1 2021-09-28 Geng Qi <gengqi@linux.alibaba.com>
3 * config/riscv/riscv.md (mulv<mode>4): Call gen_smul<mode>3_highpart.
4 (<u>mulditi3): Call <su>muldi3_highpart.
5 (<u>muldi3_highpart): Rename to <su>muldi3_highpart.
6 (<u>mulsidi3): Call <su>mulsi3_highpart.
7 (<u>mulsi3_highpart): Rename to <su>mulsi3_highpart.
9 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
11 * config/darwin.h (DSYMUTIL_SPEC): Recognize D sources.
13 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
15 * config/rs6000/darwin.h (FIXED_R13): Add for PPC64.
16 (FIRST_SAVED_GP_REGNO): Save from R13 even when it is one
19 2021-09-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21 * config/aarch64/aarch64.h (AARCH64_FL_LS64): Define
22 (AARCH64_FL_V8_7): Likewise.
23 (AARCH64_FL_FOR_ARCH8_7): Likewise.
24 * config/aarch64/aarch64-arches.def (armv8.7-a): Define.
25 * config/aarch64/aarch64-option-extensions.def (ls64): Define.
26 * doc/invoke.texi: Document the above.
28 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
30 * dbgcnt.c (dbg_cnt_counter): New.
31 * dbgcnt.h (dbg_cnt_counter): New.
32 * dumpfile.c (dump_options): Add entry for TDF_THREADING.
33 * dumpfile.h (enum dump_flag): Add TDF_THREADING.
34 * gimple-range-path.cc (DEBUG_SOLVER): Use TDF_THREADING.
35 * tree-ssa-threadupdate.c (dump_jump_thread_path): Dump out
38 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
40 * cfgcleanup.c (pass_jump::execute): Check
41 flag_expensive_optimizations.
42 (pass_jump_after_combine::gate): Same.
43 * doc/invoke.texi (-fthread-jumps): Enable for -O1.
44 * opts.c (default_options_table): Enable -fthread-jumps at -O1.
45 * tree-ssa-threadupdate.c
46 (fwd_jt_path_registry::remove_jump_threads_including): Bail unless
49 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
51 * tree-ssa-reassoc.c (biased_names): New global.
52 (propagate_bias_p): New function.
53 (loop_carried_phi): Remove.
54 (propagate_rank): Propagate bias along single uses.
55 (get_rank): Update biased_names when needed.
57 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
59 * passes.def (pass_reassoc): Rename parameter to early_p.
60 * tree-ssa-reassoc.c (reassoc_bias_loop_carried_phi_ranks_p):
62 (phi_rank): Don't bias loop-carried phi ranks
63 before vectorization pass.
64 (execute_reassoc): Add bias_loop_carried_phi_ranks_p parameter.
65 (pass_reassoc::pass_reassoc): Add bias_loop_carried_phi_ranks_p
67 (pass_reassoc::set_param): Set bias_loop_carried_phi_ranks_p
69 (pass_reassoc::execute): Pass bias_loop_carried_phi_ranks_p to
71 (pass_reassoc::bias_loop_carried_phi_ranks_p): New member.
73 2021-09-28 Jakub Jelinek <jakub@redhat.com>
76 * config/i386/i386.c (standard_80387_constant_p): Don't recognize
77 special 80387 instruction XFmode constants if flag_rounding_math.
79 2021-09-28 Richard Biener <rguenther@suse.de>
81 PR tree-optimization/100112
82 * tree-ssa-sccvn.c (visit_reference_op_load): Record the
83 referece into the hashtable twice in case last_vuse is
84 different from the original vuse on the stmt.
86 2021-09-28 Jakub Jelinek <jakub@redhat.com>
89 * gimplify.c (gimplify_adjust_omp_clauses_1): Don't call the
90 omp_finish_clause langhook on implicitly added OMP_CLAUSE_PRIVATE
91 clauses on SIMD constructs.
93 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
95 PR tree-optimization/102511
96 * gimple-range-path.cc (path_range_query::range_on_path_entry):
97 Return VARYING when nothing found.
99 2021-09-28 Hongyu Wang <hongyu.wang@intel.com>
102 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): Add
104 (VALID_SSE2_REG_VHF_MODE): Add V4HFmode and V2HFmode.
105 (VALID_MMX_REG_MODE): Add V4HFmode.
106 (SSE_REG_MODE_P): Replace VALID_AVX512FP16_REG_MODE with
107 vector mode condition.
108 * config/i386/i386.c (classify_argument): Parse V4HF/V2HF
110 (function_arg_32): Add V4HFmode.
111 (function_arg_advance_32): Likewise.
112 * config/i386/i386.md (mode): Add V4HF/V2HF.
113 (MODE_SIZE): Likewise.
114 * config/i386/mmx.md (MMXMODE): Add V4HF mode.
115 (V_32): Add V2HF mode.
116 (VHF_32_64): New mode iterator.
117 (*mov<mode>_internal): Adjust sse alternatives to support
119 (*mov<mode>_internal): Adjust sse alternatives to support
121 (<insn><mode>3): New define_insn for add/sub/mul/div.
123 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
125 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Check
127 (pass_early_thread_jumps::gate): Same.
128 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
129 Return if !flag_thread_jumps.
130 * tree-ssa-threadupdate.c
131 (jt_path_registry::register_jump_thread): Assert that
132 flag_thread_jumps is true.
134 2021-09-28 liuhongt <hongtao.liu@intel.com>
137 (simplify_context::simplify_binary_operation_1): Relax
138 condition of simplifying (vec_concat:M (vec_select op0
139 index0)(vec_select op1 index1)) to allow different modes
140 between op0 and M, but have same inner mode.
142 2021-09-28 liuhongt <hongtao.liu@intel.com>
144 * config/i386/i386-expand.c (emit_reduc_half): Handle
145 V8HF/V16HF/V32HFmode.
146 * config/i386/sse.md (REDUC_SSE_PLUS_MODE): Add V8HF.
147 (REDUC_SSE_SMINMAX_MODE): Ditto.
148 (REDUC_PLUS_MODE): Add V16HF and V32HF.
149 (REDUC_SMINMAX_MODE): Ditto.
151 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
153 * gimple-range-path.cc
154 (path_range_query::precompute_ranges_in_block): Rename to...
155 (path_range_query::compute_ranges_in_block): ...this.
156 (path_range_query::precompute_ranges): Rename to...
157 (path_range_query::compute_ranges): ...this.
158 (path_range_query::precompute_relations): Rename to...
159 (path_range_query::compute_relations): ...this.
160 (path_range_query::precompute_phi_relations): Rename to...
161 (path_range_query::compute_phi_relations): ...this.
162 * gimple-range-path.h: Rename precompute* to compute*.
163 * tree-ssa-threadbackward.c
164 (back_threader::find_taken_edge_switch): Same.
165 (back_threader::find_taken_edge_cond): Same.
166 * tree-ssa-threadedge.c
167 (hybrid_jt_simplifier::compute_ranges_from_state): Same.
168 (hybrid_jt_state::register_equivs_stmt): Inline...
169 * tree-ssa-threadedge.h: ...here.
171 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
173 * tree-vrp.c (lhs_of_dominating_assert): Remove.
174 (class vrp_jt_state): Remove.
175 (class vrp_jt_simplifier): Remove.
176 (vrp_jt_simplifier::simplify): Remove.
177 (class vrp_jump_threader): Remove.
178 (vrp_jump_threader::vrp_jump_threader): Remove.
179 (vrp_jump_threader::~vrp_jump_threader): Remove.
180 (vrp_jump_threader::before_dom_children): Remove.
181 (vrp_jump_threader::after_dom_children): Remove.
183 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
185 * passes.def (pass_vrp_threader): New.
186 * tree-pass.h (make_pass_vrp_threader): Add make_pass_vrp_threader.
187 * tree-ssa-threadedge.c (hybrid_jt_state::register_equivs_stmt): New.
188 (hybrid_jt_simplifier::hybrid_jt_simplifier): New.
189 (hybrid_jt_simplifier::simplify): New.
190 (hybrid_jt_simplifier::compute_ranges_from_state): New.
191 * tree-ssa-threadedge.h (class hybrid_jt_state): New.
192 (class hybrid_jt_simplifier): New.
193 * tree-vrp.c (execute_vrp): Remove ASSERT_EXPR based jump
195 (class hybrid_threader): New.
196 (hybrid_threader::hybrid_threader): New.
197 (hybrid_threader::~hybrid_threader): New.
198 (hybrid_threader::before_dom_children): New.
199 (hybrid_threader::after_dom_children): New.
200 (execute_vrp_threader): New.
201 (class pass_vrp_threader): New.
202 (make_pass_vrp_threader): New.
204 2021-09-27 Martin Liska <mliska@suse.cz>
206 * output.h (enum section_flag): New.
207 (SECTION_FORGET): Remove.
208 (SECTION_ENTSIZE): Make it (1UL << 8) - 1.
209 (SECTION_STYLE_MASK): Define it based on other enum
211 * varasm.c (switch_to_section): Remove unused handling of
214 2021-09-27 Martin Liska <mliska@suse.cz>
216 * common.opt: Add new variable flag_default_complex_method.
217 * opts.c (finish_options): Handle flags related to
218 x_flag_complex_method.
219 * toplev.c (process_options): Remove option handling related
220 to flag_complex_method.
222 2021-09-27 Richard Biener <rguenther@suse.de>
225 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid using
226 type_for_size, instead use int_mode_for_size.
228 2021-09-27 Andrew Pinski <apinski@marvell.com>
231 * gimplify.c (gimplify_save_expr): Return early
232 if the type of val is error_mark_node.
234 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
236 * tree-ssanames.c (ssa_name_has_boolean_range): Use
239 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
241 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
243 * tree-ssa-dom.c (cprop_operand): Convert to range_query API.
244 (cprop_into_stmt): Same.
245 (dom_opt_dom_walker::optimize_stmt): Same.
247 2021-09-27 Richard Biener <rguenther@suse.de>
249 PR tree-optimization/97351
250 PR tree-optimization/97352
251 PR tree-optimization/82426
252 * tree-vectorizer.h (dr_misalignment): Add vector type
254 (aligned_access_p): Likewise.
255 (known_alignment_for_access_p): Likewise.
256 (vect_supportable_dr_alignment): Likewise.
257 (vect_known_alignment_in_bytes): Likewise. Refactor.
258 (DR_MISALIGNMENT): Remove.
259 (vect_update_shared_vectype): Likewise.
260 * tree-vect-data-refs.c (dr_misalignment): Refactor, handle
261 a vector type with larger alignment requirement and apply
262 the negative step adjustment here.
263 (vect_calculate_target_alignment): Remove.
264 (vect_compute_data_ref_alignment): Get explicit vector type
265 argument, do not apply a negative step alignment adjustment
267 (vect_slp_analyze_node_alignment): Re-analyze alignment
268 when we re-visit the DR with a bigger desired alignment but
269 keep more precise results from smaller alignments.
270 * tree-vect-slp.c (vect_update_shared_vectype): Remove.
271 (vect_slp_analyze_node_operations_1): Do not update the
272 shared vector type on stmts.
273 * tree-vect-stmts.c (vect_analyze_stmt): Push/pop the
274 vector type of an SLP node to the representative stmt-info.
275 (vect_transform_stmt): Likewise.
277 2021-09-27 liuhongt <hongtao.liu@intel.com>
280 2021-09-09 liuhongt <hongtao.liu@intel.com>
283 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
284 (reduc_plus_scal_v4sf): .. this, New define_expand.
285 (reduc_plus_scal_v2df): .. and this, New define_expand.
287 2021-09-26 liuhongt <hongtao.liu@intel.com>
289 * doc/extend.texi (Half-Precision): Remove storage only
290 description for _Float16 w/o avx512fp16.
292 2021-09-25 Dimitar Dimitrov <dimitar@dinux.eu>
294 * config/pru/constraints.md (Rrio): New constraint.
295 * config/pru/predicates.md (regio_operand): New predicate.
296 * config/pru/pru-pragma.c (pru_register_pragmas): Register
297 the __regio_symbol address space.
298 * config/pru/pru-protos.h (pru_symref2ioregno): Declaration.
299 * config/pru/pru.c (pru_symref2ioregno): New helper function.
300 (pru_legitimate_address_p): Remove.
301 (pru_addr_space_legitimate_address_p): Use the address space
303 (pru_nongeneric_pointer_addrspace): New helper function.
304 (pru_insert_attributes): New function to validate __regio_symbol
306 (TARGET_INSERT_ATTRIBUTES): New macro.
307 (TARGET_LEGITIMATE_ADDRESS_P): Remove.
308 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): New macro.
309 * config/pru/pru.h (enum reg_class): Add REGIO_REGS class.
310 * config/pru/pru.md (*regio_readsi): New pattern to read I/O
312 (*regio_nozext_writesi): New pattern to write to I/O registers.
313 (*regio_zext_write_r30<EQS0:mode>): Ditto.
314 * doc/extend.texi: Document the new PRU Named Address Space.
316 2021-09-24 Patrick Palka <ppalka@redhat.com>
320 * real.c (encode_ieee_double): Avoid unwanted sign extension.
321 (encode_ieee_quad): Likewise.
323 2021-09-24 Vladimir Makarov <vmakarov@redhat.com>
325 PR rtl-optimization/102147
326 * ira-build.c (ira_conflict_vector_profitable_p): Make
327 profitability calculation independent of host compiler pointer and
330 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
332 * gimple-range-path.cc (path_range_query::path_range_query):
333 Move debugging header...
334 (path_range_query::precompute_ranges): ...here.
335 (path_range_query::internal_range_of_expr): Do not call
336 range_on_path_entry if NAME is defined in the current block.
338 2021-09-24 Richard Biener <rguenther@suse.de>
340 * cfghooks.c (verify_flow_info): Verify unallocated BB and
341 edge flags are not set.
343 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
345 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
347 (jt_path_registry::register_jump_thread): Call
348 cancel_invalid_paths.
349 * tree-ssa-threadupdate.h (class jt_path_registry): Add
350 cancel_invalid_paths.
352 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
354 PR tree-optimization/102400
355 * tree-ssa-sccvn.c (vn_reference_insert_pieces): Initialize
356 result_vdef to zero value.
358 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
360 PR tree-optimization/102451
361 * tree-ssa-dse.c (delete_dead_or_redundant_call): Record bb of stmt
364 2021-09-24 Hongyu Wang <hongyu.wang@intel.com>
366 * config/i386/sse.md (cond_<insn><mode>): Extend to support
368 (cond_mul<mode>): Likewise.
369 (cond_div<mode>): Likewise.
370 (cond_<code><mode>): Likewise.
371 (cond_fma<mode>): Likewise.
372 (cond_fms<mode>): Likewise.
373 (cond_fnma<mode>): Likewise.
374 (cond_fnms<mode>): Likewise.
376 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
378 PR tree-optimization/102463
379 * gimple-range-fold.cc (fold_using_range::relation_fold_and_or): If
380 there is no range-ops handler, don't look for a relation.
382 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
384 * gimple-range-cache.cc (ranger_cache::ranger_cache): Take
385 non-executable_edge flag as parameter.
386 * gimple-range-cache.h (ranger_cache): Adjust prototype.
387 * gimple-range-gori.cc (gori_compute::gori_compute): Take
388 non-executable_edge flag as parameter.
389 (gori_compute::outgoing_edge_range_p): Check new flag.
390 * gimple-range-gori.h (gori_compute): Adjust prototype.
391 * gimple-range.cc (gimple_ranger::gimple_ranger): Create new flag.
392 (gimple_ranger::range_on_edge): Check new flag.
393 * gimple-range.h (gimple_ranger::non_executable_edge_flag): New.
394 * gimple-ssa-evrp.c (rvrp_folder): Pass ranger flag to simplifer.
395 (hybrid_folder::hybrid_folder): Set ranger non-executable flag value.
396 (hybrid_folder::fold_stmt): Set flag value in the simplifer.
397 * vr-values.c (simplify_using_ranges::set_and_propagate_unexecutable):
398 Use not_executable flag if provided inmstead of EDGE_EXECUTABLE.
399 (simplify_using_ranges::simplify_switch_using_ranges): Clear
400 EDGE_EXECUTABLE like it originally did.
401 (simplify_using_ranges::cleanup_edges_and_switches): Clear any
402 NON_EXECUTABLE flags.
403 (simplify_using_ranges::simplify_using_ranges): Adjust.
404 * vr-values.h (class simplify_using_ranges): Adjust.
405 (simplify_using_ranges::set_range_query): Add non-executable flag param.
407 2021-09-23 Bill Schmidt <wschmidt@linux.ibm.com>
410 * config/rs6000/rs6000-call.c (rs6000_aggregate_candidate): Detect
411 zero-width bit fields and return indicator.
412 (rs6000_discover_homogeneous_aggregate): Diagnose when the
413 presence of a zero-width bit field changes parameter passing in
416 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
418 * gimple-range-fold.cc (fold_using_range::range_of_phi):
419 Remove dominator check.
421 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
423 * gimple-range-path.cc (path_range_query::precompute_relations):
424 Hoist edge calculations before using EDGE_SUCC.
426 2021-09-23 Jonathan Wakely <jwakely@redhat.com>
428 * configure.ac: Fix --with-multilib-list description.
429 * configure: Regenerate.
431 2021-09-23 Richard Biener <rguenther@suse.de>
433 PR tree-optimization/102448
434 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
435 Clear alignment info copied from DR_PTR_INFO.
437 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
439 * config/i386/i386-expand.c (ix86_use_mask_cmp_p): Enable
441 * config/i386/sse.md (sseintvecmodelower): Add HF vector modes.
442 (<avx512>_store<mode>_mask): Extend to support HF vector modes.
443 (vec_cmp<mode><avx512fmaskmodelower>): Likewise.
444 (vcond_mask_<mode><avx512fmaskmodelower>): Likewise.
445 (vcond<mode><mode>): New expander.
446 (vcond<mode><sseintvecmodelower>): Likewise.
447 (vcond<sseintvecmodelower><mode>): Likewise.
448 (vcondu<mode><sseintvecmodelower>): Likewise.
450 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
452 * config/i386/sse.md (extend<ssePHmodelower><mode>2):
454 (extendv4hf<mode>2): Likewise.
455 (extendv2hfv2df2): Likewise.
456 (trunc<mode><ssePHmodelower>2): Likewise.
457 (avx512fp16_vcvt<castmode>2ph_<mode>): Rename to ...
458 (trunc<mode>v4hf2): ... this, and drop constraints.
459 (avx512fp16_vcvtpd2ph_v2df): Rename to ...
460 (truncv2dfv2hf2): ... this, and likewise.
462 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
464 * config/i386/sse.md (float<floatunssuffix><mode><ssePHmodelower>2):
466 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>):
468 (float<floatunssuffix><mode>v4hf2): ... this, and drop constraints.
469 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Rename to ...
470 (float<floatunssuffix>v2div2hf2): ... this, and likewise.
472 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
474 * config/i386/i386.md (fix<fixunssuffix>_trunchf<mode>2): New expander.
475 (fixuns_trunchfhi2): Likewise.
476 (*fixuns_trunchfsi2zext): New define_insn.
477 * config/i386/sse.md (ssePHmodelower): New mode_attr.
478 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
479 New expander for same element vector fix_truncate.
480 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
481 Likewise for V4HF to V4SI/V4DI fix_truncate.
482 (fix<fixunssuffix>_truncv2hfv2di2):
483 Likeise for V2HF to V2DI fix_truncate.
485 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
487 * config/i386/i386.md (<code>hf3): New expander.
489 2021-09-23 liuhongt <hongtao.liu@intel.com>
491 * config/i386/sse.md (FMAMODEM): extend to handle FP16.
492 (VFH_SF_AVX512VL): Extend to handle HFmode.
493 (VF_SF_AVX512VL): Deleted.
495 2021-09-23 liuhongt <hongtao.liu@intel.com>
497 * config/i386/i386.md (rinthf2): New expander.
498 (nearbyinthf2): New expander.
500 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
502 * tree-ssa-dom.c (class dom_jump_threader_simplifier): Rename...
503 (class dom_jt_state): ...this and provide virtual overrides.
504 (dom_jt_state::register_equiv): New.
505 (class dom_jt_simplifier): Rename from
506 dom_jump_threader_simplifier.
507 (dom_jump_threader_simplifier::simplify): Rename...
508 (dom_jt_simplifier::simplify): ...to this.
509 (pass_dominator::execute): Use dom_jt_simplifier and
511 * tree-ssa-threadedge.c (jump_threader::jump_threader):
513 (jt_state::register_equivs_stmt): Abstract out...
514 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
516 (jump_threader::thread_around_empty_blocks): Update state.
517 (jump_threader::thread_through_normal_block): Same.
518 (jt_state::jt_state): Remove.
519 (jt_state::push): Remove pass specific bits. Keep block vector
521 (jt_state::append_path): New.
522 (jt_state::pop): Remove pass specific bits.
523 (jt_state::register_equiv): Same.
524 (jt_state::record_ranges_from_stmt): Same.
525 (jt_state::register_equivs_on_edge): Same. Rename...
526 (jt_state::register_equivs_edge): ...to this.
527 (jt_state::dump): New.
528 (jt_state::debug): New.
529 (jump_threader_simplifier::simplify): Remove.
530 (jt_state::get_path): New.
531 * tree-ssa-threadedge.h (class jt_simplifier): Make into a base
532 class. Expose common functionality as virtual methods.
533 (class jump_threader_simplifier): Same. Rename...
534 (class jt_simplifier): ...to this.
535 * tree-vrp.c (class vrp_jump_threader_simplifier): Rename...
536 (class vrp_jt_simplifier): ...to this. Provide pass specific
538 (class vrp_jt_state): New.
539 (vrp_jump_threader_simplifier::simplify): Rename...
540 (vrp_jt_simplifier::simplify): ...to this. Inline code from
541 what used to be the base class.
542 (vrp_jump_threader::vrp_jump_threader): Use vrp_jt_state and
545 2021-09-22 Tobias Burnus <tobias@codesourcery.com>
548 * doc/invoke.texi (-Wno-missing-include-dirs.): Document Fortran
551 2021-09-22 Roger Sayle <roger@nextmovesoftware.com>
552 Richard Biener <rguenther@suse.de>
554 * match.pd (negation simplifications): Implement some negation
555 folding transformations from fold-const.c's fold_negate_expr.
556 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Add a SIMPLIFY
557 argument, to control whether the op should be simplified prior
558 to looking up/assigning a value number.
559 (vn_nary_build_or_lookup): Update call to vn_nary_build_or_lookup_1.
560 (vn_nary_simplify): Likewise.
561 (visit_nary_op): Likewise, but when constructing a NEGATE_EXPR
562 now call vn_nary_build_or_lookup_1 disabling simplification.
564 2021-09-22 Jiufu Guo <guojiufu@linux.ibm.com>
566 PR tree-optimization/102087
567 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
568 Update bound/cmp/control for niter.
570 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
572 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
573 Move check for non-empty BB here.
574 (fur_source::register_outgoing_edges): ...from here.
576 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
578 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
579 Remove call to improve_range_with_equivs.
580 (path_range_query::improve_range_with_equivs): Remove
581 * gimple-range-path.h: Remove improve_range_with_equivs.
583 2021-09-22 dianhong xu <dianhong.xu@intel.com>
585 * config/i386/avx512fp16intrin.h:
586 (_mm512_mask_blend_ph): New intrinsic.
587 (_mm512_permutex2var_ph): Ditto.
588 (_mm512_permutexvar_ph): Ditto.
589 * config/i386/avx512fp16vlintrin.h:
590 (_mm256_mask_blend_ph): New intrinsic.
591 (_mm256_permutex2var_ph): Ditto.
592 (_mm256_permutexvar_ph): Ditto.
593 (_mm_mask_blend_ph): Ditto.
594 (_mm_permutex2var_ph): Ditto.
595 (_mm_permutexvar_ph): Ditto.
597 2021-09-22 dianhong xu <dianhong.xu@intel.com>
599 * config/i386/avx512fp16intrin.h: Add new intrinsics.
600 (_mm512_conj_pch): New intrinsic.
601 (_mm512_mask_conj_pch): Ditto.
602 (_mm512_maskz_conj_pch): Ditto.
603 * config/i386/avx512fp16vlintrin.h: Add new intrinsics.
604 (_mm256_conj_pch): New intrinsic.
605 (_mm256_mask_conj_pch): Ditto.
606 (_mm256_maskz_conj_pch): Ditto.
607 (_mm_conj_pch): Ditto.
608 (_mm_mask_conj_pch): Ditto.
609 (_mm_maskz_conj_pch): Ditto.
611 2021-09-22 dianhong xu <dianhong.xu@intel.com>
613 * config/i386/avx512fp16intrin.h (_MM512_REDUCE_OP): New macro
614 (_mm512_reduce_add_ph): New intrinsic.
615 (_mm512_reduce_mul_ph): Ditto.
616 (_mm512_reduce_min_ph): Ditto.
617 (_mm512_reduce_max_ph): Ditto.
618 * config/i386/avx512fp16vlintrin.h
619 (_MM256_REDUCE_OP/_MM_REDUCE_OP): New macro.
620 (_mm256_reduce_add_ph): New intrinsic.
621 (_mm256_reduce_mul_ph): Ditto.
622 (_mm256_reduce_min_ph): Ditto.
623 (_mm256_reduce_max_ph): Ditto.
624 (_mm_reduce_add_ph): Ditto.
625 (_mm_reduce_mul_ph): Ditto.
626 (_mm_reduce_min_ph): Ditto.
627 (_mm_reduce_max_ph): Ditto.
629 2021-09-22 dianhong xu <dianhong.xu@intel.com>
631 * config/i386/avx512fp16intrin.h (__m512h_u, __m256h_u,
632 __m128h_u): New typedef.
633 (_mm512_load_ph): New intrinsic.
634 (_mm256_load_ph): Ditto.
635 (_mm_load_ph): Ditto.
636 (_mm512_loadu_ph): Ditto.
637 (_mm256_loadu_ph): Ditto.
638 (_mm_loadu_ph): Ditto.
639 (_mm512_store_ph): Ditto.
640 (_mm256_store_ph): Ditto.
641 (_mm_store_ph): Ditto.
642 (_mm512_storeu_ph): Ditto.
643 (_mm256_storeu_ph): Ditto.
644 (_mm_storeu_ph): Ditto.
645 (_mm512_abs_ph): Ditto.
646 * config/i386/avx512fp16vlintrin.h
648 (_mm256_abs_ph): Ditto.
650 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
652 * config/s390/tpf.md (prologue_tpf, epilogue_tpf): Add cc clobber.
654 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
657 * config/s390/s390.c (s390_expand_insv): Emit a normal move if it
658 is actually a full copy of the source operand into the target.
659 Don't emit a strict low part move if source and target mode match.
661 2021-09-22 Jakub Jelinek <jakub@redhat.com>
664 * omp-expand.c (expand_omp_single): If region->exit is NULL,
665 assert region->entry is GIMPLE_OMP_SCOPE region and return.
667 2021-09-22 Jakub Jelinek <jakub@redhat.com>
669 * tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
670 * tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
671 arguments from 2 to 3.
672 * tree-pretty-print.c (dump_omp_clause): Print allocator() around
673 allocate clause allocator and print align if present.
674 * omp-low.c (scan_sharing_clauses): Force allocate_map entry even
675 for omp_default_mem_alloc if align modifier is present. If align
676 modifier is present, use TREE_LIST to encode both allocator and
678 (lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
679 Handle align modifier on allocator clause if present.
681 2021-09-22 liuhongt <hongtao.liu@intel.com>
683 * config/i386/i386.md (define_attr "isa"): Add
685 (define_attr "enabled"): Correspond fma_or_avx512vl to
686 TARGET_FMA || TARGET_AVX512VL.
687 * config/i386/mmx.md (fmav2sf4): Extend to AVX512 fma.
692 2021-09-22 liuhongt <hongtao.liu@intel.com>
694 * config/i386/i386.md (cstorehf3): New define_expand.
696 2021-09-22 liuhongt <hongtao.liu@intel.com>
698 * config/i386/i386.md (<rounding_insn>hf2): New expander.
699 (sse4_1_round<mode>2): Extend from MODEF to MODEFH.
700 * config/i386/sse.md (*sse4_1_round<ssescalarmodesuffix>):
701 Extend from VF_128 to VFH_128.
703 2021-09-22 liuhongt <hongtao.liu@intel.com>
705 * config/i386/i386-features.c (i386-features.c): Handle
707 * config/i386/i386.md (sqrthf2): New expander.
708 (*sqrthf2): New define_insn.
710 (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
713 2021-09-22 liuhongt <hongtao.liu@intel.com>
715 * config/i386/avx512fp16intrin.h (_mm_mask_fcmadd_sch):
717 (_mm_mask3_fcmadd_sch): Likewise.
718 (_mm_maskz_fcmadd_sch): Likewise.
719 (_mm_fcmadd_sch): Likewise.
720 (_mm_mask_fmadd_sch): Likewise.
721 (_mm_mask3_fmadd_sch): Likewise.
722 (_mm_maskz_fmadd_sch): Likewise.
723 (_mm_fmadd_sch): Likewise.
724 (_mm_mask_fcmadd_round_sch): Likewise.
725 (_mm_mask3_fcmadd_round_sch): Likewise.
726 (_mm_maskz_fcmadd_round_sch): Likewise.
727 (_mm_fcmadd_round_sch): Likewise.
728 (_mm_mask_fmadd_round_sch): Likewise.
729 (_mm_mask3_fmadd_round_sch): Likewise.
730 (_mm_maskz_fmadd_round_sch): Likewise.
731 (_mm_fmadd_round_sch): Likewise.
732 (_mm_fcmul_sch): Likewise.
733 (_mm_mask_fcmul_sch): Likewise.
734 (_mm_maskz_fcmul_sch): Likewise.
735 (_mm_fmul_sch): Likewise.
736 (_mm_mask_fmul_sch): Likewise.
737 (_mm_maskz_fmul_sch): Likewise.
738 (_mm_fcmul_round_sch): Likewise.
739 (_mm_mask_fcmul_round_sch): Likewise.
740 (_mm_maskz_fcmul_round_sch): Likewise.
741 (_mm_fmul_round_sch): Likewise.
742 (_mm_mask_fmul_round_sch): Likewise.
743 (_mm_maskz_fmul_round_sch): Likewise.
744 * config/i386/i386-builtin.def: Add corresponding new builtins.
746 (avx512fp16_fmaddcsh_v8hf_maskz<round_expand_name>): New expander.
747 (avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>): Ditto.
748 (avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>):
750 (avx512fp16_<complexopname>sh_v8hf_mask<round_name>): Ditto.
751 (avx512fp16_<complexopname>sh_v8hf<mask_scalarc_name><round_scalarcz_name>):
753 * config/i386/subst.md (mask_scalarcz_name): New.
754 (mask_scalarc_name): Ditto.
755 (mask_scalarc_operand3): Ditto.
756 (mask_scalarcz_operand4): Ditto.
757 (round_scalarcz_name): Ditto.
758 (round_scalarc_mask_operand3): Ditto.
759 (round_scalarcz_mask_operand4): Ditto.
760 (round_scalarc_mask_op3): Ditto.
761 (round_scalarcz_mask_op4): Ditto.
762 (round_scalarcz_constraint): Ditto.
763 (round_scalarcz_nimm_predicate): Ditto.
764 (mask_scalarcz): Ditto.
765 (mask_scalarc): Ditto.
766 (round_scalarcz): Ditto.
768 2021-09-22 liuhongt <hongtao.liu@intel.com>
770 * config/i386/avx512fp16intrin.h (_mm512_fcmadd_pch):
772 (_mm512_mask_fcmadd_pch): Likewise.
773 (_mm512_mask3_fcmadd_pch): Likewise.
774 (_mm512_maskz_fcmadd_pch): Likewise.
775 (_mm512_fmadd_pch): Likewise.
776 (_mm512_mask_fmadd_pch): Likewise.
777 (_mm512_mask3_fmadd_pch): Likewise.
778 (_mm512_maskz_fmadd_pch): Likewise.
779 (_mm512_fcmadd_round_pch): Likewise.
780 (_mm512_mask_fcmadd_round_pch): Likewise.
781 (_mm512_mask3_fcmadd_round_pch): Likewise.
782 (_mm512_maskz_fcmadd_round_pch): Likewise.
783 (_mm512_fmadd_round_pch): Likewise.
784 (_mm512_mask_fmadd_round_pch): Likewise.
785 (_mm512_mask3_fmadd_round_pch): Likewise.
786 (_mm512_maskz_fmadd_round_pch): Likewise.
787 (_mm512_fcmul_pch): Likewise.
788 (_mm512_mask_fcmul_pch): Likewise.
789 (_mm512_maskz_fcmul_pch): Likewise.
790 (_mm512_fmul_pch): Likewise.
791 (_mm512_mask_fmul_pch): Likewise.
792 (_mm512_maskz_fmul_pch): Likewise.
793 (_mm512_fcmul_round_pch): Likewise.
794 (_mm512_mask_fcmul_round_pch): Likewise.
795 (_mm512_maskz_fcmul_round_pch): Likewise.
796 (_mm512_fmul_round_pch): Likewise.
797 (_mm512_mask_fmul_round_pch): Likewise.
798 (_mm512_maskz_fmul_round_pch): Likewise.
799 * config/i386/avx512fp16vlintrin.h (_mm_fmadd_pch):
801 (_mm_mask_fmadd_pch): Likewise.
802 (_mm_mask3_fmadd_pch): Likewise.
803 (_mm_maskz_fmadd_pch): Likewise.
804 (_mm256_fmadd_pch): Likewise.
805 (_mm256_mask_fmadd_pch): Likewise.
806 (_mm256_mask3_fmadd_pch): Likewise.
807 (_mm256_maskz_fmadd_pch): Likewise.
808 (_mm_fcmadd_pch): Likewise.
809 (_mm_mask_fcmadd_pch): Likewise.
810 (_mm_mask3_fcmadd_pch): Likewise.
811 (_mm_maskz_fcmadd_pch): Likewise.
812 (_mm256_fcmadd_pch): Likewise.
813 (_mm256_mask_fcmadd_pch): Likewise.
814 (_mm256_mask3_fcmadd_pch): Likewise.
815 (_mm256_maskz_fcmadd_pch): Likewise.
816 (_mm_fmul_pch): Likewise.
817 (_mm_mask_fmul_pch): Likewise.
818 (_mm_maskz_fmul_pch): Likewise.
819 (_mm256_fmul_pch): Likewise.
820 (_mm256_mask_fmul_pch): Likewise.
821 (_mm256_maskz_fmul_pch): Likewise.
822 (_mm_fcmul_pch): Likewise.
823 (_mm_mask_fcmul_pch): Likewise.
824 (_mm_maskz_fcmul_pch): Likewise.
825 (_mm256_fcmul_pch): Likewise.
826 (_mm256_mask_fcmul_pch): Likewise.
827 (_mm256_maskz_fcmul_pch): Likewise.
828 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF_V8HF_V8HF,
829 V8HF_FTYPE_V16HF_V16HF_V16HF, V16HF_FTYPE_V16HF_V16HF_V16HF_UQI,
830 V32HF_FTYPE_V32HF_V32HF_V32HF_INT,
831 V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT): Add new builtin types.
832 * config/i386/i386-builtin.def: Add new builtins.
833 * config/i386/i386-expand.c: Handle new builtin types.
834 * config/i386/subst.md (SUBST_CV): New.
836 (maskc_operand3): Ditto.
838 (sdc_maskz_name): Ditto.
839 (sdc_mask_op4): Ditto.
840 (sdc_mask_op5): Ditto.
841 (sdc_mask_mode512bit_condition): Ditto.
843 (round_maskc_operand3): Ditto.
844 (round_sdc_mask_operand4): Ditto.
845 (round_maskc_op3): Ditto.
846 (round_sdc_mask_op4): Ditto.
847 (round_saeonly_sdc_mask_operand5): Ditto.
848 * config/i386/sse.md (unspec): Add complex fma unspecs.
849 (avx512fmaskcmode): New.
850 (UNSPEC_COMPLEX_F_C_MA): Ditto.
851 (UNSPEC_COMPLEX_F_C_MUL): Ditto.
852 (complexopname): Ditto.
853 (<avx512>_fmaddc_<mode>_maskz<round_expand_name>): New expander.
854 (<avx512>_fcmaddc_<mode>_maskz<round_expand_name>): Ditto.
855 (fma_<complexopname>_<mode><sdc_maskz_name><round_name>): New
857 (<avx512>_<complexopname>_<mode>_mask<round_name>): Ditto.
858 (<avx512>_<complexopname>_<mode><maskc_name><round_name>): Ditto.
860 2021-09-22 Kewen Lin <linkw@linux.ibm.com>
862 * config/rs6000/rs6000.opt (rs6000-density-pct-threshold,
863 rs6000-density-size-threshold, rs6000-density-penalty,
864 rs6000-density-load-pct-threshold,
865 rs6000-density-load-num-threshold): New parameter.
866 * config/rs6000/rs6000.c (rs6000_density_test): Adjust with
867 corresponding parameters.
869 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
871 * gimple-range-path.cc (path_range_query::defined_outside_path):
873 (path_range_query::range_on_path_entry): New.
874 (path_range_query::internal_range_of_expr): Resolve unknowns
876 (path_range_query::improve_range_with_equivs): New.
877 (path_range_query::ssa_range_in_phi): Resolve unknowns with
879 * gimple-range-path.h (class path_range_query): Add
880 defined_outside_path, range_on_path_entry, and
881 improve_range_with_equivs.
883 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
885 * gimple-range-path.cc (path_range_query::add_to_imports): New.
886 (path_range_query::add_copies_to_imports): New.
887 (path_range_query::precompute_ranges): Call
888 add_copies_to_imports.
889 * gimple-range-path.h (class path_range_query): Add prototypes
890 for add_copies_to_imports and add_to_imports.
892 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
894 * gimple-range-path.cc (path_range_query::range_defined_in_block):
897 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
899 * gimple-range-fold.h (class fur_source): Make oracle protected.
900 * gimple-range-path.cc (path_range_query::path_range_query): Add
901 resolve argument. Initialize oracle.
902 (path_range_query::~path_range_query): Delete oracle.
903 (path_range_query::range_of_stmt): Adapt to use relations.
904 (path_range_query::precompute_ranges): Pre-compute relations.
905 (class jt_fur_source): New
906 (jt_fur_source::jt_fur_source): New.
907 (jt_fur_source::register_relation): New.
908 (jt_fur_source::query_relation): New.
909 (path_range_query::precompute_relations): New.
910 (path_range_query::precompute_phi_relations): New.
911 * gimple-range-path.h (path_range_query): Add resolve argument.
912 Add oracle, precompute_relations, precompute_phi_relations.
913 * tree-ssa-threadbackward.c (back_threader::back_threader): Pass
914 resolve argument to solver.
916 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
918 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
919 Rename postfold_gcond_edges to register_outgoing_edges and
921 (fold_using_range::postfold_gcond_edges): Rename...
922 (fur_source::register_outgoing_edges): ...to this.
923 * gimple-range-fold.h (postfold_gcond_edges): Rename to
924 register_outgoing_edges and move to fur_source.
926 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
928 * gimple-range-fold.cc (fold_using_range::range_of_phi): Check
929 dom_info_available_p.
931 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
933 * gimple-range-cache.cc (non_null_ref::non_null_ref): Use create
934 and quick_grow_cleared instead of safe_grow_cleared.
936 2021-09-21 Thomas Schwinge <thomas@codesourcery.com>
939 * omp-oacc-neuter-broadcast.cc (oacc_do_neutering): Evaluate
942 2021-09-21 Richard Earnshaw <rearnsha@arm.com>
944 * configure.ac: Detect when the assembler supports new-style
945 architecture extensions.
946 * common/config/arm/arm-common.c (arm_rewrite_mcpu): Return
947 the full CPU string if the assembler can grok it.
948 (arm_rewrite_march): Likewise but for the architecture.
949 * config.in: Regenerate.
950 * configure: Regenerate.
952 2021-09-21 Richard Biener <rguenther@suse.de>
954 PR tree-optimization/102421
955 * tree-vect-loop.c (vect_dissolve_slp_only_groups): Copy and
956 adjust alignment info.
958 2021-09-21 Kewen Lin <linkw@linux.ibm.com>
960 * ipa-fnsummary.c (ipa_fn_summary_write): Remove inconsistent
963 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
965 * gimple-range-fold.cc (fold_using_range::range_of_phi): Ignore
966 undefined edges, apply an equivalence if appropriate.
967 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Return
968 UNDEFINED if EDGE_EXECUTABLE is not set.
969 * gimple-range.cc (gimple_ranger::gimple_ranger): Set all edges
970 as EXECUTABLE upon startup.
971 (gimple_ranger::range_on_edge): Return UNDEFINED for edges without
973 * vr-values.c (set_and_propagate_unexecutable): New.
974 (simplify_using_ranges::fold_cond): Call set_and_propagate.
975 (simplify_using_ranges::simplify_switch_using_ranges): Ditto.
976 * vr-values.h: Add prototype.
978 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
980 * value-relation.cc (equiv_oracle::register_initial_def): New.
981 (equiv_oracle::register_relation): Call register_initial_def.
982 (equiv_oracle::add_equiv_to_block): New. Split register_relation.
983 (relation_oracle::register_stmt): Check def block of PHI arguments.
984 * value-relation.h (equiv_oracle): Add new prototypes.
986 2021-09-20 Matthias Kretz <m.kretz@gsi.de>
988 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
989 Define __RECIPROCAL_MATH__, __NO_SIGNED_ZEROS__,
990 __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__, and
991 __ROUNDING_MATH__ according to their corresponding flags.
992 * doc/cpp.texi: Document __RECIPROCAL_MATH__,
993 __NO_SIGNED_ZEROS__, __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__,
994 and __ROUNDING_MATH__.
996 2021-09-20 Richard Biener <rguenther@suse.de>
998 * tree-vect-stmts.c (vectorizable_load): Use the vectype
1001 2021-09-20 Richard Biener <rguenther@suse.de>
1003 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
1004 Do not compute alignment of the vectorized access here.
1006 2021-09-20 Richard Biener <rguenther@suse.de>
1008 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
1009 Store -1 for runtime alias peeling iterations.
1011 2021-09-20 Richard Biener <rguenther@suse.de>
1013 * config.gcc: Obsolete hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11*.
1015 2021-09-20 Thomas Schwinge <thomas@codesourcery.com>
1017 * input.c (string_concat_db::record_string_concatenation)
1018 (string_concat_db::get_string_concatenation): Skip for
1019 'RESERVED_LOCATION_P'.
1021 2021-09-20 Richard Biener <rguenther@suse.de>
1023 PR tree-optimization/65206
1024 * tree-data-ref.h (struct data_reference): Add alt_indices,
1026 * tree-data-ref.c (free_data_ref): Release alt_indices.
1027 (dr_analyze_indices): Work on struct indices and get DR_REF as tree.
1028 (create_data_ref): Adjust.
1029 (initialize_data_dependence_relation): Split into head
1030 and tail. When the base objects fail to match up try
1031 again with pointer-based analysis of indices.
1032 * tree-vectorizer.c (vec_info_shared::check_datarefs): Do
1033 not compare the lazily computed alternate set of indices.
1035 2021-09-20 Iain Sandoe <iain@sandoe.co.uk>
1037 * gcc.c: Test for execute OK when we find the
1038 programs for assembler linker and dsymutil and those
1039 were specified at configure-time.
1041 2021-09-19 Martin Sebor <msebor@redhat.com>
1043 PR middle-end/102403
1044 * gimple-predicate-analysis.cc (predicate::init_from_control_deps):
1045 Correct a function pre/postcondition.
1047 2021-09-19 Martin Sebor <msebor@redhat.com>
1049 PR middle-end/102243
1050 * tree-ssa-strlen.c (get_range): Handle null cfun.
1052 2021-09-19 Iain Sandoe <iain@sandoe.co.uk>
1054 * config/darwin.h (LINK_COMMAND_SPEC_A): Use Darwin10
1055 unwinder shim as a convenience library.
1057 2021-09-19 Andrew Pinski <apinski@marvell.com>
1059 * doc/install.texi: Add note about
1060 binutils 2.35 is required for LTO usage.
1062 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
1064 * tree-ssa-threadbackward.c
1065 (back_threader_registry::register_path): Use push_edge.
1066 * tree-ssa-threadedge.c
1067 (jump_threader::thread_around_empty_blocks): Same.
1068 (jump_threader::thread_through_normal_block): Same.
1069 (jump_threader::thread_across_edge): Same. Also, use auto_bitmap.
1071 * tree-ssa-threadupdate.c
1072 (jt_path_registry::allocate_thread_edge): Remove.
1073 (jt_path_registry::push_edge): New.
1074 (dump_jump_thread_path): Make static.
1075 * tree-ssa-threadupdate.h (allocate_thread_edge): Remove.
1078 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
1080 * gimple-range-path.cc (path_range_query::path_range_query): Add
1082 (path_range_query::dump): Remove extern declaration of dump_ranger.
1083 * gimple-range-trace.cc (dump_ranger): Add DEBUG_FUNCTION marker.
1084 * gimple-range-trace.h (dump_ranger): Add prototype.
1086 2021-09-19 John Ericson <git@JohnEricson.me>
1088 * gcc.c (find_a_program): New function, factored out of...
1089 (find_a_file): Here.
1090 (execute): Use find_a_program when looking for programs rather
1093 2021-09-19 Matwey V. Kornilov <matwey.kornilov@gmail.com>
1095 * config/avr/avr-mcus.def: Add atmega324pb.
1096 * doc/avr-mmcu.texi: Corresponding changes.
1098 2021-09-19 Roger Sayle <roger@nextmovesoftware.com>
1101 * match.pd (cmp @0 REAL_CST@1): When @0 is also REAL_CST, apply
1102 the same transformations as to @1. For comparisons against NaN,
1103 don't check HONOR_SNANS but confirm that neither operand is a
1106 2021-09-19 Benjamin Peterson <benjamin@locrian.net>
1108 * attribs.c (make_unique_name): Delete.
1109 * attribs.h (make_unique_name): Delete.
1111 2021-09-19 Andrew Pinski <apinski@marvell.com>
1113 * lra-constraints.c (check_and_process_move): Assert
1114 that dclass and sclass are greater than or equal to NO_REGS.
1116 2021-09-18 Jakub Jelinek <jakub@redhat.com>
1118 * tree.h (OMP_CLAUSE_ORDER_UNCONSTRAINED): Define.
1119 * tree-pretty-print.c (dump_omp_clause): Print unconstrained:
1120 for OMP_CLAUSE_ORDER_UNCONSTRAINED.
1122 2021-09-18 liuhongt <hongtao.liu@intel.com>
1124 * config/i386/i386-features.c (remove_partial_avx_dependency):
1125 Restrict TARGET_USE_VECTOR_FP_CONVERTS and
1126 TARGET_USE_VECTOR_CONVERTS to conversion instructions only.
1128 2021-09-18 Jakub Jelinek <jakub@redhat.com>
1130 * gimplify.c (omp_default_clause): For C/C++ default({,first}private),
1131 if file/namespace scope variable doesn't have predetermined sharing,
1132 treat it as if there was default(none).
1134 2021-09-18 liuhongt <hongtao.liu@intel.com>
1136 * config/i386/avx512fp16intrin.h (_mm_fmadd_sh):
1138 (_mm_mask_fmadd_sh): Likewise.
1139 (_mm_mask3_fmadd_sh): Likewise.
1140 (_mm_maskz_fmadd_sh): Likewise.
1141 (_mm_fmadd_round_sh): Likewise.
1142 (_mm_mask_fmadd_round_sh): Likewise.
1143 (_mm_mask3_fmadd_round_sh): Likewise.
1144 (_mm_maskz_fmadd_round_sh): Likewise.
1145 (_mm_fnmadd_sh): Likewise.
1146 (_mm_mask_fnmadd_sh): Likewise.
1147 (_mm_mask3_fnmadd_sh): Likewise.
1148 (_mm_maskz_fnmadd_sh): Likewise.
1149 (_mm_fnmadd_round_sh): Likewise.
1150 (_mm_mask_fnmadd_round_sh): Likewise.
1151 (_mm_mask3_fnmadd_round_sh): Likewise.
1152 (_mm_maskz_fnmadd_round_sh): Likewise.
1153 (_mm_fmsub_sh): Likewise.
1154 (_mm_mask_fmsub_sh): Likewise.
1155 (_mm_mask3_fmsub_sh): Likewise.
1156 (_mm_maskz_fmsub_sh): Likewise.
1157 (_mm_fmsub_round_sh): Likewise.
1158 (_mm_mask_fmsub_round_sh): Likewise.
1159 (_mm_mask3_fmsub_round_sh): Likewise.
1160 (_mm_maskz_fmsub_round_sh): Likewise.
1161 (_mm_fnmsub_sh): Likewise.
1162 (_mm_mask_fnmsub_sh): Likewise.
1163 (_mm_mask3_fnmsub_sh): Likewise.
1164 (_mm_maskz_fnmsub_sh): Likewise.
1165 (_mm_fnmsub_round_sh): Likewise.
1166 (_mm_mask_fnmsub_round_sh): Likewise.
1167 (_mm_mask3_fnmsub_round_sh): Likewise.
1168 (_mm_maskz_fnmsub_round_sh): Likewise.
1169 * config/i386/i386-builtin-types.def
1170 (V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT): New builtin type.
1171 * config/i386/i386-builtin.def: Add new builtins.
1172 * config/i386/i386-expand.c: Handle new builtin type.
1173 * config/i386/sse.md (fmai_vmfmadd_<mode><round_name>):
1174 Ajdust to support FP16.
1175 (fmai_vmfmsub_<mode><round_name>): Ditto.
1176 (fmai_vmfnmadd_<mode><round_name>): Ditto.
1177 (fmai_vmfnmsub_<mode><round_name>): Ditto.
1178 (*fmai_fmadd_<mode>): Ditto.
1179 (*fmai_fmsub_<mode>): Ditto.
1180 (*fmai_fnmadd_<mode><round_name>): Ditto.
1181 (*fmai_fnmsub_<mode><round_name>): Ditto.
1182 (avx512f_vmfmadd_<mode>_mask<round_name>): Ditto.
1183 (avx512f_vmfmadd_<mode>_mask3<round_name>): Ditto.
1184 (avx512f_vmfmadd_<mode>_maskz<round_expand_name>): Ditto.
1185 (avx512f_vmfmadd_<mode>_maskz_1<round_name>): Ditto.
1186 (*avx512f_vmfmsub_<mode>_mask<round_name>): Ditto.
1187 (avx512f_vmfmsub_<mode>_mask3<round_name>): Ditto.
1188 (*avx512f_vmfmsub_<mode>_maskz_1<round_name>): Ditto.
1189 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
1190 (*avx512f_vmfnmsub_<mode>_mask3<round_name>): Ditto.
1191 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
1192 (*avx512f_vmfnmadd_<mode>_mask<round_name>): Renamed to ...
1193 (avx512f_vmfnmadd_<mode>_mask<round_name>) ... this, and
1194 adjust to support FP16.
1195 (avx512f_vmfnmadd_<mode>_mask3<round_name>): Ditto.
1196 (avx512f_vmfnmadd_<mode>_maskz_1<round_name>): Ditto.
1197 (avx512f_vmfnmadd_<mode>_maskz<round_expand_name>): New
1200 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
1202 * config/i386/sse.md (avx512fmaskmodelower): Extend to support
1204 (maskload<mode><avx512fmaskmodelower>): Ditto.
1205 (maskstore<mode><avx512fmaskmodelower>): Ditto.
1207 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
1209 * config/i386/i386-expand.c (ix86_expand_fp_absneg_operator):
1211 (ix86_expand_copysign): Ditto.
1212 (ix86_expand_xorsign): Ditto.
1213 * config/i386/i386.c (ix86_build_const_vector): Handle HF vector
1215 (ix86_build_signbit_mask): Ditto.
1216 (ix86_can_change_mode_class): Ditto.
1217 * config/i386/i386.md
1218 (SSEMODEF): Add HFmode.
1219 (ssevecmodef): Ditto.
1220 (<code>hf2): New define_expand.
1221 (*<code>hf2_1): New define_insn_and_split.
1222 (copysign<mode>): Extend to support HFmode under AVX512FP16.
1223 (xorsign<mode>): Ditto.
1224 * config/i386/sse.md (VFB): New mode iterator.
1225 (VFB_128_256): Ditto.
1227 (sseintvecmode2): Support HF vector mode.
1228 (<code><mode>2): Use new mode iterator.
1229 (*<code><mode>2): Ditto.
1230 (copysign<mode>3): Ditto.
1231 (xorsign<mode>3): Ditto.
1232 (<code><mode>3<mask_name>): Ditto.
1233 (<code><mode>3<mask_name>): Ditto.
1234 (<sse>_andnot<mode>3<mask_name>): Adjust for HF vector mode.
1235 (<sse>_andnot<mode>3<mask_name>): Ditto.
1236 (*<code><mode>3<mask_name>): Ditto.
1237 (*<code><mode>3<mask_name>): Ditto.
1239 2021-09-18 liuhongt <hongtao.liu@intel.com>
1241 * config/i386/avx512fp16intrin.h (_mm512_mask_fmadd_ph):
1243 (_mm512_mask3_fmadd_ph): Likewise.
1244 (_mm512_maskz_fmadd_ph): Likewise.
1245 (_mm512_fmadd_round_ph): Likewise.
1246 (_mm512_mask_fmadd_round_ph): Likewise.
1247 (_mm512_mask3_fmadd_round_ph): Likewise.
1248 (_mm512_maskz_fmadd_round_ph): Likewise.
1249 (_mm512_fnmadd_ph): Likewise.
1250 (_mm512_mask_fnmadd_ph): Likewise.
1251 (_mm512_mask3_fnmadd_ph): Likewise.
1252 (_mm512_maskz_fnmadd_ph): Likewise.
1253 (_mm512_fnmadd_round_ph): Likewise.
1254 (_mm512_mask_fnmadd_round_ph): Likewise.
1255 (_mm512_mask3_fnmadd_round_ph): Likewise.
1256 (_mm512_maskz_fnmadd_round_ph): Likewise.
1257 (_mm512_fmsub_ph): Likewise.
1258 (_mm512_mask_fmsub_ph): Likewise.
1259 (_mm512_mask3_fmsub_ph): Likewise.
1260 (_mm512_maskz_fmsub_ph): Likewise.
1261 (_mm512_fmsub_round_ph): Likewise.
1262 (_mm512_mask_fmsub_round_ph): Likewise.
1263 (_mm512_mask3_fmsub_round_ph): Likewise.
1264 (_mm512_maskz_fmsub_round_ph): Likewise.
1265 (_mm512_fnmsub_ph): Likewise.
1266 (_mm512_mask_fnmsub_ph): Likewise.
1267 (_mm512_mask3_fnmsub_ph): Likewise.
1268 (_mm512_maskz_fnmsub_ph): Likewise.
1269 (_mm512_fnmsub_round_ph): Likewise.
1270 (_mm512_mask_fnmsub_round_ph): Likewise.
1271 (_mm512_mask3_fnmsub_round_ph): Likewise.
1272 (_mm512_maskz_fnmsub_round_ph): Likewise.
1273 * config/i386/avx512fp16vlintrin.h (_mm256_fmadd_ph):
1275 (_mm256_mask_fmadd_ph): Likewise.
1276 (_mm256_mask3_fmadd_ph): Likewise.
1277 (_mm256_maskz_fmadd_ph): Likewise.
1278 (_mm_fmadd_ph): Likewise.
1279 (_mm_mask_fmadd_ph): Likewise.
1280 (_mm_mask3_fmadd_ph): Likewise.
1281 (_mm_maskz_fmadd_ph): Likewise.
1282 (_mm256_fnmadd_ph): Likewise.
1283 (_mm256_mask_fnmadd_ph): Likewise.
1284 (_mm256_mask3_fnmadd_ph): Likewise.
1285 (_mm256_maskz_fnmadd_ph): Likewise.
1286 (_mm_fnmadd_ph): Likewise.
1287 (_mm_mask_fnmadd_ph): Likewise.
1288 (_mm_mask3_fnmadd_ph): Likewise.
1289 (_mm_maskz_fnmadd_ph): Likewise.
1290 (_mm256_fmsub_ph): Likewise.
1291 (_mm256_mask_fmsub_ph): Likewise.
1292 (_mm256_mask3_fmsub_ph): Likewise.
1293 (_mm256_maskz_fmsub_ph): Likewise.
1294 (_mm_fmsub_ph): Likewise.
1295 (_mm_mask_fmsub_ph): Likewise.
1296 (_mm_mask3_fmsub_ph): Likewise.
1297 (_mm_maskz_fmsub_ph): Likewise.
1298 (_mm256_fnmsub_ph): Likewise.
1299 (_mm256_mask_fnmsub_ph): Likewise.
1300 (_mm256_mask3_fnmsub_ph): Likewise.
1301 (_mm256_maskz_fnmsub_ph): Likewise.
1302 (_mm_fnmsub_ph): Likewise.
1303 (_mm_mask_fnmsub_ph): Likewise.
1304 (_mm_mask3_fnmsub_ph): Likewise.
1305 (_mm_maskz_fnmsub_ph): Likewise.
1306 * config/i386/i386-builtin.def: Add corresponding new builtins.
1307 * config/i386/sse.md
1308 (<avx512>_fmadd_<mode>_maskz<round_expand_name>): Adjust to
1309 support HF vector modes.
1310 (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
1312 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
1313 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
1314 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
1315 (<avx512>_fmadd_<mode>_mask<round_name>): Ditto.
1316 (<avx512>_fmadd_<mode>_mask3<round_name>): Ditto.
1317 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Ditto.
1318 (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>):
1320 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
1321 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
1322 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
1323 (<avx512>_fmsub_<mode>_mask<round_name>): Ditto.
1324 (<avx512>_fmsub_<mode>_mask3<round_name>): Ditto.
1325 (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>):
1327 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
1328 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
1329 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
1330 (<avx512>_fnmadd_<mode>_mask<round_name>): Ditto.
1331 (<avx512>_fnmadd_<mode>_mask3<round_name>): Ditto.
1332 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Ditto.
1333 (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>):
1335 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
1336 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
1337 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
1338 (<avx512>_fnmsub_<mode>_mask<round_name>): Ditto.
1339 (<avx512>_fnmsub_<mode>_mask3<round_name>): Ditto.
1341 2021-09-18 liuhongt <hongtao.liu@intel.com>
1343 * config/i386/avx512fp16intrin.h (_mm512_fmaddsub_ph):
1345 (_mm512_mask_fmaddsub_ph): Likewise.
1346 (_mm512_mask3_fmaddsub_ph): Likewise.
1347 (_mm512_maskz_fmaddsub_ph): Likewise.
1348 (_mm512_fmaddsub_round_ph): Likewise.
1349 (_mm512_mask_fmaddsub_round_ph): Likewise.
1350 (_mm512_mask3_fmaddsub_round_ph): Likewise.
1351 (_mm512_maskz_fmaddsub_round_ph): Likewise.
1352 (_mm512_mask_fmsubadd_ph): Likewise.
1353 (_mm512_mask3_fmsubadd_ph): Likewise.
1354 (_mm512_maskz_fmsubadd_ph): Likewise.
1355 (_mm512_fmsubadd_round_ph): Likewise.
1356 (_mm512_mask_fmsubadd_round_ph): Likewise.
1357 (_mm512_mask3_fmsubadd_round_ph): Likewise.
1358 (_mm512_maskz_fmsubadd_round_ph): Likewise.
1359 * config/i386/avx512fp16vlintrin.h (_mm256_fmaddsub_ph):
1361 (_mm256_mask_fmaddsub_ph): Likewise.
1362 (_mm256_mask3_fmaddsub_ph): Likewise.
1363 (_mm256_maskz_fmaddsub_ph): Likewise.
1364 (_mm_fmaddsub_ph): Likewise.
1365 (_mm_mask_fmaddsub_ph): Likewise.
1366 (_mm_mask3_fmaddsub_ph): Likewise.
1367 (_mm_maskz_fmaddsub_ph): Likewise.
1368 (_mm256_fmsubadd_ph): Likewise.
1369 (_mm256_mask_fmsubadd_ph): Likewise.
1370 (_mm256_mask3_fmsubadd_ph): Likewise.
1371 (_mm256_maskz_fmsubadd_ph): Likewise.
1372 (_mm_fmsubadd_ph): Likewise.
1373 (_mm_mask_fmsubadd_ph): Likewise.
1374 (_mm_mask3_fmsubadd_ph): Likewise.
1375 (_mm_maskz_fmsubadd_ph): Likewise.
1376 * config/i386/i386-builtin.def: Add corresponding new builtins.
1377 * config/i386/sse.md (VFH_SF_AVX512VL): New mode iterator.
1378 * (<avx512>_fmsubadd_<mode>_maskz<round_expand_name>): New expander.
1379 * (<avx512>_fmaddsub_<mode>_maskz<round_expand_name>): Use
1381 * (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
1383 * (<avx512>_fmaddsub_<mode>_mask<round_name>): Ditto.
1384 * (<avx512>_fmaddsub_<mode>_mask3<round_name>): Ditto.
1385 * (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
1387 * (<avx512>_fmsubadd_<mode>_mask<round_name>): Ditto.
1388 * (<avx512>_fmsubadd_<mode>_mask3<round_name>): Ditto.
1390 2021-09-18 liuhongt <hongtao.liu@intel.com>
1393 * config/i386/i386.c (ix86_print_operand): Handle
1394 V8HF/V16HF/V32HFmode.
1395 * config/i386/i386.h (VALID_BCST_MODE_P): Add HFmode.
1396 * config/i386/sse.md (avx512bcst): Remove.
1398 2021-09-17 Martin Sebor <msebor@redhat.com>
1400 * Makefile.in (OBJS): Add gimple-predicate-analysis.o.
1401 * tree-ssa-uninit.c (max_phi_args): Move to gimple-predicate-analysis.
1402 (MASK_SET_BIT, MASK_TEST_BIT, MASK_EMPTY): Same.
1403 (check_defs): Add comment.
1404 (can_skip_redundant_opnd): Update comment.
1405 (compute_uninit_opnds_pos): Adjust to namespace change.
1406 (find_pdom): Move to gimple-predicate-analysis.cc.
1408 (struct uninit_undef_val_t): New.
1409 (is_non_loop_exit_postdominating): Move to gimple-predicate-analysis.cc.
1410 (find_control_equiv_block): Same.
1411 (MAX_NUM_CHAINS, MAX_CHAIN_LEN, MAX_POSTDOM_CHECK): Same.
1412 (MAX_SWITCH_CASES): Same.
1413 (compute_control_dep_chain): Same.
1414 (find_uninit_use): Use predicate analyzer.
1415 (struct pred_info): Move to gimple-predicate-analysis.
1416 (convert_control_dep_chain_into_preds): Same.
1417 (find_predicates): Same.
1418 (collect_phi_def_edges): Same.
1419 (warn_uninitialized_phi): Use predicate analyzer.
1420 (find_def_preds): Move to gimple-predicate-analysis.
1421 (dump_pred_info): Same.
1422 (dump_pred_chain): Same.
1423 (dump_predicates): Same.
1424 (destroy_predicate_vecs): Remove.
1425 (execute_late_warn_uninitialized): New.
1426 (get_cmp_code): Move to gimple-predicate-analysis.
1427 (is_value_included_in): Same.
1428 (value_sat_pred_p): Same.
1429 (find_matching_predicate_in_rest_chains): Same.
1430 (is_use_properly_guarded): Same.
1431 (prune_uninit_phi_opnds): Same.
1432 (find_var_cmp_const): Same.
1433 (use_pred_not_overlap_with_undef_path_pred): Same.
1434 (pred_equal_p): Same.
1435 (is_neq_relop_p): Same.
1436 (is_neq_zero_form_p): Same.
1437 (pred_expr_equal_p): Same.
1438 (is_pred_expr_subset_of): Same.
1439 (is_pred_chain_subset_of): Same.
1440 (is_included_in): Same.
1441 (is_superset_of): Same.
1443 (simplify_pred): Same.
1444 (simplify_preds_2): Same.
1445 (simplify_preds_3): Same.
1446 (simplify_preds_4): Same.
1447 (simplify_preds): Same.
1449 (push_to_worklist): Same.
1450 (get_pred_info_from_cmp): Same.
1451 (is_degenerated_phi): Same.
1452 (normalize_one_pred_1): Same.
1453 (normalize_one_pred): Same.
1454 (normalize_one_pred_chain): Same.
1455 (normalize_preds): Same.
1456 (can_one_predicate_be_invalidated_p): Same.
1457 (can_chain_union_be_invalidated_p): Same.
1458 (uninit_uses_cannot_happen): Same.
1459 (pass_late_warn_uninitialized::execute): Define.
1460 * gimple-predicate-analysis.cc: New file.
1461 * gimple-predicate-analysis.h: New file.
1463 2021-09-17 Julian Brown <julian@codesourcery.com>
1465 * config/gcn/gcn.c (gimple.h): Include.
1466 (gcn_fork_join): Emit barrier for worker-level joins.
1467 * omp-oacc-neuter-broadcast.cc (find_local_vars_to_propagate): Add
1468 writes_gang_private bitmap parameter. Set bit for blocks
1469 containing gang-private variable writes.
1470 (worker_single_simple): Don't emit barrier after predicated block.
1471 (worker_single_copy): Don't emit barrier if we're not broadcasting
1472 anything and the block contains no gang-private writes.
1473 (neuter_worker_single): Don't predicate blocks that only contain
1474 NOPs or internal marker functions. Pass has_gang_private_write
1475 argument to worker_single_copy.
1476 (oacc_do_neutering): Add writes_gang_private bitmap handling.
1478 2021-09-17 Julian Brown <julian@codesourcery.com>
1480 * config/gcn/gcn-protos.h
1481 (gcn_goacc_create_worker_broadcast_record): Update prototype.
1482 * config/gcn/gcn-tree.c (gcn_goacc_get_worker_red_decl): Use
1483 preallocated block of LDS memory. Do not cache/share decls for
1484 reduction temporaries between invocations.
1485 (gcn_goacc_reduction_teardown): Unshare VAR on second use.
1486 (gcn_goacc_create_worker_broadcast_record): Add OFFSET parameter
1487 and return temporary LDS space at that offset. Return pointer in
1489 * config/gcn/gcn.c (acc_lds_size, gang_private_hwm, lds_allocs):
1491 (ACC_LDS_SIZE): Define as acc_lds_size.
1492 (gcn_init_machine_status): Don't initialise lds_allocated,
1493 lds_allocs, reduc_decls fields of machine function struct.
1494 (gcn_option_override): Handle default size for gang-private
1495 variables and -mgang-private-size option.
1496 (gcn_expand_prologue): Use LDS_SIZE instead of LDS_SIZE-1 when
1497 initialising M0_REG.
1498 (gcn_shared_mem_layout): New function.
1499 (gcn_print_lds_decl): Update comment. Use global lds_allocs map and
1500 gang_private_hwm variable.
1501 (TARGET_GOACC_SHARED_MEM_LAYOUT): Define target hook.
1502 * config/gcn/gcn.h (machine_function): Remove lds_allocated,
1503 lds_allocs, reduc_decls. Add reduction_base, reduction_limit.
1504 * config/gcn/gcn.opt (gang_private_size_opt): New global.
1505 (mgang-private-size=): New option.
1506 * doc/tm.texi.in (TARGET_GOACC_SHARED_MEM_LAYOUT): Place
1508 * doc/tm.texi: Regenerate.
1509 * omp-oacc-neuter-broadcast.cc (targhooks.h, diagnostic-core.h):
1511 (build_sender_ref): Handle sender_decl being pointer.
1512 (worker_single_copy): Add PLACEMENT and ISOLATE_BROADCASTS
1513 parameters. Pass placement argument to
1514 create_worker_broadcast_record hook invocations. Handle
1515 sender_decl being pointer and isolate_broadcasts inserting extra
1517 (blk_offset_map_t): Add typedef.
1518 (neuter_worker_single): Add BLK_OFFSET_MAP parameter. Pass
1519 preallocated range to worker_single_copy call.
1520 (dfs_broadcast_reachable_1): New function.
1521 (idx_decl_pair_t, used_range_vec_t): New typedefs.
1522 (sort_size_descending): New function.
1523 (addr_range): New class.
1524 (splay_tree_compare_addr_range, splay_tree_free_key)
1525 (first_fit_range, merge_ranges_1, merge_ranges): New functions.
1526 (execute_omp_oacc_neuter_broadcast): Rename to...
1527 (oacc_do_neutering): ... this. Add BOUNDS_LO, BOUNDS_HI
1528 parameters. Arrange layout of shared memory for broadcast
1530 (execute_omp_oacc_neuter_broadcast): New function.
1531 (pass_omp_oacc_neuter_broadcast::gate): Remove num_workers==1
1532 handling from here. Enable pass for all OpenACC routines in order
1533 to call shared memory-layout hook.
1534 * target.def (create_worker_broadcast_record): Add OFFSET
1536 (shared_mem_layout): New hook.
1538 2021-09-17 Julian Brown <julian@codesourcery.com>
1539 Thomas Schwinge <thomas@codesourcery.com>
1541 * omp-oacc-neuter-broadcast.cc
1542 (pass_omp_oacc_neuter_broadcast::gate): Disable if num_workers is
1544 (execute_omp_oacc_neuter_broadcast): Adjust.
1546 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
1548 * value-relation.cc (class equiv_chain): Move to header file.
1549 (path_oracle::path_oracle): New.
1550 (path_oracle::~path_oracle): New.
1551 (path_oracle::register_relation): New.
1552 (path_oracle::query_relation): New.
1553 (path_oracle::reset_path): New.
1554 (path_oracle::dump): New.
1555 * value-relation.h (class equiv_chain): Move to here.
1556 (class path_oracle): New.
1558 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
1560 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a DOM
1562 * gimple-range-fold.cc (fur_depend::register_relation): Use
1563 register_stmt/edge routines.
1564 * value-relation.cc (equiv_chain::find): Relocate from equiv_oracle.
1565 (equiv_oracle::equiv_oracle): Create self equivalence cache.
1566 (equiv_oracle::~equiv_oracle): Release same.
1567 (equiv_oracle::equiv_set): Return entry from self equiv cache if there
1568 are no equivalences.
1569 (equiv_oracle::find_equiv_block): Move list find to equiv_chain.
1570 (equiv_oracle::register_relation): Rename from register_equiv.
1571 (relation_chain_head::find_relation): Relocate from dom_oracle.
1572 (relation_oracle::register_stmt): New.
1573 (relation_oracle::register_edge): New.
1574 (dom_oracle::*): Rename from relation_oracle.
1575 (dom_oracle::register_relation): Adjust to call equiv_oracle.
1576 (dom_oracle::set_one_relation): Split from register_relation.
1577 (dom_oracle::register_transitives): Consolidate 2 methods.
1578 (dom_oracle::find_relation_block): Move core to relation_chain.
1579 (dom_oracle::query_relation): Rename from find_relation_dom and adjust.
1580 * value-relation.h (class relation_oracle): New pure virtual base.
1581 (class equiv_oracle): Inherit from relation_oracle and adjust.
1582 (class dom_oracle): Rename from old relation_oracle and adjust.
1584 2021-09-17 Martin Sebor <msebor@redhat.com>
1586 PR middle-end/102200
1587 * pointer-query.cc (access_ref::inform_access): Handle MIN/MAX_EXPR.
1588 (handle_min_max_size): Change argument. Store original SSA_NAME for
1589 operands to potentially distinct (sub)objects.
1590 (compute_objsize_r): Adjust call to the above.
1592 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1594 * config/rs6000/rs6000.c (rs6000-builtins.h): New include.
1595 (rs6000_new_builtin_vectorized_function): New function.
1596 (rs6000_new_builtin_md_vectorized_function): Likewise.
1597 (rs6000_builtin_vectorized_function): Call
1598 rs6000_new_builtin_vectorized_function.
1599 (rs6000_builtin_md_vectorized_function): Call
1600 rs6000_new_builtin_md_vectorized_function.
1602 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1604 * config/rs6000/rs6000-builtin-new.def (ASSEMBLE_ACC): Add mmaint flag.
1605 (ASSEMBLE_PAIR): Likewise.
1606 (BUILD_ACC): Likewise.
1607 (DISASSEMBLE_ACC): Likewise.
1608 (DISASSEMBLE_PAIR): Likewise.
1609 (PMXVBF16GER2): Likewise.
1610 (PMXVBF16GER2NN): Likewise.
1611 (PMXVBF16GER2NP): Likewise.
1612 (PMXVBF16GER2PN): Likewise.
1613 (PMXVBF16GER2PP): Likewise.
1614 (PMXVF16GER2): Likewise.
1615 (PMXVF16GER2NN): Likewise.
1616 (PMXVF16GER2NP): Likewise.
1617 (PMXVF16GER2PN): Likewise.
1618 (PMXVF16GER2PP): Likewise.
1619 (PMXVF32GER): Likewise.
1620 (PMXVF32GERNN): Likewise.
1621 (PMXVF32GERNP): Likewise.
1622 (PMXVF32GERPN): Likewise.
1623 (PMXVF32GERPP): Likewise.
1624 (PMXVF64GER): Likewise.
1625 (PMXVF64GERNN): Likewise.
1626 (PMXVF64GERNP): Likewise.
1627 (PMXVF64GERPN): Likewise.
1628 (PMXVF64GERPP): Likewise.
1629 (PMXVI16GER2): Likewise.
1630 (PMXVI16GER2PP): Likewise.
1631 (PMXVI16GER2S): Likewise.
1632 (PMXVI16GER2SPP): Likewise.
1633 (PMXVI4GER8): Likewise.
1634 (PMXVI4GER8PP): Likewise.
1635 (PMXVI8GER4): Likewise.
1636 (PMXVI8GER4PP): Likewise.
1637 (PMXVI8GER4SPP): Likewise.
1638 (XVBF16GER2): Likewise.
1639 (XVBF16GER2NN): Likewise.
1640 (XVBF16GER2NP): Likewise.
1641 (XVBF16GER2PN): Likewise.
1642 (XVBF16GER2PP): Likewise.
1643 (XVF16GER2): Likewise.
1644 (XVF16GER2NN): Likewise.
1645 (XVF16GER2NP): Likewise.
1646 (XVF16GER2PN): Likewise.
1647 (XVF16GER2PP): Likewise.
1648 (XVF32GER): Likewise.
1649 (XVF32GERNN): Likewise.
1650 (XVF32GERNP): Likewise.
1651 (XVF32GERPN): Likewise.
1652 (XVF32GERPP): Likewise.
1653 (XVF64GER): Likewise.
1654 (XVF64GERNN): Likewise.
1655 (XVF64GERNP): Likewise.
1656 (XVF64GERPN): Likewise.
1657 (XVF64GERPP): Likewise.
1658 (XVI16GER2): Likewise.
1659 (XVI16GER2PP): Likewise.
1660 (XVI16GER2S): Likewise.
1661 (XVI16GER2SPP): Likewise.
1662 (XVI4GER8): Likewise.
1663 (XVI4GER8PP): Likewise.
1664 (XVI8GER4): Likewise.
1665 (XVI8GER4PP): Likewise.
1666 (XVI8GER4SPP): Likewise.
1667 (XXMFACC): Likewise.
1668 (XXMTACC): Likewise.
1669 (XXSETACCZ): Likewise.
1670 (ASSEMBLE_PAIR_V): Likewise.
1671 (BUILD_PAIR): Likewise.
1672 (DISASSEMBLE_PAIR_V): Likewise.
1675 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_mma_builtin):
1676 Handle RS6000_BIF_LXVP and RS6000_BIF_STXVP.
1677 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add ismmaint.
1678 (parse_bif_attrs): Handle ismmaint.
1679 (write_decls): Add bif_mmaint_bit and bif_is_mmaint.
1680 (write_bif_static_init): Handle ismmaint.
1682 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1684 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin): New
1686 (rs6000_gimple_fold_builtin): Call rs6000_gimple_fold_new_builtin.
1687 (rs6000_new_builtin_valid_without_lhs): New function.
1688 (rs6000_gimple_fold_new_mma_builtin): Likewise.
1689 (rs6000_gimple_fold_new_builtin): Likewise.
1691 2021-09-17 Thomas Schwinge <thomas@codesourcery.com>
1693 * hash-table.h (hash_table<Descriptor, Lazy, Allocator>::expand):
1694 Destruct stale Value objects.
1695 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
1698 2021-09-17 Roger Sayle <roger@nextmovesoftware.com>
1701 * match.pd (shift optimizations): Disable recent sign-changing
1702 optimization for shifts by zero, these will be folded later.
1704 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1706 * config/rs6000/rs6000-builtin-new.def (__builtin_mffsl): Move from
1707 [power9] to [always].
1709 2021-09-17 Richard Biener <rguenther@suse.de>
1711 * tree-vect-stmts.c (vectorizable_load): Do not frob
1714 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
1716 * config/i386/i386-features.c (remove_partial_avx_dependency):
1717 Also check TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY and
1718 and TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY before generating
1720 * config/i386/i386.h (TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY):
1722 (TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
1723 * config/i386/i386.md (SSE FP to FP splitters): Replace
1724 TARGET_SSE_PARTIAL_REG_DEPENDENCY with
1725 TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY.
1726 (SSE INT to FP splitter): Replace TARGET_SSE_PARTIAL_REG_DEPENDENCY
1727 with TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY.
1728 * config/i386/x86-tune.def
1729 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): New.
1730 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
1732 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
1735 * config/i386/i386-features.c (remove_partial_avx_dependency):
1736 Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
1737 before generating vxorps.
1739 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
1741 * config/i386/i386-options.c (processor_cost_table): Use
1742 tremont_cost for Tremont.
1743 * config/i386/x86-tune-costs.h (tremont_memcpy): New.
1744 (tremont_memset): Likewise.
1745 (tremont_cost): Likewise.
1746 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
1749 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
1751 * common/config/i386/i386-common.c: Use Haswell scheduling model
1753 * config/i386/i386.c (ix86_sched_init_global): Prepare for Tremont
1755 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Tremont
1757 (ix86_adjust_cost): Handle Tremont.
1758 * config/i386/x86-tune.def (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY):
1760 (X86_TUNE_USE_LEAVE): Likewise.
1761 (X86_TUNE_PUSH_MEMORY): Likewise.
1762 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
1763 (X86_TUNE_USE_CLTD): Likewise.
1764 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
1765 (X86_TUNE_AVOID_MFENCE): Likewise.
1766 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
1767 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
1768 (X86_TUNE_ACCUMULATE_OUTGOING_ARGS): Disable for Tremont.
1769 (X86_TUNE_FOUR_JUMP_LIMIT): Likewise.
1770 (X86_TUNE_OPT_AGU): Likewise.
1771 (X86_TUNE_AVOID_LEA_FOR_ADDR): Likewise.
1772 (X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE): Likewise.
1773 (X86_TUNE_EXPAND_ABS): Likewise.
1774 (X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS): Likewise.
1775 (X86_TUNE_SLOW_PSHUFB): Likewise.
1777 2021-09-17 Eric Botcazou <ebotcazou@adacore.com>
1779 PR rtl-optimization/102306
1780 * combine.c (try_combine): Abort the combination if we are about to
1781 duplicate volatile references.
1783 2021-09-17 liuhongt <hongtao.liu@intel.com>
1785 * config/i386/avx512fp16intrin.h (_mm_undefined_ph):
1787 (_mm256_undefined_ph): Likewise.
1788 (_mm512_undefined_ph): Likewise.
1789 (_mm_cvtsh_h): Likewise.
1790 (_mm256_cvtsh_h): Likewise.
1791 (_mm512_cvtsh_h): Likewise.
1792 (_mm512_castph_ps): Likewise.
1793 (_mm512_castph_pd): Likewise.
1794 (_mm512_castph_si512): Likewise.
1795 (_mm512_castph512_ph128): Likewise.
1796 (_mm512_castph512_ph256): Likewise.
1797 (_mm512_castph128_ph512): Likewise.
1798 (_mm512_castph256_ph512): Likewise.
1799 (_mm512_zextph128_ph512): Likewise.
1800 (_mm512_zextph256_ph512): Likewise.
1801 (_mm512_castps_ph): Likewise.
1802 (_mm512_castpd_ph): Likewise.
1803 (_mm512_castsi512_ph): Likewise.
1804 * config/i386/avx512fp16vlintrin.h (_mm_castph_ps):
1806 (_mm256_castph_ps): Likewise.
1807 (_mm_castph_pd): Likewise.
1808 (_mm256_castph_pd): Likewise.
1809 (_mm_castph_si128): Likewise.
1810 (_mm256_castph_si256): Likewise.
1811 (_mm_castps_ph): Likewise.
1812 (_mm256_castps_ph): Likewise.
1813 (_mm_castpd_ph): Likewise.
1814 (_mm256_castpd_ph): Likewise.
1815 (_mm_castsi128_ph): Likewise.
1816 (_mm256_castsi256_ph): Likewise.
1817 (_mm256_castph256_ph128): Likewise.
1818 (_mm256_castph128_ph256): Likewise.
1819 (_mm256_zextph128_ph256): Likewise.
1821 2021-09-17 liuhongt <hongtao.liu@intel.com>
1823 * config/i386/avx512fp16intrin.h (_mm_cvtsh_ss):
1825 (_mm_mask_cvtsh_ss): Likewise.
1826 (_mm_maskz_cvtsh_ss): Likewise.
1827 (_mm_cvtsh_sd): Likewise.
1828 (_mm_mask_cvtsh_sd): Likewise.
1829 (_mm_maskz_cvtsh_sd): Likewise.
1830 (_mm_cvt_roundsh_ss): Likewise.
1831 (_mm_mask_cvt_roundsh_ss): Likewise.
1832 (_mm_maskz_cvt_roundsh_ss): Likewise.
1833 (_mm_cvt_roundsh_sd): Likewise.
1834 (_mm_mask_cvt_roundsh_sd): Likewise.
1835 (_mm_maskz_cvt_roundsh_sd): Likewise.
1836 (_mm_cvtss_sh): Likewise.
1837 (_mm_mask_cvtss_sh): Likewise.
1838 (_mm_maskz_cvtss_sh): Likewise.
1839 (_mm_cvtsd_sh): Likewise.
1840 (_mm_mask_cvtsd_sh): Likewise.
1841 (_mm_maskz_cvtsd_sh): Likewise.
1842 (_mm_cvt_roundss_sh): Likewise.
1843 (_mm_mask_cvt_roundss_sh): Likewise.
1844 (_mm_maskz_cvt_roundss_sh): Likewise.
1845 (_mm_cvt_roundsd_sh): Likewise.
1846 (_mm_mask_cvt_roundsd_sh): Likewise.
1847 (_mm_maskz_cvt_roundsd_sh): Likewise.
1848 * config/i386/i386-builtin-types.def
1849 (V8HF_FTYPE_V2DF_V8HF_V8HF_UQI_INT,
1850 V8HF_FTYPE_V4SF_V8HF_V8HF_UQI_INT,
1851 V2DF_FTYPE_V8HF_V2DF_V2DF_UQI_INT,
1852 V4SF_FTYPE_V8HF_V4SF_V4SF_UQI_INT): Add new builtin types.
1853 * config/i386/i386-builtin.def: Add corrresponding new builtins.
1854 * config/i386/i386-expand.c: Handle new builtin types.
1855 * config/i386/sse.md (VF48_128): New mode iterator.
1856 (avx512fp16_vcvtsh2<ssescalarmodesuffix><mask_scalar_name><round_saeonly_scalar_name>):
1858 (avx512fp16_vcvt<ssescalarmodesuffix>2sh<mask_scalar_name><round_scalar_name>):
1861 2021-09-17 liuhongt <hongtao.liu@intel.com>
1863 * config/i386/avx512fp16intrin.h (_mm512_cvtph_pd):
1865 (_mm512_mask_cvtph_pd): Likewise.
1866 (_mm512_maskz_cvtph_pd): Likewise.
1867 (_mm512_cvt_roundph_pd): Likewise.
1868 (_mm512_mask_cvt_roundph_pd): Likewise.
1869 (_mm512_maskz_cvt_roundph_pd): Likewise.
1870 (_mm512_cvtxph_ps): Likewise.
1871 (_mm512_mask_cvtxph_ps): Likewise.
1872 (_mm512_maskz_cvtxph_ps): Likewise.
1873 (_mm512_cvtx_roundph_ps): Likewise.
1874 (_mm512_mask_cvtx_roundph_ps): Likewise.
1875 (_mm512_maskz_cvtx_roundph_ps): Likewise.
1876 (_mm512_cvtxps_ph): Likewise.
1877 (_mm512_mask_cvtxps_ph): Likewise.
1878 (_mm512_maskz_cvtxps_ph): Likewise.
1879 (_mm512_cvtx_roundps_ph): Likewise.
1880 (_mm512_mask_cvtx_roundps_ph): Likewise.
1881 (_mm512_maskz_cvtx_roundps_ph): Likewise.
1882 (_mm512_cvtpd_ph): Likewise.
1883 (_mm512_mask_cvtpd_ph): Likewise.
1884 (_mm512_maskz_cvtpd_ph): Likewise.
1885 (_mm512_cvt_roundpd_ph): Likewise.
1886 (_mm512_mask_cvt_roundpd_ph): Likewise.
1887 (_mm512_maskz_cvt_roundpd_ph): Likewise.
1888 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_pd):
1890 (_mm_mask_cvtph_pd): Likewise.
1891 (_mm_maskz_cvtph_pd): Likewise.
1892 (_mm256_cvtph_pd): Likewise.
1893 (_mm256_mask_cvtph_pd): Likewise.
1894 (_mm256_maskz_cvtph_pd): Likewise.
1895 (_mm_cvtxph_ps): Likewise.
1896 (_mm_mask_cvtxph_ps): Likewise.
1897 (_mm_maskz_cvtxph_ps): Likewise.
1898 (_mm256_cvtxph_ps): Likewise.
1899 (_mm256_mask_cvtxph_ps): Likewise.
1900 (_mm256_maskz_cvtxph_ps): Likewise.
1901 (_mm_cvtxps_ph): Likewise.
1902 (_mm_mask_cvtxps_ph): Likewise.
1903 (_mm_maskz_cvtxps_ph): Likewise.
1904 (_mm256_cvtxps_ph): Likewise.
1905 (_mm256_mask_cvtxps_ph): Likewise.
1906 (_mm256_maskz_cvtxps_ph): Likewise.
1907 (_mm_cvtpd_ph): Likewise.
1908 (_mm_mask_cvtpd_ph): Likewise.
1909 (_mm_maskz_cvtpd_ph): Likewise.
1910 (_mm256_cvtpd_ph): Likewise.
1911 (_mm256_mask_cvtpd_ph): Likewise.
1912 (_mm256_maskz_cvtpd_ph): Likewise.
1913 * config/i386/i386-builtin.def: Add corresponding new builtins.
1914 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
1915 * config/i386/i386-expand.c: Handle new builtin types.
1916 * config/i386/sse.md
1917 (VF4_128_8_256): New.
1918 (VF48H_AVX512VL): Ditto.
1919 (ssePHmode): Add HF vector modes.
1920 (castmode): Add new convertable modes.
1923 (avx512fp16_vcvt<castmode>2ph_<mode><mask_name><round_name>): Ditto.
1924 (avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
1925 (*avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
1926 (avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
1927 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
1928 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask_1): Ditto.
1929 (avx512fp16_float_extend_ph<mode>2<mask_name><round_saeonly_name>):
1931 (avx512fp16_float_extend_ph<mode>2<mask_name>): Ditto.
1932 (*avx512fp16_float_extend_ph<mode>2_load<mask_name>): Ditto.
1933 (avx512fp16_float_extend_phv2df2<mask_name>): Ditto.
1934 (*avx512fp16_float_extend_phv2df2_load<mask_name>): Ditto.
1936 2021-09-17 liuhongt <hongtao.liu@intel.com>
1938 * config/i386/avx512fp16intrin.h (_mm_cvttsh_i32):
1940 (_mm_cvttsh_u32): Likewise.
1941 (_mm_cvtt_roundsh_i32): Likewise.
1942 (_mm_cvtt_roundsh_u32): Likewise.
1943 (_mm_cvttsh_i64): Likewise.
1944 (_mm_cvttsh_u64): Likewise.
1945 (_mm_cvtt_roundsh_i64): Likewise.
1946 (_mm_cvtt_roundsh_u64): Likewise.
1947 * config/i386/i386-builtin.def: Add corresponding new builtins.
1948 * config/i386/sse.md
1949 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<round_saeonly_name>):
1952 2021-09-17 liuhongt <hongtao.liu@intel.com>
1954 * config/i386/avx512fp16intrin.h (_mm512_cvttph_epi32):
1956 (_mm512_mask_cvttph_epi32): Likewise.
1957 (_mm512_maskz_cvttph_epi32): Likewise.
1958 (_mm512_cvtt_roundph_epi32): Likewise.
1959 (_mm512_mask_cvtt_roundph_epi32): Likewise.
1960 (_mm512_maskz_cvtt_roundph_epi32): Likewise.
1961 (_mm512_cvttph_epu32): Likewise.
1962 (_mm512_mask_cvttph_epu32): Likewise.
1963 (_mm512_maskz_cvttph_epu32): Likewise.
1964 (_mm512_cvtt_roundph_epu32): Likewise.
1965 (_mm512_mask_cvtt_roundph_epu32): Likewise.
1966 (_mm512_maskz_cvtt_roundph_epu32): Likewise.
1967 (_mm512_cvttph_epi64): Likewise.
1968 (_mm512_mask_cvttph_epi64): Likewise.
1969 (_mm512_maskz_cvttph_epi64): Likewise.
1970 (_mm512_cvtt_roundph_epi64): Likewise.
1971 (_mm512_mask_cvtt_roundph_epi64): Likewise.
1972 (_mm512_maskz_cvtt_roundph_epi64): Likewise.
1973 (_mm512_cvttph_epu64): Likewise.
1974 (_mm512_mask_cvttph_epu64): Likewise.
1975 (_mm512_maskz_cvttph_epu64): Likewise.
1976 (_mm512_cvtt_roundph_epu64): Likewise.
1977 (_mm512_mask_cvtt_roundph_epu64): Likewise.
1978 (_mm512_maskz_cvtt_roundph_epu64): Likewise.
1979 (_mm512_cvttph_epi16): Likewise.
1980 (_mm512_mask_cvttph_epi16): Likewise.
1981 (_mm512_maskz_cvttph_epi16): Likewise.
1982 (_mm512_cvtt_roundph_epi16): Likewise.
1983 (_mm512_mask_cvtt_roundph_epi16): Likewise.
1984 (_mm512_maskz_cvtt_roundph_epi16): Likewise.
1985 (_mm512_cvttph_epu16): Likewise.
1986 (_mm512_mask_cvttph_epu16): Likewise.
1987 (_mm512_maskz_cvttph_epu16): Likewise.
1988 (_mm512_cvtt_roundph_epu16): Likewise.
1989 (_mm512_mask_cvtt_roundph_epu16): Likewise.
1990 (_mm512_maskz_cvtt_roundph_epu16): Likewise.
1991 * config/i386/avx512fp16vlintrin.h (_mm_cvttph_epi32):
1993 (_mm_mask_cvttph_epi32): Likewise.
1994 (_mm_maskz_cvttph_epi32): Likewise.
1995 (_mm256_cvttph_epi32): Likewise.
1996 (_mm256_mask_cvttph_epi32): Likewise.
1997 (_mm256_maskz_cvttph_epi32): Likewise.
1998 (_mm_cvttph_epu32): Likewise.
1999 (_mm_mask_cvttph_epu32): Likewise.
2000 (_mm_maskz_cvttph_epu32): Likewise.
2001 (_mm256_cvttph_epu32): Likewise.
2002 (_mm256_mask_cvttph_epu32): Likewise.
2003 (_mm256_maskz_cvttph_epu32): Likewise.
2004 (_mm_cvttph_epi64): Likewise.
2005 (_mm_mask_cvttph_epi64): Likewise.
2006 (_mm_maskz_cvttph_epi64): Likewise.
2007 (_mm256_cvttph_epi64): Likewise.
2008 (_mm256_mask_cvttph_epi64): Likewise.
2009 (_mm256_maskz_cvttph_epi64): Likewise.
2010 (_mm_cvttph_epu64): Likewise.
2011 (_mm_mask_cvttph_epu64): Likewise.
2012 (_mm_maskz_cvttph_epu64): Likewise.
2013 (_mm256_cvttph_epu64): Likewise.
2014 (_mm256_mask_cvttph_epu64): Likewise.
2015 (_mm256_maskz_cvttph_epu64): Likewise.
2016 (_mm_cvttph_epi16): Likewise.
2017 (_mm_mask_cvttph_epi16): Likewise.
2018 (_mm_maskz_cvttph_epi16): Likewise.
2019 (_mm256_cvttph_epi16): Likewise.
2020 (_mm256_mask_cvttph_epi16): Likewise.
2021 (_mm256_maskz_cvttph_epi16): Likewise.
2022 (_mm_cvttph_epu16): Likewise.
2023 (_mm_mask_cvttph_epu16): Likewise.
2024 (_mm_maskz_cvttph_epu16): Likewise.
2025 (_mm256_cvttph_epu16): Likewise.
2026 (_mm256_mask_cvttph_epu16): Likewise.
2027 (_mm256_maskz_cvttph_epu16): Likewise.
2028 * config/i386/i386-builtin.def: Add new builtins.
2029 * config/i386/sse.md
2030 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name><round_saeonly_name>):
2032 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name>): Ditto.
2033 (*avx512fp16_fix<fixunssuffix>_trunc<mode>2_load<mask_name>): Ditto.
2034 (avx512fp16_fix<fixunssuffix>_truncv2di2<mask_name>): Ditto.
2035 (avx512fp16_fix<fixunssuffix>_truncv2di2_load<mask_name>): Ditto.
2037 2021-09-17 liuhongt <hongtao.liu@intel.com>
2039 * config/i386/avx512fp16intrin.h (_mm_cvtsh_i32): New intrinsic.
2040 (_mm_cvtsh_u32): Likewise.
2041 (_mm_cvt_roundsh_i32): Likewise.
2042 (_mm_cvt_roundsh_u32): Likewise.
2043 (_mm_cvtsh_i64): Likewise.
2044 (_mm_cvtsh_u64): Likewise.
2045 (_mm_cvt_roundsh_i64): Likewise.
2046 (_mm_cvt_roundsh_u64): Likewise.
2047 (_mm_cvti32_sh): Likewise.
2048 (_mm_cvtu32_sh): Likewise.
2049 (_mm_cvt_roundi32_sh): Likewise.
2050 (_mm_cvt_roundu32_sh): Likewise.
2051 (_mm_cvti64_sh): Likewise.
2052 (_mm_cvtu64_sh): Likewise.
2053 (_mm_cvt_roundi64_sh): Likewise.
2054 (_mm_cvt_roundu64_sh): Likewise.
2055 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2056 * config/i386/i386-builtin.def: Add corresponding new builtins.
2057 * config/i386/i386-expand.c (ix86_expand_round_builtin):
2058 Handle new builtin types.
2059 * config/i386/sse.md
2060 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix><round_name>):
2062 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix>_2): Likewise.
2063 (avx512fp16_vcvt<floatsuffix>si2sh<rex64namesuffix><round_name>): Likewise.
2065 2021-09-16 Bill Schmidt <wschmidt@linux.ibm.com>
2067 * config/rs6000/rs6000-c.c (rs6000-builtins.h): New include.
2068 (altivec_resolve_new_overloaded_builtin): New forward decl.
2069 (rs6000_new_builtin_type_compatible): New function.
2070 (altivec_resolve_overloaded_builtin): Call
2071 altivec_resolve_new_overloaded_builtin.
2072 (altivec_build_new_resolved_builtin): New function.
2073 (altivec_resolve_new_overloaded_builtin): Likewise.
2074 * config/rs6000/rs6000-call.c (rs6000_new_builtin_is_supported):
2076 * config/rs6000/rs6000-gen-builtins.c (write_decls): Remove _p from
2077 name of rs6000_new_builtin_is_supported.
2079 2021-09-16 Uroš Bizjak <ubizjak@gmail.com>
2081 * config/i386/i386-protos.h (ix86_decompose_address):
2082 Change return type to bool.
2083 * config/i386/i386.c (ix86_decompose_address): Ditto.
2085 2021-09-16 Tobias Burnus <tobias@codesourcery.com>
2088 * config/rs6000/t-rs6000 (build/rs6000-gen-builtins.o, build/rbtree.o):
2089 Added 'build/' to target, use build/%.o rule.
2090 (build/rs6000-gen-builtins$(build_exeext)): Add 'build/' and
2091 '$(build_exeext)' to target and 'build/' for the *.o files.
2092 (rs6000-builtins.c): Update for those changes; run rs6000-gen-builtins
2095 2021-09-16 Martin Jambor <mjambor@suse.cz>
2097 * cgraph.c (cgraph_node::dump): Do not check caller count sums if
2098 the body has been removed. Remove trailing whitespace.
2100 2021-09-16 Richard Biener <rguenther@suse.de>
2102 PR middle-end/102360
2103 * internal-fn.c (expand_DEFERRED_INIT): Make pattern-init
2104 of non-memory more robust.
2106 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2108 * config/sparc/sparc-opts.h (enum sparc_processor_type): Add LEON5
2109 * config/sparc/sparc.c (struct processor_costs): Add LEON5 costs
2110 (leon5_adjust_cost): Increase cost of store with data dependency
2111 on ALU instruction and FPU anti-dependencies.
2112 (sparc_option_override): Add LEON5 costs
2113 (sparc_adjust_cost): Add LEON5 cost adjustments
2114 * config/sparc/sparc.h: Add LEON5
2115 * config/sparc/sparc.md: Include LEON5 scheduling information
2116 * config/sparc/sparc.opt: Add LEON5
2117 * doc/invoke.texi: Add LEON5
2118 * config/sparc/leon5.md: New file.
2120 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2122 * config/sparc/sparc.md (stack_protect_set32): Add NOP to prevent
2123 sensitive sequence for B2BST errata workaround.
2125 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2127 * config/sparc/sparc.c (sparc_do_work_around_errata): Do not begin
2128 functions with atomic instruction in the UT700 errata workaround.
2130 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2132 * config/sparc/sparc.c (next_active_non_empty_insn): New function
2133 that returns next active non empty assembly instruction.
2134 (sparc_do_work_around_errata): Use new function.
2136 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2138 * config/sparc/sparc.c (store_insn_p): Add predicate for store
2140 (load_insn_p): Add predicate for load attributes.
2141 (sparc_do_work_around_errata): Use new predicates.
2143 2021-09-16 Andreas Larsson <andreas@gaisler.com>
2145 * config/sparc/sparc.c (dump_target_flag_bits): Print bit names for
2148 2021-09-16 Martin Liska <mliska@suse.cz>
2150 * config/mips/netbsd.h: Fix typo in name of a macro.
2152 2021-09-16 liuhongt <hongtao.liu@intel.com>
2154 PR middle-end/102080
2155 * match.pd: Check mask type when doing cond_op related gimple
2157 * tree.c (is_truth_type_for): New function.
2158 * tree.h (is_truth_type_for): New declaration.
2160 2021-09-16 liuhongt <hongtao.liu@intel.com>
2162 * config/i386/avx512fp16intrin.h (_mm512_cvtepi32_ph): New
2164 (_mm512_mask_cvtepi32_ph): Likewise.
2165 (_mm512_maskz_cvtepi32_ph): Likewise.
2166 (_mm512_cvt_roundepi32_ph): Likewise.
2167 (_mm512_mask_cvt_roundepi32_ph): Likewise.
2168 (_mm512_maskz_cvt_roundepi32_ph): Likewise.
2169 (_mm512_cvtepu32_ph): Likewise.
2170 (_mm512_mask_cvtepu32_ph): Likewise.
2171 (_mm512_maskz_cvtepu32_ph): Likewise.
2172 (_mm512_cvt_roundepu32_ph): Likewise.
2173 (_mm512_mask_cvt_roundepu32_ph): Likewise.
2174 (_mm512_maskz_cvt_roundepu32_ph): Likewise.
2175 (_mm512_cvtepi64_ph): Likewise.
2176 (_mm512_mask_cvtepi64_ph): Likewise.
2177 (_mm512_maskz_cvtepi64_ph): Likewise.
2178 (_mm512_cvt_roundepi64_ph): Likewise.
2179 (_mm512_mask_cvt_roundepi64_ph): Likewise.
2180 (_mm512_maskz_cvt_roundepi64_ph): Likewise.
2181 (_mm512_cvtepu64_ph): Likewise.
2182 (_mm512_mask_cvtepu64_ph): Likewise.
2183 (_mm512_maskz_cvtepu64_ph): Likewise.
2184 (_mm512_cvt_roundepu64_ph): Likewise.
2185 (_mm512_mask_cvt_roundepu64_ph): Likewise.
2186 (_mm512_maskz_cvt_roundepu64_ph): Likewise.
2187 (_mm512_cvtepi16_ph): Likewise.
2188 (_mm512_mask_cvtepi16_ph): Likewise.
2189 (_mm512_maskz_cvtepi16_ph): Likewise.
2190 (_mm512_cvt_roundepi16_ph): Likewise.
2191 (_mm512_mask_cvt_roundepi16_ph): Likewise.
2192 (_mm512_maskz_cvt_roundepi16_ph): Likewise.
2193 (_mm512_cvtepu16_ph): Likewise.
2194 (_mm512_mask_cvtepu16_ph): Likewise.
2195 (_mm512_maskz_cvtepu16_ph): Likewise.
2196 (_mm512_cvt_roundepu16_ph): Likewise.
2197 (_mm512_mask_cvt_roundepu16_ph): Likewise.
2198 (_mm512_maskz_cvt_roundepu16_ph): Likewise.
2199 * config/i386/avx512fp16vlintrin.h (_mm_cvtepi32_ph): New
2201 (_mm_mask_cvtepi32_ph): Likewise.
2202 (_mm_maskz_cvtepi32_ph): Likewise.
2203 (_mm256_cvtepi32_ph): Likewise.
2204 (_mm256_mask_cvtepi32_ph): Likewise.
2205 (_mm256_maskz_cvtepi32_ph): Likewise.
2206 (_mm_cvtepu32_ph): Likewise.
2207 (_mm_mask_cvtepu32_ph): Likewise.
2208 (_mm_maskz_cvtepu32_ph): Likewise.
2209 (_mm256_cvtepu32_ph): Likewise.
2210 (_mm256_mask_cvtepu32_ph): Likewise.
2211 (_mm256_maskz_cvtepu32_ph): Likewise.
2212 (_mm_cvtepi64_ph): Likewise.
2213 (_mm_mask_cvtepi64_ph): Likewise.
2214 (_mm_maskz_cvtepi64_ph): Likewise.
2215 (_mm256_cvtepi64_ph): Likewise.
2216 (_mm256_mask_cvtepi64_ph): Likewise.
2217 (_mm256_maskz_cvtepi64_ph): Likewise.
2218 (_mm_cvtepu64_ph): Likewise.
2219 (_mm_mask_cvtepu64_ph): Likewise.
2220 (_mm_maskz_cvtepu64_ph): Likewise.
2221 (_mm256_cvtepu64_ph): Likewise.
2222 (_mm256_mask_cvtepu64_ph): Likewise.
2223 (_mm256_maskz_cvtepu64_ph): Likewise.
2224 (_mm_cvtepi16_ph): Likewise.
2225 (_mm_mask_cvtepi16_ph): Likewise.
2226 (_mm_maskz_cvtepi16_ph): Likewise.
2227 (_mm256_cvtepi16_ph): Likewise.
2228 (_mm256_mask_cvtepi16_ph): Likewise.
2229 (_mm256_maskz_cvtepi16_ph): Likewise.
2230 (_mm_cvtepu16_ph): Likewise.
2231 (_mm_mask_cvtepu16_ph): Likewise.
2232 (_mm_maskz_cvtepu16_ph): Likewise.
2233 (_mm256_cvtepu16_ph): Likewise.
2234 (_mm256_mask_cvtepu16_ph): Likewise.
2235 (_mm256_maskz_cvtepu16_ph): Likewise.
2236 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2237 * config/i386/i386-builtin.def: Add corresponding new builtins.
2238 * config/i386/i386-expand.c
2239 (ix86_expand_args_builtin): Handle new builtin types.
2240 (ix86_expand_round_builtin): Ditto.
2241 * config/i386/i386-modes.def: Declare V2HF and V6HF.
2242 * config/i386/sse.md (VI2H_AVX512VL): New.
2244 (sseintvecmode): Add HF vector modes.
2245 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode><mask_name><round_name>):
2247 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
2248 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
2249 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
2250 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
2251 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask_1): Ditto.
2252 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
2253 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
2254 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
2255 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
2256 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask_1): Ditto.
2257 * config/i386/subst.md (round_qq2phsuff): New subst_attr.
2259 2021-09-16 liuhongt <hongtao.liu@intel.com>
2261 * config/i386/avx512fp16intrin.h (_mm512_cvtph_epi32):
2263 (_mm512_mask_cvtph_epi32): Likewise.
2264 (_mm512_maskz_cvtph_epi32): Likewise.
2265 (_mm512_cvt_roundph_epi32): Likewise.
2266 (_mm512_mask_cvt_roundph_epi32): Likewise.
2267 (_mm512_maskz_cvt_roundph_epi32): Likewise.
2268 (_mm512_cvtph_epu32): Likewise.
2269 (_mm512_mask_cvtph_epu32): Likewise.
2270 (_mm512_maskz_cvtph_epu32): Likewise.
2271 (_mm512_cvt_roundph_epu32): Likewise.
2272 (_mm512_mask_cvt_roundph_epu32): Likewise.
2273 (_mm512_maskz_cvt_roundph_epu32): Likewise.
2274 (_mm512_cvtph_epi64): Likewise.
2275 (_mm512_mask_cvtph_epi64): Likewise.
2276 (_mm512_maskz_cvtph_epi64): Likewise.
2277 (_mm512_cvt_roundph_epi64): Likewise.
2278 (_mm512_mask_cvt_roundph_epi64): Likewise.
2279 (_mm512_maskz_cvt_roundph_epi64): Likewise.
2280 (_mm512_cvtph_epu64): Likewise.
2281 (_mm512_mask_cvtph_epu64): Likewise.
2282 (_mm512_maskz_cvtph_epu64): Likewise.
2283 (_mm512_cvt_roundph_epu64): Likewise.
2284 (_mm512_mask_cvt_roundph_epu64): Likewise.
2285 (_mm512_maskz_cvt_roundph_epu64): Likewise.
2286 (_mm512_cvtph_epi16): Likewise.
2287 (_mm512_mask_cvtph_epi16): Likewise.
2288 (_mm512_maskz_cvtph_epi16): Likewise.
2289 (_mm512_cvt_roundph_epi16): Likewise.
2290 (_mm512_mask_cvt_roundph_epi16): Likewise.
2291 (_mm512_maskz_cvt_roundph_epi16): Likewise.
2292 (_mm512_cvtph_epu16): Likewise.
2293 (_mm512_mask_cvtph_epu16): Likewise.
2294 (_mm512_maskz_cvtph_epu16): Likewise.
2295 (_mm512_cvt_roundph_epu16): Likewise.
2296 (_mm512_mask_cvt_roundph_epu16): Likewise.
2297 (_mm512_maskz_cvt_roundph_epu16): Likewise.
2298 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_epi32):
2300 (_mm_mask_cvtph_epi32): Likewise.
2301 (_mm_maskz_cvtph_epi32): Likewise.
2302 (_mm256_cvtph_epi32): Likewise.
2303 (_mm256_mask_cvtph_epi32): Likewise.
2304 (_mm256_maskz_cvtph_epi32): Likewise.
2305 (_mm_cvtph_epu32): Likewise.
2306 (_mm_mask_cvtph_epu32): Likewise.
2307 (_mm_maskz_cvtph_epu32): Likewise.
2308 (_mm256_cvtph_epu32): Likewise.
2309 (_mm256_mask_cvtph_epu32): Likewise.
2310 (_mm256_maskz_cvtph_epu32): Likewise.
2311 (_mm_cvtph_epi64): Likewise.
2312 (_mm_mask_cvtph_epi64): Likewise.
2313 (_mm_maskz_cvtph_epi64): Likewise.
2314 (_mm256_cvtph_epi64): Likewise.
2315 (_mm256_mask_cvtph_epi64): Likewise.
2316 (_mm256_maskz_cvtph_epi64): Likewise.
2317 (_mm_cvtph_epu64): Likewise.
2318 (_mm_mask_cvtph_epu64): Likewise.
2319 (_mm_maskz_cvtph_epu64): Likewise.
2320 (_mm256_cvtph_epu64): Likewise.
2321 (_mm256_mask_cvtph_epu64): Likewise.
2322 (_mm256_maskz_cvtph_epu64): Likewise.
2323 (_mm_cvtph_epi16): Likewise.
2324 (_mm_mask_cvtph_epi16): Likewise.
2325 (_mm_maskz_cvtph_epi16): Likewise.
2326 (_mm256_cvtph_epi16): Likewise.
2327 (_mm256_mask_cvtph_epi16): Likewise.
2328 (_mm256_maskz_cvtph_epi16): Likewise.
2329 (_mm_cvtph_epu16): Likewise.
2330 (_mm_mask_cvtph_epu16): Likewise.
2331 (_mm_maskz_cvtph_epu16): Likewise.
2332 (_mm256_cvtph_epu16): Likewise.
2333 (_mm256_mask_cvtph_epu16): Likewise.
2334 (_mm256_maskz_cvtph_epu16): Likewise.
2335 * config/i386/i386-builtin-types.def: Add new builtin types.
2336 * config/i386/i386-builtin.def: Add new builtins.
2337 * config/i386/i386-expand.c
2338 (ix86_expand_args_builtin): Handle new builtin types.
2339 (ix86_expand_round_builtin): Ditto.
2340 * config/i386/sse.md (sseintconvert): New.
2342 (UNSPEC_US_FIX_NOTRUNC): Ditto.
2343 (sseintconvertsignprefix): Ditto.
2344 (avx512fp16_vcvtph2<sseintconvertsignprefix><sseintconvert>_<mode><mask_name><round_name>):
2347 2021-09-16 liuhongt <hongtao.liu@intel.com>
2349 * config/i386/avx512fp16intrin.h: (_mm_cvtsi16_si128):
2351 (_mm_cvtsi128_si16): Likewise.
2352 (_mm_mask_load_sh): Likewise.
2353 (_mm_maskz_load_sh): Likewise.
2354 (_mm_mask_store_sh): Likewise.
2355 (_mm_move_sh): Likewise.
2356 (_mm_mask_move_sh): Likewise.
2357 (_mm_maskz_move_sh): Likewise.
2358 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2359 * config/i386/i386-builtin.def: Add corresponding new builtins.
2360 * config/i386/i386-expand.c
2361 (ix86_expand_special_args_builtin): Handle new builtin types.
2362 (ix86_expand_vector_init_one_nonzero): Adjust for FP16 target.
2363 * config/i386/sse.md (VI2F): New mode iterator.
2364 (vec_set<mode>_0): Use new mode iterator.
2365 (avx512f_mov<ssescalarmodelower>_mask): Adjust for HF vector mode.
2366 (avx512f_store<mode>_mask): Ditto.
2368 2021-09-16 Kewen Lin <linkw@linux.ibm.com>
2370 * config/rs6000/rs6000.opt (-mtoc-fusion): Remove.
2372 2021-09-15 David Edelsohn <dje.gcc@gmail.com>
2374 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
2375 Proceed if no symbol summary or the symbol alias flag is false.
2377 2021-09-15 Jakub Jelinek <jakub@redhat.com>
2381 * varasm.c (output_constructor_regular_field): Instead of assertion
2382 that array_size_for_constructor result is equal to size of
2383 TREE_TYPE (local->val) in bytes, assert that the type size is greater
2384 or equal to array_size_for_constructor result and use type size as
2387 2021-09-15 Martin Liska <mliska@suse.cz>
2390 * config/i386/vxworks.h: Use new macro TARGET_CPU_P.
2392 2021-09-15 Martin Liska <mliska@suse.cz>
2395 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
2396 Check that we have a symbol summary for a symbol.
2398 2021-09-15 Richard Biener <rguenther@suse.de>
2401 * config/rs6000/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
2402 to inherit from elfos.h
2404 2021-09-15 liuhongt <hongtao.liu@intel.com>
2407 * config/i386/i386-expand.c
2408 (ix86_expand_vector_init_interleave): Use puncklwd to pack 2
2410 (ix86_expand_vector_set): Use blendw instead of pinsrw.
2411 * config/i386/i386.c (ix86_can_change_mode_class): Adjust for
2412 AVX512FP16 which supports 16bit vector load.
2413 * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>):
2415 (avx512bw_interleave_high<mode><mask_name>): .. this, and
2416 extend to V32HFmode.
2417 (avx2_interleave_highv16hi<mask_name>): Rename to ..
2418 (avx2_interleave_high<mode><mask_name>): .. this, and extend
2420 (vec_interleave_highv8hi<mask_name>): Rename to ..
2421 (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode.
2422 (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>):
2424 (<mask_codefor>avx512bw_interleave_low<mode><mask_name>):
2425 this, and extend to V32HFmode.
2426 (avx2_interleave_lowv16hi<mask_name>): Rename to ..
2427 (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode.
2428 (vec_interleave_lowv8hi<mask_name>): Rename to ..
2429 (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode.
2430 (sse4_1_pblendw): Rename to ..
2431 (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode.
2432 (avx2_pblendph): New define_expand.
2433 (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use
2434 sseintmodesuffix instead of ssemodesuffix.
2435 (blendsuf): New mode attr.
2437 2021-09-15 Richard Biener <rguenther@suse.de>
2439 * tree-vectorizer.h (dr_misalignment): Move out of line.
2440 (dr_target_alignment): New.
2441 (DR_TARGET_ALIGNMENT): Wrap dr_target_alignment.
2442 (set_dr_target_alignment): New.
2443 (SET_DR_TARGET_ALIGNMENT): Wrap set_dr_target_alignment.
2444 * tree-vect-data-refs.c (dr_misalignment): Compute and
2445 return the group members misalignment.
2446 (vect_compute_data_ref_alignment): Use SET_DR_TARGET_ALIGNMENT.
2447 (vect_analyze_data_refs_alignment): Compute alignment only
2448 for the first element of a DR group.
2449 (vect_slp_analyze_node_alignment): Likewise.
2451 2021-09-15 Hongyu Wang <hongyu.wang@intel.com>
2453 * config/i386/avx512fp16intrin.h: Adjust all builtin calls.
2454 * config/i386/avx512fp16vlintrin.h: Likewise.
2455 * config/i386/i386-builtin.def: Adjust builtin name and
2456 enumeration to match AVX512F style.
2458 2021-09-15 Richard Biener <rguenther@suse.de>
2460 PR tree-optimization/102318
2461 * tree-vect-loop.c (vect_transform_cycle_phi): Revert
2462 previous change and do the mode conversion separately from
2463 the sign conversion.
2465 2021-09-15 Hongtao Liu <hongtao.liu@intel.com>
2466 Peter Cordes <peter@cordes.ca>
2469 * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
2470 (*vec_extract<mode><ssescalarmodelower>_valign): Output
2471 vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
2474 2021-09-15 Richard Biener <rguenther@suse.de>
2476 * config.gcc: Remove vax-*-openbsd* configuration.
2478 2021-09-15 Richard Biener <rguenther@suse.de>
2480 * config.gcc: Remove m68k-openbsd.
2482 2021-09-15 Max Filippov <jcmvbkbc@gmail.com>
2485 * config/xtensa/t-xtensa (TM_H): Add include/xtensa-config.h.
2487 2021-09-14 Peter Bergner <bergner@linux.ibm.com>
2489 * config/rs6000/mma.md (unspec): Delete UNSPEC_MMA_XXSETACCZ.
2490 (unspecv): Add UNSPECV_MMA_XXSETACCZ.
2491 (*mma_xxsetaccz): Delete.
2492 (mma_xxsetaccz): Change to define_insn. Remove operand 1.
2493 Use UNSPECV_MMA_XXSETACCZ. Update comment.
2494 * config/rs6000/rs6000.c (rs6000_rtx_costs): Use UNSPECV_MMA_XXSETACCZ.
2496 2021-09-14 Iain Sandoe <iain@sandoe.co.uk>
2498 * Makefile.in: Remove variables related to applying no-PIE
2499 to the exes on $build.
2500 * configure: Regenerate.
2501 * configure.ac: Remove configuration related to applying
2502 no-PIE to the exes on $build.
2504 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
2506 * config/arc/arc.md (doloop_end): Add missing mode.
2507 (loop_end): Likewise.
2509 2021-09-14 Jakub Jelinek <jakub@redhat.com>
2511 * gimplify.c (goa_stabilize_expr): Add depth argument, propagate
2512 it to recursive calls, for depth above 7 just gimplify or return.
2513 Perform a test even for MODIFY_EXPR, ADDR_EXPR, COMPOUND_EXPR with
2514 __builtin_clear_padding and TARGET_EXPR.
2515 (gimplify_omp_atomic): Adjust goa_stabilize_expr callers.
2517 2021-09-14 liuhongt <hongtao.liu@intel.com>
2519 * config/i386/avx512fp16intrin.h (_mm_fpclass_sh_mask):
2521 (_mm_mask_fpclass_sh_mask): Likewise.
2522 (_mm512_mask_fpclass_ph_mask): Likewise.
2523 (_mm512_fpclass_ph_mask): Likewise.
2524 (_mm_getexp_sh): Likewise.
2525 (_mm_mask_getexp_sh): Likewise.
2526 (_mm_maskz_getexp_sh): Likewise.
2527 (_mm512_getexp_ph): Likewise.
2528 (_mm512_mask_getexp_ph): Likewise.
2529 (_mm512_maskz_getexp_ph): Likewise.
2530 (_mm_getexp_round_sh): Likewise.
2531 (_mm_mask_getexp_round_sh): Likewise.
2532 (_mm_maskz_getexp_round_sh): Likewise.
2533 (_mm512_getexp_round_ph): Likewise.
2534 (_mm512_mask_getexp_round_ph): Likewise.
2535 (_mm512_maskz_getexp_round_ph): Likewise.
2536 (_mm_getmant_sh): Likewise.
2537 (_mm_mask_getmant_sh): Likewise.
2538 (_mm_maskz_getmant_sh): Likewise.
2539 (_mm512_getmant_ph): Likewise.
2540 (_mm512_mask_getmant_ph): Likewise.
2541 (_mm512_maskz_getmant_ph): Likewise.
2542 (_mm_getmant_round_sh): Likewise.
2543 (_mm_mask_getmant_round_sh): Likewise.
2544 (_mm_maskz_getmant_round_sh): Likewise.
2545 (_mm512_getmant_round_ph): Likewise.
2546 (_mm512_mask_getmant_round_ph): Likewise.
2547 (_mm512_maskz_getmant_round_ph): Likewise.
2548 * config/i386/avx512fp16vlintrin.h (_mm_mask_fpclass_ph_mask):
2550 (_mm_fpclass_ph_mask): Likewise.
2551 (_mm256_mask_fpclass_ph_mask): Likewise.
2552 (_mm256_fpclass_ph_mask): Likewise.
2553 (_mm256_getexp_ph): Likewise.
2554 (_mm256_mask_getexp_ph): Likewise.
2555 (_mm256_maskz_getexp_ph): Likewise.
2556 (_mm_getexp_ph): Likewise.
2557 (_mm_mask_getexp_ph): Likewise.
2558 (_mm_maskz_getexp_ph): Likewise.
2559 (_mm256_getmant_ph): Likewise.
2560 (_mm256_mask_getmant_ph): Likewise.
2561 (_mm256_maskz_getmant_ph): Likewise.
2562 (_mm_getmant_ph): Likewise.
2563 (_mm_mask_getmant_ph): Likewise.
2564 (_mm_maskz_getmant_ph): Likewise.
2565 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2566 * config/i386/i386-builtin.def: Add corresponding new builtins.
2567 * config/i386/i386-expand.c
2568 (ix86_expand_args_builtin): Handle new builtin types.
2569 (ix86_expand_round_builtin): Ditto.
2570 * config/i386/sse.md (vecmemsuffix): Add HF vector modes.
2571 (<avx512>_getexp<mode><mask_name><round_saeonly_name>): Adjust
2572 to support HF vector modes.
2573 (avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name):
2575 (avx512dq_fpclass<mode><mask_scalar_merge_name>): Ditto.
2576 (avx512dq_vmfpclass<mode><mask_scalar_merge_name>): Ditto.
2577 (<avx512>_getmant<mode><mask_name><round_saeonly_name>): Ditto.
2578 (avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2581 2021-09-14 liuhongt <hongtao.liu@intel.com>
2583 * config/i386/avx512fp16intrin.h (_mm512_reduce_ph):
2585 (_mm512_mask_reduce_ph): Likewise.
2586 (_mm512_maskz_reduce_ph): Likewise.
2587 (_mm512_reduce_round_ph): Likewise.
2588 (_mm512_mask_reduce_round_ph): Likewise.
2589 (_mm512_maskz_reduce_round_ph): Likewise.
2590 (_mm_reduce_sh): Likewise.
2591 (_mm_mask_reduce_sh): Likewise.
2592 (_mm_maskz_reduce_sh): Likewise.
2593 (_mm_reduce_round_sh): Likewise.
2594 (_mm_mask_reduce_round_sh): Likewise.
2595 (_mm_maskz_reduce_round_sh): Likewise.
2596 (_mm512_roundscale_ph): Likewise.
2597 (_mm512_mask_roundscale_ph): Likewise.
2598 (_mm512_maskz_roundscale_ph): Likewise.
2599 (_mm512_roundscale_round_ph): Likewise.
2600 (_mm512_mask_roundscale_round_ph): Likewise.
2601 (_mm512_maskz_roundscale_round_ph): Likewise.
2602 (_mm_roundscale_sh): Likewise.
2603 (_mm_mask_roundscale_sh): Likewise.
2604 (_mm_maskz_roundscale_sh): Likewise.
2605 (_mm_roundscale_round_sh): Likewise.
2606 (_mm_mask_roundscale_round_sh): Likewise.
2607 (_mm_maskz_roundscale_round_sh): Likewise.
2608 * config/i386/avx512fp16vlintrin.h: (_mm_reduce_ph):
2610 (_mm_mask_reduce_ph): Likewise.
2611 (_mm_maskz_reduce_ph): Likewise.
2612 (_mm256_reduce_ph): Likewise.
2613 (_mm256_mask_reduce_ph): Likewise.
2614 (_mm256_maskz_reduce_ph): Likewise.
2615 (_mm_roundscale_ph): Likewise.
2616 (_mm_mask_roundscale_ph): Likewise.
2617 (_mm_maskz_roundscale_ph): Likewise.
2618 (_mm256_roundscale_ph): Likewise.
2619 (_mm256_mask_roundscale_ph): Likewise.
2620 (_mm256_maskz_roundscale_ph): Likewise.
2621 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2622 * config/i386/i386-builtin.def: Add corresponding new builtins.
2623 * config/i386/i386-expand.c
2624 (ix86_expand_args_builtin): Handle new builtin types.
2625 (ix86_expand_round_builtin): Ditto.
2626 * config/i386/sse.md (<mask_codefor>reducep<mode><mask_name>):
2628 (<mask_codefor>reducep<mode><mask_name><round_saeonly_name>):
2629 ... this, and adjust for round operands.
2630 (reduces<mode><mask_scalar_name>): Likewise, with ...
2631 (reduces<mode><mask_scalar_name><round_saeonly_scalar_name):
2633 (<avx512>_rndscale<mode><mask_name><round_saeonly_name>):
2634 Adjust for HF vector modes.
2635 (avx512f_rndscale<mode><mask_scalar_name><round_saeonly_scalar_name>):
2637 (*avx512f_rndscale<mode><round_saeonly_name>): Ditto.
2639 2021-09-14 liuhongt <hongtao.liu@intel.com>
2641 * config/i386/avx512fp16intrin.h: (_mm512_rcp_ph):
2643 (_mm512_mask_rcp_ph): Likewise.
2644 (_mm512_maskz_rcp_ph): Likewise.
2645 (_mm_rcp_sh): Likewise.
2646 (_mm_mask_rcp_sh): Likewise.
2647 (_mm_maskz_rcp_sh): Likewise.
2648 (_mm512_scalef_ph): Likewise.
2649 (_mm512_mask_scalef_ph): Likewise.
2650 (_mm512_maskz_scalef_ph): Likewise.
2651 (_mm512_scalef_round_ph): Likewise.
2652 (_mm512_mask_scalef_round_ph): Likewise.
2653 (_mm512_maskz_scalef_round_ph): Likewise.
2654 (_mm_scalef_sh): Likewise.
2655 (_mm_mask_scalef_sh): Likewise.
2656 (_mm_maskz_scalef_sh): Likewise.
2657 (_mm_scalef_round_sh): Likewise.
2658 (_mm_mask_scalef_round_sh): Likewise.
2659 (_mm_maskz_scalef_round_sh): Likewise.
2660 * config/i386/avx512fp16vlintrin.h (_mm_rcp_ph):
2662 (_mm256_rcp_ph): Likewise.
2663 (_mm_mask_rcp_ph): Likewise.
2664 (_mm256_mask_rcp_ph): Likewise.
2665 (_mm_maskz_rcp_ph): Likewise.
2666 (_mm256_maskz_rcp_ph): Likewise.
2667 (_mm_scalef_ph): Likewise.
2668 (_mm256_scalef_ph): Likewise.
2669 (_mm_mask_scalef_ph): Likewise.
2670 (_mm256_mask_scalef_ph): Likewise.
2671 (_mm_maskz_scalef_ph): Likewise.
2672 (_mm256_maskz_scalef_ph): Likewise.
2673 * config/i386/i386-builtin.def: Add new builtins.
2674 * config/i386/sse.md (VFH_AVX512VL): New.
2675 (avx512fp16_rcp<mode>2<mask_name>): Ditto.
2676 (avx512fp16_vmrcpv8hf2<mask_scalar_name>): Ditto.
2677 (avx512f_vmscalef<mode><mask_scalar_name><round_scalar_name>):
2678 Adjust to support HF vector modes.
2679 (<avx512>_scalef<mode><mask_name><round_name>): Ditto.
2681 2021-09-14 liuhongt <hongtao.liu@intel.com>
2683 * config/i386/avx512fp16intrin.h: (_mm512_sqrt_ph):
2685 (_mm512_mask_sqrt_ph): Likewise.
2686 (_mm512_maskz_sqrt_ph): Likewise.
2687 (_mm512_sqrt_round_ph): Likewise.
2688 (_mm512_mask_sqrt_round_ph): Likewise.
2689 (_mm512_maskz_sqrt_round_ph): Likewise.
2690 (_mm512_rsqrt_ph): Likewise.
2691 (_mm512_mask_rsqrt_ph): Likewise.
2692 (_mm512_maskz_rsqrt_ph): Likewise.
2693 (_mm_rsqrt_sh): Likewise.
2694 (_mm_mask_rsqrt_sh): Likewise.
2695 (_mm_maskz_rsqrt_sh): Likewise.
2696 (_mm_sqrt_sh): Likewise.
2697 (_mm_mask_sqrt_sh): Likewise.
2698 (_mm_maskz_sqrt_sh): Likewise.
2699 (_mm_sqrt_round_sh): Likewise.
2700 (_mm_mask_sqrt_round_sh): Likewise.
2701 (_mm_maskz_sqrt_round_sh): Likewise.
2702 * config/i386/avx512fp16vlintrin.h (_mm_sqrt_ph): New intrinsic.
2703 (_mm256_sqrt_ph): Likewise.
2704 (_mm_mask_sqrt_ph): Likewise.
2705 (_mm256_mask_sqrt_ph): Likewise.
2706 (_mm_maskz_sqrt_ph): Likewise.
2707 (_mm256_maskz_sqrt_ph): Likewise.
2708 (_mm_rsqrt_ph): Likewise.
2709 (_mm256_rsqrt_ph): Likewise.
2710 (_mm_mask_rsqrt_ph): Likewise.
2711 (_mm256_mask_rsqrt_ph): Likewise.
2712 (_mm_maskz_rsqrt_ph): Likewise.
2713 (_mm256_maskz_rsqrt_ph): Likewise.
2714 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2715 * config/i386/i386-builtin.def: Add corresponding new builtins.
2716 * config/i386/i386-expand.c
2717 (ix86_expand_args_builtin): Handle new builtins.
2718 (ix86_expand_round_builtin): Ditto.
2719 * config/i386/sse.md (VF_AVX512FP16VL): New.
2720 (sqrt<mode>2): Adjust for HF vector modes.
2721 (<sse>_sqrt<mode>2<mask_name><round_name>): Likewise.
2722 (<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
2724 (<sse>_rsqrt<mode>2<mask_name>): New.
2725 (avx512fp16_vmrsqrtv8hf2<mask_scalar_name>): Likewise.
2727 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
2730 * diagnostic-spec.c (warning_suppressed_at, copy_warning): Handle
2731 'RESERVED_LOCATION_P' locations.
2732 * warning-control.cc (get_nowarn_spec, suppress_warning)
2733 (copy_warning): Likewise.
2735 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
2737 * diagnostic-spec.h (typedef xint_hash_t): Use 'location_t' instead of...
2738 (typedef key_type_t): ... this. Remove.
2739 (nowarn_map): Document.
2740 * diagnostic-spec.c (nowarn_map): Likewise.
2741 * warning-control.cc (convert_to_key): Evolve functions into...
2742 (get_location): ... these. Adjust all users.
2744 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
2746 * warning-control.cc (copy_warning): Remove 'nowarn_map' setup.
2748 2021-09-13 Jason Merrill <jason@redhat.com>
2750 * params.opt: Add destructive-interference-size and
2751 constructive-interference-size.
2752 * doc/invoke.texi: Document them.
2753 * config/aarch64/aarch64.c (aarch64_override_options_internal):
2755 * config/arm/arm.c (arm_option_override): Set them.
2756 * config/i386/i386-options.c (ix86_option_override_internal):
2759 2021-09-13 Martin Liska <mliska@suse.cz>
2760 H.J. Lu <hjl.tools@gmail.com>
2763 * common/config/i386/cpuinfo.h (cpu_indicator_init): Add support
2764 for x86-64 micro levels for __builtin_cpu_supports.
2765 * common/config/i386/i386-cpuinfo.h (enum feature_priority):
2766 Add priorities for the micro-arch levels.
2767 (enum processor_features): Add new features.
2768 * common/config/i386/i386-isas.h: Add micro-arch features.
2769 * config/i386/i386-builtins.c (get_builtin_code_for_version):
2770 Support the micro-arch levels by callsing
2771 __builtin_cpu_supports.
2772 * doc/extend.texi: Document that the levels are support by
2773 __builtin_cpu_supports.
2775 2021-09-13 Andrew Pinski <apinski@marvell.com>
2778 * config/aarch64/aarch64-builtins.c (aarch64_fold_builtin_lane_check):
2780 (aarch64_general_fold_builtin): Handle AARCH64_SIMD_BUILTIN_LANE_CHECK.
2781 (aarch64_general_gimple_fold_builtin): Likewise.
2783 2021-09-13 Andrew Pinski <apinski@marvell.com>
2785 * config.gcc: Add m32r-*-linux* and m32rle-*-linux*
2786 to the Unsupported targets list.
2787 Remove support for m32r-*-linux* and m32rle-*-linux*.
2788 * config/m32r/linux.h: Removed.
2789 * config/m32r/t-linux: Removed.
2791 2021-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2794 * config/aarch64/aarch64.c (aarch64_classify_address): Don't allow
2795 register index for SVE predicate modes.
2797 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
2799 * tree-ssa-threadbackward.c
2800 (back_threader_profitability::profitable_path_p): Remove FSM
2802 (back_threader_registry::register_path): Same.
2803 * tree-ssa-threadedge.c
2804 (jump_threader::simplify_control_stmt_condition): Same.
2805 * tree-ssa-threadupdate.c (jt_path_registry::jt_path_registry):
2806 Add backedge_threads argument.
2807 (fwd_jt_path_registry::fwd_jt_path_registry): Pass
2808 backedge_threads argument.
2809 (back_jt_path_registry::back_jt_path_registry): Same.
2810 (dump_jump_thread_path): Adjust for FSM removal.
2811 (back_jt_path_registry::rewire_first_differing_edge): Same.
2812 (back_jt_path_registry::adjust_paths_after_duplication): Same.
2813 (back_jt_path_registry::update_cfg): Same.
2814 (jt_path_registry::register_jump_thread): Same.
2815 * tree-ssa-threadupdate.h (enum jump_thread_edge_type): Remove
2817 (class back_jt_path_registry): Add backedge_threads to
2820 2021-09-13 Martin Liska <mliska@suse.cz>
2823 * asan.h (sanitize_coverage_p): Handle when fn == NULL.
2825 2021-09-13 H.J. Lu <hjl.tools@gmail.com>
2828 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): New.
2829 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
2830 (MOVE_MAX): Check TARGET_AVX256_MOVE_BY_PIECES and
2831 TARGET_AVX256_STORE_BY_PIECES instead of
2832 TARGET_AVX256_SPLIT_UNALIGNED_LOAD and
2833 TARGET_AVX256_SPLIT_UNALIGNED_STORE.
2834 (STORE_MAX_PIECES): Check TARGET_AVX256_STORE_BY_PIECES instead
2835 of TARGET_AVX256_SPLIT_UNALIGNED_STORE.
2836 * config/i386/x86-tune.def (X86_TUNE_AVX256_MOVE_BY_PIECES): New.
2837 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
2839 2021-09-13 liuhongt <hongtao.liu@intel.com>
2842 * expmed.c (extract_bit_field_using_extv): Use
2843 gen_lowpart_if_possible instead of gen_lowpart to avoid ICE.
