UNSPEC_VOLATILE_RELAX_GROUP.
3334 * config/nds32/nds32-relax-opt.c: New file.
3335 * config/nds32/nds32-predicates.c
3336 (nds32_symbol_load_store_p): New function.
3337 * config/nds32/nds32-protos.h
3338 (nds32_symbol_load_store_p): Declare function.
3339 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3340 * config/nds32/nds32.c
3341 (nds32_register_pass): New function to register pass.
3342 (nds32_register_passes): New function to register passes.
3343 * config/nds32/nds32.md (relax_group): New pattern.
3344 * config/nds32/nds32.opt (mrelax-hint): New option.
3345 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3347 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3349 * config/nds32/t-nds32: Modify files dependency.
3351 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3353 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3354 (PROFILE_HOOK): Define its implementation.
3356 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3358 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3359 type and 32-bit size.
3361 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3364 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3365 (V_128_256): New mode iterator.
3366 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3367 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3368 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3370 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3371 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3373 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3376 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3377 NaN inputs correctly.
3379 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3382 * config/rs6000/vsx.md (??r): New mode attribute.
3383 (*vsx_mov<mode>_64bit): Use it.
3384 (*vsx_mov<mode>_32bit): Likewise.
3386 2018-03-30 Martin Sebor <msebor@redhat.com>
3388 PR tree-optimization/84818
3389 * builtins.c (check_access): Use warning_n.
3391 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3394 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3396 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3399 2018-03-30 Julia Koval <julia.koval@intel.com>
3402 * x86-tune.def (movx, partial_reg_dependency): Enable for
3405 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3408 * lra-constraints.c (process_alt_operands): Move setting
3409 this_alternative_matches below.
3411 2018-03-29 Martin Liska <mliska@suse.cz>
3414 * doc/invoke.texi: Document how LTO works with debug info.
3415 Describe auto-load support of binutils. Mention 'x86-64'
3416 as valid option value of -march option.
3418 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3420 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3423 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3424 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3425 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3428 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3431 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3432 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3433 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3434 for RS6000_BTM_POWERPC64.
3435 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3436 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3437 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3442 2018-03-28 Carl Love <cel@us.ibm.com>
3445 2017-09-27 Carl Love <cel@us.ibm.com>
3447 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3448 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3449 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3452 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3454 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3455 instead of __vector bool.
3456 (_mm_max_pu8): Likewise.
3457 (_mm_min_pi16): Likewise.
3459 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3462 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3463 (DIVWEUO): Likewise.
3465 (DIVDEUO): Likewise.
3466 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3467 DIVWEUO and DIVDEUO.
3468 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3469 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3470 (div_extend): Likewise.
3471 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3473 (__builtin_divweuo): Likewise.
3474 (__builtin_divdeo): Likewise.
3475 (__builtin_divdeuo): Likewise.
3477 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3480 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3481 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3483 PR tree-optimization/82004
3484 * gimple-match-head.c (optimize_pow_to_exp): New function.
3485 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3486 Don't fold to exp if optimize_pow_to_exp is false.
3488 2018-03-28 Martin Liska <mliska@suse.cz>
3491 * calls.c (initialize_argument_information): Fix trailing space.
3492 * common.opt: Fix typo and provide better explanation for
3493 -fsanitize-coverage option.
3494 * config/i386/i386.opt: Fix typo.
3496 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3497 Martin Liska <mliska@suse.cz>
3500 * gimplify.c (asan_poison_variable): Don't do the check for
3501 gimplify_omp_ctxp here.
3502 (gimplify_decl_expr): Do it here.
3503 (gimplify_target_expr): Likewise.
3505 2018-03-28 Martin Liska <mliska@suse.cz>
3508 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3509 chkp_type_bounds_count if MPX is not enabled.
3511 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3513 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3515 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3518 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3519 function to create the function decl for complex long double
3520 multiply and divide for -mabi=ieeelongdouble.
3521 (init_float128_ieee): Call it.
3523 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3526 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3527 -fcf-protection=branch -mibt.
3528 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3530 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3533 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3535 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3538 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3539 extern array declarations.
3541 2018-03-27 Richard Biener <rguenther@suse.de>
3544 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3545 explicit single_use checks.
3547 2018-03-27 Richard Biener <rguenther@suse.de>
3549 PR tree-optimization/85082
3550 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3553 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3555 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3556 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3557 Turn on fasynchronous-unwind-tables and funwind-tables.
3559 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3562 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3563 (*bmi_blsr_<mode>_ccz): Ditto.
3565 2018-03-26 Tom de Vries <tom@codesourcery.com>
3567 PR tree-optimization/85063
3568 * omp-general.c (offloading_function_p): New function. Factor out
3570 * omp-offload.c (pass_omp_target_link::gate): ... here.
3571 * omp-general.h (offloading_function_p): Declare.
3572 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3573 with attribute omp declare target for offloading functions.
3575 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3577 PR tree-optimization/84005
3578 * tree-data-ref.h (get_base_for_alignment): Declare.
3579 * tree-data-ref.c (get_base_for_alignment_1): New function.
3580 (get_base_for_alignment): Likewise.
3581 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3582 get_base_for_alignment to find a suitable base object, instead
3583 of always using drb->base_address.
3585 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3588 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3589 known size by default.
3591 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3594 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3595 and non BLKmode operands.
3597 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3600 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3601 Clean up attributes.
3603 2018-03-23 Richard Biener <rguenther@suse.de>
3606 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3607 we are going to emit early debug for LTO.
3609 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3612 * function.c (match_asm_constraints_1): Don't optimize if input
3613 doesn't satisfy general_operand predicate for output's mode.
3616 * alias.c (write_dependence_p): Don't require for x_canonicalized
3617 non-VOIDmode if x has VOIDmode.
3620 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3621 just don't try to optimize it rather than assert it never happens.
3623 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3625 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3626 macro expansions for definition of ST_INTERNAL_<mode> and
3627 LD_INTERNAL_<mode> builtins.
3628 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3630 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3632 (altivec_expand_st_builtin): Likewise.
3633 (altivec_expand_builtin): Remove calls to deleted functions.
3634 (rs6000_address_for_altivec): Delete this function.
3635 * config/rs6000/vector.md: Remove expands for
3636 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3638 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3641 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3642 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3643 re-computing once computed.
3644 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3645 (arm_init_machine_status): Initialize
3646 machine->static_chain_stack_bytes.
3648 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3651 * doc/extend.texi: Add four new prototypes for vec_ld.
3652 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3653 definitions for more logical presentation.
3654 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3655 entries for V1TI variants of __builtin_altivec_ld builtin.
3656 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3657 handling of V1TI variant of LVX icode pattern.
3658 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3659 (rs6000_gimple_fold_builtin): Likewise.
3660 (altivec_init_builtins): Add code to define
3661 __builtin_altivec_lvx_v1ti function.
3663 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3666 * function.c (match_asm_constraints_1): Don't do the optimization
3667 if input isn't a REG, SUBREG, MEM or constant.
3669 2018-03-22 Tom de Vries <tom@codesourcery.com>
3671 PR tree-optimization/84956
3672 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3673 bb_has_abnormal_pred.
3675 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3678 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3679 DECL_INITIAL (decl) to decl at the end.
3680 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3683 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3685 * doc/extend.texi (__builtin_tgmath): Document when complex
3686 integer types are treated as _Complex _Float64.
3688 2018-03-21 Tom de Vries <tom@codesourcery.com>
3690 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3692 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3694 PR tree-optimization/84960
3695 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3696 if it is ENTRY block, move them into single succ of ENTRY in that case.
3698 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3700 PR tree-optimization/84811
3701 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3702 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3703 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3704 (known_subrange_p): Update accordingly. Cast each value involved
3705 in the size comparison, rather than casting the result of the
3708 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3710 PR tree-optimization/84982
3711 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3712 by flipping the least significant bit rather than all bits from
3713 bitpos to bitpos + bitsize - 1.
3715 2018-03-21 Nathan Sidwell <nathan@acm.org>
3717 * doc/extend.texi (Deprecated Features): Remove mention of
3718 long-deleted deprecations.
3720 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3723 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3724 * configure: Regenerate.
3726 2018-03-21 Tom de Vries <tom@codesourcery.com>
3728 PR tree-optimization/83126
3729 * tree-parloops.c (num_phis): New function.
3730 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3732 2018-03-21 Nathan Sidwell <nathan@acm.org>
3734 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3735 mention anon-struct/union members and trailing attributes.
3737 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3739 PR tree-optimization/84969
3740 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3741 builtin memset partitions if they set different rhs values.
3743 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3745 PR rtl-optimization/84989
3746 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3747 VEC_DUPLICATE with scalar result mode.
3749 2018-03-21 Martin Liska <mliska@suse.cz>
3752 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3753 not intended return statement.
3755 2018-03-21 Martin Liska <mliska@suse.cz>
3758 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3759 (chkp_find_bound_slots_1): Limit number of iterations.
3761 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3764 * Minor grammar fixes for x86 options.
3766 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3769 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3770 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3772 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3775 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3776 (altivec_lvx_<mode>_1op): Likewise.
3777 (altivec_stvx_<mode>_2op): Likewise.
3778 (altivec_stvx_<mode>_1op): Likewise.
3779 (altivec_lvx_<VM2:mode>): New define_expand.
3780 (altivec_stvx_<VM2:mode>): Likewise.
3781 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3782 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3783 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3784 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3785 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3786 (rs6000_gen_lvx): Likewise.
3787 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3788 (altivec_expand_stv_builtin): Likewise.
3789 (altivec_expand_builtin): Likewise.
3790 * config/rs6000/vector.md: Likewise.
3792 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3795 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3798 2018-03-20 Richard Biener <rguenther@suse.de>
3801 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3802 sign-conversions as zero, fall back to standard scalar_stmt
3805 2018-03-20 Martin Liska <mliska@suse.cz>
3808 * predict.c (rebuild_frequencies): Handle case when we have
3809 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3811 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3814 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3815 flag_section_anchors.
3816 * varasm.c (use_blocks_for_decl_p): Remove hack for
3817 dw2_force_const_mem.
3820 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3822 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3823 be created, use lowpart_subreg of operands[0] rather than operands[0]
3825 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3826 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3827 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3828 and n constraint instead of aarch64_shift_imm_di and Usd.
3829 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3830 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3832 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3835 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3836 to favor GPR over NEON registers.
3837 (<shift>di3_neon): Likewise.
3839 2018-03-20 Tom de Vries <tom@codesourcery.com>
3842 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3843 (nvptx_process_pars): Emit bar.sync asap and alap.
3845 2018-03-20 Tom de Vries <tom@codesourcery.com>
3848 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3849 seen_label if seen_label is already set.
3851 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3854 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3855 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3856 Use 1U instead of 1. Formatting fixes.
3859 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3860 instead of TREE_TYPE (s1) for the return value.
3862 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3864 PR tree-optimization/84946
3865 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3866 bitsize + bitsize in poly_uint64 rather than poly_int64.
3869 * dwarf2asm.c: Include fold-const.c.
3870 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3871 of decl rather than decl itself.
3873 PR rtl-optimization/84643
3874 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3876 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3879 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3880 calling assemble_variable.
3882 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3885 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3886 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3888 2018-03-19 Jim Wilson <jimw@sifive.com>
3891 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3892 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3893 (riscv_first_stack_step): Likewise.
3894 (riscv_option_override): Use STACK_BOUNDARY instead of
3896 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3898 (BIGGEST_ALIGNMENT): Set to 128.
3899 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
3900 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
3903 2018-03-19 Richard Biener <rguenther@suse.de>
3905 PR tree-optimization/84933
3906 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
3907 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
3909 2018-03-19 Richard Biener <rguenther@suse.de>
3911 PR tree-optimization/84859
3912 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
3913 (cond_if_else_store_replacement): Perform sinking operation on
3914 single-store BBs regardless of MAX_STORES_TO_SINK setting.
3915 Generalize what a BB with a single eligible store is.
3917 2018-03-19 Richard Biener <rguenther@suse.de>
3919 PR tree-optimization/84929
3920 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
3921 chrec_is_positive against non-chrec arg.
3923 2018-03-19 Tamar Christina <tamar.christina@arm.com>
3926 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
3928 2018-03-18 Martin Liska <mliska@suse.cz>
3930 PR rtl-optimization/84635
3931 * regrename.c (build_def_use): Use matches_mode only when
3934 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
3936 PR tree-optimization/84913
3937 * tree-vect-loop.c (vectorizable_reduction): Don't try to
3938 vectorize chains of COND_EXPRs.
3940 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3942 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
3944 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3946 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
3948 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
3950 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
3952 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3953 Kito Cheng <kito.cheng@gmail.com>
3955 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
3956 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
3957 (nds32_adjust_reg_alloc_order): New function.
3958 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
3960 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
3962 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
3963 nds32_print_operand, nds32_print_operand_address): Use
3964 HOST_WIDE_INT_PRINT_DEC instead.
3966 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
3968 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
3970 2018-03-17 Jakub Jelinek <jakub@redhat.com>
3973 * config/i386/i386.c (initial_ix86_tune_features,
3974 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
3976 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
3977 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
3978 rather than 1u << ix86_tune. Formatting fix.
3979 (ix86_option_override_internal): Change ix86_arch_mask from
3980 unsigned int to unsigned HOST_WIDE_INT, initialize to
3981 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
3982 (ix86_function_specific_restore): Likewise.
3984 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3987 * postreload.c (reload_combine_recognize_pattern): Perform
3988 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
3989 truncate_int_for_mode the result for the destination's mode.
3992 * hsa-gen.c (mem_type_for_type): Fix comment typo.
3993 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
3995 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3998 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4001 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4002 regno_allocno_class_array and sorted_pseudos.
4003 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4004 insns where regno is used.
4006 2018-03-16 Martin Liska <mliska@suse.cz>
4009 * multiple_target.c (create_dispatcher_calls): Redirect
4010 reference in the symbol table.
4012 2018-03-16 Martin Liska <mliska@suse.cz>
4015 * multiple_target.c (create_dispatcher_calls): Redirect also
4018 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4022 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4023 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4024 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4026 2018-03-16 Julia Koval <julia.koval@intel.com>
4028 * doc/invoke.texi (Skylake Server): Add CLWB.
4029 Cannonlake): Remove CLWB.
4031 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4033 PR tree-optimization/84841
4034 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4036 (FLOAT_ONE_CONST_TYPE): Define.
4037 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4038 (sort_by_operand_rank): Put entries with higher constant_type last
4039 rather than first to match comments.
4041 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4043 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4046 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4049 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4050 check and use address of target always.
4052 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4055 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4056 (indirect_thunk_bnd_needed): Likewise.
4057 (indirect_thunks_used): Likewise.
4058 (indirect_thunks_bnd_used): Likewise.
4059 (indirect_return_needed): New.
4060 (indirect_return_bnd_needed): Likewise.
4061 (output_indirect_thunk_function): Add a bool argument for
4063 (output_indirect_thunk_function): Don't generate alias for
4064 function return thunk.
4065 (ix86_code_end): Call output_indirect_thunk_function to generate
4066 function return thunks.
4067 (ix86_output_function_return): Set indirect_return_bnd_needed
4068 and indirect_return_needed instead of indirect_thunk_bnd_needed
4069 and indirect_thunk_needed.
4071 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4073 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4074 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4075 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4077 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4078 Paul Hua <paul.hua.gm@gmail.com>
4081 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4083 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4085 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4086 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4089 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4092 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4093 instead of GET_MODE_SIZE when comparing Units.
4095 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4098 * varasm.c (hash_section): Return an unchangeble hash value
4099 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4100 Return !aarch64_can_use_per_function_literal_pools_p ().
4102 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4105 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4106 rather than address of cmode as last argument to prepare_cmp_insn.
4108 2018-03-15 Julia Koval <julia.koval@intel.com>
4110 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4111 F_AVX512VNNI, F_AVX512BITALG): New.
4113 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4116 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4117 insn for floating-point loads and stores.
4119 2018-03-14 Carl Love <cel@us.ibm.com>
4121 * config/rs6000/rs6000-c.c: Add macro definitions for
4122 ALTIVEC_BUILTIN_VEC_PERMXOR.
4123 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4124 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4125 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4126 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4128 * config/doc/extend.texi: Add prototypes for vec_permxor.
4130 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4133 * diagnostic-show-locus.c (class layout_point): Convert m_line
4134 from int to linenum_type.
4135 (line_span::comparator): Use linenum "compare" function when
4136 comparing line numbers.
4137 (test_line_span): New function.
4138 (layout_range::contains_point): Convert param "row" from int to
4140 (layout_range::intersects_line_p): Likewise.
4141 (layout::will_show_line_p): Likewise.
4142 (layout::print_source_line): Likewise.
4143 (layout::should_print_annotation_line_p): Likewise.
4144 (layout::print_annotation_line): Likewise.
4145 (layout::print_leading_fixits): Likewise.
4146 (layout::annotation_line_showed_range_p): Likewise.
4147 (struct line_corrections): Likewise for field m_row.
4148 (line_corrections::line_corrections): Likewise for param "row".
4149 (layout::print_trailing_fixits): Likewise.
4150 (layout::get_state_at_point): Likewise.
4151 (layout::get_x_bound_for_row): Likewise.
4152 (layout::print_line): Likewise.
4153 (diagnostic_show_locus): Likewise for locals "last_line" and
4155 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4156 * input.c (selftest::test_linenum_comparisons): New function.
4157 (selftest::input_c_tests): Call it.
4158 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4159 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4160 * selftest.h (ASSERT_GT): New macro.
4161 (ASSERT_GT_AT): New macro.
4162 (ASSERT_LT): New macro.
4163 (ASSERT_LT_AT): New macro.
4165 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4167 PR rtl-optimization/84780
4168 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4170 2018-03-14 Martin Liska <mliska@suse.cz>
4172 * tree.c (record_node_allocation_statistics): Use
4173 get_stats_node_kind.
4174 (get_stats_node_kind): New function extracted from
4175 record_node_allocation_statistics.
4176 (free_node): Use get_stats_node_kind.
4178 2018-03-14 Richard Biener <rguenther@suse.de>
4180 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4181 that the value-set of ANTIC_IN doesn't grow.
4184 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4186 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4187 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4189 2018-03-14 Julia Koval <julia.koval@intel.com>
4191 * config.gcc (icelake-client, icelake-server): New.
4193 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4194 (initial_ix86_arch_features): Ditto.
4195 (PTA_SKYLAKE): Add SGX.
4196 (PTA_ICELAKE): Remove.
4197 (PTA_ICELAKE_CLIENT): New.
4198 (PTA_ICELAKE_SERVER): New.
4199 (ix86_option_override_internal): Split up icelake on icelake client and
4201 (get_builtin_code_for_version): Ditto.
4202 (fold_builtin_cpu): Ditto.
4203 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4204 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4205 * config/i386/i386.h (processor_type): Ditto.
4206 * doc/invoke.texi: Ditto.
4208 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4211 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4212 INTEGER_CST offset, add it together with bitpos / 8 and
4213 sign extend based on POINTER_SIZE.
4217 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4220 * config/i386/constraints.md (Yc): New register constraint.
4221 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4222 Use Yc constraint for alternative 2 of operand 0. Remove
4223 preferred_for_speed attribute.
4225 2018-03-14 Richard Biener <rguenther@suse.de>
4227 PR tree-optimization/84830
4228 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4229 with the old one to avoid oscillations.
4231 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4234 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4236 (assign_by_spills): Return a flag of reload assignment failure.
4237 Do not process the reload assignment failures. Do not spill other
4238 reload pseudos if they has the same reg class. Update n if
4240 (lra_assign): Add a return arg. Set up from the result of
4241 assign_by_spills call.
4242 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4243 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4244 usage_insns if it is not NULL.
4245 (spill_hard_reg_in_range): New function.
4246 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4247 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4248 function prototypes.
4249 (lra_assign): Change prototype.
4250 * lra.c (lra): Add code to deal with fails by splitting hard reg
4253 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4255 * config/riscv/riscv.opt (mrelax): New option.
4256 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4257 "norelax" when riscv_mrelax is disabled.
4258 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4260 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4263 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4264 reassociation for int modes.
4266 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4268 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4269 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4271 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4272 * config/aarch64/aarch64-sve.md
4273 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4274 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4275 (*extend<mode><Vwide>2): Rename to...
4276 (aarch64_sve_extend<mode><Vwide>2): ...this.
4277 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4278 renaming the old pattern to...
4279 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4281 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4282 define_expand, renaming the old pattern to...
4283 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4284 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4285 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4286 account when deciding which SVE instruction the optab should use.
4287 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4289 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4291 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4292 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4293 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4294 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4295 (tlsdesc_small_<mode>): Turn a define_expand and use
4296 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4297 (tlsdesc_small_advsimd_<mode>): ...this.
4298 (tlsdesc_small_sve_<mode>): New pattern.
4300 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4302 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4303 (UNSPEC_UMUL_HIGHPART): New constants.
4304 (MUL_HIGHPART): New int iteraor.
4305 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4306 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4308 (*<su>mul<mode>3_highpart): New define_insn.
4310 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4313 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4316 2018-03-13 Martin Liska <mliska@suse.cz>
4319 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4321 (sem_item_optimizer::~sem_item_optimizer): Release it.
4322 (sem_item_optimizer::merge_classes): Register variable aliases.
4323 (sem_item_optimizer::fixup_pt_set): New function.
4324 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4325 * ipa-icf.h: Declare new variables and functions.
4327 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4330 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4331 integer_pow2p@2 and test integer_pow2p in condition.
4332 (A < 0 ? C : 0): Similarly for @1.
4335 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4336 characters starting at p contain '\0' character, don't look beyond
4340 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4341 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4344 * reg-stack.c (change_stack): Change update_end var from int to
4345 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4346 also call set_block_for_insn on the newly added insns and rescan.
4349 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4350 on the last operand.
4353 * tree.c (stabilize_reference_1): Return save_expr (e) for
4354 STATEMENT_LIST even if it doesn't have side-effects.
4356 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4358 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4360 2018-03-12 Renlin Li <renlin.li@arm.com>
4362 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4363 aarch64_output_scalar_simd_mov_immediate.
4365 2018-03-12 Martin Sebor <msebor@redhat.com>
4367 PR tree-optimization/83456
4368 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4369 for perfectly overlapping calls to memcpy.
4370 (gimple_fold_builtin_memory_chk): Same.
4371 (gimple_fold_builtin_strcpy): Handle no-warning.
4372 (gimple_fold_builtin_stxcpy_chk): Same.
4373 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4375 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4377 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4378 parameter. Use it for SFmode.
4379 (rs6000_function_arg_advance_1): Adjust.
4380 (rs6000_function_arg): Adjust.
4381 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4383 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4385 PR rtl-optimization/84169
4386 PR rtl-optimization/84780
4387 * combine.c (can_combine_p): Check for a 2-insn combination whether
4388 the destination register is used between the two insns, too.
4390 2018-03-12 Richard Biener <rguenther@suse.de>
4392 PR tree-optimization/84803
4393 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4394 for refs DR analysis didn't process.
4396 2018-03-12 Richard Biener <rguenther@suse.de>
4398 PR tree-optimization/84777
4399 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4400 force-vectorize loops ignore whether we are optimizing for size.
4402 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4404 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4405 (TARGET_MD_ASM_ADJUST): Define.
4407 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4408 Kito Cheng <kito.cheng@gmail.com>
4409 Chung-Ju Wu <jasonwucj@gmail.com>
4411 * config/nds32/nds32.c (nds32_compute_stack_frame,
4412 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4413 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4414 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4415 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4416 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4417 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4418 * config/nds32/nds32.md (prologue, epilogue): Use macro
4419 NDS32_V3PUSH_AVAILABLE_P to do checking.
4421 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4424 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4425 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4426 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4427 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4428 addition of most attributes on !orig_type_die or the attribute not
4429 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4431 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4432 Chung-Ju Wu <jasonwucj@gmail.com>
4434 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4436 * config/nds32/nds32.opt (mvh): New option.
4438 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4439 Chung-Ju Wu <jasonwucj@gmail.com>
4441 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4443 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4444 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4447 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4448 Chung-Ju Wu <jasonwucj@gmail.com>
4450 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4452 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4453 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4455 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4456 Kito Cheng <kito.cheng@gmail.com>
4457 Chung-Ju Wu <jasonwucj@gmail.com>
4459 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4460 UNSPEC_FFMISM and UNSPEC_FLMISM.
4461 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4462 for ffb, ffmism and flmism.
4463 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4464 (unspec_ffmism): Ditto.
4465 (unspec_flmism): Ditto.
4466 (nds32_expand_builtin_impl): Check if string extension is available.
4467 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4468 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4470 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4473 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4476 * lra-assigns.c (assign_by_spills): Return a flag of reload
4477 assignment failure. Do not process the reload assignment
4478 failures. Do not spill other reload pseudos if they has the same
4480 (lra_assign): Add a return arg. Set up from the result of
4481 assign_by_spills call.
4482 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4483 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4484 usage_insns if it is not NULL.
4485 (spill_hard_reg_in_range): New function.
4486 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4487 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4488 function prototypes.
4489 (lra_assign): Change prototype.
4490 * lra.c (lra): Add code to deal with fails by splitting hard reg
4493 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4496 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4498 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4501 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4502 (dw_val_node): Add val_symbolic_view.
4503 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4504 (symview_upper_bound): New.
4505 (new_line_info_table): Initialize symviews_since_reset.
4506 (dwarf2out_source_line): Count symviews_since_reset and set
4507 symview_upper_bound.
4508 (dw_val_equal_p): Handle symview.
4509 (add_AT_symview): New.
4510 (print_dw_val): Handle symview.
4511 (attr_checksum, attr_checksum_ordered): Likewise.
4512 (same_dw_val_p, size_of_die): Likewise.
4513 (value_format, output_die): Likewise.
4514 (add_high_low_attributes): Use add_AT_symview for entry_view.
4515 (dwarf2out_finish): Reset symview_upper_bound, clear
4518 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4521 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4522 Add strict argument and use it.
4523 (rs6000_split_multireg_move): Update for new strict argument.
4524 (mem_operand_gpr): Disallow all non-offsettable addresses.
4525 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4527 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4530 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4531 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4532 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4535 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4536 decl, use remap_type if we want to use the type.
4538 2018-03-09 Martin Sebor <msebor@redhat.com>
4540 PR tree-optimization/84526
4541 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4543 (builtin_access::generic_overlap): Be prepared to handle non-array
4546 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4548 PR rtl-optimization/84682
4549 * lra-constraints.c (process_address_1): Check is_address flag
4550 for address constraints.
4551 (process_alt_operands): Likewise.
4552 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4553 preprocess_constraints.
4554 * recog.h (preprocess_constraints): Add oploc parameter.
4556 * recog.c (preprocess_constraints): Test address_operand for
4557 CT_ADDRESS constraints.
4559 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4562 * lra-assigns.c (assign_by_spills): Return a flag of reload
4563 assignment failure. Do not process the reload assignment
4564 failures. Do not spill other reload pseudos if they has the same
4566 (lra_assign): Add a return arg. Set up from the result of
4567 assign_by_spills call.
4568 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4569 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4570 usage_insns if it is not NULL.
4571 (spill_hard_reg_in_range): New function.
4572 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4573 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4574 function prototypes.
4575 (lra_assign): Change prototype.
4576 * lra.c (lra): Add code to deal with fails by splitting hard reg
4579 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4582 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4583 Accept complain bool parameter. Only emit errors if it is true.
4584 (arm_parse_cpu_option_name): Likewise.
4585 (arm_target_thumb_only): Adjust callers of the above.
4586 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4587 prototype to take a default true bool parameter.
4588 (arm_parse_arch_option_name): Likewise.
4590 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4591 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4595 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4596 * configure: Regenerate.
4597 * configure.ac ("linker --version-script option"): New.
4598 ("linker soname option"): New.
4600 2018-03-09 Richard Biener <rguenther@suse.de>
4602 PR tree-optimization/84775
4603 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4604 immediate uses of predicate stmts and mark them modified.
4607 PR tree-optimization/84178
4608 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4610 (version_loop_for_if_conversion): Delay update_ssa call.
4611 (tree_if_conversion): Delay update_ssa until after predicate
4614 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4617 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4618 when the function accesses prior frames.
4620 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4623 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4624 gen_llsym, otherwise call maybe_gen_llsym.
4627 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4628 has ',' character inside of it.
4630 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4633 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4634 as clobbering CC_REGNUM.
4636 2018-03-08 Richard Biener <rguenther@suse.de>
4639 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4640 (follow_copies_to_constant): Do not follow SSA names registered
4643 2018-03-08 Richard Biener <rguenther@suse.de>
4645 PR tree-optimization/84178
4646 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4648 (version_loop_for_if_conversion): Delay update_ssa call.
4649 (tree_if_conversion): Delay update_ssa until after predicate
4652 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4654 PR tree-optimization/84178
4655 * tree-if-conv.c (release_bb_predicate): Remove the
4656 the assertion that the stmts have NULL use_ops.
4657 Discard the statements, asserting that they haven't
4658 yet been added to a BB.
4660 2018-03-08 Richard Biener <rguenther@suse.de>
4662 PR tree-optimization/84746
4663 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4664 (phi_translate): Pass in destination ANTIC_OUT set.
4665 (phi_translate_1): Likewise. For a simplified result lookup
4666 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4667 (phi_translate_set): Adjust.
4668 (do_pre_regular_insertion): Likewise.
4669 (do_pre_partial_partial_insertion): Likewise.
4671 2018-03-08 Martin Liska <mliska@suse.cz>
4673 PR gcov-profile/84735
4674 * doc/gcov.texi: Document usage of profile files.
4675 * gcov-io.h: Document changes in the format.
4677 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4681 * dwarf2out.c (struct dw_line_info_table): Update comments for
4683 (FORCE_RESET_NEXT_VIEW): New.
4684 (FORCE_RESETTING_VIEW_P): New.
4685 (RESETTING_VIEW_P): Check for -1 too.
4686 (ZERO_VIEW_P): Likewise.
4687 (new_line_info_table): Force-reset next view.
4688 (dwarf2out_begin_function): Likewise.
4689 (dwarf2out_source_line): Simplify zero_view_p initialization.
4690 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4691 view directly. Omit view when omitting .loc at line 0.
4693 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4695 PR tree-optimization/84740
4696 * tree-switch-conversion.c (process_switch): Call build_constructors
4697 only if info.phi_count is non-zero.
4699 PR tree-optimization/84739
4700 * tree-tailcall.c (find_tail_calls): Check call arguments against
4701 DECL_ARGUMENTS (current_function_decl) rather than
4702 DECL_ARGUMENTS (func) when checking for tail recursion.
4704 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4706 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4707 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4708 Volker Reichelt's entry and add entries for people that perform
4709 GCC fuzzy testing and report numerous bugs.
4711 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4714 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4715 readonly data in sdata, if that is disabled.
4716 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4717 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4718 -mreadonly-in-sdata option.
4720 2018-03-07 Martin Sebor <msebor@redhat.com>
4722 PR tree-optimization/84468
4723 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4724 basic block when looking for nul assignment.
4726 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4729 * except.h (output_function_exception_table): Adjust prototype.
4730 * except.c (output_function_exception_table): Remove FNNAME parameter
4731 and add SECTION parameter. Ouput one part of the table at a time.
4732 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4733 the first part of the exception table and emit unwind directives.
4734 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4735 (i386_pe_seh_cold_init): Likewise.
4736 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4737 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4738 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4739 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4740 * config/i386/winnt.c: Include except.h.
4741 (struct seh_frame_state): Add reg_offset, after_prologue and
4742 in_cold_section fields.
4743 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4744 (i386_pe_seh_cold_init): New function.
4745 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4746 to seh->in_cold_section.
4747 (seh_emit_push): Record the offset of the push.
4748 (seh_emit_save): Record the offet of the save.
4749 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4750 Test seh->after_prologue to disregard the epilogue.
4751 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4752 (i386_pe_end_cold_function): New function.
4754 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4757 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4758 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4761 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4763 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4764 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4767 * multiple_target.c: Include tree-inline.h and intl.h.
4768 (expand_target_clones): Diagnose and fail if node->definition and
4769 !tree_versionable_function_p (node->decl).
4771 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4773 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4775 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4776 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4777 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4779 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4782 * combine.c (try_combine): Use reg_or_subregno instead of handling
4783 just paradoxical SUBREGs and REGs.
4785 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4787 * config/arc/arc.c (arc_finalize_pic): Remove function.
4788 (arc_must_save_register): We use single base PIC register, remove
4789 checks to save/restore the PIC register.
4790 (arc_expand_prologue): Likewise.
4791 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4793 (arc_verify_short): Likewise.
4794 (arc_attr_type): Likewise.
4795 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4796 (walk_stores): Likewise.
4797 (arc_address_cost): Make it static.
4798 (arc_verify_short): Likewise.
4799 (branch_dest): Likewise.
