1 2018-05-17 Richard Biener <rguenther@suse.de>
3 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
4 for pruning loop and prune defs feeding only already visited PHIs.
6 2018-05-17 Richard Biener <rguenther@suse.de>
8 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
10 2018-05-17 Bin Cheng <bin.cheng@arm.com>
11 Richard Biener <rguenther@suse.de>
13 PR tree-optimization/85793
14 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
17 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
19 * internal-fn.h (lookup_internal_fn): Declare
20 * internal-fn.c (lookup_internal_fn): New function.
21 * gimple.c (gimple_build_call_from_tree): Handle calls to
23 * gimple-pretty-print.c (dump_gimple_call): Print "." before
24 internal function names.
25 * tree-pretty-print.c (dump_generic_node): Likewise.
26 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
28 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
30 * gimple-fold.h (gimple_build): Make the function forms take
31 combined_fn rather than built_in_function.
32 (gimple_simplify): Likewise.
33 * gimple-match-head.c (gimple_simplify): Likewise.
34 * gimple-fold.c (gimple_build): Likewise.
35 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
36 rather than gimple_build_call_internal.
37 (get_initial_defs_for_reduction): Likewise.
38 (vect_create_epilog_for_reduction): Likewise.
39 (vectorizable_live_operation): Likewise.
41 2018-05-17 Martin Liska <mliska@suse.cz>
43 * gimple-ssa-sprintf.c (format_directive): Do not use
44 space in between 'G_' and '('.
46 2018-05-17 Jakub Jelinek <jakub@redhat.com>
49 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
50 even if the mask is not all ones.
53 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
55 (ix86_gimple_fold_builtin): Likewise.
58 * config/i386/i386.c: Include tree-vector-builder.h.
59 (ix86_vector_shift_count): New function.
60 (ix86_fold_builtin): Fold shift builtins by scalar count.
61 (ix86_gimple_fold_builtin): Likewise.
63 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
64 _mm512_setzero): New intrinsics.
66 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
67 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
69 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
70 code generation for cases where splatting a value is not useful.
71 * simplify-rtx.c (simplify_ternary_operation): Simplify
72 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
75 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
77 * config.gcc: Support "goldmont-plus".
78 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
80 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
81 PROCESSOR_GOLDMONT_PLUS.
82 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
83 (processor_target_table): Add "goldmont-plus".
84 (PTA_GOLDMONT_PLUS): Define.
85 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
86 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
87 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
88 (fold_builtin_cpu): Add "goldmont-plus".
89 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
90 (ix86_option_override_internal): Add "goldmont-plus".
91 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
92 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
93 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
94 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
96 2018-05-17 Richard Biener <rguenther@suse.de>
98 PR tree-optimization/85757
99 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
100 remove defs that only feed that PHI from further processing.
102 2018-05-16 Jim Wilson <jimw@sifive.com>
104 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
106 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
108 2018-05-16 Mark Wielaard <mark@klomp.org>
110 * dwarf2out.c (count_index_strings): New function.
111 (output_indirect_strings): Call count_index_strings and generate
112 header for dwarf_version >= 5.
114 2018-05-16 Mark Wielaard <mark@klomp.org>
116 * dwarf2out.c (dwarf_FORM): New function.
117 (set_indirect_string): Use dwarf_FORM.
118 (reset_indirect_string): Likewise.
119 (size_of_die): Likewise.
120 (value_format): Likewise.
121 (output_die): Likewise.
122 (add_skeleton_AT_string): Likewise.
123 (output_macinfo_op): Likewise.
124 (index_string): Likewise.
125 (output_index_string_offset): Likewise.
126 (output_index_string): Likewise.
127 (count_index_strings): Likewise.
129 2018-05-16 Carl Love <cel@us.ibm.com>
131 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
132 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
134 2018-05-16 Martin Jambor <mjambor@suse.cz>
136 * ipa-prop.c (ipa_free_all_edge_args): Remove.
137 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
139 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
141 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
142 (fnma<mode>4): Likewise.
143 (fms<mode>4): Likewise.
144 (fnms<mode>4): Likewise.
145 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
146 (aarch64_fnma<mode>4): Likewise.
147 (aarch64_fms<mode>4): Likewise.
148 (aarch64_fnms<mode>4): Likewise.
149 (aarch64_fnmadd<mode>4): Likewise.
151 2018-05-16 Jason Merrill <jason@redhat.com>
153 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
155 2018-05-16 Richard Biener <rguenther@suse.de>
157 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
158 (dump_stmt_cost): Declare.
159 (add_stmt_cost): Dump cost we add.
160 (add_stmt_costs): New function.
161 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
163 (vect_analyze_stmt): Adjust prototype.
164 (vectorizable_condition): Likewise.
165 (vectorizable_live_operation): Likewise.
166 (vectorizable_reduction): Likewise.
167 (vectorizable_induction): Likewise.
168 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
169 cost vector to pass to vectorizable_ and record afterwards.
170 (vect_model_reduction_cost): Take cost vector argument and adjust.
171 (vect_model_induction_cost): Likewise.
172 (vectorizable_reduction): Likewise.
173 (vectorizable_induction): Likewise.
174 (vectorizable_live_operation): Likewise.
175 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
176 SLP_TREE_NUMBER_OF_VEC_STMTS.
177 (vect_analyze_slp_cost_1): Remove.
178 (vect_analyze_slp_cost): Likewise.
179 (vect_slp_analyze_node_operations): Take visited args and
180 a target cost vector. Avoid processing already visited stmt sets.
181 (vect_slp_analyze_operations): Use a local cost vector to gather
182 costs and register those of non-discarded instances.
183 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
184 (vect_schedule_slp_instance): Remove copying of
185 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
187 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
188 adding cost. Record cost entry location.
189 (vect_prologue_cost_for_slp_op): Function to compute cost of
190 a constant or invariant generated for SLP vect in the prologue,
191 split out from vect_analyze_slp_cost_1.
192 (vect_model_simple_cost): Make static. Adjust for SLP costing.
193 (vect_model_promotion_demotion_cost): Likewise.
194 (vect_model_store_cost): Likewise, make static.
195 (vect_model_load_cost): Likewise.
196 (vectorizable_bswap): Add cost vector arg and adjust.
197 (vectorizable_call): Likewise.
198 (vectorizable_simd_clone_call): Likewise.
199 (vectorizable_conversion): Likewise.
200 (vectorizable_assignment): Likewise.
201 (vectorizable_shift): Likewise.
202 (vectorizable_operation): Likewise.
203 (vectorizable_store): Likewise.
204 (vectorizable_load): Likewise.
205 (vectorizable_condition): Likewise.
206 (vectorizable_comparison): Likewise.
207 (can_vectorize_live_stmts): Likewise.
208 (vect_analyze_stmt): Likewise.
209 (vect_transform_stmt): Adjust calls to vectorizable_*.
210 * tree-vectorizer.c: Include gimple-pretty-print.h.
211 (dump_stmt_cost): New function.
213 2018-05-16 Richard Biener <rguenther@suse.de>
215 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
216 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
217 * tree-ssa-dse.c: Include tree-ssa-loop.h.
218 (check_name): New callback.
219 (dse_classify_store): Track cycles via a visited bitmap of PHI
220 defs and simplify handling of in-loop and across loop dead stores
221 and properly fail for loop-variant refs. Handle byte-tracking with
222 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
225 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
227 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
228 (vect_get_mask_type_for_stmt): Likewise.
229 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
231 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
232 to determine the statement's vector type and the vector type that
233 should be used for calculating nunits. Deal with cases in which
234 the type has to be deferred.
235 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
236 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
237 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
238 (vect_determine_vf_for_stmt): New functions, split out from...
239 (vect_determine_vectorization_factor): ...here.
240 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
241 (vect_get_mask_type_for_stmt): New functions, split out from
242 vect_determine_vectorization_factor.
244 2018-05-16 Richard Biener <rguenther@suse.de>
246 * tree-cfg.c (verify_gimple_assign_ternary): Properly
247 verify the [VEC_]COND_EXPR embedded comparison.
249 2018-05-15 Martin Sebor <msebor@redhat.com>
251 PR tree-optimization/85753
252 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
253 RECORD_TYPE in addition to ARRAY_TYPE.
255 2018-05-15 Martin Sebor <msebor@redhat.com>
258 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
260 2018-05-15 Richard Biener <rguenther@suse.de>
262 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
263 add by_clobber_p one. Change algorithm to collect all defs
264 representing uses we need to walk and try reducing them to
265 a single one before failing.
266 (dse_dom_walker::dse_optimize_stmt): Adjust.
268 2018-05-13 Mark Wielaard <mark@klomp.org>
270 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
271 (size_of_loc_descr): Likewise.
272 (output_loc_operands): Likewise.
273 (output_loc_operands_raw): Likewise.
274 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
275 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
276 (hash_loc_operands): Likewise.
277 (compare_loc_operands): Likewise.
279 2018-05-14 Mark Wielaard <mark@klomp.org>
281 * dwarf2out.c (count_index_addrs): New function.
282 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
284 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
286 PR tree-optimization/83648
287 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
288 return value as malloc candidate.
290 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
293 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
294 as true in call to suggest_attribute.
296 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
298 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
301 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
303 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
305 (aarch64_crypto_aesd_fused): Likewise.
307 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
309 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
310 (movsi_aarch64): Likewise.
311 (load_pairsi): Likewise.
312 (load_pairdi): Likewise.
313 (store_pairsi): Likewise.
314 (store_pairdi): Likewise.
315 (load_pairsf): Likewise.
316 (load_pairdf): Likewise.
317 (store_pairsf): Likewise.
318 (store_pairdf): Likewise.
319 (zero_extend): Likewise.
320 (trunc): Swap alternatives.
321 (fcvt_target): Add '?' to prefer w over r.
323 2018-05-14 Jakub Jelinek <jakub@redhat.com>
326 * config/i386/i386.md: Disallow non-commutative arithmetics in
327 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
328 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
329 in the peephole2 before it.
331 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
333 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
334 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
335 (ix86_handle_option): Handle -mcldemote.
336 * config.gcc: New header.
337 * config/i386/cldemoteintrin.h: New file.
338 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
339 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
341 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
342 OPTION_MASK_ISA_CLDEMOTE.
343 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
344 (ix86_valid_target_attribute_inner_p): Ditto.
345 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
346 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
347 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
348 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
349 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
351 * config/i386/i386.opt: Add -mcldemote.
352 * config/i386/x86intrin.h: New header.
353 * doc/invoke.texi: Add -mcldemote.
355 2018-05-14 Richard Biener <rguenther@suse.de>
357 * doc/match-and-simplify.texi: Adjust :s documentation.
359 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
361 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
362 intended memcpy size.
363 (REORDER_45): Likewise.
365 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
368 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
369 * vec.c (qsort_chk): Use gcc_qsort.
370 * Makefile.in (OBJS-libcommon): Add sort.o.
371 (build/sort.o): New target. Use it...
372 (BUILD_RTL): ... here, and...
373 (build/gencfn-macros): ... here, and...
374 (build/genmatch): ... here.
376 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
377 Chung-Ju Wu <jasonwucj@gmail.com>
379 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
380 * config/nds32/nds32-graywolf.md: New file.
381 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
382 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
384 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
385 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
386 * config/nds32/nds32.md (pipeline_model): Add graywolf.
387 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
388 * config/nds32/pipelines.md: Include n15 settings.
390 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
391 Chung-Ju Wu <jasonwucj@gmail.com>
393 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
394 * config/nds32/nds32-n13.md: New file.
395 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
396 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
398 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
399 * config/nds32/nds32.md (pipeline_model): Add n13.
400 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
401 * config/nds32/pipelines.md: Include n13 settings.
403 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
404 Chung-Ju Wu <jasonwucj@gmail.com>
406 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
407 * config/nds32/nds32-n10.md: New file.
408 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
409 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
411 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
412 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
413 * config/nds32/nds32.md (pipeline_model): Add n10.
414 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
415 * config/nds32/pipelines.md: Include n10 settings.
417 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
418 Kito Cheng <kito.cheng@gmail.com>
419 Chung-Ju Wu <jasonwucj@gmail.com>
421 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
422 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
423 Add enum values for DSP extension instructions.
424 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
426 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
427 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
429 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
430 * config/nds32/nds32-dspext.md: New file for DSP implementation.
431 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
432 * config/nds32/nds32-intrinsic.md: Likewise.
433 * config/nds32/nds32_intrinsic.h: Likewise.
434 * config/nds32/nds32-md-auxiliary.c: Likewise.
435 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
436 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
437 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
438 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
439 * config/nds32/nds32-protos.h: New declarations for DSP extension.
440 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
441 TYPE_DMAC in switch statement.
442 * config/nds32/nds32.c: New checking and implementation for DSP
443 extension instructions.
444 * config/nds32/nds32.h: Likewise.
445 * config/nds32/nds32.md: Likewise.
446 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
447 * config/nds32/predicates.md: Implement new predicates for DSP
450 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
452 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
453 Reformat alternatives and attributes so it is easier to identify
454 which constraints/attributes go with which instruction.
455 (mov<mode>_hardfloat32, FMOVE64): Likewise.
456 (mov<mode>_softfloat32, FMOVE64): Likewise.
457 (mov<mode>_hardfloat64, FMOVE64): Likewise.
458 (mov<mode>_softfloat64, FMOVE64): Likewise.
460 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
462 * doc/extend.texi (PowerPC Built-in Functions): Rename this
464 (Basic PowerPC Built-in Functions): The new name of the
465 subsection previously known as "PowerPC Built-in Functions".
466 (Basic PowerPC Built-in Functions Available on all Configurations):
468 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
469 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
470 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
471 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
473 2018-05-11 Martin Jambor <mjambor@suse.cz>
476 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
479 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
482 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
484 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
486 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
487 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
488 (ix86_handle_option): Handle -mwaitpkg.
489 * config.gcc: New header.
490 * config/i386/cpuid.h (bit_WAITPKG): New bit.
491 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
492 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
494 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
495 OPTION_MASK_ISA_WAITPKG.
496 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
497 (ix86_option_override_internal): Add PTA_WAITPKG.
498 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
499 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
501 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
502 __builtin_ia32_umwait and __builtin_ia32_tpause.
503 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
504 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
505 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
506 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
507 UNSPECV_TPAUSE): New.
508 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
509 * config/i386/i386.opt: Add -mwaitpkg.
510 * config/i386/waitpkgintrin.h: New file.
511 * config/i386/x86intrin.h: New header.
512 * doc/invoke.texi: Add -mwaitpkg.
514 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
517 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
519 (cortex-m0): Use armv6s-m isa.
520 (cortex-m0plus): Likewise.
521 (cortex-m1): Likewise.
522 (cortex-m0.small-multiply): Likewise.
523 (cortex-m0plus.small-multiply): Likewise.
524 (cortex-m1.small-multiply): Likewise.
526 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
527 Jakub Jelinek <jakub@redhat.com>
529 PR tree-optimization/85692
530 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
531 source permute as well.
533 2018-05-11 Martin Liska <mliska@suse.cz>
536 * doc/extend.texi: Document LLVM style format for no_sanitize
539 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
541 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
542 mode_supports_vsx_dform_quad to mode_supports_dq_form.
543 (mode_supports_vsx_dform_quad): Likewise.
544 (mode_supports_vmx_dform): Move these functions to be next to the
545 other mode_supports functions.
546 (mode_supports_dq_form): Likewise.
547 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
548 mode_supports_dq_form.
549 (reg_offset_addressing_ok_p): Likewise.
550 (offsettable_ok_by_alignment): Likewise.
551 (rs6000_legitimate_offset_address_p): Likewise.
552 (legitimate_lo_sum_address_p): Likewise.
553 (rs6000_legitimize_address): Likewise.
554 (rs6000_legitimize_reload_address): Likewise.
555 (rs6000_secondary_reload_inner): Likewise.
556 (rs6000_preferred_reload_class): Likewise.
557 (rs6000_output_move_128bit): Likewise.
559 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
561 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
562 Generate SImode target register for null target.
563 <case IX86_BUILTIN_XGETBV>: Ditto.
564 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
565 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
567 2018-05-10 Carl Love <cel@us.ibm.com>
569 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
570 dcbtt and dcbtstt if operands[2] is 0.
572 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
575 * config/i386/sse.md (usadv64qi): New expander.
577 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
579 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
580 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
581 -maltivec=be support.
582 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
583 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
584 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
585 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
586 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
587 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
588 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
589 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
590 altivec_vsumsws): Adjust.
591 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
592 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
593 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
595 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
596 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
597 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
598 (altivec_lve<VI_char>x): Delete expand.
599 (*altivec_lve<VI_char>x_internal): Rename to...
600 (altivec_lve<VI_char>x): ... this.
601 (altivec_lvxl_<mode>): Delete expand.
602 (*altivec_lvxl_<mode>_internal): Rename to ...
603 (altivec_lvxl_<mode>): ... this.
604 (altivec_stvxl_<mode>): Delete expand.
605 (*altivec_stvxl_<mode>_internal): Rename to ...
606 (altivec_stvxl_<mode>): ... this.
607 (altivec_stve<VI_char>x): Delete expand.
608 (*altivec_stve<VI_char>x_internal): Rename to ...
609 (altivec_stve<VI_char>x): ... this.
610 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
611 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
612 reduc_plus_scal_<mode>): Adjust.
613 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
615 (rs6000_cpu_cpp_builtins): Adjust.
616 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
617 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
618 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
619 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
620 -maltivec=be support.
621 (rs6000_split_vec_extract_var): Adjust.
622 (rs6000_split_v4si_init): Adjust.
623 (swap_selector_for_mode): Delete.
624 (altivec_expand_lvx_be, altivec_expand_stvx_be,
625 altivec_expand_stvex_be): Delete.
626 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
627 -maltivec=be support.
628 (rs6000_gimple_fold_builtin): Ditto.
629 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
631 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
632 (TARGET_DIRECT_MOVE_64BIT): Adjust.
633 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
634 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
635 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
636 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
637 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
638 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
639 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
640 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
641 anonymous split): Adjust.
642 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
643 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
645 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
647 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
648 when --with-gxx-include-dir is also specified.
649 * configure: Regenerate.
651 2018-05-09 Jim Wilson <jimw@sifive.com>
654 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
655 * config/riscv/t-withmultilib: New.
656 * config/riscv/withmultilib.h: New.
657 * doc/install.texi: Document RISC-V --with-multilib-list support.
659 2018-05-09 Richard Biener <rguenther@suse.de>
661 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
663 (vect_bb_vectorization_profitable_p): Adjust. Compute
664 actual scalar cost using the cost vector and the add_stmt_cost
667 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
669 PR rtl-optimization/85645
670 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
671 in the REG_CFA_REGISTER note for LR, don't leave it empty.
673 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
675 PR rtl-optimization/85645
676 * shrink-wrap.c (spread_components): Return a boolean saying if
677 anything was changed.
678 (try_shrink_wrapping_separate): Iterate spread_components until
679 nothing changes anymore.
681 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
683 PR rtl-optimization/85645
684 * regrename.c (build_def_use): Also kill the chains that include the
685 destination of a REG_CFA_REGISTER note.
687 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
689 PR rtl-optimization/85645
690 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
691 insn that has a REG_CFA_REGISTER note.
693 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
695 * cfgexpand.c (expand_clobber): New function.
696 (expand_gimple_stmt_1): Use it.
697 * tree-vect-stmts.c (vect_clobber_variable): New function,
699 (vectorizable_simd_clone_call): ...here.
700 (vectorizable_store): Emit a clobber either side of an
701 IFN_STORE_LANES sequence.
702 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
704 2018-05-09 Tom de Vries <tom@codesourcery.com>
707 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
708 (define_insn "trap_if_false"): Add exit after trap.
710 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
712 PR rtl-optimization/85638
713 * bb-reorder.c: Include common/common-target.h.
714 (create_forwarder_block): New function extracted from...
715 (fix_up_crossing_landing_pad): ...here. Rename into...
716 (dw2_fix_up_crossing_landing_pad): ...this.
717 (sjlj_fix_up_crossing_landing_pad): New function.
718 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
719 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
720 from both partitions and exit the loop after one iteration.
722 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
725 * doc/extend.texi (PowerPC Built-in Functions): Rename this
727 (Basic PowerPC Built-in Functions): The new name of the
728 subsection previously known as "PowerPC Built-in Functions".
729 (Basic PowerPC Built-in Functions Available on all Configurations):
731 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
733 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
735 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
737 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
740 2018-05-08 Jim Wilson <jimw@sifive.com>
742 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
743 (LD_EMUL_SUFFIX): New.
746 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
748 * doc/extend.texi (PowerPC Built-in Functions): Rename this
750 (Basic PowerPC Built-in Functions): The new name of the
751 subsection previously known as "PowerPC Built-in Functions".
752 (Basic PowerPC Built-in Functions Available on all Configurations):
754 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
756 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
758 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
760 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
763 2018-05-08 Jakub Jelinek <jakub@redhat.com>
766 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
767 after cmpelim optimization.
769 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
771 * config.gcc: Support "goldmont".
772 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
773 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
775 * config/i386/i386.c (m_GOLDMONT): Define.
776 (processor_target_table): Add "goldmont".
777 (PTA_GOLDMONT): Define.
778 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
779 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
780 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
781 (fold_builtin_cpu): Add "goldmont".
782 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
783 (ix86_option_override_internal): Add "goldmont".
784 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
785 (processor_type): Add PROCESSOR_GOLDMONT.
786 * config/i386/i386.md: Add CPU "glm".
787 * config/i386/glm.md: New file.
788 * config/i386/x86-tune.def: Add m_GOLDMONT.
789 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
791 2018-05-08 Jakub Jelinek <jakub@redhat.com>
794 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
796 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
797 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
798 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
801 * config/i386/i386.c (ix86_fold_builtin): Handle
802 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
805 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
806 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
808 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
811 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
812 (check_arch): Likewise.
813 (check_fpu): Return the result rather than printing it.
814 (end arch): Fix operator precedence.
816 (END): Print the result from check_fpu.
818 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
819 Alan Hayward <alan.hayward@arm.com>
820 David Sherwood <david.sherwood@arm.com>
822 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
823 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
824 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
825 (*fcmuo<mode>_and): New patterns.
827 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
829 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
830 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
831 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
832 (cmp_op, sve_imm_con): New code attributes.
833 (SVE_COND_INT_CMP, imm_con): Delete.
834 (cmp_op): Remove above unspecs from int attribute.
835 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
837 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
838 comparison-specific unspecs.
839 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
840 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
841 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
842 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
843 (*vec_fcm<cmp_op><mode>): Rename to...
844 (*fcm<cmp_op><mode>): ...this and adjust likewise.
845 (*vec_fcmuo<mode>): Rename to...
846 (*fcmuo<mode>): ...this and adjust likewise.
847 (*pred_fcm<cmp_op><mode>): New pattern.
848 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
849 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
851 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
853 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
854 (aarch64_emit_sve_predicated_cond): New function.
855 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
856 (aarch64_emit_unspec_cond_or): Replace with...
857 (aarch64_emit_sve_or_conds): ...this new function. Use
858 aarch64_emit_sve_ptrue_op for the individual comparisons and
859 aarch64_emit_binop to OR them together.
860 (aarch64_emit_inverted_unspec_cond): Replace with...
861 (aarch64_emit_sve_inverted_cond): ...this new function. Use
862 aarch64_emit_sve_ptrue_op for the comparison and
863 aarch64_emit_unop to invert the result.
864 (aarch64_expand_sve_vec_cmp_float): Update after the above
865 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
867 2018-05-07 Nathan Sidwell <nathan@acm.org>
869 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
870 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
871 (Backwards Compatibility): Likewise.
873 2018-05-07 Luis Machado <luis.machado@linaro.org>
877 2018-05-07 Luis Machado <luis.machado@linaro.org>
879 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
880 <prefetch_dynamic_strides>: New const bool field.
881 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
882 prefetch_dynamic_strides.
883 (exynosm1_prefetch_tune): Likewise.
884 (thunderxt88_prefetch_tune): Likewise.
885 (thunderx_prefetch_tune): Likewise.
886 (thunderx2t99_prefetch_tune): Likewise.
887 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
888 (aarch64_override_options_internal): Update to set
889 PARAM_PREFETCH_DYNAMIC_STRIDES.
890 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
891 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
892 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
893 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
894 prefetch-dynamic-strides setting.
896 2018-05-07 Luis Machado <luis.machado@linaro.org>
898 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
899 <minimum_stride>: New const int field.
900 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
901 minimum_stride field.
902 (exynosm1_prefetch_tune): Likewise.
903 (thunderxt88_prefetch_tune): Likewise.
904 (thunderx_prefetch_tune): Likewise.
905 (thunderx2t99_prefetch_tune): Likewise.
906 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
907 (aarch64_override_options_internal): Update to set
908 PARAM_PREFETCH_MINIMUM_STRIDE.
909 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
910 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
911 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
912 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
913 stride is constant and is below the minimum stride threshold.
915 2018-05-07 Luis Machado <luis.machado@linaro.org>
917 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
920 2018-05-07 Luis Machado <luis.machado@linaro.org>
922 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
923 <prefetch_dynamic_strides>: New const bool field.
924 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
925 prefetch_dynamic_strides.
926 (exynosm1_prefetch_tune): Likewise.
927 (thunderxt88_prefetch_tune): Likewise.
928 (thunderx_prefetch_tune): Likewise.
929 (thunderx2t99_prefetch_tune): Likewise.
930 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
931 (aarch64_override_options_internal): Update to set
932 PARAM_PREFETCH_DYNAMIC_STRIDES.
933 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
934 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
935 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
936 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
937 prefetch-dynamic-strides setting.
939 2018-05-07 Luis Machado <luis.machado@linaro.org>
941 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
942 <minimum_stride>: New const int field.
943 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
944 minimum_stride field.
945 (exynosm1_prefetch_tune): Likewise.
946 (thunderxt88_prefetch_tune): Likewise.
947 (thunderx_prefetch_tune): Likewise.
948 (thunderx2t99_prefetch_tune): Likewise.
949 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
950 (aarch64_override_options_internal): Update to set
951 PARAM_PREFETCH_MINIMUM_STRIDE.
952 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
953 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
954 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
955 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
956 stride is constant and is below the minimum stride threshold.
958 2018-05-06 Jakub Jelinek <jakub@redhat.com>
961 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
962 the type is addressable. Don't force op into register if it has
965 2018-05-05 Roland McGrath <mcgrathr@google.com>
968 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
969 any section for which we don't know a specific type it should have,
970 regardless of name. Previously this was done only for the exact
971 names ".init_array", ".fini_array", and ".preinit_array".
972 (default_elf_asm_named_section): Add comment about
973 relationship with default_section_type_flags and SECTION_NOTYPE.
974 (get_section): Don't consider it a type conflict if one side has
975 SECTION_NOTYPE and the other doesn't, as long as neither has the
976 SECTION_BSS et al used in the default_section_type_flags logic.
978 2018-05-05 Tom de Vries <tom@codesourcery.com>
981 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
982 (workaround_barsyncs): New function.
983 (nvptx_reorg): Use workaround_barsyncs.
984 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
985 (define_expand "nvptx_membar_cta"): New define_expand.
986 (define_insn "*nvptx_membar_cta"): New insn.
988 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
990 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
991 To improve optimization opportunities.
992 * builtin-types.def: The new needed builtin types for the above.
994 2018-05-04 Richard Biener <rguenther@suse.de>
996 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
997 * gimple-ssa-store-merging.c
998 (imm_store_chain_info::output_merged_store): Remove redundant create,
999 release split_store vector contents on failure.
1000 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
1001 scalar stmt vector on cache hit.
1003 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
1005 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
1007 * config.gcc (powerpc-xilinx-eabi*): Remove.
1008 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
1010 (fusion_addis_mem_combo_load): Ditto.
1011 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
1013 (rs6000_cpu_cpp_builtins): Ditto.
1014 * config/rs6000/rs6000-linux.c
1015 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
1016 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
1017 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
1019 (rs6000_setup_reg_addr_masks): Ditto.
1020 (rs6000_init_hard_regno_mode_ok): Ditto.
1021 (rs6000_option_override_internal): Ditto.
1022 (legitimate_lo_sum_address_p): Ditto.
1023 (rs6000_legitimize_address): Ditto.
1024 (rs6000_legitimize_reload_address): Ditto.
1025 (rs6000_legitimate_address_p): Ditto.
1026 (abi_v4_pass_in_fpr): Ditto.
1027 (setup_incoming_varargs): Ditto.
1028 (rs6000_gimplify_va_arg): Ditto.
1029 (rs6000_split_multireg_move): Ditto.
1030 (rs6000_savres_strategy): Ditto.
1031 (rs6000_emit_prologue_components): Ditto.
1032 (rs6000_emit_epilogue_components): Ditto.
1033 (rs6000_emit_prologue): Ditto.
1034 (rs6000_emit_epilogue): Ditto.
1035 (rs6000_elf_file_end): Ditto.
1036 (rs6000_function_value): Ditto.
1037 (rs6000_libcall_value): Ditto.
1038 * config/rs6000/rs6000.h: Ditto.
1039 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
1040 (TARGET_MINMAX): ... this. New.
1041 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
1042 * config/rs6000/rs6000.md: Remove Xilinx FP support.
1043 (*movsi_internal1_single): Delete.
1044 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
1045 mfpu=, mxilinx-fpu): Delete.
1046 * config/rs6000/singlefp.h: Delete.
1047 * config/rs6000/sysv4.h: Remove Xilinx FP support.
1048 * config/rs6000/t-rs6000: Ditto.
1049 * config/rs6000/t-xilinx: Delete.
1050 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
1051 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
1052 (VStype_simple): Delete.
1053 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
1054 * config/rs6000/xfpu.h: Delete.
1055 * config/rs6000/xfpu.md: Delete.
1056 * config/rs6000/xilinx.h: Delete.
1057 * config/rs6000/xilinx.opt: Delete.
1058 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
1059 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
1061 2018-05-04 Tom de Vries <tom@codesourcery.com>
1064 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
1067 2018-05-04 Richard Biener <rguenther@suse.de>
1070 * tree-complex.c (update_complex_assignment): We are always in SSA form.
1071 (expand_complex_div_wide): Likewise.
1072 (expand_complex_operations_1): Likewise.
1073 (expand_complex_libcall): Preserve EH info of the original stmt.
1074 (tree_lower_complex): Handle removed blocks.
1075 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
1076 on complex multiplication and division libcall builtins.
1078 2018-05-04 Richard Biener <rguenther@suse.de>
1081 * fold-const.c (negate_expr_p): Restrict negation of operand
1082 zero of a division to when we know that can happen without
1084 (fold_negate_expr_1): Likewise.
1086 2018-05-04 Jakub Jelinek <jakub@redhat.com>
1089 * real.h (real_nextafter): Declare.
1090 * real.c (real_nextafter): New function.
1091 * fold-const-call.c (fold_const_nextafter): New function.
1092 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
1093 CASE_CFN_NEXTTOWARD.
1094 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
1095 even when arg1_mode is different from arg0_mode.
1097 2018-05-03 Nathan Sidwell <nathan@acm.org>
1099 * doc/extend.texi (Deprecated Features): Remove
1101 (Backwards Compatibility): Likewise.
1102 * doc/invoke.texi (C++ Language Options): Likewise.
1103 (C++ Dialect Options): Likewise.
1105 2018-05-03 Jakub Jelinek <jakub@redhat.com>
1108 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
1109 _mm512_mask_mullox_epi64): New intrinsics.
1111 2018-05-03 Tom de Vries <tom@codesourcery.com>
1114 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1115 dump files): Add offload-tree.
1117 2018-05-03 Richard Biener <rguenther@suse.de>
1119 PR tree-optimization/85615
1120 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
1121 to loops not nested in BBs loop father to avoid creating multi-entry
1124 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1126 PR tree-optimization/70291
1127 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
1128 arguments. Change return type to tree. Emit libcall as a new
1129 statement rather than replacing existing one when inplace_p is true.
1130 (expand_complex_multiplication_components): New function.
1131 (expand_complex_multiplication): Expand floating-point complex
1132 multiplication using the above.
1133 (expand_complex_division): Rename inner_type parameter to type.
1134 Update expand_complex_libcall call-site.
1135 (expand_complex_operations_1): Update expand_complex_multiplication
1136 and expand_complex_division call-sites.
1138 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1141 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1142 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1143 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
1144 the highest significant bit of the shift count mask is clear. In
1145 check whether and[sq]i3 is needed verify that all significant bits
1146 of the shift count other than the highest are set.
1148 2018-05-02 Tom de Vries <tom@codesourcery.com>
1151 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
1152 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
1153 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
1154 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
1155 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
1156 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
1157 __builtin_goacc_parlevel_size.
1159 2018-05-02 Richard Biener <rguenther@suse.de>
1161 PR tree-optimization/85597
1162 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
1163 do not use split vect_get_vec_defs call but call vect_get_slp_defs
1166 2018-05-02 Tom de Vries <tom@codesourcery.com>
1169 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1170 dump files): Add ltrans-tree.
1172 2018-05-02 Tom de Vries <tom@codesourcery.com>
1175 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
1176 dump files): Add wpa-ipa.
1178 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
1180 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
1181 powerpc*-*-linux*paired* target.
1182 * config/rs6000/750cl.h: Delete.
1183 * config/rs6000/paired.h: Delete.
1184 * config/rs6000/paired.md: Delete.
1185 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
1187 * config/rs6000/rs6000-builtin.def: Remove paired float support.
1188 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
1189 comment. Remove paired float support.
1190 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
1191 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1193 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
1194 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
1196 * config/rs6000/rs6000.c: Remove paired float support.
1197 (paired_expand_vector_init, paired_expand_vector_move,
1198 paired_emit_vector_compare, paired_emit_vector_cond_expr,
1199 (paired_expand_lv_builtin, paired_expand_stv_builtin,
1200 paired_expand_builtin, paired_expand_predicate_builtin,
1201 paired_init_builtins): Delete.
1202 * config/rs6000/rs6000.h: Remove paired float support.
1203 * config/rs6000/rs6000.md: Remove paired float support.
1204 (move_from_CR_ov_bit): Delete.
1205 * config/rs6000/rs6000.opt (mpaired): Delete.
1206 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
1207 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
1209 2018-05-02 Richard Biener <rguenther@suse.de>
1212 * gimplify.c (gimplify_save_expr): When in SSA form allow
1213 SAVE_EXPRs to compute to SSA vars.
1215 2018-05-02 Jakub Jelinek <jakub@redhat.com>
1218 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
1219 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
1220 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
1221 clobber operands[2], instead use a new pseudo. Formatting fixes.
1223 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
1225 PR tree-optimization/85586
1226 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
1227 exit early for statements in the same group if the accesses are
1230 2018-05-02 Tom de Vries <tom@codesourcery.com>
1233 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
1236 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
1238 PR tree-optimization/85143
1239 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
1241 2018-05-01 Tom de Vries <tom@codesourcery.com>
1244 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
1245 not found" error message.
1247 2018-05-01 Tom de Vries <tom@codesourcery.com>
1250 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
1251 * vec.c (test_ordered_remove_if): New function.
1252 (vec_c_tests): Call test_ordered_remove_if.
1253 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
1254 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
1255 * tree-vect-patterns.c (vect_pattern_recog_1): Use
1256 VEC_ORDERED_REMOVE_IF.
1258 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1260 PR tree-optimization/82665
1261 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
1262 pointer subtraction where arguments come from a memchr call.
1264 2018-05-01 Jakub Jelinek <jakub@redhat.com>
1266 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
1267 --push-state --as-needed and --pop-state instead of --as-needed and
1268 --no-as-needed if ld supports it.
1269 * configure: Regenerated.
1272 * doc/install.texi2html: Replace _002d with - and _002a with * in
1273 generated html files using sed.
1275 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1278 * gcc-rich-location.c (blank_line_before_p): New function.
1279 (use_new_line): New function.
1280 (gcc_rich_location::add_fixit_insert_formatted): New function.
1281 * gcc-rich-location.h
1282 (gcc_rich_location::add_fixit_insert_formatted): New function.
1284 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1286 * selftest.c (assert_streq): Rename "expected" and "actual" to
1287 "val1" and "val2". Extend NULL-handling to cover both inputs
1288 symmetrically, while still requiring both to be non-NULL for a pass.
1289 * selftest.h (assert_streq): Rename "expected" and "actual" to
1291 (ASSERT_EQ): Likewise.
1292 (ASSERT_EQ_AT): Likewise.
1293 (ASSERT_KNOWN_EQ): Likewise.
1294 (ASSERT_KNOWN_EQ_AT): Likewise.
1295 (ASSERT_NE): Likewise.
1296 (ASSERT_MAYBE_NE): Likewise.
1297 (ASSERT_MAYBE_NE_AT): Likewise.
1298 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
1299 the assertion to pass.
1300 (ASSERT_STREQ_AT): Likewise.
1302 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1304 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
1305 interaction with -pie.
1307 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1309 * selftest.h: Fix alphabetization of per-source-file selftest
1312 2018-04-30 Jason Merrill <jason@redhat.com>
1314 PR c++/61982 - dead stores to destroyed objects.
1315 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
1318 2018-04-30 Jason Merrill <jason@redhat.com>
1320 * tree.c (build_clobber): New.
1321 * tree.h: Declare it.
1322 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
1324 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1326 * diagnostic-show-locus.c (layout::layout): Update for
1327 location_get_source_line returning a char_span.
1328 (struct char_span): Move to input.h.
1329 (struct correction): Update for fields in char_span becoming
1331 (struct source_line): Update for location_get_source_line
1332 returning a char_span.
1333 (layout::print_line): Likewise.
1334 * edit-context.c (edited_file::print_content): Likewise.
1335 (edited_file::print_diff_hunk): Likewise.
1336 (edited_file::print_run_of_changed_lines): Likewise.
1337 (edited_file::get_num_lines): Likewise.
1338 (edited_line::edited_line): Likewise.
1339 * final.c (asm_show_source): Likewise.
1340 * input.c (location_get_source_line): Convert return type
1341 from const char * to char_span, losing the final "line_len"
1343 (dump_location_info): Update for the above.
1344 (get_substring_ranges_for_loc): Likewise. Use a char_span
1345 when handling the literal within the line.
1346 (test_reading_source_line): Update for location_get_source_line
1347 returning a char_span.
1348 * input.h (class char_span): Move here from
1349 diagnostic-show-locus.c, converting from a struct to a class.
1350 Make data members private.
1351 (char_span::operator bool): New.
1352 (char_span::length): New.
1353 (char_span::get_buffer): New.
1354 (char_span::operator[]): New.
1355 (char_span::subspan): Make const.
1356 (char_span::xstrdup): New.
1357 (location_get_source_line): Convert return type from const char *
1358 to char_span, losing the final "line_size" param.
1360 2018-04-30 Jan Hubicka <jh@suse.cz>
1362 * lto-wrapper.c (ltrans_priorities): New static var.
1363 (cmp_priority): New.
1364 (run_gcc): Read priorities and if doing parallel build order
1365 the Makefile by them.
1367 2018-04-30 David Malcolm <dmalcolm@redhat.com>
1369 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
1371 2018-04-30 Richard Biener <rguenther@suse.de>
1373 * tree-cfg.c (verify_address): Remove base argument, add
1374 flag whether to check TREE_ADDRESSABLE and do that.
1375 (verify_expr): Remove.
1376 (verify_types_in_gimple_reference): Add pieces from verify_expr.
1377 (verify_gimple_assign_single): Likewise.
1378 (verify_gimple_switch): Likewise.
1379 (verify_expr_location_1): Dereference tp once. Add (disabled)
1380 piece from verify_expr.
1381 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
1383 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1385 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
1387 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1389 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
1390 (small_data_pattern): Likewise.
1391 (arc_rewrite_small_data): Likewise.
1392 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
1393 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
1394 (get_symbol_alignment): New function.
1395 (legitimate_small_data_address_p): Likewise.
1396 (legitimate_scaled_address): Update, call
1397 legitimate_small_data_address_p.
1398 (output_sdata): New static variable.
1399 (arc_print_operand): Update how we handle small data operands.
1400 (arc_print_operand_address): Likewise.
1401 (arc_legitimate_address_p): Update, use
1402 legitimate_small_data_address_p.
1403 (arc_rewrite_small_data_p): Remove.
1404 (arc_rewrite_small_data_1): Likewise.
1405 (arc_rewrite_small_data): Likewise.
1406 (small_data_pattern): Likewise.
1407 (compact_sda_memory_operand): Update to use
1408 legitimate_small_data_address_p and get_symbol_alignment.
1409 (prepare_move_operands): Don't rewite sdata pattern.
1410 (prepare_extend_operands): Remove.
1411 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
1413 (zero_extendqisi2): Likewise.
1414 (zero_extendhisi2): Likewise.
1415 (extendqihi2): Likewise.
1416 (extendqisi2): Likewise.
1417 (extendhisi2): Likewise.
1421 * config/arc/constraints.md (Usd): Change it to memory constraint.
1423 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
1425 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
1426 as source of std instructions.
1427 * config/arc/arc.md (movsi_insn): Update pattern predicate to
1428 allow 6-bit constants as source for store instructions.
1429 (movdi_insn): Update instruction pattern to allow 6-bit constants
1430 as source for store instructions.
1432 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
1434 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
1436 2018-04-30 Nathan Sidwell <nathan@acm.org>
1437 Sandra Loosemore <sandra@codesourcery.com>
1439 * dumpfile.c (dump_open): Allow '-' for stdout.
1440 * doc/invoke.texi (Developer Options): Document dump filename
1441 determination early. Document stdin/stdout selection.
1443 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
1445 Microblaze Target: PIC data text relative
1447 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
1448 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
1450 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
1451 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
1452 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
1453 New addressing mode for data-text relative position indepenedent code.
1454 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
1455 'ADDRESS_SYMBOLIC_TXT_REL'.
1456 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
1457 (microblaze_legitimate_pic_operand): Exclude function calls from
1458 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
1459 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
1461 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1462 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
1463 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
1464 for 'address + offset'.
1465 (microblaze_expand_prologue): Add new function prologue call for
1467 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
1468 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
1469 table in case of TARGET_PIC_DATA_TEXT_REL.
1470 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
1471 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
1472 Add new macros 'UNSPEC_TEXT',
1473 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
1474 + exclude function calls from 'UNSPEC_PLT' in case of data text
1476 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1477 new target hook for generating address diff vector tables in case of
1479 * doc/tm.texi : Regenerate.
1480 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
1481 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
1482 of addr diff vector generation.
1483 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
1484 target hook definition.
1485 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
1486 Add default function for generate_pic_addr_diff_vec -> flag_pic.
1487 * doc/invoke.texi (Add new pic option): Add new microblaze pic
1488 option for data text relative.
1490 2018-04-30 Richard Biener <rguenther@suse.de>
1492 * tree-chrec.h (evolution_function_is_constant_p): Remove
1494 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
1496 2018-04-30 Richard Biener <rguenther@suse.de>
1499 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
1501 2018-04-30 Richard Biener <rguenther@suse.de>
1503 PR tree-optimization/28364
1504 PR tree-optimization/85275
1505 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
1506 copying first exit test.
1508 2018-04-28 Mark Wielaard <mark@klomp.org>
1510 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
1512 (dwarf_AT): Handle DW_AT_addr_base.
1513 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
1515 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
1518 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
1519 (*ashl<dwi>3_doubleword_mask_1): Ditto.
1520 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
1521 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
1523 2018-04-28 Richard Biener <rguenther@suse.de>
1525 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
1526 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
1527 to reflect use. Only add interesting stmts.
1529 2018-04-27 Martin Jambor <mjambor@suse.cz>
1532 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
1533 the jump function allows for passing through aggregate values.
1535 2018-04-27 David Malcolm <dmalcolm@redhat.com>
1537 * input.h (in_system_header_at): Convert from macro to inline
1539 (from_macro_expansion_at): Likewise.
1540 (from_macro_definition_at): Likewise.
1542 2018-04-27 Jeff Law <law@redhat.com>
1544 * config.gcc: Mark tile* targets as deprecated/obsolete.
1546 2018-04-27 Richard Biener <rguenther@suse.de>
1548 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
1551 2018-04-27 Richard Biener <rguenther@suse.de>
1553 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
1555 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
1557 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
1558 with Yd constraint. Set "preferred_for_speed" attribute from
1559 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
1561 (*movdi_internal): Ditto.
1562 (movti_interunit splitters): Remove
1563 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
1564 (movdi_interunit splitters): Ditto.
1565 * config/i386/constraints.md (Ye): Remove.
1566 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
1568 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1571 * config/aarch64/constraints.md (Usg): Limit to 31.
1574 2018-04-27 Jakub Jelinek <jakub@redhat.com>
1576 PR tree-optimization/85529
1577 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
1578 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
1579 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
1580 zero extension or masking of the MSB bit.
1581 (optimize_range_tests): Add FIRST_BB argument, pass it through
1582 to optimize_range_tests_var_bound.
1583 (maybe_optimize_range_tests, reassociate_bb): Adjust
1584 optimize_range_tests callers.
1586 2018-04-26 Richard Biener <rguenther@suse.de>
1587 Jakub Jelinek <jakub@redhat.com>
1589 * cgraph.h (symbol_table): Just declare debug method here.
1590 * symtab.c (symbol_table::debug): Define.
1592 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
1594 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
1596 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
1598 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
1599 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
1600 (*movdi_internal): Substitute Yi and Yj constraint with x
1601 and Ym and Yn constraint with y constraint. Update "isa"
1602 attribute and set "preferred_for_speed" attribute from
1603 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
1604 (*movsi_internal): Ditto.
1605 (*movdf_internal): Ditto.
1606 (*movsf_internal): Ditto.
1607 (*zero_extendsidi2): Ditto.
1608 * config/i386/sse.md (vec_set<mode>_0): Ditto.
1609 (sse2_loadld): Ditto.
1610 (*vec_extract<ssevecmodelower>_0): Ditto.
1611 (*vec_extractv4si_0_zext_sse4): Ditto.
1612 (vec_concatv2di): Ditto.
1613 (*vec_dup<mode>): Ditto.
1614 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
1615 * config/i386/constraints.md (Yi): Remove.
1620 2018-04-26 Nathan Sidwell <nathan@acm.org>
1622 * dumpfile.c (dump_open): New.
1623 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
1624 (dump_finish): Detect stdio/stderr by value not name.
1626 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
1628 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
1630 2018-04-26 Tom de Vries <tom@codesourcery.com>
1633 * config/nvptx/nvptx.c (verify_neutering_jumps)
1634 (verify_neutering_labels): New function
1635 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
1637 2018-04-26 Tom de Vries <tom@codesourcery.com>
1640 * config/nvptx/nvptx.c (needs_neutering_p): New function.
1641 (nvptx_single): Use needs_neutering_p to skip over insns that do not
1644 2018-04-26 Richard Biener <rguenther@suse.de>
1645 Tom de Vries <tom@codesourcery.com>
1648 * lto-streamer-out.c (output_function): Fixup loops if required to match
1649 discovery done in the reader.
1651 2018-04-26 Richard Biener <rguenther@suse.de>
1653 PR tree-optimization/85116
1654 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
1655 have a loop exit from the single latch predecessor. Remove
1656 case of header with just condition.
1657 (ch_base::copy_headers): Exclude infinite loops from any
1659 (pass_ch::execute): Record exits.