2845 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
2847 * Makefile.in (OBJS): Add value-pointer-equiv.o.
2848 * gimple-ssa-evrp.c (class ssa_equiv_stack): Move to
2849 value-pointer-equiv.*.
2850 (ssa_equiv_stack::ssa_equiv_stack): Same.
2851 (ssa_equiv_stack::enter): Same.
2852 (ssa_equiv_stack::leave): Same.
2853 (ssa_equiv_stack::push_replacement): Same.
2854 (ssa_equiv_stack::get_replacement): Same.
2855 (is_pointer_ssa): Same.
2856 (class pointer_equiv_analyzer): Same.
2857 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
2858 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Same.
2859 (pointer_equiv_analyzer::set_global_equiv): Same.
2860 (pointer_equiv_analyzer::set_cond_equiv): Same.
2861 (pointer_equiv_analyzer::get_equiv): Same.
2862 (pointer_equiv_analyzer::enter): Same.
2863 (pointer_equiv_analyzer::leave): Same.
2864 (pointer_equiv_analyzer::get_equiv_expr): Same.
2865 (pta_valueize): Same.
2866 (pointer_equiv_analyzer::visit_stmt): Same.
2867 (pointer_equiv_analyzer::visit_edge): Same.
2868 (hybrid_folder::value_of_expr): Same.
2869 (hybrid_folder::value_on_edge): Same.
2870 * value-pointer-equiv.cc: New file.
2871 * value-pointer-equiv.h: New file.
2873 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
2876 * gimple-fold.c (gimple_fold_builtin_memory_op): Allow folding
2877 memcpy if the size is not more than MOVE_MAX * MOVE_RATIO.
2879 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
2882 * config/arm/arm.md (movmisaligndi): New define_expand.
2883 * config/arm/vec-common.md (movmisalign<mode>): Iterate over VDQ mode.
2885 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
2888 * emit-rtl.c (gen_highpart): Use adjust_address to handle
2889 MEM rather than calling simplify_gen_subreg.
2891 2021-09-13 Jan-Benedict Glaw <jbglaw@ług-owl.de>
2893 * config/alpha/vms.h (INIT_CUMULATIVE_ARGS): Wrap multi-statment
2894 define into a block.
2896 2021-09-13 Richard Biener <rguenther@suse.de>
2898 * config/darwin.h (DARWIN_PREFER_DWARF): Do not define.
2899 * config/i386/darwin.h (PREFERRED_DEBUGGING_TYPE): Do not
2900 change based on DARWIN_PREFER_DWARF not being defined.
2902 2021-09-13 Richard Biener <rguenther@suse.de>
2904 * config/i386/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
2905 to inherit from elfos.h
2907 2021-09-13 Richard Biener <rguenther@suse.de>
2909 * config.gcc: Add cr16-*-* to the list of obsoleted targets.
2911 2021-09-13 Richard Biener <rguenther@suse.de>
2913 * config/avr/elf.h (PREFERRED_DEBUGGING_TYPE): Remove
2914 override, pick up DWARF2_DEBUG define from elfos.h
2916 2021-09-13 Richard Biener <rguenther@suse.de>
2918 * config/rx/rx.h (PREFERRED_DEBUGGING_TYPE): Always define to
2921 2021-09-13 Richard Biener <rguenther@suse.de>
2923 * config/alpha/vms.h (PREFERRED_DEBUGGING_TYPE): Define to
2926 2021-09-13 Richard Biener <rguenther@suse.de>
2928 * config/i386/cygming.h: Always default to DWARF2 debugging.
2929 Do not define DBX_DEBUGGING_INFO, that's done via dbxcoff.h
2931 * doc/install.texi: Document binutils 2.16 as minimum
2932 requirement for mingw.
2934 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
2936 * config/rs6000/rs6000.c (struct rs6000_cost_data): New members
2937 nstmts, nloads and extra_ctor_cost.
2938 (rs6000_density_test): Add load density related heuristics. Do
2939 extra costing on vector construction statements if need.
2940 (rs6000_init_cost): Init new members.
2941 (rs6000_update_target_cost_per_stmt): New function.
2942 (rs6000_add_stmt_cost): Factor vect_nonmem hunk out to function
2943 rs6000_update_target_cost_per_stmt and call it.
2945 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
2947 * config/rs6000/rs6000.c (struct rs6000_cost_data): Remove typedef.
2948 (rs6000_init_cost): Adjust.
2950 2021-09-13 liuhongt <hongtao.liu@intel.com>
2952 * config/i386/i386.md: (UNSPEC_COPYSIGN): Remove.
2953 (UNSPEC_XORSIGN): Ditto.
2955 2021-09-12 Roger Sayle <roger@nextmovesoftware.com>
2957 * expr.c (convert_move): Preserve SUBREG_PROMOTED_VAR_P when
2958 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
2961 2021-09-11 Aldy Hernandez <aldyh@redhat.com>
2963 * tree-ssa-threadbackward.c (class back_threader_registry): Use
2964 back_jt_path_registry.
2965 * tree-ssa-threadedge.c (jump_threader::jump_threader): Use
2966 fwd_jt_path_registry.
2967 * tree-ssa-threadedge.h (class jump_threader): Same..
2968 * tree-ssa-threadupdate.c
2969 (jump_thread_path_registry::jump_thread_path_registry): Rename...
2970 (jt_path_registry::jt_path_registry): ...to this.
2971 (jump_thread_path_registry::~jump_thread_path_registry): Rename...
2972 (jt_path_registry::~jt_path_registry): ...this.
2973 (fwd_jt_path_registry::fwd_jt_path_registry): New.
2974 (fwd_jt_path_registry::~fwd_jt_path_registry): New.
2975 (jump_thread_path_registry::allocate_thread_edge): Rename...
2976 (jt_path_registry::allocate_thread_edge): ...to this.
2977 (jump_thread_path_registry::allocate_thread_path): Rename...
2978 (jt_path_registry::allocate_thread_path): ...to this.
2979 (jump_thread_path_registry::lookup_redirection_data): Rename...
2980 (fwd_jt_path_registry::lookup_redirection_data): ...to this.
2981 (jump_thread_path_registry::thread_block_1): Rename...
2982 (fwd_jt_path_registry::thread_block_1): ...to this.
2983 (jump_thread_path_registry::thread_block): Rename...
2984 (fwd_jt_path_registry::thread_block): ...to this.
2985 (jt_path_registry::thread_through_loop_header): Rename...
2986 (fwd_jt_path_registry::thread_through_loop_header): ...to this.
2987 (jump_thread_path_registry::mark_threaded_blocks): Rename...
2988 (fwd_jt_path_registry::mark_threaded_blocks): ...to this.
2989 (jump_thread_path_registry::debug_path): Rename...
2990 (jt_path_registry::debug_path): ...to this.
2991 (jump_thread_path_registry::dump): Rename...
2992 (jt_path_registry::debug): ...to this.
2993 (jump_thread_path_registry::rewire_first_differing_edge): Rename...
2994 (back_jt_path_registry::rewire_first_differing_edge): ...to this.
2995 (jump_thread_path_registry::adjust_paths_after_duplication): Rename...
2996 (back_jt_path_registry::adjust_paths_after_duplication): ...to this.
2997 (jump_thread_path_registry::duplicate_thread_path): Rename...
2998 (back_jt_path_registry::duplicate_thread_path): ...to this. Also,
2999 drop ill-formed candidates.
3000 (jump_thread_path_registry::remove_jump_threads_including): Rename...
3001 (fwd_jt_path_registry::remove_jump_threads_including): ...to this.
3002 (jt_path_registry::thread_through_all_blocks): New.
3003 (back_jt_path_registry::update_cfg): New.
3004 (fwd_jt_path_registry::update_cfg): New.
3005 (jump_thread_path_registry::register_jump_thread): Rename...
3006 (jt_path_registry::register_jump_thread): ...to this.
3007 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
3009 (class jt_path_registry): ...here.
3010 (class fwd_jt_path_registry): New.
3011 (class back_jt_path_registry): New.
3013 2021-09-10 liuhongt <hongtao.liu@intel.com>
3016 2021-09-01 liuhongt <hongtao.liu@intel.com>
3018 * emit-rtl.c (validate_subreg): Get rid of all float-int
3021 2021-09-10 Jakub Jelinek <jakub@redhat.com>
3023 * tree-core.h (enum omp_memory_order): Add OMP_MEMORY_ORDER_MASK,
3024 OMP_FAIL_MEMORY_ORDER_UNSPECIFIED, OMP_FAIL_MEMORY_ORDER_RELAXED,
3025 OMP_FAIL_MEMORY_ORDER_ACQUIRE, OMP_FAIL_MEMORY_ORDER_RELEASE,
3026 OMP_FAIL_MEMORY_ORDER_ACQ_REL, OMP_FAIL_MEMORY_ORDER_SEQ_CST and
3027 OMP_FAIL_MEMORY_ORDER_MASK enumerators.
3028 (OMP_FAIL_MEMORY_ORDER_SHIFT): Define.
3029 * gimple-pretty-print.c (dump_gimple_omp_atomic_load,
3030 dump_gimple_omp_atomic_store): Print [weak] for weak atomic
3032 * gimple.h (enum gf_mask): Change GF_OMP_ATOMIC_MEMORY_ORDER
3033 to 6-bit mask, adjust GF_OMP_ATOMIC_NEED_VALUE value and add
3035 (gimple_omp_atomic_weak_p, gimple_omp_atomic_set_weak): New inline
3037 * tree.h (OMP_ATOMIC_WEAK): Define.
3038 * tree-pretty-print.c (dump_omp_atomic_memory_order): Adjust for
3039 fail memory order being encoded in the same enum and also print
3040 fail clause if present.
3041 (dump_generic_node): Print weak clause if OMP_ATOMIC_WEAK.
3042 * gimplify.c (goa_stabilize_expr): Add target_expr and rhs arguments,
3043 handle pre_p == NULL case as a test mode that only returns value
3044 but doesn't change gimplify nor change anything otherwise, adjust
3045 recursive calls, add MODIFY_EXPR, ADDR_EXPR, COND_EXPR, TARGET_EXPR
3046 and CALL_EXPR handling, adjust COMPOUND_EXPR handling for
3047 __builtin_clear_padding calls, for !rhs gimplify as lvalue rather
3049 (gimplify_omp_atomic): Adjust goa_stabilize_expr caller. Handle
3050 COND_EXPR rhs. Set weak flag on gimple load/store for
3052 * omp-expand.c (omp_memory_order_to_fail_memmodel): New function.
3053 (omp_memory_order_to_memmodel): Adjust for fail clause encoded
3055 (expand_omp_atomic_cas): New function.
3056 (expand_omp_atomic_pipeline): Use omp_memory_order_to_fail_memmodel
3058 (expand_omp_atomic): Attempt to optimize atomic compare and exchange
3059 using expand_omp_atomic_cas.
3061 2021-09-10 Aldy Hernandez <aldyh@redhat.com>
3062 Michael Matz <matz@suse.de>
3064 * tree-pass.h (PROP_loop_opts_done): New.
3065 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
3066 Intersect with global range.
3067 * tree-ssa-loop.c (tree_ssa_loop_done): Set PROP_loop_opts_done.
3068 * tree-ssa-threadbackward.c
3069 (back_threader_profitability::profitable_path_p): Disable
3070 threading through latches until after loop optimizations have run.
3072 2021-09-10 David Faust <david.faust@oracle.com>
3074 * doc/invoke.texi: Document BPF -mcpu, -mjmpext, -mjmp32 and -malu32
3077 2021-09-10 David Faust <david.faust@oracle.com>
3079 * config/bpf/bpf-opts.h (bpf_isa_version): New enum.
3080 * config/bpf/bpf-protos.h (bpf_expand_cbranch): New.
3081 * config/bpf/bpf.c (bpf_option_override): Handle -mcpu option.
3082 (bpf_expand_cbranch): New function.
3083 * config/bpf/bpf.md (AM mode iterator): Conditionalize support for SI
3085 (zero_extendsidi2): Only use mov32 instruction if it is available.
3086 (SIM mode iterator): Conditionalize support for SI mode.
3087 (JM mode iterator): New.
3088 (cbranchdi4): Update name, use new JM iterator. Use bpf_expand_cbranch.
3089 (*branch_on_di): Update name, use new JM iterator.
3090 * config/bpf/bpf.opt: (mjmpext): New option.
3094 (bpf_isa): New enum.
3096 2021-09-10 David Faust <david.faust@oracle.com>
3098 * config/bpf/bpf.md (zero_extendhidi2): Add new output template
3099 for register-to-register extensions.
3100 (zero_extendqidi2): Likewise.
3102 2021-09-10 Richard Biener <rguenther@suse.de>
3104 PR middle-end/102273
3105 * internal-fn.c (expand_DEFERRED_INIT): Always expand non-SSA vars.
3107 2021-09-10 Richard Biener <rguenther@suse.de>
3109 PR middle-end/102269
3110 * gimplify.c (is_var_need_auto_init): Empty types do not need
3113 2021-09-10 Richard Biener <rguenther@suse.de>
3115 * configure.ac (--with-stabs): Remove.
3116 * configure: Regenerate.
3117 * doc/install.texi: Remove --with-stabs documentation.
3119 2021-09-10 liuhongt <hongtao.liu@intel.com>
3121 * config/i386/avx512fp16intrin.h: (_mm512_cmp_ph_mask):
3123 (_mm512_mask_cmp_ph_mask): Likewise.
3124 (_mm512_cmp_round_ph_mask): Likewise.
3125 (_mm512_mask_cmp_round_ph_mask): Likewise.
3126 (_mm_cmp_sh_mask): Likewise.
3127 (_mm_mask_cmp_sh_mask): Likewise.
3128 (_mm_cmp_round_sh_mask): Likewise.
3129 (_mm_mask_cmp_round_sh_mask): Likewise.
3130 (_mm_comieq_sh): Likewise.
3131 (_mm_comilt_sh): Likewise.
3132 (_mm_comile_sh): Likewise.
3133 (_mm_comigt_sh): Likewise.
3134 (_mm_comige_sh): Likewise.
3135 (_mm_comineq_sh): Likewise.
3136 (_mm_ucomieq_sh): Likewise.
3137 (_mm_ucomilt_sh): Likewise.
3138 (_mm_ucomile_sh): Likewise.
3139 (_mm_ucomigt_sh): Likewise.
3140 (_mm_ucomige_sh): Likewise.
3141 (_mm_ucomineq_sh): Likewise.
3142 (_mm_comi_round_sh): Likewise.
3143 (_mm_comi_sh): Likewise.
3144 * config/i386/avx512fp16vlintrin.h (_mm_cmp_ph_mask): New intrinsic.
3145 (_mm_mask_cmp_ph_mask): Likewise.
3146 (_mm256_cmp_ph_mask): Likewise.
3147 (_mm256_mask_cmp_ph_mask): Likewise.
3148 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3149 * config/i386/i386-builtin.def: Add corresponding new builtins.
3150 * config/i386/i386-expand.c
3151 (ix86_expand_args_builtin): Handle new builtin types.
3152 (ix86_expand_round_builtin): Ditto.
3153 * config/i386/i386.md (ssevecmode): Add HF mode.
3154 (MODEFH): New mode iterator.
3155 * config/i386/sse.md
3156 (V48H_AVX512VL): New mode iterator to support HF vector modes.
3157 Ajdust corresponding description.
3158 (ssecmpintprefix): New.
3159 (VI12_AVX512VL): Adjust to support HF vector modes.
3160 (cmp_imm_predicate): Likewise.
3161 (<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
3163 (avx512f_vmcmp<mode>3<round_saeonly_name>): Likewise.
3164 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Likewise.
3165 (<sse>_<unord>comi<round_saeonly_name>): Likewise.
3167 2021-09-10 liuhongt <hongtao.liu@intel.com>
3169 * config/i386/avx512fp16intrin.h: (_mm512_max_ph): New intrinsic.
3170 (_mm512_mask_max_ph): Likewise.
3171 (_mm512_maskz_max_ph): Likewise.
3172 (_mm512_min_ph): Likewise.
3173 (_mm512_mask_min_ph): Likewise.
3174 (_mm512_maskz_min_ph): Likewise.
3175 (_mm512_max_round_ph): Likewise.
3176 (_mm512_mask_max_round_ph): Likewise.
3177 (_mm512_maskz_max_round_ph): Likewise.
3178 (_mm512_min_round_ph): Likewise.
3179 (_mm512_mask_min_round_ph): Likewise.
3180 (_mm512_maskz_min_round_ph): Likewise.
3181 (_mm_max_sh): Likewise.
3182 (_mm_mask_max_sh): Likewise.
3183 (_mm_maskz_max_sh): Likewise.
3184 (_mm_min_sh): Likewise.
3185 (_mm_mask_min_sh): Likewise.
3186 (_mm_maskz_min_sh): Likewise.
3187 (_mm_max_round_sh): Likewise.
3188 (_mm_mask_max_round_sh): Likewise.
3189 (_mm_maskz_max_round_sh): Likewise.
3190 (_mm_min_round_sh): Likewise.
3191 (_mm_mask_min_round_sh): Likewise.
3192 (_mm_maskz_min_round_sh): Likewise.
3193 * config/i386/avx512fp16vlintrin.h (_mm_max_ph): New intrinsic.
3194 (_mm256_max_ph): Likewise.
3195 (_mm_mask_max_ph): Likewise.
3196 (_mm256_mask_max_ph): Likewise.
3197 (_mm_maskz_max_ph): Likewise.
3198 (_mm256_maskz_max_ph): Likewise.
3199 (_mm_min_ph): Likewise.
3200 (_mm256_min_ph): Likewise.
3201 (_mm_mask_min_ph): Likewise.
3202 (_mm256_mask_min_ph): Likewise.
3203 (_mm_maskz_min_ph): Likewise.
3204 (_mm256_maskz_min_ph): Likewise.
3205 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3206 * config/i386/i386-builtin.def: Add corresponding new builtins.
3207 * config/i386/i386-expand.c
3208 (ix86_expand_args_builtin): Handle new builtin types.
3209 * config/i386/sse.md
3210 (<code><mode>3<mask_name><round_saeonly_name>): Adjust to
3211 support HF vector modes.
3212 (*<code><mode>3<mask_name><round_saeonly_name>): Likewise.
3213 (ieee_<ieee_maxmin><mode>3<mask_name><round_saeonly_name>):
3215 (<sse>_vm<code><mode>3<mask_scalar_name><round_saeonly_scalar_name>):
3217 * config/i386/subst.md (round_saeonly_mode512bit_condition):
3218 Adjust for HF vector modes.
3220 2021-09-10 Liu, Hongtao <hongtao.liu@intel.com>
3222 * config/i386/avx512fp16intrin.h (_mm_add_sh): New intrinsic.
3223 (_mm_mask_add_sh): Likewise.
3224 (_mm_maskz_add_sh): Likewise.
3225 (_mm_sub_sh): Likewise.
3226 (_mm_mask_sub_sh): Likewise.
3227 (_mm_maskz_sub_sh): Likewise.
3228 (_mm_mul_sh): Likewise.
3229 (_mm_mask_mul_sh): Likewise.
3230 (_mm_maskz_mul_sh): Likewise.
3231 (_mm_div_sh): Likewise.
3232 (_mm_mask_div_sh): Likewise.
3233 (_mm_maskz_div_sh): Likewise.
3234 (_mm_add_round_sh): Likewise.
3235 (_mm_mask_add_round_sh): Likewise.
3236 (_mm_maskz_add_round_sh): Likewise.
3237 (_mm_sub_round_sh): Likewise.
3238 (_mm_mask_sub_round_sh): Likewise.
3239 (_mm_maskz_sub_round_sh): Likewise.
3240 (_mm_mul_round_sh): Likewise.
3241 (_mm_mask_mul_round_sh): Likewise.
3242 (_mm_maskz_mul_round_sh): Likewise.
3243 (_mm_div_round_sh): Likewise.
3244 (_mm_mask_div_round_sh): Likewise.
3245 (_mm_maskz_div_round_sh): Likewise.
3246 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3247 * config/i386/i386-builtin.def: Add corresponding new builtins.
3248 * config/i386/i386-expand.c
3249 (ix86_expand_round_builtin): Handle new builtins.
3250 * config/i386/sse.md (VF_128): Change description.
3251 (<sse>_vm<plusminus_insn><mode>3<mask_scalar_name><round_scalar_name>):
3252 Adjust to support HF vector modes.
3253 (<sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name><round_scalar_name>):
3256 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
3258 * config/i386/i386-expand.c
3259 (ix86_avx256_split_vector_move_misalign): Handle V16HF mode.
3260 * config/i386/i386.c
3261 (ix86_preferred_simd_mode): Handle HF mode.
3262 * config/i386/sse.md (V_256H): New mode iterator.
3263 (avx_vextractf128<mode>): Use it.
3264 (VEC_INIT_MODE): Align vector HFmode condition to vector
3265 HImodes since there're no real HF instruction used.
3266 (VEC_INIT_HALF_MODE): Ditto.
3268 (VIHF_AVX512BW): Ditto.
3269 (*vec_extracthf): Ditto.
3270 (VEC_EXTRACT_MODE): Ditto.
3272 2021-09-10 Richard Biener <rguenther@suse.de>
3275 * config/dbx.h: Remove.
3276 * config/dbxcoff.h: Do not define PREFERRED_DEBUGGING_TYPE.
3277 * config/lynx.h: Likewise.
3279 2021-09-10 liuhongt <hongtao.liu@intel.com>
3281 * config/i386/i386-expand.c (ix86_expand_copysign): Expand
3282 right into ANDNOT + AND + IOR, using paradoxical subregs.
3283 (ix86_split_copysign_const): Remove.
3284 (ix86_split_copysign_var): Ditto.
3285 * config/i386/i386-protos.h (ix86_split_copysign_const): Dotto.
3286 (ix86_split_copysign_var): Ditto.
3287 * config/i386/i386.md (@copysign<mode>3_const): Ditto.
3288 (@copysign<mode>3_var): Ditto.
3290 2021-09-09 qing zhao <qing.zhao@oracle.com>
3292 * builtins.c (expand_builtin_memset): Make external visible.
3293 * builtins.h (expand_builtin_memset): Declare extern.
3294 * common.opt (ftrivial-auto-var-init=): New option.
3295 * doc/extend.texi: Document the uninitialized attribute.
3296 * doc/invoke.texi: Document -ftrivial-auto-var-init.
3297 * flag-types.h (enum auto_init_type): New enumerated type
3299 * gimple-fold.c (clear_padding_type): Add one new parameter.
3300 (clear_padding_union): Likewise.
3301 (clear_padding_emit_loop): Likewise.
3302 (clear_type_padding_in_mask): Likewise.
3303 (gimple_fold_builtin_clear_padding): Handle this new parameter.
3304 * gimplify.c (gimple_add_init_for_auto_var): New function.
3305 (gimple_add_padding_init_for_auto_var): New function.
3306 (is_var_need_auto_init): New function.
3307 (gimplify_decl_expr): Add initialization to automatic variables per
3309 (gimplify_call_expr): Add one new parameter for call to
3310 __builtin_clear_padding.
3311 (gimplify_init_constructor): Add padding initialization in the end.
3312 * internal-fn.c (INIT_PATTERN_VALUE): New macro.
3313 (expand_DEFERRED_INIT): New function.
3314 * internal-fn.def (DEFERRED_INIT): New internal function.
3315 * tree-cfg.c (verify_gimple_call): Verify calls to .DEFERRED_INIT.
3316 * tree-sra.c (generate_subtree_deferred_init): New function.
3317 (scan_function): Avoid setting cannot_scalarize_away_bitmap for
3318 calls to .DEFERRED_INIT.
3319 (sra_modify_deferred_init): New function.
3320 (sra_modify_function_body): Handle calls to DEFERRED_INIT specially.
3321 * tree-ssa-structalias.c (find_func_aliases_for_call): Likewise.
3322 * tree-ssa-uninit.c (warn_uninit): Handle calls to DEFERRED_INIT
3324 (check_defs): Likewise.
3325 (warn_uninitialized_vars): Likewise.
3326 * tree-ssa.c (ssa_undefined_value_p): Likewise.
3327 * tree.c (build_common_builtin_nodes): Build tree node for
3328 BUILT_IN_CLEAR_PADDING when needed.
3330 2021-09-09 Richard Biener <rguenther@suse.de>
3332 * tree-ssa-loop-im.c (fill_always_executed_in_1): Walk
3335 2021-09-09 Richard Biener <rguenther@suse.de>
3337 * tree-ssa-loop-im.c (fill_always_executed_in_1): Integrate
3338 DOM walk from get_loop_body_in_dom_order using a worklist
3341 2021-09-09 liuhongt <hongtao.liu@intel.com>
3343 * config.gcc: Add avx512fp16vlintrin.h.
3344 * config/i386/avx512fp16intrin.h: (_mm512_add_ph): New intrinsic.
3345 (_mm512_mask_add_ph): Likewise.
3346 (_mm512_maskz_add_ph): Likewise.
3347 (_mm512_sub_ph): Likewise.
3348 (_mm512_mask_sub_ph): Likewise.
3349 (_mm512_maskz_sub_ph): Likewise.
3350 (_mm512_mul_ph): Likewise.
3351 (_mm512_mask_mul_ph): Likewise.
3352 (_mm512_maskz_mul_ph): Likewise.
3353 (_mm512_div_ph): Likewise.
3354 (_mm512_mask_div_ph): Likewise.
3355 (_mm512_maskz_div_ph): Likewise.
3356 (_mm512_add_round_ph): Likewise.
3357 (_mm512_mask_add_round_ph): Likewise.
3358 (_mm512_maskz_add_round_ph): Likewise.
3359 (_mm512_sub_round_ph): Likewise.
3360 (_mm512_mask_sub_round_ph): Likewise.
3361 (_mm512_maskz_sub_round_ph): Likewise.
3362 (_mm512_mul_round_ph): Likewise.
3363 (_mm512_mask_mul_round_ph): Likewise.
3364 (_mm512_maskz_mul_round_ph): Likewise.
3365 (_mm512_div_round_ph): Likewise.
3366 (_mm512_mask_div_round_ph): Likewise.
3367 (_mm512_maskz_div_round_ph): Likewise.
3368 * config/i386/avx512fp16vlintrin.h: New header.
3369 * config/i386/i386-builtin-types.def (V16HF, V8HF, V32HF):
3370 Add new builtin types.
3371 * config/i386/i386-builtin.def: Add corresponding builtins.
3372 * config/i386/i386-expand.c
3373 (ix86_expand_args_builtin): Handle new builtin types.
3374 (ix86_expand_round_builtin): Likewise.
3375 * config/i386/immintrin.h: Include avx512fp16vlintrin.h
3376 * config/i386/sse.md (VFH): New mode_iterator.
3378 (avx512fmaskmode): Add HF vector modes.
3379 (avx512fmaskhalfmode): Likewise.
3380 (<plusminus_insn><mode>3<mask_name><round_name>): Adjust to for
3382 (*<plusminus_insn><mode>3<mask_name><round_name>): Likewise.
3383 (mul<mode>3<mask_name><round_name>): Likewise.
3384 (*mul<mode>3<mask_name><round_name>): Likewise.
3385 (div<mode>3): Likewise.
3386 (<sse>_div<mode>3<mask_name><round_name>): Likewise.
3387 * config/i386/subst.md (SUBST_V): Add HF vector modes.
3388 (SUBST_A): Likewise.
3389 (round_mode512bit_condition): Adjust for V32HFmode.
3391 2021-09-09 liuhongt <hongtao.liu@intel.com>
3394 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
3395 (reduc_plus_scal_v4sf): .. this, New define_expand.
3396 (reduc_plus_scal_v2df): .. and this, New define_expand.
3398 2021-09-09 liuhongt <hongtao.liu@intel.com>
3401 * config/i386/sse.md (*vec_extract<mode><ssescalarmodelower>_valign):
3404 2021-09-08 Jonathan Wakely <jwakely@redhat.com>
3407 * doc/trouble.texi (Copy Assignment): Fix description of
3408 behaviour and fix code in example.
3410 2021-09-08 Segher Boessenkool <segher@kernel.crashing.org>
3413 * config/rs6000/rs6000-logue.c (rs6000_emit_epilogue): For ELFv2 use
3414 r11 instead of r12 for restoring CR.
3416 2021-09-08 Jakub Jelinek <jakub@redhat.com>
3417 liuhongt <hongtao.liu@intel.com>
3420 * config/i386/i386.md (@xorsign<mode>3_1): Remove.
3421 * config/i386/i386-expand.c (ix86_expand_xorsign): Expand right away
3422 into AND with mask and XOR, using paradoxical subregs.
3423 (ix86_split_xorsign): Remove.
3424 * config/i386/i386-protos.h (ix86_split_xorsign): Remove.
3426 2021-09-08 Di Zhao <dizhao@os.amperecomputing.com>
3428 * tree-ssa-sccvn.c (vn_nary_op_insert_into): fix result compare
3430 2021-09-08 Jakub Jelinek <jakub@redhat.com>
3433 * config/i386/i386.md (xorsign<mode>3): If operands[1] is equal to
3434 operands[2], emit abs<mode>2 instead.
3435 (@xorsign<mode>3_1): Add early-clobbers for output operand, enable
3436 first alternative even for avx, add another alternative with
3437 =&Yv <- 0, Yv, Yvm constraints.
3438 * config/i386/i386-expand.c (ix86_split_xorsign): If op0 is equal
3439 to op1, emit vpandn instead.
3441 2021-09-08 liuhongt <hongtao.liu@intel.com>
3443 * config/i386/avx512fp16intrin.h (_mm_set_ph): New intrinsic.
3444 (_mm256_set_ph): Likewise.
3445 (_mm512_set_ph): Likewise.
3446 (_mm_setr_ph): Likewise.
3447 (_mm256_setr_ph): Likewise.
3448 (_mm512_setr_ph): Likewise.
3449 (_mm_set1_ph): Likewise.
3450 (_mm256_set1_ph): Likewise.
3451 (_mm512_set1_ph): Likewise.
3452 (_mm_setzero_ph): Likewise.
3453 (_mm256_setzero_ph): Likewise.
3454 (_mm512_setzero_ph): Likewise.
3455 (_mm_set_sh): Likewise.
3456 (_mm_load_sh): Likewise.
3457 (_mm_store_sh): Likewise.
3458 * config/i386/i386-builtin-types.def (V8HF): New type.
3459 (DEF_FUNCTION_TYPE (V8HF, V8HI)): New builtin function type
3460 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
3461 Support vector HFmodes.
3462 (ix86_expand_vector_init_one_nonzero): Likewise.
3463 (ix86_expand_vector_init_one_var): Likewise.
3464 (ix86_expand_vector_init_interleave): Likewise.
3465 (ix86_expand_vector_init_general): Likewise.
3466 (ix86_expand_vector_set): Likewise.
3467 (ix86_expand_vector_extract): Likewise.
3468 (ix86_expand_vector_init_concat): Likewise.
3469 (ix86_expand_sse_movcc): Handle vector HFmodes.
3470 (ix86_expand_vector_set_var): Ditto.
3471 * config/i386/i386-modes.def: Add HF vector modes in comment.
3472 * config/i386/i386.c (classify_argument): Add HF vector modes.
3473 (ix86_hard_regno_mode_ok): Allow HF vector modes for AVX512FP16.
3474 (ix86_vector_mode_supported_p): Likewise.
3475 (ix86_set_reg_reg_cost): Handle vector HFmode.
3476 (ix86_get_ssemov): Handle vector HFmode.
3477 (function_arg_advance_64): Pass unamed V16HFmode and V32HFmode
3479 (function_arg_advance_32): Pass V8HF/V16HF/V32HF by sse reg for 32bit
3481 (function_arg_advance_32): Ditto.
3482 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): New.
3483 (VALID_AVX256_REG_OR_OI_MODE): Rename to ..
3484 (VALID_AVX256_REG_OR_OI_VHF_MODE): .. this, and add V16HF.
3485 (VALID_SSE2_REG_VHF_MODE): New.
3486 (VALID_AVX512VL_128_REG_MODE): Add V8HF and TImode.
3487 (SSE_REG_MODE_P): Add vector HFmode.
3488 * config/i386/i386.md (mode): Add HF vector modes.
3489 (MODE_SIZE): Likewise.
3490 (ssemodesuffix): Add ph suffix for HF vector modes.
3491 * config/i386/sse.md (VFH_128): New mode iterator.
3492 (VMOVE): Adjust for HF vector modes.
3494 (V_256_512): Likewise.
3496 (avx512fmaskmode): Likewise.
3497 (shuffletype): Likewise.
3498 (sseinsnmode): Likewise.
3499 (ssedoublevecmode): Likewise.
3500 (ssehalfvecmode): Likewise.
3501 (ssehalfvecmodelower): Likewise.
3502 (ssePScmode): Likewise.
3503 (ssescalarmode): Likewise.
3504 (ssescalarmodelower): Likewise.
3505 (sseintprefix): Likewise.
3507 (bcstscalarsuff): Likewise.
3508 (xtg_mode): Likewise.
3509 (VI12HF_AVX512VL): New mode_iterator.
3510 (VF_AVX512FP16): Likewise.
3512 (VIHF_256): Likewise.
3513 (VIHF_AVX512BW): Likewise.
3514 (V16_256): Likewise.
3515 (V32_512): Likewise.
3516 (sseintmodesuffix): New mode_attr.
3517 (sse): Add scalar and vector HFmodes.
3518 (ssescalarmode): Add vector HFmode mapping.
3519 (ssescalarmodesuffix): Add sh suffix for HFmode.
3520 (*<sse>_vm<insn><mode>3): Use VFH_128.
3521 (*<sse>_vm<multdiv_mnemonic><mode>3): Likewise.
3522 (*ieee_<ieee_maxmin><mode>3): Likewise.
3523 (<avx512>_blendm<mode>): New define_insn.
3524 (vec_setv8hf): New define_expand.
3525 (vec_set<mode>_0): New define_insn for HF vector set.
3526 (*avx512fp16_movsh): Likewise.
3527 (avx512fp16_movsh): Likewise.
3528 (vec_extract_lo_v32hi): Rename to ...
3529 (vec_extract_lo_<mode>): ... this, and adjust to allow HF
3531 (vec_extract_hi_v32hi): Likewise.
3532 (vec_extract_hi_<mode>): Likewise.
3533 (vec_extract_lo_v16hi): Likewise.
3534 (vec_extract_lo_<mode>): Likewise.
3535 (vec_extract_hi_v16hi): Likewise.
3536 (vec_extract_hi_<mode>): Likewise.
3537 (vec_set_hi_v16hi): Likewise.
3538 (vec_set_hi_<mode>): Likewise.
3539 (vec_set_lo_v16hi): Likewise.
3540 (vec_set_lo_<mode>): Likewise.
3541 (*vec_extract<mode>_0): New define_insn_and_split for HF
3543 (*vec_extracthf): New define_insn.
3544 (VEC_EXTRACT_MODE): Add HF vector modes.
3545 (PINSR_MODE): Add V8HF.
3546 (sse2p4_1): Likewise.
3547 (pinsr_evex_isa): Likewise.
3548 (<sse2p4_1>_pinsr<ssemodesuffix>): Adjust to support
3549 insert for V8HFmode.
3550 (pbroadcast_evex_isa): Add HF vector modes.
3551 (AVX2_VEC_DUP_MODE): Likewise.
3552 (VEC_INIT_MODE): Likewise.
3553 (VEC_INIT_HALF_MODE): Likewise.
3554 (avx2_pbroadcast<mode>): Adjust to support HF vector mode
3556 (avx2_pbroadcast<mode>_1): Likewise.
3557 (<avx512>_vec_dup<mode>_1): Likewise.
3558 (<avx512>_vec_dup<mode><mask_name>): Likewise.
3559 (<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>):
3562 2021-09-08 Guo, Xuepeng <xuepeng.guo@intel.com>
3563 H.J. Lu <hongjiu.lu@intel.com>
3564 Liu Hongtao <hongtao.liu@intel.com>
3565 Wang Hongyu <hongyu.wang@intel.com>
3566 Xu Dianhong <dianhong.xu@intel.com>
3568 * common/config/i386/cpuinfo.h (get_available_features):
3569 Detect FEATURE_AVX512FP16.
3570 * common/config/i386/i386-common.c
3571 (OPTION_MASK_ISA_AVX512FP16_SET,
3572 OPTION_MASK_ISA_AVX512FP16_UNSET,
3573 OPTION_MASK_ISA2_AVX512FP16_SET,
3574 OPTION_MASK_ISA2_AVX512FP16_UNSET): New.
3575 (OPTION_MASK_ISA2_AVX512BW_UNSET,
3576 OPTION_MASK_ISA2_AVX512BF16_UNSET): Add AVX512FP16.
3577 (ix86_handle_option): Handle -mavx512fp16.
3578 * common/config/i386/i386-cpuinfo.h (enum processor_features):
3579 Add FEATURE_AVX512FP16.
3580 * common/config/i386/i386-isas.h: Add entry for AVX512FP16.
3581 * config.gcc: Add avx512fp16intrin.h.
3582 * config/i386/avx512fp16intrin.h: New intrinsic header.
3583 * config/i386/cpuid.h: Add bit_AVX512FP16.
3584 * config/i386/i386-builtin-types.def: (FLOAT16): New primitive type.
3585 * config/i386/i386-builtins.c: Support _Float16 type for i386
3587 (ix86_register_float16_builtin_type): New function.
3588 (ix86_float16_type_node): New.
3589 * config/i386/i386-c.c (ix86_target_macros_internal): Define
3591 * config/i386/i386-expand.c (ix86_expand_branch): Support
3593 (ix86_prepare_fp_compare_args): Adjust TARGET_SSE_MATH &&
3594 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
3595 (ix86_expand_fp_movcc): Ditto.
3596 * config/i386/i386-isa.def: Add PTA define for AVX512FP16.
3597 * config/i386/i386-options.c (isa2_opts): Add -mavx512fp16.
3598 (ix86_valid_target_attribute_inner_p): Add avx512fp16 attribute.
3599 * config/i386/i386.c (ix86_get_ssemov): Use
3600 vmovdqu16/vmovw/vmovsh for HFmode/HImode scalar or vector.
3601 (ix86_get_excess_precision): Use
3602 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when TARGET_AVX512FP16
3604 (sse_store_index): Use SFmode cost for HFmode cost.
3605 (inline_memory_move_cost): Add HFmode, and perfer SSE cost over
3606 GPR cost for HFmode.
3607 (ix86_hard_regno_mode_ok): Allow HImode in sse register.
3608 (ix86_mangle_type): Add manlging for _Float16 type.
3609 (inline_secondary_memory_needed): No memory is needed for
3610 16bit movement between gpr and sse reg under
3612 (ix86_multiplication_cost): Adjust TARGET_SSE_MATH &&
3613 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
3614 (ix86_division_cost): Ditto.
3615 (ix86_rtx_costs): Ditto.
3616 (ix86_add_stmt_cost): Ditto.
3617 (ix86_optab_supported_p): Ditto.
3618 * config/i386/i386.h (VALID_AVX512F_SCALAR_MODE): Add HFmode.
3619 (SSE_FLOAT_MODE_SSEMATH_OR_HF_P): Add HFmode.
3620 (PTA_SAPPHIRERAPIDS): Add PTA_AVX512FP16.
3621 * config/i386/i386.md (mode): Add HFmode.
3622 (MODE_SIZE): Add HFmode.
3623 (isa): Add avx512fp16.
3624 (enabled): Handle avx512fp16.
3625 (ssemodesuffix): Add sh suffix for HFmode.
3626 (comm): Add mult, div.
3627 (plusminusmultdiv): New code iterator.
3628 (insn): Add mult, div.
3629 (*movhf_internal): Adjust for avx512fp16 instruction.
3630 (*movhi_internal): Ditto.
3631 (*cmpi<unord>hf): New define_insn for HFmode.
3632 (*ieee_s<ieee_maxmin>hf3): Likewise.
3633 (extendhf<mode>2): Likewise.
3634 (trunc<mode>hf2): Likewise.
3635 (float<floatunssuffix><mode>hf2): Likewise.
3636 (*<insn>hf): Likewise.
3637 (cbranchhf4): New expander.
3638 (movhfcc): Likewise.
3639 (<insn>hf3): Likewise.
3642 * config/i386/i386.opt: Add mavx512fp16.
3643 * config/i386/immintrin.h: Include avx512fp16intrin.h.
3644 * doc/invoke.texi: Add mavx512fp16.
3645 * doc/extend.texi: Add avx512fp16 Usage Notes.
3647 2021-09-08 liuhongt <hongtao.liu@intel.com>
3649 * common.opt: Support -fexcess-precision=16.
3650 * config/aarch64/aarch64.c (aarch64_excess_precision): Return
3651 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when
3652 EXCESS_PRECISION_TYPE_FLOAT16.
3653 * config/arm/arm.c (arm_excess_precision): Ditto.
3654 * config/i386/i386.c (ix86_get_excess_precision): Ditto.
3655 * config/m68k/m68k.c (m68k_excess_precision): Issue an error
3656 when EXCESS_PRECISION_TYPE_FLOAT16.
3657 * config/s390/s390.c (s390_excess_precision): Ditto.
3658 * coretypes.h (enum excess_precision_type): Add
3659 EXCESS_PRECISION_TYPE_FLOAT16.
3660 * doc/tm.texi (TARGET_C_EXCESS_PRECISION): Update documents.
3661 * doc/tm.texi.in (TARGET_C_EXCESS_PRECISION): Ditto.
3662 * doc/extend.texi (Half-Precision): Document
3663 -fexcess-precision=16.
3664 * flag-types.h (enum excess_precision): Add
3665 EXCESS_PRECISION_FLOAT16.
3666 * target.def (excess_precision): Update document.
3667 * tree.c (excess_precision_type): Set excess_precision_type to
3668 EXCESS_PRECISION_FLOAT16 when -fexcess-precision=16.
3670 2021-09-08 liuhongt <hongtao.liu@intel.com>
3672 * doc/extend.texi: (@node Floating Types): Adjust the wording.
3673 (@node Half-Precision): Ditto.
3675 2021-09-07 Takayuki 'January June' Suwa <jjsuwa_sys3175@yahoo.co.jp>
3678 * config/xtensa/xtensa.c (xtensa_emit_move_sequence): Add
3679 'CONST_INT_P (src)' to the condition of the block that tries to
3680 eliminate literal when loading integer contant.
3682 2021-09-07 David Faust <david.faust@oracle.com>
3684 * doc/extend.texi (BPF Type Attributes) New node.
3685 Document new preserve_access_index attribute.
3686 Document new preserve_access_index builtin.
3687 * doc/invoke.texi: Document -mco-re and -mno-co-re options.
3689 2021-09-07 David Faust <david.faust@oracle.com>
3691 * config/bpf/bpf.c: Adjust includes.
3692 (bpf_handle_preserve_access_index_attribute): New function.
3693 (bpf_attribute_table): Use it here.
3694 (bpf_builtins): Add BPF_BUILTIN_PRESERVE_ACCESS_INDEX.
3695 (bpf_option_override): Handle "-mco-re" option.
3696 (bpf_asm_init_sections): New.
3697 (TARGET_ASM_INIT_SECTIONS): Redefine.
3698 (bpf_file_end): New.
3699 (TARGET_ASM_FILE_END): Redefine.
3700 (bpf_init_builtins): Add "__builtin_preserve_access_index".
3701 (bpf_core_compute, bpf_core_get_index): New.
3702 (is_attr_preserve_access): New.
3703 (bpf_expand_builtin): Handle new builtins.
3704 (bpf_core_newdecl, bpf_core_is_maybe_aggregate_access): New.
3705 (bpf_core_walk): New.
3706 (bpf_resolve_overloaded_builtin): New.
3707 (TARGET_RESOLVE_OVERLOADED_BUILTIN): Redefine.
3709 (pass_bpf_core_attr): New RTL pass.
3710 * config/bpf/bpf-passes.def: New file.
3711 * config/bpf/bpf-protos.h (make_pass_bpf_core_attr): New.
3712 * config/bpf/coreout.c: New file.
3713 * config/bpf/coreout.h: Likewise.
3714 * config/bpf/t-bpf (TM_H): Add $(srcdir)/config/bpf/coreout.h.
3715 (coreout.o): New rule.
3716 (PASSES_EXTRA): Add $(srcdir)/config/bpf/bpf-passes.def.
3717 * config.gcc (bpf): Add coreout.h to extra_headers.
3718 Add coreout.o to extra_objs.
3719 Add $(srcdir)/config/bpf/coreout.c to target_gtfiles.
3721 2021-09-07 David Faust <david.faust@oracle.com>
3723 * btfout.c (get_btf_id): Function is no longer static.
3724 * ctfc.h: Expose it here.
3726 2021-09-07 David Faust <david.faust@oracle.com>
3728 * ctfc.c (ctf_lookup_tree_type): New function.
3731 2021-09-07 David Faust <david.faust@oracle.com>
3733 * ctfc.c (ctf_dtd_lookup): Function is no longer static.
3734 * ctfc.h: Analogous change.
3736 2021-09-07 David Faust <david.faust@oracle.com>
3738 * dwarf2out.c (lookup_type_die): Function is no longer static.
3739 * dwarf2out.h: Expose it here.
3741 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
3743 * dwarf2ctf.c (ctf_debug_finalize): Make it static.
3744 (ctf_debug_early_finish): New definition.
3745 (ctf_debug_finish): Likewise.
3746 * dwarf2ctf.h (ctf_debug_finalize): Remove declaration.
3747 (ctf_debug_early_finish): New declaration.
3748 (ctf_debug_finish): Likewise.
3749 * dwarf2out.c (dwarf2out_finish): Invoke ctf_debug_finish.
3750 (dwarf2out_early_finish): Invoke ctf_debug_early_finish.
3752 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
3754 * config/bpf/bpf.c (bpf_option_override): For BPF backend, disable LTO
3755 support when compiling for CO-RE.
3756 * config/bpf/bpf.opt: Add new command line option -mco-re.
3758 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
3760 * flag-types.h (enum debug_info_type): Add new enum
3761 DINFO_TYPE_BTF_WITH_CORE.
3762 (BTF_WITH_CORE_DEBUG): New bitmask.
3763 * flags.h (btf_with_core_debuginfo_p): New declaration.
3764 * opts.c (btf_with_core_debuginfo_p): New definition.
3766 2021-09-07 Jason Merrill <jason@redhat.com>
3768 * tree.h (error_operand_p): Change to inline function.
3770 2021-09-07 Aldy Hernandez <aldyh@redhat.com>
3772 * tree-ssa-threadedge.c (forwarder_block_p): Rename to...
3773 (empty_block_with_phis_p): ...this.
3774 (potentially_threadable_block): Same.
3775 (jump_threader::thread_through_normal_block): Same.
3777 2021-09-07 Eric Botcazou <ebotcazou@adacore.com>
3780 * dwarf2out.c (mark_base_types): New overloaded function.
3781 (dwarf2out_early_finish): Invoke it on the COMDAT type list as well
3782 as the compilation unit, and call move_marked_base_types afterward.
3784 2021-09-07 H.J. Lu <hjl.tools@gmail.com>
3787 * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
3789 (ix86_expand_vector_convert_uns_vsivsf): Likewise.
3791 2021-09-07 Richard Biener <rguenther@suse.de>
3793 PR tree-optimization/102226
3794 * tree-vect-loop.c (vect_transform_cycle_phi): Record
3795 the converted value for the epilogue PHI use.
3797 2021-09-07 Martin Liska <mliska@suse.cz>
3799 PR gcov-profile/80223
3800 * ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
3801 options, do not inline when no_profile_instrument_function
3802 attributes are different in early inliner. It's fine to inline
3803 it after PGO instrumentation.
3805 2021-09-07 Richard Biener <rguenther@suse.de>
3807 PR tree-optimization/101555
3808 * tree-ssa-pre.c (translate_vuse_through_block): Do not
3809 perform an alias walk to determine the validity of the
3810 mem at the start of the block which is already guaranteed
3811 by means of prune_clobbered_mems.
3812 (phi_translate_1): Pass edge to translate_vuse_through_block.
3814 2021-09-07 Xionghu Luo <luoxhu@linux.ibm.com>
3817 * config/rs6000/rs6000.md (fmod<mode>3): New define_expand.
3818 (remainder<mode>3): Likewise.
3820 2021-09-07 YunQiang Su <yunqiang.su@cipunited.com>
3822 * config/mips/mips.c (mips_file_start): add .module for
3825 2021-09-06 Roger Sayle <roger@nextmovesoftware.com>
3827 * wide-int.cc (wi::clz): Reorder tests to ensure the result
3828 is zero for all negative values.
3830 2021-09-06 Tobias Burnus <tobias@codesourcery.com>
3832 * doc/invoke.texi (-foffload-options): Fix @opindex.
3834 2021-09-06 H.J. Lu <hjl.tools@gmail.com>
3837 * config/i386/i386-expand.c (ix86_split_xorsign): Use operands[2].
3838 * config/i386/i386.md (@xorsign<mode>3_1): Add non-destructive
3839 source alternative for AVX.
3841 2021-09-06 liuhongt <hongtao.liu@intel.com>
3843 PR middle-end/102182
3844 * optabs.c (expand_fix): Add from1 to avoid from being
3847 2021-09-06 Eric Botcazou <ebotcazou@adacore.com>
3849 * dwarf2out.c (modified_type_die): Deal with all array types earlier
3850 and use local variable consistently throughout the function.
3852 2021-09-06 Jakub Jelinek <jakub@redhat.com>
3854 PR tree-optimization/102207
3855 * match.pd: Don't demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
3856 were promoted from signed to wider unsigned type.
3858 2021-09-06 Andrew Pinski <apinski@marvell.com>
3860 PR tree-optimization/63184
3861 * match.pd: Add simplification of pointer_diff of two pointer_plus
3862 with addr_expr in the first operand of each pointer_plus.
3863 Add simplificatoin of ne/eq of two pointer_plus with addr_expr
3864 in the first operand of each pointer_plus.
3866 2021-09-06 Richard Biener <rguenther@suse.de>
3868 PR tree-optimization/102176
3869 * tree-vect-slp.c (vect_slp_gather_vectorized_scalar_stmts):
3871 (vect_bb_slp_scalar_cost): Use the computed set of
3872 vectorized scalar stmts instead of relying on the out-of-date
3873 and not accurate PURE_SLP_STMT.
3874 (vect_bb_vectorization_profitable_p): Compute the set
3875 of vectorized scalar stmts.
3877 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
3879 * gimple-range-path.cc (path_range_query::range_of_stmt): Remove
3880 GIMPLE_COND special casing.
3881 (path_range_query::range_defined_in_block): Use range_of_stmt
3882 instead of calling fold_range directly.
3884 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
3886 * gimple-range-path.cc (path_range_query::range_of_expr): Set
3887 m_undefined_path when appropriate.
3888 (path_range_query::internal_range_of_expr): Copy from range_of_expr.
3889 (path_range_query::unreachable_path_p): New.
3890 (path_range_query::precompute_ranges): Set m_undefined_path.
3891 * gimple-range-path.h (path_range_query::unreachable_path_p): New.
3892 (path_range_query::internal_range_of_expr): New.
3893 * tree-ssa-threadbackward.c (back_threader::find_taken_edge_cond):
3894 Use unreachable_path_p.
3896 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
3898 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
3899 Remove argument and call find_taken_edge.
3900 (back_threader::resolve_phi): Do not calculate taken edge before
3901 calling maybe_register_path.
3902 (back_threader::find_paths_to_names): Same.
3904 2021-09-05 Jeff Law <jlaw@localhost.localdomain>
3906 * config/h8300/h8300.md (QHSI2 mode iterator): New mode iterator.
3907 * config/h8300/testcompare.md (store_c): Update name, use new
3909 (store_neg_c, store_shifted_c): New patterns.
3911 2021-09-03 Segher Boessenkool <segher@kernel.crashing.org>
3914 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): On ELFv2 use r11
3915 instead of r12 for CR save, in all cases.
3917 2021-09-03 Andrew Pinski <apinski@marvell.com>
3919 * config/aarch64/aarch64-sve-builtins.cc (register_vector_type):
3920 Handle error_mark_node as the type of the type_decl.
3922 2021-09-03 Andrew Pinski <apinski@marvell.com>
3924 * config/aarch64/aarch64-builtins.c (struct aarch64_simd_type_info):
3926 (aarch64_simd_types): Likewise.
3927 (aarch64_simd_intOI_type_node): Likewise.
3928 (aarch64_simd_intCI_type_node): Likewise.
3929 (aarch64_simd_intXI_type_node): Likewise.
3930 * config/aarch64/aarch64.h (aarch64_fp16_type_node): Likewise.
3931 (aarch64_fp16_ptr_type_node): Likewise.
3932 (aarch64_bf16_type_node): Likewise.
3933 (aarch64_bf16_ptr_type_node): Likewise.
3935 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3937 * range-op.cc (operator_minus::op1_op2_relation_effect): Abstract
3939 (minus_op1_op2_relation_effect): ...here.
3940 (class operator_pointer_diff): New.
3941 (operator_pointer_diff::op1_op2_relation_effect): Call
3942 minus_op1_op2_relation_effect.
3943 (integral_table::integral_table): Add entry for POINTER_DIFF_EXPR.
3945 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3947 * tree-ssa-threadbackward.c (back_threader::thread_through_all_blocks):
3948 Add may_peel_loop_headers.
3949 (back_threader_registry::thread_through_all_blocks): Same.
3950 (try_thread_blocks): Pass may_peel_loop_headers argument.
3951 (pass_early_thread_jumps::execute): Same.
3953 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3955 * tree-ssa-threadedge.c (has_phis_p): New.
3956 (forwarder_block_p): New.
3957 (potentially_threadable_block): Call forwarder_block_p.
3958 (jump_threader::thread_around_empty_blocks): Call has_phis_p.
3959 (jump_threader::thread_through_normal_block): Call
3962 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3964 * tree-ssa-threadbackward.c (back_threader::dump): New.
3965 (back_threader::debug): New.
3966 (back_threader_profitability::profitable_path_p): Dump blocks
3967 even if we are bailing early.
3969 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3971 * tree-ssa-threadupdate.c (cancel_thread): New.
3972 (jump_thread_path_registry::thread_block_1): Use cancel_thread.
3973 (jump_thread_path_registry::mark_threaded_blocks): Same.
3974 (jump_thread_path_registry::register_jump_thread): Same.
3976 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3978 * tree-ssa-threadedge.c (jt_state::push): Only call methods for
3979 which objects are available.
3980 (jt_state::pop): Same.
3981 (jt_state::register_equiv): Same.
3982 (jt_state::register_equivs_on_edge): Same.
3984 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3986 * tree-ssa-threadedge.c (jump_threader::thread_across_edge):
3987 Move pop until after a thread is registered.
3989 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3991 * tree-ssa-threadupdate.c (debug): New.
3993 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
3995 * gimple-range-trace.cc (push_dump_file::push_dump_file): New.
3996 (push_dump_file::~push_dump_file): New.
3997 (dump_ranger): Change dump_file temporarily while dumping
3999 * gimple-range-trace.h (class push_dump_file): New.
4001 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4003 * gimple-range-trace.cc (debug_seed_ranger): Remove static.
4004 (dump_ranger): Dump function name.
4006 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4008 * gimple-range-path.cc (path_range_query::range_defined_in_block):
4009 Adjust for non-null.
4010 (path_range_query::adjust_for_non_null_uses): New.
4011 (path_range_query::precompute_ranges): Call
4012 adjust_for_non_null_uses.
4013 * gimple-range-path.h: Add m_non_null and
4014 adjust_for_non_null_uses.
4016 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4018 * gimple-range-path.cc (path_range_query::dump): Dump path
4020 (path_range_query::precompute_ranges): Dump entire path.
4022 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4024 * value-relation.cc (relation_oracle::debug): New.
4025 * value-relation.h (relation_oracle::debug): New.
4027 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4029 * tree-ssa-loop-ch.c: Remove unnecessary include file.
4031 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4033 * gimple-range-fold.cc (fold_using_range::postfold_gcond_edges):
4034 Skip statements with no defining BB.
4035 * gimple-range-path.cc (path_range_query::range_defined_in_block):
4036 Do not get confused by statements with no defining BB.
4038 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4040 * gimple-range-fold.cc (adjust_imagpart_expr): Move from
4041 gimple_range_adjustment. Add support for constants.
4042 (adjust_realpart_expr): New.
4043 (gimple_range_adjustment): Move IMAGPART_EXPR code to
4044 adjust_imagpart_expr.
4045 * range-op.cc (integral_table::integral_table): Add entry for
4048 2021-09-03 Jakub Jelinek <jakub@redhat.com>
4050 * omp-expand.c (expand_omp_atomic_pipeline): Use
4051 IFN_ATOMIC_COMPARE_EXCHANGE instead of
4052 BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_? so that memory order
4055 2021-09-03 Jakub Jelinek <jakub@redhat.com>
4058 * tree.h (DECL_FIELD_ABI_IGNORED): Changed into rvalue only macro
4059 that is false if DECL_BIT_FIELD.
4060 (SET_DECL_FIELD_ABI_IGNORED, DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD,
4061 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD): Define.
4062 * tree-streamer-out.c (pack_ts_decl_common_value_fields): For
4063 DECL_BIT_FIELD stream DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead
4064 of DECL_FIELD_ABI_IGNORED.
4065 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
4066 SET_DECL_FIELD_ABI_IGNORED instead of writing to
4067 DECL_FIELD_ABI_IGNORED and for DECL_BIT_FIELD use
4068 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead.
4069 * lto-streamer-out.c (hash_tree): For DECL_BIT_FIELD hash
4070 DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead of DECL_FIELD_ABI_IGNORED.
4072 2021-09-03 liuhongt <hongtao.liu@intel.com>
4075 * config/i386/amxbf16intrin.h : Remove macro check for __AMX_BF16__.
4076 * config/i386/amxint8intrin.h : Remove macro check for __AMX_INT8__.
4077 * config/i386/amxtileintrin.h : Remove macro check for __AMX_TILE__.
4079 2021-09-02 Martin Sebor <msebor@redhat.com>
4081 PR tree-optimization/17506
4083 * tree-ssa-uninit.c (warn_uninit): Remove conditional guarding note.
4085 2021-09-02 Richard Biener <rguenther@suse.de>
4087 * tree-ssa-loop-im.c (fill_always_executed_in_1): Refine
4088 fix for PR78185 and continue processing when leaving
4091 2021-09-02 Jakub Jelinek <jakub@redhat.com>
4093 PR tree-optimization/99591
4094 * match.pd: Demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
4097 2021-09-02 Richard Biener <rguenther@suse.de>
4100 2021-09-02 Richard Biener <rguenther@suse.de>
4102 PR tree-optimization/102155
4103 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
4104 over a part of the RPO array and do not recurse here.
4105 Dump blocks marked as always executed.
4106 (fill_always_executed_in): Walk over the RPO array and
4107 process loops whose header we run into.
4108 (loop_invariant_motion_in_fun): Compute the first RPO
4109 using rev_post_order_and_mark_dfs_back_seme in iteration
4110 order and pass that to fill_always_executed_in.
4112 2021-09-02 liuhongt <hongtao.liu@intel.com>
4114 * config/i386/i386-modes.def (FLOAT_MODE): Define ieee HFmode.
4115 * config/i386/i386.c (enum x86_64_reg_class): Add
4117 (merge_classes): Handle X86_64_SSEHF_CLASS.
4118 (examine_argument): Ditto.
4119 (construct_container): Ditto.
4120 (classify_argument): Ditto, and set HFmode/HCmode to
4122 (function_value_32): Return _FLoat16/Complex Float16 by
4124 (function_value_64): Return _Float16/Complex Float16 by SSE
4126 (ix86_print_operand): Handle CONST_DOUBLE HFmode.
4127 (ix86_secondary_reload): Require gpr as intermediate register
4128 to store _Float16 from sse register when sse4 is not
4130 (ix86_libgcc_floating_mode_supported_p): Enable _FLoat16 under
4132 (ix86_scalar_mode_supported_p): Ditto.
4133 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Defined.
4134 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add HFmode.
4135 (VALID_INT_MODE_P): Add HFmode and HCmode.
4136 * config/i386/i386.md (*pushhf_rex64): New define_insn.
4138 (*movhf_internal): Ditto.
4139 * doc/extend.texi (Half-Precision Floating Point): Documemt
4142 2021-09-02 Richard Biener <rguenther@suse.de>
4144 PR tree-optimization/102155
4145 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
4146 over a part of the RPO array and do not recurse here.
4147 Dump blocks marked as always executed.
4148 (fill_always_executed_in): Walk over the RPO array and
4149 process loops whose header we run into.
4150 (loop_invariant_motion_in_fun): Compute the first RPO
4151 using rev_post_order_and_mark_dfs_back_seme in iteration
4152 order and pass that to fill_always_executed_in.
4154 2021-09-02 YunQiang Su <syq@debian.org>
4157 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
4159 * config/mips/mips.c (mips_module_isa_name): New.
4160 mips_file_start: add .module mipsREV to all asm output
4162 2021-09-01 Jeff Law <jlaw@localhost.localdomain>
4164 PR tree-optimization/102152
4165 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Reduce a vector
4166 comparison to a scalar comparison before calling
4167 update_stmt_if_modified.
4169 2021-09-01 Andrew Pinski <apinski@marvell.com>
4172 * config/aarch64/aarch64.c (aarch64_expand_setmem):
4173 Check STRICT_ALIGNMENT before creating an overlapping
4176 2021-09-01 Martin Sebor <msebor@redhat.com>
4178 * gimple-ssa-warn-access.cc (get_size_range): Add argument.
4179 (check_access): Pass additional argument.
4180 (check_memop_access): Remove template and make a member function.
4181 (maybe_check_dealloc_call): Make a pass_waccess member function.
4182 (class pass_waccess): Add, rename, and remove members.
4183 (pass_waccess::pass_waccess): Adjust to name change.
4184 (pass_waccess::~pass_waccess): Same.
4185 (check_alloca): Make a member function.
4186 (check_alloc_size_call): Same.
4187 (check_strcat): Same.
4188 (check_strncat): Same.
4189 (check_stxcpy): Same.
4190 (check_stxncpy): Same.
4191 (check_strncmp): Same.
4192 (maybe_warn_rdwr_sizes): Rename...
4193 (pass_waccess::maybe_check_access_sizes): ...to this.
4194 (pass_waccess::check_call): Adjust to name changes.
4195 (pass_waccess::maybe_check_dealloc_call): Make a pass_waccess member
4197 (pass_waccess::execute): Adjust to name changes.
4198 * gimple-ssa-warn-access.h (check_memop_access): Remove.
4199 * pointer-query.cc (access_ref::phi): Handle null pointer.
4200 (access_ref::inform_access): Same.
4201 (pointer_query::put_ref): Modify a cached value, not a copy of it.
4202 (pointer_query::dump): New function.
4203 (compute_objsize_r): Avoid overwriting access_ref::bndrng. Cache
4205 * pointer-query.h (pointer_query::dump): Declare.
4206 * tree-ssa-strlen.c (get_range): Simplify. Use function query.
4207 (dump_strlen_info): Use function query.
4208 (printf_strlen_execute): Factor code out into pointer_query::put_ref.
4210 2021-09-01 Thomas Schwinge <thomas@codesourcery.com>
4212 * tree.c (walk_tree_1) <OMP_CLAUSE>: Simplify.
4214 2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
4216 * doc/extend.texi: Document unavailable attribute.
4217 * print-tree.c (print_node): Handle unavailable attribute.
4218 * tree-core.h (struct tree_base): Add a bit to carry unavailability.
4219 * tree.c (error_unavailable_use): New.
4220 * tree.h (TREE_UNAVAILABLE): New.
4221 (error_unavailable_use): New.
4223 2021-09-01 Jakub Jelinek <jakub@redhat.com>
4225 PR tree-optimization/102124
4226 * tree-vect-patterns.c (vect_recog_widen_op_pattern): For ORIG_CODE
4227 MINUS_EXPR, if itype is unsigned with smaller precision than type,
4228 add an extra cast to signed variant of itype to ensure sign-extension.
4230 2021-09-01 Martin Liska <mliska@suse.cz>
4232 * graph.c (draw_cfg_node_succ_edges): Do not color fallthru
4233 edges and rather use colors for TRUE and FALSE edges.
4235 2021-09-01 Richard Biener <rguenther@suse.de>
4237 PR tree-optimization/93491
4238 * tree-ssa-pre.c (compute_avail): Set BB_MAY_NOTRETURN
4239 after processing the stmt itself. Do not consider
4240 pure functions possibly not returning. Properly avoid
4241 adding possibly trapping calls to EXP_GEN when there's
4242 a preceeding possibly not returning call.
4243 * tree-ssa-sccvn.c (vn_reference_may_trap): Conservatively
4246 2021-09-01 Richard Biener <rguenther@suse.de>
4248 PR tree-optimization/102139
4249 * tree-vectorizer.h (vec_base_alignments): Adjust hash-map
4250 type to record a std::pair of the stmt-info and the innermost
4252 (dr_vec_info::group): New member.
4253 * tree-vect-data-refs.c (vect_record_base_alignment): Adjust.
4254 (vect_compute_data_ref_alignment): Verify the recorded
4255 base alignment can be used.
4256 (data_ref_pair): Remove.
4257 (dr_group_sort_cmp): Adjust.
4258 (vect_analyze_data_ref_accesses): Store the group-ID in the
4259 dr_vec_info and operate on a vector of dr_vec_infos.
4261 2021-09-01 YunQiang Su <yunqiang.su@cipunited.com>
4263 * read-md.c (md_reader::handle_enum): support value assignation.
4264 * doc/md.texi: record define_c_enum value assignation support.
4266 2021-09-01 Jakub Jelinek <jakub@redhat.com>
4268 PR tree-optimization/102141
4269 * gimple-ssa-store-merging.c (bswap_view_convert): Add BEFORE
4270 argument. If false, emit stmts after gsi instead of before, and
4272 (bswap_replace): Adjust callers. When converting output of bswap,
4273 emit VIEW_CONVERT prepratation stmts after a copy of gsi instead
4276 2021-09-01 liuhongt <hongtao.liu@intel.com>
4278 * emit-rtl.c (validate_subreg): Get rid of all float-int
4281 2021-09-01 liuhongt <hongtao.liu@intel.com>
4284 2021-08-30 liuhongt <hongtao.liu@intel.com>
4286 * expmed.c (extract_bit_field_1): Make sure we're playing with
4287 integral modes before call extract_integral_bit_field.
4288 (extract_integral_bit_field): Add a parameter of type
4289 scalar_int_mode which corresponds to of tmode.
4290 And call extract_and_convert_fixed_bit_field instead of
4291 extract_fixed_bit_field and convert_extracted_bit_field.
4292 (extract_and_convert_fixed_bit_field): New function, it's a
4293 combination of extract_fixed_bit_field and
4294 convert_extracted_bit_field.
4296 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
4298 * tree.c (walk_tree_1) <OMP_CLAUSE_TILE>: Handle three operands.
4300 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
4302 * omp-general.h (omp_is_reference): Rename to...
4303 (omp_privatize_by_reference): ... this. Adjust all users...
4304 * omp-general.c: ... here, ...
4305 * gimplify.c: ... here, ...
4306 * omp-expand.c: ... here, ...
4307 * omp-low.c: ... here.
4309 2021-08-31 Martin Sebor <msebor@redhat.com>
4311 * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Test
4312 pointer element for equality to zero, not that of the cotaining
4315 2021-08-31 Martin Sebor <msebor@redhat.com>
4317 * gcc-rich-location.h (gcc_rich_location): Make ctor explicit.
4319 2021-08-31 Martin Sebor <msebor@redhat.com>
4321 * function.h (function): Add comments.
4322 (get_range_query): Same. Add attribute returns nonnull.
4324 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
4326 * expr.c (convert_modes): Don't use subreg_promoted_mode on a
4327 SUBREG if it can't be guaranteed to a SUBREG_PROMOTED_VAR_P set.
4328 Instead use the standard (safer) is_a <scalar_int_mode> idiom.
4330 2021-08-31 Jeff Law <jlaw@localhost.localdomain>
4332 * config.gcc (cris-*-elf, cris-*-none): Remove dbxelf.h from
4334 (m32r-*-elf, m32rle-*-elf, m32r-*-linux): Likewise.
4335 (mn10300-*-*, am33_2.0-*-linux*): Likewise.
4336 (xtensa*-*-elf, xtensa*-*-linux, xtensa*-*-uclinux): Likewise.
4337 (m32c-*-elf*, m32c-*-rtems*): Likewise.
4338 * config/cris/cris.h (DBX_NO_XREFS): Remove.
4339 (DBX_CONTIN_LENGTH, DBX_CONTIN_CHAR): Likewise.
4340 * config/m32r/m32r.h (DBXOUT_SOURCE_LINE): Likewise.
4341 (DBX_DEBUGGING_INFO, DBX_CONTIN_LENGTH): Likewise.
4342 * config/mn10300/mn10300.h (DEFAULT_GDB_EXTENSIONS): Likewise.
4343 * config/mn10300/linux.h (DBX_REGISTER_NAMES): Likewise.
4345 2021-08-31 Marcel Vollweiler <marcel@codesourcery.com>
4347 * gimplify.c (gimplify_scan_omp_clauses): Error handling. 'ancestor' only
4348 allowed on target constructs and only with particular other clauses.
4349 * omp-expand.c (expand_omp_target): Output of 'sorry, not supported' if
4351 * omp-low.c (check_omp_nesting_restrictions): Error handling. No nested OpenMP
4352 structs when 'ancestor' is used.
4353 (scan_omp_1_stmt): No usage of OpenMP runtime routines in a target region when
4355 * tree-pretty-print.c (dump_omp_clause): Append 'ancestor'.
4356 * tree.h (OMP_CLAUSE_DEVICE_ANCESTOR): Define macro.
4358 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
4360 * expr.c (convert_modes): Preserve SUBREG_PROMOTED_VAR_P when
4361 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
4363 * simplify-rtx.c (simplify_unary_operation_1) [SIGN_EXTEND]:
4364 Likewise, preserve SUBREG_PROMOTED_VAR_P when creating a (wider)
4365 partial subreg from a SUBREG_PROMOTED_VAR_P subreg. Generate
4366 SIGN_EXTEND of the SUBREG_REG when a subreg would be paradoxical.
4367 [ZERO_EXTEND]: Likewise, preserve SUBREG_PROMOTED_VAR_P when
4368 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
4369 subreg. Generate ZERO_EXTEND of the SUBREG_REG when a subreg
4370 would be paradoxical.
4372 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
4374 * combine.c (combine_simplify_rtx): Avoid converting an explicit
4375 TRUNCATE into a lowpart SUBREG on !TRULY_NOOP_TRUNCATION targets.
4376 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
4378 2021-08-31 Richard Biener <rguenther@suse.de>
4380 PR tree-optimization/102142
4381 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Fix
4382 condition under which to unset the visited flag.
4384 2021-08-31 Richard Biener <rguenther@suse.de>
4386 PR middle-end/102129
4387 * tree-ssa-ter.c (find_replaceable_in_bb): Do not move
4388 possibly trapping expressions across calls.
4390 2021-08-31 Jakub Jelinek <jakub@redhat.com>
4392 PR tree-optimization/102134
4393 * tree-ssa-ccp.c (bit_value_binop) <case RSHIFT_EXPR>: If sgn is
4394 UNSIGNED and r1val | r1mask has MSB set, ensure lzcount doesn't
4397 2021-08-31 Andrew Pinski <apinski@marvell.com>
4400 * collect-utils.c (setup_signals): New declaration.
4401 * collect-utils.h (setup_signals): New function.
4402 * collect2.c (handler): Delete.
4403 (main): Instead of manually setting up the signals,
4404 just call setup_signals.
4405 * lto-wrapper.c (main): Likewise.
4407 2021-08-31 Andrew Pinski <apinski@marvell.com>
4410 * config/i386/i386-protos.h (x86_output_aligned_bss):
4411 Change align argument to unsigned type.
4412 (x86_elf_aligned_decl_common): Likewise.
4413 * config/i386/i386.c (x86_elf_aligned_decl_common): Likewise.
4414 (x86_output_aligned_bss): Likewise.
4416 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
4418 * config/mips/mips.c (mips_module_isa_name): New.
4419 mips_file_start: add .module mipsREV to all asm output
4421 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
4423 * config/mips/mips.h (struct mips_cpu_info): define enum mips_isa;
4424 use enum instead of int for 'isa' member.
4425 * config.gcc, config/mips/mips.c, config/mips/mips-cpus.def,
4426 config/mips/netbsd.h: replace hardcoded numbers with enum.
4428 2021-08-31 liuhongt <hongtao.liu@intel.com>
4430 * config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
4431 define_split to define_insn_and_split.
4432 (*avx2_eq<mode>3): Removed.
4433 (<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
4434 (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
4435 (*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
4437 (*avx2_gt<mode>3): Removed.
4438 (<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
4439 define_insn to define_expand, and adjust pattern.
4440 (UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.
4442 2021-08-30 David Malcolm <dmalcolm@redhat.com>
4445 * Makefile.in (ANALYZER_OBJS): Add analyzer/call-info.o.
4447 2021-08-30 Jason Merrill <jason@redhat.com>
4449 * doc/invoke.texi: Document -Wmissing-requires.
4451 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
4453 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Remove
4454 TARGET_EXTRA_BUILTINS guard.
4456 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
4458 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Change
4459 initialization of V2DI_type_node and unsigned_V2DI_type_node.
4461 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
4463 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Use the new
4464 decl when new_builtins_are_live.
4465 * config/rs6000/rs6000-builtin-new.def (__builtin_cfstring): New
4468 2021-08-30 Pat Haugen <pthaugen@linux.ibm.com>
4470 * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add
4471 OPTION_MASK_P10_FUSION_2STORE.
4472 (POWERPC_MASKS): Likewise.
4473 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
4474 store fusion for Power10.
4475 (is_fusable_store): New.
4476 (power10_sched_reorder): Likewise.
4477 (rs6000_sched_reorder): Do Power10 specific reordering.
4478 (rs6000_sched_reorder2): Likewise.
4479 * config/rs6000/rs6000.opt: Add new option.
4481 2021-08-30 Richard Biener <rguenther@suse.de>
4483 PR tree-optimization/102128
4484 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
4485 Move scanning for if-converted scalar code to the caller
4486 and instead delay clearing the visited flag for profitable
4488 (vect_slp_region): Cost all subgraphs before scheduling.
4489 For if-converted BB vectorization scan for scalar COND_EXPRs
4490 and do not vectorize if any found and the cost model is
4493 2021-08-30 Richard Biener <rguenther@suse.de>
4495 * common.opt (fexceptions): Mark
4496 EnabledBy(fnon-call-exceptions).
4497 * doc/invoke.texi (fnon-call-exceptions): Document this
4498 enables -fexceptions.
4500 2021-08-30 Sebastian Huber <sebastian.huber@embedded-brains.de>
4502 * tsystem.h (abort): Define abort() if inhibit_libc is defined and it
4503 is not already defined.
4505 2021-08-30 liuhongt <hongtao.liu@intel.com>
4507 * expmed.c (extract_bit_field_1): Make sure we're playing with
4508 integral modes before call extract_integral_bit_field.
4509 (extract_integral_bit_field): Add a parameter of type
4510 scalar_int_mode which corresponds to of tmode.
4511 And call extract_and_convert_fixed_bit_field instead of
4512 extract_fixed_bit_field and convert_extracted_bit_field.
4513 (extract_and_convert_fixed_bit_field): New function, it's a
4514 combination of extract_fixed_bit_field and
4515 convert_extracted_bit_field.
4517 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
4519 * config/darwin.c (darwin_libc_has_function): Do not run
4520 the checks for x86 or modern Darwin. Make sure that there
4521 is a value set for darwin_macosx_version_min before testing.
4523 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
4525 * config/i386/darwin.h (CLEAR_INSN_CACHE): New.
4527 2021-08-28 Jan Hubicka <hubicka@ucw.cz>
4529 * ipa-modref-tree.h (modref_access_node::merge): Break out
4530 logic combining offsets and logic merging ranges to ...
4531 (modref_access_node::combined_offsets): ... here
4532 (modref_access_node::update2): ... here
4533 (modref_access_node::closer_pair_p): New member function.
4534 (modref_access_node::forced_merge): New member function.
4535 (modre_ref_node::insert): Do merging when table is full.
4537 2021-08-28 YunQiang Su <yunqiang.su@cipunited.com>
4540 * config.gcc: MIPS: use N64 ABI by default if the triple end
4541 with -gnuabi64, which is used by Debian since 2013.
4543 2021-08-28 Alexandre Oliva <oliva@adacore.com>
4545 * ipa-modref.c (analyze_function): Skip debug stmts.
4546 * tree-inline.c (estimate_num_insn): Consider builtins even
4547 without a cgraph_node.
4549 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
4551 * config/h8300/bitfield.md (cstore<mode>4): Remove expander.
4552 * config/h8300/h8300.c (h8300_expand_branch): Remove function.
4553 * config/h8300/h8300-protos.h (h8300_expadn_branch): Remove prototype.
4554 * config/h8300/h8300.md (eqne): New code iterator.
4555 (geultu, geultu_to_c): Similarly.
4556 * config/h8300/testcompare.md (cstore<mode>4): Dummy expander.
4557 (store_c_<mode>, store_c_i_<mode>): New define_insn_and_splits
4558 (cmp<mode>_c): New pattern
4560 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
4562 * tree-ssa-dom.c (reduce_vector_comparison_to_scalar_comparison): New
4564 (dom_opt_dom_walker::optimize_stmt): Use it.
4566 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
4568 * config/darwin.c (finalize_ctors): Add a section-start linker-
4570 (finalize_dtors): Likewise.
4571 * config/darwin.h (MIN_LD64_INIT_TERM_START_LABELS): New.
4573 2021-08-27 Bill Schmidt <wschmidt@linux.ibm.com>
4575 * config/rs6000/rs6000-call.c (rs6000-builtins.h): New #include.
4576 (rs6000_init_builtins): Call rs6000_init_generated_builtins. Skip the
4577 old initialization logic when new builtins are enabled.
4578 * config/rs6000/rs6000-gen-builtins.c (write_decls): Rename
4579 rs6000_autoinit_builtins to rs6000_init_generated_builtins.
4580 (write_init_file): Likewise.
4582 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
4584 * configure.ac (darwin2[[0-9]]* | darwin19*): Alter use of
4585 gcc_GAS_CHECK_FEATURE to remove an extraneous parameter.
4586 (amdgcn-* | gcn-*) Likewise.
4588 2021-08-27 Anthony Sharp <anthonysharp15@gmail.com>
4590 * symbol-summary.h: Added missing template keyword.
4592 2021-08-27 Richard Biener <rguenther@suse.de>
4594 PR tree-optimization/45178
4595 * tree-ssa-dce.c (find_obviously_necessary_stmts): For
4596 infinite loops without exit do not mark control dependent
4597 edges of the latch necessary.
4599 2021-08-27 konglin1 <lingling.kong@intel.com>
4602 * config/i386/sse.md: (<avx512>scattersi<mode>): Add mask operand to
4604 (<avx512>scattersi<mode>): Likewise.
4605 (*avx512f_scattersi<VI48F:mode>): Merge mask operand to set_dest.
4606 (*avx512f_scatterdi<VI48F:mode>): Likewise
4608 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
4610 * config/rs6000/rs6000.c (rs6000_builtin_md_vectorized_function): Add
4611 support for built-in functions MISC_BUILTIN_DIVWE, MISC_BUILTIN_DIVWEU,
4612 MISC_BUILTIN_DIVDE, MISC_BUILTIN_DIVDEU, P10_BUILTIN_CFUGED,
4613 P10_BUILTIN_CNTLZDM, P10_BUILTIN_CNTTZDM, P10_BUILTIN_PDEPD and
4614 P10_BUILTIN_PEXTD on Power10.
4616 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
4618 * config/rs6000/rs6000-call.c (builtin_function_type): Add unsigned
4619 signedness for some Power10 bifs.
4621 2021-08-27 David Edelsohn <dje.gcc@gmail.com>
4624 * config/rs6000/rs6000.c (rs6000_adjust_field_align): Use
4625 computed alignment if the entire struct has attribute packed.
4627 2021-08-27 liuhongt <hongtao.liu@intel.com>
4631 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
4632 IX86_BUILTIN_SHUFPD512, IX86_BUILTIN_SHUFPS512,
4633 IX86_BUILTIN_SHUFPD256, IX86_BUILTIN_SHUFPS,
4634 IX86_BUILTIN_SHUFPS256.
4635 (ix86_masked_all_ones): New function.
4637 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
4639 * config/i386/i386.md (*btr<mode>_1): Call force_reg unconditionally.
4640 (conditional moves with memory inputs splitters): Ditto.
4641 * config/i386/sse.md (one_cmpl<mode>2): Simplify.
4643 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
4645 * ipa-modref-tree.h (modref_access_node::try_merge_with): Restart
4646 search after merging.
4648 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
4650 * config/rs6000/rs6000-overload.def: Add remaining overloads.
4652 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
4654 * config/rs6000/rs6000-builtin-new.def: Add cell stanza.
4656 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
4658 * config/rs6000/rs6000-builtin-new.def: Add ieee128-hw, dfp,
4659 crypto, and htm stanzas.
4661 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
4663 * config/rs6000/rs6000-builtin-new.def: Add mma stanza.
4665 2021-08-26 Martin Sebor <msebor@redhat.com>
4667 * tree-ssa-uninit.c (warn_uninit): Refactor and simplify.
4668 (warn_uninit_phi_uses): Remove argument from calls to warn_uninit.
4669 (warn_uninitialized_vars): Same. Reduce visibility of locals.
4670 (warn_uninitialized_phi): Same.
4672 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
4674 * tree-ssa-ccp.c (get_individual_bits): Helper function to
4675 extract the individual bits from a widest_int constant (mask).
4676 (gray_code_bit_flips): New read-only table for effiently
4677 enumerating permutations/combinations of bits.
4678 (bit_value_binop) [LROTATE_EXPR, RROTATE_EXPR]: Handle rotates
4679 by unknown counts that are guaranteed less than the target
4680 precision and four or fewer unknown bits by enumeration.
4681 [LSHIFT_EXPR, RSHIFT_EXPR]: Likewise, also handle shifts by
4682 enumeration under the same conditions. Handle remaining
4683 shifts as a mask based upon the minimum possible shift value.
4685 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
4686 Richard Biener <rguenther@suse.de>
4688 * match.pd (shift transformations): Remove a redundant
4689 !POINTER_TYPE_P check.
4691 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
4694 * config/i386/i386.md (cmove reg-reg move elimination peephole2s):
4695 Set all_regs to true in the call to replace_rtx.
4697 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
4699 * ipa-modref-tree.c (test_insert_search_collapse): Update test.
4700 * ipa-modref-tree.h (modref_base_node::insert): Be smarter when
4701 hiting --param modref-max-refs limit.
4702 (modref_tree:insert_base): Be smarter when hitting
4703 --param modref-max-bases limit. Add new parameter REF.
4704 (modref_tree:insert): Update.
4705 (modref_tree:merge): Update.
4706 * ipa-modref.c (read_modref_records): Update.
4708 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
4710 * params.opt: (modref-max-adjustments): Add full stop.
4712 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
4714 * ipa-modref-tree.h (modref_ref_node::verify): New member
4716 (modref_ref_node::insert): Use it.
4717 (modref_ref_node::try_mere_with): Fix off by one error.
4719 2021-08-26 Martin Liska <mliska@suse.cz>
4720 Stefan Kneifel <stefan.kneifel@bluewin.ch>
4722 * cgraph.h (create_version_clone_with_body): Add new parameter.
4723 * cgraphclones.c: Likewise.
4724 * multiple_target.c (create_dispatcher_calls): Do not use
4726 (create_target_clone): Likewise here.
4728 2021-08-26 Jonathan Yong <10walls@gmail.com>
4730 * doc/extend.texi: Add note about reserved priorities
4731 to the constructor attribute.
4733 2021-08-25 Martin Sebor <msebor@redhat.com>
4735 * gimple-range-cache.cc (ssa_global_cache::dump): Avoid printing
4736 range table header alone.
4737 * gimple-range.cc (gimple_ranger::export_global_ranges): Same.
4739 2021-08-25 Jan Hubicka <hubicka@ucw.cz>
4741 * doc/invoke.texi: Document --param modref-max-adjustments.
4742 * ipa-modref-tree.c (test_insert_search_collapse): Update.
4743 (test_merge): Update.
4744 * ipa-modref-tree.h (struct modref_access_node): Add adjustments;
4745 (modref_access_node::operator==): Fix handling of access ranges.
4746 (modref_access_node::contains): Constify parameter; handle also
4747 mismatched parm offsets.
4748 (modref_access_node::update): New function.
4749 (modref_access_node::merge): New function.
4750 (unspecified_modref_access_node): Update constructor.
4751 (modref_ref_node::insert_access): Add record_adjustments parameter;
4753 (modref_ref_node::try_merge_with): New private function.
4754 (modref_tree::insert): New record_adjustments parameter.
4755 (modref_tree::merge): New record_adjustments parameter.
4756 (modref_tree::copy_from): Update.
4757 * ipa-modref.c (dump_access): Dump adjustments field.
4758 (get_access): Update constructor.
4759 (record_access): Update call of insert.
4760 (record_access_lto): Update call of insert.
4761 (merge_call_side_effects): Add record_adjustments parameter.
4762 (get_access_for_fnspec): Update.
4763 (process_fnspec): Update.
4764 (analyze_call): Update.
4765 (analyze_function): Update.
4766 (read_modref_records): Update.
4767 (ipa_merge_modref_summary_after_inlining): Update.
4768 (propagate_unknown_call): Update.
4769 (modref_propagate_in_scc): Update.
4770 * params.opt (param-max-modref-adjustments=): New.
4772 2021-08-25 Michael Meissner <meissner@linux.ibm.com>
4774 * config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
4776 (xxspltiw_v4si): Use vecperm type attribute.
4777 (xxspltiw_v4si_inst): Use vecperm type attribute.
4778 (xxspltiw_v4sf_inst): Likewise.
4779 (xxspltidp_v2df): Use vecperm type attribute. Use
4780 UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
4781 (xxspltidp_v2df_inst): Likewise.
4782 (xxsplti32dx_v4si): Use vecperm type attribute.
4783 (xxsplti32dx_v4si_inst): Likewise.
4784 (xxsplti32dx_v4sf_inst): Likewise.
4785 (xxblend_<mode>): Likewise.
4786 (xxpermx): Likewise.
4787 (xxpermx_inst): Likewise.
4790 2021-08-25 Lewis Hyatt <lhyatt@gmail.com>
4793 * coretypes.h (typedef diagnostic_input_charset_callback): Declare.
4794 * diagnostic.c (diagnostic_initialize_input_context): New function.
4795 * diagnostic.h (diagnostic_initialize_input_context): Declare.
4796 * input.c (default_charset_callback): New function.
4797 (file_cache::initialize_input_context): New function.
4798 (file_cache_slot::create): Added ability to convert the input
4799 according to the input context.
4800 (file_cache::file_cache): Initialize the new input context.
4801 (class file_cache_slot): Added new m_alloc_offset member.
4802 (file_cache_slot::file_cache_slot): Initialize the new member.
4803 (file_cache_slot::~file_cache_slot): Handle potentially offset buffer.
4804 (file_cache_slot::maybe_grow): Likewise.
4805 (file_cache_slot::needs_read_p): Handle NULL fp, which is now possible.
4806 (file_cache_slot::get_next_line): Likewise.
4807 * input.h (class file_cache): Added input context member.
4809 2021-08-25 Richard Biener <rguenther@suse.de>
4811 PR tree-optimization/102046
4812 * tree-vect-slp.c (vect_build_slp_tree_2): Conservatively
4813 update ->any_pattern when swapping operands.
4815 2021-08-25 Hongyu Wang <hongyu.wang@intel.com>
4818 * config/i386/i386.c (ix86_live_on_entry): Adjust comment.
4819 (ix86_decompose_address): Remove retval check for ASHIFT,
4820 allow non-canonical zero extend if AND mask covers ASHIFT
4822 (ix86_legitimate_address_p): Adjust condition for decompose.
4823 (ix86_rtx_costs): Adjust cost for lea with non-canonical
4825 Co-Authored by: Uros Bizjak <ubizjak@gmail.com>
4827 2021-08-25 Jiufu Guo <guojiufu@linux.ibm.com>
4829 PR tree-optimization/101145
4830 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
4832 (number_of_iterations_lt): Invoke above function.
4833 (adjust_cond_for_loop_until_wrap):
4834 Merge to number_of_iterations_until_wrap.
4835 (number_of_iterations_cond): Update invokes for
4836 adjust_cond_for_loop_until_wrap and number_of_iterations_lt.
4838 2021-08-25 konglin1 <lingling.kong@intel.com>
4841 * config/i386/avx512dqintrin.h (_mm512_fpclass_ps_mask): Fix
4843 (_mm512_mask_fpclass_ps_mask): Ditto.
4845 2021-08-25 Kewen Lin <linkw@linux.ibm.com>
4847 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Remove.
4848 (vec_unpacku_hi_v8hi): Likewise.
4849 (vec_unpacku_lo_v16qi): Likewise.
4850 (vec_unpacku_lo_v8hi): Likewise.
4851 (vec_unpacku_hi_<VP_small_lc>): New define_expand.
4852 (vec_unpacku_lo_<VP_small_lc>): Likewise.
4854 2021-08-24 David Edelsohn <dje.gcc@gmail.com>
4856 * config/rs6000/aix.h (SYSTEM_IMPLICIT_EXTERN_C): Delete.
4857 * config/rs6000/aix71.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
4858 * config/rs6000/aix72.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
4859 * config/rs6000/aix73.h (TARGET_AIX_VERSION): Increase to 73.
4861 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
4863 PR middle-end/102031
4864 * simplify-rtx.c (simplify_truncation): When comparing precisions
4865 use "subreg_prec" variable, not "subreg_mode".
4867 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
4869 * config/rs6000/rs6000-builtin-new.def: Add power10 and power10-64
4872 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
4874 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Initialize
4875 various pointer type nodes.
4876 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add enum
4877 values for various pointer types.
4878 (ptr_V16QI_type_node): New macro.
4879 (ptr_V1TI_type_node): New macro.
4880 (ptr_V2DI_type_node): New macro.
4881 (ptr_V2DF_type_node): New macro.
4882 (ptr_V4SI_type_node): New macro.
4883 (ptr_V4SF_type_node): New macro.
4884 (ptr_V8HI_type_node): New macro.
4885 (ptr_unsigned_V16QI_type_node): New macro.
4886 (ptr_unsigned_V1TI_type_node): New macro.
4887 (ptr_unsigned_V8HI_type_node): New macro.
4888 (ptr_unsigned_V4SI_type_node): New macro.
4889 (ptr_unsigned_V2DI_type_node): New macro.
4890 (ptr_bool_V16QI_type_node): New macro.
4891 (ptr_bool_V8HI_type_node): New macro.
4892 (ptr_bool_V4SI_type_node): New macro.
4893 (ptr_bool_V2DI_type_node): New macro.
4894 (ptr_bool_V1TI_type_node): New macro.
4895 (ptr_pixel_type_node): New macro.
4896 (ptr_intQI_type_node): New macro.
4897 (ptr_uintQI_type_node): New macro.
4898 (ptr_intHI_type_node): New macro.
4899 (ptr_uintHI_type_node): New macro.
4900 (ptr_intSI_type_node): New macro.
4901 (ptr_uintSI_type_node): New macro.
4902 (ptr_intDI_type_node): New macro.
4903 (ptr_uintDI_type_node): New macro.
4904 (ptr_intTI_type_node): New macro.
4905 (ptr_uintTI_type_node): New macro.
4906 (ptr_long_integer_type_node): New macro.
4907 (ptr_long_unsigned_type_node): New macro.
4908 (ptr_float_type_node): New macro.
4909 (ptr_double_type_node): New macro.
4910 (ptr_long_double_type_node): New macro.
4911 (ptr_dfloat64_type_node): New macro.
4912 (ptr_dfloat128_type_node): New macro.
4913 (ptr_ieee128_type_node): New macro.
4914 (ptr_ibm128_type_node): New macro.
4915 (ptr_vector_pair_type_node): New macro.
4916 (ptr_vector_quad_type_node): New macro.
4917 (ptr_long_long_integer_type_node): New macro.
4918 (ptr_long_long_unsigned_type_node): New macro.
4920 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
4922 * config/rs6000/rs6000-builtin-new.def: Add power9-vector, power9,
4923 and power9-64 stanzas.
4925 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
4926 Tom de Vries <tdevries@suse.de>
4928 * config.gcc (nvptx-*-*): Define {c,c++}_target_objs.
4929 * config/nvptx/nvptx-protos.h (nvptx_cpu_cpp_builtins): Prototype.
4930 * config/nvptx/nvptx.h (TARGET_CPU_CPP_BUILTINS): Implement with
4931 a call to the new nvptx_cpu_cpp_builtins function in nvptx-c.c.
4932 * config/nvptx/t-nvptx (nvptx-c.o): New rule.
4933 * config/nvptx/nvptx-c.c: New source file.
4934 (nvptx_cpu_cpp_builtins): Move implementation here.
4936 2021-08-24 Martin Sebor <msebor@redhat.com>
4938 PR middle-end/101600
4939 PR middle-end/101977
4940 * gimple-ssa-warn-access.cc (maybe_warn_for_bound): Tighten up
4941 the phrasing of a warning.
4942 (check_access): Use the remaining size after subtracting any offset
4943 rather than the whole object size.
4944 * pointer-query.cc (access_ref::get_ref): Clear BASE0 flag if it's
4945 clear for any nonnull PHI argument.
4946 (compute_objsize): Clear argument.
4948 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
4950 * config/rs6000/rs6000-builtin-new.def: Add power8-vector stanza.
4952 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
4954 * config/rs6000/rs6000-builtin-new.def: Add power7 and power7-64
4957 2021-08-24 Andrew MacLeod <amacleod@redhat.com>
4959 * value-relation.cc (rr_transitive_table): New.
4960 (relation_transitive): New.
4961 (value_relation::swap): Remove.
4962 (value_relation::apply_transitive): New.
4963 (relation_oracle::relation_oracle): Allocate a new tmp bitmap.
4964 (relation_oracle::register_relation): Call register_transitives.
4965 (relation_oracle::register_transitives): New.
4966 * value-relation.h (relation_oracle): Add new temporary bitmap and
4969 2021-08-24 H.J. Lu <hjl.tools@gmail.com>
4972 * config/i386/i386-expand.c (ix86_expand_vector_move): Broadcast
4973 from integer to a pseudo vector register.
4975 2021-08-24 Richard Biener <rguenther@suse.de>
4977 PR tree-optimization/100089
4978 * tree-vectorizer.h (vect_slp_bb): Rename to ...
4979 (vect_slp_if_converted_bb): ... this and get the original
4980 loop as new argument.
4981 * tree-vectorizer.c (try_vectorize_loop_1): Revert previous fix,
4982 pass original loop to vect_slp_if_converted_bb.
4983 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
4984 If orig_loop was passed scan the not vectorized stmts
4985 for COND_EXPRs and force not profitable if found.
4986 (vect_slp_region): Pass down all SLP instances to costing
4987 if orig_loop was specified.
4988 (vect_slp_bbs): Pass through orig_loop.
4989 (vect_slp_bb): Rename to ...
4990 (vect_slp_if_converted_bb): ... this and get the original
4991 loop as new argument.
4992 (vect_slp_function): Adjust.
4994 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
4997 * config/arm/arm.md (attribute arch): Add fix_vlldm.
4998 (arch_enabled): Use it.
4999 * config/arm/vfp.md (lazy_store_multiple_insn): Add alternative to
5000 use when erratum mitigation is needed.
5002 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
5005 * config/arm/arm.opt (mfix-cmse-cve-2021-35465): New option.
5006 * doc/invoke.texi (Arm Options): Document it.
5007 * config/arm/arm-cpus.in (quirk_vlldm): New feature bit.
5008 (ALL_QUIRKS): Add quirk_vlldm.
5009 (cortex-m33): Add quirk_vlldm.
5010 (cortex-m35p, cortex-m55): Likewise.
5011 * config/arm/arm.c (arm_option_override): Enable fix_vlldm if
5012 targetting an affected CPU and not explicitly controlled on
5015 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
5017 * config/arm/vfp.md (lazy_store_multiple_insn): Rewrite as valid RTL.
5018 (lazy_load_multiple_insn): Likewise.
5020 2021-08-24 liuhongt <hongtao.liu@intel.com>
5023 * config/i386/sse.md (<avx512>_vternlog<mode><sd_maskz_name>):
5024 Enable avx512 embedded broadcast.
5025 (*<avx512>_vternlog<mode>_all): Ditto.
5026 (<avx512>_vternlog<mode>_mask): Ditto.
5028 2021-08-24 liuhongt <hongtao.liu@intel.com>
5031 * config/i386/i386.c (ix86_rtx_costs): Define cost for
5033 * config/i386/i386.h (STRIP_UNARY): New macro.
5034 * config/i386/predicates.md (reg_or_notreg_operand): New
5036 * config/i386/sse.md (*<avx512>_vternlog<mode>_all): New define_insn.
5037 (*<avx512>_vternlog<mode>_1): New pre_reload
5038 define_insn_and_split.
5039 (*<avx512>_vternlog<mode>_2): Ditto.
5040 (*<avx512>_vternlog<mode>_3): Ditto.
5041 (any_logic1,any_logic2): New code iterator.
5042 (logic_op): New code attribute.
5043 (ternlogsuffix): Extend to VNxDF and VNxSF.
5045 2021-08-24 Richard Biener <rguenther@suse.de>
5047 * doc/invoke.texi (vect-inner-loop-cost-factor): Adjust.
5048 * params.opt (--param vect-inner-loop-cost-factor): Adjust
5050 * tree-vect-loop.c (vect_analyze_loop_form): Initialize
5051 inner_loop_cost_factor to the minimum of the estimated number
5052 of iterations of the inner loop and vect-inner-loop-cost-factor.
5054 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5055 Richard Biener <rguenther@suse.de>
5057 * config/i386/i386-features.c (compute_convert_gain): Provide
5058 more accurate values for CONST_INT, when optimizing for size.
5059 * config/i386/i386.c (COSTS_N_BYTES): Move definition from here...
5060 * config/i386/i386.h (COSTS_N_BYTES): to here.
5062 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5063 Jakub Jelinek <jakub@redhat.com>
5065 PR middle-end/102029
5066 * match.pd (shift transformations): Add an additional check for
5067 !POINTER_TYPE_P in the recently added left shift transformation.
5069 2021-08-24 liuhongt <hongtao.liu@intel.com>
5071 PR tree-optimization/100089
5072 * tree-vectorizer.c (try_vectorize_loop_1): Disable slp in
5073 loop vectorizer when cost model is very-cheap.
5075 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5077 * config/rs6000/rs6000-gen-builtins.c (parse_bif_entry): Don't call
5078 asprintf, which is not available on AIX.
5080 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5082 * config.gcc (target_gtfiles): Add ./rs6000-builtins.h.
5083 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Set.
5085 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5087 * config.gcc (powerpc*-*-*): Add rs6000-builtins.o to extra_objs.
5088 * config/rs6000/rs6000-gen-builtins.c (main): Close init_file
5090 * config/rs6000/t-rs6000 (rs6000-gen-builtins.o): New target.
5091 (rbtree.o): Likewise.
5092 (rs6000-gen-builtins): Likewise.
5093 (rs6000-builtins.c): Likewise.
5094 (rs6000-builtins.h): Likewise.
5095 (rs6000.o): Add dependency.
5096 (EXTRA_HEADERS): Add rs6000-vecdefines.h.
5097 (rs6000-vecdefines.h): New target.
5098 (rs6000-builtins.o): Likewise.
5099 (rs6000-call.o): Add rs6000-builtins.h as a dependency.
5100 (rs6000-c.o): Likewise.
5102 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5105 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace):
5106 Diagnose buffer overrun.
5107 (safe_inc_pos): Fix overrun detection.
5108 (match_identifier): Diagnose buffer overrun.
5109 (match_integer): Likewise.
5110 (match_to_right_bracket): Likewise.
5112 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
5114 * ipa-modref-tree.h (modref_access_node::range_info_useful_p):
5115 Improve range compare.
5116 (modref_access_node::contains): New member function.
5117 (modref_access_node::search): Remove.
5118 (modref_access_node::insert): Be smarter about subaccesses.
5120 2021-08-23 Thomas Schwinge <thomas@codesourcery.com>
5122 * config/i386/i386-options.c (ix86_omp_device_kind_arch_isa)
5123 <omp_device_arch> [ACCEL_COMPILER]: Match "intel_mic".
5124 * config/i386/t-omp-device (omp-device-properties-i386) <arch>:
5127 2021-08-23 Jeff Law <jlaw@localhost.localdomain>
5129 * config/h8300/h8300-protos.h (h8300_expand_epilogue): Add new
5131 * config/h8300/jumpcall.md (call, call_value): Restrict to
5132 !SIBLING_CALL_P cases.
5133 (subcall, sibcall_value): New patterns & expanders.
5134 * config/h8300/proepi.md (epilogue): Pass new argument to
5135 h8300_expand_epilogue.
5136 (sibcall_epilogue): New expander.
5137 * config/h8300/h8300.c (h8300_expand_epilogue): Handle sibcall
5139 (h8300_ok_for_sibcall_p): New function.
5140 (TARGET_FUNCTION_OK_FOR_SIBCALL): define.
5142 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5144 * simplify-rtx.c (simplify_unary_operation_1): [TRUNCATE]:
5145 Handle case where the operand is already the desired mode.
5147 2021-08-23 Richard Biener <rguenther@suse.de>
5150 * tree-ssa-structalias.c (ipa_pta_execute): Check in_other_partition
5151 in addition to has_gimple_body.
5153 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
5155 PR middle-end/101949
5156 * ipa-modref.c (analyze_ssa_name_flags): Fix merging of
5159 2021-08-23 Martin Liska <mliska@suse.cz>
5161 * doc/invoke.texi: Put the option out of -mxl-mode-app-model
5164 2021-08-23 Richard Biener <rguenther@suse.de>
5166 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5167 Properly scale the inner loop cost only once.
5169 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5171 * tree-ssa-ccp.c (bit_value_binop) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]:
5172 Provide bounds for unsigned (and signed with non-negative operands)
5173 division and modulus.
5175 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5177 * simplify-rtx.c (simplify_truncation): Generalize simplification
5178 of (truncate:A (subreg:B X)).
5179 (simplify_unary_operation_1) [FLOAT_TRUNCATE, FLOAT_EXTEND,
5180 SIGN_EXTEND, ZERO_EXTEND]: Handle cases where the operand
5181 already has the desired machine mode.
5182 (test_scalar_int_ops): Add tests that useless extensions and
5183 truncations are optimized away.
5184 (test_scalar_int_ext_ops): New self-test function to confirm
5185 that truncations of extensions are correctly simplified.
5186 (test_scalar_int_ext_ops2): New self-test function to check
5187 truncations of truncations, extensions of extensions, and
5188 truncations of extensions.
5189 (test_scalar_ops): Call the above two functions with a
5190 representative sampling of integer machine modes.
5192 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5194 * match.pd (shift transformations): Change the sign of an
5195 LSHIFT_EXPR if it reduces the number of explicit conversions.
5197 2021-08-23 Jakub Jelinek <jakub@redhat.com>
5199 PR tree-optimization/86723
5200 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
5201 cast64_to_32 argument, set *cast64_to_32 to false, unless n is
5202 non-memory permutation of 64-bit src which only has bytes of
5203 0 or [5..8] and n->range is 4.
5204 (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
5205 find_bswap_or_nop_finalize caller, support bswap with some bytes
5206 zeroed, as long as at least two bytes are not zeroed.
5207 (bswap_replace): Add mask argument and handle masking of bswap
5209 (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
5210 caller, punt if cast64_to_32 or mask is not all ones.
5211 (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
5212 caller, for now punt if cast64_to_32.
5214 2021-08-23 Richard Biener <rguenther@suse.de>
5216 PR tree-optimization/79334
5217 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Record
5218 a type also for COMPONENT_REFs.
5219 (vn_reference_may_trap): Check ARRAY_REF with constant index
5220 against the array domain.
5222 2021-08-23 liuhongt <hongtao.liu@intel.com>
5225 * config/i386/sse.md (*avx512f_pshufb_truncv8hiv8qi_1): Add
5226 TARGET_AVX512BW to condition.
5228 2021-08-23 Jakub Jelinek <jakub@redhat.com>
5231 * dwarf2out.c (gen_variable_die): Add DW_AT_location for global
5232 register variables already during early_dwarf if possible.
5234 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
5236 * config/arm/arm_mve.h: Fix __arm_vctp16q return type.
5238 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
5241 * config/arm/arm.opt: Fix typo.
5242 * config/arm/t-rmprofile: Fix typo.
5244 2021-08-23 Jakub Jelinek <jakub@redhat.com>
5246 * tree.h (OMP_CLAUSE_GRAINSIZE_STRICT): Define.
5247 (OMP_CLAUSE_NUM_TASKS_STRICT): Define.
5248 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_GRAINSIZE,
5249 case OMP_CLAUSE_NUM_TASKS>: Print strict: modifier.
5250 * omp-expand.c (expand_task_call): Use GOMP_TASK_FLAG_STRICT in iflags
5251 if either grainsize or num_tasks clause has the strict modifier.
5253 2021-08-23 Martin Liska <mliska@suse.cz>
5255 * dbgcnt.def (DEBUG_COUNTER): New counter.
5256 * gimple.c (gimple_call_arg_flags): Use it in IPA PTA.
5258 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
5260 * ipa-modref.c (analyze_ssa_name_flags): Improve handling of return slot.
5262 2021-08-23 Xi Ruoyao <xry111@mengyan1223.wang>
5265 * config/mips/mips-protos.h (mips_msa_output_shift_immediate):
5267 * config/mips/mips.c (mips_msa_output_shift_immediate): New
5269 * config/mips/mips-msa.md (vashl<mode>3, vashr<mode>3,
5270 vlshr<mode>3): Call it.
5272 2021-08-22 Jan Hubicka <hubicka@ucw.cz>
5273 Martin Liska <mliska@suse.cz>
5275 PR middle-end/101949
5276 * ipa-modref.c (analyze_ssa_name_flags): Indirect call implies
5279 2021-08-21 Dragan Mladjenovic <OT_Dragan.Mladjenovic@mediatek.com>
5281 * config/mips/mips.c (mips_function_rodata_section,
5282 TARGET_ASM_FUNCTION_RODATA_SECTION): Removed.
5284 2021-08-21 John David Anglin <danglin@gcc.gnu.org>
5286 * config/pa/pa.c (pa_asm_output_aligned_common): Remove warning.
5288 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5290 * configure.ac (thread-local storage support): Remove tls_first_major
5291 and tls_first_minor. Use "$conftest_s" to check support.
5292 * configure: Regenerate.
5294 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5296 * configure.ac: Fixup formatting.
5298 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5300 * acinclude.m4 (gcc_GAS_CHECK_FEATURE): Remove third argument and ...
5301 * configure.ac: ... update all callers.
5303 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5306 * acinclude.m4 (_gcc_COMPUTE_GAS_VERSION, _gcc_GAS_VERSION_GTE_IFELSE)
5307 (gcc_GAS_VERSION_GTE_IFELSE): Remove.
5308 (gcc_GAS_CHECK_FEATURE): Do not handle in-tree case specially.
5309 * configure.ac: Remove gcc_cv_gas_major_version, gcc_cv_gas_minor_version.
5310 Remove remaining checks for in-tree assembler.
5311 * configure: Regenerate.
5313 2021-08-20 Jeff Law <jlaw@localhost.localdomain>
5315 * config/h8300/h8300.c (shift_alg_hi): Improve arithmetic shift right
5316 by 15 bits for H8/300H and H8/S. Improve logical shifts by 12
5318 (shift_alg_si): Improve arithmetic right shift by 28-30 bits for
5319 H8/300H. Improve arithmetic shift right by 15 bits for H8/S.
5320 Improve logical shifts by 27 bits for H8/S.
5321 (get_shift_alg): Corresponding changes.
5322 (h8300_option_override): Revert to loops for -Os when profitable.
5324 2021-08-20 Richard Biener <rguenther@suse.de>
5326 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
5328 (vect_analyze_data_ref_accesses): Likewise. Assign the BB
5329 index as group_id when dataref_groups were not computed.
5330 * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
5331 we advace to the next BB.
5333 2021-08-20 Jakub Jelinek <jakub@redhat.com>
5335 * omp-builtins.def (BUILT_IN_GOMP_WARNING, BUILT_IN_GOMP_ERROR): New
5338 2021-08-20 Martin Liska <mliska@suse.cz>
5340 PR gcov-profile/89961
5341 * gcov.c (make_gcov_file_name): Rewrite using std::string.
5342 (mangle_name): Simplify, do not used the second argument.
5343 (strip_extention): New function.
5344 (get_md5sum): Likewise.
5345 (get_gcov_intermediate_filename): Handle properly -p and -x
5347 (output_gcov_file): Use string type.
5348 (generate_results): Likewise.
5349 (md5sum_to_hex): Remove.
5351 2021-08-20 Michael Meissner <meissner@linux.ibm.com>
5353 * config/rs6000/altivec.md (UNSPEC_XXEVAL): Move to vsx.md.
5354 (UNSPEC_XXSPLTIW): Move to vsx.md.
5355 (UNSPEC_XXSPLTID): Move to vsx.md.
5356 (UNSPEC_XXSPLTI32DX): Move to vsx.md.
5357 (UNSPEC_XXBLEND): Move to vsx.md.
5358 (UNSPEC_XXPERMX): Move to vsx.md.
5359 (VM3): Move to vsx.md.
5360 (VM3_char): Move to vsx.md.
5361 (xxspltiw_v4si): Move to vsx.md.
5362 (xxspltiw_v4sf): Move to vsx.md.
5363 (xxspltiw_v4sf_inst): Move to vsx.md.
5364 (xxspltidp_v2df): Move to vsx.md.
5365 (xxspltidp_v2df_inst): Move to vsx.md.
5366 (xxsplti32dx_v4si_inst): Move to vsx.md.
5367 (xxsplti32dx_v4sf): Move to vsx.md.
5368 (xxsplti32dx_v4sf_inst): Move to vsx.md.
5369 (xxblend_<mode>): Move to vsx.md.
5370 (xxpermx): Move to vsx.md.
5371 (xxpermx_inst): Move to vsx.md.
5372 * config/rs6000/vsx.md (UNSPEC_XXEVAL): Move from altivec.md.
5373 (UNSPEC_XXSPLTIW): Move from altivec.md.
5374 (UNSPEC_XXSPLTID): Move from altivec.md.
5375 (UNSPEC_XXSPLTI32DX): Move from altivec.md.
5376 (UNSPEC_XXBLEND): Move from altivec.md.
5377 (UNSPEC_XXPERMX): Move from altivec.md.
5378 (VM3): Move from altivec.md.
5379 (VM3_char): Move from altivec.md.
5380 (xxspltiw_v4si): Move from altivec.md.
5381 (xxspltiw_v4sf): Move from altivec.md.
5382 (xxspltiw_v4sf_inst): Move from altivec.md.
5383 (xxspltidp_v2df): Move from altivec.md.
5384 (xxspltidp_v2df_inst): Move from altivec.md.
5385 (xxsplti32dx_v4si_inst): Move from altivec.md.
5386 (xxsplti32dx_v4sf): Move from altivec.md.
5387 (xxsplti32dx_v4sf_inst): Move from altivec.md.
5388 (xxblend_<mode>): Move from altivec.md.
5389 (xxpermx): Move from altivec.md.
5390 (xxpermx_inst): Move from altivec.md.
5392 2021-08-19 Roger Sayle <roger@nextmovesoftware.com>
5394 * tree-vect-generic.c (expand_vector_operations_1): Use either
5395 gimplify_build1 or gimplify_build2 instead of gimple_build_assign
5396 when constructing scalar splat expressions.
5398 2021-08-19 Peter Bergner <bergner@linux.ibm.com>
5401 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Cast
5402 pointer to __vector_pair *.
5404 2021-08-19 Martin Sebor <msebor@redhat.com>
5406 * gimple-range.cc: Add comments.
5407 * gimple-range.h: Same.
5409 2021-08-19 Martin Sebor <msebor@redhat.com>
5411 PR middle-end/101984
5412 * gimple-ssa-warn-access.cc (pass_waccess::execute): Also call
5415 2021-08-19 Jeff Law <jlaw@localhost.localdomain>
5417 * config.gcc (h8300-*-elf*): Do not include dbxelf.h.
5418 (h8300-*-linux*, v850-*-rtems*, v850*-elf*): Likewise.
5419 * config/v850/v850.h (DEFAULT_GDB_EXTENSIONS): Remove.
5421 2021-08-19 Jakub Jelinek <jakub@redhat.com>
5423 PR middle-end/101950
5424 * optabs.c (expand_clrsb_using_clz): New function.
5425 (expand_unop): Use it as another clrsb expansion fallback.
5427 2021-08-19 liuhongt <hongtao.liu@intel.com>
5430 2021-07-28 liuhongt <hongtao.liu@intel.com>
5433 * config/i386/i386.h (processor_costs): Add new member
5435 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
5436 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
5437 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
5438 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
5439 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
5440 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
5441 generic_cost, core_cost): Initialize integer_to_sse same value
5443 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
5444 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5445 Use integer_to_sse instead of sse_op to calculate the cost of
5448 2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
5450 * config.gcc: Include rpath.opt for Darwin.
5451 * config/darwin.h (DRIVER_SELF_SPECS): Handle -rpath.
5453 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
5456 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
5459 2021-08-18 Jonathan Wright <jonathan.wright@arm.com>
5461 * config/aarch64/arm_neon.h (vld3_lane_f64): Use float RTL
5462 pattern and type cast.
5463 (vld4_lane_f32): Use float RTL pattern.
5464 (vld4q_lane_f64): Use float type cast.
5466 2021-08-18 Jan Hubicka <hubicka@ucw.cz>
5468 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Check also
5471 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
5473 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor): Extend.
5474 (test_map_of_type_with_ctor_and_dtor_expand): Add function.
5475 (hash_map_tests_c_tests): Call it.
5477 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
5479 * ggc.h (enum ggc_collect): New.
5480 (ggc_collect): Use it.
5481 * ggc-page.c: Adjust.
5482 * ggc-common.c: Likewise.
5483 * ggc-tests.c: Likewise.
5484 * read-rtl-function.c: Likewise.
5485 * selftest-run-tests.c: Likewise.
5486 * doc/gty.texi (Invoking the garbage collector): Likewise.
5488 2021-08-18 liuhongt <hongtao.liu@intel.com>
5491 * config/i386/i386.h (TARGET_V2DF_REDUCTION_PREFER_HADDPD):
5493 * config/i386/sse.md (*sse3_haddv2df3_low): Add
5494 TARGET_V2DF_REDUCTION_PREFER_HADDPD.
5495 (*sse3_hsubv2df3_low): Ditto.
5496 * config/i386/x86-tune.def
5497 (X86_TUNE_V2DF_REDUCTION_PREFER_HADDPD): New tune.
5499 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
5501 * gimple-range-gori.cc (gori_compute::gori_compute): Enable tracing.
5502 (gori_compute::compute_operand_range): Add tracing.
5503 (gori_compute::logical_combine): Ditto.
5504 (gori_compute::compute_logical_operands): Ditto.
5505 (gori_compute::compute_operand1_range): Ditto.
5506 (gori_compute::compute_operand2_range): Ditto.
5507 (gori_compute::outgoing_edge_range_p): Ditto.
5508 * gimple-range-gori.h (class gori_compute): Add range_tracer.
5510 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
5512 * flag-types.h (enum evrp_mode): Adjust evrp-mode values.
5513 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Relocate from.
5514 * gimple-range-trace.h (DEBUG_RANGE_CACHE): Here.
5515 * params.opt (--param=evrp-mode): Adjust options.
5517 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
5519 * Makefile.in (OBJS): Add gimple-range-trace.o.
5520 * gimple-range-cache.h (enable_new_values): Remove unused prototype.
5521 * gimple-range-fold.cc: Adjust headers.
5522 * gimple-range-trace.cc: New.
5523 * gimple-range-trace.h: New.
5524 * gimple-range.cc (gimple_ranger::gimple_ranger): Enable tracer.
5525 (gimple_ranger::range_of_expr): Add tracing.
5526 (gimple_ranger::range_on_entry): Ditto.
5527 (gimple_ranger::range_on_exit): Ditto.
5528 (gimple_ranger::range_on_edge): Ditto.
5529 (gimple_ranger::fold_range_internal): Ditto.
5530 (gimple_ranger::dump_bb): Do not calculate edge range twice.
5531 (trace_ranger::*): Remove.
5532 (enable_ranger): Never create a trace_ranger.
5533 (debug_seed_ranger): Move to gimple-range-trace.cc.
5534 (dump_ranger): Ditto.
5535 (debug_ranger): Ditto.
5536 * gimple-range.h: Include gimple-range-trace.h.
5537 (range_on_entry, range_on_exit): No longer virtual.
5538 (class trace_ranger): Remove.
5539 (DEBUG_RANGE_CACHE): Move to gimple-range-trace.h.
5541 2021-08-17 Martin Sebor <msebor@redhat.com>
5543 PR middle-end/101854
5544 * builtins.c (expand_builtin_alloca): Move warning code to check_alloca
5545 in gimple-ssa-warn-access.cc.
5546 * calls.c (alloc_max_size): Move code to check_alloca.
5547 (get_size_range): Move to pointer-query.cc.
5548 (maybe_warn_alloc_args_overflow): Move to gimple-ssa-warn-access.cc.
5549 (get_attr_nonstring_decl): Move to tree.c.
5550 (fntype_argno_type): Move to gimple-ssa-warn-access.cc.
5551 (append_attrname): Same.
5552 (maybe_warn_rdwr_sizes): Same.
5553 (initialize_argument_information): Move code to
5554 gimple-ssa-warn-access.cc.
5555 * calls.h (maybe_warn_alloc_args_overflow): Move to
5556 gimple-ssa-warn-access.h.
5557 (get_attr_nonstring_decl): Move to tree.h.
5558 (maybe_warn_nonstring_arg): Move to gimple-ssa-warn-access.h.
5559 (enum size_range_flags): Move to pointer-query.h.
5560 (get_size_range): Same.
5561 * gimple-ssa-warn-access.cc (has_location): Remove unused overload
5562 to avoid Clang -Wunused-function.
5563 (get_size_range): Declare static.
5564 (maybe_emit_free_warning): Rename...
5565 (maybe_check_dealloc_call): ...to this for consistency.
5566 (class pass_waccess): Add members.
5567 (pass_waccess::~pass_waccess): Defined.
5568 (alloc_max_size): Move here from calls.c.
5569 (maybe_warn_alloc_args_overflow): Same.
5570 (check_alloca): New function.
5571 (check_alloc_size_call): New function.
5572 (check_strncat): Handle another warning flag.
5573 (pass_waccess::check_builtin): Handle alloca.
5574 (fntype_argno_type): Move here from calls.c.
5575 (append_attrname): Same.
5576 (maybe_warn_rdwr_sizes): Same.
5577 (pass_waccess::check_call): Define.
5578 (check_nonstring_args): New function.
5579 (pass_waccess::check): Call new member functions.
5580 (pass_waccess::execute): Enable ranger.
5581 * gimple-ssa-warn-access.h (get_size_range): Move here from calls.h.
5582 (maybe_warn_nonstring_arg): Same.
5583 * gimple-ssa-warn-restrict.c: Remove #include.
5584 * pointer-query.cc (get_size_range): Move here from calls.c.
5585 * pointer-query.h (enum size_range_flags): Same.
5586 (get_size_range): Same.
5587 * tree.c (get_attr_nonstring_decl): Move here from calls.c.
5588 * tree.h (get_attr_nonstring_decl): Move here from calls.h.
5590 2021-08-17 Thomas Schwinge <thomas@codesourcery.com>
5592 * ggc.h (ggc_collect): Add 'force_collect' parameter.
5593 * ggc-page.c (ggc_collect): Use that one instead of global
5594 'ggc_force_collect'. Adjust all users.
5595 * doc/gty.texi (Invoking the garbage collector): Update.
5596 * ggc-internal.h (ggc_force_collect): Remove.
5597 * ggc-common.c (ggc_force_collect): Likewise.
5598 * selftest.h (forcibly_ggc_collect): Remove.
5599 * ggc-tests.c (selftest::forcibly_ggc_collect): Likewise.
5600 * read-rtl-function.c (test_loading_labels): Adjust.
5601 * selftest-run-tests.c (run_tests): Likewise.
5603 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
5605 * config/darwin.c (darwin_file_end): Reset and reclaim the
5606 section names table at the end of compile.
5608 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
5611 * config.in: Regenerate.
5612 * config/i386/darwin.h (EXTRA_ASM_OPTS): New
5613 (ASM_SPEC): Pass options to disable branch shortening where
5615 * configure: Regenerate.
5616 * configure.ac: Detect versions of 'as' that support the
5617 optimisation which has the bug.
5619 2021-08-17 Richard Biener <rguenther@suse.de>
5621 * optabs-query.c (supports_vec_gather_load_p): Also check
5623 (supports_vec_scatter_store_p): Likewise.
5624 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Fall
5625 back to masked variants if non-masked are not supported.
5626 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
5627 When we need to use masked gather/scatter but do not have
5628 a mask set up a constant true one.
5629 * tree-vect-stmts.c (vect_check_scalar_mask): Also allow
5632 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
5634 * tree-ssa-ccp.c (bit_value_binop) [MINUS_EXPR]: Use same
5635 algorithm as PLUS_EXPR to improve subtraction bit bounds.
5636 [POINTER_DIFF_EXPR]: Treat as synonymous with MINUS_EXPR.
5638 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
5640 * tree-ssa-ccp.c (bit_value_mult_const): New helper function to
5641 calculate the mask-value pair result of a multiplication by an
5643 (bit_value_binop) [MULT_EXPR]: Call it from here for
5644 multiplications by (sparse) non-negative constants.
5646 2021-08-17 Christophe Lyon <christophe.lyon@foss.st.com>
5649 * config.gcc (gcc_cv_initfini_array): Leave undefined for
5650 uclinuxfdpiceabi targets.
5652 2021-08-17 Alexandre Oliva <oliva@adacore.com>
5654 * tree-inline.c (maybe_move_debug_stmts_to_successors): Don't
5655 reverse debug stmts.
5657 2021-08-17 Alexandre Oliva <oliva@adacore.com>
5659 * tree-cfg.c (dump_function_to_file): Use fun, not cfun.
5661 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
5663 * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete.
5664 (__LD4Q_LANE_FUNC): Likewise.
5665 (vld4_lane_u8): Define without macro.
5666 (vld4_lane_u16): Likewise.
5667 (vld4_lane_u32): Likewise.
5668 (vld4_lane_u64): Likewise.
5669 (vld4_lane_s8): Likewise.
5670 (vld4_lane_s16): Likewise.
5671 (vld4_lane_s32): Likewise.
5672 (vld4_lane_s64): Likewise.
5673 (vld4_lane_f16): Likewise.
5674 (vld4_lane_f32): Likewise.
5675 (vld4_lane_f64): Likewise.
5676 (vld4_lane_p8): Likewise.
5677 (vld4_lane_p16): Likewise.
5678 (vld4_lane_p64): Likewise.
5679 (vld4q_lane_u8): Likewise.
5680 (vld4q_lane_u16): Likewise.
5681 (vld4q_lane_u32): Likewise.
5682 (vld4q_lane_u64): Likewise.
5683 (vld4q_lane_s8): Likewise.
5684 (vld4q_lane_s16): Likewise.
5685 (vld4q_lane_s32): Likewise.
5686 (vld4q_lane_s64): Likewise.
5687 (vld4q_lane_f16): Likewise.
5688 (vld4q_lane_f32): Likewise.
5689 (vld4q_lane_f64): Likewise.
5690 (vld4q_lane_p8): Likewise.
5691 (vld4q_lane_p16): Likewise.
5692 (vld4q_lane_p64): Likewise.
5693 (vld4_lane_bf16): Likewise.
5694 (vld4q_lane_bf16): Likewise.
5696 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
5698 * config/aarch64/arm_neon.h (__LD3_LANE_FUNC): Delete.
5699 (__LD3Q_LANE_FUNC): Delete.
5700 (vld3_lane_u8): Define without macro.
5701 (vld3_lane_u16): Likewise.
5702 (vld3_lane_u32): Likewise.
5703 (vld3_lane_u64): Likewise.
5704 (vld3_lane_s8): Likewise.
5705 (vld3_lane_s16): Likewise.
5706 (vld3_lane_s32): Likewise.
5707 (vld3_lane_s64): Likewise.
5708 (vld3_lane_f16): Likewise.
5709 (vld3_lane_f32): Likewise.
5710 (vld3_lane_f64): Likewise.
5711 (vld3_lane_p8): Likewise.
5712 (vld3_lane_p16): Likewise.
5713 (vld3_lane_p64): Likewise.
5714 (vld3q_lane_u8): Likewise.
5715 (vld3q_lane_u16): Likewise.
5716 (vld3q_lane_u32): Likewise.
5717 (vld3q_lane_u64): Likewise.
5718 (vld3q_lane_s8): Likewise.
5719 (vld3q_lane_s16): Likewise.
5720 (vld3q_lane_s32): Likewise.
5721 (vld3q_lane_s64): Likewise.
5722 (vld3q_lane_f16): Likewise.
5723 (vld3q_lane_f32): Likewise.
5724 (vld3q_lane_f64): Likewise.
5725 (vld3q_lane_p8): Likewise.
5726 (vld3q_lane_p16): Likewise.
5727 (vld3q_lane_p64): Likewise.
5728 (vld3_lane_bf16): Likewise.
5729 (vld3q_lane_bf16): Likewise.
5731 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
5733 * config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Delete.
5734 (__LD2Q_LANE_FUNC): Likewise.
5735 (vld2_lane_u8): Define without macro.
5736 (vld2_lane_u16): Likewise.
5737 (vld2_lane_u32): Likewise.
5738 (vld2_lane_u64): Likewise.
5739 (vld2_lane_s8): Likewise.
5740 (vld2_lane_s16): Likewise.
5741 (vld2_lane_s32): Likewise.
5742 (vld2_lane_s64): Likewise.
5743 (vld2_lane_f16): Likewise.
5744 (vld2_lane_f32): Likewise.
5745 (vld2_lane_f64): Likewise.
5746 (vld2_lane_p8): Likewise.
5747 (vld2_lane_p16): Likewise.
5748 (vld2_lane_p64): Likewise.
5749 (vld2q_lane_u8): Likewise.
5750 (vld2q_lane_u16): Likewise.
5751 (vld2q_lane_u32): Likewise.
5752 (vld2q_lane_u64): Likewise.
5753 (vld2q_lane_s8): Likewise.
5754 (vld2q_lane_s16): Likewise.
5755 (vld2q_lane_s32): Likewise.
5756 (vld2q_lane_s64): Likewise.
5757 (vld2q_lane_f16): Likewise.
5758 (vld2q_lane_f32): Likewise.
5759 (vld2q_lane_f64): Likewise.
5760 (vld2q_lane_p8): Likewise.
5761 (vld2q_lane_p16): Likewise.
5762 (vld2q_lane_p64): Likewise.
5763 (vld2_lane_bf16): Likewise.
5764 (vld2q_lane_bf16): Likewise.
5766 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
5768 * haifa-sched.c (advance_one_cycle): Output more context-synchronization
5771 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
5773 * haifa-sched.c (enum rfs_decision, rfs_str): Add RFS_AUTOPREF.
5774 (rank_for_schedule): Use it.
5776 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
5778 PR rtl-optimization/91598
5779 * haifa-sched.c (autopref_rank_for_schedule): Prioritize "irrelevant"
5780 insns after memory reads and before memory writes.
5782 2021-08-17 Alistair_Lee <alistair.lee@arm.com>
5784 * rtl.h (CONST_VECTOR_P): New macro.
5785 * config/aarch64/aarch64.c (aarch64_get_sve_pred_bits): Use RTL
5786 code testing macros.
5787 (aarch64_ptrue_all_mode): Likewise.
5788 (aarch64_expand_mov_immediate): Likewise.
5789 (aarch64_const_vec_all_in_range_p): Likewise.
5790 (aarch64_rtx_costs): Likewise.
5791 (aarch64_legitimate_constant_p): Likewise.
5792 (aarch64_simd_valid_immediate): Likewise.
5793 (aarch64_simd_make_constant): Likewise.
5794 (aarch64_convert_mult_to_shift): Likewise.
5795 (aarch64_expand_sve_vec_perm): Likewise.
5796 (aarch64_vec_fpconst_pow_of_2): Likewise.
5798 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
5800 PR tree-optimization/101938
5801 * range-op.cc (operator_abs::op1_range): Special case
5802 -TYPE_MIN_VALUE for flag_wrapv.
5804 2021-08-17 Kewen Lin <linkw@linux.ibm.com>
5806 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Add the cost for
5809 2021-08-17 Jakub Jelinek <jakub@redhat.com>
5811 * tree.def (OMP_SCOPE): New tree code.
5812 * tree.h (OMP_SCOPE_BODY, OMP_SCOPE_CLAUSES): Define.
5813 * tree-nested.c (convert_nonlocal_reference_stmt,
5814 convert_local_reference_stmt, convert_gimple_call): Handle
5816 * tree-pretty-print.c (dump_generic_node): Handle OMP_SCOPE.
5817 * gimple.def (GIMPLE_OMP_SCOPE): New gimple code.
5818 * gimple.c (gimple_build_omp_scope): New function.
5819 (gimple_copy): Handle GIMPLE_OMP_SCOPE.
5820 * gimple.h (gimple_build_omp_scope): Declare.
5821 (gimple_has_substatements): Handle GIMPLE_OMP_SCOPE.
5822 (gimple_omp_scope_clauses, gimple_omp_scope_clauses_ptr,
5823 gimple_omp_scope_set_clauses): New inline functions.
5824 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_SCOPE.
5825 * gimple-pretty-print.c (dump_gimple_omp_scope): New function.
5826 (pp_gimple_stmt_1): Handle GIMPLE_OMP_SCOPE.
5827 * gimple-walk.c (walk_gimple_stmt): Likewise.
5828 * gimple-low.c (lower_stmt): Likewise.
5829 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
5830 (gimplify_scan_omp_clauses): For task reductions, handle OMP_SCOPE
5831 like ORT_WORKSHARE constructs. Adjust diagnostics for %<scope%>
5832 allowing task reductions. Reject inscan reductions on scope.
5833 (omp_find_stores_stmt): Handle GIMPLE_OMP_SCOPE.
5834 (gimplify_omp_workshare, gimplify_expr): Handle OMP_SCOPE.
5835 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_SCOPE.
5836 (estimate_num_insns): Likewise.
5837 * omp-low.c (build_outer_var_ref): Look through GIMPLE_OMP_SCOPE
5838 contexts if var isn't privatized there.
5839 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_SCOPE.
5840 (scan_omp_1_stmt): Likewise.
5841 (maybe_add_implicit_barrier_cancel): Look through outer
5843 (lower_omp_scope): New function.
5844 (lower_omp_task_reductions): Handle OMP_SCOPE.
5845 (lower_omp_1): Handle GIMPLE_OMP_SCOPE.
5846 (diagnose_sb_1, diagnose_sb_2): Likewise.
5847 * omp-expand.c (expand_omp_single): Support also GIMPLE_OMP_SCOPE.
5848 (expand_omp): Handle GIMPLE_OMP_SCOPE.
5849 (omp_make_gimple_edges): Likewise.
5850 * omp-builtins.def (BUILT_IN_GOMP_SCOPE_START): New built-in.
5852 2021-08-17 Richard Biener <rguenther@suse.de>
5854 PR tree-optimization/101925
5855 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Set
5856 reverse on COMPONENT_REF and ARRAY_REF according to
5857 what reverse_storage_order_for_component_p does.
5858 (vn_reference_eq): Compare reversed on reference ops.
5859 (reverse_storage_order_for_component_p): New overload.
5860 (vn_reference_lookup_3): Check reverse_storage_order_for_component_p
5861 on the reference looked up.
5863 2021-08-17 Jeff Law <jlaw@localhost.localdomain>
5865 * config/h8300/h8300.c (shift_alg_si): Avoid loops for most SImode
5867 (h8300_option_override): Use loops on H8/S more often when optimizing
5869 (get_shift_alg): Handle new "special" cases on H8/S. Simplify
5870 accordingly. Handle various arithmetic right shifts with special
5871 sequences that we couldn't handle before.
5873 2021-08-16 Jeff Law <jlaw@localhost.localdomain>
5875 * config.gcc (rl78-*-elf*): Do not include dbxelf.h.
5877 2021-08-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
5879 * config/sparc/rtemself.h (SPARC_GCOV_TYPE_SIZE): Define.
5880 * config/sparc/sparc.c (sparc_gcov_type_size): New.
5881 (TARGET_GCOV_TYPE_SIZE): Redefine if SPARC_GCOV_TYPE_SIZE is defined.
5882 * coverage.c (get_gcov_type): Use targetm.gcov_type_size().
5883 * doc/tm.texi (TARGET_GCOV_TYPE_SIZE): Add hook under "Misc".
5884 * doc/tm.texi.in: Regenerate.
5885 * target.def (gcov_type_size): New target hook.
5886 * targhooks.c (default_gcov_type_size): New.
5887 * targhooks.h (default_gcov_type_size): Declare.
5888 * tree-profile.c (gimple_gen_edge_profiler): Use precision of
5890 (gimple_gen_time_profiler): Likewise.