4800 (arc_attr_type): Likewise.
4801 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4802 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4803 (arc_final_prescan_insn): Remove inserting the nops due to
4804 hardware hazards. It is done in reorg step.
4805 (insn_length_variant_t): Remove.
4806 (insn_length_parameters_t): Likewise.
4807 (arc_insn_length_parameters): Likewise.
4808 (arc_get_insn_variants): Likewise.
4809 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4811 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4814 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4817 PR tree-optimization/84687
4818 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4820 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4822 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4824 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4825 Rename to ppc_speculation_barrier.
4826 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4827 __builtin_ppc_speculation_barrier.
4829 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4832 * combine.c (combine_simplify_rtx): Don't try to simplify if
4833 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4836 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4838 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4839 to 32 bytes when compiling for POWER9.
4841 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4844 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4845 regparm >= 3 with no arg reg available also for calls with
4846 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4849 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4851 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4853 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4856 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4858 2018-03-05 Richard Biener <rguenther@suse.de>
4860 PR tree-optimization/84486
4861 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4862 When inserting a __builtin_assume_aligned call set the LHS
4863 SSA name alignment info accordingly.
4865 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4867 PR tree-optimization/84114
4868 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4869 Avoid reassociation of FLOAT_MODE addition.
4871 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4873 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4874 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4875 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4876 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4877 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4878 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4879 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4881 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4882 __builtin_ia32_wbinvd): New builtins.
4883 (SPECIAL_ARGS2): New.
4884 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4885 (SPECIAL_ARGS2): New.
4886 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4887 (ix86_valid_target_attribute_inner_p): Ditto.
4888 (ix86_init_mmx_sse_builtins): Add special_args2.
4889 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4890 TARGET_WBNOINVD_P): New.
4891 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4892 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4893 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4894 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4895 * config/i386/pconfigintrin.h: New file.
4896 * config/i386/wbnoinvdintrin.h: Ditto.
4897 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4898 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
4900 2018-03-05 Richard Biener <rguenther@suse.de>
4902 PR tree-optimization/84670
4903 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4905 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4906 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4907 (compute_antic_aux): Only assert the number of values in ANTIC_IN
4908 doesn't grow if all successors (recursively) were visited at least
4911 2018-03-05 Richard Biener <rguenther@suse.de>
4913 PR tree-optimization/84650
4914 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
4915 if executed in the loop pipeline.
4917 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
4919 * doc/configfiles.texi (Configuration Files): Move info about
4920 conditionalizing $target-protos.h to...
4921 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
4922 differs from $target-protos.h.
4924 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
4925 Chung-Ju Wu <jasonwucj@gmail.com>
4927 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
4928 * config/nds32/nds32-multiple.md (setmemsi): Define.
4929 * config/nds32/nds32-memory-manipulation.c
4930 (nds32_gen_dup_4_byte_to_word_value): New.
4931 (emit_setmem_word_loop): New.
4932 (emit_setmem_byte_loop): New.
4933 (nds32_expand_setmem_loop): New.
4934 (nds32_expand_setmem_loop_v3m): New.
4935 (nds32_expand_setmem_unroll): New.
4936 (nds32_expand_setmem): New.
4938 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4939 Chung-Ju Wu <jasonwucj@gmail.com>
4941 * config/nds32/nds32-memory-manipulation.c
4942 (nds32_emit_load_store): New.
4943 (nds32_emit_post_inc_load_store): New.
4944 (nds32_emit_mem_move): New.
4945 (nds32_emit_mem_move_block): New.
4946 (nds32_expand_movmemsi_loop_unknown_size): New.
4947 (nds32_expand_movmemsi_loop_known_size): New.
4948 (nds32_expand_movmemsi_loop): New.
4949 (nds32_expand_movmemsi_unroll): New.
4950 (nds32_expand_movmemqi): Rename ...
4951 (nds32_expand_movmemsi): ... to this.
4952 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
4953 (movmemsi): ... to this.
4954 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
4955 (nds32_expand_movmemsi): ... to this.
4957 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4958 Monk Chiang <sh.chiang04@gmail.com>
4959 Chung-Ju Wu <jasonwucj@gmail.com>
4961 * config/nds32/nds32-protos.h
4962 (nds32_expand_load_multiple): New arguments.
4963 (nds32_expand_store_multiple): Ditto.
4964 (nds32_valid_multiple_load_store): Rename ...
4965 (nds32_valid_multiple_load_store_p): ... to this.
4966 * config/nds32/nds32-memory-manipulation.c
4967 (nds32_expand_load_multiple): Refine implementation.
4968 (nds32_expand_store_multiple): Ditto.
4969 * config/nds32/nds32-multiple.md
4970 (load_multiple): Update nds32_expand_load_multiple interface.
4971 (store_multiple): Update nds32_expand_store_multiple interface.
4972 * config/nds32/nds32-predicates.c
4973 (nds32_valid_multiple_load_store): Rename ...
4974 (nds32_valid_multiple_load_store_p): ... to this and refine
4976 * config/nds32/predicates.md
4977 (nds32_load_multiple_and_update_address_operation): New predicate.
4978 (nds32_store_multiple_and_update_address_operation): New predicate.
4980 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
4981 Chung-Ju Wu <jasonwucj@gmail.com>
4983 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
4984 (combo): New attribute.
4985 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
4987 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
4989 * config/nds32/nds32.opt: Change -mcmodel= default value.
4991 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
4992 Monk Chiang <sh.chiang04@gmail.com>
4993 Chung-Ju Wu <jasonwucj@gmail.com>
4995 * config/nds32/constants.md (unspec_element): New enum.
4996 * config/nds32/constraints.md (Umw): New constraint.
4997 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
4998 * config/nds32/nds32-intrinsic.md: Likewise.
4999 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5000 (nds32_valid_smw_lwm_base_p): New.
5001 (nds32_output_smw_single_word): New.
5002 (nds32_output_lmw_single_word): New.
5003 (nds32_expand_unaligned_load): New.
5004 (nds32_expand_unaligned_store): New.
5005 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5006 (nds32_output_smw_single_word): Declare.
5007 (nds32_output_lmw_single_word): Declare.
5008 (nds32_expand_unaligned_load): Declare.
5009 (nds32_expand_unaligned_store): Declare.
5010 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5011 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5012 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5013 NDS32_BUILTIN_UASTORE_DW.
5014 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5017 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5018 Kito Cheng <kito.cheng@gmail.com>
5019 Chung-Ju Wu <jasonwucj@gmail.com>
5021 * config/nds32/nds32-intrinsic.c
5022 (nds32_expand_builtin_null_ftype_reg): Delete.
5023 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5024 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5025 (nds32_read_argument): New.
5026 (nds32_legitimize_target): Ditto.
5027 (nds32_legitimize_argument): Ditto.
5028 (nds32_check_constant_argument): Ditto.
5029 (nds32_expand_unop_builtin): Ditto.
5030 (nds32_expand_unopimm_builtin): Ditto.
5031 (nds32_expand_binop_builtin): Ditto.
5032 (nds32_builtin_decl_impl): Ditto.
5033 (builtin_description): Ditto.
5034 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5035 (nds32_init_builtins_impl): Ditto.
5036 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5037 (nds32_builtin_decl): New.
5038 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5039 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5041 2018-03-02 Jeff Law <law@redhat.com>
5043 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5044 (redundant_insn, fill_simple_delay_slots): Likewise.
5045 (fill_slots_from_thread): Likewise.
5046 * resource.c (mark_referenced_resources): Likewise.
5047 (mark_set_resources, find_dead_or_set_registers): Likewise.
5049 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5051 * substring-locations.h (format_warning_va): Formatting fix for
5053 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5055 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5056 * substring-locations.c: Include intl.h.
5057 (format_warning_va): Turned into small wrapper around
5058 format_warning_n_va, renamed to ...
5059 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5060 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5062 (format_warning_at_substring_n): New function.
5063 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5064 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5065 format_warning_at_substring with just a shorter name instead of
5066 const function pointer.
5067 (fmtwarn_n): New function.
5068 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5069 appropriate, get rid of all the fmtstr temporaries, move conditionals
5070 with G_() wrapped string literals directly into fmtwarn arguments,
5071 cast dir.len to (int), formatting fixes.
5073 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5075 * doc/invoke.texi: Remove "Cilk Plus" references.
5077 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5078 Richard Biener <rguenther@suse.de>
5081 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5082 for error or warning attributes if CALL_FROM_THUNK_P is set.
5085 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5088 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5089 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5092 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5093 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5095 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5096 instead of -1U in last predictors element's probability member.
5098 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5101 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5102 arguments if they are comparable.
5104 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5106 PR tree-optimization/84634
5107 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5108 masks and masked_loop_p with a single loop_masks, making sure it's
5109 null for bb vectorization.
5111 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5113 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5114 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5115 loop->force_vectorize to check whether there is no alias.
5117 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5120 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5122 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5124 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5125 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5126 instead of a loop around prev_real_insn.
5127 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5131 * config/i386/i386.c (ix86_print_operand): Use conditional
5132 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5135 2018-03-02 Richard Biener <rguenther@suse.de>
5137 PR tree-optimization/84427
5138 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5139 (bitmap_set_subtract_values): Rewrite to handle multiple
5142 (prune_clobbered_mems): Likewise.
5143 (phi_translate): Take edge instead of pred/phiblock.
5144 (phi_translate_1): Likewise.
5145 (phi_translate_set): Likewise. Insert all translated
5146 exprs for a value into the set, keeping possibly multiple
5147 expressions per value.
5148 (compute_antic_aux): Adjust for phi_translate changes.
5149 When intersecting union the expressions and prune those
5150 not in the final value set, keeping possibly multiple
5151 expressions per value. Do not use value-insertion
5152 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5153 all expressions. Add verification that the value-sets
5154 only shrink during iteration.
5155 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5156 (do_pre_regular_insertion): Likewise.
5157 (do_pre_partial_partial_insertion): Likewise.
5159 2018-03-02 Richard Biener <rguenther@suse.de>
5162 * config/darwin.c (saved_debug_info_level): New static global.
5163 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5164 (darwin_asm_lto_end): Restore debug info generation settings.
5166 2018-03-01 Martin Liska <mliska@suse.cz>
5169 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5172 2018-03-01 Richard Biener <rguenther@suse.de>
5175 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5176 type annotation with LTO when debug was disabled at compile-time.
5178 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5180 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5182 (mips_final_postscan_insn): Likewise.
5184 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5186 PR rtl-optimization/84528
5187 * alias.c (init_alias_target): Add commentary.
5188 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5189 a unique base value if the frame pointer is not eliminated
5190 to the stack pointer.
5192 2018-03-01 Tom de Vries <tom@codesourcery.com>
5194 PR rtl-optimization/83327
5195 * lra-int.h (hard_regs_spilled_into): Declare.
5196 * lra.c (hard_regs_spilled_into): Define.
5197 (init_reg_info): Init hard_regs_spilled_into.
5198 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5199 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5200 (process_bb_lives): Handle hard_regs_spilled_into.
5201 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5202 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5204 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5206 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5207 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5208 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5209 * config/rs6000/aix72.h: New file.
5211 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5213 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5214 instead of warning_at with conditional singular and plural messages
5218 * stor-layout.c (update_alignment_for_field): For
5219 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5220 && !DECL_PACKED (field), do the alignment update, just use
5221 only desired_align instead of MAX (type_align, desired_align)
5223 (place_field): Don't do known_align < desired_align handling
5224 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5225 is non-NULL, instead do it after rli->prev_field handling and
5226 only if not within a bitfield word. For DECL_PACKED (field)
5227 use type_align of BITS_PER_UNIT.
5229 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5231 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5232 superfluous parentheses and trailing spaces.
5234 2018-02-28 Richard Biener <rguenther@suse.de>
5236 PR tree-optimization/84584
5237 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5238 SCoPs with fake exit edge.
5240 2018-02-28 Martin Liska <mliska@suse.cz>
5243 * timevar.c (timer::print): Fix format to properly print 100%
5246 2018-02-28 Richard Biener <rguenther@suse.de>
5249 * genmatch.c (capture_info::walk_match): Do not mark
5250 captured expressions without operands as expr_p given
5251 they act more like predicates and should be subject to
5252 "lost tail" side-effect preserving.
5254 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5256 PR rtl-optimization/81611
5257 * auto-inc-dec.c (attempt_change): Move dead note from
5258 mem_insn if it's the next use of regno
5259 (find_address): Take address use of reg holding
5260 non-incremented value. Add parm to limit search to the named
5262 (merge_in_block): Attempt to use a mem insn that is the next
5263 use of the original regno.
5265 2018-02-27 Martin Sebor <msebor@redhat.com>
5268 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5269 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5271 2018-02-27 Martin Sebor <msebor@redhat.com>
5273 PR translation/84207
5274 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5275 n argument to unsigned HOST_WIDE_INT.
5276 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5277 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5278 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5279 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5281 2018-02-27 Richard Biener <rguenther@suse.de>
5283 PR tree-optimization/84512
5284 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5285 Do not use the estimate returned from record_stmt_cost for
5286 the scalar iteration cost but sum properly using add_stmt_cost.
5288 2018-02-27 Richard Biener <rguenther@suse.de>
5290 PR tree-optimization/84466
5291 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5292 Adjust last change to less strictly validate use operands.
5294 2018-02-27 Martin Liska <mliska@suse.cz>
5296 PR gcov-profile/84548
5297 * gcov.c (process_file): Allow partial overlap and consider it
5298 also as group functions.
5299 (output_lines): Properly calculate range of lines for a group.
5301 2018-02-27 Martin Liska <mliska@suse.cz>
5303 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5304 'ggc' suffixes. Change first column width.
5305 (timer::print): Fix formatting of the column.
5307 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5309 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5310 preserve inline entry blocks for the sake of debug inline
5311 entry point markers alone.
5312 (remove_unused_locals): Suggest in comments a better place to
5313 force the preservation of inline entry blocks that are
5314 otherwise unused, but do not preserve them.
5316 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5318 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5320 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5323 * config/i386/constraints.md (Bs): Replace
5324 ix86_indirect_branch_register with
5325 TARGET_INDIRECT_BRANCH_REGISTER.
5327 * config/i386/i386.md (indirect_jump): Likewise.
5328 (tablejump): Likewise.
5329 (*sibcall_memory): Likewise.
5330 (*sibcall_value_memory): Likewise.
5331 Peepholes of indirect call and jump via memory: Likewise.
5332 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5333 (*sibcall_value_GOT_32): Likewise.
5334 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5335 (GOT_memory_operand): Likewise.
5336 (call_insn_operand): Likewise.
5337 (sibcall_insn_operand): Likewise.
5338 (GOT32_symbol_operand): Likewise.
5339 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5341 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5343 PR rtl-optimization/83496
5344 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5345 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5346 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5347 redundant insn, if any.
5348 (relax_delay_slots): Likewise.
5349 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5351 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5353 PR tree-optimization/83965
5354 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5355 that grouped statements are part of a reduction chain. Return
5356 true if the statement is not marked as a reduction itself but
5358 (vect_recog_dot_prod_pattern): Don't check whether the statement
5359 is part of a group here.
5360 (vect_recog_sad_pattern): Likewise.
5361 (vect_recog_widen_sum_pattern): Likewise.
5363 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5366 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5368 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5371 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5373 (ix86_output_indirect_function_return): New prototype.
5374 (ix86_split_simple_return_pop_internal): Likewise.
5375 * config/i386/i386.c (indirect_return_via_cx): New.
5376 (indirect_return_via_cx_bnd): Likewise.
5377 (indirect_thunk_name): Handle return va CX_REG.
5378 (output_indirect_thunk_function): Create alias for
5379 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5380 (ix86_output_indirect_jmp): Remove the bool argument.
5381 (ix86_output_indirect_function_return): New function.
5382 (ix86_split_simple_return_pop_internal): Likewise.
5383 * config/i386/i386.md (*indirect_jump): Don't pass false
5384 to ix86_output_indirect_jmp.
5385 (*tablejump_1): Likewise.
5386 (simple_return_pop_internal): Change it to define_insn_and_split.
5387 Call ix86_split_simple_return_pop_internal to split it for
5389 (simple_return_indirect_internal): Call
5390 ix86_output_indirect_function_return instead of
5391 ix86_output_indirect_jmp.
5393 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5396 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5397 memset and value initialization afterwards.
5399 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5401 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5403 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5406 * common/config/aarch64/aarch64-common.c
5407 (aarch_option_optimization_table[]): Switch
5408 off fomit-frame-pointer
5410 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5411 Chung-Ju Wu <jasonwucj@gmail.com>
5413 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5415 (store_multiple): Ditto.
5417 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5419 * config.gcc: Add --with-cpu support for nds32 target.
5420 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5421 * config/nds32/nds32.opt: Add -mcpu= option.
5423 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5425 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5426 isel=yes): Warn for these deprecated options.
5428 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5430 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5431 ISA_2_5_MASKS_EMBEDDED.
5433 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5435 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5436 p->max as pointers rather than using iterative_hash_expr.
5438 2018-02-23 Carl Love <cel@us.ibm.com>
5440 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5441 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5443 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5444 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5445 P8V_BUILTIN_VEC_VUNSIGNED2.
5447 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5450 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5451 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5453 * lra-constraints.c (curr_insn_transform): Set up
5454 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5456 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5457 * lra-eliminations.c (spill_pseudos): Ditto.
5458 (process_insn_for_elimination): Ditto.
5459 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5460 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5461 LRA_NON_CLOBBERED_ALT.
5463 2018-02-22 Martin Sebor <msebor@redhat.com>
5465 PR tree-optimization/84480
5466 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5467 to maybe_diag_stxncpy_trunc. Call it.
5468 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5469 from gimple_fold_builtin_strcpy. Print inlining stack.
5470 (handle_builtin_stxncpy): Print inlining stack.
5471 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5473 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5476 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5477 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5478 and -fcheck-pointer-bounds are used together.
5479 (indirect_thunk_prefix): New enum.
5480 (indirect_thunk_need_prefix): New function.
5481 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5482 "_nt" instead of "_bnd" for NOTRACK prefix.
5483 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5484 (output_indirect_thunk_function): Likewise.
5486 (ix86_code_end): Update output_indirect_thunk_function calls.
5487 (ix86_output_indirect_branch_via_reg): Replace
5488 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5489 (ix86_output_indirect_branch_via_push): Likewise.
5490 (ix86_output_function_return): Likewise.
5491 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5492 incompatible with -fcf-protection=branch and
5493 -fcheck-pointer-bounds.
5495 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5498 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5499 Change gcc_assert call to output_operand_lossage.
5501 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5503 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5505 2018-02-22 DJ Delorie <dj@redhat.com>
5506 Sebastian Perta <sebastian.perta@renesas.com>
5507 Oleg Endo <olegendo@gcc.gnu.org>
5509 * config/rx/rx.c (rx_rtx_costs): New function.
5510 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5512 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5514 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5516 2018-02-22 Martin Liska <mliska@suse.cz>
5519 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5520 Add "native" as a possible value.
5522 2018-02-22 Martin Liska <mliska@suse.cz>
5525 * config/i386/i386.c (ix86_option_override_internal):
5526 Add "native" as a possible value for -march and -mtune.
5528 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5531 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5532 to all type variants.
5534 PR tree-optimization/84503
5535 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5536 width as info->bitpos + info->bitsize - start.
5537 (merged_store_group::merge_overlapping): Simplify width computation.
5538 (check_no_overlap): New function.
5539 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5540 start + width and last_order of the group, fail if check_no_overlap
5542 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5543 to group if check_no_overlap fails.
5545 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5547 * config/rs6000/altivec.md: Delete contraint arguments to
5548 define_expand, define_split, and define_peephole2, and in
5549 define_insn_and_split if always unused.
5550 * config/rs6000/darwin.md: Ditto.
5551 * config/rs6000/dfp.md: Ditto.
5552 * config/rs6000/rs6000.md: Ditto.
5553 * config/rs6000/sync.md: Ditto.
5554 * config/rs6000/vector.md: Ditto.
5555 * config/rs6000/vsx.md: Ditto.
5557 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5559 * config/rs6000/altivec.md: Write output control strings as braced
5560 blocks instead of double-quoted strings.
5561 * config/rs6000/darwin.md: Ditto.
5562 * config/rs6000/rs6000.md: Ditto.
5563 * config/rs6000/vector.md: Ditto.
5564 * config/rs6000/vsx.md: Ditto.
5566 2018-02-21 Jason Merrill <jason@redhat.com>
5568 PR c++/84314 - ICE with templates and fastcall attribute.
5569 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5571 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5573 * ipa-cp.c (determine_versionability): Fix comment typos.
5575 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5578 * ipa-cp.c (determine_versionability): Do not version functions caling
5581 2018-02-21 Martin Liska <mliska@suse.cz>
5584 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5585 Add "native" as a possible value.
5586 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5587 the macro when native cpu detection is available.
5589 2018-02-21 Martin Liska <mliska@suse.cz>
5592 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5593 Add "native" as a possible value.
5594 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5595 when native cpu detection is available.
5597 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5598 Martin Sebor <msebor@redhat.com>
5600 PR tree-optimization/84478
5601 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5603 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5604 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5605 support which is conservatively correct, for 2 only stay conservative
5606 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5607 argument to the 2 argument get_range_strlen, adjust 6 arg
5608 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5610 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5611 (gimple_fold_builtin_strlen): Pass true as last argument to
5614 2018-02-20 Martin Sebor <msebor@redhat.com>
5617 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5618 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5619 (builtin_memref::builtin_memref): Factor out parts into
5620 set_base_and_offset and call it.
5622 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5625 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5626 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5627 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5628 search at the associated MODE_INT.
5630 2018-02-20 Jeff Law <law@redhat.com>
5633 PR tree-optimization/81592
5635 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5636 for range data rather than using global data.
5637 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5638 range data rather than using global data.
5639 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5640 pass it to children as needed.
5641 (struct directive::fmtresult): Similarly.
5642 (struct directive::set_width): Similarly.
5643 (struct directive::set_precision): Similarly.
5644 (format_integer, format_directive, parse_directive): Similarly.
5645 (format_none): Accept unnamed vr_values parameter.
5646 (format_percent, format_floating, format_character): Similarly.
5647 (format_string, format_plain): Similarly.
5648 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5649 the EVRP range analyzer for range data rather than using global data.
5650 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5651 gimple-ssa-evrp-analyze.h
5652 (class sprintf_dom_walker): Add after_dom_children member function.
5653 Add evrp_range_analyzer member.
5654 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5655 range analyzer as needed.
5656 (sprintf_dom_walker::after_dom_children): New member function.
5657 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5659 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5660 (evrp_range_analyzer::pop_to_marker): Likewise.
5662 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5664 PR tree-optimization/84419
5665 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5666 with the required type if its current type is compatible but
5669 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5672 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5673 after vectorization.
5675 2018-02-20 Martin Liska <mliska@suse.cz>
5678 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5679 possible values if we don't have a hint.
5681 2018-02-20 Martin Liska <mliska@suse.cz>
5685 * final.c (shorten_branches): Build align_tab array with one
5687 * opts.c (finish_options): Add alignment option limit check.
5688 (MAX_CODE_ALIGN): Likewise.
5689 (MAX_CODE_ALIGN_VALUE): Likewise.
5690 * doc/invoke.texi: Document maximum allowed option value for
5691 all -falign-* options.
5693 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5696 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5697 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5698 * var-tracking.c (emit_note_insn_var_location): Remove all references
5699 to NOTE_INSN_CALL_ARG_LOCATION.
5700 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5701 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5702 Use copy_rtx_if_shared.
5703 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5704 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5705 (dwarf2out_var_location): Remove handling of
5706 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5708 * final.c (final_scan_insn): Remove all references to
5709 NOTE_INSN_CALL_ARG_LOCATION.
5710 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5711 before dumping final insns.
5712 * except.c (emit_note_eh_region_end): Remove all references to
5713 NOTE_INSN_CALL_ARG_LOCATION.
5714 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5715 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5716 * config/arc/arc.c (hwloop_optimize): Likewise.
5717 * config/arm/arm.c (create_fix_barrier): Likewise.
5718 * config/s390/s390.c (s390_chunkify_start): Likewise.
5719 * config/sh/sh.c (find_barrier): Likewise.
5720 * config/i386/i386.c (rest_of_insert_endbranch,
5721 ix86_seh_fixup_eh_fallthru): Likewise.
5722 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5723 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5724 * config/frv/frv.c (frv_function_prologue): Likewise.
5725 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5727 (note_outside_basic_block_p): Remove all references to
5728 NOTE_INSN_CALL_ARG_LOCATION.
5729 * gengtype.c (adjust_field_rtx_def): Likewise.
5730 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5732 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5733 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5736 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5739 PR tree-optimization/84452
5740 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5741 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5744 2018-02-19 Martin Liska <mliska@suse.cz>
5747 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5749 2018-02-19 Martin Liska <mliska@suse.cz>
5750 Richard Sandiford <richard.sandiford@linaro.org>
5752 PR tree-optimization/82491
5753 * gimple-fold.c (get_base_constructor): Make earlier bail out
5756 2018-02-19 Carl Love <cel@us.ibm.com>
5758 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5759 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5761 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5762 P8V_BUILTIN_VEC_NEG.
5764 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5766 * config/rl78/rl78.md (movdf): New define expand.
5768 2018-02-19 Martin Liska <mliska@suse.cz>
5771 * doc/invoke.texi: Fix typo.
5772 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5774 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5776 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5777 handle rs6000_single_float and rs6000_double_float specially for
5780 2018-02-16 Jeff Law <law@redhat.com>
5782 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5785 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5788 * ipa-inline.c (inline_small_functions): Fix a typo.
5790 2018-02-16 Nathan Sidwell <nathan@acm.org>
5792 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5794 2018-02-16 Carl Love <cel@us.ibm.com>
5796 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5797 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5798 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5799 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5800 expansion to P8V_BUILTIN_VEC_FLOAT2.
5802 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5804 PR rtl-optimization/70023
5805 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5806 src_regno into account.
5808 2018-02-16 Carl Love <cel@us.ibm.com>
5810 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5811 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5812 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5813 * config/rs6000/rs6000.c: Remove case statements for
5814 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5815 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5816 and P9V_BUILTIN_VEC_VINSERT4B.
5817 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5818 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5819 * config/rs6000/vsx.md:
5820 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5823 2018-02-16 Carl Love <cel@us.ibm.com>
5825 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5827 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5829 * config/rs6000/rs6000-c.c: Add the definitions for
5830 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5831 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5832 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5833 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5834 definition for insert4b and define insn *insert3b_internal.
5835 * doc/extend.texi: Add documentation for vec_extract4b.
5837 2018-02-16 Nathan Sidwell <nathan@acm.org>
5839 * doc/extend.texi (Backwards Compatibility): Mention friend
5840 injection. Note for-scope is deprecated.
5841 * doc/invoke.texi (-ffriend-injection): Deprecate.
5843 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5845 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5846 that moved to I2, also allow destinations that are a paradoxical
5847 subreg (instead of a normal reg).
5849 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5852 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5855 2018-02-16 Richard Biener <rguenther@suse.de>
5857 PR tree-optimization/84037
5858 PR tree-optimization/84016
5860 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5861 Adjust vec_construct for the fact we need additional higher latency
5862 128bit inserts for AVX256 and AVX512 vector builds.
5863 (ix86_add_stmt_cost): Scale vector construction cost for
5866 2018-02-16 Richard Biener <rguenther@suse.de>
5868 PR tree-optimization/84417
5869 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5870 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5871 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5873 2018-02-16 Martin Liska <mliska@suse.cz>
5876 * internal-fn.def (ASAN_CHECK): Set proper flags.
5877 (ASAN_MARK): Likewise.
5879 2018-02-16 Julia Koval <julia.koval@intel.com>
5881 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5882 from PTA_CANNONLAKE.
5884 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5887 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5888 Use ++iter rather than iter++ for std::list iterators.
5889 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5890 defer deleting them until all nodes in the forest are processed. Do
5891 free even leaf nodes. Change to_process into auto_vec.
5894 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5895 * vec.h (vec_default_construct): Use memset instead of placement new
5896 if BROKEN_VALUE_INITIALIZATION is defined.
5897 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5898 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
5901 PR rtl-optimization/83723
5902 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
5903 * lra.c (lra_substitute_pseudo): Likewise. If true, use
5904 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
5906 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
5908 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
5910 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
5912 PR rtl-optimization/81443
5913 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
5914 from inner REGs to paradoxical SUBREGs.
5916 2018-02-16 Richard Biener <rguenther@suse.de>
5918 PR tree-optimization/84399
5919 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5920 For operands we can analyze at their definition make sure we can
5921 analyze them at each use as well.
5923 2018-02-16 Richard Biener <rguenther@suse.de>
5925 PR tree-optimization/84190
5926 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
5927 volatile accesses if the decl isn't volatile.
5929 2018-02-15 Jason Merrill <jason@redhat.com>
5931 PR c++/84314 - ICE with templates and fastcall attribute.
5932 * attribs.c (build_type_attribute_qual_variant): Don't clobber
5933 TYPE_CANONICAL on an existing type.
5935 2018-02-15 Jakub Jelinek <jakub@redhat.com>
5937 PR tree-optimization/84383
5938 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
5939 dstoff nor call operand_equal_p if dstbase is NULL.
5941 PR tree-optimization/84334
5942 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
5943 also a CONSTANT_CLASS_P, punt.
5945 2018-02-14 Jim Wilson <jimw@sifive.com>
5947 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
5948 first SMALL_OPERAND check. New local min_second_step. Move assert
5949 to where locals are set. Add TARGET_RVC support.
5950 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
5952 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
5954 * doc/invoke.texi: Correct -Wformat-overflow code sample.
5956 2018-02-14 Martin Sebor <msebor@redhat.com>
5958 PR tree-optimization/83698
5959 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
5960 arrays constrain the offset range to their bounds.
5961 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
5962 (builtin_access::overlap): Avoid setting the size of overlap if it's
5964 (maybe_diag_overlap): Also consider arrays when deciding what values
5965 of offsets to include in diagnostics.
5967 2018-02-14 Martin Sebor <msebor@redhat.com>
5970 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
5971 that correspond to the kind of a declaration.
5973 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
5976 * config/pa/pa.md: Load address of PIC label using the linkage table
5977 if the label is nonlocal.
5979 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
5981 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
5982 warning message if user requests -maltivec=be.
5983 * doc/invoke.texi: Document deprecation of -maltivec=be.
5985 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
5988 * config/rs6000/rs6000-c.c: Update definitions for
5989 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
5990 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
5992 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5995 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
5996 add _get_ssp intrinsics. Remove argument from
5997 __builtin_ia32_rdssp[d|q].
5998 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
5999 * config/i386/i386-builtin.def: Remove argument from
6000 __builtin_ia32_rdssp[d|q].
6001 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6002 ix86_expand_special_args_builtin for _rdssp[d|q].
6003 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6004 Clear register before usage.
6005 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6006 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6008 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6010 PR tree-optimization/84357
6011 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6012 operand 1 of an ARRAY_REF too.
6014 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6017 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6018 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6020 (set_of_reg): New struct.
6021 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6022 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6023 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6025 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6026 Split into bitclr, bitset, bitinvert patterns if appropriate.
6027 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6028 use rx_fuse_in_memory_bitop.
6029 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6030 to named insn, correct maximum insn length.
6032 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6035 * machmode.def: Define a complex mode for PARTIAL_INT.
6036 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6038 * doc/rtl.texi: Document CSPImode.
6039 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6041 (msp430_hard_regno_nregs_with_padding): Likewise.
6043 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6046 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6048 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6050 PR rtl-optimization/84169
6051 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6052 we generated a parallel as new i3 and we split that to new i2 and i3
6053 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6054 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6055 those to i2, not i1. Partially rewrite this scan code.
6057 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6060 * stor-layout.c (place_field): For variable length fields, adjust
6061 offset_align afterwards not just based on the field's alignment,
6062 but also on the size.
6065 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6066 of exps and logs in the use_exp2 case.
6068 2018-02-13 Jeff Law <law@redhat.com>
6070 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6073 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6076 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6080 * common.opt (gas-loc-support, gas-locview-support): New.
6081 (ginline-points, ginternal-reset-location-views): New.
6082 * doc/invoke.texi: Document them. Use @itemx where intended.
6083 (gvariable-location-views): Adjust.
6084 * target.def (reset_location_view): New.
6085 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6086 (TARGET_RESET_LOCATION_VIEW): New.
6087 * doc/tm.texi: Rebuilt.
6088 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6089 (dwarf2out_default_as_locview_support): New.
6090 (output_asm_line_debug_info): Use option variables.
6091 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6092 (output_loc_list): Likewise.
6093 (add_high_low_attributes): Check option variables.
6094 Don't output entry view attribute in strict mode.
6095 (gen_inlined_subroutine_die): Check option variables.
6096 (dwarf2out_inline_entry): Likewise.