1661 2018-04-26 Richard Biener <rguenther@suse.de>
1663 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
1664 prologue cost vector and pass it to vect_get_load_cost.
1665 (vect_get_peeling_costs_all_drs): Likewise.
1666 (vect_peeling_hash_get_lowest_cost): Likewise.
1667 (vect_enhance_data_refs_alignment): Likewise.
1669 2018-04-26 Richard Biener <rguenther@suse.de>
1672 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
1673 checking of integer<->pointer conversions.
1674 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
1675 sign-/zero-extending pointer types.
1676 (expand_omp_for_static_chunk): Likewise.
1678 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
1679 Jean Lee <xiaoyur347@gmail.com>
1681 * config/mips/mips.c (mips_asan_shadow_offset): New function.
1682 (TARGET_ASAN_SHADOW_OFFSET): Define.
1683 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
1684 true for -fsanitize=address.
1686 2018-04-25 Mark Wielaard <mark@klomp.org>
1688 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
1691 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1693 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
1694 than "alu", remove explicit "memory" and "imm_disp" attributes.
1695 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
1698 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
1699 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
1700 gen_lowpart_no_emit.
1702 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
1705 * config/i386/i386.c (ix86_expand_builtin): Change memory
1706 operand to XI, extend p0 to Pmode.
1707 * config/i386/i386.md: Change unspec volatile and operand
1708 1 mode to XI, change operand 0 mode to P.
1710 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1712 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
1713 GET_MODE_MASK before any checking.
1714 (nds32_can_use_bset_p): Likewise.
1715 (nds32_can_use_btgl_p): Likewise.
1717 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1719 * config/nds32/nds32-doubleword.md: New define_split pattern for
1720 illegal register number.
1722 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1724 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
1726 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
1728 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
1730 2018-04-25 Richard Biener <rguenther@suse.de>
1732 * lto-streamer.h (LTO_major_version): Bump to 8.
1734 2018-04-25 Jakub Jelinek <jakub@redhat.com>
1736 * BASE-VER: Set to 9.0.0.
1738 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
1740 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
1741 in __abskf2 and __powikf2.
1743 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1746 * config/aarch64/constraints.md (Usg, Usj): New constraints.
1747 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
1748 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
1749 Use the above on operand 2. Reindent.
1750 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
1752 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1755 * common/config/i386/i386-common.c (ix86_handle_option): Don't
1757 * config/i386/i386.opt (mcet): Removed.
1758 * doc/install.texi: Remove -mcet documentation.
1759 * doc/invoke.texi: Likewise.
1761 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
1764 * doc/install.texi: Remove -mcet from bootstrap-cet.
1766 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1769 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
1770 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
1774 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
1775 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
1776 containing a CONST_VECTOR.
1778 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
1780 * doc/install.texi: Update newlib dependency for nvptx.
1782 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1785 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
1786 instead of INTVAL when shifting x left.
1788 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
1790 PR tree-optimization/85478
1791 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
1792 vect_grouped_store_supported for single element vectors.
1794 2018-04-24 Richard Biener <rguenther@suse.de>
1797 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
1798 load cost increase to the case of non-constant step.
1800 2018-04-24 Jakub Jelinek <jakub@redhat.com>
1803 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
1804 destination if any_malformed_asm.
1806 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
1809 * expr.c (store_field): In the bitfield case, if the value comes from
1810 a function call and is returned in registers by means of a PARALLEL,
1811 do not change the mode of the temporary unless BLKmode and VOIDmode.
1813 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
1815 PR rtl-optimization/85423
1816 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
1817 dependencies to debug insns when the previous insn is non-debug.
1819 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1821 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
1822 enums into a single definition.
1823 (fls): Fix predicates and printing.
1826 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
1828 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
1829 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
1830 and short u6 immediate.
1831 (check_if_valid_sleep_operand): Remove.
1832 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
1834 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1836 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
1837 flag_always_save_lp condition.
1838 * config/nds32/nds32.opt (malways-save-lp): New option.
1840 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1842 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
1843 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
1844 * config/nds32/nds32.h
1845 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
1846 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
1848 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1850 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
1851 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
1853 2018-04-22 Shiva Chen <shiva0217@gmail.com>
1854 Chung-Ju Wu <jasonwucj@gmail.com>
1856 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
1858 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
1859 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
1861 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1863 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
1865 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1867 * config/nds32/nds32-protos.h (nds32_data_alignment,
1868 nds32_local_alignment): Declare.
1869 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
1870 nds32_local_alignment): New functions.
1871 (TARGET_CONSTANT_ALIGNMENT): Define.
1872 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
1874 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1876 * config/nds32/nds32.c
1877 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
1878 (TARGET_MODES_TIEABLE_P): Likewise.
1880 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
1882 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
1885 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
1886 Chung-Ju Wu <jasonwucj@gmail.com>
1888 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1889 for unaligned access.
1890 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1892 * config/nds32/nds32-intrinsic.md: Likewise.
1893 * config/nds32/nds32_intrinsic.h: Likewise.
1894 * config/nds32/nds32.h (nds32_builtins): Likewise.
1895 * config/nds32/nds32.opt (munaligned-access): New option.
1896 * config/nds32/nds32.c (nds32_asm_file_start): Display
1897 flag_unaligned_access status.
1899 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1901 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
1902 -mno-relax is present.
1903 * config/riscv/linux.h (LINK_SPEC): Ditto.
1905 2018-04-20 Martin Sebor <msebor@redhat.com>
1908 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
1910 (gimple_fold_builtin_stxcpy_chk): Same.
1911 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
1913 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
1916 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
1917 __powikf2 when long double is IEEE 128-bit.
1919 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
1921 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
1922 step to make sure stack always aligned.
1924 2018-04-20 Carl Love <cel@us.ibm.com>
1927 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
1928 size check for arg0.
1930 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
1931 Tom de Vries <tom@codesourcery.com>
1934 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
1935 Emit insns for calls too.
1936 (nvptx_find_par): Always look for worker-level predecessor insn.
1937 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
1939 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
1940 (nvptx_process_pars): Propagate frames for calls.
1942 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
1945 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
1947 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
1948 (ix86_handle_option): Don't handle OPT_mibt.
1949 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
1951 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
1953 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
1954 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
1955 (ix86_target_macros): Define __CET__ with flag_cf_protection
1956 for -fcf-protection.
1957 * config/i386/i386.c (isa2_opts): Remove -mibt.
1958 * config/i386/i386.h (TARGET_IBT): Removed.
1959 (TARGET_IBT_P): Likewise.
1960 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
1961 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
1962 * config/i386/i386.opt (mcet): Update help message.
1965 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
1966 -mcet as an alias for -mshstk.
1968 2018-04-20 Richard Biener <rguenther@suse.de>
1971 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
1972 complexity by forcing a single use of the multiply operand.
1974 2018-04-20 Martin Jambor <mjambor@suse.cz>
1977 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
1978 recursion dependency to only apply to non-clones.
1980 2018-04-20 Martin Jambor <mjambor@suse.cz>
1983 * ipa-cp.c (create_specialized_node): Check that clones of
1984 self-recursive edges exist during IPA-CP.
1986 2018-04-19 Toon Moene <toon@moene.org>
1988 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
1991 2018-04-19 Jakub Jelinek <jakub@redhat.com>
1993 PR tree-optimization/85467
1994 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
1995 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
1996 VECTOR_CST element to type.
1998 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2001 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
2002 * config/i386/i386.md (builtin_setjmp_setup): Removed.
2003 (builtin_longjmp): Likewise.
2004 (save_stack_nonlocal): New pattern.
2005 (restore_stack_nonlocal): Likewise.
2007 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2010 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2011 Replace ASM_OUTPUT_LABEL with fprintf.
2013 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
2016 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
2017 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
2018 * config/i386/i386-c.c (ix86_target_macros_internal): Also
2019 define __IBT__ and __SHSTK__ for -fcf-protection.
2020 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
2022 (ix86_trampoline_init): Likewise.
2023 (x86_output_mi_thunk): Likewise.
2024 (ix86_notrack_prefixed_insn_p): Likewise.
2025 (ix86_option_override_internal): Don't disallow -fcf-protection.
2026 * config/i386/i386.md (rdssp<mode>): Also enable for
2028 (incssp<mode>): Likewise.
2029 (nop_endbr): Likewise.
2030 * config/i386/i386.opt (mcet): Change help message to built-in
2034 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
2035 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
2036 enable CET built-in functions.
2038 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
2040 * common/config/i386/i386-common.c
2041 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
2042 OPTION_MASK_ISA_MOVDIRI_UNSET,
2043 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
2044 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
2045 * config.gcc (movdirintrin.h): New header.
2046 * config/i386/cpuid.h (bit_MOVDIRI,
2047 bit_MOVDIR64B): New bits.
2048 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
2050 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
2051 (VOID, PVOID, PCVOID)): New function types.
2052 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
2053 __builtin_ia32_directstoreu_u64,
2054 __builtin_ia32_movdir64b): New builtins.
2055 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
2056 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
2058 (ix86_valid_target_attribute_inner_p): Ditto.
2059 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
2060 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
2061 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
2062 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
2063 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
2064 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
2065 (movdiri<mode>, movdir64b_<mode>): New.
2066 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
2067 * config/i386/immintrin.h: Include movdirintrin.h.
2068 * config/i386/movdirintrin.h: New file.
2069 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
2071 2018-04-19 Richard Biener <rguenther@suse.de>
2074 * cfg.c (clear_bb_flags): When loop state says we have
2075 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
2077 2018-04-19 Richard Biener <rguenther@suse.de>
2079 PR tree-optimization/84737
2080 * tree-vect-data-refs.c (vect_copy_ref_info): New function
2081 copying restrict info.
2082 (vect_setup_realignment): Use it.
2083 * tree-vectorizer.h (vect_copy_ref_info): Declare.
2084 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
2085 the first DR to all generated stores.
2086 (vectorizable_load): Likewise for loads.
2088 2018-04-19 Jakub Jelinek <jakub@redhat.com>
2090 PR tree-optimization/85446
2091 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
2092 the integral and pointer types to have the same precision.
2094 * doc/install.texi: Document --disable-cet being the default and
2097 2018-04-18 Martin Liska <mliska@suse.cz>
2099 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
2102 2018-04-18 Martin Liska <mliska@suse.cz>
2105 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2108 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2109 arguments if they are comparable.
2111 2018-04-18 Martin Liska <mliska@suse.cz>
2114 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2117 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2120 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
2123 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
2124 ENDBR after calling __morestack.
2126 2018-04-18 David Malcolm <dmalcolm@redhat.com>
2129 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
2130 by using gcc_base_ver to generate a gcc_driver_version, and use
2131 it when generating GCC_DRIVER_NAME.
2132 * configure: Regenerate.
2134 2018-04-18 Jakub Jelinek <jakub@redhat.com>
2137 * config.gcc: Obsolete powerpc*-*-*spe*.
2139 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2142 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
2143 (stabstr_D): Change type of unum from unsigned int to
2144 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
2147 2018-04-17 Jim Wilson <jimw@sifive.com>
2150 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
2151 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
2152 Set arg_pointer_offset after using pretend_args_size.
2154 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2156 PR rtl-optimization/85431
2157 * dse.c (record_store): Ignore zero width stores.
2160 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
2161 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
2162 __builtin_stack_restore rather than after it.
2163 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
2164 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
2165 argument instead of virtual_dynamic_stack_rtx.
2167 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
2169 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
2171 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
2172 Add note to error message to explain internal mapping of overloaded
2173 built-in function name to non-overloaded built-in function name.
2174 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
2177 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2180 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
2181 where the inputs overlap with the output.
2183 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2185 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
2186 (=v, v) alternative and explicit "memory" attribute.
2187 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
2188 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2190 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
2191 "sselog1" type instead of "sselog".
2192 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
2193 "sselog". Remove explicit "memory" attribute.
2194 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
2195 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2197 (vec_extract_hi_v32hi): Merge all alternatives into one, use
2198 "sselog1" type instead of "sselog". Remove explicit "memory"
2200 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
2201 use "sselog1" type instead of "sselog". Remove explicit "memory"
2203 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
2204 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
2206 (vec_extract_hi_v64qi): Merge all alternatives into one, use
2207 "sselog1" type instead of "sselog". Remove explicit "memory"
2209 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
2210 use "sselog1" type instead of "sselog". Remove explicit "memory"
2214 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
2217 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
2220 2018-04-17 Martin Jambor <mjambor@suse.cz>
2223 * ipa-cp.c (create_specialized_node): Call
2224 expand_all_artificial_thunks if necessary.
2226 2018-04-17 Martin Liska <mliska@suse.cz>
2229 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
2230 in message, remote space in between '_G' and '('.
2232 2018-04-17 Jakub Jelinek <jakub@redhat.com>
2235 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
2236 avx512f_vmcmp<mode>3<round_saeonly_name>,
2237 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
2238 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
2239 avx512f_rndscale<mode><round_saeonly_name>,
2240 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
2241 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2242 Use %<iptr>2 instead of %2 for -masm=intel.
2243 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
2244 avx512f_vcvttss2usi<round_saeonly_name>,
2245 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
2247 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
2248 avx512f_vcvttsd2usi<round_saeonly_name>,
2249 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
2250 Use %q1 instead of %1 for -masm=intel.
2251 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
2252 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
2253 of %3 for -masm=intel.
2254 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
2256 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
2258 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
2260 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
2261 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
2263 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
2265 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
2266 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
2267 %g1 and one with %0 and %1.
2268 (avx512er_vmrcp28<mode><round_saeonly_name>,
2269 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
2271 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
2272 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
2273 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
2274 of %0 and %{%4%} for -masm=intel.
2275 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
2276 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
2277 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
2278 order of %0 and %{%5%}%{z%} for -masm=intel.
2280 2018-04-17 Jan Hubicka <jh@suse.cz>
2283 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
2285 2018-04-17 Martin Liska <mliska@suse.cz>
2288 * multiple_target.c (create_dispatcher_calls): Set apostrophes
2289 for target_clone error message. Make default implementation
2290 clone to be a local declaration.
2291 (separate_attrs): Add new argument and check for an empty
2293 (expand_target_clones): Handle it.
2294 (ipa_target_clone): Make redirection just for target_clones
2297 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
2298 Tom de Vries <tom@codesourcery.com>
2301 * omp-expand.c (expand_oacc_for): Add dummy false branch for
2302 tiled basic blocks without omp continue statements.
2304 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
2307 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
2308 vec_extract expression as having side effects to make sure it gets
2311 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
2314 * config/i386/i386.c (get_builtin_code_for_version): Check
2317 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
2320 * gcc/config.gcc: Support "skylake".
2321 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
2323 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
2324 (processor_target_table): Add "skylake".
2325 (ix86_option_override_internal): Add "skylake".
2326 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
2327 PROCESSOR_CANNONLAKE.
2328 (get_builtin_code_for_version): Fix priority for
2329 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
2330 PROCESSOR_SKYLAKE-AVX512.
2331 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
2332 (processor_type): Add PROCESSOR_SKYLAKE.
2334 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
2335 Jason Merrill <jason@redhat.com>
2338 * convert.c (convert_to_integer_1): Use direct recursion for
2339 enumeral types and types with a precision less than the number
2340 of bits in their mode.
2342 2018-04-16 Julia Koval <julia.koval@intel.com>
2345 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
2346 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
2348 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
2351 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
2352 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
2353 and -mno-direct-move.
2355 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
2358 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
2359 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
2360 Ensure negative shifts result in {0}.
2362 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
2364 PR rtl-optimization/79916
2365 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
2366 regs (if any) to define how to gnerate SD moves when LRA is in
2369 2018-04-13 Jakub Jelinek <jakub@redhat.com>
2371 PR rtl-optimization/85393
2372 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
2373 * except.c (expand_dw2_landing_pad_for_region): Make static.
2374 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
2375 a label and unconditional jump to old_bb, rather than
2376 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
2379 PR rtl-optimization/85376
2380 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
2381 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
2382 instead of a specific value.
2384 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2385 Bin Cheng <bin.cheng@arm.com>
2387 PR tree-optimization/82965
2388 PR tree-optimization/83991
2389 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
2390 by_profile_only parameter.
2391 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
2392 information if the loop was predicted to iterate too many times.
2393 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
2395 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
2398 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
2401 2018-04-13 Martin Liska <mliska@suse.cz>
2402 Jakub Jelinek <jakub@redhat.com>
2405 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
2406 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
2407 * builtins.c (expand_builtin_memory_copy_args): Use
2408 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
2409 handle dest_addr == pc_rtx.
2411 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
2414 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
2415 asked to not generate direct moves.
2416 (fix_trunc<mode>si2_stfiwx): Similar.
2417 (fix_trunc<mode>si2_internal): Similar.
2419 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2422 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
2423 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
2424 lookup if dest in some wider mode is known to be const0_rtx and
2425 if so, record permanent equivalence for it to be ZERO_EXTEND of
2426 the narrower mode destination.
2428 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2430 * lto-streamer-out.c (output_function): Revert 259346.
2431 * omp-expand.c (expand_oacc_for): Likewise.
2433 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
2435 PR rtl-optimization/85354
2436 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
2437 * sel-sched.c (sel_global_init): ... here.
2439 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
2442 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
2443 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
2444 mode for PE-COFF targets.
2445 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
2446 (i386_pe_asm_lto_end): Likewise.
2447 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
2448 (TARGET_ASM_LTO_END): Likewise.
2449 * config/i386/winnt.c (saved_debug_info_level): New static variable.
2450 (i386_pe_asm_lto_start): New function.
2451 (i386_pe_asm_lto_end): Likewise.
2453 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
2454 Richard Biener <rguenther@suse.de>
2457 * lto-streamer-out.c (output_function): Fix CFG loop state before
2459 * omp-expand.c (expand_oacc_for): Handle calls to internal
2460 functions like regular functions.
2462 2018-04-12 Richard Biener <rguenther@suse.de>
2465 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
2466 for the early LTO debug to properly generate references to it
2467 during DIE emission. Do not re-use that for the skeleton for
2469 (dwarf2out_early_finish): Likewise.
2471 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2474 * config/i386/sse.md
2475 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
2476 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
2477 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
2478 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
2479 and output is a reg, avoid creating invalid lowpart subreg, but
2480 instead split into a 512-bit move. Don't split if not AVX512VL,
2481 input is xmm16+ reg and output is a mem.
2482 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
2483 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
2484 xmm16+ reg and output is a mem.
2486 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2488 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
2489 also for flag_dwarf2_cfi_asm.
2491 2018-04-12 Jakub Jelinek <jakub@redhat.com>
2493 PR rtl-optimization/85342
2494 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
2495 a bool scalar var inside of the loop instead. Don't try to update
2496 recog_data.operand after failed apply_change_group.
2498 2018-04-12 Tom de Vries <tom@codesourcery.com>
2501 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
2502 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
2503 array with flexible array member as array without given dimension.
2504 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
2505 argument for undefined param to true.
2507 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
2510 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2511 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
2512 from PowerPC section.
2513 * config/rs6000/sysv4.opt (mcall-): Improve help text.
2514 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
2515 help text that is too long.
2516 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
2517 help text that is too long.
2518 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
2519 help text that is too long.
2521 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
2523 * config/alpha/alpha.md (stack_probe_internal): Rename
2524 from "probe_stack". Update all callers.
2526 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2528 PR rtl-optimization/84566
2529 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
2530 sched_macro_fuse_insns.
2532 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2535 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
2536 (compute_block_dependences): ... from here.
2538 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2540 PR tree-optimization/85331
2541 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
2542 from int to HOST_WIDE_INT.
2544 2018-04-11 Martin Jambor <mjambor@suse.cz>
2547 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
2548 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
2549 not the same as the source val.
2550 (cgraph_edge_brings_value_p): New parameter.
2551 (gather_edges_for_value): Pass destination value to
2552 cgraph_edge_brings_value_p.
2553 (perhaps_add_new_callers): Likewise.
2554 (get_info_about_necessary_edges): Likewise and exclude values brought
2555 only by self-recursive edges.
2556 (create_specialized_node): Redirect only clones of self-calling edges.
2557 (+self_recursive_pass_through_p): New function.
2558 (find_more_scalar_values_for_callers_subset): Use it.
2559 (find_aggregate_values_for_callers_subset): Likewise.
2560 (known_aggs_to_agg_replacement_list): Removed.
2561 (decide_whether_version_node): Re-calculate known constants for all
2562 remaining context clones.
2564 2018-04-11 Richard Biener <rguenther@suse.de>
2567 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
2568 from early DWARF output.
2569 (dwarf2out_early_finish): Output line info unconditionally into
2570 early DWARF and add reference to it.
2572 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2575 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
2576 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
2577 other than V2DFmode using iptr mode attribute.
2578 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
2580 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
2582 PR rtl-optimization/84659
2583 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
2585 2018-04-11 Jakub Jelinek <jakub@redhat.com>
2588 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
2590 (output_loc_list): Pass address of a dummy size variable even in the
2591 locview handling loop.
2592 (index_location_lists): Add comment on why skip_loc_list_entry can't
2595 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
2598 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
2601 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
2604 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
2605 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
2606 and -mstring-compare-inline-limit.
2608 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2611 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
2612 for stack clash protection in a register whenever we need it to be in
2615 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2617 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
2618 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
2620 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
2623 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
2625 (mlong-double-): Ditto.
2626 * config/rs6000/sysv4.opt (msdata=): Ditto.
2627 (mtls-size=): Ditto.
2629 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
2631 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
2632 erroneous entries for
2633 "vector int vec_ldl (int, long int *)", and
2634 "vector unsigned int vec_ldl (int, unsigned long int *)".
2635 Add comments and entries for
2636 "vector bool char vec_ldl (int, bool char *)",
2637 "vector bool short vec_ldl (int, bool short *)",
2638 "vector bool int vec_ldl (int, bool int *)",
2639 "vector bool long long vec_ldl (int, bool long long *)",
2640 "vector pixel vec_ldl (int, pixel *)",
2641 "vector long long vec_ldl (int, long long *)",
2642 "vector unsigned long long vec_ldl (int, unsigned long long *)".
2643 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
2644 type tree bool_long_long_type_node and correct definition of
2645 bool_V2DI_type_node to make reference to this new type tree.
2646 (rs6000_mangle_type): Replace erroneous reference to
2647 bool_long_type_node with bool_long_long_type_node.
2648 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
2649 comments to emphasize sign distinctions for char and int types and
2650 replace RS6000_BTI_bool_long constant with
2651 RS6000_BTI_bool_long_long constant. Also add comment to restrict
2652 use of RS6000_BTI_pixel.
2653 (bool_long_type_node): Remove this macro definition.
2654 (bool_long_long_type_node): New macro definition
2656 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2658 PR rtl-optimization/85300
2659 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
2660 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
2661 simplify_unary_operation fails.
2663 2018-04-10 Martin Liska <mliska@suse.cz>
2665 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
2666 cgraph_edge and ipa_ref.
2668 2018-04-10 Jakub Jelinek <jakub@redhat.com>
2672 * config/i386/sse.md
2673 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
2674 computation of the VEC_MERGE selector from mask.
2675 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
2676 Fix decoding of the VEC_MERGE selector into mask.
2678 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2680 PR tree-optimization/85286
2681 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
2683 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
2685 * final.c (final_1): Set insn_last_address as well as
2686 insn_current_address.
2688 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2691 * explow.c (emit_stack_probe): Call validize_mem on memory location
2692 before passing it to gen_probe_stack. Create address operand and
2693 legitimize it for the probe_stack_address case.
2695 2018-04-09 Jan Hubicka <jh@suse.cz>
2698 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
2699 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
2700 * tree.c (free_lang_data_in_type): Fix handling of binfos;
2702 (free_lang_data): Rebuild type inheritance graph.
2704 2018-04-09 Martin Sebor <msebor@redhat.com>
2706 * invoke.texi (-finline-small-functions): Mention other optimization
2708 (-findirect-inlining, -fpartial-inlining): Same.
2709 (-finline-functions-called-once): Same.
2710 (-freorder-blocks-and-partition): Same.
2712 2018-04-09 Jan Hubicka <jh@suse.cz>
2715 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
2716 jumps; choose last target that matches the criteria (i.e.
2717 no partition changes for non-crossing jumps).
2718 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
2719 support for redirecting crossing jumps to non-crossing.
2721 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
2723 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
2724 also for naked functions.
2726 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
2728 * config/arc/arc.md (add_shift): New pattern.
2729 (add_shift2): Likewise.
2730 (sub_shift): Likewise.
2731 (sub_shift_cmp0_noout): Likewise.
2732 (compare_si_ashiftsi): Likewise.
2733 (xbfu_cmp0_noout): New combine pattern.
2734 (xbfu_cmp0"): Likewise.
2735 (movsi_set_cc_insn): Place the predicable variant first.
2736 (commutative_binary_cmp0_noout): Remove clobber.
2737 (commutative_binary_cmp0): New pattern.
2738 (noncommutative_binary_cmp0): Likewise.
2739 (noncommutative_binary_cmp0_noout): Likewise.
2740 (noncommutative_binary_comparison_result_used): Removed.
2741 (rsub_cmp0): New pattern.
2742 (rsub_cmp0_noout): Likewise.
2743 (extzvsi): Changed, keep only meaningful variants.
2744 (SQH, SEZ): New iterators.
2745 (SQH_postfix): New mode attribute.
2746 (SEZ_prefix): New code attribute.
2747 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
2748 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
2749 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
2751 (noncommutative_operator): Check the availability of barrel
2754 2018-04-09 Richard Biener <rguenther@suse.de>
2756 PR tree-optimization/85284
2757 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
2758 Only use the niter constraining form of simple_iv when the exit
2761 2018-04-09 Tom de Vries <tom@codesourcery.com>
2764 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2765 (define_expand "*memory_barrier"): New define_expand.
2766 (define_insn "memory_barrier"): New insn.
2768 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2770 PR rtl-optimization/80463
2771 PR rtl-optimization/83972
2772 PR rtl-optimization/83480
2774 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
2775 correct producer for the insn.
2776 (tidy_control_flow): Fixup seqnos in case of debug insns.
2778 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2780 PR rtl-optimization/83913
2782 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
2783 different sched-times when merging exprs.
2785 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2787 PR rtl-optimization/83962
2789 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
2790 tidy_fallthru_edge and tidy_control_flow.
2792 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
2794 PR rtl-optimization/83530
2796 * sel-sched.c (force_next_insn): New global variable.
2797 (remove_insn_for_debug): When force_next_insn is true, also leave only
2798 next insn in the ready list.
2799 (sel_sched_region): When the region wasn't scheduled, make another pass
2800 over it with force_next_insn set to 1.
2802 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2804 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
2806 * config/nds32/constants.md (unspec_volatile_element): Add enum values
2807 for interrupt control.
2808 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2809 functions for interrupt control.
2810 * config/nds32/nds32-intrinsic.md: Likewise.
2811 * config/nds32/nds32_intrinsic.h: Likewise.
2812 * config/nds32/nds32.h (nds32_builtins): Likewise.
2814 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2816 * config/nds32/nds32.c (nds32_init_machine_status,
2817 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
2818 strict_aligned_p field.
2819 (nds32_expand_to_rtl_hook): New function.
2820 (TARGET_EXPAND_TO_RTL_HOOK): Define.
2821 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
2823 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2824 Chung-Ju Wu <jasonwucj@gmail.com>
2826 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
2827 * config/nds32/nds32-n7.md: New file.
2828 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
2829 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
2831 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
2832 * config/nds32/nds32.md (pipeline_model): Add n7.
2833 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
2834 * config/nds32/pipelines.md: Include n7 settings.
2836 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2837 Chung-Ju Wu <jasonwucj@gmail.com>
2839 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
2840 * config/nds32/nds32-e8.md: New file.
2841 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
2842 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
2844 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
2845 * config/nds32/nds32.md (pipeline_model): Add e8.
2846 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
2847 * config/nds32/pipelines.md: Include e8 settings.
2849 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2850 Chung-Ju Wu <jasonwucj@gmail.com>
2852 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
2853 * config/nds32/nds32-n8.md: New file.
2854 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
2855 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
2857 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
2858 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
2859 * config/nds32/nds32.md (pipeline_model): Add n8.
2860 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
2861 * config/nds32/pipelines.md: Include n8 settings.
2863 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
2864 Chung-Ju Wu <jasonwucj@gmail.com>
2866 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
2867 * config/nds32/nds32-n9-2r1w.md: New file.
2868 * config/nds32/nds32-n9-3r2w.md: New file.
2869 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
2870 nds32_register_ports): New or modify for cpu n9.
2871 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
2873 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
2874 * config/nds32/nds32-utils.c: New file.
2875 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
2876 TARGET_MUL_SLOW): Define.
2877 * config/nds32/nds32.md (pipeline_model): New attribute.
2878 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
2879 New options that support cpu n9.
2880 * config/nds32/pipelines.md: Include n9 settings.
2881 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
2883 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
2885 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
2886 information if necessary.
2887 (output_cond_branch_compare_zero): Likewise.
2888 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
2889 (nds32_target_alignment): Refine for alignment.
2890 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
2891 (FUNCTION_BOUNDARY): Modify.
2892 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
2894 * config/nds32/nds32.opt (malways-align, malign-functions): New.
2896 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
2898 * config/nds32/constants.md (unspec_volatile_element): Add values for
2899 TLB operation and data prefetch.
2900 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
2901 functions for TLB operation and data prefetch.
2902 * config/nds32/nds32-intrinsic.md: Likewise.
2903 * config/nds32/nds32_intrinsic.h: Likewise.
2904 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
2905 (nds32_print_operand): Likewise.
2906 * config/nds32/nds32.h (nds32_builtins): Likewise.
2908 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
2909 Andrew Pinski <pinsika@gcc.gnu.org>
2912 * match.pd: Use constant_boolean_node of correct type instead of
2913 boolean_true_node or boolean_false_node for simplifying
2914 pointer comparisons to zero.
2916 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2918 PR tree-optimization/80021
2919 * tree.c (verify_type_variant): Make error call in verify_variant_match
2920 translatable and remove final full stop.
2922 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2924 * config/nds32/constants.md (unspec_volatile_element): Add
2925 UNSPEC_VOLATILE_EH_RETURN.
2926 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
2927 nds32_output_stack_pop): Support dwarf exception handling process.
2928 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
2929 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
2930 exception handling process.
2931 (nds32_compute_stack_frame): Likewise.
2932 (nds32_return_addr_rtx): Likewise.
2933 (nds32_initial_elimination_offset): Likewise.
2934 (nds32_expand_prologue): Likewise.
2935 (nds32_expand_epilogue): Likewise.
2936 (nds32_dynamic_chain_address): New function.
2937 * config/nds32/nds32.h (machine_function): Add fields for dwarf
2939 (DYNAMIC_CHAIN_ADDRESS): Define.
2940 (EH_RETURN_DATA_REGNO): Define.
2941 (EH_RETURN_STACKADJ_RTX): Define.
2942 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
2943 patterns for dwarf exception handling.
2945 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2947 * config/nds32/nds32.h: Clean up obsolete macros.
2949 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2951 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2952 Add enum values for particular instructions.
2953 * config/nds32/nds32-intrinsic.c: Implementation of expanding
2954 particular intrinsic functions.
2955 * config/nds32/nds32-intrinsic.md: Likewise.
2956 * config/nds32/nds32_intrinsic.h: Likewise.
2957 * config/nds32/nds32.h (nds32_builtins): Likewise.
2958 * config/nds32/nds32.md (type): Add pbsad and pbsada.
2959 (btst, ave): New patterns for particular instructions.
2961 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2963 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2964 Add enum values for atomic load/store and memory sync.
2965 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
2967 * config/nds32/nds32-intrinsic.md: Likewise.
2968 * config/nds32/nds32_intrinsic.h: Likewise.
2969 * config/nds32/nds32.h (nds32_builtins): Likewise.
2971 2018-04-07 Jakub Jelinek <jakub@redhat.com>
2973 PR tree-optimization/85257
2974 * fold-const.c (native_encode_vector): If not all elts could fit
2975 and off is -1, return 0 rather than offset.
2976 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
2977 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
2978 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
2979 adjust buffer in native_interpret_expr call.
2981 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
2983 * config/nds32/constants.md (unspec_volatile_element): Add cache
2984 control enum values.
2985 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
2986 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
2987 * config/nds32/nds32.c (nds32_cctl_names): New.
2988 (nds32_print_operand): Handle cache control register names.
2989 * config/nds32/nds32.h (nds32_builtins): New enum values.
2990 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
2992 * config/nds32/nds32.md (type): Add mmu.
2993 * config/nds32/pipelines.md (simple_insn): Add mmu.
2995 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
2997 * config/nds32/nds32.md (type): Remove call.
2998 * config/nds32/pipelines.md (simple_insn): Likewise.
3000 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3002 * config/nds32/constants.md (unspec_volatile_element): Add
3003 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
3004 UNSPEC_VOLATILE_FMFCFG.
3005 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
3006 description for fmfcfg and fmfcsr.
3007 (bdesc_1arg): Add fmtcsr.
3008 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
3009 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
3010 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
3011 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
3012 unspec_fmfcfg): New patterns.
3013 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
3014 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
3015 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
3016 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
3017 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
3018 __nds32__fmfcfg): Define.
3020 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
3022 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
3023 intrinsic register names.
3024 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
3025 intrinsic register enum values and macros.
3027 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
3029 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
3030 for load/store addressing form.
3031 (nds32_print_operand_address): Likewise.
3033 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
3036 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
3037 based on LABEL_REF. Remove useless assertion.
3038 (pic_address_needs_scratch): Fix formatting.
3039 (sparc_legitimize_pic_address): Minor tweaks.
3040 (sparc_delegitimize_address): Adjust assertion accordingly.
3041 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
3042 into symbolic_operand.
3043 (movsi_high_pic_label_ref): Likewise.
3044 (movsi_lo_sum_pic_label_ref): Likewise.
3045 (movdi_pic_label_ref): Likewise.
3046 (movdi_high_pic_label_ref): Likewise.
3047 (movdi_lo_sum_pic_label_ref): Likewise.
3049 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
3051 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
3052 custom LIB_SPEC setup.
3054 2018-04-06 Ruslan Bukin <br@bsdpad.com>
3055 Kito Cheng <kito.cheng@gmail.com>
3057 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
3058 * config/riscv/freebsd.h: New.
3060 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3062 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
3063 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
3066 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3067 Kito Cheng <kito.cheng@gmail.com>
3069 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
3070 nds32_output_call, nds32_symbol_binds_local_p): New functions.
3071 * config/nds32/nds32-protos.h (nds32_output_call,
3072 nds32_output_return): Declare.
3073 * config/nds32/nds32.md: Refine all the call and return patterns.
3075 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3078 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
3079 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
3081 PR rtl-optimization/84872
3082 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
3083 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
3086 2018-04-06 Tamar Christina <tamar.christina@arm.com>
3088 * expr.c (copy_blkmode_to_reg): Revert 254862.
3089 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
3091 2018-04-06 Richard Biener <rguenther@suse.de>
3094 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
3095 after seeing a component reference with an adjacent field. Treat
3096 refs to arrays at struct end of external decls similar to
3097 refs to unconstrained commons.
3099 2018-04-06 Jakub Jelinek <jakub@redhat.com>
3102 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
3103 look through SAVE_EXPRs with non-side-effects argument. Adjust
3105 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
3108 2018-04-06 Richard Biener <rguenther@suse.de>
3111 * alias.c (find_base_term): New wrapper around find_base_term
3112 unwinding CSELIB_VAL_PTR changes.
3113 (find_base_term): Do not restore CSELIB_VAL_PTR during the
3116 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3118 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
3120 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
3121 constant definitions.
3122 ("nop"): lr 0,0 -> nopr r0
3123 ("nop_lr0", "nop_lr1"): New insn definitions.
3125 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
3127 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
3128 NDS32_V3PUSH_AVAILABLE_P macro.
3130 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
3131 Chung-Ju Wu <jasonwucj@gmail.com>
3133 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
3134 (nds32*-*-*): Add float and fpu_config into supported_defaults.
3135 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
3136 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
3137 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
3138 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
3139 * config/nds32/constraints.md: New constraints and checking for hard
3140 float configuration.
3141 * config/nds32/iterators.md: New mode iterator and attribute for hard
3142 float configuration.
3143 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
3145 * config/nds32/nds32-fpu.md: New file.
3146 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
3147 deal with hard float code generation.
3148 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
3150 (abi_type, float_reg_number): New enum type.
3151 * config/nds32/nds32-predicates.c: New predicates for hard float.
3152 * config/nds32/nds32-protos.h: Declare functions for hard float.
3153 * config/nds32/nds32.c: Implementation for hard float configuration.
3154 * config/nds32/nds32.h: Definitions for hard float configuration.
3155 * config/nds32/nds32.md: Include hard float machine description and
3156 modify patterns for hard float configuration.
3157 * config/nds32/nds32.opt: New options for hard float configuration.
3158 * config/nds32/predicates.md: New predicates for hard float
3161 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3163 * common/config/nds32/nds32-common.c
3164 (nds32_option_optimization_table): Enable -mreleax-hint by default.
3166 2018-04-05 Jakub Jelinek <jakub@redhat.com>
3169 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
3170 CONSTRUCTOR_ELT (ctor, ...)->value.
3172 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
3175 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
3177 2018-04-05 Tom de Vries <tom@codesourcery.com>
3180 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
3183 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3184 Kito Cheng <kito.cheng@gmail.com>
3186 * config/nds32/constraints.md (U33): Fine-tune checking condition.
3187 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
3188 * config/nds32/nds32.h (nds32_16bit_address_type): Add
3189 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
3191 2018-04-05 Shiva Chen <shiva0217@gmail.com>
3192 Kito Cheng <kito.cheng@gmail.com>
3194 * config/nds32/constraints.md (Ufe): New memory constraint.
3195 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
3196 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
3197 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
3199 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
3200 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
3202 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3204 * config/nds32/nds32.md: Use optimize_size in the condition for
3205 alu-shift instructions.
3207 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3209 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
3211 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3213 * config/nds32/nds32.md (negsi2): Refine pattern.
3215 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3216 Chung-Ju Wu <jasonwucj@gmail.com>
3218 * config/nds32/iterators.md (shift_rotate): New code iterator.
3219 (shift): New code attribute.
3220 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
3221 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
3222 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
3223 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
3224 bit-wise operations.
3225 (andsi3, *andsi3): Ditto.
3226 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
3227 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
3228 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
3229 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
3230 nds32_ior_operand, nds32_xor_operand): New predicates.
3232 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3234 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
3235 (addsi3, subsi3): ... this.
3237 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3239 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
3241 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3243 * config/nds32/nds32.md: Adjust indention.
3245 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
3247 * config/nds32/nds32.md (feature): New attribute.
3249 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
3251 * config/nds32/nds32.md (subtype): New attribute.
3253 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
3256 * config/arm/arm-builtins.c (arm_expand_builtin): Change
3257 expansion to perform a bitwise AND of the argument followed by a
3258 boolean negation of the result.
3260 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
3262 PR rtl-optimization/84878
3263 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
3264 the basic block. Assert the use reference is not artificial and that
3265 it has an associated insn.
3267 2018-04-04 Michael Matz <matz@suse.de>
3269 * builtins.c (compute_objsize): Pass correct operand
3270 to array_at_struct_end_p.
3272 2018-04-04 Richard Biener <rguenther@suse.de>
3275 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
3276 from contexts for DINFO_LEVEL_TERSE and below.
3278 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3280 * config/nds32/nds32-doubleword.md (move_<mode>): Require
3281 resiter_operand condition.
3282 * config/nds32/nds32.md (*move<mode>): Ditto.
3284 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3285 Monk Chiang <sh.chiang04@gmail.com>
3287 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
3289 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3291 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
3293 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3294 Kito Cheng <kito.cheng@gmail.com>
3296 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
3297 nds32_cond_code_str, output_cond_branch,
3298 output_cond_branch_compare_zero, nds32_expand_cbranch,
3299 nds32_expand_cstore, nds32_expand_movcc,
3300 nds32_output_cbranchsi4_equality_zero,
3301 nds32_output_cbranchsi4_equality_reg,
3302 nds32_output_cbranchsi4_equality_reg_or_const_int,
3303 nds32_output_cbranchsi4_greater_less_zero: New functions.
3304 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
3305 nds32_expand_cstore, nds32_expand_movcc,
3306 nds32_output_cbranchsi4_equality_zero,
3307 nds32_output_cbranchsi4_equality_reg,
3308 nds32_output_cbranchsi4_equality_reg_or_const_int,
3309 nds32_output_cbranchsi4_greater_less_zero): Declare.
3310 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
3311 nds32_rimm11s_operand): New predicates.
3312 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
3313 * config/nds32/nds32.md: Rewrite all the branch and conditional move
3316 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
3318 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
3319 * config/nds32/nds32.md: Ditto.
3320 * config/nds32/pipelines.md: Ditto.
3322 2018-04-04 Richard Biener <rguenther@suse.de>
3324 PR tree-optimization/85168
3325 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
3326 propagating abnormals.
3328 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3330 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
3332 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
3333 Kito Cheng <kito.cheng@gmail.com>
3335 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
3336 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
3337 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
3338 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
3339 * config/nds32/nds32.md (sibcall_internal): New.
3340 (sibcall_register): Remove.
3341 (sibcall_immediate): Remove.
3342 (sibcall_value_internal): New.
3343 (sibcall_value_register): Remove.
3344 (sibcall_value_immediate): Remove.
3345 * config/nds32/predicates.md (nds32_general_register_operand): New.
3346 (nds32_call_address_operand): New.
3348 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3350 PR rtl-optimization/85167
3351 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
3352 bb_defs if *split_p, instead preinitialize it to NULL.
3354 PR tree-optimization/85156
3355 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
3356 evaluating the argument multiple times.
3358 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
3360 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
3362 (_mm_cvtpd_ps): Likewise.
3363 (_mm_cvttpd_epi32): Likewise.
3364 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
3365 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
3366 vector, pixel, and bool following altivec.h include.
3368 2018-04-03 Martin Sebor <msebor@redhat.com>
3370 * doc/extend.texi (Common Function Attributes): Clarify.
3371 (const attribute): Likewise.
3372 (pure attribute): Likewise.
3374 2018-04-03 Jakub Jelinek <jakub@redhat.com>
3377 * config/i386/i386.c (ix86_expand_vector_set): Use
3378 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
3380 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
3382 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
3383 instructions when changing rounding bits to preserve precision bits
3384 in the x87 control word.
3386 2018-04-03 Martin Liska <mliska@suse.cz>
3388 PR tree-optimization/82491
3389 * rtl.h (strip_offset_and_add): Replace += suboffset with
3390 poly_uint64 () + suboffset.
3392 2018-03-29 Martin Liska <mliska@suse.cz>
3393 Martin Jambor <mjambor@suse.cz>
3396 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
3397 param_type is not an integral or pointer type.
3399 2018-04-03 Richard Biener <rguenther@suse.de>
3401 * sese.h (recompute_all_dominators): Remove.
3403 2018-04-02 Martin Sebor <msebor@redhat.com>
3405 * doc/invoke.texi (-Wrestrict): Fix typos.
3407 2018-04-02 Jim Wilson <jimw@sifive.com>
3409 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
3410 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
3411 (<optab>di3, <optab>si3_extend): Likewise.
3412 (<optab>si3_mask, <optab>si3_mask_1): New.
3413 (<optab>di3_mask, <optab>di3_mask_1): New.
3414 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
3415 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
3416 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
3418 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
3420 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
3423 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
3425 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
3426 (nds32_canonicalize_comparison): New function.
3428 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3429 Kito Cheng <kito.cheng@gmail.com>
3430 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3432 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
3433 * config/nds32/constants.md (unspec_volatile_element): Add
3434 UNSPEC_VOLATILE_RELAX_GROUP.
3435 * config/nds32/nds32-relax-opt.c: New file.
3436 * config/nds32/nds32-predicates.c
3437 (nds32_symbol_load_store_p): New function.
3438 * config/nds32/nds32-protos.h
3439 (nds32_symbol_load_store_p): Declare function.
3440 (make_pass_nds32_relax_opt): Declare new rtl pass function.
3441 * config/nds32/nds32.c
3442 (nds32_register_pass): New function to register pass.
3443 (nds32_register_passes): New function to register passes.
3444 * config/nds32/nds32.md (relax_group): New pattern.
3445 * config/nds32/nds32.opt (mrelax-hint): New option.
3446 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
3448 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
3450 * config/nds32/t-nds32: Modify files dependency.
3452 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3454 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
3455 (PROFILE_HOOK): Define its implementation.
3457 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
3459 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
3460 type and 32-bit size.
3462 2018-04-01 Jakub Jelinek <jakub@redhat.com>
3465 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
3466 (V_128_256): New mode iterator.
3467 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
3468 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
3469 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
3471 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
3472 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
3474 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
3477 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
3478 NaN inputs correctly.
3480 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
3483 * config/rs6000/vsx.md (??r): New mode attribute.
3484 (*vsx_mov<mode>_64bit): Use it.
3485 (*vsx_mov<mode>_32bit): Likewise.
3487 2018-03-30 Martin Sebor <msebor@redhat.com>
3489 PR tree-optimization/84818
3490 * builtins.c (check_access): Use warning_n.
3492 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3495 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
3497 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
3500 2018-03-30 Julia Koval <julia.koval@intel.com>
3503 * x86-tune.def (movx, partial_reg_dependency): Enable for
3506 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
3509 * lra-constraints.c (process_alt_operands): Move setting
3510 this_alternative_matches below.
3512 2018-03-29 Martin Liska <mliska@suse.cz>
3515 * doc/invoke.texi: Document how LTO works with debug info.
3516 Describe auto-load support of binutils. Mention 'x86-64'
3517 as valid option value of -march option.
3519 2018-03-29 Jakub Jelinek <jakub@redhat.com>
3521 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
3524 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
3525 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
3526 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
3529 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3532 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
3533 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
3534 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
3535 for RS6000_BTM_POWERPC64.
3536 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
3537 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
3538 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
3543 2018-03-28 Carl Love <cel@us.ibm.com>
3546 2017-09-27 Carl Love <cel@us.ibm.com>
3548 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
3549 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
3550 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
3553 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3555 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
3556 instead of __vector bool.
3557 (_mm_max_pu8): Likewise.
3558 (_mm_min_pi16): Likewise.
3560 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
3563 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
3564 (DIVWEUO): Likewise.
3566 (DIVDEUO): Likewise.
3567 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
3568 DIVWEUO and DIVDEUO.
3569 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
3570 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
3571 (div_extend): Likewise.
3572 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
3574 (__builtin_divweuo): Likewise.
3575 (__builtin_divdeo): Likewise.
3576 (__builtin_divdeuo): Likewise.
3578 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3581 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
3582 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
3584 PR tree-optimization/82004
3585 * gimple-match-head.c (optimize_pow_to_exp): New function.
3586 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
3587 Don't fold to exp if optimize_pow_to_exp is false.
3589 2018-03-28 Martin Liska <mliska@suse.cz>
3592 * calls.c (initialize_argument_information): Fix trailing space.
3593 * common.opt: Fix typo and provide better explanation for
3594 -fsanitize-coverage option.
3595 * config/i386/i386.opt: Fix typo.