5892 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
5894 * dwarf2out.c (add_scalar_info): Deal with DW_AT_data_bit_offset.
5896 2021-08-16 Tobias Burnus <tobias@codesourcery.com>
5898 PR middle-end/101931
5899 * omp-low.c (omp_runtime_api_call): Update for routines
5900 added in the meanwhile.
5902 2021-08-16 Martin Liska <mliska@suse.cz>
5904 PR tree-optimization/100393
5905 * tree-switch-conversion.c (group_cluster::dump): Use
5906 get_comparison_count.
5907 (jump_table_cluster::find_jump_tables): Pre-compute number of
5908 comparisons and then decrement it. Cache also max_ratio.
5909 (jump_table_cluster::can_be_handled): Change signature.
5910 * tree-switch-conversion.h (get_comparison_count): New.
5912 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
5914 * dwarf2out.c (add_data_member_location_attribute): Use GNAT
5915 encodings only when -fgnat-encodings=all is specified.
5916 (add_bound_info): Likewise.
5917 (add_byte_size_attribute): Likewise.
5918 (gen_member_die): Likewise.
5920 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
5922 * omp-oacc-neuter-broadcast.cc
5923 (execute_omp_oacc_neuter_broadcast): Plug 'par' memory leak.
5925 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
5927 * omp-oacc-neuter-broadcast.cc
5928 (execute_omp_oacc_neuter_broadcast): Clarify memory management for
5931 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
5933 * omp-oacc-neuter-broadcast.cc (field_map): Move variable into...
5934 (execute_omp_oacc_neuter_broadcast): ... here.
5935 (install_var_field, build_receiver_ref, build_sender_ref): Take
5936 'field_map_t *' parameter. Adjust all users.
5937 (worker_single_copy, neuter_worker_single): Take a
5938 'record_field_map_t *' parameter. Adjust all users.
5940 2021-08-16 liuhongt <hongtao.liu@intel.com>
5943 * config/i386/i386.md (ldexp<mode>3): Force operands[1] to
5946 2021-08-16 Martin Liska <mliska@suse.cz>
5949 * multiple_target.c (create_dispatcher_calls): Make default
5950 function local only if it is a definition.
5952 2021-08-16 Martin Liska <mliska@suse.cz>
5955 * ipa-icf-gimple.c (func_checker::compare_ssa_name): Do not
5956 consider equal SSA_NAMEs when one is a param.
5958 2021-08-16 liuhongt <hongtao.liu@intel.com>
5961 * config/i386/i386-expand.c (ix86_expand_vec_perm_vpermt2):
5962 Support vpermi2b for V32QI/V16QImode.
5963 (ix86_extract_perm_from_pool_constant): New function.
5964 (ix86_expand_vec_one_operand_perm_avx512): Support
5965 vpermw/vpermb under TARGET_AVX512BW/TARGET_AVX512VBMI.
5966 (expand_vec_perm_1): Adjust comments for upper.
5967 * config/i386/i386-protos.h (ix86_extract_perm_from_pool_constant):
5969 * config/i386/predicates.md (permvar_truncate_operand): New predicate.
5970 (pshufb_truncv4siv4hi_operand): Ditto.
5971 (pshufb_truncv8hiv8qi_operand): Ditto.
5972 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1):
5973 New pre_reload define_insn_and_split.
5974 (*avx512f_permvar_truncv8siv8hi_1): Ditto.
5975 (*avx512f_vpermvar_truncv8div8si_1): Ditto.
5976 (*avx512f_permvar_truncv32hiv32qi_1): Ditto.
5977 (*avx512f_permvar_truncv16hiv16qi_1): Ditto.
5978 (*avx512f_permvar_truncv4div4si_1): Ditto.
5979 (*avx512f_pshufb_truncv8hiv8qi_1): Ditto.
5980 (*avx512f_pshufb_truncv4siv4hi_1): Ditto.
5981 (*avx512f_pshufd_truncv2div2si_1): Ditto.
5983 2021-08-16 Kito Cheng <kito.cheng@sifive.com>
5985 * config/riscv/multilib-generator: Support code model option for
5987 * doc/install.texi: Add document of new option for
5988 --with-multilib-generator.
5990 2021-08-15 Clément Chigot <clement.chigot@atos.net>
5992 * config/rs6000/rs6000.c (xcoff_tls_exec_model_detected): New.
5993 (rs6000_legitimize_tls_address_aix): Use it.
5994 (rs6000_xcoff_file_end): Add ".ref __tls_get_addr" when
5995 xcoff_tls_exec_model_detected is true.
5997 2021-08-15 Jeff Law <jlaw@localhost.localdomain>
5999 * config/h8300/h8300.c (shift_alg_si): Retune H8/300H shifts
6000 to allow a bit more code growth, saving many dozens of cycles.
6001 (h8300_option_override): Adjus shift_alg_si if optimizing for
6003 (get_shift_alg): Use special + inline shifts for residuals
6006 2021-08-14 Stafford Horne <shorne@gmail.com>
6009 * config/or1k/or1k-opts.h: New file.
6010 * config/or1k/or1k.c (or1k_legitimize_address_1, print_reloc):
6011 Support generating gotha relocations if -mcmodel=large is
6013 * config/or1k/or1k.h (TARGET_CMODEL_SMALL, TARGET_CMODEL_LARGE):
6015 * config/or1k/or1k.opt (mcmodel=): New option.
6016 * doc/invoke.texi (OpenRISC Options): Document mcmodel.
6018 2021-08-14 Martin Sebor <msebor@redhat.com>
6020 PR middle-end/101791
6021 * gimple-ssa-warn-access.cc (new_delete_mismatch_p): Use new argument
6022 to valid_new_delete_pair_p.
6023 * tree.c (valid_new_delete_pair_p): Add argument.
6024 * tree.h (valid_new_delete_pair_p): Same.
6026 2021-08-14 Jakub Jelinek <jakub@redhat.com>
6029 * config/i386/i386-expand.c (expand_vec_perm_broadcast_1)
6030 <case E_V64QImode>: For this mode assert
6031 !TARGET_AVX512BW || d->perm[0] rather than !TARGET_AVX2 || d->perm[0].
6033 2021-08-13 Michael Meissner <meissner@linux.ibm.com>
6036 * config/rs6000/altivec.md (xxeval): Use register_predicate
6037 instead of altivec_register_predicate.
6039 2021-08-13 Martin Sebor <msebor@redhat.com>
6041 PR middle-end/101734
6042 * tree-ssa-uninit.c (maybe_warn_read_write_only): New function.
6043 (maybe_warn_operand): Call it.
6045 2021-08-13 Martin Liska <mliska@suse.cz>
6048 * attribs.c (decl_attributes): Make naked functions "noipa"
6051 2021-08-13 Martin Liska <mliska@suse.cz>
6054 * symtab.c (symtab_node::noninterposable_alias): Do not create
6055 local aliases for target_clone functions as the clonning pass
6058 2021-08-13 Martin Liska <mliska@suse.cz>
6060 * opts.c (LIVE_PATCHING_OPTION): Define.
6061 (control_options_for_live_patching): Use it in error messages.
6063 2021-08-13 Jan Hubicka <hubicka@ucw.cz>
6065 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOREAD.
6066 (implicit_const_eaf_flags, implicit_pure_eaf_flags,
6067 ignore_stores_eaf_flags): New constants.
6068 (remove_useless_eaf_flags): New function.
6069 (eaf_flags_useful_p): Use it.
6070 (deref_flags): Add EAF_NOT_RETURNED if flag is unused;
6072 (modref_lattice::init): Add EAF_NOREAD.
6073 (modref_lattice::add_escape_point): Do not reacord escape point if
6075 (modref_lattice::merge): EAF_NOESCAPE implies EAF_NODIRECTESCAPE;
6076 use remove_useless_eaf_flags.
6077 (modref_lattice::merge_deref): Use ignore_stores_eaf_flags.
6078 (modref_lattice::merge_direct_load): Add EAF_NOREAD
6079 (analyze_ssa_name_flags): Fix handling EAF_NOT_RETURNED
6080 (analyze_parms): Use remove_useless_eaf_flags.
6081 (ipa_merge_modref_summary_after_inlining): Use ignore_stores_eaf_flags.
6082 (modref_merge_call_site_flags): Add caller and ecf_flags parameter;
6083 use remove_useless_eaf_flags.
6084 (modref_propagate_flags_in_scc): Update.
6085 * ipa-modref.h: Turn eaf_flags_t back to char.
6086 * tree-core.h (EAF_NOT_RETURNED): Fix.
6087 (EAF_NOREAD): New constant
6088 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1): Check for
6090 * tree-ssa-structalias.c (handle_rhs_call): Handle new flags.
6091 (handle_pure_call): Likewise.
6093 2021-08-12 Jakub Jelinek <jakub@redhat.com>
6095 * tree.def (OMP_MASKED): New tree code.
6096 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_FILTER.
6097 * tree.h (OMP_MASKED_BODY, OMP_MASKED_CLAUSES, OMP_MASKED_COMBINED,
6098 OMP_CLAUSE_FILTER_EXPR): Define.
6099 * tree.c (omp_clause_num_ops): Add OMP_CLAUSE_FILTER entry.
6100 (omp_clause_code_name): Likewise.
6101 (walk_tree_1): Handle OMP_CLAUSE_FILTER.
6102 * tree-nested.c (convert_nonlocal_omp_clauses,
6103 convert_local_omp_clauses): Handle OMP_CLAUSE_FILTER.
6104 (convert_nonlocal_reference_stmt, convert_local_reference_stmt,
6105 convert_gimple_call): Handle GIMPLE_OMP_MASTER.
6106 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_FILTER.
6107 (dump_generic_node): Handle OMP_MASTER.
6108 * gimple.def (GIMPLE_OMP_MASKED): New gimple code.
6109 * gimple.c (gimple_build_omp_masked): New function.
6110 (gimple_copy): Handle GIMPLE_OMP_MASKED.
6111 * gimple.h (gimple_build_omp_masked): Declare.
6112 (gimple_has_substatements): Handle GIMPLE_OMP_MASKED.
6113 (gimple_omp_masked_clauses, gimple_omp_masked_clauses_ptr,
6114 gimple_omp_masked_set_clauses): New inline functions.
6115 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_MASKED.
6116 * gimple-pretty-print.c (dump_gimple_omp_masked): New function.
6117 (pp_gimple_stmt_1): Handle GIMPLE_OMP_MASKED.
6118 * gimple-walk.c (walk_gimple_stmt): Likewise.
6119 * gimple-low.c (lower_stmt): Likewise.
6120 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
6121 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_FILTER. For clauses
6122 that take one expression rather than decl or constant, force
6123 gimplification of that into a SSA_NAME or temporary unless min
6125 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_FILTER.
6126 (gimplify_expr): Handle OMP_MASKED.
6127 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_MASKED.
6128 (estimate_num_insns): Likewise.
6129 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_FILTER.
6130 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_MASKED. Adjust
6131 diagnostics for existence of masked construct.
6132 (scan_omp_1_stmt, lower_omp_master, lower_omp_1, diagnose_sb_1,
6133 diagnose_sb_2): Handle GIMPLE_OMP_MASKED.
6134 * omp-expand.c (expand_omp_synch, expand_omp, omp_make_gimple_edges):
6137 2021-08-12 Uroš Bizjak <ubizjak@gmail.com>
6140 * config/i386/i386.md (avx512f_scalef<mode>2): New insn pattern.
6141 (ldexp<mode>3): Use avx512f_scalef<mode>2.
6142 (UNSPEC_SCALEF): Move from ...
6143 * config/i386/sse.md (UNSPEC_SCALEF): ... here.
6145 2021-08-12 Jan Hubicka <hubicka@ucw.cz>
6147 * ipa-split.c (consider_split): Fix condition testing void functions.
6149 2021-08-12 Aldy Hernandez <aldyh@redhat.com>
6151 * doc/invoke.texi: Remove docs for threader-mode param.
6152 * flag-types.h (enum threader_mode): Remove.
6153 * params.opt: Remove threader-mode param.
6154 * tree-ssa-threadbackward.c (class back_threader): Remove
6155 path_is_unreachable_p.
6156 Make find_paths private.
6157 Add maybe_thread and thread_through_all_blocks.
6158 Remove reference marker for m_registry.
6159 Remove reference marker for m_profit.
6160 (back_threader::back_threader): Adjust for registry and profit not
6162 (dump_path): Move down.
6164 (class thread_jumps): Remove.
6165 (class back_threader_registry): Remove m_all_paths.
6167 (thread_jumps::thread_through_all_blocks): Move to back_threader
6169 (fsm_find_thread_path): Remove
6170 (back_threader::maybe_thread): New.
6171 (back_threader::thread_through_all_blocks): Move from
6173 (back_threader_registry::back_threader_registry): Remove
6175 (back_threader_registry::~back_threader_registry): Remove.
6176 (thread_jumps::find_taken_edge): Remove.
6177 (thread_jumps::check_subpath_and_update_thread_path): Remove.
6178 (thread_jumps::maybe_register_path): Remove.
6179 (thread_jumps::handle_phi): Remove.
6180 (handle_assignment_p): Remove.
6181 (thread_jumps::handle_assignment): Remove.
6182 (thread_jumps::fsm_find_control_statement_thread_paths): Remove.
6183 (thread_jumps::find_jump_threads_backwards): Remove.
6184 (thread_jumps::find_jump_threads_backwards_with_ranger): Remove.
6185 (try_thread_blocks): Rename find_jump_threads_backwards to
6187 (pass_early_thread_jumps::execute): Same.
6189 2021-08-12 Tobias Burnus <tobias@codesourcery.com>
6191 * tree-core.h (omp_clause_proc_bind_kind): Add
6192 OMP_CLAUSE_PROC_BIND_PRIMARY.
6193 * tree-pretty-print.c (dump_omp_clause): Add TODO comment to
6194 change 'master' to 'primary' in proc_bind for OpenMP 5.1.
6196 2021-08-12 Claudiu Zissulescu <claziss@synopsys.com>
6198 * common/config/arc/arc-common.c (arc_option_init_struct): Remove
6199 fno-common reference.
6200 * config/arc/arc.c (arc_override_options): Remove overriding of
6203 2021-08-12 Jakub Jelinek <jakub@redhat.com>
6206 * config/i386/i386-expand.c (ix86_expand_vec_one_operand_perm_avx512):
6207 If d->testing_p, return true after performing checks instead of
6208 actually expanding the insn.
6209 (expand_vec_perm_broadcast_1): Handle V32HImode - assert
6210 !TARGET_AVX512BW and return false.
6212 2021-08-12 Eric Botcazou <ebotcazou@gcc.gnu.org>
6214 * configure.ac (PE linker --disable-dynamicbase support): New check.
6215 * configure: Regenerate.
6216 * config.in: Likewise.
6217 * config/i386/mingw32.h (LINK_SPEC_DISABLE_DYNAMICBASE): New define.
6218 (LINK_SPEC): Use it.
6219 * config/i386/mingw-w64.h (LINK_SPEC_DISABLE_DYNAMICBASE): Likewise.
6220 (LINK_SPEC): Likewise.
6222 2021-08-12 liuhongt <hongtao.liu@intel.com>
6225 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_2): New
6226 post_reload define_insn_and_split.
6227 (*avx512bw_zero_extendv32qiv32hi2_2): Ditto.
6228 (*sse4_1_zero_extendv8qiv8hi2_4): Ditto.
6229 (*avx512f_zero_extendv16hiv16si2_2): Ditto.
6230 (*avx2_zero_extendv8hiv8si2_2): Ditto.
6231 (*sse4_1_zero_extendv4hiv4si2_4): Ditto.
6232 (*avx512f_zero_extendv8siv8di2_2): Ditto.
6233 (*avx2_zero_extendv4siv4di2_2): Ditto.
6234 (*sse4_1_zero_extendv2siv2di2_4): Ditto.
6235 (VI248_256, VI248_512, VI148_512, VI148_256, VI148_128): New
6238 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
6240 * config/rs6000/rs6000-builtin-new.def: Add always, power5, and
6243 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
6245 * config/rs6000/rs6000-builtin-new.def: Add vsx stanza.
6247 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
6249 * config/rs6000/rs6000-builtin-new.def: Finish altivec stanza.
6250 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
6251 initialization of pcvoid_type_node here...
6252 (altivec_init_builtins): ...from here.
6253 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add
6254 RS6000_BTI_const_ptr_void.
6255 (pcvoid_type_node): New macro.
6257 2021-08-11 Richard Biener <rguenther@suse.de>
6260 * tree-ssa-forwprop.c (pass_forwprop::execute): Do not decompose
6261 hard-register accesses.
6263 2021-08-11 Richard Biener <rguenther@suse.de>
6265 * tree-ssa-operands.c (operands_scanner::get_expr_operands):
6266 Do not look at COMPONENT_REF FIELD_DECLs TREE_THIS_VOLATILE
6267 to determine has_volatile_ops.
6269 2021-08-11 Eric Botcazou <ebotcazou@gcc.gnu.org>
6271 * cfgexpand.c (expand_used_vars): Reuse attribs local variable.
6273 2021-08-11 Jan Hubicka <hubicka@ucw.cz>
6274 Alexandre Oliva <oliva@adacore.com>
6276 * ipa-modref.c (modref_lattice::dump): Fix escape_point's min_flags
6278 (modref_lattice::merge_deref): Fix handling of indirect scape points.
6279 (update_escape_summary_1): Likewise.
6280 (update_escape_summary): Likewise.
6281 (ipa_merge_modref_summary_after_inlining): Likewise.
6283 2021-08-11 Richard Biener <rguenther@suse.de>
6285 PR middle-end/101858
6286 * fold-const.c (fold_binary_loc): Guard simplification
6287 of X < (cast) (1 << Y) to integer types.
6289 2021-08-11 Richard Biener <rguenther@suse.de>
6291 PR tree-optimization/101861
6292 * tree-vect-stmts.c (vectorizable_load): Fix error in
6293 previous change with regard to gather vectorization.
6295 2021-08-11 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
6298 * config/arm/arm_neon.h (vdup_n_s8): Replace call to builtin
6300 (vdup_n_s16): Likewise.
6301 (vdup_n_s32): Likewise.
6302 (vdup_n_s64): Likewise.
6303 (vdup_n_u8): Likewise.
6304 (vdup_n_u16): Likewise.
6305 (vdup_n_u32): Likewise.
6306 (vdup_n_u64): Likewise.
6307 (vdup_n_p8): Likewise.
6308 (vdup_n_p16): Likewise.
6309 (vdup_n_p64): Likewise.
6310 (vdup_n_f16): Likewise.
6311 (vdup_n_f32): Likewise.
6312 (vdupq_n_s8): Likewise.
6313 (vdupq_n_s16): Likewise.
6314 (vdupq_n_s32): Likewise.
6315 (vdupq_n_s64): Likewise.
6316 (vdupq_n_u8): Likewise.
6317 (vdupq_n_u16): Likewise.
6318 (vdupq_n_u32): Likewise.
6319 (vdupq_n_u64): Likewise.
6320 (vdupq_n_p8): Likewise.
6321 (vdupq_n_p16): Likewise.
6322 (vdupq_n_p64): Likewise.
6323 (vdupq_n_f16): Likewise.
6324 (vdupq_n_f32): Likewise.
6325 (vmov_n_s8): Replace call to builtin with call to corresponding
6327 (vmov_n_s16): Likewise.
6328 (vmov_n_s32): Likewise.
6329 (vmov_n_s64): Likewise.
6330 (vmov_n_u8): Likewise.
6331 (vmov_n_u16): Likewise.
6332 (vmov_n_u32): Likewise.
6333 (vmov_n_u64): Likewise.
6334 (vmov_n_p8): Likewise.
6335 (vmov_n_p16): Likewise.
6336 (vmov_n_f16): Likewise.
6337 (vmov_n_f32): Likewise.
6338 (vmovq_n_s8): Likewise.
6339 (vmovq_n_s16): Likewise.
6340 (vmovq_n_s32): Likewise.
6341 (vmovq_n_s64): Likewise.
6342 (vmovq_n_u8): Likewise.
6343 (vmovq_n_u16): Likewise.
6344 (vmovq_n_u32): Likewise.
6345 (vmovq_n_u64): Likewise.
6346 (vmovq_n_p8): Likewise.
6347 (vmovq_n_p16): Likewise.
6348 (vmovq_n_f16): Likewise.
6349 (vmovq_n_f32): Likewise.
6350 * config/arm/arm_neon_builtins.def: Remove entries for vdup_n.
6352 2021-08-11 liuhongt <hongtao.liu@intel.com>
6355 * config/i386/i386.md (ldexp<mode>3): Extend to vscalefs[sd]
6356 when TARGET_AVX512F and TARGET_SSE_MATH.
6358 2021-08-10 Jakub Jelinek <jakub@redhat.com>
6361 * config/i386/i386-expand.c (expand_vec_perm_even_odd): Return false
6362 for V32HImode if !TARGET_AVX512BW.
6363 (ix86_vectorize_vec_perm_const) <case E_V32HImode, case E_V64QImode>:
6364 If !TARGET_AVX512BW and TARGET_AVX512F and d.testing_p, don't fail
6365 early, but actually check the permutation.
6367 2021-08-10 Richard Biener <rguenther@suse.de>
6369 PR tree-optimization/101809
6370 * tree-vect-stmts.c (get_load_store_type): Allow emulated
6371 gathers with offset vector nunits being a constant multiple
6372 of the data vector nunits.
6373 (vect_get_gather_scatter_ops): Use the appropriate nunits
6374 for the offset vector defs.
6375 (vectorizable_store): Adjust call to
6376 vect_get_gather_scatter_ops.
6377 (vectorizable_load): Likewise. Handle the case of less
6378 offset vectors than data vectors.
6380 2021-08-10 Jakub Jelinek <jakub@redhat.com>
6383 * config/i386/sse.md (*avx512f_shuf_<shuffletype>64x2_1<mask_name>_1,
6384 *avx512f_shuf_<shuffletype>32x4_1<mask_name>_1): New define_insn
6387 2021-08-10 Richard Biener <rguenther@suse.de>
6389 PR tree-optimization/101801
6390 PR tree-optimization/101819
6391 * tree-vectorizer.h (vect_emulated_vector_p): Declare.
6392 * tree-vect-loop.c (vect_emulated_vector_p): New function.
6393 (vectorizable_reduction): Re-instantiate a check for emulated
6395 * tree-vect-stmts.c (vectorizable_shift): Likewise.
6396 (vectorizable_operation): Likewise. Cost emulated vector
6397 operations according to the scalar sequence synthesized by
6400 2021-08-10 Richard Biener <rguenther@suse.de>
6402 PR middle-end/101824
6403 * tree-nested.c (get_frame_field): Mark the COMPONENT_REF as
6404 volatile in case the variable was.
6406 2021-08-10 H.J. Lu <hjl.tools@gmail.com>
6409 * config/i386/constraints.md (BC): Document for integer SSE
6410 constant all bits set operand.
6411 (BF): New constraint for const floating-point all bits set
6413 * config/i386/i386.c (standard_sse_constant_p): Likewise.
6414 (standard_sse_constant_opcode): Likewise.
6415 * config/i386/sse.md (sseconstm1): New mode attribute.
6416 (mov<mode>_internal): Replace BC with <sseconstm1>.
6418 2021-08-10 liuhongt <hongtao.liu@intel.com>
6420 * config/i386/sse.md (cond_<insn><mode>): New expander.
6421 (VI248_AVX512VLBW): New mode iterator.
6422 * config/i386/predicates.md
6423 (nonimmediate_or_const_vec_dup_operand): New predicate.
6425 2021-08-09 Andrew MacLeod <amacleod@redhat.com>
6427 PR tree-optimization/101741
6428 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Check
6429 type of parameter for toupper/tolower.
6431 2021-08-09 Martin Jambor <mjambor@suse.cz>
6434 * ipa-prop.c (propagate_controlled_uses): Removed a spurious space.
6436 2021-08-09 Pat Haugen <pthaugen@linux.ibm.com>
6438 * config/rs6000/rs6000.c (is_load_insn1): Verify destination is a
6440 (is_store_insn1): Verify source is a register.
6442 2021-08-09 Uroš Bizjak <ubizjak@gmail.com>
6445 * config/i386/mmx.md (<any_logic:code>v2sf3):
6446 Rename from *mmx_<any_logic:code>v2sf3
6448 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6450 * config/nvptx/nvptx.c: Cross-reference parts adapted in
6451 'gcc/omp-oacc-neuter-broadcast.cc'.
6452 * omp-low.c: Likewise.
6453 * omp-oacc-neuter-broadcast.cc: Cross-reference parts adapted from
6456 2021-08-09 Julian Brown <julian@codesourcery.com>
6457 Kwok Cheung Yeung <kcy@codesourcery.com>
6458 Thomas Schwinge <thomas@codesourcery.com>
6460 * config/gcn/gcn.c (gcn_init_builtins): Override decls for
6461 BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START,
6462 BUILT_IN_GOACC_SINGLE_COPY_END and BUILT_IN_GOACC_BARRIER.
6463 (gcn_goacc_validate_dims): Turn on worker partitioning unconditionally.
6464 (gcn_fork_join): Update comment.
6465 * config/gcn/gcn.opt (flag_worker_partitioning): Remove.
6466 (macc_experimental_workers): Remove unused option.
6468 2021-08-09 Julian Brown <julian@codesourcery.com>
6469 Nathan Sidwell <nathan@codesourcery.com> (via 'gcc/config/nvptx/nvptx.c' master)
6470 Kwok Cheung Yeung <kcy@codesourcery.com>
6471 Thomas Schwinge <thomas@codesourcery.com>
6473 * Makefile.in (OBJS): Add omp-oacc-neuter-broadcast.o.
6474 * doc/tm.texi.in (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD):
6475 Add documentation hook.
6476 * doc/tm.texi: Regenerate.
6477 * omp-oacc-neuter-broadcast.cc: New file.
6478 * omp-builtins.def (BUILT_IN_GOACC_BARRIER)
6479 (BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START)
6480 (BUILT_IN_GOACC_SINGLE_COPY_END): New builtins.
6481 * passes.def (pass_omp_oacc_neuter_broadcast): Add pass.
6482 * target.def (goacc.create_worker_broadcast_record): Add target
6484 * tree-pass.h (make_pass_omp_oacc_neuter_broadcast): Add
6486 * config/gcn/gcn-protos.h (gcn_goacc_adjust_propagation_record):
6487 Rename prototype to...
6488 (gcn_goacc_create_worker_broadcast_record): ... this.
6489 * config/gcn/gcn-tree.c (gcn_goacc_adjust_propagation_record): Rename
6491 (gcn_goacc_create_worker_broadcast_record): ... this.
6492 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_PROPAGATION_RECORD):
6494 (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD): ... this.
6496 2021-08-09 Tejas Belagod <tejas.belagod@arm.com>
6499 * config/aarch64/aarch64-simd.md (vlshr<mode>3, vashr<mode>3): Use
6502 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6504 * Makefile.in (GTFILES): Remove '$(srcdir)/omp-offload.c'.
6506 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6508 * builtins.def (DEF_GOACC_BUILTIN, DEF_GOMP_BUILTIN): Don't
6509 consider '-foffload-abi'.
6510 * common.opt (-foffload-abi): Remove 'Var', 'Init'.
6511 * opts.c (common_handle_option) <-foffload-abi> [ACCEL_COMPILER]:
6514 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6516 * optc-gen.awk: Sanity check that 'Init' doesn't appear without
6519 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6521 * omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE): Remove.
6523 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6525 * doc/gty.texi (Files): Update.
6527 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6529 * doc/gty.texi (Files): Fix GTY header file example.
6531 2021-08-09 Roger Sayle <roger@nextmovesoftware.com>
6533 * tree-ssa-ccp.c (value_mask_to_min_max): Helper function to
6534 determine the upper and lower bounds from a mask-value pair.
6535 (bit_value_unop) [ABS_EXPR, ABSU_EXPR]: Add support for
6536 absolute value and unsigned absolute value expressions.
6537 (bit_value_binop): Initialize *VAL's precision.
6538 [LT_EXPR, LE_EXPR]: Use value_mask_to_min_max to determine
6539 upper and lower bounds of operands. Add LE_EXPR/GE_EXPR
6540 support when the operands are unknown but potentially equal.
6541 [MIN_EXPR, MAX_EXPR]: Support minimum/maximum expressions.
6543 2021-08-09 Bin Cheng <bin.cheng@linux.alibaba.com>
6545 * config/aarch64/aarch64.md
6546 (*extend<SHORT:mode><GPI:mode>2_aarch64): Use %<GPI:w>0.
6548 2021-08-08 Sergei Trofimovich <siarheit@google.com>
6550 * lra-constraints.c: Fix s/otput/output/ typo.
6552 2021-08-06 Martin Sebor <msebor@redhat.com>
6554 * builtins.c (expand_builtin_memchr): Move to gimple-ssa-warn-access.cc.
6555 (expand_builtin_strcat): Same.
6556 (expand_builtin_stpncpy): Same.
6557 (expand_builtin_strncat): Same.
6558 (check_read_access): Same.
6559 (check_memop_access): Same.
6560 (expand_builtin_strlen): Move checks to gimple-ssa-warn-access.cc.
6561 (expand_builtin_strnlen): Same.
6562 (expand_builtin_memcpy): Same.
6563 (expand_builtin_memmove): Same.
6564 (expand_builtin_mempcpy): Same.
6565 (expand_builtin_strcpy): Same.
6566 (expand_builtin_strcpy_args): Same.
6567 (expand_builtin_stpcpy_1): Same.
6568 (expand_builtin_strncpy): Same.
6569 (expand_builtin_memset): Same.
6570 (expand_builtin_bzero): Same.
6571 (expand_builtin_strcmp): Same.
6572 (expand_builtin_strncmp): Same.
6573 (expand_builtin): Remove handlers.
6574 (fold_builtin_strlen): Add a comment.
6575 * builtins.h (check_access): Move to gimple-ssa-warn-access.cc.
6576 * calls.c (maybe_warn_nonstring_arg): Same.
6577 * diagnostic-spec.c (nowarn_spec_t::nowarn_spec_t): Add warning option.
6578 * gimple-fold.c (gimple_fold_builtin_strcpy): Pass argument to callee.
6579 (gimple_fold_builtin_stpcpy): Same.
6580 * gimple-ssa-warn-access.cc (has_location): New function.
6581 (get_location): Same.
6582 (get_callee_fndecl): Same.
6585 (warn_string_no_nul): Define.
6586 (unterminated_array): Same.
6587 (check_nul_terminated_array): Same.
6588 (maybe_warn_nonstring_arg): Same.
6589 (maybe_warn_for_bound): Same.
6590 (warn_for_access): Same.
6591 (check_access): Same.
6592 (check_memop_access): Same.
6593 (check_read_access): Same.
6594 (warn_dealloc_offset): Use helper functions.
6595 (maybe_emit_free_warning): Same.
6596 (class pass_waccess): Add members.
6597 (check_strcat): New function.
6598 (check_strncat): New function.
6599 (check_stxcpy): New function.
6600 (check_stxncpy): New function.
6601 (check_strncmp): New function.
6602 (pass_waccess::check_builtin): New function.
6603 (pass_waccess::check): Call it.
6604 * gimple-ssa-warn-access.h (warn_string_no_nul): Move here from
6606 (maybe_warn_for_bound): Same.
6607 (check_access): Same.
6608 (check_memop_access): Same.
6609 (check_read_access): Same.
6610 * pointer-query.h (struct access_data): Define a ctor overload.
6612 2021-08-06 Richard Biener <rguenther@suse.de>
6614 PR tree-optimization/101801
6615 * tree-vectorizer.h (vect_worthwhile_without_simd_p): Rename...
6616 (vect_can_vectorize_without_simd_p): ... to this.
6617 * tree-vect-loop.c (vect_worthwhile_without_simd_p): Rename...
6618 (vect_can_vectorize_without_simd_p): ... to this and fold
6619 in vect_min_worthwhile_factor.
6620 (vect_min_worthwhile_factor): Remove.
6621 (vectorizable_reduction): Adjust and remove the cost part.
6622 * tree-vect-stmts.c (vectorizable_shift): Likewise.
6623 (vectorizable_operation): Likewise.
6625 2021-08-06 Uroš Bizjak <ubizjak@gmail.com>
6628 * config/i386/i386.md (cmove reg-to-reg move elimination peephole2s):
6629 Add general_gr_operand predicate to operand 3.
6631 2021-08-06 Roger Sayle <roger@nextmovesoftware.com>
6633 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern): Use
6634 CFN_BUILT_IN_CLRSB* instead of BUILT_IN_CLRSB* for consistency.
6636 2021-08-06 Tamar Christina <tamar.christina@arm.com>
6638 * config/aarch64/aarch64-sve-builtins.cc (register_svpattern,
6639 register_svprfop): Pass vec<> by pointer.
6640 * langhooks-def.h (lhd_simulate_enum_decl): Likewise.
6641 * langhooks.c (lhd_simulate_enum_decl): Likewise.
6642 * langhooks.h (struct lang_hooks_for_types): Likewise.
6644 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
6646 * config/aarch64/arm_neon.h (vst1_bf16_x2): Use
6647 __builtin_memcpy instead of constructing an additional
6648 __builtin_aarch64_simd_oi one vector at a time.
6649 (vst1q_bf16_x2): Likewise.
6650 (vst1_bf16_x3): Use __builtin_memcpy instead of constructing
6651 an additional __builtin_aarch64_simd_ci one vector at a time.
6652 (vst1q_bf16_x3): Likewise.
6653 (vst1_bf16_x4): Use __builtin_memcpy instead of a union.
6654 (vst1q_bf16_x4): Likewise.
6655 (vst2_bf16): Use __builtin_memcpy instead of constructing an
6656 additional __builtin_aarch64_simd_oi one vector at a time.
6657 (vst2q_bf16): Likewise.
6658 (vst3_bf16): Use __builtin_memcpy instead of constructing an
6659 additional __builtin_aarch64_simd_ci mode one vector at a
6661 (vst3q_bf16): Likewise.
6662 (vst4_bf16): Use __builtin_memcpy instead of constructing an
6663 additional __builtin_aarch64_simd_xi one vector at a time.
6664 (vst4q_bf16): Likewise.
6666 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
6668 * config/aarch64/arm_neon.h (__ST2_LANE_FUNC): Delete.
6669 (__ST2Q_LANE_FUNC): Delete.
6670 (vst2_lane_f16): Use __builtin_memcpy to copy vector
6671 structure instead of constructing __builtin_aarch64_simd_oi
6672 one vector at a time.
6673 (vst2_lane_f32): Likewise.
6674 (vst2_lane_f64): Likewise.
6675 (vst2_lane_p8): Likewise.
6676 (vst2_lane_p16): Likewise.
6677 (vst2_lane_p64): Likewise.
6678 (vst2_lane_s8): Likewise.
6679 (vst2_lane_s16): Likewise.
6680 (vst2_lane_s32): Likewise.
6681 (vst2_lane_s64): Likewise.
6682 (vst2_lane_u8): Likewise.
6683 (vst2_lane_u16): Likewise.
6684 (vst2_lane_u32): Likewise.
6685 (vst2_lane_u64): Likewise.
6686 (vst2_lane_bf16): Likewise.
6687 (vst2q_lane_f16): Use __builtin_memcpy to copy vector
6688 structure instead of using a union.
6689 (vst2q_lane_f32): Likewise.
6690 (vst2q_lane_f64): Likewise.
6691 (vst2q_lane_p8): Likewise.
6692 (vst2q_lane_p16): Likewise.
6693 (vst2q_lane_p64): Likewise.
6694 (vst2q_lane_s8): Likewise.
6695 (vst2q_lane_s16): Likewise.
6696 (vst2q_lane_s32): Likewise.
6697 (vst2q_lane_s64): Likewise.
6698 (vst2q_lane_u8): Likewise.
6699 (vst2q_lane_u16): Likewise.
6700 (vst2q_lane_u32): Likewise.
6701 (vst2q_lane_u64): Likewise.
6702 (vst2q_lane_bf16): Likewise.
6704 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
6706 * config/aarch64/arm_neon.h (__ST3_LANE_FUNC): Delete.
6707 (__ST3Q_LANE_FUNC): Delete.
6708 (vst3_lane_f16): Use __builtin_memcpy to copy vector
6709 structure instead of constructing __builtin_aarch64_simd_ci
6710 one vector at a time.
6711 (vst3_lane_f32): Likewise.
6712 (vst3_lane_f64): Likewise.
6713 (vst3_lane_p8): Likewise.
6714 (vst3_lane_p16): Likewise.
6715 (vst3_lane_p64): Likewise.
6716 (vst3_lane_s8): Likewise.
6717 (vst3_lane_s16): Likewise.
6718 (vst3_lane_s32): Likewise.
6719 (vst3_lane_s64): Likewise.
6720 (vst3_lane_u8): Likewise.
6721 (vst3_lane_u16): Likewise.
6722 (vst3_lane_u32): Likewise.
6723 (vst3_lane_u64): Likewise.
6724 (vst3_lane_bf16): Likewise.
6725 (vst3q_lane_f16): Use __builtin_memcpy to copy vector
6726 structure instead of using a union.
6727 (vst3q_lane_f32): Likewise.
6728 (vst3q_lane_f64): Likewise.
6729 (vst3q_lane_p8): Likewise.
6730 (vst3q_lane_p16): Likewise.
6731 (vst3q_lane_p64): Likewise.
6732 (vst3q_lane_s8): Likewise.
6733 (vst3q_lane_s16): Likewise.
6734 (vst3q_lane_s32): Likewise.
6735 (vst3q_lane_s64): Likewise.
6736 (vst3q_lane_u8): Likewise.
6737 (vst3q_lane_u16): Likewise.
6738 (vst3q_lane_u32): Likewise.
6739 (vst3q_lane_u64): Likewise.
6740 (vst3q_lane_bf16): Likewise.
6742 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
6744 * config/aarch64/arm_neon.h (__ST4_LANE_FUNC): Delete.
6745 (__ST4Q_LANE_FUNC): Delete.
6746 (vst4_lane_f16): Use __builtin_memcpy to copy vector
6747 structure instead of constructing __builtin_aarch64_simd_xi
6748 one vector at a time.
6749 (vst4_lane_f32): Likewise.
6750 (vst4_lane_f64): Likewise.
6751 (vst4_lane_p8): Likewise.
6752 (vst4_lane_p16): Likewise.
6753 (vst4_lane_p64): Likewise.
6754 (vst4_lane_s8): Likewise.
6755 (vst4_lane_s16): Likewise.
6756 (vst4_lane_s32): Likewise.
6757 (vst4_lane_s64): Likewise.
6758 (vst4_lane_u8): Likewise.
6759 (vst4_lane_u16): Likewise.
6760 (vst4_lane_u32): Likewise.
6761 (vst4_lane_u64): Likewise.
6762 (vst4_lane_bf16): Likewise.
6763 (vst4q_lane_f16): Use __builtin_memcpy to copy vector
6764 structure instead of using a union.
6765 (vst4q_lane_f32): Likewise.
6766 (vst4q_lane_f64): Likewise.
6767 (vst4q_lane_p8): Likewise.
6768 (vst4q_lane_p16): Likewise.
6769 (vst4q_lane_p64): Likewise.
6770 (vst4q_lane_s8): Likewise.
6771 (vst4q_lane_s16): Likewise.
6772 (vst4q_lane_s32): Likewise.
6773 (vst4q_lane_s64): Likewise.
6774 (vst4q_lane_u8): Likewise.
6775 (vst4q_lane_u16): Likewise.
6776 (vst4q_lane_u32): Likewise.
6777 (vst4q_lane_u64): Likewise.
6778 (vst4q_lane_bf16): Likewise.
6780 2021-08-06 Martin Liska <mliska@suse.cz>
6782 * config/rs6000/rs6000.c (rs6000_option_override_internal): When
6783 a target option is restored, it can have
6784 rs6000_long_double_type_size set to FLOAT_PRECISION_TFmode
6785 and error should not be emitted.
6787 2021-08-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
6789 * gcov-io.h (gcov_write): Declare.
6790 * gcov-io.c (gcov_write): New.
6791 (gcov_write_counter): Remove.
6792 (gcov_write_tag_length): Likewise.
6793 (gcov_write_summary): Replace gcov_write_tag_length() with calls to
6794 gcov_write_unsigned().
6795 * doc/invoke.texi (fprofile-info-section): Mention
6796 __gcov_info_to_gdca().
6798 2021-08-06 Martin Sebor <msebor@redhat.com>
6800 * dominance.c (prune_bbs_to_update_dominators): Adjust by-value vec
6801 arguments to by-reference.
6802 (iterate_fix_dominators): Same.
6803 * dominance.h (iterate_fix_dominators): Same.
6804 * ipa-prop.h: Call auto_vec::to_vec_legacy.
6805 * tree-data-ref.c (dump_data_dependence_relation): Adjust by-value vec
6806 arguments to by-reference.
6807 (debug_data_dependence_relation): Same.
6808 (dump_data_dependence_relations): Same.
6809 * tree-data-ref.h (debug_data_dependence_relation): Same.
6810 (dump_data_dependence_relations): Same.
6811 * tree-predcom.c (dump_chains): Same.
6812 (initialize_root_vars_lm): Same.
6813 (determine_unroll_factor): Same.
6814 (replace_phis_by_defined_names): Same.
6815 (insert_init_seqs): Same.
6816 (pcom_worker::tree_predictive_commoning_loop): Call
6817 auto_vec::to_vec_legacy.
6818 * tree-ssa-pre.c (insert_into_preds_of_block): Adjust by-value vec
6819 arguments to by-reference.
6820 * tree-ssa-threadbackward.c (populate_worklist): Same.
6821 (back_threader::resolve_def): Same.
6822 * tree-vect-data-refs.c (vect_check_nonzero_value): Same.
6823 (vect_enhance_data_refs_alignment): Same.
6824 (vect_check_lower_bound): Same.
6825 (vect_prune_runtime_alias_test_list): Same.
6826 (vect_permute_store_chain): Same.
6827 * tree-vect-slp-patterns.c (vect_normalize_conj_loc): Same.
6828 * tree-vect-stmts.c (vect_create_vectorized_demotion_stmts): Same.
6829 * tree-vectorizer.h (vect_permute_store_chain): Same.
6830 * vec.c (test_init): New function.
6831 (vec_c_tests): Call new function.
6832 * vec.h (vec): Declare ctors, dtor, and assignment.
6833 (auto_vec::vec_to_legacy): New function.
6834 (vec::copy): Adjust initialization.
6836 2021-08-05 H.J. Lu <hjl.tools@gmail.com>
6839 * config/i386/i386.c (ix86_can_inline_p): Ignore MASK_80387 if
6840 callee only uses GPRs.
6841 * config/i386/ia32intrin.h: Revert commit 5463cee2770.
6842 * config/i386/serializeintrin.h: Revert commit 71958f740f1.
6843 * config/i386/x86gprintrin.h: Add
6844 #pragma GCC target("general-regs-only") and #pragma GCC pop_options
6845 to disable non-GPR ISAs.
6847 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
6849 PR middle-end/101787
6850 * doc/md.texi (cond_ashl, cond_ashr, cond_lshr): Document.
6852 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
6854 * tree-vectorizer.h (vect_is_store_elt_extraction, vect_is_reduction)
6855 (vect_reduc_type, vect_embedded_comparison_type, vect_comparison_type)
6856 (vect_is_extending_load, vect_is_integer_truncation): New functions,
6857 moved from aarch64.c but given different names.
6858 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction)
6859 (aarch64_is_reduction, aarch64_reduc_type)
6860 (aarch64_embedded_comparison_type, aarch64_comparison_type)
6861 (aarch64_extending_load_p, aarch64_integer_truncation_p): Delete
6862 in favor of the above. Update callers accordingly.
6864 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
6867 * config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
6868 writing .cpu directive in asm output.
6869 * config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
6870 (arm_last_printed_arch_string): Delete.
6871 (arm_last-printed_fpu_string): Delete.
6872 (arm_configure_build_target): If use of floating-point/SIMD is
6873 disabled, remove all fp/simd related features from the target ISA.
6874 (last_arm_targ_options): New variable.
6875 (arm_print_asm_arch_directives): Add new parameters. Change order
6876 of emitted directives and handle all cases here.
6877 (arm_file_start): Always call arm_print_asm_arch_directives, move
6878 all generation of .arch/.arch_extension here.
6879 (arm_file_end): Call arm_print_asm_arch.
6880 (arm_declare_function_name): Call arm_print_asm_arch_directives
6881 instead of printing .arch/.fpu directives directly.
6883 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
6885 * config/arm/arm.c (arm_configure_build_target): Don't call
6886 arm_option_reconfigure_globals.
6887 (arm_option_restore): Call arm_option_reconfigure_globals after
6888 reconfiguring the target.
6889 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
6891 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
6893 * config/arm/arm.c (arm_configure_build_target): Ensure the target's
6894 arch_name is always set.
6896 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
6898 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
6899 of vec_select high-half from being added into Neon subtract
6902 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
6904 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
6905 of vec_select high-half from being added into Neon add cost.
6907 2021-08-05 Kewen Lin <linkw@linux.ibm.com>
6909 * cfgloop.h (loops_list::loops_list): Add one optional argument
6910 root and adjust accordingly, update loop tree walking and factor
6912 * cfgloop.c (loops_list::walk_loop_tree): ... this. New function.
6914 2021-08-05 Eric Botcazou <ebotcazou@gcc.gnu.org>
6916 PR tree-optimization/101626
6917 * tree-sra.c (propagate_subaccesses_from_rhs): Do not set the
6918 reverse scalar storage order on a pointer or vector component.
6920 2021-08-05 liuhongt <hongtao.liu@intel.com>
6922 * config/i386/sse.md (cond_<code><mode>): New expander.
6924 2021-08-05 liuhongt <hongtao.liu@intel.com>
6926 * config/i386/sse.md (cond_<code><mode>): New expander.
6928 2021-08-05 liuhongt <hongtao.liu@intel.com>
6930 * config/i386/sse.md (cond_<code><mode>): New expander.
6932 2021-08-04 David Malcolm <dmalcolm@redhat.com>
6935 * Makefile.in (ANALYZER_OBJS): Add analyzer/region-model-asm.o.
6937 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
6940 * config/i386/i386.h (STORE_MAX_PIECES): Allow 16/32/64 bytes
6941 only if TARGET_INTER_UNIT_MOVES_TO_VEC is true.
6943 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
6946 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
6947 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
6948 data with SSE register from one memory location to another.
6950 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
6952 * config/s390/s390.c (expand_perm_with_vpdi): New function.
6953 (vectorize_vec_perm_const_1): Call expand_perm_with_vpdi.
6954 * config/s390/vector.md (*vpdi1<mode>, @vpdi1<mode>): Enable a
6955 parameterized expander.
6956 (*vpdi4<mode>, @vpdi4<mode>): Likewise.
6958 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
6960 * config/s390/s390.c (MAX_VECT_LEN): Define macro.
6961 (struct expand_vec_perm_d): Define struct.
6962 (expand_perm_with_merge): New function.
6963 (vectorize_vec_perm_const_1): New function.
6964 (s390_vectorize_vec_perm_const): New function.
6965 (TARGET_VECTORIZE_VEC_PERM_CONST): Define target macro.
6967 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
6969 * config/s390/vector.md (V_HW_64): Remove mode iterator.
6970 (*vec_load_pair<mode>): Use V_HW_2 instead of V_HW_64.
6971 * config/s390/vx-builtins.md
6972 (vec_scatter_element<V_HW_2:mode>_SI): Use V_HW_2 instead of
6975 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
6977 * config/s390/s390.md (UNSPEC_VEC_PERMI): Remove constant
6979 * config/s390/vector.md (*vpdi1<mode>, *vpdi4<mode>): New pattern
6981 * config/s390/vx-builtins.md (*vec_permi<mode>): Emit generic rtx
6982 instead of an unspec.
6984 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
6986 * config/s390/s390-modes.def: Add more vector modes to support
6987 concatenation of two vectors.
6988 * config/s390/s390-protos.h (s390_expand_merge_perm_const): Add
6990 (s390_expand_merge): Likewise.
6991 * config/s390/s390.c (s390_expand_merge_perm_const): New function.
6992 (s390_expand_merge): New function.
6993 * config/s390/s390.md (UNSPEC_VEC_MERGEH, UNSPEC_VEC_MERGEL):
6994 Remove constant definitions.
6995 * config/s390/vector.md (V_HW_2): Add mode iterators.
6996 (VI_HW_4, V_HW_4): Rename VI_HW_4 to V_HW_4.
6997 (vec_2x_nelts, vec_2x_wide): New mode attributes.
6998 (*vmrhb, *vmrlb, *vmrhh, *vmrlh, *vmrhf, *vmrlf, *vmrhg, *vmrlg):
6999 New pattern definitions.
7000 (vec_widen_umult_lo_<mode>, vec_widen_umult_hi_<mode>)
7001 (vec_widen_smult_lo_<mode>, vec_widen_smult_hi_<mode>)
7002 (vec_unpacks_lo_v4sf, vec_unpacks_hi_v4sf, vec_unpacks_lo_v2df)
7003 (vec_unpacks_hi_v2df): Adjust expanders to emit non-unspec RTX for
7005 * config/s390/vx-builtins.md (V_HW_4): Remove mode iterator. Now
7007 (vec_mergeh<mode>, vec_mergel<mode>): Use s390_expand_merge to
7008 emit vec merge pattern.
7010 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
7012 * config/aarch64/aarch64.c (aarch64_strip_extend_vec_half):
7014 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent cost of
7015 vec_select high-half from being added into Neon multiply
7017 * rtlanal.c (vec_series_highpart_p): Define.
7018 * rtlanal.h (vec_series_highpart_p): Declare.
7020 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
7022 * config/aarch64/aarch64.c (aarch64_strip_duplicate_vec_elt):
7024 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent
7025 vec_select cost from being added into Neon multiply cost.
7027 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
7029 * tree-vect-loop.c (vect_better_loop_vinfo_p): Detect cases in
7030 which old_loop_vinfo is an epilogue loop that handles a constant
7031 number of iterations.
7033 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
7035 * tree-vect-loop.c (vect_analyze_loop): Print a dump message
7036 when a reanalyzed loop fails to be cheaper than the current
7039 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
7041 * config/aarch64/aarch64.c: Fix a typo.
7043 2021-08-04 Vincent Lefèvre <vincent-gcc@vinc17.net>
7045 PR gcov-profile/101773
7046 * gcov-io.c (gcov_close): Check return code of a fclose.
7048 2021-08-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
7051 * dwarf2out.c (dwarf2out_assembly_start): Emit a dummy
7052 .file statement when needed.
7054 2021-08-04 Richard Biener <rguenther@suse.de>
7056 * tree-vect-data-refs.c (vect_check_gather_scatter):
7057 Include widening conversions only when the result is
7058 still handed by native gather or the current offset
7059 size not already matches the data size.
7060 Also succeed analysis in case there's no native support,
7061 noted by a IFN_LAST ifn and a NULL decl.
7062 (vect_analyze_data_refs): Always consider gathers.
7063 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
7064 Test for no IFN gather rather than decl gather.
7065 * tree-vect-stmts.c (vect_model_load_cost): Pass in the
7066 gather-scatter info and cost emulated gathers accordingly.
7067 (vect_truncate_gather_scatter_offset): Properly test for
7069 (vect_use_strided_gather_scatters_p): Likewise.
7070 (get_load_store_type): Handle emulated gathers and its
7072 (vectorizable_load): Likewise. Emulate them by extracting
7073 scalar offsets, doing scalar loads and a vector construct.
7075 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
7078 * expr.c (op_by_pieces_d::op_by_pieces_d): Add a max_pieces
7079 argument to set m_max_size.
7080 (move_by_pieces_d): Pass MOVE_MAX_PIECES to op_by_pieces_d.
7081 (store_by_pieces_d): Pass STORE_MAX_PIECES to op_by_pieces_d.
7082 (compare_by_pieces_d): Pass COMPARE_MAX_PIECES to op_by_pieces_d.
7084 2021-08-04 Roger Sayle <roger@nextmovesoftware.com>
7085 Marc Glisse <marc.glisse@inria.fr>
7087 * match.pd (bit_ior, bit_xor): Canonicalize (X*C1)|(X*C2) and
7088 (X*C1)^(X*C2) as X*(C1+C2), and related variants, using
7089 tree_nonzero_bits to ensure that operands are bit-wise disjoint.
7091 2021-08-04 Richard Biener <rguenther@suse.de>
7093 * tree-ssa-forwprop.c (pass_forwprop::execute): Split
7094 out code to decompose vector loads ...
7095 (optimize_vector_load): ... here. Generalize it to
7096 handle intermediate widening and TARGET_MEM_REF loads
7097 and apply it to loads with a supported vector mode as well.
7099 2021-08-04 Richard Biener <rguenther@suse.de>
7101 PR tree-optimization/101756
7102 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Make sure
7103 the result of the reduction epilogue is compatible to the original
7106 2021-08-04 liuhongt <hongtao.liu@intel.com>
7109 * config/i386/i386.md (peephole2): Refine predicate from
7110 register_operand to general_reg_operand.
7112 2021-08-04 Aldy Hernandez <aldyh@redhat.com>
7114 * gimple-range-path.h (path_range_query::dump): Mark override.
7116 2021-08-04 Richard Biener <rguenther@suse.de>
7118 PR tree-optimization/101769
7119 * tree-tailcall.c (eliminate_tail_call): Add the created loop
7120 for the first recursion and return it via the new output parameter.
7121 (optimize_tail_call): Pass through new output param.
7122 (tree_optimize_tail_calls_1): After creating all latches,
7123 add the created loop to the loop tree. Do not mark loops for fixup.
7125 2021-08-04 Martin Liska <mliska@suse.cz>
7127 * doc/invoke.texi: Document threader-mode param.
7129 2021-08-04 liuhongt <hongtao.liu@intel.com>
7131 * config/i386/sse.md (cond_fma<mode>): New expander.
7132 (cond_fms<mode>): Ditto.
7133 (cond_fnma<mode>): Ditto.
7134 (cond_fnms<mode>): Ditto.
7136 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
7138 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode>): Use && instead of &.
7140 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
7142 * config/rs6000/constraints.md: Remove "e" from the list of available
7143 constraint characters.
7145 2021-08-03 Eugene Rozenfeld <erozen@microsoft.com>
7147 PR gcov-profile/71672
7148 * auto-profile.c (afdo_indirect_call): Fix setup of the historgram value for indirect calls.
7150 2021-08-03 Paul A. Clarke <pc@us.ibm.com>
7152 * config/rs6000/smmintrin.h (_mm_minpos_epu16): New.
7154 2021-08-03 H.J. Lu <hjl.tools@gmail.com>
7156 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): In 64-bit mode,
7157 try XMM31 to avoid vzeroupper.
7159 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7161 * doc/invoke.texi: Document -mtune=neoverse-512tvb and
7162 -mcpu=neoverse-512tvb.
7163 * config/aarch64/aarch64-cores.def (neoverse-512tvb): New entry.
7164 * config/aarch64/aarch64-tune.md: Regenerate.
7165 * config/aarch64/aarch64.c (neoverse512tvb_sve_vector_cost)
7166 (neoverse512tvb_sve_issue_info, neoverse512tvb_vec_issue_info)
7167 (neoverse512tvb_vector_cost, neoverse512tvb_tunings): New structures.
7168 (aarch64_adjust_body_cost_sve): Handle -mtune=neoverse-512tvb.
7169 (aarch64_adjust_body_cost): Likewise.
7171 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7173 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Only
7174 record issue information for operations that occur in the
7177 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7179 * config/aarch64/aarch64.c (aarch64_multiply_add_p): Add a vec_flags
7180 parameter. Detect cases in which an Advanced SIMD MLA would almost
7181 certainly require a MOV.
7182 (aarch64_count_ops): Update accordingly.
7184 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7186 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction): New
7187 function, split out from...
7188 (aarch64_detect_vector_stmt_subtype): ...here.
7189 (aarch64_add_stmt_cost): Treat extracting element 0 as free.
7191 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7193 * config/aarch64/aarch64-protos.h (sve_vec_cost):
7194 Add gather_load_x32_cost and gather_load_x64_cost.
7195 * config/aarch64/aarch64.c (generic_sve_vector_cost)
7196 (a64fx_sve_vector_cost, neoversev1_sve_vector_cost): Update
7197 accordingly, using the values given by the scalar_load * number
7198 of elements calculation that we used previously.
7199 (aarch64_detect_vector_stmt_subtype): Use the new fields.
7201 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7203 * config/aarch64/aarch64.c (aarch64_adjust_body_cost_sve): New
7204 function, split out from...
7205 (aarch64_adjust_body_cost): ...here.
7207 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7209 * config/aarch64/fractional-cost.h: New file.
7210 * config/aarch64/aarch64.c: Include <algorithm> (indirectly)
7211 and cost_fraction.h.
7212 (vec_cost_fraction): New typedef.
7213 (aarch64_detect_scalar_stmt_subtype): Use it for statement costs.
7214 (aarch64_detect_vector_stmt_subtype): Likewise.
7215 (aarch64_sve_adjust_stmt_cost, aarch64_adjust_stmt_cost): Likewise.
7216 (aarch64_estimate_min_cycles_per_iter): Use vec_cost_fraction
7218 (aarch64_adjust_body_cost): Likewise.
7219 (aarch64_test_cost_fraction): New function.
7220 (aarch64_run_selftests): Call it.
7222 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7224 * config/aarch64/aarch64-protos.h (tune_params::sve_width): Turn
7226 * config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Update
7228 (aarch64_estimated_poly_value): Likewise. Use the least significant
7229 set bit for the minimum and likely values. Use the most significant
7230 set bit for the maximum value.
7232 2021-08-03 liuhongt <hongtao.liu@intel.com>
7234 * config/i386/sse.md (cond_<insn><mode>): New expander.
7235 (cond_mul<mode>): Ditto.
7237 2021-08-03 Kewen Lin <linkw@linux.ibm.com>
7239 * tree-cfg.c (move_sese_region_to_fn): Fix typos on dloop.
7241 2021-08-03 liuhongt <hongtao.liu@intel.com>
7243 * config/i386/sse.md (cond_<insn><mode>):New expander.
7244 (cond_mul<mode>): Ditto.
7245 (cond_div<mode>): Ditto.
7247 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
7249 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Also
7250 check stack_realign_needed for stack realignment.
7251 (ix86_legitimate_constant_p): Always allow CONST_WIDE_INT smaller
7252 than the largest integer supported by vector register.
7253 * config/i386/i386.h (MAX_MOVE_MAX): New. Set to 64.
7254 (MOVE_MAX): Set to bytes of the largest integer supported by
7256 (STORE_MAX_PIECES): New.
7258 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
7260 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
7261 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
7262 data from one memory location to another.
7264 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
7267 * config/i386/i386.c (TARGET_GEN_MEMSET_SCRATCH_RTX): New.
7269 2021-08-02 Aldy Hernandez <aldyh@redhat.com>
7271 PR tree-optimization/101724
7272 * params.opt: Remove --param=threader-iterative.
7273 * tree-ssa-threadbackward.c (pass_thread_jumps::execute): Remove
7276 2021-08-02 Tom de Vries <tdevries@suse.de>
7278 PR middle-end/101665
7279 * doc/extend.texi (nonnull attribute): Improve documentation.
7281 2021-08-02 Andrew Pinski <apinski@marvell.com>
7283 PR rtl-optimization/101683
7284 * rtlanal.c (may_trap_p_1): Handle UNSIGNED_FIX.
7286 2021-08-02 Roger Sayle <roger@nextmovesoftware.com>
7288 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern):
7289 Renamed from cond_removal_in_popcount_clz_ctz_pattern.
7290 Add support for BSWAP, FFS, PARITY and CLRSB builtins.
7291 (tree_ssa_phiop_worker): Update call to function above.
7293 2021-08-01 H.J. Lu <hjl.tools@gmail.com>
7296 * config/i386/i386.md (bsr_rex64_1_zext): New.
7297 (combine splitter for constant - clzll): Replace gen_bsr_rex64_1
7298 with gen_bsr_rex64_1_zext.
7300 2021-07-31 Jakub Jelinek <jakub@redhat.com>
7303 * config/i386/i386.md (bsr_rex64_1, bsr_1, bsr_zext_1): New
7304 define_insn patterns.
7305 (*bsr_rex64_2, *bsr_2): New define_insn_and_split patterns.
7306 Add combine splitters for constant - clz.
7307 (clz<mode>2): Use a temporary pseudo for bsr result.
7309 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
7311 * config/rs6000/smmintrin.h (_mm_floor_pd, _mm_floor_ps,
7312 _mm_floor_sd, _mm_floor_ss): New.
7314 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
7316 * config/rs6000/smmintrin.h (_mm_ceil_pd, _mm_ceil_ps,
7317 _mm_ceil_sd, _mm_ceil_ss): New.
7319 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
7321 * config/rs6000/smmintrin.h (_mm_blend_pd, _mm_blendv_pd,
7322 _mm_blend_ps, _mm_blendv_ps): New.
7324 2021-07-30 Roger Sayle <roger@nextmovesoftware.com>
7325 Uroš Bizjak <ubizjak@gmail.com>
7327 * config/i386/i386.md (*dec_cmov<mode>): New define_insn_and_split
7328 to generate a conditional move using the carry flag after sub $1.
7329 (peephole2): Eliminate a register-to-register move by inverting
7330 the condition of a conditional move.
7332 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
7334 * config/mmix/mmix.md ("call", "call_value", "*call_real")
7335 ("*call_value_real"): Don't generate rtx mentioning the generic
7336 operands 1 and 2 to "call", and similarly for "call_value".
7337 * config/mmix/mmix.c (mmix_print_operand_punct_valid_p)
7338 (mmix_print_operand): Use '!' instead of 'p'.
7340 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
7342 * doc/md.texi (call): Correct information about operand 2.
7343 * config/mmix/mmix.md ("call", "call_value"): Remove fixed FIXMEs.
7345 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
7347 * range-op.cc (operator_trunc_mod::wi_fold): Fold constants.
7349 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
7351 * range-op.cc (operator_div::wi_fold): Return UNDEFINED for [0, 0] divisor.
7353 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
7355 * gimple-range-cache.cc (*::set_bb_range): Change const basic_block to
7357 (*::get_bb_range): Ditto.
7358 (*::bb_range_p): Ditto.
7359 * gimple-range-cache.h: Change prototypes.
7361 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
7364 * builtins.c (builtin_memcpy_read_str): Change the mode argument
7365 from scalar_int_mode to fixed_size_mode.
7366 (builtin_strncpy_read_str): Likewise.
7367 (gen_memset_value_from_prev): New function.
7368 (builtin_memset_read_str): Change the mode argument from
7369 scalar_int_mode to fixed_size_mode. Use gen_memset_value_from_prev
7370 and support CONST_VECTOR.
7371 (builtin_memset_gen_str): Likewise.
7372 (try_store_by_multiple_pieces): Use by_pieces_constfn to declare
7374 * builtins.h (builtin_strncpy_read_str): Replace scalar_int_mode
7375 with fixed_size_mode.
7376 (builtin_memset_read_str): Likewise.
7377 * expr.c (widest_int_mode_for_size): Renamed to ...
7378 (widest_fixed_size_mode_for_size): Add a bool argument to
7379 indicate if QI vector mode can be used.
7380 (by_pieces_ninsns): Call widest_fixed_size_mode_for_size
7381 instead of widest_int_mode_for_size.
7382 (pieces_addr::adjust): Change the mode argument from
7383 scalar_int_mode to fixed_size_mode.
7384 (op_by_pieces_d): Make m_len read-only. Add a bool member,
7385 m_qi_vector_mode, to indicate that QI vector mode can be used.
7386 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
7387 initialize m_qi_vector_mode. Call widest_fixed_size_mode_for_size
7388 instead of widest_int_mode_for_size.
7389 (op_by_pieces_d::get_usable_mode): Change the mode argument from
7390 scalar_int_mode to fixed_size_mode. Call
7391 widest_fixed_size_mode_for_size instead of
7392 widest_int_mode_for_size.
7393 (op_by_pieces_d::smallest_fixed_size_mode_for_size): New member
7394 function to return the smallest integer or QI vector mode.
7395 (op_by_pieces_d::run): Call widest_fixed_size_mode_for_size
7396 instead of widest_int_mode_for_size. Call
7397 smallest_fixed_size_mode_for_size instead of
7398 smallest_int_mode_for_size.
7399 (store_by_pieces_d::store_by_pieces_d): Add a bool argument to
7400 indicate that QI vector mode can be used and pass it to
7401 op_by_pieces_d::op_by_pieces_d.
7402 (can_store_by_pieces): Call widest_fixed_size_mode_for_size
7403 instead of widest_int_mode_for_size. Pass memsetp to
7404 widest_fixed_size_mode_for_size to support QI vector mode.
7405 Allow all CONST_VECTORs for memset if vec_duplicate is supported.
7406 (store_by_pieces): Pass memsetp to
7407 store_by_pieces_d::store_by_pieces_d.
7408 (clear_by_pieces_1): Removed.
7409 (clear_by_pieces): Replace clear_by_pieces_1 with
7410 builtin_memset_read_str and pass true to store_by_pieces_d to
7411 support vector mode broadcast.
7412 (string_cst_read_str): Change the mode argument from
7413 scalar_int_mode to fixed_size_mode.
7414 * expr.h (by_pieces_constfn): Change scalar_int_mode to
7416 (by_pieces_prev): Likewise.
7417 * rtl.h (lowpart_subreg_regno): New.
7418 * rtlanal.c (lowpart_subreg_regno): New. A wrapper around
7419 simplify_subreg_regno.
7420 * target.def (gen_memset_scratch_rtx): New hook.
7421 * doc/tm.texi.in: Add TARGET_GEN_MEMSET_SCRATCH_RTX.
7422 * doc/tm.texi: Regenerated.
7424 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
7427 * config/mips/mips.c (mips_atomic_assign_expand_fenv): Use
7428 TARGET_EXPR instead of MODIFY_EXPR.
7430 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
7433 * config/mips/mips-protos.h (mips_expand_vec_cmp_expr): Declare.
7434 * config/mips/mips.c (mips_expand_vec_cmp_expr): New function.
7435 * config/mips/mips-msa.md (vec_cmp<MSA:mode><mode_i>): New
7437 (vec_cmpu<IMSA:mode><mode_i>): New expander.
7439 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
7442 * config/i386/i386-options.c (ix86_option_override_internal):
7443 Don't enable LZCNT/POPCNT if they have been disabled explicitly.
7445 2021-07-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
7448 * config/arm/arm_neon.h (vld1_p64): Replace call to builtin by
7449 explicitly dereferencing __a.
7450 (vld1_s64): Likewise.
7451 (vld1_u64): Likewise.
7452 * config/arm/arm_neon_builtins.def (vld1): Remove entry for di
7453 and change to VAR13.
7455 2021-07-30 Aldy Hernandez <aldyh@redhat.com>
7457 * gimple-loop-versioning.cc (lv_dom_walker::lv_dom_walker): Remove
7458 use of m_range_analyzer.
7459 (loop_versioning::lv_dom_walker::before_dom_children): Same.
7460 (loop_versioning::lv_dom_walker::after_dom_children): Remove.
7461 (loop_versioning::prune_loop_conditions): Replace vr_values use
7462 with range_query interface.
7463 (pass_loop_versioning::execute): Use ranger.
7465 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
7468 * ipa-devirt.c (ipa_odr_read_section): Compare the precision of
7469 enum values, and emit a warning if they mismatch.
7471 2021-07-30 Kewen Lin <linkw@linux.ibm.com>
7473 * cfgloop.h (as_const): New function.
7474 (class loop_iterator): Rename to ...
7475 (class loops_list): ... this.
7476 (loop_iterator::next): Rename to ...
7477 (loops_list::Iter::fill_curr_loop): ... this and adjust.
7478 (loop_iterator::loop_iterator): Rename to ...
7479 (loops_list::loops_list): ... this and adjust.
7480 (loops_list::Iter): New class.
7481 (loops_list::iterator): New type.
7482 (loops_list::const_iterator): New type.
7483 (loops_list::begin): New function.
7484 (loops_list::end): Likewise.
7485 (loops_list::begin const): Likewise.
7486 (loops_list::end const): Likewise.
7487 (FOR_EACH_LOOP): Remove.
7488 (FOR_EACH_LOOP_FN): Remove.
7489 * cfgloop.c (flow_loops_dump): Adjust FOR_EACH_LOOP* with range-based
7490 for loop with loops_list instance.
7491 (sort_sibling_loops): Likewise.
7492 (disambiguate_loops_with_multiple_latches): Likewise.
7493 (verify_loop_structure): Likewise.
7494 * cfgloopmanip.c (create_preheaders): Likewise.
7495 (force_single_succ_latches): Likewise.
7496 * config/aarch64/falkor-tag-collision-avoidance.c
7497 (execute_tag_collision_avoidance): Likewise.
7498 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Likewise.
7499 * config/s390/s390.c (s390_adjust_loops): Likewise.
7500 * doc/loop.texi: Likewise.
7501 * gimple-loop-interchange.cc (pass_linterchange::execute): Likewise.
7502 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
7503 * gimple-loop-versioning.cc (loop_versioning::analyze_blocks): Likewise.
7504 (loop_versioning::make_versioning_decisions): Likewise.
7505 * gimple-ssa-split-paths.c (split_paths): Likewise.
7506 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
7507 * graphite.c (canonicalize_loop_form): Likewise.
7508 (graphite_transform_loops): Likewise.
7509 * ipa-fnsummary.c (analyze_function_body): Likewise.
7510 * ipa-pure-const.c (analyze_function): Likewise.
7511 * loop-doloop.c (doloop_optimize_loops): Likewise.
7512 * loop-init.c (loop_optimizer_finalize): Likewise.
7513 (fix_loop_structure): Likewise.
7514 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
7515 (move_loop_invariants): Likewise.
7516 * loop-unroll.c (decide_unrolling): Likewise.
7517 (unroll_loops): Likewise.
7518 * modulo-sched.c (sms_schedule): Likewise.
7519 * predict.c (predict_loops): Likewise.
7520 (pass_profile::execute): Likewise.
7521 * profile.c (branch_prob): Likewise.
7522 * sel-sched-ir.c (sel_finish_pipelining): Likewise.
7523 (sel_find_rgns): Likewise.
7524 * tree-cfg.c (replace_loop_annotate): Likewise.
7525 (replace_uses_by): Likewise.
7526 (move_sese_region_to_fn): Likewise.
7527 * tree-if-conv.c (pass_if_conversion::execute): Likewise.
7528 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
7529 * tree-parloops.c (parallelize_loops): Likewise.
7530 * tree-predcom.c (tree_predictive_commoning): Likewise.
7531 * tree-scalar-evolution.c (scev_initialize): Likewise.
7532 (scev_reset): Likewise.
7533 * tree-ssa-dce.c (find_obviously_necessary_stmts): Likewise.
7534 * tree-ssa-live.c (remove_unused_locals): Likewise.
7535 * tree-ssa-loop-ch.c (ch_base::copy_headers): Likewise.
7536 * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
7537 (tree_ssa_lim_initialize): Likewise.
7538 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Likewise.
7539 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize): Likewise.
7540 * tree-ssa-loop-manip.c (get_loops_exits): Likewise.
7541 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations): Likewise.
7542 (free_numbers_of_iterations_estimates): Likewise.
7543 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Likewise.
7544 * tree-ssa-loop-split.c (tree_ssa_split_loops): Likewise.
7545 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Likewise.
7546 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
7547 (pass_scev_cprop::execute): Likewise.
7548 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
7549 * tree-ssa-sccvn.c (do_rpo_vn): Likewise.
7550 * tree-ssa-threadupdate.c
7551 (jump_thread_path_registry::thread_through_all_blocks): Likewise.
7552 * tree-vectorizer.c (vectorize_loops): Likewise.
7553 * tree-vrp.c (vrp_asserts::find_assert_locations): Likewise.
7555 2021-07-29 Hans-Peter Nilsson <hp@bitrange.com>
7557 * config/mmix/mmix.c (mmix_function_arg_1): Avoid
7558 generating a VOIDmode register for e.g the
7559 function_arg_info::end_marker.
7561 2021-07-29 Jeff Law <jeffreyalaw@gmail.com>
7563 * config/h8300/h8300-modes.def: Add CCZ, CCV and CCC, drop CCZNV.
7564 * config/h8300/h8300.md (H8cc mode iterator): Add CCZ.
7565 (cc mode_attr): Similarly.
7566 (ccz subst_attr): Similarly.
7567 * config/h8300/jumpcall.md: Add new patterns for branch-on-bit.
7568 * config/h8300/testcompare.md: Remove various cc0 based patterns
7569 that had been commented out. Add pattern to set CCZ from a bit
7572 2021-07-29 Thomas Schwinge <thomas@codesourcery.com>
7573 Julian Brown <julian@codesourcery.com>
7574 Kwok Cheung Yeung <kcy@codesourcery.com>
7576 * omp-offload.c (oacc_loop_xform_head_tail, oacc_loop_process):
7577 'update_stmt' after modification.
7578 (pass_oacc_loop_designation): New function, extracted out of...
7579 (pass_oacc_device_lower): ... this.
7580 (pass_data_oacc_loop_designation, pass_oacc_loop_designation)
7581 (make_pass_oacc_loop_designation): New
7582 * passes.def: Add it.
7583 * tree-parloops.c (create_parallel_loop): Adjust.
7584 * tree-pass.h (make_pass_oacc_loop_designation): New.
7586 2021-07-29 Aldy Hernandez <aldyh@redhat.com>
7588 * flag-types.h (enum threader_mode): New.
7589 * params.opt: Add entry for --param=threader-mode.
7590 * tree-ssa-threadbackward.c (THREADER_ITERATIVE_MODE): New.
7591 (class back_threader): New.
7592 (back_threader::back_threader): New.
7593 (back_threader::~back_threader): New.
7594 (back_threader::maybe_register_path): New.
7595 (back_threader::find_taken_edge): New.
7596 (back_threader::find_taken_edge_switch): New.
7597 (back_threader::find_taken_edge_cond): New.
7598 (back_threader::resolve_def): New.
7599 (back_threader::resolve_phi): New.
7600 (back_threader::find_paths_to_names): New.
7601 (back_threader::find_paths): New.
7604 (thread_jumps::find_jump_threads_backwards): Call ranger threader.
7605 (thread_jumps::find_jump_threads_backwards_with_ranger): New.
7606 (pass_thread_jumps::execute): Abstract out code...
7607 (try_thread_blocks): ...here.
7608 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
7609 Abstract out threading candidate code to...
7610 (single_succ_to_potentially_threadable_block): ...here.
7611 * tree-ssa-threadedge.h (single_succ_to_potentially_threadable_block):
7613 * tree-ssa-threadupdate.c (register_jump_thread): Return boolean.
7614 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
7615 Return bool from register_jump_thread.
7617 2021-07-29 Andreas Krebbel <krebbel@linux.ibm.com>
7619 * target.def: in0 and in1 do not need to be registers.
7620 * doc/tm.texi: Regenerate.
7622 2021-07-29 liuhongt <hongtao.liu@intel.com>
7625 * config/i386/i386.c (ix86_widen_mult_cost): New function.
7626 (ix86_add_stmt_cost): Use ix86_widen_mult_cost for
7629 2021-07-29 Jiufu Guo <guojiufu@linux.ibm.com>
7632 * config/rs6000/rs6000.c (TARGET_PREFERRED_DOLOOP_MODE): New hook.
7633 (rs6000_preferred_doloop_mode): New hook.
7634 * doc/tm.texi: Regenerate.
7635 * doc/tm.texi.in: Add hook preferred_doloop_mode.
7636 * target.def (preferred_doloop_mode): New hook.
7637 * targhooks.c (default_preferred_doloop_mode): New hook.
7638 * targhooks.h (default_preferred_doloop_mode): New hook.
7639 * tree-ssa-loop-ivopts.c (compute_doloop_base_on_mode): New function.
7640 (add_iv_candidate_for_doloop): Call targetm.preferred_doloop_mode
7641 and compute_doloop_base_on_mode.
7643 2021-07-28 Martin Sebor <msebor@redhat.com>
7645 PR middle-end/101494
7646 * tree-ssa-uninit.c (maybe_warn_operand): Correct object offset
7647 and size computation.
7649 2021-07-28 Martin Sebor <msebor@redhat.com>
7651 PR middle-end/101601
7652 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Remove
7654 Handle pointers to functions.
7656 2021-07-28 Martin Sebor <msebor@redhat.com>
7658 * Makefile.in (OBJS): Add gimple-ssa-warn-access.o and pointer-query.o.
7659 * attribs.h (fndecl_dealloc_argno): Move fndecl_dealloc_argno to tree.h.
7660 * builtins.c (compute_objsize_r): Move to pointer-query.cc.
7661 (access_ref::access_ref): Same.
7662 (access_ref::phi): Same.
7663 (access_ref::get_ref): Same.
7664 (access_ref::size_remaining): Same.
7665 (access_ref::offset_in_range): Same.
7666 (access_ref::add_offset): Same.
7667 (access_ref::inform_access): Same.
7668 (ssa_name_limit_t::visit_phi): Same.
7669 (ssa_name_limit_t::leave_phi): Same.
7670 (ssa_name_limit_t::next): Same.
7671 (ssa_name_limit_t::next_phi): Same.
7672 (ssa_name_limit_t::~ssa_name_limit_t): Same.
7673 (pointer_query::pointer_query): Same.
7674 (pointer_query::get_ref): Same.
7675 (pointer_query::put_ref): Same.
7676 (pointer_query::flush_cache): Same.
7677 (warn_string_no_nul): Move to gimple-ssa-warn-access.cc.
7678 (check_nul_terminated_array): Same.
7679 (unterminated_array): Same.
7680 (maybe_warn_for_bound): Same.
7681 (check_read_access): Same.
7682 (warn_for_access): Same.
7683 (get_size_range): Same.
7684 (check_access): Same.
7685 (gimple_call_alloc_size): Move to tree.c.
7686 (gimple_parm_array_size): Move to pointer-query.cc.
7687 (get_offset_range): Same.
7688 (gimple_call_return_array): Same.
7689 (handle_min_max_size): Same.
7690 (handle_array_ref): Same.
7691 (handle_mem_ref): Same.
7692 (compute_objsize): Same.
7693 (gimple_call_alloc_p): Move to gimple-ssa-warn-access.cc.
7694 (call_dealloc_argno): Same.
7695 (fndecl_dealloc_argno): Same.
7696 (new_delete_mismatch_p): Same.
7697 (matching_alloc_calls_p): Same.
7698 (warn_dealloc_offset): Same.
7699 (maybe_emit_free_warning): Same.
7700 * builtins.h (check_nul_terminated_array): Move to
7701 gimple-ssa-warn-access.h.
7702 (check_nul_terminated_array): Same.
7703 (warn_string_no_nul): Same.
7704 (unterminated_array): Same.
7705 (class ssa_name_limit_t): Same.
7706 (class pointer_query): Same.
7707 (struct access_ref): Same.
7708 (class range_query): Same.
7709 (struct access_data): Same.
7710 (gimple_call_alloc_size): Same.
7711 (gimple_parm_array_size): Same.
7712 (compute_objsize): Same.
7713 (class access_data): Same.
7714 (maybe_emit_free_warning): Same.
7715 * calls.c (initialize_argument_information): Remove call to
7716 maybe_emit_free_warning.
7717 * gimple-array-bounds.cc: Include new header..
7718 * gimple-fold.c: Same.
7719 * gimple-ssa-sprintf.c: Same.
7720 * gimple-ssa-warn-restrict.c: Same.
7721 * passes.def: Add pass_warn_access.
7722 * tree-pass.h (make_pass_warn_access): Declare.
7723 * tree-ssa-strlen.c: Include new headers.
7724 * tree.c (fndecl_dealloc_argno): Move here from builtins.c.
7725 * tree.h (fndecl_dealloc_argno): Move here from attribs.h.
7726 * gimple-ssa-warn-access.cc: New file.
7727 * gimple-ssa-warn-access.h: New file.
7728 * pointer-query.cc: New file.
7729 * pointer-query.h: New file.
7731 2021-07-28 Jakub Jelinek <jakub@redhat.com>
7733 PR middle-end/101624
7734 * ubsan.c (maybe_instrument_pointer_overflow,
7735 instrument_object_size): Only test DECL_REGISTER on VAR_DECLs,
7736 PARM_DECLs or RESULT_DECLs.
7737 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
7739 2021-07-28 Jakub Jelinek <jakub@redhat.com>
7741 PR middle-end/101642
7742 * match.pd (bswap16 (x) == bswap16 (y)): Cast both operands
7743 to type of bswap16 for comparison.
7744 (bswap16 (x) == cst): Cast bswap16 operand to type of cst.
7746 2021-07-28 Richard Biener <rguenther@suse.de>
7748 PR tree-optimization/101615
7749 * tree-vect-slp.c (vect_optimize_slp): Materialize permutes
7750 at CTOR SLP graph entries.
7752 2021-07-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7754 * config/aarch64/aarch64.md (*extend<SHORT:mode><GPI:mode>2_aarch64):
7755 Add "r,w" alternative.
7757 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
7760 * config/i386/i386.c (ix86_avx_u128_mode_needed): Don't set
7761 AVX_U128_DIRTY when all bits are zero.
7763 2021-07-28 Richard Biener <rguenther@suse.de>
7765 PR tree-optimization/101615
7766 * tree-vect-slp.c (vect_optimize_slp): Pre-existing vector
7767 external nodes cannot be permuted so make them perm_out 0.
7769 2021-07-28 Andrew Stubbs <ams@codesourcery.com>
7772 * config.in: Regenerate.
7773 * config/gcn/gcn-hsa.h (A_FIJI): New define.
7774 (A_900): New define.
7775 (A_906): New define.
7776 (A_908): New define.
7777 (ASM_SPEC): Use A_FIJI, A_900, A_906 and A_908.
7778 * config/gcn/gcn.c (output_file_start): Adjust attributes according
7779 to the assembler capabilities.
7780 * config/gcn/mkoffload.c (main): Likewise.
7781 * configure: Regenerate.
7782 * configure.ac: Add tests for LLVM assembler attribute features.
7784 2021-07-28 Andrew MacLeod <amacleod@redhat.com>
7786 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Check for
7787 cond_false and cond_true on branches.
7789 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
7791 * config/aarch64/aarch64.c (aarch64_gen_adjusted_ldpstp): use
7794 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
7796 * alias.c (init_alias_analysis): Don't skip prologue/epilogue.
7798 2021-07-28 Jakub Jelinek <jakub@redhat.com>
7801 * config/i386/sse.md (vashr<mode>3): Split into vashrv8di3 expander
7802 and vashrv4di3 expander, where the latter requires just TARGET_AVX2
7803 and has special !TARGET_AVX512VL expansion.
7804 (vashrv2di3<mask_name>): Rename to ...
7805 (vashrv2di3): ... this. Change condition to TARGET_XOP || TARGET_AVX2
7806 and add special !TARGET_XOP && !TARGET_AVX512VL expansion.
7808 2021-07-28 Martin Uecker <muecker@gwdg.de>
7810 * calls.c (maybe_warn_rdwr_sizes): Correct argument
7811 numbers in warning that were switched.
7813 2021-07-28 Kewen Lin <linkw@linux.ibm.com>
7815 PR tree-optimization/101596
7816 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Fix wrong check
7817 by using new_type's precision instead.
7819 2021-07-28 liuhongt <hongtao.liu@intel.com>
7822 * config/i386/i386.h (processor_costs): Add new member
7824 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
7825 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
7826 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
7827 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
7828 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
7829 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
7830 generic_cost, core_cost): Initialize integer_to_sse same value
7832 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
7833 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7834 Use integer_to_sse instead of sse_op to calculate the cost of
7837 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
7839 * config/rs6000/rs6000-gen-builtins.c (write_ovld_static_init): New
7841 (write_init_file): Call write_ovld_static_init.
7843 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
7845 * config/rs6000/rs6000-gen-builtins.c (write_bif_static_init): New
7847 (write_init_file): Call write_bif_static_init.
7849 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
7851 * config/rs6000/rs6000-gen-builtins.c (typemap): New struct.
7852 (TYPE_MAP_SIZE): New macro.
7853 (type_map): New initialized variable.
7854 (typemap_cmp): New function.
7855 (write_type_node): Likewise.
7856 (write_fntype_init): Implement.
7858 2021-07-27 Martin Sebor <msebor@redhat.com>
7860 PR tree-optimization/101584
7861 * tree-ssa-uninit.c (builtin_call_nomodifying_p): New function.
7862 (check_defs): Call it.
7864 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
7866 * tree-ssa-dom.c (dom_jump_threader_simplifier):
7867 Put avail_exprs_stack in the class, instead of passing it to
7868 jump_threader_simplifier.
7869 (dom_jump_threader_simplifier::simplify): Add state argument.
7870 (dom_opt_dom_walker): Add state.
7871 (pass_dominator::execute): Pass state to threader.
7872 (dom_opt_dom_walker::before_dom_children): Use state.
7873 * tree-ssa-threadedge.c (jump_threader::jump_threader): Replace
7875 (jump_threader::record_temporary_equivalences_from_phis):
7876 Register equivalences through the state variable.
7877 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
7878 Record ranges in a statement through the state variable.
7879 (jump_threader::simplify_control_stmt_condition): Pass state to
7881 (jump_threader::simplify_control_stmt_condition_1): Same.
7882 (jump_threader::thread_around_empty_blocks): Remove obsolete
7884 (jump_threader::thread_through_normal_block): Record equivalences
7885 on edge through the state variable.
7886 (jump_threader::thread_across_edge): Abstract state pushing.
7887 (jt_state::jt_state): New.
7888 (jt_state::push): New.
7889 (jt_state::pop): New.
7890 (jt_state::register_equiv): New.
7891 (jt_state::record_ranges_from_stmt): New.
7892 (jt_state::register_equivs_on_edge): New.
7893 (jump_threader_simplifier::jump_threader_simplifier): Move from
7895 (jump_threader_simplifier::simplify): Add state argument.
7896 * tree-ssa-threadedge.h (class jt_state): New.
7897 (class jump_threader): Add state to constructor.
7898 (class jump_threader_simplifier): Add state to simplify. Remove
7899 avail_exprs_stack from class.
7900 * tree-vrp.c (vrp_jump_threader_simplifier::simplify): Add state
7902 (vrp_jump_threader::vrp_jump_threader): Add state.
7903 (vrp_jump_threader::~vrp_jump_threader): Cleanup state.
7905 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
7907 * Makefile.in (OBJS): Add gimple-range-path.o.
7908 * gimple-range-path.cc: New file.
7909 * gimple-range-path.h: New file.
7911 2021-07-27 Jonathan Wright <jonathan.wright@arm.com>
7913 * config/aarch64/aarch64-simd.md: Push sign/zero-extension
7914 inside vec_duplicate for all patterns.
7915 * simplify-rtx.c (simplify_context::simplify_unary_operation_1):
7916 Push sign/zero-extension inside vec_duplicate.
7918 2021-07-27 Richard Biener <rguenther@suse.de>
7920 PR tree-optimization/101573
7921 * tree-ssa-uninit.c (warn_uninit_phi_uses): New function
7922 looking at uninitialized PHI arg defs in some constrained cases.
7923 (warn_uninitialized_vars): Call it.
7924 (execute_early_warn_uninitialized): Calculate dominators.
7926 2021-07-27 Richard Biener <rguenther@suse.de>
7928 PR tree-optimization/39821
7929 * tree-vect-stmts.c (vect_model_promotion_demotion_cost): Use
7930 vector_stmt for widening arithmetic.
7931 (vectorizable_conversion): Adjust.
7933 2021-07-27 Martin Jambor <mjambor@suse.cz>
7935 * cgraph.h (ipa_replace_map): New field force_load_ref.
7936 * ipa-prop.h (ipa_param_descriptor): Reduce precision of move_cost,
7937 aded new flag load_dereferenced, adjusted comments.
7938 (ipa_get_param_dereferenced): New function.
7939 (ipa_set_param_dereferenced): Likewise.
7940 * cgraphclones.c (cgraph_node::create_virtual_clone): Follow it.
7941 * ipa-cp.c: Include gimple.h.
7942 (ipcp_discover_new_direct_edges): Take into account dereferenced flag.
7943 (get_replacement_map): New parameter force_load_ref, set the
7944 appropriate flag in ipa_replace_map if set.
7945 (struct symbol_and_index_together): New type.
7946 (adjust_refs_in_act_callers): New function.
7947 (adjust_references_in_caller): Likewise.
7948 (create_specialized_node): When appropriate, call
7949 adjust_references_in_caller and force only load references.
7950 * ipa-prop.c (load_from_dereferenced_name): New function.
7951 (ipa_analyze_controlled_uses): Also detect loads from a
7952 dereference, harden testing of call statements.
7953 (ipa_write_node_info): Stream the dereferenced flag.
7954 (ipa_read_node_info): Likewise.
7955 (ipa_set_jf_constant): Also create refdesc when jump function
7956 references a variable.
7957 (cgraph_node_for_jfunc): Rename to symtab_node_for_jfunc, work
7958 also on references of variables and return a symtab_node. Adjust
7960 (propagate_controlled_uses): Also remove references to VAR_DECLs.
7962 2021-07-27 Jakub Jelinek <jakub@redhat.com>
7964 PR middle-end/101586
7965 * gimple-fold.c (clear_padding_type): Ignore FIELD_DECLs with byte
7966 positions above or equal to sz except for diagnostics of flexible
7969 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
7971 PR tree-optimization/78888
7972 * gimple-range-fold.cc (get_letter_range): New.
7973 (fold_using_range::range_of_builtin_call): Call get_letter_range.
7975 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
7977 PR tree-optimization/78888
7978 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Add cases
7979 for CFN_BUILT_IN_TOUPPER and CFN_BUILT_IN_TOLOWER.
7981 2021-07-26 Roger Sayle <roger@nextmovesoftware.com>
7982 Marc Glisse <marc.glisse@inria.fr>
7984 * match.pd (rotate): Simplify equality/inequality of rotations.
7985 (bswap): Simplify equality/inequality tests of byte swapping.
7987 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
7989 * range-op.cc (operator_bitwise_xor::op1_op2_relation_effect):
7992 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
7994 * range-op.cc (operator_lshift::fold_range): Pass rel to
7995 base class fold_range.
7996 (operator_rshift::fold_range): Same.
7998 2021-07-26 Ashimida <ashimida@linux.alibaba.com>
8001 * toplev.h (min_align_loops_log): Remove declaration.
8002 (min_align_jumps_log, min_align_labels_log): Likewise.
8003 (min_align_functions_log): Likewise.
8005 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8007 * tree-vrp.c (vrp_simplify_cond_using_ranges): Rename vr_values
8009 (execute_vrp): Abstract out simplification of conditionals...
8010 (simplify_casted_conds): ...here.
8012 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8014 * gimple-array-bounds.cc (array_bounds_checker::get_value_range):
8015 Add gimple argument.
8016 (array_bounds_checker::check_array_ref): Same.
8017 (array_bounds_checker::check_addr_expr): Same.
8018 (array_bounds_checker::check_array_bounds): Pass statement to
8019 check_array_bounds and check_addr_expr.
8020 * gimple-array-bounds.h (check_array_bounds): Add gimple argument.
8021 (check_addr_expr): Same.
8022 (get_value_range): Same.
8024 2021-07-26 Tamar Christina <tamar.christina@arm.com>
8026 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): Rename to..
8027 (sdot_prod, udot_prod): ... This.
8028 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): Merged
8030 (<sur>dot_prod<vsi2qi>): ... this.
8031 (aarch64_<sur>dot_lane<vsi2qi>, aarch64_<sur>dot_laneq<vsi2qi>):
8032 Change operands order.
8033 (<sur>sadv16qi): Use new operands order.
8034 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32,
8035 vdotq_s32): Use new RTL ordering.
8037 2021-07-26 Tamar Christina <tamar.christina@arm.com>
8039 * config/aarch64/aarch64-builtins.c (TYPES_TERNOP_SUSS,
8040 aarch64_types_ternop_suss_qualifiers): New.
8041 * config/aarch64/aarch64-simd-builtins.def (usdot_prod): Use it.
8042 * config/aarch64/aarch64-simd.md (usdot_prod<vsi2qi>): Re-organize RTL.
8043 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Use it.
8045 2021-07-23 Jakub Jelinek <jakub@redhat.com>
8047 PR rtl-optimization/101562
8048 * expmed.c (store_integral_bit_field): Only use movstrict_optab
8049 if the operand isn't paradoxical.
8051 2021-07-23 Aldy Hernandez <aldyh@redhat.com>
8053 * gimple-array-bounds.h (class array_bounds_checker): Change
8054 ranges type to range_query.
8056 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8058 * config/aarch64/arm_neon.h (vst1_s64_x2): Use
8059 __builtin_memcpy instead of constructing
8060 __builtin_aarch64_simd_oi one vector at a time.
8061 (vst1_u64_x2): Likewise.
8062 (vst1_f64_x2): Likewise.
8063 (vst1_s8_x2): Likewise.
8064 (vst1_p8_x2): Likewise.
8065 (vst1_s16_x2): Likewise.
8066 (vst1_p16_x2): Likewise.
8067 (vst1_s32_x2): Likewise.
8068 (vst1_u8_x2): Likewise.
8069 (vst1_u16_x2): Likewise.
8070 (vst1_u32_x2): Likewise.
8071 (vst1_f16_x2): Likewise.
8072 (vst1_f32_x2): Likewise.
8073 (vst1_p64_x2): Likewise.
8074 (vst1q_s8_x2): Likewise.
8075 (vst1q_p8_x2): Likewise.
8076 (vst1q_s16_x2): Likewise.
8077 (vst1q_p16_x2): Likewise.
8078 (vst1q_s32_x2): Likewise.
8079 (vst1q_s64_x2): Likewise.
8080 (vst1q_u8_x2): Likewise.
8081 (vst1q_u16_x2): Likewise.
8082 (vst1q_u32_x2): Likewise.
8083 (vst1q_u64_x2): Likewise.
8084 (vst1q_f16_x2): Likewise.
8085 (vst1q_f32_x2): Likewise.
8086 (vst1q_f64_x2): Likewise.
8087 (vst1q_p64_x2): Likewise.
8089 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8091 * config/aarch64/arm_neon.h (vst1_s64_x3): Use
8092 __builtin_memcpy instead of constructing
8093 __builtin_aarch64_simd_ci one vector at a time.
8094 (vst1_u64_x3): Likewise.
8095 (vst1_f64_x3): Likewise.
8096 (vst1_s8_x3): Likewise.
8097 (vst1_p8_x3): Likewise.
8098 (vst1_s16_x3): Likewise.
8099 (vst1_p16_x3): Likewise.
8100 (vst1_s32_x3): Likewise.
8101 (vst1_u8_x3): Likewise.
8102 (vst1_u16_x3): Likewise.
8103 (vst1_u32_x3): Likewise.
8104 (vst1_f16_x3): Likewise.
8105 (vst1_f32_x3): Likewise.
8106 (vst1_p64_x3): Likewise.
8107 (vst1q_s8_x3): Likewise.
8108 (vst1q_p8_x3): Likewise.
8109 (vst1q_s16_x3): Likewise.
8110 (vst1q_p16_x3): Likewise.
8111 (vst1q_s32_x3): Likewise.
8112 (vst1q_s64_x3): Likewise.
8113 (vst1q_u8_x3): Likewise.
8114 (vst1q_u16_x3): Likewise.
8115 (vst1q_u32_x3): Likewise.
8116 (vst1q_u64_x3): Likewise.
8117 (vst1q_f16_x3): Likewise.
8118 (vst1q_f32_x3): Likewise.
8119 (vst1q_f64_x3): Likewise.
8120 (vst1q_p64_x3): Likewise.
8122 2021-07-23 H.J. Lu <hjl.tools@gmail.com>
8125 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): Don't return
8126 hard register when LRA is in progress.
8128 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8130 * config/aarch64/arm_neon.h (vst1_s8_x4): Use
8131 __builtin_memcpy instead of using a union.
8132 (vst1q_s8_x4): Likewise.
8133 (vst1_s16_x4): Likewise.
8134 (vst1q_s16_x4): Likewise.
8135 (vst1_s32_x4): Likewise.
8136 (vst1q_s32_x4): Likewise.
8137 (vst1_u8_x4): Likewise.
8138 (vst1q_u8_x4): Likewise.
8139 (vst1_u16_x4): Likewise.
8140 (vst1q_u16_x4): Likewise.
8141 (vst1_u32_x4): Likewise.
8142 (vst1q_u32_x4): Likewise.
8143 (vst1_f16_x4): Likewise.
8144 (vst1q_f16_x4): Likewise.
8145 (vst1_f32_x4): Likewise.
8146 (vst1q_f32_x4): Likewise.
8147 (vst1_p8_x4): Likewise.
8148 (vst1q_p8_x4): Likewise.
8149 (vst1_p16_x4): Likewise.
8150 (vst1q_p16_x4): Likewise.
8151 (vst1_s64_x4): Likewise.
8152 (vst1_u64_x4): Likewise.
8153 (vst1_p64_x4): Likewise.
8154 (vst1q_s64_x4): Likewise.
8155 (vst1q_u64_x4): Likewise.
8156 (vst1q_p64_x4): Likewise.
8157 (vst1_f64_x4): Likewise.
8158 (vst1q_f64_x4): Likewise.
8160 2021-07-23 Jonathan Wrightt <jonathan.wright@arm.com>
8162 * config/aarch64/arm_neon.h (vst2_s64): Use __builtin_memcpy
8163 instead of constructing __builtin_aarch64_simd_oi one vector
8165 (vst2_u64): Likewise.
8166 (vst2_f64): Likewise.
8167 (vst2_s8): Likewise.
8168 (vst2_p8): Likewise.
8169 (vst2_s16): Likewise.
8170 (vst2_p16): Likewise.
8171 (vst2_s32): Likewise.
8172 (vst2_u8): Likewise.
8173 (vst2_u16): Likewise.
8174 (vst2_u32): Likewise.
8175 (vst2_f16): Likewise.
8176 (vst2_f32): Likewise.
8177 (vst2_p64): Likewise.
8178 (vst2q_s8): Likewise.
8179 (vst2q_p8): Likewise.
8180 (vst2q_s16): Likewise.
8181 (vst2q_p16): Likewise.
8182 (vst2q_s32): Likewise.
8183 (vst2q_s64): Likewise.
8184 (vst2q_u8): Likewise.
8185 (vst2q_u16): Likewise.
8186 (vst2q_u32): Likewise.
8187 (vst2q_u64): Likewise.
8188 (vst2q_f16): Likewise.
8189 (vst2q_f32): Likewise.
8190 (vst2q_f64): Likewise.
8191 (vst2q_p64): Likewise.
8193 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8195 * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy
8196 instead of constructing __builtin_aarch64_simd_ci one vector
8198 (vst3_u64): Likewise.
8199 (vst3_f64): Likewise.
8200 (vst3_s8): Likewise.
8201 (vst3_p8): Likewise.
8202 (vst3_s16): Likewise.
8203 (vst3_p16): Likewise.
8204 (vst3_s32): Likewise.
8205 (vst3_u8): Likewise.
8206 (vst3_u16): Likewise.
8207 (vst3_u32): Likewise.
8208 (vst3_f16): Likewise.
8209 (vst3_f32): Likewise.
8210 (vst3_p64): Likewise.
8211 (vst3q_s8): Likewise.
8212 (vst3q_p8): Likewise.
8213 (vst3q_s16): Likewise.
8214 (vst3q_p16): Likewise.
8215 (vst3q_s32): Likewise.
8216 (vst3q_s64): Likewise.
8217 (vst3q_u8): Likewise.
8218 (vst3q_u16): Likewise.
8219 (vst3q_u32): Likewise.
8220 (vst3q_u64): Likewise.
8221 (vst3q_f16): Likewise.
8222 (vst3q_f32): Likewise.
8223 (vst3q_f64): Likewise.
8224 (vst3q_p64): Likewise.
8226 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8228 * config/aarch64/arm_neon.h (vst4_s64): Use __builtin_memcpy
8229 instead of constructing __builtin_aarch64_simd_xi one vector
8231 (vst4_u64): Likewise.
8232 (vst4_f64): Likewise.
8233 (vst4_s8): Likewise.
8234 (vst4_p8): Likewise.
8235 (vst4_s16): Likewise.
8236 (vst4_p16): Likewise.
8237 (vst4_s32): Likewise.
8238 (vst4_u8): Likewise.
8239 (vst4_u16): Likewise.
8240 (vst4_u32): Likewise.
8241 (vst4_f16): Likewise.
8242 (vst4_f32): Likewise.
8243 (vst4_p64): Likewise.
8244 (vst4q_s8): Likewise.
8245 (vst4q_p8): Likewise.
8246 (vst4q_s16): Likewise.
8247 (vst4q_p16): Likewise.
8248 (vst4q_s32): Likewise.
8249 (vst4q_s64): Likewise.
8250 (vst4q_u8): Likewise.
8251 (vst4q_u16): Likewise.
8252 (vst4q_u32): Likewise.
8253 (vst4q_u64): Likewise.
8254 (vst4q_f16): Likewise.
8255 (vst4q_f32): Likewise.
8256 (vst4q_f64): Likewise.
8257 (vst4q_p64): Likewise.
8259 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8261 * config/aarch64/arm_neon.h (vtbx4_s8): Use __builtin_memcpy
8262 instead of constructing __builtin_aarch64_simd_oi one vector
8264 (vtbx4_u8): Likewise.
8265 (vtbx4_p8): Likewise.
8267 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8269 * config/aarch64/arm_neon.h (vtbl3_s8): Use __builtin_memcpy
8270 instead of constructing __builtin_aarch64_simd_oi one vector
8272 (vtbl3_u8): Likewise.
8273 (vtbl3_p8): Likewise.
8274 (vtbl4_s8): Likewise.
8275 (vtbl4_u8): Likewise.
8276 (vtbl4_p8): Likewise.
8278 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8280 * config/aarch64/arm_neon.h (vqtbx2_s8): Use __builtin_memcpy
8281 instead of constructing __builtin_aarch64_simd_oi one vector
8283 (vqtbx2_u8): Likewise.
8284 (vqtbx2_p8): Likewise.
8285 (vqtbx2q_s8): Likewise.
8286 (vqtbx2q_u8): Likewise.
8287 (vqtbx2q_p8): Likewise.
8288 (vqtbx3_s8): Use __builtin_memcpy instead of constructing
8289 __builtin_aarch64_simd_ci one vector at a time.
8290 (vqtbx3_u8): Likewise.
8291 (vqtbx3_p8): Likewise.
8292 (vqtbx3q_s8): Likewise.
8293 (vqtbx3q_u8): Likewise.
8294 (vqtbx3q_p8): Likewise.
8295 (vqtbx4_s8): Use __builtin_memcpy instead of constructing
8296 __builtin_aarch64_simd_xi one vector at a time.
8297 (vqtbx4_u8): Likewise.
8298 (vqtbx4_p8): Likewise.
8299 (vqtbx4q_s8): Likewise.
8300 (vqtbx4q_u8): Likewise.
8301 (vqtbx4q_p8): Likewise.
8303 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8305 * config/aarch64/arm_neon.h (vqtbl2_s8): Use __builtin_memcpy
8306 instead of constructing __builtin_aarch64_simd_oi one vector
8308 (vqtbl2_u8): Likewise.
8309 (vqtbl2_p8): Likewise.
8310 (vqtbl2q_s8): Likewise.
8311 (vqtbl2q_u8): Likewise.
8312 (vqtbl2q_p8): Likewise.
8313 (vqtbl3_s8): Use __builtin_memcpy instead of constructing
8314 __builtin_aarch64_simd_ci one vector at a time.
8315 (vqtbl3_u8): Likewise.
8316 (vqtbl3_p8): Likewise.
8317 (vqtbl3q_s8): Likewise.
8318 (vqtbl3q_u8): Likewise.
8319 (vqtbl3q_p8): Likewise.
8320 (vqtbl4_s8): Use __builtin_memcpy instead of constructing
8321 __builtin_aarch64_simd_xi one vector at a time.
8322 (vqtbl4_u8): Likewise.
8323 (vqtbl4_p8): Likewise.
8324 (vqtbl4q_s8): Likewise.
8325 (vqtbl4q_u8): Likewise.
8326 (vqtbl4q_p8): Likewise.
8328 2021-07-23 Haochen Gui <guihaoc@gcc.gnu.org>
8331 * config/rs6000/rs6000.md (cstore<mode>4): Fix wrong fall through.
8333 2021-07-22 Andrew Pinski <apinski@marvell.com>
8335 PR tree-optimization/10153
8336 * tree-tailcall.c (create_tailcall_accumulator):
8337 Don't call fold_convert as the type should be correct already.
8338 (tree_optimize_tail_calls_1): Use build_{one,zero}_cst instead
8339 of integer_{one,zero}_node for the call of create_tailcall_accumulator.
8341 2021-07-22 Aldy Hernandez <aldyh@redhat.com>
8343 * gimple-range-cache.cc (non_null_ref::adjust_range): Replace
8344 varying_p check for null/non-null check.
8346 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
8348 PR tree-optimization/101511
8349 * value-relation.cc (relation_oracle::query_relation): Check if ssa1
8350 is in ssa2's equiv set, and don't trap if so.
8352 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
8354 PR tree-optimization/101497
8355 * gimple-range-fold.cc (fold_using_range::range_of_cond_expr): Check
8358 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
8360 PR tree-optimization/101496
8361 * vr-values.c (simplify_using_ranges::fold_cond): Call range_of_stmt
8362 first, then vrp_visit_cond_Stmt.
8364 2021-07-22 liuhongt <hongtao.liu@intel.com>
8366 * config/i386/i386-expand.c
8367 (ix86_broadcast_from_integer_constant): Rename to ..
8368 (ix86_broadcast_from_constant): .. this, and extend it to
8370 (ix86_expand_vector_move): Extend to float mode.
8371 * config/i386/i386-features.c
8372 (replace_constant_pool_with_broadcast): Remove.
8373 (remove_partial_avx_dependency_gate): Ditto.
8374 (constant_pool_broadcast): Ditto.
8375 (class pass_constant_pool_broadcast): Ditto.
8376 (make_pass_constant_pool_broadcast): Ditto.
8377 (remove_partial_avx_dependency): Adjust gate.
8378 * config/i386/i386-passes.def: Remove pass_constant_pool_broadcast.
8379 * config/i386/i386-protos.h
8380 (make_pass_constant_pool_broadcast): Remove.
8382 2021-07-22 liuhongt <hongtao.liu@intel.com>
8384 * config/i386/constraints.md (Wb): New constraint.
8386 * config/i386/i386.md (*ashlhi3_1): Extend to avx512 mask
8388 (*ashlqi3_1): Ditto.
8389 (*<insn><mode>3_1): Split to ..
8390 (*ashr<mode>3_1): this, ...
8391 (*lshr<mode>3_1): and this, also extend this pattern to avx512
8393 (*<insn><mode>3_1): Split to ..
8394 (*ashr<mode>3_1): this, ...
8395 (*lshrqi3_1): and this, also extend this pattern to avx512
8397 (*lshrhi3_1): And this, also extend this pattern to avx512
8399 * config/i386/sse.md (k<code><mode>): New define_split after
8400 it to convert generic shift pattern to mask shift ones.
8402 2021-07-21 Thomas Schwinge <thomas@codesourcery.com>
8403 Joseph Myers <joseph@codesourcery.com>
8404 Cesar Philippidis <cesar@codesourcery.com>
8406 * tree-core.h (omp_clause_code): Add 'OMP_CLAUSE_NOHOST'.
8407 * tree.c (omp_clause_num_ops, omp_clause_code_name, walk_tree_1):
8409 * tree-pretty-print.c (dump_omp_clause): Likewise.
8410 * omp-general.c (oacc_verify_routine_clauses): Likewise.
8411 * gimplify.c (gimplify_scan_omp_clauses)
8412 (gimplify_adjust_omp_clauses): Likewise.
8413 * tree-nested.c (convert_nonlocal_omp_clauses)
8414 (convert_local_omp_clauses): Likewise.
8415 * omp-low.c (scan_sharing_clauses): Likewise.
8416 * omp-offload.c (execute_oacc_device_lower): Update.
8418 2021-07-21 Martin Sebor <msebor@redhat.com>
8420 * tree-ssa-alias.c (walk_aliased_vdefs_1): Fix typos in a comment.
8422 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8424 * config/rs6000/rs6000-gen-builtins.c (write_init_bif_table):
8427 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8429 * config/rs6000/rs6000-gen-builtins.c (write_fntype): New
8431 (write_fntype_init): New stub function.
8432 (write_init_bif_table): Likewise.
8433 (write_init_ovld_table): New function.
8434 (write_init_file): Implement.
8436 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8438 * config/rs6000/rs6000-gen-builtins.c
8439 (write_autogenerated_header): New function.
8440 (write_decls): Likewise.
8441 (write_extern_fntype): New callback function.
8442 (write_header_file): Implement.
8444 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8446 * config/rs6000/rs6000-gen-builtins.c (write_defines_file):
8449 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8451 * config/rs6000/rs6000-gen-builtins.c (complete_vector_type): New
8453 (complete_base_type): Likewise.
8454 (construct_fntype_id): Likewise.
8455 (parse_bif_entry): Call contruct_fntype_id.
8456 (parse_ovld_entry): Likewise.
8458 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8460 * config/rs6000/rs6000-gen-builtins.c (ovld_stanza): New struct.
8461 (MAXOVLDSTANZAS): New macro.
8462 (ovld_stanzas): New variable.
8463 (curr_ovld_stanza): Likewise.
8464 (MAXOVLDS): New macro.
8465 (ovlddata): New struct.
8466 (ovlds): New variable.
8467 (curr_ovld): Likewise.
8468 (max_ovld_args): Likewise.
8469 (parse_ovld_entry): New function.
8470 (parse_ovld_stanza): Likewise.
8471 (parse_ovld): Implement.
8473 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8475 * config/rs6000/rs6000-gen-builtins.c (parse_bif_attrs):
8478 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8480 * config/rs6000/rs6000-gen-builtins.c (parse_args): New function.
8481 (parse_prototype): Implement.
8483 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8485 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): New enum.
8486 (curr_bif_stanza): New variable.
8487 (stanza_entry): New struct.
8488 (stanza_map): New initialized variable.
8489 (enable_string): Likewise.
8490 (fnkinds): New enum.
8491 (typelist): New struct.
8492 (attrinfo): Likewise.
8493 (MAXRESTROPNDS): New macro.
8494 (prototype): New struct.
8495 (MAXBIFS): New macro.
8496 (bifdata): New struct.
8497 (bifs): New variable.
8498 (curr_bif): Likewise.
8499 (bif_order): Likewise.
8500 (bif_index): Likewise.
8501 (fatal): New function.
8502 (stanza_name_to_stanza): Likewise.
8503 (parse_bif_attrs): New stub function.
8504 (parse_prototype): Likewise.
8505 (parse_bif_entry): New function.
8506 (parse_bif_stanza): Likewise.
8507 (parse_bif): Implement.
8508 (set_bif_order): New function.
8509 (create_bif_order): Implement.
8511 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8513 * config/rs6000/rs6000-gen-builtins.c (rbtree.h): New #include.
8514 (num_bifs): New variable.
8515 (num_ovld_stanzas): Likewise.
8516 (num_ovlds): Likewise.
8517 (parse_codes): New enum.
8518 (bif_rbt): New variable.
8519 (ovld_rbt): Likewise.
8520 (fntype_rbt): Likewise.
8521 (bifo_rbt): Likewise.
8522 (parse_bif): New stub function.
8523 (create_bif_order): Likewise.
8524 (parse_ovld): Likewise.
8525 (write_header_file): Likewise.
8526 (write_init_file): Likewise.
8527 (write_defines_file): Likewise.
8528 (delete_output_files): New function.
8531 2021-07-21 H.J. Lu <hjl.tools@gmail.com>
8534 * config/i386/i386-builtin.def: Remove OPTION_MASK_ISA_SSE4_2
8535 from CRC32 _builtin functions.
8537 2021-07-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
8539 * coverage.c (build_gcov_info_var_registration): Mark the object placed
8540 in the linker set as referenced so that it does not get optimized away.
8542 2021-07-21 Kito Cheng <kito.cheng@sifive.com>
8545 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
8547 * config.gcc (riscv*-*-*): Detect which python is available.
8549 2021-07-21 Jakub Jelinek <jakub@redhat.com>
8551 PR middle-end/101535
8552 * gimplify.c (omp_check_private): Properly skip ORT_TARGET_DATA
8553 contexts in which decl isn't privatized and for ORT_TARGET return
8554 false if decl is mapped.
8556 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
8558 * gimple-loop-jam.c: Include tree-ssa-sccvn.h.
8559 (tree_loop_unroll_and_jam): Run value-numbering on a loop that
8560 has been successfully unrolled.
8562 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
8564 * tree-ssa-loop-manip.c (determine_exit_conditions): Return a null
8565 exit condition if no tail loop is needed, and if the original exit
8566 condition should therefore be kept as-is.
8567 (tree_transform_and_unroll_loop): Handle that case here too.
8569 2021-07-21 Kewen Lin <linkw@linux.ibm.com>
8571 * tree-data-ref.c (free_dependence_relations): Adjust to pass vec
8573 (free_data_refs): Likewise.
8574 * tree-data-ref.h (free_dependence_relations): Likewise.
8575 (free_data_refs): Likewise.
8576 * tree-predcom.c (struct chain): Use auto_vec instead of vec for
8578 (struct component): Likewise.
8579 (pcom_worker::pcom_worker): Adjust for auto_vec and renaming changes.
8580 (pcom_worker::~pcom_worker): Likewise.
8581 (pcom_worker::release_chain): Adjust as auto_vec changes.
8582 (pcom_worker::loop): Rename to ...
8583 (pcom_worker::m_loop): ... this.
8584 (pcom_worker::datarefs): Rename to ...
8585 (pcom_worker::m_datarefs): ... this. Use auto_vec instead of vec.
8586 (pcom_worker::dependences): Rename to ...
8587 (pcom_worker::m_dependences): ... this. Use auto_vec instead of vec.
8588 (pcom_worker::chains): Rename to ...
8589 (pcom_worker::m_chains): ... this. Use auto_vec instead of vec.
8590 (pcom_worker::looparound_phis): Rename to ...
8591 (pcom_worker::m_looparound_phis): ... this. Use auto_vec instead of
8593 (pcom_worker::cache): Rename to ...
8594 (pcom_worker::m_cache): ... this. Use auto_vec instead of vec.
8595 (pcom_worker::release_chain): Adjust for auto_vec changes.
8596 (pcom_worker::release_chains): Adjust for auto_vec and renaming
8598 (release_component): Remove.
8599 (release_components): Adjust for release_component removal.
8600 (component_of): Adjust to use vec.
8601 (merge_comps): Likewise.
8602 (pcom_worker::aff_combination_dr_offset): Adjust for renaming changes.
8603 (pcom_worker::determine_offset): Likewise.
8604 (class comp_ptrs): Remove.
8605 (pcom_worker::split_data_refs_to_components): Adjust for renaming
8606 changes, for comp_ptrs removal with auto_vec.
8607 (pcom_worker::suitable_component_p): Adjust for renaming changes.
8608 (pcom_worker::filter_suitable_components): Adjust for release_component
8610 (pcom_worker::valid_initializer_p): Adjust for renaming changes.
8611 (pcom_worker::find_looparound_phi): Likewise.
8612 (pcom_worker::add_looparound_copies): Likewise.
8613 (pcom_worker::determine_roots_comp): Likewise.
8614 (pcom_worker::single_nonlooparound_use): Likewise.
8615 (pcom_worker::execute_pred_commoning_chain): Likewise.
8616 (pcom_worker::execute_pred_commoning): Likewise.
8617 (pcom_worker::try_combine_chains): Likewise.
8618 (pcom_worker::prepare_initializers_chain): Likewise.
8619 (pcom_worker::prepare_initializers): Likewise.
8620 (pcom_worker::prepare_finalizers_chain): Likewise.
8621 (pcom_worker::prepare_finalizers): Likewise.
8622 (pcom_worker::tree_predictive_commoning_loop): Likewise.
8624 2021-07-20 Martin Sebor <msebor@redhat.com>
8626 PR middle-end/101397
8627 * builtins.c (gimple_call_return_array): Add argument. Correct
8628 offsets for memchr, mempcpy, stpcpy, and stpncpy.
8629 (compute_objsize_r): Adjust offset computation for argument returning
8632 2021-07-20 Martin Sebor <msebor@redhat.com>
8634 PR middle-end/101300
8635 * tree-ssa-uninit.c (check_defs): Handle UBSAN built-ins.
8637 2021-07-20 Jeff Law <jlaw@localhost.localdomain>
8639 * function.c (assign_parm_setup_block): Use adjust_address instead
8640 of change_address to preserve MEM_EXPR and friends.
8642 2021-07-20 Martin Sebor <msebor@redhat.com>
8644 * cfgloop.h (single_likely_exit): Adjust by-value argument to
8646 * cfgloopanal.c (single_likely_exit): Same.
8647 * cgraph.h (struct cgraph_node): Same.
8648 * cgraphclones.c (cgraph_node::create_virtual_clone): Same.
8649 * genautomata.c (merge_states): Same.
8650 * genextract.c (VEC_char_to_string): Same.
8651 * genmatch.c (dt_node::gen_kids_1): Same.
8652 (walk_captures): Adjust by-value argument to by-reference.
8653 * gimple-ssa-store-merging.c (check_no_overlap): Adjust by-value argument
8654 to by-const-reference.
8655 * gimple.c (gimple_build_call_vec): Same.
8656 (gimple_build_call_internal_vec): Same.
8657 (gimple_build_switch): Same.
8658 (sort_case_labels): Same.
8659 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
8661 * gimple.h (gimple_build_call_vec): Adjust by-value argument to
8663 (gimple_build_call_internal_vec): Same.
8664 (gimple_build_switch): Same.
8665 (sort_case_labels): Same.
8666 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
8668 * haifa-sched.c (calc_priorities): Adjust by-value argument to
8670 (sched_init_luids): Same.
8671 (haifa_init_h_i_d): Same.
8672 * ipa-cp.c (ipa_get_indirect_edge_target_1): Same.
8673 (adjust_callers_for_value_intersection): Adjust by-value argument to
8675 (find_more_scalar_values_for_callers_subset): Adjust by-value argument to
8677 (find_more_contexts_for_caller_subset): Same.
8678 (find_aggregate_values_for_callers_subset): Same.
8679 (copy_useful_known_contexts): Same.
8680 * ipa-fnsummary.c (remap_edge_summaries): Same.
8681 (remap_freqcounting_predicate): Same.
8682 * ipa-inline.c (add_new_edges_to_heap): Adjust by-value argument to
8684 * ipa-predicate.c (predicate::remap_after_inlining): Adjust by-value argument
8685 to by-const-reference.
8686 * ipa-predicate.h (predicate::remap_after_inlining): Same.
8687 * ipa-prop.c (ipa_find_agg_cst_for_param): Same.
8688 * ipa-prop.h (ipa_find_agg_cst_for_param): Same.
8689 * ira-build.c (ira_loop_tree_body_rev_postorder): Same.
8690 * read-rtl.c (add_overload_instance): Same.
8691 * rtl.h (native_decode_rtx): Same.
8692 (native_decode_vector_rtx): Same.
8693 * sched-int.h (sched_init_luids): Same.
8694 (haifa_init_h_i_d): Same.
8695 * simplify-rtx.c (native_decode_vector_rtx): Same.
8696 (native_decode_rtx): Same.
8697 * tree-call-cdce.c (gen_shrink_wrap_conditions): Same.
8698 (shrink_wrap_one_built_in_call_with_conds): Same.
8699 (shrink_wrap_conditional_dead_built_in_calls): Same.
8700 * tree-data-ref.c (create_runtime_alias_checks): Same.
8701 (compute_all_dependences): Same.
8702 * tree-data-ref.h (compute_all_dependences): Same.
8703 (create_runtime_alias_checks): Same.
8704 (index_in_loop_nest): Same.
8705 * tree-if-conv.c (mask_exists): Same.
8706 * tree-loop-distribution.c (class loop_distribution): Same.
8707 (loop_distribution::create_rdg_vertices): Same.
8708 (dump_rdg_partitions): Same.
8709 (debug_rdg_partitions): Same.
8710 (partition_contains_all_rw): Same.
8711 (loop_distribution::distribute_loop): Same.
8712 * tree-parloops.c (oacc_entry_exit_ok_1): Same.
8713 (oacc_entry_exit_single_gang): Same.
8714 * tree-ssa-loop-im.c (hoist_memory_references): Same.
8715 (loop_suitable_for_sm): Same.
8716 * tree-ssa-loop-niter.c (bound_index): Same.
8717 * tree-ssa-reassoc.c (update_ops): Same.
8718 (swap_ops_for_binary_stmt): Same.
8719 (rewrite_expr_tree): Same.
8720 (rewrite_expr_tree_parallel): Same.
8721 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Same.
8722 * tree-ssa-sccvn.h (ao_ref_init_from_vn_reference): Same.
8723 * tree-ssa-structalias.c (process_all_all_constraints): Same.
8724 (make_constraints_to): Same.
8725 (handle_lhs_call): Same.
8726 (find_func_aliases_for_builtin_call): Same.
8727 (sort_fieldstack): Same.
8728 (check_for_overlaps): Same.
8729 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Same.
8730 (vect_create_cond_for_unequal_addrs): Same.
8731 (vect_create_cond_for_lower_bounds): Same.
8732 (vect_create_cond_for_alias_checks): Same.
8733 * tree-vect-slp-patterns.c (vect_validate_multiplication): Same.
8734 * tree-vect-slp.c (vect_analyze_slp_instance): Same.
8735 (vect_make_slp_decision): Same.
8736 (vect_slp_bbs): Same.
8737 (duplicate_and_interleave): Same.
8738 (vect_transform_slp_perm_load): Same.
8739 (vect_schedule_slp): Same.
8740 * tree-vectorizer.h (vect_transform_slp_perm_load): Same.
8741 (vect_schedule_slp): Same.
8742 (duplicate_and_interleave): Same.
8743 * tree.c (build_vector_from_ctor): Same.
8744 (build_vector): Same.
8745 (check_vector_cst): Same.
8746 (check_vector_cst_duplicate): Same.
8747 (check_vector_cst_fill): Same.
8748 (check_vector_cst_stepped): Same.
8749 * tree.h (build_vector_from_ctor): Same.
8751 2021-07-20 Jakub Jelinek <jakub@redhat.com>
8754 * config/rs6000/rs6000-protos.h (easy_altivec_constant): Change return
8755 type from bool to int.
8756 * config/rs6000/rs6000.c (vspltis_constant): Fix up handling the
8757 EASY_VECTOR_MSB case if either step or copies is not 1.
8758 (vspltis_shifted): Fix comment typo.
8759 (easy_altivec_constant): Change return type from bool to int, instead
8760 of returning true return byte size of the element mode that should be
8761 used to synthetize the constant.
8762 * config/rs6000/predicates.md (easy_vector_constant_msb): Require
8763 that vspltis_shifted is 0, handle the case where easy_altivec_constant
8764 assumes using different vector mode from CONST_VECTOR's mode.
8765 * config/rs6000/altivec.md (easy_vector_constant_msb splitter): Use
8766 easy_altivec_constant to determine mode in which -1 >> -1 should be
8767 performed, use rs6000_expand_vector_init instead of gen_vec_initv4sisi.
8769 2021-07-20 Richard Biener <rguenther@suse.de>
8772 * dwarf2out.h (dwarf_file_data): Add key member.
8773 * dwarf2out.c (dwarf_file_hasher::equal): Compare key.
8774 (dwarf_file_hasher::hash): Hash key.
8775 (lookup_filename): Remap the filename and store it in the
8776 filename member of dwarf_file_data when creating a new
8778 (file_name_acquire): Do not remap the filename again.
8779 (maybe_emit_file): Likewise.
8781 2021-07-20 Jonathan Wright <jonathan.wright@arm.com>
8783 * config/aarch64/aarch64-simd-builtins.def: Use two variant
8784 generators for all TBL/TBX intrinsics and rename to
8785 consistent forms: qtbl[1234] or qtbx[1234].
8786 * config/aarch64/aarch64-simd.md (aarch64_tbl1<mode>):
8788 (aarch64_qtbl1<mode>): This.
8789 (aarch64_tbx1<mode>): Rename to...
8790 (aarch64_qtbx1<mode>): This.
8791 (aarch64_tbl2v16qi): Delete.
8792 (aarch64_tbl3<mode>): Rename to...
8793 (aarch64_qtbl2<mode>): This.
8794 (aarch64_tbx4<mode>): Rename to...
8795 (aarch64_qtbx2<mode>): This.
8796 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_1): Use
8797 renamed qtbl1 and qtbl2 RTL patterns.
8798 * config/aarch64/arm_neon.h (vqtbl1_p8): Use renamed qtbl1
8800 (vqtbl1_s8): Likewise.
8801 (vqtbl1_u8): Likewise.
8802 (vqtbl1q_p8): Likewise.
8803 (vqtbl1q_s8): Likewise.
8804 (vqtbl1q_u8): Likewise.
8805 (vqtbx1_s8): Use renamed qtbx1 RTL pattern.
8806 (vqtbx1_u8): Likewise.
8807 (vqtbx1_p8): Likewise.
8808 (vqtbx1q_s8): Likewise.
8809 (vqtbx1q_u8): Likewise.
8810 (vqtbx1q_p8): Likewise.
8811 (vtbl1_s8): Use renamed qtbl1 RTL pattern.
8812 (vtbl1_u8): Likewise.
8813 (vtbl1_p8): Likewise.
8814 (vtbl2_s8): Likewise
8815 (vtbl2_u8): Likewise.
8816 (vtbl2_p8): Likewise.
8817 (vtbl3_s8): Use renamed qtbl2 RTL pattern.
8818 (vtbl3_u8): Likewise.
8819 (vtbl3_p8): Likewise.
8820 (vtbl4_s8): Likewise.
8821 (vtbl4_u8): Likewise.
8822 (vtbl4_p8): Likewise.
8823 (vtbx2_s8): Use renamed qtbx2 RTL pattern.
8824 (vtbx2_u8): Likewise.
8825 (vtbx2_p8): Likewise.
8826 (vqtbl2_s8): Use renamed qtbl2 RTL pattern.
8827 (vqtbl2_u8): Likewise.
8828 (vqtbl2_p8): Likewise.
8829 (vqtbl2q_s8): Likewise.
8830 (vqtbl2q_u8): Likewise.
8831 (vqtbl2q_p8): Likewise.
8832 (vqtbx2_s8): Use renamed qtbx2 RTL pattern.
8833 (vqtbx2_u8): Likewise.
8834 (vqtbx2_p8): Likewise.
8835 (vqtbx2q_s8): Likewise.
8836 (vqtbx2q_u8): Likewise.
8837 (vqtbx2q_p8): Likewise.
8838 (vtbx4_s8): Likewise.
8839 (vtbx4_u8): Likewise.
8840 (vtbx4_p8): Likewise.
8842 2021-07-20 Uroš Bizjak <ubizjak@gmail.com>
8845 * config/i386/sync.md (define_peephole2 atomic_storedi_fpu):
8847 (define_peephole2 atomic_loaddi_fpu): Ditto.
8849 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
8851 * config.gcc (riscv*-*-*): Detect which python is available.
8853 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
8855 * config/rs6000/vsx.md (mulhs_<mode>): Rename to...
8856 (smul<mode>3_highpart): ... this.
8857 (mulhu_<mode>): Rename to...
8858 (umul<mode>3_highpart): ... this.
8859 * config/rs6000/rs6000-builtin.def (MULHS_V2DI, MULHS_V4SI,
8860 MULHU_V2DI, MULHU_V4SI): Adjust.
8862 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
8864 PR tree-optimization/100696
8865 * internal-fn.c (first_commutative_argument): Add info for IFN_MULH.
8866 * internal-fn.def (IFN_MULH): New internal function.
8867 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Add support to
8868 recog normal multiply highpart as IFN_MULH.
8869 * config/i386/i386.c (ix86_add_stmt_cost): Adjust for combined
8872 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
8874 * config/elfos.h (CTF_DEBUGGING_INFO): New definition.
8875 (BTF_DEBUGGING_INFO): Likewise.
8876 * doc/tm.texi.in: Document the new macros.
8877 * doc/tm.texi: Regenerated.
8878 * toplev.c: Guard initialization of debug hooks.
8880 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
8882 * flags.h (ctf_debuginfo_p): New function declaration.
8883 * opts.c (ctf_debuginfo_p): New function definition.
8885 2021-07-19 Andrew Stubbs <ams@codesourcery.com>
8888 * config/gcn/gcn-hsa.h (DRIVER_SELF_SPECS): New.
8889 (ASM_SPEC): Set -mattr for xnack and sram-ecc.
8890 * config/gcn/gcn-opts.h (enum sram_ecc_type): New.
8891 * config/gcn/gcn-valu.md: Add a warning comment.
8892 * config/gcn/gcn.c (gcn_option_override): Add "sorry" for -mxnack.
8893 (output_file_start): Add xnack and sram-ecc state to ".amdgcn_target".
8894 * config/gcn/gcn.md: Add a warning comment.
8895 * config/gcn/gcn.opt: Add -mxnack and -msram-ecc.
8896 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): Remove
8898 (EF_AMDGPU_XNACK): New.
8899 (EF_AMDGPU_SRAM_ECC): New.
8901 (copy_early_debug_info): Use elf_flags.
8902 (main): Handle -mxnack and -msram-ecc options.
8903 * doc/invoke.texi: Document -mxnack and -msram-ecc.
8905 2021-07-19 Andrew Pinski <apinski@marvell.com>
8908 * config/aarch64/aarch64.md (csneg3_uxtw_insn): Rename to ...
8909 (*cs<neg_not_cs>3_uxtw_insn4): and extend to NEG_NOT.
8911 2021-07-19 Richard Biener <rguenther@suse.de>
8913 PR tree-optimization/101505
8914 * tree-vect-patterns.c (vect_determine_precisions): Walk
8915 PHIs also for loop vectorization.
8917 2021-07-19 Richard Biener <rguenther@suse.de>
8919 * gimple.h (gimple_expr_type): Remove.
8920 * doc/gimple.texi: Remove gimple_expr_type documentation.
8922 2021-07-19 Richard Biener <rguenther@suse.de>
8924 * tree-ssa-sccvn.c (vn_reference_eq): Handle NULL vr->type.
8925 (ao_ref_init_from_vn_reference): Likewise.
8926 (fully_constant_reference): Likewise.
8927 (vn_reference_lookup_call): Do not set vr->type to random
8929 * tree-ssa-pre.c (compute_avail): Do not try to PRE calls
8931 * tree-vect-generic.c (expand_vector_piecewise): Pass in
8932 whether we expanded parallel.
8933 (expand_vector_parallel): Adjust.
8934 (expand_vector_addition): Likewise.
8935 (expand_vector_comparison): Likewise.
8936 (expand_vector_operation): Likewise.
8937 (expand_vector_scalar_condition): Likewise.
8938 (expand_vector_conversion): Likewise.
8940 2021-07-19 Richard Biener <rguenther@suse.de>
8942 * tree-vrp.c (register_edge_assert_for_2): Use the
8944 (vrp_folder::fold_predicate_in): Likewise.
8945 * vr-values.c (gimple_assign_nonzero_p): Likewise.
8946 (vr_values::extract_range_from_comparison): Likewise.
8947 (vr_values::extract_range_from_ubsan_builtin): Use the
8948 type of the first operand.
8949 (vr_values::extract_range_basic): Push down type
8950 computation, use the appropriate LHS.
8951 (vr_values::extract_range_from_assignment): Use the
8954 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
8957 * common/config/i386/i386-common.c (ix86_handle_option): For
8958 -mgeneral-regs-only, enable the GPR only instructions which are
8959 enabled implicitly by SSE ISAs unless they have been disabled
8962 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
8965 * config/i386/i386.c (ix86_check_avx_upper_stores): Moved before
8966 ix86_avx_u128_mode_needed.
8967 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY if callee
8968 returns AVX register.
8970 2021-07-17 Jan Hubicka <hubicka@ucw.cz>
8972 * tree-ssa-structalias.c (handle_rhs_call): Support EAF_NOT_RETURNED.
8973 (handle_const_call): Liekise
8974 (handle_pure_call): Liekise
8976 2021-07-17 Andrew MacLeod <amacleod@redhat.com>
8978 PR tree-optimization/96542
8979 * range-op.cc (range_operator::wi_fold_in_parts): New.
8980 (range_operator::fold_range): Call wi_fold_in_parts.
8981 (operator_lshift::wi_fold): Fix broken lshift by [0,0].
8982 * range-op.h (wi_fold_in_parts): Add prototype.
8984 2021-07-16 David Malcolm <dmalcolm@redhat.com>
8986 * doc/analyzer.texi: Add __analyzer_dump_state.
8988 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
8990 * config/rs6000/rbtree.c: New file.
8991 * config/rs6000/rbtree.h: New file.
8993 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
8995 * config/rs6000/rs6000-gen-builtins.c (restriction): New enum.
8996 (typeinfo): Add restr field.
8997 (match_bracketed_pair): New function.
8998 (match_const_restriction): Implement.
9000 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9002 * config/rs6000/rs6000-gen-builtins.c (match_basetype): Implement.
9004 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9006 * config/rs6000/rs6000-gen-builtins.c (void_status): New enum.
9007 (basetype): Likewise.
9008 (typeinfo): Likewise.
9009 (handle_pointer): New function.
9010 (match_basetype): New stub function.
9011 (match_const_restriction): Likewise.