6097 (init_sections_and_labels): Likewise.
6098 (dwarf2out_early_finish): Likewise.
6099 (maybe_reset_location_view): New, from...
6100 (dwarf2out_var_location): ... here. Call it.
6101 * debug.h (dwarf2out_default_as_loc_support): Declare.
6102 (dwarf2out_default_as_locview_support): Declare.
6103 * hooks.c (hook_int_rtx_insn_0): New.
6104 * hooks.h (hook_int_rtx_insn_0): Declare.
6105 * toplev.c (process_options): Take -gas-loc-support and
6106 -gas-locview-support from dwarf2out. Enable
6107 -gvariable-location-views by default only with locview
6108 assembler support. Enable -ginternal-reset-location-views by
6109 default only if the target defines the corresponding hook.
6110 Enable -ginline-points by default if location views are
6111 enabled; force it disabled if statement frontiers are
6113 * tree-inline.c (expand_call_inline): Check option variables.
6114 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6116 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6118 PR tree-optimization/84321
6119 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6120 handling. Also check whether the anti-range contains any values
6121 that satisfy the mask; switch to a VR_RANGE if not.
6123 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6126 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6128 2018-02-13 Martin Jambor <mjambor@suse.cz>
6131 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6132 of call statements, also set location of a load to a temporary.
6134 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6136 * config/rl78/rl78.c (add_vector_labels): New function.
6137 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6138 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6139 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6140 which checks that no arguments are passed.
6141 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6142 * doc/extend.texi: Documentation for the new attribute.
6144 2018-02-13 Andreas Schwab <schwab@suse.de>
6146 * config/riscv/linux.h (CPP_SPEC): Define.
6148 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6151 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6152 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6153 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6154 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6155 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6156 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6157 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6158 temporarily for AES and PCLMUL builtins.
6160 PR tree-optimization/84339
6161 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6162 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6166 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6167 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6168 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6170 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6172 * omp-simd-clone.h: New file.
6173 * omp-simd-clone.c: Include omp-simd-clone.h.
6174 (expand_simd_clones): No longer static.
6175 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6176 cgraph.h and omp-simd-clone.h.
6177 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6178 (vect_recog_widen_shift_pattern): Formatting fix.
6179 (vect_pattern_recog_1): Don't check optab for calls.
6182 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6183 operands[2] into a REG before using gen_lowpart on it.
6185 2018-02-12 Jeff Law <law@redhat.com>
6188 * config/sh/sh.c (find_barrier): Consider a sibling call
6191 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6192 successfully back substituting a reg.
6194 2018-02-12 Richard Biener <rguenther@suse.de>
6196 PR tree-optimization/84037
6197 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6198 parameter, move visited init to caller.
6199 (vect_slp_analyze_operations): Separate cost from validity
6200 check, initialize visited once for all instances.
6201 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6203 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6205 (vect_model_promotion_demotion_cost): Likewise.
6206 (vectorizable_bswap): Guard cost modeling with !slp_node
6207 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6209 (vectorizable_call): Likewise.
6210 (vectorizable_conversion): Likewise.
6211 (vectorizable_assignment): Likewise.
6212 (vectorizable_shift): Likewise.
6213 (vectorizable_operation): Likewise.
6214 (vectorizable_store): Likewise.
6215 (vectorizable_load): Likewise.
6216 (vectorizable_condition): Likewise.
6217 (vectorizable_comparison): Likewise.
6219 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6222 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6223 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6224 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6226 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6229 * params.def (inline-min-speedup): Increase from 8 to 15.
6230 (max-inline-insns-auto): Decrease from 40 to 30.
6231 * ipa-split.c (consider_split): Add some buffer for function to
6232 be considered inlining candidate.
6233 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6236 2018-02-12 Richard Biener <rguenther@suse.de>
6238 PR tree-optimization/84037
6239 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6240 matched stmts if we cannot swap the non-matched ones.
6242 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6244 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6245 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6246 _mm_maskz_scalef_round_ss): New intrinsics.
6247 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6248 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6249 __builtin_ia32_scalefss_round): Remove.
6250 (__builtin_ia32_scalefsd_mask_round,
6251 __builtin_ia32_scalefss_mask_round): New intrinsics.
6252 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6253 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6254 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6255 "<round_constraint>")): Changed to ...
6256 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6257 "<round_scalar_constraint>")): ... this.
6258 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6259 %0, %1, %2<round_op3>}"): Changed to ...
6260 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6261 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6262 %2<round_scalar_mask_op3>}"): ... this.
6263 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6265 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6267 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6268 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6269 (_mm_maskz_sqrt_round_ss): New intrinsics.
6270 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6271 (__builtin_ia32_sqrtsd_mask_round)
6272 (__builtin_ia32_sqrtss_mask_round): New builtins.
6273 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6274 (__builtin_ia32_sqrtss_round): Remove.
6275 (__builtin_ia32_sqrtsd_mask_round)
6276 (__builtin_ia32_sqrtss_mask_round): New builtins.
6277 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6278 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6279 ((match_operand:VF_128 1 "vector_operand"
6280 "xBm,<round_constraint>")): Changed to ...
6281 ((match_operand:VF_128 1 "vector_operand"
6282 "xBm,<round_scalar_constraint>")): ... this.
6283 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6284 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6285 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6286 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6287 %<iptr>1<round_scalar_mask_op3>}): ... this.
6288 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6289 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6291 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6294 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6295 Cast vec_cmpeq result to correct type.
6296 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6297 Cast vec_cmpgt result to correct type.
6299 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6301 * final.c (final_scan_insn_1): Renamed from...
6302 (final_scan_insn): ... this. New wrapper, to recover
6303 seen from the outermost call in recursive ones.
6304 * config/sparc/sparc.c (output_return): Drop seen from call.
6305 (output_sibcall): Likewise.
6306 * config/visium/visium.c (output_branch): Likewise.
6308 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6310 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6313 2018-02-10 Alan Modra <amodra@gmail.com>
6316 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6317 Specify LR as an input.
6319 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6322 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6323 remove_member_access_dummy_vars): New functions.
6324 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6325 lower_omp_1, execute_lower_omp): Use them.
6327 PR rtl-optimization/84308
6328 * shrink-wrap.c (spread_components): Release todo vector.
6330 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6332 PR rtl-optimization/57193
6333 * ira-color.c (struct allocno_color_data): Add member
6334 conflict_allocno_hard_prefs.
6335 (update_conflict_allocno_hard_prefs): New.
6336 (bucket_allocno_compare_func): Add a preference based on
6337 conflict_allocno_hard_prefs.
6338 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6339 (color_allocnos): Remove a dead code. Initiate
6340 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6342 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6345 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6346 constraint from =wa to wa. Avoid a subreg on the output operand,
6347 instead use a pseudo and subreg it in a move.
6348 (p9_xxbrd_<mode>): Changed to ...
6349 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6350 (p9_xxbrd_v2df): New expander.
6351 (p9_xxbrw_<mode>): Changed to ...
6352 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6353 (p9_xxbrw_v4sf): New expander.
6355 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6357 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6359 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6362 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6363 multiply in 32-bit mode.
6364 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6365 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6368 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6370 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6371 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6372 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6373 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6375 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6378 * dwarf2out.c (is_trivial_indirect_ref): New function.
6379 (dwarf2out_late_global_decl): Do not generate a location
6380 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6381 and that are not defined in the current unit.
6383 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6385 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6386 instead of a libcall for UNORDERED.
6388 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6391 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6392 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6394 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6397 * config/s390/s390.c (s390_set_current_function): Invoke
6398 s390_indirect_branch_settings also if fndecl didn't change.
6400 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6402 * config/rs6000/rs6000.md (blockage): Set length to zero.
6404 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6406 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6408 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6411 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6412 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6416 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6417 PARALLEL incoming that failed vt_get_decl_and_offset check.
6420 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6421 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6423 (get_variable_section): For decls in named .bss* sections pass true as
6424 second argument to bss_initializer_p.
6426 2018-02-09 Marek Polacek <polacek@redhat.com>
6427 Jakub Jelinek <jakub@redhat.com>
6430 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6431 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6432 Sync some changes from cxx_fold_indirect_ref.
6434 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6436 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6438 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6439 (BLOCK_INLINE_ENTRY_LABEL): New.
6440 (dwarf2out_var_location): Disregard inline entry markers.
6441 (inline_entry_data): New struct.
6442 (inline_entry_data_hasher): New hashtable type.
6443 (inline_entry_data_hasher::hash): New.
6444 (inline_entry_data_hasher::equal): New.
6445 (inline_entry_data_table): New variable.
6446 (add_high_low_attributes): Add DW_AT_entry_pc and
6447 DW_AT_GNU_entry_view attributes if a pending entry is found
6448 in inline_entry_data_table. Add old entry_pc attribute only
6449 if debug nonbinding markers are disabled.
6450 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6451 markers are enabled.
6452 (block_within_block_p, dwarf2out_inline_entry): New.
6453 (dwarf2out_finish): Check that no entries remained in
6454 inline_entry_data_table.
6455 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6456 (final_scan_insn, notice_source_line): Likewise.
6457 (rest_of_clean_state): Skip inline entry markers.
6458 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6460 * gimple.c (gimple_build_debug_inline_entry): New.
6461 * gimple.h (enum gimple_debug_subcode): Add
6462 GIMPLE_DEBUG_INLINE_ENTRY.
6463 (gimple_build_debug_inline_entry): Declare.
6464 (gimple_debug_inline_entry_p): New.
6465 (gimple_debug_nonbind_marker_p): Adjust.
6466 * insn-notes.def (INLINE_ENTRY): New.
6467 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6468 inline entry marker notes.
6469 (print_insn): Likewise.
6470 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6471 (INSN_DEBUG_MARKER_KIND): Likewise.
6472 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6473 * tree-inline.c (expand_call_inline): Build and insert
6474 debug_inline_entry stmt.
6475 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6476 inline entry blocks early, if nonbind markers are enabled.
6477 (dump_scope_block): Dump fragment info.
6478 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6479 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6480 (gimple_build_debug_inline_entry): New.
6481 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6482 Enable/disable inline entry points too.
6483 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6484 (DEBUG_INSN): Describe inline entry markers.
6486 * common.opt (gvariable-location-views): New.
6487 (gvariable-location-views=incompat5): New.
6488 * config.in: Rebuilt.
6489 * configure: Rebuilt.
6490 * configure.ac: Test assembler for view support.
6491 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6492 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6493 * dwarf2out.c (var_loc_view): New typedef.
6494 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6495 (dwarf2out_locviews_in_attribute): New.
6496 (dwarf2out_locviews_in_loclist): New.
6497 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6498 (enum dw_line_info_opcode): Add LI_adv_address.
6499 (struct dw_line_info_table): Add view.
6500 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6501 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6502 (zero_view_p): New variable.
6503 (ZERO_VIEW_P): New macro.
6504 (output_asm_line_debug_info): New.
6505 (struct var_loc_node): Add view.
6506 (add_AT_view_list, AT_loc_list): New.
6507 (add_var_loc_to_decl): Add view param. Test it against last.
6508 (new_loc_list): Add view params. Record them.
6509 (AT_loc_list_ptr): Handle loc and view lists.
6510 (view_list_to_loc_list_val_node): New.
6511 (print_dw_val): Handle dw_val_class_view_list.
6512 (size_of_die): Likewise.
6513 (value_format): Likewise.
6514 (loc_list_has_views): New.
6515 (gen_llsym): Set vl_symbol too.
6516 (maybe_gen_llsym, skip_loc_list_entry): New.
6517 (dwarf2out_maybe_output_loclist_view_pair): New.
6518 (output_loc_list): Output view list or entries too.
6519 (output_view_list_offset): New.
6520 (output_die): Handle dw_val_class_view_list.
6521 (output_dwarf_version): New.
6522 (output_compilation_unit_header): Use it.
6523 (output_skeleton_debug_sections): Likewise.
6524 (output_rnglists, output_line_info): Likewise.
6525 (output_pubnames, output_aranges): Update version comments.
6526 (output_one_line_info_table): Output view numbers in asm comments.
6527 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6528 Call maybe_gen_llsym.
6529 (loc_list_from_tree_1): Adjust.
6530 (add_AT_location_description): Create view list attribute if
6531 needed, check it's absent otherwise.
6532 (convert_cfa_to_fb_loc_list): Adjust.
6533 (maybe_emit_file): Call output_asm_line_debug_info for test.
6534 (dwarf2out_var_location): Reset views as needed. Precompute
6535 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6536 attribute. Set view.
6537 (new_line_info_table): Reset next view.
6538 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6539 (dwarf2out_source_line): Likewise. Output view resets and labels to
6540 the assembler, or select appropriate line info opcodes.
6541 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6542 (optimize_string_length): Catch it. Adjust.
6543 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6544 dw_val_class_view_list, and remove it if no longer needed.
6545 (hash_loc_list): Hash view numbers.
6546 (loc_list_hasher::equal): Compare them.
6547 (optimize_location_lists): Check whether a view list symbol is
6548 needed, and whether the locview attribute is present, and
6549 whether they match. Remove the locview attribute if no longer
6551 (index_location_lists): Call skip_loc_list_entry for test.
6552 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6553 Use output_dwarf_version.
6554 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6555 (struct dw_val_node): Add val_view_list.
6556 * final.c (SEEN_NEXT_VIEW): New.
6557 (set_next_view_needed): New.
6558 (clear_next_view_needed): New.
6559 (maybe_output_next_view): New.
6560 (final_start_function): Rename to...
6561 (final_start_function_1): ... this. Take pointer to FIRST,
6562 add SEEN parameter. Emit param bindings in the initial view.
6563 (final_start_function): Reintroduce SEEN-less interface.
6564 (final): Rename to...
6565 (final_1): ... this. Take SEEN parameter. Output final pending
6566 next view at the end.
6567 (final): Reintroduce seen-less interface.
6568 (final_scan_insn): Output pending next view before switching
6569 sections or ending a block. Mark the next view as needed when
6570 outputting variable locations. Notify debug backend of section
6571 changes, and of location view changes.
6572 (rest_of_handle_final): Adjust.
6573 * toplev.c (process_options): Autodetect value for debug variable
6574 location views option. Warn on incompat5 without -gdwarf-5.
6575 * doc/invoke.texi (gvariable-location-views): New.
6576 (gvariable-location-views=incompat5): New.
6577 (gno-variable-location-views): New.
6579 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6581 PR tree-optimization/84136
6582 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6583 that the result of find_edge is non-NULL.
6585 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6588 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6589 storing integer register in SImode. Fix cost of 256 and 512
6590 byte aligned SSE register store.
6592 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6594 * config/i386/i386.c (ix86_multiplication_cost): Fix
6595 multiplication cost for TARGET_AVX512DQ.
6597 2018-02-08 Marek Polacek <polacek@redhat.com>
6599 PR tree-optimization/84238
6600 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6603 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6605 PR tree-optimization/84265
6606 * tree-vect-stmts.c (vectorizable_store): Don't treat
6607 VMAT_CONTIGUOUS accesses as grouped.
6608 (vectorizable_load): Likewise.
6610 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6612 PR tree-optimization/81635
6613 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6614 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6615 (test_round_for_mask): New functions.
6616 (wide_int_cc_tests): Call test_round_for_mask.
6617 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6618 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6619 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6620 range returned by get_range_info.
6622 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6625 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6626 * symtab.c: Include builtins.h
6627 (symtab_node::output_to_lto_symbol_table_p): Move here
6628 from lto-streamer-out.c:output_symbol_p.
6629 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6630 (output_symbol_p): Move all logic to symtab.c
6631 (produce_symtab): Update.
6633 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6635 * config/s390/s390-opts.h (enum indirect_branch): Define.
6636 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6637 (s390_indirect_branch_via_thunk)
6638 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6639 (enum s390_indirect_branch_type): Define.
6640 * config/s390/s390.c (struct s390_frame_layout, struct
6641 machine_function): Remove.
6642 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6643 (indirect_branch_table_label_no, indirect_branch_table_name):
6645 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6646 (enum s390_indirect_branch_option): Define.
6647 (s390_return_addr_from_memory): New function.
6648 (s390_handle_string_attribute): New function.
6649 (s390_attribute_table): Add new attribute handler.
6650 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6651 (s390_indirect_branch_via_thunk): New function.
6652 (s390_indirect_branch_via_inline_thunk): New function.
6653 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6654 sibling call optimization for non z10 compiles.
6655 (s390_emit_call): Force indirect branch target to be a single
6656 register. Add r1 clobber for non-z10 compiles.
6657 (s390_emit_epilogue): Emit return jump via return_use expander.
6658 (s390_reorg): Handle JUMP_INSNs as execute targets.
6659 (s390_option_override_internal): Perform validity checks for the
6660 new command line options.
6661 (s390_indirect_branch_attrvalue): New function.
6662 (s390_indirect_branch_settings): New function.
6663 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6664 (s390_output_indirect_thunk_function): New function.
6665 (s390_code_end): Implement target hook.
6666 (s390_case_values_threshold): Implement target hook.
6667 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6669 * config/s390/s390.h (struct s390_frame_layout)
6670 (struct machine_function): Move here from s390.c.
6671 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6672 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6673 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6674 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6675 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6676 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6677 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6678 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6679 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6680 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6681 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6682 (mnemonic attribute): Add values which aren't recognized
6684 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6685 pattern for branch conversion. Fix mnemonic attribute.
6686 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6687 indirect branch via thunk if requested.
6688 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6689 ("*indirect_jump"): Disable for branch conversion using out of
6691 ("indirect_jump_via_thunk<mode>_z10")
6692 ("indirect_jump_via_thunk<mode>")
6693 ("indirect_jump_via_inlinethunk<mode>_z10")
6694 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6695 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6696 ("casesi_jump_via_inlinethunk<mode>_z10")
6697 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6698 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6699 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6700 ("*indirect2_jump"): Disable for branch conversion.
6701 ("casesi_jump"): Turn into expander and expand patterns for branch
6703 ("return_use"): New expander.
6704 ("*return"): Emit return via thunk and rename it to ...
6705 ("*return<mode>"): ... this one.
6706 * config/s390/s390.opt: Add new options and and enum for the
6709 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6711 * lra-constraints.c (match_reload): Unconditionally use
6712 gen_lowpart_SUBREG, rather than selecting between that
6713 and equivalent gen_rtx_SUBREG code.
6715 2018-02-08 Richard Biener <rguenther@suse.de>
6717 PR tree-optimization/84233
6718 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6719 changed flag instead of boguously re-using phi_inserted.
6721 2018-02-08 Martin Jambor <mjambor@suse.cz>
6723 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6724 static local variables.
6726 2018-02-08 Richard Biener <rguenther@suse.de>
6728 PR tree-optimization/84278
6729 * tree-vect-stmts.c (vectorizable_store): When looking for
6730 smaller vector types to perform grouped strided loads/stores
6731 make sure the mode is supported by the target.
6732 (vectorizable_load): Likewise.
6734 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6736 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6737 Increase LDP/STP opportunities by adding adjacent callee-saves.
6739 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6741 PR rtl-optimization/84068
6742 PR rtl-optimization/83459
6743 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6745 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6747 PR tree-optimization/84224
6748 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6749 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6752 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6755 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6756 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6757 count, remove one USE.
6759 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6761 * doc/install.texi (Configuration): Document the
6762 --with-long-double-format={ibm,ieee} PowerPC configuration
6766 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6767 Convert from define_expand to be define_insn_and_split. Rework
6768 float/double/_Float128 conversions to QI/HI/SImode to work with
6769 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6770 conversions to QI/HImode types did a store and then a load to
6771 truncate the value. For conversions to VSX registers, don't split
6772 the insn, instead emit the code directly. Use the code iterator
6773 any_fix to combine signed and unsigned conversions.
6774 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6775 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6776 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6777 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6778 (fix_<mode>di2_hw): Likewise.
6779 (fixuns_<mode>di2_hw): Likewise.
6780 (fix_<mode>si2_hw): Likewise.
6781 (fixuns_<mode>si2_hw): Likewise.
6782 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6783 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6784 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6785 fix<uns>_trunc<SFDF:mode>si2_p8.
6786 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6788 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6789 (fix<uns>_<mode>_mem): Likewise.
6790 (fctiw<u>z_<mode>_mem): Likewise.
6791 (fix<uns>_<mode>_mem): Likewise.
6792 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6793 the register allocator from doing a direct move to the GPRs to do
6794 a store, and instead use the ISA 3.0 store byte/half-word from
6795 vector register instruction. For IEEE 128-bit floating point,
6796 also optimize stores of 32-bit ints.
6797 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6799 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6801 * genextract.c (push_pathstr_operand): New function to support
6803 (walk_rtx): Call push_pathstr_operand.
6804 (print_path): Support [a-zA-Z].
6806 2018-02-07 Richard Biener <rguenther@suse.de>
6808 PR tree-optimization/84037
6809 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6810 (cse_and_gimplify_to_preheader): Declare.
6811 (vect_get_place_in_interleaving_chain): Likewise.
6812 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6814 (_loop_vec_info::~_loop_vec_info): Delete it.
6815 (cse_and_gimplify_to_preheader): New function.
6816 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6817 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6818 (vectorizable_load): Likewise. For grouped stores always base
6819 the IV on the first element.
6820 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6821 condition before gimplifying.
6823 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6825 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6826 *DIV_EXPR and *MOD_EXPR.
6828 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6831 * config/i386/i386.c (ix86_option_override_internal): Mask out
6832 the CF_SET bit when checking -fcf-protection.
6834 2018-02-07 Tom de Vries <tom@codesourcery.com>
6837 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6840 2018-02-07 Richard Biener <rguenther@suse.de>
6842 PR tree-optimization/84204
6843 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6846 PR tree-optimization/84205
6847 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6848 special-case isl_ast_op_zdiv_r.
6850 PR tree-optimization/84223
6851 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6852 Only add conditions from within the region.
6853 (gather_bbs::after_dom_children): Adjust.
6855 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6858 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6859 * config/avr/avr.md: Only post-reload split REG-REG moves if
6860 either register is GENERAL_REG_P.
6862 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6864 PR tree-optimization/84235
6865 * tree-ssa-scopedtables.c
6866 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6867 if the subtraction is performed in floating point type where NaNs are
6868 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6869 build 1. Formatting fix.
6871 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6874 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6875 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6876 and skip it regardless of bb boundaries. Use CALL_P macro,
6877 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6878 unnecessarily, formatting fix.
6880 2018-02-06 Michael Collison <michael.collison@arm.com>
6882 * config/arm/thumb2.md:
6883 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6884 (*thumb_mov_notscc): Ditto.
6886 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6889 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6890 unsigned_fix, not "s".
6892 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6894 * configure.ac (gcc_fn_eh_frame_ro): New function.
6895 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6896 correct .eh_frame permissions.
6897 * configure: Regenerate.
6899 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
6901 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
6904 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6906 * config/rs6000/rs6000.c (rs6000_option_override_internal):
6907 Display warning message for -mno-speculate-indirect-jumps.
6909 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
6911 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
6913 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
6915 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
6917 PR tree-optimization/84225
6918 * tree-eh.c (find_trapping_overflow): Only call
6919 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
6921 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6924 * config/i386/i386.c: Reimplement the check of possible options
6925 -mibt/-mshstk conbination. Change error messages.
6926 * doc/invoke.texi: Fix a typo: remove extra '='.
6928 2018-02-06 Marek Polacek <polacek@redhat.com>
6930 PR tree-optimization/84228
6931 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
6933 2018-02-06 Tamar Christina <tamar.christina@arm.com>
6936 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
6937 emitted arch directives.
6938 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
6939 __ARM_FEATURE_COPROC before changing architectures.
6941 2018-02-06 Richard Biener <rguenther@suse.de>
6943 * config/i386/i386.c (print_reg): Fix typo.
6944 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
6946 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
6948 * configure: Regenerate.
6950 2018-02-05 Martin Sebor <msebor@redhat.com>
6952 PR tree-optimization/83369
6953 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
6956 2018-02-05 Martin Liska <mliska@suse.cz>
6958 * doc/invoke.texi: Cherry-pick upstream r323995.
6960 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
6962 * ira.c (ira_init_register_move_cost): Adjust comment.
6964 2018-02-05 Martin Liska <mliska@suse.cz>
6966 PR gcov-profile/84137
6967 * doc/gcov.texi: Fix typo in documentation.
6969 2018-02-05 Martin Liska <mliska@suse.cz>
6971 PR gcov-profile/83879
6972 * doc/gcov.texi: Document necessity of --dynamic-list-data when
6973 using dlopen functionality.
6975 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
6977 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
6978 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
6979 _mm_maskz_range_ss, _mm_mask_range_round_ss,
6980 _mm_maskz_range_round_ss): New intrinsics.
6981 (__builtin_ia32_rangesd128_round)
6982 (__builtin_ia32_rangess128_round): Remove.
6983 (__builtin_ia32_rangesd128_mask_round,
6984 __builtin_ia32_rangess128_mask_round): New builtins.
6985 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
6986 __builtin_ia32_rangess128_round): Remove.
6987 (__builtin_ia32_rangesd128_mask_round,
6988 __builtin_ia32_rangess128_mask_round): New builtins.
6989 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
6990 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
6991 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
6992 "<round_saeonly_constraint>")): Changed to ...
6993 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
6994 "<round_saeonly_scalar_constraint>")): ... this.
6995 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
6996 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
6997 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
6998 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
6999 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7001 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7003 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7005 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7006 Remove all values except native, 8540 and 8548.
7008 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7010 * config/i386/i386.c (ix86_output_function_return): Pass
7011 INVALID_REGNUM, instead of -1, as invalid register number to
7012 indirect_thunk_name and output_indirect_thunk.
7014 2018-02-02 Julia Koval <julia.koval@intel.com>
7016 * config.gcc: Add -march=icelake.
7017 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7018 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7019 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7020 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7021 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7022 (processor_target_table): Add icelake.
7023 (ix86_option_override_internal): Handle new PTAs.
7024 (get_builtin_code_for_version): Handle icelake.
7025 (M_INTEL_COREI7_ICELAKE): New.
7026 (fold_builtin_cpu): Handle icelake.
7027 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7028 * doc/invoke.texi: Add -march=icelake.
7030 2018-02-02 Julia Koval <julia.koval@intel.com>
7032 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7033 to wide_int_bitmask.
7034 * wide-int-bitmask.h: New.
7036 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7039 * config/i386/i386.md: Replace Pmode with word_mode in
7040 builtin_setjmp_setup and builtin_longjmp to support x32.
7042 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7046 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7048 (rs6000_supported_cpu_names): New static variable.
7049 (linux_cpu_translation_table): Likewise.
7050 (elf_platform) <cpu>: Define new static variable and use it.
7051 Translate kernel AT_PLATFORM name to canonical name if needed.
7052 Error if platform name is unknown.
7054 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7057 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7059 2018-02-01 Jeff Law <law@redhat.com>
7062 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7063 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7064 the scratch if RELEASE_VIA_POP is false.
7065 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7066 If we have to save a temporary register, decrement SIZE appropriately.
7067 Pass new arguments to release_scratch_register_on_entry.
7068 (ix86_adjust_stack_and_probe): Likewise.
7069 (ix86_emit_probe_stack_range): Pass new arguments to
7070 release_scratch_register_on_entry.
7072 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7074 PR rtl-optimization/84157
7075 * combine.c (change_zero_ext): Use REG_P predicate in
7076 front of HARD_REGISTER_P predicate.
7078 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7080 * config/avr/avr.c (avr_option_override): Move disabling of
7081 -fdelete-null-pointer-checks to...
7082 * common/config/avr/avr-common.c (avr_option_optimization_table):
7085 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7087 PR tree-optimization/81635
7088 * tree-data-ref.c (split_constant_offset_1): For types that
7089 wrap on overflow, try to use range info to prove that wrapping
7092 2018-02-01 Renlin Li <renlin.li@arm.com>
7095 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7097 (aarch64_register_move_cost): Likewise.
7098 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7100 (REG_CLASS_NAMES): Likewise.
7101 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7102 TAILCALL_ADDR_REGS. Remove IP registers.
7103 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7105 2018-02-01 Richard Biener <rguenther@suse.de>
7107 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7108 for specifying RPO order and allow NULL for that.
7109 * domwalk.c (dom_walker::dom_walker): Likewise.
7110 (dom_walker::walk): Handle NULL RPO order.
7111 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7113 (rewrite_update_dom_walker): Likewise.
7114 (mark_def_dom_walker): Likewise.
7116 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7118 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7119 (aarch64_maybe_expand_sve_subreg_move): Declare.
7120 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7121 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7123 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7124 that are semantically a reverse operation.
7125 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7126 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7127 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7129 (aarch64_can_change_mode_class): For big-endian, forbid changes
7130 between two SVE modes if they have different element sizes.
7132 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7134 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7135 the TImode handling for big-endian targets.
7137 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7139 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7140 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7142 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7143 Remove BSWAP handing for big-endian targets and use the form of
7144 LD1RQ appropariate for the mode.
7146 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7148 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7149 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7152 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7155 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7156 check for operands that need to go through aarch64_sve_reload_be.
7158 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7160 PR tree-optimization/81661
7161 PR tree-optimization/84117
7162 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7163 * tree-eh.c: Include gimplify.h.
7164 (find_trapping_overflow, replace_trapping_overflow,
7165 rewrite_to_non_trapping_overflow): New functions.
7166 * tree-vect-loop.c: Include tree-eh.h.
7167 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7168 * tree-data-ref.c: Include tree-eh.h.
7169 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7171 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7173 PR rtl-optimization/84123
7174 * combine.c (change_zero_ext): Check if hard register satisfies
7175 can_change_dest_mode before calling gen_lowpart_SUBREG.
7177 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7180 * ira.c (ira_init_register_move_cost): Remove assert.
7182 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7184 PR rtl-optimization/84071
7185 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7186 * doc/tm.texi: Regenerate.
7188 2018-01-31 Richard Biener <rguenther@suse.de>
7190 PR tree-optimization/84132
7191 * tree-data-ref.c (analyze_miv_subscript): Properly
7192 check whether evolution_function_is_affine_multivariate_p
7193 before calling gcd_of_steps_may_divide_p.
7195 2018-01-31 Julia Koval <julia.koval@intel.com>
7198 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7199 * config/i386/i386.md (rdpid_rex64) New.
7200 (rdpid): Make 32bit only.
7202 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7205 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7206 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7208 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7211 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7213 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7215 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7217 PR rtl-optimization/84071
7218 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7219 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7221 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7223 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7224 (arc_attribute_table): Add 'aux' attribute.
7225 (arc_in_small_data_p): Consider aux like variables.
7226 (arc_is_aux_reg_p): New function.
7227 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7228 (arc_get_aux_arg): New function.
7229 (prepare_move_operands): Handle aux-register access.
7230 (arc_handle_aux_attribute): New function.
7231 * doc/extend.texi (ARC Variable attributes): Add subsection.
7233 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7235 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7236 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7237 (arc_attribute_table): Add 'uncached' attribute.
7238 (arc_print_operand): Print '.di' flag for uncached memory
7240 (arc_in_small_data_p): Do not consider for small data the uncached
7242 (arc_is_uncached_mem_p): New function.
7243 * config/arc/predicates.md (compact_store_memory_operand): Check
7244 for uncached memory accesses.
7245 (nonvol_nonimm_operand): Likewise.
7246 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7248 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7251 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7252 falign-loops=): Add Optimization flag.
7254 2018-01-30 Jeff Law <law@redhat.com>
7257 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7258 INT_REGISTERS_SAVED. Check it prior to calling
7259 get_scratch_register_on_entry.
7260 (ix86_adjust_stack_and_probe): Similarly.
7261 (ix86_emit_probe_stack_range): Similarly.
7262 (ix86_expand_prologue): Corresponding changes.
7264 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7267 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7268 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7270 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7273 * lra-constraints.c (curr_insn_transform): Process AND in the
7276 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7278 PR rtl-optimization/83986
7279 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7280 dependence against last_pending_memory_flush in addition to
7283 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7285 PR tree-optimization/81611
7286 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7289 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7292 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7295 2018-01-30 Richard Biener <rguenther@suse.de>
7296 Jakub Jelinek <jakub@redhat.com>
7298 PR tree-optimization/84111
7299 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7300 inner loops added during recursion, as they don't have up-to-date
7303 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7306 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7307 (can_inline_edge_by_limits_p): ... here.
7308 (can_early_inline_edge_p, check_callers,
7309 update_caller_keys, update_callee_keys, recursive_inlining,
7310 add_new_edges_to_heap, speculation_useful_p,
7311 inline_small_functions,
7312 inline_small_functions, flatten_function,
7313 inline_to_all_callers_1): Update.
7315 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7317 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7320 2018-01-30 Richard Biener <rguenther@suse.de>
7322 PR tree-optimization/83008
7323 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7324 invariant and constant vector uses in stmts when they need
7327 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7330 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7331 * configure: Regenerate.