3597 2018-03-28 Jakub Jelinek <jakub@redhat.com>
3598 Martin Liska <mliska@suse.cz>
3601 * gimplify.c (asan_poison_variable): Don't do the check for
3602 gimplify_omp_ctxp here.
3603 (gimplify_decl_expr): Do it here.
3604 (gimplify_target_expr): Likewise.
3606 2018-03-28 Martin Liska <mliska@suse.cz>
3609 * config/i386/i386.c (ix86_function_arg_advance): Do not call
3610 chkp_type_bounds_count if MPX is not enabled.
3612 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
3614 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
3616 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
3619 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
3620 function to create the function decl for complex long double
3621 multiply and divide for -mabi=ieeelongdouble.
3622 (init_float128_ieee): Call it.
3624 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
3627 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
3628 -fcf-protection=branch -mibt.
3629 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
3631 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3634 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
3636 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
3639 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
3640 extern array declarations.
3642 2018-03-27 Richard Biener <rguenther@suse.de>
3645 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
3646 explicit single_use checks.
3648 2018-03-27 Richard Biener <rguenther@suse.de>
3650 PR tree-optimization/85082
3651 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
3654 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3656 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
3657 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
3658 Turn on fasynchronous-unwind-tables and funwind-tables.
3660 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
3663 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
3664 (*bmi_blsr_<mode>_ccz): Ditto.
3666 2018-03-26 Tom de Vries <tom@codesourcery.com>
3668 PR tree-optimization/85063
3669 * omp-general.c (offloading_function_p): New function. Factor out
3671 * omp-offload.c (pass_omp_target_link::gate): ... here.
3672 * omp-general.h (offloading_function_p): Declare.
3673 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
3674 with attribute omp declare target for offloading functions.
3676 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
3678 PR tree-optimization/84005
3679 * tree-data-ref.h (get_base_for_alignment): Declare.
3680 * tree-data-ref.c (get_base_for_alignment_1): New function.
3681 (get_base_for_alignment): Likewise.
3682 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
3683 get_base_for_alignment to find a suitable base object, instead
3684 of always using drb->base_address.
3686 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3689 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
3690 known size by default.
3692 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
3695 * lra-constraints.c (process_alt_operands): Don't match BLKmode
3696 and non BLKmode operands.
3698 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3701 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
3702 Clean up attributes.
3704 2018-03-23 Richard Biener <rguenther@suse.de>
3707 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
3708 we are going to emit early debug for LTO.
3710 2018-03-23 Jakub Jelinek <jakub@redhat.com>
3713 * function.c (match_asm_constraints_1): Don't optimize if input
3714 doesn't satisfy general_operand predicate for output's mode.
3717 * alias.c (write_dependence_p): Don't require for x_canonicalized
3718 non-VOIDmode if x has VOIDmode.
3721 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
3722 just don't try to optimize it rather than assert it never happens.
3724 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3726 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
3727 macro expansions for definition of ST_INTERNAL_<mode> and
3728 LD_INTERNAL_<mode> builtins.
3729 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
3731 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
3733 (altivec_expand_st_builtin): Likewise.
3734 (altivec_expand_builtin): Remove calls to deleted functions.
3735 (rs6000_address_for_altivec): Delete this function.
3736 * config/rs6000/vector.md: Remove expands for
3737 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
3739 2018-03-22 Sudakshina Das <sudi.das@arm.com>
3742 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
3743 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
3744 re-computing once computed.
3745 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
3746 (arm_init_machine_status): Initialize
3747 machine->static_chain_stack_bytes.
3749 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
3752 * doc/extend.texi: Add four new prototypes for vec_ld.
3753 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
3754 definitions for more logical presentation.
3755 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
3756 entries for V1TI variants of __builtin_altivec_ld builtin.
3757 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
3758 handling of V1TI variant of LVX icode pattern.
3759 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
3760 (rs6000_gimple_fold_builtin): Likewise.
3761 (altivec_init_builtins): Add code to define
3762 __builtin_altivec_lvx_v1ti function.
3764 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3767 * function.c (match_asm_constraints_1): Don't do the optimization
3768 if input isn't a REG, SUBREG, MEM or constant.
3770 2018-03-22 Tom de Vries <tom@codesourcery.com>
3772 PR tree-optimization/84956
3773 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
3774 bb_has_abnormal_pred.
3776 2018-03-22 Jakub Jelinek <jakub@redhat.com>
3779 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
3780 DECL_INITIAL (decl) to decl at the end.
3781 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
3784 2018-03-21 Joseph Myers <joseph@codesourcery.com>
3786 * doc/extend.texi (__builtin_tgmath): Document when complex
3787 integer types are treated as _Complex _Float64.
3789 2018-03-21 Tom de Vries <tom@codesourcery.com>
3791 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
3793 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3795 PR tree-optimization/84960
3796 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
3797 if it is ENTRY block, move them into single succ of ENTRY in that case.
3799 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
3801 PR tree-optimization/84811
3802 * poly-int.h (poly_span_traits): Remove the T3 parameter and
3803 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
3804 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
3805 (known_subrange_p): Update accordingly. Cast each value involved
3806 in the size comparison, rather than casting the result of the
3809 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3811 PR tree-optimization/84982
3812 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
3813 by flipping the least significant bit rather than all bits from
3814 bitpos to bitpos + bitsize - 1.
3816 2018-03-21 Nathan Sidwell <nathan@acm.org>
3818 * doc/extend.texi (Deprecated Features): Remove mention of
3819 long-deleted deprecations.
3821 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3824 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
3825 * configure: Regenerate.
3827 2018-03-21 Tom de Vries <tom@codesourcery.com>
3829 PR tree-optimization/83126
3830 * tree-parloops.c (num_phis): New function.
3831 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
3833 2018-03-21 Nathan Sidwell <nathan@acm.org>
3835 * doc/extend.texi (Deprecated Features): Update deprecated flags,
3836 mention anon-struct/union members and trailing attributes.
3838 2018-03-21 Bin Cheng <bin.cheng@arm.com>
3840 PR tree-optimization/84969
3841 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
3842 builtin memset partitions if they set different rhs values.
3844 2018-03-21 Jakub Jelinek <jakub@redhat.com>
3846 PR rtl-optimization/84989
3847 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
3848 VEC_DUPLICATE with scalar result mode.
3850 2018-03-21 Martin Liska <mliska@suse.cz>
3853 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
3854 not intended return statement.
3856 2018-03-21 Martin Liska <mliska@suse.cz>
3859 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
3860 (chkp_find_bound_slots_1): Limit number of iterations.
3862 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
3865 * Minor grammar fixes for x86 options.
3867 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3870 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
3871 holding REG_CFA_RESTORE notes, instead turn them into a USE.
3873 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
3876 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
3877 (altivec_lvx_<mode>_1op): Likewise.
3878 (altivec_stvx_<mode>_2op): Likewise.
3879 (altivec_stvx_<mode>_1op): Likewise.
3880 (altivec_lvx_<VM2:mode>): New define_expand.
3881 (altivec_stvx_<VM2:mode>): Likewise.
3882 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
3883 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3884 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
3885 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
3886 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
3887 (rs6000_gen_lvx): Likewise.
3888 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
3889 (altivec_expand_stv_builtin): Likewise.
3890 (altivec_expand_builtin): Likewise.
3891 * config/rs6000/vector.md: Likewise.
3893 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3896 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
3899 2018-03-20 Richard Biener <rguenther@suse.de>
3902 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
3903 sign-conversions as zero, fall back to standard scalar_stmt
3906 2018-03-20 Martin Liska <mliska@suse.cz>
3909 * predict.c (rebuild_frequencies): Handle case when we have
3910 PROFILE_ABSENT, but flag_guess_branch_prob is false.
3912 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3915 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
3916 flag_section_anchors.
3917 * varasm.c (use_blocks_for_decl_p): Remove hack for
3918 dw2_force_const_mem.
3921 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
3923 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
3924 be created, use lowpart_subreg of operands[0] rather than operands[0]
3926 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
3927 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
3928 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
3929 and n constraint instead of aarch64_shift_imm_di and Usd.
3930 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
3931 (*aarch64_<optab>_reg_minus<mode>3): ... this.
3933 2018-03-20 Sudakshina Das <sudi.das@arm.com>
3936 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
3937 to favor GPR over NEON registers.
3938 (<shift>di3_neon): Likewise.
3940 2018-03-20 Tom de Vries <tom@codesourcery.com>
3943 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
3944 (nvptx_process_pars): Emit bar.sync asap and alap.
3946 2018-03-20 Tom de Vries <tom@codesourcery.com>
3949 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
3950 seen_label if seen_label is already set.
3952 2018-03-20 Jakub Jelinek <jakub@redhat.com>
3955 * config/i386/i386.c (fold_builtin_cpu): For features above 31
3956 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
3957 Use 1U instead of 1. Formatting fixes.
3960 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
3961 instead of TREE_TYPE (s1) for the return value.
3963 2018-03-19 Jakub Jelinek <jakub@redhat.com>
3965 PR tree-optimization/84946
3966 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
3967 bitsize + bitsize in poly_uint64 rather than poly_int64.
3970 * dwarf2asm.c: Include fold-const.c.
3971 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
3972 of decl rather than decl itself.
3974 PR rtl-optimization/84643
3975 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
3977 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
3980 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
3981 calling assemble_variable.
3983 2018-03-19 Sudakshina Das <sudi.das@arm.com>
3986 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
3987 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
3989 2018-03-19 Jim Wilson <jimw@sifive.com>
3992 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
3993 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
3994 (riscv_first_stack_step): Likewise.
3995 (riscv_option_override): Use STACK_BOUNDARY instead of
3997 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
3999 (BIGGEST_ALIGNMENT): Set to 128.
4000 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
4001 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
4004 2018-03-19 Richard Biener <rguenther@suse.de>
4006 PR tree-optimization/84933
4007 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
4008 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
4010 2018-03-19 Richard Biener <rguenther@suse.de>
4012 PR tree-optimization/84859
4013 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
4014 (cond_if_else_store_replacement): Perform sinking operation on
4015 single-store BBs regardless of MAX_STORES_TO_SINK setting.
4016 Generalize what a BB with a single eligible store is.
4018 2018-03-19 Richard Biener <rguenther@suse.de>
4020 PR tree-optimization/84929
4021 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
4022 chrec_is_positive against non-chrec arg.
4024 2018-03-19 Tamar Christina <tamar.christina@arm.com>
4027 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
4029 2018-03-18 Martin Liska <mliska@suse.cz>
4031 PR rtl-optimization/84635
4032 * regrename.c (build_def_use): Use matches_mode only when
4035 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
4037 PR tree-optimization/84913
4038 * tree-vect-loop.c (vectorizable_reduction): Don't try to
4039 vectorize chains of COND_EXPRs.
4041 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4043 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
4045 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4047 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
4049 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
4051 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
4053 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4054 Kito Cheng <kito.cheng@gmail.com>
4056 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
4057 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
4058 (nds32_adjust_reg_alloc_order): New function.
4059 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
4061 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
4063 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
4064 nds32_print_operand, nds32_print_operand_address): Use
4065 HOST_WIDE_INT_PRINT_DEC instead.
4067 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
4069 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
4071 2018-03-17 Jakub Jelinek <jakub@redhat.com>
4074 * config/i386/i386.c (initial_ix86_tune_features,
4075 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
4077 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
4078 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
4079 rather than 1u << ix86_tune. Formatting fix.
4080 (ix86_option_override_internal): Change ix86_arch_mask from
4081 unsigned int to unsigned HOST_WIDE_INT, initialize to
4082 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
4083 (ix86_function_specific_restore): Likewise.
4085 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4088 * postreload.c (reload_combine_recognize_pattern): Perform
4089 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
4090 truncate_int_for_mode the result for the destination's mode.
4093 * hsa-gen.c (mem_type_for_type): Fix comment typo.
4094 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
4096 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4099 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
4102 * lra-assigns.c (lra_split_hard_reg_for): Don't use
4103 regno_allocno_class_array and sorted_pseudos.
4104 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
4105 insns where regno is used.
4107 2018-03-16 Martin Liska <mliska@suse.cz>
4110 * multiple_target.c (create_dispatcher_calls): Redirect
4111 reference in the symbol table.
4113 2018-03-16 Martin Liska <mliska@suse.cz>
4116 * multiple_target.c (create_dispatcher_calls): Redirect also
4119 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4123 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
4124 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
4125 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
4127 2018-03-16 Julia Koval <julia.koval@intel.com>
4129 * doc/invoke.texi (Skylake Server): Add CLWB.
4130 Cannonlake): Remove CLWB.
4132 2018-03-16 Jakub Jelinek <jakub@redhat.com>
4134 PR tree-optimization/84841
4135 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
4137 (FLOAT_ONE_CONST_TYPE): Define.
4138 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
4139 (sort_by_operand_rank): Put entries with higher constant_type last
4140 rather than first to match comments.
4142 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
4144 * config/nios2/nios2.md (movsi_internal): Fix thinko in
4147 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4150 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
4151 check and use address of target always.
4153 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
4156 * config/i386/i386.c (indirect_thunk_needed): Update comments.
4157 (indirect_thunk_bnd_needed): Likewise.
4158 (indirect_thunks_used): Likewise.
4159 (indirect_thunks_bnd_used): Likewise.
4160 (indirect_return_needed): New.
4161 (indirect_return_bnd_needed): Likewise.
4162 (output_indirect_thunk_function): Add a bool argument for
4164 (output_indirect_thunk_function): Don't generate alias for
4165 function return thunk.
4166 (ix86_code_end): Call output_indirect_thunk_function to generate
4167 function return thunks.
4168 (ix86_output_function_return): Set indirect_return_bnd_needed
4169 and indirect_return_needed instead of indirect_thunk_bnd_needed
4170 and indirect_thunk_needed.
4172 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
4174 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
4175 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
4176 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
4178 2018-03-15 David Malcolm <dmalcolm@redhat.com>
4179 Paul Hua <paul.hua.gm@gmail.com>
4182 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
4184 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
4186 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
4187 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
4190 2018-03-15 Tamar Christina <tamar.christina@arm.com>
4193 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
4194 instead of GET_MODE_SIZE when comparing Units.
4196 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
4199 * varasm.c (hash_section): Return an unchangeble hash value
4200 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
4201 Return !aarch64_can_use_per_function_literal_pools_p ().
4203 2018-03-15 Jakub Jelinek <jakub@redhat.com>
4206 * optabs.c (emit_conditional_move): Pass address of cmode's copy
4207 rather than address of cmode as last argument to prepare_cmp_insn.
4209 2018-03-15 Julia Koval <julia.koval@intel.com>
4211 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
4212 F_AVX512VNNI, F_AVX512BITALG): New.
4214 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
4217 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
4218 insn for floating-point loads and stores.
4220 2018-03-14 Carl Love <cel@us.ibm.com>
4222 * config/rs6000/rs6000-c.c: Add macro definitions for
4223 ALTIVEC_BUILTIN_VEC_PERMXOR.
4224 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
4225 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
4226 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
4227 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
4229 * config/doc/extend.texi: Add prototypes for vec_permxor.
4231 2018-03-14 David Malcolm <dmalcolm@redhat.com>
4234 * diagnostic-show-locus.c (class layout_point): Convert m_line
4235 from int to linenum_type.
4236 (line_span::comparator): Use linenum "compare" function when
4237 comparing line numbers.
4238 (test_line_span): New function.
4239 (layout_range::contains_point): Convert param "row" from int to
4241 (layout_range::intersects_line_p): Likewise.
4242 (layout::will_show_line_p): Likewise.
4243 (layout::print_source_line): Likewise.
4244 (layout::should_print_annotation_line_p): Likewise.
4245 (layout::print_annotation_line): Likewise.
4246 (layout::print_leading_fixits): Likewise.
4247 (layout::annotation_line_showed_range_p): Likewise.
4248 (struct line_corrections): Likewise for field m_row.
4249 (line_corrections::line_corrections): Likewise for param "row".
4250 (layout::print_trailing_fixits): Likewise.
4251 (layout::get_state_at_point): Likewise.
4252 (layout::get_x_bound_for_row): Likewise.
4253 (layout::print_line): Likewise.
4254 (diagnostic_show_locus): Likewise for locals "last_line" and
4256 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
4257 * input.c (selftest::test_linenum_comparisons): New function.
4258 (selftest::input_c_tests): Call it.
4259 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
4260 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
4261 * selftest.h (ASSERT_GT): New macro.
4262 (ASSERT_GT_AT): New macro.
4263 (ASSERT_LT): New macro.
4264 (ASSERT_LT_AT): New macro.
4266 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
4268 PR rtl-optimization/84780
4269 * combine.c (distribute_links): Don't make a link based on pc_rtx.
4271 2018-03-14 Martin Liska <mliska@suse.cz>
4273 * tree.c (record_node_allocation_statistics): Use
4274 get_stats_node_kind.
4275 (get_stats_node_kind): New function extracted from
4276 record_node_allocation_statistics.
4277 (free_node): Use get_stats_node_kind.
4279 2018-03-14 Richard Biener <rguenther@suse.de>
4281 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
4282 that the value-set of ANTIC_IN doesn't grow.
4285 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
4287 (BB_VISITED_WITH_VISITED_SUCCS): New define.
4288 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
4290 2018-03-14 Julia Koval <julia.koval@intel.com>
4292 * config.gcc (icelake-client, icelake-server): New.
4294 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
4295 (initial_ix86_arch_features): Ditto.
4296 (PTA_SKYLAKE): Add SGX.
4297 (PTA_ICELAKE): Remove.
4298 (PTA_ICELAKE_CLIENT): New.
4299 (PTA_ICELAKE_SERVER): New.
4300 (ix86_option_override_internal): Split up icelake on icelake client and
4302 (get_builtin_code_for_version): Ditto.
4303 (fold_builtin_cpu): Ditto.
4304 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
4305 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
4306 * config/i386/i386.h (processor_type): Ditto.
4307 * doc/invoke.texi: Ditto.
4309 2018-03-14 Jakub Jelinek <jakub@redhat.com>
4312 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
4313 INTEGER_CST offset, add it together with bitpos / 8 and
4314 sign extend based on POINTER_SIZE.
4318 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
4321 * config/i386/constraints.md (Yc): New register constraint.
4322 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
4323 Use Yc constraint for alternative 2 of operand 0. Remove
4324 preferred_for_speed attribute.
4326 2018-03-14 Richard Biener <rguenther@suse.de>
4328 PR tree-optimization/84830
4329 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
4330 with the old one to avoid oscillations.
4332 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
4335 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
4337 (assign_by_spills): Return a flag of reload assignment failure.
4338 Do not process the reload assignment failures. Do not spill other
4339 reload pseudos if they has the same reg class. Update n if
4341 (lra_assign): Add a return arg. Set up from the result of
4342 assign_by_spills call.
4343 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4344 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4345 usage_insns if it is not NULL.
4346 (spill_hard_reg_in_range): New function.
4347 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4348 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4349 function prototypes.
4350 (lra_assign): Change prototype.
4351 * lra.c (lra): Add code to deal with fails by splitting hard reg
4354 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
4356 * config/riscv/riscv.opt (mrelax): New option.
4357 * config/riscv/riscv.c (riscv_file_start): Emit ".option
4358 "norelax" when riscv_mrelax is disabled.
4359 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
4361 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4364 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
4365 reassociation for int modes.
4367 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4369 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
4370 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
4372 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
4373 * config/aarch64/aarch64-sve.md
4374 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
4375 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
4376 (*extend<mode><Vwide>2): Rename to...
4377 (aarch64_sve_extend<mode><Vwide>2): ...this.
4378 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
4379 renaming the old pattern to...
4380 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
4382 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
4383 define_expand, renaming the old pattern to...
4384 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
4385 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
4386 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
4387 account when deciding which SVE instruction the optab should use.
4388 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
4390 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4392 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
4393 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
4394 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
4395 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
4396 (tlsdesc_small_<mode>): Turn a define_expand and use
4397 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
4398 (tlsdesc_small_advsimd_<mode>): ...this.
4399 (tlsdesc_small_sve_<mode>): New pattern.
4401 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
4403 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
4404 (UNSPEC_UMUL_HIGHPART): New constants.
4405 (MUL_HIGHPART): New int iteraor.
4406 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
4407 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
4409 (*<su>mul<mode>3_highpart): New define_insn.
4411 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4414 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4417 2018-03-13 Martin Liska <mliska@suse.cz>
4420 * (sem_item_optimizer::sem_item_optimizer): Initialize new
4422 (sem_item_optimizer::~sem_item_optimizer): Release it.
4423 (sem_item_optimizer::merge_classes): Register variable aliases.
4424 (sem_item_optimizer::fixup_pt_set): New function.
4425 (sem_item_optimizer::fixup_points_to_sets): Likewise.
4426 * ipa-icf.h: Declare new variables and functions.
4428 2018-03-13 Jakub Jelinek <jakub@redhat.com>
4431 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
4432 integer_pow2p@2 and test integer_pow2p in condition.
4433 (A < 0 ? C : 0): Similarly for @1.
4436 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
4437 characters starting at p contain '\0' character, don't look beyond
4441 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
4442 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
4445 * reg-stack.c (change_stack): Change update_end var from int to
4446 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
4447 also call set_block_for_insn on the newly added insns and rescan.
4450 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
4451 on the last operand.
4454 * tree.c (stabilize_reference_1): Return save_expr (e) for
4455 STATEMENT_LIST even if it doesn't have side-effects.
4457 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
4459 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
4461 2018-03-12 Renlin Li <renlin.li@arm.com>
4463 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
4464 aarch64_output_scalar_simd_mov_immediate.
4466 2018-03-12 Martin Sebor <msebor@redhat.com>
4468 PR tree-optimization/83456
4469 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
4470 for perfectly overlapping calls to memcpy.
4471 (gimple_fold_builtin_memory_chk): Same.
4472 (gimple_fold_builtin_strcpy): Handle no-warning.
4473 (gimple_fold_builtin_stxcpy_chk): Same.
4474 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
4476 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4478 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
4479 parameter. Use it for SFmode.
4480 (rs6000_function_arg_advance_1): Adjust.
4481 (rs6000_function_arg): Adjust.
4482 (rs6000_gimplify_va_arg): Pass false for that new parameter.
4484 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
4486 PR rtl-optimization/84169
4487 PR rtl-optimization/84780
4488 * combine.c (can_combine_p): Check for a 2-insn combination whether
4489 the destination register is used between the two insns, too.
4491 2018-03-12 Richard Biener <rguenther@suse.de>
4493 PR tree-optimization/84803
4494 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
4495 for refs DR analysis didn't process.
4497 2018-03-12 Richard Biener <rguenther@suse.de>
4499 PR tree-optimization/84777
4500 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
4501 force-vectorize loops ignore whether we are optimizing for size.
4503 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
4505 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
4506 (TARGET_MD_ASM_ADJUST): Define.
4508 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
4509 Kito Cheng <kito.cheng@gmail.com>
4510 Chung-Ju Wu <jasonwucj@gmail.com>
4512 * config/nds32/nds32.c (nds32_compute_stack_frame,
4513 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
4514 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
4515 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
4516 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
4517 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
4518 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
4519 * config/nds32/nds32.md (prologue, epilogue): Use macro
4520 NDS32_V3PUSH_AVAILABLE_P to do checking.
4522 2018-03-11 Jakub Jelinek <jakub@redhat.com>
4525 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
4526 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
4527 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
4528 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
4529 addition of most attributes on !orig_type_die or the attribute not
4530 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
4532 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4533 Chung-Ju Wu <jasonwucj@gmail.com>
4535 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
4537 * config/nds32/nds32.opt (mvh): New option.
4539 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4540 Chung-Ju Wu <jasonwucj@gmail.com>
4542 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
4544 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
4545 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
4548 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
4549 Chung-Ju Wu <jasonwucj@gmail.com>
4551 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
4553 * config/nds32/nds32-multiple.md (strlensi): New pattern.
4554 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
4556 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
4557 Kito Cheng <kito.cheng@gmail.com>
4558 Chung-Ju Wu <jasonwucj@gmail.com>
4560 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
4561 UNSPEC_FFMISM and UNSPEC_FLMISM.
4562 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
4563 for ffb, ffmism and flmism.
4564 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
4565 (unspec_ffmism): Ditto.
4566 (unspec_flmism): Ditto.
4567 (nds32_expand_builtin_impl): Check if string extension is available.
4568 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
4569 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
4571 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
4574 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4577 * lra-assigns.c (assign_by_spills): Return a flag of reload
4578 assignment failure. Do not process the reload assignment
4579 failures. Do not spill other reload pseudos if they has the same
4581 (lra_assign): Add a return arg. Set up from the result of
4582 assign_by_spills call.
4583 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4584 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4585 usage_insns if it is not NULL.
4586 (spill_hard_reg_in_range): New function.
4587 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4588 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4589 function prototypes.
4590 (lra_assign): Change prototype.
4591 * lra.c (lra): Add code to deal with fails by splitting hard reg
4594 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
4597 * config/i386/i386.opt: Replace Enforcment with Enforcement.
4599 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
4602 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
4603 (dw_val_node): Add val_symbolic_view.
4604 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
4605 (symview_upper_bound): New.
4606 (new_line_info_table): Initialize symviews_since_reset.
4607 (dwarf2out_source_line): Count symviews_since_reset and set
4608 symview_upper_bound.
4609 (dw_val_equal_p): Handle symview.
4610 (add_AT_symview): New.
4611 (print_dw_val): Handle symview.
4612 (attr_checksum, attr_checksum_ordered): Likewise.
4613 (same_dw_val_p, size_of_die): Likewise.
4614 (value_format, output_die): Likewise.
4615 (add_high_low_attributes): Use add_AT_symview for entry_view.
4616 (dwarf2out_finish): Reset symview_upper_bound, clear
4619 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
4622 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
4623 Add strict argument and use it.
4624 (rs6000_split_multireg_move): Update for new strict argument.
4625 (mem_operand_gpr): Disallow all non-offsettable addresses.
4626 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
4628 2018-03-09 Jakub Jelinek <jakub@redhat.com>
4631 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
4632 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
4633 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
4636 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
4637 decl, use remap_type if we want to use the type.
4639 2018-03-09 Martin Sebor <msebor@redhat.com>
4641 PR tree-optimization/84526
4642 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
4644 (builtin_access::generic_overlap): Be prepared to handle non-array
4647 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
4649 PR rtl-optimization/84682
4650 * lra-constraints.c (process_address_1): Check is_address flag
4651 for address constraints.
4652 (process_alt_operands): Likewise.
4653 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
4654 preprocess_constraints.
4655 * recog.h (preprocess_constraints): Add oploc parameter.
4657 * recog.c (preprocess_constraints): Test address_operand for
4658 CT_ADDRESS constraints.
4660 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
4663 * lra-assigns.c (assign_by_spills): Return a flag of reload
4664 assignment failure. Do not process the reload assignment
4665 failures. Do not spill other reload pseudos if they has the same
4667 (lra_assign): Add a return arg. Set up from the result of
4668 assign_by_spills call.
4669 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
4670 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
4671 usage_insns if it is not NULL.
4672 (spill_hard_reg_in_range): New function.
4673 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
4674 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
4675 function prototypes.
4676 (lra_assign): Change prototype.
4677 * lra.c (lra): Add code to deal with fails by splitting hard reg
4680 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4683 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
4684 Accept complain bool parameter. Only emit errors if it is true.
4685 (arm_parse_cpu_option_name): Likewise.
4686 (arm_target_thumb_only): Adjust callers of the above.
4687 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
4688 prototype to take a default true bool parameter.
4689 (arm_parse_arch_option_name): Likewise.
4691 2018-03-09 David Malcolm <dmalcolm@redhat.com>
4692 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
4696 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
4697 * configure: Regenerate.
4698 * configure.ac ("linker --version-script option"): New.
4699 ("linker soname option"): New.
4701 2018-03-09 Richard Biener <rguenther@suse.de>
4703 PR tree-optimization/84775
4704 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
4705 immediate uses of predicate stmts and mark them modified.
4708 PR tree-optimization/84178
4709 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4711 (version_loop_for_if_conversion): Delay update_ssa call.
4712 (tree_if_conversion): Delay update_ssa until after predicate
4715 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4718 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
4719 when the function accesses prior frames.
4721 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4724 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
4725 gen_llsym, otherwise call maybe_gen_llsym.
4728 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
4729 has ',' character inside of it.
4731 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4734 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
4735 as clobbering CC_REGNUM.
4737 2018-03-08 Richard Biener <rguenther@suse.de>
4740 * tree-scalar-evolution.c: Include tree-into-ssa.h.
4741 (follow_copies_to_constant): Do not follow SSA names registered
4744 2018-03-08 Richard Biener <rguenther@suse.de>
4746 PR tree-optimization/84178
4747 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
4749 (version_loop_for_if_conversion): Delay update_ssa call.
4750 (tree_if_conversion): Delay update_ssa until after predicate
4753 2018-03-08 David Malcolm <dmalcolm@redhat.com>
4755 PR tree-optimization/84178
4756 * tree-if-conv.c (release_bb_predicate): Remove the
4757 the assertion that the stmts have NULL use_ops.
4758 Discard the statements, asserting that they haven't
4759 yet been added to a BB.
4761 2018-03-08 Richard Biener <rguenther@suse.de>
4763 PR tree-optimization/84746
4764 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
4765 (phi_translate): Pass in destination ANTIC_OUT set.
4766 (phi_translate_1): Likewise. For a simplified result lookup
4767 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
4768 (phi_translate_set): Adjust.
4769 (do_pre_regular_insertion): Likewise.
4770 (do_pre_partial_partial_insertion): Likewise.
4772 2018-03-08 Martin Liska <mliska@suse.cz>
4774 PR gcov-profile/84735
4775 * doc/gcov.texi: Document usage of profile files.
4776 * gcov-io.h: Document changes in the format.
4778 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
4782 * dwarf2out.c (struct dw_line_info_table): Update comments for
4784 (FORCE_RESET_NEXT_VIEW): New.
4785 (FORCE_RESETTING_VIEW_P): New.
4786 (RESETTING_VIEW_P): Check for -1 too.
4787 (ZERO_VIEW_P): Likewise.
4788 (new_line_info_table): Force-reset next view.
4789 (dwarf2out_begin_function): Likewise.
4790 (dwarf2out_source_line): Simplify zero_view_p initialization.
4791 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
4792 view directly. Omit view when omitting .loc at line 0.
4794 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4796 PR tree-optimization/84740
4797 * tree-switch-conversion.c (process_switch): Call build_constructors
4798 only if info.phi_count is non-zero.
4800 PR tree-optimization/84739
4801 * tree-tailcall.c (find_tail_calls): Check call arguments against
4802 DECL_ARGUMENTS (current_function_decl) rather than
4803 DECL_ARGUMENTS (func) when checking for tail recursion.
4805 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4807 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
4808 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
4809 Volker Reichelt's entry and add entries for people that perform
4810 GCC fuzzy testing and report numerous bugs.
4812 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
4815 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
4816 readonly data in sdata, if that is disabled.
4817 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
4818 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
4819 -mreadonly-in-sdata option.
4821 2018-03-07 Martin Sebor <msebor@redhat.com>
4823 PR tree-optimization/84468
4824 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
4825 basic block when looking for nul assignment.
4827 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
4830 * except.h (output_function_exception_table): Adjust prototype.
4831 * except.c (output_function_exception_table): Remove FNNAME parameter
4832 and add SECTION parameter. Ouput one part of the table at a time.
4833 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
4834 the first part of the exception table and emit unwind directives.
4835 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
4836 (i386_pe_seh_cold_init): Likewise.
4837 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
4838 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
4839 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
4840 (ix86_output_call_insn): Emit a nop in one more case for SEH.
4841 * config/i386/winnt.c: Include except.h.
4842 (struct seh_frame_state): Add reg_offset, after_prologue and
4843 in_cold_section fields.
4844 (i386_pe_seh_end_prologue): Set seh->after_prologue.
4845 (i386_pe_seh_cold_init): New function.
4846 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
4847 to seh->in_cold_section.
4848 (seh_emit_push): Record the offset of the push.
4849 (seh_emit_save): Record the offet of the save.
4850 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
4851 Test seh->after_prologue to disregard the epilogue.
4852 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
4853 (i386_pe_end_cold_function): New function.
4855 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4858 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
4859 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
4862 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
4864 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
4865 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
4868 * multiple_target.c: Include tree-inline.h and intl.h.
4869 (expand_target_clones): Diagnose and fail if node->definition and
4870 !tree_versionable_function_p (node->decl).
4872 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
4874 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
4876 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
4877 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
4878 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
4880 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4883 * combine.c (try_combine): Use reg_or_subregno instead of handling
4884 just paradoxical SUBREGs and REGs.
4886 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
4888 * config/arc/arc.c (arc_finalize_pic): Remove function.
4889 (arc_must_save_register): We use single base PIC register, remove
4890 checks to save/restore the PIC register.
4891 (arc_expand_prologue): Likewise.
4892 * config/arc/arc-protos.h (arc_set_default_type_attributes):
4894 (arc_verify_short): Likewise.
4895 (arc_attr_type): Likewise.
4896 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
4897 (walk_stores): Likewise.
4898 (arc_address_cost): Make it static.
4899 (arc_verify_short): Likewise.
4900 (branch_dest): Likewise.
4901 (arc_attr_type): Likewise.
4902 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
4903 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
4904 (arc_final_prescan_insn): Remove inserting the nops due to
4905 hardware hazards. It is done in reorg step.
4906 (insn_length_variant_t): Remove.
4907 (insn_length_parameters_t): Likewise.
4908 (arc_insn_length_parameters): Likewise.
4909 (arc_get_insn_variants): Likewise.
4910 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
4912 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4915 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
4918 PR tree-optimization/84687
4919 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
4921 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
4923 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4925 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
4926 Rename to ppc_speculation_barrier.
4927 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
4928 __builtin_ppc_speculation_barrier.
4930 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4933 * combine.c (combine_simplify_rtx): Don't try to simplify if
4934 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
4937 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
4939 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
4940 to 32 bytes when compiling for POWER9.
4942 2018-03-05 Jakub Jelinek <jakub@redhat.com>
4945 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
4946 regparm >= 3 with no arg reg available also for calls with
4947 flag_force_indirect_call. Pass decl to ix86_function_regparm.
4950 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
4952 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
4954 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
4957 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
4959 2018-03-05 Richard Biener <rguenther@suse.de>
4961 PR tree-optimization/84486
4962 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
4963 When inserting a __builtin_assume_aligned call set the LHS
4964 SSA name alignment info accordingly.
4966 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
4968 PR tree-optimization/84114
4969 * config/aarch64/aarch64.c (aarch64_reassociation_width)
4970 Avoid reassociation of FLOAT_MODE addition.
4972 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
4974 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
4975 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
4976 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
4977 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
4978 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
4979 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
4980 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
4982 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
4983 __builtin_ia32_wbinvd): New builtins.
4984 (SPECIAL_ARGS2): New.
4985 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
4986 (SPECIAL_ARGS2): New.
4987 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
4988 (ix86_valid_target_attribute_inner_p): Ditto.
4989 (ix86_init_mmx_sse_builtins): Add special_args2.
4990 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
4991 TARGET_WBNOINVD_P): New.
4992 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
4993 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
4994 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
4995 * config/i386/immintrin.h (_wbinvd): New intrinsic.
4996 * config/i386/pconfigintrin.h: New file.
4997 * config/i386/wbnoinvdintrin.h: Ditto.
4998 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
4999 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
5001 2018-03-05 Richard Biener <rguenther@suse.de>
5003 PR tree-optimization/84670
5004 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
5006 (BB_VISITED_WITH_VISITED_SUCCS): New define.
5007 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
5008 (compute_antic_aux): Only assert the number of values in ANTIC_IN
5009 doesn't grow if all successors (recursively) were visited at least
5012 2018-03-05 Richard Biener <rguenther@suse.de>
5014 PR tree-optimization/84650
5015 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
5016 if executed in the loop pipeline.
5018 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
5020 * doc/configfiles.texi (Configuration Files): Move info about
5021 conditionalizing $target-protos.h to...
5022 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
5023 differs from $target-protos.h.
5025 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
5026 Chung-Ju Wu <jasonwucj@gmail.com>
5028 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
5029 * config/nds32/nds32-multiple.md (setmemsi): Define.
5030 * config/nds32/nds32-memory-manipulation.c
5031 (nds32_gen_dup_4_byte_to_word_value): New.
5032 (emit_setmem_word_loop): New.
5033 (emit_setmem_byte_loop): New.
5034 (nds32_expand_setmem_loop): New.
5035 (nds32_expand_setmem_loop_v3m): New.
5036 (nds32_expand_setmem_unroll): New.
5037 (nds32_expand_setmem): New.
5039 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5040 Chung-Ju Wu <jasonwucj@gmail.com>
5042 * config/nds32/nds32-memory-manipulation.c
5043 (nds32_emit_load_store): New.
5044 (nds32_emit_post_inc_load_store): New.
5045 (nds32_emit_mem_move): New.
5046 (nds32_emit_mem_move_block): New.
5047 (nds32_expand_movmemsi_loop_unknown_size): New.
5048 (nds32_expand_movmemsi_loop_known_size): New.
5049 (nds32_expand_movmemsi_loop): New.
5050 (nds32_expand_movmemsi_unroll): New.
5051 (nds32_expand_movmemqi): Rename ...
5052 (nds32_expand_movmemsi): ... to this.
5053 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
5054 (movmemsi): ... to this.
5055 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
5056 (nds32_expand_movmemsi): ... to this.
5058 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5059 Monk Chiang <sh.chiang04@gmail.com>
5060 Chung-Ju Wu <jasonwucj@gmail.com>
5062 * config/nds32/nds32-protos.h
5063 (nds32_expand_load_multiple): New arguments.
5064 (nds32_expand_store_multiple): Ditto.
5065 (nds32_valid_multiple_load_store): Rename ...
5066 (nds32_valid_multiple_load_store_p): ... to this.
5067 * config/nds32/nds32-memory-manipulation.c
5068 (nds32_expand_load_multiple): Refine implementation.
5069 (nds32_expand_store_multiple): Ditto.
5070 * config/nds32/nds32-multiple.md
5071 (load_multiple): Update nds32_expand_load_multiple interface.
5072 (store_multiple): Update nds32_expand_store_multiple interface.
5073 * config/nds32/nds32-predicates.c
5074 (nds32_valid_multiple_load_store): Rename ...
5075 (nds32_valid_multiple_load_store_p): ... to this and refine
5077 * config/nds32/predicates.md
5078 (nds32_load_multiple_and_update_address_operation): New predicate.
5079 (nds32_store_multiple_and_update_address_operation): New predicate.
5081 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
5082 Chung-Ju Wu <jasonwucj@gmail.com>
5084 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
5085 (combo): New attribute.
5086 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
5088 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
5090 * config/nds32/nds32.opt: Change -mcmodel= default value.
5092 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
5093 Monk Chiang <sh.chiang04@gmail.com>
5094 Chung-Ju Wu <jasonwucj@gmail.com>
5096 * config/nds32/constants.md (unspec_element): New enum.
5097 * config/nds32/constraints.md (Umw): New constraint.
5098 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
5099 * config/nds32/nds32-intrinsic.md: Likewise.
5100 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
5101 (nds32_valid_smw_lwm_base_p): New.
5102 (nds32_output_smw_single_word): New.
5103 (nds32_output_lmw_single_word): New.
5104 (nds32_expand_unaligned_load): New.
5105 (nds32_expand_unaligned_store): New.
5106 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
5107 (nds32_output_smw_single_word): Declare.
5108 (nds32_output_lmw_single_word): Declare.
5109 (nds32_expand_unaligned_load): Declare.
5110 (nds32_expand_unaligned_store): Declare.
5111 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
5112 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
5113 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
5114 NDS32_BUILTIN_UASTORE_DW.
5115 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
5118 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
5119 Kito Cheng <kito.cheng@gmail.com>
5120 Chung-Ju Wu <jasonwucj@gmail.com>
5122 * config/nds32/nds32-intrinsic.c
5123 (nds32_expand_builtin_null_ftype_reg): Delete.
5124 (nds32_expand_builtin_reg_ftype_imm): Ditto.
5125 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
5126 (nds32_read_argument): New.
5127 (nds32_legitimize_target): Ditto.
5128 (nds32_legitimize_argument): Ditto.
5129 (nds32_check_constant_argument): Ditto.
5130 (nds32_expand_unop_builtin): Ditto.
5131 (nds32_expand_unopimm_builtin): Ditto.
5132 (nds32_expand_binop_builtin): Ditto.
5133 (nds32_builtin_decl_impl): Ditto.
5134 (builtin_description): Ditto.
5135 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
5136 (nds32_init_builtins_impl): Ditto.
5137 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
5138 (nds32_builtin_decl): New.
5139 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
5140 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
5142 2018-03-02 Jeff Law <law@redhat.com>
5144 * reorg.c (stop_search_p): Handle DEBUG_INSN.
5145 (redundant_insn, fill_simple_delay_slots): Likewise.
5146 (fill_slots_from_thread): Likewise.
5147 * resource.c (mark_referenced_resources): Likewise.
5148 (mark_set_resources, find_dead_or_set_registers): Likewise.
5150 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5152 * substring-locations.h (format_warning_va): Formatting fix for
5154 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
5156 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
5157 * substring-locations.c: Include intl.h.
5158 (format_warning_va): Turned into small wrapper around
5159 format_warning_n_va, renamed to ...
5160 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
5161 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
5163 (format_warning_at_substring_n): New function.
5164 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
5165 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
5166 format_warning_at_substring with just a shorter name instead of
5167 const function pointer.
5168 (fmtwarn_n): New function.
5169 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
5170 appropriate, get rid of all the fmtstr temporaries, move conditionals
5171 with G_() wrapped string literals directly into fmtwarn arguments,
5172 cast dir.len to (int), formatting fixes.
5174 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
5176 * doc/invoke.texi: Remove "Cilk Plus" references.
5178 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5179 Richard Biener <rguenther@suse.de>
5182 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
5183 for error or warning attributes if CALL_FROM_THUNK_P is set.
5186 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5189 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
5190 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
5193 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
5194 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
5196 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
5197 instead of -1U in last predictors element's probability member.
5199 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5202 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5203 arguments if they are comparable.
5205 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5207 PR tree-optimization/84634
5208 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
5209 masks and masked_loop_p with a single loop_masks, making sure it's
5210 null for bb vectorization.
5212 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
5214 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5215 (vect_analyze_data_ref_access): Use loop->safe_len rather than
5216 loop->force_vectorize to check whether there is no alias.
5218 2018-03-02 Jakub Jelinek <jakub@redhat.com>
5221 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
5223 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
5225 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
5226 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
5227 instead of a loop around prev_real_insn.
5228 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
5232 * config/i386/i386.c (ix86_print_operand): Use conditional
5233 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
5236 2018-03-02 Richard Biener <rguenther@suse.de>
5238 PR tree-optimization/84427
5239 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
5240 (bitmap_set_subtract_values): Rewrite to handle multiple
5243 (prune_clobbered_mems): Likewise.
5244 (phi_translate): Take edge instead of pred/phiblock.
5245 (phi_translate_1): Likewise.
5246 (phi_translate_set): Likewise. Insert all translated
5247 exprs for a value into the set, keeping possibly multiple
5248 expressions per value.
5249 (compute_antic_aux): Adjust for phi_translate changes.
5250 When intersecting union the expressions and prune those
5251 not in the final value set, keeping possibly multiple
5252 expressions per value. Do not use value-insertion
5253 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
5254 all expressions. Add verification that the value-sets
5255 only shrink during iteration.
5256 (compute_partial_antic_aux): Adjust for the phi_translate changes.
5257 (do_pre_regular_insertion): Likewise.
5258 (do_pre_partial_partial_insertion): Likewise.
5260 2018-03-02 Richard Biener <rguenther@suse.de>
5263 * config/darwin.c (saved_debug_info_level): New static global.
5264 (darwin_asm_lto_start): Disable debug info generation for LTO out.
5265 (darwin_asm_lto_end): Restore debug info generation settings.
5267 2018-03-01 Martin Liska <mliska@suse.cz>
5270 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
5273 2018-03-01 Richard Biener <rguenther@suse.de>
5276 * dwarf2out.c (gen_variable_die): Properly handle late VLA
5277 type annotation with LTO when debug was disabled at compile-time.
5279 2018-03-01 Matthew Fortune <mfortune@gmail.com>
5281 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
5283 (mips_final_postscan_insn): Likewise.
5285 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
5287 PR rtl-optimization/84528
5288 * alias.c (init_alias_target): Add commentary.
5289 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
5290 a unique base value if the frame pointer is not eliminated
5291 to the stack pointer.
5293 2018-03-01 Tom de Vries <tom@codesourcery.com>
5295 PR rtl-optimization/83327
5296 * lra-int.h (hard_regs_spilled_into): Declare.
5297 * lra.c (hard_regs_spilled_into): Define.
5298 (init_reg_info): Init hard_regs_spilled_into.
5299 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
5300 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
5301 (process_bb_lives): Handle hard_regs_spilled_into.
5302 (lra_create_live_ranges_1): Before doing liveness propagation, clear
5303 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
5305 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
5307 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
5308 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
5309 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
5310 * config/rs6000/aix72.h: New file.
5312 2018-02-28 Jakub Jelinek <jakub@redhat.com>
5314 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
5315 instead of warning_at with conditional singular and plural messages
5319 * stor-layout.c (update_alignment_for_field): For
5320 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
5321 && !DECL_PACKED (field), do the alignment update, just use
5322 only desired_align instead of MAX (type_align, desired_align)
5324 (place_field): Don't do known_align < desired_align handling
5325 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
5326 is non-NULL, instead do it after rli->prev_field handling and
5327 only if not within a bitfield word. For DECL_PACKED (field)
5328 use type_align of BITS_PER_UNIT.
5330 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
5332 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
5333 superfluous parentheses and trailing spaces.
5335 2018-02-28 Richard Biener <rguenther@suse.de>
5337 PR tree-optimization/84584
5338 * graphite-scop-detection.c (scop_detection::add_scop): Discard
5339 SCoPs with fake exit edge.
5341 2018-02-28 Martin Liska <mliska@suse.cz>
5344 * timevar.c (timer::print): Fix format to properly print 100%
5347 2018-02-28 Richard Biener <rguenther@suse.de>
5350 * genmatch.c (capture_info::walk_match): Do not mark
5351 captured expressions without operands as expr_p given
5352 they act more like predicates and should be subject to
5353 "lost tail" side-effect preserving.
5355 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
5357 PR rtl-optimization/81611
5358 * auto-inc-dec.c (attempt_change): Move dead note from
5359 mem_insn if it's the next use of regno
5360 (find_address): Take address use of reg holding
5361 non-incremented value. Add parm to limit search to the named
5363 (merge_in_block): Attempt to use a mem insn that is the next
5364 use of the original regno.
5366 2018-02-27 Martin Sebor <msebor@redhat.com>
5369 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
5370 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
5372 2018-02-27 Martin Sebor <msebor@redhat.com>
5374 PR translation/84207
5375 * diagnostic-core.h (warning_n, error_n, inform_n): Change
5376 n argument to unsigned HOST_WIDE_INT.