9012 (match_type): New function.
9014 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9016 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace): New
9018 (advance_line): Likewise.
9019 (safe_inc_pos): Likewise.
9020 (match_identifier): Likewise.
9021 (match_integer): Likewise.
9022 (match_to_right_bracket): Likewise.
9024 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9026 * config/rs6000/rs6000-gen-builtins.c (bif_file): New variable.
9027 (ovld_file): Likewise.
9028 (header_file): Likewise.
9029 (init_file): Likewise.
9030 (defines_file): Likewise.
9031 (pgm_path): Likewise.
9032 (bif_path): Likewise.
9033 (ovld_path): Likewise.
9034 (header_path): Likewise.
9035 (init_path): Likewise.
9036 (defines_path): Likewise.
9037 (LINELEN): New macro.
9038 (linebuf): New variable.
9042 (bif_diag): New function.
9043 (ovld_diag): Likewise.
9045 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9047 * config/rs6000/rs6000-builtin-new.def: New.
9048 * config/rs6000/rs6000-overload.def: New.
9050 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9052 * config/rs6000/rs6000-gen-builtins.c: New.
9054 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9056 * Makefile.in (EXTRA_GTYPE_DEPS): New variable.
9057 (s-gtype): Depend on EXTRA_GTYPE_DEPS.
9058 * gengtype-state.c (state_writer::write_state_file_list): Add a
9059 parameter to the fileslist expression for the number of build
9061 (read_state_files_list): Detect build headers and strip the
9062 initial "./" or ".\" from their names.
9063 * gengtype.c (build_headers): New global variable.
9064 (num_build_headers): Likewise.
9065 (open_base_files): Emit #include for each build header.
9066 (main): Detect and count build headers.
9067 * gengtype.h (build_headers): New extern variable.
9068 (num_build_headers): Likewise.
9070 2021-07-16 Richard Biener <rguenther@suse.de>
9072 * gimple-ssa-store-merging.c (verify_symbolic_number_p): Use
9073 the type of the LHS.
9074 (find_bswap_or_nop_1): Likewise.
9075 (find_bswap_or_nop): Likewise.
9076 * tree-vectorizer.h (vect_get_smallest_scalar_type): Adjust
9078 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
9079 Remove unused parameters, pass in the scalar type. Fix
9080 internal store function handling.
9081 * tree-vect-stmts.c (vect_analyze_stmt): Remove assert.
9082 (vect_get_vector_types_for_stmt): Move down check for
9083 existing vector stmt after we've determined a scalar type.
9084 Pass down the used scalar type to vect_get_smallest_scalar_type.
9085 * tree-vect-generic.c (expand_vector_condition): Use
9086 the type of the LHS.
9087 (expand_vector_scalar_condition): Likewise.
9088 (expand_vector_operations_1): Likewise.
9089 * tree-vect-patterns.c (vect_widened_op_tree): Likewise.
9090 (vect_recog_dot_prod_pattern): Likewise.
9091 (vect_recog_sad_pattern): Likewise.
9092 (vect_recog_widen_op_pattern): Likewise.
9093 (vect_recog_widen_sum_pattern): Likewise.
9094 (vect_recog_mixed_size_cond_pattern): Likewise.
9096 2021-07-16 Jan Hubicka <hubicka@ucw.cz>
9098 * ipa-modref.c (struct escape_entry): Use eaf_fleags_t.
9099 (dump_eaf_flags): Dump EAF_NOT_RETURNED
9100 (eaf_flags_useful_p): Use eaf_fleags_t; handle const functions
9101 and EAF_NOT_RETURNED.
9102 (modref_summary::useful_p): Likewise.
9103 (modref_summary_lto::useful_p): Likewise.
9104 (struct) modref_summary_lto: Use eaf_fleags_t.
9105 (deref_flags): Handle EAF_NOT_RETURNED.
9106 (struct escape_point): Use min_flags.
9107 (modref_lattice::init): Add EAF_NOT_RETURNED.
9108 (merge_call_lhs_flags): Ignore EAF_NOT_RETURNED functions
9109 (analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
9111 (analyze_parms): Also analyze const functions; update conition on
9113 (modref_write): Update streaming.
9114 (read_section): Update streaming.
9115 (remap_arg_flags): Use eaf_flags_t.
9116 (modref_merge_call_site_flags): Hanlde EAF_NOT_RETURNED.
9117 * ipa-modref.h: (eaf_flags_t): New typedef.
9118 (struct modref_summary): Use eaf_flags_t.
9119 * tree-core.h (EAF_NOT_RETURNED): New constant.
9121 2021-07-16 Richard Biener <rguenther@suse.de>
9123 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Use
9124 the type of the LHS.
9125 (gimple_assign_nonnegative_warnv_p): Likewise.
9126 (gimple_call_nonnegative_warnv_p): Likewise. Return false
9127 if the call has no LHS.
9128 * gimple.c (gimple_could_trap_p_1): Use the type of the LHS.
9129 * tree-eh.c (stmt_could_throw_1_p): Likewise.
9130 * tree-inline.c (insert_init_stmt): Likewise.
9131 * tree-ssa-loop-niter.c (get_val_for): Likewise.
9132 * tree-outof-ssa.c (ssa_is_replaceable_p): Use the type of
9134 * tree-ssa-sccvn.c (init_vn_nary_op_from_stmt): Take a
9135 gassign *. Use the type of the lhs.
9136 (vn_nary_op_lookup_stmt): Adjust.
9137 (vn_nary_op_insert_stmt): Likewise.
9139 2021-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
9141 * config/s390/predicates.md (bras_sym_operand): Accept all
9142 functions in 64-bit mode, use UNSPEC_PLT31.
9143 (larl_operand): Use UNSPEC_PLT31.
9144 * config/s390/s390.c (s390_loadrelative_operand_p): Likewise.
9145 (legitimize_pic_address): Likewise.
9146 (s390_emit_tls_call_insn): Mark __tls_get_offset as function,
9148 (s390_delegitimize_address): Use UNSPEC_PLT31.
9149 (s390_output_addr_const_extra): Likewise.
9150 (print_operand): Add @PLT to TLS calls, handle %K.
9151 (s390_function_profiler): Mark __fentry__/_mcount as function,
9152 use %K, use UNSPEC_PLT31.
9153 (s390_output_mi_thunk): Use only UNSPEC_GOT, use %K.
9154 (s390_emit_call): Use UNSPEC_PLT31.
9155 (s390_emit_tpf_eh_return): Mark __tpf_eh_return as function.
9156 * config/s390/s390.md (UNSPEC_PLT31): Rename from UNSPEC_PLT.
9157 (*movdi_64): Use %K.
9158 (reload_base_64): Likewise.
9159 (*sibcall_brc): Likewise.
9160 (*sibcall_brcl): Likewise.
9161 (*sibcall_value_brc): Likewise.
9162 (*sibcall_value_brcl): Likewise.
9165 (*bras_r): Likewise.
9166 (*brasl_r): Likewise.
9167 (*bras_tls): Likewise.
9168 (*brasl_tls): Likewise.
9169 (main_base_64): Likewise.
9170 (reload_base_64): Likewise.
9171 (@split_stack_call<mode>): Likewise.
9173 2021-07-16 Richard Biener <rguenther@suse.de>
9175 PR tree-optimization/101467
9176 * tree-vect-stmts.c (vect_gen_while): Properly guard
9177 make_temp_ssa_name usage.
9179 2021-07-16 Cooper Qu <cooper.qu@linux.alibaba.com>
9181 * config.gcc: Don't use forked print-sysroot-suffix.sh and
9182 t-sysroot-suffix for C-SKY.
9183 * config/csky/print-sysroot-suffix.sh: Delete.
9184 * config/csky/t-csky-linux: Delete.
9185 * config/csky/t-sysroot-suffix: Define MULTILIB_DIRNAMES
9186 instead of CSKY_MULTILIB_DIRNAMES.
9188 2021-07-16 Richard Biener <rguenther@suse.de>
9190 * tree-vect-loop.c (vect_transform_cycle_phi): Correct sign
9191 conversion issues with the partial reduction of the reused
9194 2021-07-16 Richard Biener <rguenther@suse.de>
9196 * config/i386/i386-options.c (ix86_option_override_internal): Set
9197 param_vect_partial_vector_usage to zero if not set.
9199 2021-07-15 Uroš Bizjak <ubizjak@gmail.com>
9202 * config/i386/i386.h (VALID_SSE_REG_MODE): Add TDmode.
9203 (VALID_INT_MODE_P): Add SDmode and DDmode.
9204 Add TDmode for TARGET_64BIT.
9205 (VALID_DFP_MODE_P): Remove.
9206 * config/i386/i386.c (ix86_hard_regno_mode_ok):
9207 Do not use VALID_DFP_MODE_P.
9209 2021-07-15 Andrew MacLeod <amacleod@redhat.com>
9211 * gimple-range-fold.cc (adjust_pointer_diff_expr): Use
9213 (fold_using_range::fold_stmt): Ditto.
9214 (fold_using_range::range_of_range_op): Ditto.
9215 (fold_using_range::range_of_phi): Ditto.
9216 (fold_using_range::range_of_call): Ditto.
9217 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
9218 (fold_using_range::range_of_builtin_call): Ditto.
9219 (fold_using_range::range_of_cond_expr): Ditto.
9220 * gimple-range-fold.h (gimple_range_type): New.
9222 2021-07-15 Martin Sebor <msebor@redhat.com>
9225 * tree-ssa-strlen.c (handle_assign): New function.
9226 (maybe_warn_overflow): Add argument.
9227 (nonzero_bytes_for_type): New function.
9228 (count_nonzero_bytes): Handle more tree types. Call
9229 nonzero_bytes_for_tye.
9230 (count_nonzero_bytes): Handle types.
9231 (handle_store): Handle stores from function calls.
9232 (strlen_check_and_optimize_call): Move code to handle_assign. Call
9233 it for assignments from function calls.
9235 2021-07-15 David Malcolm <dmalcolm@redhat.com>
9240 * doc/invoke.texi: Add -Wanalyzer-use-of-uninitialized-value.
9242 2021-07-15 David Malcolm <dmalcolm@redhat.com>
9244 * doc/invoke.texi (-fdump-analyzer-exploded-paths): New.
9246 2021-07-15 Martin Sebor <msebor@redhat.com>
9250 * fold-const.c (operand_compare::operand_equal_p): Handle OEP_DECL_NAME.
9251 (operand_compare::verify_hash_value): Same.
9252 * tree-core.h (OEP_DECL_NAME): New.
9254 2021-07-15 Martin Jambor <mjambor@suse.cz>
9256 * profile-count.h (profile_count::value): Change the return type to
9258 * gimple-pretty-print.c (dump_gimple_bb_header): Adjust print
9260 * tree-cfg.c (dump_function_to_file): Likewise.
9262 2021-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
9265 * config/rs6000/rs6000-p8swap.c (has_part_mult): New.
9266 (rs6000_analyze_swaps): Insns containing a subreg of a mult are
9269 2021-07-15 Richard Biener <rguenther@suse.de>
9271 * tree-vectorizer.h (vect_gen_while): Match up with
9273 * tree-vect-stmts.c (vect_gen_while): Adjust API to that
9274 of vect_gen_while_not.
9275 (vect_gen_while_not): Adjust.
9276 * tree-vect-loop-manip.c (vect_set_loop_controls_directly): Likewise.
9278 2021-07-15 Aldy Hernandez <aldyh@redhat.com>
9280 * gimple-range-cache.cc (non_null_ref::adjust_range): New.
9281 (ranger_cache::range_of_def): Call adjust_range.
9282 (ranger_cache::entry_range): Same.
9283 * gimple-range-cache.h (non_null_ref::adjust_range): New.
9284 * gimple-range.cc (gimple_ranger::range_of_expr): Call
9286 (gimple_ranger::range_on_entry): Same.
9288 2021-07-15 Tamar Christina <tamar.christina@arm.com>
9291 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9293 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
9295 2021-07-15 Tamar Christina <tamar.christina@arm.com>
9298 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9300 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
9301 (sdot_prod, udot_prod): ...These.
9302 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
9303 (aarch64_<sur>dot<vsi2qi>): Rename to...
9304 (<sur>dot_prod<vsi2qi>): ...This.
9305 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
9308 2021-07-15 Jakub Jelinek <jakub@redhat.com>
9310 PR middle-end/101437
9311 * gimplify.c (gimplify_expr): Throw away volatile reads from empty
9312 types even if they have non-BLKmode TYPE_MODE.
9314 2021-07-15 Richard Biener <rguenther@suse.de>
9317 * gcc.c (process_command): Process -gtoggle like process_options
9318 would after parsing options.
9320 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
9322 * cfgexpand.c (expand_asm_loc): Adjust.
9323 (expand_asm_stmt): Likewise.
9324 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Likewise.
9325 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
9326 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
9327 * config/avr/avr.c (avr_md_asm_adjust): Likewise.
9328 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
9329 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
9330 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
9331 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
9332 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
9333 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
9334 * config/s390/s390.c (s390_md_asm_adjust): Likewise.
9335 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
9336 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
9337 * doc/tm.texi: Regenerate.
9338 * target.def: Add location argument to md_asm_adjust.
9340 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
9342 * tree-diagnostic.c (diagnostic_report_current_function): Use the
9343 diagnostic's location, not input_location.
9345 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
9347 * cfgexpand.c (tree_conflicts_with_clobbers_p): Pass location to
9349 (expand_asm_stmt): Likewise.
9351 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
9353 * config/rs6000/rs6000.c (adjacent_mem_locations): Return the lower
9354 addressed memory rtx, if any.
9355 (rs6000_split_multireg_move): Fix code formatting.
9356 Handle MMA build built-ins with operands in adjacent memory locations.
9358 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
9360 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Move to later
9363 2021-07-14 Jason Merrill <jason@redhat.com>
9365 * sel-sched-ir.h (get_all_loop_exits): Use auto_vec.
9367 2021-07-14 Jason Merrill <jason@redhat.com>
9369 * doc/invoke.texi: -fdelete-dead-exceptions is on by default for
9372 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9374 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
9375 Remove erroneous line.
9377 2021-07-14 Andrew MacLeod <amacleod@redhat.com>
9379 * params.opt (param_evrp_mode): Change default.
9381 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9383 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
9384 (sdot_prod, udot_prod): ...These.
9385 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
9386 (aarch64_<sur>dot<vsi2qi>): Rename to...
9387 (<sur>dot_prod<vsi2qi>): ...This.
9388 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
9391 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9393 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
9395 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9397 * doc/sourcebuild.texi (arm_v8_2a_i8mm_neon_hw): Document.
9399 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9401 * config/arm/neon.md (usdot_prod<vsi2qi>): New.
9403 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9405 * config/aarch64/aarch64-simd.md (aarch64_usdot<vsi2qi>): Rename to...
9406 (usdot_prod<vsi2qi>): ... This.
9407 * config/aarch64/aarch64-simd-builtins.def (usdot): Rename to...
9408 (usdot_prod): ...This.
9409 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Likewise.
9410 * config/aarch64/aarch64-sve.md (@aarch64_<sur>dot_prod<vsi2qi>):
9412 (@<sur>dot_prod<vsi2qi>): ...This.
9413 * config/aarch64/aarch64-sve-builtins-base.cc
9414 (svusdot_impl::expand): Use it.
9416 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9418 * optabs.def (usdot_prod_optab): New.
9419 * doc/md.texi: Document it and clarify other dot prod optabs.
9420 * optabs-tree.h (enum optab_subtype): Add optab_vector_mixed_sign.
9421 * optabs-tree.c (optab_for_tree_code): Support usdot_prod_optab.
9422 * optabs.c (expand_widen_pattern_expr): Likewise.
9423 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
9424 * tree-vect-loop.c (vectorizable_reduction): Query dot-product kind.
9425 * tree-vect-patterns.c (vect_supportable_direct_optab_p): Take optional
9427 (vect_widened_op_tree): Optionally ignore
9429 (vect_recog_dot_prod_pattern): Support usdot_prod_optab.
9431 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
9434 * config/i386/driver-i386.c (host_detect_local_cpu): Check
9435 "arch [32|64]" and "tune [32|64]" for 32-bit and 64-bit codegen.
9436 Enable UINTR only for 64-bit codegen.
9437 * config/i386/i386-options.c
9438 (ix86_option_override_internal::DEF_PTA): Skip PTA_UINTR if not
9440 * config/i386/i386.h (ARCH_ARG): New.
9441 (CC1_CPU_SPEC): Pass "[arch|tune] 32" for 32-bit codegen and
9442 "[arch|tune] 64" for 64-bit codegen.
9444 2021-07-14 Richard Biener <rguenther@suse.de>
9446 PR tree-optimization/101445
9447 * tree-vect-stmts.c (vectorizable_load): Do the gap adjustment
9448 of the IV in the correct direction for negative stride
9451 2021-07-14 Jakub Jelinek <jakub@redhat.com>
9454 * godump.c (godump_str_hash): New type.
9455 (godump_container::pot_dummy_types): Use string_hash instead of
9456 ptr_hash in the hash_set.
9458 2021-07-14 Richard Biener <rguenther@suse.de>
9460 * tree-vect-loop.c (vect_find_reusable_accumulator): Handle
9461 vector types where the old vector type has a multiple of
9462 the new vector type elements.
9463 (vect_create_partial_epilog): New function, split out from...
9464 (vect_create_epilog_for_reduction): ... here.
9465 (vect_transform_cycle_phi): Reduce the re-used accumulator
9466 to the new vector type.
9468 2021-07-14 Alexandre Oliva <oliva@adacore.com>
9470 * tree-ssa-alias.c (attr_fnspec::verify): Fix index in
9471 non-'t'-sized arg check.
9473 2021-07-14 Alexandre Oliva <oliva@adacore.com>
9475 * tree-cfg.c (cleanup_dead_labels_eh): Update
9476 post_landing_pad label upon change of landing pad block's
9478 (cleanup_dead_labels): Check that a removed label is not that
9481 2021-07-13 Jonathan Wright <jonathan.wright@arm.com>
9483 * combine.c (combine_simplify_rtx): Add vec_select -> subreg
9485 * config/aarch64/aarch64.md (*zero_extend<SHORT:mode><GPI:mode>2_aarch64):
9486 Add Neon to general purpose register case for zero-extend
9488 * config/arm/vfp.md (*arm_movsi_vfp): Remove "*" from *t -> r
9489 case to prevent some cases opting to go through memory.
9490 * cse.c (fold_rtx): Add vec_select -> subreg simplification.
9491 * rtl.c (rtvec_series_p): Define predicate to determine
9492 whether a vector contains a linear series of integers.
9493 * rtl.h (rtvec_series_p): Define.
9494 * rtlanal.c (vec_series_lowpart_p): Define predicate to
9495 determine if a vector selection is equivalent to the low part
9497 * rtlanal.h (vec_series_lowpart_p): Define.
9498 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
9499 Add vec_select -> subreg simplification.
9501 2021-07-13 Paul A. Clarke <pc@us.ibm.com>
9503 * config/rs6000/smmintrin.h (_mm_testz_si128, _mm_testc_si128,
9504 _mm_testnzc_si128, _mm_test_all_ones, _mm_test_all_zeros,
9505 _mm_test_mix_ones_zeros): New.
9507 2021-07-13 Roger Sayle <roger@nextmovesoftware.com>
9508 Richard Biener <rguenther@suse.de>
9510 * gimple.c (gimple_could_trap_p_1): Make S argument a
9511 "const gimple*". Preserve constness in call to
9512 gimple_asm_volatile_p.
9513 (gimple_could_trap_p): Make S argument a "const gimple*".
9514 * gimple.h (gimple_could_trap_p_1, gimple_could_trap_p):
9515 Update function prototypes.
9517 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9519 * tree-vectorizer.h (vect_reusable_accumulator): New structure.
9520 (_loop_vec_info::main_loop_edge): New field.
9521 (_loop_vec_info::skip_main_loop_edge): Likewise.
9522 (_loop_vec_info::skip_this_loop_edge): Likewise.
9523 (_loop_vec_info::reusable_accumulators): Likewise.
9524 (_stmt_vec_info::reduc_scalar_results): Likewise.
9525 (_stmt_vec_info::reused_accumulator): Likewise.
9526 (vect_get_main_loop_result): Declare.
9527 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize
9528 reduc_scalar_inputs.
9529 (vec_info::free_stmt_vec_info): Free reduc_scalar_inputs.
9530 * tree-vect-loop-manip.c (vect_get_main_loop_result): New function.
9531 (vect_do_peeling): Fill an epilogue loop's main_loop_edge,
9532 skip_main_loop_edge and skip_this_loop_edge fields.
9533 * tree-vect-loop.c (INCLUDE_ALGORITHM): Define.
9534 (vect_emit_reduction_init_stmts): New function.
9535 (get_initial_def_for_reduction): Use it.
9536 (get_initial_defs_for_reduction): Likewise. Change the vinfo
9537 parameter to a loop_vec_info.
9538 (vect_create_epilog_for_reduction): Store the scalar results
9539 in the reduc_info. If an epilogue loop is reusing an accumulator
9540 from the main loop, and if the epilogue loop can also be skipped,
9541 try to place the reduction code in the join block. Record
9542 accumulators that could potentially be reused by epilogue loops.
9543 (vect_transform_cycle_phi): When vectorizing epilogue loops,
9544 try to reuse accumulators from the main loop. Record the initial
9545 value in reduc_info for non-SLP reductions too.
9547 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9549 * tree-vect-loop.c (get_initial_def_for_reduction): Remove
9550 adjustment handling. Take the neutral value as an argument,
9551 in place of the code argument.
9552 (vect_transform_cycle_phi): Update accordingly. Handle the
9553 initial values of cond reductions separately from code reductions.
9554 Choose the adjustment here rather than in
9555 get_initial_def_for_reduction. Sink the splat of vec_initial_def.
9557 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9559 * tree-vect-loop.c (neutral_op_for_slp_reduction): Replace with...
9560 (neutral_op_for_reduction): ...this, providing a more general
9562 (vect_create_epilog_for_reduction): Update accordingly.
9563 (vectorizable_reduction): Likewise.
9564 (vect_transform_cycle_phi): Likewise.
9566 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9568 * tree-vect-loop.c (get_initial_def_for_reduction): Take the
9569 reduc_info instead of the original stmt_vec_info.
9570 (vect_transform_cycle_phi): Update accordingly.
9572 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9574 * tree-vect-loop.c (get_initial_defs_for_reduction): Take the
9575 reduc_info as an additional parameter.
9576 (vect_transform_cycle_phi): Update accordingly.
9578 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9580 * tree-vectorizer.h: Include tree-ssa-operands.h.
9581 (vect_phi_initial_value): New function.
9582 * tree-vect-loop.c (neutral_op_for_slp_reduction): Use it.
9583 (get_initial_defs_for_reduction, info_for_reduction): Likewise.
9584 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
9585 (vect_transform_cycle_phi, vectorizable_induction): Likewise.
9587 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9589 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
9590 the phi results to vectype after creating them. Remove later
9591 conversion code that thus becomes redundant.
9593 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9595 * tree-vect-loop.c (vect_create_epilog_for_reduction): Replace
9596 the new_phis vector with a reduc_inputs vector. Combine handling
9597 of reduction chains and ncopies > 1.
9599 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9601 * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
9602 scalar_results to group_size elements after reducing down from
9603 N*group_size elements. Construct an array_slice of the live-out
9604 stmts and assert that there is one stmt per scalar result.
9606 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9608 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
9609 nested_in_vect_loop and use double_reduc everywhere. Remove dead
9610 assignment to "loop".
9612 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9614 * internal-fn.c (vectorized_internal_fn_supported_p): Handle
9615 vector types first. For scalar types, consider both the preferred
9616 vector mode and the alternative vector modes.
9617 * optabs-query.c (can_vec_mask_load_store_p): Use the same
9618 structure as above, in particular using related_vector_mode
9619 for modes provided by autovectorize_vector_modes.
9621 2021-07-13 Jakub Jelinek <jakub@redhat.com>
9622 Richard Biener <rguenther@suse.de>
9624 PR tree-optimization/101419
9625 * tree-pass.h (PROP_objsz): Define.
9626 (make_pass_early_object_sizes): Declare.
9627 * passes.def (pass_all_early_optimizations): Rename pass_object_sizes
9628 there to pass_early_object_sizes, drop parameter.
9629 (pass_all_optimizations): Move pass_object_sizes right after pass_ccp,
9630 drop parameter, move pass_post_ipa_warn right after that.
9631 * tree-object-size.c (pass_object_sizes::execute): Rename to...
9632 (object_sizes_execute): ... this. Add insert_min_max_p argument.
9633 (pass_data_object_sizes): Move after object_sizes_execute.
9634 (pass_object_sizes): Likewise. In execute method call
9635 object_sizes_execute, drop set_pass_param method and insert_min_max_p
9636 non-static data member and its initializer in the ctor.
9637 (pass_data_early_object_sizes, pass_early_object_sizes,
9638 make_pass_early_object_sizes): New.
9639 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Use
9640 (cfun->curr_properties & PROP_objsz) instead of cfun->after_inlining.
9642 2021-07-13 Kito Cheng <kito.cheng@sifive.com>
9645 * config/riscv/constraints.md ("S"): Update description and remove
9647 * doc/md.texi (Machine Constraints): Document the 'S' constraints
9650 2021-07-13 Richard Biener <rguenther@suse.de>
9653 2021-07-12 Richard Biener <rguenther@suse.de>
9655 * tree-vect-slp.c (vect_slp_region): Show the number of
9656 SLP graph entries in the optimization message.
9658 2021-07-13 Michael Meissner <meissner@linux.ibm.com>
9660 * config/rs6000/altivec.md (xxspltiw_v4sf): Change local variable
9662 * config/rs6000/rs6000-protos.h (rs6000_const_f32_to_i32): Change
9663 return type to long.
9664 * config/rs6000/rs6000.c (rs6000_const_f32_to_i32): Change return
9667 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
9669 * gimple-range-fold.cc (fold_using_range::range_of_builtin_ubsan_call):
9670 Query relation between the 2 operands and use it.
9672 2021-07-12 Sergei Trofimovich <siarheit@google.com>
9674 * doc/cfg.texi: Fix s/ei_safe_safe/ei_safe_edge/ typo.
9676 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
9679 * config/i386/predicates.md (vec_setm_sse41_operand):
9680 Rename from vec_setm_operand.
9681 (vec_setm_avx2_operand): New predicate.
9682 * config/i386/sse.md (vec_set<V_128:mode>): Use V_128 mode iterator.
9683 Use vec_setm_sse41_operand as operand 2 predicate.
9684 (vec_set<V_256_512:mode): New expander.
9685 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_sse41_operand
9686 as operand 2 predicate.
9688 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
9690 PR tree-optimization/101335
9691 * range-op.cc (operator_cast::lhs_op1_relation): Delete.
9693 2021-07-12 Andrew Pinski <apinski@marvell.com>
9695 * tree-ssa-phiopt.c (match_simplify_replacement): Move
9696 insert of the sequence before the movement of the
9697 statement. Check if to see if the statement is used
9698 outside of the original phi to see if we should move it.
9700 2021-07-12 Richard Biener <rguenther@suse.de>
9702 * dump-context.h (debug_dump_context::debug_dump_context):
9703 Add FILE * parameter defaulted to stderr.
9704 * dumpfile.c (debug_dump_context::debug_dump_context): Adjust.
9705 * tree-vect-slp.c (dot_slp_tree): New functions.
9707 2021-07-12 Richard Biener <rguenther@suse.de>
9709 PR tree-optimization/101373
9710 * tree-ssa-pre.c (prune_clobbered_mems): Also prune trapping
9711 references when the BB may not return.
9712 (compute_avail): Pass in the function we're working on and
9713 replace cfun references with it. Externally throwing
9714 const calls also possibly terminate the function.
9715 (pass_pre::execute): Pass down the function we're working on.
9716 * gcse.c (compute_hash_table_work): Externally throwing
9717 const/pure calls also need record_last_mem_set_info.
9718 * postreload-gcse.c (record_opr_changes): Looping or externally
9719 throwing const/pure calls also need record_last_mem_set_info.
9721 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
9723 * recog.c (memory_address_addr_space_p): Change the type to bool.
9724 Return true/false instead of 1/0.
9725 (offsettable_memref_p): Ditto.
9726 (offsettable_nonstrict_memref_p): Ditto.
9727 (offsettable_address_addr_space_p): Ditto.
9728 Change the type of addressp indirect function to bool.
9729 * recog.h (memory_address_addr_space_p): Change the type to bool.
9730 (strict_memory_address_addr_space_p): Ditto.
9731 (offsettable_memref_p): Ditto.
9732 (offsettable_nonstrict_memref_p): Ditto.
9733 (offsettable_address_addr_space_p): Ditto.
9734 * reload.c (maybe_memory_address_addr_space_p): Ditto.
9735 (strict_memory_address_addr_space_p): Change the type to bool.
9736 Return true/false instead of 1/0.
9737 (maybe_memory_address_addr_space_p): Change the type to bool.
9739 2021-07-12 Richard Biener <rguenther@suse.de>
9741 * tree-vect-slp.c (vect_slp_region): Show the number of
9742 SLP graph entries in the optimization message.
9744 2021-07-12 Richard Biener <rguenther@suse.de>
9746 PR tree-optimization/101394
9747 * tree-ssa-pre.c (do_pre_regular_insertion): Avoid inserting
9748 copies from abnormals for a full redundancy.
9750 2021-07-12 Richard Biener <rguenther@suse.de>
9752 PR middle-end/101423
9753 * gimple.c (gimple_could_trap_p_1): Internal function calls
9755 * tree-eh.c (tree_could_trap_p): Likewise.
9757 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
9760 * config/arm/arm_neon.h (vmul_n_u32): Replace call to builtin with
9762 (vmulq_n_u32): Likewise.
9763 (vmul_n_f32): Gate __a * __b on __FAST_MATH__.
9764 (vmulq_n_f32): Likewise.
9765 (vmul_n_f16): Likewise.
9766 (vmulq_n_f16): Likewise.
9768 2021-07-12 Martin Liska <mliska@suse.cz>
9771 * gcc.c (check_offload_target_name): Call
9772 candidates_list_and_hint only if we have a candidate.
9774 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
9777 * config/arm/neon.md (vec_init): Move to ...
9778 * config/arm/vec-common.md (vec_init): ... here.
9779 Change the pattern's mode to VDQX and gate it on VALID_MVE_MODE.
9781 2021-07-12 Roger Sayle <roger@nextmovesoftware.com>
9783 PR tree-optimization/101403
9784 * match.pd ((T)bswap(X)>>C): Correctly handle cases where
9785 signedness of the shift is not the same as the signedness of
9788 2021-07-09 Roger Sayle <roger@nextmovesoftware.com>
9789 Uroš Bizjak <ubizjak@gmail.com>
9791 * config/i386/i386.md (*divmodsi4_const): Optimize SImode
9792 divmod of a constant numerator with new define_insn_and_split.
9794 2021-07-09 Iain Sandoe <iain@sandoe.co.uk>
9797 * config/i386/i386-expand.c (ix86_expand_call): If a call is
9798 to a non-local-binding, or local but to a public symbol, then
9799 assume that it might be indirected via the lazy symbol binder.
9800 Mark R10 and R10 as clobbered in that case.
9802 2021-07-09 Eric Botcazou <ebotcazou@adacore.com>
9805 * gcc.c (ASM_DEBUG_DWARF_OPTION): Set again to --gdwarf2 in
9806 the case where HAVE_AS_WORKING_DWARF_N_FLAG is not defined
9807 and HAVE_LD_BROKEN_PE_DWARF5 is defined.
9809 2021-07-09 Uroš Bizjak <ubizjak@gmail.com>
9811 * config/i386/i386.md (*udivmodsi4_pow2_zext_1): Limit the
9812 log2 range of operands[3] to [1,31].
9813 (*udivmodsi4_pow2_zext_2): Ditto. Correct insn RTX pattern.
9815 2021-07-09 Sergei Trofimovich <siarheit@google.com>
9817 * doc/md.texi: Don't split @smallexample in multiple @groups.
9819 2021-07-09 Sergei Trofimovich <siarheit@google.com>
9821 * doc/md.texi: Add missing 'see' word.
9823 2021-07-09 Andrew Pinski <apinski@marvell.com>
9825 * tree-ssa-phiopt.c (phiopt_early_allow): Change arguments
9826 to take sequence and gimple_match_op. Accept the case where
9827 op is a SSA_NAME and one statement in the sequence.
9828 Also allow constants.
9829 (gimple_simplify_phiopt): Always pass a sequence to resimplify.
9830 Update call to phiopt_early_allow. Discard the sequence if not
9833 2021-07-09 Xi Ruoyao <xry111@mengyan1223.wang>
9838 * config/mips/mips.c (mips_const_insns): Use MSA_SUPPORTED_MODE_P
9839 instead of ISA_HAS_MSA.
9840 (mips_expand_vec_unpack): Likewise.
9841 (mips_expand_vector_init): Likewise.
9843 2021-07-09 Kewen Lin <linkw@linux.ibm.com>
9845 * config/rs6000/vsx.md (mods_<mode>): Rename to...
9846 (mod<mode>3): ... this.
9847 (modu_<mode>): Rename to...
9848 (umod<mode>3): ... this.
9849 * config/rs6000/rs6000-builtin.def (MODS_V2DI, MODS_V4SI, MODU_V2DI,
9852 2021-07-08 Jeff Law <jeffreyalaw@gmail.com>
9854 * config/h8300/shiftrotate.md (variable shifts): Expose condition
9855 code handling for the test before the loop.
9857 2021-07-08 Martin Jambor <mjambor@suse.cz>
9860 * ipa-sra.c (class isra_call_summary): New member
9861 m_before_any_store, initialize it in the constructor.
9862 (isra_call_summary::dump): Dump the new field.
9863 (ipa_sra_call_summaries::duplicate): Copy it.
9864 (process_scan_results): Set it.
9865 (isra_write_edge_summary): Stream it.
9866 (isra_read_edge_summary): Likewise.
9867 (param_splitting_across_edge): Only override
9868 safe_to_import_accesses if m_before_any_store is set.
9870 2021-07-08 Martin Sebor <msebor@redhat.com>
9873 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref):
9874 Use Object Size Type 0 instead of 1.
9876 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
9878 * tree-vect-loop.c (vectorizable_reduction): Remove always-true
9881 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
9883 * match.pd: Simplify an extend-operate-truncate sequence involving
9886 2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
9887 Richard Biener <rguenther@suse.de>
9889 PR tree-optimization/40210
9890 * match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
9891 (x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
9892 when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
9894 2021-07-08 Uroš Bizjak <ubizjak@gmail.com>
9897 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
9899 * config/i386/mmx.md (V_32): New mode iterator.
9900 (mov<V_32:mode>): Use V_32 mode iterator.
9901 (*mov<V_32:mode>_internal): Ditto.
9902 (*push<V_32:mode>2_rex64): Ditto.
9903 (*push<V_32:mode>2): Ditto.
9904 (movmisalign<V_32:mode>): Ditto.
9905 (mmx_<any_shiftrt:insn>v1si3): New insn pattern.
9906 (sse4_1_<any_extend:code>v2qiv2hi2): Ditto.
9907 (vec_unpacks_lo_v4qi): New expander.
9908 (vec_unpacks_hi_v4qi): Ditto.
9909 (vec_unpacku_lo_v4qi): Ditto.
9910 (vec_unpacku_hi_v4qi): Ditto.
9911 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add V1SImode.
9912 (VALID_INT_MODE_P): Ditto.
9914 2021-07-08 Michael Meissner <meissner@linux.ibm.com>
9917 * config/rs6000/rs6000.md (udivti3): New insn.
9919 (umodti3): New insn.
9922 2021-07-07 Martin Sebor <msebor@redhat.com>
9924 PR tree-optimization/100137
9925 PR tree-optimization/99121
9926 PR tree-optimization/97027
9927 * builtins.c (access_ref::access_ref): Also set offmax.
9928 (access_ref::offset_in_range): Define new function.
9929 (access_ref::add_offset): Set offmax.
9930 (access_ref::inform_access): Handle access_none.
9931 (handle_mem_ref): Clear ostype.
9932 (compute_objsize_r): Handle ASSERT_EXPR.
9933 * builtins.h (struct access_ref): Add offmax member.
9934 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Use
9935 compute_objsize() and simplify.
9937 2021-07-07 Peter Bergner <bergner@linux.ibm.com>
9939 * config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
9940 and VSX_BUILTIN_STXVP.
9942 2021-07-07 Martin Sebor <msebor@redhat.com>
9945 * config/aarch64/aarch64.c (aarch64_simd_lane_bounds): Remove
9946 a stray %K from error_at() missed in r12-2088.
9948 2021-07-07 Richard Biener <rguenther@suse.de>
9950 PR tree-optimization/99728
9951 * tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
9953 (mem_refs_may_alias_p): Add assert we handled aggregate
9955 (sm_seq_valid_bb): Give up when running into aggregate copies.
9956 (ref_indep_loop_p): Handle aggregate copies as never
9957 being invariant themselves but allow other refs to be
9958 disambiguated against them.
9959 (can_sm_ref_p): Do not try to apply store-motion to aggregate
9962 2021-07-06 Indu Bhagat <indu.bhagat@oracle.com>
9965 * dwarf2ctf.c (ctf_get_AT_data_member_location): Multiply by 8 to get
9968 2021-07-06 Martin Sebor <msebor@redhat.com>
9970 * gimple-pretty-print.c (percent_G_format): Remove.
9971 * tree-diagnostic.c (default_tree_printer): Remove calls.
9972 * tree-pretty-print.c (percent_K_format): Remove.
9973 * tree-pretty-print.h (percent_K_format): Remove.
9975 2021-07-06 Martin Sebor <msebor@redhat.com>
9977 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_builtin):
9978 Remove %K and use error_at.
9979 (aarch64_expand_fcmla_builtin): Same.
9980 (aarch64_expand_builtin_tme): Same.
9981 (aarch64_expand_builtin_memtag): Same.
9982 * config/arm/arm-builtins.c (arm_expand_acle_builtin): Same.
9983 (arm_expand_builtin): Same.
9984 * config/arm/arm.c (bounds_check): Same.
9986 2021-07-06 Martin Sebor <msebor@redhat.com>
9988 * builtins.c (warn_string_no_nul): Remove %G.
9989 (maybe_warn_for_bound): Same.
9990 (warn_for_access): Same.
9991 (check_access): Same.
9992 (check_strncat_sizes): Same.
9993 (expand_builtin_strncat): Same.
9994 (expand_builtin_strncmp): Same.
9995 (expand_builtin): Same.
9996 (expand_builtin_object_size): Same.
9997 (warn_dealloc_offset): Same.
9998 (maybe_emit_free_warning): Same.
9999 * calls.c (maybe_warn_alloc_args_overflow): Same.
10000 (maybe_warn_nonstring_arg): Same.
10001 (maybe_warn_rdwr_sizes): Same.
10002 * expr.c (expand_expr_real_1): Remove %K.
10003 * gimple-fold.c (gimple_fold_builtin_strncpy): Remove %G.
10004 (gimple_fold_builtin_strncat): Same.
10005 * gimple-ssa-sprintf.c (format_directive): Same.
10006 (handle_printf_call): Same.
10007 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Same.
10008 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
10009 (maybe_diag_access_bounds): Same. Call gimple_location.
10010 (check_bounds_or_overlap): Same.
10011 * trans-mem.c (ipa_tm_scan_irr_block): Remove %K. Simplify.
10012 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Remove %G.
10013 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
10014 (maybe_diag_stxncpy_trunc): Same.
10015 (handle_builtin_stxncpy_strncat): Same.
10016 (maybe_warn_pointless_strcmp): Same.
10017 * tree-ssa-uninit.c (maybe_warn_operand): Same.
10019 2021-07-06 Uroš Bizjak <ubizjak@gmail.com>
10022 * config/i386/predicates.md (vec_setm_operand): Enable
10023 register_operand for TARGET_SSE4_1.
10024 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_operand
10025 as operand 2 predicate. Call ix86_expand_vector_set_var
10026 for non-constant index operand.
10027 (vec_setv4qi): Use vec_setm_mmx_operand as operand 2 predicate.
10028 Call ix86_expand_vector_set_var for non-constant index operand.
10030 2021-07-06 Jeff Law <jeffreyalaw@gmail.com>
10032 * config/h8300/jumpcall.md (*branch): When possible, generate
10033 the comparison in CCZN mode.
10034 * config/h8300/predicates.md (simple_memory_operand): Reject all
10035 auto-increment addressing modes.
10037 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
10039 PR bootstrap/100246
10040 * config/i386/i386.h (struct stringop_algs): Define a CTOR for
10043 2021-07-06 Richard Biener <rguenther@suse.de>
10045 * doc/md.texi (vec_fmaddsub<mode>4): Document.
10046 (vec_fmsubadd<mode>4): Likewise.
10047 * optabs.def (vec_fmaddsub$a4): Add.
10048 (vec_fmsubadd$a4): Likewise.
10049 * internal-fn.def (IFN_VEC_FMADDSUB): Add.
10050 (IFN_VEC_FMSUBADD): Likewise.
10051 * tree-vect-slp-patterns.c (addsub_pattern::recognize):
10052 Refactor to handle IFN_VEC_FMADDSUB and IFN_VEC_FMSUBADD.
10053 (addsub_pattern::build): Likewise.
10054 * tree-vect-slp.c (vect_optimize_slp): CFN_VEC_FMADDSUB
10055 and CFN_VEC_FMSUBADD are not transparent for permutes.
10056 * config/i386/sse.md (vec_fmaddsub<mode>4): New expander.
10057 (vec_fmsubadd<mode>4): Likewise.
10059 2021-07-06 Richard Biener <rguenther@suse.de>
10061 * doc/invoke.texi (fmove-loop-stores): Document.
10062 * common.opt (fmove-loop-stores): New option.
10063 * opts.c (default_options_table): Enable -fmove-loop-stores
10064 at -O1 but not -Og.
10065 * tree-ssa-loop-im.c (pass_lim::execute): Pass
10066 flag_move_loop_stores instead of true to
10067 loop_invariant_motion_in_fun.
10069 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
10071 * doc/install.texi: Document --with-dsymutil.
10073 2021-07-06 Andrew Pinski <apinski@marvell.com>
10075 PR tree-optimization/101256
10076 * dbgcnt.def (phiopt_edge_range): New counter.
10077 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
10078 Check to make sure the new name is defined in the same
10079 bb as the conditional before duplicating range info.
10080 Also add debug counter.
10082 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
10084 PR rtl-optimization/100328
10085 * config/i386/i386-options.c (ix86_option_override_internal):
10086 Set param_ira_consider_dup_in_all_alts to 0.
10088 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
10090 PR rtl-optimization/100328
10091 * doc/invoke.texi (ira-consider-dup-in-all-alts): Document new
10093 * ira.c (ira_get_dup_out_num): Adjust as parameter
10094 param_ira_consider_dup_in_all_alts.
10095 * params.opt (ira-consider-dup-in-all-alts): New.
10096 * ira-conflicts.c (process_regs_for_copy): Add one parameter
10097 single_input_op_has_cstr_p.
10098 (get_freq_for_shuffle_copy): New function.
10099 (add_insn_allocno_copies): Adjust as single_input_op_has_cstr_p.
10100 * ira-int.h (ira_get_dup_out_num): Add one bool parameter.
10102 2021-07-05 Jeff Law <jeffreyalaw@gmail.com>
10104 * config/h8300/shiftrotate.md (shift-by-variable patterns): Update to
10105 generate condition code aware RTL directly.
10107 2021-07-05 Andrew Pinski <apinski@marvell.com>
10109 PR tree-optimization/101039
10110 * match.pd (A CMP 0 ? A : -A): New patterns.
10111 * tree-ssa-phiopt.c (abs_replacement): Delete function.
10112 (tree_ssa_phiopt_worker): Don't call abs_replacement.
10113 Update comment about abs_replacement.
10115 2021-07-05 Andrew Pinski <apinski@marvell.com>
10117 * tree-ssa-phiopt.c (gimple_simplify_phiopt):
10118 If "A ? B : C" fails to simplify, try "(!A) ? C : B".
10120 2021-07-05 Andrew Pinski <apinski@marvell.com>
10122 * tree-ssa-phiopt.c (match_simplify_replacement):
10123 Add early_p argument. Call gimple_simplify_phiopt
10124 instead of gimple_simplify.
10125 (tree_ssa_phiopt_worker): Update call to
10126 match_simplify_replacement and allow unconditionally.
10127 (phiopt_early_allow): New function.
10128 (gimple_simplify_phiopt): New function.
10130 2021-07-05 Andrew Pinski <apinski@marvell.com>
10132 PR middle-end/101237
10133 * fold-const.c (negate_expr_p): Remove call to element_mode
10134 and TREE_MODE/TREE_TYPE when calling HONOR_SIGNED_ZEROS,
10135 HONOR_SIGN_DEPENDENT_ROUNDING, and HONOR_SNANS.
10136 (fold_negate_expr_1): Likewise.
10137 (const_unop): Likewise.
10138 (fold_cond_expr_with_comparison): Likewise.
10139 (fold_binary_loc): Likewise.
10140 (fold_ternary_loc): Likewise.
10141 (tree_call_nonnegative_warnv_p): Likewise.
10142 * match.pd (-(A + B) -> (-B) - A): Likewise.
10144 2021-07-05 Iain Sandoe <iain@sandoe.co.uk>
10146 * configure.ac: Handle --with-dsymutil in the same way as we
10147 do for the assembler and linker. (DEFAULT_DSYMUTIL): New.
10148 Extract the type and version for the dsymutil configured or
10149 found by the default searches.
10150 * config.in: Regenerated.
10151 * configure: Regenerated.
10152 * collect2.c (do_dsymutil): Handle locating dsymutil in the
10153 same way as for the assembler and linker.
10154 * config/darwin.h (DSYMUTIL): Delete.
10155 * gcc.c: Report a configured dsymutil correctly.
10156 * exec-tool.in: Allow for dsymutil.
10158 2021-07-05 Uroš Bizjak <ubizjak@gmail.com>
10160 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
10161 Handle V4QI and V2HI modes.
10162 (expand_vec_perm_blend): Allow 4-byte vector modes with TARGET_SSE4_1.
10163 Handle V4QI mode. Emit mmx_pblendvb32 for 4-byte modes.
10164 (expand_vec_perm_pshufb): Rewrite to use switch statemets.
10165 Handle 4-byte dual operands with TARGET_XOP and single operands
10166 with TARGET_SSSE3. Emit mmx_ppermv32 for TARGET_XOP and
10167 mmx_pshufbv4qi3 for TARGET_SSSE3.
10168 (expand_vec_perm_pblendv): Allow 4-byte vector modes with TARGET_SSE4_1.
10169 (expand_vec_perm_interleave2): Allow 4-byte vector modes.
10170 (expand_vec_perm_pshufb2): Allow 4-byte vector modes with TARGET_SSSE3.
10171 (expand_vec_perm_even_odd_1): Handle V4QI mode.
10172 (expand_vec_perm_broadcast_1): Handle V4QI mode.
10173 (ix86_vectorize_vec_perm_const): Handle V4QI mode.
10174 * config/i386/mmx.md (mmx_ppermv32): New insn pattern.
10175 (mmx_pshufbv4qi3): Ditto.
10176 (*mmx_pblendw32): Ditto.
10177 (*mmx_pblendw64): Rename from *mmx_pblendw.
10178 (mmx_punpckhbw_low): New insn_and_split pattern.
10179 (mmx_punpcklbw_low): Ditto.
10181 2021-07-05 Richard Biener <rguenther@suse.de>
10183 * tree-vect-loop-manip.c (vect_loop_versioning): Do not
10184 set LOOP_C_INFINITE on the vectorized loop.
10186 2021-07-05 Richard Biener <rguenther@suse.de>
10188 PR middle-end/101291
10189 * cfgloopmanip.c (loop_version): Set the loop copy of the
10190 versioned loop to the new loop.
10192 2021-07-04 Iain Sandoe <iain@sandoe.co.uk>
10195 * config.gcc: Ensure that Darwin biarch definitions are
10196 added before i386.h.
10197 * config/i386/darwin.h (TARGET_64BIT): Remove.
10198 (PR80556_WORKAROUND): New.
10199 (REAL_LIBGCC_SPEC): Amend to use PR80556_WORKAROUND.
10200 (DARWIN_SUBARCH_SPEC): New.
10201 * config/i386/darwin32-biarch.h (TARGET_64BIT_DEFAULT,
10202 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
10203 (REAL_LIBGCC_SPEC): Remove.
10204 * config/i386/darwin64-biarch.h (TARGET_64BIT_DEFAULT,
10205 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
10206 (REAL_LIBGCC_SPEC): Remove.
10208 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
10210 PR middle-end/101294
10211 * expr.c (store_constructor): Don't use vec_duplicate on vector.
10213 2021-07-02 Martin Sebor <msebor@redhat.com>
10215 PR middle-end/98871
10216 PR middle-end/98512
10217 * diagnostic.c (get_any_inlining_info): New.
10218 (update_effective_level_from_pragmas): Handle inlining context.
10219 (diagnostic_enabled): Same.
10220 (diagnostic_report_diagnostic): Same.
10221 * diagnostic.h (struct diagnostic_info): Add ctor.
10222 (struct diagnostic_context): Add new member.
10223 * tree-diagnostic.c (set_inlining_locations): New.
10224 (tree_diagnostics_defaults): Set new callback pointer.
10226 2021-07-02 Peter Bergner <bergner@linux.ibm.com>
10228 * config/rs6000/rs6000-builtin.def (BU_MMA_PAIR_LD, BU_MMA_PAIR_ST):
10230 (__builtin_vsx_lxvp, __builtin_vsx_stxvp): New built-ins.
10231 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Expand
10232 lxvp and stxvp built-ins.
10233 (mma_init_builtins): Handle lxvp and stxvp built-ins.
10234 (builtin_function_type): Likewise.
10235 * doc/extend.texi (__builtin_vsx_lxvp, __builtin_mma_stxvp): Document.
10237 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10239 * config/h8300/h8300-protos.h (compute_a_shift_cc): Accept
10240 additional argument for the code.
10241 * config/h8300/h8300.c (compute_a_shift_cc): Accept additional
10242 argument for the code. Just return if the ZN bits are useful or
10243 not rather than the old style CC_* enums.
10244 * config/h8300/shiftrotate.md (shiftqi_noscratch): Move before
10245 more generic shiftqi patterns.
10246 (shifthi_noscratch, shiftsi_noscratch): Similarly.
10247 (shiftqi_noscratch_set_flags): New pattern.
10248 (shifthi_noscratch_set_flags, shiftsi_noscratch_set_flags): Likewise.
10250 2021-07-02 Andrew MacLeod <amacleod@redhat.com>
10252 PR tree-optimization/101223
10253 * range-op.cc (build_lt): Add -1 for signed values.
10254 (built_gt): Subtract -1 for signed values.
10256 2021-07-02 David Faust <david.faust@oracle.com>
10258 * btfout.c (get_btf_kind): Support BTF_KIND_FLOAT.
10259 (btf_asm_type): Likewise.
10261 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10263 * config/h8300/h8300-protos.h (output_a_shift): Make first argument
10264 an array of rtx rather than a pointer to rtx. Add code argument.
10265 (compute_a_shift_length): Similarly.
10266 * config/h8300/h8300.c (h8300_shift_costs): Adjust now that the
10267 shift itself isn't an operand. Create dummy operand[0] to carry
10268 a mode and pass a suitable rtx code to compute_a_shift_length.
10269 (get_shift_alg): Adjust operand number of clobber in output templates.
10270 (output_a_shift): Make first argument an array of rtx rather than
10271 a pointer to rtx. Add code argument for the type of shift.
10272 Adjust now that the shift itself is no longer an operand.
10273 (compute_a_shift_length): Similarly.
10274 * config/h8300/shiftrotate.md (shiftqi, shifthi, shiftsi): Use an
10275 iterator rather than nshift_operator.
10276 (shiftqi_noscratch, shifthi_noscratch, shiftsi_noscratch): Likewise.
10277 (shiftqi_clobber_flags): Adjust to API changes in output_a_shift
10278 and compute_a_shift_length.
10279 (shiftqi_noscratch_clobber_flags): Likewise.
10280 (shifthi_noscratch_clobber_flags): Likewise.
10281 (shiftsi_noscratch_clobber_flags): Likewise.
10283 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
10286 * config/darwin.h (DSYMUTIL_SPEC): Do not try to run
10287 dsymutil for BTF/CTF.
10289 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
10292 * config/darwin.h (CTF_INFO_SECTION_NAME): Update the
10293 segment to include BTF.
10294 (BTF_INFO_SECTION_NAME): New.
10296 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10298 * config/m32r/m32r-protos.h (call_operand): Adjust return type.
10299 (small_data_operand, memreg_operand, small_insn_p): Likewise.
10300 * config/m32r/m32r.c (call_operand): Adjust return type.
10301 (small_data_operand, memreg_operand): Likewise.
10303 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10305 * config/frv/frv-protos.h (integer_register_operand): Adjust return
10307 (frv_load_operand, gpr_or_fpr_operand, gpr_no_subreg_operand): Likewise.
10308 (fpr_or_int6_operand, gpr_or_int_operand); Likewise.
10309 (gpr_or_int12_operand, gpr_or_int10_operand); Likewise.
10310 (move_source_operand, move_destination_operand): Likewise.
10311 (condexec_source_operand, condexec_dest_operand): Likewise.
10312 (lr_operand, gpr_or_memory_operand, fpr_or_memory_operand): Likewise.
10313 (reg_or_0_operand, fcc_operand, icc_operand, cc_operand): Likewise.
10314 (fcr_operand, icr_operand, cr_operand, call_operand): Likewise.
10315 (fpr_operand, even_reg_operand, odd_reg_operand): Likewise.
10316 (even_gpr_operand, odd_gpr_operand, quad_fpr_operand): Likewise.
10317 (even_fpr_operand, odd_fpr_operand): Likewise.
10318 (dbl_memory_one_insn_operand, dbl_memory_two_insn_operand): Likewise.
10319 (int12_operand, int6_operand, int5_operand, uint5_operand): Likewise.
10320 (uint4_operand, uint1_operand, int_2word_operand): Likewise
10321 (upper_int16_operand, uint16_operand, symbolic_operand): Likewise.
10322 (relational_operator, float_relational_operator): Likewise.
10323 (ccr_eqne_operator, minmax_operator): Likewise.
10324 (condexec_si_binary_operator, condexec_si_media_operator): Likewise.
10325 (condexec_si_divide_operator, condexec_si_unary_operator): Likewise.
10326 (condexec_sf_conv_operator, condexec_sf_add_operator): Likewise.
10327 (intop_compare_operator, acc_operand, even_acc_operand): Likewise.
10328 (quad_acc_operand, accg_operand): Likewise.
10330 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10332 * config/stormy16/stormy16-protos.h (xstormy16_below_100_symbol): Change
10333 return type to a bool.
10334 (nonimmediate_nonstack_operand): Likewise.
10335 (xstormy16_splittable_below100_operand): Likewise.
10336 * config/stormy16/stormy16.c (xstormy16_below_100_symbol): Fix
10338 (xstormy16_splittable_below100_operand): Likewise.
10340 2021-07-02 Richard Biener <rguenther@suse.de>
10342 PR tree-optimization/101293
10343 * tree-ssa-loop-im.c (mem_ref_hasher::equal): Compare MEM_REF bases
10344 with combined offsets.
10345 (gather_mem_refs_stmt): Hash MEM_REFs as if their offset were
10346 combined with the rest of the offset.
10348 2021-07-02 Eric Botcazou <ebotcazou@adacore.com>
10350 * config/i386/i386.c (asm_preferred_eh_data_format): Always use the
10351 PIC encodings for PE-COFF targets.
10353 2021-07-02 Jakub Jelinek <jakub@redhat.com>
10356 * config/i386/i386-expand.c (ix86_broadcast_from_integer_constant):
10357 Return nullptr for TImode inner mode.
10359 2021-07-02 Richard Biener <rguenther@suse.de>
10361 PR tree-optimization/101280
10362 PR tree-optimization/101173
10363 * gimple-loop-interchange.cc
10364 (tree_loop_interchange::valid_data_dependences): Properly
10365 guard all dependence checks with DDR_REVERSED_P or its
10368 2021-07-02 Hongyu Wang <hongyu.wang@intel.com>
10370 * config/i386/i386-expand.c (ix86_expand_builtin):
10371 Add branch to clear odata when ZF is set for asedecenc_expand
10372 and wideaesdecenc_expand.
10374 2021-07-02 Eugene Rozenfeld <erozen@microsoft.com>
10376 * config/i386/gcc-auto-profile: regenerate
10378 2021-07-02 liuhongt <hongtao.liu@intel.com>
10380 * config/i386/sse.md (trunc<mode><pmov_dst_4>2): Refined to ..
10381 (trunc<mode><pmov_dst_4_lower>2): this.
10383 2021-07-01 David Malcolm <dmalcolm@redhat.com>
10385 * diagnostic.h (diagnostic_context::m_file_cache): New field.
10386 * input.c (class fcache): Rename to...
10387 (class file_cache_slot): ...this, making most members private and
10388 prefixing fields with "m_".
10389 (file_cache_slot::get_file_path): New accessor.
10390 (file_cache_slot::get_use_count): New accessor.
10391 (file_cache_slot::missing_trailing_newline_p): New accessor.
10392 (file_cache_slot::inc_use_count): New.
10393 (fcache_buffer_size): Move to...
10394 (file_cache_slot::buffer_size): ...here.
10395 (fcache_line_record_size): Move to...
10396 (file_cache_slot::line_record_size): ...here.
10397 (fcache_tab): Delete, in favor of global_dc->m_file_cache.
10398 (fcache_tab_size): Move to file_cache::num_file_slots.
10399 (diagnostic_file_cache_init): Update for move of fcache_tab
10400 to global_dc->m_file_cache.
10401 (diagnostic_file_cache_fini): Likewise.
10402 (lookup_file_in_cache_tab): Convert to...
10403 (file_cache::lookup_file): ...this.
10404 (diagnostics_file_cache_forcibly_evict_file): Update for move of
10405 fcache_tab to global_dc->m_file_cache, moving most of
10406 implementation to...
10407 (file_cache::forcibly_evict_file): ...this new function and...
10408 (file_cache_slot::evict): ...this new function.
10409 (evicted_cache_tab_entry): Convert to...
10410 (file_cache::evicted_cache_tab_entry): ...this.
10411 (add_file_to_cache_tab): Convert to...
10412 (file_cache::add_file): ...this, moving bulk of implementation
10414 (file_cache_slot::create): ..this new function.
10415 (file_cache::file_cache): New.
10416 (file_cache::~file_cache): New.
10417 (lookup_or_add_file_to_cache_tab): Convert to...
10418 (file_cache::lookup_or_add_file): ..this new function.
10419 (fcache::fcache): Rename to...
10420 (file_cache_slot::file_cache_slot): ...this, adding "m_" prefixes
10422 (fcache::~fcache): Rename to...
10423 (file_cache_slot::~file_cache_slot): ...this, adding "m_" prefixes
10425 (needs_read): Convert to...
10426 (file_cache_slot::needs_read_p): ...this.
10427 (needs_grow): Convert to...
10428 (file_cache_slot::needs_grow_p): ...this.
10429 (maybe_grow): Convert to...
10430 (file_cache_slot::maybe_grow): ...this.
10431 (read_data): Convert to...
10432 (file_cache_slot::read_data): ...this.
10433 (maybe_read_data): Convert to...
10434 (file_cache_slot::maybe_read_data): ...this.
10435 (get_next_line): Convert to...
10436 (file_cache_slot::get_next_line): ...this.
10437 (goto_next_line): Convert to...
10438 (file_cache_slot::goto_next_line): ...this.
10439 (read_line_num): Convert to...
10440 (file_cache_slot::read_line_num): ...this.
10441 (location_get_source_line): Update for moving of globals to
10442 global_dc->m_file_cache.
10443 (location_missing_trailing_newline): Likewise.
10444 * input.h (class file_cache_slot): New forward decl.
10445 (class file_cache): New.
10447 2021-07-01 Michael Meissner <meissner@linux.ibm.com>
10449 * config/rs6000/rs6000.c (rs6000_maybe_emit_fp_cmove): Add IEEE
10450 128-bit floating point conditional move support.
10451 (have_compare_and_set_mask): Add IEEE 128-bit floating point
10453 * config/rs6000/rs6000.md (mov<mode>cc, IEEE128 iterator): New insn.
10454 (mov<mode>cc_p10, IEEE128 iterator): New insn.
10455 (mov<mode>cc_invert_p10, IEEE128 iterator): New insn.
10456 (fpmask<mode>, IEEE128 iterator): New insn.
10457 (xxsel<mode>, IEEE128 iterator): New insn.
10459 2021-07-01 Iain Sandoe <iain@sandoe.co.uk>
10462 * config/darwin.h (CTF_INFO_SECTION_NAME): New.
10464 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
10466 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
10468 * config/i386/i386-protos.h (ix86_expand_vector_init_duplicate):
10470 * config/i386/sse.md (INT_BROADCAST_MODE): New mode iterator.
10471 (vec_duplicate<mode>): New expander.
10473 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
10476 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
10478 (ix86_byte_broadcast): New function.
10479 (ix86_convert_const_wide_int_to_broadcast): Likewise.
10480 (ix86_expand_move): Convert CONST_WIDE_INT to broadcast if mode
10481 size is 16 bytes or bigger.
10482 (ix86_broadcast_from_integer_constant): New function.
10483 (ix86_expand_vector_move): Convert CONST_WIDE_INT and CONST_VECTOR
10484 to broadcast if mode size is 16 bytes or bigger.
10485 * config/i386/i386-protos.h (ix86_gen_scratch_sse_rtx): New
10487 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): New function.
10489 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10491 * config/i386/predicates.md (ix86_endbr_immediate_operand):
10492 Return true/false instead of 1/0.
10493 (movq_parallel): Ditto.
10495 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10497 * recog.c (general_operand): Return true/false instead of 1/0.
10498 (register_operand): Ditto.
10499 (immediate_operand): Ditto.
10500 (const_int_operand): Ditto.
10501 (const_scalar_int_operand): Ditto.
10502 (const_double_operand): Ditto.
10503 (push_operand): Ditto.
10504 (pop_operand): Ditto.
10505 (memory_operand): Ditto.
10506 (indirect_operand): Ditto.
10508 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10510 * genpreds.c (write_predicate_subfunction):
10511 Change the type of written subfunction to bool.
10512 (write_one_predicate_function):
10513 Change the type of written function to bool.
10514 (write_tm_preds_h): Ditto.
10515 * recog.h (*insn_operand_predicate_fn): Change the type to bool.
10516 * recog.c (general_operand): Change the type to bool.
10517 (address_operand): Ditto.
10518 (register_operand): Ditto.
10519 (pmode_register_operand): Ditto.
10520 (scratch_operand): Ditto.
10521 (immediate_operand): Ditto.
10522 (const_int_operand): Ditto.
10523 (const_scalar_int_operand): Ditto.
10524 (const_double_operand): Ditto.
10525 (nonimmediate_operand): Ditto.
10526 (nonmemory_operand): Ditto.
10527 (push_operand): Ditto.
10528 (pop_operand): Ditto.
10529 (memory_operand): Ditto.
10530 (indirect_operand): Ditto.
10531 (ordered_comparison_operator): Ditto.
10532 (comparison_operator): Ditto.
10533 * config/i386/i386-expand.c (ix86_expand_sse_cmp):
10534 Change the type of indirect predicate function to bool.
10535 * config/rs6000/rs6000.c (easy_vector_constant):
10536 Change the type to bool.
10537 * config/mips/mips-protos.h (m16_based_address_p):
10538 Change the type of operand 3 to bool.
10540 2021-07-01 Richard Biener <rguenther@suse.de>
10542 PR tree-optimization/101280
10543 PR tree-optimization/101173
10544 * gimple-loop-interchange.cc
10545 (tree_loop_interchange::valid_data_dependences): Revert
10546 previous change and instead correctly handle DDR_REVERSED_P
10549 2021-07-01 Richard Biener <rguenther@suse.de>
10551 PR tree-optimization/101278
10552 * tree-ssa-dse.c (dse_classify_store): First check for
10553 uses, then ignore stmt for chaining purposes.
10555 2021-07-01 Richard Biener <rguenther@suse.de>
10557 PR tree-optimization/100778
10558 * tree-vect-slp.c (vect_schedule_slp_node): Do not place trapping
10559 vectorized ops ahead of their scalar BB.
10561 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10564 * config/i386/i386.md (*nabs<dwi>2_doubleword):
10565 New insn_and_split pattern.
10566 (*nabs<dwi>2_1): Ditto.
10567 * config/i386/i386-features.c
10568 (general_scalar_chain::compute_convert_gain):
10569 Handle (NEG (ABS (...))) RTX. Rewrite src code
10570 scanner as switch statement.
10571 (general_scalar_chain::convert_insn):
10572 Handle (NEG (ABS (...))) RTX.
10573 (general_scalar_to_vector_candidate_p):
10574 Detect (NEG (ABS (...))) RTX. Reorder case statements
10575 for (AND (NOT (...) ...)) fallthrough.
10577 2021-07-01 Richard Biener <rguenther@suse.de>
10579 PR tree-optimization/101178
10580 * tree-vect-slp.c (slpg_vertex::materialize): Remove.
10581 (slpg::perm_in): Add.
10582 (slpg::get_perm_in): Remove.
10583 (slpg::get_perm_materialized): Add.
10584 (vect_optimize_slp): Handle VEC_PERM nodes more optimally
10585 during permute propagation and materialization.
10587 2021-07-01 Jakub Jelinek <jakub@redhat.com>
10590 * dwarf2out.c (loc_list_from_tree_1): Handle COMPOUND_LITERAL_EXPR.
10592 2021-07-01 Jakub Jelinek <jakub@redhat.com>
10594 PR middle-end/94366
10595 * omp-low.c (lower_rec_input_clauses): Rename is_fp_and_or to
10596 is_truth_op, set it for TRUTH_*IF_EXPR regardless of new_var's type,
10597 use boolean_type_node instead of integer_type_node as NE_EXPR type.
10598 (lower_reduction_clauses): Likewise.
10600 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
10602 * config/gcn/gcn.c: Include dwarf2.h.
10603 (gcn_addr_space_debug): New function.
10604 (TARGET_ADDR_SPACE_DEBUG): New hook.
10606 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
10608 * common/config/gcn/gcn-common.c
10609 (gcn_option_optimization_table): Change OPT_fomit_frame_pointer to -O3.
10610 * config/gcn/gcn.c (gcn_expand_prologue): Prefer the frame pointer
10612 (gcn_expand_prologue): Prefer the frame pointer when emitting CFI.
10613 (gcn_frame_pointer_rqd): New function.
10614 (TARGET_FRAME_POINTER_REQUIRED): New hook.
10616 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
10618 * config/gcn/gcn.c (move_callee_saved_registers): Emit CFI notes for
10619 prologue register saves.
10620 (gcn_debug_unwind_info): Use UI_DWARF2.
10621 (gcn_dwarf_register_number): Map DWARF_LINK_REGISTER to DWARF PC.
10622 (gcn_dwarf_register_span): DWARF_LINK_REGISTER doesn't span.
10623 * config/gcn/gcn.h: (DWARF_FRAME_RETURN_COLUMN): New define.
10624 (DWARF_LINK_REGISTER): New define.
10625 (FIRST_PSEUDO_REGISTER): Increment.
10626 (FIXED_REGISTERS): Add entry for DWARF_LINK_REGISTER.
10627 (CALL_USED_REGISTERS): Likewise.
10628 (REGISTER_NAMES): Likewise.
10630 2021-06-30 Richard Biener <rguenther@suse.de>
10632 PR tree-optimization/101267
10633 * tree-vect-stmts.c (vect_check_scalar_mask): Adjust
10634 API and use SLP compatible interface of vect_is_simple_use.
10635 Reject not vectorized SLP defs for callers that do not support
10637 (vect_check_store_rhs): Handle masked stores and pass down
10638 the appropriate operator index.
10639 (vectorizable_call): Adjust.
10640 (vectorizable_store): Likewise.
10641 (vectorizable_load): Likewise. Handle SLP pecularity of
10643 (vect_is_simple_use): Remove special-casing of masked stores.
10645 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
10647 * common.opt (foffload): Remove help as Driver only.
10648 * gcc.c (display_help): Add -foffload.
10650 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
10652 * gcc.c (close_at_file, execute): Replace alloca by XALLOCAVEC.
10653 (check_offload_target_name): Fix splitting OFFLOAD_TARGETS into
10654 a candidate list; better inform no offload target is configured
10655 and fix hint extraction when passed target is not '\0' at [len].
10656 * common.opt (foffload): Add tailing '.'.
10657 (foffload-options): Likewise; fix flag name in the help string.
10659 2021-06-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
10662 * config/arm/arm_neon.h: Move vabs intrinsics before vcage_f32.
10663 (vcage_f32): Gate comparison on __FAST_MATH__.
10664 (vcageq_f32): Likewise.
10665 (vcale_f32): Likewise.
10666 (vcaleq_f32): Likewise.
10667 (vcagt_f32): Likewise.
10668 (vcagtq_f32): Likewise.
10669 (vcalt_f32): Likewise.
10670 (vcaltq_f32): Likewise.
10671 (vcage_f16): Likewise.
10672 (vcageq_f16): Likewise.
10673 (vcale_f16): Likewise.
10674 (vcaleq_f16): Likewise.
10675 (vcagt_f16): Likewise.
10676 (vcagtq_f16): Likewise.
10677 (vcalt_f16): Likewise.
10678 (vcaltq_f16): Likewise.
10680 2021-06-30 Richard Biener <rguenther@suse.de>
10682 PR tree-optimization/101264
10683 * tree-vect-slp.c (vect_optimize_slp): Propagate the
10684 computed perm_in to all "any" permute successors
10685 we cannot de-duplicate immediately.
10687 2021-06-30 liuhongt <hongtao.liu@intel.com>
10690 * config/i386/sse.md
10691 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>):
10693 (avx512f_sfixupimm<mode><maskz_scalar_name><round_saeonly_name>):
10695 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>"): Refined.
10696 * config/i386/subst.md (maskz_scalar): New define_subst.
10697 (maskz_scalar_name): New subst_attr.
10698 (maskz_scalar_op5): Ditto.
10699 (round_saeonly_maskz_scalar_op5): Ditto.
10700 (round_saeonly_maskz_scalar_operand5): Ditto.
10702 2021-06-30 David Edelsohn <dje.gcc@gmail.com>
10704 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
10705 Increase code CSECT alignment to at least 32 bytes.
10706 * config/rs6000/xcoff.h (TEXT_SECTION_ASM_OP): Add 32 byte
10707 alignment designation.
10709 2021-06-29 Sergei Trofimovich <siarheit@google.com>
10711 * doc/generic.texi: Fix s/net yet/not yet/ typo.
10713 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
10715 PR tree-optimization/101254
10716 * range-op.cc (operator_minus::op1_op2_relation_effect): Check for
10717 wrapping/non-wrapping when setting the result range.
10719 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
10721 * value-query.cc (gimple_range_global): Allow phis.
10723 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
10725 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): Use stmt.
10726 (simplify_using_ranges::op_with_boolean_value_range_p): Add a
10727 statement for location context.
10728 (check_for_binary_op_overflow): Ditto.
10729 (simplify_using_ranges::get_vr_for_comparison): Ditto.
10730 (simplify_using_ranges::compare_name_with_value): Ditto.
10731 (simplify_using_ranges::compare_names): Ditto.
10732 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Ditto.
10733 (simplify_using_ranges::simplify_truth_ops_using_ranges): Ditto.
10734 (simplify_using_ranges::simplify_min_or_max_using_ranges): Ditto.
10735 (simplify_using_ranges::simplify_internal_call_using_ranges): Ditto.
10736 (simplify_using_ranges::two_valued_val_range_p): Ditto.
10737 (simplify_using_ranges::simplify): Ditto.
10738 * vr-values.h: Adjust prototypes.
10740 2021-06-29 Uroš Bizjak <ubizjak@gmail.com>
10743 * config/i386/mmx.md (vec_addsubv2sf3): New insn pattern.
10745 2021-06-29 Julian Brown <julian@codesourcery.com>
10747 * config/gcn/gcn.c (gcn_init_libfuncs): New function.
10748 (TARGET_INIT_LIBFUNCS): Define target hook using above function.
10749 * config/gcn/gcn.h (UNITS_PER_WORD): Define to 8 for IN_LIBGCC2, 4
10751 (LIBGCC2_UNITS_PER_WORD, BITS_PER_WORD): Remove definitions.
10752 (MAX_FIXED_MODE_SIZE): Change to 128.
10754 2021-06-29 Julian Brown <julian@codesourcery.com>
10756 * config/gcn/gcn.md (UNSPEC_FLBIT_INT): New unspec constant.
10757 (s_mnemonic): Add clrsb.
10758 (gcn_flbit<mode>_int): Add insn pattern for SImode/DImode.
10759 (clrsb<mode>2): Add expander for SImode/DImode.
10761 2021-06-29 Julian Brown <julian@codesourcery.com>
10763 * config/gcn/gcn.md (<su>mulsidi3, <su>mulsidi3_reg, <su>mulsidi3_imm,
10764 muldi3): Add patterns.
10766 2021-06-29 Julian Brown <julian@codesourcery.com>
10768 * config/gcn/gcn.md (<su>mulsi3_highpart): Change to expander.
10769 (<su>mulsi3_highpart_reg, <su>mulsi3_highpart_imm): New patterns.
10771 2021-06-29 Julian Brown <julian@codesourcery.com>
10773 * config/gcn/gcn.md (mulsi3): Make s_mulk_i32 variant clobber SCC.
10775 2021-06-29 Joseph Myers <joseph@codesourcery.com>
10777 * btfout.c, ctfout.c: Include "memmodel.h".
10779 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
10781 * gcc.c (check_offload_target_name): Cast len argument to
10782 %q.*s to 'int'; avoid -Wstringop-truncation warning.
10784 2021-06-29 Richard Biener <rguenther@suse.de>
10786 * tree-vect-slp.c (vect_optimize_slp): Forward propagate
10787 to "any" permute nodes and relax "any" permute proapgation
10788 during iterative backward propagation.
10790 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
10793 * common.opt (-foffload=): Update description.
10794 (-foffload-options=): New.
10795 * doc/invoke.texi (C Language Options): Document
10796 -foffload and -foffload-options.
10797 * gcc.c (check_offload_target_name): New, split off from
10798 handle_foffload_option.
10799 (check_foffload_target_names): New.
10800 (handle_foffload_option): Handle -foffload=default.
10801 (driver_handle_option): Update for -foffload-options.
10802 * lto-opts.c (lto_write_options): Use -foffload-options
10803 instead of -foffload.
10804 * lto-wrapper.c (merge_and_complain, append_offload_options):
10806 * opts.c (common_handle_option): Likewise.
10808 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
10810 * doc/invoke.texi (C Language Options): Sort options
10811 alphabetically in optlist and also the description itself.
10812 Remove leftover -fallow-single-precision from and add missing
10813 -fgnu-tm to the optlist.
10815 2021-06-29 Richard Biener <rguenther@suse.de>
10817 * tree-vect-slp.c (slpg_vertex::visited): Remove.
10818 (vect_slp_perms_eq): Handle -1 permutes.
10819 (vect_optimize_slp): Rewrite permute propagation.
10821 2021-06-29 Jakub Jelinek <jakub@redhat.com>
10824 * match.pd ((intptr_t)x eq/ne CST to x eq/ne (typeof x) CST): Don't
10825 perform the optimization in GENERIC when sanitizing and x has a
10828 2021-06-29 Richard Biener <rguenther@suse.de>
10830 PR tree-optimization/101242
10831 * tree-vect-slp.c (vect_slp_build_vertices): Force-add
10832 PHIs with not represented initial values as leafs.
10834 2021-06-29 Jan-Benedict Glaw <jbglaw@getslash.de>
10836 * config/pdp11/pdp11.h (ASM_OUTPUT_SKIP): Fix signedness warning.
10837 * config/pdp11/pdp11.c (pdp11_asm_print_operand_punct_valid_p): Remove
10838 "register" keyword.
10839 (pdp11_initial_elimination_offset) Remove unused variable.
10840 (pdp11_cmp_length) Ditto.
10841 (pdp11_insn_cost): Ditto, and fix signedness warning.
10843 2021-06-29 David Edelsohn <dje.gcc@gmail.com>
10845 * btfout.c: Include tm_p.h.
10848 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
10850 * config/bpf/bpf.c (bpf_expand_prologue): Do not mark insns as
10852 (bpf_expand_epilogue): Likewise.
10853 * config/bpf/bpf.h (DWARF2_FRAME_INFO): Define to 0.
10854 Do not define DBX_DEBUGGING_INFO.
10856 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
10858 * doc/invoke.texi: Document the CTF and BTF debug info options.
10860 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
10861 David Faust <david.faust@oracle.com>
10862 Jose E. Marchesi <jose.marchesi@oracle.com>
10863 Weimin Pan <weimin.pan@oracle.com>
10865 * Makefile.in: Add ctfc.*, ctfout.c and btfout.c files to
10866 GTFILES. Add new object files.
10867 * common.opt: Add CTF and BTF debug info options.
10868 * btfout.c: New file.
10869 * ctfc.c: Likewise.
10870 * ctfc.h: Likewise.
10871 * ctfout.c: Likewise.
10872 * dwarf2ctf.c: Likewise.
10873 * dwarf2ctf.h: Likewise.
10874 * dwarf2cfi.c (dwarf2out_do_frame): Acknowledge CTF_DEBUG and
10876 * dwarf2out.c (dwarf2out_source_line): Likewise.
10877 (dwarf2out_finish): Skip emitting DWARF if CTF or BTF are to
10879 (debug_format_do_cu): New function.
10880 (dwarf2out_early_finish): Traverse DIEs and emit CTF/BTF for
10882 Include dwarf2ctf.c.
10883 * final.c (dwarf2_debug_info_emitted_p): Acknowledge DWARF-based debug
10885 * flag-types.h (enum debug_info_type): Add CTF_DEBUG and BTF_DEBUG.
10886 (CTF_DEBUG): New bitmask.
10887 (BTF_DEBUG): Likewise.
10888 (enum ctf_debug_info_levels): New enum.
10889 * gengtype.c (open_base_files): Handle ctfc.h.
10890 (main): Handle uint32_t type.
10891 * flags.h (btf_debuginfo_p): New definition.
10892 (dwarf_based_debuginfo_p): Likewise.
10893 * opts.c (debug_type_names): Add entries for CTF and BTF.
10894 (btf_debuginfo_p): New function.
10895 (dwarf_based_debuginfo_p): Likewise.
10896 (common_handle_option): Handle -gctfN and -gbtf options.
10897 (set_debug_level): Set CTF_DEBUG, BTF_DEBUG whenever appropriate.
10898 * toplev.c (process_options): Inform the user and ignore -gctfLEVEL if
10901 2021-06-28 Jose E. Marchesi <jose.marchesi@oracle.com>
10903 * dwarf2out.c (AT_class): Function is no longer static.
10904 (AT_int): Likewise.
10905 (AT_unsigned): Likewise.
10906 (AT_loc): Likewise.
10907 (get_AT): Likewise.
10908 (get_AT_string): Likewise.
10909 (get_AT_flag): Likewise.
10910 (get_AT_unsigned): Likewise.
10911 (get_AT_ref): Likewise.
10912 (new_die_raw): Likewise.
10913 (lookup_decl_die): Likewise.
10914 (base_type_die): Likewise.
10915 (add_name_attribute): Likewise.
10916 (add_AT_int): Likewise.
10917 (add_AT_unsigned): Likewise.
10918 (add_AT_loc): Likewise.
10919 (dw_get_die_tag): New function.
10920 (dw_get_die_child): Likewise.
10921 (dw_get_die_sib): Likewise.
10922 (struct dwarf_file_data): Move from here to dwarf2out.h
10923 (struct dw_attr_struct): Likewise.
10924 * dwarf2out.h: Analogous changes.
10926 2021-06-28 Martin Jambor <mjambor@suse.cz>
10929 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
10930 members m_dead_stmts and m_dead_ssas.
10931 * ipa-param-manipulation.c
10932 (ipa_param_body_adjustments::mark_dead_statements): New function.
10933 (ipa_param_body_adjustments::common_initialization): Call it on
10934 all removed but not split parameters.
10935 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
10937 (ipa_param_body_adjustments::modify_call_stmt): Remove arguments that
10939 * tree-inline.c (remap_gimple_stmt): Do not copy dead statements, reset
10940 dead debug statements.
10941 (copy_phis_for_bb): Do not copy dead PHI nodes.
10943 2021-06-28 Martin Jambor <mjambor@suse.cz>
10946 * symtab-clones.h (clone_info): Removed member param_adjustments.
10947 * ipa-param-manipulation.h: Adjust initial comment to reflect how we
10948 deal with pass-through splits now.
10949 (ipa_param_performed_split): Removed.
10950 (ipa_param_adjustments::modify_call): Adjusted parameters.
10951 (class ipa_param_body_adjustments): Adjusted parameters of
10952 register_replacement, modify_gimple_stmt and modify_call_stmt.
10953 (ipa_verify_edge_has_no_modifications): Declare.
10954 (ipa_edge_modifications_finalize): Declare.
10955 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Remove
10956 performed_splits processing, pas only edge to padjs->modify_call,
10957 check that call arguments were not modified if they should not have
10959 * cgraphclones.c (cgraph_node::create_clone): Do not copy performed
10961 * ipa-param-manipulation.c (struct pass_through_split_map): New type.
10962 (ipa_edge_modification_info): Likewise.
10963 (ipa_edge_modification_sum): Likewise.
10964 (ipa_edge_modifications): New edge summary.
10965 (ipa_verify_edge_has_no_modifications): New function.
10966 (transitive_split_p): Removed.
10967 (transitive_split_map): Likewise.
10968 (init_transitive_splits): Likewise.
10969 (ipa_param_adjustments::modify_call): Adjusted to use the new edge
10970 summary instead of performed_splits.
10971 (ipa_param_body_adjustments::register_replacement): Drop dummy
10972 parameter, set base_index of the created ipa_param_body_replacement.
10973 (phi_arg_will_live_p): New function.
10974 (ipa_param_body_adjustments::common_initialization): Do not create
10975 IPA_SRA dummy decls.
10976 (simple_tree_swap_info): Removed.
10977 (remap_split_decl_to_dummy): Likewise.
10978 (record_argument_state_1): New function.
10979 (record_argument_state): Likewise.
10980 (ipa_param_body_adjustments::modify_call_stmt): New parameter
10981 orig_stmt. Do not work with dummy decls, save necessary info about
10982 changes to ipa_edge_modifications.
10983 (ipa_param_body_adjustments::modify_gimple_stmt): New parameter
10984 orig_stmt, pass it to modify_call_stmt.
10985 (ipa_param_body_adjustments::modify_cfun_body): Adjust call to
10986 modify_gimple_stmt.
10987 (ipa_edge_modifications_finalize): New function.
10988 * tree-inline.c (remap_gimple_stmt): Pass original statement to
10989 modify_gimple_stmt.
10990 (copy_phis_for_bb): Do not copy dead PHI nodes.
10991 (expand_call_inline): Do not remap performed_splits.
10992 (update_clone_info): Likewise.
10993 * toplev.c: Include ipa-param-manipulation.h.
10994 (toplev::finalize): Call ipa_edge_modifications_finalize.
10996 2021-06-28 Andrew Pinski <apinski@marvell.com>
10998 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Duplicate range
10999 info if we're the only things setting the target PHI.
11000 (value_replacement): Don't duplicate range here.
11001 (minmax_replacement): Likewise.
11003 2021-06-28 Richard Biener <rguenther@suse.de>
11005 PR tree-optimization/101229
11006 * gimple-walk.c (gimple_walk_op): Handle PHIs.
11008 2021-06-28 Martin Liska <mliska@suse.cz>
11010 * config/v850/v850.c (construct_dispose_instruction): Allocate
11012 (construct_prepare_instruction): Likewise.
11014 2021-06-28 Martin Liska <mliska@suse.cz>
11016 * config/v850/v850.c (v850_option_override): Build default
11018 (v850_can_inline_p): New. Allow MASK_PROLOG_FUNCTION to be
11019 ignored for inlining.
11020 (TARGET_CAN_INLINE_P): New.
11022 2021-06-28 Richard Biener <rguenther@suse.de>
11024 PR tree-optimization/101207
11025 * tree-vect-slp.c (vect_optimize_slp): Do BB reduction
11026 permute eliding for load permutations properly.
11028 2021-06-28 Richard Biener <rguenther@suse.de>
11030 PR tree-optimization/101173
11031 * gimple-loop-interchange.cc
11032 (tree_loop_interchange::valid_data_dependences): Disallow outer
11033 loop dependence distance of zero.
11035 2021-06-28 liuhongt <hongtao.liu@intel.com>
11038 * config/i386/sse.md (*avx_cmp<mode>3_lt): New
11039 define_insn_and_split.
11040 (*avx_cmp<mode>3_ltint): Ditto.
11041 (*avx2_pcmp<mode>3_3): Ditto.
11042 (*avx2_pcmp<mode>3_4): Ditto.
11043 (*avx2_pcmp<mode>3_5): Ditto.
11045 2021-06-28 liuhongt <hongtao.liu@intel.com>
11047 * config/i386/i386-builtin.def (IX86_BUILTIN_BLENDVPD256,
11048 IX86_BUILTIN_BLENDVPS256, IX86_BUILTIN_PBLENDVB256,
11049 IX86_BUILTIN_BLENDVPD, IX86_BUILTIN_BLENDVPS,
11050 IX86_BUILTIN_PBLENDVB128): Replace icode with
11052 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold blendv
11054 * config/i386/sse.md (*<sse4_1_avx2>_pblendvb_lt_subreg_not):
11055 New pre_reload splitter.
11057 2021-06-27 Andrew Pinski <apinski@marvell.com>
11059 PR middle-end/101230
11060 * fold-const.c (fold_ternary_loc): Check
11061 the return value of invert_tree_comparison.
11063 2021-06-27 David Edelsohn <dje.gcc@gmail.com>
11065 * config.gcc: Add SPDX License Identifier.
11066 (powerpc-ibm-aix789): Default to aix73.h.
11067 (powerpc-ibm-aix7.2.*.*): New stanza.
11068 * config/rs6000/aix72.h: Add SPDX License Identifier.
11069 * config/rs6000/aix73.h: New file.
11071 2021-06-26 Jason Merrill <jason@redhat.com>
11073 * except.c: #include "dwarf2.h" instead of "dwarf2out.h".
11075 2021-06-26 Andrew Pinski <apinski@marvell.com>
11077 * genmatch.c (lower_cond): Copy for_subst_vec
11078 for the simplify also.
11079 (lower): Swap the order for lower_for and lower_cond.
11081 2021-06-26 Andrew Pinski <apinski@marvell.com>
11083 * tree-ssa-phiopt.c (match_simplify_replacement): Reset
11084 flow senatitive info on the moved ssa set.
11086 2021-06-26 Andrew Pinski <apinski@marvell.com>
11088 * fold-const.c (fold_cond_expr_with_comparison):
11089 Exand arg0 into comp_code, arg00, and arg01.
11090 (fold_ternary_loc): Use invert_tree_comparison
11091 instead of fold_invert_truthvalue for the case
11092 where we have A CMP B ? C : A.
11094 2021-06-25 Martin Sebor <msebor@redhat.com>
11096 PR middle-end/101216
11097 * calls.c (maybe_warn_rdwr_sizes): Use the no_warning constant.
11099 2021-06-25 Jeff Law <jeffreyalaw@gmail.com>
11101 * config/h8300/h8300.c (select_cc_mode): Handle ASHIFTRT and LSHIFTRT.
11103 2021-06-25 Richard Biener <rguenther@suse.de>
11105 PR tree-optimization/101202
11106 * tree-vect-slp.c (vect_optimize_slp): Explicitely handle
11109 2021-06-25 Richard Biener <rguenther@suse.de>
11111 * tree-vect-slp-patterns.c (addsub_pattern::build): Copy
11112 STMT_VINFO_REDUC_DEF from the original representative.
11114 2021-06-25 Martin Sebor <msebor@redhat.com>
11116 * builtins.c (warn_string_no_nul): Replace uses of TREE_NO_WARNING,
11117 gimple_no_warning_p and gimple_set_no_warning with
11118 warning_suppressed_p, and suppress_warning.
11120 (maybe_warn_for_bound): Same.
11121 (warn_for_access): Same.
11122 (check_access): Same.
11123 (expand_builtin_strncmp): Same.
11124 (fold_builtin_varargs): Same.
11125 * calls.c (maybe_warn_nonstring_arg): Same.
11126 (maybe_warn_rdwr_sizes): Same.
11127 * cfgexpand.c (expand_call_stmt): Same.
11128 * cgraphunit.c (check_global_declaration): Same.
11129 * fold-const.c (fold_undefer_overflow_warnings): Same.
11130 (fold_truth_not_expr): Same.
11131 (fold_unary_loc): Same.
11132 (fold_checksum_tree): Same.
11133 * gimple-array-bounds.cc (array_bounds_checker::check_array_ref): Same.
11134 (array_bounds_checker::check_mem_ref): Same.
11135 (array_bounds_checker::check_addr_expr): Same.
11136 (array_bounds_checker::check_array_bounds): Same.
11137 * gimple-expr.c (copy_var_decl): Same.
11138 * gimple-fold.c (gimple_fold_builtin_strcpy): Same.
11139 (gimple_fold_builtin_strncat): Same.
11140 (gimple_fold_builtin_stxcpy_chk): Same.
11141 (gimple_fold_builtin_stpcpy): Same.
11142 (gimple_fold_builtin_sprintf): Same.
11143 (fold_stmt_1): Same.
11144 * gimple-ssa-isolate-paths.c (diag_returned_locals): Same.
11145 * gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Same.
11146 * gimple-ssa-sprintf.c (handle_printf_call): Same.
11147 * gimple-ssa-store-merging.c (imm_store_chain_info::output_merged_store): Same.
11148 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
11149 * gimple-ssa-warn-restrict.h: Adjust declarations.
11150 (maybe_diag_access_bounds): Replace uses of TREE_NO_WARNING,
11151 gimple_no_warning_p and gimple_set_no_warning with
11152 warning_suppressed_p, and suppress_warning.
11153 (check_call): Same.
11154 (check_bounds_or_overlap): Same.
11155 * gimple.c (gimple_build_call_from_tree): Same.
11156 * gimplify.c (gimplify_return_expr): Same.
11157 (gimplify_cond_expr): Same.
11158 (gimplify_modify_expr_complex_part): Same.
11159 (gimplify_modify_expr): Same.
11160 (gimple_push_cleanup): Same.
11161 (gimplify_expr): Same.
11162 * omp-expand.c (expand_omp_for_generic): Same.
11163 (expand_omp_taskloop_for_outer): Same.
11164 * omp-low.c (lower_rec_input_clauses): Same.
11165 (lower_lastprivate_clauses): Same.
11166 (lower_send_clauses): Same.
11167 (lower_omp_target): Same.
11168 * tree-cfg.c (pass_warn_function_return::execute): Same.
11169 * tree-complex.c (create_one_component_var): Same.
11170 * tree-inline.c (remap_gimple_op_r): Same.
11171 (copy_tree_body_r): Same.
11172 (declare_return_variable): Same.
11173 (expand_call_inline): Same.
11174 * tree-nested.c (lookup_field_for_decl): Same.
11175 * tree-sra.c (create_access_replacement): Same.
11176 (generate_subtree_copies): Same.
11177 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Same.
11178 * tree-ssa-forwprop.c (combine_cond_expr_cond): Same.
11179 * tree-ssa-loop-ch.c (ch_base::copy_headers): Same.
11180 * tree-ssa-loop-im.c (execute_sm): Same.
11181 * tree-ssa-phiopt.c (cond_store_replacement): Same.
11182 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
11183 (handle_builtin_strcpy): Same.
11184 (maybe_diag_stxncpy_trunc): Same.
11185 (handle_builtin_stxncpy_strncat): Same.
11186 (handle_builtin_strcat): Same.
11187 * tree-ssa-uninit.c (get_no_uninit_warning): Same.
11188 (set_no_uninit_warning): Same.
11189 (uninit_undefined_value_p): Same.
11190 (warn_uninit): Same.
11191 (maybe_warn_operand): Same.
11192 * tree-vrp.c (compare_values_warnv): Same.
11193 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr): Same.
11194 (test_for_singularity): Same.
11195 * gimple.h (warning_suppressed_p): New function.
11196 (suppress_warning): Same.
11197 (copy_no_warning): Same.
11198 (gimple_set_block): Call gimple_set_location.
11199 (gimple_set_location): Call copy_warning.
11201 2021-06-25 Martin Sebor <msebor@redhat.com>
11203 * tree.h (warning_suppressed_at, copy_warning,
11204 warning_suppressed_p, suppress_warning): New functions.
11206 2021-06-25 Martin Sebor <msebor@redhat.com>
11208 * Makefile.in (OBJS-libcommon): Add diagnostic-spec.o.
11209 * gengtype.c (open_base_files): Add diagnostic-spec.h.
11210 * diagnostic-spec.c: New file.
11211 * diagnostic-spec.h: New file.
11212 * tree.h (no_warning, all_warnings, suppress_warning_at): New
11214 * warning-control.cc: New file.
11216 2021-06-25 liuhongt <hongtao.liu@intel.com>
11219 * config/i386/i386.c (x86_order_regs_for_local_alloc):
11222 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11224 PR tree-optimization/101189
11225 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Pass
11226 LHS range of condition to postfold routine.
11227 (fold_using_range::postfold_gcond_edges): Only process the TRUE or
11228 FALSE edge if the LHS range supports it being taken.
11229 * gimple-range-fold.h (postfold_gcond_edges): Add range parameter.
11231 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11233 * value-relation.cc (equiv_oracle::dump): Do not dump NULL blocks.
11234 (relation_oracle::find_relation_block): Check correct bitmap.
11235 (relation_oracle::dump): Do not dump NULL blocks.
11237 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11239 * gimple-range-cache.cc (ranger_cache::propagate_cache): Call
11240 range_on_edge instead of manually calculating.
11242 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11244 * range-op.cc: Fix comment.
11246 2021-06-24 Uroš Bizjak <ubizjak@gmail.com>
11249 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
11250 Handle V8QI and V4HI modes.
11251 * config/i386/mmx.md (sse4_1_<any_extend:code>v4qiv4hi2):
11253 (sse4_1_<any_extend:code>v4qiv4hi2): Ditto.
11254 (mmxpackmode): New mode attribute.
11255 (vec_pack_trunc_<mmxpackmode:mode>): New expander.
11256 (mmxunpackmode): New mode attribute.
11257 (vec_unpacks_lo_<mmxunpackmode:mode>): New expander.
11258 (vec_unpacks_hi_<mmxunpackmode:mode>): Ditto.
11259 (vec_unpacku_lo_<mmxunpackmode:mode>): Ditto.
11260 (vec_unpacku_hi_<mmxunpackmode:mode>): Ditto.
11261 * config/i386/i386.md (extsuffix): Move from ...
11262 * config/i386/sse.md: ... here.
11264 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
11266 * dwarf2out.c (dwarf2out_assembly_start): Emit .file 0 marker here..
11267 (dwarf2out_finish): ...instead of here.
11269 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
11271 * configure.ac (--gdwarf-5 option): Use objdump instead of readelf.
11272 (working --gdwarf-4/--gdwarf-5 for all sources): Likewise.
11273 (--gdwarf-4 not refusing generated .debug_line): Adjust for Windows.
11274 * configure: Regenerate.
11276 2021-06-24 Richard Biener <rguenther@suse.de>
11278 * config/i386/sse.md (vec_addsubv4df3, vec_addsubv2df3,
11279 vec_addsubv8sf3, vec_addsubv4sf3): Merge into ...
11280 (vec_addsub<mode>3): ... using a new addsub_cst mode attribute.
11282 2021-06-24 Richard Biener <rguenther@suse.de>
11284 * config/i386/sse.md (avx_addsubv4df3): Rename to
11286 (avx_addsubv8sf3): Rename to vec_addsubv8sf3.
11287 (sse3_addsubv2df3): Rename to vec_addsubv2df3.
11288 (sse3_addsubv4sf3): Rename to vec_addsubv4sf3.
11289 * config/i386/i386-builtin.def: Adjust.
11290 * internal-fn.def (VEC_ADDSUB): New internal optab fn.
11291 * optabs.def (vec_addsub_optab): New optab.
11292 * tree-vect-slp-patterns.c (class addsub_pattern): New.
11293 (slp_patterns): Add addsub_pattern.
11294 * tree-vect-slp.c (vect_optimize_slp): Disable propagation
11295 across CFN_VEC_ADDSUB.
11296 * tree-vectorizer.h (vect_pattern::vect_pattern): Make
11298 * doc/md.texi (vec_addsub<mode>3): Document.
11300 2021-06-24 Jakub Jelinek <jakub@redhat.com>
11302 PR middle-end/101170
11303 * df-scan.c (df_ref_record): For paradoxical big-endian SUBREGs
11304 where regno + subreg_regno_offset wraps around use 0 as starting
11307 2021-06-24 Jakub Jelinek <jakub@redhat.com>
11309 PR middle-end/101172
11310 * stor-layout.c (finish_bitfield_representative): If nextf has
11311 error_mark_node type, set repr type to error_mark_node too.
11313 2021-06-24 Ilya Leoshkevich <iii@linux.ibm.com>
11315 * config/s390/s390.c (s390_function_profiler): Ignore labelno
11317 * config/s390/s390.h (NO_PROFILE_COUNTERS): Define.
11319 2021-06-24 Richard Biener <rguenther@suse.de>
11321 * tree-vect-slp.c (vect_optimize_slp): Do not propagate
11322 across operations that have different semantics on different
11325 2021-06-24 Jakub Jelinek <jakub@redhat.com>
11327 * tree.h (OMP_CLAUSE_MAP_IN_REDUCTION): Document meaning for OpenMP.
11328 * gimplify.c (gimplify_scan_omp_clauses): For OpenMP map clauses
11329 with OMP_CLAUSE_MAP_IN_REDUCTION flag partially defer gimplification
11330 of non-decl OMP_CLAUSE_DECL. For OMP_CLAUSE_IN_REDUCTION on
11331 OMP_TARGET user outer_ctx instead of ctx for placeholders and
11332 initializer/combiner gimplification.
11333 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_MAP_IN_REDUCTION
11334 on target constructs.
11335 (lower_rec_input_clauses): Likewise.
11336 (lower_omp_target): Likewise.
11337 * omp-expand.c (expand_omp_target): Temporarily ignore nowait clause
11338 on target if in_reduction is present.
11340 2021-06-24 Kewen Lin <linkw@linux.ibm.com>
11342 * tree-predcom.c (class pcom_worker): New class.
11343 (release_chain): Renamed to...
11344 (pcom_worker::release_chain): ...this.
11345 (release_chains): Renamed to...
11346 (pcom_worker::release_chains): ...this.
11347 (aff_combination_dr_offset): Renamed to...
11348 (pcom_worker::aff_combination_dr_offset): ...this.
11349 (determine_offset): Renamed to...
11350 (pcom_worker::determine_offset): ...this.
11351 (class comp_ptrs): New class.
11352 (split_data_refs_to_components): Renamed to...
11353 (pcom_worker::split_data_refs_to_components): ...this,
11354 and update with class comp_ptrs.
11355 (suitable_component_p): Renamed to...
11356 (pcom_worker::suitable_component_p): ...this.
11357 (filter_suitable_components): Renamed to...
11358 (pcom_worker::filter_suitable_components): ...this.
11359 (valid_initializer_p): Renamed to...
11360 (pcom_worker::valid_initializer_p): ...this.
11361 (find_looparound_phi): Renamed to...
11362 (pcom_worker::find_looparound_phi): ...this.
11363 (add_looparound_copies): Renamed to...
11364 (pcom_worker::add_looparound_copies): ...this.
11365 (determine_roots_comp): Renamed to...
11366 (pcom_worker::determine_roots_comp): ...this.
11367 (determine_roots): Renamed to...
11368 (pcom_worker::determine_roots): ...this.
11369 (single_nonlooparound_use): Renamed to...
11370 (pcom_worker::single_nonlooparound_use): ...this.
11371 (remove_stmt): Renamed to...
11372 (pcom_worker::remove_stmt): ...this.
11373 (execute_pred_commoning_chain): Renamed to...
11374 (pcom_worker::execute_pred_commoning_chain): ...this.
11375 (execute_pred_commoning): Renamed to...
11376 (pcom_worker::execute_pred_commoning): ...this.
11377 (struct epcc_data): New member worker.
11378 (execute_pred_commoning_cbck): Call execute_pred_commoning
11379 with pcom_worker pointer.
11380 (find_use_stmt): Renamed to...
11381 (pcom_worker::find_use_stmt): ...this.
11382 (find_associative_operation_root): Renamed to...
11383 (pcom_worker::find_associative_operation_root): ...this.
11384 (find_common_use_stmt): Renamed to...
11385 (pcom_worker::find_common_use_stmt): ...this.
11386 (combinable_refs_p): Renamed to...
11387 (pcom_worker::combinable_refs_p): ...this.
11388 (reassociate_to_the_same_stmt): Renamed to...
11389 (pcom_worker::reassociate_to_the_same_stmt): ...this.
11390 (stmt_combining_refs): Renamed to...
11391 (pcom_worker::stmt_combining_refs): ...this.
11392 (combine_chains): Renamed to...
11393 (pcom_worker::combine_chains): ...this.
11394 (try_combine_chains): Renamed to...
11395 (pcom_worker::try_combine_chains): ...this.
11396 (prepare_initializers_chain): Renamed to...
11397 (pcom_worker::prepare_initializers_chain): ...this.
11398 (prepare_initializers): Renamed to...
11399 (pcom_worker::prepare_initializers): ...this.
11400 (prepare_finalizers_chain): Renamed to...
11401 (pcom_worker::prepare_finalizers_chain): ...this.
11402 (prepare_finalizers): Renamed to...
11403 (pcom_worker::prepare_finalizers): ...this.
11404 (tree_predictive_commoning_loop): Renamed to...
11405 (pcom_worker::tree_predictive_commoning_loop): ...this, adjust
11406 some calls and remove some cleanup code.
11407 (tree_predictive_commoning): Adjusted to use pcom_worker instance.
11408 (static variable looparound_phis): Remove.
11409 (static variable name_expansions): Remove.
11411 2021-06-24 Richard Biener <rguenther@suse.de>
11413 * tree-vect-slp.c (slpg_vertex): New struct.
11414 (vect_slp_build_vertices): Adjust.
11415 (vect_optimize_slp): Likewise. Maintain an outgoing permute
11416 and a materialized one.
11418 2021-06-24 Richard Biener <rguenther@suse.de>
11420 PR tree-optimization/101105
11421 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
11422 Only ignore steps when they are equal or scalar order is preserved.
11424 2021-06-24 liuhongt <hongtao.liu@intel.com>
11427 * config/i386/i386-expand.c (ix86_expand_vec_interleave):
11428 Adjust comments for ix86_expand_vecop_qihi2.
11429 (ix86_expand_vecmul_qihi): Renamed to ..
11430 (ix86_expand_vecop_qihi2): Adjust function prototype to
11431 support shift operation, add static to definition.
11432 (ix86_expand_vec_shift_qihi_constant): Add static to definition.
11433 (ix86_expand_vecop_qihi): Call ix86_expand_vecop_qihi2 and
11434 ix86_expand_vec_shift_qihi_constant.
11435 * config/i386/i386-protos.h (ix86_expand_vecmul_qihi): Deleted.
11436 (ix86_expand_vec_shift_qihi_constant): Deleted.
11437 * config/i386/sse.md (VI12_256_512_AVX512VL): New mode
11439 (mulv8qi3): Call ix86_expand_vecop_qihi directly, add
11440 condition TARGET_64BIT.
11441 (mul<mode>3): Ditto.
11442 (<insn><mode>3): Ditto.
11443 (vlshr<mode>3): Extend to support avx512 vlshr.
11444 (v<insn><mode>3): New expander for
11446 (v<insn>v8qi3): Ditto.
11447 (vashrv8hi3<mask_name>): Renamed to ..
11448 (vashr<mode>3): And extend to support V16QImode for avx512.
11449 (vashrv16qi3): Deleted.
11450 (vashrv2di3<mask_name>): Extend expander to support avx512
11453 2021-06-23 Dimitar Dimitrov <dimitar@dinux.eu>
11455 * doc/lto.texi (Design Overview): Update that slim objects are
11458 2021-06-23 Aaron Sawdey <acsawdey@linux.ibm.com>
11460 * config/rs6000/rs6000-cpus.def: Take OPTION_MASK_PCREL_OPT out
11461 of OTHER_POWER10_MASKS so it will not be enabled by default.
11463 2021-06-23 Richard Biener <rguenther@suse.de>
11464 Martin Jambor <mjambor@suse.cz>
11466 * tree-inline.c (setup_one_parameter): Set TREE_READONLY of the
11467 param replacement unconditionally. Adjust comment.
11469 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11471 * Makefile.in (OBJS): Add gimple-range-fold.o
11472 * gimple-range-fold.cc: New.
11473 * gimple-range-fold.h: New.
11474 * gimple-range-gori.cc (gimple_range_calc_op1): Move to here.
11475 (gimple_range_calc_op2): Ditto.
11476 * gimple-range-gori.h: Move prototypes to here.
11477 * gimple-range.cc: Adjust include files.
11478 (fur_source:fur_source): Relocate to gimple-range-fold.cc.
11479 (fur_source::get_operand): Ditto.
11480 (fur_source::get_phi_operand): Ditto.
11481 (fur_source::query_relation): Ditto.
11482 (fur_source::register_relation): Ditto.
11483 (class fur_edge): Ditto.
11484 (fur_edge::fur_edge): Ditto.
11485 (fur_edge::get_operand): Ditto.
11486 (fur_edge::get_phi_operand): Ditto.
11487 (fur_stmt::fur_stmt): Ditto.
11488 (fur_stmt::get_operand): Ditto.
11489 (fur_stmt::get_phi_operand): Ditto.
11490 (fur_stmt::query_relation): Ditto.
11491 (class fur_depend): Relocate to gimple-range-fold.h.
11492 (fur_depend::fur_depend): Relocate to gimple-range-fold.cc.
11493 (fur_depend::register_relation): Ditto.
11494 (fur_depend::register_relation): Ditto.
11495 (class fur_list): Ditto.
11496 (fur_list::fur_list): Ditto.
11497 (fur_list::get_operand): Ditto.
11498 (fur_list::get_phi_operand): Ditto.
11499 (fold_range): Ditto.
11500 (adjust_pointer_diff_expr): Ditto.
11501 (gimple_range_adjustment): Ditto.
11502 (gimple_range_base_of_assignment): Ditto.
11503 (gimple_range_operand1): Ditto.
11504 (gimple_range_operand2): Ditto.
11505 (gimple_range_calc_op1): Relocate to gimple-range-gori.cc.
11506 (gimple_range_calc_op2): Ditto.
11507 (fold_using_range::fold_stmt): Relocate to gimple-range-fold.cc.
11508 (fold_using_range::range_of_range_op): Ditto.
11509 (fold_using_range::range_of_address): Ditto.
11510 (fold_using_range::range_of_phi): Ditto.
11511 (fold_using_range::range_of_call): Ditto.
11512 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
11513 (fold_using_range::range_of_builtin_call): Ditto.
11514 (fold_using_range::range_of_cond_expr): Ditto.
11515 (fold_using_range::range_of_ssa_name_with_loop_info): Ditto.
11516 (fold_using_range::relation_fold_and_or): Ditto.
11517 (fold_using_range::postfold_gcond_edges): Ditto.
11518 * gimple-range.h: Add gimple-range-fold.h to include files. Change
11519 GIMPLE_RANGE_STMT_H to GIMPLE_RANGE_H.
11520 (gimple_range_handler): Relocate to gimple-range-fold.h.
11521 (gimple_range_ssa_p): Ditto.
11522 (range_compatible_p): Ditto.
11523 (class fur_source): Ditto.
11524 (class fur_stmt): Ditto.
11525 (class fold_using_range): Ditto.
11526 (gimple_range_calc_op1): Relocate to gimple-range-gori.h
11527 (gimple_range_calc_op2): Ditto.
11529 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11531 PR tree-optimization/101148
11532 PR tree-optimization/101014
11533 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
11534 (ranger_cache::~ranger_cache): Adjust.
11535 (ranger_cache::block_range): Check if propagation disallowed.
11536 (ranger_cache::propagate_cache): Disallow propagation if new value
11537 can't be stored properly.
11538 * gimple-range-cache.h (ranger_cache::m_propfail): New member.
11540 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11542 * gimple-range-cache.cc (class ssa_block_ranges): Adjust prototype.
11543 (sbr_vector::set_bb_range): Return true.
11544 (class sbr_sparse_bitmap): Adjust.
11545 (sbr_sparse_bitmap::set_bb_range): Return value.
11546 (block_range_cache::set_bb_range): Return value.
11547 (ranger_cache::propagate_cache): Use return value to print msg.
11548 * gimple-range-cache.h (class block_range_cache): Adjust.
11550 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11552 * gimple-range.cc (dump_bb): Use range_on_edge from the cache.
11554 2021-06-23 Jeff Law <jeffreyalaw@gmail.com>
11556 * config/h8300/logical.md (<code><mode>3<ccnz>): Use <cczn>
11557 so this pattern can be used for test/compare removal. Pass
11558 current insn to compute_logical_op_length and output_logical_op.
11559 * config/h8300/h8300.c (compute_logical_op_cc): Remove.
11560 (h8300_and_costs): Add argument to compute_logical_op_length.
11561 (output_logical_op): Add new argument. Use it to determine if the
11562 condition codes are used and adjust the output accordingly.
11563 (compute_logical_op_length): Add new argument and update length
11564 computations when condition codes are used.
11565 * config/h8300/h8300-protos.h (compute_logical_op_length): Update
11567 (output_logical_op): Likewise.
11569 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
11572 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
11573 Handle 64bit modes for TARGET_XOP. Use indirect gen_* functions.
11574 * config/i386/mmx.md (mmx_ppermv64): New insn pattern.
11575 * config/i386/i386.md (unspec): Move UNSPEC_XOP_PERMUTE from ...
11576 * config/i386/sse.md (unspec): ... here.
11578 2021-06-23 Martin Liska <mliska@suse.cz>
11581 * optc-save-gen.awk: Put back arm_fp16_format to
11584 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
11587 * config/i386/i386.md (bsr_rex64): Add zero-flag setting RTX.
11590 (clz<mode>2): Update RTX pattern for additions.
11592 2021-06-23 Jakub Jelinek <jakub@redhat.com>
11594 PR middle-end/101167
11595 * omp-low.c (lower_omp_regimplify_p): Regimplify also PARM_DECLs
11596 and RESULT_DECLs that have DECL_HAS_VALUE_EXPR_P set.
11598 2021-06-22 Sergei Trofimovich <siarheit@google.com>
11600 * doc/rtl.texi: drop unbalanced parenthesis.
11602 2021-06-22 Richard Biener <rguenther@suse.de>
11604 PR middle-end/101156
11605 * gimplify.c (gimplify_expr): Remove premature incorrect
11608 2021-06-22 Jakub Jelinek <jakub@redhat.com>
11610 PR tree-optimization/101159
11611 * tree-vect-patterns.c (vect_recog_popcount_pattern): Fix some
11614 2021-06-22 Jakub Jelinek <jakub@redhat.com>
11616 PR middle-end/101160
11617 * function.c (assign_parms): For decl_result with TYPE_EMPTY_P type
11618 clear crtl->return_rtx instead of keeping it referencing a pseudo.
11620 2021-06-22 Jakub Jelinek <jakub@redhat.com>
11621 Andrew Pinski <apinski@marvell.com>
11623 PR tree-optimization/101162
11624 * fold-const.c (range_check_type): Handle OFFSET_TYPE like pointer
11627 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11629 * range-op.cc (range_relational_tests): New.
11630 (range_op_tests): Call range_relational_tests.
11632 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11634 * range-op.cc (operator_cast::lhs_op1_relation): New.
11635 (operator_identity::lhs_op1_relation): Mew.
11637 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11639 * range-op.cc (operator_minus::op1_op2_relation_effect): New.
11641 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11643 * range-op.cc (operator_plus::lhs_op1_relation): New.
11644 (operator_plus::lhs_op2_relation): New.
11646 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11648 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a
11649 relation_oracle if dominators exist.
11650 (ranger_cache::~ranger_cache): Dispose of oracle.
11651 (ranger_cache::dump_bb): Dump oracle.
11652 * gimple-range.cc (fur_source::fur_source): New.
11653 (fur_source::get_operand): Use mmeber query.
11654 (fur_source::get_phi_operand): Use member_query.
11655 (fur_source::query_relation): New.
11656 (fur_source::register_dependency): Delete.
11657 (fur_source::register_relation): New.
11658 (fur_edge::fur_edge): Adjust.
11659 (fur_edge::get_phi_operand): Fix comment.
11660 (fur_edge::query): Delete.
11661 (fur_stmt::fur_stmt): Adjust.
11662 (fur_stmt::query): Delete.
11663 (fur_depend::fur_depend): Adjust.
11664 (fur_depend::register_relation): New.
11665 (fur_depend::register_relation): New.
11666 (fur_list::fur_list): Adjust.
11667 (fur_list::get_operand): Use member query.
11668 (fold_using_range::range_of_range_op): Process and query relations.
11669 (fold_using_range::range_of_address): Adjust dependency call.
11670 (fold_using_range::range_of_phi): Ditto.
11671 (gimple_ranger::gimple_ranger): New. Use ranger_ache oracle.
11672 (fold_using_range::relation_fold_and_or): New.
11673 (fold_using_range::postfold_gcond_edges): New.
11674 * gimple-range.h (class gimple_ranger): Adjust.
11675 (class fur_source): Adjust members.
11676 (class fur_stmt): Ditto.
11677 (class fold_using_range): Ditto.
11679 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11681 * range-op.cc (range_operator::wi_fold): Apply relation effect.
11682 (range_operator::fold_range): Adjust and apply relation effect.
11683 (*::fold_range): Add relation parameters.
11684 (*::op1_range): Ditto.
11685 (*::op2_range): Ditto.
11686 (range_operator::lhs_op1_relation): New.
11687 (range_operator::lhs_op2_relation): New.
11688 (range_operator::op1_op2_relation): New.
11689 (range_operator::op1_op2_relation_effect): New.
11690 (relop_early_resolve): New.
11691 (operator_equal::op1_op2_relation): New.
11692 (operator_equal::fold_range): Call relop_early_resolve.
11693 (operator_not_equal::op1_op2_relation): New.
11694 (operator_not_equal::fold_range): Call relop_early_resolve.
11695 (operator_lt::op1_op2_relation): New.
11696 (operator_lt::fold_range): Call relop_early_resolve.
11697 (operator_le::op1_op2_relation): New.
11698 (operator_le::fold_range): Call relop_early_resolve.
11699 (operator_gt::op1_op2_relation): New.
11700 (operator_gt::fold_range): Call relop_early_resolve.
11701 (operator_ge::op1_op2_relation): New.
11702 (operator_ge::fold_range): Call relop_early_resolve.
11703 * range-op.h (class range_operator): Adjust parameters and methods.
11705 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11707 * Makefile.in (OBJS): Add value-relation.o.
11708 * gimple-range.h: Adjust include files.
11709 * tree-data-ref.c: Adjust include file order.
11710 * value-query.cc (range_query::get_value_range): Default to no oracle.
11711 (range_query::query_relation): New.
11712 (range_query::query_relation): New.
11713 * value-query.h (class range_query): Adjust.
11714 * value-relation.cc: New.
11715 * value-relation.h: New.
11717 2021-06-22 Richard Biener <rguenther@suse.de>
11719 PR tree-optimization/101151
11720 * tree-ssa-sink.c (statement_sink_location): Expand irreducible
11723 2021-06-22 Jojo R <rjiejie@linux.alibaba.com>
11725 * config/riscv/riscv.c (thead_c906_tune_info): New.
11726 (riscv_tune_info_table): Use new tune.
11728 2021-06-22 Richard Biener <rguenther@suse.de>
11730 PR tree-optimization/101158
11731 * tree-vect-slp.c (vect_build_slp_tree_1): Move same operand
11732 checking after checking for matching operation.
11734 2021-06-22 Richard Biener <rguenther@suse.de>
11736 PR tree-optimization/101159
11737 * tree-vect-patterns.c (vect_recog_popcount_pattern): Add
11738 missing NULL vectype check.
11740 2021-06-22 Richard Biener <rguenther@suse.de>
11742 PR tree-optimization/101154
11743 * tree-vect-slp.c (vect_build_slp_tree_2): Fix out-of-bound access.
11745 2021-06-22 Jakub Jelinek <jakub@redhat.com>
11748 * config/i386/i386-protos.h (ix86_last_zero_store_uid): Declare.
11749 * config/i386/i386-expand.c (ix86_last_zero_store_uid): New variable.
11750 * config/i386/i386.c (ix86_expand_prologue): Clear it.
11751 * config/i386/i386.md (peephole2s for 1/2/4 stores of const0_rtx):
11752 Remove "" from match_operand. Emit new insns using emit_move_insn and
11753 set ix86_last_zero_store_uid to INSN_UID of the last store.
11754 Add peephole2s for 1/2/4 stores of const0_rtx following previous
11757 2021-06-22 Martin Liska <mliska@suse.cz>
11759 * auto-profile.c (AUTO_PROFILE_VERSION): Bump as string format
11762 2021-06-22 Martin Liska <mliska@suse.cz>
11764 * gcov-io.h: Remove padding entries.
11766 2021-06-22 liuhongt <hongtao.liu@intel.com>
11768 PR tree-optimization/97770
11769 * tree-vect-patterns.c (vect_recog_popcount_pattern):
11771 (vect_recog_func vect_vect_recog_func_ptrs): Add new pattern.
11773 2021-06-22 liuhongt <hongtao.liu@intel.com>
11776 * config/i386/i386-builtin.def (BDESC): Adjust builtin name.
11777 * config/i386/sse.md (<avx512>_expand<mode>_mask): Rename to ..
11778 (expand<mode>_mask): this ..
11779 (*expand<mode>_mask): New pre_reload splitter to transform
11780 v{,p}expand* to vmov* when mask is zero, all ones, or has all
11781 ones in it's lower part, otherwise still generate
11784 2021-06-22 liuhongt <hongtao.liu@intel.com>
11787 * config/i386/i386-expand.c
11788 (ix86_expand_special_args_builtin): Keep constm1_operand only
11789 if it satisfies insn's operand predicate.
11791 2021-06-21 Jason Merrill <jason@redhat.com>
11794 * df-scan.c (df_ref_record): Check that regno < endregno.
11795 * function.c (assign_parms, expand_function_end): Do nothing with a
11796 TYPE_EMPTY_P result.
11798 2021-06-21 Richard Biener <rguenther@suse.de>
11800 PR tree-optimization/101120
11801 * tree-vect-data-refs.c (bump_vector_ptr): Fold the
11803 * tree-vect-slp.c (vect_transform_slp_perm_load): Add
11804 DR chain DCE capability.
11805 * tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
11806 * tree-vect-stmts.c (vectorizable_load): Remove unused
11807 loads in the DR chain for SLP.
11809 2021-06-21 Jakub Jelinek <jakub@redhat.com>
11811 PR inline-asm/100785
11812 * gimplify.c (gimplify_asm_expr): Don't diagnose errors if
11813 output or input operands were already error_mark_node.
11814 * cfgexpand.c (expand_asm_stmt): If errors are emitted,
11815 remove all inputs, outputs and clobbers from the asm and
11816 set template to "".
11818 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
11820 * config/arm/arm_neon.h (vceq_s8): Replace builtin with __a == __b.
11821 (vceq_s16): Likewise.
11822 (vceq_s32): Likewise.
11823 (vceq_u8): Likewise.
11824 (vceq_u16): Likewise.
11825 (vceq_u32): Likewise.
11826 (vceq_p8): Likewise.
11827 (vceqq_s8): Likewise.
11828 (vceqq_s16): Likewise.
11829 (vceqq_s32): Likewise.
11830 (vceqq_u8): Likewise.
11831 (vceqq_u16): Likewise.
11832 (vceqq_u32): Likewise.
11833 (vceqq_p8): Likewise.
11834 (vceq_f32): Gate __a == __b on __FAST_MATH__.
11835 (vceqq_f32): Likewise.
11836 (vceq_f16): Likewise.
11837 (vceqq_f16): Likewise.
11839 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
11842 * config/arm/iterators.md (NEON_VACMP): Remove.
11843 * config/arm/neon.md (neon_vca<cmp_op><mode>): Use GLTE instead of GTGE
11845 (neon_vca<cmp_op><mode>_insn): Likewise.
11846 (neon_vca<cmp_op_unsp><mode>_insn_unspec): Use NEON_VAGLTE instead of
11849 2021-06-21 Richard Biener <rguenther@suse.de>
11851 PR tree-optimization/101121
11852 * tree-vect-slp.c (vect_build_slp_tree_2): To not fail fatally
11853 when we just lack a stmt with the desired op when doing permutation.
11854 (vect_build_slp_tree): When caching a failed SLP build attempt
11855 assert that at least one lane is marked as not matching.
11857 2021-06-21 liuhongt <hongtao.liu@intel.com>
11860 * config/i386/i386.md: (*anddi_1): Disparage slightly the mask
11861 register alternative.
11862 (*and<mode>_1): Ditto.
11864 (*andn<mode>_1): Ditto.
11865 (*<code><mode>_1): Ditto.
11866 (*<code>qi_1): Ditto.
11867 (*one_cmpl<mode>2_1): Ditto.
11868 (*one_cmplsi2_1_zext): Ditto.
11869 (*one_cmplqi2_1): Ditto.
11870 * config/i386/i386.c (x86_order_regs_for_local_alloc): Change
11871 the order of mask registers to be before general registers.
11873 2021-06-21 Roger Sayle <roger@nextmovesoftware.com>
11876 * config/i386/i386.md: New define_peephole2s to shrink writing
11877 1, 2 or 4 consecutive zeros to memory when optimizing for size.
11879 2021-06-18 Jeff Law <jeffreyalaw@gmail.com>
11881 * config/h8300/h8300.c (h8300_select_cc_mode): Handle SYMBOL_REF.
11882 * config/h8300/logical.md (<code><mode>3 logcial expander): Generate
11883 more efficient code when the source can be trivially simplified.
11885 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
11887 * gimple-range-cache.cc (ranger_cache::range_of_def): Calculate
11888 a range if global is not available.
11889 (ranger_cache::entry_range): Fallback to range_of_def.
11890 * gimple-range-cache.h (range_of_def): Adjust prototype.
11892 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
11894 PR tree-optimization/101014
11895 * gimple-range-cache.cc (ranger_cache::ranger_cache): Remove poor
11897 (ranger_cache::~ranger_cache): Ditto.
11898 (ranger_cache::enable_new_values): Delete.
11899 (ranger_cache::push_poor_value): Delete.
11900 (ranger_cache::range_of_def): Remove poor value processing.
11901 (ranger_cache::entry_range): Ditto.
11902 (ranger_cache::fill_block_cache): Ditto.
11903 * gimple-range-cache.h (class ranger_cache): Remove poor value members.
11904 * gimple-range.cc (gimple_ranger::range_of_expr): Remove call.
11905 * gimple-range.h (class gimple_ranger): Adjust.
11907 2021-06-18 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
11910 * common/config/arm/arm-common.c (arm_canon_arch_option_1): New function
11911 derived from arm_canon_arch.
11912 (arm_canon_arch_option): Call it.
11913 (arm_canon_arch_multilib_option): New function.
11914 * config/arm/arm-cpus.in (IGNORE_FOR_MULTILIB): New fgroup.
11915 * config/arm/arm.h (arm_canon_arch_multilib_option): New prototype.
11916 (CANON_ARCH_MULTILIB_SPEC_FUNCTION): New macro.
11917 (MULTILIB_ARCH_CANONICAL_SPECS): New macro.
11918 (DRIVER_SELF_SPECS): Add MULTILIB_ARCH_CANONICAL_SPECS.
11919 * config/arm/arm.opt (mlibarch): New option.
11920 * config/arm/t-rmprofile (MULTILIB_MATCHES): For armv8*-m, replace use
11921 of march on RHS with mlibarch.
11923 2021-06-18 Marcel Vollweiler <marcel@codesourcery.com>
11925 * config.in: Regenerate.
11926 * config/gcn/gcn.c (print_operand_address): Fix for global_load assembler
11928 * configure: Regenerate.
11929 * configure.ac: Fix for global_load assembler functions.
11931 2021-06-18 Richard Biener <rguenther@suse.de>
11933 PR tree-optimization/101112
11934 * tree-vect-slp.c (vect_slp_linearize_chain): Fix condition
11935 to lookup a pattern stmt def.
11937 2021-06-18 Jakub Jelinek <jakub@redhat.com>
11939 PR middle-end/101062
11940 * stor-layout.c (finish_bitfield_layout): Don't add bitfield
11941 representatives in QUAL_UNION_TYPE.
11943 2021-06-18 Andrew Pinski <apinski@marvell.com>
11945 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
11946 Add counting of how many times it is done.
11947 (factor_out_conditional_conversion): Likewise.
11948 (match_simplify_replacement): Likewise.
11949 (value_replacement): Likewise.
11950 (spaceship_replacement): Likewise.
11951 (cond_store_replacement): Likewise.
11952 (cond_if_else_store_replacement_1): Likewise.
11953 (hoist_adjacent_loads): Likewise.
11955 2021-06-18 Andrew Pinski <apinski@marvell.com>
11957 * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
11958 types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and CNONJ_EXPR.
11959 (verify_gimple_assign_binary): Reject point and offset types on
11960 MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
11961 FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
11962 FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.
11964 2021-06-18 Michael Meissner <meissner@linux.ibm.com>
11966 * config/rs6000/rs6000.c (rs6000_emit_minmax): Add support for ISA
11967 3.1 IEEE 128-bit floating point xsmaxcqp/xsmincqp instructions.
11968 * config/rs6000/rs6000.md (s<minmax><mode>3, IEEE128 iterator):
11971 2021-06-17 Aaron Sawdey <acsawdey@linux.ibm.com>
11973 * config/rs6000/genfusion.pl (gen_logical_addsubf): Add
11974 earlyclobber to alts 0/1.
11975 (gen_addadd): Add earlyclobber to alts 0/1.
11976 * config/rs6000/fusion.md: Regenerate file.
11978 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
11980 * cfgloopanal.c (get_loop_hot_path): Make path an auto_vec.
11982 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
11984 * gimple-range-cache.cc: Comment cleanups.
11985 * gimple-range-gori.cc: Comment cleanups.
11986 * gimple-range.cc: Comment/spacing cleanups
11987 * value-range.h: Comment cleanups.
11989 2021-06-17 H.J. Lu <hjl.tools@gmail.com>
11992 * calls.c (expand_call): Replace PUSH_ARGS with
11993 targetm.calls.push_argument (0).
11994 (emit_library_call_value_1): Likewise.
11995 * defaults.h (PUSH_ARGS): Removed.
11996 (PUSH_ARGS_REVERSED): Replace PUSH_ARGS with
11997 targetm.calls.push_argument (0).
11998 * expr.c (block_move_libcall_safe_for_call_parm): Likewise.
11999 (emit_push_insn): Pass the number bytes to push to
12000 targetm.calls.push_argument and pass 0 if ARGS_ADDR is 0.
12001 * hooks.c (hook_bool_uint_true): New.
12002 * hooks.h (hook_bool_uint_true): Likewise.
12003 * rtlanal.c (nonzero_bits1): Replace PUSH_ARGS with
12004 targetm.calls.push_argument (0).
12005 * target.def (push_argument): Add a targetm.calls hook.
12006 * targhooks.c (default_push_argument): New.
12007 * targhooks.h (default_push_argument): Likewise.
12008 * config/bpf/bpf.h (PUSH_ARGS): Removed.
12009 * config/cr16/cr16.c (TARGET_PUSH_ARGUMENT): New.
12010 * config/cr16/cr16.h (PUSH_ARGS): Removed.
12011 * config/i386/i386.c (ix86_push_argument): New.
12012 (TARGET_PUSH_ARGUMENT): Likewise.
12013 * config/i386/i386.h (PUSH_ARGS): Removed.
12014 * config/m32c/m32c.c (TARGET_PUSH_ARGUMENT): New.
12015 * config/m32c/m32c.h (PUSH_ARGS): Removed.
12016 * config/nios2/nios2.h (PUSH_ARGS): Likewise.
12017 * config/pru/pru.h (PUSH_ARGS): Likewise.
12018 * doc/tm.texi.in: Remove PUSH_ARGS documentation. Add
12019 TARGET_PUSH_ARGUMENT hook.
12020 * doc/tm.texi: Regenerated.
12022 2021-06-17 Uroš Bizjak <ubizjak@gmail.com>
12025 * config/i386/i386-expand.c (expand_vector_set_var):
12026 Handle V2FS mode remapping. Pass TARGET_MMX_WITH_SSE to
12027 ix86_expand_vector_init_duplicate.
12028 (ix86_expand_vector_init_duplicate): Emit insv_1 for
12029 QImode for !TARGET_PARTIAL_REG_STALL.
12030 * config/i386/predicates.md (vec_setm_mmx_operand): New predicate.
12031 * config/i386/mmx.md (vec_setv2sf): Use vec_setm_mmx_operand
12032 as operand 2 predicate. Call ix86_expand_vector_set_var
12033 for non-constant index operand.
12034 (vec_setv2si): Ditto.
12035 (vec_setv4hi): Ditto.
12036 (vec_setv8qi): ditto.
12038 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
12040 PR tree-optimization/100790
12041 * gimple-range.cc (range_of_builtin_call): Cleanup clz and ctz
12044 2021-06-17 Martin Liska <mliska@suse.cz>
12046 * doc/invoke.texi: Use consistently -O1 instead of -O.
12048 2021-06-17 Martin Liska <mliska@suse.cz>
12050 * gcov-io.h: Update documentation entry about string format.
12052 2021-06-17 Marius Hillenbrand <mhillen@linux.ibm.com>
12055 * config/s390/vecintrin.h (vec_doublee): Fix to use
12056 __builtin_s390_vflls.
12057 (vec_floate): Fix to use __builtin_s390_vflrd.
12059 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12061 * dominance.c (get_dominated_to_depth): Return auto_vec<basic_block>.
12062 * dominance.h (get_dominated_to_depth): Likewise.
12063 (get_all_dominated_blocks): Likewise.
12064 * cfgcleanup.c (delete_unreachable_blocks): Adjust.
12065 * gcse.c (hoist_code): Likewise.
12066 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
12067 * tree-parloops.c (oacc_entry_exit_ok): Likewise.
12068 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Likewise.
12069 * tree-ssa-phiprop.c (pass_phiprop::execute): Likewise.
12071 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12073 * dominance.c (get_dominated_by_region): Return auto_vec<basic_block>.
12074 * dominance.h (get_dominated_by_region): Likewise.
12075 * tree-cfg.c (gimple_duplicate_sese_region): Adjust.
12076 (gimple_duplicate_sese_tail): Likewise.
12077 (move_sese_region_to_fn): Likewise.
12079 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12081 * dominance.c (get_dominated_by): Return auto_vec<basic_block>.
12082 * dominance.h (get_dominated_by): Likewise.
12083 * auto-profile.c (afdo_find_equiv_class): Adjust.
12084 * cfgloopmanip.c (duplicate_loop_to_header_edge): Likewise.
12085 * loop-unroll.c (unroll_loop_runtime_iterations): Likewise.
12086 * tree-cfg.c (test_linear_chain): Likewise.
12087 (test_diamond): Likewise.
12089 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12091 * cfgloop.h (get_loop_hot_path): Return auto_vec<basic_block>.
12092 * cfgloopanal.c (get_loop_hot_path): Likewise.
12093 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
12095 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12097 * cgraph.c (cgraph_node::collect_callers): Return
12098 auto_vec<cgraph_edge *>.
12099 * cgraph.h (cgraph_node::collect_callers): Likewise.
12100 * ipa-cp.c (create_specialized_node): Adjust.
12101 (decide_about_value): Likewise.
12102 (decide_whether_version_node): Likewise.
12103 * ipa-sra.c (process_isra_node_results): Likewise.
12105 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12107 * vec.h (vl_ptr>::using_auto_storage): Handle null m_vec.
12108 (auto_vec<T, 0>::auto_vec): Define move constructor, and delete copy
12110 (auto_vec<T, 0>::operator=): Define move assignment and delete copy
12113 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
12115 * gimple-range.cc (debug_seed_ranger): New.
12116 (dump_ranger): New.
12117 (debug_ranger): New.
12119 2021-06-17 Richard Biener <rguenther@suse.de>
12121 PR tree-optimization/54400
12122 * tree-vectorizer.h (enum slp_instance_kind): Add
12123 slp_inst_kind_bb_reduc.
12124 (reduction_fn_for_scalar_code): Declare.
12125 * tree-vect-data-refs.c (vect_slp_analyze_instance_dependence):
12126 Check SLP_INSTANCE_KIND instead of looking at the
12128 (vect_slp_analyze_instance_alignment): Likewise.
12129 * tree-vect-loop.c (reduction_fn_for_scalar_code): Export.
12130 * tree-vect-slp.c (vect_slp_linearize_chain): Split out
12131 chain linearization from vect_build_slp_tree_2 and generalize
12132 for the use of BB reduction vectorization.
12133 (vect_build_slp_tree_2): Adjust accordingly.