7333 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7335 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7337 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7338 Use gen_rtx_REG rather than gen_lowpart.
7340 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7342 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7343 rather than 0 when creating partial subregs.
7345 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7347 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7350 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7353 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7354 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7355 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7356 flags. This restores the settings used before the 2017-07-24.
7357 Turning off pre increment/decrement/modify allows IVOPTS to
7358 optimize DF/SF loops where the index is an int.
7360 2018-01-29 Richard Biener <rguenther@suse.de>
7361 Kelvin Nilsen <kelvin@gcc.gnu.org>
7364 * tree-vect-stmts.c (vectorizable_call): Don't call
7365 targetm.vectorize_builtin_md_vectorized_function if callee is
7368 2018-01-22 Carl Love <cel@us.ibm.com>
7370 * doc/extend.tex: Fix typo in second arg in
7371 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7373 2018-01-29 Richard Biener <rguenther@suse.de>
7375 PR tree-optimization/84086
7376 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7377 (flush_ssaname_freelist): When SSA names were released reset
7378 the SCEV hash table.
7380 2018-01-29 Richard Biener <rguenther@suse.de>
7382 PR tree-optimization/84057
7383 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7384 removed paths when removing edges.
7386 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7388 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7389 -mfunction-return=@var{choice}.
7391 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7394 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7396 (layout::print_source_line): Likewise.
7397 (test_get_line_width_without_trailing_whitespace): Add test cases.
7399 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7402 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7405 2018-01-26 Jim Wilson <jimw@sifive.com>
7407 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7409 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7412 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7414 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7415 and CMP + SUB-immediate -> SUBS.
7417 2018-01-26 Martin Sebor <msebor@redhat.com>
7419 PR tree-optimization/83896
7420 * tree-ssa-strlen.c (get_string_len): Rename...
7421 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7422 Avoid assuming length is constant.
7423 (handle_char_store): Use HOST_WIDE_INT for string length.
7425 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7428 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7429 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7431 2018-01-26 Richard Biener <rguenther@suse.de>
7433 PR rtl-optimization/84003
7434 * dse.c (record_store): Only record redundant stores when
7435 the earlier store aliases at least all accesses the later one does.
7437 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7439 PR rtl-optimization/83985
7440 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7441 REG_CFA_RESTORE insns.
7442 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7443 REG_CFA_RESTORE insns here.
7446 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7447 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7449 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7451 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7452 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7453 (arc_init): Likewise.
7454 (arc_override_options): Likewise.
7455 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7457 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7459 * config/arc/arc.h (TARGET_DBNZ): Define.
7460 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7461 properly set the tune attribute.
7462 (dbnz): Use TARGET_DBNZ guard.
7463 * config/arc/arc.opt (mtune): Add core3 option.
7465 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7467 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7468 recognize new pic like addresses.
7469 (arc_delegitimize_address): Clean up.
7471 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7473 * config/arc/arc-arches.def: Option mrf16 valid for all
7475 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7476 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7477 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7478 * config/arc/arc-tables.opt: Regenerate.
7479 * config/arc/arc.c (arc_conditional_register_usage): Handle
7480 reduced register file case.
7481 (arc_file_start): Set must have build attributes.
7482 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7484 * config/arc/arc.opt (mrf16): Add new option.
7485 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7486 * config/arc/genmultilib.awk: Handle new mrf16 option.
7487 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7488 * config/arc/t-multilib: Regenerate.
7489 * doc/invoke.texi (ARC Options): Document mrf16 option.
7491 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7493 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7494 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7495 (arc_attribute_table): Add 'secure_call' attribute.
7496 (arc_print_operand): Print secure call operand.
7497 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7499 (arc_is_secure_call_p): New function. * config/arc/arc.md
7500 (call_i): Add support for sjli instruction.
7501 (call_value_i): Likewise.
7502 * config/arc/constraints.md (Csc): New constraint.
7504 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7505 John Eric Martin <John.Martin@emmicro-us.com>
7507 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7508 * config/arc/arc.c (_arc_jli_section): New struct.
7509 (arc_jli_section): New type.
7510 (rc_jli_sections): New static variable.
7511 (arc_handle_jli_attribute): New function.
7512 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7513 (arc_file_end): New function.
7514 (TARGET_ASM_FILE_END): Define.
7515 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7516 (arc_add_jli_section): New function.
7517 (jli_call_scan): Likewise.
7518 (arc_reorg): Call jli_call_scan.
7519 (arc_output_addsi): Remove 'S' from printing asm operand.
7520 (arc_is_jli_call_p): New function.
7521 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7523 (movhi_insn): Likewise.
7524 (movsi_insn): Likewise.
7525 (movsi_set_cc_insn): Likewise.
7526 (loadqi_update): Likewise.
7527 (load_zeroextendqisi_update): Likewise.
7528 (load_signextendqisi_update): Likewise.
7529 (loadhi_update): Likewise.
7530 (load_zeroextendhisi_update): Likewise.
7531 (load_signextendhisi_update): Likewise.
7532 (loadsi_update): Likewise.
7533 (loadsf_update): Likewise.
7534 (movsicc_insn): Likewise.
7535 (bset_insn): Likewise.
7536 (bxor_insn): Likewise.
7537 (bclr_insn): Likewise.
7538 (bmsk_insn): Likewise.
7539 (bicsi3_insn): Likewise.
7540 (cmpsi_cc_c_insn): Likewise.
7541 (movsi_ne): Likewise.
7542 (movsi_cond_exec): Likewise.
7543 (clrsbsi2): Likewise.
7554 (call_i): Remove 'S' asm letter, add jli instruction.
7555 (call_value_i): Likewise.
7556 * config/arc/arc.op (mjli-always): New option.
7557 * config/arc/constraints.md (Cji): New constraint.
7558 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7560 (subsf3_fpx): Likewise.
7561 (mulsf3_fpx): Likewise.
7562 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7564 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7566 * doc/invoke.texi (ARC): Document mjli-always option.
7568 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7570 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7571 avoid addition with 0 and use incw and decw where possible.
7573 2018-01-26 Richard Biener <rguenther@suse.de>
7575 PR tree-optimization/81082
7576 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7577 association if it requires casting to unsigned.
7578 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7579 from fold_plusminus_mult_expr to catch important cases late when
7580 range info is available.
7582 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7584 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7585 * configure.ac (hidden_linkonce): New test.
7586 * configure: Regenerate.
7587 * config.in: Regenerate.
7589 2018-01-26 Julia Koval <julia.koval@intel.com>
7591 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7592 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7593 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7594 _mm_mask_bitshuffle_epi64_mask): Fix type.
7595 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7596 USI_FTYPE_V4DI_V4DI_USI): Remove.
7597 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7598 __builtin_ia32_vpshufbitqmb256_mask,
7599 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7600 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7601 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7603 2018-01-26 Alan Modra <amodra@gmail.com>
7606 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7607 UNSPEC_VBPERMQ. Sort other unspecs.
7609 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7611 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7613 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7616 * predict.c (drop_profile): Do not push/pop cfun; update also
7618 (handle_missing_profiles): Fix logic looking for zero profiles.
7620 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7623 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7624 on functions with #pragma omp declare simd or functions with simd
7626 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7627 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7628 Remove trailing \n from warning_at calls.
7630 2018-01-25 Tom de Vries <tom@codesourcery.com>
7633 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7634 for neutered workers.
7636 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7639 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7640 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7642 2018-01-24 Jeff Law <law@redhat.com>
7645 * i386.c (get_probe_interval): Move to earlier point.
7646 (ix86_compute_frame_layout): If -fstack-clash-protection and
7647 the frame is larger than the probe interval, then use pushes
7648 to save registers rather than reg->mem moves.
7649 (ix86_expand_prologue): Remove conditional for int_registers_saved
7652 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7655 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7656 min/max for never referenced object.
7658 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7661 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7663 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7664 attributes from DECL_ATTRIBUTES (decl) without affecting
7665 DECL_ATTRIBUTES (current_function_decl).
7666 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7667 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7669 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7671 PR tree-optimization/83979
7672 * fold-const.c (fold_comparison): Use constant_boolean_node
7673 instead of boolean_{true,false}_node.
7675 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7677 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7680 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7682 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7683 Simplify the clause that sets the length attribute.
7684 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7685 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7686 clause that sets the length attribute.
7687 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7689 2018-01-24 Tom de Vries <tom@codesourcery.com>
7692 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7693 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7694 Add strict parameter.
7695 (prevent_branch_around_nothing): Insert dummy insn between branch to
7696 label and label with no ptx insn inbetween.
7697 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7699 2018-01-24 Tom de Vries <tom@codesourcery.com>
7702 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7703 for neutered threads in warp.
7704 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7706 2018-01-24 Richard Biener <rguenther@suse.de>
7708 PR tree-optimization/83176
7709 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7712 2018-01-24 Richard Biener <rguenther@suse.de>
7714 PR tree-optimization/82819
7715 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7716 code generating pluses that are no-ops in the target precision.
7718 2018-01-24 Richard Biener <rguenther@suse.de>
7721 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7723 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7725 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7726 to merge probabilities.
7727 * predict.c (probably_never_executed): Also mark as cold functions
7728 with global 0 profile and guessed local profile.
7729 * profile-count.c (profile_probability::combine_with_count): New
7731 * profile-count.h (profile_probability::operator*,
7732 profile_probability::operator*=, profile_probability::operator/,
7733 profile_probability::operator/=): Reduce precision to adjusted
7734 and set value to guessed on contradictory divisions.
7735 (profile_probability::combine_with_freq): Remove.
7736 (profile_probability::combine_wiht_count): Declare.
7737 (profile_count::force_nonzero):: Set to adjusted.
7738 (profile_count::probability_in):: Set quality to adjusted.
7739 * tree-ssa-tail-merge.c (replace_block_by): Use
7742 2018-01-23 Andrew Waterman <andrew@sifive.com>
7743 Jim Wilson <jimw@sifive.com>
7745 * config/riscv/riscv.c (riscv_stack_boundary): New.
7746 (riscv_option_override): Set riscv_stack_boundary. Handle
7747 riscv_preferred_stack_boundary_arg.
7748 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7749 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7750 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7751 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7752 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7753 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7755 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7758 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7759 of struct ix86_frame.
7760 (ix86_expand_epilogue): Likewise. Add a local variable for
7761 the reg_save_offset field in struct ix86_frame.
7763 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7765 PR tree-optimization/82604
7766 * tree-loop-distribution.c (enum partition_kind): New enum item
7767 PKIND_PARTIAL_MEMSET.
7768 (partition_builtin_p): Support above new enum item.
7769 (generate_code_for_partition): Ditto.
7770 (compute_access_range): Differentiate cases that equality can be
7771 proven at all loops, the innermost loops or no loops.
7772 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7773 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7774 (finalize_partitions, distribute_loop): Don't fuse partition of
7775 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7776 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7779 2018-01-23 Martin Liska <mliska@suse.cz>
7781 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7782 order to ignore the predictor.
7783 (PRED_POLYMORPHIC_CALL): Likewise.
7784 (PRED_RECURSIVE_CALL): Likewise.
7786 2018-01-23 Martin Liska <mliska@suse.cz>
7788 * tree-profile.c (tree_profiling): Print function header to
7789 aware reader which function we are working on.
7790 * value-prof.c (gimple_find_values_to_profile): Do not print
7791 not interesting value histograms.
7793 2018-01-23 Martin Liska <mliska@suse.cz>
7795 * profile-count.h (enum profile_quality): Add
7796 profile_uninitialized as the first value. Do not number values
7797 as they are zero based.
7798 (profile_count::verify): Update sanity check.
7799 (profile_probability::verify): Likewise.
7801 2018-01-23 Nathan Sidwell <nathan@acm.org>
7803 * doc/invoke.texi (ffor-scope): Deprecate.
7805 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7807 PR tree-optimization/83510
7808 * domwalk.c (set_all_edges_as_executable): New function.
7809 (dom_walker::dom_walker): Convert bool param
7810 "skip_unreachable_blocks" to enum reachability. Move setup of
7811 edge flags to set_all_edges_as_executable and only do it when
7812 reachability is REACHABLE_BLOCKS.
7813 * domwalk.h (enum dom_walker::reachability): New enum.
7814 (dom_walker::dom_walker): Convert bool param
7815 "skip_unreachable_blocks" to enum reachability.
7816 (set_all_edges_as_executable): New decl.
7817 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7818 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7820 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7821 but converting true to REACHABLE_BLOCKS.
7822 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7824 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7825 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7826 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7828 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7829 if check_all_array_refs will be called.
7831 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7833 * tree.c (selftest::test_location_wrappers): Add more test
7836 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7838 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7839 (selftest::test_bit_in_range): Likewise.
7841 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7844 * doc/sourcebuild.texi (vect_float): Say that the selector
7845 only describes the situation when -funsafe-math-optimizations is on.
7846 (vect_float_strict): Document.
7848 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7850 PR tree-optimization/83965
7851 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7852 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7853 instead of checking only for a reduction.
7854 (vect_recog_widen_sum_pattern): Likewise.
7856 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7858 * predict.c (probably_never_executed): Only use precise profile info.
7859 (compute_function_frequency): Skip after inlining hack since we now
7860 have quality checking.
7862 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7864 * profile-count.h (profile_probability::very_unlikely,
7865 profile_probability::unlikely, profile_probability::even): Set
7866 precision to guessed.
7868 2018-01-23 Richard Biener <rguenther@suse.de>
7870 PR tree-optimization/83963
7871 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7872 Properly terminate dominator walk when crossing the exit edge not
7873 when visiting its source block.
7875 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7878 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7879 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7881 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7883 PR tree-optimization/83957
7884 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7885 semicolon after for body surrounded by braces.
7887 PR tree-optimization/83081
7888 * profile-count.h (profile_probability::split): New method.
7889 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7890 Use profile_probability::split.
7891 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7892 when splitting a single conditional jump into 2.
7894 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7896 PR tree-optimization/69452
7897 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
7900 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7902 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
7903 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
7904 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
7906 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7908 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
7909 * config/rl78/rl78.md (movdi): New define_expand.
7910 * config/rl78/rl78.c (rl78_split_movdi): New function.
7912 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
7915 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
7917 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
7918 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
7919 128-bit to produce an UNSPEC move to get the double word with the
7920 signbit and then a shift directly to do signbit.
7921 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
7922 implementation with a new version that just does either a direct
7923 move or a regular move. Move memory interface to separate insns.
7924 Move insns so they are next to the expander.
7925 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
7926 with signbit move. Split big and little endian case.
7927 (signbit<mode>2_dm_mem_le): Likewise.
7928 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
7929 (signbit<mode>2_dm2): Likewise.
7931 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7933 * config/rl78/rl78.md (anddi3): New define_expand.
7935 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7937 * config/rl78/rl78.md (umindi3): New define_expand.
7939 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7941 * config/rl78/rl78.md (smindi3): New define_expand.
7943 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7945 * config/rl78/rl78.md (smaxdi3): New define_expand.
7947 2018-01-22 Carl Love <cel@us.ibm.com>
7949 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
7950 LVX_V1TI): Add macro expansion.
7951 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
7952 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
7953 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
7954 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
7955 Change check to determine if the instruction is a byte reversing
7956 entry. Fix typo in comment.
7957 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
7958 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
7959 Add def_builtin calls for new builtins.
7960 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
7961 Add define_insn expansion.
7963 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7965 * config/rl78/rl78.md (umaxdi3): New define_expand.
7967 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
7969 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
7970 for non-QImode registers.
7972 2018-01-22 Richard Biener <rguenther@suse.de>
7974 PR tree-optimization/83963
7975 * graphite-scop-detection.c (scop_detection::get_sese): Delay
7976 including the loop exit block.
7977 (scop_detection::merge_sese): Likewise.
7978 (scop_detection::add_scop): Do it here instead.
7980 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7982 * doc/sourcebuild.texi (arm_softfloat): Document.
7984 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
7987 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
7988 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
7989 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
7991 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7992 David Edelsohn <dje.gcc@gmail.com>
7995 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7996 Change "crset eq" to "crset 2".
7997 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7998 (*call_indirect_aix<mode>_nospec): Likewise.
7999 (*call_value_indirect_aix<mode>_nospec): Likewise.
8000 (*call_indirect_elfv2<mode>_nospec): Likewise.
8001 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8002 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8003 change assembly output from . to $.
8004 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8005 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8006 (*tablejump<mode>_internal1_nospec): Likewise.
8008 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8011 * config/sh/sh_optimize_sett_clrt.cc:
8012 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8014 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8016 PR tree-optimization/83940
8017 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8018 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8019 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8020 use it to pass back the definition type.
8021 (vect_check_store_rhs): Likewise rhs_dt_out.
8022 (vect_build_gather_load_calls): Add a mask_dt argument and use
8023 it instead of a call to vect_is_simple_use.
8024 (vectorizable_store): Update calls to vect_check_load_store_mask
8025 and vect_check_store_rhs. Use the dt returned by the latter instead
8026 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8027 instead of calls to vect_is_simple_use. Pass the scalar rather
8028 than the vector operand to vect_is_simple_use when handling
8029 second and subsequent copies of an rhs value.
8030 (vectorizable_load): Update calls to vect_check_load_store_mask
8031 and vect_build_gather_load_calls. Use the cached mask_dt and
8032 gs_info.offset_dt instead of calls to vect_is_simple_use.
8034 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8037 * tree-emutls.c: Include gimplify.h.
8038 (lower_emutls_2): New function.
8039 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8040 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8041 it before further processing.
8044 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8045 UINTVAL (trueop1) instead of INTVAL (op1).
8047 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8051 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8052 INCOMING_FRAME_SP_OFFSET if not defined.
8053 (scan_trace): Add ENTRY argument. If true and
8054 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8055 emit a note to adjust the CFA offset.
8056 (create_cfi_notes): Adjust scan_trace callers.
8057 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8058 INCOMING_FRAME_SP_OFFSET in the CIE.
8059 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8060 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8062 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8063 * doc/tm.texi: Regenerated.
8065 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8067 PR rtl-optimization/83147
8068 * lra-constraints.c (remove_inheritance_pseudos): Use
8069 lra_substitute_pseudo_within_insn.
8071 2018-01-19 Tom de Vries <tom@codesourcery.com>
8072 Cesar Philippidis <cesar@codesourcery.com>
8075 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8077 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8080 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8081 spaces for function labels.
8083 2018-01-19 Martin Liska <mliska@suse.cz>
8085 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8086 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8087 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8088 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8089 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8090 (PRED_CONST_RETURN): Change from 69 to 65.
8091 (PRED_NULL_RETURN): Change from 91 to 71.
8092 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8093 (PRED_LOOP_GUARD): Change from 66 to 73.
8095 2018-01-19 Martin Liska <mliska@suse.cz>
8097 * predict.c (predict_insn_def): Add new assert.
8098 (struct branch_predictor): Change type to signed integer.
8099 (test_prediction_value_range): Amend test to cover
8101 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8102 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8103 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8104 (PRED_LOOP_IV_COMPARE): Likewise.
8105 * predict.h (PROB_UNINITIALIZED): Define new constant.
8107 2018-01-19 Martin Liska <mliska@suse.cz>
8109 * predict.c (dump_prediction): Add new format for
8110 analyze_brprob.py script which is enabled with -details
8112 * profile-count.h (precise_p): New function.
8114 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8116 PR tree-optimization/83922
8117 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8118 there are no statements that need masking.
8119 (vect_active_double_reduction_p): New function.
8120 (vect_analyze_loop_operations): Use it when handling phis that
8121 are not in the loop header.
8123 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8125 PR tree-optimization/83914
8126 * tree-vect-loop.c (vectorizable_induction): Don't convert
8127 init_expr or apply the peeling adjustment for inductions
8128 that are nested within the vectorized loop.
8130 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8132 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8135 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8139 * function.h (gimplify_parameters): Add gimple_seq * argument.
8140 * function.c: Include gimple.h and options.h.
8141 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8142 for the added local temporaries if needed.
8143 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8144 if there are any parameter cleanups, wrap whole body into a
8145 try/finally with the cleanups.
8147 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8150 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8151 Use GET_MODE_CLASS for scalar floating point.
8153 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8157 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8158 Fix call of call_cgraph_insertion_hooks.
8160 2018-01-18 Martin Sebor <msebor@redhat.com>
8162 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8164 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8167 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8170 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8173 * common.opt: (-ffile-prefix-map): New option.
8174 * opts.c (common_handle_option): Defer it.
8175 * opts-global.c (handle_common_deferred_options): Handle it.
8176 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8177 * file-prefix-map.h: New file.
8178 (remap_debug_filename, add_debug_prefix_map): ...here.
8179 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8180 * final.c (debug_prefix_map, add_debug_prefix_map
8181 remap_debug_filename): Move to...
8182 * file-prefix-map.c: New file.
8183 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8184 generalize, get rid of alloca(), use strrchr() instead of strchr().
8185 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8186 Implement in terms of add_prefix_map().
8187 (remap_macro_filename, remap_debug_filename): Implement in term of
8189 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8190 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8191 * dbxout.c: Include file-prefix-map.h.
8192 * varasm.c: Likewise.
8193 * vmsdbgout.c: Likewise.
8194 * xcoffout.c: Likewise.
8195 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8196 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8197 * doc/invoke.texi (-ffile-prefix-map): Document.
8198 (-fdebug-prefix-map): Update description.
8200 2018-01-18 Martin Liska <mliska@suse.cz>
8202 * config/i386/i386.c (indirect_thunk_name): Document that also
8204 (output_indirect_thunk): Document why both instructions
8205 (pause and lfence) are generated.
8207 2018-01-18 Richard Biener <rguenther@suse.de>
8209 PR tree-optimization/83887
8210 * graphite-scop-detection.c
8211 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8212 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8213 (scop_detection::merge_sese): Re-implement with a flood-fill
8214 algorithm that properly finds a SESE region if it exists.
8216 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8219 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8220 pointer_diff optimizations use view_convert instead of convert.
8222 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8224 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8225 Generate different code for -mno-speculate-indirect-jumps.
8226 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8227 (*call_indirect_aix<mode>): Disable for
8228 -mno-speculate-indirect-jumps.
8229 (*call_indirect_aix<mode>_nospec): New define_insn.
8230 (*call_value_indirect_aix<mode>): Disable for
8231 -mno-speculate-indirect-jumps.
8232 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8233 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8234 -mno-speculate-indirect-jumps.
8235 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8237 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8239 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8240 long double type, set the flags for noting the default long double
8241 type, even if we don't pass or return a long double type.
8243 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8246 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8249 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8251 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8252 support for merge[hl].
8253 (fold_mergehl_helper): New helper function.
8254 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8255 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8256 (altivec_vmrglw_direct): Add xxmrglw insn.
8258 2018-01-17 Andrew Waterman <andrew@sifive.com>
8260 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8261 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8263 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8266 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8267 call the lto_location_cache before reading the
8268 DECL_SOURCE_LOCATION of the types.
8270 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8271 Richard Sandiford <richard.sandiford@linaro.org>
8273 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8274 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8275 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8276 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8277 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8279 * config/aarch64/constraints.md (aarch64_movti_operand):
8281 * config/aarch64/predicates.md (Uti): Add new constraint.
8283 2018-01-17 Carl Love <cel@us.ibm.com>
8285 * config/rs6000/vsx.md (define_expand xl_len_r,
8286 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8287 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8289 (define_expand, define_insn): Move the shift left from the
8290 define_insn to the define_expand for lxvl and stxvl instructions.
8291 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8292 and XL_LEN_R definitions to PURE.
8294 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8296 * config/i386/i386.c (indirect_thunk_name): Declare regno
8297 as unsigned int. Compare regno with INVALID_REGNUM.
8298 (output_indirect_thunk): Ditto.
8299 (output_indirect_thunk_function): Ditto.
8300 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8301 in the call to output_indirect_thunk_function.
8303 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8306 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8307 rather than the size of inner_type to determine the stack slot size
8308 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8310 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8313 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8316 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8318 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8319 endian Linux systems to optionally enable multilibs for selecting
8320 the long double type if the user configured an explicit type.
8321 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8322 have no long double multilibs if not defined.
8323 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8324 warn if the user used -mabi={ieee,ibm}longdouble and we built
8325 multilibs for long double.
8326 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8327 appropriate multilib option.
8328 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8330 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8331 for building long double multilibs.
8332 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8334 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8336 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8339 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8341 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8344 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8347 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8350 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8352 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8353 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8355 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8357 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8358 different rtl trees depending on TARGET_64BIT.
8359 (rs6000_gen_lvx): Likewise.
8361 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8363 * config/visium/visium.md (nop): Tweak comment.
8364 (hazard_nop): Likewise.
8366 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8368 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8369 -mspeculate-indirect-jumps.
8370 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8371 for -mno-speculate-indirect-jumps.
8372 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8373 (*call_value_indirect_elfv2<mode>): Disable for
8374 -mno-speculate-indirect-jumps.
8375 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8376 (indirect_jump): Emit different RTL for
8377 -mno-speculate-indirect-jumps.
8378 (*indirect_jump<mode>): Disable for
8379 -mno-speculate-indirect-jumps.
8380 (*indirect_jump<mode>_nospec): New define_insn.
8381 (tablejump): Emit different RTL for
8382 -mno-speculate-indirect-jumps.
8383 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8384 (tablejumpsi_nospec): New define_expand.
8385 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8386 (tablejumpdi_nospec): New define_expand.
8387 (*tablejump<mode>_internal1): Disable for
8388 -mno-speculate-indirect-jumps.
8389 (*tablejump<mode>_internal1_nospec): New define_insn.
8390 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8393 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8395 * caller-save.c (insert_save): Drop unnecessary parameter. All
8398 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8399 Richard Biener <rguenth@suse.de>
8402 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8403 return early, inline manually is_gimple_sizepos. Make sure if we
8404 call gimplify_expr we don't end up with a gimple constant.
8405 * tree.c (variably_modified_type_p): Don't return true for
8406 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8407 * gimplify.h (is_gimple_sizepos): Remove.
8409 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8411 PR tree-optimization/83857
8412 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8413 vectorizable_live_operation for pure SLP statements.
8414 (vectorizable_live_operation): Handle PHIs.
8416 2018-01-16 Richard Biener <rguenther@suse.de>
8418 PR tree-optimization/83867
8419 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8420 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8422 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8425 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8426 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8427 If off is not INTEGER_CST, issue a may not be aligned warning
8428 rather than isn't aligned. Use isn%'t rather than isn't.
8429 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8431 <case MULT_EXPR>: Improve the case when bottom and one of the
8432 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8433 operand, in that case check if the other operand is multiple of
8434 bottom divided by the INTEGER_CST operand.
8436 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8439 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8440 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8441 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8442 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8443 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8444 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8445 (pa_function_arg_size): New function.
8447 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8449 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8450 in a separate statement.
8452 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8454 PR tree-optimization/83847
8455 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8456 group gathers and scatters.
8458 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8460 PR rtl-optimization/86620
8461 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8463 PR rtl-optimization/83213
8464 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8465 to last if both are JUMP_INSNs.
8467 PR tree-optimization/83843
8468 * gimple-ssa-store-merging.c
8469 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8470 store_immediate_info for bswap/nop orig_stores.
8472 2018-01-15 Andrew Waterman <andrew@sifive.com>
8474 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8476 <UDIV>: Increase cost if !TARGET_DIV.
8478 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8480 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8481 (define_attr "cr_logical_3op"): New.
8482 (cceq_ior_compare): Adjust.
8483 (cceq_ior_compare_complement): Adjust.
8484 (*cceq_rev_compare): Adjust.
8485 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8486 (is_cracked_insn): Adjust.
8487 (insn_must_be_first_in_group): Adjust.
8488 * config/rs6000/40x.md: Adjust.
8489 * config/rs6000/440.md: Adjust.
8490 * config/rs6000/476.md: Adjust.
8491 * config/rs6000/601.md: Adjust.
8492 * config/rs6000/603.md: Adjust.
8493 * config/rs6000/6xx.md: Adjust.
8494 * config/rs6000/7450.md: Adjust.
8495 * config/rs6000/7xx.md: Adjust.
8496 * config/rs6000/8540.md: Adjust.
8497 * config/rs6000/cell.md: Adjust.
8498 * config/rs6000/e300c2c3.md: Adjust.
8499 * config/rs6000/e500mc.md: Adjust.
8500 * config/rs6000/e500mc64.md: Adjust.
8501 * config/rs6000/e5500.md: Adjust.
8502 * config/rs6000/e6500.md: Adjust.
8503 * config/rs6000/mpc.md: Adjust.
8504 * config/rs6000/power4.md: Adjust.
8505 * config/rs6000/power5.md: Adjust.
8506 * config/rs6000/power6.md: Adjust.
8507 * config/rs6000/power7.md: Adjust.
8508 * config/rs6000/power8.md: Adjust.
8509 * config/rs6000/power9.md: Adjust.
8510 * config/rs6000/rs64.md: Adjust.
8511 * config/rs6000/titan.md: Adjust.
8513 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8515 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8516 ix86_indirect_branch_register logic.
8518 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8520 * config/i386/constraints.md (Bs): Update
8521 ix86_indirect_branch_register check. Don't check
8522 ix86_indirect_branch_register with GOT_memory_operand.
8524 * config/i386/predicates.md (GOT_memory_operand): Don't check
8525 ix86_indirect_branch_register here.
8526 (GOT32_symbol_operand): Likewise.
8528 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8530 * config/i386/predicates.md (constant_call_address_operand):
8531 Rewrite ix86_indirect_branch_register logic.
8532 (sibcall_insn_operand): Likewise.
8534 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8536 * config/i386/constraints.md (Bs): Replace
8537 ix86_indirect_branch_thunk_register with
8538 ix86_indirect_branch_register.
8540 * config/i386/i386.md (indirect_jump): Likewise.
8541 (tablejump): Likewise.
8542 (*sibcall_memory): Likewise.
8543 (*sibcall_value_memory): Likewise.
8544 Peepholes of indirect call and jump via memory: Likewise.
8545 * config/i386/i386.opt: Likewise.
8546 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8547 (GOT_memory_operand): Likewise.
8548 (call_insn_operand): Likewise.
8549 (sibcall_insn_operand): Likewise.
8550 (GOT32_symbol_operand): Likewise.
8552 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8555 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8556 type rather than type addr's type points to.
8557 (expand_omp_atomic_mutex): Likewise.
8558 (expand_omp_atomic): Likewise.
8560 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8563 * config/i386/i386.c (output_indirect_thunk_function): Use
8564 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8565 for __x86_return_thunk.
8567 2018-01-15 Richard Biener <rguenther@suse.de>
8570 * expmed.c (extract_bit_field_1): Fix typo.
8572 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8575 * config/arm/iterators.md (VF): New mode iterator.
8576 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8577 Remove integer-related logic from pattern.
8578 (neon_vabd<mode>_3): Likewise.
8580 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8583 * common.opt (fstrict-overflow): No longer an alias.
8584 (fwrapv-pointer): New option.
8585 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8586 also for pointer types based on flag_wrapv_pointer.
8587 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8588 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8589 opts->x_flag_wrapv got set.
8590 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8591 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8592 POINTER_TYPE_OVERFLOW_UNDEFINED.
8593 * match.pd: Likewise in address comparison pattern.
8594 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8596 2018-01-15 Richard Biener <rguenther@suse.de>
8599 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8600 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8601 Reset type names to their identifier if their TYPE_DECL doesn't
8602 have linkage (and thus is used for ODR and devirt).
8603 (save_debug_info_for_decl): Remove.
8604 (save_debug_info_for_type): Likewise.
8605 (add_tree_to_fld_list): Adjust.
8606 * tree-pretty-print.c (dump_generic_node): Make dumping of
8607 type names more robust.
8609 2018-01-15 Richard Biener <rguenther@suse.de>
8611 * BASE-VER: Bump to 8.0.1.
8613 2018-01-14 Martin Sebor <msebor@redhat.com>
8616 * builtins.c (check_access): Avoid warning when the no-warning bit
8619 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8621 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8622 * ira-color (allocno_hard_regs_compare): Likewise.
8624 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8627 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8628 Use .pushsection/.popsection.
8630 2018-01-14 Martin Sebor <msebor@redhat.com>
8633 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8635 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8637 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8638 entry from extra_headers.
8639 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8640 extra_headers, make the list bitwise identical to the i?86-*-* one.
8642 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8644 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8645 -mcmodel=large with -mindirect-branch=thunk,
8646 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8647 -mfunction-return=thunk-extern.
8648 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8649 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8650 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8652 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8654 * config/i386/i386.c (print_reg): Print the name of the full
8655 integer register without '%'.
8656 (ix86_print_operand): Handle 'V'.
8657 * doc/extend.texi: Document 'V' modifier.
8659 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8661 * config/i386/constraints.md (Bs): Disallow memory operand for
8662 -mindirect-branch-register.
8664 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8665 (GOT_memory_operand): Likewise.
8666 (call_insn_operand): Likewise.