5377 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
5378 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
5379 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
5380 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
5382 2018-02-27 Richard Biener <rguenther@suse.de>
5384 PR tree-optimization/84512
5385 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5386 Do not use the estimate returned from record_stmt_cost for
5387 the scalar iteration cost but sum properly using add_stmt_cost.
5389 2018-02-27 Richard Biener <rguenther@suse.de>
5391 PR tree-optimization/84466
5392 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
5393 Adjust last change to less strictly validate use operands.
5395 2018-02-27 Martin Liska <mliska@suse.cz>
5397 PR gcov-profile/84548
5398 * gcov.c (process_file): Allow partial overlap and consider it
5399 also as group functions.
5400 (output_lines): Properly calculate range of lines for a group.
5402 2018-02-27 Martin Liska <mliska@suse.cz>
5404 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
5405 'ggc' suffixes. Change first column width.
5406 (timer::print): Fix formatting of the column.
5408 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
5410 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
5411 preserve inline entry blocks for the sake of debug inline
5412 entry point markers alone.
5413 (remove_unused_locals): Suggest in comments a better place to
5414 force the preservation of inline entry blocks that are
5415 otherwise unused, but do not preserve them.
5417 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5419 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
5421 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5424 * config/i386/constraints.md (Bs): Replace
5425 ix86_indirect_branch_register with
5426 TARGET_INDIRECT_BRANCH_REGISTER.
5428 * config/i386/i386.md (indirect_jump): Likewise.
5429 (tablejump): Likewise.
5430 (*sibcall_memory): Likewise.
5431 (*sibcall_value_memory): Likewise.
5432 Peepholes of indirect call and jump via memory: Likewise.
5433 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
5434 (*sibcall_value_GOT_32): Likewise.
5435 * config/i386/predicates.md (indirect_branch_operand): Likewise.
5436 (GOT_memory_operand): Likewise.
5437 (call_insn_operand): Likewise.
5438 (sibcall_insn_operand): Likewise.
5439 (GOT32_symbol_operand): Likewise.
5440 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
5442 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5444 PR rtl-optimization/83496
5445 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
5446 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
5447 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
5448 redundant insn, if any.
5449 (relax_delay_slots): Likewise.
5450 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
5452 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
5454 PR tree-optimization/83965
5455 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
5456 that grouped statements are part of a reduction chain. Return
5457 true if the statement is not marked as a reduction itself but
5459 (vect_recog_dot_prod_pattern): Don't check whether the statement
5460 is part of a group here.
5461 (vect_recog_sad_pattern): Likewise.
5462 (vect_recog_widen_sum_pattern): Likewise.
5464 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
5467 * final.c (rest_of_clean_state): Also look for calls inside sequences.
5469 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
5472 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
5474 (ix86_output_indirect_function_return): New prototype.
5475 (ix86_split_simple_return_pop_internal): Likewise.
5476 * config/i386/i386.c (indirect_return_via_cx): New.
5477 (indirect_return_via_cx_bnd): Likewise.
5478 (indirect_thunk_name): Handle return va CX_REG.
5479 (output_indirect_thunk_function): Create alias for
5480 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
5481 (ix86_output_indirect_jmp): Remove the bool argument.
5482 (ix86_output_indirect_function_return): New function.
5483 (ix86_split_simple_return_pop_internal): Likewise.
5484 * config/i386/i386.md (*indirect_jump): Don't pass false
5485 to ix86_output_indirect_jmp.
5486 (*tablejump_1): Likewise.
5487 (simple_return_pop_internal): Change it to define_insn_and_split.
5488 Call ix86_split_simple_return_pop_internal to split it for
5490 (simple_return_indirect_internal): Call
5491 ix86_output_indirect_function_return instead of
5492 ix86_output_indirect_jmp.
5494 2018-02-26 Jakub Jelinek <jakub@redhat.com>
5497 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
5498 memset and value initialization afterwards.
5500 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
5502 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
5504 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5507 * common/config/aarch64/aarch64-common.c
5508 (aarch_option_optimization_table[]): Switch
5509 off fomit-frame-pointer
5511 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5512 Chung-Ju Wu <jasonwucj@gmail.com>
5514 * config/nds32/nds32-multiple.md (load_multiple): Disallow
5516 (store_multiple): Ditto.
5518 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
5520 * config.gcc: Add --with-cpu support for nds32 target.
5521 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
5522 * config/nds32/nds32.opt: Add -mcpu= option.
5524 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
5526 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
5527 isel=yes): Warn for these deprecated options.
5529 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
5531 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
5532 ISA_2_5_MASKS_EMBEDDED.
5534 2018-02-23 Jakub Jelinek <jakub@redhat.com>
5536 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
5537 p->max as pointers rather than using iterative_hash_expr.
5539 2018-02-23 Carl Love <cel@us.ibm.com>
5541 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
5542 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
5544 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
5545 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
5546 P8V_BUILTIN_VEC_VUNSIGNED2.
5548 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
5551 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
5552 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
5554 * lra-constraints.c (curr_insn_transform): Set up
5555 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
5557 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
5558 * lra-eliminations.c (spill_pseudos): Ditto.
5559 (process_insn_for_elimination): Ditto.
5560 * lra-lives.c (reg_early_clobber_p): Use the new macros.
5561 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
5562 LRA_NON_CLOBBERED_ALT.
5564 2018-02-22 Martin Sebor <msebor@redhat.com>
5566 PR tree-optimization/84480
5567 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
5568 to maybe_diag_stxncpy_trunc. Call it.
5569 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
5570 from gimple_fold_builtin_strcpy. Print inlining stack.
5571 (handle_builtin_stxncpy): Print inlining stack.
5572 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
5574 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
5577 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
5578 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
5579 and -fcheck-pointer-bounds are used together.
5580 (indirect_thunk_prefix): New enum.
5581 (indirect_thunk_need_prefix): New function.
5582 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
5583 "_nt" instead of "_bnd" for NOTRACK prefix.
5584 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
5585 (output_indirect_thunk_function): Likewise.
5587 (ix86_code_end): Update output_indirect_thunk_function calls.
5588 (ix86_output_indirect_branch_via_reg): Replace
5589 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
5590 (ix86_output_indirect_branch_via_push): Likewise.
5591 (ix86_output_function_return): Likewise.
5592 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
5593 incompatible with -fcf-protection=branch and
5594 -fcheck-pointer-bounds.
5596 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5599 * config/aarch64/aarch64.c (aarch64_print_address_internal):
5600 Change gcc_assert call to output_operand_lossage.
5602 2018-02-22 Steve Ellcey <sellcey@cavium.com>
5604 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
5606 2018-02-22 DJ Delorie <dj@redhat.com>
5607 Sebastian Perta <sebastian.perta@renesas.com>
5608 Oleg Endo <olegendo@gcc.gnu.org>
5610 * config/rx/rx.c (rx_rtx_costs): New function.
5611 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
5613 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
5615 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
5617 2018-02-22 Martin Liska <mliska@suse.cz>
5620 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
5621 Add "native" as a possible value.
5623 2018-02-22 Martin Liska <mliska@suse.cz>
5626 * config/i386/i386.c (ix86_option_override_internal):
5627 Add "native" as a possible value for -march and -mtune.
5629 2018-02-22 Jakub Jelinek <jakub@redhat.com>
5632 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
5633 to all type variants.
5635 PR tree-optimization/84503
5636 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
5637 width as info->bitpos + info->bitsize - start.
5638 (merged_store_group::merge_overlapping): Simplify width computation.
5639 (check_no_overlap): New function.
5640 (imm_store_chain_info::try_coalesce_bswap): Compute expected
5641 start + width and last_order of the group, fail if check_no_overlap
5643 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
5644 to group if check_no_overlap fails.
5646 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5648 * config/rs6000/altivec.md: Delete contraint arguments to
5649 define_expand, define_split, and define_peephole2, and in
5650 define_insn_and_split if always unused.
5651 * config/rs6000/darwin.md: Ditto.
5652 * config/rs6000/dfp.md: Ditto.
5653 * config/rs6000/rs6000.md: Ditto.
5654 * config/rs6000/sync.md: Ditto.
5655 * config/rs6000/vector.md: Ditto.
5656 * config/rs6000/vsx.md: Ditto.
5658 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
5660 * config/rs6000/altivec.md: Write output control strings as braced
5661 blocks instead of double-quoted strings.
5662 * config/rs6000/darwin.md: Ditto.
5663 * config/rs6000/rs6000.md: Ditto.
5664 * config/rs6000/vector.md: Ditto.
5665 * config/rs6000/vsx.md: Ditto.
5667 2018-02-21 Jason Merrill <jason@redhat.com>
5669 PR c++/84314 - ICE with templates and fastcall attribute.
5670 * attribs.c (build_type_attribute_qual_variant): Remove assert.
5672 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5674 * ipa-cp.c (determine_versionability): Fix comment typos.
5676 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
5679 * ipa-cp.c (determine_versionability): Do not version functions caling
5682 2018-02-21 Martin Liska <mliska@suse.cz>
5685 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
5686 Add "native" as a possible value.
5687 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
5688 the macro when native cpu detection is available.
5690 2018-02-21 Martin Liska <mliska@suse.cz>
5693 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
5694 Add "native" as a possible value.
5695 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
5696 when native cpu detection is available.
5698 2018-02-21 Jakub Jelinek <jakub@redhat.com>
5699 Martin Sebor <msebor@redhat.com>
5701 PR tree-optimization/84478
5702 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
5704 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
5705 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
5706 support which is conservatively correct, for 2 only stay conservative
5707 for maxlen. Formatting and comment capitalization fixes. Add STRICT
5708 argument to the 2 argument get_range_strlen, adjust 6 arg
5709 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
5711 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
5712 (gimple_fold_builtin_strlen): Pass true as last argument to
5715 2018-02-20 Martin Sebor <msebor@redhat.com>
5718 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
5719 (builtin_memref::set_base_and_offset): Same. Handle inner references.
5720 (builtin_memref::builtin_memref): Factor out parts into
5721 set_base_and_offset and call it.
5723 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5726 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
5727 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
5728 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
5729 search at the associated MODE_INT.
5731 2018-02-20 Jeff Law <law@redhat.com>
5734 PR tree-optimization/81592
5736 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
5737 for range data rather than using global data.
5738 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
5739 range data rather than using global data.
5740 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
5741 pass it to children as needed.
5742 (struct directive::fmtresult): Similarly.
5743 (struct directive::set_width): Similarly.
5744 (struct directive::set_precision): Similarly.
5745 (format_integer, format_directive, parse_directive): Similarly.
5746 (format_none): Accept unnamed vr_values parameter.
5747 (format_percent, format_floating, format_character): Similarly.
5748 (format_string, format_plain): Similarly.
5749 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
5750 the EVRP range analyzer for range data rather than using global data.
5751 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
5752 gimple-ssa-evrp-analyze.h
5753 (class sprintf_dom_walker): Add after_dom_children member function.
5754 Add evrp_range_analyzer member.
5755 (sprintf_dom_walker::before_dom_children): Call into the EVRP
5756 range analyzer as needed.
5757 (sprintf_dom_walker::after_dom_children): New member function.
5758 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
5760 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5761 (evrp_range_analyzer::pop_to_marker): Likewise.
5763 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
5765 PR tree-optimization/84419
5766 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
5767 with the required type if its current type is compatible but
5770 2018-02-20 Jakub Jelinek <jakub@redhat.com>
5773 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
5774 after vectorization.
5776 2018-02-20 Martin Liska <mliska@suse.cz>
5779 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
5780 possible values if we don't have a hint.
5782 2018-02-20 Martin Liska <mliska@suse.cz>
5786 * final.c (shorten_branches): Build align_tab array with one
5788 * opts.c (finish_options): Add alignment option limit check.
5789 (MAX_CODE_ALIGN): Likewise.
5790 (MAX_CODE_ALIGN_VALUE): Likewise.
5791 * doc/invoke.texi: Document maximum allowed option value for
5792 all -falign-* options.
5794 2018-02-19 Jakub Jelinek <jakub@redhat.com>
5797 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
5798 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
5799 * var-tracking.c (emit_note_insn_var_location): Remove all references
5800 to NOTE_INSN_CALL_ARG_LOCATION.
5801 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
5802 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
5803 Use copy_rtx_if_shared.
5804 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
5805 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
5806 (dwarf2out_var_location): Remove handling of
5807 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
5809 * final.c (final_scan_insn): Remove all references to
5810 NOTE_INSN_CALL_ARG_LOCATION.
5811 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
5812 before dumping final insns.
5813 * except.c (emit_note_eh_region_end): Remove all references to
5814 NOTE_INSN_CALL_ARG_LOCATION.
5815 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
5816 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
5817 * config/arc/arc.c (hwloop_optimize): Likewise.
5818 * config/arm/arm.c (create_fix_barrier): Likewise.
5819 * config/s390/s390.c (s390_chunkify_start): Likewise.
5820 * config/sh/sh.c (find_barrier): Likewise.
5821 * config/i386/i386.c (rest_of_insert_endbranch,
5822 ix86_seh_fixup_eh_fallthru): Likewise.
5823 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
5824 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
5825 * config/frv/frv.c (frv_function_prologue): Likewise.
5826 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
5828 (note_outside_basic_block_p): Remove all references to
5829 NOTE_INSN_CALL_ARG_LOCATION.
5830 * gengtype.c (adjust_field_rtx_def): Likewise.
5831 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
5833 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
5834 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
5837 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
5840 PR tree-optimization/84452
5841 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
5842 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
5845 2018-02-19 Martin Liska <mliska@suse.cz>
5848 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
5850 2018-02-19 Martin Liska <mliska@suse.cz>
5851 Richard Sandiford <richard.sandiford@linaro.org>
5853 PR tree-optimization/82491
5854 * gimple-fold.c (get_base_constructor): Make earlier bail out
5857 2018-02-19 Carl Love <cel@us.ibm.com>
5859 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
5860 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
5862 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
5863 P8V_BUILTIN_VEC_NEG.
5865 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
5867 * config/rl78/rl78.md (movdf): New define expand.
5869 2018-02-19 Martin Liska <mliska@suse.cz>
5872 * doc/invoke.texi: Fix typo.
5873 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
5875 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
5877 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
5878 handle rs6000_single_float and rs6000_double_float specially for
5881 2018-02-16 Jeff Law <law@redhat.com>
5883 * config/rx/rx.c (add_pop_cfi_notes): New function.;
5886 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5889 * ipa-inline.c (inline_small_functions): Fix a typo.
5891 2018-02-16 Nathan Sidwell <nathan@acm.org>
5893 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
5895 2018-02-16 Carl Love <cel@us.ibm.com>
5897 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
5898 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
5899 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
5900 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
5901 expansion to P8V_BUILTIN_VEC_FLOAT2.
5903 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
5905 PR rtl-optimization/70023
5906 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
5907 src_regno into account.
5909 2018-02-16 Carl Love <cel@us.ibm.com>
5911 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
5912 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
5913 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
5914 * config/rs6000/rs6000.c: Remove case statements for
5915 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
5916 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
5917 and P9V_BUILTIN_VEC_VINSERT4B.
5918 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
5919 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
5920 * config/rs6000/vsx.md:
5921 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
5924 2018-02-16 Carl Love <cel@us.ibm.com>
5926 * config/rs6000/altivec.h: Add builtin names vec_extract4b
5928 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
5930 * config/rs6000/rs6000-c.c: Add the definitions for
5931 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
5932 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
5933 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
5934 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
5935 definition for insert4b and define insn *insert3b_internal.
5936 * doc/extend.texi: Add documentation for vec_extract4b.
5938 2018-02-16 Nathan Sidwell <nathan@acm.org>
5940 * doc/extend.texi (Backwards Compatibility): Mention friend
5941 injection. Note for-scope is deprecated.
5942 * doc/invoke.texi (-ffriend-injection): Deprecate.
5944 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
5946 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
5947 that moved to I2, also allow destinations that are a paradoxical
5948 subreg (instead of a normal reg).
5950 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
5953 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
5956 2018-02-16 Richard Biener <rguenther@suse.de>
5958 PR tree-optimization/84037
5959 PR tree-optimization/84016
5961 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5962 Adjust vec_construct for the fact we need additional higher latency
5963 128bit inserts for AVX256 and AVX512 vector builds.
5964 (ix86_add_stmt_cost): Scale vector construction cost for
5967 2018-02-16 Richard Biener <rguenther@suse.de>
5969 PR tree-optimization/84417
5970 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
5971 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
5972 (non_rewritable_lvalue_p): Likewise, use poly-ints.
5974 2018-02-16 Martin Liska <mliska@suse.cz>
5977 * internal-fn.def (ASAN_CHECK): Set proper flags.
5978 (ASAN_MARK): Likewise.
5980 2018-02-16 Julia Koval <julia.koval@intel.com>
5982 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
5983 from PTA_CANNONLAKE.
5985 2018-02-16 Jakub Jelinek <jakub@redhat.com>
5988 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
5989 Use ++iter rather than iter++ for std::list iterators.
5990 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
5991 defer deleting them until all nodes in the forest are processed. Do
5992 free even leaf nodes. Change to_process into auto_vec.
5995 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
5996 * vec.h (vec_default_construct): Use memset instead of placement new
5997 if BROKEN_VALUE_INITIALIZATION is defined.
5998 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
5999 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
6002 PR rtl-optimization/83723
6003 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
6004 * lra.c (lra_substitute_pseudo): Likewise. If true, use
6005 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
6007 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
6009 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
6011 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
6013 PR rtl-optimization/81443
6014 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
6015 from inner REGs to paradoxical SUBREGs.
6017 2018-02-16 Richard Biener <rguenther@suse.de>
6019 PR tree-optimization/84399
6020 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
6021 For operands we can analyze at their definition make sure we can
6022 analyze them at each use as well.
6024 2018-02-16 Richard Biener <rguenther@suse.de>
6026 PR tree-optimization/84190
6027 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
6028 volatile accesses if the decl isn't volatile.
6030 2018-02-15 Jason Merrill <jason@redhat.com>
6032 PR c++/84314 - ICE with templates and fastcall attribute.
6033 * attribs.c (build_type_attribute_qual_variant): Don't clobber
6034 TYPE_CANONICAL on an existing type.
6036 2018-02-15 Jakub Jelinek <jakub@redhat.com>
6038 PR tree-optimization/84383
6039 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
6040 dstoff nor call operand_equal_p if dstbase is NULL.
6042 PR tree-optimization/84334
6043 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
6044 also a CONSTANT_CLASS_P, punt.
6046 2018-02-14 Jim Wilson <jimw@sifive.com>
6048 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
6049 first SMALL_OPERAND check. New local min_second_step. Move assert
6050 to where locals are set. Add TARGET_RVC support.
6051 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
6053 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
6055 * doc/invoke.texi: Correct -Wformat-overflow code sample.
6057 2018-02-14 Martin Sebor <msebor@redhat.com>
6059 PR tree-optimization/83698
6060 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
6061 arrays constrain the offset range to their bounds.
6062 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
6063 (builtin_access::overlap): Avoid setting the size of overlap if it's
6065 (maybe_diag_overlap): Also consider arrays when deciding what values
6066 of offsets to include in diagnostics.
6068 2018-02-14 Martin Sebor <msebor@redhat.com>
6071 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
6072 that correspond to the kind of a declaration.
6074 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
6077 * config/pa/pa.md: Load address of PIC label using the linkage table
6078 if the label is nonlocal.
6080 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
6082 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
6083 warning message if user requests -maltivec=be.
6084 * doc/invoke.texi: Document deprecation of -maltivec=be.
6086 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
6089 * config/rs6000/rs6000-c.c: Update definitions for
6090 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
6091 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
6093 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6096 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
6097 add _get_ssp intrinsics. Remove argument from
6098 __builtin_ia32_rdssp[d|q].
6099 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
6100 * config/i386/i386-builtin.def: Remove argument from
6101 __builtin_ia32_rdssp[d|q].
6102 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
6103 ix86_expand_special_args_builtin for _rdssp[d|q].
6104 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
6105 Clear register before usage.
6106 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
6107 Add documentation for new _get_ssp and _inc_ssp intrinsics.
6109 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
6111 PR tree-optimization/84357
6112 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
6113 operand 1 of an ARRAY_REF too.
6115 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
6118 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
6119 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6121 (set_of_reg): New struct.
6122 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
6123 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
6124 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
6126 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
6127 Split into bitclr, bitset, bitinvert patterns if appropriate.
6128 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
6129 use rx_fuse_in_memory_bitop.
6130 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
6131 to named insn, correct maximum insn length.
6133 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
6136 * machmode.def: Define a complex mode for PARTIAL_INT.
6137 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
6139 * doc/rtl.texi: Document CSPImode.
6140 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
6142 (msp430_hard_regno_nregs_with_padding): Likewise.
6144 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
6147 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
6149 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
6151 PR rtl-optimization/84169
6152 * combine.c (try_combine): New variable split_i2i3. Set it to true if
6153 we generated a parallel as new i3 and we split that to new i2 and i3
6154 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
6155 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
6156 those to i2, not i1. Partially rewrite this scan code.
6158 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6161 * stor-layout.c (place_field): For variable length fields, adjust
6162 offset_align afterwards not just based on the field's alignment,
6163 but also on the size.
6166 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
6167 of exps and logs in the use_exp2 case.
6169 2018-02-13 Jeff Law <law@redhat.com>
6171 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
6174 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
6177 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
6181 * common.opt (gas-loc-support, gas-locview-support): New.
6182 (ginline-points, ginternal-reset-location-views): New.
6183 * doc/invoke.texi: Document them. Use @itemx where intended.
6184 (gvariable-location-views): Adjust.
6185 * target.def (reset_location_view): New.
6186 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
6187 (TARGET_RESET_LOCATION_VIEW): New.
6188 * doc/tm.texi: Rebuilt.
6189 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
6190 (dwarf2out_default_as_locview_support): New.
6191 (output_asm_line_debug_info): Use option variables.
6192 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
6193 (output_loc_list): Likewise.
6194 (add_high_low_attributes): Check option variables.
6195 Don't output entry view attribute in strict mode.
6196 (gen_inlined_subroutine_die): Check option variables.
6197 (dwarf2out_inline_entry): Likewise.
6198 (init_sections_and_labels): Likewise.
6199 (dwarf2out_early_finish): Likewise.
6200 (maybe_reset_location_view): New, from...
6201 (dwarf2out_var_location): ... here. Call it.
6202 * debug.h (dwarf2out_default_as_loc_support): Declare.
6203 (dwarf2out_default_as_locview_support): Declare.
6204 * hooks.c (hook_int_rtx_insn_0): New.
6205 * hooks.h (hook_int_rtx_insn_0): Declare.
6206 * toplev.c (process_options): Take -gas-loc-support and
6207 -gas-locview-support from dwarf2out. Enable
6208 -gvariable-location-views by default only with locview
6209 assembler support. Enable -ginternal-reset-location-views by
6210 default only if the target defines the corresponding hook.
6211 Enable -ginline-points by default if location views are
6212 enabled; force it disabled if statement frontiers are
6214 * tree-inline.c (expand_call_inline): Check option variables.
6215 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
6217 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
6219 PR tree-optimization/84321
6220 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
6221 handling. Also check whether the anti-range contains any values
6222 that satisfy the mask; switch to a VR_RANGE if not.
6224 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
6227 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
6229 2018-02-13 Martin Jambor <mjambor@suse.cz>
6232 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
6233 of call statements, also set location of a load to a temporary.
6235 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
6237 * config/rl78/rl78.c (add_vector_labels): New function.
6238 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
6239 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
6240 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
6241 which checks that no arguments are passed.
6242 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
6243 * doc/extend.texi: Documentation for the new attribute.
6245 2018-02-13 Andreas Schwab <schwab@suse.de>
6247 * config/riscv/linux.h (CPP_SPEC): Define.
6249 2018-02-13 Jakub Jelinek <jakub@redhat.com>
6252 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
6253 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
6254 OPTION_MASK_ISA_AES as first argument to def_builtin_const
6255 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
6256 instead of OPTION_MASK_ISA_PCLMUL as first argument to
6257 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
6258 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
6259 temporarily for AES and PCLMUL builtins.
6261 PR tree-optimization/84339
6262 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
6263 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
6267 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
6268 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
6269 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
6271 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
6273 * omp-simd-clone.h: New file.
6274 * omp-simd-clone.c: Include omp-simd-clone.h.
6275 (expand_simd_clones): No longer static.
6276 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
6277 cgraph.h and omp-simd-clone.h.
6278 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
6279 (vect_recog_widen_shift_pattern): Formatting fix.
6280 (vect_pattern_recog_1): Don't check optab for calls.
6283 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
6284 operands[2] into a REG before using gen_lowpart on it.
6286 2018-02-12 Jeff Law <law@redhat.com>
6289 * config/sh/sh.c (find_barrier): Consider a sibling call
6292 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
6293 successfully back substituting a reg.
6295 2018-02-12 Richard Biener <rguenther@suse.de>
6297 PR tree-optimization/84037
6298 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
6299 parameter, move visited init to caller.
6300 (vect_slp_analyze_operations): Separate cost from validity
6301 check, initialize visited once for all instances.
6302 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
6304 * tree-vect-stmts.c (vect_model_simple_cost): Make early
6306 (vect_model_promotion_demotion_cost): Likewise.
6307 (vectorizable_bswap): Guard cost modeling with !slp_node
6308 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
6310 (vectorizable_call): Likewise.
6311 (vectorizable_conversion): Likewise.
6312 (vectorizable_assignment): Likewise.
6313 (vectorizable_shift): Likewise.
6314 (vectorizable_operation): Likewise.
6315 (vectorizable_store): Likewise.
6316 (vectorizable_load): Likewise.
6317 (vectorizable_condition): Likewise.
6318 (vectorizable_comparison): Likewise.
6320 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
6323 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
6324 (ASAN_MARK): Fix fnspec to account for return value, change pointer
6325 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
6327 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6330 * params.def (inline-min-speedup): Increase from 8 to 15.
6331 (max-inline-insns-auto): Decrease from 40 to 30.
6332 * ipa-split.c (consider_split): Add some buffer for function to
6333 be considered inlining candidate.
6334 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
6337 2018-02-12 Richard Biener <rguenther@suse.de>
6339 PR tree-optimization/84037
6340 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
6341 matched stmts if we cannot swap the non-matched ones.
6343 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6345 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
6346 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
6347 _mm_maskz_scalef_round_ss): New intrinsics.
6348 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
6349 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
6350 __builtin_ia32_scalefss_round): Remove.
6351 (__builtin_ia32_scalefsd_mask_round,
6352 __builtin_ia32_scalefss_mask_round): New intrinsics.
6353 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
6354 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
6355 ((match_operand:VF_128 2 "<round_nimm_predicate>"
6356 "<round_constraint>")): Changed to ...
6357 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
6358 "<round_scalar_constraint>")): ... this.
6359 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
6360 %0, %1, %2<round_op3>}"): Changed to ...
6361 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
6362 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
6363 %2<round_scalar_mask_op3>}"): ... this.
6364 * config/i386/subst.md (round_scalar_nimm_predicate): New.
6366 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
6368 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
6369 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
6370 (_mm_maskz_sqrt_round_ss): New intrinsics.
6371 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
6372 (__builtin_ia32_sqrtsd_mask_round)
6373 (__builtin_ia32_sqrtss_mask_round): New builtins.
6374 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
6375 (__builtin_ia32_sqrtss_round): Remove.
6376 (__builtin_ia32_sqrtsd_mask_round)
6377 (__builtin_ia32_sqrtss_mask_round): New builtins.
6378 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
6379 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
6380 ((match_operand:VF_128 1 "vector_operand"
6381 "xBm,<round_constraint>")): Changed to ...
6382 ((match_operand:VF_128 1 "vector_operand"
6383 "xBm,<round_scalar_constraint>")): ... this.
6384 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
6385 %0, %2, %<iptr>1<round_op3>}): Changed to ...
6386 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
6387 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
6388 %<iptr>1<round_scalar_mask_op3>}): ... this.
6389 ((set_attr "prefix" "<round_prefix>")): Changed to ...
6390 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
6392 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
6395 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
6396 Cast vec_cmpeq result to correct type.
6397 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
6398 Cast vec_cmpgt result to correct type.
6400 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
6402 * final.c (final_scan_insn_1): Renamed from...
6403 (final_scan_insn): ... this. New wrapper, to recover
6404 seen from the outermost call in recursive ones.
6405 * config/sparc/sparc.c (output_return): Drop seen from call.
6406 (output_sibcall): Likewise.
6407 * config/visium/visium.c (output_branch): Likewise.
6409 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
6411 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
6414 2018-02-10 Alan Modra <amodra@gmail.com>
6417 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
6418 Specify LR as an input.
6420 2018-02-10 Jakub Jelinek <jakub@redhat.com>
6423 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
6424 remove_member_access_dummy_vars): New functions.
6425 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
6426 lower_omp_1, execute_lower_omp): Use them.
6428 PR rtl-optimization/84308
6429 * shrink-wrap.c (spread_components): Release todo vector.
6431 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
6433 PR rtl-optimization/57193
6434 * ira-color.c (struct allocno_color_data): Add member
6435 conflict_allocno_hard_prefs.
6436 (update_conflict_allocno_hard_prefs): New.
6437 (bucket_allocno_compare_func): Add a preference based on
6438 conflict_allocno_hard_prefs.
6439 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
6440 (color_allocnos): Remove a dead code. Initiate
6441 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
6443 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6446 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
6447 constraint from =wa to wa. Avoid a subreg on the output operand,
6448 instead use a pseudo and subreg it in a move.
6449 (p9_xxbrd_<mode>): Changed to ...
6450 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
6451 (p9_xxbrd_v2df): New expander.
6452 (p9_xxbrw_<mode>): Changed to ...
6453 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
6454 (p9_xxbrw_v4sf): New expander.
6456 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6458 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
6460 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
6463 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
6464 multiply in 32-bit mode.
6465 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
6466 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
6469 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
6471 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
6472 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
6473 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
6474 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
6476 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
6479 * dwarf2out.c (is_trivial_indirect_ref): New function.
6480 (dwarf2out_late_global_decl): Do not generate a location
6481 attribute for variables that have a non-trivial DECL_VALUE_EXPR
6482 and that are not defined in the current unit.
6484 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6486 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
6487 instead of a libcall for UNORDERED.
6489 2018-02-09 Tamar Christina <tamar.christina@arm.com>
6492 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
6493 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
6495 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6498 * config/s390/s390.c (s390_set_current_function): Invoke
6499 s390_indirect_branch_settings also if fndecl didn't change.
6501 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6503 * config/rs6000/rs6000.md (blockage): Set length to zero.
6505 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
6507 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
6509 2018-02-09 Jakub Jelinek <jakub@redhat.com>
6512 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
6513 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
6517 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
6518 PARALLEL incoming that failed vt_get_decl_and_offset check.
6521 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
6522 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
6524 (get_variable_section): For decls in named .bss* sections pass true as
6525 second argument to bss_initializer_p.
6527 2018-02-09 Marek Polacek <polacek@redhat.com>
6528 Jakub Jelinek <jakub@redhat.com>
6531 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
6532 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
6533 Sync some changes from cxx_fold_indirect_ref.
6535 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
6537 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
6539 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
6540 (BLOCK_INLINE_ENTRY_LABEL): New.
6541 (dwarf2out_var_location): Disregard inline entry markers.
6542 (inline_entry_data): New struct.
6543 (inline_entry_data_hasher): New hashtable type.
6544 (inline_entry_data_hasher::hash): New.
6545 (inline_entry_data_hasher::equal): New.
6546 (inline_entry_data_table): New variable.
6547 (add_high_low_attributes): Add DW_AT_entry_pc and
6548 DW_AT_GNU_entry_view attributes if a pending entry is found
6549 in inline_entry_data_table. Add old entry_pc attribute only
6550 if debug nonbinding markers are disabled.
6551 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
6552 markers are enabled.
6553 (block_within_block_p, dwarf2out_inline_entry): New.
6554 (dwarf2out_finish): Check that no entries remained in
6555 inline_entry_data_table.
6556 * final.c (reemit_insn_block_notes): Handle inline entry notes.
6557 (final_scan_insn, notice_source_line): Likewise.
6558 (rest_of_clean_state): Skip inline entry markers.
6559 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
6561 * gimple.c (gimple_build_debug_inline_entry): New.
6562 * gimple.h (enum gimple_debug_subcode): Add
6563 GIMPLE_DEBUG_INLINE_ENTRY.
6564 (gimple_build_debug_inline_entry): Declare.
6565 (gimple_debug_inline_entry_p): New.
6566 (gimple_debug_nonbind_marker_p): Adjust.
6567 * insn-notes.def (INLINE_ENTRY): New.
6568 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
6569 inline entry marker notes.
6570 (print_insn): Likewise.
6571 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
6572 (INSN_DEBUG_MARKER_KIND): Likewise.
6573 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
6574 * tree-inline.c (expand_call_inline): Build and insert
6575 debug_inline_entry stmt.
6576 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
6577 inline entry blocks early, if nonbind markers are enabled.
6578 (dump_scope_block): Dump fragment info.
6579 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
6580 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
6581 (gimple_build_debug_inline_entry): New.
6582 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
6583 Enable/disable inline entry points too.
6584 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
6585 (DEBUG_INSN): Describe inline entry markers.
6587 * common.opt (gvariable-location-views): New.
6588 (gvariable-location-views=incompat5): New.
6589 * config.in: Rebuilt.
6590 * configure: Rebuilt.
6591 * configure.ac: Test assembler for view support.
6592 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
6593 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
6594 * dwarf2out.c (var_loc_view): New typedef.
6595 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
6596 (dwarf2out_locviews_in_attribute): New.
6597 (dwarf2out_locviews_in_loclist): New.
6598 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
6599 (enum dw_line_info_opcode): Add LI_adv_address.
6600 (struct dw_line_info_table): Add view.
6601 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
6602 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
6603 (zero_view_p): New variable.
6604 (ZERO_VIEW_P): New macro.
6605 (output_asm_line_debug_info): New.
6606 (struct var_loc_node): Add view.
6607 (add_AT_view_list, AT_loc_list): New.
6608 (add_var_loc_to_decl): Add view param. Test it against last.
6609 (new_loc_list): Add view params. Record them.
6610 (AT_loc_list_ptr): Handle loc and view lists.
6611 (view_list_to_loc_list_val_node): New.
6612 (print_dw_val): Handle dw_val_class_view_list.
6613 (size_of_die): Likewise.
6614 (value_format): Likewise.
6615 (loc_list_has_views): New.
6616 (gen_llsym): Set vl_symbol too.
6617 (maybe_gen_llsym, skip_loc_list_entry): New.
6618 (dwarf2out_maybe_output_loclist_view_pair): New.
6619 (output_loc_list): Output view list or entries too.
6620 (output_view_list_offset): New.
6621 (output_die): Handle dw_val_class_view_list.
6622 (output_dwarf_version): New.
6623 (output_compilation_unit_header): Use it.
6624 (output_skeleton_debug_sections): Likewise.
6625 (output_rnglists, output_line_info): Likewise.
6626 (output_pubnames, output_aranges): Update version comments.
6627 (output_one_line_info_table): Output view numbers in asm comments.
6628 (dw_loc_list): Determine current endview, pass it to new_loc_list.
6629 Call maybe_gen_llsym.
6630 (loc_list_from_tree_1): Adjust.
6631 (add_AT_location_description): Create view list attribute if
6632 needed, check it's absent otherwise.
6633 (convert_cfa_to_fb_loc_list): Adjust.
6634 (maybe_emit_file): Call output_asm_line_debug_info for test.
6635 (dwarf2out_var_location): Reset views as needed. Precompute
6636 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
6637 attribute. Set view.
6638 (new_line_info_table): Reset next view.
6639 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
6640 (dwarf2out_source_line): Likewise. Output view resets and labels to
6641 the assembler, or select appropriate line info opcodes.
6642 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
6643 (optimize_string_length): Catch it. Adjust.
6644 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
6645 dw_val_class_view_list, and remove it if no longer needed.
6646 (hash_loc_list): Hash view numbers.
6647 (loc_list_hasher::equal): Compare them.
6648 (optimize_location_lists): Check whether a view list symbol is
6649 needed, and whether the locview attribute is present, and
6650 whether they match. Remove the locview attribute if no longer
6652 (index_location_lists): Call skip_loc_list_entry for test.
6653 (dwarf2out_finish): Call output_asm_line_debug_info for test.
6654 Use output_dwarf_version.
6655 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
6656 (struct dw_val_node): Add val_view_list.
6657 * final.c (SEEN_NEXT_VIEW): New.
6658 (set_next_view_needed): New.
6659 (clear_next_view_needed): New.
6660 (maybe_output_next_view): New.
6661 (final_start_function): Rename to...
6662 (final_start_function_1): ... this. Take pointer to FIRST,
6663 add SEEN parameter. Emit param bindings in the initial view.
6664 (final_start_function): Reintroduce SEEN-less interface.
6665 (final): Rename to...
6666 (final_1): ... this. Take SEEN parameter. Output final pending
6667 next view at the end.
6668 (final): Reintroduce seen-less interface.
6669 (final_scan_insn): Output pending next view before switching
6670 sections or ending a block. Mark the next view as needed when
6671 outputting variable locations. Notify debug backend of section
6672 changes, and of location view changes.
6673 (rest_of_handle_final): Adjust.
6674 * toplev.c (process_options): Autodetect value for debug variable
6675 location views option. Warn on incompat5 without -gdwarf-5.
6676 * doc/invoke.texi (gvariable-location-views): New.
6677 (gvariable-location-views=incompat5): New.
6678 (gno-variable-location-views): New.
6680 2018-02-08 David Malcolm <dmalcolm@redhat.com>
6682 PR tree-optimization/84136
6683 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
6684 that the result of find_edge is non-NULL.
6686 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6689 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
6690 storing integer register in SImode. Fix cost of 256 and 512
6691 byte aligned SSE register store.
6693 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
6695 * config/i386/i386.c (ix86_multiplication_cost): Fix
6696 multiplication cost for TARGET_AVX512DQ.
6698 2018-02-08 Marek Polacek <polacek@redhat.com>
6700 PR tree-optimization/84238
6701 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
6704 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6706 PR tree-optimization/84265
6707 * tree-vect-stmts.c (vectorizable_store): Don't treat
6708 VMAT_CONTIGUOUS accesses as grouped.
6709 (vectorizable_load): Likewise.
6711 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6713 PR tree-optimization/81635
6714 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
6715 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
6716 (test_round_for_mask): New functions.
6717 (wide_int_cc_tests): Call test_round_for_mask.
6718 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
6719 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
6720 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
6721 range returned by get_range_info.
6723 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
6726 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
6727 * symtab.c: Include builtins.h
6728 (symtab_node::output_to_lto_symbol_table_p): Move here
6729 from lto-streamer-out.c:output_symbol_p.
6730 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
6731 (output_symbol_p): Move all logic to symtab.c
6732 (produce_symtab): Update.
6734 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6736 * config/s390/s390-opts.h (enum indirect_branch): Define.
6737 * config/s390/s390-protos.h (s390_return_addr_from_memory)
6738 (s390_indirect_branch_via_thunk)
6739 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
6740 (enum s390_indirect_branch_type): Define.
6741 * config/s390/s390.c (struct s390_frame_layout, struct
6742 machine_function): Remove.
6743 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
6744 (indirect_branch_table_label_no, indirect_branch_table_name):
6746 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
6747 (enum s390_indirect_branch_option): Define.
6748 (s390_return_addr_from_memory): New function.
6749 (s390_handle_string_attribute): New function.
6750 (s390_attribute_table): Add new attribute handler.
6751 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
6752 (s390_indirect_branch_via_thunk): New function.
6753 (s390_indirect_branch_via_inline_thunk): New function.
6754 (s390_function_ok_for_sibcall): When jumping via thunk disallow
6755 sibling call optimization for non z10 compiles.
6756 (s390_emit_call): Force indirect branch target to be a single
6757 register. Add r1 clobber for non-z10 compiles.
6758 (s390_emit_epilogue): Emit return jump via return_use expander.
6759 (s390_reorg): Handle JUMP_INSNs as execute targets.
6760 (s390_option_override_internal): Perform validity checks for the
6761 new command line options.
6762 (s390_indirect_branch_attrvalue): New function.
6763 (s390_indirect_branch_settings): New function.
6764 (s390_set_current_function): Invoke s390_indirect_branch_settings.
6765 (s390_output_indirect_thunk_function): New function.
6766 (s390_code_end): Implement target hook.
6767 (s390_case_values_threshold): Implement target hook.
6768 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
6770 * config/s390/s390.h (struct s390_frame_layout)
6771 (struct machine_function): Move here from s390.c.
6772 (TARGET_INDIRECT_BRANCH_NOBP_RET)
6773 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
6774 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
6775 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
6776 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
6777 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
6778 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
6779 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
6780 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
6781 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
6782 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
6783 (mnemonic attribute): Add values which aren't recognized
6785 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
6786 pattern for branch conversion. Fix mnemonic attribute.
6787 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
6788 indirect branch via thunk if requested.
6789 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
6790 ("*indirect_jump"): Disable for branch conversion using out of
6792 ("indirect_jump_via_thunk<mode>_z10")
6793 ("indirect_jump_via_thunk<mode>")
6794 ("indirect_jump_via_inlinethunk<mode>_z10")
6795 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
6796 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
6797 ("casesi_jump_via_inlinethunk<mode>_z10")
6798 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
6799 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
6800 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
6801 ("*indirect2_jump"): Disable for branch conversion.
6802 ("casesi_jump"): Turn into expander and expand patterns for branch
6804 ("return_use"): New expander.
6805 ("*return"): Emit return via thunk and rename it to ...
6806 ("*return<mode>"): ... this one.
6807 * config/s390/s390.opt: Add new options and and enum for the
6810 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
6812 * lra-constraints.c (match_reload): Unconditionally use
6813 gen_lowpart_SUBREG, rather than selecting between that
6814 and equivalent gen_rtx_SUBREG code.
6816 2018-02-08 Richard Biener <rguenther@suse.de>
6818 PR tree-optimization/84233
6819 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
6820 changed flag instead of boguously re-using phi_inserted.
6822 2018-02-08 Martin Jambor <mjambor@suse.cz>
6824 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
6825 static local variables.
6827 2018-02-08 Richard Biener <rguenther@suse.de>
6829 PR tree-optimization/84278
6830 * tree-vect-stmts.c (vectorizable_store): When looking for
6831 smaller vector types to perform grouped strided loads/stores
6832 make sure the mode is supported by the target.
6833 (vectorizable_load): Likewise.
6835 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6837 * config/aarch64/aarch64.c (aarch64_components_for_bb):
6838 Increase LDP/STP opportunities by adding adjacent callee-saves.
6840 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
6842 PR rtl-optimization/84068
6843 PR rtl-optimization/83459
6844 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
6846 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
6848 PR tree-optimization/84224
6849 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
6850 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
6853 2018-02-07 Iain Sandoe <iain@codesourcery.com>
6856 * config/rs6000/altivec.md (*restore_world): Remove LR use.
6857 * config/rs6000/predicates.md (restore_world_operation): Adjust op
6858 count, remove one USE.
6860 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
6862 * doc/install.texi (Configuration): Document the
6863 --with-long-double-format={ibm,ieee} PowerPC configuration
6867 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
6868 Convert from define_expand to be define_insn_and_split. Rework
6869 float/double/_Float128 conversions to QI/HI/SImode to work with
6870 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
6871 conversions to QI/HImode types did a store and then a load to
6872 truncate the value. For conversions to VSX registers, don't split
6873 the insn, instead emit the code directly. Use the code iterator
6874 any_fix to combine signed and unsigned conversions.
6875 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
6876 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
6877 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6878 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
6879 (fix_<mode>di2_hw): Likewise.
6880 (fixuns_<mode>di2_hw): Likewise.
6881 (fix_<mode>si2_hw): Likewise.
6882 (fixuns_<mode>si2_hw): Likewise.
6883 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
6884 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
6885 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
6886 fix<uns>_trunc<SFDF:mode>si2_p8.
6887 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
6889 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
6890 (fix<uns>_<mode>_mem): Likewise.
6891 (fctiw<u>z_<mode>_mem): Likewise.
6892 (fix<uns>_<mode>_mem): Likewise.
6893 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
6894 the register allocator from doing a direct move to the GPRs to do
6895 a store, and instead use the ISA 3.0 store byte/half-word from
6896 vector register instruction. For IEEE 128-bit floating point,
6897 also optimize stores of 32-bit ints.
6898 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
6900 2018-02-07 Alan Hayward <alan.hayward@arm.com>
6902 * genextract.c (push_pathstr_operand): New function to support
6904 (walk_rtx): Call push_pathstr_operand.
6905 (print_path): Support [a-zA-Z].
6907 2018-02-07 Richard Biener <rguenther@suse.de>
6909 PR tree-optimization/84037
6910 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
6911 (cse_and_gimplify_to_preheader): Declare.
6912 (vect_get_place_in_interleaving_chain): Likewise.
6913 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6915 (_loop_vec_info::~_loop_vec_info): Delete it.
6916 (cse_and_gimplify_to_preheader): New function.
6917 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
6918 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
6919 (vectorizable_load): Likewise. For grouped stores always base
6920 the IV on the first element.
6921 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
6922 condition before gimplifying.
6924 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6926 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
6927 *DIV_EXPR and *MOD_EXPR.
6929 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
6932 * config/i386/i386.c (ix86_option_override_internal): Mask out
6933 the CF_SET bit when checking -fcf-protection.
6935 2018-02-07 Tom de Vries <tom@codesourcery.com>
6938 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
6941 2018-02-07 Richard Biener <rguenther@suse.de>
6943 PR tree-optimization/84204
6944 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
6947 PR tree-optimization/84205
6948 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
6949 special-case isl_ast_op_zdiv_r.
6951 PR tree-optimization/84223
6952 * graphite-scop-detection.c (gather_bbs::before_dom_children):
6953 Only add conditions from within the region.
6954 (gather_bbs::after_dom_children): Adjust.
6956 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
6959 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
6960 * config/avr/avr.md: Only post-reload split REG-REG moves if
6961 either register is GENERAL_REG_P.
6963 2018-02-07 Jakub Jelinek <jakub@redhat.com>
6965 PR tree-optimization/84235
6966 * tree-ssa-scopedtables.c
6967 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
6968 if the subtraction is performed in floating point type where NaNs are
6969 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
6970 build 1. Formatting fix.
6972 2018-02-06 Jakub Jelinek <jakub@redhat.com>
6975 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
6976 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
6977 and skip it regardless of bb boundaries. Use CALL_P macro,
6978 don't test INSN_P (insn) together with CALL_P or JUMP_P check
6979 unnecessarily, formatting fix.
6981 2018-02-06 Michael Collison <michael.collison@arm.com>
6983 * config/arm/thumb2.md:
6984 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
6985 (*thumb_mov_notscc): Ditto.
6987 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
6990 * config/rs6000/rs6000.md (su code attribute): Use "u" for
6991 unsigned_fix, not "s".
6993 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6995 * configure.ac (gcc_fn_eh_frame_ro): New function.
6996 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
6997 correct .eh_frame permissions.
6998 * configure: Regenerate.