12134 (vect_optimize_slp): Elide permutes at the root of BB reduction
12136 (vectorizable_bb_reduc_epilogue): New function.
12137 (vect_slp_prune_covered_roots): Likewise.
12138 (vect_slp_analyze_operations): Use them.
12139 (vect_slp_check_for_constructors): Recognize associatable
12140 chains for BB reduction vectorization.
12141 (vectorize_slp_instance_root_stmt): Generate code for the
12142 BB reduction epilogue.
12144 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
12146 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Check with
12148 (gori_compute::may_recompute_p): New.
12149 (gori_compute::outgoing_edge_range_p): Perform recomputations.
12150 * gimple-range-gori.h (class gori_compute): Add prototype.
12152 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
12154 * gimple-range-cache.cc (ranger_cache::range_on_edge): Always return
12155 true when a range can be calculated.
12156 * gimple-range.cc (gimple_ranger::dump_bb): Check has_edge_range_p.
12158 2021-06-16 Martin Sebor <msebor@redhat.com>
12160 * doc/invoke.texi (-Wmismatched-dealloc, -Wmismatched-new-delete):
12161 Correct documented defaults.
12163 2021-06-16 Andrew MacLeod <amacleod@redhat.com>
12165 * gimple-range-cache.cc (ranger_cache::ranger_cache): Initialize
12166 m_new_value_p directly.
12168 2021-06-16 Uroš Bizjak <ubizjak@gmail.com>
12171 * config/i386/i386-expand.c (expand_vec_perm_2perm_pblendv):
12172 Handle 64bit modes for TARGET_SSE4_1.
12173 (expand_vec_perm_pshufb2): Handle 64bit modes for TARGET_SSSE3.
12174 (expand_vec_perm_even_odd_pack): Handle V4HI mode.
12175 (expand_vec_perm_even_odd_1) <case E_V4HImode>: Expand via
12176 expand_vec_perm_pshufb2 for TARGET_SSSE3 and via
12177 expand_vec_perm_even_odd_pack for TARGET_SSE4_1.
12178 * config/i386/mmx.md (mmx_packusdw): New insn pattern.
12180 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12182 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn<mode>):
12183 Change to an expander that emits the correct instruction
12184 depending on endianness.
12185 (aarch64_<sur><addsub>hn<mode>_insn_le): Define.
12186 (aarch64_<sur><addsub>hn<mode>_insn_be): Define.
12188 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12190 * config/aarch64/aarch64-simd-builtins.def: Split generator
12191 for aarch64_<su>qmovn builtins into scalar and vector
12193 * config/aarch64/aarch64-simd.md (aarch64_<su>qmovn<mode>_insn_le):
12195 (aarch64_<su>qmovn<mode>_insn_be): Define.
12196 (aarch64_<su>qmovn<mode>): Split into scalar and vector
12197 variants. Change vector variant to an expander that emits the
12198 correct instruction depending on endianness.
12200 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12202 * config/aarch64/aarch64-simd-builtins.def: Split generator
12203 for aarch64_sqmovun builtins into scalar and vector variants.
12204 * config/aarch64/aarch64-simd.md (aarch64_sqmovun<mode>):
12205 Split into scalar and vector variants. Change vector variant
12206 to an expander that emits the correct instruction depending
12208 (aarch64_sqmovun<mode>_insn_le): Define.
12209 (aarch64_sqmovun<mode>_insn_be): Define.
12211 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12213 * config/aarch64/aarch64-simd.md (aarch64_xtn<mode>_insn_le):
12214 Define - modeling zero-high-half semantics.
12215 (aarch64_xtn<mode>): Change to an expander that emits the
12216 appropriate instruction depending on endianness.
12217 (aarch64_xtn<mode>_insn_be): Define - modeling zero-high-half
12219 (aarch64_xtn2<mode>_le): Rename to...
12220 (aarch64_xtn2<mode>_insn_le): This.
12221 (aarch64_xtn2<mode>_be): Rename to...
12222 (aarch64_xtn2<mode>_insn_be): This.
12223 (vec_pack_trunc_<mode>): Emit truncation instruction instead
12225 * config/aarch64/iterators.md (Vnarrowd): Add Vnarrowd mode
12226 attribute iterator.
12228 2021-06-16 Martin Jambor <mjambor@suse.cz>
12230 PR tree-optimization/100453
12231 * tree-sra.c (create_access): Disqualify any const candidates
12232 which are written to.
12233 (sra_modify_expr): Do not store sub-replacements back to a const base.
12234 (handle_unscalarized_data_in_subtree): Likewise.
12235 (sra_modify_assign): Likewise. Earlier, use TREE_READONLy test
12236 instead of constant_decl_p.
12238 2021-06-16 Jakub Jelinek <jakub@redhat.com>
12240 PR middle-end/101062
12241 * stor-layout.c (finish_bitfield_representative): For fields in unions
12242 assume nextf is always NULL.
12243 (finish_bitfield_layout): Compute bit field representatives also in
12244 unions, but handle it as if each bitfield was the only field in the
12247 2021-06-16 Richard Biener <rguenther@suse.de>
12249 PR tree-optimization/101088
12250 * tree-ssa-loop-im.c (sm_seq_valid_bb): Only look for
12251 supported refs on edges. Do not assert same ref but
12252 different kind stores are unsuported but mark them so.
12253 (hoist_memory_references): Only look for supported refs
12256 2021-06-16 Roger Sayle <roger@nextmovesoftware.com>
12258 PR rtl-optimization/46235
12259 * config/i386/i386.md: New define_split for bt followed by cmov.
12260 (*bt<mode>_setcqi): New define_insn_and_split for bt followed by setc.
12261 (*bt<mode>_setncqi): New define_insn_and_split for bt then setnc.
12262 (*bt<mode>_setnc<mode>): New define_insn_and_split for bt followed
12263 by setnc with zero extension.
12265 2021-06-16 Richard Biener <rguenther@suse.de>
12267 PR tree-optimization/101083
12268 * tree-vect-slp.c (vect_slp_build_two_operator_nodes): Get
12269 vectype as argument.
12270 (vect_build_slp_tree_2): Adjust.
12272 2021-06-15 Martin Sebor <msebor@redhat.com>
12274 PR middle-end/100876
12275 * builtins.c: (gimple_call_return_array): Account for size_t
12276 mangling as either unsigned int or unsigned long
12278 2021-06-15 Jeff Law <jeffreyalaw@gmail.com>
12280 * compare-elim.c (try_eliminate_compare): Run DCE to clean things
12281 up before eliminating comparisons.
12283 2021-06-15 Aldy Hernandez <aldyh@redhat.com>
12285 * range-op.cc (operator_bitwise_or::wi_fold): Make sure
12286 nonzero|X is nonzero.
12287 (range_op_bitwise_and_tests): Add tests for above.
12289 2021-06-15 Carl Love <cel@us.ibm.com>
12292 * config/rs6000/rs6000-builtin.def (VCMPEQUT): Fix the ICODE for the
12294 (VRLQ, VSLQ, VSRQ, VSRAQ): Remove unused BU_P10_OVERLOAD_2
12297 2021-06-15 Tobias Burnus <tobias@codesourcery.com>
12300 * gimplify.c (enum gimplify_defaultmap_kind): Add GDMK_SCALAR_TARGET.
12301 (struct gimplify_omp_ctx): Extend defaultmap array by one.
12302 (new_omp_context): Init defaultmap[GDMK_SCALAR_TARGET].
12303 (omp_notice_variable): Update type classification for Fortran.
12304 (gimplify_scan_omp_clauses): Update calls for new argument; handle
12305 GDMK_SCALAR_TARGET; for Fortran, GDMK_POINTER avoid GOVD_MAP_0LEN_ARRAY.
12306 * langhooks-def.h (lhd_omp_scalar_p): Add 'ptr_ok' argument.
12307 * langhooks.c (lhd_omp_scalar_p): Likewise.
12308 (LANG_HOOKS_OMP_ALLOCATABLE_P, LANG_HOOKS_OMP_SCALAR_TARGET_P): New.
12309 (LANG_HOOKS_DECLS): Add them.
12310 * langhooks.h (struct lang_hooks_for_decls): Add new hooks, update
12311 omp_scalar_p pointer type to include the new bool argument.
12313 2021-06-15 David Malcolm <dmalcolm@redhat.com>
12315 * doc/analyzer.texi
12316 (Special Functions for Debugging the Analyzer): Add
12317 __analyzer_dump_capacity.
12319 2021-06-15 Jakub Jelinek <jakub@redhat.com>
12322 * expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
12323 case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.
12325 2021-06-15 Richard Biener <rguenther@suse.de>
12327 * cfgloopanal.c (mark_irreducible_loops): Use a dominance
12328 check to identify loop latches.
12329 * cfgloop.c (verify_loop_structure): Likewise.
12330 * loop-init.c (apply_loop_flags): Allow marked irreducible
12331 regions even with multiple latches.
12332 * predict.c (rebuild_frequencies): Simplify.
12334 2021-06-15 Richard Biener <rguenther@suse.de>
12336 * tree-ssa-threadupdate.c
12337 (jump_thread_path_registry::mark_threaded_blocks): Assert we
12338 have marked irreducible regions.
12340 2021-06-14 Martin Sebor <msebor@redhat.com>
12343 * builtins.c (gimple_call_return_array): Check for attribute fn spec.
12344 Handle calls to placement new.
12345 (ndecl_dealloc_argno): Avoid placement delete.
12347 2021-06-14 Peter Bergner <bergner@linux.ibm.com>
12350 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
12351 create_tmp_reg_or_ssa_name().
12353 2021-06-14 Andrew MacLeod <amacleod@redhat.com>
12355 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
12356 (ranger_cache::enable_new_values): Set to specified value and
12357 return the old value.
12358 (ranger_cache::disable_new_values): Delete.
12359 (ranger_cache::fill_block_cache): Disable non 1st order derived
12361 * gimple-range-cache.h (ranger_cache): Adjust prototypes.
12362 * gimple-range.cc (gimple_ranger::range_of_expr): Adjust.
12364 2021-06-14 Uroš Bizjak <ubizjak@gmail.com>
12367 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
12368 Return true early when testing with V2HImode.
12369 * config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.
12371 2021-06-14 Christophe Lyon <christophe.lyon@linaro.org>
12373 * config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
12374 (mve_vec_unpack<US>_hi_<mode>): New pattern.
12375 (@mve_vec_pack_trunc_lo_<mode>): New pattern.
12376 (mve_vmovntq_<supf><mode>): Prefix with '@'.
12377 * config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
12379 (vec_unpack<US>_lo_<mode>): Likewise.
12380 (vec_pack_trunc_<mode>): Rename to
12381 neon_quad_vec_pack_trunc_<mode>.
12382 * config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
12384 (vec_unpack<US>_lo_<mode>): New.
12385 (vec_pack_trunc_<mode>): New.
12387 2021-06-14 Richard Biener <rguenther@suse.de>
12389 PR tree-optimization/100934
12390 * tree-ssa-dom.c (pass_dominator::execute): Properly
12391 mark irreducible regions.
12393 2021-06-14 Martin Liska <mliska@suse.cz>
12395 * doc/invoke.texi: Put r{...} on the same line as @item.
12397 2021-06-14 Martin Liska <mliska@suse.cz>
12399 * doc/invoke.texi: Add missing newline.
12401 2021-06-14 Martin Liska <mliska@suse.cz>
12403 * doc/invoke.texi: Remove '+' charasters.
12405 2021-06-14 Claudiu Zissulescu <claziss@synopsys.com>
12407 * config.gcc (arc): Add support for with_cpu option.
12408 * config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.
12410 2021-06-14 Richard Biener <rguenther@suse.de>
12412 PR tree-optimization/101031
12413 * tree-ssa-strlen.c (maybe_invalidate): Increment max_size
12414 instead of size when accounting for a possibly string
12417 2021-06-14 Martin Liska <mliska@suse.cz>
12419 * gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].
12421 2021-06-14 Aldy Hernandez <aldyh@redhat.com>
12423 * value-query.cc (gimple_range_global): Call get_range_global
12424 if called after inlining.
12426 2021-06-13 Uroš Bizjak <ubizjak@gmail.com>
12429 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
12430 Emit constant permutation insn directly from here.
12432 2021-06-13 Trevor Saunders <tbsaunde@tbsaunde.org>
12434 * attribs.c (find_attribute_namespace): Iterate over vec<> with
12436 * auto-profile.c (afdo_find_equiv_class): Likewise.
12437 * gcc.c (do_specs_vec): Likewise.
12438 (do_spec_1): Likewise.
12439 (driver::set_up_specs): Likewise.
12440 * gimple-loop-jam.c (any_access_function_variant_p): Likewise.
12441 * gimple-ssa-store-merging.c (compatible_load_p): Likewise.
12442 (imm_store_chain_info::try_coalesce_bswap): Likewise.
12443 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
12444 (get_location_for_stmts): Likewise.
12445 * graphite-poly.c (print_iteration_domains): Likewise.
12446 (free_poly_bb): Likewise.
12447 (remove_gbbs_in_scop): Likewise.
12448 (free_scop): Likewise.
12449 (dump_gbb_cases): Likewise.
12450 (dump_gbb_conditions): Likewise.
12451 (print_pdrs): Likewise.
12452 (print_scop): Likewise.
12453 * ifcvt.c (cond_move_process_if_block): Likewise.
12454 * lower-subreg.c (decompose_multiword_subregs): Likewise.
12455 * regcprop.c (pass_cprop_hardreg::execute): Likewise.
12456 * sanopt.c (sanitize_rewrite_addressable_params): Likewise.
12457 * sel-sched-dump.c (dump_insn_vector): Likewise.
12458 * store-motion.c (store_ops_ok): Likewise.
12459 (store_killed_in_insn): Likewise.
12460 * timevar.c (timer::named_items::print): Likewise.
12461 * tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
12462 (cleanup_tree_cfg_noloop): Likewise.
12463 * tree-data-ref.c (dump_data_references): Likewise.
12464 (print_dir_vectors): Likewise.
12465 (print_dist_vectors): Likewise.
12466 (dump_data_dependence_relations): Likewise.
12467 (dump_dist_dir_vectors): Likewise.
12468 (dump_ddrs): Likewise.
12469 (create_runtime_alias_checks): Likewise.
12470 (free_subscripts): Likewise.
12471 (save_dist_v): Likewise.
12472 (save_dir_v): Likewise.
12473 (invariant_access_functions): Likewise.
12474 (same_access_functions): Likewise.
12475 (access_functions_are_affine_or_constant_p): Likewise.
12476 (find_data_references_in_stmt): Likewise.
12477 (graphite_find_data_references_in_stmt): Likewise.
12478 (free_dependence_relations): Likewise.
12479 (free_data_refs): Likewise.
12480 * tree-inline.c (copy_debug_stmts): Likewise.
12481 * tree-into-ssa.c (dump_currdefs): Likewise.
12482 (rewrite_update_phi_arguments): Likewise.
12483 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
12484 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
12486 (vect_slp_analyze_node_dependences): Likewise.
12487 (vect_slp_analyze_instance_dependence): Likewise.
12488 (vect_record_base_alignments): Likewise.
12489 (vect_get_peeling_costs_all_drs): Likewise.
12490 (vect_peeling_supportable): Likewise.
12491 * tree-vectorizer.c (vec_info::~vec_info): Likewise.
12492 (vec_info::free_stmt_vec_infos): Likewise.
12494 2021-06-13 Jeff Law <jeffreyalaw@gmail.com>
12496 * config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
12497 (andqi3_1<cczn>): Removed.
12498 (<ors>qi3_1): Do not split for IOR/XOR a single bit.
12499 (H8/SX bit logicals): Split out from other patterns.
12500 * config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
12501 mulqihi3_const_clobber_flags.
12502 (mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly
12504 2021-06-13 H.J. Lu <hjl.tools@gmail.com>
12507 * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
12508 to true if red zone is used.
12509 (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
12510 ix86_red_zone_used.
12511 * config/i386/i386.h (machine_function): Add red_zone_used.
12512 (ix86_red_zone_size): Removed.
12513 (ix86_red_zone_used): New.
12514 * config/i386/i386.md (peephole2 patterns): Replace
12515 ix86_red_zone_size with ix86_red_zone_used.
12517 2021-06-12 Jason Merrill <jason@redhat.com>
12519 * doc/extend.texi (unused variable attribute): Applies to
12520 structure fields as well.
12522 2021-06-12 Eugene Rozenfeld <erozen@microsoft.com>
12524 * auto-profile.c (read_profile): fix a typo in an error string
12526 2021-06-11 Thomas Schwinge <thomas@codesourcery.com>
12528 * tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
12530 * tree-pretty-print.c (dump_omp_clauses): Update.
12531 (dump_generic_node) <OMP_CLAUSE>: Use it.
12533 2021-06-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
12536 * config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
12537 int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
12538 the polymorphic variants matching code.
12539 (__arm_vld1q_z): Likewise.
12540 (__arm_vld2q): Likewise.
12541 (__arm_vld4q): Likewise.
12542 (__arm_vldrbq_gather_offset): Likewise.
12543 (__arm_vldrbq_gather_offset_z): Likewise.
12545 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
12547 PR tree-optimization/96392
12548 * fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.
12550 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
12552 PR tree-optimization/96392
12553 * fold-const.c (fold_real_zero_addition_p): Take both arguments
12554 of the addition or subtraction, not just the zero. Use this
12555 other argument in tests for signaling NaNs and signed zeros.
12556 (tree_expr_maybe_real_minus_zero_p): New predicate.
12557 * fold-const.h (fold_real_zero_addition_p): Update prototype.
12558 (tree_expr_maybe_real_minus_zero_p): New function prototype.
12559 * match.pd: Update calls to fold_real_zero_addition_p.
12560 Replace HONOR_NANS with tree_expr_maybe_nan_p.
12561 Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
12562 Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
12563 * tree-ssa-reassoc.c (eliminate_using_constants): Update
12564 call to fold_real_zero_addition_p.
12566 2021-06-11 Richard Biener <rguenther@suse.de>
12568 PR tree-optimization/101025
12569 * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
12570 all refs that require dependence checking.
12572 2021-06-11 Richard Biener <rguenther@suse.de>
12574 PR tree-optimization/101028
12575 * tree-vect-slp.c (vect_build_slp_tree_2): When SLP
12576 reassoc discovery fails fatally, mark appropriate lanes
12579 2021-06-11 Richard Biener <rguenther@suse.de>
12581 PR tree-optimization/101026
12582 * tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
12583 have a representative for the associated chain nodes.
12585 2021-06-11 Jakub Jelinek <jakub@redhat.com>
12587 PR rtl-optimization/101008
12588 * simplify-rtx.c (relational_result): New function.
12589 (simplify_logical_relational_operation,
12590 simplify_relational_operation): Use it.
12592 2021-06-11 Jakub Jelinek <jakub@redhat.com>
12595 * config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.
12597 2021-06-11 Uroš Bizjak <ubizjak@gmail.com>
12600 * config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
12601 false if the permutation can be implemented with constant
12602 permutation instruction in wider mode.
12603 (canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
12604 Handle V8QImode and V4HImode.
12606 2021-06-11 Martin Liska <mliska@suse.cz>
12608 PR gcov-profile/100788
12609 * common.opt: Add new option.
12610 * coverage.c (coverage_begin_function): Emit warning instead on
12611 the internal compiler error.
12612 * doc/invoke.texi: Document the option.
12613 * toplev.c (process_options): Enable it by default.
12615 2021-06-11 Richard Biener <rguenther@suse.de>
12617 PR middle-end/101009
12618 * tree-data-ref.c (build_classic_dist_vector_1): Make sure
12619 to set *init_b to true when we encounter a constant equal
12621 (compute_affine_dependence): Also dump the actual DR_REF.
12623 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
12625 PR tree-optimization/100984
12626 * gimple-ssa-evrp.c (ssa_equiv_stack): Use auto_vec for
12627 replacements table.
12628 (ssa_equiv_stack::~ssa_equiv_stack): Remove.
12630 2021-06-11 Kewen Lin <linkw@linux.ibm.com>
12632 * config/rs6000/rs6000.md
12633 (floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
12634 define_insn_and_split.
12636 2021-06-11 Richard Biener <rguenther@suse.de>
12638 * tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
12639 to sort operands of the associative chain.
12641 2021-06-11 Richard Biener <rguenther@suse.de>
12643 * system.h (gcc_stablesort_r): Declare.
12644 * sort.cc (gcc_sort_r): Support stable sort.
12645 (gcc_stablesort_r): Define.
12646 * vec.h (vec<>::stablesort): Add.
12648 2021-06-10 Uroš Bizjak <ubizjak@gmail.com>
12651 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
12652 Handle V2SF mode. Emit SHUFPS to fixup unpack-high for V2SF mode.
12653 (expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
12654 (expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
12655 (expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
12656 (expand_vec_perm_interleave2): Handle 64bit modes.
12657 (expand_vec_perm_even_odd_pack): Handle V8QI mode.
12658 (expand_vec_perm_even_odd_1): Ditto.
12659 (ix86_vectorize_vec_perm_const): Ditto.
12660 * config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
12661 * config/i386/sse.md: ... here.
12662 * config/i386/mmx.md (*vec_interleave_lowv2sf):
12663 New insn_and_split pattern.
12664 (*vec_interleave_highv2sf): Ditto.
12665 (mmx_pshufbv8qi3): New insn pattern.
12666 (*mmx_pblendw): Ditto.
12668 2021-06-10 Peter Bergner <bergner@linux.ibm.com>
12670 * config/rs6000/rs6000-builtin.def (build_pair): New built-in.
12671 (build_acc): Likewise.
12672 * config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
12673 source operands in little-endian mode.
12674 (rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
12675 (mma_init_builtins): Likewise.
12676 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
12677 ordering for the MMA assemble and build source operands.
12678 * doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
12680 (__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
12683 2021-06-10 Jeff Law <jeffreyalaw@gmail.com>
12685 * config/h8300/h8300.c (select_cc_mode): Handle MEM. Use
12687 * config/h8300/extensions.md: Replace _clobber_flags patterns
12690 2021-06-10 Robin Dapp <rdapp@linux.ibm.com>
12692 * config/s390/vector.md (vcond_mask_<mode><mode>): Change to
12693 (vcond_mask_<mode><tointvec>): this.
12695 2021-06-10 Andrew Stubbs <ams@codesourcery.com>
12696 Thomas Schwinge <thomas@codesourcery.com>
12698 * omp-builtins.def (BUILT_IN_GOACC_ENTER_EXIT_DATA): Split into...
12699 (BUILT_IN_GOACC_ENTER_DATA, BUILT_IN_GOACC_EXIT_DATA): ... these.
12700 * gimple.h (enum gf_mask): Split
12701 'GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA' into
12702 'GF_OMP_TARGET_KIND_OACC_ENTER_DATA' and
12703 'GF_OMP_TARGET_KIND_OACC_EXIT_DATA'.
12704 (is_gimple_omp_oacc): Update.
12705 * gimple-pretty-print.c (dump_gimple_omp_target): Likewise.
12706 * gimplify.c (gimplify_omp_target_update): Likewise.
12707 * omp-expand.c (expand_omp_target, build_omp_regions_1)
12708 (omp_make_gimple_edges): Likewise.
12709 * omp-low.c (check_omp_nesting_restrictions, lower_omp_target):
12712 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
12714 * value-query.cc (value_query::value_on_edge): Rename name to
12716 (range_query::range_on_edge): Same.
12717 (range_query::value_of_expr): Same.
12718 (range_query::value_on_edge): Same.
12719 * value-query.h (class value_query): Same.
12720 (class range_query): Same.
12722 2021-06-10 Richard Biener <rguenther@suse.de>
12724 PR tree-optimization/101003
12725 * tree-vect-slp.c (vect_build_slp_tree_2): Appropriately
12726 use the pattern stmt defs when linearizing a chain.
12728 2021-06-10 Jakub Jelinek <jakub@redhat.com>
12731 * ifcvt.c (noce_get_alt_condition, noce_try_abs): Use
12732 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
12734 2021-06-10 Clement Chigot <clement.chigot@atos.net>
12736 * config/rs6000/aix71.h (ASM_CPU_SPEC): Add Power10 directive.
12737 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
12739 2021-06-09 Andrew Pinski <apinski@marvell.com>
12741 PR tree-optimization/100925
12742 * match.pd (a ? CST1 : CST2): Limit transformations
12743 that would produce a negative to integeral types only.
12744 Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.
12746 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
12749 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
12751 * doc/tm.texi: Correctly update.
12753 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
12755 * doc/tm.texi: Correctly update.
12757 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
12760 * doc/tm.texi.in (Trampolines): Add a missing blank line.
12762 2021-06-09 Paul Eggert <eggert@cs.ucla.edu>
12765 * doc/invoke.texi (Code Gen Options); Document that -fno-trampolines
12766 and -ftrampolines work only with Ada.
12767 * doc/tm.texi.in (Trampolines): Likewise.
12768 * doc/tm.texi: Regenerated.
12770 2021-06-09 Carl Love <cel@us.ibm.com>
12772 * config/rs6000/altivec.h (vec_signextll, vec_signexti, vec_signextq):
12773 Add define for new builtins.
12774 * config/rs6000/altivec.md(altivec_vreveti2): Add define_expand.
12775 * config/rs6000/rs6000-builtin.def (VSIGNEXTI, VSIGNEXTLL): Add
12776 overloaded builtin definitions.
12777 (VSIGNEXTSB2W, VSIGNEXTSH2W, VSIGNEXTSB2D, VSIGNEXTSH2D,VSIGNEXTSW2D,
12778 VSIGNEXTSD2Q): Add builtin expansions.
12779 (SIGNEXT): Add P10 overload definition.
12780 * config/rs6000/rs6000-call.c (P9V_BUILTIN_VEC_VSIGNEXTI, P9V_BUILTIN_VEC_VSIGNEXTLL,
12781 P10_BUILTIN_VEC_SIGNEXT): Add overloaded argument definitions.
12782 * config/rs6000/vsx.md (vsx_sign_extend_v2di_v1ti): Add define_insn.
12783 (vsignextend_v2di_v1ti, vsignextend_qi_<mode>, vsignextend_hi_<mode>,
12784 vsignextend_si_v2di)[VIlong]: Add define_expand.
12785 Make define_insn vsx_sign_extend_si_v2di visible.
12786 * doc/extend.texi: Add documentation for the vec_signexti,
12787 vec_signextll builtins and vec_signextq.
12789 2021-06-09 Carl Love <cel@us.ibm.com>
12791 * config/rs6000/rs6000.c (__fixkfti, __fixunskfti, __floattikf,
12792 __floatuntikf): Names changed to __fixkfti_sw, __fixunskfti_sw,
12793 __floattikf_sw, __floatuntikf_sw respectively.
12794 * config/rs6000/rs6000.md (floatti<mode>2, floatunsti<mode>2,
12795 fix_trunc<mode>ti2, fixuns_trunc<mode>ti2): Add
12796 define_insn for mode IEEE 128.
12798 2021-06-09 Carl Love <cel@us.ibm.com>
12800 * config/rs6000/altivec.md (altivec_vslq, altivec_vsrq):
12801 Rename to altivec_vslq_<mode>, altivec_vsrq_<mode>, mode VEC_TI.
12802 * config/rs6000/vector.md (VEC_TI): Was named VSX_TI in vsx.md.
12803 (vashlv1ti3): Change to vashl<mode>3, mode VEC_TI.
12804 (vlshrv1ti3): Change to vlshr<mode>3, mode VEC_TI.
12805 * config/rs6000/vsx.md (VSX_TI): Remove define_mode_iterator. Update
12806 uses of VSX_TI to VEC_TI.
12808 2021-06-09 Carl Love <cel@us.ibm.com>
12810 * config/rs6000/dfp.md (floattitd2, fixtdti2): New define_insns.
12812 2021-06-09 Carl Love <cel@us.ibm.com>
12814 * config/rs6000/altivec.h (vec_dive, vec_mod): Add define for new
12816 * config/rs6000/altivec.md (UNSPEC_VMULEUD, UNSPEC_VMULESD,
12817 UNSPEC_VMULOUD, UNSPEC_VMULOSD): New unspecs.
12818 (altivec_eqv1ti, altivec_gtv1ti, altivec_gtuv1ti, altivec_vmuleud,
12819 altivec_vmuloud, altivec_vmulesd, altivec_vmulosd, altivec_vrlq,
12820 altivec_vrlqmi, altivec_vrlqmi_inst, altivec_vrlqnm,
12821 altivec_vrlqnm_inst, altivec_vslq, altivec_vsrq, altivec_vsraq,
12822 altivec_vcmpequt_p, altivec_vcmpgtst_p, altivec_vcmpgtut_p): New
12824 (vec_widen_umult_even_v2di, vec_widen_smult_even_v2di,
12825 vec_widen_umult_odd_v2di, vec_widen_smult_odd_v2di, altivec_vrlqmi,
12826 altivec_vrlqnm): New define_expands.
12827 * config/rs6000/rs6000-builtin.def (VCMPEQUT_P, VCMPGTST_P,
12828 VCMPGTUT_P): Add macro expansions.
12829 (BU_P10V_AV_P): Add builtin predicate definition.
12830 (VCMPGTUT, VCMPGTST, VCMPEQUT, CMPNET, CMPGE_1TI,
12831 CMPGE_U1TI, CMPLE_1TI, CMPLE_U1TI, VNOR_V1TI_UNS, VNOR_V1TI, VCMPNET_P,
12832 VCMPAET_P, VMULEUD, VMULESD, VMULOUD, VMULOSD, VRLQ,
12833 VSLQ, VSRQ, VSRAQ, VRLQNM, DIV_V1TI, UDIV_V1TI, DIVES_V1TI, DIVEU_V1TI,
12834 MODS_V1TI, MODU_V1TI, VRLQMI): New macro expansions.
12835 (VRLQ, VSLQ, VSRQ, VSRAQ, DIVE, MOD): New overload expansions.
12836 * config/rs6000/rs6000-call.c (P10_BUILTIN_VCMPEQUT,
12837 P10V_BUILTIN_CMPGE_1TI, P10V_BUILTIN_CMPGE_U1TI,
12838 P10V_BUILTIN_VCMPGTUT, P10V_BUILTIN_VCMPGTST,
12839 P10V_BUILTIN_CMPLE_1TI, P10V_BUILTIN_VCMPLE_U1TI,
12840 P10V_BUILTIN_DIV_V1TI, P10V_BUILTIN_UDIV_V1TI,
12841 P10V_BUILTIN_VMULESD, P10V_BUILTIN_VMULEUD,
12842 P10V_BUILTIN_VMULOSD, P10V_BUILTIN_VMULOUD,
12843 P10V_BUILTIN_VNOR_V1TI, P10V_BUILTIN_VNOR_V1TI_UNS,
12844 P10V_BUILTIN_VRLQ, P10V_BUILTIN_VRLQMI,
12845 P10V_BUILTIN_VRLQNM, P10V_BUILTIN_VSLQ,
12846 P10V_BUILTIN_VSRQ, P10V_BUILTIN_VSRAQ,
12847 P10V_BUILTIN_VCMPGTUT_P, P10V_BUILTIN_VCMPGTST_P,
12848 P10V_BUILTIN_VCMPEQUT_P, P10V_BUILTIN_VCMPGTUT_P,
12849 P10V_BUILTIN_VCMPGTST_P, P10V_BUILTIN_CMPNET,
12850 P10V_BUILTIN_VCMPNET_P, P10V_BUILTIN_VCMPAET_P,
12851 P10V_BUILTIN_DIVES_V1TI, P10V_BUILTIN_MODS_V1TI,
12852 P10V_BUILTIN_MODU_V1TI):
12853 New overloaded definitions.
12854 (rs6000_gimple_fold_builtin) [P10V_BUILTIN_VCMPEQUT,
12855 P10V_BUILTIN_CMPNET, P10V_BUILTIN_CMPGE_1TI,
12856 P10V_BUILTIN_CMPGE_U1TI, P10V_BUILTIN_VCMPGTUT,
12857 P10V_BUILTIN_VCMPGTST, P10V_BUILTIN_CMPLE_1TI,
12858 P10V_BUILTIN_CMPLE_U1TI]: New case statements.
12859 (rs6000_init_builtins) [bool_V1TI_type_node, int_ftype_int_v1ti_v1ti]:
12861 (altivec_init_builtins): New E_V1TImode case statement.
12862 (builtin_function_type)[P10_BUILTIN_128BIT_VMULEUD,
12863 P10_BUILTIN_128BIT_VMULOUD, P10_BUILTIN_128BIT_DIVEU_V1TI,
12864 P10_BUILTIN_128BIT_MODU_V1TI, P10_BUILTIN_CMPGE_U1TI,
12865 P10_BUILTIN_VCMPGTUT, P10_BUILTIN_VCMPEQUT]: New case statements.
12866 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute) [E_TImode,
12867 E_V1TImode]: New case statements.
12868 * config/rs6000/rs6000.h (rs6000_builtin_type_index): New enum
12869 value RS6000_BTI_bool_V1TI.
12870 * config/rs6000/vector.md (vector_gtv1ti,vector_nltv1ti,
12871 vector_gtuv1ti, vector_nltuv1ti, vector_ngtv1ti, vector_ngtuv1ti,
12872 vector_eq_v1ti_p, vector_ne_v1ti_p, vector_ae_v1ti_p,
12873 vector_gt_v1ti_p, vector_gtu_v1ti_p, vrotlv1ti3, vashlv1ti3,
12874 vlshrv1ti3, vashrv1ti3): New define_expands.
12875 * config/rs6000/vsx.md (UNSPEC_VSX_DIVSQ, UNSPEC_VSX_DIVUQ,
12876 UNSPEC_VSX_DIVESQ, UNSPEC_VSX_DIVEUQ, UNSPEC_VSX_MODSQ,
12877 UNSPEC_VSX_MODUQ): New unspecs.
12878 (mulv2di3, vsx_div_v1ti, vsx_udiv_v1ti, vsx_dives_v1ti,
12879 vsx_diveu_v1ti, vsx_mods_v1ti, vsx_modu_v1ti, xxswapd_v1ti): New
12881 (vcmpnet): New define_expand.
12882 * doc/extend.texi: Add documentation for the new builtins vec_rl,
12883 vec_rlmi, vec_rlnm, vec_sl, vec_sr, vec_sra, vec_mule, vec_mulo,
12884 vec_div, vec_dive, vec_mod, vec_cmpeq, vec_cmpne, vec_cmpgt, vec_cmplt,
12885 vec_cmpge, vec_cmple, vec_all_eq, vec_all_ne, vec_all_gt, vec_all_lt,
12886 vec_all_ge, vec_all_le, vec_any_eq, vec_any_ne, vec_any_gt, vec_any_lt,
12887 vec_any_ge, vec_any_le.
12889 2021-06-09 Carl Love <cel@us.ibm.com>
12891 * config/rs6000/altivec.md (altivec_vrl<VI_char>mi): Fix
12892 bug in argument generation.
12894 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
12896 * config/arm/iterators.md (<supf>): Remove VCLZQ_U, VCLZQ_S.
12898 * config/arm/mve.md (mve_vclzq_<supf><mode>): Add '@' prefix,
12899 remove <supf> iterator.
12900 (mve_vclzq_u<mode>): New.
12901 * config/arm/neon.md (clz<mode>2): Rename to neon_vclz<mode>.
12902 (neon_vclz<mode): Move to ...
12903 * config/arm/unspecs.md (VCLZQ_U, VCLZQ_S): Remove.
12904 * config/arm/vec-common.md: ... here. Add support for MVE.
12906 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
12908 * config/arm/mve.md (mve_vhaddq_<supf><mode>): Prefix with '@'.
12909 (@mve_vrhaddq_<supf><mode): Likewise.
12910 * config/arm/neon.md (neon_v<r>hadd<sup><mode>): Likewise.
12911 * config/arm/vec-common.md (avg<mode>3_floor, uavg<mode>3_floor)
12912 (avg<mode>3_ceil", uavg<mode>3_ceil): New patterns.
12914 2021-06-09 imba-tjd <109224573@qq.com>
12916 * doc/invoke.texi: Fix typo.
12918 2021-06-09 Roger Sayle <roger@nextmovesoftware.com>
12920 PR middle-end/53267
12921 * fold-const-call.c (fold_const_call_sss) [CASE_CFN_FMOD]:
12922 Support evaluation of fmod/fmodf/fmodl at compile-time.
12924 2021-06-09 Richard Biener <rguenther@suse.de>
12926 PR tree-optimization/100981
12927 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
12928 gimple_get_lhs to also handle calls.
12929 * tree-vect-slp-patterns.c (complex_pattern::build): Transfer
12932 2021-06-09 Richard Biener <rguenther@suse.de>
12934 PR tree-optimization/97832
12935 * tree-vectorizer.h (_slp_tree::failed): New.
12936 * tree-vect-slp.c (_slp_tree::_slp_tree): Initialize
12938 (_slp_tree::~_slp_tree): Free failed.
12939 (vect_build_slp_tree): Retain failed nodes and record
12940 matches in them, copying that back out when running
12941 into a cached fail. Dump start and end of discovery.
12942 (dt_sort_cmp): New.
12943 (vect_build_slp_tree_2): Handle associatable chains
12944 together doing more aggressive operand swapping.
12946 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
12949 * config.gcc (gcc_cv_initfini_array): Set to yes for Linux and
12951 * doc/install.texi: Require glibc 2.1 and binutils 2.12 for
12952 Linux and GNU targets.
12954 2021-06-09 Richard Biener <rguenther@suse.de>
12956 * tree-vect-stmts.c (vect_is_simple_use): Always get dt
12959 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
12961 * config/arc/arc.md (loop_end): Change it to
12962 define_insn_and_split.
12964 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
12966 * config/arc/arc.md (maddhisi4): Use VMAC2H instruction.
12967 (machi): New pattern.
12968 (umaddhisi4): Use VMAC2HU instruction.
12969 (umachi): New pattern.
12971 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
12973 * config/arc/arc-protos.h (arc_split_move_p): New prototype.
12974 * config/arc/arc.c (arc_split_move_p): New function.
12975 (arc_split_move): Clean up.
12976 * config/arc/arc.md (movdi_insn): Clean up, use arc_split_move_p.
12977 (movdf_insn): Likewise.
12978 * config/arc/simdext.md (mov<VWH>_insn): Likewise.
12980 2021-06-09 Uroš Bizjak <ubizjak@gmail.com>
12983 * config/i386/i386.c (print_operand_address_as): Rename "no_rip"
12984 argument to "raw". Do not emit segment overrides when "raw" is true.
12986 2021-06-09 Martin Liska <mliska@suse.cz>
12988 * doc/gcov.texi: Create a proper JSON files.
12989 * doc/invoke.texi: Remove dots in order to make it a valid
12992 2021-06-09 Xionghu Luo <luoxhu@linux.ibm.com>
12994 * config/rs6000/rs6000-p8swap.c (pattern_is_rotate64): New.
12995 (insn_is_load_p): Use pattern_is_rotate64.
12996 (insn_is_swap_p): Likewise.
12997 (quad_aligned_load_p): Likewise.
12998 (const_load_sequence_p): Likewise.
12999 (replace_swapped_aligned_load): Likewise.
13000 (recombine_lvx_pattern): Likewise.
13001 (recombine_stvx_pattern): Likewise.
13003 2021-06-09 Andrew MacLeod <amacleod@redhat.com>
13005 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Use a
13006 fur_stmt source record.
13007 * gimple-range.cc (fur_source::get_operand): Generic range query.
13008 (fur_source::get_phi_operand): New.
13009 (fur_source::register_dependency): New.
13010 (fur_source::query): New.
13011 (class fur_edge): New. Edge source for operands.
13012 (fur_edge::fur_edge): New.
13013 (fur_edge::get_operand): New.
13014 (fur_edge::get_phi_operand): New.
13015 (fur_edge::query): New.
13016 (fur_stmt::fur_stmt): New.
13017 (fur_stmt::get_operand): New.
13018 (fur_stmt::get_phi_operand): New.
13019 (fur_stmt::query): New.
13020 (class fur_depend): New. Statement source and process dependencies.
13021 (fur_depend::fur_depend): New.
13022 (fur_depend::register_dependency): New.
13023 (class fur_list): New. List source for operands.
13024 (fur_list::fur_list): New.
13025 (fur_list::get_operand): New.
13026 (fur_list::get_phi_operand): New.
13027 (fold_range): New. Instantiate appropriate fur_source class and fold.
13028 (fold_using_range::range_of_range_op): Use new API.
13029 (fold_using_range::range_of_address): Ditto.
13030 (fold_using_range::range_of_phi): Ditto.
13031 (imple_ranger::fold_range_internal): Use fur_depend class.
13032 (fold_using_range::range_of_ssa_name_with_loop_info): Use new API.
13033 * gimple-range.h (class fur_source): Now a base class.
13034 (class fur_stmt): New.
13035 (fold_range): New prototypes.
13036 (fur_source::fur_source): Delete.
13038 2021-06-08 Andrew Pinski <apinski@marvell.com>
13040 PR tree-optimization/25290
13041 * tree-ssa-phiopt.c (xor_replacement): Delete.
13042 (tree_ssa_phiopt_worker): Delete use of xor_replacement.
13043 (match_simplify_replacement): Allow one cheap preparation
13044 statement that can be moved to before the if.
13046 2021-06-08 Pat Haugen <pthaugen@linux.ibm.com>
13048 * config/rs6000/power10.md (power10-fused-load, power10-fused-store,
13049 power10-fused_alu, power10-fused-vec, power10-fused-branch): New.
13051 2021-06-08 Jeff Law <jeffreyalaw@gmail.com>
13053 * config/h8300/logical.md (andqi3_1): Move BCLR case into define_insn_and_split.
13054 Create length attribute on define_insn_and_split. Only split for cases which we
13056 (andqi3_1<cczn>): Renamed from andqi3_1_clobber_flags. Only handle AND here and
13057 fix length computation.
13058 (b<code><mode>msx): Combine QImode and HImode H8/SX patterns using iterator.
13060 2021-06-08 Richard Biener <rguenther@suse.de>
13062 PR tree-optimization/100923
13063 * tree-ssa-sccvn.c (valueize_refs_1): Take a pointer to
13064 the operand vector to be valueized.
13065 (valueize_refs): Likewise.
13066 (valueize_shared_reference_ops_from_ref): Adjust.
13067 (valueize_shared_reference_ops_from_call): Likewise.
13068 (vn_reference_lookup_3): Likewise.
13069 (vn_reference_lookup_pieces): Likewise. Re-valueize
13070 with honoring availability when we are about to create
13071 the ao_ref and valueized before.
13072 (vn_reference_lookup): Likewise.
13073 (vn_reference_insert_pieces): Adjust.
13075 2021-06-08 Richard Biener <rguenther@suse.de>
13077 * tree-vectorizer.h (_slp_instance::root_stmt): Change to...
13078 (_slp_instance::root_stmts): ... a vector.
13079 (SLP_INSTANCE_ROOT_STMT): Rename to ...
13080 (SLP_INSTANCE_ROOT_STMTS): ... this.
13081 (slp_root::root): Change to...
13082 (slp_root::roots): ... a vector.
13083 (slp_root::slp_root): Adjust.
13084 * tree-vect-slp.c (_slp_instance::location): Adjust.
13085 (vect_free_slp_instance): Release the root stmt vector.
13086 (vect_build_slp_instance): Adjust.
13087 (vect_analyze_slp): Likewise.
13088 (_bb_vec_info::~_bb_vec_info): Likewise.
13089 (vect_slp_analyze_operations): Likewise.
13090 (vect_bb_vectorization_profitable_p): Likewise. Adjust
13091 costs for the root stmt.
13092 (vect_slp_check_for_constructors): Gather all BIT_INSERT_EXPRs
13094 (vect_slp_analyze_bb_1): Simplify by marking all root stmts
13096 (vectorize_slp_instance_root_stmt): Adjust.
13097 (vect_schedule_slp): Likewise.
13099 2021-06-08 Aldy Hernandez <aldyh@redhat.com>
13101 * gimple-ssa-evrp.c (class ssa_equiv_stack): New.
13102 (ssa_equiv_stack::ssa_equiv_stack): New.
13103 (ssa_equiv_stack::~ssa_equiv_stack): New.
13104 (ssa_equiv_stack::enter): New.
13105 (ssa_equiv_stack::leave): New.
13106 (ssa_equiv_stack::push_replacement): New.
13107 (ssa_equiv_stack::get_replacement): New.
13108 (is_pointer_ssa): New.
13109 (class pointer_equiv_analyzer): New.
13110 (pointer_equiv_analyzer::pointer_equiv_analyzer): New.
13111 (pointer_equiv_analyzer::~pointer_equiv_analyzer): New.
13112 (pointer_equiv_analyzer::set_global_equiv): New.
13113 (pointer_equiv_analyzer::set_cond_equiv): New.
13114 (pointer_equiv_analyzer::get_equiv): New.
13115 (pointer_equiv_analyzer::enter): New.
13116 (pointer_equiv_analyzer::leave): New.
13117 (pointer_equiv_analyzer::get_equiv_expr): New.
13118 (pta_valueize): New.
13119 (pointer_equiv_analyzer::visit_stmt): New.
13120 (pointer_equiv_analyzer::visit_edge): New.
13121 (hybrid_folder::value_of_expr): Call PTA.
13122 (hybrid_folder::value_on_edge): Same.
13123 (hybrid_folder::pre_fold_bb): New.
13124 (hybrid_folder::post_fold_bb): New.
13125 (hybrid_folder::pre_fold_stmt): New.
13126 (rvrp_folder::pre_fold_bb): New.
13127 (rvrp_folder::post_fold_bb): New.
13128 (rvrp_folder::pre_fold_stmt): New.
13129 (rvrp_folder::value_of_expr): Call PTA.
13130 (rvrp_folder::value_on_edge): Same.
13132 2021-06-08 Jakub Jelinek <jakub@redhat.com>
13135 * tree-inline.c (copy_tree_body_r): For OMP_CLAUSE_DEPEND don't
13136 check TREE_CODE if OMP_CLAUSE_DECL is NULL.
13138 2021-06-08 Richard Biener <rguenther@suse.de>
13140 PR middle-end/100951
13141 * tree-vect-generic.c (expand_vector_piecewise): Build a
13142 VECTOR_CST if all elements are constant.
13143 (expand_vector_condition): Likewise.
13144 (lower_vec_perm): Likewise.
13145 (expand_vector_conversion): Likewise.
13147 2021-06-08 Martin Liska <mliska@suse.cz>
13149 * doc/invoke.texi: Document new param evrp-sparse-threshold.
13151 2021-06-08 Martin Liska <mliska@suse.cz>
13153 * genautomata.c (create_automata): Fix typo.
13155 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
13157 PR tree-optimization/100794
13158 * tree-predcom.c (tree_predictive_commoning_loop): Add parameter
13159 allow_unroll_p and only allow unrolling when it's true.
13160 (tree_predictive_commoning): Add parameter allow_unroll_p and
13162 (run_tree_predictive_commoning): Likewise.
13163 (pass_predcom::gate): Check flag_tree_loop_vectorize and
13164 global_options_set.x_flag_predictive_commoning.
13165 (pass_predcom::execute): Adjust for allow_unroll_p.
13167 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
13169 * tree-predcom.c (execute_pred_commoning): Remove update_ssa call.
13170 (tree_predictive_commoning_loop): Factor some cleanup stuffs into
13171 lambda function cleanup, remove scev_reset call, and adjust return
13173 (tree_predictive_commoning): Adjust for different changed values,
13174 only set flag TODO_update_ssa_only_virtuals if changed.
13175 (pass_data pass_data_predcom): Remove TODO_update_ssa_only_virtuals
13176 from todo_flags_finish.
13178 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
13180 * gimple-range-cache.cc (class sbr_sparse_bitmap): New.
13181 (sbr_sparse_bitmap::sbr_sparse_bitmap): New.
13182 (sbr_sparse_bitmap::bitmap_set_quad): New.
13183 (sbr_sparse_bitmap::bitmap_get_quad): New.
13184 (sbr_sparse_bitmap::set_bb_range): New.
13185 (sbr_sparse_bitmap::get_bb_range): New.
13186 (sbr_sparse_bitmap::bb_range_p): New.
13187 (block_range_cache::block_range_cache): initialize bitmap obstack.
13188 (block_range_cache::~block_range_cache): Destruct obstack.
13189 (block_range_cache::set_bb_range): Decide when to utilze the
13190 sparse on entry cache.
13191 * gimple-range-cache.h (block_range_cache): Add bitmap obstack.
13192 * params.opt (-param=evrp-sparse-threshold): New.
13194 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
13196 * bitmap.c (bitmap_set_aligned_chunk): New.
13197 (bitmap_get_aligned_chunk): New.
13198 (test_aligned_chunk): New.
13199 (bitmap_c_tests): Call test_aligned_chunk.
13200 * bitmap.h (bitmap_set_aligned_chunk, bitmap_get_aligned_chunk): New.
13202 2021-06-07 Uroš Bizjak <ubizjak@gmail.com>
13205 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
13207 (ix86_expand_vector_init_one_nonzero): Ditto.
13208 (ix86_expand_vector_init_one_var): Ditto.
13209 (ix86_expand_vector_init_general): Ditto.
13210 * config/i386/mmx.md (vec_initv4qiqi): New expander.
13212 2021-06-07 Jeff Law <jeffreyalaw@gmail.com>
13214 * config/h8300/movepush.md: Change most _clobber_flags
13215 patterns to instead use <cczn> subst.
13216 (movsi_cczn): New pattern with usable CC cases split out.
13217 (movsi_h8sx_cczn): Likewise.
13219 2021-06-07 Martin Liska <mliska@suse.cz>
13221 * common/common-target.def: Split long lines and replace them
13223 * target.def: Likewise.
13224 * doc/tm.texi: Re-generated.
13226 2021-06-07 Jakub Jelinek <jakub@redhat.com>
13229 * fold-const.c (fold_read_from_vector): Return NULL if trying to
13230 read from a CONSTRUCTOR with vector type elements.
13232 2021-06-07 Jakub Jelinek <jakub@redhat.com>
13234 PR middle-end/100898
13235 * tree-inline.c (copy_bb): Only use gimple_call_arg_ptr if memcpy
13236 should copy any arguments. Don't call gimple_call_num_args
13237 on id->call_stmt or call_stmt more than once.
13239 2021-06-07 liuhongt <hongtao.liu@intel.com>
13242 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): Refine
13244 (<insn>v4siv4di2): Delete constraints for define_expand.
13246 2021-06-07 liuhongt <hongtao.liu@intel.com>
13249 * config/i386/i386-expand.c (ix86_expand_builtin): Remove
13250 assignment of cfun->machine->has_explicit_vzeroupper.
13251 * config/i386/i386-features.c
13252 (ix86_add_reg_usage_to_vzerouppers): Delete.
13253 (ix86_add_reg_usage_to_vzeroupper): Ditto.
13254 (rest_of_handle_insert_vzeroupper): Remove
13255 ix86_add_reg_usage_to_vzerouppers, add df_analyze at the end
13257 (gate): Remove cfun->machine->has_explicit_vzeroupper.
13258 * config/i386/i386-protos.h (ix86_expand_avx_vzeroupper):
13260 * config/i386/i386.c (ix86_insn_callee_abi): New function.
13261 (ix86_initialize_callee_abi): Ditto.
13262 (ix86_expand_avx_vzeroupper): Ditto.
13263 (ix86_hard_regno_call_part_clobbered): Adjust for vzeroupper
13265 (TARGET_INSN_CALLEE_ABI): Define as ix86_insn_callee_abi.
13266 (ix86_emit_mode_set): Call ix86_expand_avx_vzeroupper
13268 * config/i386/i386.h (struct GTY(()) machine_function): Delete
13269 has_explicit_vzeroupper.
13270 * config/i386/i386.md (enum unspec): New member
13272 (ABI_DEFAULT,ABI_VZEROUPPER,ABI_UNKNOWN): New
13273 define_constants for insn callee abi index.
13274 * config/i386/predicates.md (vzeroupper_pattern): Adjust.
13275 * config/i386/sse.md (UNSPECV_VZEROUPPER): Deleted.
13276 (avx_vzeroupper): Call ix86_expand_avx_vzeroupper.
13277 (*avx_vzeroupper): Rename to ..
13278 (avx_vzeroupper_callee_abi): .. this, and adjust pattern as
13279 call_insn which has a special vzeroupper ABI.
13280 (*avx_vzeroupper_1): Deleted.
13282 2021-06-07 liuhongt <hongtao.liu@intel.com>
13285 * df-scan.c (df_get_call_refs): When call_insn is a fake call,
13286 it won't use stack pointer reg.
13287 * final.c (leaf_function_p): When call_insn is a fake call, it
13288 won't affect caller as a leaf function.
13289 * reg-stack.c (callee_clobbers_any_stack_reg): New.
13290 (subst_stack_regs): When call_insn doesn't clobber any stack
13291 reg, don't clear the arguments.
13292 * rtl.c (shallow_copy_rtx): Don't clear flag used when orig is
13294 * shrink-wrap.c (requires_stack_frame_p): No need for stack
13295 frame for a fake call.
13296 * rtl.h (FAKE_CALL_P): New macro.
13298 2021-06-06 Eric Botcazou <ebotcazou@adacore.com>
13300 * config/sparc/sparc-protos.h (order_regs_for_local_alloc): Rename
13302 (sparc_order_regs_for_local_alloc): ...this.
13303 (sparc_leaf_reg_remap): Declare.
13304 * config/sparc/sparc.h (ADJUST_REG_ALLOC_ORDER): Adjust.
13305 (LEAF_REG_REMAP): Reimplement as call to sparc_leaf_reg_remap.
13306 * config/sparc/sparc.c (leaf_reg_remap): Delete.
13307 (order_regs_for_local_alloc): Rename to...
13308 (sparc_order_regs_for_local_alloc): ...this.
13309 (sparc_leaf_reg_remap): New function.
13310 (sparc_conditional_register_usage): Do not modify leaf_reg_remap.
13312 2021-06-06 David Edelsohn <dje.gcc@gmail.com>
13314 * config/rs6000/rs6000.c (rs6000_xcoff_asm_output_aligned_decl_common):
13315 Use assemble_name to output BSS section name.
13317 2021-06-06 Uroš Bizjak <ubizjak@gmail.com>
13319 * config/i386/constraints.md (Bs):
13320 Remove boolean operators from match_test RTX.
13323 (M): Use "mode" variable instead of GET_MODE (op) in match_test RTX.
13326 2021-06-06 Martin Liska <mliska@suse.cz>
13328 * doc/extend.texi: Add missing @headitem.
13329 * doc/invoke.texi: Likewise.
13330 * doc/objc.texi: Likewise.
13332 2021-06-06 Martin Liska <mliska@suse.cz>
13334 * genhooks.c (emit_findices): Remove unused function.
13335 (emit_documentation): Do not call emit_findices
13336 and do not search for @Fcode directives.
13338 2021-06-06 Martin Liska <mliska@suse.cz>
13340 * doc/invoke.texi: Remove extra character.
13342 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13344 * config/sh/sh.md (doloop_end_split): Fix empty split condition.
13346 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13348 * config/sparc/sparc.md (*snedi<W:mode>_zero_vis3,
13349 *neg_snedi<W:mode>_zero_subxc, *plus_snedi<W:mode>_zero,
13350 *plus_plus_snedi<W:mode>_zero, *minus_snedi<W:mode>_zero,
13351 *minus_minus_snedi<W:mode>_zero): Fix empty split condition.
13353 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13355 * config/or1k/or1k.md (*movdi): Fix empty split condition.
13357 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13359 * config/mips/mips.md (<anonymous>, bswapsi2, bswapdi2): Fix empty
13362 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13364 * config/m68k/m68k.md (*zero_extend_inc, *zero_extend_dec,
13365 *zero_extendsidi2): Fix empty split condition.
13367 2021-06-05 Jeff Law <jeffreyalaw@gmail.com>
13369 * config/h8300/addsub.md: Fix split condition in define_insn_and_split
13371 * config/h8300/bitfield.md: Likewise.
13372 * config/h8300/combiner.md: Likewise.
13373 * config/h8300/divmod.md: Likewise.
13374 * config/h8300/extensions.md: Likewise.
13375 * config/h8300/jumpcall.md: Likewise.
13376 * config/h8300/movepush.md: Likewise.
13377 * config/h8300/multiply.md: Likewise.
13378 * config/h8300/other.md: Likewise.
13379 * config/h8300/shiftrotate.md: Likewise.
13380 * config/h8300/logical.md: Likewise. Fix split pattern to use
13381 code iterator that somehow slipped through.
13383 2021-06-04 Tobias Burnus <tobias@codesourcery.com>
13385 PR middle-end/100905
13386 * tree-nested.c (convert_nonlocal_omp_clauses,
13387 convert_local_omp_clauses): Handle OMP_CLAUSE_BIND.
13389 2021-06-04 Martin Sebor <msebor@redhat.com>
13391 PR middle-end/100732
13392 * gimple-fold.c (gimple_fold_builtin_sprintf): Avoid folding calls
13393 with either source or destination argument of invalid type.
13394 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Avoid checking
13395 calls with arguments of invalid type.
13397 2021-06-04 Martin Sebor <msebor@redhat.com>
13399 * attribs.c (init_attr_rdwr_indices): Use VLA bounds in the expected
13401 (attr_access::vla_bounds): Also handle VLA bounds.
13403 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
13405 * config/i386/predicates.md (GOT_memory_operand):
13406 Implement using match_code RTXes.
13407 (GOT32_symbol_operand): Ditto.
13409 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
13412 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
13414 (ix86_expand_vector_init_general): Ditto.
13415 Use SImode instead of word_mode for logic operations
13416 when GET_MODE_SIZE (mode) < UNITS_PER_WORD.
13417 (expand_vec_perm_even_odd_1): Assert that V2HI mode should be
13418 implemented by expand_vec_perm_1.
13419 (expand_vec_perm_broadcast_1): Assert that V2HI and V4HI modes
13420 should be implemented using standard shuffle patterns.
13421 (ix86_vectorize_vec_perm_const): Handle V2HImode. Add V4HI and
13422 V2HI modes to modes, implementable with shuffle for one operand.
13423 * config/i386/mmx.md (*punpckwd): New insn_and_split pattern.
13424 (*pshufw_1): New insn pattern.
13425 (*vec_dupv2hi): Ditto.
13426 (vec_initv2hihi): New expander.
13428 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
13430 * config/arm/vfp.md (no_literal_pool_df_immediate,
13431 no_literal_pool_sf_immediate): Fix empty split condition.
13433 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
13435 * config/i386/i386.md (*load_tp_x32_zext, *add_tp_x32_zext,
13436 *tls_dynamic_gnu2_combine_32): Fix empty split condition.
13437 * config/i386/sse.md (*<sse2_avx2>_pmovmskb_lt,
13438 *<sse2_avx2>_pmovmskb_zext_lt, *sse2_pmovmskb_ext_lt,
13439 *<sse4_1_avx2>_pblendvb_lt): Likewise.
13441 2021-06-04 Jakub Jelinek <jakub@redhat.com>
13444 * config/i386/i386-expand.c (ix86_expand_vector_init): Handle
13445 concatenation from half-sized modes with TImode elements.
13447 2021-06-04 Claudiu Zissulescu <claziss@synopsys.com>
13449 * config/arc/arc.c (arc_override_options): Disable millicode
13450 thunks when RF16 is on.
13452 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
13454 * config/rs6000/rs6000.h (PROMOTE_MODE): Remove.
13456 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
13458 * config/rs6000/rs6000-call.c (rs6000_promote_function_mode):
13459 Replace PROMOTE_MODE marco with its content.
13461 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
13463 * config/cris/cris.md (*addi_reload): Fix empty split condition.
13465 2021-06-03 Jim Wilson <jimw@sifive.com>
13467 * config.gcc (riscv*-*-*): If --with-riscv-attribute not used,
13468 turn it on for all riscv targets.
13470 2021-06-03 Uroš Bizjak <ubizjak@gmail.com>
13473 * config/i386/i386-expand.c (ix86_expand_vector_set):
13474 Handle V2HI and V4QI modes.
13475 (ix86_expand_vector_extract): Ditto.
13476 * config/i386/mmx.md (*pinsrw): New insn pattern.
13479 (*pextrw_zext): Ditto.
13481 (*pextrb_zext): Ditto.
13482 (vec_setv2hi): New expander.
13483 (vec_extractv2hihi): Ditto.
13484 (vec_setv4qi): Ditto.
13485 (vec_extractv4qiqi): Ditto.
13486 (vec_setv8qi): Enable only for TARGET_SSE4_1.
13487 (vec_extractv8qiqi): Ditto.
13489 2021-06-03 Aaron Sawdey <acsawdey@linux.ibm.com>
13491 * config/rs6000/genfusion.pl (gen_logical_addsubf): Fix input
13492 order to subf instruction.
13493 * config/rs6000/fusion.md: Regenerate.
13495 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
13497 * calls.c (get_size_range): Use range_of_expr instead of
13498 determine_value_range.
13499 * tree-affine.c (expr_to_aff_combination): Same.
13500 * tree-data-ref.c (split_constant_offset): Same.
13501 * tree-vrp.c (determine_value_range_1): Remove.
13502 (determine_value_range): Remove.
13503 * tree-vrp.h (determine_value_range): Remove.
13505 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
13507 * function-tests.c (test_ranges): Call gimple_range_tests.
13508 * gimple-range-cache.cc (ranger_cache::range_of_expr): Pass stmt
13510 * gimple-range.cc (fur_source::get_operand): Do not call
13511 get_tree_range or gimple_range_global.
13513 (get_tree_range): Move to value-query.cc.
13514 Call get_arith_expr_range.
13515 (gimple_ranger::range_of_expr): Add argument to get_tree_range.
13516 Include gimple-range-tests.cc.
13517 * gimple-range.h (fold_range): Add argument.
13518 (get_tree_range): Remove.
13519 * selftest.h (gimple_range_tests): New.
13520 * value-query.cc (global_range_query::range_of_expr): Add
13522 (range_query::get_tree_range): Move from gimple-range.cc.
13523 * value-query.h (class range_query): Add get_tree_range and
13524 get_arith_expr_range. Make fur_source a friend.
13525 * vr-values.c (vr_values::range_of_expr): Pass stmt to
13527 * gimple-range-tests.cc: New file.
13529 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
13531 * gimple-range.cc (gimple_ranger::export_global_ranges): Call
13532 update_global_range.
13533 * value-query.cc (update_global_range): New.
13534 * value-query.h (update_global_range): New.
13536 2021-06-03 David Malcolm <dmalcolm@redhat.com>
13538 * diagnostic-show-locus.c (diagnostic_show_locus): Don't reject
13539 printing the same location twice if there are fix-it hints,
13540 multiple locations, or a label.
13542 2021-06-03 Andre Vieira <andre.simoesdiasvieira@arm.com>
13544 * tree-vect-loop.c (vect_transform_loop): Use main loop's various'
13545 thresholds to narrow the upper bound on epilogue iterations.
13547 2021-06-03 Christophe Lyon <christophe.lyon@linaro.org>
13549 * config/arm/mve.md (mve_vabsq_f<mode>): Use 'abs' instead of unspec.
13550 (mve_vabsq_s<mode>): Likewise.
13551 * config/arm/neon.md (abs<mode>2): Rename to neon_abs<mode>2.
13552 * config/arm/unspecs.md (VABSQ_F, VABSQ_S): Delete.
13553 * config/arm/vec-common.md (neg<mode>2): Rename to
13554 <absneg_str><mode>2.
13556 2021-06-03 Claudiu Zissulescu <claziss@synopsys.com>
13558 * common/config/arc/arc-common.c (arc_option_optimization_table):
13559 Remove malign-call.
13560 * config/arc/arc.c (arc_unalign_branch_p): Remove unused function.
13561 * config/arc/arc.h (TARGET_MIXED_CODE): Remove macro.
13562 (INDEX_REG_CLASS): Only refer to GENERAL_REGS.
13563 * config/arc/arc.md (abssi2_mixed): Remove pattern.
13564 * config/arc/arc.opt (munalign-prob-threshold): Mark it obsolete.
13565 (malign-call): Likewise.
13566 (mmixed-code): Likewise.
13567 * doc/invoke.texi (ARC): Update doc.
13569 2021-06-03 Martin Liska <mliska@suse.cz>
13571 * common.opt: Use proper Enum values.
13572 * opts.c (COVERAGE_SANITIZER_OPT): Remove.
13573 (parse_sanitizer_options): Handle only sanitizer_opts.
13574 (common_handle_option): Just assign value.
13576 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13579 * tree-inline.c (inline_forbidden_p): Remove test on return type.
13581 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13583 * dwarf2out.c (loc_list_from_tree_1) <FUNCTION_DECL>: Also generate
13584 DW_OP_GNU_variable_value referencing an existing DIE at file scope.
13585 (type_byte_size): Inline into...
13586 (add_byte_size_attribute): ...this and call add_scalar_info.
13588 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13590 * dwarf2out.c (mem_loc_descriptor) <UDIV>: Fix typo.
13591 (typed_binop_from_tree): New function.
13592 (loc_list_from_tree_1) <EXACT_DIV_EXPR>: For an unsigned type,
13593 turn a divide by a power of 2 into a shift.
13594 <CEIL_DIV_EXPR>: For an unsigned type, use a signed divide if the
13595 size of the mode is lower than DWARF2_ADDR_SIZE; otherwise, do a
13596 typed divide by calling typed_binop_from_tree.
13598 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13600 * dwarf2out.c (scompare_loc_descriptor): Fix head comment.
13601 (is_handled_procedure_type): Likewise.
13602 (struct loc_descr_context): Add strict_signedness field.
13603 (resolve_args_picking_1): Deal with DW_OP_[GNU_]deref_type,
13604 DW_OP_[GNU_]convert and DW_OP_[GNU_]reinterpret.
13605 (resolve_args_picking): Minor tweak.
13606 (function_to_dwarf_procedure): Initialize strict_signedness field.
13607 (type_byte_size): Likewise.
13608 (field_byte_offset): Likewise.
13609 (gen_descr_array_type_die): Likewise.
13610 (gen_variant_part): Likewise.
13611 (loc_list_from_tree_1) <CALL_EXPR>: Tidy up and set strict_signedness
13612 to true when a context is present before evaluating the arguments.
13613 <COND_EXPR>: Do not generate a useless comparison with zero.
13614 When dereferencing an address, if strict_signedness is true and the
13615 type is small and signed, use DW_OP_deref_type to do the dereference
13616 and then DW_OP_convert to convert back to the generic type.
13618 2021-06-03 Jakub Jelinek <jakub@redhat.com>
13621 * tree-inline.c (copy_tree_body_r): Handle iterators on
13622 OMP_CLAUSE_AFFINITY or OMP_CLAUSE_DEPEND.
13624 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
13626 * config/arc/arc.md (*bbit_di): Remove.
13628 2021-06-02 Christoph Muellner <cmuellner@gcc.gnu.org>
13630 PR rtl-optimization/100264
13631 * ree.c (get_sub_rtx): Ignore SET expressions without register
13632 destinations and remove assertion, as it is not valid anymore
13633 with this new behaviour.
13634 (merge_def_and_ext): Eliminate destination check for register
13635 as such SET expressions can't occur anymore.
13636 (combine_reaching_defs): Likewise.
13638 2021-06-02 Jakub Jelinek <jakub@redhat.com>
13641 * config/xtensa/xtensa.h (LEAF_REG_REMAP): Cast REGNO to int to avoid
13642 -Wtype-limits warnings.
13643 (DWARF_FRAME_REGISTER): Rewrite into ternary operator with addition
13644 in operands to avoid -Wsign-compare warnings.
13646 2021-06-02 Pat Haugen <pthaugen@linux.ibm.com>
13648 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): Use
13651 2021-06-02 Vineet Gupta <vgupta@synopsys.com>
13653 * config/arc/arc.h (TARGET_CPU_DEFAULT): Change to hs38_linux.
13655 2021-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
13657 * config/s390/s390.md(*ashrdi3_31<setcc><cconly>): Use a single
13659 * config/s390/subst.md(cconly_subst): Use a single constraint
13660 in (match_scratch).
13662 2021-06-02 Martin Liska <mliska@suse.cz>
13664 * ipa-icf.h: Use auto_vec for memory_access_types.
13666 2021-06-02 Jeff Law <jeffreyalaw@gmail.com>
13668 * config/h8300/h8300-protos.h (compute_a_shift_length): Drop unused
13669 argument from prototype.
13670 (output_logical_op): Add rtx_code argument.
13671 (compute_logical_op_length): Likewise.
13672 * config/h8300/h8300.c (h8300_and_costs): Pass additional argument
13673 to compute_a_shift_length.
13674 (output_logical_op); New argument with the rtx code rather than
13675 extracting it from an operand. Handle QImode too.
13676 (compute_logical_op_length): Similary.
13677 (compute_a_shift_length): Drop unused argument.
13678 * config/h8300/h8300.md (logicals): New code iterator.
13679 * config/h8300/logical.md (<code><mode>3 expander): Combine
13680 the "and" expander with the "ior"/"xor" expander.
13681 (bclr<mode>msx): Combine the QI/HI mode patterns.
13682 (<logical><mode>3 insns): Use code iterator rather than match_operator.
13683 Handle QImode as well. Update call to output_logical_op and
13684 compute_logical_op_length to pass in rtx_code
13685 Fix split condition on all define_insn_and_split patterns.
13686 (one_cmpl<mode>2<cczn>): Use <cczn> to support both clobbering
13687 the flags and setting ZN via existing define_subst.
13688 * config/h8300/shiftrotate.md: Drop unused argument from
13689 calls to compute_a_shift_length.
13690 Signed-off-by: Jeff Law <jeffreyalaw@gmail.com>
13692 2021-06-01 Andrew Pinski <apinski@marvell.com>
13694 PR tree-optimization/25290
13695 * tree-ssa-phiopt.c (match_simplify_replacement):
13697 (tree_ssa_phiopt_worker): Use match_simplify_replacement.
13698 (two_value_replacement): Change the comment about
13699 conditional_replacement.
13700 (conditional_replacement): Delete.
13702 2021-06-01 Andrew Pinski <apinski@marvell.com>
13704 PR tree-optimization/95481
13705 * tree-tailcall.c (find_tail_calls): Handle empty typed
13708 2021-06-01 Andrew Pinski <apinski@marvell.com>
13710 * gimplify.c (zero_sized_field_decl): Delete
13711 (zero_sized_type): Delete
13712 (gimplify_init_ctor_eval): Use is_empty_type instead
13713 of zero_sized_field_decl.
13714 (gimplify_modify_expr): Use is_empty_type instead of
13717 2021-06-01 Jason Merrill <jason@redhat.com>
13720 * tree.h (CALL_FROM_NEW_OR_DELETE_P): Adjust comment.
13722 2021-06-01 Jason Merrill <jason@redhat.com>
13725 * diagnostic.h (warning_enabled_at): Declare.
13726 * diagnostic.c (diagnostic_enabled): Factor out from...
13727 (diagnostic_report_diagnostic): ...here.
13728 (warning_enabled_at): New.
13730 2021-06-01 Aldy Hernandez <aldyh@redhat.com>
13732 * gimple-ssa-evrp.c: Enable exporting of global ranges.
13734 2021-06-01 Martin Liska <mliska@suse.cz>
13737 * doc/invoke.texi: Mention that -fgcse-after-reload
13738 is enabled with -O3.
13740 2021-06-01 liuhongt <hongtao.liu@intel.com>
13742 PR tree-optimization/98365
13743 * tree-if-conv.c (strip_nop_cond_scalar_reduction): New function.
13744 (is_cond_scalar_reduction): Handle nop_expr in cond scalar reduction.
13745 (convert_scalar_cond_reduction): Ditto.
13746 (predicate_scalar_phi): Ditto.
13748 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
13750 PR tree-optimization/100781
13751 * gimple-range-cache.cc (ranger_cache::ranger_cache): Enable new
13752 value calculation by default.
13753 (ranger_cache::enable_new_values): New.
13754 (ranger_cache::disable_new_values): New.
13755 (ranger_cache::push_poor_value): Check if new values are allowed.
13756 * gimple-range-cache.h (class ranger_cache): New member/methods.
13757 * gimple-range.cc (gimple_ranger::range_of_expr): Check for debug
13758 statement, and disable/renable new value calculation.
13760 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
13762 * gimple-range-cache.cc (ranger_cache::ssa_range_in_bb): Delete.
13763 (ranger_cache::range_of_def): New.
13764 (ranger_cache::entry_range): New.
13765 (ranger_cache::exit_range): New.
13766 (ranger_cache::range_of_expr): Adjust.
13767 (ranger_cache::range_on_edge): Adjust.
13768 (ranger_cache::propagate_cache): Call exit_range directly.
13769 * gimple-range-cache.h (class ranger_cache): Adjust.
13771 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
13773 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust for
13774 gori_compute being a member rather than base class.
13775 dervied call to member call.
13776 (ranger_cache::dump): No longer dump gori_map.
13777 (ranger_cache::dump_bb): New.
13778 (ranger_cache::get_non_stale_global_range): Adjust for gori_compute
13779 being a member rather than base class.
13780 (ranger_cache::set_global_range): Ditto.
13781 (ranger_cache::ssa_range_in_bb): Ditto.
13782 (ranger_cache::range_of_expr): New.
13783 (ranger_cache::range_on_edge): New.
13784 (ranger_cache::block_range): Adjust for gori_computes. Debug changes.
13785 (ranger_cache::propagate_cache): Adjust debugging output.
13786 (ranger_cache::fill_block_cache): Adjust for gori_computes. Debug
13788 * gimple-range-cache.h (class ranger_cache): Make gori_compute a
13789 member, and inherit from range_query instead.
13790 (ranger_cache::dump_bb): New. split from dump.
13791 * gimple-range-gori.cc (gori_compute::ssa_range_in_bb): Delete.
13792 (gori_compute::expr_range_at_stmt): Delete.
13793 (gori_compute::compute_name_range_op): Delete.
13794 (gori_compute::compute_operand_range_switch): Add fur_source.
13795 (gori_compute::compute_operand_range): Add fur_source param, inline
13796 old compute_name_range_op and optimize_logical_operands.
13797 (struct tf_range): Delete.
13798 (gori_compute::logical_combine): Adjust
13799 (gori_compute::optimize_logical_operands): Delete.
13800 (gori_compute::compute_logical_operands_in_chain): Delete.
13801 (gori_compute::compute_logical_operands): Adjust.
13802 (gori_compute::compute_operand1_range): Adjust to fur_source.
13803 (gori_compute::compute_operand2_range): Ditto.
13804 (gori_compute::compute_operand1_and_operand2_range): Ditto.
13805 (gori_compute::outgoing_edge_range_p): Add range_query parameter,
13806 and adjust to fur_source.
13807 * gimple-range-gori.h (class gori_compute): Simplify and adjust to
13808 range_query and fur_source.
13809 * gimple-range.cc (gimple_ranger::range_on_edge): Query range_on_edge
13810 from the ranger_cache..
13811 (gimple_ranger::fold_range_internal): Adjust to base class change of
13813 (gimple_ranger::dump_bb): Adjust dump.
13814 * gimple-range.h (gimple_ranger):export gori computes object.
13816 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
13818 PR tree-optimization/100774
13819 * gimple-range-cache.cc (ranger_cache::get_non_stale_global_range):
13820 Constant values are also not stale.
13821 (ranger_cache::set_global_range): Range invariant values should also
13822 have the correct timestamp.
13824 2021-05-31 Martin Liska <mliska@suse.cz>
13826 * tree-streamer-in.c (unpack_ts_function_decl_value_fields):
13827 Unpack FUNCTION_DECL_DECL_TYPE.
13828 * tree-streamer-out.c (pack_ts_function_decl_value_fields):
13829 Stream FUNCTION_DECL_DECL_TYPE instead of
13830 DECL_IS_OPERATOR_NEW_P.
13831 * tree.h (set_function_decl_type): Use FUNCTION_DECL_DECL_TYPE
13833 (DECL_IS_OPERATOR_NEW_P): Likewise.
13834 (DECL_IS_OPERATOR_DELETE_P): Likewise.
13835 (DECL_LAMBDA_FUNCTION_P): Likewise.
13837 2021-05-31 Richard Biener <rguenther@suse.de>
13840 * internal-fn.c (expand_SHUFFLEVECTOR): Define.
13841 * internal-fn.def (SHUFFLEVECTOR): New.
13842 * internal-fn.h (expand_SHUFFLEVECTOR): Declare.
13843 * doc/extend.texi: Document __builtin_shufflevector.
13845 2021-05-31 Peter Bergner <bergner@linux.ibm.com>
13848 * config/rs6000/predicates.md(mma_assemble_input_operand): Allow
13849 indexed form addresses.
13851 2021-05-29 Jeff Law <jlaw@tachyum.com>
13853 * config/h8300/h8300.c (h8300_emit_stack_adjustment): Drop unused
13854 parameter. Call callers fixed.
13856 (output_plussi): Add FALLTHRU markers.
13857 (h8300_shift_needs_scratch_p): Add gcc_unreachable marker.
13859 2021-05-29 Jakub Jelinek <jakub@redhat.com>
13861 PR middle-end/99928
13862 * gimplify.c (gimplify_scan_omp_clauses): For taskloop simd
13863 combined with parallel, make sure to add shared clause to
13864 parallel for explicit linear clause.
13866 2021-05-29 Aldy Hernandez <aldyh@redhat.com>
13868 PR tree-optimization/100787
13869 * gimple-ssa-evrp.c: Disable exporting of global ranges.
13871 2021-05-28 Jason Merrill <jason@redhat.com>
13873 * tree-iterator.h (struct tree_stmt_iterator): Add operator++,
13874 operator--, operator*, operator==, and operator!=.
13875 (class tsi_range): New.
13877 2021-05-28 Richard Biener <rguenther@suse.de>
13879 PR tree-optimization/100778
13880 * tree-vect-slp.c (vect_build_slp_tree_1): Prevent possibly
13881 trapping ops in different BBs.
13883 2021-05-28 Richard Biener <rguenther@suse.de>
13886 * tree-inline.c (copy_bb): When processing __builtin_va_arg_pack
13887 copy fntype from original call.
13889 2021-05-28 Martin Liska <mliska@suse.cz>
13891 PR gcov-profile/100751
13892 * doc/gcov.texi: Revert partially a hunk that was wrong.
13894 2021-05-28 Cooper Qu <cooper.qu@linux.alibaba.com>
13896 * config/csky/csky-linux-elf.h (HAVE_sync_compare_and_swapqi):
13898 (HAVE_sync_compare_and_swaphi): Likewise.
13899 (HAVE_sync_compare_and_swapsi): Likewise.
13901 2021-05-28 Jakub Jelinek <jakub@redhat.com>
13903 PR middle-end/99928
13904 * tree.h (OMP_CLAUSE_MAP_IMPLICIT): Define.
13906 2021-05-28 Tobias Burnus <tobias@codesourcery.com>
13908 * gimplify.c (gimplify_omp_affinity): New.
13909 (gimplify_scan_omp_clauses): Call it; remove affinity clause afterwards.
13910 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_AFFINITY.
13911 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_AFFINITY.
13912 * tree.c (omp_clause_num_ops, omp_clause_code_name): Add clause.
13913 (walk_tree_1): Handle OMP_CLAUSE_AFFINITY.
13915 2021-05-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
13916 Richard Biener <rguenther@suse.de>
13918 * match.pd <popcount & / + pattern matching>:
13919 When generating popcount directly fails, try doing it in two halves.
13921 2021-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
13923 * Makefile.in (generated_files): Add gimple-match.c and
13926 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
13928 * gensupport.c (alter_predicate_for_insn): Handle MATCH_DUP.
13930 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
13932 * gensupport.c (alter_constraints): Add MATCH_SCRATCH case.
13934 2021-05-28 Kewen Lin <linkw@linux.ibm.com>
13936 PR tree-optimization/99398
13937 * tree-ssa-forwprop.c (simplify_permutation): Optimize some cases
13938 where the fed operands are CTOR/CST and propagated through
13939 VIEW_CONVERT_EXPR. Call vec_perm_indices::new_shrunk_vector.
13940 * vec-perm-indices.c (vec_perm_indices::new_shrunk_vector): New
13942 * vec-perm-indices.h (vec_perm_indices::new_shrunk_vector): New
13945 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
13947 * config/i386/mmx.md (addv2sf3): Do not call
13948 ix86_fixup_binary_operands_no_copy.
13951 (<smaxmin:code>v2sf3): Ditto.
13952 (<plusminus:insn><MMXMODEI:mode>3): Ditto.
13953 (<plusminus:insn><VI_32:mode>3): Remove expander.
13954 (<plusminus:insn><VI_32:mode>3): Rename from
13955 "*<plusminus:insn><VI_32:mode>3".
13956 (mulv4hi): Do not call ix86_fixup_binary_operands_no_copy.
13957 (mulv2hi3): Remove expander.
13958 (mulv2hi3): Rename from *mulv2hi3.
13959 (<s>mulv2hi3_highpart): Remove expander.
13960 (<s>mulv2hi3_highpart): Rename from *<s>mulv2hi3_highpart.
13961 (<smaxmin:code><MMXMODE14:mode>3): Rename from
13962 "*mmx_<smaxmin:code><MMXMODE14:mode>3".
13963 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Remove expander.
13964 (SMAXMIN_MMXMODEI): Remove mode iterator.
13965 (<smaxmin:code>v4hi3): New expander.
13966 (<smaxmin:code>v4qi3): Rename from *<smaxmin:code>v4qi3.
13967 (<smaxmin:code>v2hi3): Rename from *<smaxmin:code>v2hi3.
13968 (<smaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
13969 (SMAXMIN_VI_32): Remove mode iterator.
13970 (<umaxmin:code><MMXMODE24:mode>3): Rename from
13971 "*mmx_<umaxmin:code><MMXMODE24:mode>3".
13972 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Remove expander.
13973 (UMAXMIN_MMXMODEI): Remove mode iterator.
13974 (<umaxmin:code>v8qi3): New expander.
13975 (<umaxmin:code>v4qi3): Rename from *<umaxmin:code>v4qi3.
13976 (<umaxmin:code>v2hi3): Rename from *<umaxmin:code>v2hi3.
13977 (<umaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
13978 (UMAXMIN_VI_32): Remove mode iterator.
13979 (<any_shift:insn>v2hi3): Remove expander.
13980 (<any_shift:insn>v2hi3): Rename from *<any_shift:insn>v2hi3.
13981 (<any_logic:code><MMXMODEI:mode>3): Do not call
13982 ix86_fixup_binary_operands_no_copy.
13983 (<any_logic:code><VI_32:mode>3): Remove expander.
13984 (<any_logic:code><VI_32:mode>3): Rename from
13985 "*<any_logic:code><VI_32:mode>3".
13986 (uavg<mode>3_ceil): Do not call ix86_fixup_binary_operands_no_copy.
13987 * config/i386/sse.md (div<VF2:mode>3): Do not call
13988 ix86_fixup_binary_operands_no_copy.
13989 (div<VF1:mode>3): Ditto.
13990 (<maxmin:code><VI8_AVX2_AVX512F:mode>3): Ditto.
13991 (smulhrsv4hi3): Ditto.
13992 (smulhrsv2hi3): Ditto.
13994 2021-05-27 Martin Sebor <msebor@redhat.com>
13996 * ggc.h (gt_ggc_mx): Add overloads for all integers.
13998 * hash-map.h (class hash_map): Add pch_nx_helper overloads for all
14000 (hash_map::operator==): New function.
14002 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
14005 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
14006 For TARGET_XOP bypass SSE comparisons for all supported vector modes.
14007 * config/i386/mmx.md (*xop_maskcmp<MMXMODEI:mode>3): New insn pattern.
14008 (*xop_maskcmp<VI_32:mode>3): Ditto.
14009 (*xop_maskcmp_uns<MMXMODEI:mode>3): Ditto.
14010 (*xop_maskcmp_uns<VI_32:mode>3): Ditto.
14012 2021-05-27 Richard Earnshaw <rearnsha@arm.com>
14015 * config/arm/arm.c (arm_configure_build_target): Remove parameter
14016 opts_set, directly check opts parameters for being non-null.
14017 (arm_option_restore): Update call to arm_configure_build_target.
14018 (arm_option_override): Likewise.
14019 (arm_can_inline_p): Likewise.
14020 (arm_valid_target_attribute_tree): Likewise.
14021 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
14022 * config/arm/arm-protos.h (arm_configure_build_target): Adjust
14025 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
14027 * vr-values.c (simplify_conversion_using_ranges): Use
14028 get_range_query instead of get_global_range_query.
14030 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
14032 * gimple-range.cc (get_range_global): Move to value-query.cc.
14033 (gimple_range_global): Same.
14034 (get_global_range_query): Same.
14035 (global_range_query::range_of_expr): Same.
14036 * gimple-range.h (class global_range_query): Move to
14038 (gimple_range_global): Same.
14039 * tree-ssanames.c (get_range_info): Move to value-query.cc.
14040 (get_ptr_nonnull): Same.
14041 * tree-ssanames.h (get_range_info): Remove.
14042 (get_ptr_nonnull): Remove.
14043 * value-query.cc (get_ssa_name_range_info): Move from
14045 (get_ssa_name_ptr_info_nonnull): Same.
14046 (get_range_global): Move from gimple-range.cc.
14047 (gimple_range_global): Same.
14048 (get_global_range_query): Same.
14049 (global_range_query::range_of_expr): Same.
14050 * value-query.h (class global_range_query): Move from
14052 (gimple_range_global): Same.
14054 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
14057 * config/i386/mmx.md (uavgv4qi3_ceil): New insn pattern.
14058 (uavgv2hi3_ceil): Ditto.
14060 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
14063 * doc/extend.texi (scalar_storage_order): Rephrase slightly.
14065 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14067 * tree-ssanames.c (get_range_info): Merge both copies of
14068 get_range_info into one that works with irange.
14069 * tree-ssanames.h (get_range_info): Remove version that works on
14072 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14074 * builtins.c (check_nul_terminated_array): Convert to get_range_query.
14075 (expand_builtin_strnlen): Same.
14076 (determine_block_size): Same.
14077 * fold-const.c (expr_not_equal_to): Same.
14078 * gimple-fold.c (size_must_be_zero_p): Same.
14079 * gimple-match-head.c: Include gimple-range.h.
14080 * gimple-pretty-print.c (dump_ssaname_info): Convert to get_range_query.
14081 * gimple-ssa-warn-restrict.c
14082 (builtin_memref::extend_offset_range): Same.
14083 * graphite-sese-to-poly.c (add_param_constraints): Same.
14084 * internal-fn.c (get_min_precision): Same.
14085 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): Same.
14086 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Same.
14088 * tree-data-ref.c (split_constant_offset): Same.
14089 (dr_step_indicator): Same.
14090 * tree-dfa.c (get_ref_base_and_extent): Same.
14091 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
14092 * tree-ssa-loop-niter.c (refine_value_range_using_guard): Same.
14093 (determine_value_range): Same.
14094 (record_nonwrapping_iv): Same.
14095 (infer_loop_bounds_from_signedness): Same.
14096 (scev_var_range_cant_overflow): Same.
14097 * tree-ssa-phiopt.c (two_value_replacement): Same.
14098 * tree-ssa-pre.c (insert_into_preds_of_block): Same.
14099 * tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Same.
14100 * tree-ssa-strlen.c (handle_builtin_stxncpy_strncat): Same.
14102 (dump_strlen_info): Same.
14103 (set_strlen_range): Same.
14104 (maybe_diag_stxncpy_trunc): Same.
14105 (get_len_or_size): Same.
14106 (handle_integral_assign): Same.
14107 * tree-ssa-structalias.c (find_what_p_points_to): Same.
14108 * tree-ssa-uninit.c (find_var_cmp_const): Same.
14109 * tree-switch-conversion.c (bit_test_cluster::emit): Same.
14110 * tree-vect-patterns.c (vect_get_range_info): Same.
14111 (vect_recog_divmod_pattern): Same.
14112 * tree-vrp.c (intersect_range_with_nonzero_bits): Same.
14113 (register_edge_assert_for_2): Same.
14114 (determine_value_range_1): Same.
14115 * tree.c (get_range_pos_neg): Same.
14116 * vr-values.c (vr_values::get_lattice_entry): Same.
14117 (vr_values::update_value_range): Same.
14118 (simplify_conversion_using_ranges): Same.
14120 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14122 * gimple-ssa-warn-alloca.c (alloca_call_type): Use
14123 get_range_query instead of query argument.
14124 (pass_walloca::execute): Enable and disable global ranger.
14126 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14128 * gimple-ssa-evrp.c (rvrp_folder::rvrp_folder): Call
14130 (rvrp_folder::~rvrp_folder): Call disable_ranger.
14131 (hybrid_folder::hybrid_folder): Call enable_ranger.
14132 (hybrid_folder::~hybrid_folder): Call disable_ranger.
14134 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14136 * function.c (allocate_struct_function): Set cfun->x_range_query.
14137 * function.h (struct function): Declare x_range_query.
14138 (get_range_query): New.
14139 (get_global_range_query): New.
14140 * gimple-range-cache.cc (ssa_global_cache::ssa_global_cache):
14141 Remove call to safe_grow_cleared.
14142 * gimple-range.cc (get_range_global): New.
14143 (gimple_range_global): Move from gimple-range.h.
14144 (get_global_range_query): New.
14145 (global_range_query::range_of_expr): New.
14146 (enable_ranger): New.
14147 (disable_ranger): New.
14148 * gimple-range.h (gimple_range_global): Move to gimple-range.cc.
14149 (class global_range_query): New.
14150 (enable_ranger): New.
14151 (disable_ranger): New.
14152 * gimple-ssa-evrp.c (evrp_folder::~evrp_folder): Rename
14153 dump_all_value_ranges to dump.
14154 * tree-vrp.c (vrp_prop::finalize): Same.
14155 * value-query.cc (range_query::dump): New.
14156 * value-query.h (range_query::dump): New.
14157 * vr-values.c (vr_values::dump_all_value_ranges): Rename to...
14158 (vr_values::dump): ...this.
14159 * vr-values.h (class vr_values): Rename dump_all_value_ranges to
14160 dump and make virtual.
14162 2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
14164 * config/i386/i386.c (ix86_autovectorize_vector_modes):
14165 Add V4QImode and V16QImode for TARGET_SSE2.
14166 * doc/sourcebuild.texi (Vector-specific attributes):
14167 Add vect64 and vect32 description.
14169 2021-05-26 Bernd Edlinger <bernd.edlinger@hotmail.de>
14171 * gimple-range-gori.cc (range_def_chain::register_dependency):
14172 Resize m_def_chain when needed.
14174 2021-05-26 Christophe Lyon <christophe.lyon@linaro.org>
14176 * config/arm/mve.md (mve_vaddvq_<supf><mode>): Prefix with '@'.
14177 * config/arm/neon.md (reduc_plus_scal_<mode>): Move to ..
14178 * config/arm/vec-common.md: .. here. Add support for MVE.
14180 2021-05-26 Jakub Jelinek <jakub@redhat.com>
14182 * config/epiphany/epiphany.c (epiphany_print_operand_address): Remove
14184 * config/microblaze/microblaze.c (microblaze_legitimize_address,
14186 microblaze_option_override, print_operand): Likewise.
14187 * config/microblaze/microblaze.md (call_internal_plt,
14188 call_value_intern_plt, call_value_intern): Likewise.
14189 * config/arm/aout.h (ASM_OUTPUT_ALIGN): Likewise.
14190 * config/iq2000/iq2000.md (call_internal1, call_value_internal1,
14191 call_value_multiple_internal1): Likewise.
14192 * config/bfin/bfin.c (symbolic_reference_mentioned_p): Likewise.
14194 2021-05-26 Jan-Benedict Glaw <jbglaw@lug-owl.de>
14196 * config/arc/arc.c (arc_address_cost, arc_print_operand_address,
14197 arc_ccfsm_advance, symbolic_reference_mentioned_p,
14198 arc_raw_symbolic_reference_mentioned_p): Remove register
14201 2021-05-26 Jakub Jelinek <jakub@redhat.com>
14204 * omp-low.c: Include omp-offload.h.
14205 (create_omp_child_function): If current_function_decl has
14206 "omp declare target" attribute and is_gimple_omp_offloaded,
14207 remove that attribute from the copy of attribute list and
14208 add "omp target entrypoint" attribute instead.
14209 (lower_omp_target): Mark .omp_data_sizes.* and .omp_data_kinds.*
14210 variables for offloading if in omp_maybe_offloaded_ctx.
14211 * omp-offload.c (pass_omp_target_link::execute): Nullify second
14212 argument to GOMP_target_data_ext in offloaded code.
14214 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14216 * config/csky/csky.c (csky_can_change_mode_class): Delete.
14217 For csky, HF/SF mode use the low bits of VREGS.
14219 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
14221 * gimplify.c (gimplify_decl_expr): Do not clear TREE_READONLY on a
14222 DECL which is a reference for OMP.
14224 2021-05-26 Martin Liska <mliska@suse.cz>
14226 PR gcov-profile/100751
14227 * doc/gcov.texi: Document that __gcov_dump can be called just
14228 once and that __gcov_reset resets run-time counters.
14230 2021-05-26 Martin Liska <mliska@suse.cz>
14232 * doc/install.texi: Port relevant part from install-old.texi
14233 and re-generate list of CPUs and systems.
14235 2021-05-26 Martin Liska <mliska@suse.cz>
14237 * Makefile.in: Remove it.
14238 * doc/include/fdl.texi: Update next/previous chapters.
14239 * doc/install.texi: Likewise.
14240 * doc/install-old.texi: Removed.
14242 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14244 * config/csky/csky.c (ck810_legitimate_index_p): Support
14245 "base + index" with DF mode.
14246 * config/csky/constraints.md ("Y"): New constraint for memory operands
14247 without index register.
14248 * config/csky/csky_insn_fpuv2.md (fpuv3_movdf): Use "Y" instead of "m"
14249 when mov between memory and general registers, and lower their priority.
14250 * config/csky/csky_insn_fpuv3.md (fpuv2_movdf): Likewise.
14252 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14254 * config/csky/csky.c (TARGET_PROMOTE_PROTOTYPES): Delete.
14256 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14258 * config/csky/csky.md (untyped_call): Emit clobber for return
14259 registers to mark them used.
14261 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14263 * config/csky/csky.md (cskyv2_sextend_ldbs): New.
14265 2021-05-26 Andrew Pinski <apinski@marvell.com>
14267 * match.pd (x < 0 ? ~y : y): New patterns.
14269 2021-05-26 Andrew Pinski <apinski@marvell.com>
14271 * match.pd (A?CST1:CST2): Add simplifcations for A?0:+-1, A?+-1:0,
14272 A?POW2:0 and A?0:POW2.
14274 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14276 * gimple-range-gori.cc (class logical_stmt_cache): Delete
14277 (logical_stmt_cache::logical_stmt_cache ): Delete.
14278 (logical_stmt_cache::~logical_stmt_cache): Delete.
14279 (logical_stmt_cache::cache_entry::dump): Delete.
14280 (logical_stmt_cache::get_range): Delete.
14281 (logical_stmt_cache::cached_name ): Delete.
14282 (logical_stmt_cache::same_cached_name): Delete.
14283 (logical_stmt_cache::cacheable_p): Delete.
14284 (logical_stmt_cache::slot_diagnostics ): Delete.
14285 (logical_stmt_cache::dump): Delete.
14286 (gori_compute_cache::gori_compute_cache): Delete.
14287 (gori_compute_cache::~gori_compute_cache): Delete.
14288 (gori_compute_cache::compute_operand_range): Delete.
14289 (gori_compute_cache::cache_stmt): Delete.
14290 * gimple-range-gori.h (gori_compute::compute_operand_range): Remove
14292 (class gori_compute_cache): Delete.
14294 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14296 * gimple-range.cc (fold_using_range::range_of_range_op): Use m_gori
14298 (fold_using_range::range_of_address): Adjust.
14299 (fold_using_range::range_of_phi): Adjust.
14300 * gimple-range.h (class fur_source): Adjust.
14301 (fur_source::fur_source): Adjust.
14303 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14305 * gimple-range-gori.cc (gori_compute::expr_range_at_stmt): Rename
14306 from expr_range_in_bb and adjust.
14307 (gori_compute::compute_name_range_op): Adjust.
14308 (gori_compute::optimize_logical_operands): Adjust.
14309 (gori_compute::compute_logical_operands_in_chain): Adjust.
14310 (gori_compute::compute_operand1_range): Adjust.
14311 (gori_compute::compute_operand2_range): Adjust.
14312 (ori_compute_cache::cache_stmt): Adjust.
14313 * gimple-range-gori.h (gori_compute): Rename prototype.
14315 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14317 * gimple-range.cc (gimple_ranger::range_of_expr): Non-null should be
14318 checked only after range_of_stmt, not range_on_entry.
14319 (gimple_ranger::range_on_entry): Check for non-null in any
14320 predecessor block, if it is not already non-null.
14321 (gimple_ranger::range_on_exit): DOnt check for non-null after
14322 range on entry call.
14323 (gimple_ranger::dump_bb): New. Split from dump.
14324 (gimple_ranger::dump): Adjust.
14325 * gimple-range.h (class gimple_ranger): Adjust.
14327 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14329 * gimple-range-cache.cc (struct range_timestamp): Delete.
14330 (class temporal_cache): Adjust.
14331 (temporal_cache::get_timestamp): Delete.
14332 (temporal_cache::set_dependency): Delete.
14333 (temporal_cache::temporal_value): Adjust.
14334 (temporal_cache::current_p): Take dependencies as params.
14335 (temporal_cache::set_timestamp): Adjust.
14336 (temporal_cache::set_always_current): Adjust.
14337 (ranger_cache::get_non_stale_global_range): Adjust.
14338 (ranger_cache::register_dependency): Delete.
14339 * gimple-range-cache.h (class range_cache): Adjust.
14341 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14343 * gimple-range-gori.cc (range_def_chain::range_def_chain): init
14345 (range_def_chain::~range_def_chain): Dispose of obstack rather than
14346 each individual bitmap.
14347 (range_def_chain::set_import): New.
14348 (range_def_chain::get_imports): New.
14349 (range_def_chain::chain_import_p): New.
14350 (range_def_chain::register_dependency): Rename from build_def_chain
14352 (range_def_chain::def_chain_in_bitmap_p): New.
14353 (range_def_chain::add_def_chain_to_bitmap): New.
14354 (range_def_chain::has_def_chain): Just check first depenedence.
14355 (range_def_chain::get_def_chain): Process imports, use generic
14356 register_dependency routine.
14357 (range_def_chain::dump): New.
14358 (gori_map::gori_map): Allocate import list.
14359 (gori_map::~gori_map): Release imports.
14360 (gori_map::exports): Check for past allocated block size.
14361 (gori_map::imports): New.
14362 (gori_map::def_chain_in_export_p): Delete.
14363 (gori_map::is_import_p): New.
14364 (gori_map::maybe_add_gori): Handle imports.
14365 (gori_map::dump): Adjust output, add imports.
14366 (gori_compute::has_edge_range_p): Remove def_chain_in_export call.
14367 (gori_export_iterator::gori_export_iterator): New.
14368 (gori_export_iterator::next): New.
14369 (gori_export_iterator::get_name): New.
14370 * gimple-range-gori.h (range_def_chain): Add imports and direct
14371 dependecies via struct rdc.
14372 (range_def_chain::depend1): New.
14373 (range_def_chain::depend2): New.
14374 (class gori_map): Adjust.
14375 (FOR_EACH_GORI_IMPORT_NAME): New.
14376 (FOR_EACH_GORI_EXPORT_NAME): New.
14377 (class gori_export_iterator): New.
14379 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14381 * gimple-range-cache.cc (ranger_cache::ranger_cache): Move initial
14382 export cache filling to here.
14383 * gimple-range-gori.cc (gori_compute::gori_compute) : From Here.
14385 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14387 * gimple-range-gori.cc (range_def_chain): Move to gimple-range-gori.h.
14388 (gori_map): Move to gimple-range-gori.h.
14389 (gori_compute::gori_compute): Adjust.
14390 (gori_compute::~gori_compute): Delete.
14391 (gori_compute::compute_operand_range_switch): Adjust.
14392 (gori_compute::compute_operand_range): Adjust.
14393 (gori_compute::compute_logical_operands): Adjust.
14394 (gori_compute::has_edge_range_p ): Adjust.
14395 (gori_compute::set_range_invariant): Delete.
14396 (gori_compute::dump): Adjust.
14397 (gori_compute::outgoing_edge_range_p): Adjust.
14398 * gimple-range-gori.h (class range_def_chain): Relocate here.
14399 (class gori_map): Relocate here.
14400 (class gori_compute): Inherit from gori_map, and adjust.
14402 2021-05-25 Aldy Hernandez <aldyh@redhat.com>
14404 * value-range.cc (range_tests_legacy): Use
14405 build_nonstandard_integer_type instead of int and short.
14407 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
14409 * gimplify.c (gimplify_decl_expr): Clear TREE_READONLY on the DECL
14410 when really creating an initialization statement for it.
14412 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
14414 * tree-inline.c (setup_one_parameter): Fix thinko in new condition.
14416 2021-05-25 Kito Cheng <kito.cheng@sifive.com>
14418 * config/riscv/riscv.h (ASM_SPEC): Pass -mno-relax.
14420 2021-05-25 Martin Liska <mliska@suse.cz>
14422 PR tree-optimization/92860
14424 * optc-save-gen.awk: Remove exceptions.
14426 2021-05-25 Martin Liska <mliska@suse.cz>
14428 * asan.h (sanitize_coverage_p): New function.
14429 * doc/extend.texi: Document it.
14430 * fold-const.c (fold_range_test): Use sanitize_flags_p
14431 instead of flag_sanitize_coverage.
14432 (fold_truth_andor): Likewise.
14433 * sancov.c: Likewise.
14434 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
14435 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Handle
14436 -fsanitize-coverage when inlining.
14438 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
14440 * config/csky/csky-modes.def : Fix copyright.
14442 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
14444 * config/csky/csky-modes.def : Amend copyright.
14445 * config/csky/csky_insn_fpuv2.md : Likewise.
14446 * config/csky/csky_insn_fpuv3.md : Likewise.
14448 2021-05-25 Richard Biener <rguenther@suse.de>
14450 PR middle-end/100727
14451 * calls.c (initialize_argument_information): Explicitely test
14452 for WITH_SIZE_EXPR.
14453 * gimple-expr.c (mark_addressable): Skip outer WITH_SIZE_EXPR.
14455 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14457 * config/csky/csky.h (FRAME_POINTER_REGNUM): Use
14458 HARD_FRAME_POINTER_REGNUM and FRAME_POINTER_REGNUM instead of
14459 the signle definition. The signle definition may not work well
14460 at simplify_subreg_regno().
14461 (HARD_FRAME_POINTER_REGNUM): New.
14462 (ELIMINABLE_REGS): Add for HARD_FRAME_POINTER_REGNUM.
14463 * config/csky/csky.c (get_csky_live_regs, csky_can_eliminate,
14464 csky_initial_elimination_offset, csky_expand_prologue,
14465 csky_expand_epilogue): Add for HARD_FRAME_POINTER_REGNUM.
14467 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14469 * config/csky/csky.c (csky_option_override):
14470 Init csky_arch_isa_features[] in advance, so TARGET_DSP
14471 and TARGET_DIV can be set well.
14473 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14475 * config/csky/constraints.md ("l", "h"): Delete.
14476 * config/csky/csky.h (reg_class, REG_CLASS_NAMES,
14477 REG_CLASS_CONTENTS): Delete LO_REGS and HI_REGS.
14478 * config/csky/csky.c (regno_reg_classm,
14479 csky_secondary_reload, csky_register_move_cost):
14480 Use HILO_REGS instead of LO_REGS and HI_REGS.
14482 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14484 * config/csky/constraints.md ("W"): New constriant for mem operand
14485 with base reg, index register.
14486 ("Q"): Renamed and modified "csky_valid_fpuv2_mem_operand" to
14487 "csky_valid_mem_constraint_operand" to deal with both "Q" and "W"
14489 ("Dv"): New constraint for const double value that can be used at
14491 * config/csky/csky-modes.def (HFmode): New mode.
14492 * config/csky/csky-protos.h (csky_valid_fpuv2_mem_operand): Rename
14493 to "csky_valid_mem_constraint_operand" and support new constraint
14495 (csky_get_movedouble_length): New.
14496 (fpuv3_output_move): New.
14497 (fpuv3_const_double): New.
14498 * config/csky/csky.c (csky_option_override): New arch CK860 with fpv3.
14499 (decompose_csky_address): Refine.
14500 (csky_print_operand): New "CONST_DOUBLE" operand.
14501 (csky_output_move): Support fpv3 instructions.
14502 (csky_get_movedouble_length): New.
14503 (fpuv3_output_move): New.
14504 (fpuv3_const_double): New.
14505 (csky_emit_compare): Cover float comparsion.
14506 (csky_emit_compare_float): Refine.
14507 (csky_vaild_fpuv2_mem_operand): Rename to
14508 "csky_valid_mem_constraint_operand" and support new constraint "W".
14509 (ck860_rtx_costs): New.
14510 (csky_rtx_costs): Add the cost calculation of CK860.
14511 (regno_reg_class): New vregs for fpuv3.
14512 (csky_dbx_regno): Likewise.
14513 (csky_cpu_cpp_builtins): New builtin macro for fpuv3.
14514 (csky_conditional_register_usage): Suporrot fpuv3.
14515 (csky_dwarf_register_span): Suporrot fpuv3.
14516 (csky_init_builtins, csky_mangle_type): Support "__fp16" type.
14517 (ck810_legitimate_index_p): Support fp16.
14518 * config/csky/csky.h (TARGET_TLS): ADD CK860.
14519 (CSKY_VREG_P, CSKY_VREG_LO_P, CSKY_VREG_HI_P): Support fpuv3.
14520 (TARGET_SINGLE_FPU): Support fpuv3.
14521 (TARGET_SUPPORT_FPV3): New.
14522 (FIRST_PSEUDO_REGISTER): Change to 202 to hold the new fpuv3 registers.
14523 (FIXED_REGISTERS, CALL_REALLY_USED_REGISTERS, REGISTER_NAMES,
14524 REG_CLASS_CONTENTS): Support fpuv3.
14525 * config/csky/csky.md (movsf): Move to cksy_insn_fpu.md and refine.
14526 (csky_movsf_fpv2): Likewise.
14527 (ck801_movsf): Likewise.
14528 (csky_movsf): Likewise.
14530 (csky_movdf_fpv2): Likewise.
14531 (ck801_movdf): Likewise.
14532 (csky_movdf): Likewise.
14533 (movsicc): Refine. Use "comparison_operatior" instead of
14534 "ordered_comparison_operatior".
14535 (addsicc): Likewise.
14536 (CSKY_FIRST_VFP3_REGNUM, CSKY_LAST_VFP3_REGNUM): New constant.
14537 (call_value_internal_vh): New.
14538 * config/csky/csky_cores.def (CK860): New arch and cpu.
14543 * config/csky/csky_insn_fpu.md: Refactor. Separate all float patterns
14544 into emit-patterns and match-patterns, remain the emit-patterns here,
14545 and move the match-patterns to csky_insn_fpuv2.md or
14546 csky_insn_fpuv3.md.
14547 * config/csky/csky_insn_fpuv2.md: New file for fpuv2 instructions.
14548 * config/csky/csky_insn_fpuv3.md: New file and new patterns for fpuv3
14550 * config/csky/csky_isa.def (fcr): New.
14555 (CK860): New definition for ck860.
14556 * config/csky/csky_tables.opt (ck860): New processors ck860,
14557 ck860f. And new arch ck860.
14562 * config/csky/predicates.md (csky_float_comparsion_operator): Delete
14563 "geu", "gtu", "leu", "ltu", which will never appear at float comparison.
14564 * config/csky/t-csky-elf: Support 860.
14565 * config/csky/t-csky-linux: Likewise.
14566 * doc/md.texi: Add "Q" and "W" constraints for C-SKY.
14568 2021-05-24 Aaron Sawdey <acsawdey@linux.ibm.com>
14570 * config/rs6000/genfusion.pl (gen_logical_addsubf): Refactor to
14571 add generation of logical-add and add-logical fusion pairs.
14572 * config/rs6000/rs6000-cpus.def: Add new fusion to ISA 3.1 mask
14574 * config/rs6000/rs6000.c (rs6000_option_override_internal): Turn on
14575 logical-add and add-logical fusion by default.
14576 * config/rs6000/rs6000.opt: Add -mpower10-fusion-logical-add and
14577 -mpower10-fusion-add-logical options.
14578 * config/rs6000/fusion.md: Regenerate file.
14580 2021-05-24 Aldy Hernandez <aldyh@redhat.com>
14582 * value-range.cc (irange::legacy_equal_p): Check type when
14583 comparing VR_VARYING types.
14584 (range_tests_legacy): Test comparing VARYING ranges of different
14587 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
14589 * config/aarch64/aarch64.c (neoversen1_tunings):
14590 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
14592 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
14594 * config/aarch64/aarch64.c (aarch64_classify_symbol): Use GOT for
14595 extern weak symbols. Limit symbol offsets for non-GOT symbols with
14598 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
14600 * config/arm/neon.md (vec_load_lanesxi<mode>)
14601 (vec_store_lanexoi<mode>): Move ...
14602 * config/arm/vec-common.md: here.
14604 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
14606 * config/arm/neon.md (vec_load_lanesoi<mode>)
14607 (vec_store_lanesoi<mode>): Move ...
14608 * config/arm/vec-common.md: here.
14610 2021-05-24 liuhongt <hongtao.liu@intel.com>
14613 * config/i386/i386.c (ix86_gimple_fold_builtin): Replacing
14614 stmt with GIMPLE_NOP when lhs doesn't exist.
14616 2021-05-23 Uroš Bizjak <ubizjak@gmail.com>
14619 * config/i386/mmx.md (*push<VI_32:mode>2_rex64):
14620 New instruction pattern.
14621 (*push<VI_32:mode>2): Ditto.
14622 (push splitter for SSE registers): New splitter.
14624 2021-05-23 Andrew Pinski <apinski@marvell.com>
14626 * match.pd ((A & C) != 0 ? D : 0): Limit to non pointer types.
14628 2021-05-22 Aaron Sawdey <acsawdey@linux.ibm.com>
14630 * config/rs6000/genfusion.pl (gen_addadd): Fix incorrect attr types.
14631 * config/rs6000/fusion.md: Regenerate file.
14633 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
14635 * config/rs6000/genfusion.pl (gen_addadd): New function.
14636 * config/rs6000/fusion.md: Regenerate file.
14637 * config/rs6000/rs6000-cpus.def: Add
14638 OPTION_MASK_P10_FUSION_2ADD to masks.
14639 * config/rs6000/rs6000.c (rs6000_option_override_internal):
14640 Handle default value of OPTION_MASK_P10_FUSION_2ADD.
14641 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2add.
14643 2021-05-21 Jakub Jelinek <jakub@redhat.com>
14645 PR middle-end/99928
14646 * tree.h (OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET): Define.
14647 * gimplify.c (enum gimplify_omp_var_data): Fix up
14648 GOVD_MAP_HAS_ATTACHMENTS value, add GOVD_FIRSTPRIVATE_IMPLICIT.
14649 (omp_lastprivate_for_combined_outer_constructs): If combined target
14650 has GOVD_FIRSTPRIVATE_IMPLICIT set for the decl, change it to
14651 GOVD_MAP | GOVD_SEEN.
14652 (gimplify_scan_omp_clauses): Set GOVD_FIRSTPRIVATE_IMPLICIT for
14653 firstprivate clauses with OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT.
14654 (gimplify_adjust_omp_clauses): For firstprivate clauses with
14655 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT either clear that bit and
14656 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET too, or remove it and
14657 let it be replaced by implicit map clause.
14659 2021-05-21 Jakub Jelinek <jakub@redhat.com>
14661 PR middle-end/99928
14662 * gimplify.c (omp_lastprivate_for_combined_outer_constructs): New
14664 (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LASTPRIVATE>: Use it.
14665 (gimplify_omp_for): Likewise.
14667 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14669 PR middle-end/90115
14670 * omp-low.c (oacc_privatization_candidate_p): Reject 'static',
14671 'external' in blocks.
14673 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14675 PR middle-end/90115
14676 * flag-types.h (enum openacc_privatization): New.
14677 * params.opt (-param=openacc-privatization): New.
14678 * doc/invoke.texi (openacc-privatization): Document it.
14679 * omp-general.h (get_openacc_privatization_dump_flags): New
14681 * omp-low.c (oacc_privatization_candidate_p): Add diagnostics.
14682 * omp-offload.c (execute_oacc_device_lower)
14683 <IFN_UNIQUE_OACC_PRIVATE>: Re-work diagnostics.
14684 * target.def (goacc.adjust_private_decl): Add 'location_t'
14686 * doc/tm.texi: Regenerate.
14687 * config/gcn/gcn-protos.h (gcn_goacc_adjust_private_decl): Adjust.
14688 * config/gcn/gcn-tree.c (gcn_goacc_adjust_private_decl): Likewise.
14689 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl):
14690 Likewise. Preserve it for...
14691 (nvptx_goacc_expand_var_decl): ... use here.
14693 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14695 * doc/sourcebuild.texi (Other attributes): Document '__OPTIMIZE__'
14698 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14700 PR middle-end/90115
14701 * omp-low.c (oacc_privatization_candidate_p): New function.
14702 (oacc_privatization_scan_clause_chain)
14703 (oacc_privatization_scan_decl_chain): Use it. Also
14704 'gcc_checking_assert' that we're not seeing duplicates.
14706 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14708 PR middle-end/90115
14709 * omp-offload.c (execute_oacc_device_lower): Skip processing if no
14712 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14714 PR middle-end/90115
14715 * omp-offload.c (execute_oacc_device_lower): Explain.
14717 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14719 PR middle-end/90115
14720 * omp-offload.c (execute_oacc_device_lower)
14721 <IFN_UNIQUE_OACC_PRIVATE>: Diagnose and handle for 'level == -1'
14723 * internal-fn.c (expand_UNIQUE): Don't expect
14724 'IFN_UNIQUE_OACC_PRIVATE'.
14726 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14728 PR middle-end/90115
14729 * omp-low.c (lower_omp_for): Don't evaluate OpenMP 'for' clauses.
14731 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14733 PR middle-end/90115
14734 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl)
14735 (nvptx_goacc_expand_var_decl): Tighten.
14737 2021-05-21 Julian Brown <julian@codesourcery.com>
14738 Chung-Lin Tang <cltang@codesourcery.com>
14739 Thomas Schwinge <thomas@codesourcery.com>
14741 PR middle-end/90115
14742 * doc/tm.texi.in (TARGET_GOACC_EXPAND_VAR_DECL)
14743 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Add documentation hooks.
14744 * doc/tm.texi: Regenerate.
14745 * expr.c (expand_expr_real_1): Expand decls using the
14746 expand_var_decl OpenACC hook if defined.
14747 * internal-fn.c (expand_UNIQUE): Handle IFN_UNIQUE_OACC_PRIVATE.
14748 * internal-fn.h (IFN_UNIQUE_CODES): Add OACC_PRIVATE.
14749 * omp-low.c (omp_context): Add oacc_privatization_candidates
14751 (lower_oacc_reductions): Add PRIVATE_MARKER parameter. Insert
14753 (lower_oacc_head_tail): Add PRIVATE_MARKER parameter. Modify
14754 private marker's gimple call arguments, and pass it to
14755 lower_oacc_reductions.
14756 (oacc_privatization_scan_clause_chain)
14757 (oacc_privatization_scan_decl_chain, lower_oacc_private_marker):
14759 (lower_omp_for, lower_omp_target, lower_omp_1): Use these.
14760 * omp-offload.c (convert.h): Include.
14761 (oacc_loop_xform_head_tail): Treat private-variable markers like
14762 fork/join when transforming head/tail sequences.
14763 (struct var_decl_rewrite_info): Add struct.
14764 (oacc_rewrite_var_decl, is_sync_builtin_call): New functions.
14765 (execute_oacc_device_lower): Support rewriting gang-private
14766 variables using target hook, and fix up addr_expr and var_decl
14768 * target.def (adjust_private_decl, expand_var_decl): New hooks.
14769 * config/gcn/gcn-protos.h (gcn_goacc_adjust_gangprivate_decl):
14771 (gcn_goacc_adjust_private_decl): ...this.
14772 * config/gcn/gcn-tree.c (gcn_goacc_adjust_gangprivate_decl):
14774 (gcn_goacc_adjust_private_decl): ...this. Add LEVEL parameter.
14775 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_GANGPRIVATE_DECL): Rename
14776 definition using gcn_goacc_adjust_gangprivate_decl...
14777 (TARGET_GOACC_ADJUST_PRIVATE_DECL): ...to this, using
14778 gcn_goacc_adjust_private_decl.
14779 * config/nvptx/nvptx.c (tree-pretty-print.h): Include.
14780 (gang_private_shared_size): New global variable.
14781 (gang_private_shared_align): Likewise.
14782 (gang_private_shared_sym): Likewise.
14783 (gang_private_shared_hmap): Likewise.
14784 (nvptx_option_override): Initialize these.
14785 (nvptx_file_end): Output gang_private_shared_sym.
14786 (nvptx_goacc_adjust_private_decl, nvptx_goacc_expand_var_decl):
14788 (nvptx_set_current_function): Clear gang_private_shared_hmap.
14789 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Define hook.
14790 (TARGET_GOACC_EXPAND_VAR_DECL): Likewise.
14792 2021-05-21 H.J. Lu <hjl.tools@gmail.com>
14794 * config/i386/i386-modes.def (MAX_BITSIZE_MODE_ANY_INT): Removed.
14796 2021-05-21 Richard Biener <rguenther@suse.de>
14797 H.J. Lu <hjl.tools@gmail.com>
14799 PR middle-end/90773
14800 * expr.c (expand_constructor): Elide expand_constructor if
14801 move by pieces is preferred.
14803 2021-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14805 * config/aarch64/aarch64-builtins.c (aarch64_call_properties):
14806 Take a flag and mode value as arguments.
14807 (aarch64_modifies_global_state_p): Likewise.
14808 (aarch64_reads_global_state_p): Likewise.
14809 (aarch64_could_trap_p): Likewise.
14810 (aarch64_get_attributes): Likewise.
14811 (aarch64_init_simd_builtins): Adjust callsite of above.
14812 (aarch64_init_fcmla_laneq_builtins): Use aarch64_get_attributes to get
14813 function attributes to apply to builtins.
14814 (aarch64_init_crc32_builtins): Likewise.
14815 (aarch64_init_builtin_rsqrt): Likewise.
14817 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
14819 * config/rs6000/rs6000.md (define_attr "type"): Add types for fusion.
14820 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10): Use new fusion types.
14821 (gen_2logical): Use new fusion types.
14822 * config/rs6000/fusion.md: Regenerate.
14824 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
14827 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
14828 Handle V4QI and V2HI modes.
14829 (ix86_expand_sse_movcc): Ditto.
14830 * config/i386/mmx.md (*<sat_plusminus:insn><VI_32:mode>3):
14831 New instruction pattern.
14832 (*eq<VI_32:mode>3): Ditto.
14833 (*gt<VI_32:mode>3): Ditto.
14834 (*xop_pcmov_<VI_32:mode>): Ditto.
14835 (mmx_pblendvb32): Ditto.
14836 (mmx_pblendvb64): Rename from mmx_pblendvb.
14837 (vec_cmp<VI_32:mode><VI_32:mode>): New expander.
14838 (vec_cmpu<VI_32:mode><VI_32:mode>): Ditto.
14839 (vcond<VI_32:mode><VI_32:mode>): Ditto.
14840 (vcondu<VI_32:mode><VI_32:mode>): Ditto.
14841 (vcond_mask_<VI_32:mode><VI_32:mode>): Ditto.
14843 2021-05-21 Jakub Jelinek <jakub@redhat.com>
14845 PR tree-optimization/94589
14846 * tree-ssa-phiopt.c (spaceship_replacement): For integral rhs1 and
14847 rhs2, treat x <= 4 equivalently to x < 5 etc. In cmp1 and cmp2 (if
14848 not the same as cmp3) treat <= the same as < and >= the same as >.
14849 Don't require that cond2_phi_edge is true edge, instead take
14850 false/true edges into account based on cmp1/cmp2 comparison kinds.
14852 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
14855 * config/i386/mmx.md (SMAXMIN_MMXMODEI): New mode iterator.
14856 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Macroize expander
14857 from <smaxmin:code>v4hi3> and <smaxmin:code><MMXMODE14:mode>3
14858 using SMAXMIN_MMXMODEI mode iterator.
14859 (*<smaxmin:code>v4qi3): New insn pattern.
14860 (*<smaxmin:code>v2hi3): Ditto.
14861 (SMAXMIN_VI_32): New mode iterator.
14862 (<smaxmin:code><SMAXMIN_VI_32>mode3): New expander.
14863 (UMAXMIN_MMXMODEI): New mode iterator.
14864 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Macroize expander
14865 from <umaxmin:code>v8qi3> and <umaxmin:code><MMXMODE24:mode>3
14866 using UMAXMIN_MMXMODEI mode iterator.
14867 (*<umaxmin:code>v4qi3): New insn pattern.
14868 (*<umaxmin:code>v2hi3): Ditto.
14869 (UMAXMIN_VI_32): New mode iterator.
14870 (<umaxmin:code><UMAXMIN_VI_32>mode3): New expander.
14871 (abs<VI_32:mode>2): New insn pattern.
14872 (ssse3_abs<MMXMODEI:mode>2, abs<MMXMODEI:mode>2): Move from ...
14873 * config/i386/sse.md: ... here.
14875 2021-05-20 Clement Chigot <clement.chigot@atos.net>
14876 David Edelsohn <dje.gcc@gmail.com>
14878 * collect2.c (scan_prog_file): Issue non-fatal warning for
14881 2021-05-20 Jonathan Wakely <jwakely@redhat.com>
14883 * doc/invoke.texi (-Wno-c++11-extensions)
14884 (-Wno-c++14-extensions, -Wno-c++17-extensions)
14885 (-Wno-c++20-extensions, -Wno-c++23-extensions): Document
14888 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
14890 * config/c6x/c6x.c (c6x_output_file_unwind): Use dwarf_debuginfo_p.
14891 * config/darwin.c (darwin_override_options): Likewise.
14892 * config/i386/cygming.h (DBX_REGISTER_NUMBER): Likewise.
14893 * config/i386/darwin.h (DBX_REGISTER_NUMBER): Likewise.
14894 (DWARF2_FRAME_REG_OUT): Likewise.
14895 * config/mips/mips.c (mips_output_filename): Likewise.
14896 * config/rs6000/rs6000.c (rs6000_xcoff_declare_function_name):
14898 (rs6000_dbx_register_number): Likewise.
14899 * dbxout.c: Include flags.h.
14900 * dwarf2cfi.c (cfi_label_required_p): Likewise.
14901 (dwarf2out_do_frame): Likewise.
14902 * except.c: Include flags.h.
14903 * final.c (dwarf2_debug_info_emitted_p): Likewise.
14904 (final_scan_insn_1): Likewise.
14905 * flags.h (dwarf_debuginfo_p): New function declaration.
14906 * opts.c (dwarf_debuginfo_p): New function definition.
14907 * targhooks.c (default_debug_unwind_info): Use dwarf_debuginfo_p.
14908 * toplev.c (process_options): Likewise.
14910 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
14912 * common.opt: Change type to support bitmasks.
14913 * flag-types.h (enum debug_info_type): Rename enumerator constants.
14914 (NO_DEBUG): New bitmask.
14915 (DBX_DEBUG): Likewise.
14916 (DWARF2_DEBUG): Likewise.
14917 (XCOFF_DEBUG): Likewise.
14918 (VMS_DEBUG): Likewise.
14919 (VMS_AND_DWARF2_DEBUG): Likewise.
14920 * flags.h (debug_set_to_format): New function declaration.
14921 (debug_set_count): Likewise.
14922 (debug_set_names): Likewise.
14923 * opts.c (debug_type_masks): Array of bitmasks for debug formats.
14924 (debug_set_to_format): New function definition.
14925 (debug_set_count): Likewise.
14926 (debug_set_names): Likewise.
14927 (set_debug_level): Update access to debug_type_names.
14928 * toplev.c: Likewise.
14930 2021-05-20 Martin Sebor <msebor@redhat.com>
14932 PR middle-end/100684
14933 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Handle C++ lambda.
14935 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
14938 * config/i386/i386.md (isa): Remove x64_bmi.
14939 (enabled): Remove x64_bmi.
14940 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
14941 Remove general register alternative.
14942 (*andnot<VI_32:mode>3): Ditto.
14943 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
14944 (*<any_logic:code><VI_32:mode>3): Ditto.
14946 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
14948 * config/arm/arm.c: Include head files tree-vectorizer.h and
14951 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
14954 * config/i386/mmx.md (Yv_Yw): Revert adding V4QI and V2HI modes.
14955 (*<plusminus:insn><VI_32:mode>3): Use Yw instad of <Yv_Yw> constrint.
14956 (<s>mulv4hi3_highpart): New expander.
14957 (*<s>mulv2hi3_highpart): New insn pattern.
14958 (<s>mulv2hi3_higpart): New expander.
14959 (*<any_shift:insn>v2hi3): New insn pattern.
14960 (<any_shift:insn>v2hi3): New expander.
14961 * config/i386/sse.md (smulhrsv2hi3): New expander.
14962 (*smulhrsv2hi3): New insn pattern.
14964 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
14966 * doc/invoke.texi (vect-inner-loop-cost-factor): Document new
14968 * params.opt (vect-inner-loop-cost-factor): New.
14969 * targhooks.c (default_add_stmt_cost): Replace hardcoded factor
14970 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR, include head file
14971 tree-vectorizer.h and its required ones.
14972 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Replace
14973 hardcoded factor 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR.
14974 * config/arm/arm.c (arm_add_stmt_cost): Likewise.
14975 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
14976 * config/rs6000/rs6000.c (rs6000_add_stmt_cost): Likewise.
14977 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
14979 (_loop_vec_info::_loop_vec_info): Init inner_loop_cost_factor.
14980 * tree-vectorizer.h (_loop_vec_info): Add inner_loop_cost_factor.
14981 (LOOP_VINFO_INNER_LOOP_COST_FACTOR): New macro.
14983 2021-05-20 Christophe Lyon <christophe.lyon@linaro.org>
14984 Torbjörn Svensson <torbjorn.svensson@st.com>
14987 * doc/cpp.texi (Common Predefined Macros): Document __FILE_NAME__.
14989 2021-05-20 Jakub Jelinek <jakub@redhat.com>
14991 PR middle-end/99928
14992 * gimplify.c (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LINEAR>: For
14993 explicit linear clause when combined with target, make it map(tofrom:)
14994 instead of no clause or firstprivate.
14996 2021-05-20 Jakub Jelinek <jakub@redhat.com>
14998 PR tree-optimization/94589
14999 * match.pd ((X & Y) == X -> (X & ~Y) == 0): Simplify even in presence
15000 of integral conversions.
15002 2021-05-19 Andrew MacLeod <amacleod@redhat.com>
15004 * gimple-range.cc (fur_source::get_operand): New.
15005 (gimple_range_fold): Delete.
15006 (fold_using_range::fold_stmt): Move from gimple_ranger::calc_stmt.
15007 (fold_using_range::range_of_range_op): Move from gimple_ranger.
15008 (fold_using_range::range_of_address): Ditto.
15009 (fold_using_range::range_of_phi): Ditto.
15010 (fold_using_range::range_of_call): Ditto.
15011 (fold_using_range::range_of_builtin_ubsan_call): Move from
15012 range_of_builtin_ubsan_call.
15013 (fold_using_range::range_of_builtin_call): Move from
15014 range_of_builtin_call.
15015 (gimple_ranger::range_of_builtin_call): Delete.
15016 (fold_using_range::range_of_cond_expr): Move from gimple_ranger.
15017 (gimple_ranger::fold_range_internal): New.
15018 (gimple_ranger::range_of_stmt): Use new fold_using_range API.
15019 (fold_using_range::range_of_ssa_name_with_loop_info): Move from
15020 gimple_ranger. Improve ranges of SSA_NAMES when possible.
15021 * gimple-range.h (gimple_ranger): Remove various range_of routines.
15022 (class fur_source): New.
15023 (class fold_using_range): New.
15024 (fur_source::fur_source): New.
15026 * vr-values.c (vr_values::extract_range_basic): Use fold_using_range
15027 instead of range_of_builtin_call.
15029 2021-05-19 Jonathan Wakely <jwakely@redhat.com>
15031 * doc/cpp.texi (Common Predefined Macros): Update documentation
15032 for the __GXX_EXPERIMENTAL_CXX0X__ macro.
15034 2021-05-19 Alex Coplan <alex.coplan@arm.com>
15037 * config/arm/arm.md (nonsecure_call_internal): Always ensure
15038 callee's address is in a register.
15040 2021-05-19 Geng Qi <gengqi@linux.alibaba.com>
15042 * common/config/riscv/riscv-common.c
15043 (riscv_subset_list::parsing_subset_version): Properly parse the letter
15045 (riscv_subset_list::parse_std_ext,
15046 riscv_subset_list::parse_multiletter_ext): To handle errors generated
15047 in riscv_subset_list::parsing_subset_version.
15049 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15051 * config/aarch64/aarch64-simd.md: Use "neon_move_narrow_q"
15052 type attribute in patterns generating XTN(2).
15054 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15056 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_pack_trunc_<mode>):
15057 Remove as duplicate of...
15058 (aarch64_xtn<mode>): This.
15059 (aarch64_xtn2<mode>_le): Move position in file.
15060 (aarch64_xtn2<mode>_be): Move position in file.
15061 (aarch64_xtn2<mode>): Move position in file.
15062 (vec_pack_trunc_<mode>): Define as an expander.
15064 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15066 * config/aarch64/aarch64-simd-builtins.def: Split builtin
15067 generation for aarch64_<sur>q<r>shr<u>n_n<mode> pattern into
15068 separate scalar and vector generators.
15069 * config/aarch64/aarch64-simd.md
15070 (aarch64_<sur>q<r>shr<u>n_n<mode>): Define as an expander and
15072 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_le): This and...
15073 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_be): This.
15074 * config/aarch64/iterators.md: Define SD_HSDI iterator.
15076 2021-05-19 Jonathn Wright <jonathan.wright@arm.com>
15078 * config/aarch64/aarch64-simd.md: Use UNSPEC_SQXTUN instead
15080 * config/aarch64/iterators.md: Remove UNSPEC_SQXTUN2.
15082 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15084 * config/aarch64/aarch64-simd.md (aarch64_<sur>q<r>shr<u>n2_n<mode>):
15085 Implement as an expand emitting a big/little endian
15086 instruction pattern.
15087 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_le): Define.
15088 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_be): Define.
15090 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15092 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn2<mode>):
15093 Implement as an expand emitting a big/little endian
15094 instruction pattern.
15095 (aarch64_<sur><addsub>hn2<mode>_insn_le): Define.
15096 (aarch64_<sur><addsub>hn2<mode>_insn_be): Define.
15097 * config/aarch64/iterators.md: Remove UNSPEC_[R]ADDHN2 and
15098 UNSPEC_[R]SUBHN2 unspecs and ADDSUBHN2 iterator.
15100 2021-05-19 Richard Biener <rguenther@suse.de>
15102 PR middle-end/100672
15103 * fold-const.c (fold_negate_expr_1): Use element_precision.
15104 (negate_expr_p): Likewise.
15106 2021-05-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
15108 * config/aarch64/iterators.md (SVE_PRED_LOAD): New iterator.
15109 (pred_load): New int attribute.
15110 * config/aarch64/aarch64-sve.md
15111 (aarch64_load_<ANY_EXTEND:optab><SVE_HSDI:mode><SVE_PARTIAL_I:mode>): Use
15112 SVE_PRED_LOAD enum iterator and corresponding pred_load attribute.
15113 * config/aarch64/aarch64-sve-builtins-base.cc (expand): Update call to
15114 code_for_aarch64_load.
15116 2021-05-19 Richard Biener <rguenther@suse.de>
15118 * cfgexpand.c (discover_nonconstant_array_refs_r): Make
15119 sure TARGET_MEM_REF bases are expanded as memory.
15120 * tree-ssa-operands.c (operands_scanner::get_tmr_operands):
15121 Do not mark TARGET_MEM_REF bases addressable.
15122 * tree-ssa.c (non_rewritable_mem_ref_base): Handle
15123 TARGET_MEM_REF bases as never rewritable.
15124 * gimple-walk.c (walk_stmt_load_store_addr_ops): Do not
15125 walk TARGET_MEM_REF bases as address-takens.
15126 * tree-ssa-dce.c (ref_may_be_aliased): Handle TARGET_MEM_REF.
15128 2021-05-19 Richard Biener <rguenther@suse.de>
15130 * builtins.c (get_object_alignment_1): Strip outer
15132 * tree-dfa.c (get_ref_base_and_extent): Handle outer
15133 WITH_SIZE_EXPR for size processing and process the
15135 * tree-ssa-alias.c (ao_ref_base_alias_set): Strip
15136 outer WITH_SIZE_EXPR.
15137 (ao_ref_base_alias_ptr_type): Likewise.
15138 (refs_may_alias_p_2): Allow WITH_SIZE_EXPR in ref->ref
15139 and handle that accordingly, stripping it for the
15140 core alias workers.
15141 * tree.c (get_base_address): Handle WITH_SIZE_EXPR by
15142 looking through it instead of returning NULL.
15144 2021-05-19 Jakub Jelinek <jakub@redhat.com>
15146 PR middle-end/100576
15147 * builtins.c (check_read_access): Convert bound to size_type_node if
15150 2021-05-19 Richard Biener <rguenther@suse.de>
15152 * tree-cfg.c (verify_types_in_gimple_min_lval): Inline...