8667 (sibcall_insn_operand): Likewise.
8668 (GOT32_symbol_operand): Likewise.
8669 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8670 for -mindirect-branch-register.
8671 (tablejump): Likewise.
8672 (*sibcall_memory): Likewise.
8673 (*sibcall_value_memory): Likewise.
8674 Disallow peepholes of indirect call and jump via memory for
8675 -mindirect-branch-register.
8676 (*call_pop): Replace m with Bw.
8677 (*call_value_pop): Likewise.
8678 (*sibcall_pop_memory): Replace m with Bs.
8679 * config/i386/i386.opt (mindirect-branch-register): New option.
8680 * doc/invoke.texi: Document -mindirect-branch-register option.
8682 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8684 * config/i386/i386-protos.h (ix86_output_function_return): New.
8685 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8686 set function_return_type.
8687 (indirect_thunk_name): Add ret_p to indicate thunk for function
8689 (output_indirect_thunk_function): Pass false to
8690 indirect_thunk_name.
8691 (ix86_output_indirect_branch_via_reg): Likewise.
8692 (ix86_output_indirect_branch_via_push): Likewise.
8693 (output_indirect_thunk_function): Create alias for function
8694 return thunk if regno < 0.
8695 (ix86_output_function_return): New function.
8696 (ix86_handle_fndecl_attribute): Handle function_return.
8697 (ix86_attribute_table): Add function_return.
8698 * config/i386/i386.h (machine_function): Add
8699 function_return_type.
8700 * config/i386/i386.md (simple_return_internal): Use
8701 ix86_output_function_return.
8702 (simple_return_internal_long): Likewise.
8703 * config/i386/i386.opt (mfunction-return=): New option.
8704 (indirect_branch): Mention -mfunction-return=.
8705 * doc/extend.texi: Document function_return function attribute.
8706 * doc/invoke.texi: Document -mfunction-return= option.
8708 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8710 * config/i386/i386-opts.h (indirect_branch): New.
8711 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8712 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8713 with local indirect jump when converting indirect call and jump.
8714 (ix86_set_indirect_branch_type): New.
8715 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8716 (indirectlabelno): New.
8717 (indirect_thunk_needed): Likewise.
8718 (indirect_thunk_bnd_needed): Likewise.
8719 (indirect_thunks_used): Likewise.
8720 (indirect_thunks_bnd_used): Likewise.
8721 (INDIRECT_LABEL): Likewise.
8722 (indirect_thunk_name): Likewise.
8723 (output_indirect_thunk): Likewise.
8724 (output_indirect_thunk_function): Likewise.
8725 (ix86_output_indirect_branch_via_reg): Likewise.
8726 (ix86_output_indirect_branch_via_push): Likewise.
8727 (ix86_output_indirect_branch): Likewise.
8728 (ix86_output_indirect_jmp): Likewise.
8729 (ix86_code_end): Call output_indirect_thunk_function if needed.
8730 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8732 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8733 (ix86_attribute_table): Add indirect_branch.
8734 * config/i386/i386.h (machine_function): Add indirect_branch_type
8735 and has_local_indirect_jump.
8736 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8738 (tablejump): Likewise.
8739 (*indirect_jump): Use ix86_output_indirect_jmp.
8740 (*tablejump_1): Likewise.
8741 (simple_return_indirect_internal): Likewise.
8742 * config/i386/i386.opt (mindirect-branch=): New option.
8743 (indirect_branch): New.
8746 (thunk-inline): Likewise.
8747 (thunk-extern): Likewise.
8748 * doc/extend.texi: Document indirect_branch function attribute.
8749 * doc/invoke.texi: Document -mindirect-branch= option.
8751 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8754 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8756 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8758 * ipa-inline.c (want_inline_small_function_p): Return false if
8759 inlining has already failed with CIF_FINAL_ERROR.
8760 (update_caller_keys): Call want_inline_small_function_p before
8762 (update_callee_keys): Likewise.
8764 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8766 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8768 (rs6000_quadword_masked_address_p): Likewise.
8769 (quad_aligned_load_p): Likewise.
8770 (quad_aligned_store_p): Likewise.
8771 (const_load_sequence_p): Add comment to describe the outer-most loop.
8772 (mimic_memory_attributes_and_flags): New function.
8773 (rs6000_gen_stvx): Likewise.
8774 (replace_swapped_aligned_store): Likewise.
8775 (rs6000_gen_lvx): Likewise.
8776 (replace_swapped_aligned_load): Likewise.
8777 (replace_swapped_load_constant): Capitalize argument name in
8778 comment describing this function.
8779 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8780 and stores that access quad-word aligned addresses and replace
8781 with stvx or lvx instructions when appropriate.
8782 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8783 New function prototype.
8784 (rs6000_quadword_masked_address_p): Likewise.
8785 (rs6000_gen_lvx): Likewise.
8786 (rs6000_gen_stvx): Likewise.
8787 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8788 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8789 when memory address is aligned.
8790 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8791 this split to select lvx instruction when memory address is aligned.
8792 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8793 instruction when memory address is aligned.
8794 (*vsx_le_perm_load_v16qi): Likewise.
8795 (four unnamed splitters): Modify to select the stvx instruction
8796 when memory is aligned.
8798 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8800 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8801 which appears in the queue multiple times.
8803 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8804 Alan Hayward <alan.hayward@arm.com>
8805 David Sherwood <david.sherwood@arm.com>
8807 * tree-vectorizer.h (vec_lower_bound): New structure.
8808 (_loop_vec_info): Add check_nonzero and lower_bounds.
8809 (LOOP_VINFO_CHECK_NONZERO): New macro.
8810 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8811 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8812 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8813 fields. Make seg_len the distance travelled, not including the
8815 (dr_direction_indicator): Declare.
8816 (dr_zero_step_indicator): Likewise.
8817 (dr_known_forward_stride_p): Likewise.
8818 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8820 (runtime_alias_check_p): Allow runtime alias checks with
8822 (operator ==): Compare access_size and align.
8823 (prune_runtime_alias_test_list): Rework for new distinction between
8824 the access_size and seg_len.
8825 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8827 (get_segment_min_max): New function.
8828 (create_intersect_range_checks): Use it.
8829 (dr_step_indicator): New function.
8830 (dr_direction_indicator): Likewise.
8831 (dr_zero_step_indicator): Likewise.
8832 (dr_known_forward_stride_p): Likewise.
8833 * tree-loop-distribution.c (data_ref_segment_size): Return
8834 DR_STEP * (niters - 1).
8835 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8837 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8838 (vect_preserves_scalar_order_p): New function, split out from...
8839 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8840 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8841 (vect_vfa_access_size): New function.
8842 (vect_vfa_align): Likewise.
8843 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8844 (dump_lower_bound): New function.
8845 (vect_check_lower_bound): Likewise.
8846 (vect_small_gap_p): Likewise.
8847 (vectorizable_with_step_bound_p): Likewise.
8848 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8849 depencies if the vectorization factor is 1. Convert the checks
8850 for nonzero steps into checks on the bounds of DR_STEP. Try using
8851 a bunds check for variable steps if the minimum required step is
8852 relatively small. Update calls to the dr_with_seg_len
8853 constructor and to vect_compile_time_alias.
8854 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8856 (vect_loop_versioning): Call it.
8857 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8859 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8861 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8862 Alan Hayward <alan.hayward@arm.com>
8863 David Sherwood <david.sherwood@arm.com>
8865 * doc/sourcebuild.texi (vect_scatter_store): Document.
8866 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8868 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8870 * genopinit.c (main): Add supports_vec_scatter_store and
8871 supports_vec_scatter_store_cached to target_optabs.
8872 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8873 IFN_MASK_SCATTER_STORE.
8874 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8876 * internal-fn.h (internal_store_fn_p): Declare.
8877 (internal_fn_stored_value_index): Likewise.
8878 * internal-fn.c (scatter_store_direct): New macro.
8879 (expand_scatter_store_optab_fn): New function.
8880 (direct_scatter_store_optab_supported_p): New macro.
8881 (internal_store_fn_p): New function.
8882 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8883 IFN_MASK_SCATTER_STORE.
8884 (internal_fn_mask_index): Likewise.
8885 (internal_fn_stored_value_index): New function.
8886 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8888 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8889 * optabs-query.c (supports_vec_scatter_store_p): New function.
8890 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8891 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8892 true for scatter stores.
8893 (vect_gather_scatter_fn_p): Handle scatter stores too.
8894 (vect_check_gather_scatter): Consider using scatter stores if
8895 supports_vec_scatter_store_p.
8896 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8898 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8899 internal_fn_stored_value_index.
8900 (check_load_store_masking): Handle scatter stores too.
8901 (vect_get_store_rhs): Make public.
8902 (vectorizable_call): Use internal_store_fn_p.
8903 (vectorizable_store): Handle scatter store internal functions.
8904 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
8905 when deciding whether the end of the group has been reached.
8906 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
8907 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
8908 (mask_scatter_store<mode>): New insns.
8910 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8911 Alan Hayward <alan.hayward@arm.com>
8912 David Sherwood <david.sherwood@arm.com>
8914 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
8915 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
8916 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
8918 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
8919 Use vect_truncate_gather_scatter_offset if we can't treat the
8920 operation as a normal gather load or scatter store.
8921 (get_group_load_store_type): Take the gather_scatter_info
8922 as argument. Try using a gather load or scatter store for
8923 single-element groups.
8924 (get_load_store_type): Update calls to get_group_load_store_type
8925 and vect_use_strided_gather_scatters_p.
8927 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8928 Alan Hayward <alan.hayward@arm.com>
8929 David Sherwood <david.sherwood@arm.com>
8931 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
8932 optional tree argument.
8933 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
8935 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
8936 but continue to use the current value as a fallback.
8937 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
8938 to compare the updates.
8939 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
8940 (get_load_store_type): Use it when handling a strided access.
8941 (vect_get_strided_load_store_ops): New function.
8942 (vect_get_data_ptr_increment): Likewise.
8943 (vectorizable_load): Handle strided gather loads. Always pass
8944 a step to vect_create_data_ref_ptr and bump_vector_ptr.
8946 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8947 Alan Hayward <alan.hayward@arm.com>
8948 David Sherwood <david.sherwood@arm.com>
8950 * doc/md.texi (gather_load@var{m}): Document.
8951 (mask_gather_load@var{m}): Likewise.
8952 * genopinit.c (main): Add supports_vec_gather_load and
8953 supports_vec_gather_load_cached to target_optabs.
8954 * optabs-tree.c (init_tree_optimization_optabs): Use
8955 ggc_cleared_alloc to allocate target_optabs.
8956 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
8957 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
8959 * internal-fn.h (internal_load_fn_p): Declare.
8960 (internal_gather_scatter_fn_p): Likewise.
8961 (internal_fn_mask_index): Likewise.
8962 (internal_gather_scatter_fn_supported_p): Likewise.
8963 * internal-fn.c (gather_load_direct): New macro.
8964 (expand_gather_load_optab_fn): New function.
8965 (direct_gather_load_optab_supported_p): New macro.
8966 (direct_internal_fn_optab): New function.
8967 (internal_load_fn_p): Likewise.
8968 (internal_gather_scatter_fn_p): Likewise.
8969 (internal_fn_mask_index): Likewise.
8970 (internal_gather_scatter_fn_supported_p): Likewise.
8971 * optabs-query.c (supports_at_least_one_mode_p): New function.
8972 (supports_vec_gather_load_p): Likewise.
8973 * optabs-query.h (supports_vec_gather_load_p): Declare.
8974 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
8975 and memory_type field.
8976 (NUM_PATTERNS): Bump to 15.
8977 * tree-vect-data-refs.c: Include internal-fn.h.
8978 (vect_gather_scatter_fn_p): New function.
8979 (vect_describe_gather_scatter_call): Likewise.
8980 (vect_check_gather_scatter): Try using internal functions for
8981 gather loads. Recognize existing calls to a gather load function.
8982 (vect_analyze_data_refs): Consider using gather loads if
8983 supports_vec_gather_load_p.
8984 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
8985 (vect_get_gather_scatter_offset_type): Likewise.
8986 (vect_convert_mask_for_vectype): Likewise.
8987 (vect_add_conversion_to_patterm): Likewise.
8988 (vect_try_gather_scatter_pattern): Likewise.
8989 (vect_recog_gather_scatter_pattern): New pattern recognizer.
8990 (vect_vect_recog_func_ptrs): Add it.
8991 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
8992 internal_fn_mask_index and internal_gather_scatter_fn_p.
8993 (check_load_store_masking): Take the gather_scatter_info as an
8994 argument and handle gather loads.
8995 (vect_get_gather_scatter_ops): New function.
8996 (vectorizable_call): Check internal_load_fn_p.
8997 (vectorizable_load): Likewise. Handle gather load internal
8999 (vectorizable_store): Update call to check_load_store_masking.
9000 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9001 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9002 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9003 (aarch64_gather_scale_operand_d): New predicates.
9004 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9005 (mask_gather_load<mode>): New insns.
9007 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9008 Alan Hayward <alan.hayward@arm.com>
9009 David Sherwood <david.sherwood@arm.com>
9011 * optabs.def (fold_left_plus_optab): New optab.
9012 * doc/md.texi (fold_left_plus_@var{m}): Document.
9013 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9014 * internal-fn.c (fold_left_direct): Define.
9015 (expand_fold_left_optab_fn): Likewise.
9016 (direct_fold_left_optab_supported_p): Likewise.
9017 * fold-const-call.c (fold_const_fold_left): New function.
9018 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9019 * tree-parloops.c (valid_reduction_p): New function.
9020 (gather_scalar_reductions): Use it.
9021 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9022 (vect_finish_replace_stmt): Declare.
9023 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9024 (needs_fold_left_reduction_p): New function, split out from...
9025 (vect_is_simple_reduction): ...here. Accept reductions that
9026 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9027 (vect_force_simple_reduction): Also store the reduction type in
9028 the assignment's STMT_VINFO_REDUC_TYPE.
9029 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9030 (merge_with_identity): New function.
9031 (vect_expand_fold_left): Likewise.
9032 (vectorize_fold_left_reduction): Likewise.
9033 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9034 scalar phi in place for it. Check for target support and reject
9035 cases that would reassociate the operation. Defer the transform
9036 phase to vectorize_fold_left_reduction.
9037 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9038 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9039 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9041 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9043 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9044 call to ifc_temp_var.
9046 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9047 Alan Hayward <alan.hayward@arm.com>
9048 David Sherwood <david.sherwood@arm.com>
9050 * target.def (legitimize_address_displacement): Take the original
9051 offset as a poly_int.
9052 * targhooks.h (default_legitimize_address_displacement): Update
9054 * targhooks.c (default_legitimize_address_displacement): Likewise.
9055 * doc/tm.texi: Regenerate.
9056 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9057 as an argument, moving assert of ad->disp == ad->disp_term to...
9058 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9059 Try calling targetm.legitimize_address_displacement before expanding
9060 the address rather than afterwards, and adjust for the new interface.
9061 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9062 Match the new hook interface. Handle SVE addresses.
9063 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9066 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9068 * Makefile.in (OBJS): Add early-remat.o.
9069 * target.def (select_early_remat_modes): New hook.
9070 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9071 * doc/tm.texi: Regenerate.
9072 * targhooks.h (default_select_early_remat_modes): Declare.
9073 * targhooks.c (default_select_early_remat_modes): New function.
9074 * timevar.def (TV_EARLY_REMAT): New timevar.
9075 * passes.def (pass_early_remat): New pass.
9076 * tree-pass.h (make_pass_early_remat): Declare.
9077 * early-remat.c: New file.
9078 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9080 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9082 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9083 Alan Hayward <alan.hayward@arm.com>
9084 David Sherwood <david.sherwood@arm.com>
9086 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9087 vfm1 with a bound_epilog parameter.
9088 (vect_do_peeling): Update calls accordingly, and move the prologue
9089 call earlier in the function. Treat the base bound_epilog as 0 for
9090 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9091 this base when peeling for gaps.
9092 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9093 with fully-masked loops.
9094 (vect_estimate_min_profitable_iters): Handle the single peeled
9095 iteration in that case.
9097 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9098 Alan Hayward <alan.hayward@arm.com>
9099 David Sherwood <david.sherwood@arm.com>
9101 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9102 single-element interleaving even if the size is not a power of 2.
9103 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9104 accesses for single-element interleaving if the group size is
9107 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9108 Alan Hayward <alan.hayward@arm.com>
9109 David Sherwood <david.sherwood@arm.com>
9111 * doc/md.texi (fold_extract_last_@var{m}): Document.
9112 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9113 * optabs.def (fold_extract_last_optab): New optab.
9114 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9115 * internal-fn.c (fold_extract_direct): New macro.
9116 (expand_fold_extract_optab_fn): Likewise.
9117 (direct_fold_extract_optab_supported_p): Likewise.
9118 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9119 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9120 EXTRACT_LAST_REDUCTION.
9121 (get_initial_def_for_reduction): Do not create an initial vector
9122 for EXTRACT_LAST_REDUCTION reductions.
9123 (vectorizable_reduction): Leave the scalar phi in place for
9124 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9125 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9126 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9127 transform phase to vectorizable_condition.
9128 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9130 (vect_finish_stmt_generation): ...here.
9131 (vect_finish_replace_stmt): New function.
9132 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9133 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9135 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9137 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9138 Alan Hayward <alan.hayward@arm.com>
9139 David Sherwood <david.sherwood@arm.com>
9141 * doc/md.texi (extract_last_@var{m}): Document.
9142 * optabs.def (extract_last_optab): New optab.
9143 * internal-fn.def (EXTRACT_LAST): New internal function.
9144 * internal-fn.c (cond_unary_direct): New macro.
9145 (expand_cond_unary_optab_fn): Likewise.
9146 (direct_cond_unary_optab_supported_p): Likewise.
9147 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9148 loops using EXTRACT_LAST.
9149 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9150 (extract_last_<mode>): ...this optab.
9151 (vec_extract<mode><Vel>): Update accordingly.
9153 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9154 Alan Hayward <alan.hayward@arm.com>
9155 David Sherwood <david.sherwood@arm.com>
9157 * target.def (empty_mask_is_expensive): New hook.
9158 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9159 * doc/tm.texi: Regenerate.
9160 * targhooks.h (default_empty_mask_is_expensive): Declare.
9161 * targhooks.c (default_empty_mask_is_expensive): New function.
9162 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9163 if the target says that empty masks are expensive.
9164 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9166 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9168 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9169 Alan Hayward <alan.hayward@arm.com>
9170 David Sherwood <david.sherwood@arm.com>
9172 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9173 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9174 (vect_use_loop_mask_for_alignment_p): New function.
9175 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9176 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9177 niters_skip argument. Make sure that the first niters_skip elements
9178 of the first iteration are inactive.
9179 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9180 Update call to vect_set_loop_masks_directly.
9181 (get_misalign_in_elems): New function, split out from...
9182 (vect_gen_prolog_loop_niters): ...here.
9183 (vect_update_init_of_dr): Take a code argument that specifies whether
9184 the adjustment should be added or subtracted.
9185 (vect_update_init_of_drs): Likewise.
9186 (vect_prepare_for_masked_peels): New function.
9187 (vect_do_peeling): Skip prologue peeling if we're using a mask
9188 instead. Update call to vect_update_inits_of_drs.
9189 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9191 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9192 alignment. Do not include the number of peeled iterations in
9193 the minimum threshold in that case.
9194 (vectorizable_induction): Adjust the start value down by
9195 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9196 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9197 Take the number of skipped iterations into account when calculating
9199 * tree-vect-stmts.c (vect_gen_while_not): New function.
9201 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9202 Alan Hayward <alan.hayward@arm.com>
9203 David Sherwood <david.sherwood@arm.com>
9205 * doc/sourcebuild.texi (vect_fully_masked): Document.
9206 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9208 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9210 (vect_analyze_loop_2): ...here. Don't check the vectorization
9211 factor against the number of loop iterations if the loop is
9214 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9215 Alan Hayward <alan.hayward@arm.com>
9216 David Sherwood <david.sherwood@arm.com>
9218 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9219 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9220 (dump_groups): Update accordingly.
9221 (iv_use::mem_type): New member variable.
9222 (address_p): New function.
9223 (record_use): Add a mem_type argument and initialize the new
9225 (record_group_use): Add a mem_type argument. Use address_p.
9226 Remove obsolete null checks of base_object. Update call to record_use.
9227 (find_interesting_uses_op): Update call to record_group_use.
9228 (find_interesting_uses_cond): Likewise.
9229 (find_interesting_uses_address): Likewise.
9230 (get_mem_type_for_internal_fn): New function.
9231 (find_address_like_use): Likewise.
9232 (find_interesting_uses_stmt): Try find_address_like_use before
9233 calling find_interesting_uses_op.
9234 (addr_offset_valid_p): Use the iv mem_type field as the type
9235 of the addressed memory.
9236 (add_autoinc_candidates): Likewise.
9237 (get_address_cost): Likewise.
9238 (split_small_address_groups_p): Use address_p.
9239 (split_address_groups): Likewise.
9240 (add_iv_candidate_for_use): Likewise.
9241 (autoinc_possible_for_pair): Likewise.
9242 (rewrite_groups): Likewise.
9243 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9244 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9245 (get_alias_ptr_type_for_ptr_address): New function.
9246 (rewrite_use_address): Rewrite address uses in calls that were
9247 identified by find_address_like_use.
9249 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9250 Alan Hayward <alan.hayward@arm.com>
9251 David Sherwood <david.sherwood@arm.com>
9253 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9255 * gimple-expr.h (is_gimple_addressable: Likewise.
9256 * gimple-expr.c (is_gimple_address): Likewise.
9257 * internal-fn.c (expand_call_mem_ref): New function.
9258 (expand_mask_load_optab_fn): Use it.
9259 (expand_mask_store_optab_fn): Likewise.
9261 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9262 Alan Hayward <alan.hayward@arm.com>
9263 David Sherwood <david.sherwood@arm.com>
9265 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9266 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9267 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9268 (cond_umax@var{mode}): Document.
9269 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9270 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9271 (cond_umin_optab, cond_umax_optab): New optabs.
9272 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9273 (COND_IOR, COND_XOR): New internal functions.
9274 * internal-fn.h (get_conditional_internal_fn): Declare.
9275 * internal-fn.c (cond_binary_direct): New macro.
9276 (expand_cond_binary_optab_fn): Likewise.
9277 (direct_cond_binary_optab_supported_p): Likewise.
9278 (get_conditional_internal_fn): New function.
9279 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9280 Cope with reduction statements that are vectorized as calls rather
9282 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9283 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9284 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9285 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9286 (UNSPEC_COND_EOR): New unspecs.
9287 (optab): Add mappings for them.
9288 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9289 (sve_int_op, sve_fp_op): New int attributes.
9291 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9292 Alan Hayward <alan.hayward@arm.com>
9293 David Sherwood <david.sherwood@arm.com>
9295 * optabs.def (while_ult_optab): New optab.
9296 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9297 * internal-fn.def (WHILE_ULT): New internal function.
9298 * internal-fn.h (direct_internal_fn_supported_p): New override
9299 that takes two types as argument.
9300 * internal-fn.c (while_direct): New macro.
9301 (expand_while_optab_fn): New function.
9302 (convert_optab_supported_p): Likewise.
9303 (direct_while_optab_supported_p): New macro.
9304 * wide-int.h (wi::udiv_ceil): New function.
9305 * tree-vectorizer.h (rgroup_masks): New structure.
9306 (vec_loop_masks): New typedef.
9307 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9309 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9310 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9311 (vect_max_vf): New function.
9312 (slpeel_make_loop_iterate_ntimes): Delete.
9313 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9314 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9315 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9316 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9317 internal-fn.h, stor-layout.h and optabs-query.h.
9318 (vect_set_loop_mask): New function.
9319 (add_preheader_seq): Likewise.
9320 (add_header_seq): Likewise.
9321 (interleave_supported_p): Likewise.
9322 (vect_maybe_permute_loop_masks): Likewise.
9323 (vect_set_loop_masks_directly): Likewise.
9324 (vect_set_loop_condition_masked): Likewise.
9325 (vect_set_loop_condition_unmasked): New function, split out from
9326 slpeel_make_loop_iterate_ntimes.
9327 (slpeel_make_loop_iterate_ntimes): Rename to..
9328 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9329 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9330 (vect_do_peeling): Update call accordingly.
9331 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9333 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9334 mask_compare_type, can_fully_mask_p and fully_masked_p.
9335 (release_vec_loop_masks): New function.
9336 (_loop_vec_info): Use it to free the loop masks.
9337 (can_produce_all_loop_masks_p): New function.
9338 (vect_get_max_nscalars_per_iter): Likewise.
9339 (vect_verify_full_masking): Likewise.
9340 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9341 retries, and free the mask rgroups before retrying. Check loop-wide
9342 reasons for disallowing fully-masked loops. Make the final decision
9343 about whether use a fully-masked loop or not.
9344 (vect_estimate_min_profitable_iters): Do not assume that peeling
9345 for the number of iterations will be needed for fully-masked loops.
9346 (vectorizable_reduction): Disable fully-masked loops.
9347 (vectorizable_live_operation): Likewise.
9348 (vect_halve_mask_nunits): New function.
9349 (vect_double_mask_nunits): Likewise.
9350 (vect_record_loop_mask): Likewise.
9351 (vect_get_loop_mask): Likewise.
9352 (vect_transform_loop): Handle the case in which the final loop
9353 iteration might handle a partial vector. Call vect_set_loop_condition
9354 instead of slpeel_make_loop_iterate_ntimes.
9355 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9356 (check_load_store_masking): New function.
9357 (prepare_load_store_mask): Likewise.
9358 (vectorizable_store): Handle fully-masked loops.
9359 (vectorizable_load): Likewise.
9360 (supportable_widening_operation): Use vect_halve_mask_nunits for
9362 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9363 (vect_gen_while): New function.
9364 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9365 (aarch64_uqdec<mode>): New insn.
9367 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9368 Alan Hayward <alan.hayward@arm.com>
9369 David Sherwood <david.sherwood@arm.com>
9371 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9372 (reduc_xor_scal_optab): New optabs.
9373 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9374 (reduc_xor_scal_@var{m}): Document.
9375 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9376 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9378 * fold-const-call.c (fold_const_call): Handle them.
9379 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9380 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9381 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9382 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9383 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9384 (UNSPEC_XORV): New unspecs.
9385 (optab): Add entries for them.
9386 (BITWISEV): New int iterator.
9387 (bit_reduc_op): New int attributes.
9389 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9390 Alan Hayward <alan.hayward@arm.com>
9391 David Sherwood <david.sherwood@arm.com>
9393 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9394 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9395 * optabs.def (vec_shl_insert_optab): New optab.
9396 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9397 (duplicate_and_interleave): Likewise.
9398 * tree-vect-loop.c: Include internal-fn.h.
9399 (neutral_op_for_slp_reduction): New function, split out from
9400 get_initial_defs_for_reduction.
9401 (get_initial_def_for_reduction): Handle option 2 for variable-length
9402 vectors by loading the neutral value into a vector and then shifting
9403 the initial value into element 0.
9404 (get_initial_defs_for_reduction): Replace the code argument with
9405 the neutral value calculated by neutral_op_for_slp_reduction.
9406 Use gimple_build_vector for constant-length vectors.
9407 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9408 but the first group_size elements have a neutral value.
9409 Use duplicate_and_interleave otherwise.
9410 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9411 Update call to get_initial_defs_for_reduction. Handle SLP
9412 reductions for variable-length vectors by creating one vector
9413 result for each scalar result, with the elements associated
9414 with other scalar results stubbed out with the neutral value.
9415 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9416 Require IFN_VEC_SHL_INSERT for double reductions on
9417 variable-length vectors, or SLP reductions that have
9418 a neutral value. Require can_duplicate_and_interleave_p
9419 support for variable-length unchained SLP reductions if there
9420 is no neutral value, such as for MIN/MAX reductions. Also require
9421 the number of vector elements to be a multiple of the number of
9422 SLP statements when doing variable-length unchained SLP reductions.
9423 Update call to vect_create_epilog_for_reduction.
9424 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9425 and remove initial values.
9426 (duplicate_and_interleave): Make public.
9427 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9428 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9430 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9431 Alan Hayward <alan.hayward@arm.com>
9432 David Sherwood <david.sherwood@arm.com>
9434 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9435 (can_duplicate_and_interleave_p): New function.
9436 (vect_get_and_check_slp_defs): Take the vector of statements
9437 rather than just the current one. Remove excess parentheses.
9438 Restriction rejectinon of vect_constant_def and vect_external_def
9439 for variable-length vectors to boolean types, or types for which
9440 can_duplicate_and_interleave_p is false.
9441 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9442 (duplicate_and_interleave): New function.
9443 (vect_get_constant_vectors): Use gimple_build_vector for
9444 constant-length vectors and suitable variable-length constant
9445 vectors. Use duplicate_and_interleave for other variable-length
9446 vectors. Don't defer the update when inserting new statements.
9448 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9449 Alan Hayward <alan.hayward@arm.com>
9450 David Sherwood <david.sherwood@arm.com>
9452 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9453 min_profitable_iters doesn't go negative.
9455 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9456 Alan Hayward <alan.hayward@arm.com>
9457 David Sherwood <david.sherwood@arm.com>
9459 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9460 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9461 * optabs.def (vec_mask_load_lanes_optab): New optab.
9462 (vec_mask_store_lanes_optab): Likewise.
9463 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9464 (MASK_STORE_LANES): Likewise.
9465 * internal-fn.c (mask_load_lanes_direct): New macro.
9466 (mask_store_lanes_direct): Likewise.
9467 (expand_mask_load_optab_fn): Handle masked operations.
9468 (expand_mask_load_lanes_optab_fn): New macro.
9469 (expand_mask_store_optab_fn): Handle masked operations.
9470 (expand_mask_store_lanes_optab_fn): New macro.
9471 (direct_mask_load_lanes_optab_supported_p): Likewise.
9472 (direct_mask_store_lanes_optab_supported_p): Likewise.
9473 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9475 (vect_load_lanes_supported): Likewise.
9476 * tree-vect-data-refs.c (strip_conversion): New function.
9477 (can_group_stmts_p): Likewise.
9478 (vect_analyze_data_ref_accesses): Use it instead of checking
9479 for a pair of assignments.
9480 (vect_store_lanes_supported): Take a masked_p parameter.
9481 (vect_load_lanes_supported): Likewise.
9482 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9483 vect_store_lanes_supported and vect_load_lanes_supported.
9484 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9485 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9486 parameter. Don't allow gaps for masked accesses.
9487 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9488 and vect_load_lanes_supported.
9489 (get_load_store_type): Take a masked_p parameter and update
9490 call to get_group_load_store_type.
9491 (vectorizable_store): Update call to get_load_store_type.
9492 Handle IFN_MASK_STORE_LANES.
9493 (vectorizable_load): Update call to get_load_store_type.
9494 Handle IFN_MASK_LOAD_LANES.
9496 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9497 Alan Hayward <alan.hayward@arm.com>
9498 David Sherwood <david.sherwood@arm.com>
9500 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9502 * config/aarch64/aarch64-protos.h
9503 (aarch64_sve_struct_memory_operand_p): Declare.
9504 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9505 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9506 (VPRED, vpred): Handle SVE structure modes.
9507 * config/aarch64/constraints.md (Utx): New constraint.
9508 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9509 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9510 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9511 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9512 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9513 structure modes. Split into pieces after RA.
9514 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9515 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9517 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9518 SVE structure modes.
9519 (aarch64_classify_address): Likewise.
9520 (sizetochar): Move earlier in file.
9521 (aarch64_print_operand): Handle SVE register lists.
9522 (aarch64_array_mode): New function.
9523 (aarch64_sve_struct_memory_operand_p): Likewise.
9524 (TARGET_ARRAY_MODE): Redefine.
9526 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9527 Alan Hayward <alan.hayward@arm.com>
9528 David Sherwood <david.sherwood@arm.com>
9530 * target.def (array_mode): New target hook.
9531 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9532 * doc/tm.texi: Regenerate.
9533 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9534 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9535 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9537 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9540 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9541 Alan Hayward <alan.hayward@arm.com>
9542 David Sherwood <david.sherwood@arm.com>
9544 * fold-const.c (fold_binary_loc): Check the argument types
9545 rather than the result type when testing for a vector operation.
9547 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9549 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9550 * doc/tm.texi: Regenerate.
9552 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9553 Alan Hayward <alan.hayward@arm.com>
9554 David Sherwood <david.sherwood@arm.com>
9556 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9557 (sve): Document new AArch64 extension.
9558 * doc/md.texi (w): Extend the description of the AArch64
9559 constraint to include SVE vectors.
9560 (Upl, Upa): Document new AArch64 predicate constraints.
9561 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9563 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9564 (msve-vector-bits=): New option.
9565 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9566 SVE when these are disabled.
9567 (sve): New extension.