7000 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
7002 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
7005 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7007 * config/rs6000/rs6000.c (rs6000_option_override_internal):
7008 Display warning message for -mno-speculate-indirect-jumps.
7010 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
7012 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
7014 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
7016 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
7018 PR tree-optimization/84225
7019 * tree-eh.c (find_trapping_overflow): Only call
7020 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
7022 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7025 * config/i386/i386.c: Reimplement the check of possible options
7026 -mibt/-mshstk conbination. Change error messages.
7027 * doc/invoke.texi: Fix a typo: remove extra '='.
7029 2018-02-06 Marek Polacek <polacek@redhat.com>
7031 PR tree-optimization/84228
7032 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
7034 2018-02-06 Tamar Christina <tamar.christina@arm.com>
7037 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
7038 emitted arch directives.
7039 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
7040 __ARM_FEATURE_COPROC before changing architectures.
7042 2018-02-06 Richard Biener <rguenther@suse.de>
7044 * config/i386/i386.c (print_reg): Fix typo.
7045 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
7047 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
7049 * configure: Regenerate.
7051 2018-02-05 Martin Sebor <msebor@redhat.com>
7053 PR tree-optimization/83369
7054 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
7057 2018-02-05 Martin Liska <mliska@suse.cz>
7059 * doc/invoke.texi: Cherry-pick upstream r323995.
7061 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
7063 * ira.c (ira_init_register_move_cost): Adjust comment.
7065 2018-02-05 Martin Liska <mliska@suse.cz>
7067 PR gcov-profile/84137
7068 * doc/gcov.texi: Fix typo in documentation.
7070 2018-02-05 Martin Liska <mliska@suse.cz>
7072 PR gcov-profile/83879
7073 * doc/gcov.texi: Document necessity of --dynamic-list-data when
7074 using dlopen functionality.
7076 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
7078 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
7079 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
7080 _mm_maskz_range_ss, _mm_mask_range_round_ss,
7081 _mm_maskz_range_round_ss): New intrinsics.
7082 (__builtin_ia32_rangesd128_round)
7083 (__builtin_ia32_rangess128_round): Remove.
7084 (__builtin_ia32_rangesd128_mask_round,
7085 __builtin_ia32_rangess128_mask_round): New builtins.
7086 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
7087 __builtin_ia32_rangess128_round): Remove.
7088 (__builtin_ia32_rangesd128_mask_round,
7089 __builtin_ia32_rangess128_mask_round): New builtins.
7090 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
7091 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
7092 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
7093 "<round_saeonly_constraint>")): Changed to ...
7094 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
7095 "<round_saeonly_scalar_constraint>")): ... this.
7096 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
7097 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
7098 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
7099 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
7100 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
7102 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
7104 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
7106 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
7107 Remove all values except native, 8540 and 8548.
7109 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
7111 * config/i386/i386.c (ix86_output_function_return): Pass
7112 INVALID_REGNUM, instead of -1, as invalid register number to
7113 indirect_thunk_name and output_indirect_thunk.
7115 2018-02-02 Julia Koval <julia.koval@intel.com>
7117 * config.gcc: Add -march=icelake.
7118 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
7119 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
7120 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
7121 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
7122 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
7123 (processor_target_table): Add icelake.
7124 (ix86_option_override_internal): Handle new PTAs.
7125 (get_builtin_code_for_version): Handle icelake.
7126 (M_INTEL_COREI7_ICELAKE): New.
7127 (fold_builtin_cpu): Handle icelake.
7128 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
7129 * doc/invoke.texi: Add -march=icelake.
7131 2018-02-02 Julia Koval <julia.koval@intel.com>
7133 * config/i386/i386.c (ix86_option_override_internal): Change flags type
7134 to wide_int_bitmask.
7135 * wide-int-bitmask.h: New.
7137 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
7140 * config/i386/i386.md: Replace Pmode with word_mode in
7141 builtin_setjmp_setup and builtin_longjmp to support x32.
7143 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
7147 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
7149 (rs6000_supported_cpu_names): New static variable.
7150 (linux_cpu_translation_table): Likewise.
7151 (elf_platform) <cpu>: Define new static variable and use it.
7152 Translate kernel AT_PLATFORM name to canonical name if needed.
7153 Error if platform name is unknown.
7155 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
7158 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
7160 2018-02-01 Jeff Law <law@redhat.com>
7163 * config/i386/i386.c (release_scratch_register_on_entry): Add new
7164 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
7165 the scratch if RELEASE_VIA_POP is false.
7166 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
7167 If we have to save a temporary register, decrement SIZE appropriately.
7168 Pass new arguments to release_scratch_register_on_entry.
7169 (ix86_adjust_stack_and_probe): Likewise.
7170 (ix86_emit_probe_stack_range): Pass new arguments to
7171 release_scratch_register_on_entry.
7173 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
7175 PR rtl-optimization/84157
7176 * combine.c (change_zero_ext): Use REG_P predicate in
7177 front of HARD_REGISTER_P predicate.
7179 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
7181 * config/avr/avr.c (avr_option_override): Move disabling of
7182 -fdelete-null-pointer-checks to...
7183 * common/config/avr/avr-common.c (avr_option_optimization_table):
7186 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7188 PR tree-optimization/81635
7189 * tree-data-ref.c (split_constant_offset_1): For types that
7190 wrap on overflow, try to use range info to prove that wrapping
7193 2018-02-01 Renlin Li <renlin.li@arm.com>
7196 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
7198 (aarch64_register_move_cost): Likewise.
7199 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
7201 (REG_CLASS_NAMES): Likewise.
7202 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
7203 TAILCALL_ADDR_REGS. Remove IP registers.
7204 * config/aarch64/aarch64.md (Ucs): Update register constraint.
7206 2018-02-01 Richard Biener <rguenther@suse.de>
7208 * domwalk.h (dom_walker::dom_walker): Add additional constructor
7209 for specifying RPO order and allow NULL for that.
7210 * domwalk.c (dom_walker::dom_walker): Likewise.
7211 (dom_walker::walk): Handle NULL RPO order.
7212 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
7214 (rewrite_update_dom_walker): Likewise.
7215 (mark_def_dom_walker): Likewise.
7217 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7219 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
7220 (aarch64_maybe_expand_sve_subreg_move): Declare.
7221 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
7222 * config/aarch64/predicates.md (aarch64_any_register_operand): New
7224 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
7225 that are semantically a reverse operation.
7226 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
7227 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
7228 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
7230 (aarch64_can_change_mode_class): For big-endian, forbid changes
7231 between two SVE modes if they have different element sizes.
7233 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7235 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
7236 the TImode handling for big-endian targets.
7238 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7240 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
7241 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
7243 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
7244 Remove BSWAP handing for big-endian targets and use the form of
7245 LD1RQ appropariate for the mode.
7247 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7249 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
7250 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
7253 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
7256 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
7257 check for operands that need to go through aarch64_sve_reload_be.
7259 2018-02-01 Jakub Jelinek <jakub@redhat.com>
7261 PR tree-optimization/81661
7262 PR tree-optimization/84117
7263 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
7264 * tree-eh.c: Include gimplify.h.
7265 (find_trapping_overflow, replace_trapping_overflow,
7266 rewrite_to_non_trapping_overflow): New functions.
7267 * tree-vect-loop.c: Include tree-eh.h.
7268 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
7269 * tree-data-ref.c: Include tree-eh.h.
7270 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
7272 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
7274 PR rtl-optimization/84123
7275 * combine.c (change_zero_ext): Check if hard register satisfies
7276 can_change_dest_mode before calling gen_lowpart_SUBREG.
7278 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
7281 * ira.c (ira_init_register_move_cost): Remove assert.
7283 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7285 PR rtl-optimization/84071
7286 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
7287 * doc/tm.texi: Regenerate.
7289 2018-01-31 Richard Biener <rguenther@suse.de>
7291 PR tree-optimization/84132
7292 * tree-data-ref.c (analyze_miv_subscript): Properly
7293 check whether evolution_function_is_affine_multivariate_p
7294 before calling gcd_of_steps_may_divide_p.
7296 2018-01-31 Julia Koval <julia.koval@intel.com>
7299 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
7300 * config/i386/i386.md (rdpid_rex64) New.
7301 (rdpid): Make 32bit only.
7303 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
7306 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
7307 an IDENTIFIER_NODE for FUNCTION_TYPE's.
7309 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7312 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7314 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7316 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
7318 PR rtl-optimization/84071
7319 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
7320 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
7322 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7324 * config/arc/arc.c (arc_handle_aux_attribute): New function.
7325 (arc_attribute_table): Add 'aux' attribute.
7326 (arc_in_small_data_p): Consider aux like variables.
7327 (arc_is_aux_reg_p): New function.
7328 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
7329 (arc_get_aux_arg): New function.
7330 (prepare_move_operands): Handle aux-register access.
7331 (arc_handle_aux_attribute): New function.
7332 * doc/extend.texi (ARC Variable attributes): Add subsection.
7334 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
7336 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
7337 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
7338 (arc_attribute_table): Add 'uncached' attribute.
7339 (arc_print_operand): Print '.di' flag for uncached memory
7341 (arc_in_small_data_p): Do not consider for small data the uncached
7343 (arc_is_uncached_mem_p): New function.
7344 * config/arc/predicates.md (compact_store_memory_operand): Check
7345 for uncached memory accesses.
7346 (nonvol_nonimm_operand): Likewise.
7347 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
7349 2018-01-31 Jakub Jelinek <jakub@redhat.com>
7352 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
7353 falign-loops=): Add Optimization flag.
7355 2018-01-30 Jeff Law <law@redhat.com>
7358 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
7359 INT_REGISTERS_SAVED. Check it prior to calling
7360 get_scratch_register_on_entry.
7361 (ix86_adjust_stack_and_probe): Similarly.
7362 (ix86_emit_probe_stack_range): Similarly.
7363 (ix86_expand_prologue): Corresponding changes.
7365 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7368 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
7369 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
7371 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
7374 * lra-constraints.c (curr_insn_transform): Process AND in the
7377 2018-01-30 Jakub Jelinek <jakub@redhat.com>
7379 PR rtl-optimization/83986
7380 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
7381 dependence against last_pending_memory_flush in addition to
7384 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
7386 PR tree-optimization/81611
7387 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
7390 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7393 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
7396 2018-01-30 Richard Biener <rguenther@suse.de>
7397 Jakub Jelinek <jakub@redhat.com>
7399 PR tree-optimization/84111
7400 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
7401 inner loops added during recursion, as they don't have up-to-date
7404 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7407 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
7408 (can_inline_edge_by_limits_p): ... here.
7409 (can_early_inline_edge_p, check_callers,
7410 update_caller_keys, update_callee_keys, recursive_inlining,
7411 add_new_edges_to_heap, speculation_useful_p,
7412 inline_small_functions,
7413 inline_small_functions, flatten_function,
7414 inline_to_all_callers_1): Update.
7416 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
7418 * profile-count.c (profile_count::combine_with_ipa_count): Handle
7421 2018-01-30 Richard Biener <rguenther@suse.de>
7423 PR tree-optimization/83008
7424 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
7425 invariant and constant vector uses in stmts when they need
7428 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7431 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
7432 * configure: Regenerate.
7434 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7436 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
7438 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
7439 Use gen_rtx_REG rather than gen_lowpart.
7441 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7443 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
7444 rather than 0 when creating partial subregs.
7446 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
7448 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
7451 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
7454 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
7455 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
7456 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
7457 flags. This restores the settings used before the 2017-07-24.
7458 Turning off pre increment/decrement/modify allows IVOPTS to
7459 optimize DF/SF loops where the index is an int.
7461 2018-01-29 Richard Biener <rguenther@suse.de>
7462 Kelvin Nilsen <kelvin@gcc.gnu.org>
7465 * tree-vect-stmts.c (vectorizable_call): Don't call
7466 targetm.vectorize_builtin_md_vectorized_function if callee is
7469 2018-01-22 Carl Love <cel@us.ibm.com>
7471 * doc/extend.tex: Fix typo in second arg in
7472 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
7474 2018-01-29 Richard Biener <rguenther@suse.de>
7476 PR tree-optimization/84086
7477 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
7478 (flush_ssaname_freelist): When SSA names were released reset
7479 the SCEV hash table.
7481 2018-01-29 Richard Biener <rguenther@suse.de>
7483 PR tree-optimization/84057
7484 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
7485 removed paths when removing edges.
7487 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
7489 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
7490 -mfunction-return=@var{choice}.
7492 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
7495 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
7497 (layout::print_source_line): Likewise.
7498 (test_get_line_width_without_trailing_whitespace): Add test cases.
7500 2018-01-27 Jakub Jelinek <jakub@redhat.com>
7503 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
7506 2018-01-26 Jim Wilson <jimw@sifive.com>
7508 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
7510 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
7513 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7515 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
7516 and CMP + SUB-immediate -> SUBS.
7518 2018-01-26 Martin Sebor <msebor@redhat.com>
7520 PR tree-optimization/83896
7521 * tree-ssa-strlen.c (get_string_len): Rename...
7522 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
7523 Avoid assuming length is constant.
7524 (handle_char_store): Use HOST_WIDE_INT for string length.
7526 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
7529 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
7530 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
7532 2018-01-26 Richard Biener <rguenther@suse.de>
7534 PR rtl-optimization/84003
7535 * dse.c (record_store): Only record redundant stores when
7536 the earlier store aliases at least all accesses the later one does.
7538 2018-01-26 Jakub Jelinek <jakub@redhat.com>
7540 PR rtl-optimization/83985
7541 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
7542 REG_CFA_RESTORE insns.
7543 (delete_unmarked_insns): Don't ignore separate shrink wrapping
7544 REG_CFA_RESTORE insns here.
7547 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
7548 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
7550 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7552 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
7553 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
7554 (arc_init): Likewise.
7555 (arc_override_options): Likewise.
7556 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
7558 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
7560 * config/arc/arc.h (TARGET_DBNZ): Define.
7561 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
7562 properly set the tune attribute.
7563 (dbnz): Use TARGET_DBNZ guard.
7564 * config/arc/arc.opt (mtune): Add core3 option.
7566 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7568 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
7569 recognize new pic like addresses.
7570 (arc_delegitimize_address): Clean up.
7572 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7574 * config/arc/arc-arches.def: Option mrf16 valid for all
7576 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
7577 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
7578 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
7579 * config/arc/arc-tables.opt: Regenerate.
7580 * config/arc/arc.c (arc_conditional_register_usage): Handle
7581 reduced register file case.
7582 (arc_file_start): Set must have build attributes.
7583 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
7585 * config/arc/arc.opt (mrf16): Add new option.
7586 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
7587 * config/arc/genmultilib.awk: Handle new mrf16 option.
7588 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
7589 * config/arc/t-multilib: Regenerate.
7590 * doc/invoke.texi (ARC Options): Document mrf16 option.
7592 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7594 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
7595 * config/arc/arc.c (arc_handle_secure_attribute): New function.
7596 (arc_attribute_table): Add 'secure_call' attribute.
7597 (arc_print_operand): Print secure call operand.
7598 (arc_function_ok_for_sibcall): Don't optimize tail calls when
7600 (arc_is_secure_call_p): New function. * config/arc/arc.md
7601 (call_i): Add support for sjli instruction.
7602 (call_value_i): Likewise.
7603 * config/arc/constraints.md (Csc): New constraint.
7605 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
7606 John Eric Martin <John.Martin@emmicro-us.com>
7608 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
7609 * config/arc/arc.c (_arc_jli_section): New struct.
7610 (arc_jli_section): New type.
7611 (rc_jli_sections): New static variable.
7612 (arc_handle_jli_attribute): New function.
7613 (arc_attribute_table): Add jli_always and jli_fixed attribute.
7614 (arc_file_end): New function.
7615 (TARGET_ASM_FILE_END): Define.
7616 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
7617 (arc_add_jli_section): New function.
7618 (jli_call_scan): Likewise.
7619 (arc_reorg): Call jli_call_scan.
7620 (arc_output_addsi): Remove 'S' from printing asm operand.
7621 (arc_is_jli_call_p): New function.
7622 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
7624 (movhi_insn): Likewise.
7625 (movsi_insn): Likewise.
7626 (movsi_set_cc_insn): Likewise.
7627 (loadqi_update): Likewise.
7628 (load_zeroextendqisi_update): Likewise.
7629 (load_signextendqisi_update): Likewise.
7630 (loadhi_update): Likewise.
7631 (load_zeroextendhisi_update): Likewise.
7632 (load_signextendhisi_update): Likewise.
7633 (loadsi_update): Likewise.
7634 (loadsf_update): Likewise.
7635 (movsicc_insn): Likewise.
7636 (bset_insn): Likewise.
7637 (bxor_insn): Likewise.
7638 (bclr_insn): Likewise.
7639 (bmsk_insn): Likewise.
7640 (bicsi3_insn): Likewise.
7641 (cmpsi_cc_c_insn): Likewise.
7642 (movsi_ne): Likewise.
7643 (movsi_cond_exec): Likewise.
7644 (clrsbsi2): Likewise.
7655 (call_i): Remove 'S' asm letter, add jli instruction.
7656 (call_value_i): Likewise.
7657 * config/arc/arc.op (mjli-always): New option.
7658 * config/arc/constraints.md (Cji): New constraint.
7659 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
7661 (subsf3_fpx): Likewise.
7662 (mulsf3_fpx): Likewise.
7663 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
7665 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
7667 * doc/invoke.texi (ARC): Document mjli-always option.
7669 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
7671 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
7672 avoid addition with 0 and use incw and decw where possible.
7674 2018-01-26 Richard Biener <rguenther@suse.de>
7676 PR tree-optimization/81082
7677 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
7678 association if it requires casting to unsigned.
7679 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
7680 from fold_plusminus_mult_expr to catch important cases late when
7681 range info is available.
7683 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7685 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
7686 * configure.ac (hidden_linkonce): New test.
7687 * configure: Regenerate.
7688 * config.in: Regenerate.
7690 2018-01-26 Julia Koval <julia.koval@intel.com>
7692 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
7693 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
7694 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
7695 _mm_mask_bitshuffle_epi64_mask): Fix type.
7696 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
7697 USI_FTYPE_V4DI_V4DI_USI): Remove.
7698 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
7699 __builtin_ia32_vpshufbitqmb256_mask,
7700 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
7701 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
7702 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
7704 2018-01-26 Alan Modra <amodra@gmail.com>
7707 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
7708 UNSPEC_VBPERMQ. Sort other unspecs.
7710 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
7712 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
7714 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
7717 * predict.c (drop_profile): Do not push/pop cfun; update also
7719 (handle_missing_profiles): Fix logic looking for zero profiles.
7721 2018-01-25 Jakub Jelinek <jakub@redhat.com>
7724 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
7725 on functions with #pragma omp declare simd or functions with simd
7727 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
7728 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
7729 Remove trailing \n from warning_at calls.
7731 2018-01-25 Tom de Vries <tom@codesourcery.com>
7734 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7735 for neutered workers.
7737 2018-01-24 Joseph Myers <joseph@codesourcery.com>
7740 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
7741 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
7743 2018-01-24 Jeff Law <law@redhat.com>
7746 * i386.c (get_probe_interval): Move to earlier point.
7747 (ix86_compute_frame_layout): If -fstack-clash-protection and
7748 the frame is larger than the probe interval, then use pushes
7749 to save registers rather than reg->mem moves.
7750 (ix86_expand_prologue): Remove conditional for int_registers_saved
7753 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
7756 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
7757 min/max for never referenced object.
7759 2018-01-24 Jakub Jelinek <jakub@redhat.com>
7762 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
7764 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
7765 attributes from DECL_ATTRIBUTES (decl) without affecting
7766 DECL_ATTRIBUTES (current_function_decl).
7767 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
7768 functions with non-NULL DECL_ABSTRACT_ORIGIN.
7770 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
7772 PR tree-optimization/83979
7773 * fold-const.c (fold_comparison): Use constant_boolean_node
7774 instead of boolean_{true,false}_node.
7776 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
7778 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
7781 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7783 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7784 Simplify the clause that sets the length attribute.
7785 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7786 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
7787 clause that sets the length attribute.
7788 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7790 2018-01-24 Tom de Vries <tom@codesourcery.com>
7793 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
7794 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
7795 Add strict parameter.
7796 (prevent_branch_around_nothing): Insert dummy insn between branch to
7797 label and label with no ptx insn inbetween.
7798 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
7800 2018-01-24 Tom de Vries <tom@codesourcery.com>
7803 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
7804 for neutered threads in warp.
7805 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
7807 2018-01-24 Richard Biener <rguenther@suse.de>
7809 PR tree-optimization/83176
7810 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
7813 2018-01-24 Richard Biener <rguenther@suse.de>
7815 PR tree-optimization/82819
7816 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
7817 code generating pluses that are no-ops in the target precision.
7819 2018-01-24 Richard Biener <rguenther@suse.de>
7822 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
7824 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7826 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
7827 to merge probabilities.
7828 * predict.c (probably_never_executed): Also mark as cold functions
7829 with global 0 profile and guessed local profile.
7830 * profile-count.c (profile_probability::combine_with_count): New
7832 * profile-count.h (profile_probability::operator*,
7833 profile_probability::operator*=, profile_probability::operator/,
7834 profile_probability::operator/=): Reduce precision to adjusted
7835 and set value to guessed on contradictory divisions.
7836 (profile_probability::combine_with_freq): Remove.
7837 (profile_probability::combine_wiht_count): Declare.
7838 (profile_count::force_nonzero):: Set to adjusted.
7839 (profile_count::probability_in):: Set quality to adjusted.
7840 * tree-ssa-tail-merge.c (replace_block_by): Use
7843 2018-01-23 Andrew Waterman <andrew@sifive.com>
7844 Jim Wilson <jimw@sifive.com>
7846 * config/riscv/riscv.c (riscv_stack_boundary): New.
7847 (riscv_option_override): Set riscv_stack_boundary. Handle
7848 riscv_preferred_stack_boundary_arg.
7849 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
7850 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
7851 (STACK_BOUNDARY): Set to riscv_stack_boundary.
7852 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
7853 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
7854 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
7856 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
7859 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
7860 of struct ix86_frame.
7861 (ix86_expand_epilogue): Likewise. Add a local variable for
7862 the reg_save_offset field in struct ix86_frame.
7864 2018-01-23 Bin Cheng <bin.cheng@arm.com>
7866 PR tree-optimization/82604
7867 * tree-loop-distribution.c (enum partition_kind): New enum item
7868 PKIND_PARTIAL_MEMSET.
7869 (partition_builtin_p): Support above new enum item.
7870 (generate_code_for_partition): Ditto.
7871 (compute_access_range): Differentiate cases that equality can be
7872 proven at all loops, the innermost loops or no loops.
7873 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
7874 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
7875 (finalize_partitions, distribute_loop): Don't fuse partition of
7876 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
7877 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
7880 2018-01-23 Martin Liska <mliska@suse.cz>
7882 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
7883 order to ignore the predictor.
7884 (PRED_POLYMORPHIC_CALL): Likewise.
7885 (PRED_RECURSIVE_CALL): Likewise.
7887 2018-01-23 Martin Liska <mliska@suse.cz>
7889 * tree-profile.c (tree_profiling): Print function header to
7890 aware reader which function we are working on.
7891 * value-prof.c (gimple_find_values_to_profile): Do not print
7892 not interesting value histograms.
7894 2018-01-23 Martin Liska <mliska@suse.cz>
7896 * profile-count.h (enum profile_quality): Add
7897 profile_uninitialized as the first value. Do not number values
7898 as they are zero based.
7899 (profile_count::verify): Update sanity check.
7900 (profile_probability::verify): Likewise.
7902 2018-01-23 Nathan Sidwell <nathan@acm.org>
7904 * doc/invoke.texi (ffor-scope): Deprecate.
7906 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7908 PR tree-optimization/83510
7909 * domwalk.c (set_all_edges_as_executable): New function.
7910 (dom_walker::dom_walker): Convert bool param
7911 "skip_unreachable_blocks" to enum reachability. Move setup of
7912 edge flags to set_all_edges_as_executable and only do it when
7913 reachability is REACHABLE_BLOCKS.
7914 * domwalk.h (enum dom_walker::reachability): New enum.
7915 (dom_walker::dom_walker): Convert bool param
7916 "skip_unreachable_blocks" to enum reachability.
7917 (set_all_edges_as_executable): New decl.
7918 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
7919 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
7921 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
7922 but converting true to REACHABLE_BLOCKS.
7923 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
7925 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
7926 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
7927 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
7929 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
7930 if check_all_array_refs will be called.
7932 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7934 * tree.c (selftest::test_location_wrappers): Add more test
7937 2018-01-23 David Malcolm <dmalcolm@redhat.com>
7939 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
7940 (selftest::test_bit_in_range): Likewise.
7942 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7945 * doc/sourcebuild.texi (vect_float): Say that the selector
7946 only describes the situation when -funsafe-math-optimizations is on.
7947 (vect_float_strict): Document.
7949 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
7951 PR tree-optimization/83965
7952 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7953 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
7954 instead of checking only for a reduction.
7955 (vect_recog_widen_sum_pattern): Likewise.
7957 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7959 * predict.c (probably_never_executed): Only use precise profile info.
7960 (compute_function_frequency): Skip after inlining hack since we now
7961 have quality checking.
7963 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
7965 * profile-count.h (profile_probability::very_unlikely,
7966 profile_probability::unlikely, profile_probability::even): Set
7967 precision to guessed.
7969 2018-01-23 Richard Biener <rguenther@suse.de>
7971 PR tree-optimization/83963
7972 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
7973 Properly terminate dominator walk when crossing the exit edge not
7974 when visiting its source block.
7976 2018-01-23 Jakub Jelinek <jakub@redhat.com>
7979 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
7980 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
7982 2018-01-22 Jakub Jelinek <jakub@redhat.com>
7984 PR tree-optimization/83957
7985 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
7986 semicolon after for body surrounded by braces.
7988 PR tree-optimization/83081
7989 * profile-count.h (profile_probability::split): New method.
7990 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
7991 Use profile_probability::split.
7992 (do_compare_rtx_and_jump): Fix adjustment of probabilities
7993 when splitting a single conditional jump into 2.
7995 2018-01-22 David Malcolm <dmalcolm@redhat.com>
7997 PR tree-optimization/69452
7998 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
8001 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8003 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
8004 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
8005 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
8007 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8009 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
8010 * config/rl78/rl78.md (movdi): New define_expand.
8011 * config/rl78/rl78.c (rl78_split_movdi): New function.
8013 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
8016 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
8018 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
8019 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
8020 128-bit to produce an UNSPEC move to get the double word with the
8021 signbit and then a shift directly to do signbit.
8022 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
8023 implementation with a new version that just does either a direct
8024 move or a regular move. Move memory interface to separate insns.
8025 Move insns so they are next to the expander.
8026 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
8027 with signbit move. Split big and little endian case.
8028 (signbit<mode>2_dm_mem_le): Likewise.
8029 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
8030 (signbit<mode>2_dm2): Likewise.
8032 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8034 * config/rl78/rl78.md (anddi3): New define_expand.
8036 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8038 * config/rl78/rl78.md (umindi3): New define_expand.
8040 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8042 * config/rl78/rl78.md (smindi3): New define_expand.
8044 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8046 * config/rl78/rl78.md (smaxdi3): New define_expand.
8048 2018-01-22 Carl Love <cel@us.ibm.com>
8050 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
8051 LVX_V1TI): Add macro expansion.
8052 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
8053 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
8054 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
8055 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
8056 Change check to determine if the instruction is a byte reversing
8057 entry. Fix typo in comment.
8058 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
8059 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
8060 Add def_builtin calls for new builtins.
8061 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
8062 Add define_insn expansion.
8064 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8066 * config/rl78/rl78.md (umaxdi3): New define_expand.
8068 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
8070 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
8071 for non-QImode registers.
8073 2018-01-22 Richard Biener <rguenther@suse.de>
8075 PR tree-optimization/83963
8076 * graphite-scop-detection.c (scop_detection::get_sese): Delay
8077 including the loop exit block.
8078 (scop_detection::merge_sese): Likewise.
8079 (scop_detection::add_scop): Do it here instead.
8081 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8083 * doc/sourcebuild.texi (arm_softfloat): Document.
8085 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
8088 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
8089 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
8090 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
8092 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8093 David Edelsohn <dje.gcc@gmail.com>
8096 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8097 Change "crset eq" to "crset 2".
8098 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8099 (*call_indirect_aix<mode>_nospec): Likewise.
8100 (*call_value_indirect_aix<mode>_nospec): Likewise.
8101 (*call_indirect_elfv2<mode>_nospec): Likewise.
8102 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
8103 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
8104 change assembly output from . to $.
8105 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8106 (indirect_jump<mode>_nospec): Change assembly output from . to $.
8107 (*tablejump<mode>_internal1_nospec): Likewise.
8109 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
8112 * config/sh/sh_optimize_sett_clrt.cc:
8113 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
8115 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
8117 PR tree-optimization/83940
8118 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
8119 offset_dt to vect_constant_def rather than vect_unknown_def_type.
8120 (vect_check_load_store_mask): Add a mask_dt_out parameter and
8121 use it to pass back the definition type.
8122 (vect_check_store_rhs): Likewise rhs_dt_out.
8123 (vect_build_gather_load_calls): Add a mask_dt argument and use
8124 it instead of a call to vect_is_simple_use.
8125 (vectorizable_store): Update calls to vect_check_load_store_mask
8126 and vect_check_store_rhs. Use the dt returned by the latter instead
8127 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
8128 instead of calls to vect_is_simple_use. Pass the scalar rather
8129 than the vector operand to vect_is_simple_use when handling
8130 second and subsequent copies of an rhs value.
8131 (vectorizable_load): Update calls to vect_check_load_store_mask
8132 and vect_build_gather_load_calls. Use the cached mask_dt and
8133 gs_info.offset_dt instead of calls to vect_is_simple_use.
8135 2018-01-20 Jakub Jelinek <jakub@redhat.com>
8138 * tree-emutls.c: Include gimplify.h.
8139 (lower_emutls_2): New function.
8140 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
8141 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
8142 it before further processing.
8145 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
8146 UINTVAL (trueop1) instead of INTVAL (op1).
8148 2018-01-19 Jakub Jelinek <jakub@redhat.com>
8152 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
8153 INCOMING_FRAME_SP_OFFSET if not defined.
8154 (scan_trace): Add ENTRY argument. If true and
8155 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
8156 emit a note to adjust the CFA offset.
8157 (create_cfi_notes): Adjust scan_trace callers.
8158 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
8159 INCOMING_FRAME_SP_OFFSET in the CIE.
8160 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
8161 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
8163 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
8164 * doc/tm.texi: Regenerated.
8166 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8168 PR rtl-optimization/83147
8169 * lra-constraints.c (remove_inheritance_pseudos): Use
8170 lra_substitute_pseudo_within_insn.
8172 2018-01-19 Tom de Vries <tom@codesourcery.com>
8173 Cesar Philippidis <cesar@codesourcery.com>
8176 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
8178 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
8181 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
8182 spaces for function labels.
8184 2018-01-19 Martin Liska <mliska@suse.cz>
8186 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
8187 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
8188 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
8189 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
8190 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
8191 (PRED_CONST_RETURN): Change from 69 to 65.
8192 (PRED_NULL_RETURN): Change from 91 to 71.
8193 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
8194 (PRED_LOOP_GUARD): Change from 66 to 73.
8196 2018-01-19 Martin Liska <mliska@suse.cz>
8198 * predict.c (predict_insn_def): Add new assert.
8199 (struct branch_predictor): Change type to signed integer.
8200 (test_prediction_value_range): Amend test to cover
8202 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
8203 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
8204 (PRED_LOOP_ITERATIONS_MAX): Likewise.
8205 (PRED_LOOP_IV_COMPARE): Likewise.
8206 * predict.h (PROB_UNINITIALIZED): Define new constant.
8208 2018-01-19 Martin Liska <mliska@suse.cz>
8210 * predict.c (dump_prediction): Add new format for
8211 analyze_brprob.py script which is enabled with -details
8213 * profile-count.h (precise_p): New function.
8215 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8217 PR tree-optimization/83922
8218 * tree-vect-loop.c (vect_verify_full_masking): Return false if
8219 there are no statements that need masking.
8220 (vect_active_double_reduction_p): New function.
8221 (vect_analyze_loop_operations): Use it when handling phis that
8222 are not in the loop header.
8224 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
8226 PR tree-optimization/83914
8227 * tree-vect-loop.c (vectorizable_induction): Don't convert
8228 init_expr or apply the peeling adjustment for inductions
8229 that are nested within the vectorized loop.
8231 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8233 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
8236 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8240 * function.h (gimplify_parameters): Add gimple_seq * argument.
8241 * function.c: Include gimple.h and options.h.
8242 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
8243 for the added local temporaries if needed.
8244 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
8245 if there are any parameter cleanups, wrap whole body into a
8246 try/finally with the cleanups.
8248 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
8251 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
8252 Use GET_MODE_CLASS for scalar floating point.
8254 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8258 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
8259 Fix call of call_cgraph_insertion_hooks.
8261 2018-01-18 Martin Sebor <msebor@redhat.com>
8263 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
8265 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
8268 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
8271 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
8274 * common.opt: (-ffile-prefix-map): New option.
8275 * opts.c (common_handle_option): Defer it.
8276 * opts-global.c (handle_common_deferred_options): Handle it.
8277 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
8278 * file-prefix-map.h: New file.
8279 (remap_debug_filename, add_debug_prefix_map): ...here.
8280 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
8281 * final.c (debug_prefix_map, add_debug_prefix_map
8282 remap_debug_filename): Move to...
8283 * file-prefix-map.c: New file.
8284 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
8285 generalize, get rid of alloca(), use strrchr() instead of strchr().
8286 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
8287 Implement in terms of add_prefix_map().
8288 (remap_macro_filename, remap_debug_filename): Implement in term of
8290 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
8291 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
8292 * dbxout.c: Include file-prefix-map.h.
8293 * varasm.c: Likewise.
8294 * vmsdbgout.c: Likewise.
8295 * xcoffout.c: Likewise.
8296 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
8297 * doc/cppopts.texi (-fmacro-prefix-map): Document.
8298 * doc/invoke.texi (-ffile-prefix-map): Document.
8299 (-fdebug-prefix-map): Update description.
8301 2018-01-18 Martin Liska <mliska@suse.cz>
8303 * config/i386/i386.c (indirect_thunk_name): Document that also
8305 (output_indirect_thunk): Document why both instructions
8306 (pause and lfence) are generated.
8308 2018-01-18 Richard Biener <rguenther@suse.de>
8310 PR tree-optimization/83887
8311 * graphite-scop-detection.c
8312 (scop_detection::get_nearest_dom_with_single_entry): Remove.
8313 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
8314 (scop_detection::merge_sese): Re-implement with a flood-fill
8315 algorithm that properly finds a SESE region if it exists.
8317 2018-01-18 Jakub Jelinek <jakub@redhat.com>
8320 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
8321 pointer_diff optimizations use view_convert instead of convert.
8323 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8325 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
8326 Generate different code for -mno-speculate-indirect-jumps.
8327 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
8328 (*call_indirect_aix<mode>): Disable for
8329 -mno-speculate-indirect-jumps.
8330 (*call_indirect_aix<mode>_nospec): New define_insn.
8331 (*call_value_indirect_aix<mode>): Disable for
8332 -mno-speculate-indirect-jumps.
8333 (*call_value_indirect_aix<mode>_nospec): New define_insn.
8334 (*sibcall_nonlocal_sysv<mode>): Generate different code for
8335 -mno-speculate-indirect-jumps.
8336 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
8338 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
8340 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
8341 long double type, set the flags for noting the default long double
8342 type, even if we don't pass or return a long double type.
8344 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
8347 * ipa-inline.c (flatten_function): Do not overwrite final inlining
8350 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
8352 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
8353 support for merge[hl].
8354 (fold_mergehl_helper): New helper function.
8355 (tree-vector-builder.h): New #include for tree_vector_builder usage.
8356 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
8357 (altivec_vmrglw_direct): Add xxmrglw insn.
8359 2018-01-17 Andrew Waterman <andrew@sifive.com>
8361 * config/riscv/riscv.c (riscv_conditional_register_usage): If
8362 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
8364 2018-01-17 David Malcolm <dmalcolm@redhat.com>
8367 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
8368 call the lto_location_cache before reading the
8369 DECL_SOURCE_LOCATION of the types.
8371 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
8372 Richard Sandiford <richard.sandiford@linaro.org>
8374 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
8375 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
8376 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
8377 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
8378 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
8380 * config/aarch64/constraints.md (aarch64_movti_operand):
8382 * config/aarch64/predicates.md (Uti): Add new constraint.
8384 2018-01-17 Carl Love <cel@us.ibm.com>
8386 * config/rs6000/vsx.md (define_expand xl_len_r,
8387 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
8388 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
8390 (define_expand, define_insn): Move the shift left from the
8391 define_insn to the define_expand for lxvl and stxvl instructions.
8392 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
8393 and XL_LEN_R definitions to PURE.
8395 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
8397 * config/i386/i386.c (indirect_thunk_name): Declare regno
8398 as unsigned int. Compare regno with INVALID_REGNUM.
8399 (output_indirect_thunk): Ditto.
8400 (output_indirect_thunk_function): Ditto.
8401 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
8402 in the call to output_indirect_thunk_function.
8404 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
8407 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
8408 rather than the size of inner_type to determine the stack slot size
8409 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
8411 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
8414 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
8417 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
8419 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
8420 endian Linux systems to optionally enable multilibs for selecting
8421 the long double type if the user configured an explicit type.
8422 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
8423 have no long double multilibs if not defined.
8424 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
8425 warn if the user used -mabi={ieee,ibm}longdouble and we built
8426 multilibs for long double.
8427 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
8428 appropriate multilib option.
8429 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
8431 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
8432 for building long double multilibs.
8433 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
8435 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
8437 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
8440 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
8442 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
8445 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
8448 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
8451 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8453 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
8454 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
8456 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8458 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
8459 different rtl trees depending on TARGET_64BIT.
8460 (rs6000_gen_lvx): Likewise.
8462 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
8464 * config/visium/visium.md (nop): Tweak comment.
8465 (hazard_nop): Likewise.
8467 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8469 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
8470 -mspeculate-indirect-jumps.
8471 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
8472 for -mno-speculate-indirect-jumps.
8473 (*call_indirect_elfv2<mode>_nospec): New define_insn.
8474 (*call_value_indirect_elfv2<mode>): Disable for
8475 -mno-speculate-indirect-jumps.
8476 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
8477 (indirect_jump): Emit different RTL for
8478 -mno-speculate-indirect-jumps.
8479 (*indirect_jump<mode>): Disable for
8480 -mno-speculate-indirect-jumps.
8481 (*indirect_jump<mode>_nospec): New define_insn.
8482 (tablejump): Emit different RTL for
8483 -mno-speculate-indirect-jumps.
8484 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
8485 (tablejumpsi_nospec): New define_expand.
8486 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
8487 (tablejumpdi_nospec): New define_expand.
8488 (*tablejump<mode>_internal1): Disable for
8489 -mno-speculate-indirect-jumps.
8490 (*tablejump<mode>_internal1_nospec): New define_insn.
8491 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
8494 2018-01-16 Artyom Skrobov tyomitch@gmail.com
8496 * caller-save.c (insert_save): Drop unnecessary parameter. All
8499 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8500 Richard Biener <rguenth@suse.de>
8503 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
8504 return early, inline manually is_gimple_sizepos. Make sure if we
8505 call gimplify_expr we don't end up with a gimple constant.
8506 * tree.c (variably_modified_type_p): Don't return true for
8507 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
8508 * gimplify.h (is_gimple_sizepos): Remove.
8510 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8512 PR tree-optimization/83857
8513 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
8514 vectorizable_live_operation for pure SLP statements.
8515 (vectorizable_live_operation): Handle PHIs.
8517 2018-01-16 Richard Biener <rguenther@suse.de>
8519 PR tree-optimization/83867
8520 * tree-vect-stmts.c (vect_transform_stmt): Precompute
8521 nested_in_vect_loop_p since the scalar stmt may get invalidated.
8523 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8526 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
8527 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
8528 If off is not INTEGER_CST, issue a may not be aligned warning
8529 rather than isn't aligned. Use isn%'t rather than isn't.
8530 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
8532 <case MULT_EXPR>: Improve the case when bottom and one of the
8533 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
8534 operand, in that case check if the other operand is multiple of
8535 bottom divided by the INTEGER_CST operand.
8537 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8540 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
8541 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
8542 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
8543 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
8544 * config/pa/pa.c (pa_function_arg_advance): Likewise.
8545 (pa_function_arg, pa_arg_partial_bytes): Likewise.
8546 (pa_function_arg_size): New function.
8548 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8550 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
8551 in a separate statement.
8553 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
8555 PR tree-optimization/83847
8556 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
8557 group gathers and scatters.
8559 2018-01-16 Jakub Jelinek <jakub@redhat.com>
8561 PR rtl-optimization/86620
8562 * params.def (max-sched-ready-insns): Bump minimum value to 1.
8564 PR rtl-optimization/83213
8565 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
8566 to last if both are JUMP_INSNs.
8568 PR tree-optimization/83843
8569 * gimple-ssa-store-merging.c
8570 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
8571 store_immediate_info for bswap/nop orig_stores.
8573 2018-01-15 Andrew Waterman <andrew@sifive.com>
8575 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
8577 <UDIV>: Increase cost if !TARGET_DIV.
8579 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
8581 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
8582 (define_attr "cr_logical_3op"): New.
8583 (cceq_ior_compare): Adjust.
8584 (cceq_ior_compare_complement): Adjust.
8585 (*cceq_rev_compare): Adjust.
8586 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
8587 (is_cracked_insn): Adjust.
8588 (insn_must_be_first_in_group): Adjust.
8589 * config/rs6000/40x.md: Adjust.
8590 * config/rs6000/440.md: Adjust.
8591 * config/rs6000/476.md: Adjust.
8592 * config/rs6000/601.md: Adjust.
8593 * config/rs6000/603.md: Adjust.
8594 * config/rs6000/6xx.md: Adjust.
8595 * config/rs6000/7450.md: Adjust.
8596 * config/rs6000/7xx.md: Adjust.
8597 * config/rs6000/8540.md: Adjust.
8598 * config/rs6000/cell.md: Adjust.
8599 * config/rs6000/e300c2c3.md: Adjust.
8600 * config/rs6000/e500mc.md: Adjust.
8601 * config/rs6000/e500mc64.md: Adjust.
8602 * config/rs6000/e5500.md: Adjust.
8603 * config/rs6000/e6500.md: Adjust.
8604 * config/rs6000/mpc.md: Adjust.
8605 * config/rs6000/power4.md: Adjust.
8606 * config/rs6000/power5.md: Adjust.
8607 * config/rs6000/power6.md: Adjust.
8608 * config/rs6000/power7.md: Adjust.
8609 * config/rs6000/power8.md: Adjust.
8610 * config/rs6000/power9.md: Adjust.
8611 * config/rs6000/rs64.md: Adjust.
8612 * config/rs6000/titan.md: Adjust.
8614 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8616 * config/i386/predicates.md (indirect_branch_operand): Rewrite
8617 ix86_indirect_branch_register logic.
8619 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8621 * config/i386/constraints.md (Bs): Update
8622 ix86_indirect_branch_register check. Don't check
8623 ix86_indirect_branch_register with GOT_memory_operand.
8625 * config/i386/predicates.md (GOT_memory_operand): Don't check
8626 ix86_indirect_branch_register here.
8627 (GOT32_symbol_operand): Likewise.
8629 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8631 * config/i386/predicates.md (constant_call_address_operand):
8632 Rewrite ix86_indirect_branch_register logic.
8633 (sibcall_insn_operand): Likewise.
8635 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8637 * config/i386/constraints.md (Bs): Replace
8638 ix86_indirect_branch_thunk_register with
8639 ix86_indirect_branch_register.
8641 * config/i386/i386.md (indirect_jump): Likewise.
8642 (tablejump): Likewise.
8643 (*sibcall_memory): Likewise.
8644 (*sibcall_value_memory): Likewise.
8645 Peepholes of indirect call and jump via memory: Likewise.
8646 * config/i386/i386.opt: Likewise.
8647 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8648 (GOT_memory_operand): Likewise.
8649 (call_insn_operand): Likewise.
8650 (sibcall_insn_operand): Likewise.
8651 (GOT32_symbol_operand): Likewise.
8653 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8656 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
8657 type rather than type addr's type points to.
8658 (expand_omp_atomic_mutex): Likewise.
8659 (expand_omp_atomic): Likewise.
8661 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
8664 * config/i386/i386.c (output_indirect_thunk_function): Use
8665 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
8666 for __x86_return_thunk.
8668 2018-01-15 Richard Biener <rguenther@suse.de>
8671 * expmed.c (extract_bit_field_1): Fix typo.
8673 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8676 * config/arm/iterators.md (VF): New mode iterator.
8677 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
8678 Remove integer-related logic from pattern.
8679 (neon_vabd<mode>_3): Likewise.
8681 2018-01-15 Jakub Jelinek <jakub@redhat.com>
8684 * common.opt (fstrict-overflow): No longer an alias.
8685 (fwrapv-pointer): New option.
8686 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
8687 also for pointer types based on flag_wrapv_pointer.
8688 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
8689 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
8690 opts->x_flag_wrapv got set.
8691 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
8692 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
8693 POINTER_TYPE_OVERFLOW_UNDEFINED.
8694 * match.pd: Likewise in address comparison pattern.
8695 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
8697 2018-01-15 Richard Biener <rguenther@suse.de>
8700 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
8701 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
8702 Reset type names to their identifier if their TYPE_DECL doesn't
8703 have linkage (and thus is used for ODR and devirt).
8704 (save_debug_info_for_decl): Remove.
8705 (save_debug_info_for_type): Likewise.
8706 (add_tree_to_fld_list): Adjust.
8707 * tree-pretty-print.c (dump_generic_node): Make dumping of
8708 type names more robust.
8710 2018-01-15 Richard Biener <rguenther@suse.de>
8712 * BASE-VER: Bump to 8.0.1.
8714 2018-01-14 Martin Sebor <msebor@redhat.com>
8717 * builtins.c (check_access): Avoid warning when the no-warning bit
8720 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
8722 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
8723 * ira-color (allocno_hard_regs_compare): Likewise.
8725 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
8728 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
8729 Use .pushsection/.popsection.
8731 2018-01-14 Martin Sebor <msebor@redhat.com>
8734 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
8736 2018-01-14 Jakub Jelinek <jakub@redhat.com>
8738 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
8739 entry from extra_headers.
8740 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
8741 extra_headers, make the list bitwise identical to the i?86-*-* one.
8743 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8745 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
8746 -mcmodel=large with -mindirect-branch=thunk,
8747 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
8748 -mfunction-return=thunk-extern.
8749 * doc/invoke.texi: Document -mcmodel=large is incompatible with
8750 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
8751 -mfunction-return=thunk and -mfunction-return=thunk-extern.
8753 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8755 * config/i386/i386.c (print_reg): Print the name of the full
8756 integer register without '%'.
8757 (ix86_print_operand): Handle 'V'.
8758 * doc/extend.texi: Document 'V' modifier.
8760 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8762 * config/i386/constraints.md (Bs): Disallow memory operand for
8763 -mindirect-branch-register.
8765 * config/i386/predicates.md (indirect_branch_operand): Likewise.
8766 (GOT_memory_operand): Likewise.