15153 (verify_types_in_gimple_reference): ... here. Sanitize.
15154 (verify_gimple_call): Verify references in LHS and arguments.
15155 (verify_gimple_assign_single): Reject WITH_SIZE_EXPR.
15157 2021-05-19 Uroš Bizjak <ubizjak@gmail.com>
15159 * config/i386/i386.h (VALID_INT_MODE_P):
15160 Add V8QI, V4HI and V2SI modes for TARGET_64BIT.
15161 * config/i386/i386.md (isa): Add x64_bmi.
15162 (enabled): Handle x64_bmi.
15163 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
15164 Add alternative using 64bit general registers.
15165 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
15167 2021-05-19 Jakub Jelinek <jakub@redhat.com>
15169 PR middle-end/99928
15170 * tree.h (OMP_MASTER_COMBINED): Define.
15171 * gimplify.c (gimplify_scan_omp_clauses): Rewrite lastprivate
15172 handling for outer combined/composite constructs to a loop.
15173 Handle lastprivate on combined target.
15174 (gimplify_expr): Formatting fix.
15176 2021-05-19 Xionghu Luo <luoxhu@linux.ibm.com>
15178 * passes.def: Add sink_code pass before store_merging.
15179 * tree-ssa-sink.c (pass_sink_code:clone): New.
15181 2021-05-18 Bill Schmidt <wschmidt@linux.ibm.com>
15183 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Remove call to
15184 rs6000_special_adjust_field_align_p.
15185 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
15186 * config/rs6000/rs6000-call.c (rs6000_function_arg_boundary):
15187 Remove ABI warning.
15188 (rs6000_function_arg): Likewise.
15189 * config/rs6000/rs6000-protos.h
15190 (rs6000_special_adjust_field_align_p): Remove prototype.
15191 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
15193 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Remove call to
15194 rs6000_special_adjust_field_align_p.
15196 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
15199 * config/i386/i386.h (VALID_SSE2_REG_MODE):
15200 Add V4QI and V2HI modes.
15201 (VALID_INT_MODE_P): Ditto.
15202 * config/i386/mmx.md (VI_32): New mode iterator.
15203 (mmxvecsize): Handle V4QI and V2HI.
15205 (mov<VI_32:mode>): New expander.
15206 (*mov<mode>_internal): New insn pattern.
15207 (movmisalign<VI_32:mode>): New expander.
15208 (neg<VI_32:mode>): New expander.
15209 (<plusminus:insn><VI_32:mode>3): New expander.
15210 (*<plusminus:insn><VI_32:mode>3): New insn pattern.
15211 (mulv2hi3): New expander.
15212 (*mulv2hi3): New insn pattern.
15213 (one_cmpl<VI_32:mode>2): New expander.
15214 (*andnot<VI_32:mode>3): New insn pattern.
15215 (<any_logic:code><VI_32:mode>3): New expander.
15216 (*<any_logic:code><VI_32:mode>3): New insn pattern.
15218 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
15220 * config/i386/sse.md (<any_extend:insn>v4qiv4di2):
15221 Fix a mode mismatch with operand 1.
15223 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
15226 * config/i386/i386-expand.c (split_double_mode): Return
15227 temporary register when simplify_gen_subreg fails with
15228 the high half od the paradoxical subreg.
15230 2021-05-18 Richard Biener <rguenther@suse.de>
15232 * cfgexpand.c (expand_one_var): Pass in forced_stack_var
15233 and honor it when expanding.
15234 (expand_used_vars_for_block): Pass through forced_stack_var.
15235 (expand_used_vars): Likewise.
15236 (discover_nonconstant_array_refs_r): Set bits in
15237 forced_stack_vars instead of marking vars TREE_ADDRESSABLE.
15238 (avoid_type_punning_on_regs): Likewise.
15239 (discover_nonconstant_array_refs): Likewise.
15240 (pass_expand::execute): Create and pass down forced_stack_var
15241 bitmap. For parameters and returns temporarily set
15242 TREE_ADDRESSABLE when expand_function_start.
15244 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
15246 * doc/sourcebuild.texi: Document 'dg-note'.
15248 2021-05-18 Tobias Burnus <tobias@codesourcery.com>
15251 * configure: Regenerate.
15252 * configure.ac (BUILD_CFLAG, BUILD_CXXFLAGS): Add $(CFLAGS-$@).
15254 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
15256 * gimple.h (is_gimple_omp_oacc): Tighten.
15257 * omp-low.c (check_omp_nesting_restrictions): Adjust.
15259 2021-05-18 Richard Biener <rguenther@suse.de>
15261 * tree-ssa-operands.c (mark_address_taken): Simplify.
15263 2021-05-18 Martin Liska <mliska@suse.cz>
15265 * config/gcn/mkoffload.c (STR): Redefine.
15266 * config/i386/intelmic-mkoffload.c (STR): Likewise.
15267 * config/nvptx/mkoffload.c (STR): Likewise.
15269 2021-05-18 Martin Liska <mliska@suse.cz>
15271 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
15272 Use startswith function instead of strncmp.
15273 * common/config/bfin/bfin-common.c (bfin_handle_option): Likewise.
15274 * common/config/riscv/riscv-common.c (riscv_subset_list::parse): Likewise.
15275 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): Likewise.
15276 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Likewise.
15277 * config/alpha/alpha.c (alpha_elf_section_type_flags): Likewise.
15278 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
15279 * config/arm/arm.c (arm_file_start): Likewise.
15280 (arm_valid_target_attribute_rec): Likewise.
15281 (thumb1_md_asm_adjust): Likewise.
15282 * config/arm/driver-arm.c (host_detect_local_cpu): Likewise.
15283 * config/avr/avr.c (STR_PREFIX_P): Likewise.
15284 (avr_set_current_function): Likewise.
15285 (avr_handle_addr_attribute): Likewise.
15286 (avr_asm_output_aligned_decl_common): Likewise.
15287 (avr_asm_named_section): Likewise.
15288 (avr_section_type_flags): Likewise.
15289 (avr_asm_select_section): Likewise.
15290 * config/c6x/c6x.c (c6x_in_small_data_p): Likewise.
15291 (c6x_section_type_flags): Likewise.
15292 * config/darwin-c.c (darwin_cfstring_ref_p): Likewise.
15293 (darwin_objc_declare_unresolved_class_reference): Likewise.
15294 (darwin_objc_declare_class_definition): Likewise.
15295 * config/darwin.c (indirect_data): Likewise.
15296 (darwin_encode_section_info): Likewise.
15297 (darwin_objc2_section): Likewise.
15298 (darwin_objc1_section): Likewise.
15299 (machopic_select_section): Likewise.
15300 (darwin_globalize_label): Likewise.
15301 (darwin_label_is_anonymous_local_objc_name): Likewise.
15302 (darwin_asm_named_section): Likewise.
15303 (darwin_asm_output_dwarf_offset): Likewise.
15304 * config/frv/frv.c (frv_string_begins_with): Likewise.
15305 (frv_in_small_data_p): Likewise.
15306 * config/gcn/mkoffload.c (STR): Likewise.
15308 * config/i386/i386-builtins.c (get_builtin_code_for_version): Likewise.
15309 * config/i386/i386-options.c (ix86_option_override_internal): Likewise.
15310 * config/i386/i386.c (x86_64_elf_section_type_flags): Likewise.
15311 (ix86_md_asm_adjust): Likewise.
15312 * config/i386/intelmic-mkoffload.c (STR): Likewise.
15313 * config/i386/winnt.c (i386_pe_asm_named_section): Likewise.
15314 (i386_pe_file_end): Likewise.
15315 * config/ia64/ia64.c (ia64_in_small_data_p): Likewise.
15316 (ia64_section_type_flags): Likewise.
15317 * config/mips/driver-native.c (host_detect_local_cpu): Likewise.
15318 * config/mips/mips.c (mips_handle_interrupt_attr): Likewise.
15319 (mips16_stub_function_p): Likewise.
15320 (mips_function_rodata_section): Likewise.
15321 * config/msp430/msp430.c (msp430_mcu_name): Likewise.
15322 (msp430_function_section): Likewise.
15323 (msp430_section_type_flags): Likewise.
15324 (msp430_expand_helper): Likewise.
15325 * config/nios2/nios2.c (nios2_small_section_name_p): Likewise.
15326 (nios2_valid_target_attribute_rec): Likewise.
15327 * config/nvptx/mkoffload.c (process): Likewise.
15329 * config/pa/som.h: Likewise.
15330 * config/pdp11/pdp11.c (pdp11_output_ident): Likewise.
15331 * config/riscv/riscv.c (riscv_elf_select_rtx_section): Likewise.
15332 * config/rs6000/rs6000.c (VTABLE_NAME_P): Likewise.
15333 (rs6000_inner_target_options): Likewise.
15334 * config/s390/driver-native.c (s390_host_detect_local_cpu): Likewise.
15335 * config/sparc/driver-sparc.c (host_detect_local_cpu): Likewise.
15336 * config/vax/vax.c (vax_output_int_move): Likewise.
15337 * config/vms/vms-ld.c (startswith): Likewise.
15338 (process_args): Likewise.
15340 * config/vms/vms.c: Likewise.
15342 2021-05-18 Jakub Jelinek <jakub@redhat.com>
15344 PR rtl-optimization/100590
15345 * regcprop.c (copyprop_hardreg_forward_1): Only DCE dead sets if
15346 they are NONJUMP_INSN_P.
15348 2021-05-18 Jakub Jelinek <jakub@redhat.com>
15351 * function.c (push_dummy_function): Set DECL_ARTIFICIAL and
15352 DECL_ASSEMBLER_NAME on the fn_decl.
15354 2021-05-18 Jakub Jelinek <jakub@redhat.com>
15356 PR tree-optimization/94589
15357 * tree-ssa-phiopt.c (spaceship_replacement): Pattern match
15358 phi result used in (res & ~1) == 0 comparison as res >= 0 as
15359 res == 2 would be UB with -ffinite-math-only.
15361 2021-05-18 Martin Liska <mliska@suse.cz>
15363 * Makefile.in: genversion.o should depend on DATESTAMP.
15365 2021-05-18 Claudiu Zissulescu <claziss@synopsys.com>
15367 * config/arc/simdext.md (negv2si2): Remove round bracket.
15369 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
15371 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Define
15372 _Bool as macro expanding to _Bool.
15374 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
15377 * tree.c (build_reference_type_for_mode)
15378 (build_pointer_type_for_mode): Pick pointer mode if MODE argument
15380 (build_reference_type, build_pointer_type): Invoke
15381 build_*_type_for_mode with VOIDmode.
15383 2021-05-17 Andrew MacLeod <amacleod@redhat.com>
15385 PR tree-optimization/100512
15386 * gimple-range-cache.cc (ranger_cache::set_global_range): Mark const
15387 and non-zero pointer ranges as invariant.
15388 * gimple-range.cc (gimple_ranger::range_of_stmt): Remove pointer
15389 processing from here.
15391 2021-05-17 Tom de Vries <tdevries@suse.de>
15394 * config/nvptx/nvptx-protos.h (nvptx_output_atomic_insn): Declare
15395 * config/nvptx/nvptx.c (nvptx_output_barrier)
15396 (nvptx_output_atomic_insn): New function.
15397 (nvptx_print_operand): Add support for 'B'.
15398 * config/nvptx/nvptx.md: Use nvptx_output_atomic_insn for atomic
15401 2021-05-17 Aldy Hernandez <aldyh@redhat.com>
15403 PR tree-optimization/100349
15404 * vr-values.c (bounds_of_var_in_loop): Bail if scev returns
15407 2021-05-17 Tamar Christina <tamar.christina@arm.com>
15409 * config/aarch64/driver-aarch64.c (DEFAULT_ARCH): New.
15410 (host_detect_local_cpu): Use it.
15412 2021-05-17 Martin Liska <mliska@suse.cz>
15414 * doc/invoke.texi: Add 2 missing dots.
15416 2021-05-17 Marius Hillenbrand <mhillen@linux.ibm.com>
15418 PR bootstrap/100552
15419 * configure.ac: Replace pattern substitution with call to sed.
15420 * configure: Regenerate.
15422 2021-05-17 Richard Biener <rguenther@suse.de>
15424 PR middle-end/100582
15425 * tree.c (array_at_struct_end_p): Get to the base of the
15426 reference before looking for the underlying decl.
15428 2021-05-17 Joern Rennecke <joern.rennecke@embecosm.com>
15430 * genoutput.c (validate_insn_alternatives) Make "wrong number of
15431 alternatives" message more specific, and remove assumption on where
15434 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
15436 * config/arm/iterators.md (V16): New iterator.
15437 (VH_cvtto): New iterator.
15438 (v_cmp_result): Added V4HF and V8HF support.
15439 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result>): Use VDQWH.
15440 (vcond<mode><mode>): Likewise.
15441 (vcond_mask_<mode><v_cmp_result>): Likewise.
15442 (vcond<VH_cvtto><mode>): New expander.
15444 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
15446 * config/arm/arm-protos.h (arm_expand_vector_compare): Update
15448 * config/arm/arm.c (arm_expand_vector_compare): Add support for
15450 (arm_expand_vcond): Likewise.
15451 * config/arm/iterators.md (supf): Remove VCMPNEQ_S, VCMPEQQ_S,
15452 VCMPEQQ_N_S, VCMPNEQ_N_S.
15453 (VCMPNEQ, VCMPEQQ, VCMPEQQ_N, VCMPNEQ_N): Remove.
15454 * config/arm/mve.md (@mve_vcmp<mve_cmp_op>q_<mode>): Add '@' prefix.
15455 (@mve_vcmp<mve_cmp_op>q_f<mode>): Likewise.
15456 (@mve_vcmp<mve_cmp_op>q_n_f<mode>): Likewise.
15457 (@mve_vpselq_<supf><mode>): Likewise.
15458 (@mve_vpselq_f<mode>"): Likewise.
15459 * config/arm/neon.md (vec_cmp<mode><v_cmp_result): Enable for MVE
15460 and move to vec-common.md.
15461 (vec_cmpu<mode><mode>): Likewise.
15462 (vcond<mode><mode>): Likewise.
15463 (vcond<V_cvtto><mode>): Likewise.
15464 (vcondu<mode><v_cmp_result>): Likewise.
15465 (vcond_mask_<mode><v_cmp_result>): Likewise.
15466 * config/arm/unspecs.md (VCMPNEQ_U, VCMPNEQ_S, VCMPEQQ_S)
15467 (VCMPEQQ_N_S, VCMPNEQ_N_S, VCMPEQQ_U, CMPEQQ_N_U, VCMPNEQ_N_U)
15468 (VCMPGEQ_N_S, VCMPGEQ_S, VCMPGTQ_N_S, VCMPGTQ_S, VCMPLEQ_N_S)
15469 (VCMPLEQ_S, VCMPLTQ_N_S, VCMPLTQ_S, VCMPCSQ_N_U, VCMPCSQ_U)
15470 (VCMPHIQ_N_U, VCMPHIQ_U): Remove.
15471 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result): Moved
15473 (vec_cmpu<mode><mode>): Likewise.
15474 (vcond<mode><mode>): Likewise.
15475 (vcond<V_cvtto><mode>): Likewise.
15476 (vcondu<mode><v_cmp_result>): Likewise.
15477 (vcond_mask_<mode><v_cmp_result>): Likewise. Added unsafe math
15480 2021-05-17 liuhongt <hongtao.liu@intel.com>
15483 * config/i386/i386.c (ix86_gimple_fold_builtin): Use
15484 gsi_insert_seq_before instead.
15486 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
15488 * doc/sourcebuild.texi (arm_qbit_ok): Rename into...
15489 (arm_sat_ok): ...this.
15491 2021-05-17 Martin Liska <mliska@suse.cz>
15493 * lto-wrapper.c (merge_flto_options): Factor out a new function.
15494 (merge_and_complain): Use it.
15495 (run_gcc): Merge also linker command line -flto=foo argument
15498 2021-05-16 Christophe Lyon <christophe.lyon@linaro.org>
15500 * config/arm/arm.h (CPP_SPEC): Remove error message about
15501 -mlittle-endian/-mbig-endian conflict.
15503 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
15505 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
15506 __ROP_PROTECT__ if -mrop-protect is selected.
15508 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
15510 * config/rs6000/rs6000-internal.h (rs6000_stack): Add
15511 rop_hash_save_offset and rop_hash_size.
15512 * config/rs6000/rs6000-logue.c (rs6000_stack_info): Compute
15513 rop_hash_size and rop_hash_save_offset.
15514 (debug_stack_info): Dump rop_hash_save_offset and rop_hash_size.
15515 (rs6000_emit_prologue): Emit hashst[p] in prologue.
15516 (rs6000_emit_epilogue): Emit hashchk[p] in epilogue.
15517 * config/rs6000/rs6000.md (unspec): Add UNSPEC_HASHST and
15519 (hashst): New define_insn.
15520 (hashchk): Likewise.
15522 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
15524 * config/rs6000/rs6000.c (rs6000_option_override_internal):
15525 Disable shrink wrap when inserting ROP-protect instructions.
15526 * config/rs6000/rs6000.opt (mrop-protect): New option.
15527 (mprivileged): Likewise.
15528 * doc/invoke.texi: Document mrop-protect and mprivileged.
15530 2021-05-15 Hans-Peter Nilsson <hp@axis.com>
15532 * reorg.c (fill_slots_from_thread): Reinstate code typoed out in
15535 2021-05-15 Martin Jambor <mjambor@suse.cz>
15538 2021-05-13 Martin Jambor <mjambor@suse.cz>
15540 PR tree-optimization/100453
15541 * tree-sra.c (sra_modify_assign): All const base accesses do not
15542 need refreshing, not just those from decl_pool.
15543 (sra_modify_assign): Do not refresh into a const base decl.
15545 2021-05-15 Jakub Jelinek <jakub@redhat.com>
15547 PR rtl-optimization/100342
15548 * regcprop.c (copy_value): When copying a source reg in a wider
15549 mode than it has recorded for the value, adjust recorded destination
15550 mode too or punt if !REG_CAN_CHANGE_MODE_P.
15552 2021-05-14 Jason Merrill <jason@redhat.com>
15554 * intl.h: Add comments.
15556 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15558 * config/aarch64/aarch64-simd.md
15559 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Split into...
15560 (aarch64_sqdmlsl2_lane<mode>_internal): ... This...
15561 (aarch64_sqdmlal2_lane<mode>_internal): ... And this.
15562 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Split into ...
15563 (aarch64_sqdmlsl2_laneq<mode>_internal): ... This...
15564 (aarch64_sqdmlal2_laneq<mode>_internal): ... And this.
15565 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>_internal): Split into...
15566 (aarch64_sqdmlsl2_n<mode>_internal): ... This...
15567 (aarch64_sqdmlal2_n<mode>_internal): ... And this.
15569 2021-05-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15572 * config/arm/arm_neon.h (vtst_s8): Replace call to vtst builtin with it's
15573 boolean logic equivalent.
15574 (vtst_s16): Likewise.
15575 (vtst_s32): Likewise.
15576 (vtst_u8): Likewise.
15577 (vtst_u16): Likewise.
15578 (vtst_u32): Likewise.
15579 (vtst_p8): Likewise.
15580 (vtst_p16): Likewise.
15581 (vtstq_s8): Likewise.
15582 (vtstq_s16): Likewise.
15583 (vtstq_s32): Likewise.
15584 (vtstq_u8): Likewise.
15585 (vtstq_u16): Likewise.
15586 (vtstq_u32): Likewise.
15587 (vtstq_p8): Likewise.
15588 (vtstq_p16): Likewise.
15589 * config/arm/arm_neon_builtins.def: Remove entry for vtst.
15590 * config/arm/neon.md (neon_vtst<mode>): Remove pattern.
15592 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15594 * config/aarch64/aarch64-simd.md (aarch64_sqdmlal2<mode>): Merge into...
15595 (aarch64_sqdml<SBINQOPS:as>l2<mode>): ... This.
15596 (aarch64_sqdmlsl2<mode>): Delete.
15597 (aarch64_sqdmlal2_lane<mode>): Merge this...
15598 (aarch64_sqdmlsl2_lane<mode>): ... And this...
15599 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>): ... Into this.
15600 (aarch64_sqdmlal2_laneq<mode>): Merge this...
15601 (aarch64_sqdmlsl2_laneq<mode>): ... And this...
15602 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>): ... Into this.
15603 (aarch64_sqdmlal2_n<mode>): Merge this...
15604 (aarch64_sqdmlsl2_n<mode>): ... And this...
15605 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>): ... Into this.
15607 2021-05-13 Martin Sebor <msebor@redhat.com>
15609 PR middle-end/100574
15610 * builtins.c (access_ref::get_ref): Improve detection of PHIs with
15611 all null arguments.
15613 2021-05-13 Martin Sebor <msebor@redhat.com>
15615 PR tree-optimization/93100
15616 PR middle-end/98583
15617 * tree-ssa-uninit.c (check_defs): Exclude intrinsic functions that
15618 don't modify referenced objects.
15620 2021-05-13 Martin Jambor <mjambor@suse.cz>
15622 PR tree-optimization/100453
15623 * tree-sra.c (sra_modify_assign): All const base accesses do not
15624 need refreshing, not just those from decl_pool.
15625 (sra_modify_assign): Do not refresh into a const base decl.
15627 2021-05-13 Martin Liska <mliska@suse.cz>
15629 * tree-ssa-dom.c: Remove m_simplifier.
15631 2021-05-13 Richard Earnshaw <rearnsha@arm.com>
15634 * config/arm/arm.c (arm_canonicalize_comparison): Correctly
15635 canonicalize DImode inequality comparisons against the
15636 maximum integral value.
15638 2021-05-13 Jakub Jelinek <jakub@redhat.com>
15640 PR tree-optimization/98856
15641 * config/i386/i386.c (ix86_shift_rotate_cost): Add CODE argument.
15642 Expect V2DI and V4DI arithmetic right shifts to be emulated.
15643 (ix86_rtx_costs, ix86_add_stmt_cost): Adjust ix86_shift_rotate_cost
15645 * config/i386/i386-expand.c (expand_vec_perm_2perm_interleave,
15646 expand_vec_perm_2perm_pblendv): New functions.
15647 (ix86_expand_vec_perm_const_1): Use them.
15648 * config/i386/sse.md (ashr<mode>3<mask_name>): Rename to ...
15649 (<mask_codefor>ashr<mode>3<mask_name>): ... this.
15650 (ashr<mode>3): New define_expand with VI248_AVX512BW iterator.
15651 (ashrv4di3): New define_expand.
15652 (ashrv2di3): Change condition to TARGET_SSE2, handle !TARGET_XOP
15653 and !TARGET_AVX512VL expansion.
15655 2021-05-13 Uroš Bizjak <ubizjak@gmail.com>
15658 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Force mode
15659 sizes < 16 to a register when constructing vpcmov pattern.
15660 * config/i386/mmx.md (*xop_pcmov_<mode>): Use MMXMODE124 mode.
15662 2021-05-13 Martin Liska <mliska@suse.cz>
15664 * gcov-io.c (gcov_write_block): Remove.
15665 (gcov_write_words): Likewise.
15666 (gcov_read_words): Re-implement using gcov_read_bytes.
15667 (gcov_allocate): Remove.
15668 (GCOV_BLOCK_SIZE): Likewise.
15669 (struct gcov_var): Remove most of the fields.
15670 (gcov_position): Implement with ftell.
15671 (gcov_rewrite): Remove setting of start and offset fields.
15672 (from_file): Re-format.
15673 (gcov_open): Remove setbuf call. It should not be needed.
15674 (gcov_close): Remove internal buffer handling.
15675 (gcov_magic): Use __builtin_bswap32.
15676 (gcov_write_counter): Use directly gcov_write_unsigned.
15677 (gcov_write_string): Use direct fwrite and do not round
15679 (gcov_seek): Use directly fseek.
15680 (gcov_write_tag): Use gcov_write_unsigned directly.
15681 (gcov_write_length): Likewise.
15682 (gcov_write_tag_length): Likewise.
15683 (gcov_read_bytes): Use directly fread.
15684 (gcov_read_unsigned): Use gcov_read_words.
15685 (gcov_read_counter): Likewise.
15686 (gcov_read_string): Use gcov_read_bytes.
15687 * gcov-io.h (GCOV_WORD_SIZE): Adjust to reflect
15688 that size is not in bytes, but words (4B).
15689 (GCOV_TAG_FUNCTION_LENGTH): Likewise.
15690 (GCOV_TAG_ARCS_LENGTH): Likewise.
15691 (GCOV_TAG_ARCS_NUM): Likewise.
15692 (GCOV_TAG_COUNTER_LENGTH): Likewise.
15693 (GCOV_TAG_COUNTER_NUM): Likewise.
15694 (GCOV_TAG_SUMMARY_LENGTH): Likewise.
15696 2021-05-13 liuhongt <hongtao.liu@intel.com>
15699 * config/i386/sse.md (ssedoublevecmode): Add attribute for
15700 V64QI/V32HI/V16SI/V4DI.
15701 (ssehalfvecmode): Add attribute for V2DI/V2DF.
15702 (*vec_concatv4si_0): Extend to VI124_128.
15703 (*vec_concat<mode>_0): New pre-reload splitter.
15704 * config/i386/predicates.md (movq_parallel): New predicate.
15706 2021-05-13 Alexandre Oliva <oliva@adacore.com>
15708 * targhooks.c (default_zero_call_used_regs): Retry using
15709 successfully-zeroed registers as sources.
15711 2021-05-12 Tobias Burnus <tobias@codesourcery.com>
15713 * omp-low.c (finish_taskreg_scan): Use the proper detach decl.
15715 2021-05-12 Aldy Hernandez <aldyh@redhat.com>
15718 * gimple-range.cc (range_of_builtin_call): Skip out on
15719 processing __builtin_clz when varying.
15721 2021-05-12 Tom de Vries <tdevries@suse.de>
15724 * config/nvptx/nvptx-opts.h (enum ptx_version): New enum.
15725 * config/nvptx/nvptx.c (nvptx_file_start): Print .version according
15726 to ptx_version_option.
15727 * config/nvptx/nvptx.h (TARGET_PTX_6_3): Define.
15728 * config/nvptx/nvptx.md (define_insn "nvptx_shuffle<mode>")
15729 (define_insn "nvptx_vote_ballot"): Use sync variant for
15731 * config/nvptx/nvptx.opt (ptx_version): Add enum.
15732 (mptx): Add option.
15733 * doc/invoke.texi (Nvidia PTX Options): Add mptx item.
15735 2021-05-12 Richard Biener <rguenther@suse.de>
15737 PR tree-optimization/100566
15738 * tree-ssa-sccvn.c (dominated_by_p_w_unex): Properly handle
15739 allow_back for all edge queries.
15741 2021-05-12 liuhongt <hongtao.liu@intel.com>
15744 * config/i386/sse.md (<sse4_1_avx2>_pblendvb): Add
15745 splitters for pblendvb of NOT mask register.
15747 2021-05-12 Richard Biener <rguenther@suse.de>
15749 PR tree-optimization/100519
15750 * tree-ssa-reassoc.c (can_associate_p): Split into...
15751 (can_associate_op_p): ... this
15752 (can_associate_type_p): ... and this.
15753 (is_reassociable_op): Call can_associate_op_p.
15754 (break_up_subtract_bb): Call the appropriate predicates.
15755 (reassociate_bb): Likewise.
15757 2021-05-12 Martin Liska <mliska@suse.cz>
15759 * lto-wrapper.c (merge_and_complain): Merge -flto=arg options.
15760 (run_gcc): Use -flto argument detection for merged
15763 2021-05-12 Martin Liska <mliska@suse.cz>
15765 * lto-wrapper.c (print_lto_docs_link): New function.
15766 (run_gcc): Print warning about missing job server detection
15767 after we know NR of partitions. Do the same for -flto{,=1}.
15768 * opts.c (get_option_html_page): Support -flto option.
15770 2021-05-12 Martin Liska <mliska@suse.cz>
15772 * lto-wrapper.c (get_options_from_collect_gcc_options): Change
15774 (append_option): Remove.
15775 (find_option): Rework to use the vector type.
15776 (remove_option): Remove.
15777 (merge_and_complain): Use vectors for cl_decoded_option data
15779 (append_compiler_options): Likewise.
15780 (append_diag_options): Likewise.
15781 (append_linker_options): Likewise.
15782 (append_offload_options): Likewise.
15783 (compile_offload_image): Likewise.
15784 (compile_images_for_offload_targets): Likewise.
15785 (find_and_merge_options): Likewise.
15786 (run_gcc): Likewise.
15788 2021-05-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
15791 * dwarf2out.c (dwarf2out_finish): Set
15792 have_multiple_function_sections with multi-range text_section.
15794 2021-05-12 Martin Liska <mliska@suse.cz>
15796 PR bootstrap/100560
15797 * Makefile.in: Remove version.h from linker command line.
15799 2021-05-12 Richard Biener <rguenther@suse.de>
15801 PR middle-end/100547
15802 * rtl.h (rtvec_alloc): Make argument size_t.
15803 * rtl.c (rtvec_alloc): Verify the count is less than INT_MAX.
15805 2021-05-12 Jakub Jelinek <jakub@redhat.com>
15807 PR middle-end/100508
15808 * cfgexpand.c (expand_debug_expr): For DEBUG_EXPR_DECL with vector
15809 type, don't reuse DECL_RTL if it has different mode, instead force
15810 creation of a new DEBUG_EXPR.
15812 2021-05-12 Jakub Jelinek <jakub@redhat.com>
15813 Marc Glisse <marc.glisse@inria.fr>
15815 PR tree-optimization/94589
15816 * match.pd ((X & Y) == X -> (X & ~Y) == 0,
15817 (X | Y) == Y -> (X & ~Y) == 0): New GIMPLE simplifications.
15819 2021-05-12 Uroš Bizjak <ubizjak@gmail.com>
15822 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Handle V2SF mode.
15823 * config/i386/mmx.md (MMXMODE124): New mode iterator.
15825 (mmxintvecmode): New mode attribute.
15826 (mmxintvecmodelower): Ditto.
15827 (*mmx_maskcmpv2sf3_comm): New insn pattern.
15828 (*mmx_maskcmpv2sf3): Ditto.
15829 (vec_cmpv2sfv2si): New expander.
15830 (vcond<V2FI:mode>v2si): Ditto.
15831 (mmx_vlendvps): New insn pattern.
15832 (vcond<MMXMODE124:mode><MMXMODEI:mode>): Also handle V2SFmode.
15833 (vcondu<MMXMODE124:mode><MMXMODEI:mode>): Ditto.
15834 (vcond_mask_<mode><mmxintvecmodelower>): Ditto.
15836 2021-05-11 Martin Sebor <msebor@redhat.com>
15838 PR middle-end/21433
15839 * expr.c (expand_expr_real_1): Replace unreachable code with an assert.
15841 2021-05-11 Richard Biener <rguenther@suse.de>
15843 * gimple-fold.c (gimple_fold_call): Do not call
15844 maybe_fold_reference on call arguments or the static chain.
15845 (fold_stmt_1): Do not call maybe_fold_reference on GIMPLE_ASM
15848 2021-05-11 Martin Liska <mliska@suse.cz>
15850 * builtins.def (DEF_HSAIL_BUILTIN): Remove.
15851 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
15852 (DEF_HSAIL_SAT_BUILTIN): Likewise.
15853 (DEF_HSAIL_INTR_BUILTIN): Likewise.
15854 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
15855 * doc/frontends.texi: Remove BRIG.
15856 * doc/install.texi: Likewise.
15857 * doc/invoke.texi: Likewise.
15858 * doc/standards.texi: Likewise.
15859 * brig-builtins.def: Removed.
15860 * brig/ChangeLog: Removed.
15861 * brig/Make-lang.in: Removed.
15862 * brig/brig-builtins.h: Removed.
15863 * brig/brig-c.h: Removed.
15864 * brig/brig-lang.c: Removed.
15865 * brig/brigfrontend/brig-arg-block-handler.cc: Removed.
15866 * brig/brigfrontend/brig-atomic-inst-handler.cc: Removed.
15867 * brig/brigfrontend/brig-basic-inst-handler.cc: Removed.
15868 * brig/brigfrontend/brig-branch-inst-handler.cc: Removed.
15869 * brig/brigfrontend/brig-cmp-inst-handler.cc: Removed.
15870 * brig/brigfrontend/brig-code-entry-handler.cc: Removed.
15871 * brig/brigfrontend/brig-code-entry-handler.h: Removed.
15872 * brig/brigfrontend/brig-comment-handler.cc: Removed.
15873 * brig/brigfrontend/brig-control-handler.cc: Removed.
15874 * brig/brigfrontend/brig-copy-move-inst-handler.cc: Removed.
15875 * brig/brigfrontend/brig-cvt-inst-handler.cc: Removed.
15876 * brig/brigfrontend/brig-fbarrier-handler.cc: Removed.
15877 * brig/brigfrontend/brig-function-handler.cc: Removed.
15878 * brig/brigfrontend/brig-function.cc: Removed.
15879 * brig/brigfrontend/brig-function.h: Removed.
15880 * brig/brigfrontend/brig-inst-mod-handler.cc: Removed.
15881 * brig/brigfrontend/brig-label-handler.cc: Removed.
15882 * brig/brigfrontend/brig-lane-inst-handler.cc: Removed.
15883 * brig/brigfrontend/brig-machine.c: Removed.
15884 * brig/brigfrontend/brig-machine.h: Removed.
15885 * brig/brigfrontend/brig-mem-inst-handler.cc: Removed.
15886 * brig/brigfrontend/brig-module-handler.cc: Removed.
15887 * brig/brigfrontend/brig-queue-inst-handler.cc: Removed.
15888 * brig/brigfrontend/brig-seg-inst-handler.cc: Removed.
15889 * brig/brigfrontend/brig-signal-inst-handler.cc: Removed.
15890 * brig/brigfrontend/brig-to-generic.cc: Removed.
15891 * brig/brigfrontend/brig-to-generic.h: Removed.
15892 * brig/brigfrontend/brig-util.cc: Removed.
15893 * brig/brigfrontend/brig-util.h: Removed.
15894 * brig/brigfrontend/brig-variable-handler.cc: Removed.
15895 * brig/brigfrontend/hsa-brig-format.h: Removed.
15896 * brig/brigfrontend/phsa.h: Removed.
15897 * brig/brigspec.c: Removed.
15898 * brig/config-lang.in: Removed.
15899 * brig/gccbrig.texi: Removed.
15900 * brig/lang-specs.h: Removed.
15901 * brig/lang.opt: Removed.
15903 2021-05-11 Richard Biener <rguenther@suse.de>
15906 * ipa-param-manipulation.c
15907 (ipa_param_body_adjustments::modify_call_stmt): Avoid
15908 altering SSA_NAME_DEF_STMT by adjusting the calls LHS
15909 via gimple_call_lhs_ptr.
15911 2021-05-11 Alex Coplan <alex.coplan@arm.com>
15914 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear):
15915 Avoid emitting CFA adjusts on the sp if we have the fp.
15917 2021-05-11 Richard Sandiford <richard.sandiford@arm.com>
15919 * config/aarch64/iterators.md (VMUL_CHANGE_NLANES): Delete.
15920 (VMULD): New iterator.
15921 (VCOND): Handle V4HF and V8HF.
15922 (VCONQ): Fix entry for V2SF.
15923 * config/aarch64/aarch64-simd.md (mul_lane<mode>3): Use VMULD
15924 instead of VMUL. Use a 64-bit vector mode for the indexed operand.
15925 (*aarch64_mul3_elt_<vswap_width_name><mode>): Merge with...
15926 (mul_laneq<mode>3): ...this define_insn. Use VMUL instead of VDQSF.
15927 Use a 128-bit vector mode for the indexed operand. Use stype for
15928 the scheduling type.
15930 2021-05-11 Richard Biener <rguenther@suse.de>
15932 * gimple-fold.c (maybe_fold_reference): Only return
15933 is_gimple_min_invariant values.
15935 2021-05-11 Richard Biener <rguenther@suse.de>
15937 PR middle-end/100509
15938 * gimple-fold.c (fold_gimple_assign): Only call
15939 get_symbol_constant_value on register type symbols.
15941 2021-05-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
15942 Joe Ramsay <joe.ramsay@arm.com>
15945 * config/arm/arm_mve.h (__arm_vstrwq_scatter_offset): Fix wrong arguments.
15946 (__arm_vcmpneq): Remove duplicate definition.
15947 (__arm_vstrwq_scatter_offset_p): Likewise.
15948 (__arm_vmaxq_x): Likewise.
15949 (__arm_vmlsdavaq): Likewise.
15950 (__arm_vmlsdavaxq): Likewise.
15951 (__arm_vmlsdavq_p): Likewise.
15952 (__arm_vmlsdavxq_p): Likewise.
15953 (__arm_vrmlaldavhaq): Likewise.
15954 (__arm_vstrbq_p): Likewise.
15955 (__arm_vstrbq_scatter_offset): Likewise.
15956 (__arm_vstrbq_scatter_offset_p): Likewise.
15957 (__arm_vstrdq_scatter_offset): Likewise.
15958 (__arm_vstrdq_scatter_offset_p): Likewise.
15959 (__arm_vstrdq_scatter_shifted_offset): Likewise.
15960 (__arm_vstrdq_scatter_shifted_offset_p): Likewise.
15962 2021-05-11 Jakub Jelinek <jakub@redhat.com>
15964 PR middle-end/100471
15965 * omp-low.c (lower_omp_task_reductions): For OMP_TASKLOOP, if data
15966 is 0, bypass the reduction loop including
15967 GOMP_taskgroup_reduction_unregister call.
15969 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
15971 * config/rs6000/rs6000.c (struct rs6000_cost_data): New member
15972 costing_for_scalar.
15973 (rs6000_density_test): Early return if costing_for_scalar is true.
15974 (rs6000_init_cost): Init costing_for_scalar of rs6000_cost_data.
15976 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
15978 * doc/tm.texi: Regenerated.
15979 * target.def (init_cost): Add new parameter costing_for_scalar.
15980 * targhooks.c (default_init_cost): Adjust for new parameter.
15981 * targhooks.h (default_init_cost): Likewise.
15982 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
15983 (vect_compute_single_scalar_iteration_cost): Likewise.
15984 (vect_analyze_loop_2): Likewise.
15985 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
15986 (vect_bb_vectorization_profitable_p): Likewise.
15987 * tree-vectorizer.h (init_cost): Likewise.
15988 * config/aarch64/aarch64.c (aarch64_init_cost): Likewise.
15989 * config/i386/i386.c (ix86_init_cost): Likewise.
15990 * config/rs6000/rs6000.c (rs6000_init_cost): Likewise.
15992 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
15994 * config/rs6000/rs6000.c (rs6000_vect_nonmem): Renamed to
15995 vect_nonmem and moved into...
15996 (struct rs6000_cost_data): ...here.
15997 (rs6000_init_cost): Use vect_nonmem of cost_data instead.
15998 (rs6000_add_stmt_cost): Likewise.
15999 (rs6000_finish_cost): Likewise.
16001 2021-05-10 Eric Botcazou <ebotcazou@adacore.com>
16003 * range-op.cc (get_bool_state): Adjust head comment.
16004 (operator_not_equal::op1_range): Fix comment.
16005 (operator_bitwise_xor::op1_range): Remove call to gcc_unreachable.
16007 2021-05-10 Martin Sebor <msebor@redhat.com>
16009 PR middle-end/100425
16010 PR middle-end/100510
16011 * gimple-ssa-warn-alloca.c (pass_walloca::firast_time_p): Rename...
16012 (pass_walloca::xlimit_certain_p): ...to this.
16013 (pass_walloca::gate): Execute for any kind of handled warning.
16014 (pass_walloca::execute): Avoid issuing "maybe" and "unbounded"
16015 warnings when xlimit_certain_p is set.
16017 2021-05-10 Pat Haugen <pthaugen@linux.ibm.com>
16019 * config/rs6000/rs6000.c (rs6000_ira_change_pseudo_allocno_class):
16020 Return ALTIVEC_REGS if that is best_class.
16021 (rs6000_compute_pressure_classes): Add ALTIVEC_REGS.
16023 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16025 * config/arm/arm.h (CPP_SPEC): Remove error message about
16028 2021-05-10 Martin Jambor <mjambor@suse.cz>
16030 * ipa-prop.h (IPA_NODE_REF): Removed.
16031 (IPA_NODE_REF_GET_CREATE): Likewise.
16032 (IPA_EDGE_REF): Likewise.
16033 (IPA_EDGE_REF_GET_CREATE): Likewise.
16034 (IS_VALID_JUMP_FUNC_INDEX): Likewise.
16035 * ipa-cp.c (print_all_lattices): Replaced IPA_NODE_REF with a direct
16036 use of ipa_node_params_sum.
16037 (ipcp_versionable_function_p): Likewise.
16038 (push_node_to_stack): Likewise.
16039 (pop_node_from_stack): Likewise.
16040 (set_single_call_flag): Replaced two IPA_NODE_REF with one single
16041 direct use of ipa_node_params_sum.
16042 (initialize_node_lattices): Replaced IPA_NODE_REF with a direct use of
16043 ipa_node_params_sum.
16044 (ipa_context_from_jfunc): Replaced IPA_EDGE_REF with a direct use of
16046 (ipcp_verify_propagated_values): Replaced IPA_NODE_REF with a direct
16047 use of ipa_node_params_sum.
16048 (self_recursively_generated_p): Likewise.
16049 (propagate_scalar_across_jump_function): Likewise.
16050 (propagate_context_across_jump_function): Replaced IPA_EDGE_REF with a
16051 direct use of ipa_edge_args_sum, moved the lookup after the early
16052 exit. Replaced IPA_NODE_REF with a direct use of ipa_node_params_sum.
16053 (propagate_bits_across_jump_function): Replaced IPA_NODE_REF with
16054 direct uses of ipa_node_params_sum.
16055 (propagate_vr_across_jump_function): Likewise.
16056 (propagate_aggregate_lattice): Likewise.
16057 (propagate_aggs_across_jump_function): Likewise.
16058 (propagate_constants_across_call): Likewise, also replaced
16059 IPA_EDGE_REF with a direct use of ipa_edge_args_sum.
16060 (good_cloning_opportunity_p): Replaced IPA_NODE_REF with a direct use
16061 of ipa_node_params_sum.
16062 (estimate_local_effects): Likewise.
16063 (add_all_node_vals_to_toposort): Likewise.
16064 (propagate_constants_topo): Likewise.
16065 (ipcp_propagate_stage): Likewise.
16066 (ipcp_discover_new_direct_edges): Likewise.
16067 (calls_same_node_or_its_all_contexts_clone_p): Likewise.
16068 (cgraph_edge_brings_value_p): Likewise (in both overloaded functions).
16069 (get_info_about_necessary_edges): Likewise.
16070 (want_remove_some_param_p): Likewise.
16071 (create_specialized_node): Likewise.
16072 (self_recursive_pass_through_p): Likewise.
16073 (self_recursive_agg_pass_through_p): Likewise.
16074 (find_more_scalar_values_for_callers_subset): Likewise and also
16075 replaced IPA_EDGE_REF with direct uses of ipa_edge_args_sum, in one
16076 case replacing two of those with a single query.
16077 (find_more_contexts_for_caller_subset): Likewise for the
16078 ipa_polymorphic_call_context overload.
16079 (intersect_aggregates_with_edge): Replaced IPA_EDGE_REF with a direct
16080 use of ipa_edge_args_sum. Replaced IPA_NODE_REF with direct uses of
16081 ipa_node_params_sum.
16082 (find_aggregate_values_for_callers_subset): Likewise, also reusing
16083 results of ipa_edge_args_sum->get.
16084 (cgraph_edge_brings_all_scalars_for_node): Replaced IPA_NODE_REF with
16085 direct uses of ipa_node_params_sum, replaced IPA_EDGE_REF with a
16086 direct use of ipa_edge_args_sum.
16087 (cgraph_edge_brings_all_agg_vals_for_node): Likewise, moved node
16088 summary query after the early exit and reused the result later.
16089 (decide_about_value): Replaced IPA_NODE_REF with a direct use of
16090 ipa_node_params_sum.
16091 (decide_whether_version_node): Likewise. Removed re-querying for
16092 summaries after cloning.
16093 (spread_undeadness): Replaced IPA_NODE_REF with a direct use of
16094 ipa_node_params_sum.
16095 (has_undead_caller_from_outside_scc_p): Likewise, reusing results of
16097 (identify_dead_nodes): Likewise.
16098 (ipcp_store_bits_results): Replaced IPA_NODE_REF with direct uses of
16099 ipa_node_params_sum.
16100 (ipcp_store_vr_results): Likewise.
16101 * ipa-fnsummary.c (evaluate_properties_for_edge): Likewise.
16102 (ipa_fn_summary_t::duplicate): Likewise.
16103 (analyze_function_body): Likewise.
16104 (estimate_calls_size_and_time): Likewise.
16105 (ipa_cached_call_context::duplicate_from): Likewise.
16106 (ipa_call_context::equal_to): Likewise.
16107 (remap_edge_params): Likewise.
16108 (ipa_merge_fn_summary_after_inlining): Likewise.
16109 (inline_read_section): Likewise.
16110 * ipa-icf.c (sem_function::param_used_p): Likewise.
16111 * ipa-modref.c (compute_parm_map): Likewise.
16112 (compute_parm_map): Replaced IPA_EDGE_REF with a direct use of
16114 (get_access_for_fnspec): Replaced IPA_NODE_REF with a direct use of
16115 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
16117 * ipa-profile.c (check_argument_count): Likewise.
16118 * ipa-prop.c (ipa_alloc_node_params): Replaced IPA_NODE_REF_GET_CREATE
16119 with a direct use of ipa_node_params_sum.
16120 (ipa_initialize_node_params): Likewise.
16121 (ipa_print_node_jump_functions_for_edge): Replaced IPA_EDGE_REF with a
16122 direct use of ipa_edge_args_sum and reused the query result.
16123 (ipa_compute_jump_functions_for_edge): Replaced IPA_NODE_REF with a
16124 direct use of ipa_node_params_sum and replaced IPA_EDGE_REF with a
16125 direct use of ipa_edge_args_sum.
16126 (ipa_note_param_call): Replaced IPA_NODE_REF with a direct use of
16127 ipa_node_params_sum and reused the result of the query.
16128 (ipa_analyze_node): Likewise.
16129 (ipa_analyze_controlled_uses): Replaced IPA_NODE_REF with a direct use
16130 of ipa_node_params_sum.
16131 (update_jump_functions_after_inlining): Replaced IPA_EDGE_REF with
16132 direct uses of ipa_edge_args_sum.
16133 (update_indirect_edges_after_inlining): Replaced IPA_NODE_REF with
16134 direct uses of ipa_node_params_sum and replaced IPA_EDGE_REF with a
16135 direct use of ipa_edge_args_sum. Removed superficial re-querying the
16137 (propagate_controlled_uses): Replaced IPA_NODE_REF with direct uses of
16138 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
16140 (ipa_propagate_indirect_call_infos): Replaced IPA_EDGE_REF with a
16141 direct use of ipa_edge_args_sum.
16142 (ipa_edge_args_sum_t::duplicate): Replaced IPA_NODE_REF with a direct
16143 use of ipa_node_params_sum.
16144 (ipa_print_node_params): Likewise.
16145 (ipa_write_node_info): Likewise and also replaced IPA_EDGE_REF with
16146 direct uses of ipa_edge_args_sum.
16147 (ipa_read_edge_info): Replaced IPA_EDGE_REF with a direct use of
16149 (ipa_read_node_info): Replaced IPA_NODE_REF with a direct use of
16150 ipa_node_params_sum.
16151 (ipa_prop_write_jump_functions): Likewise. Move variable node to the
16152 scopes where it is used.
16154 2021-05-10 Uroš Bizjak <ubizjak@gmail.com>
16156 * config/i386/i386-expand.c (ix86_expand_sse_movcc)
16157 <case E_V2SImode>: Force op_true to register.
16159 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16161 * config/arm/iterators.md (MVE_FP_COMPARISONS): New.
16162 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_f<mode>)
16163 (mve_vcmp<mve_cmp_op>q_n_f<mode>): New, merge all vcmp_*f*
16165 (mve_vcmpeqq_f<mode>, mve_vcmpeqq_n_f<mode>, mve_vcmpgeq_f<mode>)
16166 (mve_vcmpgeq_n_f<mode>, mve_vcmpgtq_f<mode>)
16167 (mve_vcmpgtq_n_f<mode>, mve_vcmpleq_f<mode>)
16168 (mve_vcmpleq_n_f<mode>, mve_vcmpltq_f<mode>)
16169 (mve_vcmpltq_n_f<mode>, mve_vcmpneq_f<mode>)
16170 (mve_vcmpneq_n_f<mode>): Remove.
16171 * config/arm/unspecs.md (VCMPEQQ_F, VCMPEQQ_N_F, VCMPGEQ_F)
16172 (VCMPGEQ_N_F, VCMPGTQ_F, VCMPGTQ_N_F, VCMPLEQ_F, VCMPLEQ_N_F)
16173 (VCMPLTQ_F, VCMPLTQ_N_F, VCMPNEQ_F, VCMPNEQ_N_F): Remove.
16175 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16177 * config/arm/iterators.md (MVE_COMPARISONS): New.
16179 (mve_cmp_type): New.
16180 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_<mode>): New, merge all
16182 (mve_vcmpneq_<mode>, mve_vcmpcsq_n_<mode>, mve_vcmpcsq_<mode>)
16183 (mve_vcmpeqq_n_<mode>, mve_vcmpeqq_<mode>, mve_vcmpgeq_n_<mode>)
16184 (mve_vcmpgeq_<mode>, mve_vcmpgtq_n_<mode>, mve_vcmpgtq_<mode>)
16185 (mve_vcmphiq_n_<mode>, mve_vcmphiq_<mode>, mve_vcmpleq_n_<mode>)
16186 (mve_vcmpleq_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
16187 (mve_vcmpneq_n_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
16188 (mve_vcmpneq_n_<mode>): Remove.
16190 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16192 * config/arm/arm_mve.h (__arm_vcmp*): Remove 's' suffix.
16193 * config/arm/arm_mve_builtins.def (vcmp*): Remove 's' suffix.
16194 * config/arm/mve.md (mve_vcmp*): Remove 's' suffix in pattern
16197 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16199 * config/arm/arm_mve_builtins.def (vcmpneq_u): Remove.
16200 (vcmpneq_n_u): Likewise.
16201 (vcmpeqq_u,): Likewise.
16202 (vcmpeqq_n_u): Likewise.
16203 * config/arm/iterators.md (supf): Remove VCMPNEQ_U, VCMPEQQ_U,
16204 VCMPEQQ_N_U and VCMPNEQ_N_U.
16205 * config/arm/mve.md (mve_vcmpneq): Remove <supf> iteration.
16206 (mve_vcmpeqq_n): Likewise.
16207 (mve_vcmpeqq): Likewise.
16208 (mve_vcmpneq_n): Likewise.
16210 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16212 * config/arm/arm_mve.h (__arm_vcmpeq*u*, __arm_vcmpne*u*): Call
16213 the 's' version of the builtin.
16215 2021-05-10 Richard Biener <rguenther@suse.de>
16217 PR tree-optimization/100492
16218 * tree-loop-distribution.c (find_seed_stmts_for_distribution):
16219 Find nothing when the loop contains an irreducible region.
16221 2021-05-10 Richard Biener <rguenther@suse.de>
16223 PR middle-end/100464
16225 * gimple-fold.c (canonicalize_constructor_val): Do not set
16228 2021-05-10 Richard Biener <rguenther@suse.de>
16230 PR tree-optimization/100434
16231 * tree-ssa-dse.c (initialize_ao_ref_for_dse): Handle
16233 (dse_optimize_stmt): Handle call LHS by dropping the
16234 LHS or the whole call if it doesn't have other
16236 (pass_dse::execute): Adjust.
16238 2021-05-10 Martin Liska <mliska@suse.cz>
16240 * Makefile.in: Add missing genversion rule.
16242 2021-05-10 Alex Coplan <alex.coplan@arm.com>
16245 * config/arm/mve.md (*mve_mov<mode>): Simplify output code. Use
16246 vldrw.u32 and vstrw.32 for V2D[IF]mode loads and stores.
16248 2021-05-10 Martin Liska <mliska@suse.cz>
16250 * builtins.c (is_builtin_name): Use startswith
16251 function instead of strncmp.
16252 * collect2.c (main): Likewise.
16253 (has_lto_section): Likewise.
16254 (scan_libraries): Likewise.
16255 * coverage.c (coverage_checksum_string): Likewise.
16256 (coverage_init): Likewise.
16257 * dwarf2out.c (is_cxx): Likewise.
16258 (gen_compile_unit_die): Likewise.
16259 * gcc-ar.c (main): Likewise.
16260 * gcc.c (init_spec): Likewise.
16261 (read_specs): Likewise.
16262 (execute): Likewise.
16263 (check_live_switch): Likewise.
16264 * genattrtab.c (write_attr_case): Likewise.
16265 (IS_ATTR_GROUP): Likewise.
16266 * gencfn-macros.c (main): Likewise.
16267 * gengtype.c (type_for_name): Likewise.
16268 (gen_rtx_next): Likewise.
16269 (get_file_langdir): Likewise.
16270 (write_local): Likewise.
16271 * genmatch.c (get_operator): Likewise.
16272 (get_operand_type): Likewise.
16273 (expr::gen_transform): Likewise.
16274 * genoutput.c (validate_optab_operands): Likewise.
16275 * incpath.c (add_sysroot_to_chain): Likewise.
16276 * langhooks.c (lang_GNU_C): Likewise.
16277 (lang_GNU_CXX): Likewise.
16278 (lang_GNU_Fortran): Likewise.
16279 (lang_GNU_OBJC): Likewise.
16280 * lto-wrapper.c (run_gcc): Likewise.
16281 * omp-general.c (omp_max_simt_vf): Likewise.
16282 * omp-low.c (omp_runtime_api_call): Likewise.
16283 * opts-common.c (parse_options_from_collect_gcc_options): Likewise.
16284 * read-rtl-function.c (function_reader::read_rtx_operand_r): Likewise.
16285 * real.c (real_from_string): Likewise.
16286 * selftest.c (assert_str_startswith): Likewise.
16287 * timevar.c (timer::validate_phases): Likewise.
16288 * tree.c (get_file_function_name): Likewise.
16289 * ubsan.c (ubsan_use_new_style_p): Likewise.
16290 * varasm.c (default_function_rodata_section): Likewise.
16291 (incorporeal_function_p): Likewise.
16292 (default_section_type_flags): Likewise.
16293 * system.h (startswith): Define startswith.
16295 2021-05-10 Martin Liska <mliska@suse.cz>
16297 * bitmap.h (class auto_bitmap): Remove
16298 __cplusplus >= 201103.
16299 * config/aarch64/aarch64.c: Likewise.
16300 * gimple-ssa-store-merging.c (store_immediate_info::store_immediate_info):
16302 * sbitmap.h: Likewise.
16304 2021-05-10 Martin Liska <mliska@suse.cz>
16306 * Makefile.in: Rename gcov-iov to genversion and depend
16307 on version.h (instead of gcov-iov.h).
16308 * gcov-io.h: Include version.h instread of gcov-iov.h.
16309 * gengtype-state.c (read_state_version): Likewise.
16310 * gcov-iov.c: Moved to...
16311 * genversion.c: ...here.
16312 * lto-streamer.h (LTO_major_version): Define it with
16314 * version.c: Removed.
16315 * version.h: Removed.
16317 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16319 * config/arc/arc.md (UNSPEC_ARC_DMPYWH): Define.
16320 * config/arc/simdext.md (VCT): Add predicates for iterator
16323 (voptab): Likewise.
16324 (vec_widen_<V_US>mult_hi_v4hi): Change pattern predicate.
16325 (<voptab>v2si3): New patterns.
16327 (reduc_plus_scal_v4hi): Likewise.
16328 (reduc_plus_scal_v2si): Likewise.
16329 (vec_duplicatev2si): Likewise.
16330 (vec_duplicatev4hi): Likewise.
16332 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16334 * config/arc/simdext.md: Format and cleanup file.
16336 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16338 * config/arc/simdext.md (movmisalignv2hi): Allow misaligned access
16339 only when munaligned-access option is on.
16340 (movmisalign<mode>): Likewise.
16342 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16344 * common/config/arc/arc-common.c (arc_handle_option): Remove dot
16346 * config/arc/arc.c (arc_reorg): Remove underscore from string.
16348 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16350 * config/arc/arc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
16351 (CTZ_DEFINED_VALUE_AT_ZERO): Likewise.
16352 * config/arc/arc.md (clrsbsi2): Cleanup pattern.
16353 (norm_f): Likewise.
16356 (clzsi2): Use fls instruction when available.
16357 (arc_clzsi2): Likewise.
16359 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16361 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add r26 and r27.
16363 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16365 * doc/extend.texi (__builtin_arc_sr): Swap arguments.
16367 2021-05-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
16369 PR middle-end/100467
16370 * toplev.c (compile_file): Call insn_locations_init before
16371 targetm.asm_out.code_end.
16373 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
16376 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
16378 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
16380 2021-05-07 Jakub Jelinek <jakub@redhat.com>
16381 Andrew Stubbs <amd@codesourcery.com>
16384 * builtins.c (try_store_by_multiple_pieces): Use force_operand for
16385 emit_move_insn operands.
16387 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
16389 * cfgexpand.c (expand_gimple_basic_block): Do not inherit a current
16390 location for the outgoing edges of an empty block.
16391 * dwarf2out.c (add_subscript_info): Retrieve the bounds and index
16392 type by means of the get_array_descr_info langhook, if it is set and
16393 returns true. Remove obsolete code dealing with unnamed subtypes.
16395 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16397 * gimple-range-cache.cc (ssa_block_ranges): Virtualize.
16398 (sbr_vector): Renamed from ssa_block_cache.
16399 (sbr_vector::sbr_vector): Allocate from obstack abd initialize.
16400 (ssa_block_ranges::~ssa_block_ranges): Remove.
16401 (sbr_vector::set_bb_range): Use varying and undefined cached values.
16402 (ssa_block_ranges::set_bb_varying): Remove.
16403 (sbr_vector::get_bb_range): Adjust assert.
16404 (sbr_vector::bb_range_p): Adjust assert.
16405 (~block_range_cache): No freeing loop required.
16406 (block_range_cache::get_block_ranges): Remove.
16407 (block_range_cache::set_bb_range): Inline get_block_ranges.
16408 (block_range_cache::set_bb_varying): Remove.
16409 * gimple-range-cache.h (set_bb_varying): Remove prototype.
16410 * value-range.h (irange_allocator::get_memory): New.
16412 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16414 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Search
16415 dominator tree is available and requested.
16416 (ranger_cache::ssa_range_in_bb): Don't search dom tree here.
16417 (ranger_cache::fill_block_cache): Don't search dom tree here either.
16418 * gimple-range-cache.h (non_null_deref_p): Add dom_search param.
16420 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16422 * gimple-range.cc (gimple_ranger::range_on_exit): Handle block with
16423 only PHI nodes better.
16425 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16427 * gimple-range-edge.h (gimple_outgoing_range): Rename from
16429 (gcond_edge_range): Export prototype.
16430 * gimple-range-edge.cc (gcond_edge_range): New.
16431 (gimple_outgoing_range::edge_range_p): Use gcond_edge_range.
16432 * gimple-range-gori.h (gori_compute): Use gimple_outgoing_range.
16434 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16436 * gimple-range-edge.cc (outgoing_range::calc_switch_ranges): Compute
16437 default range into a temp and allocate only what is needed.
16439 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16441 * range-op.cc (operator_trunc_mod::wi_fold): x % 0 is UNDEFINED.
16443 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16445 * gimple-range.h (gimple_range_global): Pick up parameter initial
16446 values, and use-before defined locals are UNDEFINED.
16448 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
16450 * doc/extend.texi (scalar_storage_order): Mention effect on pointer
16452 * tree.h (reverse_storage_order_for_component_p): Return false if
16453 the type is a pointer.
16455 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
16457 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
16459 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
16462 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
16463 Handle V8QI, V4HI and V2SI modes.
16464 * config/i386/mmx.md (mmx_pblendvb): New insn pattern.
16465 * config/i386/sse.md (unspec): Move UNSPEC_BLENDV ...
16466 * config/i386/i386.md (unspec): ... here.
16468 2021-05-07 Tobias Burnus <tobias@codesourcery.com>
16469 Tom de Vries <tdevries@suse.de>
16471 * omp-low.c (lower_rec_simd_input_clauses): Set max_vf = 1 if
16472 a truth_value_p reduction variable is nonintegral.
16474 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
16477 * config/i386/i386-expand.c (ix86_use_mask_cmp_p):
16478 Return false for mode sizes < 16.
16480 2021-05-07 Jakub Jelinek <jakub@redhat.com>
16483 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn.
16485 2021-05-06 Martin Jambor <mjambor@suse.cz>
16487 * ipa-sra.c (ipa_sra_dump_all_summaries): Dump edge summaries even
16488 when there is no function summary.
16489 (ipa_sra_summarize_function): produce edge summaries even when
16492 2021-05-06 Tom Tromey <tom@tromey.com>
16494 * godump.c (string_hash_eq): Remove.
16495 (go_finish): Use htab_eq_string.
16497 2021-05-06 Tom Tromey <tom@tromey.com>
16499 * gengtype-state.c (read_state): Use htab_eq_string.
16500 (string_eq): Remove.
16502 2021-05-06 Tom Tromey <tom@tromey.com>
16504 * gensupport.c (htab_eq_string): Remove.
16506 2021-05-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
16509 * debug.h (gcc_debug_hooks): Add set_ignored_loc function pointer.
16510 * dwarf2out.h (dw_fde_node::ignored_debug): New data item.
16511 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Add dummy
16512 set_ignored_loc callbacks.
16513 * debug.c (do_nothing_debug_hooks): Likewise.
16514 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
16515 * dwarf2out.c (text_section_used, cold_text_section_used): Remove.
16516 (in_text_section_p, last_text_label, last_cold_label,
16517 switch_text_ranges, switch_cold_ranges): New data items.
16518 (dwarf2out_note_section_used): Remove.
16519 (dwarf2out_begin_prologue): Set fde->ignored_debug and
16521 (mark_ignored_debug_section): New helper function.
16522 (dwarf2out_end_epilogue, dwarf2out_switch_text_section): Call
16523 mark_ignored_debug_section.
16524 (dwarf2_debug_hooks): Use dwarf2out_set_ignored_loc.
16525 (dwarf2_lineno_debug_hooks): Use dummy for set_ignored_loc.
16526 (size_of_aranges): Adjust formula for multi-part text ranges size.
16527 (output_aranges): Output multi-part text ranges.
16528 (dwarf2out_set_ignored_loc): New callback function.
16529 (dwarf2out_finish): Output multi-part text ranges.
16530 (dwarf2out_c_finalize): Clear new data items.
16531 * final.c (final_start_function_1): Call set_ignored_loc callback.
16532 (final_scan_insn_1): Likewise.
16533 * ggc-page.c (gt_ggc_mx): New helper function.
16534 * stringpool.c (gt_pch_nx): Likewise.
16536 2021-05-06 Richard Biener <rguenther@suse.de>
16538 * timevar.def (TV_TREE_INSERT_PHI_NODES): Remove.
16539 (TV_TREE_SSA_REWRITE_BLOCKS): Likewise.
16540 (TV_TREE_INTO_SSA): New.
16541 * tree-into-ssa.c (insert_phi_nodes): Do not account separately.
16542 (rewrite_blocks): Likewise.
16543 (pass_data_build_ssa): Account to TV_TREE_INTO_SSA.
16545 2021-05-06 Jakub Jelinek <jakub@redhat.com>
16547 * tree-ssa-phiopt.c (value_replacement, minmax_replacement,
16548 abs_replacement, xor_replacement,
16549 cond_removal_in_popcount_clz_ctz_pattern,
16550 replace_phi_edge_with_variable): Change type of phi argument from
16551 gimple * to gphi *.
16553 2021-05-06 Richard Biener <rguenther@suse.de>
16555 * tree-ssa-loop-split.c (split_loop): Delay updating SSA form.
16556 Output an opt-info message.
16557 (do_split_loop_on_cond): Likewise.
16558 (tree_ssa_split_loops): Update SSA form here.
16560 2021-05-06 Richard Biener <rguenther@suse.de>
16562 * tree-inline.c (tree_function_versioning): Fix DECL_BY_REFERENCE
16563 return variable removal.
16565 2021-05-06 Marius Hillenbrand <mhillen@linux.ibm.com>
16567 * config/s390/s390-builtins.def (O_M5, O1_M5, ...): Remove unused macros.
16568 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
16569 (s390_vec_permi_dbl, s390_vpdi): Use the O3_U2 type for the immediate
16571 * config/s390/s390.c (s390_const_operand_ok): Remove unused
16574 2021-05-06 Jakub Jelinek <jakub@redhat.com>
16576 PR tree-optimization/94589
16577 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Call
16578 spaceship_replacement.
16579 (cond_only_block_p, spaceship_replacement): New functions.
16581 2021-05-06 Richard Biener <rguenther@suse.de>
16584 * tree-emutls.c (gen_emutls_addr): Pass in whether we're
16585 dealing with a debug use and only query existing addresses
16587 (lower_emutls_1): Avoid splitting out addresses for debug
16588 stmts, reset the debug stmt when we fail to find existing
16590 (lower_emutls_phi_arg): Set wi.stmt.
16592 2021-05-06 Christoph Muellner <cmuellner@gcc.gnu.org>
16595 * config/riscv/riscv.c (riscv_block_move_loop): Use cbranch helper.
16596 * config/riscv/riscv.md (cbranch<mode>4): Generate helpers.
16597 (stack_protect_test): Use cbranch helper.
16599 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
16602 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
16603 always return the establisher frame for __builtin_frame_address (0).
16605 2021-05-05 Ivan Sorokin <vanyacpp@gmail.com>
16608 * config/i386/i386-builtins.c (ix86_cpu_model_type_node): New.
16609 (ix86_cpu_model_var): Likewise.
16610 (ix86_cpu_features2_type_node): Likewise.
16611 (ix86_cpu_features2_var): Likewise.
16612 (fold_builtin_cpu): Cache __cpu_model and __cpu_features2 with
16615 2021-05-05 Martin Sebor <msebor@redhat.com>
16617 * passes.def (pass_warn_printf): Run after SSA.
16619 2021-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16621 * config/arm/neon.md (neon_vtst_combine<mode>): New pattern.
16622 * config/arm/predicates.md (minus_one_operand): New predicate.
16624 2021-05-05 Jeff Law <jlaw@tachyum.com>
16626 * config/avr/avr.md: Remove references to CC_STATUS_INIT.
16628 2021-05-05 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
16630 PR rtl-optimization/100263
16631 * postreload.c (move2add_valid_value_p): Ensure register can
16634 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
16636 PR rtl-optimization/100411
16637 * cfgcleanup.c (try_crossjump_to_edge): Also skip end of prologue
16638 and beginning of function markers.
16640 2021-05-05 Jeff Law <jlaw@tachyum.com>
16642 * config/cr16/cr16.h (NOTICE_UPDATE_CC): Remove.
16643 * config/cr16/cr16.c (notice_update_cc): Remove.
16644 * config/cr16/cr16-protos.h (notice_update_cc): Remove.
16646 2021-05-05 Uroš Bizjak <ubizjak@gmail.com>
16649 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
16650 Handle V8QI, V4HI and V2SI modes.
16651 * config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
16652 (ix86_build_signbit_mask): Ditto.
16653 * config/i386/mmx.md (MMXMODE14): New mode iterator.
16654 (<smaxmin:code><MMXMODE14:mode>3): New expander.
16655 (*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
16656 (<umaxmin:code><MMXMODE24:mode>3): New expander.
16657 (*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
16658 (vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
16659 (vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
16660 (vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
16661 (vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
16662 (vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
16664 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
16666 * dwarf2out.c (loc_list_from_tree_1) <DECL>: During early DWARF, do
16667 not expand the VALUE_EXPR of variables put in the non-local frame.
16668 * gimplify.c (gimplify_type_sizes) <RECORD_TYPE>: If the type is not
16669 to be ignored for debug info, ensure its variable offsets are not.
16671 2021-05-05 Richard Biener <rguenther@suse.de>
16673 PR tree-optimization/79333
16674 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
16675 Fold stmt following SSA edges.
16677 2021-05-05 Richard Biener <rguenther@suse.de>
16679 PR middle-end/100394
16680 * calls.c (expand_call): Preserve possibly throwing calls.
16681 * cfgexpand.c (expand_call_stmt): When a call can throw signal
16682 RTL expansion there are side-effects.
16683 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Simplify,
16684 mark all possibly throwing stmts necessary unless we can elide
16686 * tree-ssa-dse.c (pass_dse::execute): Preserve exceptions unless
16687 -fdelete-dead-exceptions.
16688 * tree.h (DECL_PURE_P): Add note about exceptions.
16690 2021-05-05 Alexandre Oliva <oliva@adacore.com>
16692 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Make it
16695 2021-05-04 David Edelsohn <dje.gcc@gmail.com>
16697 * config/rs6000/rs6000-call.c (rs6000_output_mi_thunk): Use
16698 get_fnname_from_decl for name of thunk.
16699 * config/rs6000/rs6000.c (rs6000_declare_alias): Use assemble_name
16700 and ASM_OUTPUT_LABEL.
16701 (rs6000_xcoff_declare_function_name): Use assemble_name and
16703 (rs6000_xcoff_declare_object_name): Use ASM_OUTPUT_LABEL.
16704 (rs6000_xcoff_encode_section_info): Don't add mapping class
16705 for aliases. Always add [DS] mapping class to primary
16707 (rs6000_asm_weaken_decl): Don't explicitly add [DS].
16709 2021-05-04 Martin Sebor <msebor@redhat.com>
16711 PR middle-end/100307
16712 * builtins.c (compute_objsize_r): Clear base0 for pointers.
16714 2021-05-04 Jeff Law <jlaw@tachyum.com>
16716 * config/bfin/bfin.h (NOTICE_UPDATE_CC): Remove.
16718 2021-05-04 Segher Boessenkool <segher@kernel.crashing.org>
16720 * caller-save.c: Remove CC0.
16721 * cfgcleanup.c: Remove CC0.
16722 * cfgrtl.c: Remove CC0.
16723 * combine.c: Remove CC0.
16724 * compare-elim.c: Remove CC0.
16725 * conditions.h: Remove CC0.
16726 * config/h8300/h8300.h: Remove CC0.
16727 * config/h8300/h8300-protos.h: Remove CC0.
16728 * config/h8300/peepholes.md: Remove CC0.
16729 * config/i386/x86-tune-sched.c: Remove CC0.
16730 * config/m68k/m68k.c: Remove CC0.
16731 * config/rl78/rl78.c: Remove CC0.
16732 * config/sparc/sparc.c: Remove CC0.
16733 * config/xtensa/xtensa.c: Remove CC0.
16734 (gen_conditional_move): Use pc_rtx instead of cc0_rtx in a piece of
16735 RTL where that is used as a placeholder only.
16736 * cprop.c: Remove CC0.
16737 * cse.c: Remove CC0.
16738 * cselib.c: Remove CC0.
16739 * df-problems.c: Remove CC0.
16740 * df-scan.c: Remove CC0.
16741 * doc/md.texi: Remove CC0. Adjust an example.
16742 * doc/rtl.texi: Remove CC0. Adjust an example.
16743 * doc/tm.texi: Regenerate.
16744 * doc/tm.texi.in: Remove CC0.
16745 * emit-rtl.c: Remove CC0.
16746 * final.c: Remove CC0.
16747 * fwprop.c: Remove CC0.
16748 * gcse-common.c: Remove CC0.
16749 * gcse.c: Remove CC0.
16750 * genattrtab.c: Remove CC0.
16751 * genconfig.c: Remove CC0.
16752 * genemit.c: Remove CC0.
16753 * genextract.c: Remove CC0.
16754 * gengenrtl.c: Remove CC0.
16755 * genrecog.c: Remove CC0.
16756 * haifa-sched.c: Remove CC0.
16757 * ifcvt.c: Remove CC0.
16758 * ira-costs.c: Remove CC0.
16759 * ira.c: Remove CC0.
16760 * jump.c: Remove CC0.
16761 * loop-invariant.c: Remove CC0.
16762 * lra-constraints.c: Remove CC0.
16763 * lra-eliminations.c: Remove CC0.
16764 * optabs.c: Remove CC0.
16765 * postreload-gcse.c: Remove CC0.
16766 * postreload.c: Remove CC0.
16767 * print-rtl.c: Remove CC0.
16768 * read-rtl-function.c: Remove CC0.
16769 * reg-notes.def: Remove CC0.
16770 * reg-stack.c: Remove CC0.
16771 * reginfo.c: Remove CC0.
16772 * regrename.c: Remove CC0.
16773 * reload.c: Remove CC0.
16774 * reload1.c: Remove CC0.
16775 * reorg.c: Remove CC0.
16776 * resource.c: Remove CC0.
16777 * rtl.c: Remove CC0.
16778 * rtl.def: Remove CC0.
16779 * rtl.h: Remove CC0.
16780 * rtlanal.c: Remove CC0.
16781 * sched-deps.c: Remove CC0.
16782 * sched-rgn.c: Remove CC0.
16783 * shrink-wrap.c: Remove CC0.
16784 * simplify-rtx.c: Remove CC0.
16785 * system.h: Remove CC0. Poison NOTICE_UPDATE_CC, CC_STATUS_MDEP_INIT,
16786 CC_STATUS_MDEP, and CC_STATUS.
16787 * target.def: Remove CC0.
16788 * valtrack.c: Remove CC0.
16789 * var-tracking.c: Remove CC0.
16791 2021-05-04 Richard Biener <rguenther@suse.de>
16793 PR tree-optimization/100414
16794 * tree-ssa-phiopt.c (get_non_trapping): Do not compute dominance
16796 (tree_ssa_phiopt_worker): But unconditionally here.
16798 2021-05-04 Tobias Burnus <tobias@codesourcery.com>
16800 * omp-low.c (lower_rec_input_clauses, lower_reduction_clauses): Handle
16801 && and || with floating-point and complex arguments.
16803 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
16805 * tree-inline.c (insert_debug_decl_map): Delete.
16806 (copy_debug_stmt): Minor tweak.
16807 (setup_one_parameter): Do not use a variable if the value is either
16808 a read-only DECL or a non-addressable local variable in the caller.
16809 In this case, insert the debug-only variable in the map manually.
16810 (expand_call_inline): Do not generate a CLOBBER for these values.
16811 * tree-inline.h (debug_map): Minor tweak.
16813 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
16815 * builtins.c (builtin_with_linkage_p): Return true for stp[n]cpy.
16816 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Tidy up.
16818 2021-05-04 Richard Biener <rguenther@suse.de>
16820 PR tree-optimization/100329
16821 * tree-ssa-reassoc.c (can_reassociate_p): Do not reassociate
16823 (insert_stmt_after): Assert we're not running into asm goto.
16825 2021-05-04 Richard Biener <rguenther@suse.de>
16827 PR tree-optimization/100398
16828 * tree-ssa-dse.c (pass_dse::execute): Preserve control
16831 2021-05-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16833 * builtins.c (try_store_by_multiple_pieces): Fix constfun's prototype.
16835 2021-05-04 Alexandre Oliva <oliva@adacore.com>
16837 * builtins.c (try_store_by_multiple_pieces): New.
16838 (expand_builtin_memset_args): Use it. If target_char_cast
16839 fails, proceed as for non-constant val. Pass len's ctz to...
16840 * expr.c (clear_storage_hints): ... this. Try store by
16841 multiple pieces after setmem.
16842 (clear_storage): Adjust.
16843 * expr.h (clear_storage_hints): Likewise.
16844 (try_store_by_multiple_pieces): Declare.
16845 * passes.def: Replace the last copy_prop with ccp.
16847 2021-05-03 Tom de Vries <tdevries@suse.de>
16850 * omp-low.c (lower_rec_input_clauses): Disable SIMT for user-defined
16853 2021-05-03 Richard Biener <rguenther@suse.de>
16855 * tree-ssa-dse.c (dse_classify_store): Track two PHI defs.
16857 2021-05-03 Richard Biener <rguenther@suse.de>
16859 * tree-ssa-dse.c: Do not include domwalk.h but cfganal.h.
16860 (dse_dom_walker): Remove.
16861 (dse_dom_walker::dse_optimize_stmt): Rename...
16862 (dse_optimize_stmt): ... to this, pass in live_bytes sbitmap.
16863 (dse_dom_walker::before_dom_children): Inline ...
16864 (pass_dse::execute): ... here. Perform a reverse program
16867 2021-05-03 H.J. Lu <hjl.tools@gmail.com>
16870 * configure: Regenerated.
16872 2021-05-03 Ilya Leoshkevich <iii@linux.ibm.com>
16875 * config/s390/s390.c (s390_hard_fp_reg_p): New function.
16876 (s390_md_asm_adjust): Handle hard registers.
16878 2021-05-03 Jakub Jelinek <jakub@redhat.com>
16880 PR tree-optimization/100382
16881 * tree-ssa-dse.c: Include tree-eh.h.
16882 (dse_dom_walker::before_dom_children): Don't remove stmts if
16883 stmt_unremovable_because_of_non_call_eh_p is true.
16885 2021-05-02 David Edelsohn <dje.gcc@gmail.com>
16887 * varasm.c (compute_reloc_for_var): Split out from...
16888 (get_variable_section): Use it.
16889 * output.h (compute_reloc_for_var): Declare.
16890 * config/rs6000/rs6000-protos.h
16891 (rs6000_xcoff_asm_output_aligned_decl_common): Change alignment to
16893 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address_aix):
16894 Don't append storage mapping class to symbol.
16895 (rs6000_xcoff_asm_named_section): Add BS and UL mapping classes.
16896 Don't convert TLS BSS to common.
16897 (rs6000_xcoff_unique_section): Don't fall back to select_secton.
16898 (rs6000_xcoff_section_type_flags): Add SECTION_BSS if DECL is
16900 (rs6000_xcoff_asm_globalize_decl_name): Don't strip storage
16902 (rs6000_xcoff_asm_output_aligned_decl_common): Align is unsigned int.
16903 If align is 0 from TLS class, use the same rules as varasm.c
16904 If not common, switch to BSS section manually.
16905 If common, emit appropriate comm or lcomm directive.
16906 (rs6000_xcoff_encode_section_info): Add logic to append all
16907 storage mapping classes.
16908 (rs6000_asm_weaken_decl): Adjust for qualname symbols.
16909 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
16910 rs6000_xcoff_asm_output_aligned_decl_common.
16911 (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
16912 rs6000_xcoff_asm_output_aligned_decl_common.
16913 (ASM_OUTPUT_TLS_COMMON): Use
16914 rs6000_xcoff_asm_output_aligned_decl_common.
16916 2021-05-02 Jakub Jelinek <jakub@redhat.com>
16919 * config/nvptx/nvptx.c (nvptx_sese_pseudo): Use nullptr instead of 0
16920 as first argument of pseudo_node_t constructors.
16922 2021-05-02 Jakub Jelinek <jakub@redhat.com>
16925 * config/i386/t-i386 (TM_H): Add $(srcdir)/config/i386/i386-isa.def.
16927 2021-05-01 Aldy Hernandez <aldyh@redhat.com>
16929 * value-range.cc (DEFINE_INT_RANGE_GC_STUBS): Remove.
16930 (gt_pch_nx (int_range<1> *&)): New.
16931 (gt_ggc_mx (int_range<1> *&)): New.
16932 * value-range.h (class irange): Add GTY support for
16935 2021-05-01 Geng Qi <gengqi@linux.alibaba.com>
16937 * doc/options.texi (Negative): Change either or to both and.
16939 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
16941 * config/aarch64/aarch64-simd-builtins.def: Add
16942 float_ml[as][q]_laneq builtin generator macros.
16943 * config/aarch64/aarch64-simd.md (mul_laneq<mode>3): Define.
16944 (aarch64_float_mla_laneq<mode>): Define.
16945 (aarch64_float_mls_laneq<mode>): Define.
16946 * config/aarch64/arm_neon.h (vmla_laneq_f32): Use RTL builtin
16947 instead of GCC vector extensions.
16948 (vmlaq_laneq_f32): Likewise.
16949 (vmls_laneq_f32): Likewise.
16950 (vmlsq_laneq_f32): Likewise.
16952 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
16954 * config/aarch64/aarch64-simd-builtins.def: Add
16955 float_ml[as]_lane builtin generator macros.
16956 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt<mode>):
16958 (mul_lane<mode>3): This, and re-order arguments.
16959 (aarch64_float_mla_lane<mode>): Define.
16960 (aarch64_float_mls_lane<mode>): Define.
16961 * config/aarch64/arm_neon.h (vmla_lane_f32): Use RTL builtin
16962 instead of GCC vector extensions.
16963 (vmlaq_lane_f32): Likewise.
16964 (vmls_lane_f32): Likewise.
16965 (vmlsq_lane_f32): Likewise.
16967 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
16969 * config/aarch64/aarch64-simd-builtins.def: Add float_ml[as]
16970 builtin generator macros.
16971 * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>):
16973 (aarch64_float_mls<mode>): Define.
16974 * config/aarch64/arm_neon.h (vmla_f32): Use RTL builtin
16975 instead of relying on GCC vector extensions.
16976 (vmla_f64): Likewise.
16977 (vmlaq_f32): Likewise.
16978 (vmlaq_f64): Likewise.
16979 (vmls_f32): Likewise.
16980 (vmls_f64): Likewise.
16981 (vmlsq_f32): Likewise.
16982 (vmlsq_f64): Likewise.
16983 * config/aarch64/iterators.md: Define VDQF_DF mode iterator.
16985 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
16987 * config/aarch64/aarch64-simd-builtins.def: Add
16988 float_ml[as]_n_builtin generator macros.
16989 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt_from_dup<mode>):
16991 (mul_n<mode>3): This, and re-order arguments.
16992 (aarch64_float_mla_n<mode>): Define.
16993 (aarch64_float_mls_n<mode>): Define.
16994 * config/aarch64/arm_neon.h (vmla_n_f32): Use RTL builtin
16995 instead of inline asm.
16996 (vmlaq_n_f32): Likewise.
16997 (vmls_n_f32): Likewise.
16998 (vmlsq_n_f32): Likewise.
17000 2021-04-30 Jonathan Wright <joanthan.wright@arm.com>
17002 * config/aarch64/aarch64-simd-builtins.def: Add pmull[2]
17003 builtin generator macros.
17004 * config/aarch64/aarch64-simd.md (aarch64_pmullv8qi): Define.
17005 (aarch64_pmull_hiv16qi_insn): Define.
17006 (aarch64_pmull_hiv16qi): Define.
17007 * config/aarch64/arm_neon.h (vmull_high_p8): Use RTL builtin
17008 instead of inline asm.
17009 (vmull_p8): Likewise.
17011 2021-04-30 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
17013 * config/avr/avr.md: Adjust peepholes to match and
17014 generate parallels with clobber of REG_CC.
17015 (mov<mode>_insn): Rename to mov<mode>_insn_split.
17016 (*mov<mode>_insn): Rename to mov<mode>_insn.
17018 2021-04-30 David Edelsohn <dje.gcc@gmail.com>
17020 * varasm.c (use_blocks_for_decl_p): Don't use section anchors
17021 for VAR_DECLs if -fdata-sections enabled.
17023 2021-04-30 Michael Meissner <meissner@linux.ibm.com>
17025 PR bootstrap/100327
17026 * config/rs6000/rs6000.c
17027 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Define.
17028 (rs6000_libgcc_floating_mode_supported_p): New target hook.
17030 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
17032 * tree-ssa-threadbackward.c (class thread_jumps): Split out code
17034 (class back_threader_registry): ...to here...
17035 (class back_threader_profitability): ...and here...
17036 (thread_jumps::thread_through_all_blocks): Remove argument.
17037 (back_threader_registry::back_threader_registry): New.
17038 (back_threader_registry::~back_threader_registry): New.
17039 (back_threader_registry::thread_through_all_blocks): New.
17040 (thread_jumps::profitable_jump_thread_path): Move from here...
17041 (back_threader_profitability::profitable_path_p): ...to here.
17042 (thread_jumps::find_taken_edge): New.
17043 (thread_jumps::convert_and_register_current_path): Move...
17044 (back_threader_registry::register_path): ...to here.
17045 (thread_jumps::register_jump_thread_path_if_profitable): Move...
17046 (thread_jumps::maybe_register_path): ...to here.
17047 (thread_jumps::handle_phi): Call find_taken_edge and
17048 maybe_register_path.
17049 (thread_jumps::handle_assignment): Same.
17050 (thread_jumps::fsm_find_control_statement_thread_paths): Remove
17051 tree argument to handle_phi and handle_assignment.
17052 (thread_jumps::find_jump_threads_backwards): Set m_name. Remove
17053 set of m_speed_p and m_max_threaded_paths.
17054 (pass_thread_jumps::execute): Remove second argument from
17055 find_jump_threads_backwards.
17056 (pass_early_thread_jumps::execute): Same.
17058 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
17060 * tree-ssa-dom.c (class dom_jump_threader_simplifier): New.
17061 (class dom_opt_dom_walker): Initialize some class variables.
17062 (pass_dominator::execute): Pass evrp_range_analyzer and
17063 dom_jump_threader_simplifier to dom_opt_dom_walker.
17064 Adjust for some functions moving into classes.
17065 (simplify_stmt_for_jump_threading): Adjust and move to...
17066 (jump_threader_simplifier::simplify): ...here.
17067 (dom_opt_dom_walker::before_dom_children): Adjust for
17068 m_evrp_range_analyzer.
17069 (dom_opt_dom_walker::after_dom_children): Remove x_vr_values hack.
17070 (test_for_singularity): Place in dom_opt_dom_walker class.
17071 (dom_opt_dom_walker::optimize_stmt): The argument
17072 evrp_range_analyzer is now a class field.
17073 * tree-ssa-threadbackward.c (class thread_jumps): Add m_registry.
17074 (thread_jumps::thread_through_all_blocks): New.
17075 (thread_jumps::convert_and_register_current_path): Use m_registry.
17076 (pass_thread_jumps::execute): Adjust for thread_through_all_blocks
17077 being in the threader class.
17078 (pass_early_thread_jumps::execute): Same.
17079 * tree-ssa-threadedge.c (threadedge_initialize_values): Move...
17080 (jump_threader::jump_threader): ...here.
17081 (threadedge_finalize_values): Move...
17082 (jump_threader::~jump_threader): ...here.
17083 (jump_threader::remove_jump_threads_including): New.
17084 (jump_threader::thread_through_all_blocks): New.
17085 (record_temporary_equivalences_from_phis): Move...
17086 (jump_threader::record_temporary_equivalences_from_phis): ...here.
17087 (record_temporary_equivalences_from_stmts_at_dest): Move...
17088 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
17090 (simplify_control_stmt_condition_1): Move to jump_threader class.
17091 (simplify_control_stmt_condition): Move...
17092 (jump_threader::simplify_control_stmt_condition): ...here.
17093 (thread_around_empty_blocks): Move...
17094 (jump_threader::thread_around_empty_blocks): ...here.
17095 (thread_through_normal_block): Move...
17096 (jump_threader::thread_through_normal_block): ...here.
17097 (thread_across_edge): Move...
17098 (jump_threader::thread_across_edge): ...here.
17099 (thread_outgoing_edges): Move...
17100 (jump_threader::thread_outgoing_edges): ...here.
17101 * tree-ssa-threadedge.h: Move externally facing functings...
17102 (class jump_threader): ...here...
17103 (class jump_threader_simplifier): ...and here.
17104 * tree-ssa-threadupdate.c (struct redirection_data): Remove comment.
17105 (jump_thread_path_allocator::jump_thread_path_allocator): New.
17106 (jump_thread_path_allocator::~jump_thread_path_allocator): New.
17107 (jump_thread_path_allocator::allocate_thread_edge): New.
17108 (jump_thread_path_allocator::allocate_thread_path): New.
17109 (jump_thread_path_registry::jump_thread_path_registry): New.
17110 (jump_thread_path_registry::~jump_thread_path_registry): New.
17111 (jump_thread_path_registry::allocate_thread_edge): New.
17112 (jump_thread_path_registry::allocate_thread_path): New.
17113 (dump_jump_thread_path): Make extern.
17114 (debug (const vec<jump_thread_edge *> &path)): New.
17115 (struct removed_edges): Move to tree-ssa-threadupdate.h.
17116 (struct thread_stats_d): Remove.
17117 (remove_ctrl_stmt_and_useless_edges): Make static.
17118 (lookup_redirection_data): Move...
17119 (jump_thread_path_registry::lookup_redirection_data): ...here.
17120 (ssa_redirect_edges): Make static.
17121 (thread_block_1): Move...
17122 (jump_thread_path_registry::thread_block_1): ...here.
17123 (thread_block): Move...
17124 (jump_thread_path_registry::thread_block): ...here.
17125 (thread_through_loop_header): Move...
17126 (jump_thread_path_registry::thread_through_loop_header): ...here.
17127 (mark_threaded_blocks): Move...
17128 (jump_thread_path_registry::mark_threaded_blocks): ...here.
17129 (debug_path): Move...
17130 (jump_thread_path_registry::debug_path): ...here.
17131 (debug_all_paths): Move...
17132 (jump_thread_path_registry::dump): ..here.
17133 (rewire_first_differing_edge): Move...
17134 (jump_thread_path_registry::rewire_first_differing_edge): ...here.
17135 (adjust_paths_after_duplication): Move...
17136 (jump_thread_path_registry::adjust_paths_after_duplication): ...here.
17137 (duplicate_thread_path): Move...
17138 (jump_thread_path_registry::duplicate_thread_path): ..here.
17139 (remove_jump_threads_including): Move...
17140 (jump_thread_path_registry::remove_jump_threads_including): ...here.
17141 (thread_through_all_blocks): Move to...
17142 (jump_thread_path_registry::thread_through_all_blocks): ...here.
17143 (delete_jump_thread_path): Remove.
17144 (register_jump_thread): Move...
17145 (jump_thread_path_registry::register_jump_thread): ...here.
17146 * tree-ssa-threadupdate.h: Move externally facing functions...
17147 (class jump_thread_path_allocator): ...here...
17148 (class jump_thread_path_registry): ...and here.
17149 (thread_through_all_blocks): Remove.
17150 (struct removed_edges): New.
17151 (register_jump_thread): Remove.
17152 (remove_jump_threads_including): Remove.
17153 (delete_jump_thread_path): Remove.
17154 (remove_ctrl_stmt_and_useless_edges): Remove.
17155 (free_dom_edge_info): New prototype.
17156 * tree-vrp.c: Remove x_vr_values hack.
17157 (class vrp_jump_threader_simplifier): New.
17158 (vrp_jump_threader_simplifier::simplify): New.
17159 (vrp_jump_threader::vrp_jump_threader): Adjust method signature.
17160 Remove m_dummy_cond.
17161 Instantiate m_simplifier and m_threader.
17162 (vrp_jump_threader::thread_through_all_blocks): New.
17163 (vrp_jump_threader::simplify_stmt): Remove.
17164 (vrp_jump_threader::after_dom_children): Do not set m_dummy_cond.
17165 Remove x_vr_values hack.
17166 (execute_vrp): Adjust for thread_through_all_blocks being in a
17169 2021-04-30 Christophe Lyon <christophe.lyon@linaro.org>
17171 * genflags.c (gen_insn): Print failed expansion string.
17173 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
17175 * expr.c (alignment_for_piecewise_move): Call mode_for_size
17176 without limit to MAX_FIXED_MODE_SIZE.
17178 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
17180 PR middle-end/90773
17181 * builtins.c (builtin_memset_gen_str): Don't use return from
17182 simplify_gen_subreg.
17184 2021-04-30 Uroš Bizjak <ubizjak@gmail.com>
17187 * config/i386/i386.md (*add<mode>3_carry_0r): New insn pattern.
17188 (*addsi3_carry_zext_0r): Ditto.
17189 (*sub<mode>3_carry_0): Ditto.
17190 (*subsi3_carry_zext_0r): Ditto.
17191 * config/i386/predicates.md (ix86_carry_flag_unset_operator):
17193 * config/i386/i386.c (ix86_rtx_costs) <case PLUS, case MINUS>:
17194 Also consider ix86_carry_flag_unset_operator to calculate
17195 the cost of adc/sbb insn.
17197 2021-04-30 Roman Zhuykov <zhroma@ispras.ru>
17199 PR rtl-optimization/100225
17200 PR rtl-optimization/84878
17201 * modulo-sched.c (sms_schedule): Use note_stores to skip loops
17202 where we have an instruction which touches (writes) any hard
17203 register from df->regular_block_artificial_uses set.
17204 Allow not-single-set instruction only right before basic block
17207 2021-04-30 Geng Qi <gengqi@linux.alibaba.com>
17209 * config/riscv/riscv.opt (march=,mabi=): Negative itself.
17211 2021-04-30 LevyHsu <admin@levyhsu.com>
17213 * config/riscv/riscv.c (riscv_min_arithmetic_precision): New.
17214 * config/riscv/riscv.h (TARGET_MIN_ARITHMETIC_PRECISION): New.
17215 * config/riscv/riscv.md (addv<mode>4, uaddv<mode>4): New.
17216 (subv<mode>4, usubv<mode>4, mulv<mode>4, umulv<mode>4): New.
17218 2021-04-29 Alexandre Oliva <oliva@adacore.com>
17220 * config.gcc: Merged x86 and x86_64 cpu_type-setting cases.
17222 2021-04-29 Alexandre Oliva <oliva@adacore.com>
17224 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Rename to...
17225 (ASM_OUTPUT_MAX_SKIP_ALIGN): ... this. Enclose in do/while(0).
17226 * config/i386/i386.c: Adjust.
17227 * config/i386/i386.md: Adjust.
17228 * config/i386/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Drop.
17229 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17230 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17231 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17232 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17233 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17234 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17235 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17236 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17237 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17238 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
17240 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
17242 * config/i386/i386-expand.c (ix86_expand_int_compare):
17243 Swap operands of GTU and LEU comparison to emit carry flag comparison.
17244 * config/i386/i386.md (*add<mode>3_carry_0): Change insn
17245 predicate to allow more combine opportunities with memory operands.
17246 (*sub<mode>3_carry_0): Ditto.
17248 2021-04-29 Richard Sandiford <richard.sandiford@arm.com>
17250 PR rtl-optimization/100303
17251 * rtl-ssa/accesses.cc (function_info::make_use_available): Take a
17252 boolean that indicates whether the use will only be used in
17253 debug instructions. Treat it in the same way that existing
17254 cross-EBB debug references would be handled if so.
17255 (function_info::make_uses_available): Likewise.
17256 * rtl-ssa/functions.h (function_info::make_uses_available): Update
17257 prototype accordingly.
17258 (function_info::make_uses_available): Likewise.
17259 * fwprop.c (try_fwprop_subst): Update call accordingly.
17261 2021-04-29 Jeff Law <jlaw@tachyum.com>
17263 * config/nios2/nios2-protos.h (nios2_fpu_insn_enabled): Move outside
17266 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
17267 Richard Biener <rguenther@suse.de>
17270 * config/i386/i386-builtin.def (IX86_BUILTIN_MASKLOADPD)
17271 (IX86_BUILTIN_MASKLOADPS, IX86_BUILTIN_MASKLOADPD256)
17272 (IX86_BUILTIN_MASKLOADPS256, IX86_BUILTIN_MASKLOADD)
17273 (IX86_BUILTIN_MASKLOADQ, IX86_BUILTIN_MASKLOADD256)
17274 (IX86_BUILTIN_MASKLOADQ256): Move from SPECIAL_ARGS
17275 to PURE_ARGS category.
17276 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
17277 Handle PURE_ARGS category.
17278 * config/i386/i386-expand.c (ix86_expand_builtin): Ditto.
17280 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
17282 * configure.ac: Check for the presence of sys/locking.h header and
17283 for whether _LK_LOCK is supported by _locking.
17284 * configure: Regenerate.
17285 * config.in: Likewise.
17286 * gcov-io.h: Define GCOV_LOCKED_WITH_LOCKING if HOST_HAS_LK_LOCK.
17287 * gcov-io.c (gcov_open): Add support for GCOV_LOCKED_WITH_LOCKING.
17288 * system.h: Include <sys/locking.h> if HAVE_SYS_LOCKING_H.
17290 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
17292 * config/i386/predicates.md (fcmov_comparison_operator):
17293 Do not check for trivial FP comparison operator.
17294 <case GEU, case LTU>: Allow CCGZmode.
17295 <case GTU, case LEU>: Do not allow CCCmode.
17296 (ix86_comparison_operator) <case GTU, case LEU>: Allow only CCmode.
17297 (ix86_carry_flag_operator): Match only LTU and UNLT code.
17298 Do not check for trivial FP comparison operator. Allow CCGZmode.
17300 2021-04-29 Tom de Vries <tdevries@suse.de>
17302 * omp-expand.c (expand_omp_simd): Add step_orig, and replace uses of
17303 fd->loop.step by either step or orig_step.
17305 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
17307 * config/sparc/sparc.c (gen_load_pcrel_sym): Delete.
17308 (load_got_register): Do the PIC dance here.
17309 (sparc_legitimize_tls_address): Simplify.
17310 (sparc_emit_probe_stack_range): Likewise.
17311 (sparc32_initialize_trampoline): Likewise.
17312 (sparc64_initialize_trampoline): Likewise.
17313 * config/sparc/sparc.md (load_pcrel_sym<P:mode>): Add @ marker.
17314 (probe_stack_range<P:mode>): Likewise.
17315 (flush<P:mode>): Likewise.
17316 (tgd_hi22<P:mode>): Likewise.
17317 (tgd_lo10<P:mode>): Likewise.
17318 (tgd_add<P:mode>): Likewise.
17319 (tgd_call<P:mode>): Likewise.
17320 (tldm_hi22<P:mode>): Likewise.
17321 (tldm_lo10<P:mode>): Likewise.
17322 (tldm_add<P:mode>): Likewise.
17323 (tldm_call<P:mode>): Likewise.
17324 (tldo_hix22<P:mode>): Likewise.
17325 (tldo_lox10<P:mode>): Likewise.
17326 (tldo_add<P:mode>): Likewise.
17327 (tie_hi22<P:mode>): Likewise.
17328 (tie_lo10<P:mode>): Likewise.
17329 (tie_add<P:mode>): Likewise.
17330 (tle_hix22<P:mode>): Likewise.
17331 (tle_lox10<P:mode>): Likewise.
17332 (stack_protect_setsi): Rename to...
17333 (stack_protect_set32): ...this.
17334 (stack_protect_setdi): Rename to...
17335 (stack_protect_set64): ...this.
17336 (stack_protect_set): Adjust calls to above.
17337 (stack_protect_testsi): Rename to...
17338 (stack_protect_test32): ...this.
17339 (stack_protect_testdi): Rename to...
17340 (stack_protect_test64): ...this.
17341 (stack_protect_test): Adjust calls to above.
17343 2021-04-29 H.J. Lu <hjl.tools@gmail.com>
17345 PR middle-end/90773
17346 * builtins.c (builtin_memcpy_read_str): Add a dummy argument.
17347 (builtin_strncpy_read_str): Likewise.
17348 (builtin_memset_read_str): Add an argument for the previous RTL
17349 information and generate the new RTL from the previous RTL info.
17350 (builtin_memset_gen_str): Likewise.
17351 * builtins.h (builtin_strncpy_read_str): Update the prototype.
17352 (builtin_memset_read_str): Likewise.
17353 * expr.c (by_pieces_ninsns): If targetm.overlap_op_by_pieces_p()
17354 returns true, round up size and alignment to the widest integer
17355 mode for maximum size.
17356 (pieces_addr::adjust): Add a pointer to by_pieces_prev argument
17357 and pass it to m_constfn.
17358 (op_by_pieces_d): Add m_push and m_overlap_op_by_pieces.
17359 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
17360 initialize m_push. Initialize m_overlap_op_by_pieces with
17361 targetm.overlap_op_by_pieces_p ().
17362 (op_by_pieces_d::run): Pass the previous RTL information to
17363 pieces_addr::adjust and generate overlapping operations if
17364 m_overlap_op_by_pieces is true.
17366 (move_by_pieces_d::move_by_pieces_d): Updated for op_by_pieces_d
17368 (store_by_pieces_d::store_by_pieces_d): Updated for op_by_pieces_d
17370 (can_store_by_pieces): Use by_pieces_constfn on constfun.
17371 (store_by_pieces): Use by_pieces_constfn on constfun. Updated
17372 for op_by_pieces_d change.
17373 (clear_by_pieces_1): Add a dummy argument.
17374 (clear_by_pieces): Updated for op_by_pieces_d change.
17375 (compare_by_pieces_d::compare_by_pieces_d): Likewise.
17376 (string_cst_read_str): Add a dummy argument.
17377 * expr.h (by_pieces_constfn): Add a dummy argument.
17378 (by_pieces_prev): New.
17379 * target.def (overlap_op_by_pieces_p): New target hook.
17380 * config/i386/i386.c (TARGET_OVERLAP_OP_BY_PIECES_P): New.
17381 * doc/tm.texi.in: Add TARGET_OVERLAP_OP_BY_PIECES_P.
17382 * doc/tm.texi: Regenerated.
17384 2021-04-29 Richard Biener <rguenther@suse.de>
17386 PR tree-optimization/100253
17387 * tree-vect-stmts.c (vectorizable_load): Do not assume
17388 element alignment when DR_MISALIGNMENT is -1.
17389 (vectorizable_store): Likewise.
17391 2021-04-29 Jakub Jelinek <jakub@redhat.com>
17394 * config/aarch64/aarch64.c (aarch64_add_offset_1_temporaries): Use
17395 absu_hwi instead of abs_hwi.
17397 2021-04-29 Richard Biener <rguenther@suse.de>
17399 PR middle-end/38474
17400 * tree-ssa-structalias.c (add_graph_edge): Avoid direct
17401 forwarding when indirect forwarding through ESCAPED
17404 2021-04-29 Tom de Vries <tdevries@suse.de>
17407 * internal-fn.c (expand_GOMP_SIMT_ENTER_ALLOC)
17408 (expand_GOMP_SIMT_LAST_LANE, expand_GOMP_SIMT_ORDERED_PRED)
17409 (expand_GOMP_SIMT_VOTE_ANY, expand_GOMP_SIMT_XCHG_BFLY)
17410 (expand_GOMP_SIMT_XCHG_IDX): Ensure target is assigned to.
17412 2021-04-29 Richard Biener <rguenther@suse.de>
17414 PR tree-optimization/99912
17415 * tree-ssa-dse.c (dse_dom_walker::m_need_cfg_cleanup): New.
17416 (dse_dom_walker::todo): Likewise.
17417 (dse_dom_walker::dse_optimize_stmt): Move VDEF check to the
17419 (dse_dom_walker::before_dom_children): Remove trivially
17420 dead SSA defs and schedule CFG cleanup if we removed all
17422 (pass_dse::execute): Get TODO as computed by the DOM walker
17423 and return it. Wipe dominator info earlier.
17425 2021-04-29 Richard Biener <rguenther@suse.de>
17428 * ipa-prop.c (ipcp_modif_dom_walker::before_dom_children):
17429 Track blocks to cleanup EH in new m_need_eh_cleanup.
17430 (ipcp_modif_dom_walker::cleanup_eh): New.
17431 (ipcp_transform_function): Release dominator info before
17434 2021-04-29 Martin Sebor <msebor@redhat.com>
17436 PR middle-end/100250
17437 * attribs.c (attr_access::array_as_string): Avoid dereferencing
17438 a pointer when it's null.
17440 2021-04-29 Martin Sebor <msebor@redhat.com>
17442 * Makefile.in (OBJS): Add ipa-free-lang-data.o.
17443 * ipa-free-lang-data.cc: New file.
17444 * tree.c: Move pass free_lang_data to file above.
17445 (build_array_type_1): Declare extern.
17446 * tree.h (build_array_type_1): Declare.
17448 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17450 * config/aarch64/aarch64-simd-builtins.def: Modify comment to
17451 make consistent with updated RTL pattern.
17452 * config/aarch64/aarch64-simd.md (aarch64_<sur>qmovn<mode>):
17453 Implement using ss_truncate and us_truncate rather than
17455 * config/aarch64/iterators.md: Remove redundant unspecs and
17456 iterator: UNSPEC_[SU]QXTN and SUQMOVN respectively.
17458 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17460 * config/aarch64/arm_acle.h (__attribute__): Make intrinsic
17461 attributes consistent with those defined in arm_neon.h.
17463 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17465 * config/aarch64/arm_fp16.h (__attribute__): Make intrinsic
17466 attributes consistent with those defined in arm_neon.h.
17468 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17470 * config/aarch64/aarch64-simd-builtins.def: Add
17471 float_trunc_rodd builtin generator macros.
17472 * config/aarch64/aarch64-simd.md (aarch64_float_trunc_rodd_df):
17474 (aarch64_float_trunc_rodd_lo_v2sf): Define.
17475 (aarch64_float_trunc_rodd_hi_v4sf_le): Define.
17476 (aarch64_float_trunc_rodd_hi_v4sf_be): Define.
17477 (aarch64_float_trunc_rodd_hi_v4sf): Define.
17478 * config/aarch64/arm_neon.h (vcvtx_f32_f64): Use RTL builtin
17479 instead of inline asm.
17480 (vcvtx_high_f32_f64): Likewise.
17481 (vcvtxd_f32_f64): Likewise.
17482 * config/aarch64/iterators.md: Add FCVTXN unspec.
17484 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17486 * config/aarch64/aarch64-simd-builtins.def: Add tbx1 builtin
17488 * config/aarch64/aarch64-simd.md (aarch64_tbx1<mode>):
17490 * config/aarch64/arm_neon.h (vqtbx1_s8): USE RTL builtin
17491 instead of inline asm.
17492 (vqtbx1_u8): Likewise.
17493 (vqtbx1_p8): Likewise.
17494 (vqtbx1q_s8): Likewise.
17495 (vqtbx1q_u8): Likewise.
17496 (vqtbx1q_p8): Likewise.
17497 (vtbx2_s8): Likewise.
17498 (vtbx2_u8): Likewise.
17499 (vtbx2_p8): Likewise.
17501 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17503 * config/aarch64/aarch64-simd-builtins.def: Add tbl1 builtin
17505 * config/aarch64/arm_neon.h (vqtbl1_p8): Use RTL builtin
17506 instead of inline asm.
17507 (vqtbl1_s8): Likewise.
17508 (vqtbl1_u8): Likewise.
17509 (vqtbl1q_p8): Likewise.
17510 (vqtbl1q_s8): Likewise.
17511 (vqtbl1q_u8): Likewise.
17512 (vtbl1_s8): Likewise.
17513 (vtbl1_u8): Likewise.
17514 (vtbl1_p8): Likewise.
17515 (vtbl2_s8): Likewise.
17516 (vtbl2_u8): Likewise.
17517 (vtbl2_p8): Likewise.
17519 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17521 * config/aarch64/aarch64-simd-builtins.def: Add polynomial
17522 ssri_n buitin generator macro.
17523 * config/aarch64/arm_neon.h (vsri_n_p8): Use RTL builtin
17524 instead of inline asm.
17525 (vsri_n_p16): Likewise.
17526 (vsri_n_p64): Likewise.
17527 (vsriq_n_p8): Likewise.
17528 (vsriq_n_p16): Likewise.
17529 (vsriq_n_p64): Likewise.
17531 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17533 * config/aarch64/aarch64-simd-builtins.def: Use VALLP mode
17534 iterator for polynomial ssli_n builtin generator macro.
17535 * config/aarch64/arm_neon.h (vsli_n_p8): Use RTL builtin
17536 instead of inline asm.
17537 (vsli_n_p16): Likewise.
17538 (vsliq_n_p8): Likewise.
17539 (vsliq_n_p16): Likewise.
17540 * config/aarch64/iterators.md: Define VALLP mode iterator.
17542 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17544 * config/aarch64/aarch64-simd-builtins.def: Use VDQV_L
17545 iterator to generate [su]adalp RTL builtins.
17546 * config/aarch64/aarch64-simd.md: Use VDQV_L iterator in
17547 [su]adalp RTL pattern.
17548 * config/aarch64/arm_neon.h (vpadal_s32): Use RTL builtin
17549 instead of inline asm.
17550 (vpadal_u32): Likewise.
17552 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17554 * config/aarch64/aarch64-simd-builtins.def: Add [su]addlp
17555 builtin generator macros.
17556 * config/aarch64/aarch64-simd.md (aarch64_<su>addlp<mode>):
17558 * config/aarch64/arm_neon.h (vpaddl_s8): Use RTL builtin
17559 instead of inline asm.
17560 (vpaddl_s16): Likewise.
17561 (vpaddl_s32): Likewise.
17562 (vpaddl_u8): Likewise.
17563 (vpaddl_u16): Likewise.
17564 (vpaddl_u32): Likewise.
17565 (vpaddlq_s8): Likewise.
17566 (vpaddlq_s16): Likewise.
17567 (vpaddlq_s32): Likewise.
17568 (vpaddlq_u8): Likewise.
17569 (vpaddlq_u16): Likewise.
17570 (vpaddlq_u32): Liwewise.
17571 * config/aarch64/iterators.md: Define [SU]ADDLP unspecs with
17572 appropriate attributes.
17574 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17576 * config/aarch64/aarch64-simd-builtins.def: Use VDQ_I iterator
17577 for aarch64_addp<mode> builtin macro generator.
17578 * config/aarch64/aarch64-simd.md: Use VDQ_I iterator in
17579 aarch64_addp<mode> RTL pattern.
17580 * config/aarch64/arm_neon.h (vpaddq_s8): Use RTL builtin
17581 instead of inline asm.
17582 (vpaddq_s16): Likewise.
17583 (vpaddq_s32): Likewise.
17584 (vpaddq_s64): Likewise.
17585 (vpaddq_u8): Likewise.
17586 (vpaddq_u16): Likewise.
17587 (vpaddq_u32): Likewise.
17588 (vpaddq_u64): Likewise.
17590 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17592 * config/aarch64/aarch64-simd-builtins.def: Add sq[r]dmulh_n
17593 builtin generator macros.
17594 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_n<mode>):
17596 * config/aarch64/arm_neon.h (vqdmulh_n_s16): Use RTL builtin
17597 instead of inline asm.
17598 (vqdmulh_n_s32): Likewise.
17599 (vqdmulhq_n_s16): Likewise.
17600 (vqdmulhq_n_s32): Likewise.
17601 (vqrdmulh_n_s16): Likewise.
17602 (vqrdmulh_n_s32): Likewise.
17603 (vqrdmulhq_n_s16): Likewise.
17604 (vqrdmulhq_n_s32): Likewise.
17606 2021-04-28 Tobias Burnus <tobias@codesourcery.com>
17608 * doc/install.texi (--enable-offload-defaulted): Document.
17610 2021-04-28 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
17612 * config/avr/avr-dimode.md: Turn existing patterns into
17613 define_insn_and_split style patterns where the splitter
17614 adds a clobber of the condition code register. Drop "cc"
17615 attribute. Add new patterns to match output of
17617 * config/avr/avr-fixed.md: Likewise.
17618 * config/avr/avr.c (cc_reg_rtx): New.
17619 (avr_parallel_insn_from_insns): Adjust insn count
17620 for removal of set of cc0.
17621 (avr_is_casesi_sequence): Likewise.
17622 (avr_casei_sequence_check_operands): Likewise.
17623 (avr_optimize_casesi): Likewise. Also insert
17624 new insns after jump_insn.
17625 (avr_pass_casesi::avr_rest_of_handle_casesi): Adjust
17626 for removal of set of cc0.
17627 (avr_init_expanders): Initialize cc_reg_rtx.
17628 (avr_regno_reg_class): Handle REG_CC.
17629 (cond_string): Remove usage of CC_OVERFLOW_UNUSABLE.
17630 (avr_notice_update_cc): Remove function.
17631 (ret_cond_branch): Remove usage of CC_OVERFLOW_UNUSABLE.
17632 (compare_condition): Adjust for PARALLEL with
17634 (out_shift_with_cnt): Likewise.
17635 (ashlhi3_out): Likewise.
17636 (ashrhi3_out): Likewise.
17637 (lshrhi3_out): Likewise.
17638 (avr_class_max_nregs): Return single reg for REG_CC.
17639 (avr_compare_pattern): Check for REG_CC instead
17641 (avr_reorg_remove_redundant_compare): Likewise.
17642 (avr_reorg):Adjust for PARALLEL with REG_CC clobber.
17643 (avr_hard_regno_nregs): Return single reg for REG_CC.
17644 (avr_hard_regno_mode_ok): Allow only CCmode for REG_CC.
17645 (avr_md_asm_adjust): Clobber REG_CC.
17646 (TARGET_HARD_REGNO_NREGS): Define.
17647 (TARGET_CLASS_MAX_NREGS): Define.
17648 (TARGET_MD_ASM_ADJUST): Define.
17649 * config/avr/avr.h (FIRST_PSEUDO_REGISTER): Adjust
17651 (enum reg_class): Add CC_REG class.
17652 (NOTICE_UPDATE_CC): Remove.
17653 (CC_OVERFLOW_UNUSABLE): Remove.
17654 (CC_NO_CARRY): Remove.
17655 * config/avr/avr.md: Turn existing patterns into
17656 define_insn_and_split style patterns where the splitter
17657 adds a clobber of the condition code register. Drop "cc"
17658 attribute. Add new patterns to match output of
17660 (sez): Remove unused pattern.
17662 2021-04-28 Richard Earnshaw <rearnsha@arm.com>
17665 * config/arm/arm.c (arm_hard_regno_mode_ok): Only allow VPR to be
17668 2021-04-28 Richard Sandiford <richard.sandiford@arm.com>
17671 * config/aarch64/constraints.md (Utq): Require the address to
17672 be valid for both the element mode and for V2DImode.
17674 2021-04-28 Jakub Jelinek <jakub@redhat.com>
17675 Tobias Burnus <tobias@codesourcery.com>
17677 * configure.ac (OFFLOAD_DEFAULTED): AC_DEFINE if offload-defaulted.
17678 * gcc.c (process_command): New variable.
17679 (driver::maybe_putenv_OFFLOAD_TARGETS): If OFFLOAD_DEFAULTED,
17680 set it if -foffload is defaulted.
17681 * lto-wrapper.c (OFFLOAD_TARGET_DEFAULT_ENV): Define.
17682 (compile_offload_image): If OFFLOAD_DEFAULTED and
17683 OFFLOAD_TARGET_DEFAULT is in the environment, don't fail
17684 if corresponding mkoffload can't be found.
17685 (compile_images_for_offload_targets): Likewise. Free and clear
17686 offload_names if no valid offload is found.
17687 * config.in: Regenerate.
17688 * configure: Regenerate.
17690 2021-04-28 Richard Biener <rguenther@suse.de>
17692 PR tree-optimization/100292
17693 * tree-vect-generic.c (expand_vector_condition): Do not fold
17696 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
17698 * config/rs6000/aix.h (SUBTARGET_DRIVER_SELF_SPECS): New.
17699 * config/rs6000/aix64.opt (m64): New.
17702 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
17704 * config/vax/vax.c (print_operand_address, vax_address_cost_1)
17705 (index_term_p): Handle ASHIFT too.
17707 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
17709 * config/vax/builtins.md (jbb<ccss>i<mode>): Remove operand #3.
17710 (sync_lock_test_and_set<mode>): Adjust accordingly.
17711 (sync_lock_release<mode>): Likewise.
17713 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
17715 * config/vax/vax-protos.h (adjacent_operands_p): Remove
17717 * config/vax/vax.c (adjacent_operands_p): Remove.
17719 2021-04-27 Maciej W. Rozycki <macro@linux-mips.org>
17721 * ifcvt.c (dead_or_predicable) [!IFCVT_MODIFY_TESTS]: Fall
17722 through to the non-conditional execution case if getting the
17723 condition for conditional execution has failed.
17725 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
17727 PR middle-end/100284
17728 * gimple.c (gimple_could_trap_p_1): Remove VEC_COND_EXPR test.
17729 * tree-eh.c (operation_could_trap_p): Handle VEC_COND_EXPR rather
17730 than asserting on it.
17732 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
17734 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): Protect
17735 with TARGET_AIX_OS.
17737 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
17740 * calls.c (precompute_register_parameters): Additionally test
17741 targetm.precompute_tls_p to pre-compute argument.
17742 * config/rs6000/aix.h (TARGET_PRECOMPUTE_TLS_P): Define.
17743 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): New.
17744 * target.def (precompute_tls_p): New.
17745 * doc/tm.texi.in (TARGET_PRECOMPUTE_TLS_P): Add hook documentation.
17746 * doc/tm.texi: Regenerated.
17748 2021-04-27 Jakub Jelinek <jakub@redhat.com>
17751 * config/aarch64/aarch64.c (aarch64_print_operand): Cast -UINTVAL
17752 back to HOST_WIDE_INT.
17754 2021-04-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
17757 * simplify-rtx.c (simplify_context::simplify_subreg): Check the
17758 memory alignment for the outer mode.
17760 2021-04-27 H.J. Lu <hjl.tools@gmail.com>
17762 PR middle-end/90773
17763 * expr.c (op_by_pieces_d::get_usable_mode): New member function.
17764 (op_by_pieces_d::run): Cange a while loop to a do-while loop.
17766 2021-04-27 Alex Coplan <alex.coplan@arm.com>
17769 * config/arm/arm.c (arm_split_compare_and_swap): Fix up codegen
17770 with negative immediates: ensure we expand cbranchsi4_scratch
17771 correctly and ensure we satisfy its constraints.
17772 * config/arm/sync.md
17773 (@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): Don't
17774 attempt to tie two output operands together with constraints;
17775 collapse two alternatives.
17776 (@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
17777 * config/arm/thumb1.md (cbranchsi4_neg_late): New.
17779 2021-04-27 Jakub Jelinek <jakub@redhat.com>
17782 * config/aarch64/predicates.md (aarch64_sub_immediate,
17783 aarch64_plus_immediate): Use -UINTVAL instead of -INTVAL.
17784 * config/aarch64/aarch64.md (casesi, rotl<mode>3): Likewise.
17785 * config/aarch64/aarch64.c (aarch64_print_operand,
17786 aarch64_split_atomic_op, aarch64_expand_subvti): Likewise.
17788 2021-04-27 Jakub Jelinek <jakub@redhat.com>
17790 PR tree-optimization/100239
17791 * tree-vect-generic.c (lower_vec_perm): Don't accept constant
17792 permutations with all indices from the first zero element as vec_shl.
17794 2021-04-27 Jakub Jelinek <jakub@redhat.com>
17796 PR rtl-optimization/100254
17797 * cfgcleanup.c (outgoing_edges_match): Check REG_EH_REGION on
17798 last1 and last2 insns rather than BB_END (bb1) and BB_END (bb2) insns.
17800 2021-04-27 Richard Biener <rguenther@suse.de>
17802 PR tree-optimization/99912
17803 * passes.def: Add comment about new TODO_remove_unused_locals.
17804 * tree-stdarg.c (pass_data_stdarg): Run TODO_remove_unused_locals
17807 2021-04-27 Richard Biener <rguenther@suse.de>
17809 PR tree-optimization/99912
17810 * passes.def (pass_all_optimizations): Add pass_dse before
17811 the first pass_dce, move the first pass_dse before the
17812 pass_dce following pass_pre.
17814 2021-04-27 Jakub Jelinek <jakub@redhat.com>
17816 PR tree-optimization/95527
17817 * generic-match-head.c: Include tm.h.
17818 * gimple-match-head.c: Include tm.h.
17819 * match.pd (CLZ == INTEGER_CST): Don't use
17820 #ifdef CLZ_DEFINED_VALUE_AT_ZERO, only test CLZ_DEFINED_VALUE_AT_ZERO
17821 if clz == CFN_CLZ. Add missing val declaration.
17822 (CTZ cmp CST): New simplifications.
17824 2021-04-27 Jakub Jelinek <jakub@redhat.com>
17826 PR tree-optimization/96696
17827 * expr.c (expand_expr_divmod): New function.
17828 (expand_expr_real_2) <case TRUNC_DIV_EXPR>: Use it for truncations and
17829 divisions. Formatting fixes.
17830 <case MULT_EXPR>: Optimize x / y * y as x - x % y if the latter is
17833 2021-04-27 Martin Jambor <mjambor@suse.cz>
17836 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
17837 If removing a call statement LHS SSA name, release it.
17839 2021-04-27 Richard Earnshaw <rearnsha@arm.com>
17842 * config/arm/arm.c (THUMB2_WORK_REGS): Check PIC_OFFSET_TABLE_REGNUM
17843 is valid before including it in the mask.
17845 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
17848 * config/aarch64/aarch64.c (aarch64_comp_type_attributes): Handle
17851 2021-04-27 Richard Biener <rguenther@suse.de>
17853 PR tree-optimization/100051
17854 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Add
17855 disambiguator based on access size vs. decl size.
17857 2021-04-27 Richard Biener <rguenther@suse.de>
17859 PR tree-optimization/100278
17860 * tree-ssa-pre.c (compute_avail): Give up when we cannot
17861 adjust TBAA beacuse of mismatching bases.
17863 2021-04-27 Jakub Jelinek <jakub@redhat.com>
17866 * config/i386/i386.md (*<insn><mode>3_mask, *<insn><mode>3_mask_1):
17867 For any_rotate define_insn_split and following splitters, use
17868 SWI iterator instead of SWI48.
17870 2021-04-27 Richard Biener <rguenther@suse.de>
17872 PR tree-optimization/99776
17873 * match.pd (bit_field_ref (ctor)): Relax element extract
17874 type compatibility checks.
17876 2021-04-27 Cui,Lili <lili.cui@intel.com>
17878 * common/config/i386/i386-common.c (processor_names):
17879 Sync processor_names with processor_type.
17880 * config/i386/i386-options.c (processor_cost_table):
17881 Sync processor_cost_table with processor_type.
17883 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
17885 * value-range.cc (irange::irange_set_1bit_anti_range): Add assert.
17886 (irange::set): Call irange_set_1bit_anti_range for handling all
17887 1-bit ranges. Fall through on ~[MIN,MAX].
17889 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
17891 * value-range.cc (irange::legacy_num_pairs): Remove.
17892 (irange::invert): Change gcc_assert to gcc_checking_assert.
17893 * value-range.h (irange::num_pairs): Adjust for a cached
17894 num_pairs(). Also, rename all gcc_assert's to
17895 gcc_checking_assert's.
17897 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
17899 * value-range.cc (irange::operator=): Set m_kind.
17900 (irange::copy_to_legacy): Handle varying and undefined sources
17901 as a legacy copy since they can be easily copied.
17902 (irange::irange_set): Set m_kind.
17903 (irange::irange_set_anti_range): Same.
17904 (irange::set): Rename normalize_min_max to normalize_kind.
17905 (irange::verify_range): Adjust for multi-ranges having the
17907 (irange::irange_union): Set m_kind.
17908 (irange::irange_intersect): Same.
17909 (irange::invert): Same.
17910 * value-range.h (irange::kind): Always return m_kind.
17911 (irange::varying_p): Rename to...
17912 (irange::varying_comptaible_p): ...this.
17913 (irange::undefined_p): Only look at m_kind.
17914 (irange::irange): Always set VR_UNDEFINED if applicable.
17915 (irange::set_undefined): Always set VR_UNDEFINED.
17916 (irange::set_varying): Always set m_kind to VR_VARYING.
17917 (irange::normalize_min_max): Rename to...
17918 (irange::normalize_kind): ...this.
17920 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
17922 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
17923 Adjust for constant_p including varying_p.
17924 * tree-vrp.c (vrp_prop::finalize): Same.
17925 (determine_value_range): Same.
17926 * vr-values.c (vr_values::range_of_expr): Same.
17927 * value-range.cc (irange::symbolic_p): Do not check varying_p.
17928 (irange::constant_p): Same.
17930 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
17932 * value-range.cc (irange::legacy_lower_bound): Replace
17933 !undefined_p check with num_ranges > 0.
17934 (irange::legacy_upper_bound): Same.
17935 * value-range.h (irange::type): Same.
17936 (irange::lower_bound): Same.
17937 (irange::upper_bound): Same.
17939 2021-04-26 Richard Biener <rguenther@suse.de>
17941 PR tree-optimization/99956
17942 * gimple-loop-interchange.cc (compute_access_stride):
17943 Try instantiating the access in a shallower loop nest
17944 if instantiating failed.
17945 (compute_access_strides): Pass adjustable loop_nest
17946 to compute_access_stride.
17948 2021-04-26 Christophe Lyon <christophe.lyon@linaro.org>
17950 * doc/sourcebuild.texi (arm_cmse_hw): Document.
17952 2021-04-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17954 * config/aarch64/iterators.md (vwcore): Handle V4BF, V8BF.
17956 2021-04-26 Thomas Schwinge <thomas@codesourcery.com>
17957 Nathan Sidwell <nathan@codesourcery.com>
17958 Tom de Vries <vries@codesourcery.com>
17959 Julian Brown <julian@codesourcery.com>
17960 Kwok Cheung Yeung <kcy@codesourcery.com>
17962 * omp-offload.c (oacc_validate_dims): Implement
17963 '-Wopenacc-parallelism'.
17964 * doc/invoke.texi (-Wopenacc-parallelism): Document.
17966 2021-04-26 Richard Biener <rguenther@suse.de>
17968 * tree-cfg.h (gimplify_build1): Remove.
17969 (gimplify_build2): Likewise.
17970 (gimplify_build3): Likewise.
17971 * tree-cfg.c (gimplify_build1): Move to tree-vect-generic.c.
17972 (gimplify_build2): Likewise.
17973 (gimplify_build3): Likewise.
17974 * tree-vect-generic.c (gimplify_build1): Move from tree-cfg.c.
17976 (gimplify_build2): Likewise.
17977 (gimplify_build3): Likewise.
17978 (tree_vec_extract): Use resimplify with following SSA edges.
17979 (expand_vector_parallel): Avoid passing NULL size/bitpos
17980 to tree_vec_extract.
17981 * expr.c (store_constructor): Deal with zero-element CTORs.
17982 * match.pd (bit_field_ref <vector CTOR>): Make sure to
17983 produce vector constants when possible.
17985 2021-04-26 Richard Biener <rguenther@suse.de>
17987 * tree-complex.c: Include gimple-fold.h.
17988 (expand_complex_addition): Use gimple_build.
17989 (expand_complex_multiplication_components): Likewise.
17990 (expand_complex_multiplication): Likewise.
17991 (expand_complex_div_straight): Likewise.
17992 (expand_complex_div_wide): Likewise.
17993 (expand_complex_division): Likewise.
17994 (expand_complex_conjugate): Likewise.
17995 (expand_complex_comparison): Likewise.
17997 2021-04-26 Richard Biener <rguenther@suse.de>
17999 * tree-ssa-phiopt.c (two_value_replacement): Remove use
18000 of legacy gimplify_buildN API.
18002 2021-04-26 Richard Biener <rguenther@suse.de>
18004 PR tree-optimization/99473
18005 * tree-ssa-phiopt.c (cond_store_replacement): Handle all
18008 2021-04-26 Richard Biener <rguenther@suse.de>
18010 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin):
18011 Use replace_call_with_value.
18013 2021-04-26 Richard Biener <rguenther@suse.de>
18015 * tree-ssa-propagate.h (valid_gimple_rhs_p): Remove.
18016 (update_gimple_call): Likewise.
18017 (update_call_from_tree): Likewise.
18018 * tree-ssa-propagate.c (valid_gimple_rhs_p): Remove.
18019 (valid_gimple_call_p): Likewise.
18020 (move_ssa_defining_stmt_for_defs): Likewise.
18021 (finish_update_gimple_call): Likewise.
18022 (update_gimple_call): Likewise.
18023 (update_call_from_tree): Likewise.
18024 (propagate_tree_value_into_stmt): Use replace_call_with_value.
18025 * gimple-fold.h (update_gimple_call): Declare.
18026 * gimple-fold.c (valid_gimple_rhs_p): Move here from
18027 tree-ssa-propagate.c.
18028 (update_gimple_call): Likewise.
18029 (valid_gimple_call_p): Likewise.
18030 (finish_update_gimple_call): Likewise, and simplify.
18031 (gimplify_and_update_call_from_tree): Implement
18032 update_call_from_tree functionality, avoid excessive
18033 push/pop_gimplify_context.
18034 (gimple_fold_builtin): Use only gimplify_and_update_call_from_tree.
18035 (gimple_fold_call): Likewise.
18036 * gimple-ssa-sprintf.c (try_substitute_return_value): Likewise.
18037 * tree-ssa-ccp.c (ccp_folder::fold_stmt): Likewise.
18038 (pass_fold_builtins::execute): Likewise.
18039 (optimize_stack_restore): Use replace_call_with_value.
18040 * tree-cfg.c (fold_loop_internal_call): Likewise.
18041 * tree-ssa-dce.c (maybe_optimize_arith_overflow): Use
18042 only gimplify_and_update_call_from_tree.
18043 * tree-ssa-strlen.c (handle_builtin_strlen): Likewise.
18044 (handle_builtin_strchr): Likewise.
18045 * tsan.c: Include gimple-fold.h instead of tree-ssa-propagate.h.
18047 2021-04-26 Jakub Jelinek <jakub@redhat.com>
18050 * vmsdbgout.c (ASM_OUTPUT_DEBUG_STRING, vmsdbgout_begin_block,
18051 vmsdbgout_end_block, lookup_filename, vmsdbgout_source_line): Remove
18054 2021-04-25 liuhongt <hongtao.liu@intel.com>
18057 * config/i386/i386-builtin.def (BDESC): Change the icode of
18058 the following builtins to CODE_FOR_nothing.
18059 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
18060 IX86_BUILTIN_PCMPEQB128, IX86_BUILTIN_PCMPEQW128,
18061 IX86_BUILTIN_PCMPEQD128, IX86_BUILTIN_PCMPEQQ,
18062 IX86_BUILTIN_PCMPEQB256, IX86_BUILTIN_PCMPEQW256,
18063 IX86_BUILTIN_PCMPEQD256, IX86_BUILTIN_PCMPEQQ256,
18064 IX86_BUILTIN_PCMPGTB128, IX86_BUILTIN_PCMPGTW128,
18065 IX86_BUILTIN_PCMPGTD128, IX86_BUILTIN_PCMPGTQ,
18066 IX86_BUILTIN_PCMPGTB256, IX86_BUILTIN_PCMPGTW256,
18067 IX86_BUILTIN_PCMPGTD256, IX86_BUILTIN_PCMPGTQ256.
18068 * config/i386/sse.md (avx2_eq<mode>3): Deleted.
18069 (sse2_eq<mode>3): Ditto.
18070 (sse4_1_eqv2di3): Ditto.
18071 (sse2_gt<mode>3): Rename to ..
18072 (*sse2_gt<mode>3): .. this.
18074 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
18077 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
18080 * config/darwin.c (darwin_binds_local_p): Assume that any
18081 public symbol might be interposed for PIC code. Update function
18082 header comment to reflect current Darwin capability.
18084 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
18087 * config/darwin.c (darwin_binds_local_p): Assume that any
18088 public symbol might be interposed for PIC code. Update function
18089 header comment to reflect current Darwin capability.
18091 2021-04-24 Richard Sandiford <richard.sandiford@arm.com>
18093 * doc/sourcebuild.texi: Document no-opts and any-opts target
18096 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
18098 * config/bpf/bpf.h (ASM_OUTPUT_ALIGNED_BSS): Use .type and .lcomm.
18100 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
18102 * config/bpf/bpf.h (FUNCTION_BOUNDARY): Set to 64.
18104 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
18107 * config/i386/i386-options.c (ix86_option_override_internal):
18108 Error out when -m96bit-long-double is used with 64bit targets.
18109 * config/i386/i386.md (*pushxf_rounded): Remove pattern.
18111 2021-04-23 Martin Liska <mliska@suse.cz>
18113 * lto-wrapper.c: Remove FIXME about usage of
18114 hardware_concurrency. The function is not on par with
18117 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
18120 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
18121 Copy operand 3 to operand 4. Use sse_reg_operand
18122 as operand 3 predicate.
18123 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): Ditto.
18124 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
18125 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
18126 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2):
18127 Copy operand 1 to operand 0.
18128 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): Ditto.
18129 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
18130 (LDX_ATOMIC/LDX_ATOMIC FP store peephole2 with mem blockage): Ditto.
18132 2021-04-23 Alex Coplan <alex.coplan@arm.com>
18134 PR rtl-optimization/100230
18135 * early-remat.c (early_remat::sort_candidates): Use delete[]
18136 instead of delete for array allocated with new[].
18138 2021-04-23 Richard Biener <rguenther@suse.de>
18140 * genmatch.c (lower_cond): Remove VEC_COND_EXPR special-casing.
18141 (capture_info::capture_info): Likewise.
18142 (capture_info::walk_match): Likewise.
18143 (expr::gen_transform): Likewise.
18144 (dt_simplify::gen_1): Likewise.
18145 * gimple-match-head.c (maybe_resimplify_conditional_op):
18146 Remove VEC_COND_EXPR special-casing.
18147 (gimple_simplify): Likewise.
18148 * gimple.c (gimple_could_trap_p_1): Adjust.
18149 * tree-ssa-pre.c (compute_avail): Allow VEC_COND_EXPR
18150 to participate in PRE.
18152 2021-04-23 Richard Biener <rguenther@suse.de>
18154 * cfganal.c (connect_infinite_loops_to_exit): First call
18155 add_noreturn_fake_exit_edges.
18156 * ipa-sra.c (process_scan_results): Do not call the now redundant
18157 add_noreturn_fake_exit_edges.