9568 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9569 modes. Adjust their number of units based on aarch64_sve_vg.
9570 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9571 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9572 aarch64_addr_query_type.
9573 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9574 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9575 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9576 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9577 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9578 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9579 (aarch64_simd_imm_zero_p): Delete.
9580 (aarch64_check_zero_based_sve_index_immediate): Declare.
9581 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9582 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9583 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9584 (aarch64_sve_float_mul_immediate_p): Likewise.
9585 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9587 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9588 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9589 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9590 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9591 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9592 (aarch64_regmode_natural_size): Likewise.
9593 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9594 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9596 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9597 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9598 for VG and the SVE predicate registers.
9599 (V_ALIASES): Add a "z"-prefixed alias.
9600 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9601 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9602 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9603 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9604 (REG_CLASS_NAMES): Add entries for them.
9605 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9606 and the predicate registers.
9607 (aarch64_sve_vg): Declare.
9608 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9609 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9610 (REGMODE_NATURAL_SIZE): Define.
9611 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9613 * config/aarch64/aarch64.c: Include cfgrtl.h.
9614 (simd_immediate_info): Add a constructor for series vectors,
9615 and an associated step field.
9616 (aarch64_sve_vg): New variable.
9617 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9618 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9619 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9620 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9621 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9622 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9623 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9624 (aarch64_get_mask_mode): New functions.
9625 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9626 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9627 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9628 predicate modes and predicate registers. Explicitly restrict
9629 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9630 to store a vector mode if it is recognized by
9631 aarch64_classify_vector_mode.
9632 (aarch64_regmode_natural_size): New function.
9633 (aarch64_hard_regno_caller_save_mode): Return the original mode
9635 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9636 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9637 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9638 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9640 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9641 does not overlap dest if the function is frame-related. Handle
9643 (aarch64_split_add_offset): New function.
9644 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9645 them aarch64_add_offset.
9646 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9647 and update call to aarch64_sub_sp.
9648 (aarch64_add_cfa_expression): New function.
9649 (aarch64_expand_prologue): Pass extra temporary registers to the
9650 functions above. Handle the case in which we need to emit new
9651 DW_CFA_expressions for registers that were originally saved
9652 relative to the stack pointer, but now have to be expressed
9653 relative to the frame pointer.
9654 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9656 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9657 IP0 and IP1 values for SVE frames.
9658 (aarch64_expand_vec_series): New function.
9659 (aarch64_expand_sve_widened_duplicate): Likewise.
9660 (aarch64_expand_sve_const_vector): Likewise.
9661 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9662 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9663 into the register, rather than emitting a SET directly.
9664 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9665 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9666 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9667 (offset_9bit_signed_scaled_p): New functions.
9668 (aarch64_replicate_bitmask_imm): New function.
9669 (aarch64_bitmask_imm): Use it.
9670 (aarch64_cannot_force_const_mem): Reject expressions involving
9671 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9672 (aarch64_classify_index): Handle SVE indices, by requiring
9673 a plain register index with a scale that matches the element size.
9674 (aarch64_classify_address): Handle SVE addresses. Assert that
9675 the mode of the address is VOIDmode or an integer mode.
9676 Update call to aarch64_classify_symbol.
9677 (aarch64_classify_symbolic_expression): Update call to
9678 aarch64_classify_symbol.
9679 (aarch64_const_vec_all_in_range_p): New function.
9680 (aarch64_print_vector_float_operand): Likewise.
9681 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9682 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9683 and the FP immediates 1.0 and 0.5.
9684 (aarch64_print_address_internal): Handle SVE addresses.
9685 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9686 (aarch64_regno_regclass): Handle predicate registers.
9687 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9689 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9690 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9691 (aarch64_convert_sve_vector_bits): New function.
9692 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9693 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9695 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9696 Handle SVE vector and predicate modes. Accept VL-based constants
9697 that need only one temporary register, and VL offsets that require
9698 no temporary registers.
9699 (aarch64_conditional_register_usage): Mark the predicate registers
9700 as fixed if SVE isn't available.
9701 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9702 Return true for SVE vector and predicate modes.
9703 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9704 rather than an unsigned int. Handle SVE modes.
9705 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9707 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9709 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9710 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9711 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9712 (aarch64_sve_float_mul_immediate_p): New functions.
9713 (aarch64_sve_valid_immediate): New function.
9714 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9715 Explicitly reject structure modes. Check for INDEX constants.
9716 Handle PTRUE and PFALSE constants.
9717 (aarch64_check_zero_based_sve_index_immediate): New function.
9718 (aarch64_simd_imm_zero_p): Delete.
9719 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9720 vector modes. Accept constants in the range of CNT[BHWD].
9721 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9722 ask for an Advanced SIMD mode.
9723 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9724 (aarch64_simd_vector_alignment): Handle SVE predicates.
9725 (aarch64_vectorize_preferred_vector_alignment): New function.
9726 (aarch64_simd_vector_alignment_reachable): Use it instead of
9728 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9729 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9731 (MAX_VECT_LEN): Delete.
9732 (expand_vec_perm_d): Add a vec_flags field.
9733 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9734 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9735 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9737 (aarch64_evpc_rev): Rename to...
9738 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9739 (aarch64_evpc_rev_global): New function.
9740 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9741 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9743 (aarch64_evpc_sve_tbl): New function.
9744 (aarch64_expand_vec_perm_const_1): Update after rename of
9745 aarch64_evpc_rev. Handle SVE permutes too, trying
9746 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9747 than aarch64_evpc_tbl.
9748 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9749 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9750 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9751 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9752 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9753 (aarch64_expand_sve_vcond): New functions.
9754 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9755 of aarch64_vector_mode_p.
9756 (aarch64_dwarf_poly_indeterminate_value): New function.
9757 (aarch64_compute_pressure_classes): Likewise.
9758 (aarch64_can_change_mode_class): Likewise.
9759 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9760 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9761 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9762 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9763 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9764 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9765 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9766 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9768 (Dn, Dl, Dr): Accept const as well as const_vector.
9769 (Dz): Likewise. Compare against CONST0_RTX.
9770 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9771 of "vector" where appropriate.
9772 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9773 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9774 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9775 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9776 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9777 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9778 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9779 (v_int_equiv): Extend to SVE modes.
9780 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9782 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9783 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9784 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9785 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9786 (SVE_COND_FP_CMP): New int iterators.
9787 (perm_hilo): Handle the new unpack unspecs.
9788 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9790 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9791 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9792 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9793 (aarch64_equality_operator, aarch64_constant_vector_operand)
9794 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9795 (aarch64_sve_nonimmediate_operand): Likewise.
9796 (aarch64_sve_general_operand): Likewise.
9797 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9798 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9799 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9800 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9801 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9802 (aarch64_sve_float_arith_immediate): Likewise.
9803 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9804 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9805 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9806 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9807 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9808 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9809 (aarch64_sve_float_arith_operand): Likewise.
9810 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9811 (aarch64_sve_float_mul_operand): Likewise.
9812 (aarch64_sve_vec_perm_operand): Likewise.
9813 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9814 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9815 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9816 as well as const_vector.
9817 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9818 in file. Use CONST0_RTX and CONSTM1_RTX.
9819 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9820 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9821 Use aarch64_simd_imm_zero.
9822 * config/aarch64/aarch64-sve.md: New file.
9823 * config/aarch64/aarch64.md: Include it.
9824 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9825 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9826 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9827 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9828 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9829 (sve): New attribute.
9830 (enabled): Disable instructions with the sve attribute unless
9832 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9833 aarch64_expand_mov_immediate.
9834 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9835 CNT[BHSD] immediates.
9836 (movti): Split CONST_POLY_INT moves into two halves.
9837 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9838 Split additions that need a temporary here if the destination
9839 is the stack pointer.
9840 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9841 (*add<mode>3_poly_1): New instruction.
9842 (set_clobber_cc): New expander.
9844 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9846 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9847 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9848 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9849 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9850 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9851 Change innermode from fixed_mode_size to machine_mode.
9852 (simplify_subreg): Update call accordingly. Handle a constant-sized
9853 subreg of a variable-length CONST_VECTOR.
9855 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9856 Alan Hayward <alan.hayward@arm.com>
9857 David Sherwood <david.sherwood@arm.com>
9859 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9860 (add_offset_to_base): New function, split out from...
9861 (create_mem_ref): ...here. When handling a scale other than 1,
9862 check first whether the address is valid without the offset.
9863 Add it into the base if so, leaving the index and scale as-is.
9865 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9868 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9869 fold_for_warn before checking if arg2 is INTEGER_CST.
9871 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9873 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9874 (store_multiple_operation): Delete.
9875 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9876 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9877 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9878 guarded by TARGET_STRING.
9879 (rs6000_output_load_multiple): Delete.
9880 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9881 OPTION_MASK_STRING / TARGET_STRING handling.
9882 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9883 (const rs6000_opt_masks) <"string">: Change mask to 0.
9884 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9885 (MASK_STRING): Delete.
9886 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9888 (load_multiple): Delete.
9895 (store_multiple): Delete.
9902 (movmemsi_8reg): Delete.
9903 (corresponding unnamed define_insn): Delete.
9904 (movmemsi_6reg): Delete.
9905 (corresponding unnamed define_insn): Delete.
9906 (movmemsi_4reg): Delete.
9907 (corresponding unnamed define_insn): Delete.
9908 (movmemsi_2reg): Delete.
9909 (corresponding unnamed define_insn): Delete.
9910 (movmemsi_1reg): Delete.
9911 (corresponding unnamed define_insn): Delete.
9912 * config/rs6000/rs6000.opt (mno-string): New.
9913 (mstring): Replace by deprecation warning stub.
9914 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
9916 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9918 * regrename.c (regrename_do_replace): If replacing the same
9919 reg multiple times, try to reuse last created gen_raw_REG.
9922 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
9923 main to workaround a bug in GDB.
9925 2018-01-12 Tom de Vries <tom@codesourcery.com>
9928 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
9930 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
9932 PR rtl-optimization/80481
9933 * ira-color.c (get_cap_member): New function.
9934 (allocnos_conflict_by_live_ranges_p): Use it.
9935 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
9936 (setup_slot_coalesced_allocno_live_ranges): Ditto.
9938 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
9941 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
9942 (*saddl_se_1): Ditto.
9944 (*ssubl_se_1): Ditto.
9946 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9948 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
9949 rather than wi::to_widest for DR_INITs.
9950 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
9951 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
9952 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
9954 (vect_analyze_group_access_1): Note that here.
9956 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9958 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
9959 polynomial type sizes.
9961 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
9963 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
9964 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
9965 (gimple_add_tmp_var): Likewise.
9967 2018-01-12 Martin Liska <mliska@suse.cz>
9969 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
9970 (gimple_alloc_sizes): Likewise.
9971 (dump_gimple_statistics): Use PRIu64 in printf format.
9972 * gimple.h: Change uint64_t to int.
9974 2018-01-12 Martin Liska <mliska@suse.cz>
9976 * tree-core.h: Use uint64_t instead of int.
9977 * tree.c (tree_node_counts): Likewise.
9978 (tree_node_sizes): Likewise.
9979 (dump_tree_statistics): Use PRIu64 in printf format.
9981 2018-01-12 Martin Liska <mliska@suse.cz>
9983 * Makefile.in: As qsort_chk is implemented in vec.c, add
9984 vec.o to linkage of gencfn-macros.
9985 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
9986 passing the info to record_node_allocation_statistics.
9987 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
9989 * ggc-common.c (struct ggc_usage): Add operator== and use
9990 it in operator< and compare function.
9991 * mem-stats.h (struct mem_usage): Likewise.
9992 * vec.c (struct vec_usage): Remove operator< and compare
9993 function. Can be simply inherited.
9995 2018-01-12 Martin Jambor <mjambor@suse.cz>
9998 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
9999 * tree-ssa-math-opts.c: Include domwalk.h.
10000 (convert_mult_to_fma_1): New function.
10001 (fma_transformation_info): New type.
10002 (fma_deferring_state): Likewise.
10003 (cancel_fma_deferring): New function.
10004 (result_of_phi): Likewise.
10005 (last_fma_candidate_feeds_initial_phi): Likewise.
10006 (convert_mult_to_fma): Added deferring logic, split actual
10007 transformation to convert_mult_to_fma_1.
10008 (math_opts_dom_walker): New type.
10009 (math_opts_dom_walker::after_dom_children): New method, body moved
10010 here from pass_optimize_widening_mul::execute, added deferring logic
10012 (pass_optimize_widening_mul::execute): Moved most of code to
10013 math_opts_dom_walker::after_dom_children.
10014 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10015 * config/i386/i386.c (ix86_option_override_internal): Added
10016 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10018 2018-01-12 Richard Biener <rguenther@suse.de>
10021 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10022 inline instance vars.
10024 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10027 * config/rx/rx.c (rx_is_restricted_memory_address):
10028 Handle SUBREG case.
10030 2018-01-12 Richard Biener <rguenther@suse.de>
10032 PR tree-optimization/80846
10033 * target.def (split_reduction): New target hook.
10034 * targhooks.c (default_split_reduction): New function.
10035 * targhooks.h (default_split_reduction): Declare.
10036 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10037 target requests first reduce vectors by combining low and high
10039 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10040 (get_vectype_for_scalar_type_and_size): Export.
10041 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10042 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10043 * doc/tm.texi: Regenerate.
10044 * config/i386/i386.c (ix86_split_reduction): Implement
10045 TARGET_VECTORIZE_SPLIT_REDUCTION.
10047 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10050 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10051 in PIC mode except for TARGET_VXWORKS_RTP.
10052 * config/sparc/sparc.c: Include cfgrtl.h.
10053 (TARGET_INIT_PIC_REG): Define.
10054 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10055 (sparc_pic_register_p): New predicate.
10056 (sparc_legitimate_address_p): Use it.
10057 (sparc_legitimize_pic_address): Likewise.
10058 (sparc_delegitimize_address): Likewise.
10059 (sparc_mode_dependent_address_p): Likewise.
10060 (gen_load_pcrel_sym): Remove 4th parameter.
10061 (load_got_register): Adjust call to above. Remove obsolete stuff.
10062 (sparc_expand_prologue): Do not call load_got_register here.
10063 (sparc_flat_expand_prologue): Likewise.
10064 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10065 (sparc_use_pseudo_pic_reg): New function.
10066 (sparc_init_pic_reg): Likewise.
10067 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10068 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10070 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10072 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10073 Add item for branch_cost.
10075 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10077 PR rtl-optimization/83565
10078 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10079 not extend the result to a larger mode for rotate operations.
10080 (num_sign_bit_copies1): Likewise.
10082 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10085 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10087 Use values-Xc.o for -pedantic.
10088 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10090 2018-01-12 Martin Liska <mliska@suse.cz>
10093 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10095 (possible_polymorphic_call_targets): Use it.
10096 (ipa_devirt): Likewise.
10098 2018-01-12 Martin Liska <mliska@suse.cz>
10100 * profile-count.h (enum profile_quality): Use 0 as invalid
10101 enum value of profile_quality.
10103 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10105 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10106 -mext-string options.
10108 2018-01-12 Richard Biener <rguenther@suse.de>
10110 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10111 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10112 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10114 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10116 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10118 * configure.ac (--with-long-double-format): Add support for the
10119 configuration option to change the default long double format on
10121 * config.gcc (powerpc*-linux*-*): Likewise.
10122 * configure: Regenerate.
10123 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10124 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10125 used without modification.
10127 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10129 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10130 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10131 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10132 MISC_BUILTIN_SPEC_BARRIER.
10133 (rs6000_init_builtins): Likewise.
10134 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10136 (speculation_barrier): New define_insn.
10137 * doc/extend.texi: Document __builtin_speculation_barrier.
10139 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10142 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10143 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10144 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10146 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10147 integral modes instead of "ss" and "sd".
10148 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10149 vectors with 32-bit and 64-bit elements.
10150 (vecdupssescalarmodesuffix): New mode attribute.
10151 (vec_dup<mode>): Use it.
10153 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10156 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10157 frame if argument is passed on stack.
10159 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10162 * ree.c (combine_reaching_defs): Optimize also
10163 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10164 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10166 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10168 PR middle-end/83189
10169 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10171 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10173 PR middle-end/83718
10174 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10175 after they are computed.
10177 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10179 PR tree-optimization/83695
10180 * gimple-loop-linterchange.cc
10181 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10182 reset cached scev information after interchange.
10183 (pass_linterchange::execute): Remove call to scev_reset_htab.
10185 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10187 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10188 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10189 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10190 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10191 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10192 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10193 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10194 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10195 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10196 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10197 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10198 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10199 (V_lane_reg): Likewise.
10200 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10202 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10203 (vfmal_lane_low<mode>_intrinsic,
10204 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10205 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10206 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10207 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10208 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10209 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10211 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10213 * config/arm/arm-cpus.in (fp16fml): New feature.
10214 (ALL_SIMD): Add fp16fml.
10215 (armv8.2-a): Add fp16fml as an option.
10216 (armv8.3-a): Likewise.
10217 (armv8.4-a): Add fp16fml as part of fp16.
10218 * config/arm/arm.h (TARGET_FP16FML): Define.
10219 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10221 * config/arm/arm-modes.def (V2HF): Define.
10222 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10223 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10224 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10225 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10226 vfmsl_low, vfmsl_high): New set of builtins.
10227 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10228 (vfml_op): New code attribute.
10229 (VFMLHALVES): New int iterator.
10230 (VFML, VFMLSEL): New mode attributes.
10231 (V_reg): Define mapping for V2HF.
10232 (V_hi, V_lo): New mode attributes.
10233 (VF_constraint): Likewise.
10234 (vfml_half, vfml_half_selector): New int attributes.
10235 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10237 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10238 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10240 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10241 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10242 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10243 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10245 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10246 Document new effective target and option set.
10248 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10250 * config/arm/arm-cpus.in (armv8_4): New feature.
10251 (ARMv8_4a): New fgroup.
10252 (armv8.4-a): New arch.
10253 * config/arm/arm-tables.opt: Regenerate.
10254 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10255 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10256 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10257 Add matching rules for -march=armv8.4-a and extensions.
10258 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10260 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10263 * config/rx/rx.md (BW): New mode attribute.
10264 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10266 2018-01-11 Richard Biener <rguenther@suse.de>
10268 PR tree-optimization/83435
10269 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10270 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10271 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10273 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10274 Alan Hayward <alan.hayward@arm.com>
10275 David Sherwood <david.sherwood@arm.com>
10277 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10279 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10280 (aarch64_print_address_internal): Use it to check for a zero offset.
10282 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10283 Alan Hayward <alan.hayward@arm.com>
10284 David Sherwood <david.sherwood@arm.com>
10286 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10287 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10288 Return a poly_int64 rather than a HOST_WIDE_INT.
10289 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10290 rather than a HOST_WIDE_INT.
10291 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10292 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10293 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10294 final_offset from HOST_WIDE_INT to poly_int64.
10295 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10296 to_constant when getting the number of units in an Advanced SIMD
10298 (aarch64_builtin_vectorized_function): Check for a constant number
10300 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10302 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10303 attribute instead of GET_MODE_NUNITS.
10304 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10305 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10306 GET_MODE_SIZE for fixed-size registers.
10307 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10308 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10309 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10310 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10311 (aarch64_print_operand, aarch64_print_address_internal)
10312 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10313 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10314 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10315 Handle polynomial GET_MODE_SIZE.
10316 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10317 wider than SImode without modification.
10318 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10319 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10320 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10321 passing and returning SVE modes.
10322 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10323 rather than GEN_INT.
10324 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10325 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10326 (aarch64_allocate_and_probe_stack_space): Likewise.
10327 (aarch64_layout_frame): Cope with polynomial offsets.
10328 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10329 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10330 polynomial offsets.
10331 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10332 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10333 poly_int64 rather than a HOST_WIDE_INT.
10334 (aarch64_get_separate_components, aarch64_process_components)
10335 (aarch64_expand_prologue, aarch64_expand_epilogue)
10336 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10337 (aarch64_anchor_offset): New function, split out from...
10338 (aarch64_legitimize_address): ...here.
10339 (aarch64_builtin_vectorization_cost): Handle polynomial
10340 TYPE_VECTOR_SUBPARTS.
10341 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10343 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10344 number of elements from the PARALLEL rather than the mode.
10345 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10346 rather than GET_MODE_BITSIZE.
10347 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10348 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10349 (aarch64_expand_vec_perm_const_1): Handle polynomial
10350 d->perm.length () and d->perm elements.
10351 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10352 Apply to_constant to d->perm elements.
10353 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10354 polynomial CONST_VECTOR_NUNITS.
10355 (aarch64_move_pointer): Take amount as a poly_int64 rather
10357 (aarch64_progress_pointer): Avoid temporary variable.
10358 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10359 the mode attribute instead of GET_MODE.
10361 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10362 Alan Hayward <alan.hayward@arm.com>
10363 David Sherwood <david.sherwood@arm.com>
10365 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10366 x exists before using it.
10367 (aarch64_add_constant_internal): Rename to...
10368 (aarch64_add_offset_1): ...this. Replace regnum with separate
10369 src and dest rtxes. Handle the case in which they're different,
10370 including when the offset is zero. Replace scratchreg with an rtx.
10371 Use 2 additions if there is no spare register into which we can
10372 move a 16-bit constant.
10373 (aarch64_add_constant): Delete.
10374 (aarch64_add_offset): Replace reg with separate src and dest
10375 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10376 Use aarch64_add_offset_1.
10377 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10378 an rtx rather than an int. Take the delta as a poly_int64
10379 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10380 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10381 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10382 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10383 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10384 and aarch64_add_sp.
10385 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10386 aarch64_add_constant.
10388 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10390 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10391 Use scalar_float_mode.
10393 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10395 * config/aarch64/aarch64-simd.md
10396 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10397 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10398 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10399 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10400 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10401 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10402 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10403 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10404 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10405 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10407 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10410 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10411 targ_options->x_arm_arch_string is non NULL.
10413 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10415 * config/aarch64/aarch64.h
10416 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10418 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10421 * expmed.c (emit_store_flag_force): Swap if const op0
10422 and change VOIDmode to mode of op0.
10424 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10426 PR rtl-optimization/83761
10427 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10428 than bytes to mode_for_size.
10430 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10432 PR middle-end/83189
10433 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10434 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10437 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10439 PR middle-end/83575
10440 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10441 when in layout mode.
10442 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10443 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10446 2018-01-10 Michael Collison <michael.collison@arm.com>
10448 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10449 * config/aarch64/aarch64-option-extension.def: Add
10450 AARCH64_OPT_EXTENSION of 'fp16fml'.
10451 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10452 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10453 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10454 * config/aarch64/constraints.md (Ui7): New constraint.
10455 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10456 (VFMLA_SEL_W): Ditto.
10459 (VFMLA16_LOW): New int iterator.
10460 (VFMLA16_HIGH): Ditto.
10461 (UNSPEC_FMLAL): New unspec.
10462 (UNSPEC_FMLSL): Ditto.
10463 (UNSPEC_FMLAL2): Ditto.
10464 (UNSPEC_FMLSL2): Ditto.
10465 (f16mac): New code attribute.
10466 * config/aarch64/aarch64-simd-builtins.def
10467 (aarch64_fmlal_lowv2sf): Ditto.
10468 (aarch64_fmlsl_lowv2sf): Ditto.
10469 (aarch64_fmlalq_lowv4sf): Ditto.
10470 (aarch64_fmlslq_lowv4sf): Ditto.
10471 (aarch64_fmlal_highv2sf): Ditto.
10472 (aarch64_fmlsl_highv2sf): Ditto.
10473 (aarch64_fmlalq_highv4sf): Ditto.
10474 (aarch64_fmlslq_highv4sf): Ditto.
10475 (aarch64_fmlal_lane_lowv2sf): Ditto.
10476 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10477 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10478 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10479 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10480 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10481 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10482 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10483 (aarch64_fmlal_lane_highv2sf): Ditto.
10484 (aarch64_fmlsl_lane_highv2sf): Ditto.
10485 (aarch64_fmlal_laneq_highv2sf): Ditto.
10486 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10487 (aarch64_fmlalq_lane_highv4sf): Ditto.
10488 (aarch64_fmlsl_lane_highv4sf): Ditto.
10489 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10490 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10491 * config/aarch64/aarch64-simd.md:
10492 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10493 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10494 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10495 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10496 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10497 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10498 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10499 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10500 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10501 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10502 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10503 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10504 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10505 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10506 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10507 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10508 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10509 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10510 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10511 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10512 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10513 (vfmlsl_low_u32): Ditto.
10514 (vfmlalq_low_u32): Ditto.
10515 (vfmlslq_low_u32): Ditto.
10516 (vfmlal_high_u32): Ditto.
10517 (vfmlsl_high_u32): Ditto.
10518 (vfmlalq_high_u32): Ditto.
10519 (vfmlslq_high_u32): Ditto.
10520 (vfmlal_lane_low_u32): Ditto.
10521 (vfmlsl_lane_low_u32): Ditto.
10522 (vfmlal_laneq_low_u32): Ditto.
10523 (vfmlsl_laneq_low_u32): Ditto.
10524 (vfmlalq_lane_low_u32): Ditto.
10525 (vfmlslq_lane_low_u32): Ditto.
10526 (vfmlalq_laneq_low_u32): Ditto.
10527 (vfmlslq_laneq_low_u32): Ditto.
10528 (vfmlal_lane_high_u32): Ditto.
10529 (vfmlsl_lane_high_u32): Ditto.
10530 (vfmlal_laneq_high_u32): Ditto.
10531 (vfmlsl_laneq_high_u32): Ditto.
10532 (vfmlalq_lane_high_u32): Ditto.
10533 (vfmlslq_lane_high_u32): Ditto.
10534 (vfmlalq_laneq_high_u32): Ditto.
10535 (vfmlslq_laneq_high_u32): Ditto.
10536 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10537 (AARCH64_FL_FOR_ARCH8_4): New.
10538 (AARCH64_ISA_F16FML): New ISA flag.
10539 (TARGET_F16FML): New feature flag for fp16fml.
10540 (doc/invoke.texi): Document new fp16fml option.
10542 2018-01-10 Michael Collison <michael.collison@arm.com>
10544 * config/aarch64/aarch64-builtins.c:
10545 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10546 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10547 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10548 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10549 (AARCH64_ISA_SHA3): New ISA flag.
10550 (TARGET_SHA3): New feature flag for sha3.
10551 * config/aarch64/iterators.md (sha512_op): New int attribute.
10552 (CRYPTO_SHA512): New int iterator.
10553 (UNSPEC_SHA512H): New unspec.
10554 (UNSPEC_SHA512H2): Ditto.
10555 (UNSPEC_SHA512SU0): Ditto.
10556 (UNSPEC_SHA512SU1): Ditto.
10557 * config/aarch64/aarch64-simd-builtins.def
10558 (aarch64_crypto_sha512hqv2di): New builtin.
10559 (aarch64_crypto_sha512h2qv2di): Ditto.
10560 (aarch64_crypto_sha512su0qv2di): Ditto.
10561 (aarch64_crypto_sha512su1qv2di): Ditto.
10562 (aarch64_eor3qv8hi): Ditto.
10563 (aarch64_rax1qv2di): Ditto.
10564 (aarch64_xarqv2di): Ditto.
10565 (aarch64_bcaxqv8hi): Ditto.
10566 * config/aarch64/aarch64-simd.md:
10567 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10568 (aarch64_crypto_sha512su0qv2di): Ditto.
10569 (aarch64_crypto_sha512su1qv2di): Ditto.
10570 (aarch64_eor3qv8hi): Ditto.
10571 (aarch64_rax1qv2di): Ditto.
10572 (aarch64_xarqv2di): Ditto.
10573 (aarch64_bcaxqv8hi): Ditto.
10574 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10575 (vsha512h2q_u64): Ditto.
10576 (vsha512su0q_u64): Ditto.
10577 (vsha512su1q_u64): Ditto.
10578 (veor3q_u16): Ditto.
10579 (vrax1q_u64): Ditto.
10580 (vxarq_u64): Ditto.
10581 (vbcaxq_u16): Ditto.
10582 * config/arm/types.md (crypto_sha512): New type attribute.
10583 (crypto_sha3): Ditto.
10584 (doc/invoke.texi): Document new sha3 option.
10586 2018-01-10 Michael Collison <michael.collison@arm.com>
10588 * config/aarch64/aarch64-builtins.c:
10589 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10590 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10591 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10592 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10593 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10594 (AARCH64_ISA_SM4): New ISA flag.
10595 (TARGET_SM4): New feature flag for sm4.
10596 * config/aarch64/aarch64-simd-builtins.def
10597 (aarch64_sm3ss1qv4si): Ditto.
10598 (aarch64_sm3tt1aq4si): Ditto.
10599 (aarch64_sm3tt1bq4si): Ditto.
10600 (aarch64_sm3tt2aq4si): Ditto.
10601 (aarch64_sm3tt2bq4si): Ditto.
10602 (aarch64_sm3partw1qv4si): Ditto.
10603 (aarch64_sm3partw2qv4si): Ditto.
10604 (aarch64_sm4eqv4si): Ditto.
10605 (aarch64_sm4ekeyqv4si): Ditto.
10606 * config/aarch64/aarch64-simd.md:
10607 (aarch64_sm3ss1qv4si): Ditto.
10608 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10609 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10610 (aarch64_sm4eqv4si): Ditto.
10611 (aarch64_sm4ekeyqv4si): Ditto.
10612 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10613 (sm3part_op): Ditto.
10614 (CRYPTO_SM3TT): Ditto.
10615 (CRYPTO_SM3PART): Ditto.
10616 (UNSPEC_SM3SS1): New unspec.
10617 (UNSPEC_SM3TT1A): Ditto.
10618 (UNSPEC_SM3TT1B): Ditto.
10619 (UNSPEC_SM3TT2A): Ditto.
10620 (UNSPEC_SM3TT2B): Ditto.
10621 (UNSPEC_SM3PARTW1): Ditto.
10622 (UNSPEC_SM3PARTW2): Ditto.
10623 (UNSPEC_SM4E): Ditto.
10624 (UNSPEC_SM4EKEY): Ditto.
10625 * config/aarch64/constraints.md (Ui2): New constraint.
10626 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10627 * config/arm/types.md (crypto_sm3): New type attribute.
10628 (crypto_sm4): Ditto.
10629 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10630 (vsm3tt1aq_u32): Ditto.
10631 (vsm3tt1bq_u32): Ditto.
10632 (vsm3tt2aq_u32): Ditto.
10633 (vsm3tt2bq_u32): Ditto.
10634 (vsm3partw1q_u32): Ditto.
10635 (vsm3partw2q_u32): Ditto.
10636 (vsm4eq_u32): Ditto.
10637 (vsm4ekeyq_u32): Ditto.
10638 (doc/invoke.texi): Document new sm4 option.
10640 2018-01-10 Michael Collison <michael.collison@arm.com>
10642 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10643 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10644 (AARCH64_FL_FOR_ARCH8_4): New.
10645 (AARCH64_FL_V8_4): New flag.
10646 (doc/invoke.texi): Document new armv8.4-a option.
10648 2018-01-10 Michael Collison <michael.collison@arm.com>
10650 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10651 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10652 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10653 * config/aarch64/aarch64-option-extension.def: Add
10654 AARCH64_OPT_EXTENSION of 'sha2'.
10655 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10656 (crypto): Disable sha2 and aes if crypto disabled.
10657 (crypto): Enable aes and sha2 if enabled.
10658 (simd): Disable sha2 and aes if simd disabled.
10659 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10661 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10662 (TARGET_SHA2): New feature flag for sha2.
10663 (TARGET_AES): New feature flag for aes.
10664 * config/aarch64/aarch64-simd.md:
10665 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10666 conditional on TARGET_AES.
10667 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10668 (aarch64_crypto_sha1hsi): Make pattern conditional
10670 (aarch64_crypto_sha1hv4si): Ditto.
10671 (aarch64_be_crypto_sha1hv4si): Ditto.
10672 (aarch64_crypto_sha1su1v4si): Ditto.
10673 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10674 (aarch64_crypto_sha1su0v4si): Ditto.
10675 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10676 (aarch64_crypto_sha256su0v4si): Ditto.
10677 (aarch64_crypto_sha256su1v4si): Ditto.
10678 (doc/invoke.texi): Document new aes and sha2 options.
10680 2018-01-10 Martin Sebor <msebor@redhat.com>
10682 PR tree-optimization/83781
10683 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10686 2018-01-11 Martin Sebor <msebor@gmail.com>
10687 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10689 PR tree-optimization/83501
10690 PR tree-optimization/81703
10692 * tree-ssa-strlen.c (get_string_cst): Rename...
10693 (get_string_len): ...to this. Handle global constants.
10694 (handle_char_store): Adjust.
10696 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10697 Jim Wilson <jimw@sifive.com>
10699 * config/riscv/riscv-protos.h (riscv_output_return): New.