8767 (call_insn_operand): Likewise.
8768 (sibcall_insn_operand): Likewise.
8769 (GOT32_symbol_operand): Likewise.
8770 * config/i386/i386.md (indirect_jump): Call convert_memory_address
8771 for -mindirect-branch-register.
8772 (tablejump): Likewise.
8773 (*sibcall_memory): Likewise.
8774 (*sibcall_value_memory): Likewise.
8775 Disallow peepholes of indirect call and jump via memory for
8776 -mindirect-branch-register.
8777 (*call_pop): Replace m with Bw.
8778 (*call_value_pop): Likewise.
8779 (*sibcall_pop_memory): Replace m with Bs.
8780 * config/i386/i386.opt (mindirect-branch-register): New option.
8781 * doc/invoke.texi: Document -mindirect-branch-register option.
8783 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8785 * config/i386/i386-protos.h (ix86_output_function_return): New.
8786 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
8787 set function_return_type.
8788 (indirect_thunk_name): Add ret_p to indicate thunk for function
8790 (output_indirect_thunk_function): Pass false to
8791 indirect_thunk_name.
8792 (ix86_output_indirect_branch_via_reg): Likewise.
8793 (ix86_output_indirect_branch_via_push): Likewise.
8794 (output_indirect_thunk_function): Create alias for function
8795 return thunk if regno < 0.
8796 (ix86_output_function_return): New function.
8797 (ix86_handle_fndecl_attribute): Handle function_return.
8798 (ix86_attribute_table): Add function_return.
8799 * config/i386/i386.h (machine_function): Add
8800 function_return_type.
8801 * config/i386/i386.md (simple_return_internal): Use
8802 ix86_output_function_return.
8803 (simple_return_internal_long): Likewise.
8804 * config/i386/i386.opt (mfunction-return=): New option.
8805 (indirect_branch): Mention -mfunction-return=.
8806 * doc/extend.texi: Document function_return function attribute.
8807 * doc/invoke.texi: Document -mfunction-return= option.
8809 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
8811 * config/i386/i386-opts.h (indirect_branch): New.
8812 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
8813 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
8814 with local indirect jump when converting indirect call and jump.
8815 (ix86_set_indirect_branch_type): New.
8816 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
8817 (indirectlabelno): New.
8818 (indirect_thunk_needed): Likewise.
8819 (indirect_thunk_bnd_needed): Likewise.
8820 (indirect_thunks_used): Likewise.
8821 (indirect_thunks_bnd_used): Likewise.
8822 (INDIRECT_LABEL): Likewise.
8823 (indirect_thunk_name): Likewise.
8824 (output_indirect_thunk): Likewise.
8825 (output_indirect_thunk_function): Likewise.
8826 (ix86_output_indirect_branch_via_reg): Likewise.
8827 (ix86_output_indirect_branch_via_push): Likewise.
8828 (ix86_output_indirect_branch): Likewise.
8829 (ix86_output_indirect_jmp): Likewise.
8830 (ix86_code_end): Call output_indirect_thunk_function if needed.
8831 (ix86_output_call_insn): Call ix86_output_indirect_branch if
8833 (ix86_handle_fndecl_attribute): Handle indirect_branch.
8834 (ix86_attribute_table): Add indirect_branch.
8835 * config/i386/i386.h (machine_function): Add indirect_branch_type
8836 and has_local_indirect_jump.
8837 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
8839 (tablejump): Likewise.
8840 (*indirect_jump): Use ix86_output_indirect_jmp.
8841 (*tablejump_1): Likewise.
8842 (simple_return_indirect_internal): Likewise.
8843 * config/i386/i386.opt (mindirect-branch=): New option.
8844 (indirect_branch): New.
8847 (thunk-inline): Likewise.
8848 (thunk-extern): Likewise.
8849 * doc/extend.texi: Document indirect_branch function attribute.
8850 * doc/invoke.texi: Document -mindirect-branch= option.
8852 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
8855 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
8857 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
8859 * ipa-inline.c (want_inline_small_function_p): Return false if
8860 inlining has already failed with CIF_FINAL_ERROR.
8861 (update_caller_keys): Call want_inline_small_function_p before
8863 (update_callee_keys): Likewise.
8865 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
8867 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
8869 (rs6000_quadword_masked_address_p): Likewise.
8870 (quad_aligned_load_p): Likewise.
8871 (quad_aligned_store_p): Likewise.
8872 (const_load_sequence_p): Add comment to describe the outer-most loop.
8873 (mimic_memory_attributes_and_flags): New function.
8874 (rs6000_gen_stvx): Likewise.
8875 (replace_swapped_aligned_store): Likewise.
8876 (rs6000_gen_lvx): Likewise.
8877 (replace_swapped_aligned_load): Likewise.
8878 (replace_swapped_load_constant): Capitalize argument name in
8879 comment describing this function.
8880 (rs6000_analyze_swaps): Add a third pass to search for vector loads
8881 and stores that access quad-word aligned addresses and replace
8882 with stvx or lvx instructions when appropriate.
8883 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
8884 New function prototype.
8885 (rs6000_quadword_masked_address_p): Likewise.
8886 (rs6000_gen_lvx): Likewise.
8887 (rs6000_gen_stvx): Likewise.
8888 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
8889 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
8890 when memory address is aligned.
8891 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
8892 this split to select lvx instruction when memory address is aligned.
8893 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
8894 instruction when memory address is aligned.
8895 (*vsx_le_perm_load_v16qi): Likewise.
8896 (four unnamed splitters): Modify to select the stvx instruction
8897 when memory is aligned.
8899 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
8901 * predict.c (determine_unlikely_bbs): Handle correctly BBs
8902 which appears in the queue multiple times.
8904 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8905 Alan Hayward <alan.hayward@arm.com>
8906 David Sherwood <david.sherwood@arm.com>
8908 * tree-vectorizer.h (vec_lower_bound): New structure.
8909 (_loop_vec_info): Add check_nonzero and lower_bounds.
8910 (LOOP_VINFO_CHECK_NONZERO): New macro.
8911 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
8912 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
8913 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
8914 fields. Make seg_len the distance travelled, not including the
8916 (dr_direction_indicator): Declare.
8917 (dr_zero_step_indicator): Likewise.
8918 (dr_known_forward_stride_p): Likewise.
8919 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
8921 (runtime_alias_check_p): Allow runtime alias checks with
8923 (operator ==): Compare access_size and align.
8924 (prune_runtime_alias_test_list): Rework for new distinction between
8925 the access_size and seg_len.
8926 (create_intersect_range_checks_index): Likewise. Cope with polynomial
8928 (get_segment_min_max): New function.
8929 (create_intersect_range_checks): Use it.
8930 (dr_step_indicator): New function.
8931 (dr_direction_indicator): Likewise.
8932 (dr_zero_step_indicator): Likewise.
8933 (dr_known_forward_stride_p): Likewise.
8934 * tree-loop-distribution.c (data_ref_segment_size): Return
8935 DR_STEP * (niters - 1).
8936 (compute_alias_check_pairs): Update call to the dr_with_seg_len
8938 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
8939 (vect_preserves_scalar_order_p): New function, split out from...
8940 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
8941 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
8942 (vect_vfa_access_size): New function.
8943 (vect_vfa_align): Likewise.
8944 (vect_compile_time_alias): Take access_size_a and access_b arguments.
8945 (dump_lower_bound): New function.
8946 (vect_check_lower_bound): Likewise.
8947 (vect_small_gap_p): Likewise.
8948 (vectorizable_with_step_bound_p): Likewise.
8949 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
8950 depencies if the vectorization factor is 1. Convert the checks
8951 for nonzero steps into checks on the bounds of DR_STEP. Try using
8952 a bunds check for variable steps if the minimum required step is
8953 relatively small. Update calls to the dr_with_seg_len
8954 constructor and to vect_compile_time_alias.
8955 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
8957 (vect_loop_versioning): Call it.
8958 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
8960 (vect_estimate_min_profitable_iters): Account for any bounds checks.
8962 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8963 Alan Hayward <alan.hayward@arm.com>
8964 David Sherwood <david.sherwood@arm.com>
8966 * doc/sourcebuild.texi (vect_scatter_store): Document.
8967 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
8969 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
8971 * genopinit.c (main): Add supports_vec_scatter_store and
8972 supports_vec_scatter_store_cached to target_optabs.
8973 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
8974 IFN_MASK_SCATTER_STORE.
8975 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
8977 * internal-fn.h (internal_store_fn_p): Declare.
8978 (internal_fn_stored_value_index): Likewise.
8979 * internal-fn.c (scatter_store_direct): New macro.
8980 (expand_scatter_store_optab_fn): New function.
8981 (direct_scatter_store_optab_supported_p): New macro.
8982 (internal_store_fn_p): New function.
8983 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
8984 IFN_MASK_SCATTER_STORE.
8985 (internal_fn_mask_index): Likewise.
8986 (internal_fn_stored_value_index): New function.
8987 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
8989 * optabs-query.h (supports_vec_scatter_store_p): Declare.
8990 * optabs-query.c (supports_vec_scatter_store_p): New function.
8991 * tree-vectorizer.h (vect_get_store_rhs): Declare.
8992 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
8993 true for scatter stores.
8994 (vect_gather_scatter_fn_p): Handle scatter stores too.
8995 (vect_check_gather_scatter): Consider using scatter stores if
8996 supports_vec_scatter_store_p.
8997 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
8999 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9000 internal_fn_stored_value_index.
9001 (check_load_store_masking): Handle scatter stores too.
9002 (vect_get_store_rhs): Make public.
9003 (vectorizable_call): Use internal_store_fn_p.
9004 (vectorizable_store): Handle scatter store internal functions.
9005 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
9006 when deciding whether the end of the group has been reached.
9007 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
9008 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
9009 (mask_scatter_store<mode>): New insns.
9011 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9012 Alan Hayward <alan.hayward@arm.com>
9013 David Sherwood <david.sherwood@arm.com>
9015 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
9016 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
9017 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
9019 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
9020 Use vect_truncate_gather_scatter_offset if we can't treat the
9021 operation as a normal gather load or scatter store.
9022 (get_group_load_store_type): Take the gather_scatter_info
9023 as argument. Try using a gather load or scatter store for
9024 single-element groups.
9025 (get_load_store_type): Update calls to get_group_load_store_type
9026 and vect_use_strided_gather_scatters_p.
9028 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9029 Alan Hayward <alan.hayward@arm.com>
9030 David Sherwood <david.sherwood@arm.com>
9032 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
9033 optional tree argument.
9034 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
9036 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
9037 but continue to use the current value as a fallback.
9038 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
9039 to compare the updates.
9040 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
9041 (get_load_store_type): Use it when handling a strided access.
9042 (vect_get_strided_load_store_ops): New function.
9043 (vect_get_data_ptr_increment): Likewise.
9044 (vectorizable_load): Handle strided gather loads. Always pass
9045 a step to vect_create_data_ref_ptr and bump_vector_ptr.
9047 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9048 Alan Hayward <alan.hayward@arm.com>
9049 David Sherwood <david.sherwood@arm.com>
9051 * doc/md.texi (gather_load@var{m}): Document.
9052 (mask_gather_load@var{m}): Likewise.
9053 * genopinit.c (main): Add supports_vec_gather_load and
9054 supports_vec_gather_load_cached to target_optabs.
9055 * optabs-tree.c (init_tree_optimization_optabs): Use
9056 ggc_cleared_alloc to allocate target_optabs.
9057 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
9058 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
9060 * internal-fn.h (internal_load_fn_p): Declare.
9061 (internal_gather_scatter_fn_p): Likewise.
9062 (internal_fn_mask_index): Likewise.
9063 (internal_gather_scatter_fn_supported_p): Likewise.
9064 * internal-fn.c (gather_load_direct): New macro.
9065 (expand_gather_load_optab_fn): New function.
9066 (direct_gather_load_optab_supported_p): New macro.
9067 (direct_internal_fn_optab): New function.
9068 (internal_load_fn_p): Likewise.
9069 (internal_gather_scatter_fn_p): Likewise.
9070 (internal_fn_mask_index): Likewise.
9071 (internal_gather_scatter_fn_supported_p): Likewise.
9072 * optabs-query.c (supports_at_least_one_mode_p): New function.
9073 (supports_vec_gather_load_p): Likewise.
9074 * optabs-query.h (supports_vec_gather_load_p): Declare.
9075 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
9076 and memory_type field.
9077 (NUM_PATTERNS): Bump to 15.
9078 * tree-vect-data-refs.c: Include internal-fn.h.
9079 (vect_gather_scatter_fn_p): New function.
9080 (vect_describe_gather_scatter_call): Likewise.
9081 (vect_check_gather_scatter): Try using internal functions for
9082 gather loads. Recognize existing calls to a gather load function.
9083 (vect_analyze_data_refs): Consider using gather loads if
9084 supports_vec_gather_load_p.
9085 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
9086 (vect_get_gather_scatter_offset_type): Likewise.
9087 (vect_convert_mask_for_vectype): Likewise.
9088 (vect_add_conversion_to_patterm): Likewise.
9089 (vect_try_gather_scatter_pattern): Likewise.
9090 (vect_recog_gather_scatter_pattern): New pattern recognizer.
9091 (vect_vect_recog_func_ptrs): Add it.
9092 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
9093 internal_fn_mask_index and internal_gather_scatter_fn_p.
9094 (check_load_store_masking): Take the gather_scatter_info as an
9095 argument and handle gather loads.
9096 (vect_get_gather_scatter_ops): New function.
9097 (vectorizable_call): Check internal_load_fn_p.
9098 (vectorizable_load): Likewise. Handle gather load internal
9100 (vectorizable_store): Update call to check_load_store_masking.
9101 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
9102 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
9103 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
9104 (aarch64_gather_scale_operand_d): New predicates.
9105 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
9106 (mask_gather_load<mode>): New insns.
9108 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9109 Alan Hayward <alan.hayward@arm.com>
9110 David Sherwood <david.sherwood@arm.com>
9112 * optabs.def (fold_left_plus_optab): New optab.
9113 * doc/md.texi (fold_left_plus_@var{m}): Document.
9114 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
9115 * internal-fn.c (fold_left_direct): Define.
9116 (expand_fold_left_optab_fn): Likewise.
9117 (direct_fold_left_optab_supported_p): Likewise.
9118 * fold-const-call.c (fold_const_fold_left): New function.
9119 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
9120 * tree-parloops.c (valid_reduction_p): New function.
9121 (gather_scalar_reductions): Use it.
9122 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
9123 (vect_finish_replace_stmt): Declare.
9124 * tree-vect-loop.c (fold_left_reduction_fn): New function.
9125 (needs_fold_left_reduction_p): New function, split out from...
9126 (vect_is_simple_reduction): ...here. Accept reductions that
9127 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
9128 (vect_force_simple_reduction): Also store the reduction type in
9129 the assignment's STMT_VINFO_REDUC_TYPE.
9130 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
9131 (merge_with_identity): New function.
9132 (vect_expand_fold_left): Likewise.
9133 (vectorize_fold_left_reduction): Likewise.
9134 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
9135 scalar phi in place for it. Check for target support and reject
9136 cases that would reassociate the operation. Defer the transform
9137 phase to vectorize_fold_left_reduction.
9138 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
9139 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
9140 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
9142 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9144 * tree-if-conv.c (predicate_mem_writes): Remove redundant
9145 call to ifc_temp_var.
9147 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9148 Alan Hayward <alan.hayward@arm.com>
9149 David Sherwood <david.sherwood@arm.com>
9151 * target.def (legitimize_address_displacement): Take the original
9152 offset as a poly_int.
9153 * targhooks.h (default_legitimize_address_displacement): Update
9155 * targhooks.c (default_legitimize_address_displacement): Likewise.
9156 * doc/tm.texi: Regenerate.
9157 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
9158 as an argument, moving assert of ad->disp == ad->disp_term to...
9159 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
9160 Try calling targetm.legitimize_address_displacement before expanding
9161 the address rather than afterwards, and adjust for the new interface.
9162 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
9163 Match the new hook interface. Handle SVE addresses.
9164 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
9167 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9169 * Makefile.in (OBJS): Add early-remat.o.
9170 * target.def (select_early_remat_modes): New hook.
9171 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
9172 * doc/tm.texi: Regenerate.
9173 * targhooks.h (default_select_early_remat_modes): Declare.
9174 * targhooks.c (default_select_early_remat_modes): New function.
9175 * timevar.def (TV_EARLY_REMAT): New timevar.
9176 * passes.def (pass_early_remat): New pass.
9177 * tree-pass.h (make_pass_early_remat): Declare.
9178 * early-remat.c: New file.
9179 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
9181 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
9183 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9184 Alan Hayward <alan.hayward@arm.com>
9185 David Sherwood <david.sherwood@arm.com>
9187 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
9188 vfm1 with a bound_epilog parameter.
9189 (vect_do_peeling): Update calls accordingly, and move the prologue
9190 call earlier in the function. Treat the base bound_epilog as 0 for
9191 fully-masked loops and retain vf - 1 for other loops. Add 1 to
9192 this base when peeling for gaps.
9193 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
9194 with fully-masked loops.
9195 (vect_estimate_min_profitable_iters): Handle the single peeled
9196 iteration in that case.
9198 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9199 Alan Hayward <alan.hayward@arm.com>
9200 David Sherwood <david.sherwood@arm.com>
9202 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
9203 single-element interleaving even if the size is not a power of 2.
9204 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
9205 accesses for single-element interleaving if the group size is
9208 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9209 Alan Hayward <alan.hayward@arm.com>
9210 David Sherwood <david.sherwood@arm.com>
9212 * doc/md.texi (fold_extract_last_@var{m}): Document.
9213 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
9214 * optabs.def (fold_extract_last_optab): New optab.
9215 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
9216 * internal-fn.c (fold_extract_direct): New macro.
9217 (expand_fold_extract_optab_fn): Likewise.
9218 (direct_fold_extract_optab_supported_p): Likewise.
9219 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
9220 * tree-vect-loop.c (vect_model_reduction_cost): Handle
9221 EXTRACT_LAST_REDUCTION.
9222 (get_initial_def_for_reduction): Do not create an initial vector
9223 for EXTRACT_LAST_REDUCTION reductions.
9224 (vectorizable_reduction): Leave the scalar phi in place for
9225 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
9226 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
9227 epilogue code for EXTRACT_LAST_REDUCTION and defer the
9228 transform phase to vectorizable_condition.
9229 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
9231 (vect_finish_stmt_generation): ...here.
9232 (vect_finish_replace_stmt): New function.
9233 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
9234 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
9236 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
9238 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9239 Alan Hayward <alan.hayward@arm.com>
9240 David Sherwood <david.sherwood@arm.com>
9242 * doc/md.texi (extract_last_@var{m}): Document.
9243 * optabs.def (extract_last_optab): New optab.
9244 * internal-fn.def (EXTRACT_LAST): New internal function.
9245 * internal-fn.c (cond_unary_direct): New macro.
9246 (expand_cond_unary_optab_fn): Likewise.
9247 (direct_cond_unary_optab_supported_p): Likewise.
9248 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
9249 loops using EXTRACT_LAST.
9250 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
9251 (extract_last_<mode>): ...this optab.
9252 (vec_extract<mode><Vel>): Update accordingly.
9254 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9255 Alan Hayward <alan.hayward@arm.com>
9256 David Sherwood <david.sherwood@arm.com>
9258 * target.def (empty_mask_is_expensive): New hook.
9259 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
9260 * doc/tm.texi: Regenerate.
9261 * targhooks.h (default_empty_mask_is_expensive): Declare.
9262 * targhooks.c (default_empty_mask_is_expensive): New function.
9263 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
9264 if the target says that empty masks are expensive.
9265 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
9267 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
9269 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9270 Alan Hayward <alan.hayward@arm.com>
9271 David Sherwood <david.sherwood@arm.com>
9273 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
9274 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
9275 (vect_use_loop_mask_for_alignment_p): New function.
9276 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
9277 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
9278 niters_skip argument. Make sure that the first niters_skip elements
9279 of the first iteration are inactive.
9280 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
9281 Update call to vect_set_loop_masks_directly.
9282 (get_misalign_in_elems): New function, split out from...
9283 (vect_gen_prolog_loop_niters): ...here.
9284 (vect_update_init_of_dr): Take a code argument that specifies whether
9285 the adjustment should be added or subtracted.
9286 (vect_update_init_of_drs): Likewise.
9287 (vect_prepare_for_masked_peels): New function.
9288 (vect_do_peeling): Skip prologue peeling if we're using a mask
9289 instead. Update call to vect_update_inits_of_drs.
9290 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9292 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
9293 alignment. Do not include the number of peeled iterations in
9294 the minimum threshold in that case.
9295 (vectorizable_induction): Adjust the start value down by
9296 LOOP_VINFO_MASK_SKIP_NITERS iterations.
9297 (vect_transform_loop): Call vect_prepare_for_masked_peels.
9298 Take the number of skipped iterations into account when calculating
9300 * tree-vect-stmts.c (vect_gen_while_not): New function.
9302 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9303 Alan Hayward <alan.hayward@arm.com>
9304 David Sherwood <david.sherwood@arm.com>
9306 * doc/sourcebuild.texi (vect_fully_masked): Document.
9307 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
9309 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
9311 (vect_analyze_loop_2): ...here. Don't check the vectorization
9312 factor against the number of loop iterations if the loop is
9315 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9316 Alan Hayward <alan.hayward@arm.com>
9317 David Sherwood <david.sherwood@arm.com>
9319 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
9320 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
9321 (dump_groups): Update accordingly.
9322 (iv_use::mem_type): New member variable.
9323 (address_p): New function.
9324 (record_use): Add a mem_type argument and initialize the new
9326 (record_group_use): Add a mem_type argument. Use address_p.
9327 Remove obsolete null checks of base_object. Update call to record_use.
9328 (find_interesting_uses_op): Update call to record_group_use.
9329 (find_interesting_uses_cond): Likewise.
9330 (find_interesting_uses_address): Likewise.
9331 (get_mem_type_for_internal_fn): New function.
9332 (find_address_like_use): Likewise.
9333 (find_interesting_uses_stmt): Try find_address_like_use before
9334 calling find_interesting_uses_op.
9335 (addr_offset_valid_p): Use the iv mem_type field as the type
9336 of the addressed memory.
9337 (add_autoinc_candidates): Likewise.
9338 (get_address_cost): Likewise.
9339 (split_small_address_groups_p): Use address_p.
9340 (split_address_groups): Likewise.
9341 (add_iv_candidate_for_use): Likewise.
9342 (autoinc_possible_for_pair): Likewise.
9343 (rewrite_groups): Likewise.
9344 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
9345 (determine_group_iv_cost): Update after split of USE_ADDRESS.
9346 (get_alias_ptr_type_for_ptr_address): New function.
9347 (rewrite_use_address): Rewrite address uses in calls that were
9348 identified by find_address_like_use.
9350 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9351 Alan Hayward <alan.hayward@arm.com>
9352 David Sherwood <david.sherwood@arm.com>
9354 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
9356 * gimple-expr.h (is_gimple_addressable: Likewise.
9357 * gimple-expr.c (is_gimple_address): Likewise.
9358 * internal-fn.c (expand_call_mem_ref): New function.
9359 (expand_mask_load_optab_fn): Use it.
9360 (expand_mask_store_optab_fn): Likewise.
9362 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9363 Alan Hayward <alan.hayward@arm.com>
9364 David Sherwood <david.sherwood@arm.com>
9366 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
9367 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
9368 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
9369 (cond_umax@var{mode}): Document.
9370 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
9371 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
9372 (cond_umin_optab, cond_umax_optab): New optabs.
9373 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
9374 (COND_IOR, COND_XOR): New internal functions.
9375 * internal-fn.h (get_conditional_internal_fn): Declare.
9376 * internal-fn.c (cond_binary_direct): New macro.
9377 (expand_cond_binary_optab_fn): Likewise.
9378 (direct_cond_binary_optab_supported_p): Likewise.
9379 (get_conditional_internal_fn): New function.
9380 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
9381 Cope with reduction statements that are vectorized as calls rather
9383 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
9384 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
9385 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
9386 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
9387 (UNSPEC_COND_EOR): New unspecs.
9388 (optab): Add mappings for them.
9389 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
9390 (sve_int_op, sve_fp_op): New int attributes.
9392 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9393 Alan Hayward <alan.hayward@arm.com>
9394 David Sherwood <david.sherwood@arm.com>
9396 * optabs.def (while_ult_optab): New optab.
9397 * doc/md.texi (while_ult@var{m}@var{n}): Document.
9398 * internal-fn.def (WHILE_ULT): New internal function.
9399 * internal-fn.h (direct_internal_fn_supported_p): New override
9400 that takes two types as argument.
9401 * internal-fn.c (while_direct): New macro.
9402 (expand_while_optab_fn): New function.
9403 (convert_optab_supported_p): Likewise.
9404 (direct_while_optab_supported_p): New macro.
9405 * wide-int.h (wi::udiv_ceil): New function.
9406 * tree-vectorizer.h (rgroup_masks): New structure.
9407 (vec_loop_masks): New typedef.
9408 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
9410 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
9411 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
9412 (vect_max_vf): New function.
9413 (slpeel_make_loop_iterate_ntimes): Delete.
9414 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
9415 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
9416 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
9417 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
9418 internal-fn.h, stor-layout.h and optabs-query.h.
9419 (vect_set_loop_mask): New function.
9420 (add_preheader_seq): Likewise.
9421 (add_header_seq): Likewise.
9422 (interleave_supported_p): Likewise.
9423 (vect_maybe_permute_loop_masks): Likewise.
9424 (vect_set_loop_masks_directly): Likewise.
9425 (vect_set_loop_condition_masked): Likewise.
9426 (vect_set_loop_condition_unmasked): New function, split out from
9427 slpeel_make_loop_iterate_ntimes.
9428 (slpeel_make_loop_iterate_ntimes): Rename to..
9429 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
9430 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
9431 (vect_do_peeling): Update call accordingly.
9432 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
9434 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9435 mask_compare_type, can_fully_mask_p and fully_masked_p.
9436 (release_vec_loop_masks): New function.
9437 (_loop_vec_info): Use it to free the loop masks.
9438 (can_produce_all_loop_masks_p): New function.
9439 (vect_get_max_nscalars_per_iter): Likewise.
9440 (vect_verify_full_masking): Likewise.
9441 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
9442 retries, and free the mask rgroups before retrying. Check loop-wide
9443 reasons for disallowing fully-masked loops. Make the final decision
9444 about whether use a fully-masked loop or not.
9445 (vect_estimate_min_profitable_iters): Do not assume that peeling
9446 for the number of iterations will be needed for fully-masked loops.
9447 (vectorizable_reduction): Disable fully-masked loops.
9448 (vectorizable_live_operation): Likewise.
9449 (vect_halve_mask_nunits): New function.
9450 (vect_double_mask_nunits): Likewise.
9451 (vect_record_loop_mask): Likewise.
9452 (vect_get_loop_mask): Likewise.
9453 (vect_transform_loop): Handle the case in which the final loop
9454 iteration might handle a partial vector. Call vect_set_loop_condition
9455 instead of slpeel_make_loop_iterate_ntimes.
9456 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
9457 (check_load_store_masking): New function.
9458 (prepare_load_store_mask): Likewise.
9459 (vectorizable_store): Handle fully-masked loops.
9460 (vectorizable_load): Likewise.
9461 (supportable_widening_operation): Use vect_halve_mask_nunits for
9463 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
9464 (vect_gen_while): New function.
9465 * config/aarch64/aarch64.md (umax<mode>3): New expander.
9466 (aarch64_uqdec<mode>): New insn.
9468 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9469 Alan Hayward <alan.hayward@arm.com>
9470 David Sherwood <david.sherwood@arm.com>
9472 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
9473 (reduc_xor_scal_optab): New optabs.
9474 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
9475 (reduc_xor_scal_@var{m}): Document.
9476 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
9477 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
9479 * fold-const-call.c (fold_const_call): Handle them.
9480 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
9481 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
9482 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
9483 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
9484 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
9485 (UNSPEC_XORV): New unspecs.
9486 (optab): Add entries for them.
9487 (BITWISEV): New int iterator.
9488 (bit_reduc_op): New int attributes.
9490 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9491 Alan Hayward <alan.hayward@arm.com>
9492 David Sherwood <david.sherwood@arm.com>
9494 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
9495 * internal-fn.def (VEC_SHL_INSERT): New internal function.
9496 * optabs.def (vec_shl_insert_optab): New optab.
9497 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
9498 (duplicate_and_interleave): Likewise.
9499 * tree-vect-loop.c: Include internal-fn.h.
9500 (neutral_op_for_slp_reduction): New function, split out from
9501 get_initial_defs_for_reduction.
9502 (get_initial_def_for_reduction): Handle option 2 for variable-length
9503 vectors by loading the neutral value into a vector and then shifting
9504 the initial value into element 0.
9505 (get_initial_defs_for_reduction): Replace the code argument with
9506 the neutral value calculated by neutral_op_for_slp_reduction.
9507 Use gimple_build_vector for constant-length vectors.
9508 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
9509 but the first group_size elements have a neutral value.
9510 Use duplicate_and_interleave otherwise.
9511 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
9512 Update call to get_initial_defs_for_reduction. Handle SLP
9513 reductions for variable-length vectors by creating one vector
9514 result for each scalar result, with the elements associated
9515 with other scalar results stubbed out with the neutral value.
9516 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
9517 Require IFN_VEC_SHL_INSERT for double reductions on
9518 variable-length vectors, or SLP reductions that have
9519 a neutral value. Require can_duplicate_and_interleave_p
9520 support for variable-length unchained SLP reductions if there
9521 is no neutral value, such as for MIN/MAX reductions. Also require
9522 the number of vector elements to be a multiple of the number of
9523 SLP statements when doing variable-length unchained SLP reductions.
9524 Update call to vect_create_epilog_for_reduction.
9525 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
9526 and remove initial values.
9527 (duplicate_and_interleave): Make public.
9528 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
9529 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
9531 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9532 Alan Hayward <alan.hayward@arm.com>
9533 David Sherwood <david.sherwood@arm.com>
9535 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
9536 (can_duplicate_and_interleave_p): New function.
9537 (vect_get_and_check_slp_defs): Take the vector of statements
9538 rather than just the current one. Remove excess parentheses.
9539 Restriction rejectinon of vect_constant_def and vect_external_def
9540 for variable-length vectors to boolean types, or types for which
9541 can_duplicate_and_interleave_p is false.
9542 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
9543 (duplicate_and_interleave): New function.
9544 (vect_get_constant_vectors): Use gimple_build_vector for
9545 constant-length vectors and suitable variable-length constant
9546 vectors. Use duplicate_and_interleave for other variable-length
9547 vectors. Don't defer the update when inserting new statements.
9549 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9550 Alan Hayward <alan.hayward@arm.com>
9551 David Sherwood <david.sherwood@arm.com>
9553 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
9554 min_profitable_iters doesn't go negative.
9556 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9557 Alan Hayward <alan.hayward@arm.com>
9558 David Sherwood <david.sherwood@arm.com>
9560 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
9561 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
9562 * optabs.def (vec_mask_load_lanes_optab): New optab.
9563 (vec_mask_store_lanes_optab): Likewise.
9564 * internal-fn.def (MASK_LOAD_LANES): New internal function.
9565 (MASK_STORE_LANES): Likewise.
9566 * internal-fn.c (mask_load_lanes_direct): New macro.
9567 (mask_store_lanes_direct): Likewise.
9568 (expand_mask_load_optab_fn): Handle masked operations.
9569 (expand_mask_load_lanes_optab_fn): New macro.
9570 (expand_mask_store_optab_fn): Handle masked operations.
9571 (expand_mask_store_lanes_optab_fn): New macro.
9572 (direct_mask_load_lanes_optab_supported_p): Likewise.
9573 (direct_mask_store_lanes_optab_supported_p): Likewise.
9574 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
9576 (vect_load_lanes_supported): Likewise.
9577 * tree-vect-data-refs.c (strip_conversion): New function.
9578 (can_group_stmts_p): Likewise.
9579 (vect_analyze_data_ref_accesses): Use it instead of checking
9580 for a pair of assignments.
9581 (vect_store_lanes_supported): Take a masked_p parameter.
9582 (vect_load_lanes_supported): Likewise.
9583 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
9584 vect_store_lanes_supported and vect_load_lanes_supported.
9585 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
9586 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
9587 parameter. Don't allow gaps for masked accesses.
9588 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
9589 and vect_load_lanes_supported.
9590 (get_load_store_type): Take a masked_p parameter and update
9591 call to get_group_load_store_type.
9592 (vectorizable_store): Update call to get_load_store_type.
9593 Handle IFN_MASK_STORE_LANES.
9594 (vectorizable_load): Update call to get_load_store_type.
9595 Handle IFN_MASK_LOAD_LANES.
9597 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9598 Alan Hayward <alan.hayward@arm.com>
9599 David Sherwood <david.sherwood@arm.com>
9601 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
9603 * config/aarch64/aarch64-protos.h
9604 (aarch64_sve_struct_memory_operand_p): Declare.
9605 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
9606 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
9607 (VPRED, vpred): Handle SVE structure modes.
9608 * config/aarch64/constraints.md (Utx): New constraint.
9609 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
9610 (aarch64_sve_struct_nonimmediate_operand): New predicates.
9611 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
9612 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
9613 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
9614 structure modes. Split into pieces after RA.
9615 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
9616 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
9618 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
9619 SVE structure modes.
9620 (aarch64_classify_address): Likewise.
9621 (sizetochar): Move earlier in file.
9622 (aarch64_print_operand): Handle SVE register lists.
9623 (aarch64_array_mode): New function.
9624 (aarch64_sve_struct_memory_operand_p): Likewise.
9625 (TARGET_ARRAY_MODE): Redefine.
9627 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9628 Alan Hayward <alan.hayward@arm.com>
9629 David Sherwood <david.sherwood@arm.com>
9631 * target.def (array_mode): New target hook.
9632 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
9633 * doc/tm.texi: Regenerate.
9634 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
9635 * hooks.c (hook_optmode_mode_uhwi_none): New function.
9636 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
9638 * stor-layout.c (mode_for_array): Likewise. Support polynomial
9641 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9642 Alan Hayward <alan.hayward@arm.com>
9643 David Sherwood <david.sherwood@arm.com>
9645 * fold-const.c (fold_binary_loc): Check the argument types
9646 rather than the result type when testing for a vector operation.
9648 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9650 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
9651 * doc/tm.texi: Regenerate.
9653 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9654 Alan Hayward <alan.hayward@arm.com>
9655 David Sherwood <david.sherwood@arm.com>
9657 * doc/invoke.texi (-msve-vector-bits=): Document new option.
9658 (sve): Document new AArch64 extension.
9659 * doc/md.texi (w): Extend the description of the AArch64
9660 constraint to include SVE vectors.
9661 (Upl, Upa): Document new AArch64 predicate constraints.
9662 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
9664 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
9665 (msve-vector-bits=): New option.
9666 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
9667 SVE when these are disabled.
9668 (sve): New extension.
9669 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
9670 modes. Adjust their number of units based on aarch64_sve_vg.
9671 (MAX_BITSIZE_MODE_ANY_MODE): Define.
9672 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
9673 aarch64_addr_query_type.
9674 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
9675 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
9676 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
9677 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
9678 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
9679 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
9680 (aarch64_simd_imm_zero_p): Delete.
9681 (aarch64_check_zero_based_sve_index_immediate): Declare.
9682 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9683 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9684 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9685 (aarch64_sve_float_mul_immediate_p): Likewise.
9686 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9688 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
9689 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
9690 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
9691 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
9692 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
9693 (aarch64_regmode_natural_size): Likewise.
9694 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
9695 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
9697 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
9698 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
9699 for VG and the SVE predicate registers.
9700 (V_ALIASES): Add a "z"-prefixed alias.
9701 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
9702 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
9703 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
9704 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
9705 (REG_CLASS_NAMES): Add entries for them.
9706 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
9707 and the predicate registers.
9708 (aarch64_sve_vg): Declare.
9709 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
9710 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
9711 (REGMODE_NATURAL_SIZE): Define.
9712 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
9714 * config/aarch64/aarch64.c: Include cfgrtl.h.
9715 (simd_immediate_info): Add a constructor for series vectors,
9716 and an associated step field.
9717 (aarch64_sve_vg): New variable.
9718 (aarch64_dbx_register_number): Handle VG and the predicate registers.
9719 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
9720 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
9721 (VEC_ANY_DATA, VEC_STRUCT): New constants.
9722 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
9723 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
9724 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
9725 (aarch64_get_mask_mode): New functions.
9726 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
9727 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
9728 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
9729 predicate modes and predicate registers. Explicitly restrict
9730 GPRs to modes of 16 bytes or smaller. Only allow FP registers
9731 to store a vector mode if it is recognized by
9732 aarch64_classify_vector_mode.
9733 (aarch64_regmode_natural_size): New function.
9734 (aarch64_hard_regno_caller_save_mode): Return the original mode
9736 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
9737 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
9738 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
9739 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
9741 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
9742 does not overlap dest if the function is frame-related. Handle
9744 (aarch64_split_add_offset): New function.
9745 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
9746 them aarch64_add_offset.
9747 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
9748 and update call to aarch64_sub_sp.
9749 (aarch64_add_cfa_expression): New function.
9750 (aarch64_expand_prologue): Pass extra temporary registers to the
9751 functions above. Handle the case in which we need to emit new
9752 DW_CFA_expressions for registers that were originally saved
9753 relative to the stack pointer, but now have to be expressed
9754 relative to the frame pointer.
9755 (aarch64_output_mi_thunk): Pass extra temporary registers to the
9757 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
9758 IP0 and IP1 values for SVE frames.
9759 (aarch64_expand_vec_series): New function.
9760 (aarch64_expand_sve_widened_duplicate): Likewise.
9761 (aarch64_expand_sve_const_vector): Likewise.
9762 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
9763 Handle SVE constants. Use emit_move_insn to move a force_const_mem
9764 into the register, rather than emitting a SET directly.
9765 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
9766 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
9767 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
9768 (offset_9bit_signed_scaled_p): New functions.
9769 (aarch64_replicate_bitmask_imm): New function.
9770 (aarch64_bitmask_imm): Use it.
9771 (aarch64_cannot_force_const_mem): Reject expressions involving
9772 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
9773 (aarch64_classify_index): Handle SVE indices, by requiring
9774 a plain register index with a scale that matches the element size.
9775 (aarch64_classify_address): Handle SVE addresses. Assert that
9776 the mode of the address is VOIDmode or an integer mode.
9777 Update call to aarch64_classify_symbol.
9778 (aarch64_classify_symbolic_expression): Update call to
9779 aarch64_classify_symbol.
9780 (aarch64_const_vec_all_in_range_p): New function.
9781 (aarch64_print_vector_float_operand): Likewise.
9782 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
9783 "vN" for FP registers with SVE modes. Handle (const ...) vectors
9784 and the FP immediates 1.0 and 0.5.
9785 (aarch64_print_address_internal): Handle SVE addresses.
9786 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
9787 (aarch64_regno_regclass): Handle predicate registers.
9788 (aarch64_secondary_reload): Handle big-endian reloads of SVE
9790 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
9791 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
9792 (aarch64_convert_sve_vector_bits): New function.
9793 (aarch64_override_options): Use it to handle -msve-vector-bits=.
9794 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
9796 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
9797 Handle SVE vector and predicate modes. Accept VL-based constants
9798 that need only one temporary register, and VL offsets that require
9799 no temporary registers.
9800 (aarch64_conditional_register_usage): Mark the predicate registers
9801 as fixed if SVE isn't available.
9802 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
9803 Return true for SVE vector and predicate modes.
9804 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
9805 rather than an unsigned int. Handle SVE modes.
9806 (aarch64_preferred_simd_mode): Update call accordingly. Handle
9808 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
9810 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
9811 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
9812 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
9813 (aarch64_sve_float_mul_immediate_p): New functions.
9814 (aarch64_sve_valid_immediate): New function.
9815 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
9816 Explicitly reject structure modes. Check for INDEX constants.
9817 Handle PTRUE and PFALSE constants.
9818 (aarch64_check_zero_based_sve_index_immediate): New function.
9819 (aarch64_simd_imm_zero_p): Delete.
9820 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
9821 vector modes. Accept constants in the range of CNT[BHWD].
9822 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
9823 ask for an Advanced SIMD mode.
9824 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
9825 (aarch64_simd_vector_alignment): Handle SVE predicates.
9826 (aarch64_vectorize_preferred_vector_alignment): New function.
9827 (aarch64_simd_vector_alignment_reachable): Use it instead of
9829 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
9830 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
9832 (MAX_VECT_LEN): Delete.
9833 (expand_vec_perm_d): Add a vec_flags field.
9834 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
9835 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
9836 (aarch64_evpc_ext): Don't apply a big-endian lane correction
9838 (aarch64_evpc_rev): Rename to...
9839 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
9840 (aarch64_evpc_rev_global): New function.
9841 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
9842 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
9844 (aarch64_evpc_sve_tbl): New function.
9845 (aarch64_expand_vec_perm_const_1): Update after rename of
9846 aarch64_evpc_rev. Handle SVE permutes too, trying
9847 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
9848 than aarch64_evpc_tbl.
9849 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
9850 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
9851 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
9852 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
9853 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
9854 (aarch64_expand_sve_vcond): New functions.
9855 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
9856 of aarch64_vector_mode_p.
9857 (aarch64_dwarf_poly_indeterminate_value): New function.
9858 (aarch64_compute_pressure_classes): Likewise.
9859 (aarch64_can_change_mode_class): Likewise.
9860 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
9861 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
9862 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
9863 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
9864 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
9865 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
9866 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
9867 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
9869 (Dn, Dl, Dr): Accept const as well as const_vector.
9870 (Dz): Likewise. Compare against CONST0_RTX.
9871 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
9872 of "vector" where appropriate.
9873 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
9874 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
9875 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
9876 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
9877 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
9878 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
9879 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
9880 (v_int_equiv): Extend to SVE modes.
9881 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
9883 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
9884 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
9885 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
9886 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
9887 (SVE_COND_FP_CMP): New int iterators.
9888 (perm_hilo): Handle the new unpack unspecs.
9889 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
9891 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
9892 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
9893 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
9894 (aarch64_equality_operator, aarch64_constant_vector_operand)
9895 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
9896 (aarch64_sve_nonimmediate_operand): Likewise.
9897 (aarch64_sve_general_operand): Likewise.
9898 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
9899 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
9900 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
9901 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
9902 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
9903 (aarch64_sve_float_arith_immediate): Likewise.
9904 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
9905 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
9906 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
9907 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
9908 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
9909 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
9910 (aarch64_sve_float_arith_operand): Likewise.
9911 (aarch64_sve_float_arith_with_sub_operand): Likewise.
9912 (aarch64_sve_float_mul_operand): Likewise.
9913 (aarch64_sve_vec_perm_operand): Likewise.
9914 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
9915 (aarch64_mov_operand): Accept const_poly_int and const_vector.
9916 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
9917 as well as const_vector.
9918 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
9919 in file. Use CONST0_RTX and CONSTM1_RTX.
9920 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
9921 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
9922 Use aarch64_simd_imm_zero.
9923 * config/aarch64/aarch64-sve.md: New file.
9924 * config/aarch64/aarch64.md: Include it.
9925 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
9926 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
9927 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
9928 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
9929 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
9930 (sve): New attribute.
9931 (enabled): Disable instructions with the sve attribute unless
9933 (movqi, movhi): Pass CONST_POLY_INT operaneds through
9934 aarch64_expand_mov_immediate.
9935 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
9936 CNT[BHSD] immediates.
9937 (movti): Split CONST_POLY_INT moves into two halves.
9938 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
9939 Split additions that need a temporary here if the destination
9940 is the stack pointer.
9941 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
9942 (*add<mode>3_poly_1): New instruction.
9943 (set_clobber_cc): New expander.
9945 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9947 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
9948 parameter and use it instead of GET_MODE_SIZE (innermode). Use
9949 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
9950 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
9951 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
9952 Change innermode from fixed_mode_size to machine_mode.
9953 (simplify_subreg): Update call accordingly. Handle a constant-sized
9954 subreg of a variable-length CONST_VECTOR.
9956 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
9957 Alan Hayward <alan.hayward@arm.com>
9958 David Sherwood <david.sherwood@arm.com>
9960 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
9961 (add_offset_to_base): New function, split out from...
9962 (create_mem_ref): ...here. When handling a scale other than 1,
9963 check first whether the address is valid without the offset.
9964 Add it into the base if so, leaving the index and scale as-is.
9966 2018-01-12 Jakub Jelinek <jakub@redhat.com>
9969 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
9970 fold_for_warn before checking if arg2 is INTEGER_CST.
9972 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
9974 * config/rs6000/predicates.md (load_multiple_operation): Delete.
9975 (store_multiple_operation): Delete.
9976 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
9977 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
9978 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
9979 guarded by TARGET_STRING.
9980 (rs6000_output_load_multiple): Delete.
9981 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
9982 OPTION_MASK_STRING / TARGET_STRING handling.
9983 (print_operand) <'N', 'O'>: Add comment that these are unused now.
9984 (const rs6000_opt_masks) <"string">: Change mask to 0.
9985 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
9986 (MASK_STRING): Delete.
9987 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
9989 (load_multiple): Delete.
9996 (store_multiple): Delete.
10003 (movmemsi_8reg): Delete.
10004 (corresponding unnamed define_insn): Delete.
10005 (movmemsi_6reg): Delete.
10006 (corresponding unnamed define_insn): Delete.
10007 (movmemsi_4reg): Delete.
10008 (corresponding unnamed define_insn): Delete.
10009 (movmemsi_2reg): Delete.
10010 (corresponding unnamed define_insn): Delete.
10011 (movmemsi_1reg): Delete.
10012 (corresponding unnamed define_insn): Delete.
10013 * config/rs6000/rs6000.opt (mno-string): New.
10014 (mstring): Replace by deprecation warning stub.
10015 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
10017 2018-01-12 Jakub Jelinek <jakub@redhat.com>
10019 * regrename.c (regrename_do_replace): If replacing the same
10020 reg multiple times, try to reuse last created gen_raw_REG.
10023 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
10024 main to workaround a bug in GDB.
10026 2018-01-12 Tom de Vries <tom@codesourcery.com>
10029 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
10031 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
10033 PR rtl-optimization/80481
10034 * ira-color.c (get_cap_member): New function.
10035 (allocnos_conflict_by_live_ranges_p): Use it.
10036 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
10037 (setup_slot_coalesced_allocno_live_ranges): Ditto.
10039 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
10042 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
10043 (*saddl_se_1): Ditto.
10044 (*ssubsi_1): Ditto.
10045 (*ssubl_se_1): Ditto.
10047 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10049 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
10050 rather than wi::to_widest for DR_INITs.
10051 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
10052 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
10053 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
10055 (vect_analyze_group_access_1): Note that here.
10057 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10059 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
10060 polynomial type sizes.
10062 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
10064 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
10065 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
10066 (gimple_add_tmp_var): Likewise.
10068 2018-01-12 Martin Liska <mliska@suse.cz>
10070 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
10071 (gimple_alloc_sizes): Likewise.
10072 (dump_gimple_statistics): Use PRIu64 in printf format.