18158 * predict.c (tree_estimate_probability): Likewise.
18159 (rebuild_frequencies): Likewise.
18160 * store-motion.c (one_store_motion_pass): Likewise.
18162 2021-04-23 Richard Biener <rguenther@suse.de>
18164 PR tree-optimization/100222
18165 * predict.c (pass_profile::execute): Remove redundant call to
18166 mark_irreducible_loops.
18167 (report_predictor_hitrates): Likewise.
18169 2021-04-23 Richard Biener <rguenther@suse.de>
18171 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Avoid
18172 valid_gimple_rhs_p by instead gimplifying to one.
18174 2021-04-23 Richard Biener <rguenther@suse.de>
18176 PR tree-optimization/99971
18177 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
18178 Always use TBAA for loads.
18180 2021-04-23 liuhongt <hongtao.liu@intel.com>
18183 * config/i386/i386-options.c (ix86_option_override_internal):
18184 Clear MASK_AVX256_SPLIT_UNALIGNED_LOAD/STORE in x_target_flags
18185 when X86_TUNE_AVX256_UNALIGNED_LOAD/STORE_OPTIMAL is enabled
18186 by target attribute.
18188 2021-04-23 David Edelsohn <dje.gcc@gmail.com>
18190 * config/rs6000/aix71.h (PREFERRED_DEBUGGING_TYPE): Change to
18192 * config/rs6000/aix72.h (PREFERRED_DEBUGGING_TYPE): Same.
18194 2021-04-22 David Edelsohn <dje.gcc@gmail.com>
18196 * config.gcc (powerpc-ibm-aix6.*): Remove.
18197 * config/rs6000/aix61.h: Delete.
18199 2021-04-22 Martin Liska <mliska@suse.cz>
18201 PR testsuite/100159
18202 PR testsuite/100192
18203 * builtins.c (expand_builtin): Fix typos and missing comments.
18204 * dwarf2out.c (gen_subprogram_die): Likewise.
18205 (gen_struct_or_union_type_die): Likewise.
18207 2021-04-22 Uroš Bizjak <ubizjak@gmail.com>
18210 * config/i386/i386-expand.c (ix86_expand_convert_uns_sidf_sse):
18211 Remove the sign with FE_DOWNWARD, where x - x = -0.0.
18213 2021-04-21 Iain Sandoe <iain@sandoe.co.uk>
18215 * config/i386/darwin.h (TARGET_64BIT): Remove definition
18216 based on TARGET_ISA_64BIT.
18217 (TARGET_64BIT_P): Remove definition based on
18218 TARGET_ISA_64BIT_P().
18220 2021-04-21 Martin Liska <mliska@suse.cz>
18223 2021-04-21 Martin Liska <mliska@suse.cz>
18225 * lto-wrapper.c (cpuset_popcount): Remove.
18226 (init_num_threads): Remove and use hardware_concurrency.
18228 2021-04-21 Martin Liska <mliska@suse.cz>
18231 * main.c (main): Call toplev::finalize in CHECKING_P mode.
18232 * ipa-modref.c (ipa_modref_c_finalize): summaries are NULL
18233 when incremental LTO linking happens.
18235 2021-04-21 Martin Liska <mliska@suse.cz>
18237 * lto-wrapper.c (run_gcc): When -flto=jobserver is used, but the
18238 makeserver cannot be detected, then use -flto=N fallback.
18240 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
18242 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): When cross-compiling,
18243 default to yes for aarch64-linux-gnu.
18244 * configure: Regenerate.
18246 2021-04-21 Martin Liska <mliska@suse.cz>
18248 * lto-wrapper.c (cpuset_popcount): Remove.
18249 (init_num_threads): Remove and use hardware_concurrency.
18251 2021-04-21 Martin Liska <mliska@suse.cz>
18253 * config/i386/i386.c: Remove superfluous || TARGET_MACHO
18254 which remains to be '(... || 0)' and clang complains about it.
18255 * dwarf2out.c (AT_vms_delta): Declare conditionally.
18256 (add_AT_vms_delta): Likewise.
18257 * tree.c (fld_simplified_type): Use rather more common pattern
18258 for disabling of something (#if 0).
18259 (get_tree_code_name): Likewise.
18260 (verify_type_variant): Likewise.
18262 2021-04-21 Martin Liska <mliska@suse.cz>
18264 * config/i386/i386-expand.c (decide_alignment): Use newly named
18265 macro TARGET_CPU_P.
18266 * config/i386/i386.c (ix86_decompose_address): Likewise.
18267 (ix86_address_cost): Likewise.
18268 (ix86_lea_outperforms): Likewise.
18269 (ix86_avoid_lea_for_addr): Likewise.
18270 (ix86_add_stmt_cost): Likewise.
18271 * config/i386/i386.h (TARGET_*): Remove.
18272 (TARGET_CPU_P): New macro.
18273 * config/i386/i386.md: Use newly named macro TARGET_CPU_P.
18274 * config/i386/x86-tune-sched-atom.c (do_reorder_for_imul): Likewise.
18275 (swap_top_of_ready_list): Likewise.
18276 (ix86_atom_sched_reorder): Likewise.
18277 * config/i386/x86-tune-sched-bd.c (ix86_bd_has_dispatch): Likewise.
18278 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Likewise.
18280 2021-04-21 Martin Liska <mliska@suse.cz>
18282 * config/i386/i386-options.c (TARGET_EXPLICIT_NO_SAHF_P):
18284 (SET_TARGET_NO_SAHF): Likewise.
18285 (TARGET_EXPLICIT_PREFETCH_SSE_P): Likewise.
18286 (SET_TARGET_PREFETCH_SSE): Likewise.
18287 (TARGET_EXPLICIT_NO_TUNE_P): Likewise.
18288 (SET_TARGET_NO_TUNE): Likewise.
18289 (TARGET_EXPLICIT_NO_80387_P): Likewise.
18290 (SET_TARGET_NO_80387): Likewise.
18292 * config/i386/i386.h (TARGET_*): Remove.
18293 * opth-gen.awk: Generate new used macros.
18295 2021-04-21 Martin Liska <mliska@suse.cz>
18297 * config/i386/i386.h (PTA_*): Remove.
18298 (enum pta_flag): New.
18299 (DEF_PTA): Generate PTA_* values from i386-isa.def.
18300 * config/i386/i386-isa.def: New file.
18302 2021-04-21 Alex Coplan <alex.coplan@arm.com>
18305 * config/aarch64/aarch64-bti-insert.c (aarch64_bti_j_insn_p): New.
18306 (rest_of_insert_bti): Avoid inserting duplicate bti j insns for
18307 jump table targets.
18309 2021-04-21 H.J. Lu <hjl.tools@gmail.com>
18311 * config.gcc: Install mwaitintrin.h for i[34567]86-*-* and
18312 x86_64-*-* targets.
18313 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_MWAIT_SET):
18315 (OPTION_MASK_ISA2_MWAIT_UNSET): Likewise.
18316 (ix86_handle_option): Handle -mmwait.
18317 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
18318 Replace OPTION_MASK_ISA_SSE3 with OPTION_MASK_ISA2_MWAIT on
18319 __builtin_ia32_monitor and __builtin_ia32_mwait.
18320 * config/i386/i386-options.c (isa2_opts): Add -mmwait.
18321 (ix86_valid_target_attribute_inner_p): Likewise.
18322 (ix86_option_override_internal): Enable mwait/monitor
18323 instructions for -msse3.
18324 * config/i386/i386.h (TARGET_MWAIT): New.
18325 (TARGET_MWAIT_P): Likewise.
18326 * config/i386/i386.opt: Add -mmwait.
18327 * config/i386/mwaitintrin.h: New file.
18328 * config/i386/pmmintrin.h: Include <mwaitintrin.h>.
18329 * config/i386/sse.md (sse3_mwait): Replace TARGET_SSE3 with
18331 (@sse3_monitor_<mode>): Likewise.
18332 * config/i386/x86gprintrin.h: Include <mwaitintrin.h>.
18333 * doc/extend.texi: Document mwait target attribute.
18334 * doc/invoke.texi: Document -mmwait.
18336 2021-04-21 Martin Liska <mliska@suse.cz>
18338 * config/i386/i386-options.c (DEF_ENUM): Remove it.
18339 * config/i386/i386-opts.h (DEF_ENUM): Likewise.
18340 * config/i386/stringop.def (DEF_ENUM): Likewise.
18342 2021-04-21 Martin Liska <mliska@suse.cz>
18344 * tree-cfg.c (gimple_verify_flow_info): Use qD instead
18345 of print_generic_expr.
18347 2021-04-21 Jakub Jelinek <jakub@redhat.com>
18349 PR rtl-optimization/100148
18350 * cprop.c (constprop_register): Use next_nondebug_insn instead of
18353 2021-04-21 Martin Liska <mliska@suse.cz>
18356 * cgraphunit.c (cgraph_node::analyze): Remove duplicate
18357 free_dominance_info calls.
18359 2021-04-21 Richard Biener <rguenther@suse.de>
18361 * gimple-fold.c (maybe_fold_reference): Remove is_lhs
18362 parameter (and assume it to be false).
18363 (fold_gimple_assign): Adjust, remove all callers of
18364 maybe_fold_reference calling it with is_lhs true.
18365 (gimple_fold_call): Likewise.
18366 (fold_stmt_1): Likewise.
18368 2021-04-21 Richard Biener <rguenther@suse.de>
18370 * fold-const.c (pedantic_non_lvalue_loc): Remove.
18371 (fold_binary_loc): Adjust.
18372 (fold_ternary_loc): Likewise.
18374 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
18376 PR middle-end/100130
18377 * varasm.c (get_block_for_decl): Make sure that any use of the
18378 retain attribute matches the section's retain flag.
18379 (switch_to_section): Check for retain mismatches even when
18380 changing sections, but do not warn if the given decl is the
18381 section's named.decl.
18382 (output_object_block): Pass the first decl in the block (if any)
18383 to switch_to_section.
18385 2021-04-20 H.J. Lu <hjl.tools@gmail.com>
18387 * config/i386/i386-c.c (ix86_target_macros_internal): Define
18388 __CRC32__ for -mcrc32.
18389 * config/i386/i386-options.c (ix86_option_override_internal):
18390 Enable crc32 instruction for -msse4.2.
18391 * config/i386/i386.md (sse4_2_crc32<mode>): Remove TARGET_SSE4_2
18393 (sse4_2_crc32di): Likewise.
18394 * config/i386/ia32intrin.h: Use crc32 target option for CRC32
18397 2021-04-20 Segher Boessenkool <segher@kernel.crashing.org>
18400 * config/rs6000/rs6000.c (rs6000_machine_from_flags): Do not consider
18403 2021-04-20 Martin Liska <mliska@suse.cz>
18405 * doc/invoke.texi: Fix typo.
18406 * params.opt: Likewise.
18408 2021-04-20 Martin Liska <mliska@suse.cz>
18410 * doc/invoke.texi: Document new param.
18412 2021-04-19 Andrew MacLeod <amacleod@redhat.com>
18414 PR tree-optimization/100081
18415 * gimple-range-cache.h (ranger_cache): Inherit from gori_compute
18416 rather than gori_compute_cache.
18417 * gimple-range-gori.cc (is_gimple_logical_p): Move to top of file.
18418 (range_def_chain::m_logical_depth): New member.
18419 (range_def_chain::range_def_chain): Initialize m_logical_depth.
18420 (range_def_chain::get_def_chain): Don't build defchains through more
18421 than LOGICAL_LIMIT logical expressions.
18422 * params.opt (param_ranger_logical_depth): New.
18424 2021-04-19 Richard Earnshaw <rearnsha@arm.com>
18427 * config/arm/arm.c (arm_configure_build_target): Do not strip
18428 extended FPU/SIMD feature bits from the target ISA when -mfpu
18429 is specified (partial revert of r11-8168).
18431 2021-04-19 Thomas Schwinge <thomas@codesourcery.com>
18433 * params.opt (-param=openacc-kernels=): Add.
18434 * omp-oacc-kernels-decompose.cc
18435 (pass_omp_oacc_kernels_decompose::gate): Use it.
18436 * doc/invoke.texi (-fopenacc-kernels=@var{mode}): Move...
18437 (--param): ... here, 'openacc-kernels'.
18439 2021-04-19 Martin Liska <mliska@suse.cz>
18442 * gengtype.c (finish_root_table): Align function arguments
18443 in between declaration and definition.
18445 2021-04-19 Eric Botcazou <ebotcazou@adacore.com>
18447 * config/i386/winnt.c (i386_pe_seh_cold_init): Properly deal with
18448 frames larger than the SEH maximum frame size.
18450 2021-04-18 Segher Boessenkool <segher@kernel.crashing.org>
18452 PR rtl-optimization/99927
18453 * combine.c (distribute_notes) [REG_UNUSED]: If the register already
18454 is dead, just drop it.
18456 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
18459 * config/i386/winnt-d.c (TARGET_D_TEMPLATES_ALWAYS_COMDAT): Define.
18460 * doc/tm.texi: Regenerate.
18461 * doc/tm.texi.in (D language and ABI): Add @hook for
18462 TARGET_D_TEMPLATES_ALWAYS_COMDAT.
18464 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
18466 * config/darwin-d.c (darwin_d_handle_target_object_format): New
18468 (darwin_d_register_target_info): New function.
18469 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18470 * config/dragonfly-d.c (dragonfly_d_handle_target_object_format): New
18472 (dragonfly_d_register_target_info): New function.
18473 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18474 * config/freebsd-d.c (freebsd_d_handle_target_object_format): New
18476 (freebsd_d_register_target_info): New function.
18477 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18478 * config/glibc-d.c (glibc_d_handle_target_object_format): New
18480 (glibc_d_register_target_info): New function.
18481 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18482 * config/i386/i386-d.c (ix86_d_handle_target_object_format): New
18484 (ix86_d_register_target_info): Add ix86_d_handle_target_object_format
18485 as handler for objectFormat key.
18486 * config/i386/winnt-d.c (winnt_d_handle_target_object_format): New
18488 (winnt_d_register_target_info): New function.
18489 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18490 * config/netbsd-d.c (netbsd_d_handle_target_object_format): New
18492 (netbsd_d_register_target_info): New function.
18493 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18494 * config/openbsd-d.c (openbsd_d_handle_target_object_format): New
18496 (openbsd_d_register_target_info): New function.
18497 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18498 * config/pa/pa-d.c (pa_d_handle_target_object_format): New function.
18499 (pa_d_register_target_info): Add pa_d_handle_target_object_format as
18500 handler for objectFormat key.
18501 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_object_format): New
18503 (rs6000_d_register_target_info): Add
18504 rs6000_d_handle_target_object_format as handler for objectFormat key.
18505 * config/sol2-d.c (solaris_d_handle_target_object_format): New
18507 (solaris_d_register_target_info): New function.
18508 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18510 2021-04-16 Jakub Jelinek <jakub@redhat.com>
18513 * config/aarch64/aarch64.c (aarch64_function_arg_alignment): Change
18514 abi_break argument from bool * to unsigned *, store there the pre-GCC 9
18516 (aarch64_layout_arg, aarch64_gimplify_va_arg_expr): Adjust callers.
18517 (aarch64_function_arg_regno_p): Likewise. Only emit -Wpsabi note if
18518 the old and new alignment after applying MIN/MAX to it is different.
18520 2021-04-16 Tamar Christina <tamar.christina@arm.com>
18523 * config/aarch64/aarch64-sve.md (@aarch64_sve_trn1_conv<mode>): New.
18524 * config/aarch64/aarch64.c (aarch64_expand_sve_const_pred_trn): Use new
18526 * config/aarch64/iterators.md (UNSPEC_TRN1_CONV): New.
18528 2021-04-16 Bill Schmidt <wschmidt@linux.ibm.com>
18530 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Revise
18531 this section and its subsections.
18533 2021-04-16 Jakub Jelinek <jakub@redhat.com>
18536 * config/aarch64/aarch64.md (*neg_asr_si2_extr, *extrsi5_insn_di): New
18537 define_insn patterns.
18539 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
18541 PR rtl-optimization/98689
18542 * reg-notes.def (UNTYPED_CALL): New note.
18543 * combine.c (distribute_notes): Handle it.
18544 * emit-rtl.c (try_split): Likewise.
18545 * rtlanal.c (rtx_properties::try_to_add_insn): Likewise. Assume
18546 that calls with the note implicitly set all return value registers.
18547 * builtins.c (expand_builtin_apply): Add a REG_UNTYPED_CALL
18550 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
18552 PR rtl-optimization/99596
18553 * rtlanal.c (rtx_properties::try_to_add_insn): Don't add global
18554 register accesses for const calls. Assume that pure functions
18555 can only read from global registers. Ignore cases in which
18556 the stack pointer has been marked global.
18558 2021-04-16 Jakub Jelinek <jakub@redhat.com>
18561 * tree-vect-loop.c (vect_transform_loop): Don't remove just
18562 dead scalar .MASK_LOAD calls, but also dead .COND_* calls - replace
18563 them by their last argument.
18565 2021-04-15 Martin Liska <mliska@suse.cz>
18567 * doc/invoke.texi: Other params don't use it, remove it.
18569 2021-04-15 Richard Biener <rguenther@suse.de>
18571 * gimple-builder.h: Add deprecation note.
18573 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
18576 * attribs.h (restrict_type_identity_attributes_to): Declare.
18577 * attribs.c (restrict_type_identity_attributes_to): New function.
18579 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
18582 * attribs.h (affects_type_identity_attributes): Declare.
18583 * attribs.c (remove_attributes_matching): New function.
18584 (affects_type_identity_attributes): Likewise.
18586 2021-04-15 Jakub Jelinek <jakub@redhat.com>
18589 * config/aarch64/aarch64.md (*<LOGICAL:optab>_<SHIFT:optab><mode>3):
18590 Add combine splitters for *<LOGICAL:optab>_ashl<mode>3 with
18591 ZERO_EXTEND, SIGN_EXTEND or AND.
18593 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
18595 PR rtl-optimization/99929
18596 * rtl.h (same_vector_encodings_p): New function.
18597 * cse.c (exp_equiv_p): Check that CONST_VECTORs have the same encoding.
18598 * cselib.c (rtx_equal_for_cselib_1): Likewise.
18599 * jump.c (rtx_renumbered_equal_p): Likewise.
18600 * lra-constraints.c (operands_match_p): Likewise.
18601 * reload.c (operands_match_p): Likewise.
18602 * rtl.c (rtx_equal_p_cb, rtx_equal_p): Likewise.
18604 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
18606 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Print
18607 more information about variable-length CONST_VECTORs.
18609 2021-04-14 Vladimir N. Makarov <vmakarov@redhat.com>
18611 PR rtl-optimization/100066
18612 * lra-constraints.c (split_reg): Check paradoxical_subreg_p for
18613 ordered modes when choosing splitting mode for hard reg.
18615 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
18618 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector_sel):
18620 (aarch64_expand_sve_const_vector): Use it for nelts_per_pattern==2.
18622 2021-04-14 Andreas Krebbel <krebbel@linux.ibm.com>
18624 * config/s390/s390-builtins.def (O_M5, O_M12, ...): Add new macros
18625 for mask operand types.
18626 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
18627 (s390_vec_permi_dbl, s390_vpdi): Use the M5 type for the immediate
18629 (s390_vec_msum_u128, s390_vmslg): Use the M12 type for the
18631 * config/s390/s390.c (s390_const_operand_ok): Check the new
18632 operand types and generate a list of valid values.
18634 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
18636 * doc/tm.texi: Regenerate.
18637 * doc/tm.texi.in (D language and ABI): Add @hook for
18638 TARGET_D_REGISTER_OS_TARGET_INFO.
18640 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
18642 * config/aarch64/aarch64-d.c (aarch64_d_handle_target_float_abi): New
18644 (aarch64_d_register_target_info): New function.
18645 * config/aarch64/aarch64-protos.h (aarch64_d_register_target_info):
18647 * config/aarch64/aarch64.h (TARGET_D_REGISTER_CPU_TARGET_INFO):
18649 * config/arm/arm-d.c (arm_d_handle_target_float_abi): New function.
18650 (arm_d_register_target_info): New function.
18651 * config/arm/arm-protos.h (arm_d_register_target_info): Declare.
18652 * config/arm/arm.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18653 * config/i386/i386-d.c (ix86_d_handle_target_float_abi): New function.
18654 (ix86_d_register_target_info): New function.
18655 * config/i386/i386-protos.h (ix86_d_register_target_info): Declare.
18656 * config/i386/i386.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18657 * config/mips/mips-d.c (mips_d_handle_target_float_abi): New function.
18658 (mips_d_register_target_info): New function.
18659 * config/mips/mips-protos.h (mips_d_register_target_info): Declare.
18660 * config/mips/mips.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18661 * config/pa/pa-d.c (pa_d_handle_target_float_abi): New function.
18662 (pa_d_register_target_info): New function.
18663 * config/pa/pa-protos.h (pa_d_register_target_info): Declare.
18664 * config/pa/pa.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18665 * config/riscv/riscv-d.c (riscv_d_handle_target_float_abi): New
18667 (riscv_d_register_target_info): New function.
18668 * config/riscv/riscv-protos.h (riscv_d_register_target_info): Declare.
18669 * config/riscv/riscv.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18670 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_float_abi): New
18672 (rs6000_d_register_target_info): New function.
18673 * config/rs6000/rs6000-protos.h (rs6000_d_register_target_info):
18675 * config/rs6000/rs6000.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18676 * config/s390/s390-d.c (s390_d_handle_target_float_abi): New function.
18677 (s390_d_register_target_info): New function.
18678 * config/s390/s390-protos.h (s390_d_register_target_info): Declare.
18679 * config/s390/s390.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18680 * config/sparc/sparc-d.c (sparc_d_handle_target_float_abi): New
18682 (sparc_d_register_target_info): New function.
18683 * config/sparc/sparc-protos.h (sparc_d_register_target_info): Declare.
18684 * config/sparc/sparc.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18685 * doc/tm.texi: Regenerate.
18686 * doc/tm.texi.in (D language and ABI): Add @hook for
18687 TARGET_D_REGISTER_CPU_TARGET_INFO.
18689 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
18691 * config/i386/i386-d.c (ix86_d_has_stdcall_convention): New function.
18692 * config/i386/i386-protos.h (ix86_d_has_stdcall_convention): Declare.
18693 * config/i386/i386.h (TARGET_D_HAS_STDCALL_CONVENTION): Define.
18694 * doc/tm.texi: Regenerate.
18695 * doc/tm.texi.in (D language and ABI): Add @hook for
18696 TARGET_D_HAS_STDCALL_CONVENTION.
18698 2021-04-14 Richard Biener <rguenther@suse.de>
18700 * tree-cfg.c (verify_gimple_assign_ternary): Verify that
18701 VEC_COND_EXPRs have a gimple_val condition.
18702 * tree-ssa-propagate.c (valid_gimple_rhs_p): VEC_COND_EXPR
18703 can no longer have a GENERIC condition.
18705 2021-04-14 Richard Earnshaw <rearnsha@arm.com>
18708 * config/arm/arm.c (arm_configure_build_target): Strip isa_all_fpbits
18709 from the isa_delta when -mfpu has been used.
18710 (arm_options_perform_arch_sanity_checks): It's the architecture that
18711 lacks an FPU not the processor.
18713 2021-04-13 Richard Biener <rguenther@suse.de>
18715 PR tree-optimization/100053
18716 * tree-ssa-sccvn.c (vn_nary_op_get_predicated_value): Do
18717 not use optimistic dominance queries for backedges to validate
18719 (dominated_by_p_w_unex): Add parameter to ignore executable
18720 state on backedges.
18721 (rpo_elim::eliminate_avail): Adjust.
18723 2021-04-13 Jakub Jelinek <jakub@redhat.com>
18726 * config/aarch64/aarch64.md (*aarch64_bfxil<mode>_extr,
18727 *aarch64_bfxilsi_extrdi): New define_insn patterns.
18729 2021-04-13 Jakub Jelinek <jakub@redhat.com>
18732 * simplify-rtx.c (simplify_immed_subreg): For MODE_COMPOSITE_P
18733 outermode, return NULL if the result doesn't encode back to the
18734 original byte sequence.
18735 (simplify_gen_subreg): Don't create SUBREGs from constants to
18736 MODE_COMPOSITE_P outermode.
18738 2021-04-12 Jakub Jelinek <jakub@redhat.com>
18740 PR rtl-optimization/99905
18741 * combine.c (expand_compound_operation): If pos + len > modewidth,
18742 perform the right shift by pos in inner_mode and then convert to mode,
18743 instead of trying to simplify a shift of rtx with inner_mode by pos
18744 as if it was a shift in mode.
18746 2021-04-12 Jakub Jelinek <jakub@redhat.com>
18749 * combine.c (simplify_and_const_int_1): Don't optimize varop
18750 away if it has side-effects.
18752 2021-04-12 Martin Liska <mliska@suse.cz>
18754 * doc/extend.texi: Escape @smallexample content.
18756 2021-04-12 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
18758 * config/s390/s390.md ("*movdi_31", "*movdi_64"): Add
18759 alternative in order to load a DFP zero.
18761 2021-04-12 Martin Liska <mliska@suse.cz>
18763 * doc/extend.texi: Be more precise in documentation
18764 of symver attribute.
18766 2021-04-12 Martin Liska <mliska@suse.cz>
18769 * gimplify.c (gimplify_expr): Right now, we unpoison all
18770 variables before a goto <dest>. We should not do it if we are
18773 2021-04-12 Cui,Lili <lili.cui@intel.com>
18775 * common/config/i386/cpuinfo.h (get_intel_cpu): Handle
18777 * common/config/i386/i386-common.c (processor_names): Add
18779 (processor_alias_table): Add rocketlake.
18780 * common/config/i386/i386-cpuinfo.h (processor_subtypes): Add
18781 INTEL_COREI7_ROCKETLAKE.
18782 * config.gcc: Add -march=rocketlake.
18783 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
18785 * config/i386/i386-options.c (m_ROCKETLAKE) : Define.
18786 (processor_cost_table): Add rocketlake cost.
18787 * config/i386/i386.h (ix86_size_cost) : Define
18789 (processor_type) : Add PROCESSOR_ROCKETLAKE.
18790 (PTA_ROCKETLAKE): Ditto.
18791 * doc/extend.texi: Add rocketlake.
18792 * doc/invoke.texi: Add rocketlake.
18794 2021-04-12 Cui,Lili <lili.cui@intel.com>
18796 * config/i386/i386.h (PTA_ALDERLAKE): Change alderlake ISA list.
18797 * config/i386/i386-options.c (m_CORE_AVX2): Add m_ALDERLAKE.
18798 * common/config/i386/cpuinfo.h (get_intel_cpu): Add AlderLake model.
18799 * doc/invoke.texi: Change alderlake ISA list.
18801 2021-04-11 Hafiz Abid Qadeer <abidh@codesourcery.com>
18803 PR middle-end/98088
18804 * omp-expand.c (expand_oacc_collapse_init): Update condition in
18807 2021-04-10 H.J. Lu <hjl.tools@gmail.com>
18810 * config/i386/serializeintrin.h (_serialize): Defined as macro.
18812 2021-04-10 Jakub Jelinek <jakub@redhat.com>
18815 * expr.c (expand_expr_addr_expr_1): Test is_global_var rather than
18816 just TREE_STATIC on COMPOUND_LITERAL_EXPR_DECLs.
18818 2021-04-10 Jakub Jelinek <jakub@redhat.com>
18820 PR middle-end/99989
18821 * gimple-ssa-warn-alloca.c
18822 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit to
18823 0 with integer precision unconditionally.
18825 2021-04-10 Jakub Jelinek <jakub@redhat.com>
18827 PR rtl-optimization/98601
18828 * rtlanal.c (rtx_addr_can_trap_p_1): Allow in assert unknown size
18829 not just for BLKmode, but also for VOIDmode. For STRICT_ALIGNMENT
18830 unaligned_mems handle VOIDmode like BLKmode.
18832 2021-04-10 Jan Hubicka <hubicka@ucw.cz>
18835 * tree.c (free_lang_data_in_decl): Do not release body of
18836 declare_variant_alt.
18838 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
18840 * config/aarch64/aarch64.c (aarch64_option_restore): If the
18841 architecture was specified explicitly and the tuning wasn't,
18842 tune for the architecture rather than the configured default CPU.
18844 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
18846 * config/aarch64/aarch64.md (tlsdesc_small_sve_<mode>): Use X30
18847 as the temporary register.
18849 2021-04-09 Martin Liska <mliska@suse.cz>
18851 * doc/extend.texi: Move non-target attributes on the top level.
18853 2021-04-09 Martin Liska <mliska@suse.cz>
18855 * doc/invoke.texi: Document minimum and maximum value of the
18856 argument for both supported compression algorithms.
18858 2021-04-08 David Edelsohn <dje.gcc@gmail.com>
18860 * config/rs6000/rs6000.c (rs6000_xcoff_select_section): Select
18861 TLS BSS before TLS data.
18862 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .comm.
18864 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
18866 * doc/sourcebuild.texi (stdint_types_mbig_endian): Document.
18868 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
18870 * match.pd: Extend vec_cond folds to handle shifts.
18872 2021-04-08 Maciej W. Rozycki <macro@orcam.me.uk>
18874 * config/vax/vax.md: Fix comment for `*bit<mode>' pattern's
18877 2021-04-08 Alex Coplan <alex.coplan@arm.com>
18880 * config/arm/iterators.md (MVE_vecs): New.
18881 (V_elem): Also handle V2DF.
18882 * config/arm/mve.md (*mve_mov<mode>): Rename to ...
18883 (*mve_vdup<mode>): ... this. Remove second alternative since
18884 vec_duplicate of const_int is not canonical RTL, and we don't
18885 want to match symbol_refs.
18886 (*mve_vec_duplicate<mode>): Delete (pattern is redundant).
18888 2021-04-08 Xionghu Luo <luoxhu@linux.ibm.com>
18890 * fold-const.c (fold_single_bit_test): Fix typo.
18891 * print-rtl.c (print_rtx_insn_vec): Call print_rtl_single
18894 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
18896 PR tree-optimization/97513
18897 * tree-vect-slp.c (vect_add_slp_permutation): New function,
18899 (vectorizable_slp_permutation): ...here. Detect cases in which
18900 all VEC_PERM_EXPRs are guaranteed to have the same stepped
18901 permute vector and only generate one permute vector for that case.
18902 Extend that case to handle variable-length vectors.
18904 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
18906 PR tree-optimization/99873
18907 * tree-vect-slp.c (vect_slp_prefer_store_lanes_p): New function.
18908 (vect_build_slp_instance): Don't split store groups that could
18909 use IFN_STORE_LANES.
18911 2021-04-07 Jakub Jelinek <jakub@redhat.com>
18914 * varasm.c (output_constant_pool_contents): Don't strip name encoding
18915 from XSTR (desc->sym, 0) or from label before passing those to
18918 2021-04-07 Richard Biener <rguenther@suse.de>
18920 PR tree-optimization/99954
18921 * tree-loop-distribution.c: Include tree-affine.h.
18922 (generate_memcpy_builtin): Try using tree-affine to prove
18924 (loop_distribution::classify_builtin_ldst): Always classify
18927 2021-04-07 Richard Biener <rguenther@suse.de>
18929 PR tree-optimization/99947
18930 * tree-vect-loop.c (vectorizable_induction): Pre-allocate
18931 steps vector to avoid pushing elements from the reallocated
18934 2021-04-07 Richard Biener <rguenther@suse.de>
18936 * tree-ssa-sccvn.h (print_vn_reference_ops): Declare.
18937 * tree-ssa-pre.c (print_pre_expr): Factor out VN reference operand
18939 * tree-ssa-sccvn.c (print_vn_reference_ops): ... into this new
18941 (debug_vn_reference_ops): New.
18943 2021-04-07 Bin Cheng <bin.cheng@linux.alibaba.com>
18945 PR tree-optimization/98736
18946 * tree-loop-distribution.c
18947 * (loop_distribution::bb_top_order_init):
18948 Compute RPO with programing order preserved by calling function
18949 rev_post_order_and_mark_dfs_back_seme.
18951 2021-04-06 Vladimir N. Makarov <vmakarov@redhat.com>
18954 * lra-constraints.c (split_reg): Don't check paradoxical_subreg_p.
18955 * lra-lives.c (clear_sparseset_regnos, regnos_in_sparseset_p): New
18957 (process_bb_lives): Don't update biggest mode of hard reg for
18958 implicit in multi-register group. Use the new functions for
18959 updating dead_set and unused_set by register notes.
18961 2021-04-06 Xianmiao Qu <xianmiao_qu@c-sky.com>
18963 * config/csky/csky_pipeline_ck802.md : Use insn reservation name
18966 2021-04-06 H.J. Lu <hjl.tools@gmail.com>
18968 * config/i386/x86-tune-costs.h (skylake_memcpy): Updated.
18969 (skylake_memset): Likewise.
18970 (skylake_cost): Change CLEAR_RATIO to 17.
18971 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
18972 Replace m_CANNONLAKE, m_ICELAKE_CLIENT, m_ICELAKE_SERVER,
18973 m_TIGERLAKE and m_SAPPHIRERAPIDS with m_SKYLAKE and m_CORE_AVX512.
18975 2021-04-06 Richard Biener <rguenther@suse.de>
18977 PR tree-optimization/99880
18978 * tree-vect-loop.c (maybe_set_vectorized_backedge_value): Only
18979 set vectorized defs of relevant PHIs.
18981 2021-04-06 Richard Biener <rguenther@suse.de>
18983 PR tree-optimization/99924
18984 * tree-vect-slp.c (vect_bb_partition_graph_r): Do not mark
18985 nodes w/o scalar stmts as visited.
18987 2021-04-06 Alex Coplan <alex.coplan@arm.com>
18990 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Also use base
18991 PCS for [su]fix_optab.
18993 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
18995 * config/darwin.c (machopic_legitimize_pic_address): Check
18996 that the current pic register is one of the hard reg set
18997 before setting liveness.
18999 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
19001 * config/darwin.c (machopic_legitimize_pic_address): Fix
19002 whitespace, remove unused code.
19004 2021-04-03 Jakub Jelinek <jakub@redhat.com>
19006 PR tree-optimization/99882
19007 * gimple-ssa-store-merging.c (bswap_view_convert): Handle val with
19010 2021-04-03 Jakub Jelinek <jakub@redhat.com>
19012 PR rtl-optimization/99863
19013 * dse.c (replace_read): Drop regs_live argument. Instead of
19014 regs_live, use store_insn->fixed_regs_live if non-NULL,
19015 otherwise punt if insns sequence clobbers or sets any hard
19018 2021-04-03 Jakub Jelinek <jakub@redhat.com>
19021 * targhooks.h (default_print_patchable_function_entry_1): Declare.
19022 * targhooks.c (default_print_patchable_function_entry_1): New function,
19023 copied from default_print_patchable_function_entry with an added flags
19025 (default_print_patchable_function_entry): Rewritten into a small
19026 wrapper around default_print_patchable_function_entry_1.
19027 * config/rs6000/rs6000.c (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
19029 (rs6000_print_patchable_function_entry): New function.
19031 2021-04-02 Eric Botcazou <ebotcazou@adacore.com>
19033 * doc/invoke.texi (fdelete-dead-exceptions): Minor tweak.
19035 2021-04-01 Jason Merrill <jason@redhat.com>
19038 * common.opt: Document v15 and v16.
19040 2021-04-01 Richard Biener <rguenther@suse.de>
19042 PR tree-optimization/99863
19043 * gimplify.c (gimplify_init_constructor): Recompute vector
19046 2021-04-01 Jakub Jelinek <jakub@redhat.com>
19048 * doc/extend.texi (symver attribute): Fix up syntax errors
19051 2021-04-01 Jakub Jelinek <jakub@redhat.com>
19053 PR tree-optimization/96573
19054 * gimple-ssa-store-merging.c (init_symbolic_number): Handle
19055 also pointer types.
19057 2021-04-01 Richard Biener <rguenther@suse.de>
19059 PR tree-optimization/99856
19060 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Promote
19061 precision to vector element precision.
19063 2021-04-01 Martin Jambor <mjambor@suse.cz>
19065 PR tree-optimization/97009
19066 * tree-sra.c (access_or_its_child_written): New function.
19067 (propagate_subaccesses_from_rhs): Use it instead of a simple grp_write
19070 2021-03-31 Jan Hubicka <hubicka@ucw.cz>
19073 * cif-code.def (USES_COMDAT_LOCAL): Make CIF_FINAL_NORMAL.
19075 2021-03-31 Pat Haugen <pthaugen@linux.ibm.com>
19078 * config/rs6000/altivec.md (xxspltiw_v4si, xxspltiw_v4sf_inst,
19079 xxspltidp_v2df_inst, xxsplti32dx_v4si_inst, xxsplti32dx_v4sf_inst,
19080 xxblend_<mode>, xxpermx_inst, xxeval): Mark prefixed.
19081 * config/rs6000/mma.md (mma_<vvi4i4i8>, mma_<avvi4i4i8>,
19082 mma_<vvi4i4i2>, mma_<avvi4i4i2>, mma_<vvi4i4>, mma_<avvi4i4>,
19083 mma_<pvi4i2>, mma_<apvi4i2>, mma_<vvi4i4i4>, mma_<avvi4i4i4>):
19085 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Adjust test.
19086 * config/rs6000/rs6000.md (define_attr "maybe_prefixed"): New.
19087 (define_attr "prefixed"): Update initializer.
19089 2021-03-31 Jakub Jelinek <jakub@redhat.com>
19092 * dwarf2out.c (debug_ranges_dwo_section): New variable.
19093 (DW_RANGES_IDX_SKELETON): Define.
19094 (struct dw_ranges): Add begin_entry and end_entry members.
19095 (DEBUG_DWO_RNGLISTS_SECTION): Define.
19096 (add_ranges_num): Adjust r initializer for addition of *_entry
19098 (add_ranges_by_labels): For -gsplit-dwarf and force_direct,
19099 set idx to DW_RANGES_IDX_SKELETON.
19100 (use_distinct_base_address_for_range): New function.
19101 (index_rnglists): Don't set r->idx if it is equal to
19102 DW_RANGES_IDX_SKELETON. Initialize r->begin_entry and
19103 r->end_entry for -gsplit-dwarf if those will be needed by
19105 (output_rnglists): Add DWO argument. If true, switch to
19106 debug_ranges_dwo_section rather than debug_ranges_section.
19107 Adjust l1/l2 label indexes. Only output the offset table when
19108 dwo is true and don't include in there the skeleton range
19109 entry if present. For -gsplit-dwarf, skip ranges that belong
19110 to the other rnglists section. Change return type from void
19111 to bool and return true if there are any range entries for
19112 the other section. For dwarf_split_debug_info use
19113 DW_RLE_startx_endx, DW_RLE_startx_length and DW_RLE_base_addressx
19114 entries instead of DW_RLE_start_end, DW_RLE_start_length and
19115 DW_RLE_base_address. Use use_distinct_base_address_for_range.
19116 (init_sections_and_labels): Initialize debug_ranges_dwo_section
19117 if -gsplit-dwarf and DWARF >= 5. Adjust ranges_section_label
19118 and range_base_label indexes.
19119 (dwarf2out_finish): Call index_rnglists earlier before finalizing
19120 .debug_addr. Never emit DW_AT_rnglists_base attribute. For
19121 -gsplit-dwarf and DWARF >= 5 call output_rnglists up to twice
19122 with different dwo arguments.
19123 (dwarf2out_c_finalize): Clear debug_ranges_dwo_section.
19125 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19127 PR tree-optimization/98268
19128 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Call
19129 recompute_tree_invariant_for_addr_expr after successfully
19130 folding a TARGET_MEM_REF that occurs inside an ADDR_EXPR.
19132 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19134 PR tree-optimization/99726
19135 * tree-data-ref.c (create_intersect_range_checks_index): Bail
19136 out if there is more than one access function SCEV for the loop
19139 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19141 PR rtl-optimization/97141
19142 PR rtl-optimization/98726
19143 * emit-rtl.c (valid_for_const_vector_p): Return true for
19145 * rtx-vector-builder.h (rtx_vector_builder::step): Return a
19146 poly_wide_int instead of a wide_int.
19147 (rtx_vector_builder::apply_set): Take a poly_wide_int instead
19149 * rtx-vector-builder.c (rtx_vector_builder::apply_set): Likewise.
19150 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Return
19151 false for CONST_VECTORs that cannot be forced to memory.
19152 * config/aarch64/aarch64-simd.md (mov<mode>): If a CONST_VECTOR
19153 is too complex to force to memory, build it up from individual
19156 2021-03-31 Jan Hubicka <jh@suse.cz>
19159 * cgraph.c (cgraph_node::release_body): Fix overactive check.
19161 2021-03-31 Christophe Lyon <christophe.lyon@linaro.org>
19164 * config/arm/vec-common.md (mul<mode>3): Disable on iwMMXT, expect
19167 2021-03-31 H.J. Lu <hjl.tools@gmail.com>
19169 * config/i386/i386-expand.c (expand_set_or_cpymem_via_rep):
19170 For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, don't convert QImode
19172 (decide_alg): For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, use
19173 "rep movsb/stosb" only for known sizes.
19174 * config/i386/i386-options.c (processor_cost_table): Use Ice
19175 Lake cost for Cannon Lake, Ice Lake, Tiger Lake, Sapphire
19176 Rapids and Alder Lake.
19177 * config/i386/i386.h (TARGET_PREFER_KNOWN_REP_MOVSB_STOSB): New.
19178 * config/i386/x86-tune-costs.h (icelake_memcpy): New.
19179 (icelake_memset): Likewise.
19180 (icelake_cost): Likewise.
19181 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
19184 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19187 * config/aarch64/aarch64.c
19188 (aarch64_vectorize_preferred_vector_alignment): Query the size
19189 of the provided SVE vector; do not assume that all SVE vectors
19190 have the same size.
19192 2021-03-31 Jan Hubicka <jh@suse.cz>
19195 * cgraph.c (cgraph_node::release_body): Remove all callers and
19197 * cgraphclones.c (cgraph_node::materialize_clone): Do not do it here.
19198 * cgraphunit.c (cgraph_node::expand): And here.
19200 2021-03-31 Martin Liska <mliska@suse.cz>
19202 * ipa-modref.c (analyze_ssa_name_flags): Fix coding style
19203 and one negated condition.
19205 2021-03-31 Jakub Jelinek <jakub@redhat.com>
19206 Richard Sandiford <richard.sandiford@arm.com>
19209 * config/aarch64/aarch64.md (*add<mode>3_poly_1): Swap Uai and Uav
19210 constraints on operands[2] and similarly 0 and rk constraints
19211 on operands[1] corresponding to that.
19213 2021-03-31 Jakub Jelinek <jakub@redhat.com>
19216 * configure.ac (HAVE_LD_BROKEN_PE_DWARF5): New AC_DEFINE if PECOFF
19217 linker doesn't support DWARF sections new in DWARF5.
19218 * config/i386/i386-options.c (ix86_option_override_internal): Default
19219 to dwarf_version 4 if HAVE_LD_BROKEN_PE_DWARF5 for TARGET_PECOFF
19221 * config.in: Regenerated.
19222 * configure: Regenerated.
19224 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19227 * config/aarch64/aarch64.c (aarch64_analyze_loop_vinfo): Check for
19228 available issue_info before using it.
19230 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19233 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): Do not allow zero
19236 2021-03-30 Xionghu Luo <luoxhu@linux.ibm.com>
19239 * config/rs6000/altivec.md (altivec_lvsl_reg): Change to ...
19240 (altivec_lvsl_reg_<mode>): ... this.
19241 (altivec_lvsr_reg): Change to ...
19242 (altivec_lvsr_reg_<mode>): ... this.
19243 * config/rs6000/predicates.md (vec_set_index_operand): New.
19244 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
19245 Enable 32bit variable vec_insert for all TARGET_VSX.
19246 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
19247 Enable 32bit variable vec_insert for p9 and above.
19248 (rs6000_expand_vector_set_var_p8): Rename to ...
19249 (rs6000_expand_vector_set_var_p7): ... this.
19250 (rs6000_expand_vector_set): Use TARGET_VSX and adjust assert
19252 * config/rs6000/vector.md (vec_set<mode>): Use vec_set_index_operand.
19253 * config/rs6000/vsx.md (xl_len_r): Use gen_altivec_lvsl_reg_di and
19254 gen_altivec_lvsr_reg_di.
19256 2021-03-30 H.J. Lu <hjl.tools@gmail.com>
19259 * config/i386/ia32intrin.h (__rdtsc): Defined as macro.
19260 (__rdtscp): Likewise.
19262 2021-03-30 Tamar Christina <tamar.christina@arm.com>
19264 PR tree-optimization/99825
19265 * tree-vect-slp-patterns.c (vect_check_evenodd_blend):
19266 Reject non-mult 2 lanes.
19268 2021-03-30 Richard Earnshaw <rearnsha@arm.com>
19271 * config/arm/arm.c (arm_file_start): Fix emission of
19272 Tag_ABI_VFP_args attribute.
19274 2021-03-30 Richard Biener <rguenther@suse.de>
19276 PR tree-optimization/99824
19277 * stor-layout.c (set_min_and_max_values_for_integral_type):
19278 Assert the precision is within the bounds of
19279 WIDE_INT_MAX_PRECISION.
19280 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
19281 the outermost component ref only to lower the access size
19282 and initialize that from the access type.
19284 2021-03-30 Richard Sandiford <richard.sandiford@arm.com>
19287 * config/aarch64/aarch64.md (mov<mode>): Pass multi-instruction
19288 CONST_INTs to aarch64_expand_mov_immediate when called after RA.
19290 2021-03-30 Mihailo Stojanovic <mihailo.stojanovic@typhoon-hil.com>
19292 * config/aarch64/aarch64.md
19293 (<optab>_trunc<fcvt_target><GPI:mode>2): Set the "arch"
19294 attribute to disambiguate between SIMD and FP variants of the
19297 2021-03-29 Jan Hubicka <hubicka@ucw.cz>
19299 * ipa-modref.c (merge_call_lhs_flags): Correct handling of deref.
19300 (analyze_ssa_name_flags): Fix typo in comment.
19302 2021-03-29 Alex Coplan <alex.coplan@arm.com>
19305 * config/aarch64/aarch64-sve-builtins.cc
19306 (function_builder::add_function): Add placeholder_p argument, use
19307 placeholder decls if this is set.
19308 (function_builder::add_unique_function): Instead of conditionally adding
19309 direct overloads, unconditionally add either a direct overload or a
19311 (function_builder::add_overloaded_function): Set placeholder_p if we're
19312 using C++ overloads. Use the obstack for string storage instead
19313 of relying on the tree nodes.
19314 (function_builder::add_overloaded_functions): Don't return early for
19315 m_direct_overloads: we need to add placeholders.
19316 * config/aarch64/aarch64-sve-builtins.h
19317 (function_builder::add_function): Add placeholder_p argument.
19319 2021-03-29 Richard Biener <rguenther@suse.de>
19321 PR tree-optimization/99807
19322 * tree-vect-slp.c (vect_slp_analyze_node_operations_1): Move
19323 assert below VEC_PERM handling.
19325 2021-03-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19328 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>): Use
19329 aarch64_simd_or_scalar_imm_zero to match zeroes. Remove pattern
19330 matching const_int 0.
19331 (move_lo_quad_internal_be_<mode>): Likewise.
19332 (move_lo_quad_<mode>): Update for the above.
19333 * config/aarch64/iterators.md (VQ_2E): Delete.
19335 2021-03-29 Jakub Jelinek <jakub@redhat.com>
19337 PR tree-optimization/99777
19338 * fold-const.c (extract_muldiv_1): For conversions, punt on casts from
19339 types other than scalar integral types.
19341 2021-03-28 David Edelsohn <dje.gcc@gmail.com>
19343 * config/rs6000/rs6000.c (rs6000_output_dwarf_dtprel): Do not add
19344 XCOFF TLS reloc decorations.
19346 2021-03-28 Gerald Pfeifer <gerald@pfeifer.com>
19348 * doc/analyzer.texi (Analyzer Internals): Update link to
19349 "A Memory Model for Static Analysis of C Programs".
19351 2021-03-26 David Edelsohn <dje.gcc@gmail.com>
19353 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Call function.
19354 * config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align):
19356 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align): New.
19357 (rs6000_special_round_type_align): Recursively check innermost first
19360 2021-03-26 Jakub Jelinek <jakub@redhat.com>
19363 * dwarf2out.h (struct dw_fde_node): Add rule18 member.
19364 * dwarf2cfi.c (dwarf2out_frame_debug_expr): When handling (set hfp sp)
19365 assignment with drap_reg active, queue reg save for hfp with offset 0
19366 and flush queued reg saves. When handling a push with rule18,
19367 defer queueing reg save for hfp and just assert the offset is 0.
19368 (scan_trace): Assert that fde->rule18 is false.
19370 2021-03-26 Vladimir Makarov <vmakarov@redhat.com>
19373 * ira-costs.c (record_reg_classes): Put case with
19374 CT_RELAXED_MEMORY adjacent to one with CT_MEMORY.
19375 * ira.c (ira_setup_alts): Ditto.
19376 * lra-constraints.c (process_alt_operands): Ditto.
19377 * recog.c (asm_operand_ok): Ditto.
19378 * reload.c (find_reloads): Ditto.
19380 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19382 * config/aarch64/aarch64-protos.h
19383 (cpu_addrcost_table::post_modify_ld3_st3): New member variable.
19384 (cpu_addrcost_table::post_modify_ld4_st4): Likewise.
19385 * config/aarch64/aarch64.c (generic_addrcost_table): Update
19386 accordingly, using the same costs as for post_modify.
19387 (exynosm1_addrcost_table, xgene1_addrcost_table): Likewise.
19388 (thunderx2t99_addrcost_table, thunderx3t110_addrcost_table):
19389 (tsv110_addrcost_table, qdf24xx_addrcost_table): Likewise.
19390 (a64fx_addrcost_table): Likewise.
19391 (neoversev1_addrcost_table): New.
19392 (neoversev1_tunings): Use neoversev1_addrcost_table.
19393 (aarch64_address_cost): Use the new post_modify costs for CImode
19396 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19398 * config/aarch64/aarch64.opt
19399 (-param=aarch64-loop-vect-issue-rate-niters=): New parameter.
19400 * doc/invoke.texi: Document it.
19401 * config/aarch64/aarch64-protos.h (aarch64_base_vec_issue_info)
19402 (aarch64_scalar_vec_issue_info, aarch64_simd_vec_issue_info)
19403 (aarch64_advsimd_vec_issue_info, aarch64_sve_vec_issue_info)
19404 (aarch64_vec_issue_info): New structures.
19405 (cpu_vector_cost): Write comments above the variables rather
19407 (cpu_vector_cost::issue_info): New member variable.
19408 * config/aarch64/aarch64.c: Include gimple-pretty-print.h
19409 and tree-ssa-loop-niter.h.
19410 (generic_vector_cost, a64fx_vector_cost, qdf24xx_vector_cost)
19411 (thunderx_vector_cost, tsv110_vector_cost, cortexa57_vector_cost)
19412 (exynosm1_vector_cost, xgene1_vector_cost, thunderx2t99_vector_cost)
19413 (thunderx3t110_vector_cost): Initialize issue_info to null.
19414 (neoversev1_scalar_issue_info, neoversev1_advsimd_issue_info)
19415 (neoversev1_sve_issue_info, neoversev1_vec_issue_info): New structures.
19416 (neoversev1_vector_cost): Use them.
19417 (aarch64_vec_op_count, aarch64_sve_op_count): New structures.
19418 (aarch64_vector_costs::saw_sve_only_op): New member variable.
19419 (aarch64_vector_costs::num_vector_iterations): Likewise.
19420 (aarch64_vector_costs::scalar_ops): Likewise.
19421 (aarch64_vector_costs::advsimd_ops): Likewise.
19422 (aarch64_vector_costs::sve_ops): Likewise.
19423 (aarch64_vector_costs::seen_loads): Likewise.
19424 (aarch64_simd_vec_costs_for_flags): New function.
19425 (aarch64_analyze_loop_vinfo): Initialize num_vector_iterations.
19426 Count the number of predicate operations required by SVE WHILE
19428 (aarch64_comparison_type, aarch64_multiply_add_p): New functions.
19429 (aarch64_sve_only_stmt_p, aarch64_in_loop_reduction_latency): Likewise.
19430 (aarch64_count_ops): Likewise.
19431 (aarch64_add_stmt_cost): Record whether see an SVE operation
19432 that cannot currently be implementing using Advanced SIMD.
19433 Record issue information about the scalar, Advanced SIMD
19434 and (where relevant) SVE versions of a loop.
19435 (aarch64_vec_op_count::dump): New function.
19436 (aarch64_sve_op_count::dump): Likewise.
19437 (aarch64_estimate_min_cycles_per_iter): Likewise.
19438 (aarch64_adjust_body_cost): If issue information is available,
19439 try to compare the issue rates of the various loop implementations
19440 and increase or decrease the vector body cost accordingly.
19442 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19444 * config/aarch64/aarch64.c (aarch64_detect_vector_stmt_subtype):
19445 Assume a zero cost for induction phis.
19447 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19449 * config/aarch64/aarch64.c (aarch64_embedded_comparison_type): New
19451 (aarch64_adjust_stmt_cost): Add the costs of embedded scalar and
19452 vector comparisons.
19454 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19456 * config/aarch64/aarch64.c (aarch64_detect_scalar_stmt_subtype):
19458 (aarch64_add_stmt_cost): Call it.
19460 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19462 * config/aarch64/aarch64-tuning-flags.def (matched_vector_throughput):
19463 New tuning parameter.
19464 * config/aarch64/aarch64.c (neoversev1_tunings): Use it.
19465 (aarch64_estimated_sve_vq): New function.
19466 (aarch64_vector_costs::analyzed_vinfo): New member variable.
19467 (aarch64_vector_costs::is_loop): Likewise.
19468 (aarch64_vector_costs::unrolled_advsimd_niters): Likewise.
19469 (aarch64_vector_costs::unrolled_advsimd_stmts): Likewise.
19470 (aarch64_record_potential_advsimd_unrolling): New function.
19471 (aarch64_analyze_loop_vinfo, aarch64_analyze_bb_vinfo): Likewise.
19472 (aarch64_add_stmt_cost): Call aarch64_analyze_loop_vinfo or
19473 aarch64_analyze_bb_vinfo on the first use of a costs structure.
19474 Detect whether we're vectorizing a loop for SVE that might be
19475 completely unrolled if it used Advanced SIMD instead.
19476 (aarch64_adjust_body_cost_for_latency): New function.
19477 (aarch64_finish_cost): Call it.
19479 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19481 * config/aarch64/aarch64.c (aarch64_vector_costs): New structure.
19482 (aarch64_init_cost): New function.
19483 (aarch64_add_stmt_cost): Use aarch64_vector_costs instead of
19484 the default unsigned[3].
19485 (aarch64_finish_cost, aarch64_destroy_cost_data): New functions.
19486 (TARGET_VECTORIZE_INIT_COST): Override.
19487 (TARGET_VECTORIZE_FINISH_COST): Likewise.
19488 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
19490 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19492 * config/aarch64/aarch64.c (neoversev1_advsimd_vector_cost)
19493 (neoversev1_sve_vector_cost): New cost structures.
19494 (neoversev1_vector_cost): Likewise.
19495 (neoversev1_tunings): Use them. Enable use_new_vector_costs.
19497 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19499 * config/aarch64/aarch64-protos.h
19500 (sve_vec_cost::scatter_store_elt_cost): New member variable.
19501 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
19502 accordingly, taking the cost from the cost of a scalar_store.
19503 (a64fx_sve_vector_cost): Likewise.
19504 (aarch64_detect_vector_stmt_subtype): Detect scatter stores.
19506 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19508 * config/aarch64/aarch64-protos.h
19509 (simd_vec_cost::store_elt_extra_cost): New member variable.
19510 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
19511 accordingly, using the vec_to_scalar cost for the new field.
19512 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
19513 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
19514 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
19515 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
19516 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
19517 (thunderx3t110_advsimd_vector_cost): Likewise.
19518 (aarch64_detect_vector_stmt_subtype): Detect single-element stores.
19520 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19522 * config/aarch64/aarch64-protos.h (simd_vec_cost::ld2_st2_permute_cost)
19523 (simd_vec_cost::ld3_st3_permute_cost): New member variables.
19524 (simd_vec_cost::ld4_st4_permute_cost): Likewise.
19525 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
19526 accordingly, using zero for the new costs.
19527 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
19528 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
19529 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
19530 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
19531 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
19532 (thunderx3t110_advsimd_vector_cost): Likewise.
19533 (aarch64_ld234_st234_vectors): New function.
19534 (aarch64_adjust_stmt_cost): Likewise.
19535 (aarch64_add_stmt_cost): Call aarch64_adjust_stmt_cost if using
19536 the new vector costs.
19538 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19540 * config/aarch64/aarch64-protos.h (sve_vec_cost): Turn into a
19541 derived class of simd_vec_cost. Add information about CLAST[AB]
19542 and FADDA instructions.
19543 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
19544 accordingly, using the vec_to_scalar costs for the new fields.
19545 (a64fx_sve_vector_cost): Likewise.
19546 (aarch64_reduc_type): New function.
19547 (aarch64_sve_in_loop_reduction_latency): Likewise.
19548 (aarch64_detect_vector_stmt_subtype): Take a vinfo parameter.
19549 Use aarch64_sve_in_loop_reduction_latency to handle SVE reductions
19550 that occur in the loop body.
19551 (aarch64_add_stmt_cost): Update call accordingly.
19553 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19555 * config/aarch64/aarch64-tuning-flags.def (use_new_vector_costs):
19557 * config/aarch64/aarch64-protos.h (simd_vec_cost): Put comments
19558 above the fields rather than to the right.
19559 (simd_vec_cost::reduc_i8_cost): New member variable.
19560 (simd_vec_cost::reduc_i16_cost): Likewise.
19561 (simd_vec_cost::reduc_i32_cost): Likewise.
19562 (simd_vec_cost::reduc_i64_cost): Likewise.
19563 (simd_vec_cost::reduc_f16_cost): Likewise.
19564 (simd_vec_cost::reduc_f32_cost): Likewise.
19565 (simd_vec_cost::reduc_f64_cost): Likewise.
19566 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
19567 accordingly, using the vec_to_scalar_cost for the new fields.
19568 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
19569 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
19570 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
19571 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
19572 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
19573 (thunderx3t110_advsimd_vector_cost): Likewise.
19574 (aarch64_use_new_vector_costs_p): New function.
19575 (aarch64_simd_vec_costs): New function, split out from...
19576 (aarch64_builtin_vectorization_cost): ...here.
19577 (aarch64_is_reduction): New function.
19578 (aarch64_detect_vector_stmt_subtype): Likewise.
19579 (aarch64_add_stmt_cost): Call aarch64_detect_vector_stmt_subtype if
19580 using the new vector costs.
19582 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19585 * tree-emutls.c (get_emutls_init_templ_addr): Mark initializer of weak
19586 TLS declarations as public.
19588 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19590 * config/aarch64/aarch64-d.c (IN_TARGET_CODE): Define.
19591 * config/arm/arm-d.c (IN_TARGET_CODE): Likewise.
19592 * config/i386/i386-d.c (IN_TARGET_CODE): Likewise.
19593 * config/mips/mips-d.c (IN_TARGET_CODE): Likewise.
19594 * config/pa/pa-d.c (IN_TARGET_CODE): Likewise.
19595 * config/riscv/riscv-d.c (IN_TARGET_CODE): Likewise.
19596 * config/rs6000/rs6000-d.c (IN_TARGET_CODE): Likewise.
19597 * config/s390/s390-d.c (IN_TARGET_CODE): Likewise.
19598 * config/sparc/sparc-d.c (IN_TARGET_CODE): Likewise.
19600 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19603 * config.gcc (*-*-cygwin*): Add winnt-d.o
19604 (*-*-mingw*): Likewise.
19605 * config/i386/cygwin.h (EXTRA_TARGET_D_OS_VERSIONS): New macro.
19606 * config/i386/mingw32.h (EXTRA_TARGET_D_OS_VERSIONS): Likewise.
19607 * config/i386/t-cygming: Add winnt-d.o.
19608 * config/i386/winnt-d.c: New file.
19610 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19612 * config/freebsd-d.c: Include memmodel.h.
19614 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19617 * config.gcc (*-*-openbsd*): Add openbsd-d.o.
19618 * config/t-openbsd: Add openbsd-d.o.
19619 * config/openbsd-d.c: New file.
19621 2021-03-25 Stam Markianos-Wright <stam.markianos-wright@arm.com>
19623 PR tree-optimization/96974
19624 * tree-vect-stmts.c (vect_get_vector_types_for_stmt): Replace assert
19625 with graceful exit.
19627 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
19630 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
19634 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
19635 always_inline in system headers.
19637 2021-03-25 Kewen Lin <linkw@linux.ibm.com>
19639 * tree-vect-loop.c (vect_model_reduction_cost): Init inside_cost.
19641 2021-03-25 Jakub Jelinek <jakub@redhat.com>
19644 * tree-core.h (enum operand_equal_flag): Add OEP_ADDRESS_OF_SAME_FIELD.
19645 * fold-const.c (operand_compare::operand_equal_p): Don't compare
19646 field offsets if OEP_ADDRESS_OF_SAME_FIELD.
19648 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
19652 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
19653 always_inline in system headers.
19655 2021-03-25 Richard Biener <rguenther@suse.de>
19657 PR tree-optimization/99746
19658 * tree-vect-slp-patterns.c (complex_pattern::build): Do not mark
19659 the scalar stmt as patterned. Instead set up required things
19662 2021-03-25 Xionghu Luo <luoxhu@linux.ibm.com>
19664 * config/rs6000/rs6000.c (power8_costs): Change l2 cache
19667 2021-03-24 Martin Liska <mliska@suse.cz>
19670 * common/config/i386/i386-common.c (ARRAY_SIZE): Fix off-by-one
19672 * config/i386/i386-options.c (ix86_option_override_internal):
19673 Add run-time assert.
19675 2021-03-24 Martin Jambor <mjambor@suse.cz>
19678 * ipa-cp.c (initialize_node_lattices): Mark as bottom all
19679 parameters with unknown type.
19680 (ipacp_value_safe_for_type): New function.
19681 (propagate_vals_across_arith_jfunc): Verify that the constant type
19682 can be used for a type of the formal parameter.
19683 (propagate_vals_across_ancestor): Likewise.
19684 (propagate_scalar_across_jump_function): Likewise. Pass the type
19685 also to propagate_vals_across_ancestor.
19687 2021-03-24 Christophe Lyon <christophe.lyon@linaro.org>
19690 * config/arm/mve.md (movmisalign<mode>_mve_store): Use Ux
19692 (movmisalign<mode>_mve_load): Likewise.
19694 2021-03-24 Jakub Jelinek <jakub@redhat.com>
19697 * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
19698 movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.
19700 2021-03-24 Alexandre Oliva <oliva@adacore.com>
19702 * doc/sourcebuild.texi (sysconf): New effective target.
19704 2021-03-24 Alexandre Oliva <oliva@adacore.com>
19706 * config/i386/predicates.md (reg_or_const_vec_operand): New.
19707 * config/i386/sse.md (ssse3_pshufbv8qi3): Add an expander for
19708 the now *-prefixed insn_and_split, turn the splitter const vec
19709 into an input for the insn, making it an ignored immediate for
19710 non-split cases, and loaded into the scratch register
19713 2021-03-23 Vladimir N. Makarov <vmakarov@redhat.com>
19716 * config/aarch64/constraints.md (Utq, UOb, UOh, UOw, UOd, UOty):
19717 Use define_relaxed_memory_constraint for them.
19719 2021-03-23 Iain Sandoe <iain@sandoe.co.uk>
19722 * config/host-darwin.c (darwin_gt_pch_use_address): Add a
19723 colon to the diagnostic message.
19725 2021-03-23 Ilya Leoshkevich <iii@linux.ibm.com>
19727 * fwprop.c (fwprop_propagation::fwprop_propagation): Look at
19729 (try_fwprop_subst_note): Use set_info instead of insn_info.
19730 (try_fwprop_subst_pattern): Likewise.
19731 (try_fwprop_subst_notes): Likewise.
19732 (try_fwprop_subst): Likewise.
19733 (forward_propagate_subreg): Likewise.
19734 (forward_propagate_and_simplify): Likewise.
19735 (forward_propagate_into): Likewise.
19736 * rtl-ssa/accesses.h (set_info::single_nondebug_use) New
19738 (set_info::single_nondebug_insn_use): Likewise.
19739 (set_info::single_phi_use): Likewise.
19740 * rtl-ssa/member-fns.inl (set_info::single_nondebug_use) New
19742 (set_info::single_nondebug_insn_use): Likewise.
19743 (set_info::single_phi_use): Likewise.
19745 2021-03-23 Christophe Lyon <christophe.lyon@linaro.org>
19747 * doc/sourcebuild.texi (arm_dsp_ok, arm_dsp): Document.
19749 2021-03-23 Jakub Jelinek <jakub@redhat.com>
19752 * config/aarch64/aarch64.c (aarch64_add_offset): Tell
19753 expand_mult to perform an unsigned rather than a signed
19756 2021-03-23 H.J. Lu <hjl.tools@gmail.com>
19759 * config/i386/cpuid.h (__cpuid): Add __volatile__.
19760 (__cpuid_count): Likewise.
19762 2021-03-23 Richard Biener <rguenther@suse.de>
19764 PR tree-optimization/99721
19765 * tree-vect-slp.c (vect_slp_analyze_node_operations):
19766 Make sure we can schedule the node.
19768 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
19770 * config/riscv/riscv.c (riscv_subword): Take endianness into
19771 account when calculating the byte offset.
19773 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
19775 * config/riscv/predicates.md (subreg_lowpart_operator): New predicate
19776 * config/riscv/riscv.md (*addsi3_extended2, *subsi3_extended2)
19777 (*negsi2_extended2, *mulsi3_extended2, *<optab>si3_mask)
19778 (*<optab>si3_mask_1, *<optab>di3_mask, *<optab>di3_mask_1)
19779 (*<optab>si3_extend_mask, *<optab>si3_extend_mask_1): Use
19780 new predicate "subreg_lowpart_operator"
19782 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
19784 * config/riscv/riscv.c (riscv_swap_instruction): New function
19785 to byteswap an SImode rtx containing an instruction.
19786 (riscv_trampoline_init): Byteswap the generated instructions
19789 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
19791 * common/config/riscv/riscv-common.c
19792 (TARGET_DEFAULT_TARGET_FLAGS): Set default endianness.
19793 * config.gcc (riscv32be-*, riscv64be-*): Set
19794 TARGET_BIG_ENDIAN_DEFAULT to 1.
19795 * config/riscv/elf.h (LINK_SPEC): Change -melf* value
19796 depending on default endianness.
19797 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
19798 * config/riscv/linux.h (LINK_SPEC): Likewise.
19799 * config/riscv/riscv.c (TARGET_DEFAULT_TARGET_FLAGS): Set
19800 default endianness.
19801 * config/riscv/riscv.h (DEFAULT_ENDIAN_SPEC): New macro.
19803 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
19805 * config/riscv/elf.h (LINK_SPEC): Pass linker endianness flag.
19806 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
19807 * config/riscv/linux.h (LINK_SPEC): Likewise.
19808 * config/riscv/riscv.h (ASM_SPEC): Pass -mbig-endian and
19810 (BYTES_BIG_ENDIAN): Handle big endian.
19811 (WORDS_BIG_ENDIAN): Define to BYTES_BIG_ENDIAN.
19812 * config/riscv/riscv.opt (-mbig-endian, -mlittle-endian): New
19814 * doc/invoke.texi (-mbig-endian, -mlittle-endian): Document.
19816 2021-03-23 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
19818 * regcprop.c (find_oldest_value_reg): Ask target whether
19819 different mode is fine for replacement register.
19821 2021-03-23 Aldy Hernandez <aldyh@redhat.com>
19823 PR tree-optimization/99296
19824 * value-range.cc (irange::irange_set_1bit_anti_range): New.
19825 (irange::irange_set_anti_range): Call irange_set_1bit_anti_range
19826 * value-range.h (irange::irange_set_1bit_anti_range): New.
19828 2021-03-22 Vladimir N. Makarov <vmakarov@redhat.com>
19831 * config/aarch64/constraints.md (UtQ): Use
19832 define_relaxed_memory_constraint for it.
19833 * doc/md.texi (define_relaxed_memory_constraint): Describe it.
19834 * genoutput.c (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
19835 * genpreds.c (constraint_data): Add bitfield is_relaxed_memory.
19836 (have_relaxed_memory_constraints): New static var.
19837 (relaxed_memory_start, relaxed_memory_end): Ditto.
19838 (add_constraint): Add arg is_relaxed_memory. Check name for
19839 relaxed memory. Set up is_relaxed_memory in constraint_data and
19840 have_relaxed_memory_constraints. Adjust calls.
19841 (choose_enum_order): Process relaxed memory.
19842 (write_tm_preds_h): Ditto.
19843 (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
19844 * gensupport.c (process_rtx): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
19845 * ira-costs.c (record_reg_classes): Process CT_RELAXED_MEMORY.
19846 * ira-lives.c (single_reg_class): Use
19847 insn_extra_relaxed_memory_constraint.
19848 * ira.c (ira_setup_alts): CT_RELAXED_MEMORY.
19849 * lra-constraints.c (valid_address_p): Use
19850 insn_extra_relaxed_memory_constraint instead of other memory
19852 (process_alt_operands): Process CT_RELAXED_MEMORY.
19853 (curr_insn_transform): Use insn_extra_relaxed_memory_constraint.
19854 * recog.c (asm_operand_ok, preprocess_constraints): Process
19856 * reload.c (find_reloads): Ditto.
19857 * rtl.def (DEFINE_RELAXED_MEMORY_CONSTRAINT): New.
19858 * stmt.c (parse_input_constraint): Use
19859 insn_extra_relaxed_memory_constraint.
19861 2021-03-22 Segher Boessenkool <segher@kernel.crashing.org>
19864 * ubsan.c (ubsan_instrument_float_cast): Don't test for unordered if
19867 2021-03-22 Alex Coplan <alex.coplan@arm.com>
19870 * config/arm/arm-protos.h (neon_make_constant): Add generate
19871 argument to guard emitting insns, default to true.
19872 * config/arm/arm.c (arm_legitimate_constant_p_1): Reject
19873 CONST_VECTORs which neon_make_constant can't handle.
19874 (neon_vdup_constant): Add generate argument, avoid emitting
19875 insns if it's not set.
19876 (neon_make_constant): Plumb new generate argument through.
19877 * config/arm/constraints.md (Ui): New. Use it...
19878 * config/arm/mve.md (*mve_mov<mode>): ... here.
19879 * config/arm/vec-common.md (movv8hf): Use neon_make_constant to
19880 synthesize constants.
19882 2021-03-22 Richard Biener <rguenther@suse.de>
19884 * debug.h: Add deprecation warning.
19886 2021-03-22 Richard Biener <rguenther@suse.de>
19888 PR tree-optimization/99694
19889 * tree-ssa-sccvn.c (visit_phi): Ignore edges with the
19892 2021-03-22 Kito Cheng <kito.cheng@sifive.com>
19895 * config/riscv/riscv.c (riscv_expand_block_move): Get RTL value
19896 after type checking.
19898 2021-03-22 Jakub Jelinek <jakub@redhat.com>
19902 * dwarf2out.c (get_full_len): Use get_precision rather than
19904 (add_const_value_attribute): Make sure add_AT_wide argument has
19905 precision prec rather than some very wide one.
19907 2021-03-22 Kewen Lin <linkw@linux.ibm.com>
19909 * config/rs6000/rs6000.md (*rotldi3_insert_sf,
19910 *mov<SFDF:mode><SFDF2:mode>cc_p9, floatsi<mode>2_lfiwax,
19911 floatsi<mode>2_lfiwax_mem, floatunssi<mode>2_lfiwzx,
19912 floatunssi<mode>2_lfiwzx_mem, *floatsidf2_internal,
19913 *floatunssidf2_internal, fix_trunc<mode>si2_stfiwx,
19914 fix_trunc<mode>si2_internal, fixuns_trunc<mode>si2_stfiwx,
19915 *round32<mode>2_fprs, *roundu32<mode>2_fprs,
19916 *fix_trunc<mode>si2_internal): Fix empty split condition.
19917 * config/rs6000/vsx.md (*vsx_le_undo_permute_<mode>,
19918 vsx_reduc_<VEC_reduc_name>_v2df, vsx_reduc_<VEC_reduc_name>_v4sf,
19919 *vsx_reduc_<VEC_reduc_name>_v2df_scalar,
19920 *vsx_reduc_<VEC_reduc_name>_v4sf_scalar): Likewise.
19922 2021-03-22 Xionghu Luo <luoxhu@linux.ibm.com>
19925 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
19926 Convert idx to DImode.
19927 (rs6000_expand_vector_set_var_p8): Likewise.
19929 2021-03-21 Jakub Jelinek <jakub@redhat.com>
19932 * dwarf2out.c (insert_float): Change return type from void to
19933 unsigned, handle GET_MODE_SIZE (mode) == 2 and return element size.
19934 (mem_loc_descriptor, loc_descriptor, add_const_value_attribute):
19937 2021-03-20 H.J. Lu <hjl.tools@gmail.com>
19940 * config/i386/i386.c (construct_container): Check cfun != NULL
19941 before accessing silent_p.
19943 2021-03-20 Ahamed Husni <ahamedhusni73@gmail.com>
19945 * asan.c: Fix typos in comments.
19947 2021-03-20 Vladimir N. Makarov <vmakarov@redhat.com>
19949 PR rtl-optimization/99680
19950 * lra-constraints.c (skip_contraint_modifiers): Rename to skip_constraint_modifiers.
19951 (process_address_1): Check empty constraint before using
19954 2021-03-19 Pat Haugen <pthaugen@linux.ibm.com>
19956 * config/rs6000/rs6000.c (power10_cost): New.
19957 (rs6000_option_override_internal): Set Power10 costs.
19958 (rs6000_issue_rate): Set Power10 issue rate.
19959 * config/rs6000/power10.md: Rewrite for Power10.
19961 2021-03-19 Vladimir N. Makarov <vmakarov@redhat.com>
19964 * lra-constraints.c (process_address_1): Don't use unknown
19965 constraint for address constraint.
19967 2021-03-19 Iain Sandoe <iain@sandoe.co.uk>
19970 * config.gcc (powerpc-*-darwin8): Delete the reference to
19971 the now removed darwin8.h.
19973 2021-03-19 Olivier Hainque <hainque@adacore.com>
19976 * config/vxworksae.h (VX_CPU_PREFIX): Define.
19978 2021-03-19 John David Anglin <danglin@gcc.gnu.org>
19980 * config/pa/pa.c (import_milli): Use memcpy instead of strncpy.
19982 2021-03-19 Tamar Christina <tamar.christina@arm.com>
19984 PR tree-optimization/99656
19985 * tree-vect-slp-patterns.c (linear_loads_p,
19986 complex_add_pattern::matches, is_eq_or_top,
19987 vect_validate_multiplication, complex_mul_pattern::matches,
19988 complex_fms_pattern::matches): Remove complex_perm_kinds_t.
19989 * tree-vectorizer.h: (complex_load_perm_t): Removed.
19990 (slp_tree_to_load_perm_map_t): Use complex_perm_kinds_t instead of
19991 complex_load_perm_t.
19993 2021-03-19 H.J. Lu <hjl.tools@gmail.com>
19996 * config/i386/i386-options.c (ix86_init_machine_status): Set
19998 * config/i386/i386.c (init_cumulative_args): Set silent_p to
20000 (construct_container): Return early for return and argument
20001 errors if silent_p is true.
20002 * config/i386/i386.h (machine_function): Add silent_p.
20004 2021-03-19 Jakub Jelinek <jakub@redhat.com>
20007 * config/arm/constraints.md (Ds): New constraint.
20008 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): Use w,Ds
20009 constraint instead of w,Dm.
20011 2021-03-19 Andrew Stubbs <ams@codesourcery.com>
20013 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Fix quotes
20016 2021-03-19 Eric Botcazou <ebotcazou@adacore.com>
20018 PR middle-end/99641
20019 * fold-const.c (native_encode_initializer) <CONSTRUCTOR>: For an
20020 array type, do the computation of the current position in sizetype.
20022 2021-03-18 Vladimir N. Makarov <vmakarov@redhat.com>
20025 * lra-constraints.c (process_address_1): Use lookup_constraint
20026 only for a single constraint.
20028 2021-03-18 Martin Sebor <msebor@redhat.com>
20030 PR middle-end/99502
20031 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): Rename...
20032 (inbounds_memaccess_p): ...to this. Check the ending offset of
20033 the accessed member.
20035 2021-03-18 Andrew Stubbs <ams@codesourcery.com>
20037 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Add %< and
20038 %> quote markers to error messages.
20039 (gcn_goacc_validate_dims): Likewise.
20040 (gcn_conditional_register_usage): Remove exclaimation mark from error
20042 (gcn_vectorize_vec_perm_const): Ensure perm is fully uninitialized.
20044 2021-03-18 Jan Hubicka <hubicka@ucw.cz>
20046 * config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
20049 2021-03-18 Sinan Lin <sinan@isrc.iscas.ac.cn>
20050 Kito Cheng <kito.cheng@sifive.com>
20052 * config/riscv/riscv.c (riscv_block_move_straight): Change type
20053 to unsigned HOST_WIDE_INT for parameter and local variable with
20054 HOST_WIDE_INT type.
20055 (riscv_adjust_block_mem): Ditto.
20056 (riscv_block_move_loop): Ditto.
20057 (riscv_expand_block_move): Ditto.
20059 2021-03-18 Nick Clifton <nickc@redhat.com>
20061 * config/v850/v850.c (construct_restore_jr): Increase static
20063 (construct_save_jarl): Likewise.
20064 * config/v850/v850.h (DWARF2_DEBUGGING_INFO): Define.
20066 2021-03-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20068 * config/aarch64/aarch64.c (aarch64_adjust_generic_arch_tuning): Define.
20069 (aarch64_override_options_internal): Use it.
20070 (generic_tunings): Add AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS to
20073 2021-03-17 Sandra Loosemore <sandra@codesourcery.com>
20075 * config/nios2/nios2.c (nios2_custom_check_insns): Clean up
20076 error message format issues.
20077 (nios2_option_override): Likewise.
20078 (nios2_expand_fpu_builtin): Likewise.
20079 (nios2_init_custom_builtins): Adjust to avoid bogus strncpy
20080 truncation warning.
20081 (nios2_expand_custom_builtin): More error message format fixes.
20082 (nios2_expand_rdwrctl_builtin): Likewise.
20083 (nios2_expand_rdprs_builtin): Likewise.
20084 (nios2_expand_eni_builtin): Likewise.
20085 (nios2_expand_builtin): Likewise.
20086 (nios2_register_custom_code): Likewise.
20087 (nios2_valid_target_attribute_rec): Likewise.
20088 (nios2_add_insn_asm): Fix uninitialized variable warning.
20090 2021-03-17 Jan Hubicka <jh@suse.cz>
20092 * config/i386/x86-tune-costs.h (struct processor_costs): Update costs
20093 of gather to match reality.
20094 * config/i386/x86-tune.def (X86_TUNE_USE_GATHER): Enable for znver3.
20096 2021-03-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20098 * config/aarch64/aarch64-builtins.c (aarch64_expand_rng_builtin): Use EQ
20099 to compare against CC_REG rather than NE.
20101 2021-03-17 H.J. Lu <hjl.tools@gmail.com>
20104 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support
20105 inline assembly statements.
20106 (ix86_print_operand): Update 'P' handling for -fno-plt.
20108 2021-03-17 Tamar Christina <tamar.christina@arm.com>
20111 * config/aarch64/aarch64.c
20112 (aarch64_simd_clone_compute_vecsize_and_simdlen): Remove unused var.
20114 2021-03-16 Segher Boessenkool <segher@kernel.crashing.org>
20117 * config/rs6000/predicates.md (branch_comparison_operator): Allow
20118 ordered and unordered for CCFPmode, if flag_finite_math_only.
20120 2021-03-16 Jakub Jelinek <jakub@redhat.com>
20123 * config/i386/i386-expand.c (ix86_split_lea_for_addr): Emit a MULT
20124 rather than ASHIFT.
20125 * config/i386/i386.md (mult by 1248 into ashift): New splitter.
20127 2021-03-16 Martin Liska <mliska@suse.cz>
20130 * optc-save-gen.awk: Add flag_ipa_ra to exceptions for
20131 cl_optimization_compare function.
20133 2021-03-16 Ilya Leoshkevich <iii@linux.ibm.com>
20135 * config/s390/s390.c (f_constraint_p): Treat "fv" constraints
20138 2021-03-16 Jakub Jelinek <jakub@redhat.com>
20141 * config/i386/i386.h (struct machine_function): Add
20142 has_explicit_vzeroupper bitfield.
20143 * config/i386/i386-expand.c (ix86_expand_builtin): Set
20144 cfun->machine->has_explicit_vzeroupper when expanding
20145 IX86_BUILTIN_VZEROUPPER.
20146 * config/i386/i386-features.c (rest_of_handle_insert_vzeroupper):
20147 Do the mode switching only when TARGET_VZEROUPPER, expensive
20148 optimizations turned on and not optimizing for size.
20149 (pass_insert_vzeroupper::gate): Enable even when
20150 cfun->machine->has_explicit_vzeroupper is set.
20152 2021-03-16 Jakub Jelinek <jakub@redhat.com>
20155 * config/aarch64/aarch64.c
20156 (aarch64_simd_clone_compute_vecsize_and_simdlen): If not a function
20157 definition, walk TYPE_ARG_TYPES list if non-NULL for argument types
20158 instead of DECL_ARGUMENTS. Ignore types for uniform arguments.
20160 2021-03-15 Richard Biener <rguenther@suse.de>
20162 PR tree-optimization/98834
20163 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle missing
20164 subsetting by truncating the access size.
20166 2021-03-15 Jan Hubicka <hubicka@ucw.cz>
20168 * config/i386/i386-options.c (processor_cost_table): Add znver3_cost.
20169 * config/i386/x86-tune-costs.h (znver3_cost): New gobal variable; copy
20172 2021-03-15 Martin Liska <mliska@suse.cz>
20174 * spellcheck.c: Add missing comma in initialization.
20176 2021-03-14 Uroš Bizjak <ubizjak@gmail.com>
20178 * config/i386/sse.md (*vec_extract<mode>): Merge alternative 0 with
20179 alternative 2 and alternative 1 with alternative 3 using
20180 YW register constraint.
20181 (*vec_extract<PEXTR_MODE12:mode>_zext): Merge alternatives
20182 using YW register constraint.
20183 (*vec_extractv16qi_zext): Ditto.
20184 (*vec_extractv4si): Merge alternatives 4 and 5
20185 using Yw register constraint.
20186 (*ssse3_palignr<mode>_perm): Use Yw instead of v for alternative 3.
20188 2021-03-13 Martin Sebor <msebor@redhat.com>
20190 PR tree-optimization/99489
20191 * builtins.c (gimple_call_alloc_size): Fail gracefully when argument
20192 is not a call statement.
20194 2021-03-13 Jakub Jelinek <jakub@redhat.com>
20196 PR tree-optimization/99544
20197 * match.pd (X + (X << C) -> X * (1 + (1 << C))): Don't simplify
20198 if for vector types multiplication can't be done in type's mode.
20200 2021-03-12 Eric Botcazou <ebotcazou@adacore.com>
20203 * config/sparc/constraints.md (w): Rename to...
20204 (W): ... this and ditch previous implementation.
20205 * config/sparc/sparc.md (*movdi_insn_sp64): Replace W with m.
20206 (*movdf_insn_sp64): Likewise.
20207 (*mov<VM64:mode>_insn_sp64): Likewise.
20208 * config/sparc/sync.md (*atomic_compare_and_swap<mode>_1): Replace
20210 (atomic_compare_and_swap_leon3_1): Likewise.
20211 (*atomic_compare_and_swapdi_v8plus): Likewise.
20212 * config/sparc/sparc.c (memory_ok_for_ldd): Remove useless test on
20213 architecture and add missing address validity check during LRA.
20215 2021-03-12 Tobias Burnus <tobias@codesourcery.com>
20218 * gimplify.c (omp_add_variable): Handle NULL_TREE as size
20219 occuring for assumed-size arrays in use_device_{ptr,addr}.
20221 2021-03-12 Jakub Jelinek <jakub@redhat.com>
20224 * config/i386/constraints.md (YW): New internal constraint.
20225 * config/i386/sse.md (v_Yw): Add V4TI, V2TI, V1TI and TI cases.
20226 (*<sse2_avx2>_<insn><mode>3<mask_name>,
20227 *<sse2_avx2>_uavg<mode>3<mask_name>, *abs<mode>2,
20228 *<s>mul<mode>3_highpart<mask_name>): Use <v_Yw> instead of v in
20230 (<sse2_avx2>_psadbw): Use YW instead of v in constraints.
20231 (*avx2_pmaddwd, *sse2_pmaddwd, *<code>v8hi3, *<code>v16qi3,
20232 avx2_pmaddubsw256, ssse3_pmaddubsw128): Merge last two alternatives
20233 into one, use Yw instead of former x,v.
20234 (ashr<mode>3, <insn><mode>3): Use <v_Yw> instead of x in constraints of
20235 the last alternative.
20236 (<sse2_avx2>_packsswb<mask_name>, <sse2_avx2>_packssdw<mask_name>,
20237 <sse2_avx2>_packuswb<mask_name>, <sse4_1_avx2>_packusdw<mask_name>,
20238 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, <ssse3_avx2>_palignr<mode>,
20239 <ssse3_avx2>_pshufb<mode>3<mask_name>): Merge last two alternatives
20240 into one, use <v_Yw> instead of former x,v.
20241 (avx2_interleave_highv32qi<mask_name>,
20242 vec_interleave_highv16qi<mask_name>): Use Yw instead of v in
20243 constraints. Add && <mask_avx512bw_condition> to condition.
20244 (avx2_interleave_lowv32qi<mask_name>,
20245 vec_interleave_lowv16qi<mask_name>,
20246 avx2_interleave_highv16hi<mask_name>,
20247 vec_interleave_highv8hi<mask_name>,
20248 avx2_interleave_lowv16hi<mask_name>, vec_interleave_lowv8hi<mask_name>,
20249 avx2_pshuflw_1<mask_name>, sse2_pshuflw_1<mask_name>,
20250 avx2_pshufhw_1<mask_name>, sse2_pshufhw_1<mask_name>,
20251 avx2_<code>v16qiv16hi2<mask_name>, sse4_1_<code>v8qiv8hi2<mask_name>,
20252 *sse4_1_<code>v8qiv8hi2<mask_name>_1, <sse2_avx2>_<insn><mode>3): Use
20253 Yw instead of v in constraints.
20254 * config/i386/mmx.md (Yv_Yw): New define_mode_attr.
20255 (*mmx_<insn><mode>3, mmx_ashr<mode>3, mmx_<insn><mode>3): Use <Yv_Yw>
20256 instead of Yv in constraints.
20257 (*mmx_<insn><mode>3, *mmx_mulv4hi3, *mmx_smulv4hi3_highpart,
20258 *mmx_umulv4hi3_highpart, *mmx_pmaddwd, *mmx_<code>v4hi3,
20259 *mmx_<code>v8qi3, mmx_pack<s_trunsuffix>swb, mmx_packssdw,
20260 mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
20261 *mmx_uavgv8qi3, *mmx_uavgv4hi3, mmx_psadbw): Use Yw instead of Yv in
20263 (*mmx_pinsrw, *mmx_pinsrb, *mmx_pextrw, *mmx_pextrw_zext, *mmx_pextrb,
20264 *mmx_pextrb_zext): Use YW instead of Yv in constraints.
20265 (*mmx_eq<mode>3, mmx_gt<mode>3): Use x instead of Yv in constraints.
20266 (mmx_andnot<mode>3, *mmx_<code><mode>3): Split last alternative into
20267 two, one with just x, another isa avx512vl with v.
20269 2021-03-12 Martin Liska <mliska@suse.cz>
20271 * doc/invoke.texi: Add missing param documentation.
20273 2021-03-11 David Malcolm <dmalcolm@redhat.com>
20276 * Makefile.in (ANALYZER_OBJS): Add analyzer/feasible-graph.o and
20277 analyzer/trimmed-graph.o.
20278 * doc/analyzer.texi (Analyzer Paths): Rewrite description of
20279 feasibility checking to reflect new implementation.
20280 * doc/invoke.texi (-fdump-analyzer-feasibility): Document new
20282 * shortest-paths.h (shortest_paths::get_shortest_distance): New.
20284 2021-03-11 David Malcolm <dmalcolm@redhat.com>
20286 * digraph.cc (selftest::test_shortest_paths): Update
20287 shortest_paths init for new param. Add test of
20288 SPS_TO_GIVEN_TARGET.
20289 * shortest-paths.h (enum shortest_path_sense): New.
20290 (shortest_paths::shortest_paths): Add "sense" param.
20291 Update for renamings. Generalize to use "sense" param.
20292 (shortest_paths::get_shortest_path): Rename param.
20293 (shortest_paths::m_sense): New field.
20294 (shortest_paths::m_prev): Rename...
20295 (shortest_paths::m_best_edge): ...to this.
20296 (shortest_paths::get_shortest_path): Update for renamings.
20297 Conditionalize flipping of path on sense of traversal.
20299 2021-03-11 David Malcolm <dmalcolm@redhat.com>
20301 * digraph.cc (selftest::test_shortest_paths): Add test coverage
20302 for paths from B and C.
20303 * shortest-paths.h (shortest_paths::shortest_paths): Handle
20304 unreachable nodes, rather than asserting.
20306 2021-03-11 David Edelsohn <dje.gcc@gmail.com>
20309 * config/rs6000/rs6000.c (rs6000_xcoff_file_start): Don't create
20310 xcoff_tbss_section_name.
20311 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .lcomm.
20312 * xcoffout.c (xcoff_tbss_section_name): Delete.
20313 * xcoffout.h (xcoff_tbss_section_name): Delete.
20315 2021-03-11 Richard Biener <rguenther@suse.de>
20317 PR tree-optimization/99523
20318 * tree-cfg.c (dump_function_to_file): Dump SSA names
20319 w/o identifier to the decls section as well, not only those
20320 without a VAR_DECL.
20322 2021-03-11 Jakub Jelinek <jakub@redhat.com>
20325 * ipa-icf-gimple.c (func_checker::compare_gimple_call): For internal
20326 function calls with lhs fail if the lhs don't have compatible types.
20328 2021-03-11 Hans-Peter Nilsson <hp@axis.com>
20330 * config/cris/cris.h (HARD_FRAME_POINTER_REGNUM): Define.
20331 Change FRAME_POINTER_REGNUM to correspond to a new faked
20332 register faked_fp, part of GENNONACR_REGS like faked_ap.
20333 (CRIS_FAKED_REGS_CONTENTS): New helper macro.
20334 (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS, CALL_USED_REGISTERS):
20335 (REG_ALLOC_ORDER, REG_CLASS_CONTENTS, REGNO_OK_FOR_BASE_P)
20336 (ELIMINABLE_REGS, REGISTER_NAMES): Adjust accordingly.
20337 * config/cris/cris.md (CRIS_FP_REGNUM): Renumber to new faked
20339 (CRIS_REAL_FP_REGNUM): New constant.
20340 * config/cris/cris.c (cris_reg_saved_in_regsave_area): Check
20341 for HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM.
20342 (cris_initial_elimination_offset): Handle elimination changes
20343 to HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM
20344 and add one from FRAME_POINTER_REGNUM to
20345 HARD_FRAME_POINTER_REGNUM.
20346 (cris_expand_prologue, cris_expand_epilogue): Emit code for
20347 hard_frame_pointer_rtx instead of frame_pointer_rtx.
20349 2021-03-10 David Edelsohn <dje.gcc@gmail.com>
20352 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Add check for DCmode.
20353 * config/rs6000/rs6000.c (rs6000_special_round_type_align): Same.
20355 2021-03-10 Vladimir N. Makarov <vmakarov@redhat.com>
20358 * lra-constraints.c (process_address_1): Don't check unknown
20359 constraint, use X for empty constraint.
20361 2021-03-10 Alex Coplan <alex.coplan@arm.com>
20363 * config/aarch64/aarch64.c (aarch64_vfp_is_call_or_return_candidate):
20364 Fix typo in comment describing "is_ha" argument.
20366 2021-03-10 John David Anglin <danglin@gcc.gnu.org>
20368 * doc/sourcebuild.texi: Document LRA target selector.
20370 2021-03-10 David Malcolm <dmalcolm@redhat.com>
20372 * doc/ux.texi: Add subsection contrasting interactive versus
20373 batch usage of GCC.
20375 2021-03-10 Joel Hutton <joel.hutton@arm.com>
20378 * tree-vect-stmts.c (vectorizable_store): Fix scatter store mask
20380 (vectorizable_load): Fix gather load mask check condition.
20382 2021-03-10 Richard Biener <rguenther@suse.de>
20384 PR tree-optimization/99510
20385 * tree.c (check_aligned_type): Check that the candidate
20386 has TYPE_USER_ALIGN set instead of matching with the
20389 2021-03-10 Eric Botcazou <ebotcazou@adacore.com>
20391 * config/sparc/sparc.c (sparc_regmode_natural_size): Return 4 for
20392 float and vector integer modes only if the mode is not larger.
20394 2021-03-10 Hans-Peter Nilsson <hp@axis.com>
20396 * config/cris/cris.h (DWARF_FRAME_REGISTERS): Define.
20398 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
20400 * ira.c (ira_setup_alts, ira_get_dup_out_num): Process digital
20402 * ira-lives.c (single_reg_class): Ditto.
20404 2021-03-09 Sebastian Huber <sebastian.huber@embedded-brains.de>
20406 * config.gcc (aarch64-*-rtems*): Include general rtems.h after
20407 the architecture-specific rtems.h.
20408 (aarch64-*-rtems*): Likewise.
20409 (arm*-*-rtems*): Likewise.
20410 (epiphany-*-rtems*): Likewise.
20411 (riscv*-*-rtems*): Likewise.
20413 2021-03-09 Jakub Jelinek <jakub@redhat.com>
20415 PR tree-optimization/99305
20416 * tree-ssa-phiopt.c (conditional_replacement): Test integer_pow2p
20417 before integer_all_onesp instead of vice versa.
20419 2021-03-09 Richard Earnshaw <rearnsha@arm.com>
20421 * common/config/arm/arm-common.c (arm_config_default): Change type
20422 of 'i' to unsigned.
20424 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
20427 * lra-constraints.c (process_address_1): Process constraint 'g'
20428 separately and digital constraints containing more one digit.
20430 2021-03-09 Nick Clifton <nickc@redhat.com>
20432 * config/rx/rx.h (DBX_DEBUGGING_INFO): Define.
20433 (DWARF"_DEBUGGING_INFO): Define.
20435 2021-03-09 Eric Botcazou <ebotcazou@adacore.com>
20438 * calls.c (initialize_argument_information): When the argument
20439 is passed by reference, do not make a copy in a thunk only if
20440 the argument is already in memory. Remove redundant test for
20441 the case of callee copy.
20443 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
20446 * lra-constraints.c (process_address_1): Process 0..9 constraints
20447 in process_address_1.
20449 2021-03-09 Andreas Krebbel <krebbel@linux.ibm.com>
20451 * config/s390/s390.c (struct s390_processor processor_table):
20452 Binutils name string must not be empty.
20454 2021-03-09 Claudiu Zissulescu <claziss@synopsys.com>
20456 * config/arc/arc.c (arc_attr_type): Remove function.
20458 2021-03-09 Martin Liska <mliska@suse.cz>
20461 * config/i386/i386-options.c (ix86_option_override_internal):
20462 Set isa_flags for OPTS argument and not for the global
20465 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
20467 * config/rs6000/predicates.md (ds_form_mem_operand): Check
20470 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
20473 * config/rs6000/predicates.md (ds_form_mem_operand) New
20475 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10) Use
20476 ds_form_mem_operand in ld/lwa patterns.
20477 * config/rs6000/fusion.md: Regenerate file.
20479 2021-03-08 Martin Sebor <msebor@redhat.com>
20481 PR middle-end/98266
20482 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): New function.
20483 (array_bounds_checker::check_array_bounds): Call it.
20485 2021-03-08 Martin Sebor <msebor@redhat.com>
20487 PR middle-end/97631
20488 * tree-ssa-strlen.c (maybe_warn_overflow): Test rawmem.
20489 (handle_builtin_stxncpy_strncat): Rename locals. Determine
20490 destination size from allocation calls. Issue a more appropriate
20492 (handle_builtin_memcpy): Pass true as rawmem to maybe_warn_overflow.
20493 (handle_builtin_memset): Same.
20495 2021-03-08 Peter Bergner <bergner@linux.ibm.com>
20498 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_permute): Add an assert
20499 to ensure we do not have an Altivec style address.
20500 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): Disable if passed
20501 an Altivec style address.
20502 (*vsx_le_perm_store_<mode>): Likewise.
20503 (splitters after *vsx_le_perm_store_<mode>): Likewise.
20504 (vsx_load_<mode>): Disable special expander if passed an Altivec
20506 (vsx_store_<mode>): Likewise.
20508 2021-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20511 * config/aarch64/predicates.md (aarch64_simd_shift_imm_vec_qi): Define.
20512 (aarch64_simd_shift_imm_vec_hi): Likewise.
20513 (aarch64_simd_shift_imm_vec_si): Likewise.
20514 (aarch64_simd_shift_imm_vec_di): Likewise.
20515 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le): Use
20516 predicate from above.
20517 (aarch64_shrn<mode>_insn_be): Likewise.
20518 (aarch64_rshrn<mode>_insn_le): Likewise.
20519 (aarch64_rshrn<mode>_insn_be): Likewise.
20520 (aarch64_shrn2<mode>_insn_le): Likewise.
20521 (aarch64_shrn2<mode>_insn_be): Likewise.
20522 (aarch64_rshrn2<mode>_insn_le): Likewise.
20523 (aarch64_rshrn2<mode>_insn_be): Likewise.
20525 2021-03-08 Vladimir N. Makarov <vmakarov@redhat.com>
20528 * lra-constraints.c (skip_contraint_modifiers): New function.
20529 (process_address_1): Use it before lookup_constraint call.
20531 2021-03-08 Martin Liska <mliska@suse.cz>
20534 * config/i386/i386-options.c (ix86_option_override_internal):
20535 Enable UINTR and HRESET for -march that supports it.
20537 2021-03-08 Ilya Leoshkevich <iii@linux.ibm.com>
20539 * config/s390/s390.c (f_constraint_p): New function.
20540 (s390_md_asm_adjust): Implement TARGET_MD_ASM_ADJUST.
20541 (TARGET_MD_ASM_ADJUST): Likewise.
20543 2021-03-08 Tobias Burnus <tobias@codesourcery.com>
20546 * tree-nested.c (convert_local_reference_stmt): Avoid calling
20547 lookup_field_for_decl for Fortran module (= namespace context).
20549 2021-03-08 Andreas Krebbel <krebbel@linux.ibm.com>
20551 * config/s390/s390.c (s390_expand_vec_compare): Implement <0
20552 comparison with arithmetic right shift.
20553 (s390_expand_vcond): No need for a force_reg anymore.
20554 s390_vec_compare will do it.
20555 * config/s390/vector.md ("vec_cmp<mode><tointvec>"): Accept also
20556 immediate operands.
20558 2021-03-07 Jakub Jelinek <jakub@redhat.com>
20561 * config/i386/constraints.md (Yw): Use SSE_REGS if TARGET_SSE
20562 but TARGET_AVX512BW or TARGET_AVX512VL is not set. Adjust description
20564 * config/i386/sse.md (v_Yw): New define_mode_attr.
20565 (*<insn><mode>3, *mul<mode>3<mask_name>, *avx2_<code><mode>3,
20566 *sse4_1_<code><mode>3<mask_name>): Use <v_Yw> instead of v
20568 * config/i386/mmx.md (mmx_pshufw_1, *vec_dupv4hi): Use Yw instead of
20569 xYw in constraints.
20571 2021-03-06 Julian Brown <julian@codesourcery.com>
20573 * tree-pretty-print.c (dump_generic_node): Emit non-generic
20574 address space info for aggregates.
20576 2021-03-06 Hans-Peter Nilsson <hp@axis.com>
20578 * config/cris/cris.h (MAX_FIXED_MODE_SIZE): Don't define.
20580 2021-03-05 Jakub Jelinek <jakub@redhat.com>
20582 PR middle-end/99322
20583 * tree-cfg.c (bb_to_omp_idx): New variable.
20584 (execute_build_cfg): Release the bb_to_omp_idx vector after
20585 cleanup_tree_cfg returns.
20586 (handle_abnormal_edges): Remove bb_to_omp_idx argument, adjust
20587 for bb_to_omp_idx being a vec<int> instead of pointer to array
20589 (make_edges): Remove bb_to_omp_idx local variable, don't pass
20590 it to handle_abnormal_edges, adjust for bb_to_omp_idx being a
20591 vec<int> instead of pointer to array of ints and don't free/release
20593 (remove_bb): When removing a bb and placing forced label somewhere
20594 else, ensure it is put into the same OpenMP region during cfg
20595 pass if possible or to entry successor as fallback. Unregister
20596 bb from bb_to_omp_idx.
20598 2021-03-05 Vladimir N. Makarov <vmakarov@redhat.com>
20601 * lra-constraints.c (process_address_1): Skip decomposing address
20602 for asm insn operand with unknown constraint.
20604 2021-03-05 Martin Jambor <mjambor@suse.cz>
20607 * cgraph.c (cgraph_edge::set_call_stmt): Do not update all
20608 corresponding speculative edges if we are about to resolve
20609 sepculation. Make edge direct (and so resolve speculations) before
20610 removing it from call_site_hash.
20611 (cgraph_edge::make_direct): Relax the initial assert to allow calling
20612 the function on speculative direct edges.
20614 2021-03-05 Eric Botcazou <ebotcazou@adacore.com>
20616 PR rtl-optimization/99376
20617 * rtlanal.c (nonzero_bits1) <arithmetic operators>: If the number
20618 of low-order zero bits is too large, set the result to 0 directly.
20620 2021-03-04 Jakub Jelinek <jakub@redhat.com>
20622 PR middle-end/93235
20623 * expmed.c (store_bit_field_using_insv): Return false of xop0 is a
20624 SUBREG and a SUBREG to op_mode can't be created.
20626 2021-03-04 Alex Coplan <alex.coplan@arm.com>
20629 * config/aarch64/aarch64-sve-builtins.cc
20630 (function_resolver::require_vector_type): Handle error_mark_node.
20632 2021-03-04 Ilya Leoshkevich <iii@linux.ibm.com>
20634 * cfgexpand.c (expand_asm_loc): Pass new parameter.
20635 (expand_asm_stmt): Likewise.
20636 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Add new
20638 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
20639 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
20640 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
20641 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
20642 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
20643 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
20644 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
20645 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
20646 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
20647 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
20648 * doc/tm.texi (md_asm_adjust): Likewise.
20649 * target.def (md_asm_adjust): Likewise.
20651 2021-03-04 Richard Biener <rguenther@suse.de>
20653 PR middle-end/97855
20654 * tree-pretty-print.c: Poison pp_printf.
20655 (dump_decl_name): Avoid use of pp_printf.
20656 (dump_block_node): Likewise.
20657 (dump_generic_node): Likewise.
20659 2021-03-04 Martin Sebor <msebor@redhat.com>
20661 PR middle-end/96963
20662 PR middle-end/94655
20663 * builtins.c (handle_array_ref): New helper.
20664 (handle_mem_ref): New helper.
20665 (compute_objsize_r): Factor out ARRAY_REF and MEM_REF handling
20666 into new helper functions. Correct a workaround for vectorized
20669 2021-03-03 Pat Haugen <pthaugen@linux.ibm.com>
20671 * config/rs6000/dfp.md (extendddtd2, trunctddd2, *cmp<mode>_internal1,
20672 floatditd2, ftrunc<mode>2, fix<mode>di2, dfp_ddedpd_<mode>,
20673 dfp_denbcd_<mode>, dfp_dxex_<mode>, dfp_diex_<mode>,
20674 *dfp_sgnfcnc_<mode>, dfp_dscli_<mode>, dfp_dscri_<mode>): Update size
20675 attribute for Power10.
20676 * config/rs6000/mma.md (*movoo): Likewise.
20677 * config/rs6000/rs6000.md (define_attr "size"): Add 256.
20678 (define_mode_attr bits): Add DD/TD modes.
20679 * config/rs6000/sync.md (load_quadpti, store_quadpti, load_lockedpti,
20680 store_conditionalpti): Update size attribute for Power10.
20682 2021-03-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20685 * config/sparc/t-sparc (tree-ssanames.o-warn): Don't error for
20686 -Wuninitialized, -Wmaybe-uninitialized.
20687 (wide-int.o-warn): Likewise.
20689 2021-03-03 Richard Earnshaw <rearnsha@arm.com>
20691 * common/config/arm/arm-common.c: Include configargs.h.
20692 (arm_config_default): New function.
20693 (arm_target_mode): Renamed from arm_target_thumb_only. Handle
20694 processors that do not support Thumb. Take into account the
20695 --with-mode configuration setting for selecting the default.
20696 * config/arm/arm.h (OPTION_DEFAULT_SPECS): Remove entry for 'mode'.
20697 (TARGET_MODE_SPEC_FUNCTIONS): Update for function name change.
20699 2021-03-03 Martin Liska <mliska@suse.cz>
20701 PR gcov-profile/97461
20702 * gcov-io.h (GCOV_PREALLOCATED_KVP): Remove.
20704 2021-03-03 Eric Botcazou <ebotcazou@adacore.com>
20707 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
20708 point back the hard frame pointer to its default location when the
20709 frame is larger than SEH_MAX_FRAME_SIZE.
20711 2021-03-03 Jakub Jelinek <jakub@redhat.com>
20714 * config/i386/predicates.md (logic_operator): New define_predicate.
20715 * config/i386/i386.md (mov + mem using comm arith peephole2):
20716 Punt if operands[1] is EXT_REX_SSE_REGNO_P, AVX512BW is not enabled
20717 and the inner mode is [QH]Imode.
20719 2021-03-03 Jakub Jelinek <jakub@redhat.com>
20722 * dwarf2out.c (dw_loc_list_struct): Add end_entry member.
20723 (new_loc_list): Clear end_entry.
20724 (output_loc_list): Only use DW_LLE_startx_length for -gsplit-dwarf
20725 if HAVE_AS_LEB128, otherwise use DW_LLE_startx_endx. Fix comment
20727 (index_location_lists): For dwarf_version >= 5 without HAVE_AS_LEB128,
20728 initialize also end_entry.
20730 2021-03-03 Jakub Jelinek <jakub@redhat.com>
20733 * cfgrtl.c (fixup_partitions): When changing some bbs from hot to cold
20734 partitions, if in non-layout mode after reorder_blocks also move
20735 affected blocks to ensure a single partition transition.
20737 2021-03-03 Jason Merrill <jason@redhat.com>
20740 * cgraphunit.c (process_function_and_variable_attributes): Don't
20741 warn about flatten on an alias if the target also has it.
20742 * cgraph.h (symtab_node::get_alias_target_tree): New.
20744 2021-03-02 David Edelsohn <dje.gcc@gmail.com>
20746 * config/rs6000/rs6000.md (tls_get_tpointer_internal): Prepend
20747 period to symbol name.
20748 (tls_get_addr_internal<mode>): Same.
20750 2021-03-02 David Malcolm <dmalcolm@redhat.com>
20753 * diagnostic-show-locus.c
20754 (selftest::test_one_liner_many_fixits_2): Fix accidental usage of
20757 2021-03-02 Martin Sebor <msebor@redhat.com>
20759 PR middle-end/99276
20760 * builtins.c (warn_for_access): Remove stray warning text.
20762 2021-03-02 Martin Sebor <msebor@redhat.com>
20764 PR middle-end/99295
20765 * doc/extend.texi (attribute malloc): Reword and clarify nonaliasing
20768 2021-03-02 Jakub Jelinek <jakub@redhat.com>
20771 * dwarf2out.c (output_macinfo_op): Use DW_MACRO_*_str* even with
20772 -gdwarf-5 -gstrict-dwarf. For -gsplit-dwarf -gdwarf-5 use
20773 DW_MACRO_*_strx instead of DW_MACRO_*_strp. Handle
20774 DW_MACRO_define_strx and DW_MACRO_undef_strx.
20775 (save_macinfo_strings): Use DW_MACRO_*_str* even with
20776 -gdwarf-5 -gstrict-dwarf. Handle DW_MACRO_define_strx and
20777 DW_MACRO_undef_strx.
20779 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
20781 * config/s390/s390-builtin-types.def (BT_FN_V4SF_V8HI_UINT): New
20783 (BT_FN_V8HI_V8HI_UINT): Likewise.
20784 (BT_FN_V8HI_V4SF_V4SF_UINT): Likewise.
20785 * config/s390/s390-builtins.def (B_NNPA): New macro definition.
20786 (s390_vclfnhs, s390_vclfnls, s390_vcrnfs, s390_vcfn, s390_vcnf):
20787 New builtin definitions.
20788 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Bump
20789 vector extension version.
20790 * config/s390/s390.c (s390_expand_builtin): Check if builtins are
20791 available with current -march level.
20792 * config/s390/s390.md (UNSPEC_NNPA_VCLFNHS_V8HI)
20793 (UNSPEC_NNPA_VCLFNLS_V8HI, UNSPEC_NNPA_VCRNFS_V8HI)
20794 (UNSPEC_NNPA_VCFN_V8HI, UNSPEC_NNPA_VCNF_V8HI): New constants.
20795 * config/s390/vecintrin.h (vec_extend_to_fp32_hi): New macro.
20796 (vec_extend_to_fp32_lo): Likewise.
20797 (vec_round_from_fp32): Likewise.
20798 (vec_convert_to_fp16): Likewise.
20799 (vec_convert_from_fp16): Likewise.
20800 * config/s390/vx-builtins.md (vclfnhs_v8hi): New insn pattern.
20801 (vclfnls_v8hi): Likewise.
20802 (vcrnfs_v8hi): Likewise.
20803 (vcfn_v8hi): Likewise.
20804 (vcnf_v8hi): Likewise.
20806 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
20808 * common/config/s390/s390-common.c (processor_flags_table): New entry.
20809 * config.gcc: Enable arch14 for --with-arch and --with-tune.
20810 * config/s390/driver-native.c (s390_host_detect_local_cpu): Pick
20811 arch14 for unknown CPU models.
20812 * config/s390/s390-opts.h (enum processor_type): Add PROCESSOR_ARCH14.
20813 * config/s390/s390.c (s390_issue_rate): Add case for PROCESSOR_ARCH14.
20814 (s390_get_sched_attrmask): Likewise.
20815 (s390_get_unit_mask): Likewise.
20816 * config/s390/s390.h (enum processor_flags): Add PF_NNPA and PF_ARCH14.
20817 (TARGET_CPU_ARCH14, TARGET_CPU_ARCH14_P, TARGET_CPU_NNPA)
20818 (TARGET_CPU_NNPA_P, TARGET_ARCH14, TARGET_ARCH14_P, TARGET_NNPA)
20819 (TARGET_NNPA_P): New macro definitions.
20820 * config/s390/s390.md ("cpu_facility", "enabled"): Add arch14 and nnpa.
20821 * config/s390/s390.opt: Add PROCESSOR_ARCH14.
20823 2021-03-02 Jakub Jelinek <jakub@redhat.com>
20825 PR middle-end/95757
20826 * tree-vrp.c (register_edge_assert_for): Remove superfluous ()s around
20827 condition. Call register_edge_assert_for_1 for == 0, != 0, == 1 and
20828 != 1 comparisons if name is lhs of a comparison.
20830 2021-03-01 Iain Sandoe <iain@sandoe.co.uk>
20834 * config/darwin-protos.h (darwin_should_restore_cfa_state): New.
20835 * config/darwin.c (darwin_should_restore_cfa_state): New.
20836 * config/darwin.h (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New.
20837 * doc/tm.texi: Regenerated.
20838 * doc/tm.texi.in: Document TARGET_ASM_SHOULD_RESTORE_CFA_STATE.
20839 * dwarf2cfi.c (connect_traces): If the target requests, restore
20840 the CFA expression after a DW_CFA_restore.
20841 * target.def (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New hook.
20843 2021-03-01 Martin Liska <mliska@suse.cz>
20846 * optc-save-gen.awk: Add 4 more exceptions.
20848 2021-03-01 Nathan Sidwell <nathan@acm.org>
20851 * tree.h (TYPE_ALIGN_RAW): New accessor.
20852 (TYPE_ALIGN): Use it.
20854 2021-03-01 Jan Hubicka <jh@suse.cz>
20857 * ipa-fnsummary.c (compute_fn_summary): Fix sanity check.
20859 2021-03-01 Eric Botcazou <ebotcazou@adacore.com>
20862 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
20863 point the hard frame pointer to the SSE register save area instead
20864 of the general register save area. Perform only minimal adjustment
20865 for small frames if it is initially not correctly aligned.
20866 (ix86_expand_prologue): Remove early saves for a SEH target.
20867 * config/i386/winnt.c (struct seh_frame_state): Document constraint.
20869 2021-02-28 Jakub Jelinek <jakub@redhat.com>
20872 * ipa.c (symbol_table::remove_unreachable_nodes): Fix a comment
20873 typo - referneced -> referenced.
20874 * tree.c (component_ref_size): Fix comment typo -
20875 refernce -> reference.
20876 * tree-ssa-alias.c (access_path_may_continue_p): Fix comment typo -
20877 traling -> trailing.
20878 (aliasing_component_refs_p): Fix comment typos -
20879 refernce -> reference and refernece -> reference and
20880 traling -> trailing.
20881 (nonoverlapping_refs_since_match_p): Fix comment typo -
20882 referneces -> references.
20883 * doc/invoke.texi (--param modref-max-bases): Fix a typo -
20884 referneces -> references.
20886 2021-02-27 Iain Sandoe <iain@sandoe.co.uk>
20888 * config/host-darwin.c (darwin_gt_pch_use_address): Modify
20889 diagnostic message to avoid use of a contraction and format
20892 2021-02-27 Jakub Jelinek <jakub@redhat.com>
20895 * gcse.c (gcse_or_cprop_is_too_expensive): Use %wu instead of
20896 HOST_WIDE_INT_PRINT_UNSIGNED in warning format string.
20897 * ipa-devirt.c (ipa_odr_read_section): Use %wd instead of
20898 HOST_WIDE_INT_PRINT_DEC in inform format string. Fix comment
20901 2021-02-26 Richard Biener <rguenther@suse.de>
20903 PR middle-end/99281
20904 * expr.c (store_field): For calls with return-slot optimization
20905 and addressable return type expand the store directly.
20907 2021-02-26 Richard Biener <rguenther@suse.de>
20910 * builtins.c (warn_string_no_nul): Fix diagnostic formatting.
20912 2021-02-26 Peter Bergner <bergner@linux.ibm.com>
20915 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Replace assert
20918 2021-02-26 Aaron Sawdey <acsawdey@linux.ibm.com>
20920 * config.gcc: Add rs6000-pcrel-opt.o.
20921 * config/rs6000/rs6000-pcrel-opt.c: New file.
20922 * config/rs6000/pcrel-opt.md: New file.
20923 * config/rs6000/predicates.md: Add d_form_memory predicate.
20924 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_PCREL_OPT.
20925 * config/rs6000/rs6000-passes.def: Add pass_pcrel_opt.
20926 * config/rs6000/rs6000-protos.h: Add reg_to_non_prefixed(),
20927 pcrel_opt_valid_mem_p(), output_pcrel_opt_reloc(),
20928 and make_pass_pcrel_opt().
20929 * config/rs6000/rs6000.c (reg_to_non_prefixed): Make global.
20930 (rs6000_option_override_internal): Add pcrel-opt.
20931 (rs6000_delegitimize_address): Support pcrel-opt.
20932 (rs6000_opt_masks): Add pcrel-opt.
20933 (pcrel_opt_valid_mem_p): New function.
20934 (reg_to_non_prefixed): Make global.
20935 (rs6000_asm_output_opcode): Reset prepend_p_to_next_insn.
20936 (output_pcrel_opt_reloc): New function.
20937 * config/rs6000/rs6000.md (loads_extern_addr): New attr.
20938 (pcrel_extern_addr): Set loads_extern_addr.
20939 Add include for pcrel-opt.md.
20940 * config/rs6000/rs6000.opt: Add -mpcrel-opt.
20941 * config/rs6000/t-rs6000: Add rules for pcrel-opt.c and
20944 2021-02-26 YunQiang Su <yunqiang.su@cipunited.com>
20947 * config/mips/mips.c (mips_expand_ext_as_unaligned_load):
20948 If TARGET_64BIT and dest is SUBREG, we check the width, if it
20949 equal to SImode, we use SImode operation, just like what we are
20952 2021-02-26 Marek Polacek <polacek@redhat.com>
20954 * builtins.c (warn_for_access): Fix typos.
20956 2021-02-25 Iain Sandoe <iain@sandoe.co.uk>
20958 * config/aarch64/aarch64.md (<optab>_rol<mode>3): Add a '#'
20959 mark in front of the immediate quantity.
20960 (<optab>_rolsi3_uxtw): Likewise.
20962 2021-02-25 Richard Earnshaw <rearnsha@arm.com>
20965 * config/arm/thumb2.md (nonsecure_call_reg_thumb2_fpcxt): New pattern.
20966 (nonsecure_call_value_reg_thumb2_fpcxt): Likewise.
20967 (nonsecure_call_reg_thumb2): Restrict to using r4 for the callee
20968 address and disable when the FPCXT is not available.
20969 (nonsecure_call_value_reg_thumb2): Likewise.
20971 2021-02-25 Nathan Sidwell <nathan@acm.org>
20974 * doc/invoke.texi (flang-info-module-cmi): Renamed option.
20976 2021-02-25 Tamar Christina <tamar.christina@arm.com>
20978 * tree-vect-slp.c (optimize_load_redistribution_1): Abort on NULL nodes.
20980 2021-02-25 Richard Biener <rguenther@suse.de>
20982 PR tree-optimization/99253
20983 * tree-vect-loop.c (check_reduction_path): First compute
20984 code, then verify out-of-loop uses.
20986 2021-02-25 Jakub Jelinek <jakub@redhat.com>
20989 * match.pd ((T)(A) + CST -> (T)(A + CST)): Add :s to convert.
20991 2021-02-25 Jakub Jelinek <jakub@redhat.com>
20993 PR tree-optimization/80635
20994 * tree-vrp.c (vrp_simplify_cond_using_ranges): Also handle
20995 VIEW_CONVERT_EXPR if modes are the same, innerop is integral and
20996 has mode precision.
20998 2021-02-25 Richard Biener <rguenther@suse.de>
21000 * tree-vect-slp.c (optimize_load_redistribution_1): Delay
21001 load_map population.
21002 (vect_match_slp_patterns_2): Revert part of last change.
21003 (vect_analyze_slp): Do not interleave optimize_load_redistribution
21004 with pattern detection but do it afterwards. Dump the
21005 whole SLP graph after pattern recognition and load
21006 redistribution optimization finished.
21008 2021-02-24 Jakub Jelinek <jakub@redhat.com>
21011 * omp-low.c (struct omp_context): Add teams_nested_p and
21012 nonteams_nested_p members.
21013 (scan_omp_target): Diagnose teams nested inside of target with other
21014 directives strictly nested inside of the same target.
21015 (check_omp_nesting_restrictions): Set ctx->teams_nested_p or
21016 ctx->nonteams_nested_p as needed.
21018 2021-02-24 Vladimir N. Makarov <vmakarov@redhat.com>
21020 PR inline-asm/99123
21021 * lra-constraints.c (uses_hard_regs_p): Don't use decompose_mem_address.
21023 2021-02-24 Hans-Peter Nilsson <hp@axis.com>
21025 * config/cris/cris.c (cris_expand_prologue): Set
21026 current_function_static_stack_size, if flag_stack_usage_info.
21028 2021-02-24 Pat Haugen <pthaugen@linux.ibm.com>
21030 * config/rs6000/rs6000.c (next_insn_prefixed_p): Rename.
21031 (rs6000_final_prescan_insn): Adjust.
21032 (rs6000_asm_output_opcode): Likewise.
21034 2021-02-24 Martin Sebor <msebor@redhat.com>
21036 PR middle-end/97172
21037 * attribs.c (attr_access::free_lang_data): Clear attribute arg spec
21038 from function arguments.
21040 2021-02-24 Tamar Christina <tamar.christina@arm.com>
21042 PR tree-optimization/99220
21043 * tree-vect-slp.c (optimize_load_redistribution_1): Remove
21044 node from cache when it's about to be deleted.
21046 2021-02-24 Jakub Jelinek <jakub@redhat.com>
21048 PR tree-optimization/99225
21049 * fold-const.c (fold_binary_loc) <case NE_EXPR>: In (x & (1 << y)) != 0
21050 to ((x >> y) & 1) != 0 simplifications use build_one_cst instead of
21051 build_int_cst (..., 1). Formatting fixes.
21053 2021-02-24 Tamar Christina <tamar.christina@arm.com>
21055 PR tree-optimization/99149
21056 * tree-vect-slp-patterns.c (vect_detect_pair_op): Don't recreate the
21058 (vect_slp_reset_pattern): Remove.
21059 (complex_fma_pattern::matches): Remove call to vect_slp_reset_pattern.
21060 (complex_mul_pattern::build, complex_fma_pattern::build,
21061 complex_fms_pattern::build): Fix ref counts.
21062 * tree-vect-slp.c (vect_free_slp_tree): Undo SLP only pattern relevancy
21063 when node is being deleted.
21064 (vect_match_slp_patterns_2): Correct result of cache hit on patterns.
21065 (vect_schedule_slp): Invalidate SLP_TREE_REPRESENTATIVE of removed
21067 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize value.
21069 2021-02-24 Matthias Klose <doko@ubuntu.com>
21072 2020-12-07 Matthias Klose <doko@ubuntu.com>
21074 * genextract.c (print_header): Undefine ENABLE_RTL_CHECKING
21075 and ENABLE_RTL_FLAG_CHECKING.
21077 2021-02-24 Richard Biener <rguenther@suse.de>
21080 * builtins.c (fold_builtin_next_arg): Avoid NULL arg.
21082 2021-02-23 Peter Bergner <bergner@linux.ibm.com>
21084 * config/rs6000/mma.md (mma_assemble_pair): Rename from this...
21085 (vsx_assemble_pair): ...to this.
21086 (*mma_assemble_pair): Rename from this...
21087 (*vsx_assemble_pair): ...to this.
21088 (mma_disassemble_pair): Rename from this...
21089 (vsx_disassemble_pair): ...to this.
21090 (*mma_disassemble_pair): Rename from this...
21091 (*vsx_disassemble_pair): ...to this.
21092 * config/rs6000/rs6000-builtin.def (BU_MMA_V2, BU_MMA_V3,
21093 BU_COMPAT): New macros.
21094 (mma_assemble_pair): Rename from this...
21095 (vsx_assemble_pair): ...to this.
21096 (mma_disassemble_pair): Rename from this...
21097 (vsx_disassemble_pair): ...to this.
21098 (mma_assemble_pair): New compatibility built-in.
21099 (mma_disassemble_pair): Likewise.
21100 * config/rs6000/rs6000-call.c (struct builtin_compatibility): New.
21101 (RS6000_BUILTIN_COMPAT): Define.
21102 (bdesc_compat): New.
21103 (mma_expand_builtin): Use VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
21104 (rs6000_gimple_fold_mma_builtin): Use MMA_BUILTIN_DISASSEMBLE_PAIR
21105 and VSX_BUILTIN_ASSEMBLE_PAIR.
21106 (rs6000_init_builtins): Register compatibility built-ins.
21107 (mma_init_builtins): Use VSX_BUILTIN_ASSEMBLE_PAIR,
21108 VSX_BUILTIN_ASSEMBLE_PAIR_INTERNAL, VSX_BUILTIN_DISASSEMBLE_PAIR and
21109 VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
21110 * doc/extend.texi (__builtin_mma_assemble_pair): Rename from this...
21111 (__builtin_vsx_assemble_pair): ...to this.
21112 (__builtin_mma_disassemble_pair): Rename from this...
21113 (__builtin_vsx_disassemble_pair): ...to this.
21115 2021-02-23 Martin Liska <mliska@suse.cz>
21118 * ipa-icf.c (sem_variable::merge): Do not merge 2 variables
21119 with different alignment. That leads to an invalid red zone
21120 size allocated in runtime.
21122 2021-02-23 Jakub Jelinek <jakub@redhat.com>
21124 PR tree-optimization/99204
21125 * fold-const.c (fold_read_from_constant_string): Check that
21126 tree_fits_uhwi_p (index) rather than just that index is INTEGER_CST.
21128 2021-02-23 Segher Boessenkool <segher@kernel.crashing.org>
21129 Kewen Lin <linkw@gcc.gnu.org>
21131 * config/rs6000/rs6000.md (*rotl<mode>3_insert_3): Renamed to...
21132 (rotl<mode>3_insert_3): ...this.
21133 (plus_ior_xor): New code_iterator.
21134 (define_split for GPR rl*imi): New splitter.
21135 * config/rs6000/vsx.md (vsx_init_v4si): Use gen_rotldi3_insert_3
21136 for integer merging.
21138 2021-02-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21140 * config/aarch64/aarch64-tuning-flags.def (cse_sve_vl_constants):
21142 * config/aarch64/aarch64.md (add<mode>3): Force CONST_POLY_INT immediates
21143 into a register when the above is enabled.
21144 * config/aarch64/aarch64.c (neoversev1_tunings):
21145 AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
21146 (aarch64_rtx_costs): Use AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
21148 2021-02-22 Hans-Peter Nilsson <hp@axis.com>
21150 * config/cris/cris.c (cris_print_operand) <'T'>: Change
21151 valid operand from is now an addi mult-value to shift-value.
21152 * config/cris/cris.md (*addi): Change expression of scaled
21153 operand from mult to ashift.
21154 * config/cris/cris.md (*addi_reload): New insn_and_split.
21156 2021-02-22 John David Anglin <danglin@gcc.gnu.org>
21159 * config/pa/pa.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Define as
21160 hook_bool_const_tree_hwi_hwi_const_tree_true.
21161 (pa_asm_output_mi_thunk): Add support for nonzero vcall_offset.
21163 2021-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
21165 PR rtl-optimization/98791
21166 * ira-conflicts.c (process_regs_for_copy): Don't create allocno copies
21167 for unordered modes.
21169 2021-02-22 Martin Liska <mliska@suse.cz>
21171 * tree-inline.c (inline_forbidden_p): Set
21172 inline_forbidden_reason.
21174 2021-02-22 Richard Biener <rguenther@suse.de>
21176 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Dump
21179 2021-02-22 Richard Biener <rguenther@suse.de>
21181 PR tree-optimization/99165
21182 * gimple-ssa-store-merging.c (pass_store_merging::process_store):
21183 Accumulate changed to ret.
21185 2021-02-21 Uros Bizjak <ubizjak@gmail.com>
21188 2020-12-09 Uroš Bizjak <ubizjak@gmail.com>
21190 * config/i386/i386.h (REG_ALLOC_ORDER): Remove
21192 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
21195 * config/s390/vector.md (trunctf<DFP_ALL:mode>2_vr): New
21197 (trunctf<DFP_ALL:mode>2): Likewise.
21198 (trunctdtf2_vr): Likewise.
21199 (trunctdtf2): Likewise.
21200 (extend<DFP_ALL:mode>tf2_vr): Likewise.
21201 (extend<DFP_ALL:mode>tf2): Likewise.
21202 (extendtftd2_vr): Likewise.
21203 (extendtftd2): Likewise.
21205 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
21207 * config/s390/vector.md (*fprx2_to_tf): Rename to fprx2_to_tf,
21208 add memory alternative.
21209 (tf_to_fprx2): New pattern.
21211 2021-02-19 Martin Sebor <msebor@redhat.com>
21214 * attribs.c (init_attr_rdwr_indices): Guard vblist use.
21215 (attr_access::free_lang_data): Remove a spurious test.
21217 2021-02-19 Nathan Sidwell <nathan@acm.org>
21219 * doc/invoke.texi (flang-info-module-read): Document.
21221 2021-02-19 Martin Liska <mliska@suse.cz>
21223 PR translation/99167
21224 * params.opt: Fix typo.
21226 2021-02-19 Richard Biener <rguenther@suse.de>
21228 PR middle-end/99122
21229 * tree-inline.c (inline_forbidden_p): Do not inline functions
21230 with VLA arguments or return value.
21232 2021-02-19 Jakub Jelinek <jakub@redhat.com>
21235 * config/arm/arm.md (*stack_protect_combined_set_insn,
21236 *stack_protect_combined_test_insn): If force_const_mem result
21237 is not valid general operand, force its address into the destination
21240 2021-02-19 Jakub Jelinek <jakub@redhat.com>
21243 * tree-cfg.c (gimple_merge_blocks): If bb a starts with eh landing
21244 pad or non-local label, put FORCED_LABELs from bb b after that label
21245 rather than before it.
21247 2021-02-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
21250 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3): Use
21251 expand_vector_broadcast' to emit the vec_duplicate operand.
21253 2021-02-18 Vladimir N. Makarov <vmakarov@redhat.com>
21255 PR rtl-optimization/96264
21256 * lra-remat.c (reg_overlap_for_remat_p): Check also output insn
21259 2021-02-18 H.J. Lu <hjl.tools@gmail.com>
21262 * varasm.c (get_section): Replace SUPPORTS_SHF_GNU_RETAIN with
21263 looking up the retain attribute.
21264 (resolve_unique_section): Likewise.
21265 (get_variable_section): Likewise.
21266 (switch_to_section): Likewise. Warn when a symbol without the
21267 retain attribute and a symbol with the retain attribute are
21268 placed in the section with the same name, instead of the used
21270 * doc/extend.texi: Document the "retain" attribute.
21272 2021-02-18 Nathan Sidwell <nathan@acm.org>
21275 * doc/invoke.texi (flang-info-include-translate): Document header
21278 2021-02-18 Richard Biener <rguenther@suse.de>
21280 PR middle-end/99122
21281 * ipa-fnsummary.c (analyze_function_body): Set
21282 CIF_FUNCTION_NOT_INLINABLE for VLA parameter calls.
21283 * tree-inline.c (insert_init_debug_bind): Pass NULL for
21284 error_mark_node values.
21285 (force_value_to_type): Do not build V_C_Es for WITH_SIZE_EXPR
21287 (setup_one_parameter): Delay force_value_to_type until when
21290 2021-02-18 Hans-Peter Nilsson <hp@axis.com>
21292 PR tree-optimization/99142
21293 * match.pd (clz cmp 0): Gate replacement on single_use of clz result.
21295 2021-02-18 Jakub Jelinek <jakub@redhat.com>
21297 * wide-int-bitmask.h (wide_int_bitmask::wide_int_bitmask (),
21298 wide_int_bitmask::wide_int_bitmask (uint64_t),
21299 wide_int_bitmask::wide_int_bitmask (uint64_t, uint64_t),
21300 wide_int_bitmask::operator ~ () const,
21301 wide_int_bitmask::operator | (wide_int_bitmask) const,
21302 wide_int_bitmask::operator & (wide_int_bitmask) const): Use constexpr
21304 * config/i386/i386.h (PTA_3DNOW, PTA_3DNOW_A, PTA_64BIT, PTA_ABM,
21305 PTA_AES, PTA_AVX, PTA_BMI, PTA_CX16, PTA_F16C, PTA_FMA, PTA_FMA4,
21306 PTA_FSGSBASE, PTA_LWP, PTA_LZCNT, PTA_MMX, PTA_MOVBE, PTA_NO_SAHF,
21307 PTA_PCLMUL, PTA_POPCNT, PTA_PREFETCH_SSE, PTA_RDRND, PTA_SSE, PTA_SSE2,
21308 PTA_SSE3, PTA_SSE4_1, PTA_SSE4_2, PTA_SSE4A, PTA_SSSE3, PTA_TBM,
21309 PTA_XOP, PTA_AVX2, PTA_BMI2, PTA_RTM, PTA_HLE, PTA_PRFCHW, PTA_RDSEED,
21310 PTA_ADX, PTA_FXSR, PTA_XSAVE, PTA_XSAVEOPT, PTA_AVX512F, PTA_AVX512ER,
21311 PTA_AVX512PF, PTA_AVX512CD, PTA_NO_TUNE, PTA_SHA, PTA_PREFETCHWT1,
21312 PTA_CLFLUSHOPT, PTA_XSAVEC, PTA_XSAVES, PTA_AVX512DQ, PTA_AVX512BW,
21313 PTA_AVX512VL, PTA_AVX512IFMA, PTA_AVX512VBMI, PTA_CLWB, PTA_MWAITX,
21314 PTA_CLZERO, PTA_NO_80387, PTA_PKU, PTA_AVX5124VNNIW, PTA_AVX5124FMAPS,
21315 PTA_AVX512VPOPCNTDQ, PTA_SGX, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES,
21316 PTA_AVX512VBMI2, PTA_VPCLMULQDQ, PTA_AVX512BITALG, PTA_RDPID,
21317 PTA_PCONFIG, PTA_WBNOINVD, PTA_AVX512VP2INTERSECT, PTA_PTWRITE,
21318 PTA_AVX512BF16, PTA_WAITPKG, PTA_MOVDIRI, PTA_MOVDIR64B, PTA_ENQCMD,
21319 PTA_CLDEMOTE, PTA_SERIALIZE, PTA_TSXLDTRK, PTA_AMX_TILE, PTA_AMX_INT8,
21320 PTA_AMX_BF16, PTA_UINTR, PTA_HRESET, PTA_KL, PTA_WIDEKL, PTA_AVXVNNI,
21321 PTA_X86_64_BASELINE, PTA_X86_64_V2, PTA_X86_64_V3, PTA_X86_64_V4,
21322 PTA_CORE2, PTA_NEHALEM, PTA_WESTMERE, PTA_SANDYBRIDGE, PTA_IVYBRIDGE,
21323 PTA_HASWELL, PTA_BROADWELL, PTA_SKYLAKE, PTA_SKYLAKE_AVX512,
21324 PTA_CASCADELAKE, PTA_COOPERLAKE, PTA_CANNONLAKE, PTA_ICELAKE_CLIENT,
21325 PTA_ICELAKE_SERVER, PTA_TIGERLAKE, PTA_SAPPHIRERAPIDS, PTA_ALDERLAKE,
21326 PTA_KNL, PTA_BONNELL, PTA_SILVERMONT, PTA_GOLDMONT, PTA_GOLDMONT_PLUS,
21327 PTA_TREMONT, PTA_KNM): Use constexpr instead of const.
21329 2021-02-18 Jakub Jelinek <jakub@redhat.com>
21331 PR middle-end/99109
21332 * gimple-array-bounds.cc (build_zero_elt_array_type): Rename to ...