10700 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10701 (riscv_attribute_table, riscv_output_return),
10702 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10703 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10704 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10706 (riscv_expand_prologue): Add early return for naked function.
10707 (riscv_expand_epilogue): Likewise.
10708 (riscv_function_ok_for_sibcall): Return false for naked function.
10709 (riscv_set_current_function): New.
10710 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10711 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10712 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10713 * doc/extend.texi (RISC-V Function Attributes): New.
10715 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10717 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10718 check for 128-bit long double before checking TCmode.
10719 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10720 128-bit long doubles before checking TFmode or TCmode.
10721 (FLOAT128_IBM_P): Likewise.
10723 2018-01-10 Martin Sebor <msebor@redhat.com>
10725 PR tree-optimization/83671
10726 * builtins.c (c_strlen): Unconditionally return zero for the empty
10728 Use -Warray-bounds for warnings.
10729 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10730 for non-constant array indices with COMPONENT_REF, arrays of
10731 arrays, and pointers to arrays.
10732 (gimple_fold_builtin_strlen): Determine and set length range for
10733 non-constant character arrays.
10735 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10737 PR middle-end/81897
10738 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10741 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10743 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10745 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10748 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10749 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10750 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10751 indexed_or_indirect_operand predicate.
10752 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10753 (*vsx_le_perm_load_v8hi): Likewise.
10754 (*vsx_le_perm_load_v16qi): Likewise.
10755 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10756 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10757 (*vsx_le_perm_store_v8hi): Likewise.
10758 (*vsx_le_perm_store_v16qi): Likewise.
10759 (eight unnamed splitters): Likewise.
10761 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10763 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10764 * config/rs6000/emmintrin.h: Likewise.
10765 * config/rs6000/mmintrin.h: Likewise.
10766 * config/rs6000/xmmintrin.h: Likewise.
10768 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10771 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10773 * tree.c (tree_nop_conversion): Return true for location wrapper
10775 (maybe_wrap_with_location): New function.
10776 (selftest::check_strip_nops): New function.
10777 (selftest::test_location_wrappers): New function.
10778 (selftest::tree_c_tests): Call it.
10779 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10780 (maybe_wrap_with_location): New decl.
10781 (EXPR_LOCATION_WRAPPER_P): New macro.
10782 (location_wrapper_p): New inline function.
10783 (tree_strip_any_location_wrapper): New inline function.
10785 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10788 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10789 stack_realign_offset for the largest alignment of stack slot
10791 (ix86_find_max_used_stack_alignment): New function.
10792 (ix86_finalize_stack_frame_flags): Use it. Set
10793 max_used_stack_alignment if we don't realign stack.
10794 * config/i386/i386.h (machine_function): Add
10795 max_used_stack_alignment.
10797 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10799 * config/arm/arm.opt (-mbranch-cost): New option.
10800 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10803 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10806 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10807 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10809 2018-01-10 Richard Biener <rguenther@suse.de>
10812 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10813 early out so it also covers the case where we have a non-NULL
10816 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10818 PR tree-optimization/83753
10819 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10820 for non-strided grouped accesses if the number of elements is 1.
10822 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10825 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10826 * i386.h (TARGET_USE_GATHER): Define.
10827 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10829 2018-01-10 Martin Liska <mliska@suse.cz>
10832 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10833 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10835 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10836 CLEANUP_NO_PARTITIONING is not set.
10838 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10840 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10841 for vectors, as a partial revert of r254296.
10842 * rtl.h (const_vec_p): Delete.
10843 (const_vec_duplicate_p): Don't test for vector CONSTs.
10844 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10845 * expmed.c (make_tree): Likewise.
10848 * common.md (E, F): Use CONSTANT_P instead of checking for
10850 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10851 checking for CONST_VECTOR.
10853 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10855 PR middle-end/83575
10856 * predict.c (force_edge_cold): Handle in more sane way edges
10857 with no prediction.
10859 2018-01-09 Carl Love <cel@us.ibm.com>
10861 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10863 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10864 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10865 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10866 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10867 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10868 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10869 * config/rs6000/rs6000-protos.h: Add extern defition for
10870 rs6000_generate_float2_double_code.
10871 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10873 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10874 (float2_v2df): Add define_expand.
10876 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10879 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10880 op_mode in the force_to_mode call.
10882 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10884 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10885 instead of checking each element individually.
10886 (aarch64_evpc_uzp): Likewise.
10887 (aarch64_evpc_zip): Likewise.
10888 (aarch64_evpc_ext): Likewise.
10889 (aarch64_evpc_rev): Likewise.
10890 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10891 instead of checking each element individually. Return true without
10893 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10894 whether all selected elements come from the same input, instead of
10895 checking each element individually. Remove calls to gen_rtx_REG,
10896 start_sequence and end_sequence and instead assert that no rtl is
10899 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10901 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
10902 order of HIGH and CONST checks.
10904 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10906 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
10907 if the destination isn't an SSA_NAME.
10909 2018-01-09 Richard Biener <rguenther@suse.de>
10911 PR tree-optimization/83668
10912 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
10914 (canonicalize_loop_form): ... here, renamed from ...
10915 (canonicalize_loop_closed_ssa_form): ... this and amended to
10916 swap successor edges for loop exit blocks to make us use
10917 the RPO order we need for initial schedule generation.
10919 2018-01-09 Joseph Myers <joseph@codesourcery.com>
10921 PR tree-optimization/64811
10922 * match.pd: When optimizing comparisons with Inf, avoid
10923 introducing or losing exceptions from comparisons with NaN.
10925 2018-01-09 Martin Liska <mliska@suse.cz>
10928 * asan.c (shadow_mem_size): Add gcc_assert.
10930 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
10932 Don't save registers in main().
10935 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
10936 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
10937 * config/avr/avr.c (avr_set_current_function): Don't error if
10938 naked, OS_task or OS_main are specified at the same time.
10939 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
10941 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
10943 * common/config/avr/avr-common.c (avr_option_optimization_table):
10944 Switch on -mmain-is-OS_task for optimizing compilations.
10946 2018-01-09 Richard Biener <rguenther@suse.de>
10948 PR tree-optimization/83572
10949 * graphite.c: Include cfganal.h.
10950 (graphite_transform_loops): Connect infinite loops to exit
10951 and remove fake edges at the end.
10953 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10955 * ipa-inline.c (edge_badness): Revert accidental checkin.
10957 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10960 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
10961 symbols; not inline clones.
10963 2018-01-09 Jakub Jelinek <jakub@redhat.com>
10966 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
10967 hard registers. Formatting fixes.
10969 PR preprocessor/83722
10970 * gcc.c (try_generate_repro): Pass
10971 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
10972 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
10975 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
10976 Kito Cheng <kito.cheng@gmail.com>
10978 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
10979 (riscv_leaf_function_p): Delete.
10980 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
10982 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10984 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
10986 (do_ifelse): New function.
10987 (do_isel): New function.
10988 (do_sub3): New function.
10989 (do_add3): New function.
10990 (do_load_mask_compare): New function.
10991 (do_overlap_load_compare): New function.
10992 (expand_compare_loop): New function.
10993 (expand_block_compare): Call expand_compare_loop() when appropriate.
10994 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
10995 option description.
10996 (-mblock-compare-inline-loop-limit): New option.
10998 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11001 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11002 Reverse order of second and third operands in first alternative.
11003 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11004 of first and second elements in UNSPEC_VPERMR vector.
11005 (altivec_expand_vec_perm_le): Likewise.
11007 2018-01-08 Jeff Law <law@redhat.com>
11009 PR rtl-optimizatin/81308
11010 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11011 (process_switch): If group_case_labels makes a change, then set
11013 (pass_convert_switch::execute): If a switch is converted, then
11014 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11016 PR rtl-optimization/81308
11017 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11020 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11022 PR target/83663 - Revert r255946
11023 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11024 generation for cases where splatting a value is not useful.
11025 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11026 across a vec_duplicate and a paradoxical subreg forming a vector
11027 mode to a vec_concat.
11029 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11031 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11032 -march=armv8.3-a variants.
11033 * config/arm/t-multilib: Likewise.
11034 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11036 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11038 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11040 (cceq_ior_compare_complement): Give it a name so I can use it, and
11041 change boolean_or_operator predicate to boolean_operator so it can
11042 be used to generate a crand.
11043 (eqne): New code iterator.
11044 (bd/bd_neg): New code_attrs.
11045 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11046 a single define_insn.
11047 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11048 decrement (bdnzt/bdnzf/bdzt/bdzf).
11049 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11050 with the new names of the branch decrement patterns, and added the
11051 names of the branch decrement conditional patterns.
11053 2018-01-08 Richard Biener <rguenther@suse.de>
11055 PR tree-optimization/83563
11056 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11059 2018-01-08 Richard Biener <rguenther@suse.de>
11061 PR middle-end/83713
11062 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11064 2018-01-08 Richard Biener <rguenther@suse.de>
11066 PR tree-optimization/83685
11067 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11068 references to abnormals.
11070 2018-01-08 Richard Biener <rguenther@suse.de>
11073 * dwarf2out.c (output_indirect_strings): Handle empty
11074 skeleton_debug_str_hash.
11075 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11077 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11079 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11080 (emit_store_direct): Likewise.
11081 (arc_trampoline_adjust_address): Likewise.
11082 (arc_asm_trampoline_template): New function.
11083 (arc_initialize_trampoline): Use asm_trampoline_template.
11084 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11085 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11086 * config/arc/arc.md (flush_icache): Delete pattern.
11088 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11090 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11091 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11094 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11097 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11098 by not USED_FOR_TARGET.
11099 (make_pass_resolve_sw_modes): Likewise.
11101 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11103 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11106 2018-01-08 Richard Biener <rguenther@suse.de>
11108 PR middle-end/83580
11109 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11111 2018-01-08 Richard Biener <rguenther@suse.de>
11113 PR middle-end/83517
11114 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11116 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11118 PR middle-end/81897
11119 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11120 basic blocks with a small number of successors.
11121 (convert_control_dep_chain_into_preds): Improve handling of
11123 (dump_predicates): Split apart into...
11124 (dump_pred_chain): ...here...
11125 (dump_pred_info): ...and here.
11126 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11127 (can_chain_union_be_invalidated_p): Improve check for invalidation
11129 (uninit_uses_cannot_happen): Avoid unnecessary if
11130 convert_control_dep_chain_into_preds yielded nothing.
11132 2018-01-06 Martin Sebor <msebor@redhat.com>
11134 PR tree-optimization/83640
11135 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11136 subtracting negative offset from size.
11137 (builtin_access::overlap): Adjust offset bounds of the access to fall
11138 within the size of the object if possible.
11140 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11142 PR rtl-optimization/83699
11143 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11144 extract_bit_field_as_subreg to cases in which the extracted
11145 value is also a vector.
11147 * lra-constraints.c (process_alt_operands): Test for the equivalence
11148 substitutions when detecting a possible reload cycle.
11150 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11153 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11154 by default if flag_selective_schedling{,2}. Formatting fixes.
11156 PR rtl-optimization/83682
11157 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11158 if it has non-VECTOR_MODE element mode.
11159 (vec_duplicate_p): Likewise.
11161 PR middle-end/83694
11162 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11163 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11165 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11168 * config/i386/i386-builtin.def
11169 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11170 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11171 Require also OPTION_MASK_ISA_AVX512F in addition to
11172 OPTION_MASK_ISA_GFNI.
11173 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11174 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11175 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11176 to OPTION_MASK_ISA_GFNI.
11177 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11178 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11179 OPTION_MASK_ISA_AVX512BW.
11180 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11181 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11182 addition to OPTION_MASK_ISA_GFNI.
11183 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11184 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11185 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11186 to OPTION_MASK_ISA_GFNI.
11187 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11188 a requirement for all ISAs rather than any of them with a few
11190 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11192 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11193 bitmasks to be enabled with 3 exceptions, instead of requiring any
11194 enabled ISA with lots of exceptions.
11195 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11196 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11197 Change avx512bw in isa attribute to avx512f.
11198 * config/i386/sgxintrin.h: Add license boilerplate.
11199 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11200 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11201 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11202 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11204 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11205 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11206 temporarily sse2 rather than sse if not enabled already.
11209 * config/i386/sse.md (VI248_VLBW): Rename to ...
11210 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11211 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11212 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11213 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11214 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11215 mode iterator instead of VI248_VLBW.
11217 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11219 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11220 (record_modified): Skip clobbers; add debug output.
11221 (param_change_prob): Use sreal frequencies.
11223 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11225 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11226 punt for user-aligned variables.
11228 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11230 * tree-chrec.c (chrec_contains_symbols): Return true for
11233 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11236 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11237 of (x|y) == x for BICS pattern.
11239 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11241 PR tree-optimization/83605
11242 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11243 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11246 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11248 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11249 * config/epiphany/rtems.h: New file.
11251 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11252 Uros Bizjak <ubizjak@gmail.com>
11255 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11256 QIreg_operand instead of register_operand predicate.
11257 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11258 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11259 comments instead of -fmitigate[-_]rop.
11261 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11264 * cgraphunit.c (symbol_table::compile): Switch to text_section
11265 before calling assembly_start debug hook.
11266 * run-rtl-passes.c (run_rtl_passes): Likewise.
11269 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11271 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11272 range_int_cst_p rather than !symbolic_range_p before calling
11273 extract_range_from_multiplicative_op_1.
11275 2018-01-04 Jeff Law <law@redhat.com>
11277 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11278 redundant test in assertion.
11280 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11282 * doc/rtl.texi: Document machine_mode wrapper classes.
11284 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11286 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11287 using tree_to_uhwi.
11289 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11291 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11292 the VEC_PERM_EXPR fold to fail.
11294 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11297 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11298 to switched_sections.
11300 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11303 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11304 test for d.testing.
11306 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11309 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11310 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11312 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11315 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11316 is BLKmode and bitpos not zero or mode change is needed.
11318 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11321 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11324 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11327 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11328 instead of MULT rtx. Update all corresponding splitters.
11329 (*saddl_se): Ditto.
11330 (*ssub<modesuffix>): Ditto.
11331 (*ssubl_se): Ditto.
11332 (*cmp_sadd_di): Update split patterns.
11333 (*cmp_sadd_si): Ditto.
11334 (*cmp_sadd_sidi): Ditto.
11335 (*cmp_ssub_di): Ditto.
11336 (*cmp_ssub_si): Ditto.
11337 (*cmp_ssub_sidi): Ditto.
11338 * config/alpha/predicates.md (const23_operand): New predicate.
11339 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11340 Look for ASHIFT, not MULT inner operand.
11341 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11343 2018-01-04 Martin Liska <mliska@suse.cz>
11345 PR gcov-profile/83669
11346 * gcov.c (output_intermediate_file): Add version to intermediate
11348 * doc/gcov.texi: Document new field 'version' in intermediate
11349 file format. Fix location of '-k' option of gcov command.
11351 2018-01-04 Martin Liska <mliska@suse.cz>
11354 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11356 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11358 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11360 2018-01-03 Martin Sebor <msebor@redhat.com>
11362 PR tree-optimization/83655
11363 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11364 checking calls with invalid arguments.
11366 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11368 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11369 (vectorizable_mask_load_store): Delete.
11370 (vectorizable_call): Return false for masked loads and stores.
11371 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11372 instead of gimple_assign_rhs1.
11373 (vectorizable_load): Handle IFN_MASK_LOAD.
11374 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11376 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11378 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11380 (vectorizable_mask_load_store): ...here.
11381 (vectorizable_load): ...and here.
11383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11385 * tree-vect-stmts.c (vect_build_all_ones_mask)
11386 (vect_build_zero_merge_argument): New functions, split out from...
11387 (vectorizable_load): ...here.
11389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11391 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11393 (vectorizable_mask_load_store): ...here.
11394 (vectorizable_store): ...and here.
11396 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11398 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11400 (vectorizable_mask_load_store): ...here.
11402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11404 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11405 (vect_model_store_cost): Take a vec_load_store_type instead of a
11407 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11408 (vect_model_store_cost): Take a vec_load_store_type instead of a
11410 (vectorizable_mask_load_store): Update accordingly.
11411 (vectorizable_store): Likewise.
11412 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11414 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11416 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11417 IFN_MASK_LOAD calls here rather than...
11418 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11420 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11421 Alan Hayward <alan.hayward@arm.com>
11422 David Sherwood <david.sherwood@arm.com>
11424 * expmed.c (extract_bit_field_1): For vector extracts,
11425 fall back to extract_bit_field_as_subreg if vec_extract
11428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11429 Alan Hayward <alan.hayward@arm.com>
11430 David Sherwood <david.sherwood@arm.com>
11432 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11433 they are variable or constant sized.
11434 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11435 slots for constant-sized data.
11437 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11438 Alan Hayward <alan.hayward@arm.com>
11439 David Sherwood <david.sherwood@arm.com>
11441 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11442 handling COND_EXPRs with boolean comparisons, try to find a better
11443 basis for the mask type than the boolean itself.
11445 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11447 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11448 is calculated and how it can be overridden.
11449 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11450 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11452 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11455 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11456 Alan Hayward <alan.hayward@arm.com>
11457 David Sherwood <david.sherwood@arm.com>
11459 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11460 Remove the mode argument.
11461 (aarch64_simd_valid_immediate): Remove the mode and inverse
11463 * config/aarch64/iterators.md (bitsize): New iterator.
11464 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11465 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11466 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11467 aarch64_simd_valid_immediate.
11468 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11469 (aarch64_reg_or_bic_imm): Likewise.
11470 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11471 with an insn_type enum and msl with a modifier_type enum.
11472 Replace element_width with a scalar_mode. Change the shift
11473 to unsigned int. Add constructors for scalar_float_mode and
11474 scalar_int_mode elements.
11475 (aarch64_vect_float_const_representable_p): Delete.
11476 (aarch64_can_const_movi_rtx_p)
11477 (aarch64_simd_scalar_immediate_valid_for_move)
11478 (aarch64_simd_make_constant): Update call to
11479 aarch64_simd_valid_immediate.
11480 (aarch64_advsimd_valid_immediate_hs): New function.
11481 (aarch64_advsimd_valid_immediate): Likewise.
11482 (aarch64_simd_valid_immediate): Remove mode and inverse
11483 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11484 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11485 and aarch64_float_const_representable_p on the result.
11486 (aarch64_output_simd_mov_immediate): Remove mode argument.
11487 Update call to aarch64_simd_valid_immediate and use of
11488 simd_immediate_info.
11489 (aarch64_output_scalar_simd_mov_immediate): Update call
11492 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11493 Alan Hayward <alan.hayward@arm.com>
11494 David Sherwood <david.sherwood@arm.com>
11496 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11497 (mode_nunits): Likewise CONST_MODE_NUNITS.
11498 * machmode.def (ADJUST_NUNITS): Document.
11499 * genmodes.c (mode_data::need_nunits_adj): New field.
11500 (blank_mode): Update accordingly.
11501 (adj_nunits): New variable.
11502 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11504 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11505 listed in adj_nunits.
11506 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11507 listed in adj_nunits. Don't emit case statements for such modes.
11508 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11509 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11510 nothing if adj_nunits is nonnull.
11511 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11512 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11513 (emit_mode_fbit): Update use of print_maybe_const_decl.
11514 (emit_move_size): Likewise. Treat the array as non-const
11516 (emit_mode_adjustments): Handle adj_nunits.
11518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11520 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11521 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11522 (VECTOR_MODES): Use it.
11523 (make_vector_modes): Take the prefix as an argument.
11525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11526 Alan Hayward <alan.hayward@arm.com>
11527 David Sherwood <david.sherwood@arm.com>
11529 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11530 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11531 for MODE_VECTOR_BOOL.
11532 * machmode.def (VECTOR_BOOL_MODE): Document.
11533 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11534 (make_vector_bool_mode): New function.
11535 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11537 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11538 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11540 * stor-layout.c (int_mode_for_mode): Likewise.
11541 * tree.c (build_vector_type_for_mode): Likewise.
11542 * varasm.c (output_constant_pool_2): Likewise.
11543 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11544 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11545 for MODE_VECTOR_BOOL.
11546 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11547 of mode class checks.
11548 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11549 instead of a list of mode class checks.
11550 (expand_vector_scalar_condition): Likewise.
11551 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11553 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11554 Alan Hayward <alan.hayward@arm.com>
11555 David Sherwood <david.sherwood@arm.com>
11557 * machmode.h (mode_size): Change from unsigned short to
11559 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11560 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11561 or if measurement_type is not polynomial.
11562 (fixed_size_mode::includes_p): Check for constant-sized modes.
11563 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11564 return a poly_uint16 rather than an unsigned short.
11565 (emit_mode_size): Change the type of mode_size from unsigned short
11566 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11567 (emit_mode_adjustments): Cope with polynomial vector sizes.
11568 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11570 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11572 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11573 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11574 * caller-save.c (setup_save_areas): Likewise.
11575 (replace_reg_with_saved_mem): Likewise.
11576 * calls.c (emit_library_call_value_1): Likewise.
11577 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11578 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11579 (gen_lowpart_for_combine): Likewise.
11580 * convert.c (convert_to_integer_1): Likewise.
11581 * cse.c (equiv_constant, cse_insn): Likewise.
11582 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11583 (cselib_subst_to_values): Likewise.
11584 * dce.c (word_dce_process_block): Likewise.
11585 * df-problems.c (df_word_lr_mark_ref): Likewise.
11586 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11587 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11588 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11589 (rtl_for_decl_location): Likewise.
11590 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11591 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11592 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11593 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11594 (expand_expr_real_1): Likewise.
11595 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11596 (pad_below): Likewise.
11597 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11598 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11599 * ira.c (get_subreg_tracking_sizes): Likewise.
11600 * ira-build.c (ira_create_allocno_objects): Likewise.
11601 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11602 (ira_sort_regnos_for_alter_reg): Likewise.
11603 * ira-costs.c (record_operand_costs): Likewise.
11604 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11605 (resolve_simple_move): Likewise.
11606 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11607 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11608 (lra_constraints): Likewise.
11609 (CONST_POOL_OK_P): Reject variable-sized modes.
11610 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11611 (add_pseudo_to_slot, lra_spill): Likewise.
11612 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11613 * optabs-query.c (get_best_extraction_insn): Likewise.
11614 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11615 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11616 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11617 * recog.c (offsettable_address_addr_space_p): Likewise.
11618 * regcprop.c (maybe_mode_change): Likewise.
11619 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11620 * regrename.c (build_def_use): Likewise.
11621 * regstat.c (dump_reg_info): Likewise.
11622 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11623 (find_reloads, find_reloads_subreg_address): Likewise.
11624 * reload1.c (eliminate_regs_1): Likewise.
11625 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11626 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11627 (simplify_binary_operation_1, simplify_subreg): Likewise.
11628 * targhooks.c (default_function_arg_padding): Likewise.
11629 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11630 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11631 (verify_gimple_assign_ternary): Likewise.
11632 * tree-inline.c (estimate_move_cost): Likewise.
11633 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11634 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11635 (get_address_cost_ainc): Likewise.
11636 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11637 (vect_supportable_dr_alignment): Likewise.
11638 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11639 (vectorizable_reduction): Likewise.
11640 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11641 (vectorizable_operation, vectorizable_load): Likewise.
11642 * tree.c (build_same_sized_truth_vector_type): Likewise.
11643 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11644 * var-tracking.c (emit_note_insn_var_location): Likewise.
11645 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11646 (ADDR_VEC_ALIGN): Likewise.
11648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11649 Alan Hayward <alan.hayward@arm.com>
11650 David Sherwood <david.sherwood@arm.com>
11652 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11654 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11655 or if measurement_type is polynomial.
11656 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11657 * combine.c (make_extraction): Likewise.
11658 * dse.c (find_shift_sequence): Likewise.
11659 * dwarf2out.c (mem_loc_descriptor): Likewise.
11660 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11661 (extract_bit_field, extract_low_bits): Likewise.
11662 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11663 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11664 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11665 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11666 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11667 * reload.c (find_reloads): Likewise.
11668 * reload1.c (alter_reg): Likewise.
11669 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11670 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11671 * tree-if-conv.c (predicate_mem_writes): Likewise.
11672 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11673 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11674 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11675 * valtrack.c (dead_debug_insert_temp): Likewise.
11676 * varasm.c (mergeable_constant_section): Likewise.
11677 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11679 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11680 Alan Hayward <alan.hayward@arm.com>
11681 David Sherwood <david.sherwood@arm.com>
11683 * expr.c (expand_assignment): Cope with polynomial mode sizes
11684 when assigning to a CONCAT.
11686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11687 Alan Hayward <alan.hayward@arm.com>
11688 David Sherwood <david.sherwood@arm.com>
11690 * machmode.h (mode_precision): Change from unsigned short to
11692 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11694 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11695 or if measurement_type is not polynomial.
11696 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11697 in which the mode is already known to be a scalar_int_mode.
11698 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11699 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11701 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11702 for GET_MODE_PRECISION.
11703 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11704 for GET_MODE_PRECISION.
11705 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11707 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11708 (expand_field_assignment, make_extraction): Likewise.
11709 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11710 (get_last_value): Likewise.
11711 * convert.c (convert_to_integer_1): Likewise.
11712 * cse.c (cse_insn): Likewise.
11713 * expr.c (expand_expr_real_1): Likewise.
11714 * lra-constraints.c (simplify_operand_subreg): Likewise.
11715 * optabs-query.c (can_atomic_load_p): Likewise.
11716 * optabs.c (expand_atomic_load): Likewise.
11717 (expand_atomic_store): Likewise.
11718 * ree.c (combine_reaching_defs): Likewise.
11719 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11720 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11721 * tree.h (type_has_mode_precision_p): Likewise.
11722 * ubsan.c (instrument_si_overflow): Likewise.
11724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11725 Alan Hayward <alan.hayward@arm.com>
11726 David Sherwood <david.sherwood@arm.com>
11728 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11729 polynomial numbers of units.
11730 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11731 (valid_vector_subparts_p): New function.
11732 (build_vector_type): Remove temporary shim and take the number
11733 of units as a poly_uint64 rather than an int.
11734 (build_opaque_vector_type): Take the number of units as a
11735 poly_uint64 rather than an int.
11736 * tree.c (build_vector_from_ctor): Handle polynomial
11737 TYPE_VECTOR_SUBPARTS.
11738 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11739 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11740 (build_vector_from_val): If the number of units is variable,
11741 use build_vec_duplicate_cst for constant operands and
11742 VEC_DUPLICATE_EXPR otherwise.
11743 (make_vector_type): Remove temporary is_constant ().
11744 (build_vector_type, build_opaque_vector_type): Take the number of
11745 units as a poly_uint64 rather than an int.
11746 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11748 * cfgexpand.c (expand_debug_expr): Likewise.
11749 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11750 (store_constructor, expand_expr_real_1): Likewise.
11751 (const_scalar_mask_from_tree): Likewise.
11752 * fold-const-call.c (fold_const_reduction): Likewise.
11753 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11754 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11755 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11756 (fold_relational_const): Likewise.
11757 (native_interpret_vector): Likewise. Change the size from an
11758 int to an unsigned int.
11759 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11760 TYPE_VECTOR_SUBPARTS.
11761 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11762 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11763 duplicating a non-constant operand into a variable-length vector.
11764 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11765 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11766 * ipa-icf.c (sem_variable::equals): Likewise.
11767 * match.pd: Likewise.
11768 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11769 * print-tree.c (print_node): Likewise.
11770 * stor-layout.c (layout_type): Likewise.
11771 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11772 * tree-cfg.c (verify_gimple_comparison): Likewise.
11773 (verify_gimple_assign_binary): Likewise.
11774 (verify_gimple_assign_ternary): Likewise.
11775 (verify_gimple_assign_single): Likewise.
11776 * tree-pretty-print.c (dump_generic_node): Likewise.
11777 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11778 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11779 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11780 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11781 (vect_shift_permute_load_chain): Likewise.
11782 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11783 (expand_vector_condition, optimize_vector_constructor): Likewise.
11784 (lower_vec_perm, get_compute_type): Likewise.
11785 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11786 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11787 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11788 (vect_recog_mask_conversion_pattern): Likewise.
11789 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11790 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11791 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11792 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11793 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11794 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11795 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11796 (supportable_widening_operation): Likewise.
11797 (supportable_narrowing_operation): Likewise.
11798 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11800 * varasm.c (output_constant): Likewise.
11802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11803 Alan Hayward <alan.hayward@arm.com>
11804 David Sherwood <david.sherwood@arm.com>
11806 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11807 so that both the length == 3 and length != 3 cases set up their
11808 own permute vectors. Add comments explaining why we know the
11809 number of elements is constant.
11810 (vect_permute_load_chain): Likewise.
11812 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11813 Alan Hayward <alan.hayward@arm.com>
11814 David Sherwood <david.sherwood@arm.com>
11816 * machmode.h (mode_nunits): Change from unsigned char to
11818 (ONLY_FIXED_SIZE_MODES): New macro.
11819 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11820 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11821 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11823 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11824 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11825 or if measurement_type is not polynomial.
11826 * genmodes.c (ZERO_COEFFS): New macro.
11827 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11829 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11830 Use ZERO_COEFFS when emitting initializers.
11831 * data-streamer.h (bp_pack_poly_value): New function.
11832 (bp_unpack_poly_value): Likewise.
11833 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11834 for GET_MODE_NUNITS.
11835 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11836 for GET_MODE_NUNITS.
11837 * tree.c (make_vector_type): Remove temporary shim and make
11838 the real function take the number of units as a poly_uint64
11839 rather than an int.
11840 (build_vector_type_for_mode): Handle polynomial nunits.
11841 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11842 * emit-rtl.c (const_vec_series_p_1): Likewise.
11843 (gen_rtx_CONST_VECTOR): Likewise.
11844 * fold-const.c (test_vec_duplicate_folding): Likewise.
11845 * genrecog.c (validate_pattern): Likewise.
11846 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11847 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11848 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11849 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11850 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11851 * rtlanal.c (subreg_get_info): Likewise.
11852 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11853 (vect_grouped_load_supported): Likewise.
11854 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11855 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11856 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11857 (simplify_const_unary_operation, simplify_binary_operation_1)
11858 (simplify_const_binary_operation, simplify_ternary_operation)
11859 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11860 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11861 instead of CONST_VECTOR_NUNITS.
11862 * varasm.c (output_constant_pool_2): Likewise.
11863 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11864 explicit-encoded elements in the XVEC for variable-length vectors.
11866 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11868 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11870 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11871 Alan Hayward <alan.hayward@arm.com>
11872 David Sherwood <david.sherwood@arm.com>
11874 * coretypes.h (fixed_size_mode): Declare.
11875 (fixed_size_mode_pod): New typedef.
11876 * builtins.h (target_builtins::x_apply_args_mode)
11877 (target_builtins::x_apply_result_mode): Change type to
11878 fixed_size_mode_pod.
11879 * builtins.c (apply_args_size, apply_result_size, result_vector)
11880 (expand_builtin_apply_args_1, expand_builtin_apply)
11881 (expand_builtin_return): Update accordingly.
11883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11885 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11886 * cselib.c (cselib_hash_rtx): Likewise.
11887 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11888 CONST_VECTOR encoding.
11890 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11891 Jeff Law <law@redhat.com>
11894 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11895 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11896 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11897 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
11900 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
11901 explicitly probe *sp in a noreturn function if there were any callee
11902 register saves or frame pointer is needed.
11904 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11907 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
11908 BLKmode for ternary, binary or unary expressions.
11911 * var-tracking.c (delete_vta_debug_insn): New inline function.
11912 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
11913 insns from get_insns () to NULL instead of each bb separately.
11914 Use delete_vta_debug_insn. No longer static.
11915 (vt_debug_insns_local, variable_tracking_main_1): Adjust
11916 delete_vta_debug_insns callers.
11917 * rtl.h (delete_vta_debug_insns): Declare.
11918 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
11919 instead of variable_tracking_main.
11921 2018-01-03 Martin Sebor <msebor@redhat.com>
11923 PR tree-optimization/83603
11924 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
11925 arguments past the endof the argument list in functions declared
11926 without a prototype.
11927 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
11928 Avoid checking when arguments are null.
11930 2018-01-03 Martin Sebor <msebor@redhat.com>
11933 * doc/extend.texi (attribute const): Fix a typo.
11934 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
11935 issuing -Wsuggest-attribute for void functions.
11937 2018-01-03 Martin Sebor <msebor@redhat.com>
11939 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
11940 offset_int::from instead of wide_int::to_shwi.
11941 (maybe_diag_overlap): Remove assertion.
11942 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
11943 * gimple-ssa-sprintf.c (format_directive): Same.
11944 (parse_directive): Same.
11945 (sprintf_dom_walker::compute_format_length): Same.
11946 (try_substitute_return_value): Same.