10073 * gimple.h: Change uint64_t to int.
10075 2018-01-12 Martin Liska <mliska@suse.cz>
10077 * tree-core.h: Use uint64_t instead of int.
10078 * tree.c (tree_node_counts): Likewise.
10079 (tree_node_sizes): Likewise.
10080 (dump_tree_statistics): Use PRIu64 in printf format.
10082 2018-01-12 Martin Liska <mliska@suse.cz>
10084 * Makefile.in: As qsort_chk is implemented in vec.c, add
10085 vec.o to linkage of gencfn-macros.
10086 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
10087 passing the info to record_node_allocation_statistics.
10088 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
10090 * ggc-common.c (struct ggc_usage): Add operator== and use
10091 it in operator< and compare function.
10092 * mem-stats.h (struct mem_usage): Likewise.
10093 * vec.c (struct vec_usage): Remove operator< and compare
10094 function. Can be simply inherited.
10096 2018-01-12 Martin Jambor <mjambor@suse.cz>
10099 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
10100 * tree-ssa-math-opts.c: Include domwalk.h.
10101 (convert_mult_to_fma_1): New function.
10102 (fma_transformation_info): New type.
10103 (fma_deferring_state): Likewise.
10104 (cancel_fma_deferring): New function.
10105 (result_of_phi): Likewise.
10106 (last_fma_candidate_feeds_initial_phi): Likewise.
10107 (convert_mult_to_fma): Added deferring logic, split actual
10108 transformation to convert_mult_to_fma_1.
10109 (math_opts_dom_walker): New type.
10110 (math_opts_dom_walker::after_dom_children): New method, body moved
10111 here from pass_optimize_widening_mul::execute, added deferring logic
10113 (pass_optimize_widening_mul::execute): Moved most of code to
10114 math_opts_dom_walker::after_dom_children.
10115 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
10116 * config/i386/i386.c (ix86_option_override_internal): Added
10117 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
10119 2018-01-12 Richard Biener <rguenther@suse.de>
10122 * dwarf2out.c (gen_variable_die): Do not reset old_die for
10123 inline instance vars.
10125 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
10128 * config/rx/rx.c (rx_is_restricted_memory_address):
10129 Handle SUBREG case.
10131 2018-01-12 Richard Biener <rguenther@suse.de>
10133 PR tree-optimization/80846
10134 * target.def (split_reduction): New target hook.
10135 * targhooks.c (default_split_reduction): New function.
10136 * targhooks.h (default_split_reduction): Declare.
10137 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
10138 target requests first reduce vectors by combining low and high
10140 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
10141 (get_vectype_for_scalar_type_and_size): Export.
10142 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
10143 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
10144 * doc/tm.texi: Regenerate.
10145 * config/i386/i386.c (ix86_split_reduction): Implement
10146 TARGET_VECTORIZE_SPLIT_REDUCTION.
10148 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10151 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
10152 in PIC mode except for TARGET_VXWORKS_RTP.
10153 * config/sparc/sparc.c: Include cfgrtl.h.
10154 (TARGET_INIT_PIC_REG): Define.
10155 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
10156 (sparc_pic_register_p): New predicate.
10157 (sparc_legitimate_address_p): Use it.
10158 (sparc_legitimize_pic_address): Likewise.
10159 (sparc_delegitimize_address): Likewise.
10160 (sparc_mode_dependent_address_p): Likewise.
10161 (gen_load_pcrel_sym): Remove 4th parameter.
10162 (load_got_register): Adjust call to above. Remove obsolete stuff.
10163 (sparc_expand_prologue): Do not call load_got_register here.
10164 (sparc_flat_expand_prologue): Likewise.
10165 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
10166 (sparc_use_pseudo_pic_reg): New function.
10167 (sparc_init_pic_reg): Likewise.
10168 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10169 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
10171 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
10173 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
10174 Add item for branch_cost.
10176 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10178 PR rtl-optimization/83565
10179 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
10180 not extend the result to a larger mode for rotate operations.
10181 (num_sign_bit_copies1): Likewise.
10183 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10186 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
10188 Use values-Xc.o for -pedantic.
10189 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
10191 2018-01-12 Martin Liska <mliska@suse.cz>
10194 * ipa-devirt.c (final_warning_record::grow_type_warnings):
10196 (possible_polymorphic_call_targets): Use it.
10197 (ipa_devirt): Likewise.
10199 2018-01-12 Martin Liska <mliska@suse.cz>
10201 * profile-count.h (enum profile_quality): Use 0 as invalid
10202 enum value of profile_quality.
10204 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
10206 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
10207 -mext-string options.
10209 2018-01-12 Richard Biener <rguenther@suse.de>
10211 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
10212 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
10213 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
10215 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
10217 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
10219 * configure.ac (--with-long-double-format): Add support for the
10220 configuration option to change the default long double format on
10222 * config.gcc (powerpc*-linux*-*): Likewise.
10223 * configure: Regenerate.
10224 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
10225 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
10226 used without modification.
10228 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10230 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
10231 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
10232 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
10233 MISC_BUILTIN_SPEC_BARRIER.
10234 (rs6000_init_builtins): Likewise.
10235 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
10237 (speculation_barrier): New define_insn.
10238 * doc/extend.texi: Document __builtin_speculation_barrier.
10240 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10243 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
10244 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
10245 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
10247 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
10248 integral modes instead of "ss" and "sd".
10249 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
10250 vectors with 32-bit and 64-bit elements.
10251 (vecdupssescalarmodesuffix): New mode attribute.
10252 (vec_dup<mode>): Use it.
10254 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
10257 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
10258 frame if argument is passed on stack.
10260 2018-01-11 Jakub Jelinek <jakub@redhat.com>
10263 * ree.c (combine_reaching_defs): Optimize also
10264 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
10265 reg2=any_extend(exp); reg1=reg2;, formatting fix.
10267 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10269 PR middle-end/83189
10270 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
10272 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
10274 PR middle-end/83718
10275 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
10276 after they are computed.
10278 2018-01-11 Bin Cheng <bin.cheng@arm.com>
10280 PR tree-optimization/83695
10281 * gimple-loop-linterchange.cc
10282 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
10283 reset cached scev information after interchange.
10284 (pass_linterchange::execute): Remove call to scev_reset_htab.
10286 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10288 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
10289 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
10290 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
10291 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
10292 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
10293 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
10294 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
10295 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
10296 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
10297 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
10298 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
10299 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
10300 (V_lane_reg): Likewise.
10301 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
10303 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
10304 (vfmal_lane_low<mode>_intrinsic,
10305 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
10306 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
10307 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
10308 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
10309 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
10310 vfmsl_lane_high<mode>_intrinsic): New define_insns.
10312 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10314 * config/arm/arm-cpus.in (fp16fml): New feature.
10315 (ALL_SIMD): Add fp16fml.
10316 (armv8.2-a): Add fp16fml as an option.
10317 (armv8.3-a): Likewise.
10318 (armv8.4-a): Add fp16fml as part of fp16.
10319 * config/arm/arm.h (TARGET_FP16FML): Define.
10320 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
10322 * config/arm/arm-modes.def (V2HF): Define.
10323 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
10324 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
10325 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
10326 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
10327 vfmsl_low, vfmsl_high): New set of builtins.
10328 * config/arm/iterators.md (PLUSMINUS): New code iterator.
10329 (vfml_op): New code attribute.
10330 (VFMLHALVES): New int iterator.
10331 (VFML, VFMLSEL): New mode attributes.
10332 (V_reg): Define mapping for V2HF.
10333 (V_hi, V_lo): New mode attributes.
10334 (VF_constraint): Likewise.
10335 (vfml_half, vfml_half_selector): New int attributes.
10336 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
10338 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
10339 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
10341 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
10342 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
10343 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
10344 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
10346 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
10347 Document new effective target and option set.
10349 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10351 * config/arm/arm-cpus.in (armv8_4): New feature.
10352 (ARMv8_4a): New fgroup.
10353 (armv8.4-a): New arch.
10354 * config/arm/arm-tables.opt: Regenerate.
10355 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
10356 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
10357 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
10358 Add matching rules for -march=armv8.4-a and extensions.
10359 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
10361 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
10364 * config/rx/rx.md (BW): New mode attribute.
10365 (sync_lock_test_and_setsi): Add mode suffix to insn output.
10367 2018-01-11 Richard Biener <rguenther@suse.de>
10369 PR tree-optimization/83435
10370 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
10371 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
10372 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
10374 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10375 Alan Hayward <alan.hayward@arm.com>
10376 David Sherwood <david.sherwood@arm.com>
10378 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
10380 (aarch64_classify_address): Initialize it. Track polynomial offsets.
10381 (aarch64_print_address_internal): Use it to check for a zero offset.
10383 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10384 Alan Hayward <alan.hayward@arm.com>
10385 David Sherwood <david.sherwood@arm.com>
10387 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
10388 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
10389 Return a poly_int64 rather than a HOST_WIDE_INT.
10390 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
10391 rather than a HOST_WIDE_INT.
10392 * config/aarch64/aarch64.h (aarch64_frame): Protect with
10393 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
10394 hard_fp_offset, frame_size, initial_adjust, callee_offset and
10395 final_offset from HOST_WIDE_INT to poly_int64.
10396 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
10397 to_constant when getting the number of units in an Advanced SIMD
10399 (aarch64_builtin_vectorized_function): Check for a constant number
10401 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
10403 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
10404 attribute instead of GET_MODE_NUNITS.
10405 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
10406 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
10407 GET_MODE_SIZE for fixed-size registers.
10408 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
10409 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
10410 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
10411 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
10412 (aarch64_print_operand, aarch64_print_address_internal)
10413 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
10414 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
10415 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
10416 Handle polynomial GET_MODE_SIZE.
10417 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
10418 wider than SImode without modification.
10419 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
10420 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
10421 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
10422 passing and returning SVE modes.
10423 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
10424 rather than GEN_INT.
10425 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
10426 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
10427 (aarch64_allocate_and_probe_stack_space): Likewise.
10428 (aarch64_layout_frame): Cope with polynomial offsets.
10429 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
10430 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
10431 polynomial offsets.
10432 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
10433 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
10434 poly_int64 rather than a HOST_WIDE_INT.
10435 (aarch64_get_separate_components, aarch64_process_components)
10436 (aarch64_expand_prologue, aarch64_expand_epilogue)
10437 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
10438 (aarch64_anchor_offset): New function, split out from...
10439 (aarch64_legitimize_address): ...here.
10440 (aarch64_builtin_vectorization_cost): Handle polynomial
10441 TYPE_VECTOR_SUBPARTS.
10442 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
10444 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
10445 number of elements from the PARALLEL rather than the mode.
10446 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
10447 rather than GET_MODE_BITSIZE.
10448 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
10449 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
10450 (aarch64_expand_vec_perm_const_1): Handle polynomial
10451 d->perm.length () and d->perm elements.
10452 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
10453 Apply to_constant to d->perm elements.
10454 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
10455 polynomial CONST_VECTOR_NUNITS.
10456 (aarch64_move_pointer): Take amount as a poly_int64 rather
10458 (aarch64_progress_pointer): Avoid temporary variable.
10459 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
10460 the mode attribute instead of GET_MODE.
10462 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10463 Alan Hayward <alan.hayward@arm.com>
10464 David Sherwood <david.sherwood@arm.com>
10466 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
10467 x exists before using it.
10468 (aarch64_add_constant_internal): Rename to...
10469 (aarch64_add_offset_1): ...this. Replace regnum with separate
10470 src and dest rtxes. Handle the case in which they're different,
10471 including when the offset is zero. Replace scratchreg with an rtx.
10472 Use 2 additions if there is no spare register into which we can
10473 move a 16-bit constant.
10474 (aarch64_add_constant): Delete.
10475 (aarch64_add_offset): Replace reg with separate src and dest
10476 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
10477 Use aarch64_add_offset_1.
10478 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
10479 an rtx rather than an int. Take the delta as a poly_int64
10480 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
10481 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
10482 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
10483 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
10484 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
10485 and aarch64_add_sp.
10486 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
10487 aarch64_add_constant.
10489 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10491 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
10492 Use scalar_float_mode.
10494 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10496 * config/aarch64/aarch64-simd.md
10497 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
10498 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
10499 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
10500 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
10501 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
10502 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
10503 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
10504 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
10505 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
10506 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
10508 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10511 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
10512 targ_options->x_arm_arch_string is non NULL.
10514 2018-01-11 Tamar Christina <tamar.christina@arm.com>
10516 * config/aarch64/aarch64.h
10517 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
10519 2018-01-11 Sudakshina Das <sudi.das@arm.com>
10522 * expmed.c (emit_store_flag_force): Swap if const op0
10523 and change VOIDmode to mode of op0.
10525 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
10527 PR rtl-optimization/83761
10528 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
10529 than bytes to mode_for_size.
10531 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10533 PR middle-end/83189
10534 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
10535 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
10538 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10540 PR middle-end/83575
10541 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
10542 when in layout mode.
10543 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
10544 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
10547 2018-01-10 Michael Collison <michael.collison@arm.com>
10549 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
10550 * config/aarch64/aarch64-option-extension.def: Add
10551 AARCH64_OPT_EXTENSION of 'fp16fml'.
10552 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10553 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
10554 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
10555 * config/aarch64/constraints.md (Ui7): New constraint.
10556 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
10557 (VFMLA_SEL_W): Ditto.
10560 (VFMLA16_LOW): New int iterator.
10561 (VFMLA16_HIGH): Ditto.
10562 (UNSPEC_FMLAL): New unspec.
10563 (UNSPEC_FMLSL): Ditto.
10564 (UNSPEC_FMLAL2): Ditto.
10565 (UNSPEC_FMLSL2): Ditto.
10566 (f16mac): New code attribute.
10567 * config/aarch64/aarch64-simd-builtins.def
10568 (aarch64_fmlal_lowv2sf): Ditto.
10569 (aarch64_fmlsl_lowv2sf): Ditto.
10570 (aarch64_fmlalq_lowv4sf): Ditto.
10571 (aarch64_fmlslq_lowv4sf): Ditto.
10572 (aarch64_fmlal_highv2sf): Ditto.
10573 (aarch64_fmlsl_highv2sf): Ditto.
10574 (aarch64_fmlalq_highv4sf): Ditto.
10575 (aarch64_fmlslq_highv4sf): Ditto.
10576 (aarch64_fmlal_lane_lowv2sf): Ditto.
10577 (aarch64_fmlsl_lane_lowv2sf): Ditto.
10578 (aarch64_fmlal_laneq_lowv2sf): Ditto.
10579 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
10580 (aarch64_fmlalq_lane_lowv4sf): Ditto.
10581 (aarch64_fmlsl_lane_lowv4sf): Ditto.
10582 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
10583 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
10584 (aarch64_fmlal_lane_highv2sf): Ditto.
10585 (aarch64_fmlsl_lane_highv2sf): Ditto.
10586 (aarch64_fmlal_laneq_highv2sf): Ditto.
10587 (aarch64_fmlsl_laneq_highv2sf): Ditto.
10588 (aarch64_fmlalq_lane_highv4sf): Ditto.
10589 (aarch64_fmlsl_lane_highv4sf): Ditto.
10590 (aarch64_fmlalq_laneq_highv4sf): Ditto.
10591 (aarch64_fmlsl_laneq_highv4sf): Ditto.
10592 * config/aarch64/aarch64-simd.md:
10593 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
10594 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10595 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
10596 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
10597 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
10598 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
10599 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
10600 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
10601 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
10602 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
10603 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
10604 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
10605 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
10606 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
10607 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
10608 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
10609 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
10610 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
10611 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
10612 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
10613 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
10614 (vfmlsl_low_u32): Ditto.
10615 (vfmlalq_low_u32): Ditto.
10616 (vfmlslq_low_u32): Ditto.
10617 (vfmlal_high_u32): Ditto.
10618 (vfmlsl_high_u32): Ditto.
10619 (vfmlalq_high_u32): Ditto.
10620 (vfmlslq_high_u32): Ditto.
10621 (vfmlal_lane_low_u32): Ditto.
10622 (vfmlsl_lane_low_u32): Ditto.
10623 (vfmlal_laneq_low_u32): Ditto.
10624 (vfmlsl_laneq_low_u32): Ditto.
10625 (vfmlalq_lane_low_u32): Ditto.
10626 (vfmlslq_lane_low_u32): Ditto.
10627 (vfmlalq_laneq_low_u32): Ditto.
10628 (vfmlslq_laneq_low_u32): Ditto.
10629 (vfmlal_lane_high_u32): Ditto.
10630 (vfmlsl_lane_high_u32): Ditto.
10631 (vfmlal_laneq_high_u32): Ditto.
10632 (vfmlsl_laneq_high_u32): Ditto.
10633 (vfmlalq_lane_high_u32): Ditto.
10634 (vfmlslq_lane_high_u32): Ditto.
10635 (vfmlalq_laneq_high_u32): Ditto.
10636 (vfmlslq_laneq_high_u32): Ditto.
10637 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
10638 (AARCH64_FL_FOR_ARCH8_4): New.
10639 (AARCH64_ISA_F16FML): New ISA flag.
10640 (TARGET_F16FML): New feature flag for fp16fml.
10641 (doc/invoke.texi): Document new fp16fml option.
10643 2018-01-10 Michael Collison <michael.collison@arm.com>
10645 * config/aarch64/aarch64-builtins.c:
10646 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
10647 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10648 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
10649 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
10650 (AARCH64_ISA_SHA3): New ISA flag.
10651 (TARGET_SHA3): New feature flag for sha3.
10652 * config/aarch64/iterators.md (sha512_op): New int attribute.
10653 (CRYPTO_SHA512): New int iterator.
10654 (UNSPEC_SHA512H): New unspec.
10655 (UNSPEC_SHA512H2): Ditto.
10656 (UNSPEC_SHA512SU0): Ditto.
10657 (UNSPEC_SHA512SU1): Ditto.
10658 * config/aarch64/aarch64-simd-builtins.def
10659 (aarch64_crypto_sha512hqv2di): New builtin.
10660 (aarch64_crypto_sha512h2qv2di): Ditto.
10661 (aarch64_crypto_sha512su0qv2di): Ditto.
10662 (aarch64_crypto_sha512su1qv2di): Ditto.
10663 (aarch64_eor3qv8hi): Ditto.
10664 (aarch64_rax1qv2di): Ditto.
10665 (aarch64_xarqv2di): Ditto.
10666 (aarch64_bcaxqv8hi): Ditto.
10667 * config/aarch64/aarch64-simd.md:
10668 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
10669 (aarch64_crypto_sha512su0qv2di): Ditto.
10670 (aarch64_crypto_sha512su1qv2di): Ditto.
10671 (aarch64_eor3qv8hi): Ditto.
10672 (aarch64_rax1qv2di): Ditto.
10673 (aarch64_xarqv2di): Ditto.
10674 (aarch64_bcaxqv8hi): Ditto.
10675 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
10676 (vsha512h2q_u64): Ditto.
10677 (vsha512su0q_u64): Ditto.
10678 (vsha512su1q_u64): Ditto.
10679 (veor3q_u16): Ditto.
10680 (vrax1q_u64): Ditto.
10681 (vxarq_u64): Ditto.
10682 (vbcaxq_u16): Ditto.
10683 * config/arm/types.md (crypto_sha512): New type attribute.
10684 (crypto_sha3): Ditto.
10685 (doc/invoke.texi): Document new sha3 option.
10687 2018-01-10 Michael Collison <michael.collison@arm.com>
10689 * config/aarch64/aarch64-builtins.c:
10690 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
10691 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10692 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
10693 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
10694 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
10695 (AARCH64_ISA_SM4): New ISA flag.
10696 (TARGET_SM4): New feature flag for sm4.
10697 * config/aarch64/aarch64-simd-builtins.def
10698 (aarch64_sm3ss1qv4si): Ditto.
10699 (aarch64_sm3tt1aq4si): Ditto.
10700 (aarch64_sm3tt1bq4si): Ditto.
10701 (aarch64_sm3tt2aq4si): Ditto.
10702 (aarch64_sm3tt2bq4si): Ditto.
10703 (aarch64_sm3partw1qv4si): Ditto.
10704 (aarch64_sm3partw2qv4si): Ditto.
10705 (aarch64_sm4eqv4si): Ditto.
10706 (aarch64_sm4ekeyqv4si): Ditto.
10707 * config/aarch64/aarch64-simd.md:
10708 (aarch64_sm3ss1qv4si): Ditto.
10709 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
10710 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
10711 (aarch64_sm4eqv4si): Ditto.
10712 (aarch64_sm4ekeyqv4si): Ditto.
10713 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
10714 (sm3part_op): Ditto.
10715 (CRYPTO_SM3TT): Ditto.
10716 (CRYPTO_SM3PART): Ditto.
10717 (UNSPEC_SM3SS1): New unspec.
10718 (UNSPEC_SM3TT1A): Ditto.
10719 (UNSPEC_SM3TT1B): Ditto.
10720 (UNSPEC_SM3TT2A): Ditto.
10721 (UNSPEC_SM3TT2B): Ditto.
10722 (UNSPEC_SM3PARTW1): Ditto.
10723 (UNSPEC_SM3PARTW2): Ditto.
10724 (UNSPEC_SM4E): Ditto.
10725 (UNSPEC_SM4EKEY): Ditto.
10726 * config/aarch64/constraints.md (Ui2): New constraint.
10727 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
10728 * config/arm/types.md (crypto_sm3): New type attribute.
10729 (crypto_sm4): Ditto.
10730 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
10731 (vsm3tt1aq_u32): Ditto.
10732 (vsm3tt1bq_u32): Ditto.
10733 (vsm3tt2aq_u32): Ditto.
10734 (vsm3tt2bq_u32): Ditto.
10735 (vsm3partw1q_u32): Ditto.
10736 (vsm3partw2q_u32): Ditto.
10737 (vsm4eq_u32): Ditto.
10738 (vsm4ekeyq_u32): Ditto.
10739 (doc/invoke.texi): Document new sm4 option.
10741 2018-01-10 Michael Collison <michael.collison@arm.com>
10743 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
10744 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
10745 (AARCH64_FL_FOR_ARCH8_4): New.
10746 (AARCH64_FL_V8_4): New flag.
10747 (doc/invoke.texi): Document new armv8.4-a option.
10749 2018-01-10 Michael Collison <michael.collison@arm.com>
10751 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
10752 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
10753 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
10754 * config/aarch64/aarch64-option-extension.def: Add
10755 AARCH64_OPT_EXTENSION of 'sha2'.
10756 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
10757 (crypto): Disable sha2 and aes if crypto disabled.
10758 (crypto): Enable aes and sha2 if enabled.
10759 (simd): Disable sha2 and aes if simd disabled.
10760 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
10762 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
10763 (TARGET_SHA2): New feature flag for sha2.
10764 (TARGET_AES): New feature flag for aes.
10765 * config/aarch64/aarch64-simd.md:
10766 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
10767 conditional on TARGET_AES.
10768 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
10769 (aarch64_crypto_sha1hsi): Make pattern conditional
10771 (aarch64_crypto_sha1hv4si): Ditto.
10772 (aarch64_be_crypto_sha1hv4si): Ditto.
10773 (aarch64_crypto_sha1su1v4si): Ditto.
10774 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
10775 (aarch64_crypto_sha1su0v4si): Ditto.
10776 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
10777 (aarch64_crypto_sha256su0v4si): Ditto.
10778 (aarch64_crypto_sha256su1v4si): Ditto.
10779 (doc/invoke.texi): Document new aes and sha2 options.
10781 2018-01-10 Martin Sebor <msebor@redhat.com>
10783 PR tree-optimization/83781
10784 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
10787 2018-01-11 Martin Sebor <msebor@gmail.com>
10788 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10790 PR tree-optimization/83501
10791 PR tree-optimization/81703
10793 * tree-ssa-strlen.c (get_string_cst): Rename...
10794 (get_string_len): ...to this. Handle global constants.
10795 (handle_char_store): Adjust.
10797 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
10798 Jim Wilson <jimw@sifive.com>
10800 * config/riscv/riscv-protos.h (riscv_output_return): New.
10801 * config/riscv/riscv.c (struct machine_function): New naked_p field.
10802 (riscv_attribute_table, riscv_output_return),
10803 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
10804 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
10805 (riscv_compute_frame_info): Only compute frame->mask if not a naked
10807 (riscv_expand_prologue): Add early return for naked function.
10808 (riscv_expand_epilogue): Likewise.
10809 (riscv_function_ok_for_sibcall): Return false for naked function.
10810 (riscv_set_current_function): New.
10811 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
10812 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
10813 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
10814 * doc/extend.texi (RISC-V Function Attributes): New.
10816 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
10818 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
10819 check for 128-bit long double before checking TCmode.
10820 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
10821 128-bit long doubles before checking TFmode or TCmode.
10822 (FLOAT128_IBM_P): Likewise.
10824 2018-01-10 Martin Sebor <msebor@redhat.com>
10826 PR tree-optimization/83671
10827 * builtins.c (c_strlen): Unconditionally return zero for the empty
10829 Use -Warray-bounds for warnings.
10830 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
10831 for non-constant array indices with COMPONENT_REF, arrays of
10832 arrays, and pointers to arrays.
10833 (gimple_fold_builtin_strlen): Determine and set length range for
10834 non-constant character arrays.
10836 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
10838 PR middle-end/81897
10839 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
10842 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
10844 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
10846 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10849 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
10850 VECTOR_MEM_ALTIVEC_OR_VSX_P.
10851 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
10852 indexed_or_indirect_operand predicate.
10853 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
10854 (*vsx_le_perm_load_v8hi): Likewise.
10855 (*vsx_le_perm_load_v16qi): Likewise.
10856 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
10857 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
10858 (*vsx_le_perm_store_v8hi): Likewise.
10859 (*vsx_le_perm_store_v16qi): Likewise.
10860 (eight unnamed splitters): Likewise.
10862 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
10864 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
10865 * config/rs6000/emmintrin.h: Likewise.
10866 * config/rs6000/mmintrin.h: Likewise.
10867 * config/rs6000/xmmintrin.h: Likewise.
10869 2018-01-10 David Malcolm <dmalcolm@redhat.com>
10872 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
10874 * tree.c (tree_nop_conversion): Return true for location wrapper
10876 (maybe_wrap_with_location): New function.
10877 (selftest::check_strip_nops): New function.
10878 (selftest::test_location_wrappers): New function.
10879 (selftest::tree_c_tests): Call it.
10880 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
10881 (maybe_wrap_with_location): New decl.
10882 (EXPR_LOCATION_WRAPPER_P): New macro.
10883 (location_wrapper_p): New inline function.
10884 (tree_strip_any_location_wrapper): New inline function.
10886 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
10889 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
10890 stack_realign_offset for the largest alignment of stack slot
10892 (ix86_find_max_used_stack_alignment): New function.
10893 (ix86_finalize_stack_frame_flags): Use it. Set
10894 max_used_stack_alignment if we don't realign stack.
10895 * config/i386/i386.h (machine_function): Add
10896 max_used_stack_alignment.
10898 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
10900 * config/arm/arm.opt (-mbranch-cost): New option.
10901 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
10904 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
10907 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
10908 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
10910 2018-01-10 Richard Biener <rguenther@suse.de>
10913 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
10914 early out so it also covers the case where we have a non-NULL
10917 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10919 PR tree-optimization/83753
10920 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
10921 for non-strided grouped accesses if the number of elements is 1.
10923 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
10926 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
10927 * i386.h (TARGET_USE_GATHER): Define.
10928 * x86-tune.def (X86_TUNE_USE_GATHER): New.
10930 2018-01-10 Martin Liska <mliska@suse.cz>
10933 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
10934 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
10936 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
10937 CLEANUP_NO_PARTITIONING is not set.
10939 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
10941 * doc/rtl.texi: Remove documentation of (const ...) wrappers
10942 for vectors, as a partial revert of r254296.
10943 * rtl.h (const_vec_p): Delete.
10944 (const_vec_duplicate_p): Don't test for vector CONSTs.
10945 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
10946 * expmed.c (make_tree): Likewise.
10949 * common.md (E, F): Use CONSTANT_P instead of checking for
10951 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
10952 checking for CONST_VECTOR.
10954 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
10956 PR middle-end/83575
10957 * predict.c (force_edge_cold): Handle in more sane way edges
10958 with no prediction.
10960 2018-01-09 Carl Love <cel@us.ibm.com>
10962 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
10964 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
10965 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
10966 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
10967 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
10968 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
10969 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
10970 * config/rs6000/rs6000-protos.h: Add extern defition for
10971 rs6000_generate_float2_double_code.
10972 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
10974 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
10975 (float2_v2df): Add define_expand.
10977 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
10980 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
10981 op_mode in the force_to_mode call.
10983 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
10985 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
10986 instead of checking each element individually.
10987 (aarch64_evpc_uzp): Likewise.
10988 (aarch64_evpc_zip): Likewise.
10989 (aarch64_evpc_ext): Likewise.
10990 (aarch64_evpc_rev): Likewise.
10991 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
10992 instead of checking each element individually. Return true without
10994 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
10995 whether all selected elements come from the same input, instead of
10996 checking each element individually. Remove calls to gen_rtx_REG,
10997 start_sequence and end_sequence and instead assert that no rtl is
11000 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11002 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
11003 order of HIGH and CONST checks.
11005 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
11007 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
11008 if the destination isn't an SSA_NAME.
11010 2018-01-09 Richard Biener <rguenther@suse.de>
11012 PR tree-optimization/83668
11013 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
11015 (canonicalize_loop_form): ... here, renamed from ...
11016 (canonicalize_loop_closed_ssa_form): ... this and amended to
11017 swap successor edges for loop exit blocks to make us use
11018 the RPO order we need for initial schedule generation.
11020 2018-01-09 Joseph Myers <joseph@codesourcery.com>
11022 PR tree-optimization/64811
11023 * match.pd: When optimizing comparisons with Inf, avoid
11024 introducing or losing exceptions from comparisons with NaN.
11026 2018-01-09 Martin Liska <mliska@suse.cz>
11029 * asan.c (shadow_mem_size): Add gcc_assert.
11031 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
11033 Don't save registers in main().
11036 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
11037 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
11038 * config/avr/avr.c (avr_set_current_function): Don't error if
11039 naked, OS_task or OS_main are specified at the same time.
11040 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
11042 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
11044 * common/config/avr/avr-common.c (avr_option_optimization_table):
11045 Switch on -mmain-is-OS_task for optimizing compilations.
11047 2018-01-09 Richard Biener <rguenther@suse.de>
11049 PR tree-optimization/83572
11050 * graphite.c: Include cfganal.h.
11051 (graphite_transform_loops): Connect infinite loops to exit
11052 and remove fake edges at the end.
11054 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11056 * ipa-inline.c (edge_badness): Revert accidental checkin.
11058 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
11061 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
11062 symbols; not inline clones.
11064 2018-01-09 Jakub Jelinek <jakub@redhat.com>
11067 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
11068 hard registers. Formatting fixes.
11070 PR preprocessor/83722
11071 * gcc.c (try_generate_repro): Pass
11072 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
11073 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
11076 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
11077 Kito Cheng <kito.cheng@gmail.com>
11079 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
11080 (riscv_leaf_function_p): Delete.
11081 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
11083 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11085 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
11087 (do_ifelse): New function.
11088 (do_isel): New function.
11089 (do_sub3): New function.
11090 (do_add3): New function.
11091 (do_load_mask_compare): New function.
11092 (do_overlap_load_compare): New function.
11093 (expand_compare_loop): New function.
11094 (expand_block_compare): Call expand_compare_loop() when appropriate.
11095 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
11096 option description.
11097 (-mblock-compare-inline-loop-limit): New option.
11099 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11102 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
11103 Reverse order of second and third operands in first alternative.
11104 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
11105 of first and second elements in UNSPEC_VPERMR vector.
11106 (altivec_expand_vec_perm_le): Likewise.
11108 2018-01-08 Jeff Law <law@redhat.com>
11110 PR rtl-optimizatin/81308
11111 * tree-switch-conversion.c (cfg_altered): New file scoped static.
11112 (process_switch): If group_case_labels makes a change, then set
11114 (pass_convert_switch::execute): If a switch is converted, then
11115 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
11117 PR rtl-optimization/81308
11118 * recog.c (split_all_insns): Conditionally cleanup the CFG after
11121 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
11123 PR target/83663 - Revert r255946
11124 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
11125 generation for cases where splatting a value is not useful.
11126 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
11127 across a vec_duplicate and a paradoxical subreg forming a vector
11128 mode to a vec_concat.
11130 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11132 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
11133 -march=armv8.3-a variants.
11134 * config/arm/t-multilib: Likewise.
11135 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
11137 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11139 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
11141 (cceq_ior_compare_complement): Give it a name so I can use it, and
11142 change boolean_or_operator predicate to boolean_operator so it can
11143 be used to generate a crand.
11144 (eqne): New code iterator.
11145 (bd/bd_neg): New code_attrs.
11146 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
11147 a single define_insn.
11148 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
11149 decrement (bdnzt/bdnzf/bdzt/bdzf).
11150 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
11151 with the new names of the branch decrement patterns, and added the
11152 names of the branch decrement conditional patterns.
11154 2018-01-08 Richard Biener <rguenther@suse.de>
11156 PR tree-optimization/83563
11157 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
11160 2018-01-08 Richard Biener <rguenther@suse.de>
11162 PR middle-end/83713
11163 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
11165 2018-01-08 Richard Biener <rguenther@suse.de>
11167 PR tree-optimization/83685
11168 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
11169 references to abnormals.
11171 2018-01-08 Richard Biener <rguenther@suse.de>
11174 * dwarf2out.c (output_indirect_strings): Handle empty
11175 skeleton_debug_str_hash.
11176 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
11178 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11180 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
11181 (emit_store_direct): Likewise.
11182 (arc_trampoline_adjust_address): Likewise.
11183 (arc_asm_trampoline_template): New function.
11184 (arc_initialize_trampoline): Use asm_trampoline_template.
11185 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
11186 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
11187 * config/arc/arc.md (flush_icache): Delete pattern.
11189 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
11191 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
11192 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
11195 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11198 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
11199 by not USED_FOR_TARGET.
11200 (make_pass_resolve_sw_modes): Likewise.
11202 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
11204 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
11207 2018-01-08 Richard Biener <rguenther@suse.de>
11209 PR middle-end/83580
11210 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
11212 2018-01-08 Richard Biener <rguenther@suse.de>
11214 PR middle-end/83517
11215 * match.pd ((t * 2) / 2) -> t): Add missing :c.
11217 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
11219 PR middle-end/81897
11220 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
11221 basic blocks with a small number of successors.
11222 (convert_control_dep_chain_into_preds): Improve handling of
11224 (dump_predicates): Split apart into...
11225 (dump_pred_chain): ...here...
11226 (dump_pred_info): ...and here.
11227 (can_one_predicate_be_invalidated_p): Add debugging printfs.
11228 (can_chain_union_be_invalidated_p): Improve check for invalidation
11230 (uninit_uses_cannot_happen): Avoid unnecessary if
11231 convert_control_dep_chain_into_preds yielded nothing.
11233 2018-01-06 Martin Sebor <msebor@redhat.com>
11235 PR tree-optimization/83640
11236 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
11237 subtracting negative offset from size.
11238 (builtin_access::overlap): Adjust offset bounds of the access to fall
11239 within the size of the object if possible.
11241 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
11243 PR rtl-optimization/83699
11244 * expmed.c (extract_bit_field_1): Restrict the vector usage of
11245 extract_bit_field_as_subreg to cases in which the extracted
11246 value is also a vector.
11248 * lra-constraints.c (process_alt_operands): Test for the equivalence
11249 substitutions when detecting a possible reload cycle.
11251 2018-01-06 Jakub Jelinek <jakub@redhat.com>
11254 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
11255 by default if flag_selective_schedling{,2}. Formatting fixes.
11257 PR rtl-optimization/83682
11258 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
11259 if it has non-VECTOR_MODE element mode.
11260 (vec_duplicate_p): Likewise.
11262 PR middle-end/83694
11263 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
11264 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
11266 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11269 * config/i386/i386-builtin.def
11270 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
11271 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
11272 Require also OPTION_MASK_ISA_AVX512F in addition to
11273 OPTION_MASK_ISA_GFNI.
11274 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
11275 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
11276 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
11277 to OPTION_MASK_ISA_GFNI.
11278 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
11279 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
11280 OPTION_MASK_ISA_AVX512BW.
11281 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
11282 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
11283 addition to OPTION_MASK_ISA_GFNI.
11284 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
11285 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
11286 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
11287 to OPTION_MASK_ISA_GFNI.
11288 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
11289 a requirement for all ISAs rather than any of them with a few
11291 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
11293 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
11294 bitmasks to be enabled with 3 exceptions, instead of requiring any
11295 enabled ISA with lots of exceptions.
11296 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
11297 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
11298 Change avx512bw in isa attribute to avx512f.
11299 * config/i386/sgxintrin.h: Add license boilerplate.
11300 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
11301 to __AVX512F__ and __AVX512VL to __AVX512VL__.
11302 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
11303 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
11305 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
11306 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
11307 temporarily sse2 rather than sse if not enabled already.
11310 * config/i386/sse.md (VI248_VLBW): Rename to ...
11311 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
11312 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
11313 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
11314 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
11315 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
11316 mode iterator instead of VI248_VLBW.
11318 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
11320 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
11321 (record_modified): Skip clobbers; add debug output.
11322 (param_change_prob): Use sreal frequencies.
11324 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11326 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
11327 punt for user-aligned variables.
11329 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
11331 * tree-chrec.c (chrec_contains_symbols): Return true for
11334 2018-01-05 Sudakshina Das <sudi.das@arm.com>
11337 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
11338 of (x|y) == x for BICS pattern.
11340 2018-01-05 Jakub Jelinek <jakub@redhat.com>
11342 PR tree-optimization/83605
11343 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
11344 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
11347 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
11349 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
11350 * config/epiphany/rtems.h: New file.
11352 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11353 Uros Bizjak <ubizjak@gmail.com>
11356 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
11357 QIreg_operand instead of register_operand predicate.
11358 * config/i386/i386.c (ix86_rop_should_change_byte_p,
11359 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
11360 comments instead of -fmitigate[-_]rop.
11362 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11365 * cgraphunit.c (symbol_table::compile): Switch to text_section
11366 before calling assembly_start debug hook.
11367 * run-rtl-passes.c (run_rtl_passes): Likewise.
11370 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11372 * tree-vrp.c (extract_range_from_binary_expr_1): Check
11373 range_int_cst_p rather than !symbolic_range_p before calling
11374 extract_range_from_multiplicative_op_1.
11376 2018-01-04 Jeff Law <law@redhat.com>
11378 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
11379 redundant test in assertion.
11381 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11383 * doc/rtl.texi: Document machine_mode wrapper classes.
11385 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11387 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
11388 using tree_to_uhwi.
11390 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11392 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
11393 the VEC_PERM_EXPR fold to fail.
11395 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11398 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
11399 to switched_sections.
11401 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11404 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
11405 test for d.testing.
11407 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
11410 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
11411 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
11413 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11416 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
11417 is BLKmode and bitpos not zero or mode change is needed.
11419 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
11422 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
11425 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
11428 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
11429 instead of MULT rtx. Update all corresponding splitters.
11430 (*saddl_se): Ditto.
11431 (*ssub<modesuffix>): Ditto.
11432 (*ssubl_se): Ditto.
11433 (*cmp_sadd_di): Update split patterns.
11434 (*cmp_sadd_si): Ditto.
11435 (*cmp_sadd_sidi): Ditto.
11436 (*cmp_ssub_di): Ditto.
11437 (*cmp_ssub_si): Ditto.
11438 (*cmp_ssub_sidi): Ditto.
11439 * config/alpha/predicates.md (const23_operand): New predicate.
11440 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
11441 Look for ASHIFT, not MULT inner operand.
11442 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
11444 2018-01-04 Martin Liska <mliska@suse.cz>
11446 PR gcov-profile/83669
11447 * gcov.c (output_intermediate_file): Add version to intermediate
11449 * doc/gcov.texi: Document new field 'version' in intermediate
11450 file format. Fix location of '-k' option of gcov command.
11452 2018-01-04 Martin Liska <mliska@suse.cz>
11455 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
11457 2018-01-04 Jakub Jelinek <jakub@redhat.com>
11459 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
11461 2018-01-03 Martin Sebor <msebor@redhat.com>
11463 PR tree-optimization/83655
11464 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
11465 checking calls with invalid arguments.
11467 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11469 * tree-vect-stmts.c (vect_get_store_rhs): New function.
11470 (vectorizable_mask_load_store): Delete.
11471 (vectorizable_call): Return false for masked loads and stores.
11472 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
11473 instead of gimple_assign_rhs1.
11474 (vectorizable_load): Handle IFN_MASK_LOAD.
11475 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
11477 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11479 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
11481 (vectorizable_mask_load_store): ...here.
11482 (vectorizable_load): ...and here.
11484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11486 * tree-vect-stmts.c (vect_build_all_ones_mask)
11487 (vect_build_zero_merge_argument): New functions, split out from...
11488 (vectorizable_load): ...here.
11490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11492 * tree-vect-stmts.c (vect_check_store_rhs): New function,
11494 (vectorizable_mask_load_store): ...here.
11495 (vectorizable_store): ...and here.
11497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11499 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
11501 (vectorizable_mask_load_store): ...here.
11503 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11505 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
11506 (vect_model_store_cost): Take a vec_load_store_type instead of a
11508 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
11509 (vect_model_store_cost): Take a vec_load_store_type instead of a
11511 (vectorizable_mask_load_store): Update accordingly.
11512 (vectorizable_store): Likewise.
11513 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
11515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11517 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
11518 IFN_MASK_LOAD calls here rather than...
11519 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
11521 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11522 Alan Hayward <alan.hayward@arm.com>
11523 David Sherwood <david.sherwood@arm.com>
11525 * expmed.c (extract_bit_field_1): For vector extracts,
11526 fall back to extract_bit_field_as_subreg if vec_extract
11529 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11530 Alan Hayward <alan.hayward@arm.com>
11531 David Sherwood <david.sherwood@arm.com>
11533 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
11534 they are variable or constant sized.
11535 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
11536 slots for constant-sized data.
11538 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11539 Alan Hayward <alan.hayward@arm.com>
11540 David Sherwood <david.sherwood@arm.com>
11542 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
11543 handling COND_EXPRs with boolean comparisons, try to find a better
11544 basis for the mask type than the boolean itself.
11546 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11548 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
11549 is calculated and how it can be overridden.
11550 * genmodes.c (max_bitsize_mode_any_mode): New variable.
11551 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
11553 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
11556 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11557 Alan Hayward <alan.hayward@arm.com>
11558 David Sherwood <david.sherwood@arm.com>
11560 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
11561 Remove the mode argument.
11562 (aarch64_simd_valid_immediate): Remove the mode and inverse
11564 * config/aarch64/iterators.md (bitsize): New iterator.
11565 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
11566 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
11567 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
11568 aarch64_simd_valid_immediate.
11569 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
11570 (aarch64_reg_or_bic_imm): Likewise.
11571 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
11572 with an insn_type enum and msl with a modifier_type enum.
11573 Replace element_width with a scalar_mode. Change the shift
11574 to unsigned int. Add constructors for scalar_float_mode and
11575 scalar_int_mode elements.
11576 (aarch64_vect_float_const_representable_p): Delete.
11577 (aarch64_can_const_movi_rtx_p)
11578 (aarch64_simd_scalar_immediate_valid_for_move)
11579 (aarch64_simd_make_constant): Update call to
11580 aarch64_simd_valid_immediate.
11581 (aarch64_advsimd_valid_immediate_hs): New function.
11582 (aarch64_advsimd_valid_immediate): Likewise.
11583 (aarch64_simd_valid_immediate): Remove mode and inverse
11584 arguments. Rewrite to use the above. Use const_vec_duplicate_p
11585 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
11586 and aarch64_float_const_representable_p on the result.
11587 (aarch64_output_simd_mov_immediate): Remove mode argument.
11588 Update call to aarch64_simd_valid_immediate and use of
11589 simd_immediate_info.
11590 (aarch64_output_scalar_simd_mov_immediate): Update call
11593 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11594 Alan Hayward <alan.hayward@arm.com>
11595 David Sherwood <david.sherwood@arm.com>
11597 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
11598 (mode_nunits): Likewise CONST_MODE_NUNITS.
11599 * machmode.def (ADJUST_NUNITS): Document.
11600 * genmodes.c (mode_data::need_nunits_adj): New field.
11601 (blank_mode): Update accordingly.
11602 (adj_nunits): New variable.
11603 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
11605 (emit_mode_size_inline): Set need_bytesize_adj for all modes
11606 listed in adj_nunits.
11607 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
11608 listed in adj_nunits. Don't emit case statements for such modes.
11609 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
11610 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
11611 nothing if adj_nunits is nonnull.
11612 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
11613 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
11614 (emit_mode_fbit): Update use of print_maybe_const_decl.
11615 (emit_move_size): Likewise. Treat the array as non-const
11617 (emit_mode_adjustments): Handle adj_nunits.
11619 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11621 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
11622 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
11623 (VECTOR_MODES): Use it.
11624 (make_vector_modes): Take the prefix as an argument.
11626 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11627 Alan Hayward <alan.hayward@arm.com>
11628 David Sherwood <david.sherwood@arm.com>
11630 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
11631 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
11632 for MODE_VECTOR_BOOL.
11633 * machmode.def (VECTOR_BOOL_MODE): Document.
11634 * genmodes.c (VECTOR_BOOL_MODE): New macro.
11635 (make_vector_bool_mode): New function.
11636 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
11638 * lto-streamer-in.c (lto_input_mode_table): Likewise.
11639 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
11641 * stor-layout.c (int_mode_for_mode): Likewise.
11642 * tree.c (build_vector_type_for_mode): Likewise.
11643 * varasm.c (output_constant_pool_2): Likewise.
11644 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
11645 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
11646 for MODE_VECTOR_BOOL.
11647 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
11648 of mode class checks.
11649 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
11650 instead of a list of mode class checks.
11651 (expand_vector_scalar_condition): Likewise.
11652 (type_for_widest_vector_mode): Handle BImode as an inner mode.
11654 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11655 Alan Hayward <alan.hayward@arm.com>
11656 David Sherwood <david.sherwood@arm.com>
11658 * machmode.h (mode_size): Change from unsigned short to
11660 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
11661 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11662 or if measurement_type is not polynomial.
11663 (fixed_size_mode::includes_p): Check for constant-sized modes.
11664 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
11665 return a poly_uint16 rather than an unsigned short.
11666 (emit_mode_size): Change the type of mode_size from unsigned short
11667 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
11668 (emit_mode_adjustments): Cope with polynomial vector sizes.
11669 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11671 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11673 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
11674 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
11675 * caller-save.c (setup_save_areas): Likewise.
11676 (replace_reg_with_saved_mem): Likewise.
11677 * calls.c (emit_library_call_value_1): Likewise.
11678 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
11679 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
11680 (gen_lowpart_for_combine): Likewise.
11681 * convert.c (convert_to_integer_1): Likewise.
11682 * cse.c (equiv_constant, cse_insn): Likewise.
11683 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
11684 (cselib_subst_to_values): Likewise.
11685 * dce.c (word_dce_process_block): Likewise.
11686 * df-problems.c (df_word_lr_mark_ref): Likewise.
11687 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
11688 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
11689 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
11690 (rtl_for_decl_location): Likewise.
11691 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
11692 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
11693 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
11694 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
11695 (expand_expr_real_1): Likewise.
11696 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
11697 (pad_below): Likewise.
11698 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11699 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
11700 * ira.c (get_subreg_tracking_sizes): Likewise.
11701 * ira-build.c (ira_create_allocno_objects): Likewise.
11702 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
11703 (ira_sort_regnos_for_alter_reg): Likewise.
11704 * ira-costs.c (record_operand_costs): Likewise.
11705 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
11706 (resolve_simple_move): Likewise.
11707 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
11708 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
11709 (lra_constraints): Likewise.
11710 (CONST_POOL_OK_P): Reject variable-sized modes.
11711 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
11712 (add_pseudo_to_slot, lra_spill): Likewise.
11713 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11714 * optabs-query.c (get_best_extraction_insn): Likewise.
11715 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11716 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
11717 (expand_mult_highpart, valid_multiword_target_p): Likewise.
11718 * recog.c (offsettable_address_addr_space_p): Likewise.
11719 * regcprop.c (maybe_mode_change): Likewise.
11720 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
11721 * regrename.c (build_def_use): Likewise.
11722 * regstat.c (dump_reg_info): Likewise.
11723 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
11724 (find_reloads, find_reloads_subreg_address): Likewise.
11725 * reload1.c (eliminate_regs_1): Likewise.
11726 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
11727 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
11728 (simplify_binary_operation_1, simplify_subreg): Likewise.
11729 * targhooks.c (default_function_arg_padding): Likewise.
11730 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
11731 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
11732 (verify_gimple_assign_ternary): Likewise.
11733 * tree-inline.c (estimate_move_cost): Likewise.
11734 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11735 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
11736 (get_address_cost_ainc): Likewise.
11737 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
11738 (vect_supportable_dr_alignment): Likewise.
11739 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11740 (vectorizable_reduction): Likewise.
11741 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
11742 (vectorizable_operation, vectorizable_load): Likewise.
11743 * tree.c (build_same_sized_truth_vector_type): Likewise.
11744 * valtrack.c (cleanup_auto_inc_dec): Likewise.
11745 * var-tracking.c (emit_note_insn_var_location): Likewise.
11746 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
11747 (ADDR_VEC_ALIGN): Likewise.
11749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11750 Alan Hayward <alan.hayward@arm.com>
11751 David Sherwood <david.sherwood@arm.com>
11753 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
11755 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
11756 or if measurement_type is polynomial.
11757 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
11758 * combine.c (make_extraction): Likewise.
11759 * dse.c (find_shift_sequence): Likewise.
11760 * dwarf2out.c (mem_loc_descriptor): Likewise.
11761 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
11762 (extract_bit_field, extract_low_bits): Likewise.
11763 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
11764 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
11765 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
11766 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
11767 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
11768 * reload.c (find_reloads): Likewise.
11769 * reload1.c (alter_reg): Likewise.
11770 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
11771 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
11772 * tree-if-conv.c (predicate_mem_writes): Likewise.
11773 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
11774 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
11775 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
11776 * valtrack.c (dead_debug_insert_temp): Likewise.
11777 * varasm.c (mergeable_constant_section): Likewise.
11778 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
11780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11781 Alan Hayward <alan.hayward@arm.com>
11782 David Sherwood <david.sherwood@arm.com>
11784 * expr.c (expand_assignment): Cope with polynomial mode sizes
11785 when assigning to a CONCAT.
11787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11788 Alan Hayward <alan.hayward@arm.com>
11789 David Sherwood <david.sherwood@arm.com>
11791 * machmode.h (mode_precision): Change from unsigned short to
11793 (mode_to_precision): Return a poly_uint16 rather than an unsigned
11795 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
11796 or if measurement_type is not polynomial.
11797 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
11798 in which the mode is already known to be a scalar_int_mode.
11799 * genmodes.c (emit_mode_precision): Change the type of mode_precision
11800 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
11802 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11803 for GET_MODE_PRECISION.
11804 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11805 for GET_MODE_PRECISION.
11806 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
11808 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
11809 (expand_field_assignment, make_extraction): Likewise.
11810 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
11811 (get_last_value): Likewise.
11812 * convert.c (convert_to_integer_1): Likewise.
11813 * cse.c (cse_insn): Likewise.
11814 * expr.c (expand_expr_real_1): Likewise.
11815 * lra-constraints.c (simplify_operand_subreg): Likewise.
11816 * optabs-query.c (can_atomic_load_p): Likewise.
11817 * optabs.c (expand_atomic_load): Likewise.
11818 (expand_atomic_store): Likewise.
11819 * ree.c (combine_reaching_defs): Likewise.
11820 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
11821 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
11822 * tree.h (type_has_mode_precision_p): Likewise.
11823 * ubsan.c (instrument_si_overflow): Likewise.
11825 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11826 Alan Hayward <alan.hayward@arm.com>
11827 David Sherwood <david.sherwood@arm.com>
11829 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
11830 polynomial numbers of units.
11831 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
11832 (valid_vector_subparts_p): New function.
11833 (build_vector_type): Remove temporary shim and take the number
11834 of units as a poly_uint64 rather than an int.
11835 (build_opaque_vector_type): Take the number of units as a
11836 poly_uint64 rather than an int.
11837 * tree.c (build_vector_from_ctor): Handle polynomial
11838 TYPE_VECTOR_SUBPARTS.
11839 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
11840 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
11841 (build_vector_from_val): If the number of units is variable,
11842 use build_vec_duplicate_cst for constant operands and
11843 VEC_DUPLICATE_EXPR otherwise.
11844 (make_vector_type): Remove temporary is_constant ().
11845 (build_vector_type, build_opaque_vector_type): Take the number of
11846 units as a poly_uint64 rather than an int.
11847 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
11849 * cfgexpand.c (expand_debug_expr): Likewise.
11850 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
11851 (store_constructor, expand_expr_real_1): Likewise.
11852 (const_scalar_mask_from_tree): Likewise.
11853 * fold-const-call.c (fold_const_reduction): Likewise.
11854 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
11855 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
11856 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
11857 (fold_relational_const): Likewise.
11858 (native_interpret_vector): Likewise. Change the size from an
11859 int to an unsigned int.
11860 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
11861 TYPE_VECTOR_SUBPARTS.
11862 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
11863 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
11864 duplicating a non-constant operand into a variable-length vector.
11865 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
11866 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
11867 * ipa-icf.c (sem_variable::equals): Likewise.
11868 * match.pd: Likewise.
11869 * omp-simd-clone.c (simd_clone_subparts): Likewise.
11870 * print-tree.c (print_node): Likewise.
11871 * stor-layout.c (layout_type): Likewise.
11872 * targhooks.c (default_builtin_vectorization_cost): Likewise.
11873 * tree-cfg.c (verify_gimple_comparison): Likewise.
11874 (verify_gimple_assign_binary): Likewise.
11875 (verify_gimple_assign_ternary): Likewise.
11876 (verify_gimple_assign_single): Likewise.
11877 * tree-pretty-print.c (dump_generic_node): Likewise.
11878 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11879 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
11880 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
11881 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
11882 (vect_shift_permute_load_chain): Likewise.
11883 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
11884 (expand_vector_condition, optimize_vector_constructor): Likewise.
11885 (lower_vec_perm, get_compute_type): Likewise.
11886 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11887 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
11888 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
11889 (vect_recog_mask_conversion_pattern): Likewise.
11890 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
11891 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
11892 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11893 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
11894 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
11895 (vectorizable_shift, vectorizable_operation, vectorizable_store)
11896 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
11897 (supportable_widening_operation): Likewise.
11898 (supportable_narrowing_operation): Likewise.
11899 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
11901 * varasm.c (output_constant): Likewise.
11903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11904 Alan Hayward <alan.hayward@arm.com>
11905 David Sherwood <david.sherwood@arm.com>
11907 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
11908 so that both the length == 3 and length != 3 cases set up their
11909 own permute vectors. Add comments explaining why we know the
11910 number of elements is constant.
11911 (vect_permute_load_chain): Likewise.
11913 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11914 Alan Hayward <alan.hayward@arm.com>
11915 David Sherwood <david.sherwood@arm.com>
11917 * machmode.h (mode_nunits): Change from unsigned char to
11919 (ONLY_FIXED_SIZE_MODES): New macro.
11920 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
11921 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
11922 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
11924 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
11925 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
11926 or if measurement_type is not polynomial.
11927 * genmodes.c (ZERO_COEFFS): New macro.
11928 (emit_mode_nunits_inline): Make mode_nunits_inline return a
11930 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
11931 Use ZERO_COEFFS when emitting initializers.
11932 * data-streamer.h (bp_pack_poly_value): New function.
11933 (bp_unpack_poly_value): Likewise.
11934 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
11935 for GET_MODE_NUNITS.
11936 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
11937 for GET_MODE_NUNITS.
11938 * tree.c (make_vector_type): Remove temporary shim and make
11939 the real function take the number of units as a poly_uint64
11940 rather than an int.
11941 (build_vector_type_for_mode): Handle polynomial nunits.
11942 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
11943 * emit-rtl.c (const_vec_series_p_1): Likewise.
11944 (gen_rtx_CONST_VECTOR): Likewise.
11945 * fold-const.c (test_vec_duplicate_folding): Likewise.
11946 * genrecog.c (validate_pattern): Likewise.
11947 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
11948 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
11949 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
11950 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
11951 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
11952 * rtlanal.c (subreg_get_info): Likewise.
11953 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11954 (vect_grouped_load_supported): Likewise.
11955 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
11956 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
11957 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11958 (simplify_const_unary_operation, simplify_binary_operation_1)
11959 (simplify_const_binary_operation, simplify_ternary_operation)
11960 (test_vector_ops_duplicate, test_vector_ops): Likewise.
11961 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
11962 instead of CONST_VECTOR_NUNITS.
11963 * varasm.c (output_constant_pool_2): Likewise.
11964 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
11965 explicit-encoded elements in the XVEC for variable-length vectors.
11967 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11969 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
11971 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11972 Alan Hayward <alan.hayward@arm.com>
11973 David Sherwood <david.sherwood@arm.com>
11975 * coretypes.h (fixed_size_mode): Declare.
11976 (fixed_size_mode_pod): New typedef.
11977 * builtins.h (target_builtins::x_apply_args_mode)
11978 (target_builtins::x_apply_result_mode): Change type to
11979 fixed_size_mode_pod.
11980 * builtins.c (apply_args_size, apply_result_size, result_vector)
11981 (expand_builtin_apply_args_1, expand_builtin_apply)
11982 (expand_builtin_return): Update accordingly.
11984 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11986 * cse.c (hash_rtx_cb): Hash only the encoded elements.
11987 * cselib.c (cselib_hash_rtx): Likewise.
11988 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
11989 CONST_VECTOR encoding.
11991 2018-01-03 Jakub Jelinek <jakub@redhat.com>
11992 Jeff Law <law@redhat.com>
11995 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
11996 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
11997 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
11998 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
12001 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
12002 explicitly probe *sp in a noreturn function if there were any callee
12003 register saves or frame pointer is needed.
12005 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12008 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
12009 BLKmode for ternary, binary or unary expressions.
12012 * var-tracking.c (delete_vta_debug_insn): New inline function.
12013 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
12014 insns from get_insns () to NULL instead of each bb separately.
12015 Use delete_vta_debug_insn. No longer static.
12016 (vt_debug_insns_local, variable_tracking_main_1): Adjust
12017 delete_vta_debug_insns callers.
12018 * rtl.h (delete_vta_debug_insns): Declare.
12019 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
12020 instead of variable_tracking_main.
12022 2018-01-03 Martin Sebor <msebor@redhat.com>
12024 PR tree-optimization/83603
12025 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
12026 arguments past the endof the argument list in functions declared
12027 without a prototype.
12028 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
12029 Avoid checking when arguments are null.
12031 2018-01-03 Martin Sebor <msebor@redhat.com>
12034 * doc/extend.texi (attribute const): Fix a typo.
12035 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
12036 issuing -Wsuggest-attribute for void functions.
12038 2018-01-03 Martin Sebor <msebor@redhat.com>
12040 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
12041 offset_int::from instead of wide_int::to_shwi.
12042 (maybe_diag_overlap): Remove assertion.
12043 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
12044 * gimple-ssa-sprintf.c (format_directive): Same.
12045 (parse_directive): Same.
12046 (sprintf_dom_walker::compute_format_length): Same.
12047 (try_substitute_return_value): Same.
12049 2018-01-03 Jeff Law <law@redhat.com>
12051 PR middle-end/83654
12052 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
12053 non-constant residual for zero at runtime and avoid probing in
12054 that case. Reorganize code for trailing problem to mirror handling
12057 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12059 PR tree-optimization/83501
12060 * tree-ssa-strlen.c (get_string_cst): New.
12061 (handle_char_store): Call get_string_cst.
12063 2018-01-03 Martin Liska <mliska@suse.cz>
12065 PR tree-optimization/83593
12066 * tree-ssa-strlen.c: Include tree-cfg.h.
12067 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
12068 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
12069 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
12071 (strlen_dom_walker::before_dom_children): Call
12072 gimple_purge_dead_eh_edges. Dump tranformation with details
12074 (strlen_dom_walker::before_dom_children): Update call by adding
12075 new argument cleanup_eh.
12076 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
12078 2018-01-03 Martin Liska <mliska@suse.cz>
12081 * cif-code.def (VARIADIC_THUNK): New enum value.
12082 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
12085 2018-01-03 Jan Beulich <jbeulich@suse.com>
12087 * sse.md (mov<mode>_internal): Tighten condition for when to use
12088 vmovdqu<ssescalarsize> for TI and OI modes.
12090 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12092 Update copyright years.
12094 2018-01-03 Martin Liska <mliska@suse.cz>
12097 * ipa-visibility.c (function_and_variable_visibility): Skip
12098 functions with noipa attribure.
12100 2018-01-03 Jakub Jelinek <jakub@redhat.com>
12102 * gcc.c (process_command): Update copyright notice dates.
12103 * gcov-dump.c (print_version): Ditto.
12104 * gcov.c (print_version): Ditto.
12105 * gcov-tool.c (print_version): Ditto.
12106 * gengtype.c (create_file): Ditto.
12107 * doc/cpp.texi: Bump @copying's copyright year.
12108 * doc/cppinternals.texi: Ditto.
12109 * doc/gcc.texi: Ditto.
12110 * doc/gccint.texi: Ditto.
12111 * doc/gcov.texi: Ditto.
12112 * doc/install.texi: Ditto.
12113 * doc/invoke.texi: Ditto.
12115 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12117 * vector-builder.h (vector_builder::m_full_nelts): Change from
12118 unsigned int to poly_uint64.
12119 (vector_builder::full_nelts): Update prototype accordingly.
12120 (vector_builder::new_vector): Likewise.
12121 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
12122 (vector_builder::operator ==): Likewise.
12123 (vector_builder::finalize): Likewise.
12124 * int-vector-builder.h (int_vector_builder::int_vector_builder):
12125 Take the number of elements as a poly_uint64 rather than an
12127 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
12128 from unsigned int to poly_uint64.
12129 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
12130 (vec_perm_indices::new_vector): Likewise.
12131 (vec_perm_indices::length): Likewise.
12132 (vec_perm_indices::nelts_per_input): Likewise.
12133 (vec_perm_indices::input_nelts): Likewise.
12134 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
12135 number of elements per input as a poly_uint64 rather than an
12136 unsigned int. Use the original encoding for variable-length
12137 vectors, rather than clamping each individual element.
12138 For the second and subsequent elements in each pattern,
12139 clamp the step and base before clamping their sum.
12140 (vec_perm_indices::series_p): Handle polynomial element counts.
12141 (vec_perm_indices::all_in_range_p): Likewise.
12142 (vec_perm_indices_to_tree): Likewise.
12143 (vec_perm_indices_to_rtx): Likewise.
12144 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
12145 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
12146 (tree_vector_builder::new_binary_operation): Handle polynomial
12147 element counts. Return false if we need to know the number
12148 of elements at compile time.
12149 * fold-const.c (fold_vec_perm): Punt if the number of elements
12150 isn't known at compile time.
12152 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12154 * vec-perm-indices.h (vec_perm_builder): Change element type
12155 from HOST_WIDE_INT to poly_int64.
12156 (vec_perm_indices::element_type): Update accordingly.
12157 (vec_perm_indices::clamp): Handle polynomial element_types.
12158 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12159 (vec_perm_indices::all_in_range_p): Likewise.
12160 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
12162 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
12163 polynomial vec_perm_indices element types.
12164 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
12165 * fold-const.c (fold_vec_perm): Likewise.
12166 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
12167 * tree-vect-generic.c (lower_vec_perm): Likewise.
12168 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12169 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
12170 element type to HOST_WIDE_INT.
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 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
12177 rather than an int. Use plus_constant.
12178 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
12179 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
12181 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12182 Alan Hayward <alan.hayward@arm.com>
12183 David Sherwood <david.sherwood@arm.com>
12185 * calls.c (emit_call_1, expand_call): Change struct_value_size from
12186 a HOST_WIDE_INT to a poly_int64.
12188 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12189 Alan Hayward <alan.hayward@arm.com>
12190 David Sherwood <david.sherwood@arm.com>
12192 * calls.c (load_register_parameters): Cope with polynomial
12193 mode sizes. Require a constant size for BLKmode parameters
12194 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
12195 forces a parameter to be padded at the lsb end in order to
12196 fill a complete number of words, require the parameter size
12197 to be ordered wrt UNITS_PER_WORD.
12199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12200 Alan Hayward <alan.hayward@arm.com>
12201 David Sherwood <david.sherwood@arm.com>
12203 * reload1.c (spill_stack_slot_width): Change element type
12204 from unsigned int to poly_uint64_pod.
12205 (alter_reg): Treat mode sizes as polynomial.
12207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12208 Alan Hayward <alan.hayward@arm.com>
12209 David Sherwood <david.sherwood@arm.com>
12211 * reload.c (complex_word_subreg_p): New function.
12212 (reload_inner_reg_of_subreg, push_reload): Use it.
12214 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12215 Alan Hayward <alan.hayward@arm.com>
12216 David Sherwood <david.sherwood@arm.com>
12218 * lra-constraints.c (process_alt_operands): Reject matched
12219 operands whose sizes aren't ordered.
12220 (match_reload): Refer to this check here.
12222 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12223 Alan Hayward <alan.hayward@arm.com>
12224 David Sherwood <david.sherwood@arm.com>
12226 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
12227 that the mode size is in the set {1, 2, 4, 8, 16}.
12229 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12230 Alan Hayward <alan.hayward@arm.com>
12231 David Sherwood <david.sherwood@arm.com>
12233 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
12234 Use plus_constant instead of gen_rtx_PLUS.
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 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
12241 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
12242 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
12243 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
12244 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
12245 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
12246 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
12247 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
12248 * config/i386/i386.c (ix86_push_rounding): ...this new function.
12249 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
12251 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
12252 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
12253 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
12254 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
12255 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
12256 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
12257 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
12258 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
12259 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
12260 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
12262 * expr.c (emit_move_resolve_push): Treat the input and result
12263 of PUSH_ROUNDING as a poly_int64.
12264 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
12265 (emit_push_insn): Likewise.
12266 * lra-eliminations.c (mark_not_eliminable): Likewise.
12267 * recog.c (push_operand): Likewise.
12268 * reload1.c (elimination_effects): Likewise.
12269 * rtlanal.c (nonzero_bits1): Likewise.
12270 * calls.c (store_one_arg): Likewise. Require the padding to be
12271 known at compile time.
12273 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12274 Alan Hayward <alan.hayward@arm.com>
12275 David Sherwood <david.sherwood@arm.com>
12277 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
12278 Use plus_constant instead of gen_rtx_PLUS.
12280 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12281 Alan Hayward <alan.hayward@arm.com>
12282 David Sherwood <david.sherwood@arm.com>
12284 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
12285 rather than an int.
12287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12288 Alan Hayward <alan.hayward@arm.com>
12289 David Sherwood <david.sherwood@arm.com>
12291 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
12292 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
12293 via stack temporaries. Treat the mode size as polynomial too.
12295 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12296 Alan Hayward <alan.hayward@arm.com>
12297 David Sherwood <david.sherwood@arm.com>
12299 * expr.c (expand_expr_real_2): When handling conversions involving
12300 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
12301 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
12302 as a poly_uint64 too.
12304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12305 Alan Hayward <alan.hayward@arm.com>
12306 David Sherwood <david.sherwood@arm.com>
12308 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
12310 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12311 Alan Hayward <alan.hayward@arm.com>
12312 David Sherwood <david.sherwood@arm.com>
12314 * combine.c (can_change_dest_mode): Handle polynomial
12315 REGMODE_NATURAL_SIZE.
12316 * expmed.c (store_bit_field_1): Likewise.
12317 * expr.c (store_constructor): Likewise.
12318 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
12319 and polynomial REGMODE_NATURAL_SIZE.
12320 (gen_lowpart_common): Likewise.
12321 * reginfo.c (record_subregs_of_mode): Likewise.
12322 * rtlanal.c (read_modify_subreg_p): Likewise.
12324 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12325 Alan Hayward <alan.hayward@arm.com>
12326 David Sherwood <david.sherwood@arm.com>
12328 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
12329 numbers of elements.
12331 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12332 Alan Hayward <alan.hayward@arm.com>
12333 David Sherwood <david.sherwood@arm.com>
12335 * match.pd: Cope with polynomial numbers of vector elements.
12337 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12338 Alan Hayward <alan.hayward@arm.com>
12339 David Sherwood <david.sherwood@arm.com>
12341 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
12342 in a POINTER_PLUS_EXPR.
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 * omp-simd-clone.c (simd_clone_subparts): New function.
12349 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
12350 (ipa_simd_modify_function_body): Likewise.
12352 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12353 Alan Hayward <alan.hayward@arm.com>
12354 David Sherwood <david.sherwood@arm.com>
12356 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
12357 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
12358 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
12359 (expand_vector_condition, vector_element): Likewise.
12360 (subparts_gt): New function.
12361 (get_compute_type): Use subparts_gt.
12362 (count_type_subparts): Delete.
12363 (expand_vector_operations_1): Use subparts_gt instead of
12364 count_type_subparts.
12366 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12367 Alan Hayward <alan.hayward@arm.com>
12368 David Sherwood <david.sherwood@arm.com>
12370 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
12371 (vect_compile_time_alias): ...this new function. Do the calculation
12372 on poly_ints rather than trees.
12373 (vect_prune_runtime_alias_test_list): Update call accordingly.
12375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12376 Alan Hayward <alan.hayward@arm.com>
12377 David Sherwood <david.sherwood@arm.com>
12379 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
12381 (vect_schedule_slp_instance): Likewise.
12383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12384 Alan Hayward <alan.hayward@arm.com>
12385 David Sherwood <david.sherwood@arm.com>
12387 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
12388 constant and extern definitions for variable-length vectors.
12389 (vect_get_constant_vectors): Note that the number of units
12390 is known to be constant.
12392 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12393 Alan Hayward <alan.hayward@arm.com>
12394 David Sherwood <david.sherwood@arm.com>
12396 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
12397 of units as polynomial. Choose between WIDE and NARROW based
12400 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12401 Alan Hayward <alan.hayward@arm.com>
12402 David Sherwood <david.sherwood@arm.com>
12404 * tree-vect-stmts.c (simd_clone_subparts): New function.
12405 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
12407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12408 Alan Hayward <alan.hayward@arm.com>
12409 David Sherwood <david.sherwood@arm.com>
12411 * tree-vect-stmts.c (vectorizable_call): Treat the number of
12412 vectors as polynomial. Use build_index_vector for
12413 IFN_GOMP_SIMD_LANE.
12415 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12416 Alan Hayward <alan.hayward@arm.com>
12417 David Sherwood <david.sherwood@arm.com>
12419 * tree-vect-stmts.c (get_load_store_type): Treat the number of
12420 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
12421 for variable-length vectors.
12422 (vectorizable_mask_load_store): Treat the number of units as
12423 polynomial, asserting that it is constant if the condition has
12424 already been enforced.
12425 (vectorizable_store, vectorizable_load): Likewise.
12427 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12428 Alan Hayward <alan.hayward@arm.com>
12429 David Sherwood <david.sherwood@arm.com>
12431 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
12432 of units as polynomial. Punt if we can't tell at compile time
12433 which vector contains the final result.
12435 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12436 Alan Hayward <alan.hayward@arm.com>
12437 David Sherwood <david.sherwood@arm.com>
12439 * tree-vect-loop.c (vectorizable_induction): Treat the number
12440 of units as polynomial. Punt on SLP inductions. Use an integer
12441 VEC_SERIES_EXPR for variable-length integer reductions. Use a
12442 cast of such a series for variable-length floating-point
12445 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12446 Alan Hayward <alan.hayward@arm.com>
12447 David Sherwood <david.sherwood@arm.com>
12449 * tree.h (build_index_vector): Declare.
12450 * tree.c (build_index_vector): New function.
12451 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
12452 of units as polynomial, forcibly converting it to a constant if
12453 vectorizable_reduction has already enforced the condition.
12454 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
12455 to create a {1,2,3,...} vector.
12456 (vectorizable_reduction): Treat the number of units as polynomial.
12457 Choose vectype_in based on the largest scalar element size rather
12458 than the smallest number of units. Enforce the restrictions
12461 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12462 Alan Hayward <alan.hayward@arm.com>
12463 David Sherwood <david.sherwood@arm.com>
12465 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
12466 number of units as polynomial.
12468 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12469 Alan Hayward <alan.hayward@arm.com>
12470 David Sherwood <david.sherwood@arm.com>
12472 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
12473 * target.def (autovectorize_vector_sizes): Return the vector sizes
12474 by pointer, using vector_sizes rather than a bitmask.
12475 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
12476 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
12477 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
12479 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
12480 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
12481 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
12482 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
12483 * omp-general.c (omp_max_vf): Likewise.
12484 * omp-low.c (omp_clause_aligned_alignment): Likewise.
12485 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
12486 * tree-vect-loop.c (vect_analyze_loop): Likewise.
12487 * tree-vect-slp.c (vect_slp_bb): Likewise.
12488 * doc/tm.texi: Regenerate.
12489 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
12491 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
12492 the vector size as a poly_uint64 rather than an unsigned int.
12493 (current_vector_size): Change from an unsigned int to a poly_uint64.
12494 (get_vectype_for_scalar_type): Update accordingly.
12495 * tree.h (build_truth_vector_type): Take the size and number of
12496 units as a poly_uint64 rather than an unsigned int.
12497 (build_vector_type): Add a temporary overload that takes
12498 the number of units as a poly_uint64 rather than an unsigned int.
12499 * tree.c (make_vector_type): Likewise.
12500 (build_truth_vector_type): Take the number of units as a poly_uint64
12501 rather than an unsigned int.
12503 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12504 Alan Hayward <alan.hayward@arm.com>
12505 David Sherwood <david.sherwood@arm.com>
12507 * target.def (get_mask_mode): Take the number of units and length
12508 as poly_uint64s rather than unsigned ints.
12509 * targhooks.h (default_get_mask_mode): Update accordingly.
12510 * targhooks.c (default_get_mask_mode): Likewise.
12511 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
12512 * doc/tm.texi: Regenerate.
12514 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12515 Alan Hayward <alan.hayward@arm.com>
12516 David Sherwood <david.sherwood@arm.com>
12518 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
12519 * omp-general.c (omp_max_vf): Likewise.
12520 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
12521 (expand_omp_simd): Handle polynomial safelen.
12522 * omp-low.c (omplow_simd_context): Add a default constructor.
12523 (omplow_simd_context::max_vf): Change from int to poly_uint64.
12524 (lower_rec_simd_input_clauses): Update accordingly.
12525 (lower_rec_input_clauses): Likewise.
12527 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12528 Alan Hayward <alan.hayward@arm.com>
12529 David Sherwood <david.sherwood@arm.com>
12531 * tree-vectorizer.h (vect_nunits_for_cost): New function.
12532 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
12533 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
12534 (vect_analyze_slp_cost): Likewise.
12535 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
12536 (vect_model_load_cost): Likewise.
12538 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12539 Alan Hayward <alan.hayward@arm.com>
12540 David Sherwood <david.sherwood@arm.com>
12542 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
12543 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
12544 from an unsigned int * to a poly_uint64_pod *.
12545 (calculate_unrolling_factor): New function.
12546 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
12548 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12549 Alan Hayward <alan.hayward@arm.com>
12550 David Sherwood <david.sherwood@arm.com>
12552 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
12553 from an unsigned int to a poly_uint64.
12554 (_loop_vec_info::slp_unrolling_factor): Likewise.
12555 (_loop_vec_info::vectorization_factor): Change from an int
12557 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
12558 (vect_get_num_vectors): New function.
12559 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
12560 (vect_get_num_copies): Use vect_get_num_vectors.
12561 (vect_analyze_data_ref_dependences): Change max_vf from an int *
12562 to an unsigned int *.
12563 (vect_analyze_data_refs): Change min_vf from an int * to a
12565 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12566 than an unsigned HOST_WIDE_INT.
12567 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
12568 (vect_analyze_data_ref_dependence): Change max_vf from an int *
12569 to an unsigned int *.
12570 (vect_analyze_data_ref_dependences): Likewise.
12571 (vect_compute_data_ref_alignment): Handle polynomial vf.
12572 (vect_enhance_data_refs_alignment): Likewise.
12573 (vect_prune_runtime_alias_test_list): Likewise.
12574 (vect_shift_permute_load_chain): Likewise.
12575 (vect_supportable_dr_alignment): Likewise.
12576 (dependence_distance_ge_vf): Take the vectorization factor as a
12577 poly_uint64 rather than an unsigned HOST_WIDE_INT.
12578 (vect_analyze_data_refs): Change min_vf from an int * to a
12580 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
12581 vfm1 as a poly_uint64 rather than an int. Make the same change
12582 for the returned bound_scalar.
12583 (vect_gen_vector_loop_niters): Handle polynomial vf.
12584 (vect_do_peeling): Likewise. Update call to
12585 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
12586 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
12588 * tree-vect-loop.c (vect_determine_vectorization_factor)
12589 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
12590 (vect_get_known_peeling_cost): Likewise.
12591 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
12592 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
12593 (vect_transform_loop): Likewise. Use the lowest possible VF when
12594 updating the upper bounds of the loop.
12595 (vect_min_worthwhile_factor): Make static. Return an unsigned int
12596 rather than an int.
12597 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
12598 polynomial unroll factors.
12599 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
12600 (vect_make_slp_decision): Likewise.
12601 (vect_supported_load_permutation_p): Likewise, and polynomial
12603 (vect_analyze_slp_cost): Handle polynomial vf.
12604 (vect_slp_analyze_node_operations): Likewise.
12605 (vect_slp_analyze_bb_1): Likewise.
12606 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
12607 than an unsigned HOST_WIDE_INT.
12608 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
12609 (vectorizable_load): Handle polynomial vf.
12610 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
12612 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
12614 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12615 Alan Hayward <alan.hayward@arm.com>
12616 David Sherwood <david.sherwood@arm.com>
12618 * match.pd: Handle bit operations involving three constants
12619 and try to fold one pair.
12621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
12623 * tree-vect-loop-manip.c: Include gimple-fold.h.
12624 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
12625 niters_maybe_zero parameters. Handle other cases besides a step of 1.
12626 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
12627 Add a path that uses a step of VF instead of 1, but disable it
12629 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
12630 and niters_no_overflow parameters. Update calls to
12631 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
12632 Create a new SSA name if the latter choses to use a ste other
12633 than zero, and return it via niters_vector_mult_vf_var.
12634 * tree-vect-loop.c (vect_transform_loop): Update calls to
12635 vect_do_peeling, vect_gen_vector_loop_niters and
12636 slpeel_make_loop_iterate_ntimes.
12637 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
12638 (vect_gen_vector_loop_niters): Update declarations after above changes.
12640 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
12642 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
12643 128-bit round to integer instructions.
12644 (ceil<mode>2): Likewise.
12645 (btrunc<mode>2): Likewise.
12646 (round<mode>2): Likewise.
12648 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
12650 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
12651 unaligned VSX load/store on P8/P9.
12652 (expand_block_clear): Allow the use of unaligned VSX
12653 load/store on P8/P9.
12655 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12657 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
12659 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
12660 swap associated with both a load and a store.
12662 2018-01-02 Andrew Waterman <andrew@sifive.com>
12664 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
12665 * config/riscv/riscv.md (clear_cache): Use it.
12667 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
12669 * web.c: Remove out-of-date comment.
12671 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12673 * expr.c (fixup_args_size_notes): Check that any existing
12674 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
12675 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
12676 (emit_single_push_insn): ...here.
12678 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12680 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
12681 (const_vector_encoded_nelts): New function.
12682 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
12683 (const_vector_int_elt, const_vector_elt): Declare.
12684 * emit-rtl.c (const_vector_int_elt_1): New function.
12685 (const_vector_elt): Likewise.
12686 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
12687 of CONST_VECTOR_ELT.
12689 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12691 * expr.c: Include rtx-vector-builder.h.
12692 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
12693 directly on the tree encoding.
12694 (const_vector_from_tree): Likewise.
12695 * optabs.c: Include rtx-vector-builder.h.
12696 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
12697 sequence of "u" values.
12698 * vec-perm-indices.c: Include rtx-vector-builder.h.
12699 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
12700 directly on the vec_perm_indices encoding.
12702 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12704 * doc/rtl.texi (const_vector): Describe new encoding scheme.
12705 * Makefile.in (OBJS): Add rtx-vector-builder.o.
12706 * rtx-vector-builder.h: New file.
12707 * rtx-vector-builder.c: Likewise.
12708 * rtl.h (rtx_def::u2): Add a const_vector field.
12709 (CONST_VECTOR_NPATTERNS): New macro.
12710 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
12711 (CONST_VECTOR_DUPLICATE_P): Likewise.
12712 (CONST_VECTOR_STEPPED_P): Likewise.
12713 (CONST_VECTOR_ENCODED_ELT): Likewise.
12714 (const_vec_duplicate_p): Check for a duplicated vector encoding.
12715 (unwrap_const_vec_duplicate): Likewise.
12716 (const_vec_series_p): Check for a non-duplicated vector encoding.
12717 Say that the function only returns true for integer vectors.
12718 * emit-rtl.c: Include rtx-vector-builder.h.
12719 (gen_const_vec_duplicate_1): Delete.
12720 (gen_const_vector): Call gen_const_vec_duplicate instead of
12721 gen_const_vec_duplicate_1.
12722 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
12723 (gen_const_vec_duplicate): Use rtx_vector_builder.
12724 (gen_const_vec_series): Likewise.
12725 (gen_rtx_CONST_VECTOR): Likewise.
12726 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
12727 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12728 Build a new vector rather than modifying a CONST_VECTOR in-place.
12729 (handle_special_swappables): Update call accordingly.
12730 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
12731 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
12732 Build a new vector rather than modifying a CONST_VECTOR in-place.
12733 (handle_special_swappables): Update call accordingly.
12735 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12737 * simplify-rtx.c (simplify_const_binary_operation): Use
12738 CONST_VECTOR_ELT instead of XVECEXP.
12740 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12742 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
12743 the selector elements to be different from the data elements
12744 if the selector is a VECTOR_CST.
12745 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
12746 ssizetype for the selector.
12748 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12750 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
12751 before testing each element individually.
12752 * tree-vect-generic.c (lower_vec_perm): Likewise.
12754 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12756 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
12757 * selftest-run-tests.c (selftest::run_tests): Call it.
12758 * vector-builder.h (vector_builder::operator ==): New function.
12759 (vector_builder::operator !=): Likewise.
12760 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
12761 (vec_perm_indices::all_from_input_p): New function.
12762 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
12763 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
12764 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
12765 instead of reading the VECTOR_CST directly. Detect whether both
12766 vector inputs are the same before constructing the vec_perm_indices,
12767 and update the number of inputs argument accordingly. Use the
12768 utility functions added above. Only construct sel2 if we need to.
12770 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12772 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
12773 the broadcast of the low byte.
12774 (expand_mult_highpart): Use an explicit encoding for the permutes.
12775 * optabs-query.c (can_mult_highpart_p): Likewise.
12776 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
12777 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12778 (vectorizable_bswap): Likewise.
12779 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
12780 explicit encoding for the power-of-2 permutes.
12781 (vect_permute_store_chain): Likewise.
12782 (vect_grouped_load_supported): Likewise.
12783 (vect_permute_load_chain): Likewise.
12785 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12787 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
12788 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
12789 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
12790 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
12791 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
12792 (vect_gen_perm_mask_any): Likewise.
12794 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12796 * int-vector-builder.h: New file.
12797 * vec-perm-indices.h: Include int-vector-builder.h.
12798 (vec_perm_indices): Redefine as an int_vector_builder.
12799 (auto_vec_perm_indices): Delete.
12800 (vec_perm_builder): Redefine as a stand-alone class.
12801 (vec_perm_indices::vec_perm_indices): New function.
12802 (vec_perm_indices::clamp): Likewise.
12803 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
12804 (vec_perm_indices::new_vector): New function.
12805 (vec_perm_indices::new_expanded_vector): Update for new
12806 vec_perm_indices class.
12807 (vec_perm_indices::rotate_inputs): New function.
12808 (vec_perm_indices::all_in_range_p): Operate directly on the
12809 encoded form, without computing elided elements.
12810 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
12811 encoding. Update for new vec_perm_indices class.
12812 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
12813 the given vec_perm_builder.
12814 (expand_vec_perm_var): Update vec_perm_builder constructor.
12815 (expand_mult_highpart): Use vec_perm_builder instead of
12816 auto_vec_perm_indices.
12817 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
12818 vec_perm_indices instead of auto_vec_perm_indices. Use a single
12819 or double series encoding as appropriate.
12820 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
12821 vec_perm_indices instead of auto_vec_perm_indices.
12822 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
12823 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
12824 (vect_permute_store_chain): Likewise.
12825 (vect_grouped_load_supported): Likewise.
12826 (vect_permute_load_chain): Likewise.
12827 (vect_shift_permute_load_chain): Likewise.
12828 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
12829 (vect_transform_slp_perm_load): Likewise.
12830 (vect_schedule_slp_instance): Likewise.
12831 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
12832 (vectorizable_mask_load_store): Likewise.
12833 (vectorizable_bswap): Likewise.
12834 (vectorizable_store): Likewise.
12835 (vectorizable_load): Likewise.
12836 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
12837 vec_perm_indices instead of auto_vec_perm_indices. Use
12838 tree_to_vec_perm_builder to read the vector from a tree.
12839 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
12840 vec_perm_builder instead of a vec_perm_indices.
12841 (have_whole_vector_shift): Use vec_perm_builder and
12842 vec_perm_indices instead of auto_vec_perm_indices. Leave the
12843 truncation to calc_vec_perm_mask_for_shift.
12844 (vect_create_epilog_for_reduction): Likewise.
12845 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
12846 from auto_vec_perm_indices to vec_perm_indices.
12847 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12848 instead of changing individual elements.
12849 (aarch64_vectorize_vec_perm_const): Use new_vector to install
12850 the vector in d.perm.
12851 * config/arm/arm.c (expand_vec_perm_d::perm): Change
12852 from auto_vec_perm_indices to vec_perm_indices.
12853 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
12854 instead of changing individual elements.
12855 (arm_vectorize_vec_perm_const): Use new_vector to install
12856 the vector in d.perm.
12857 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
12858 Update vec_perm_builder constructor.
12859 (rs6000_expand_interleave): Likewise.
12860 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
12861 (rs6000_expand_interleave): Likewise.
12863 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12865 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
12866 to qimode could truncate the indices.
12867 * optabs.c (expand_vec_perm_var): Likewise.
12869 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
12871 * Makefile.in (OBJS): Add vec-perm-indices.o.
12872 * vec-perm-indices.h: New file.
12873 * vec-perm-indices.c: Likewise.
12874 * target.h (vec_perm_indices): Replace with a forward class
12876 (auto_vec_perm_indices): Move to vec-perm-indices.h.
12877 * optabs.h: Include vec-perm-indices.h.
12878 (expand_vec_perm): Delete.
12879 (selector_fits_mode_p, expand_vec_perm_var): Declare.
12880 (expand_vec_perm_const): Declare.
12881 * target.def (vec_perm_const_ok): Replace with...
12882 (vec_perm_const): ...this new hook.
12883 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
12884 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
12885 * doc/tm.texi: Regenerate.
12886 * optabs.def (vec_perm_const): Delete.
12887 * doc/md.texi (vec_perm_const): Likewise.
12888 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
12889 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
12890 expand_vec_perm for constant permutation vectors. Assert that
12891 the mode of variable permutation vectors is the integer equivalent
12892 of the mode that is being permuted.
12893 * optabs-query.h (selector_fits_mode_p): Declare.
12894 * optabs-query.c: Include vec-perm-indices.h.
12895 (selector_fits_mode_p): New function.
12896 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
12897 is defined, instead of checking whether the vec_perm_const_optab
12898 exists. Use targetm.vectorize.vec_perm_const instead of
12899 targetm.vectorize.vec_perm_const_ok. Check whether the indices
12900 fit in the vector mode before using a variable permute.
12901 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
12902 vec_perm_indices instead of an rtx.
12903 (expand_vec_perm): Replace with...
12904 (expand_vec_perm_const): ...this new function. Take the selector
12905 as a vec_perm_indices rather than an rtx. Also take the mode of
12906 the selector. Update call to shift_amt_for_vec_perm_mask.
12907 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
12908 Use vec_perm_indices::new_expanded_vector to expand the original
12909 selector into bytes. Check whether the indices fit in the vector
12910 mode before using a variable permute.
12911 (expand_vec_perm_var): Make global.
12912 (expand_mult_highpart): Use expand_vec_perm_const.
12913 * fold-const.c: Includes vec-perm-indices.h.
12914 * tree-ssa-forwprop.c: Likewise.
12915 * tree-vect-data-refs.c: Likewise.
12916 * tree-vect-generic.c: Likewise.
12917 * tree-vect-loop.c: Likewise.
12918 * tree-vect-slp.c: Likewise.
12919 * tree-vect-stmts.c: Likewise.
12920 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
12922 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
12923 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
12924 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
12925 (aarch64_vectorize_vec_perm_const): ...this new function.
12926 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12927 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12928 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
12929 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
12930 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12931 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12932 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
12934 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
12935 check for NEON modes.
12936 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
12937 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
12938 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
12939 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
12941 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
12942 the old VEC_PERM_CONST conditions.
12943 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
12944 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
12945 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
12946 (ia64_vectorize_vec_perm_const_ok): Merge into...
12947 (ia64_vectorize_vec_perm_const): ...this new function.
12948 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
12949 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
12950 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
12951 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
12952 * config/mips/mips.c (mips_expand_vec_perm_const)
12953 (mips_vectorize_vec_perm_const_ok): Merge into...
12954 (mips_vectorize_vec_perm_const): ...this new function.
12955 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
12956 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
12957 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
12958 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
12959 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
12960 (rs6000_expand_vec_perm_const): Delete.
12961 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
12963 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12964 (altivec_expand_vec_perm_const_le): Take each operand individually.
12965 Operate on constant selectors rather than rtxes.
12966 (altivec_expand_vec_perm_const): Likewise. Update call to
12967 altivec_expand_vec_perm_const_le.
12968 (rs6000_expand_vec_perm_const): Delete.
12969 (rs6000_vectorize_vec_perm_const_ok): Delete.
12970 (rs6000_vectorize_vec_perm_const): New function.
12971 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12972 an element count and rtx array.
12973 (rs6000_expand_extract_even): Update call accordingly.
12974 (rs6000_expand_interleave): Likewise.
12975 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
12976 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
12977 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
12978 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
12979 (rs6000_expand_vec_perm_const): Delete.
12980 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
12981 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12982 (altivec_expand_vec_perm_const_le): Take each operand individually.
12983 Operate on constant selectors rather than rtxes.
12984 (altivec_expand_vec_perm_const): Likewise. Update call to
12985 altivec_expand_vec_perm_const_le.
12986 (rs6000_expand_vec_perm_const): Delete.
12987 (rs6000_vectorize_vec_perm_const_ok): Delete.
12988 (rs6000_vectorize_vec_perm_const): New function. Remove stray
12989 reference to the SPE evmerge intructions.
12990 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
12991 an element count and rtx array.
12992 (rs6000_expand_extract_even): Update call accordingly.
12993 (rs6000_expand_interleave): Likewise.
12994 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
12995 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
12997 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
12999 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13001 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
13002 vector mode and that that mode matches the mode of the data
13004 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
13005 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
13006 directly using expand_vec_perm_1 when forcing selectors into
13008 (expand_vec_perm_var): New function, split out from expand_vec_perm.
13010 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13012 * optabs-query.h (can_vec_perm_p): Delete.
13013 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
13014 * optabs-query.c (can_vec_perm_p): Split into...
13015 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
13016 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
13017 particular selector is valid.
13018 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13019 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13020 (vect_grouped_load_supported): Likewise.
13021 (vect_shift_permute_load_chain): Likewise.
13022 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
13023 (vect_transform_slp_perm_load): Likewise.
13024 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13025 (vectorizable_bswap): Likewise.
13026 (vect_gen_perm_mask_checked): Likewise.
13027 * fold-const.c (fold_ternary_loc): Likewise. Don't take
13028 implementations of variable permutation vectors into account
13029 when deciding which selector to use.
13030 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
13031 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
13032 with a false third argument.
13033 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
13034 to test whether the constant selector is valid and can_vec_perm_var_p
13035 to test whether a variable selector is valid.
13037 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13039 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
13040 * optabs-query.c (can_vec_perm_p): Likewise.
13041 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
13042 instead of vec_perm_indices.
13043 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
13044 (vect_gen_perm_mask_checked): Likewise,
13045 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
13046 (vect_gen_perm_mask_checked): Likewise,
13048 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
13050 * optabs-query.h (qimode_for_vec_perm): Declare.
13051 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
13052 (qimode_for_vec_perm): ...this new function.
13053 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
13055 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13057 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
13058 does not have a conditional at the top.
13060 2018-01-02 Richard Biener <rguenther@suse.de>
13062 * ipa-inline.c (big_speedup_p): Fix expression.
13064 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13067 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
13070 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
13074 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
13075 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
13076 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
13077 cond_taken_branch_cost 3->4.
13079 2018-01-01 Jakub Jelinek <jakub@redhat.com>
13081 PR tree-optimization/83581
13082 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
13083 TODO_cleanup_cfg if any changes have been made.
13085 PR middle-end/83608
13086 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
13087 convert_modes if target mode has the right side, but different mode
13090 PR middle-end/83609
13091 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
13092 last argument when extracting from CONCAT. If either from_real or
13093 from_imag is NULL, use expansion through memory. If result is not
13094 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
13095 the parts directly to inner mode, if even that fails, use expansion
13098 PR middle-end/83623
13099 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
13100 check for bswap in mode rather than HImode and use that in expand_unop
13103 Copyright (C) 2018 Free Software Foundation, Inc.
13105 Copying and distribution of this file, with or without modification,
13106 are permitted in any medium without royalty provided the copyright
13107 notice and this notice are preserved.