11948 2018-01-03 Jeff Law <law@redhat.com>
11950 PR middle-end/83654
11951 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
11952 non-constant residual for zero at runtime and avoid probing in
11953 that case. Reorganize code for trailing problem to mirror handling
11956 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11958 PR tree-optimization/83501
11959 * tree-ssa-strlen.c (get_string_cst): New.
11960 (handle_char_store): Call get_string_cst.
11962 2018-01-03 Martin Liska <mliska@suse.cz>
11964 PR tree-optimization/83593
11965 * tree-ssa-strlen.c: Include tree-cfg.h.
11966 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
11967 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
11968 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
11970 (strlen_dom_walker::before_dom_children): Call
11971 gimple_purge_dead_eh_edges. Dump tranformation with details
11973 (strlen_dom_walker::before_dom_children): Update call by adding
11974 new argument cleanup_eh.
11975 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
11977 2018-01-03 Martin Liska <mliska@suse.cz>
11980 * cif-code.def (VARIADIC_THUNK): New enum value.
11981 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
11984 2018-01-03 Jan Beulich <jbeulich@suse.com>
11986 * sse.md (mov<mode>_internal): Tighten condition for when to use
11987 vmovdqu<ssescalarsize> for TI and OI modes.
11989 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11991 Update copyright years.
11993 2018-01-03 Martin Liska <mliska@suse.cz>
11996 * ipa-visibility.c (function_and_variable_visibility): Skip
11997 functions with noipa attribure.
11999 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12001 * gcc.c (process_command): Update copyright notice dates.
12002 * gcov-dump.c (print_version): Ditto.
12003 * gcov.c (print_version): Ditto.
12004 * gcov-tool.c (print_version): Ditto.
12005 * gengtype.c (create_file): Ditto.
12006 * doc/cpp.texi: Bump @copying's copyright year.
12007 * doc/cppinternals.texi: Ditto.
12008 * doc/gcc.texi: Ditto.
12009 * doc/gccint.texi: Ditto.
12010 * doc/gcov.texi: Ditto.
12011 * doc/install.texi: Ditto.
12012 * doc/invoke.texi: Ditto.
12014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12016 * vector-builder.h (vector_builder::m_full_nelts): Change from
12017 unsigned int to poly_uint64.
12018 (vector_builder::full_nelts): Update prototype accordingly.
12019 (vector_builder::new_vector): Likewise.
12020 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12021 (vector_builder::operator ==): Likewise.
12022 (vector_builder::finalize): Likewise.
12023 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12024 Take the number of elements as a poly_uint64 rather than an
12026 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12027 from unsigned int to poly_uint64.
12028 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12029 (vec_perm_indices::new_vector): Likewise.
12030 (vec_perm_indices::length): Likewise.
12031 (vec_perm_indices::nelts_per_input): Likewise.
12032 (vec_perm_indices::input_nelts): Likewise.
12033 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12034 number of elements per input as a poly_uint64 rather than an
12035 unsigned int. Use the original encoding for variable-length
12036 vectors, rather than clamping each individual element.
12037 For the second and subsequent elements in each pattern,
12038 clamp the step and base before clamping their sum.
12039 (vec_perm_indices::series_p): Handle polynomial element counts.
12040 (vec_perm_indices::all_in_range_p): Likewise.
12041 (vec_perm_indices_to_tree): Likewise.
12042 (vec_perm_indices_to_rtx): Likewise.
12043 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12044 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12045 (tree_vector_builder::new_binary_operation): Handle polynomial
12046 element counts. Return false if we need to know the number
12047 of elements at compile time.
12048 * fold-const.c (fold_vec_perm): Punt if the number of elements
12049 isn't known at compile time.
12051 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12053 * vec-perm-indices.h (vec_perm_builder): Change element type
12054 from HOST_WIDE_INT to poly_int64.
12055 (vec_perm_indices::element_type): Update accordingly.
12056 (vec_perm_indices::clamp): Handle polynomial element_types.
12057 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12058 (vec_perm_indices::all_in_range_p): Likewise.
12059 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12061 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12062 polynomial vec_perm_indices element types.
12063 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12064 * fold-const.c (fold_vec_perm): Likewise.
12065 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12066 * tree-vect-generic.c (lower_vec_perm): Likewise.
12067 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12068 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12069 element type to HOST_WIDE_INT.
12071 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12072 Alan Hayward <alan.hayward@arm.com>
12073 David Sherwood <david.sherwood@arm.com>
12075 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12076 rather than an int. Use plus_constant.
12077 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12078 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12080 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12081 Alan Hayward <alan.hayward@arm.com>
12082 David Sherwood <david.sherwood@arm.com>
12084 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12085 a HOST_WIDE_INT to a poly_int64.
12087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12088 Alan Hayward <alan.hayward@arm.com>
12089 David Sherwood <david.sherwood@arm.com>
12091 * calls.c (load_register_parameters): Cope with polynomial
12092 mode sizes. Require a constant size for BLKmode parameters
12093 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12094 forces a parameter to be padded at the lsb end in order to
12095 fill a complete number of words, require the parameter size
12096 to be ordered wrt UNITS_PER_WORD.
12098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12099 Alan Hayward <alan.hayward@arm.com>
12100 David Sherwood <david.sherwood@arm.com>
12102 * reload1.c (spill_stack_slot_width): Change element type
12103 from unsigned int to poly_uint64_pod.
12104 (alter_reg): Treat mode sizes as polynomial.
12106 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12107 Alan Hayward <alan.hayward@arm.com>
12108 David Sherwood <david.sherwood@arm.com>
12110 * reload.c (complex_word_subreg_p): New function.
12111 (reload_inner_reg_of_subreg, push_reload): Use it.
12113 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12114 Alan Hayward <alan.hayward@arm.com>
12115 David Sherwood <david.sherwood@arm.com>
12117 * lra-constraints.c (process_alt_operands): Reject matched
12118 operands whose sizes aren't ordered.
12119 (match_reload): Refer to this check here.
12121 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12122 Alan Hayward <alan.hayward@arm.com>
12123 David Sherwood <david.sherwood@arm.com>
12125 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12126 that the mode size is in the set {1, 2, 4, 8, 16}.
12128 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12129 Alan Hayward <alan.hayward@arm.com>
12130 David Sherwood <david.sherwood@arm.com>
12132 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12133 Use plus_constant instead of gen_rtx_PLUS.
12135 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12136 Alan Hayward <alan.hayward@arm.com>
12137 David Sherwood <david.sherwood@arm.com>
12139 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12140 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12141 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12142 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12143 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12144 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12145 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12146 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12147 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12148 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12150 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12151 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12152 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12153 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12154 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12155 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12156 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12157 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12158 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12159 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12161 * expr.c (emit_move_resolve_push): Treat the input and result
12162 of PUSH_ROUNDING as a poly_int64.
12163 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12164 (emit_push_insn): Likewise.
12165 * lra-eliminations.c (mark_not_eliminable): Likewise.
12166 * recog.c (push_operand): Likewise.
12167 * reload1.c (elimination_effects): Likewise.
12168 * rtlanal.c (nonzero_bits1): Likewise.
12169 * calls.c (store_one_arg): Likewise. Require the padding to be
12170 known at compile time.
12172 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12173 Alan Hayward <alan.hayward@arm.com>
12174 David Sherwood <david.sherwood@arm.com>
12176 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12177 Use plus_constant instead of gen_rtx_PLUS.
12179 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12180 Alan Hayward <alan.hayward@arm.com>
12181 David Sherwood <david.sherwood@arm.com>
12183 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12184 rather than an int.
12186 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12187 Alan Hayward <alan.hayward@arm.com>
12188 David Sherwood <david.sherwood@arm.com>
12190 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12191 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12192 via stack temporaries. Treat the mode size as polynomial too.
12194 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12195 Alan Hayward <alan.hayward@arm.com>
12196 David Sherwood <david.sherwood@arm.com>
12198 * expr.c (expand_expr_real_2): When handling conversions involving
12199 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12200 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12201 as a poly_uint64 too.
12203 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12204 Alan Hayward <alan.hayward@arm.com>
12205 David Sherwood <david.sherwood@arm.com>
12207 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12210 Alan Hayward <alan.hayward@arm.com>
12211 David Sherwood <david.sherwood@arm.com>
12213 * combine.c (can_change_dest_mode): Handle polynomial
12214 REGMODE_NATURAL_SIZE.
12215 * expmed.c (store_bit_field_1): Likewise.
12216 * expr.c (store_constructor): Likewise.
12217 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12218 and polynomial REGMODE_NATURAL_SIZE.
12219 (gen_lowpart_common): Likewise.
12220 * reginfo.c (record_subregs_of_mode): Likewise.
12221 * rtlanal.c (read_modify_subreg_p): Likewise.
12223 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12224 Alan Hayward <alan.hayward@arm.com>
12225 David Sherwood <david.sherwood@arm.com>
12227 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12228 numbers of elements.
12230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12231 Alan Hayward <alan.hayward@arm.com>
12232 David Sherwood <david.sherwood@arm.com>
12234 * match.pd: Cope with polynomial numbers of vector elements.
12236 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12237 Alan Hayward <alan.hayward@arm.com>
12238 David Sherwood <david.sherwood@arm.com>
12240 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12241 in a POINTER_PLUS_EXPR.
12243 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12244 Alan Hayward <alan.hayward@arm.com>
12245 David Sherwood <david.sherwood@arm.com>
12247 * omp-simd-clone.c (simd_clone_subparts): New function.
12248 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12249 (ipa_simd_modify_function_body): Likewise.
12251 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12252 Alan Hayward <alan.hayward@arm.com>
12253 David Sherwood <david.sherwood@arm.com>
12255 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12256 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12257 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12258 (expand_vector_condition, vector_element): Likewise.
12259 (subparts_gt): New function.
12260 (get_compute_type): Use subparts_gt.
12261 (count_type_subparts): Delete.
12262 (expand_vector_operations_1): Use subparts_gt instead of
12263 count_type_subparts.
12265 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12266 Alan Hayward <alan.hayward@arm.com>
12267 David Sherwood <david.sherwood@arm.com>
12269 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12270 (vect_compile_time_alias): ...this new function. Do the calculation
12271 on poly_ints rather than trees.
12272 (vect_prune_runtime_alias_test_list): Update call accordingly.
12274 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12275 Alan Hayward <alan.hayward@arm.com>
12276 David Sherwood <david.sherwood@arm.com>
12278 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12280 (vect_schedule_slp_instance): Likewise.
12282 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12283 Alan Hayward <alan.hayward@arm.com>
12284 David Sherwood <david.sherwood@arm.com>
12286 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12287 constant and extern definitions for variable-length vectors.
12288 (vect_get_constant_vectors): Note that the number of units
12289 is known to be constant.
12291 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12292 Alan Hayward <alan.hayward@arm.com>
12293 David Sherwood <david.sherwood@arm.com>
12295 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12296 of units as polynomial. Choose between WIDE and NARROW based
12299 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12300 Alan Hayward <alan.hayward@arm.com>
12301 David Sherwood <david.sherwood@arm.com>
12303 * tree-vect-stmts.c (simd_clone_subparts): New function.
12304 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12306 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12307 Alan Hayward <alan.hayward@arm.com>
12308 David Sherwood <david.sherwood@arm.com>
12310 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12311 vectors as polynomial. Use build_index_vector for
12312 IFN_GOMP_SIMD_LANE.
12314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12315 Alan Hayward <alan.hayward@arm.com>
12316 David Sherwood <david.sherwood@arm.com>
12318 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12319 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12320 for variable-length vectors.
12321 (vectorizable_mask_load_store): Treat the number of units as
12322 polynomial, asserting that it is constant if the condition has
12323 already been enforced.
12324 (vectorizable_store, vectorizable_load): Likewise.
12326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12327 Alan Hayward <alan.hayward@arm.com>
12328 David Sherwood <david.sherwood@arm.com>
12330 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12331 of units as polynomial. Punt if we can't tell at compile time
12332 which vector contains the final result.
12334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12335 Alan Hayward <alan.hayward@arm.com>
12336 David Sherwood <david.sherwood@arm.com>
12338 * tree-vect-loop.c (vectorizable_induction): Treat the number
12339 of units as polynomial. Punt on SLP inductions. Use an integer
12340 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12341 cast of such a series for variable-length floating-point
12344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12345 Alan Hayward <alan.hayward@arm.com>
12346 David Sherwood <david.sherwood@arm.com>
12348 * tree.h (build_index_vector): Declare.
12349 * tree.c (build_index_vector): New function.
12350 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12351 of units as polynomial, forcibly converting it to a constant if
12352 vectorizable_reduction has already enforced the condition.
12353 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12354 to create a {1,2,3,...} vector.
12355 (vectorizable_reduction): Treat the number of units as polynomial.
12356 Choose vectype_in based on the largest scalar element size rather
12357 than the smallest number of units. Enforce the restrictions
12360 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12361 Alan Hayward <alan.hayward@arm.com>
12362 David Sherwood <david.sherwood@arm.com>
12364 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12365 number of units as polynomial.
12367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12368 Alan Hayward <alan.hayward@arm.com>
12369 David Sherwood <david.sherwood@arm.com>
12371 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12372 * target.def (autovectorize_vector_sizes): Return the vector sizes
12373 by pointer, using vector_sizes rather than a bitmask.
12374 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12375 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12376 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12378 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12379 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12380 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12381 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12382 * omp-general.c (omp_max_vf): Likewise.
12383 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12384 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12385 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12386 * tree-vect-slp.c (vect_slp_bb): Likewise.
12387 * doc/tm.texi: Regenerate.
12388 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12390 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12391 the vector size as a poly_uint64 rather than an unsigned int.
12392 (current_vector_size): Change from an unsigned int to a poly_uint64.
12393 (get_vectype_for_scalar_type): Update accordingly.
12394 * tree.h (build_truth_vector_type): Take the size and number of
12395 units as a poly_uint64 rather than an unsigned int.
12396 (build_vector_type): Add a temporary overload that takes
12397 the number of units as a poly_uint64 rather than an unsigned int.
12398 * tree.c (make_vector_type): Likewise.
12399 (build_truth_vector_type): Take the number of units as a poly_uint64
12400 rather than an unsigned int.
12402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12403 Alan Hayward <alan.hayward@arm.com>
12404 David Sherwood <david.sherwood@arm.com>
12406 * target.def (get_mask_mode): Take the number of units and length
12407 as poly_uint64s rather than unsigned ints.
12408 * targhooks.h (default_get_mask_mode): Update accordingly.
12409 * targhooks.c (default_get_mask_mode): Likewise.
12410 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12411 * doc/tm.texi: Regenerate.
12413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12414 Alan Hayward <alan.hayward@arm.com>
12415 David Sherwood <david.sherwood@arm.com>
12417 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12418 * omp-general.c (omp_max_vf): Likewise.
12419 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12420 (expand_omp_simd): Handle polynomial safelen.
12421 * omp-low.c (omplow_simd_context): Add a default constructor.
12422 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12423 (lower_rec_simd_input_clauses): Update accordingly.
12424 (lower_rec_input_clauses): Likewise.
12426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12427 Alan Hayward <alan.hayward@arm.com>
12428 David Sherwood <david.sherwood@arm.com>
12430 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12431 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12432 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12433 (vect_analyze_slp_cost): Likewise.
12434 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12435 (vect_model_load_cost): Likewise.
12437 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12438 Alan Hayward <alan.hayward@arm.com>
12439 David Sherwood <david.sherwood@arm.com>
12441 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12442 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12443 from an unsigned int * to a poly_uint64_pod *.
12444 (calculate_unrolling_factor): New function.
12445 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12447 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12448 Alan Hayward <alan.hayward@arm.com>
12449 David Sherwood <david.sherwood@arm.com>
12451 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12452 from an unsigned int to a poly_uint64.
12453 (_loop_vec_info::slp_unrolling_factor): Likewise.
12454 (_loop_vec_info::vectorization_factor): Change from an int
12456 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12457 (vect_get_num_vectors): New function.
12458 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12459 (vect_get_num_copies): Use vect_get_num_vectors.
12460 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12461 to an unsigned int *.
12462 (vect_analyze_data_refs): Change min_vf from an int * to a
12464 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12465 than an unsigned HOST_WIDE_INT.
12466 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12467 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12468 to an unsigned int *.
12469 (vect_analyze_data_ref_dependences): Likewise.
12470 (vect_compute_data_ref_alignment): Handle polynomial vf.
12471 (vect_enhance_data_refs_alignment): Likewise.
12472 (vect_prune_runtime_alias_test_list): Likewise.
12473 (vect_shift_permute_load_chain): Likewise.
12474 (vect_supportable_dr_alignment): Likewise.
12475 (dependence_distance_ge_vf): Take the vectorization factor as a
12476 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12477 (vect_analyze_data_refs): Change min_vf from an int * to a
12479 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12480 vfm1 as a poly_uint64 rather than an int. Make the same change
12481 for the returned bound_scalar.
12482 (vect_gen_vector_loop_niters): Handle polynomial vf.
12483 (vect_do_peeling): Likewise. Update call to
12484 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12485 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12487 * tree-vect-loop.c (vect_determine_vectorization_factor)
12488 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12489 (vect_get_known_peeling_cost): Likewise.
12490 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12491 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12492 (vect_transform_loop): Likewise. Use the lowest possible VF when
12493 updating the upper bounds of the loop.
12494 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12495 rather than an int.
12496 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12497 polynomial unroll factors.
12498 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12499 (vect_make_slp_decision): Likewise.
12500 (vect_supported_load_permutation_p): Likewise, and polynomial
12502 (vect_analyze_slp_cost): Handle polynomial vf.
12503 (vect_slp_analyze_node_operations): Likewise.
12504 (vect_slp_analyze_bb_1): Likewise.
12505 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12506 than an unsigned HOST_WIDE_INT.
12507 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12508 (vectorizable_load): Handle polynomial vf.
12509 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12511 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12514 Alan Hayward <alan.hayward@arm.com>
12515 David Sherwood <david.sherwood@arm.com>
12517 * match.pd: Handle bit operations involving three constants
12518 and try to fold one pair.
12520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12522 * tree-vect-loop-manip.c: Include gimple-fold.h.
12523 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12524 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12525 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12526 Add a path that uses a step of VF instead of 1, but disable it
12528 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12529 and niters_no_overflow parameters. Update calls to
12530 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12531 Create a new SSA name if the latter choses to use a ste other
12532 than zero, and return it via niters_vector_mult_vf_var.
12533 * tree-vect-loop.c (vect_transform_loop): Update calls to
12534 vect_do_peeling, vect_gen_vector_loop_niters and
12535 slpeel_make_loop_iterate_ntimes.
12536 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12537 (vect_gen_vector_loop_niters): Update declarations after above changes.
12539 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12541 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12542 128-bit round to integer instructions.
12543 (ceil<mode>2): Likewise.
12544 (btrunc<mode>2): Likewise.
12545 (round<mode>2): Likewise.
12547 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12549 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12550 unaligned VSX load/store on P8/P9.
12551 (expand_block_clear): Allow the use of unaligned VSX
12552 load/store on P8/P9.
12554 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12556 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12558 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12559 swap associated with both a load and a store.
12561 2018-01-02 Andrew Waterman <andrew@sifive.com>
12563 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12564 * config/riscv/riscv.md (clear_cache): Use it.
12566 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12568 * web.c: Remove out-of-date comment.
12570 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12572 * expr.c (fixup_args_size_notes): Check that any existing
12573 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12574 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12575 (emit_single_push_insn): ...here.
12577 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12579 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12580 (const_vector_encoded_nelts): New function.
12581 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12582 (const_vector_int_elt, const_vector_elt): Declare.
12583 * emit-rtl.c (const_vector_int_elt_1): New function.
12584 (const_vector_elt): Likewise.
12585 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12586 of CONST_VECTOR_ELT.
12588 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12590 * expr.c: Include rtx-vector-builder.h.
12591 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12592 directly on the tree encoding.
12593 (const_vector_from_tree): Likewise.
12594 * optabs.c: Include rtx-vector-builder.h.
12595 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12596 sequence of "u" values.
12597 * vec-perm-indices.c: Include rtx-vector-builder.h.
12598 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12599 directly on the vec_perm_indices encoding.
12601 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12603 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12604 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12605 * rtx-vector-builder.h: New file.
12606 * rtx-vector-builder.c: Likewise.
12607 * rtl.h (rtx_def::u2): Add a const_vector field.
12608 (CONST_VECTOR_NPATTERNS): New macro.
12609 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12610 (CONST_VECTOR_DUPLICATE_P): Likewise.
12611 (CONST_VECTOR_STEPPED_P): Likewise.
12612 (CONST_VECTOR_ENCODED_ELT): Likewise.
12613 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12614 (unwrap_const_vec_duplicate): Likewise.
12615 (const_vec_series_p): Check for a non-duplicated vector encoding.
12616 Say that the function only returns true for integer vectors.
12617 * emit-rtl.c: Include rtx-vector-builder.h.
12618 (gen_const_vec_duplicate_1): Delete.
12619 (gen_const_vector): Call gen_const_vec_duplicate instead of
12620 gen_const_vec_duplicate_1.
12621 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12622 (gen_const_vec_duplicate): Use rtx_vector_builder.
12623 (gen_const_vec_series): Likewise.
12624 (gen_rtx_CONST_VECTOR): Likewise.
12625 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12626 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12627 Build a new vector rather than modifying a CONST_VECTOR in-place.
12628 (handle_special_swappables): Update call accordingly.
12629 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12630 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12631 Build a new vector rather than modifying a CONST_VECTOR in-place.
12632 (handle_special_swappables): Update call accordingly.
12634 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12636 * simplify-rtx.c (simplify_const_binary_operation): Use
12637 CONST_VECTOR_ELT instead of XVECEXP.
12639 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12641 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12642 the selector elements to be different from the data elements
12643 if the selector is a VECTOR_CST.
12644 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12645 ssizetype for the selector.
12647 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12649 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12650 before testing each element individually.
12651 * tree-vect-generic.c (lower_vec_perm): Likewise.
12653 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12655 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12656 * selftest-run-tests.c (selftest::run_tests): Call it.
12657 * vector-builder.h (vector_builder::operator ==): New function.
12658 (vector_builder::operator !=): Likewise.
12659 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12660 (vec_perm_indices::all_from_input_p): New function.
12661 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12662 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12663 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12664 instead of reading the VECTOR_CST directly. Detect whether both
12665 vector inputs are the same before constructing the vec_perm_indices,
12666 and update the number of inputs argument accordingly. Use the
12667 utility functions added above. Only construct sel2 if we need to.
12669 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12671 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12672 the broadcast of the low byte.
12673 (expand_mult_highpart): Use an explicit encoding for the permutes.
12674 * optabs-query.c (can_mult_highpart_p): Likewise.
12675 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12676 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12677 (vectorizable_bswap): Likewise.
12678 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12679 explicit encoding for the power-of-2 permutes.
12680 (vect_permute_store_chain): Likewise.
12681 (vect_grouped_load_supported): Likewise.
12682 (vect_permute_load_chain): Likewise.
12684 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12686 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12687 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12688 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12689 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12690 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12691 (vect_gen_perm_mask_any): Likewise.
12693 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12695 * int-vector-builder.h: New file.
12696 * vec-perm-indices.h: Include int-vector-builder.h.
12697 (vec_perm_indices): Redefine as an int_vector_builder.
12698 (auto_vec_perm_indices): Delete.
12699 (vec_perm_builder): Redefine as a stand-alone class.
12700 (vec_perm_indices::vec_perm_indices): New function.
12701 (vec_perm_indices::clamp): Likewise.
12702 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12703 (vec_perm_indices::new_vector): New function.
12704 (vec_perm_indices::new_expanded_vector): Update for new
12705 vec_perm_indices class.
12706 (vec_perm_indices::rotate_inputs): New function.
12707 (vec_perm_indices::all_in_range_p): Operate directly on the
12708 encoded form, without computing elided elements.
12709 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12710 encoding. Update for new vec_perm_indices class.
12711 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12712 the given vec_perm_builder.
12713 (expand_vec_perm_var): Update vec_perm_builder constructor.
12714 (expand_mult_highpart): Use vec_perm_builder instead of
12715 auto_vec_perm_indices.
12716 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12717 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12718 or double series encoding as appropriate.
12719 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12720 vec_perm_indices instead of auto_vec_perm_indices.
12721 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12722 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12723 (vect_permute_store_chain): Likewise.
12724 (vect_grouped_load_supported): Likewise.
12725 (vect_permute_load_chain): Likewise.
12726 (vect_shift_permute_load_chain): Likewise.
12727 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12728 (vect_transform_slp_perm_load): Likewise.
12729 (vect_schedule_slp_instance): Likewise.
12730 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12731 (vectorizable_mask_load_store): Likewise.
12732 (vectorizable_bswap): Likewise.
12733 (vectorizable_store): Likewise.
12734 (vectorizable_load): Likewise.
12735 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12736 vec_perm_indices instead of auto_vec_perm_indices. Use
12737 tree_to_vec_perm_builder to read the vector from a tree.
12738 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12739 vec_perm_builder instead of a vec_perm_indices.
12740 (have_whole_vector_shift): Use vec_perm_builder and
12741 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12742 truncation to calc_vec_perm_mask_for_shift.
12743 (vect_create_epilog_for_reduction): Likewise.
12744 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12745 from auto_vec_perm_indices to vec_perm_indices.
12746 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12747 instead of changing individual elements.
12748 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12749 the vector in d.perm.
12750 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12751 from auto_vec_perm_indices to vec_perm_indices.
12752 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12753 instead of changing individual elements.
12754 (arm_vectorize_vec_perm_const): Use new_vector to install
12755 the vector in d.perm.
12756 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12757 Update vec_perm_builder constructor.
12758 (rs6000_expand_interleave): Likewise.
12759 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12760 (rs6000_expand_interleave): Likewise.
12762 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12764 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12765 to qimode could truncate the indices.
12766 * optabs.c (expand_vec_perm_var): Likewise.
12768 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12770 * Makefile.in (OBJS): Add vec-perm-indices.o.
12771 * vec-perm-indices.h: New file.
12772 * vec-perm-indices.c: Likewise.
12773 * target.h (vec_perm_indices): Replace with a forward class
12775 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12776 * optabs.h: Include vec-perm-indices.h.
12777 (expand_vec_perm): Delete.
12778 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12779 (expand_vec_perm_const): Declare.
12780 * target.def (vec_perm_const_ok): Replace with...
12781 (vec_perm_const): ...this new hook.
12782 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12783 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12784 * doc/tm.texi: Regenerate.
12785 * optabs.def (vec_perm_const): Delete.
12786 * doc/md.texi (vec_perm_const): Likewise.
12787 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12788 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12789 expand_vec_perm for constant permutation vectors. Assert that
12790 the mode of variable permutation vectors is the integer equivalent
12791 of the mode that is being permuted.
12792 * optabs-query.h (selector_fits_mode_p): Declare.
12793 * optabs-query.c: Include vec-perm-indices.h.
12794 (selector_fits_mode_p): New function.
12795 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12796 is defined, instead of checking whether the vec_perm_const_optab
12797 exists. Use targetm.vectorize.vec_perm_const instead of
12798 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12799 fit in the vector mode before using a variable permute.
12800 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12801 vec_perm_indices instead of an rtx.
12802 (expand_vec_perm): Replace with...
12803 (expand_vec_perm_const): ...this new function. Take the selector
12804 as a vec_perm_indices rather than an rtx. Also take the mode of
12805 the selector. Update call to shift_amt_for_vec_perm_mask.
12806 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12807 Use vec_perm_indices::new_expanded_vector to expand the original
12808 selector into bytes. Check whether the indices fit in the vector
12809 mode before using a variable permute.
12810 (expand_vec_perm_var): Make global.
12811 (expand_mult_highpart): Use expand_vec_perm_const.
12812 * fold-const.c: Includes vec-perm-indices.h.
12813 * tree-ssa-forwprop.c: Likewise.
12814 * tree-vect-data-refs.c: Likewise.
12815 * tree-vect-generic.c: Likewise.
12816 * tree-vect-loop.c: Likewise.
12817 * tree-vect-slp.c: Likewise.
12818 * tree-vect-stmts.c: Likewise.
12819 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12821 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12822 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12823 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12824 (aarch64_vectorize_vec_perm_const): ...this new function.
12825 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12826 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12827 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12828 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12829 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12830 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12831 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12833 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12834 check for NEON modes.
12835 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12836 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12837 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12838 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12840 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12841 the old VEC_PERM_CONST conditions.
12842 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12843 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12844 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12845 (ia64_vectorize_vec_perm_const_ok): Merge into...
12846 (ia64_vectorize_vec_perm_const): ...this new function.
12847 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12848 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12849 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12850 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12851 * config/mips/mips.c (mips_expand_vec_perm_const)
12852 (mips_vectorize_vec_perm_const_ok): Merge into...
12853 (mips_vectorize_vec_perm_const): ...this new function.
12854 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12855 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12856 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12857 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12858 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12859 (rs6000_expand_vec_perm_const): Delete.
12860 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12862 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12863 (altivec_expand_vec_perm_const_le): Take each operand individually.
12864 Operate on constant selectors rather than rtxes.
12865 (altivec_expand_vec_perm_const): Likewise. Update call to
12866 altivec_expand_vec_perm_const_le.
12867 (rs6000_expand_vec_perm_const): Delete.
12868 (rs6000_vectorize_vec_perm_const_ok): Delete.
12869 (rs6000_vectorize_vec_perm_const): New function.
12870 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12871 an element count and rtx array.
12872 (rs6000_expand_extract_even): Update call accordingly.
12873 (rs6000_expand_interleave): Likewise.
12874 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12875 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12876 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12877 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12878 (rs6000_expand_vec_perm_const): Delete.
12879 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12880 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12881 (altivec_expand_vec_perm_const_le): Take each operand individually.
12882 Operate on constant selectors rather than rtxes.
12883 (altivec_expand_vec_perm_const): Likewise. Update call to
12884 altivec_expand_vec_perm_const_le.
12885 (rs6000_expand_vec_perm_const): Delete.
12886 (rs6000_vectorize_vec_perm_const_ok): Delete.
12887 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12888 reference to the SPE evmerge intructions.
12889 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12890 an element count and rtx array.
12891 (rs6000_expand_extract_even): Update call accordingly.
12892 (rs6000_expand_interleave): Likewise.
12893 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12894 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12896 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12898 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12900 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
12901 vector mode and that that mode matches the mode of the data
12903 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
12904 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
12905 directly using expand_vec_perm_1 when forcing selectors into
12907 (expand_vec_perm_var): New function, split out from expand_vec_perm.
12909 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12911 * optabs-query.h (can_vec_perm_p): Delete.
12912 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
12913 * optabs-query.c (can_vec_perm_p): Split into...
12914 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
12915 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
12916 particular selector is valid.
12917 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12918 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12919 (vect_grouped_load_supported): Likewise.
12920 (vect_shift_permute_load_chain): Likewise.
12921 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12922 (vect_transform_slp_perm_load): Likewise.
12923 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12924 (vectorizable_bswap): Likewise.
12925 (vect_gen_perm_mask_checked): Likewise.
12926 * fold-const.c (fold_ternary_loc): Likewise. Don't take
12927 implementations of variable permutation vectors into account
12928 when deciding which selector to use.
12929 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
12930 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
12931 with a false third argument.
12932 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
12933 to test whether the constant selector is valid and can_vec_perm_var_p
12934 to test whether a variable selector is valid.
12936 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12938 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
12939 * optabs-query.c (can_vec_perm_p): Likewise.
12940 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
12941 instead of vec_perm_indices.
12942 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
12943 (vect_gen_perm_mask_checked): Likewise,
12944 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
12945 (vect_gen_perm_mask_checked): Likewise,
12947 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12949 * optabs-query.h (qimode_for_vec_perm): Declare.
12950 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
12951 (qimode_for_vec_perm): ...this new function.
12952 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
12954 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12956 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
12957 does not have a conditional at the top.
12959 2018-01-02 Richard Biener <rguenther@suse.de>
12961 * ipa-inline.c (big_speedup_p): Fix expression.
12963 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12966 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
12969 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
12973 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
12974 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
12975 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
12976 cond_taken_branch_cost 3->4.
12978 2018-01-01 Jakub Jelinek <jakub@redhat.com>
12980 PR tree-optimization/83581
12981 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
12982 TODO_cleanup_cfg if any changes have been made.
12984 PR middle-end/83608
12985 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
12986 convert_modes if target mode has the right side, but different mode
12989 PR middle-end/83609
12990 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
12991 last argument when extracting from CONCAT. If either from_real or
12992 from_imag is NULL, use expansion through memory. If result is not
12993 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
12994 the parts directly to inner mode, if even that fails, use expansion
12997 PR middle-end/83623
12998 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
12999 check for bswap in mode rather than HImode and use that in expand_unop
13002 Copyright (C) 2018 Free Software Foundation, Inc.
13004 Copying and distribution of this file, with or without modification,
13005 are permitted in any medium without royalty provided the copyright
13006 notice and this notice are preserved.