21333 (build_printable_array_type): ... this. Add nelts argument. For
21334 overaligned eltype, use TYPE_MAIN_VARIANT (eltype) instead. If
21335 nelts, call build_array_type_nelts.
21336 (array_bounds_checker::check_mem_ref): Use build_printable_array_type
21337 instead of build_zero_elt_array_type and build_array_type_nelts.
21339 2021-02-18 Jakub Jelinek <jakub@redhat.com>
21342 * config/i386/i386.c (distance_non_agu_define): Don't call
21343 extract_insn_cached here.
21344 (ix86_lea_outperforms): Save and restore recog_data around call
21345 to distance_non_agu_define and distance_agu_use.
21346 (ix86_ok_to_clobber_flags): Remove.
21347 (ix86_avoid_lea_for_add): Don't call ix86_ok_to_clobber_flags.
21348 (ix86_avoid_lea_for_addr): Likewise. Adjust function comment.
21349 * config/i386/i386.md (*lea<mode>): Change from define_insn_and_split
21350 into define_insn. Move the splitting to define_peephole2 and
21351 check there using peep2_regno_dead_p if FLAGS_REG is dead.
21353 2021-02-17 Julian Brown <julian@codesourcery.com>
21355 * gimplify.c (gimplify_scan_omp_clauses): Handle ATTACH_DETACH
21358 2021-02-17 Xi Ruoyao <xry111@mengyan1223.wang>
21361 * config/mips/mips.c (mips_symbol_insns): Do not use
21362 MSA_SUPPORTED_MODE_P if mode is MAX_MACHINE_MODE.
21364 2021-02-16 Vladimir N. Makarov <vmakarov@redhat.com>
21366 PR inline-asm/98096
21367 * stmt.c (resolve_operand_name_1): Take inout operands into account
21368 for access to labels by names.
21369 * doc/extend.texi: Describe counting operands for accessing labels.
21371 2021-02-16 Richard Biener <rguenther@suse.de>
21373 PR tree-optimization/38474
21374 * tree-ssa-structalias.c (variable_info::address_taken): New.
21375 (new_var_info): Initialize address_taken.
21376 (process_constraint): Set address_taken.
21377 (solve_constraints): Use the new address_taken flag rather
21378 than is_reg_var for sorting variables.
21379 (dump_constraint): Dump the variable number if the name
21382 2021-02-16 Jakub Jelinek <jakub@redhat.com>
21385 * tree-vect-stmts.c (vectorizable_simd_clone_call): For num_calls != 1
21386 multiply by 4096 and for inbranch by 8192.
21387 * config/i386/i386.c (ix86_simd_clone_usable): For TARGET_AVX512F,
21388 return 3, 2 or 1 for mangle letters 'b', 'c' or 'd'.
21390 2021-02-15 Maya Rashish <coypu@sdf.org>
21392 * config/aarch64/aarch64.c (aarch64_init_builtins):
21393 Call SUBTARGET_INIT_BUILTINS.
21395 2021-02-15 Peter Bergner <bergner@linux.ibm.com>
21397 PR rtl-optimization/98872
21398 * init-regs.c (initialize_uninitialized_regs): Skip initialization
21399 if CONST0_RTX is NULL.
21401 2021-02-15 Richard Sandiford <richard.sandiford@arm.com>
21403 PR rtl-optimization/98863
21404 * rtl-ssa/functions.h (function_info::bb_live_out_info): Delete.
21405 (function_info::build_info): Turn into a declaration, moving the
21406 definition to internals.h.
21407 (function_info::bb_walker): Declare.
21408 (function_info::create_reg_use): Likewise.
21409 (function_info::calculate_potential_phi_regs): Take a build_info
21411 (function_info::place_phis, function_info::create_ebbs): Declare.
21412 (function_info::calculate_ebb_live_in_for_debug): Likewise.
21413 (function_info::populate_backedge_phis): Delete.
21414 (function_info::start_block, function_info::end_block): Declare.
21415 (function_info::populate_phi_inputs): Delete.
21416 (function_info::m_potential_phi_regs): Move information to build_info.
21417 * rtl-ssa/internals.h: New file.
21418 (function_info::bb_phi_info): New class.
21419 (function_info::build_info): Moved from functions.h.
21420 Add a constructor and destructor.
21421 (function_info::build_info::ebb_use): Delete.
21422 (function_info::build_info::ebb_def): Likewise.
21423 (function_info::build_info::bb_live_out): Likewise.
21424 (function_info::build_info::tmp_ebb_live_in_for_debug): New variable.
21425 (function_info::build_info::potential_phi_regs): Likewise.
21426 (function_info::build_info::potential_phi_regs_for_debug): Likewise.
21427 (function_info::build_info::ebb_def_regs): Likewise.
21428 (function_info::build_info::bb_phis): Likewise.
21429 (function_info::build_info::bb_mem_live_out): Likewise.
21430 (function_info::build_info::bb_to_rpo): Likewise.
21431 (function_info::build_info::def_stack): Likewise.
21432 (function_info::build_info::old_def_stack_limit): Likewise.
21433 * rtl-ssa/internals.inl (function_info::build_info::record_reg_def):
21434 Remove the regno argument. Push the previous definition onto the
21435 definition stack where necessary.
21436 * rtl-ssa/accesses.cc: Include internals.h.
21437 * rtl-ssa/changes.cc: Likewise.
21438 * rtl-ssa/blocks.cc: Likewise.
21439 (function_info::build_info::build_info): Define.
21440 (function_info::build_info::~build_info): Likewise.
21441 (function_info::bb_walker): New class.
21442 (function_info::bb_walker::bb_walker): Define.
21443 (function_info::add_live_out_use): Convert a logarithmic-complexity
21444 test into a linear one. Allow the same definition to be passed
21446 (function_info::calculate_potential_phi_regs): Moved from
21447 functions.cc. Take a build_info parameter and store the
21448 information there instead.
21449 (function_info::place_phis): New function.
21450 (function_info::add_entry_block_defs): Update call to record_reg_def.
21451 (function_info::calculate_ebb_live_in_for_debug): New function.
21452 (function_info::add_phi_nodes): Use bb_phis to decide which
21453 registers need phi nodes and initialize ebb_def_regs accordingly.
21454 Do not add degenerate phis here.
21455 (function_info::add_artificial_accesses): Use create_reg_use.
21456 Assert that all definitions are listed in the DF LR sets.
21457 Update call to record_reg_def.
21458 (function_info::record_block_live_out): Record live-out register
21459 values in the phis of successor blocks. Use the live-out set
21460 when processing the last block in an EBB, instead of always
21461 using the live-in sets of successor blocks. AND the live sets
21462 with the set of registers that have been defined in the EBB,
21463 rather than with all potential phi registers. Cope correctly
21464 with branches back to the start of the current EBB.
21465 (function_info::start_block): New function.
21466 (function_info::end_block): Likewise.
21467 (function_info::populate_phi_inputs): Likewise.
21468 (function_info::create_ebbs): Likewise.
21469 (function_info::process_all_blocks): Rewrite into a multi-phase
21471 * rtl-ssa/functions.cc: Include internals.h.
21472 (function_info::calculate_potential_phi_regs): Move to blocks.cc.
21473 (function_info::init_function_data): Remove caller.
21474 * rtl-ssa/insns.cc: Include internals.h
21475 (function_info::create_reg_use): New function. Lazily any
21476 degenerate phis needed by the linear RPO view.
21477 (function_info::record_use): Use create_reg_use. When processing
21478 debug uses, use potential_phi_regs and test it before checking
21479 whether the register is live on entry to the current EBB. Lazily
21480 calculate ebb_live_in_for_debug.
21481 (function_info::record_call_clobbers): Update call to record_reg_def.
21482 (function_info::record_def): Likewise.
21484 2021-02-15 Martin Liska <mliska@suse.cz>
21486 * toplev.c (init_asm_output): Free output of
21487 gen_command_line_string function.
21488 (process_options): Likewise.
21490 2021-02-15 Martin Liska <mliska@suse.cz>
21492 * params.opt: Add 2 missing Param keywords.
21494 2021-02-15 Eric Botcazou <ebotcazou@adacore.com>
21496 * df-core.c (df_worklist_dataflow_doublequeue): Use proper cast.
21498 2021-02-15 Jakub Jelinek <jakub@redhat.com>
21500 PR tree-optimization/99079
21501 * match.pd (A % (pow2pcst << N) -> A & ((pow2pcst << N) - 1)): Remove
21502 useless tree_nop_conversion_p (type, TREE_TYPE (@3)) check. Instead
21503 require both type and TREE_TYPE (@1) to be integral types and either
21504 type having smaller or equal precision, or TREE_TYPE (@1) being
21505 unsigned type, or type being signed type. If TREE_TYPE (@1)
21506 doesn't have wrapping overflow, perform the subtraction of one in
21509 2021-02-14 Jan Hubicka <hubicka@ucw.cz>
21510 Richard Biener <rguether@suse.de>
21513 * ipa-reference.c (ipa_init): Only conditinally initialize
21514 reference_vars_to_consider.
21515 (propagate): Conditionally deninitialize reference_vars_to_consider.
21516 (ipa_reference_write_optimization_summary): Sanity check that
21517 reference_vars_to_consider is not allocated.
21519 2021-02-13 Levy Hsu <admin@levyhsu.com>
21522 * config/riscv/riscv-shorten-memrefs.c (pass_shorten_memrefs): Add
21523 extend parameter to get_si_mem_base_reg declaration.
21524 (get_si_mem_base_reg): Add extend parameter. Set it.
21525 (analyze): Pass extend arg to get_si_mem_base_reg.
21526 (transform): Likewise. Use it when rewriting mems.
21527 * config/riscv/riscv.c (riscv_legitimize_move): Check for subword
21528 loads and emit sign/zero extending load followed by subreg move.
21530 2021-02-13 Jim Wilson <jimw@sifive.com>
21533 * config/riscv/riscv.c (riscv_compressed_lw_address_p): Drop early
21534 exit when !reload_completed. Only perform check for compressed reg
21535 if reload_completed.
21536 (riscv_rtx_costs): In MEM case, when optimizing for size and
21537 shorten memrefs, if not compressible, then increase cost.
21539 2021-02-13 Jakub Jelinek <jakub@redhat.com>
21541 PR rtl-optimization/98439
21542 * recog.c (pass_split_before_regstack::gate): Enable even when
21543 pass_split_before_sched2 is enabled if -fselective-scheduling2 is
21546 2021-02-13 Jakub Jelinek <jakub@redhat.com>
21549 * config/i386/mmx.md (*mmx_pshufd_1): Add a combine splitter for
21550 swap of V2SImode elements in memory into DImode memory rotate by 32.
21552 2021-02-12 Martin Sebor <msebor@redhat.com>
21554 * tree-pretty-print.c (print_generic_expr_to_str): Update comment.
21556 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
21558 * rtl-ssa/accesses.cc (function_info::make_use_available): Use
21559 m_temp_obstack rather than m_obstack to allocate the temporary use.
21561 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
21563 * df-problems.c (df_lr_bb_local_compute): Treat partial definitions
21564 as read-modify operations.
21566 2021-02-12 Richard Biener <rguenther@suse.de>
21568 PR middle-end/38474
21569 * ipa-fnsummary.c (unmodified_parm_1): Only walk when
21570 fbi->aa_walk_budget is bigger than zero. Update
21571 fbi->aa_walk_budget.
21572 (param_change_prob): Likewise.
21573 * ipa-prop.c (detect_type_change_from_memory_writes):
21574 Properly account walk_aliased_vdefs.
21575 (parm_preserved_before_stmt_p): Canonicalize updates.
21576 (parm_ref_data_preserved_p): Likewise.
21577 (parm_ref_data_pass_through_p): Likewise.
21578 (determine_known_aggregate_parts): Account own alias queries.
21580 2021-02-12 Martin Liska <mliska@suse.cz>
21582 * opts-common.c (decode_cmdline_option): Release werror_arg.
21583 * opts.c (gen_producer_string): Release output of
21584 gen_command_line_string.
21586 2021-02-12 Richard Biener <rguenther@suse.de>
21588 PR tree-optimization/38474
21589 * params.opt (-param=max-store-chains-to-track=): New param.
21590 (-param=max-stores-to-track=): Likewise.
21591 * doc/invoke.texi (max-store-chains-to-track): Document.
21592 (max-stores-to-track): Likewise.
21593 * gimple-ssa-store-merging.c (pass_store_merging::m_n_chains):
21595 (pass_store_merging::m_n_stores): Likewise.
21596 (pass_store_merging::terminate_and_process_chain): Update
21597 m_n_stores and m_n_chains.
21598 (pass_store_merging::process_store): Likewise. Terminate
21599 oldest chains if the number of stores or chains get too large.
21600 (imm_store_chain_info::terminate_and_process_chain): Dump
21603 2021-02-11 Eric Botcazou <ebotcazou@adacore.com>
21605 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
21606 the cold section, emit a nop before the directive if the previous
21607 active instruction can throw.
21609 2021-02-11 Peter Bergner <bergner@linux.ibm.com>
21612 * config/rs6000/predicates.md (mma_assemble_input_operand): Restrict
21613 memory addresses that are legal for quad word accesses.
21615 2021-02-11 Andrea Corallo <andrea.corallo@arm.com>
21618 * config/arm/thumb2.md (*doloop_end_internal): Generate
21619 alternative sequence to handle long range branches.
21621 2021-02-11 Joel Hutton <joel.hutton@arm.com>
21623 PR tree-optimization/98772
21624 * optabs-tree.c (supportable_half_widening_operation): New function
21625 to check for supportable V8QI->V8HI widening patterns.
21626 * optabs-tree.h (supportable_half_widening_operation): New function.
21627 * tree-vect-stmts.c (vect_create_half_widening_stmts): New function
21628 to create promotion stmts for V8QI->V8HI widening patterns.
21629 (vectorizable_conversion): Add case for V8QI->V8HI.
21631 2021-02-11 Richard Biener <rguenther@suse.de>
21633 * sparseset.h (SPARSESET_ELT_BITS): Remove.
21634 (SPARSESET_ELT_TYPE): Use unsigned int.
21635 * fwprop.c: Do not include sparseset.h.
21637 2021-02-10 Jakub Jelinek <jakub@redhat.com>
21640 * varasm.c (declare_weak): For -fsyntax-only, allow even
21641 TREE_ASM_WRITTEN function decls.
21643 2021-02-10 Jakub Jelinek <jakub@redhat.com>
21646 * config/i386/sse.md (fix<fixunssuffix>_truncv2sfv2di2,
21647 <insn>v8qiv8hi2, <insn>v8qiv8si2, <insn>v4qiv4si2, <insn>v4hiv4si2,
21648 <insn>v8qiv8di2, <insn>v4qiv4di2, <insn>v2qiv2di2, <insn>v4hiv4di2,
21649 <insn>v2hiv2di2, <insn>v2siv2di2): Force operands[1] into REG before
21650 calling simplify_gen_subreg on it.
21652 2021-02-10 Martin Liska <mliska@suse.cz>
21654 * config/nvptx/nvptx.c (nvptx_option_override): Use
21655 flag_patchable_function_entry instead of the removed
21656 function_entry_patch_area_size.
21658 2021-02-10 Martin Liska <mliska@suse.cz>
21660 PR tree-optimization/99002
21661 PR tree-optimization/99026
21662 * gimple-if-to-switch.cc (if_chain::is_beneficial): Fix memory
21663 leak when adjacent cases are merged.
21664 * tree-switch-conversion.c (switch_decision_tree::analyze_switch_statement): Use
21666 (make_pass_lower_switch): Remove trailing whitespace.
21667 * tree-switch-conversion.h (release_clusters): New.
21669 2021-02-10 Richard Biener <rguenther@suse.de>
21671 PR rtl-optimization/99054
21672 * cfgrtl.c (rtl-optimization/99054): Return an auto_vec.
21673 (fixup_partitions): Adjust.
21674 (rtl_verify_edges): Likewise.
21676 2021-02-10 Jakub Jelinek <jakub@redhat.com>
21678 PR middle-end/99007
21679 * gimplify.c (gimplify_scan_omp_clauses): For MEM_REF on reductions,
21680 temporarily disable gimplify_ctxp->into_ssa around gimplify_expr
21683 2021-02-10 Richard Biener <rguenther@suse.de>
21686 * ipa-pure-const.c (propagate_malloc): Use an auto_vec<>
21689 2021-02-10 Richard Biener <rguenther@suse.de>
21691 PR tree-optimization/99024
21692 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Only
21693 clear loop->aux if it is associated with the destroyed loop_vinfo.
21695 2021-02-10 Martin Liska <mliska@suse.cz>
21697 PR tree-optimization/99002
21698 * gimple-if-to-switch.cc (find_conditions): Fix memory leak
21701 2021-02-10 Martin Liska <mliska@suse.cz>
21704 * ipa-icf.c (sem_item::add_reference): Fix memory leak when
21705 a reference exists.
21707 2021-02-10 Jakub Jelinek <jakub@redhat.com>
21710 * dwarf2out.c (prune_unused_types_walk): Mark DW_TAG_variable DIEs
21711 at class scope for DWARF5+.
21713 2021-02-09 Eric Botcazou <ebotcazou@adacore.com>
21715 PR rtl-optimization/96015
21716 * reorg.c (skip_consecutive_labels): Minor comment tweaks.
21717 (relax_delay_slots): When deleting a jump to the next active
21718 instruction over a barrier, first delete the barrier if the
21719 jump is the only way to reach the target label.
21721 2021-02-09 Andre Vieira <andre.simoesdiasvieira@arm.com>
21723 * config/aarch64/aarch64-cost-tables.h: Add entries for vect.mul.
21724 * config/aarch64/aarch64.c (aarch64_rtx_mult_cost): Use vect.mul for
21725 vector multiplies and vect.alu for SSRA.
21726 * config/arm/aarch-common-protos.h (struct vector_cost_table): Define
21727 vect.mul cost field.
21728 * config/arm/aarch-cost-tables.h: Add entries for vect.mul.
21729 * config/arm/arm.c: Likewise.
21731 2021-02-09 Richard Biener <rguenther@suse.de>
21733 PR tree-optimization/98863
21734 * tree-ssa-sccvn.h (vn_avail::next_undo): Add.
21735 * tree-ssa-sccvn.c (last_pushed_avail): New global.
21736 (rpo_elim::eliminate_push_avail): Chain pushed avails.
21737 (unwind_state::avail_top): Add.
21738 (do_unwind): Rewrite unwinding of avail entries.
21739 (do_rpo_vn): Initialize last_pushed_avail and
21740 avail_top of the undo state.
21742 2021-02-09 Jakub Jelinek <jakub@redhat.com>
21744 PR middle-end/99004
21745 * calls.c (maybe_warn_rdwr_sizes): Change s0 and s1 type from
21746 const char * to char * and free those pointers after use.
21748 2021-02-09 Richard Biener <rguenther@suse.de>
21750 PR tree-optimization/99017
21751 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Allow
21752 zero vector cost entries.
21754 2021-02-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
21756 PR middle-end/98974
21757 * tree-vect-stmts.c (vectorizable_condition): Remove shadow vec_num
21758 parameter in vectorizable_condition.
21760 2021-02-08 Richard Biener <rguenther@suse.de>
21763 * tree.c (walk_tree_1): Walk VECTOR_CST elements.
21765 2021-02-08 Martin Liska <mliska@suse.cz>
21768 * cfgexpand.c (pass_expand::execute): Parse per-function option
21769 flag_patchable_function_entry and use it.
21770 * common.opt: Remove function_entry_patch_area_size and
21771 function_entry_patch_area_start global variables.
21772 * opts.c (parse_and_check_patch_area): New function.
21773 (common_handle_option): Use it.
21774 * opts.h (parse_and_check_patch_area): New function.
21775 * toplev.c (process_options): Parse and use
21776 function_entry_patch_area_size.
21778 2021-02-08 Martin Sebor <msebor@redhat.com>
21780 * doc/extend.texi (attribute malloc): Correct typos.
21782 2021-02-05 Nathan Sidwell <nathan@acm.org>
21785 * gcc.c (driver::maybe_run_linker): Check for input file
21786 accessibility if not linking.
21788 2021-02-05 Richard Biener <rguenther@suse.de>
21790 PR tree-optimization/98855
21791 * tree-vectorizer.h (add_stmt_cost): New overload.
21792 * tree-vect-slp.c (li_cost_vec_cmp): New.
21793 (vect_bb_slp_scalar_cost): Cost individual loop regions
21794 separately. Account for the scalar instance root stmt.
21796 2021-02-05 Tom de Vries <tdevries@suse.de>
21799 * tree-switch-conversion.c (jump_table_cluster::emit): Add loc
21801 (bit_test_cluster::emit): Reuse location_t for newly created
21803 (switch_decision_tree::try_switch_expansion): Preserve
21805 * tree-switch-conversion.h: Change function signatures.
21807 2021-02-05 Jakub Jelinek <jakub@redhat.com>
21810 * config/i386/i386-options.c (m_NONE, m_ALL): Define.
21811 * config/i386/x86-tune.def (X86_TUNE_BRANCH_PREDICTION_HINTS,
21812 X86_TUNE_PROMOTE_QI_REGS): Use m_NONE instead of 0U.
21813 (X86_TUNE_QIMODE_MATH): Use m_ALL instead of ~0U.
21815 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21817 * config/aarch64/aarch64-simd-builtins.def (get_high): Define builtin.
21818 * config/aarch64/aarch64-simd.md (aarch64_get_high<mode>): Define.
21819 * config/aarch64/arm_neon.h (__GET_HIGH): Delete.
21820 (vget_high_f16): Reimplement using new builtin.
21821 (vget_high_f32): Likewise.
21822 (vget_high_f64): Likewise.
21823 (vget_high_p8): Likewise.
21824 (vget_high_p16): Likewise.
21825 (vget_high_p64): Likewise.
21826 (vget_high_s8): Likewise.
21827 (vget_high_s16): Likewise.
21828 (vget_high_s32): Likewise.
21829 (vget_high_s64): Likewise.
21830 (vget_high_u8): Likewise.
21831 (vget_high_u16): Likewise.
21832 (vget_high_u32): Likewise.
21833 (vget_high_u64): Likewise.
21835 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21837 * config/aarch64/aarch64-simd-builtins.def (get_low): Define builtin.
21838 * config/aarch64/aarch64-simd.md (aarch64_get_low<mode>): Define.
21839 * config/aarch64/arm_neon.h (__GET_LOW): Delete.
21840 (vget_low_f16): Reimplement using new builtin.
21841 (vget_low_f32): Likewise.
21842 (vget_low_f64): Likewise.
21843 (vget_low_p8): Likewise.
21844 (vget_low_p16): Likewise.
21845 (vget_low_p64): Likewise.
21846 (vget_low_s8): Likewise.
21847 (vget_low_s16): Likewise.
21848 (vget_low_s32): Likewise.
21849 (vget_low_s64): Likewise.
21850 (vget_low_u8): Likewise.
21851 (vget_low_u16): Likewise.
21852 (vget_low_u32): Likewise.
21853 (vget_low_u64): Likewise.
21855 2021-02-05 Kito Cheng <kito.cheng@sifive.com>
21857 * gcc.c (print_multilib_info): Check all required argument is provided
21860 2021-02-05 liuhongt <hongtao.liu@intel.com>
21863 * config/i386/i386-expand.c (ix86_expand_sse_cmp): Don't
21864 generate integer mask comparison for 128/256-bits vector when
21865 op_true/op_false is NULL_RTX or CONSTM1_RTX/CONST0_RTX. Also
21866 delete redundant !maskcmp condition.
21867 (ix86_expand_int_vec_cmp): Ditto but no redundant deletion
21869 (ix86_expand_sse_movcc): Delete definition of maskcmp, add the
21870 condition directly to if (maskcmp), add extra check for
21871 cmpmode, it should be MODE_INT.
21872 (ix86_expand_fp_vec_cmp): Pass NULL to ix86_expand_sse_cmp's
21873 parameters op_true/op_false.
21874 (ix86_use_mask_cmp_p): New.
21876 2021-02-05 liuhongt <hongtao.liu@intel.com>
21879 * config/i386/x86-tune.def (X86_TUNE_AVX256_UNALIGNED_LOAD_OPTIMAL):
21880 Remove m_GENERIC from ~list.
21881 (X86_TUNE_AVX256_UNALIGNED_STORE_OPTIMAL): Ditto.
21883 2021-02-04 David Malcolm <dmalcolm@redhat.com>
21886 * diagnostic-show-locus.c (compatible_locations_p): Require
21887 locations in the same macro map to be either both from the
21888 macro definition, or both from the macro arguments.
21890 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
21892 * config/aarch64/aarch64-simd-builtins.def: Add
21893 [su]mull_hi_lane[q] builtin generator macros.
21894 * config/aarch64/aarch64-simd.md
21895 (aarch64_<su>mull_hi_lane<mode>_insn): Define.
21896 (aarch64_<su>mull_hi_lane<mode>): Define.
21897 (aarch64_<su>mull_hi_laneq<mode>_insn): Define.
21898 (aarch64_<su>mull_hi_laneq<mode>): Define.
21899 * config/aarch64/arm_neon.h (vmull_high_lane_s16): Use RTL
21900 builtin instead of inline asm.
21901 (vmull_high_lane_s32): Likewise.
21902 (vmull_high_lane_u16): Likewise.
21903 (vmull_high_lane_u32): Likewise.
21904 (vmull_high_laneq_s16): Likewise.
21905 (vmull_high_laneq_s32): Likewise.
21906 (vmull_high_laneq_u16): Likewise.
21907 (vmull_high_laneq_u32): Liekwise.
21909 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
21911 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_hi_n
21912 builtin generator macros.
21913 * config/aarch64/aarch64-simd.md
21914 (aarch64_<su>mull_hi_n<mode>_insn): Define.
21915 (aarch64_<su>mull_hi_n<mode>): Define.
21916 * config/aarch64/arm_neon.h (vmull_high_n_s16): Use RTL builtin
21917 instead of inline asm.
21918 (vmull_high_n_s32): Likewise.
21919 (vmull_high_n_u16): Likewise.
21920 (vmull_high_n_u32): Likewise.
21922 2021-02-04 Richard Biener <rguenther@suse.de>
21924 PR tree-optimization/98855
21925 * tree-vect-loop.c (vectorizable_phi): Do not cost
21926 single-argument PHIs.
21927 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Likewise.
21928 * tree-vect-stmts.c (vectorizable_bswap): Also perform
21929 costing for SLP operation.
21931 2021-02-04 Martin Liska <mliska@suse.cz>
21933 * doc/extend.texi: Mention -mprefer-vector-width in target
21936 2021-02-03 Martin Sebor <msebor@redhat.com>
21938 PR tree-optimization/98937
21939 * tree-ssa-strlen.c (strlen_dom_walker::~strlen_dom_walker): Define.
21940 Flush pointer_query cache.
21942 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
21944 * config/rs6000/genfusion.pl (gen_2logical): Add missing
21945 fixes based on patch review.
21946 * config/rs6000/fusion.md: Regenerate file.
21948 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
21950 * config/rs6000/t-rs6000: Comment out auto generation of
21953 2021-02-03 Andrew Stubbs <ams@codesourcery.com>
21955 * config/gcn/gcn-opts.h (enum processor_type): Add PROCESSOR_GFX908.
21956 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Add gfx908.
21957 (output_file_start): Add gfx908.
21958 * config/gcn/gcn.opt (gpu_type): Add gfx908.
21959 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Add march=gfx908.
21960 (MULTILIB_DIRNAMES): Add gfx908.
21961 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): New define.
21962 (main): Recognize gfx908.
21963 * config/gcn/t-omp-device: Add gfx908.
21965 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
21967 * config/aarch64/aarch64-simd-builtins.def: Add
21968 [su]mlsl_hi_lane[q] builtin macro generators.
21969 * config/aarch64/aarch64-simd.md
21970 (aarch64_<su>mlsl_hi_lane<mode>_insn): Define.
21971 (aarch64_<su>mlsl_hi_lane<mode>): Define.
21972 (aarch64_<su>mlsl_hi_laneq<mode>_insn): Define.
21973 (aarch64_<su>mlsl_hi_laneq<mode>): Define.
21974 * config/aarch64/arm_neon.h (vmlsl_high_lane_s16): Use RTL
21975 builtin instead of inline asm.
21976 (vmlsl_high_lane_s32): Likewise.
21977 (vmlsl_high_lane_u16): Likewise.
21978 (vmlsl_high_lane_u32): Likewise.
21979 (vmlsl_high_laneq_s16): Likewise.
21980 (vmlsl_high_laneq_s32): Likewise.
21981 (vmlsl_high_laneq_u16): Likewise.
21982 (vmlsl_high_laneq_u32): Likewise.
21983 (vmlal_high_laneq_u32): Likewise.
21985 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
21987 * config/aarch64/aarch64-simd-builtins.def: Add
21988 [su]mlal_hi_lane[q] builtin generator macros.
21989 * config/aarch64/aarch64-simd.md
21990 (aarch64_<su>mlal_hi_lane<mode>_insn): Define.
21991 (aarch64_<su>mlal_hi_lane<mode>): Define.
21992 (aarch64_<su>mlal_hi_laneq<mode>_insn): Define.
21993 (aarch64_<su>mlal_hi_laneq<mode>): Define.
21994 * config/aarch64/arm_neon.h (vmlal_high_lane_s16): Use RTL
21995 builtin instead of inline asm.
21996 (vmlal_high_lane_s32): Likewise.
21997 (vmlal_high_lane_u16): Likewise.
21998 (vmlal_high_lane_u32): Likewise.
21999 (vmlal_high_laneq_s16): Likewise.
22000 (vmlal_high_laneq_s32): Likewise.
22001 (vmlal_high_laneq_u16): Likewise.
22002 (vmlal_high_laneq_u32): Likewise.
22004 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22006 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_hi_n
22007 builtin generator macros.
22008 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_hi_n<mode>_insn):
22010 (aarch64_<su>mlsl_hi_n<mode>): Define.
22011 * config/aarch64/arm_neon.h (vmlsl_high_n_s16): Use RTL builtin
22012 instead of inline asm.
22013 (vmlsl_high_n_s32): Likewise.
22014 (vmlsl_high_n_u16): Likewise.
22015 (vmlsl_high_n_u32): Likewise.
22017 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22019 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_hi_n
22020 builtin generator macros.
22021 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_hi_n<mode>_insn):
22023 (aarch64_<su>mlal_hi_n<mode>): Define.
22024 * config/aarch64/arm_neon.h (vmlal_high_n_s16): Use RTL builtin
22025 instead of inline asm.
22026 (vmlal_high_n_s32): Likewise.
22027 (vmlal_high_n_u16): Likewise.
22028 (vmlal_high_n_u32): Likewise.
22030 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22032 * config/aarch64/aarch64-simd-builtins.def: Add RTL builtin
22034 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal_hi<mode>):
22036 (aarch64_<su>mlal_hi<mode>_insn): This.
22037 (aarch64_<su>mlal_hi<mode>): Define.
22038 * config/aarch64/arm_neon.h (vmlal_high_s8): Use RTL builtin
22039 instead of inline asm.
22040 (vmlal_high_s16): Likewise.
22041 (vmlal_high_s32): Likewise.
22042 (vmlal_high_u8): Likewise.
22043 (vmlal_high_u16): Likewise.
22044 (vmlal_high_u32): Likewise.
22046 2021-02-03 Ilya Leoshkevich <iii@linux.ibm.com>
22048 * lra-spills.c (remove_pseudos): Call lra_update_insn_recog_data()
22049 after calling alter_subreg() on a (mem).
22051 2021-02-03 Martin Liska <mliska@suse.cz>
22054 * lto-streamer-out.c (produce_lto_section): Fill up missing
22056 * lto-streamer.h (struct lto_section): Add _padding field.
22058 2021-02-03 Richard Biener <rguenther@suse.de>
22060 * lto-streamer.c (lto_get_section_name): Free temporary
22062 * tree-loop-distribution.c
22063 (loop_distribution::merge_dep_scc_partitions): Free edge data.
22065 2021-02-03 Jakub Jelinek <jakub@redhat.com>
22067 PR middle-end/97487
22068 * ifcvt.c (noce_can_force_operand): New function.
22069 (noce_emit_move_insn): Use it.
22070 (noce_try_sign_mask): Likewise. Formatting fix.
22072 2021-02-03 Jakub Jelinek <jakub@redhat.com>
22074 PR middle-end/97971
22075 * lra-constraints.c (process_alt_operands): For inline asm, don't call
22076 fatal_insn, but instead return false.
22078 2021-02-03 Jakub Jelinek <jakub@redhat.com>
22080 PR tree-optimization/98287
22081 * config/i386/mmx.md (<insn><mode>3): For shifts don't enable expander
22084 2021-02-03 Tamar Christina <tamar.christina@arm.com>
22086 PR tree-optimization/98928
22087 * tree-vect-loop.c (vect_analyze_loop_2): Change
22088 STMT_VINFO_SLP_VECT_ONLY to STMT_VINFO_SLP_VECT_ONLY_PATTERN.
22089 * tree-vect-slp-patterns.c (complex_pattern::build): Likewise.
22090 * tree-vectorizer.h (STMT_VINFO_SLP_VECT_ONLY_PATTERN): New.
22091 (class _stmt_vec_info): Add slp_vect_pattern_only_p.
22093 2021-02-02 Richard Biener <rguenther@suse.de>
22095 * gimple-loop-interchange.cc (prepare_data_references):
22097 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
22098 * tree-ssa-loop-im.c (hoist_memory_references): Likewise.
22099 * tree-vect-stmts.c (vectorizable_condition): Do not
22101 (vectorizable_comparison): Likewise.
22103 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22105 * config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
22106 * config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
22107 * config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
22108 (vrsqrteq_u32): Likewise.
22110 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22112 * config/aarch64/aarch64-simd-builtins.def (sqxtun2): Define builtin.
22113 * config/aarch64/aarch64-simd.md (aarch64_sqxtun2<mode>_le): Define.
22114 (aarch64_sqxtun2<mode>_be): Likewise.
22115 (aarch64_sqxtun2<mode>): Likewise.
22116 * config/aarch64/arm_neon.h (vqmovun_high_s16): Reimplement using builtin.
22117 (vqmovun_high_s32): Likewise.
22118 (vqmovun_high_s64): Likewise.
22119 * config/aarch64/iterators.md (UNSPEC_SQXTUN2): Define.
22121 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22123 * config/aarch64/aarch64-simd-builtins.def (bfdot_lane, bfdot_laneq): Use
22125 (bfmlalb_lane, bfmlalt_lane, bfmlalb_lane_q, bfmlalt_lane_q): Use FP flags.
22127 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22129 * config/aarch64/aarch64-simd-builtins.def (fcmla_lane0, fcmla_lane90,
22130 fcmla_lane180, fcmla_lane270, fcmlaq_lane0, fcmlaq_lane90, fcmlaq_lane180,
22131 fcmlaq_lane270, scvtf, ucvtf, fcvtzs, fcvtzu, scvtfsi, scvtfdi, ucvtfsi,
22132 ucvtfdi, fcvtzshf, fcvtzuhf, fmlal_lane_low, fmlsl_lane_low,
22133 fmlal_laneq_low, fmlsl_laneq_low, fmlalq_lane_low, fmlslq_lane_low,
22134 fmlalq_laneq_low, fmlslq_laneq_low, fmlal_lane_high, fmlsl_lane_high,
22135 fmlal_laneq_high, fmlsl_laneq_high, fmlalq_lane_high, fmlslq_lane_high,
22136 fmlalq_laneq_high, fmlslq_laneq_high): Use FP flags.
22138 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22140 * config/aarch64/aarch64-builtins.c (FLAG_LOAD): Define.
22141 * config/aarch64/aarch64-simd-builtins.def (ld1x2, ld2, ld3, ld4, ld2r,
22142 ld3r, ld4r, ld1, ld1x3, ld1x4): Use LOAD flags.
22144 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22146 * config/aarch64/aarch64-simd-builtins.def (combine, zip1, zip2,
22147 uzp1, uzp2, trn1, trn2, simd_bsl): Use AUTO_FP flags.
22149 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22151 * config/aarch64/aarch64-simd-builtins.def (clrsb, clz, ctz, popcount,
22152 vec_smult_lane_, vec_smlal_lane_, vec_smult_laneq_, vec_smlal_laneq_,
22153 vec_umult_lane_, vec_umlal_lane_, vec_umult_laneq_, vec_umlal_laneq_,
22154 ashl, sshl, ushl, srshl, urshl, sdot_lane, udot_lane, sdot_laneq,
22155 udot_laneq, usdot_lane, usdot_laneq, sudot_lane, sudot_laneq, ashr,
22156 ashr_simd, lshr, lshr_simd, srshr_n, urshr_n, ssra_n, usra_n, srsra_n,
22157 ursra_n, sshll_n, ushll_n, sshll2_n, ushll2_n, ssri_n, usri_n, ssli_n,
22158 ssli_n, usli_n, bswap, rbit, simd_bsl, eor3q, rax1q, xarq, bcaxq): Use
22159 NONE builtin flags.
22161 2021-02-02 Jakub Jelinek <jakub@redhat.com>
22163 PR tree-optimization/98848
22164 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
22165 STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
22167 2021-02-02 Kito Cheng <kito.cheng@sifive.com>
22170 * expr.c: Check mode before calling store_expr.
22172 2021-02-02 Christophe Lyon <christophe.lyon@linaro.org>
22174 * config/arm/iterators.md (supf): Remove VORNQ_S and VORNQ_U.
22176 * config/arm/mve.md (mve_vornq_s<mode>): New entry for vorn
22177 instruction using expression ior.
22178 (mve_vornq_u<mode>): New expander.
22179 (mve_vornq_f<mode>): Use ior code instead of unspec.
22180 * config/arm/unspecs.md (VORNQ_S, VORNQ_U, VORNQ_F): Remove.
22182 2021-02-02 Alexandre Oliva <oliva@adacore.com>
22184 * tree-nested.c (convert_nonlocal_reference_op): Move
22185 current_function_decl restore after re-gimplification.
22186 (convert_local_reference_op): Likewise.
22188 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22190 * config/aarch64/aarch64-simd-builtins.def (rshrn, rshrn2):
22192 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>_insn_le):
22194 (aarch64_rshrn<mode>_insn_be): Likewise.
22195 (aarch64_rshrn<mode>): Likewise.
22196 (aarch64_rshrn2<mode>_insn_le): Likewise.
22197 (aarch64_rshrn2<mode>_insn_be): Likewise.
22198 (aarch64_rshrn2<mode>): Likewise.
22199 * config/aarch64/aarch64.md (unspec): Add UNSPEC_RSHRN.
22200 * config/aarch64/arm_neon.h (vrshrn_high_n_s16): Reimplement
22202 (vrshrn_high_n_s32): Likewise.
22203 (vrshrn_high_n_s64): Likewise.
22204 (vrshrn_high_n_u16): Likewise.
22205 (vrshrn_high_n_u32): Likewise.
22206 (vrshrn_high_n_u64): Likewise.
22207 (vrshrn_n_s16): Likewise.
22208 (vrshrn_n_s32): Likewise.
22209 (vrshrn_n_s64): Likewise.
22210 (vrshrn_n_u16): Likewise.
22211 (vrshrn_n_u32): Likewise.
22212 (vrshrn_n_u64): Likewise.
22214 2021-02-01 Sergei Trofimovich <siarheit@google.com>
22216 PR tree-optimization/98499
22217 * ipa-modref.c (analyze_ssa_name_flags): treat RVO
22218 conservatively and assume all possible side-effects.
22220 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22222 * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi,
22223 vec_unpacku_hi_): Define builtins.
22224 * config/aarch64/arm_neon.h (vmovl_high_s8): Reimplement using
22226 (vmovl_high_s16): Likewise.
22227 (vmovl_high_s32): Likewise.
22228 (vmovl_high_u8): Likewise.
22229 (vmovl_high_u16): Likewise.
22230 (vmovl_high_u32): Likewise.
22232 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22234 * config/aarch64/aarch64-simd-builtins.def (sabdl, uabdl):
22236 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl<mode>): New
22238 * config/aarch64/aarch64.md (unspec): Define UNSPEC_SABDL,
22240 * config/aarch64/arm_neon.h (vabdl_s8): Reimplemet using
22242 (vabdl_s16): Likewise.
22243 (vabdl_s32): Likewise.
22244 (vabdl_u8): Likewise.
22245 (vabdl_u16): Likewise.
22246 (vabdl_u32): Likewise.
22247 * config/aarch64/iterators.md (ABDL): New int iterator.
22248 (sur): Handle UNSPEC_SABDL, UNSPEC_UABDL.
22250 2021-02-01 Martin Sebor <msebor@redhat.com>
22252 * tree.h (BLOCK_VARS): Add comment.
22253 (BLOCK_SUBBLOCKS): Same.
22254 (BLOCK_SUPERCONTEXT): Same.
22255 (BLOCK_ABSTRACT_ORIGIN): Same.
22256 (inlined_function_outer_scope_p): Same.
22258 2021-02-01 Martin Sebor <msebor@redhat.com>
22260 PR middle-end/97172
22261 * attribs.c (attr_access::free_lang_data): Define new function.
22262 * attribs.h (attr_access::free_lang_data): Declare new function.
22264 2021-02-01 Richard Biener <rguenther@suse.de>
22266 * vec.h (auto_vec::auto_vec): Add memory stat parameters
22268 * bitmap.h (auto_bitmap::auto_bitmap): Likewise.
22270 2021-02-01 Tamar Christina <tamar.christina@arm.com>
22272 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>,
22273 aarch64_<su>mlsl<mode>, aarch64_<su>mlsl_n<mode>): Flip mult operands.
22275 2021-02-01 Richard Biener <rguenther@suse.de>
22277 PR rtl-optimization/98863
22278 * config/i386/i386-features.c (convert_scalars_to_vector):
22279 Set DF_RD_PRUNE_DEAD_DEFS.
22281 2021-01-31 Eric Botcazou <ebotcazou@adacore.com>
22283 * system.h (SIZE_MAX): Define if not already defined.
22285 2021-01-30 Aaron Sawdey <acsawdey@linux.ibm.com>
22287 * config/rs6000/genfusion.pl (gen_2logical): New function to
22288 generate patterns for logical-logical fusion.
22289 * config/rs6000/fusion.md: Regenerated patterns.
22290 * config/rs6000/rs6000-cpus.def: Add
22291 OPTION_MASK_P10_FUSION_2LOGICAL.
22292 * config/rs6000/rs6000.c (rs6000_option_override_internal):
22293 Enable logical-logical fusion for p10.
22294 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2logical.
22296 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
22298 * config/rs6000/rs6000.opt: Add periods to new AIX options.
22300 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
22302 * config/rs6000/rs6000.opt (mabi=vec-extabi): New.
22303 (mabi=vec-default): New.
22304 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
22305 __EXTABI__ for AIX Vector extended ABI.
22306 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print AIX Vector
22308 (conditional_register_usage): If AIX vec_extabi enabled, vs20-vs31
22310 * doc/invoke.texi (PowerPC mabi): Add AIX vec-extabi and vec-default.
22312 2021-01-30 Jakub Jelinek <jakub@redhat.com>
22314 * config/i386/i386-features.c (remove_partial_avx_dependency): Clear
22315 DF_DEFER_INSN_RESCAN after calling df_process_deferred_rescans.
22317 2021-01-29 Vladimir N. Makarov <vmakarov@redhat.com>
22320 * lra-constraints.c (in_class_p): Don't narrow class only for REG
22323 2021-01-29 Will Schmidt <will_schmidt@vnet.ibm.com>
22325 * config/rs6000/rs6000-call.c (rs6000_expand_binup_builtin): Add
22326 clauses for CODE_FOR_vsx_xvcvuxddp_scale and
22327 CODE_FOR_vsx_xvcvsxddp_scale to the parameter checking code.
22329 2021-01-29 Andrew MacLeod <amacleod@redhat.com>
22331 PR tree-optimization/98866
22332 * gimple-range-gori.h (gori_compute:set_range_invariant): New.
22333 * gimple-range-gori.cc (gori_map::set_range_invariant): New.
22334 (gori_map::m_maybe_invariant): Rename from all_outgoing.
22335 (gori_map::gori_map): Rename all_outgoing to m_maybe_invariant.
22336 (gori_map::is_export_p): Ditto.
22337 (gori_map::calculate_gori): Ditto.
22338 (gori_compute::set_range_invariant): New.
22339 * gimple-range.cc (gimple_ranger::range_of_stmt): Set range
22340 invariant for pointers evaluating to [1, +INF].
22342 2021-01-29 Richard Biener <rguenther@suse.de>
22344 PR rtl-optimization/98863
22345 * config/i386/i386-features.c (remove_partial_avx_dependency):
22346 Do not perform DF analysis.
22347 (pass_data_remove_partial_avx_dependency): Remove
22350 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
22352 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_n
22353 builtin generator macros.
22354 * config/aarch64/aarch64-simd.md (aarch64_<su>mull_n<mode>):
22356 * config/aarch64/arm_neon.h (vmull_n_s16): Use RTL builtin
22357 instead of inline asm.
22358 (vmull_n_s32): Likewise.
22359 (vmull_n_u16): Likewise.
22360 (vmull_n_u32): Likewise.
22362 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22364 * config/aarch64/aarch64-simd-builtins.def (sabdl2, uabdl2):
22366 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
22368 (aarch64_<sur>abdl2<mode>): ... This.
22369 (<sur>sadv16qi): Adjust use of above.
22370 * config/aarch64/arm_neon.h (vabdl_high_s8): Reimplement using
22372 (vabdl_high_s16): Likewise.
22373 (vabdl_high_s32): Likewise.
22374 (vabdl_high_u8): Likewise.
22375 (vabdl_high_u16): Likewise.
22376 (vabdl_high_u32): Likewise.
22378 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22380 * config/aarch64/aarch64-simd-builtins.def (sabal2): Define
22382 (uabal2): Likewise.
22383 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal2<mode>): New
22385 * config/aarch64/aarch64.md (unspec): Add UNSPEC_SABAL2 and
22387 * config/aarch64/arm_neon.h (vabal_high_s8): Reimplement using
22389 (vabal_high_s16): Likewise.
22390 (vabal_high_s32): Likewise.
22391 (vabal_high_u8): Likewise.
22392 (vabal_high_u16): Likewise.
22393 (vabal_high_u32): Likewise.
22394 * config/aarch64/iterators.md (ABAL2): New mode iterator.
22395 (sur): Handle UNSPEC_SABAL2, UNSPEC_UABAL2.
22397 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22399 * config/aarch64/aarch64-simd-builtins.def (sabal): Define
22402 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal<mode>_4):
22404 (aarch64_<sur>abal<mode>): ... This
22405 (<sur>sadv16qi): Adust use of the above.
22406 * config/aarch64/arm_neon.h (vabal_s8): Reimplement using
22408 (vabal_s16): Likewise.
22409 (vabal_s32): Likewise.
22410 (vabal_u8): Likewise.
22411 (vabal_u16): Likewise.
22412 (vabal_u32): Likewise.
22414 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22416 * config/aarch64/aarch64-simd-builtins.def (saddlv, uaddlv):
22418 * config/aarch64/aarch64-simd.md (aarch64_<su>addlv<mode>):
22420 * config/aarch64/arm_neon.h (vaddlv_s8): Reimplement using
22422 (vaddlv_s16): Likewise.
22423 (vaddlv_u8): Likewise.
22424 (vaddlv_u16): Likewise.
22425 (vaddlvq_s8): Likewise.
22426 (vaddlvq_s16): Likewise.
22427 (vaddlvq_s32): Likewise.
22428 (vaddlvq_u8): Likewise.
22429 (vaddlvq_u16): Likewise.
22430 (vaddlvq_u32): Likewise.
22431 (vaddlv_s32): Likewise.
22432 (vaddlv_u32): Likewise.
22433 * config/aarch64/iterators.md (VDQV_L): New mode iterator.
22434 (unspec): Add UNSPEC_SADDLV, UNSPEC_UADDLV.
22435 (Vwstype): New mode attribute.
22437 (VWIDE_S): Likewise.
22438 (USADDLV): New int iterator.
22439 (su): Handle UNSPEC_SADDLV, UNSPEC_UADDLV.
22441 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
22443 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_lane[q]
22444 builtin generator macros.
22445 * config/aarch64/aarch64-simd.md (aarch64_vec_<su>mlsl_lane<Qlane>):
22447 * config/aarch64/arm_neon.h (vmlsl_lane_s16): Use RTL builtin
22448 instead of inline asm.
22449 (vmlsl_lane_s32): Likewise.
22450 (vmlsl_lane_u16): Likewise.
22451 (vmlsl_lane_u32): Likewise.
22452 (vmlsl_laneq_s16): Likewise.
22453 (vmlsl_laneq_s32): Likewise.
22454 (vmlsl_laneq_u16): Likewise.
22455 (vmlsl_laneq_u32): Likewise.
22457 2021-01-29 Richard Biener <rguenther@suse.de>
22459 * doc/invoke.texi (--param max-gcse-memory): Document unit
22461 * gcse.c (gcse_or_cprop_is_too_expensive): Adjust.
22462 * params.opt (--param max-gcse-memory): Adjust default and
22463 document unit of size.
22465 2021-01-29 Richard Biener <rguenther@suse.de>
22467 PR rtl-optimization/98863
22468 * gcse.c (gcse_or_cprop_is_too_expensive): Use unsigned
22469 HOST_WIDE_INT for the memory estimate.
22471 2021-01-29 Bin Cheng <bin.cheng@linux.alibaba.com>
22472 Richard Biener <rguenther@suse.de>
22474 PR tree-optimization/97627
22475 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
22476 Do not analyze fake edges.
22478 2021-01-29 Richard Biener <rguenther@suse.de>
22480 PR rtl-optimization/98144
22481 * df.h (df_mir_bb_info): Add con_visited member.
22482 * df-problems.c (df_mir_alloc): Initialize con_visited,
22483 do not fully populate IN and OUT.
22484 (df_mir_reset): Likewise.
22485 (df_mir_confluence_0): Set con_visited.
22486 (df_mir_confluence_n): Properly handle implicitely
22487 fully populated IN and OUT as designated by con_visited
22488 and update con_visited accordingly.
22490 2021-01-29 Jakub Jelinek <jakub@redhat.com>
22493 * config/arm/vec-common.md (mve_vshlq_<supf><mode>,
22494 vashl<mode>3, vashr<mode>3, vlshr<mode>3): Add
22495 && !TARGET_REALLY_IWMMXT to conditions.
22497 2021-01-29 Jakub Jelinek <jakub@redhat.com>
22500 * cfgbuild.c (find_bb_boundaries): Reset debug_insn when seeing
22503 2021-01-28 Marek Polacek <polacek@redhat.com>
22506 * stor-layout.c (finalize_type_size): If we reset TYPE_USER_ALIGN in
22507 the main variant, maybe reset it in its variants too.
22508 * tree.c (check_base_type): Return true only if TYPE_USER_ALIGN match.
22509 (check_aligned_type): Check if TYPE_USER_ALIGN match.
22511 2021-01-28 Christophe Lyon <christophe.lyon@linaro.org>
22514 * config/arm/arm.c (arm_rtx_costs_internal): Adjust cost of vector
22515 of constant zero for comparisons.
22517 2021-01-28 Michael Meissner <meissner@linux.ibm.com>
22519 * config/rs6000/rs6000.c (rs6000_mangle_decl_assembler_name): Add
22520 support for mapping built-in function names for long double
22521 built-in functions if long double is IEEE 128-bit.
22523 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
22525 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_n
22526 builtin generator macros.
22527 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_n<mode>):
22529 * config/aarch64/arm_neon.h (vmlsl_n_s16): Use RTL builtin
22530 instead of inline asm.
22531 (vmlsl_n_s32): Likewise.
22532 (vmlsl_n_u16): Likewise.
22533 (vmlsl_n_u32): Likewise.
22535 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
22537 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_n
22538 builtin generator macros.
22539 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>):
22541 * config/aarch64/arm_neon.h (vmlal_n_s16): Use RTL builtin
22542 instead of inline asm.
22543 (vmlal_n_s32): Likewise.
22544 (vmlal_n_u16): Likewise.
22545 (vmlal_n_u32): Likewise.
22547 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22549 * config/aarch64/aarch64-simd-builtins.def (shrn2): Define
22551 * config/aarch64/aarch64-simd.md (aarch64_shrn2<mode>_insn_le):
22553 (aarch64_shrn2<mode>_insn_be): Likewise.
22554 (aarch64_shrn2<mode>): Likewise.
22555 * config/aarch64/arm_neon.h (vshrn_high_n_s16): Reimlplement
22557 (vshrn_high_n_s32): Likewise.
22558 (vshrn_high_n_s64): Likewise.
22559 (vshrn_high_n_u16): Likewise.
22560 (vshrn_high_n_u32): Likewise.
22561 (vshrn_high_n_u64): Likewise.
22563 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22565 * config/aarch64/aarch64-simd-builtins.def (shrn): Define
22567 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le):
22569 (aarch64_shrn<mode>_insn_be): Likewise.
22570 (aarch64_shrn<mode>): Likewise.
22571 * config/aarch64/arm_neon.h (vshrn_n_s16): Reimplement using
22573 (vshrn_n_s32): Likewise.
22574 (vshrn_n_s64): Likewise.
22575 (vshrn_n_u16): Likewise.
22576 (vshrn_n_u32): Likewise.
22577 (vshrn_n_u64): Likewise.
22578 * config/aarch64/iterators.md (vn_mode): New mode attribute.
22580 2021-01-28 Richard Biener <rguenther@suse.de>
22582 PR rtl-optimization/80960
22583 * dse.c (check_mem_read_rtx): Call get_addr on the
22586 2021-01-28 Xionghu Luo <luoxhu@linux.ibm.com>
22587 David Edelsohn <dje.gcc@gmail.com>
22590 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22591 Don't generate VIEW_CONVERT_EXPR for fcode ALTIVEC_BUILTIN_VEC_INSERT
22593 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
22595 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Remove the
22596 wrapper call rs6000_expand_vector_set_var for cleanup. Call
22597 rs6000_expand_vector_set_var_p9 and rs6000_expand_vector_set_var_p8
22599 (rs6000_expand_vector_set_var): Delete.
22600 (rs6000_expand_vector_set_var_p9): Make static.
22601 (rs6000_expand_vector_set_var_p8): Make static.
22603 2021-01-28 Xing GUO <higuoxing@gmail.com>
22605 * common/config/riscv/riscv-common.c
22606 (riscv_subset_list::parsing_subset_version): Fix -march option parsing
22607 when `p` extension exists.
22609 2021-01-27 Vladimir N. Makarov <vmakarov@redhat.com>
22611 PR rtl-optimization/97684
22612 * ira.c (ira): Call ira_set_pseudo_classes before
22613 update_equiv_regs when it is necessary.
22615 2021-01-27 Jakub Jelinek <jakub@redhat.com>
22618 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Use
22619 %w0, %w1 and %2 instead of %0, %1 and %2.
22621 2021-01-27 Aaron Sawdey <acsawdey@linux.ibm.com>
22623 * config/rs6000/genfusion.pl: New script to generate
22624 define_insn_and_split patterns so combine can arrange fused
22625 instructions next to each other.
22626 * config/rs6000/fusion.md: New file, generated fused instruction
22627 patterns for combine.
22628 * config/rs6000/predicates.md (const_m1_to_1_operand): New predicate.
22629 (non_update_memory_operand): New predicate.
22630 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_P10_FUSION and
22631 OPTION_MASK_P10_FUSION_LD_CMPI to ISA_3_1_MASKS_SERVER and
22633 * config/rs6000/rs6000-protos.h (address_is_non_pfx_d_or_x): Add
22635 * config/rs6000/rs6000.c (rs6000_option_override_internal):
22636 Automatically set OPTION_MASK_P10_FUSION and
22637 OPTION_MASK_P10_FUSION_LD_CMPI if target is power10.
22638 (rs600_opt_masks): Allow -mpower10-fusion
22639 in function attributes.
22640 (address_is_non_pfx_d_or_x): New function.
22641 * config/rs6000/rs6000.h: Add MASK_P10_FUSION.
22642 * config/rs6000/rs6000.md: Include fusion.md.
22643 * config/rs6000/rs6000.opt: Add -mpower10-fusion
22644 and -mpower10-fusion-ld-cmpi.
22645 * config/rs6000/t-rs6000: Add dependencies involving fusion.md.
22647 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
22649 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal
22650 builtin generator macros.
22651 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal<mode>):
22653 (aarch64_<su>mlal<mode>): This.
22654 * config/aarch64/arm_neon.h (vmlal_s8): Use RTL builtin
22655 instead of inline asm.
22656 (vmlal_s16): Likewise.
22657 (vmlal_s32): Likewise.
22658 (vmlal_u8): Likewise.
22659 (vmlal_u16): Likewise.
22660 (vmlal_u32): Likewise.
22662 2021-01-27 Richard Biener <rguenther@suse.de>
22664 PR tree-optimization/98854
22665 * tree-vect-slp.c (vect_build_slp_tree_2): Also build
22666 PHIs from scalars when the number of CTORs matches the
22667 number of children.
22669 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
22671 * config/aarch64/aarch64-simd-builtins.def: Add mls_n builtin
22673 * config/aarch64/aarch64-simd.md (*aarch64_mls_elt_merge<mode>):
22675 (aarch64_mls_n<mode>): This.
22676 * config/aarch64/arm_neon.h (vmls_n_s16): Use RTL builtin
22678 (vmls_n_s32): Likewise.
22679 (vmls_n_u16): Likewise.
22680 (vmls_n_u32): Likewise.
22681 (vmlsq_n_s16): Likewise.
22682 (vmlsq_n_s32): Likewise.
22683 (vmlsq_n_u16): Likewise.
22684 (vmlsq_n_u32): Likewise.
22686 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
22688 * config/aarch64/aarch64-simd-builtins.def: Add mls builtin
22690 * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather
22692 (vmls_s16): Likewise.
22693 (vmls_s32): Likewise.
22694 (vmls_u8): Likewise.
22695 (vmls_u16): Likewise.
22696 (vmls_u32): Likewise.
22697 (vmlsq_s8): Likewise.
22698 (vmlsq_s16): Likewise.
22699 (vmlsq_s32): Likewise.
22700 (vmlsq_u8): Likewise.
22701 (vmlsq_u16): Likewise.
22702 (vmlsq_u32): Likewise.
22704 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
22706 * config/aarch64/aarch64-simd-builtins.def: Add mla_n builtin
22708 * config/aarch64/aarch64-simd.md (*aarch64_mla_elt_merge<mode>):
22710 (aarch64_mla_n<mode>): This.
22711 * config/aarch64/arm_neon.h (vmla_n_s16): Use RTL builtin
22713 (vmla_n_s32): Likewise.
22714 (vmla_n_u16): Likewise.
22715 (vmla_n_u32): Likewise.
22716 (vmlaq_n_s16): Likewise.
22717 (vmlaq_n_s32): Likewise.
22718 (vmlaq_n_u16): Likewise.
22719 (vmlaq_n_u32): Likewise.
22721 2021-01-27 liuhongt <hongtao.liu@intel.com>
22724 * config/i386/sse.md (sse2_gt<mode>3): Drop !TARGET_XOP in condition.
22725 (*sse2_eq<mode>3): Ditto.
22727 2021-01-27 Jakub Jelinek <jakub@redhat.com>
22729 * tree-pass.h (PROP_trees): Rename to ...
22730 (PROP_gimple): ... this.
22731 * cfgexpand.c (pass_data_expand): Replace PROP_trees with PROP_gimple.
22732 * passes.c (execute_function_dump, execute_function_todo,
22733 execute_one_ipa_transform_pass, execute_one_pass): Likewise.
22734 * varpool.c (ctor_for_folding): Likewise.
22736 2021-01-27 Jakub Jelinek <jakub@redhat.com>
22738 PR tree-optimization/97260
22739 * varpool.c: Include tree-pass.h.
22740 (ctor_for_folding): In GENERIC return DECL_INITIAL for TREE_READONLY
22741 non-TREE_SIDE_EFFECTS automatic variables.
22743 2021-01-26 Paul Fee <paul.f.fee@gmail.com>
22745 * doc/cpp.texi (__cplusplus): Document value for -std=c++23
22747 * doc/invoke.texi: Document -std=c++23 and -std=gnu++23.
22748 * dwarf2out.c (highest_c_language): Recognise C++20 and C++23.
22749 (gen_compile_unit_die): Recognise C++23.
22751 2021-01-26 Jakub Jelinek <jakub@redhat.com>
22754 * dwarf2asm.c (dw2_assemble_integer): Cast DWARF2_ADDR_SIZE to int
22757 2021-01-26 Jakub Jelinek <jakub@redhat.com>
22760 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
22761 Use UINTVAL (shft_amnt) and UINTVAL (mask) instead of INTVAL (shft_amnt)
22762 and INTVAL (mask). Add && INTVAL (mask) > 0 condition.
22764 2021-01-26 Richard Biener <rguenther@suse.de>
22766 * gimple-pretty-print.c (dump_binary_rhs): Handle
22767 VEC_WIDEN_{PLUS,MINUS}_{LO,HI}_EXPR.
22769 2021-01-26 Richard Biener <rguenther@suse.de>
22771 PR middle-end/98726
22772 * tree.h (vector_cst_int_elt): Remove.
22773 * tree.c (vector_cst_int_elt): Use poly_wide_int for computations,
22776 2021-01-26 Andrew Stubbs <ams@codesourcery.com>
22778 * config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
22779 for V64DFmode min/max reductions.
22781 2021-01-26 Jakub Jelinek <jakub@redhat.com>
22783 * dwarf2asm.c (dw2_assemble_integer): Handle size twice as large
22784 as DWARF2_ADDR_SIZE if x is not a scalar int by emitting it as
22785 two halves, one with x and the other with const0_rtx, ordered
22786 depending on endianity.
22788 2021-01-26 Alexandre Oliva <oliva@adacore.com>
22790 * gimplify.c (gimplify_decl_expr): Skip asan marking calls for
22791 temporaries not seen in binding block, and not about to be
22792 added as gimple variables.
22794 2021-01-25 Martin Sebor <msebor@redhat.com>
22797 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust warning text.
22799 2021-01-25 Martin Liska <mliska@suse.cz>
22801 * value-prof.c (get_nth_most_common_value): Use %s instead
22804 2021-01-25 Jakub Jelinek <jakub@redhat.com>
22807 * configure.ac (HAVE_AS_GDWARF_5_DEBUG_FLAG): Only define if
22808 readelf -wi is able to read the emitted .debug_info back.
22809 * configure: Regenerated.
22811 2021-01-25 Martin Liska <mliska@suse.cz>
22813 PR gcov-profile/98739
22814 * common.opt: Add missing sign symbol.
22815 * value-prof.c (get_nth_most_common_value): Restore handling
22816 of PROFILE_REPRODUCIBILITY_PARALLEL_RUNS and
22817 PROFILE_REPRODUCIBILITY_MULTITHREADED.
22819 2021-01-25 Richard Biener <rguenther@suse.de>
22821 PR middle-end/98807
22822 * tree.c (vector_element_bits): Always use precision of
22823 the element type for boolean vectors.
22825 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
22827 * config/rtems.h (STARTFILE_SPEC): Remove qnolinkcmds.
22828 (ENDFILE_SPEC): Evaluate qnolinkcmds.
22830 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
22832 * config/rtems.h (STARTFILE_SPEC): Remove nostdlib and
22833 nostartfiles handling since this is already done by
22834 LINK_COMMAND_SPEC. Evaluate qnolinkcmds.
22835 (ENDFILE_SPEC): Remove nostdlib and nostartfiles handling since this
22836 is already done by LINK_COMMAND_SPEC.
22837 (LIB_SPECS): Remove nostdlib and nodefaultlibs handling since
22838 this is already done by LINK_COMMAND_SPEC. Remove qnolinkcmds
22841 2021-01-25 Jakub Jelinek <jakub@redhat.com>
22844 * fold-const-call.c (host_size_t_cst_p): Renamed to ...
22845 (size_t_cst_p): ... this. Check and store unsigned HOST_WIDE_INT
22846 value rather than host size_t.
22847 (fold_const_call): Change type of s2 from size_t to
22848 unsigned HOST_WIDE_INT. Use size_t_cst_p instead of
22849 host_size_t_cst_p. For strncmp calls, pass MIN (s2, SIZE_MAX)
22850 instead of s2 as last argument.
22852 2021-01-25 Tamar Christina <tamar.christina@arm.com>
22854 * config/arm/iterators.md (rotsplit1, rotsplit2, conj_op, fcmac1,
22855 VCMLA_OP, VCMUL_OP): New.
22856 * config/arm/mve.md (mve_vcmlaq<mve_rot><mode>): Support vec_dup 0.
22857 * config/arm/neon.md (cmul<conj_op><mode>3): New.
22858 * config/arm/unspecs.md (UNSPEC_VCMLA_CONJ, UNSPEC_VCMLA180_CONJ,
22859 UNSPEC_VCMUL_CONJ): New.
22860 * config/arm/vec-common.md (cmul<conj_op><mode>3, arm_vcmla<rot><mode>,
22861 cml<fcmac1><conj_op><mode>4): New.
22863 2021-01-23 Jakub Jelinek <jakub@redhat.com>
22866 * config/rs6000/mmintrin.h (__m64): Add __may_alias__ attribute.
22868 2021-01-22 Jonathan Wright <jonathan.wright@arm.com>
22870 * config/aarch64/aarch64-simd-builtins.def: Add mla builtin
22872 * config/aarch64/arm_neon.h (vmla_s8): Use RTL builtin rather
22874 (vmla_s16): Likewise.
22875 (vmla_s32): Likewise.
22876 (vmla_u8): Likewise.
22877 (vmla_u16): Likewise.
22878 (vmla_u32): Likewise.
22879 (vmlaq_s8): Likewise.
22880 (vmlaq_s16): Likewise.
22881 (vmlaq_s32): Likewise.
22882 (vmlaq_u8): Likewise.
22883 (vmlaq_u16): Likewise.
22884 (vmlaq_u32): Likewise.
22886 2021-01-22 David Malcolm <dmalcolm@redhat.com>
22888 * doc/invoke.texi (GCC_EXTRA_DIAGNOSTIC_OUTPUT): Add @findex
22891 2021-01-22 Jakub Jelinek <jakub@redhat.com>
22894 * dwarf2out.c (output_file_names): For -gdwarf-5, if there are no
22895 filenames to emit, still emit the required 0 index directory and
22896 filename entries that match DW_AT_comp_dir and DW_AT_name of the
22899 2021-01-22 Marek Polacek <polacek@redhat.com>
22902 * doc/invoke.texi: Update C++ ABI Version 15 description.
22904 2021-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22906 PR tree-optimization/98766
22907 * tree-ssa-math-opts.c (convert_mult_to_fma): Use maybe_le when
22908 comparing against type size with param_avoid_fma_max_bits.
22910 2021-01-22 Richard Biener <rguenther@suse.de>
22912 PR middle-end/98793
22913 * tree.c (vector_element_bits): Key single-bit bool vector on
22914 integer mode rather than not vector mode.
22916 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
22919 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22920 Generate ARRAY_REF(VIEW_CONVERT_EXPR) for P8 and later
22922 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): Update
22923 to call different path for P8 and P9.
22924 (rs6000_expand_vector_set_var_p9): New function.
22925 (rs6000_expand_vector_set_var_p8): New function.
22927 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
22931 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22932 Ajdust variable index vec_insert from address dereference to
22933 ARRAY_REF(VIEW_CONVERT_EXPR) tree expression.
22934 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
22936 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): New function.
22938 2021-01-22 Martin Liska <mliska@suse.cz>
22940 PR gcov-profile/98739
22941 * profile.c (compute_value_histograms): Drop time profile for
22942 -fprofile-reproducible=multithreaded.
22944 2021-01-22 Nathan Sidwell <nathan@acm.org>
22946 * gcc.c (process_command): Don't check OPT_SPECIAL_input_file
22949 2021-01-22 Richard Biener <rguenther@suse.de>
22951 PR middle-end/98773
22952 * tree-data-ref.c (initalize_matrix_A): Revert previous
22953 change, retaining failing on HOST_WIDE_INT_MIN CHREC_RIGHT.
22955 2021-01-22 Jakub Jelinek <jakub@redhat.com>
22957 PR tree-optimization/90248
22958 * match.pd (X cmp 0.0 ? 1.0 : -1.0 -> copysign(1, +-X),
22959 X cmp 0.0 ? -1.0 : +1.0 -> copysign(1, -+X)): Remove
22961 (X * (X cmp 0.0 ? 1.0 : -1.0) -> +-abs(X),
22962 X * (X cmp 0.0 ? -1.0 : 1.0) -> +-abs(X)): New simplifications.
22964 2021-01-22 Jakub Jelinek <jakub@redhat.com>
22966 PR tree-optimization/98255
22967 * tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
22968 extend index - low_bound from sizetype's precision rather than index
22970 (get_addr_base_and_unit_offset_1): Likewise.
22971 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
22972 * gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
22974 2021-01-22 Richard Biener <rguenther@suse.de>
22976 PR tree-optimization/98786
22977 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Avoid
22978 adding new uses of abnormals. Verify we deal with a conditional
22981 2021-01-22 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
22984 * optc-save-gen.awk: Add arm_fp16_format to checked_options.
22986 2021-01-22 liuhongt <hongtao.liu@intel.com>
22990 * config/i386/sse.md (VI_128_256): New mode iterator.
22991 (*avx_cmp<mode>3_1, *avx_cmp<mode>3_2, *avx_cmp<mode>3_3,
22992 *avx_cmp<mode>3_4, *avx2_eq<mode>3, *avx2_pcmp<mode>3_1,
22993 *avx2_pcmp<mode>3_2, *avx2_gt<mode>3): New
22994 define_insn_and_split to lower avx512 vector comparison to avx
22995 version when dest is vector.
22996 (*<avx512>_cmp<mode>3,*<avx512>_cmp<mode>3,*<avx512>_ucmp<mode>3):
22997 define_insn_and_split for negating the comparison result.
22998 * config/i386/predicates.md (float_vector_all_ones_operand):
23000 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Use
23001 general NOT operator without UNSPEC_MASKOP.
23003 2021-01-21 Vladimir N. Makarov <vmakarov@redhat.com>
23005 PR rtl-optimization/98777
23006 * lra-int.h (lra_pmode_pseudo): New extern.
23007 * lra.c (lra_pmode_pseudo): New global.
23009 * lra-eliminations.c (eliminate_regs_in_insn): Use it.
23011 2021-01-21 Ilya Leoshkevich <iii@linux.ibm.com>
23013 * fwprop.c (fwprop_propagation::classify_result): Allow
23014 (subreg (mem)) simplifications.
23016 2021-01-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23018 * config/aarch64/aarch64-simd.md (aarch64_sqdml<SBINQOPS:as>l<mode>):
23020 (aarch64_sqdmlal<mode>): ... This...
23021 (aarch64_sqdmlsl<mode>): ... And this.
23022 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Split into...
23023 (aarch64_sqdmlal_lane<mode>): ... This...
23024 (aarch64_sqdmlsl_lane<mode>): ... And this.
23025 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Split into...
23026 (aarch64_sqdmlsl_laneq<mode>): ... This...
23027 (aarch64_sqdmlal_laneq<mode>): ... And this.
23028 (aarch64_sqdml<SBINQOPS:as>l_n<mode>): Split into...
23029 (aarch64_sqdmlsl_n<mode>): ... This...
23030 (aarch64_sqdmlal_n<mode>): ... And this.
23031 (aarch64_sqdml<SBINQOPS:as>l2<mode>_internal): Split into...
23032 (aarch64_sqdmlal2<mode>_internal): ... This...
23033 (aarch64_sqdmlsl2<mode>_internal): ... And this.
23035 2021-01-21 Christophe Lyon <christophe.lyon@linaro.org>
23037 * config/arm/arm_mve.h (__arm_vcmpneq_s8): Fix return type.
23039 2021-01-21 Andrea Corallo <andrea.corallo@arm.com>
23042 * doc/sourcebuild.texi (arm_thumb2_no_arm_v8_1_lob): Document.
23044 2021-01-21 liuhongt <hongtao.liu@intel.com>
23046 PR rtl-optimization/98694
23047 * regcprop.c (copy_value): If SRC had been assigned a mode
23048 narrower than the copy, we can't link DEST into the chain even
23049 they have same hard_regno_nregs(i.e. HImode/SImode in i386
23052 2021-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23054 * config/aarch64/aarch64-simd.md (aarch64_get_lane<mode>):
23055 Convert to define_insn_and_split. Split into simple move when moving
23058 2021-01-20 Segher Boessenkool <segher@kernel.crashing.org>
23060 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Change assert.
23061 Adjust comment. Simplify code.
23063 2021-01-20 Jakub Jelinek <jakub@redhat.com>
23066 * dwarf2out.c (reset_indirect_string): Also reset indirect strings
23067 with DW_FORM_line_strp form.
23068 (prune_unused_types_update_strings): Don't add into debug_str_hash
23069 indirect strings with DW_FORM_line_strp form.
23070 (adjust_name_comp_dir): New function.
23071 (dwarf2out_finish): Call it on CU DIEs after resetting
23072 debug_line_str_hash.
23074 2021-01-20 Vladimir N. Makarov <vmakarov@redhat.com>
23076 PR rtl-optimization/98722
23077 * lra-eliminations.c (eliminate_regs_in_insn): Check that target
23078 has no 3-op add insn to transform insns containing two pluses.
23080 2021-01-20 Richard Biener <rguenther@suse.de>
23082 * hwint.h (add_hwi): New function.
23083 (mul_hwi): Likewise.
23084 * tree-data-ref.c (initialize_matrix_A): Properly translate
23085 tree constants and avoid HOST_WIDE_INT_MIN.
23086 (lambda_matrix_row_add): Avoid undefined integer overflow
23087 and return true on such overflow.
23088 (lambda_matrix_right_hermite): Handle overflow from
23089 lambda_matrix_row_add gracefully. Simplify previous fix.
23090 (analyze_subscript_affine_affine): Likewise.
23092 2021-01-20 Eugene Rozenfeld <erozen@microsoft.com>
23094 PR tree-optimization/96674
23095 * match.pd: New patterns: x < y || y == XXX_MIN --> x <= y - 1
23096 x >= y && y != XXX_MIN --> x > y - 1
23098 2021-01-20 Richard Sandiford <richard.sandiford@arm.com>
23100 PR tree-optimization/98535
23101 * tree-vect-slp.c (duplicate_and_interleave): Use quick_grow_cleared.
23102 If the high and low permutes are the same, remove the high permutes
23103 from the working set and only continue with the low ones.
23105 2021-01-20 Jakub Jelinek <jakub@redhat.com>
23107 PR tree-optimization/98721
23108 * builtins.c (access_ref::inform_access): Don't assume
23109 SSA_NAME_IDENTIFIER must be non-NULL. Print messages about
23110 object whenever allocfn is NULL, rather than only when DECL_P
23111 is true. Use %qE instead of %qD for that. Formatting fixes.
23113 2021-01-20 Richard Biener <rguenther@suse.de>
23115 PR tree-optimization/98758
23116 * tree-data-ref.c (int_divides_p): Use lambda_int arguments.
23117 (lambda_matrix_right_hermite): Avoid undefinedness with
23118 signed integer abs and multiplication.
23119 (analyze_subscript_affine_affine): Use lambda_int.
23121 2021-01-20 David Malcolm <dmalcolm@redhat.com>
23124 * dwarf2out.c (output_line_info): Rename static variable
23125 "generation", moving it out of the function to...
23126 (output_line_info_generation): New.
23127 (init_sections_and_labels): Likewise, renaming the variable to...
23128 (init_sections_and_labels_generation): New.
23129 (dwarf2out_c_finalize): Reset the new variables.
23131 2021-01-19 Martin Sebor <msebor@redhat.com>
23133 PR middle-end/98664
23134 * tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
23135 all functions, even if they're not declared artificial or inline.
23136 * tree.c (tree_inlined_location): Use macro expansion location
23137 only if scope traversal fails to expose one.
23139 2021-01-19 Richard Sandiford <richard.sandiford@arm.com>
23141 PR rtl-optimization/92294
23142 * alias.c (compare_base_symbol_refs): Take an extra parameter
23143 and add the distance between two symbols to it. Enshrine in
23144 comments that -1 means "either 0 or 1, but we can't tell
23145 which at compile time".
23146 (memrefs_conflict_p): Update call accordingly.
23147 (rtx_equal_for_memref_p): Likewise. Take the distance between symbols
23150 2021-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23152 * config/aarch64/aarch64-simd-builtins.def (sqshl, uqshl,
23153 sqrshl, uqrshl, sqadd, uqadd, sqsub, uqsub, suqadd, usqadd, sqmovn,
23154 uqmovn, sqxtn2, uqxtn2, sqabs, sqneg, sqdmlal, sqdmlsl, sqdmlal_lane,
23155 sqdmlsl_lane, sqdmlal_laneq, sqdmlsl_laneq, sqdmlal_n, sqdmlsl_n,
23156 sqdmlal2, sqdmlsl2, sqdmlal2_lane, sqdmlsl2_lane, sqdmlal2_laneq,
23157 sqdmlsl2_laneq, sqdmlal2_n, sqdmlsl2_n, sqdmull, sqdmull_lane,
23158 sqdmull_laneq, sqdmull_n, sqdmull2, sqdmull2_lane, sqdmull2_laneq,
23159 sqdmull2_n, sqdmulh, sqrdmulh, sqdmulh_lane, sqdmulh_laneq,
23160 sqrdmulh_lane, sqrdmulh_laneq, sqshrun_n, sqrshrun_n, sqshrn_n,
23161 uqshrn_n, sqrshrn_n, uqrshrn_n, sqshlu_n, sqshl_n, uqshl_n, sqrdmlah,
23162 sqrdmlsh, sqrdmlah_lane, sqrdmlsh_lane, sqrdmlah_laneq, sqrdmlsh_laneq,
23163 sqmovun): Use NONE flags.
23165 2021-01-19 Richard Biener <rguenther@suse.de>
23168 * ipa-modref.c (analyze_stmt): Only record a summary for a
23171 2021-01-19 Richard Biener <rguenther@suse.de>
23173 PR middle-end/98638
23174 * tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
23176 2021-01-19 Daniel Hellstrom <daniel@gaisler.com>
23178 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
23179 built-in define __FIX_LEON3FT_TN0018.
23181 2021-01-19 Richard Biener <rguenther@suse.de>
23184 * tree-inline.c (tree_function_versioning): Set input_location
23185 to UNKNOWN_LOCATION throughout the function.
23187 2021-01-19 Tobias Burnus <tobias@codesourcery.com>
23190 * omp-low.c (lower_omp_target): Handle nonpointer is_device_ptr.
23192 2021-01-19 Martin Jambor <mjambor@suse.cz>
23195 * ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
23196 whether non-call exceptions allow removal of a statement.
23197 (isra_analyze_call): Pass the appropriate function to
23198 ssa_name_only_returned_p.
23200 2021-01-19 Geng Qi <gengqi@linux.alibaba.com>
23202 * config/riscv/arch-canonicalize (longext_sort): New function for
23203 sorting 'multi-letter'.
23204 * config/riscv/multilib-generator: Adjusting the loop of 'alt' in
23205 'alts'. The 'arch' may not be the first of 'alts'.
23206 (_expand_combination): Add underline for the 'ext' without '*'.
23207 This is because, a single-letter extension can always be treated well
23208 with a '_' prefix, but it cannot be separated out if it is appended
23211 2021-01-18 Vladimir N. Makarov <vmakarov@redhat.com>
23214 * ira.c (ira): Skip abnormal critical edge splitting.
23216 2021-01-18 Jakub Jelinek <jakub@redhat.com>
23218 PR tree-optimization/98727
23219 * tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
23220 second .MUL_OVERFLOW operand for signed multiplication with overflow
23221 checking if the second operand of multiplication is not constant.
23223 2021-01-18 David Edelsohn <dje.gcc@gmail.com>
23225 * doc/invoke.texi (-gdwarf): TPF defaults to version 2 and AIX
23226 defaults to version 4.
23228 2021-01-18 David Malcolm <dmalcolm@redhat.com>
23230 * attribs.h (fndecl_dealloc_argno): New decl.
23231 * builtins.c (call_dealloc_argno): Split out second half of
23233 (fndecl_dealloc_argno): New.
23234 * doc/extend.texi (Common Function Attributes): Document the
23235 interaction between the analyzer and the malloc attribute.
23236 * doc/invoke.texi (Static Analyzer Options): Likewise.
23238 2021-01-17 David Edelsohn <dje.gcc@gmail.com>
23240 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Override
23241 dwarf_version to 4.
23242 * config/rs6000/aix72.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
23244 2021-01-17 Martin Jambor <mjambor@suse.cz>
23247 * cgraph.c (clone_of_p): Check also former_clone_of as we climb
23250 2021-01-17 Mark Wielaard <mark@klomp.org>
23252 * common.opt (gdwarf-): Init(5).
23253 * doc/invoke.texi (-gdwarf): Document default to 5.
23255 2021-01-16 Kwok Cheung Yeung <kcy@codesourcery.com>
23257 * builtin-types.def
23258 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
23260 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
23261 ...this. Add extra argument.
23262 * gimplify.c (omp_default_clause): Ensure that event handle is
23263 firstprivate in a task region.
23264 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_DETACH.
23265 (gimplify_adjust_omp_clauses): Likewise.
23266 * omp-builtins.def (BUILT_IN_GOMP_TASK): Change function type to
23267 BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR.
23268 * omp-expand.c (expand_task_call): Add GOMP_TASK_FLAG_DETACH to flags
23269 if detach clause specified. Add detach argument when generating
23271 * omp-low.c (scan_sharing_clauses): Setup data environment for detach
23273 (finish_taskreg_scan): Move field for variable containing the event
23274 handle to the front of the struct.
23275 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_DETACH. Fix
23277 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
23278 OMP_CLAUSE_DETACH clause.
23279 (convert_local_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
23280 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_DETACH.
23281 * tree.c (omp_clause_num_ops): Add entry for OMP_CLAUSE_DETACH.
23283 (omp_clause_code_name): Add entry for OMP_CLAUSE_DETACH. Fix
23285 (walk_tree_1): Handle OMP_CLAUSE_DETACH.
23287 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
23289 * config/nios2/t-rtems: Reset all MULTILIB_* variables. Shorten
23290 multilib directory names. Use MULTILIB_REQUIRED instead of
23291 MULTILIB_EXCEPTIONS. Add -mhw-mul -mhw-mulx -mhw-div
23292 -mcustom-fpu-cfg=fph2 multilib.
23294 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
23296 * config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
23297 (nios2_init_fpu_configs): Provide register values for new
23298 -mcustom-fpu-cfg=fph2 option variant.
23299 * doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
23302 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
23304 * config/nios2/nios2.c (nios2_custom_check_insns): Remove
23305 custom instruction warnings.
23307 2021-01-16 Jakub Jelinek <jakub@redhat.com>
23309 PR tree-optimization/96669
23310 * match.pd ((CST << x) & 1 -> x == 0): New simplification.
23312 2021-01-16 Jakub Jelinek <jakub@redhat.com>
23314 PR tree-optimization/96271
23315 * passes.def: Pass false argument to first two pass_cd_dce
23316 instances and true to last instance. Add comment that
23317 last instance rewrites no longer addressed locals.
23318 * tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
23320 (pass_cd_dce::set_pass_param): New method.
23321 (pass_cd_dce::execute): Return TODO_update_address_taken from
23322 last cd_dce instance.
23324 2021-01-15 Carl Love <cel@us.ibm.com>
23326 * config/rs6000/altivec.h (vec_mulh, vec_div, vec_dive, vec_mod):
23328 * config/rs6000/altivec.md (VIlong): Move define to file vsx.md.
23329 * config/rs6000/rs6000-builtin.def (DIVES_V4SI, DIVES_V2DI,
23330 DIVEU_V4SI, DIVEU_V2DI, DIVS_V4SI, DIVS_V2DI, DIVU_V4SI,
23331 DIVU_V2DI, MODS_V2DI, MODS_V4SI, MODU_V2DI, MODU_V4SI,
23332 MULHS_V2DI, MULHS_V4SI, MULHU_V2DI, MULHU_V4SI, MULLD_V2DI):
23333 Add builtin define.
23334 (MULH, DIVE, MOD): Add new BU_P10_OVERLOAD_2 definitions.
23335 * config/rs6000/rs6000-call.c (VSX_BUILTIN_VEC_DIV,
23336 VSX_BUILTIN_VEC_DIVE, P10_BUILTIN_VEC_MOD, P10_BUILTIN_VEC_MULH):
23337 New overloaded definitions.
23338 (builtin_function_type) [P10V_BUILTIN_DIVEU_V4SI,
23339 P10V_BUILTIN_DIVEU_V2DI, P10V_BUILTIN_DIVU_V4SI,
23340 P10V_BUILTIN_DIVU_V2DI, P10V_BUILTIN_MODU_V2DI,
23341 P10V_BUILTIN_MODU_V4SI, P10V_BUILTIN_MULHU_V2DI,
23342 P10V_BUILTIN_MULHU_V4SI]: Add case
23343 statement for builtins.
23344 * config/rs6000/rs6000.md (bits): Add new attribute sizes V4SI, V2DI.
23345 * config/rs6000/vsx.md (VIlong): Moved from config/rs6000/altivec.md.
23346 (UNSPEC_VDIVES, UNSPEC_VDIVEU): New unspec definitions.
23347 (vsx_mul_v2di): Add if TARGET_POWER10 statement.
23348 (vsx_udiv_v2di): Add if TARGET_POWER10 statement.
23349 (dives_<mode>, diveu_<mode>, div<mode>3, uvdiv<mode>3,
23350 mods_<mode>, modu_<mode>, mulhs_<mode>, mulhu_<mode>, mulv2di3):
23351 Add define_insn, mode is VIlong.
23352 * doc/extend.texi (vec_mulh, vec_mul, vec_div, vec_dive, vec_mod):
23353 Add builtin descriptions.
23355 2021-01-15 Eric Botcazou <ebotcazou@adacore.com>
23357 * final.c (final_start_function_1): Reset force_source_line.
23359 2021-01-15 Jakub Jelinek <jakub@redhat.com>
23361 PR tree-optimization/96669
23362 * match.pd (((1 << A) & 1) != 0 -> A == 0,
23363 ((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
23364 possibly different power of two constants and to right shift too.
23366 2021-01-15 Jakub Jelinek <jakub@redhat.com>
23368 PR tree-optimization/96681
23369 * match.pd ((x < 0) ^ (y < 0) to (x ^ y) < 0): New simplification.
23370 ((x >= 0) ^ (y >= 0) to (x ^ y) < 0): Likewise.
23371 ((x < 0) ^ (y >= 0) to (x ^ y) >= 0): Likewise.
23372 ((x >= 0) ^ (y < 0) to (x ^ y) >= 0): Likewise.
23374 2021-01-15 Alexandre Oliva <oliva@adacore.com>
23376 * opts.c (gen_command_line_string): Exclude -dumpbase-ext.
23378 2021-01-15 Tamar Christina <tamar.christina@arm.com>
23380 * config/aarch64/aarch64-simd.md (cml<fcmac1><conj_op><mode>4,
23381 cmul<conj_op><mode>3): New.
23382 * config/aarch64/iterators.md (UNSPEC_FCMUL,
23383 UNSPEC_FCMUL180, UNSPEC_FCMLA_CONJ, UNSPEC_FCMLA180_CONJ,
23384 UNSPEC_CMLA_CONJ, UNSPEC_CMLA180_CONJ, UNSPEC_CMUL, UNSPEC_CMUL180,
23385 FCMLA_OP, FCMUL_OP, conj_op, rotsplit1, rotsplit2, fcmac1, sve_rot1,
23386 sve_rot2, SVE2_INT_CMLA_OP, SVE2_INT_CMUL_OP, SVE2_INT_CADD_OP): New.
23387 (rot): Add UNSPEC_FCMUL, UNSPEC_FCMUL180.
23388 (rot_op): Renamed to conj_op.
23389 * config/aarch64/aarch64-sve.md (cml<fcmac1><conj_op><mode>4,
23390 cmul<conj_op><mode>3): New.
23391 * config/aarch64/aarch64-sve2.md (cml<fcmac1><conj_op><mode>4,
23392 cmul<conj_op><mode>3): New.
23394 2021-01-15 David Malcolm <dmalcolm@redhat.com>
23398 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
23399 Escape the tempfile name when constructing the expected output.
23401 2021-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23403 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlsl_hi<mode>):
23405 (aarch64_<su>mlsl_hi<mode>): ... This.
23406 (aarch64_<su>mlsl_hi<mode>): Define.
23407 (*aarch64_<su>mlsl<mode): Rename to...
23408 (aarch64_<su>mlsl<mode): ... This.
23409 * config/aarch64/aarch64-simd-builtins.def (smlsl, umlsl,
23410 smlsl_hi, umlsl_hi): Define builtins.
23411 * config/aarch64/arm_neon.h (vmlsl_high_s8, vmlsl_high_s16,
23412 vmlsl_high_s32, vmlsl_high_u8, vmlsl_high_u16, vmlsl_high_u32,
23413 vmlsl_s8, vmlsl_s16, vmlsl_s32, vmlsl_u8,
23414 vmlsl_u16, vmlsl_u32): Reimplement with builtins.
23416 2021-01-15 Uroš Bizjak <ubizjak@gmail.com>
23418 * config/i386/i386-c.c (ix86_target_macros):
23419 Use cpp_define_formatted for __SIZEOF_FLOAT80__ definition.
23421 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
23424 * config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
23425 * Makefile.in (RTL_SSA_H): New variable.
23426 * config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
23427 * config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
23428 * config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
23430 * config/aarch64/aarch64-cc-fusion.cc: New file.
23432 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
23434 * recog.h (insn_change_watermark::~insn_change_watermark): Avoid
23435 calling cancel_changes for changes that no longer exist.
23437 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
23439 * rtl-ssa/functions.h (function_info::ref_defs): Rename to...
23440 (function_info::reg_defs): ...this.
23441 * rtl-ssa/member-fns.inl (function_info::ref_defs): Rename to...
23442 (function_info::reg_defs): ...this.
23444 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23447 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
23449 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23452 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23455 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
23457 2021-01-15 Richard Biener <rguenther@suse.de>
23459 PR tree-optimization/96376
23460 * tree-vect-stmts.c (get_load_store_type): Disregard alignment
23461 for VMAT_INVARIANT.
23463 2021-01-15 Martin Liska <mliska@suse.cz>
23465 * doc/install.texi: Document that some tests need pytest module.
23466 * doc/sourcebuild.texi: Likewise.
23468 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23471 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
23473 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23475 * config/arm/mve.md (mve_vshrq_n_s<mode>_imm): New entry.
23476 (mve_vshrq_n_u<mode>_imm): Likewise.
23477 * config/arm/neon.md (vashr<mode>3, vlshr<mode>3): Move to ...
23478 * config/arm/vec-common.md: ... here.
23480 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23482 * config/arm/mve.md (mve_vshlq_<supf><mode>): Move to
23484 * config/arm/neon.md (vashl<mode>3): Delete.
23485 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): New.
23486 (vasl<mode>3): New expander.
23488 2021-01-15 Richard Biener <rguenther@suse.de>
23490 PR tree-optimization/98685
23491 * tree-vect-slp.c (vect_schedule_slp_node): Refactor handling
23492 of vector extern defs.
23494 2021-01-14 David Malcolm <dmalcolm@redhat.com>
23497 * diagnostic.c (diagnostic_kind_text): Break out this array
23499 (diagnostic_build_prefix): ...here.
23500 (fancy_abort): Detect when diagnostic_initialize has not yet been
23501 called and fall back to a minimal implementation of printing the
23502 ICE, rather than segfaulting in internal_error.
23504 2021-01-14 David Malcolm <dmalcolm@redhat.com>
23506 * diagnostic.c (diagnostic_initialize): Eliminate
23507 parseable_fixits_p in favor of initializing extra_output_kind from
23508 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
23509 (convert_column_unit): New function, split out from...
23510 (diagnostic_converted_column): ...this.
23511 (print_parseable_fixits): Add "column_unit" and "tabstop" params.
23512 Use them to call convert_column_unit on the column values.
23513 (diagnostic_report_diagnostic): Eliminate conditional on
23514 parseable_fixits_p in favor of a switch statement on
23515 extra_output_kind, passing the appropriate values to the new
23516 params of print_parseable_fixits.
23517 (selftest::test_print_parseable_fixits_none): Update for new
23518 params of print_parseable_fixits.
23519 (selftest::test_print_parseable_fixits_insert): Likewise.
23520 (selftest::test_print_parseable_fixits_remove): Likewise.
23521 (selftest::test_print_parseable_fixits_replace): Likewise.
23522 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
23524 (selftest::diagnostic_c_tests): Call it.
23525 * diagnostic.h (enum diagnostics_extra_output_kind): New.
23526 (diagnostic_context::parseable_fixits_p): Delete field in favor
23528 (diagnostic_context::extra_output_kind): ...this new field.
23529 * doc/invoke.texi (Environment Variables): Add
23530 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
23531 * opts.c (common_handle_option): Update handling of
23532 OPT_fdiagnostics_parseable_fixits for change to diagnostic_context
23535 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23537 * tree-vect-slp-patterns.c (class complex_operations_pattern,
23538 complex_operations_pattern::matches,
23539 complex_operations_pattern::recognize,
23540 complex_operations_pattern::build): New.
23541 (slp_patterns): Use it.
23543 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23545 * internal-fn.def (COMPLEX_FMS, COMPLEX_FMS_CONJ): New.
23546 * optabs.def (cmls_optab, cmls_conj_optab): New.
23547 * doc/md.texi: Document them.
23548 * tree-vect-slp-patterns.c (class complex_fms_pattern,
23549 complex_fms_pattern::matches, complex_fms_pattern::recognize,
23550 complex_fms_pattern::build): New.
23552 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23554 * internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
23555 * optabs.def (cmla_optab, cmla_conj_optab): New.
23556 * doc/md.texi: Document them.
23557 * tree-vect-slp-patterns.c (vect_match_call_p,
23558 class complex_fma_pattern, vect_slp_reset_pattern,
23559 complex_fma_pattern::matches, complex_fma_pattern::recognize,
23560 complex_fma_pattern::build): New.
23562 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23564 * internal-fn.def (COMPLEX_MUL, COMPLEX_MUL_CONJ): New.
23565 * optabs.def (cmul_optab, cmul_conj_optab): New.
23566 * doc/md.texi: Document them.
23567 * tree-vect-slp-patterns.c (vect_match_call_complex_mla,
23568 vect_normalize_conj_loc, is_eq_or_top, vect_validate_multiplication,
23569 vect_build_combine_node, class complex_mul_pattern,
23570 complex_mul_pattern::matches, complex_mul_pattern::recognize,
23571 complex_mul_pattern::build): New.
23573 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23575 * tree-vect-slp.c (optimize_load_redistribution_1): New.
23576 (optimize_load_redistribution, vect_is_slp_load_node): New.
23577 (vect_match_slp_patterns): Use it.
23579 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23581 * tree-vect-slp-patterns.c (complex_add_pattern::build):
23584 2021-01-14 Thomas Schwinge <thomas@codesourcery.com>
23586 * config/gcn/mkoffload.c (main): Create an offload image only in
23587 64-bit configurations.
23589 2021-01-14 H.J. Lu <hjl.tools@gmail.com>
23592 * config/i386/i386-options.c (ix86_option_override_internal):
23593 Issue an error for -fcf-protection with CF_BRANCH when compiling
23594 for 32-bit non-TARGET_CMOV targets.
23596 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
23599 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p):
23600 Remove declaration and initialization of shadow variable "ret".
23601 (ix86_option_override_internal): Remove delcaration of
23602 shadow variable "i". Redeclare shadowed variable to unsigned.
23603 * common/config/i386/i386-common.c (pta_size): Redeclare to unsigned.
23604 * config/i386/i386-builtins.c (get_builtin_code_for_version):
23605 Update for redeclaration.
23606 * config/i386/i386.h (pta_size): Ditto.
23608 2021-01-14 Richard Biener <rguenther@suse.de>
23610 PR tree-optimization/98674
23611 * tree-data-ref.c (base_supports_access_fn_components_p): New.
23612 (initialize_data_dependence_relation): For two bases without
23613 possible access fns resort to type size equality when determining
23614 shape compatibility.
23616 2021-01-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23619 * config/arm/arm_neon.h: Replace calls to __builtin_vcge* by
23620 <=, >= operators in vcle and vcge intrinsics respectively.
23621 * config/arm/arm_neon_builtins.def: Remove entry for
23624 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
23627 * config/i386/i386-options.c (ix86_function_specific_save):
23628 Remove redundant assignment to opts->x_ix86_branch_cost.
23629 * config/i386/i386.c (ix86_prefetch_sse):
23630 Rename from x86_prefetch_sse. Update all uses.
23631 * config/i386/i386.h: Update for rename.
23632 * config/i386/i386-options.h: Ditto.
23634 2021-01-14 Jakub Jelinek <jakub@redhat.com>
23637 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3,
23638 *sse4_1_zero_extendv4hiv4si2_3, *sse4_1_zero_extendv2siv2di2_3):
23639 Use Bm instead of m for non-avx. Add isa attribute.
23641 2021-01-14 Jakub Jelinek <jakub@redhat.com>
23643 PR tree-optimization/96688
23644 * match.pd (~(X >> Y) -> ~X >> Y): New simplification if
23645 ~X can be simplified.
23647 2021-01-14 Richard Sandiford <richard.sandiford@arm.com>
23649 * tree-vect-stmts.c (vect_model_load_cost): Account for unused
23650 IFN_LOAD_LANES results.
23652 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23654 * config/aarch64/aarch64-simd.md (aarch64_<su>xtl<mode>):
23656 (aarch64_xtn<mode>): Likewise.
23657 * config/aarch64/aarch64-simd-builtins.def (sxtl, uxtl, xtn):
23660 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
23662 (vmovl_s16): Likewise.
23663 (vmovl_s32): Likewise.
23664 (vmovl_u8): Likewise.
23665 (vmovl_u16): Likewise.
23666 (vmovl_u32): Likewise.
23667 (vmovn_s16): Likewise.
23668 (vmovn_s32): Likewise.
23669 (vmovn_s64): Likewise.
23670 (vmovn_u16): Likewise.
23671 (vmovn_u32): Likewise.
23672 (vmovn_u64): Likewise.
23674 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23676 * config/aarch64/aarch64-simd.md (aarch64_<su>qxtn2<mode>_le):
23678 (aarch64_<su>qxtn2<mode>_be): Likewise.
23679 (aarch64_<su>qxtn2<mode>): Likewise.
23680 * config/aarch64/aarch64-simd-builtins.def (sqxtn2, uqxtn2):
23682 * config/aarch64/iterators.md (SAT_TRUNC): Define code_iterator.
23683 (su): Handle ss_truncate and us_truncate.
23684 * config/aarch64/arm_neon.h (vqmovn_high_s16): Reimplement using
23686 (vqmovn_high_s32): Likewise.
23687 (vqmovn_high_s64): Likewise.
23688 (vqmovn_high_u16): Likewise.
23689 (vqmovn_high_u32): Likewise.
23690 (vqmovn_high_u64): Likewise.
23692 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23694 * config/aarch64/aarch64-simd.md (aarch64_xtn2<mode>_le):
23696 (aarch64_xtn2<mode>_be): Likewise.
23697 (aarch64_xtn2<mode>): Likewise.
23698 * config/aarch64/aarch64-simd-builtins.def (xtn2): Define
23700 * config/aarch64/arm_neon.h (vmovn_high_s16): Reimplement using
23702 (vmovn_high_s32): Likewise.
23703 (vmovn_high_s64): Likewise.
23704 (vmovn_high_u16): Likewise.
23705 (vmovn_high_u32): Likewise.
23706 (vmovn_high_u64): Likewise.
23708 2021-01-13 Stafford Horne <shorne@gmail.com>
23710 * config/or1k/or1k.h (ASM_PREFERRED_EH_DATA_FORMAT): New macro.
23712 2021-01-13 Stafford Horne <shorne@gmail.com>
23714 * config/or1k/linux.h (TARGET_ASM_FILE_END): Define macro.
23716 2021-01-13 Stafford Horne <shorne@gmail.com>
23718 * config/or1k/or1k.h (TARGET_CPU_CPP_BUILTINS): Add builtin
23719 define for __or1k_hard_float__.
23721 2021-01-13 Stafford Horne <shorne@gmail.com>
23723 * config/or1k/or1k.h (NO_PROFILE_COUNTERS): Define as 1.
23724 (PROFILE_HOOK): Define to call _mcount.
23725 (FUNCTION_PROFILER): Change from abort to no-op.
23727 2021-01-13 Jakub Jelinek <jakub@redhat.com>
23729 PR tree-optimization/96691
23730 * match.pd ((~X | C) ^ D -> (X | C) ^ (~D ^ C),
23731 (~X & C) ^ D -> (X & C) ^ (D ^ C)): New simplifications if
23732 (~D ^ C) or (D ^ C) can be simplified.
23734 2021-01-13 Richard Biener <rguenther@suse.de>
23736 PR tree-optimization/92645
23737 * match.pd (BIT_FIELD_REF to conversion): Delay canonicalization
23738 until after vector lowering.
23740 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
23742 * config/aarch64/aarch64-sve.md (fnma<mode>4): Extend from SVE_FULL_I
23744 (@aarch64_pred_fnma<mode>, cond_fnma<mode>, *cond_fnma<mode>_2)
23745 (*cond_fnma<mode>_4, *cond_fnma<mode>_any): Likewise.
23747 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
23749 * config/aarch64/aarch64-sve.md (fma<mode>4): Extend from SVE_FULL_I
23751 (@aarch64_pred_fma<mode>, cond_fma<mode>, *cond_fma<mode>_2)
23752 (*cond_fma<mode>_4, *cond_fma<mode>_any): Likewise.
23754 2021-01-13 Richard Biener <rguenther@suse.de>
23756 PR tree-optimization/92645
23757 * tree-vect-slp.c (vect_build_slp_tree_1): Relax supported
23758 BIT_FIELD_REF argument.
23759 (vect_build_slp_tree_2): Record the desired vector type
23760 on the external vector def.
23761 (vectorizable_slp_permutation): Handle required punning
23762 of existing vector defs.
23764 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
23766 * rtl-ssa/accesses.h (def_lookup): Fix order of comparison results.
23768 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
23770 * config/sh/sh.md (movsf_ie): Remove operands[2] test.
23772 2021-01-13 Samuel Thibault <samuel.thibault@ens-lyon.org>
23774 * config.gcc [$target == *-*-gnu*]: Enable
23775 'default_gnu_indirect_function'.
23777 2021-01-13 Jakub Jelinek <jakub@redhat.com>
23780 * optabs.c (expand_vec_perm_const): Don't force v0 and v1 into
23781 registers before calling targetm.vectorize.vec_perm_const, only after
23783 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const): Handle
23784 two argument permutation when one operand is zero vector and only
23785 after that force operands into registers.
23786 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_1): New
23787 define_insn_and_split pattern.
23788 (*avx512bw_zero_extendv32qiv32hi2_1): Likewise.
23789 (*avx512f_zero_extendv16hiv16si2_1): Likewise.
23790 (*avx2_zero_extendv8hiv8si2_1): Likewise.
23791 (*avx512f_zero_extendv8siv8di2_1): Likewise.
23792 (*avx2_zero_extendv4siv4di2_1): Likewise.
23793 * config/mips/mips.c (mips_vectorize_vec_perm_const): Force operands
23795 * config/arm/arm.c (arm_vectorize_vec_perm_const): Likewise.
23796 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Likewise.
23797 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const): Likewise.
23798 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const): Likewise.
23799 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const): Likewise.
23800 * config/gcn/gcn.c (gcn_vectorize_vec_perm_const): Likewise. Use std::swap.
23802 2021-01-13 Martin Liska <mliska@suse.cz>
23804 PR tree-optimization/98455
23805 * gimple-if-to-switch.cc (condition_info::record_phi_mapping):
23806 Record also virtual PHIs.
23807 (pass_if_to_switch::execute): Return TODO_cleanup_cfg only
23810 2021-01-13 Jonathan Wakely <jwakely@redhat.com>
23812 * doc/invoke.texi (C++ Modules): Fix typos.
23814 2021-01-13 Richard Biener <rguenther@suse.de>
23816 PR tree-optimization/98640
23817 * tree-ssa-sccvn.c (visit_nary_op): Do not try to
23818 handle plus or minus from a truncated operand to be
23821 2021-01-13 Jakub Jelinek <jakub@redhat.com>
23824 * config/i386/i386.md (*btr<mode>_1, *btr<mode>_2): New
23825 define_insn_and_split patterns.
23826 (splitter after *btr<mode>_2): New splitter.
23828 2021-01-13 Martin Liska <mliska@suse.cz>
23831 * cgraphunit.c (analyze_functions): Remove dead code.
23833 2021-01-13 Qian Jianhua <qianjh@cn.fujitsu.com>
23835 * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
23836 * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
23837 (a64fx_regmove_cost, a64fx_vector_cost): New.
23838 (a64fx_tunings): Use the new added cost tables.
23840 2021-01-13 Jakub Jelinek <jakub@redhat.com>
23843 * config/i386/predicates.md (pmovzx_parallel): New predicate.
23844 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): New
23845 define_insn_and_split pattern.
23846 (*sse4_1_zero_extendv4hiv4si2_3): Likewise.
23847 (*sse4_1_zero_extendv2siv2di2_3): Likewise.
23849 2021-01-13 Julian Brown <julian@codesourcery.com>
23851 * config/gcn/gcn.c (gcn_conditional_register_usage): Remove dead code
23852 to fix v0 register.
23854 2021-01-13 Julian Brown <julian@codesourcery.com>
23856 * config/gcn/gcn.c (gcn_md_reorg): Fix case where EXEC reg is live
23859 2021-01-13 Julian Brown <julian@codesourcery.com>
23861 * config/gcn/gcn-valu.md (recip<mode>2<exec>, recip<mode>2): Use unspec
23862 for reciprocal-approximation instructions.
23863 (div<mode>3): Use fused multiply-accumulate operations for reciprocal
23864 refinement and division result.
23865 * config/gcn/gcn.md (UNSPEC_RCP): New unspec constant.
23867 2021-01-13 Julian Brown <julian@codesourcery.com>
23869 * config/gcn/gcn-valu.md (subdf): Rename to...
23872 2021-01-12 Martin Liska <mliska@suse.cz>
23874 * gcov.c (source_info::debug): Fix printf format for 32-bit hosts.
23876 2021-01-12 Andrea Corallo <andrea.corallo@arm.com>
23878 * function-abi.h: Fix typo.
23880 2021-01-12 Christophe Lyon <christophe.lyon@linaro.org>
23884 * config/arm/arm.h (ARM_HAVE_NEON_V8QI_LDST): New macro.
23885 (ARM_HAVE_NEON_V16QI_LDST, ARM_HAVE_NEON_V4HI_LDST): Likewise.
23886 (ARM_HAVE_NEON_V8HI_LDST, ARM_HAVE_NEON_V2SI_LDST): Likewise.
23887 (ARM_HAVE_NEON_V4SI_LDST, ARM_HAVE_NEON_V4HF_LDST): Likewise.
23888 (ARM_HAVE_NEON_V8HF_LDST, ARM_HAVE_NEON_V4BF_LDST): Likewise.
23889 (ARM_HAVE_NEON_V8BF_LDST, ARM_HAVE_NEON_V2SF_LDST): Likewise.
23890 (ARM_HAVE_NEON_V4SF_LDST, ARM_HAVE_NEON_DI_LDST): Likewise.
23891 (ARM_HAVE_NEON_V2DI_LDST): Likewise.
23892 (ARM_HAVE_V8QI_LDST, ARM_HAVE_V16QI_LDST): Likewise.
23893 (ARM_HAVE_V4HI_LDST, ARM_HAVE_V8HI_LDST): Likewise.
23894 (ARM_HAVE_V2SI_LDST, ARM_HAVE_V4SI_LDST, ARM_HAVE_V4HF_LDST): Likewise.
23895 (ARM_HAVE_V8HF_LDST, ARM_HAVE_V4BF_LDST, ARM_HAVE_V8BF_LDST): Likewise.
23896 (ARM_HAVE_V2SF_LDST, ARM_HAVE_V4SF_LDST, ARM_HAVE_DI_LDST): Likewise.
23897 (ARM_HAVE_V2DI_LDST): Likewise.
23898 * config/arm/mve.md (*movmisalign<mode>_mve_store): New pattern.
23899 (*movmisalign<mode>_mve_load): New pattern.
23900 * config/arm/neon.md (movmisalign<mode>): Move to ...
23901 * config/arm/vec-common.md: ... here.
23903 2021-01-12 Vladimir N. Makarov <vmakarov@redhat.com>
23906 * lra-eliminations.c (eliminate_regs_in_insn): Add transformation
23907 of pattern 'plus (plus (hard reg, const), pseudo)'.
23909 2021-01-12 Richard Biener <rguenther@suse.de>
23911 PR tree-optimization/98550
23912 * tree-vect-slp.c (vect_record_max_nunits): Check whether
23913 the group size is a multiple of the vector element count.
23914 (vect_build_slp_tree_1): When we need to fail because
23915 the vector type choosen causes unrolling do so lazily
23916 without affecting matches only at the end to guide group splitting.
23918 2021-01-12 Martin Liska <mliska@suse.cz>
23921 * optc-save-gen.awk: Compare also n_target_save vars with
23924 2021-01-12 Martin Liska <mliska@suse.cz>
23926 * gcov.c (source_info::debug): New.
23927 (print_usage): Add --debug (-D) option.
23928 (process_args): Likewise.
23929 (generate_results): Call src->debug after
23930 accumulate_line_counts.
23931 (read_graph_file): Properly assign id for EXIT_BLOCK.
23932 * profile.c (branch_prob): Dump function body before it is
23935 2021-01-12 Jakub Jelinek <jakub@redhat.com>
23937 PR tree-optimization/98629
23938 * tree-ssa-math-opts.c (arith_overflow_check_p): Don't update use_stmt
23939 unless returning non-zero.
23941 2021-01-12 Jakub Jelinek <jakub@redhat.com>
23943 PR tree-optimization/95731
23944 * tree-ssa-reassoc.c (optimize_range_tests_cmp_bitwise): Also optimize
23945 x < 0 && y < 0 && z < 0 into (x | y | z) < 0 for signed x, y, z.
23946 (optimize_range_tests): Call optimize_range_tests_cmp_bitwise
23947 only after optimize_range_tests_var_bound.
23949 2021-01-12 Jakub Jelinek <jakub@redhat.com>
23951 * configure.ac: Ensure c/Make-lang.in comes first in @all_lang_makefrags@.
23952 * configure: Regenerated.
23954 2021-01-12 liuhongt <hongtao.liu@intel.com>
23957 * config/i386/i386-builtins.h (BUILTIN_DESC_SWAP_OPERANDS):
23959 * config/i386/i386-expand.c (ix86_expand_sse_comi): Delete
23962 2021-01-12 Alexandre Oliva <oliva@adacore.com>
23964 * ssa-iterators.h (end_imm_use_stmt_traverse): Forward
23966 (auto_end_imm_use_stmt_traverse): New struct.
23967 (FOR_EACH_IMM_USE_STMT): Use it.
23968 (BREAK_FROM_IMM_USE_STMT, RETURN_FROM_IMM_USE_STMT): Remove,
23970 * gimple-ssa-strength-reduction.c: ... here, ...
23971 * graphite-scop-detection.c: ... here, ...
23972 * ipa-modref.c, ipa-pure-const.c, ipa-sra.c: ... here, ...
23973 * tree-predcom.c, tree-ssa-ccp.c: ... here, ...
23974 * tree-ssa-dce.c, tree-ssa-dse.c: ... here, ...
23975 * tree-ssa-loop-ivopts.c, tree-ssa-math-opts.c: ... here, ...
23976 * tree-ssa-phiprop.c, tree-ssa.c: ... here, ...
23977 * tree-vect-slp.c: ... and here, ...
23978 * doc/tree-ssa.texi: ... and the example here.
23980 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
23982 * config/aarch64/aarch64-sve.md (sdiv_pow2<mode>3): Extend from
23983 SVE_FULL_I to SVE_I. Generate an UNSPEC_PRED_X.
23984 (*sdiv_pow2<mode>3): New pattern.
23985 (@cond_<sve_int_op><mode>): Extend from SVE_FULL_I to SVE_I.
23986 Wrap the ASRD in an UNSPEC_PRED_X.
23987 (*cond_<sve_int_op><mode>_2): Likewise. Replace the UNSPEC_PRED_X
23988 predicate with a constant PTRUE, if it isn't already.
23989 (*cond_<sve_int_op><mode>_z): Replace with...
23990 (*cond_<sve_int_op><mode>_any): ...this new pattern.
23992 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
23994 * config/aarch64/aarch64-sve.md (*cond_bic<mode>_2): Extend from
23995 SVE_FULL_I to SVE_I.
23996 (*cond_bic<mode>_any): Likewise.
23998 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24000 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart)
24001 (@aarch64_pred_<MUL_HIGHPART:optab><mode>): Extend from SVE_FULL_I
24004 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24006 * config/aarch64/aarch64-sve.md (<su>abd<mode>_3): Extend from
24007 SVE_FULL_I to SVE_I.
24008 (*aarch64_cond_<su>abd<mode>_2): Likewise.
24009 (*aarch64_cond_<su>abd<mode>_any): Likewise.
24010 (@aarch64_pred_<su>abd<mode>): Likewise. Use UNSPEC_PRED_X
24011 for the max and min but not for the minus.
24012 (*aarch64_cond_<su>abd<mode>_3): New pattern.
24014 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24016 * config/aarch64/iterators.md (SVE_24I): New iterator.
24017 * config/aarch64/aarch64-sve.md (*aarch64_adr<mode>_shift): Extend from
24018 SVE_FULL_SDI to SVE_24I. Use containers rather than elements.
24020 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24022 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_BINARY:optab><mode>)
24023 (*cond_<SVE_INT_BINARY:optab><mode>_2): Extend from SVE_FULL_I
24025 (*cond_<SVE_INT_BINARY:optab><mode>_3): Likewise.
24026 (*cond_<SVE_INT_BINARY:optab><mode>_any): Likewise.
24027 (*cond_<SVE_INT_BINARY:optab><mode>_2_const): Likewise.
24028 (*cond_<SVE_INT_BINARY:optab><mode>_any_const): Likewise.
24030 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24032 * config/aarch64/aarch64-sve.md (<SVE_INT_BINARY_IMM:optab><mode>3)
24033 (@aarch64_pred_<SVE_INT_BINARY_IMM:optab><mode>)
24034 (*post_ra_<SVE_INT_BINARY_IMM:optab><mode>3): Extend from SVE_FULL_I
24037 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24039 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3)
24040 (v<ASHIFT:optab><mode>3, @aarch64_pred_<optab><mode>)
24041 (*post_ra_v<ASHIFT:optab><mode>3): Extend from SVE_FULL_I to SVE_I.
24043 2021-01-11 Martin Liska <mliska@suse.cz>
24046 * symtab-clones.h (clone_info::release): Release
24047 symtab::m_clones with ggc_delete as it's a GGC memory.
24049 2021-01-11 Matthias Klose <doko@ubuntu.com>
24051 * Makefile.in (LINK_PROGRESS): Show the link target.
24053 2021-01-11 Richard Biener <rguenther@suse.de>
24055 PR tree-optimization/91403
24056 * tree-vect-data-refs.c (vect_analyze_group_access_1): Cap
24057 single-element interleaving group size at 4096 elements.
24059 2021-01-11 Richard Biener <rguenther@suse.de>
24061 PR tree-optimization/98526
24062 * tree-vect-loop.c (vect_model_reduction_cost): Remove costing
24063 of the actual reduction op for the regular case.
24064 (vectorizable_reduction): Cost the stmts
24065 vect_transform_reduction produces here.
24067 2021-01-11 Andreas Krebbel <krebbel@linux.ibm.com>
24069 * tree-ssa-forwprop.c (simplify_vector_constructor): For
24070 big-endian, use UNPACK[_FLOAT]_HI.
24072 2021-01-11 Tamar Christina <tamar.christina@arm.com>
24074 * tree-vect-slp-patterns.c (class complex_pattern,
24075 class complex_add_pattern): Add parameters to matches.
24076 (complex_add_pattern::build): Free memory.
24077 (complex_add_pattern::matches): Move validation end of match.
24078 (complex_add_pattern::recognize): Likewise.
24080 2021-01-11 Tamar Christina <tamar.christina@arm.com>
24082 * tree-vect-slp-patterns.c (linear_loads_p): Fix externals.
24084 2021-01-11 Tamar Christina <tamar.christina@arm.com>
24086 * tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
24088 2021-01-11 Jakub Jelinek <jakub@redhat.com>
24090 PR tree-optimization/95867
24091 * tree-ssa-math-opts.h: New header.
24092 * tree-ssa-math-opts.c: Include tree-ssa-math-opts.h.
24093 (powi_as_mults): No longer static. Use build_one_cst instead of
24094 build_real. Formatting fix.
24095 * tree-ssa-reassoc.c: Include tree-ssa-math-opts.h.
24096 (attempt_builtin_powi): Handle multiplication reassociation without
24097 powi_fndecl using powi_as_mults.
24098 (reassociate_bb): For integral types don't require
24099 -funsafe-math-optimizations to call attempt_builtin_powi.
24101 2021-01-11 Jakub Jelinek <jakub@redhat.com>
24103 PR tree-optimization/95852
24104 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): Change
24105 mul_stmts parameter type to vec<gimple *> &. Before cond_stmt
24106 allow in the bb any of the stmts in that vector, div_stmt and
24107 up to 3 cast stmts.
24108 (arith_cast_equal_p): New function.
24109 (arith_overflow_check_p): Add cast_stmt argument, handle signed
24110 multiply overflow checks.
24111 (match_arith_overflow): Adjust caller. Handle signed multiply
24114 2021-01-11 Jakub Jelinek <jakub@redhat.com>
24116 PR tree-optimization/95852
24117 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): New function.
24118 (uaddsub_overflow_check_p): Renamed to ...
24119 (arith_overflow_check_p): ... this. Handle also multiplication
24120 with overflow check.
24121 (match_uaddsub_overflow): Renamed to ...
24122 (match_arith_overflow): ... this. Add cfg_changed argument. Handle
24123 also multiplication with overflow check. Adjust function comment.
24124 (math_opts_dom_walker::after_dom_children): Adjust callers. Call
24125 match_arith_overflow also for MULT_EXPR.
24127 2021-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24129 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
24130 __builtin_convertvector.
24131 (vmovl_s16): Likewise.
24132 (vmovl_s32): Likewise.
24133 (vmovl_u8): Likewise.
24134 (vmovl_u16): Likewise.
24135 (vmovl_u32): Likewise.
24136 (vmovn_s16): Likewise.
24137 (vmovn_s32): Likewise.
24138 (vmovn_s64): Likewise.
24139 (vmovn_u16): Likewise.
24140 (vmovn_u32): Likewise.
24141 (vmovn_u64): Likewise.
24143 2021-01-11 Martin Liska <mliska@suse.cz>
24145 * gimple-if-to-switch.cc (struct condition_info): Use auto_var.
24146 (if_chain::is_beneficial): Delete clusters
24147 (find_conditions): Make second argument of conditions_in_bbs a
24148 pointer so that we control over it's lifetime.
24149 (pass_if_to_switch::execute): Delete them.
24151 2021-01-11 Kewen Lin <linkw@linux.ibm.com>
24153 * ira.c (move_unallocated_pseudos): Check other_reg and skip if
24156 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24158 * config/vax/vax.md (cc): Remove mode attribute.
24159 (subst_<cc>, subst_f<cc>): Rename to...
24160 (subst_<mode>, subst_f<VAXccnz:mode>): ... these respectively.
24161 (*cbranch<VAXint:mode>4_<VAXcc:mode>): Update for `cc' removal.
24162 (*cbranch<VAXfp:mode>4_<VAXccnz:mode>): Likewise.
24163 (*branch_<mode>, *branch_<mode>_reversed): Likewise.
24165 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24167 * config/vax/vax.md (subst_f<cc>): Add mode to operands and
24168 `const_double_zero'.
24170 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24172 * config/pdp11/pdp11.md (PDPfp): New mode iterator.
24173 (fcc_cc, fcc_ccnz): Use it. Add mode to `const_double_zero' and
24176 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24178 * genemit.c (gen_exp) <CONST_DOUBLE>: Handle `const_double_zero'
24180 * read-rtl.c (rtx_reader::read_rtx_code): Handle machine mode
24181 with `const_double_zero'.
24182 * doc/rtl.texi (Constant Expression Types): Document it.
24184 2021-01-09 Jakub Jelinek <jakub@redhat.com>
24187 * tree-cfg.c (verify_gimple_assign_binary): Allow lhs of
24188 POINTER_DIFF_EXPR to be any integral type.
24190 2021-01-09 Jakub Jelinek <jakub@redhat.com>
24192 PR rtl-optimization/98603
24193 * function.c (instantiate_virtual_regs_in_insn): For asm goto
24194 with impossible constraints, drop all SETs, CLOBBERs, drop PARALLEL
24195 if any, set ASM_OPERANDS mode to VOIDmode and change
24196 ASM_OPERANDS_OUTPUT_CONSTRAINT and ASM_OPERANDS_OUTPUT_IDX.
24198 2021-01-09 Alexandre Oliva <oliva@gnu.org>
24201 * final.c (notice_source_line): Narrow down the condition to
24202 skip a line-0 marker.
24204 2021-01-08 Sergei Trofimovich <siarheit@google.com>
24206 * ipa-modref.c (merge_call_side_effects): Fix
24207 linebreak split by reordering two print calls.
24209 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
24211 * config/s390/vector.md (*tf_to_fprx2_0): Rename from
24212 "*mov_tf_to_fprx2_0" for consistency, fix constraint.
24213 (*tf_to_fprx2_1): Rename from "*mov_tf_to_fprx2_1" for
24214 consistency, fix constraint.
24216 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
24218 * config/s390/s390-c.c (s390_def_or_undef_macro): Accept
24219 callables instead of mask values.
24220 (struct target_flag_set_p): New predicate.
24221 (s390_cpu_cpp_builtins_internal): Define or undefine
24222 __LONG_DOUBLE_VX__ macro.
24224 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
24227 * config/i386/i386.c (x86_function_profiler): Use R10 and R11
24228 to call mcount in large model with PIC for NO_PROFILE_COUNTERS
24231 2021-01-08 Richard Biener <rguenther@suse.de>
24233 * tree-ssa-sccvn.c (pass_fre::execute): Reset the SCEV hash table.
24235 2021-01-08 Richard Biener <rguenther@suse.de>
24237 * tree-vect-slp.c (scalar_stmts_to_slp_tree_map_t): Fix.
24238 (vect_build_slp_tree): On cache hit release the matched
24239 scalar stmts vector.
24240 * tree-vect-stmts.c (vectorizable_store): Properly free
24241 vec_oprnds before possibly gathering them again.
24243 2021-01-08 Richard Biener <rguenther@suse.de>
24245 PR tree-optimization/98544
24246 * tree-vect-slp.c (vect_optimize_slp): Always materialize
24247 permutes at a permute node.
24249 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
24252 * config/i386/i386.c (x86_function_profiler): Use R10 to call
24253 mcount in large model. Sorry for large model with PIC.
24255 2021-01-08 Jakub Jelinek <jakub@redhat.com>
24258 * config/i386/i386.opt (ix86_cmodel, ix86_incoming_stack_boundary_arg,
24259 ix86_pmode, ix86_preferred_stack_boundary_arg, ix86_regparm,
24260 ix86_veclibabi_type): Remove x_ prefix, use TargetVariable instead of
24261 TargetSave and initialize for variables with enum types.
24262 (mfentry, mstack-protector-guard-reg=, mstack-protector-guard-offset=,
24263 mstack-protector-guard-symbol=): Add Save.
24264 * config/i386/i386-options.c (ix86_function_specific_save,
24265 ix86_function_specific_restore): Don't save or restore x_ix86_cmodel,
24266 x_ix86_incoming_stack_boundary_arg, x_ix86_pmode,
24267 x_ix86_preferred_stack_boundary_arg, x_ix86_regparm,
24268 x_ix86_veclibabi_type.
24270 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
24272 * config/aarch64/aarch64-sve.md (*cnot<mode>): Extend from
24273 SVE_FULL_I to SVE_I.
24274 (*cond_cnot<mode>_2, *cond_cnot<mode>_any): Likewise.
24276 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
24278 * config/aarch64/aarch64-sve.md (*cond_uxt<mode>_2): Extend from
24279 SVE_FULL_I to SVE_I.
24280 (*cond_uxt<mode>_any): Likewise.
24282 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24284 * config/aarch64/iterators.md (Vwhalf): New iterator.
24285 * config/aarch64/aarch64-simd.md (aarch64_<sur>adalp<mode>_3):
24287 (aarch64_<sur>adalp<mode>): ... This. Make more
24289 (<sur>sadv16qi): Adjust callsite of the above.
24290 * config/aarch64/aarch64-simd-builtins.def (sadalp, uadalp): New
24292 * config/aarch64/arm_neon.h (vpadal_s8): Reimplement using
24294 (vpadal_s16): Likewise.
24295 (vpadal_u8): Likewise.
24296 (vpadal_u16): Likewise.
24297 (vpadalq_s8): Likewise.
24298 (vpadalq_s16): Likewise.
24299 (vpadalq_s32): Likewise.
24300 (vpadalq_u8): Likewise.
24301 (vpadalq_u16): Likewise.
24302 (vpadalq_u32): Likewise.
24304 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24306 * config/aarch64/aarch64-simd.md (aarch64_<su>abd<mode>_3):
24308 (aarch64_<su>abd<mode>): ... This.
24309 (<sur>sadv16qi): Adjust callsite of the above.
24310 * config/aarch64/aarch64-simd-builtins.def (sabd, uabd): Define
24312 * config/aarch64/arm_neon.h (vabd_s8): Reimplement using
24314 (vabd_s16): Likewise.
24315 (vabd_s32): Likewise.
24316 (vabd_u8): Likewise.
24317 (vabd_u16): Likewise.
24318 (vabd_u32): Likewise.
24319 (vabdq_s8): Likewise.
24320 (vabdq_s16): Likewise.
24321 (vabdq_s32): Likewise.
24322 (vabdq_u8): Likewise.
24323 (vabdq_u16): Likewise.
24324 (vabdq_u32): Likewise.
24326 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24328 * config/aarch64/aarch64-simd-builtins.def (saba, uaba): Define
24330 * config/aarch64/arm_neon.h (vaba_s8): Implement using builtin.
24331 (vaba_s16): Likewise.
24332 (vaba_s32): Likewise.
24333 (vaba_u8): Likewise.
24334 (vaba_u16): Likewise.
24335 (vaba_u32): Likewise.
24336 (vabaq_s8): Likewise.
24337 (vabaq_s16): Likewise.
24338 (vabaq_s32): Likewise.
24339 (vabaq_u8): Likewise.
24340 (vabaq_u16): Likewise.
24341 (vabaq_u32): Likewise.
24343 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24345 * config/aarch64/aarch64-simd.md (aba<mode>_3): Rename to...
24346 (aarch64_<su>aba<mode>): ... This. Handle uaba as well.
24347 Change RTL pattern to match.
24349 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
24351 * common/config/riscv/riscv-common.c (riscv_current_subset_list): New.
24352 * config/riscv/riscv-c.c (riscv-subset.h): New.
24353 (INCLUDE_STRING): Define.
24354 (riscv_cpu_cpp_builtins): Add new style architecture extension
24356 * config/riscv/riscv-subset.h (riscv_subset_list::begin): New.
24357 (riscv_subset_list::end): New.
24358 (riscv_current_subset_list): New.
24360 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
24362 * common/config/riscv/riscv-common.c (RISCV_DONT_CARE_VERSION):
24363 Move to riscv-subset.h.
24364 (struct riscv_subset_t): Ditto.
24365 (class riscv_subset_list): Ditto.
24366 * config/riscv/riscv-subset.h (RISCV_DONT_CARE_VERSION): Move
24367 from riscv-common.c.
24368 (struct riscv_subset_t): Ditto.
24369 (class riscv_subset_list): Ditto.
24370 * config/riscv/t-riscv ($(common_out_file)): Add file
24373 2021-01-07 Jakub Jelinek <jakub@redhat.com>
24376 * config/i386/i386.md (*bmi_blsi_<mode>_cmp, *bmi_blsi_<mode>_ccno):
24377 New define_insn patterns.
24379 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
24381 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_UNARY:optab><mode>)
24382 (*cond_<SVE_INT_UNARY:optab><mode>_2): Extend from SVE_FULL_I to SVE_I.
24383 (*cond_<SVE_INT_UNARY:optab><mode>_any): Likewise.
24385 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
24387 PR tree-optimization/98560
24388 * internal-fn.def (IFN_VCONDU, IFN_VCONDEQ): Use type vec_cond.
24389 * internal-fn.c (vec_cond_mask_direct): Get the data mode from
24391 (vec_cond_direct): Likewise argument 2.
24392 (vec_condu_direct, vec_condeq_direct): Delete.
24393 (expand_vect_cond_optab_fn): Rename to...
24394 (expand_vec_cond_optab_fn): ...this, replacing old macro.
24395 (expand_vec_condu_optab_fn, expand_vec_condeq_optab_fn): Delete.
24396 (expand_vect_cond_mask_optab_fn): Rename to...
24397 (expand_vec_cond_mask_optab_fn): ...this, replacing old macro.
24398 (direct_vec_cond_mask_optab_supported_p): Treat the optab as a
24400 (direct_vec_cond_optab_supported_p): Likewise.
24401 (direct_vec_condu_optab_supported_p): Delete.
24402 (direct_vec_condeq_optab_supported_p): Delete.
24403 * gimple-isel.cc: Include internal-fn.h.
24404 (gimple_expand_vec_cond_expr): Check that IFN_VCONDEQ is supported
24407 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
24409 PR tree-optimization/98560
24410 * gimple-isel.cc (gimple_expand_vec_cond_expr): If we fail to use
24411 IFN_VCOND{,U,EQ}, fall back on IFN_VCOND_MASK.
24413 2021-01-07 Uroš Bizjak <ubizjak@gmail.com>
24415 * config/i386/i386.md (insn): Merge from plusminus_insn, shift_insn,
24416 rotate_insn and optab code attributes.
24417 Update all uses to merged code attribute.
24418 * config/i386/sse.md: Update all uses to merged code attribute.
24419 * config/i386/mmx.md: Update all uses to merged code attribute.
24421 2021-01-07 Jakub Jelinek <jakub@redhat.com>
24423 PR tree-optimization/98568
24424 * gimple-ssa-store-merging.c (bswap_view_convert): New function.
24425 (bswap_replace): Use it.
24427 2021-01-06 Vladimir N. Makarov <vmakarov@redhat.com>
24429 PR rtl-optimization/97978
24430 * lra-int.h (lra_hard_reg_split_p): New external.
24431 * lra.c (lra_hard_reg_split_p): New global.
24432 (lra): Set up lra_hard_reg_split_p after splitting a hard reg.
24433 * lra-assigns.c (lra_assign): Don't check allocation correctness
24434 after hard reg splitting.
24436 2021-01-06 Martin Sebor <msebor@redhat.com>
24439 * builtins.c (new_delete_mismatch_p): New overload.
24440 (new_delete_mismatch_p (tree, tree)): Call it.
24442 2021-01-06 Alexandre Oliva <oliva@adacore.com>
24444 * Makefile.in (T_GLIMITS_H): New.
24445 (stmp-int-hdrs): Depend on it, use it.
24446 * config/t-vxworks (T_GLIMITS_H): Override it.
24447 (vxw-glimits.h): New.
24449 2021-01-06 Richard Biener <rguenther@suse.de>
24451 PR tree-optimization/98513
24452 * value-range.cc (intersect_ranges): Compare the upper bounds
24453 for the expected relation.
24455 2021-01-06 Gerald Pfeifer <gerald@pfeifer.com>
24458 2020-12-28 Gerald Pfeifer <gerald@pfeifer.com>
24460 * doc/standards.texi (HSAIL): Remove section.
24462 2021-01-05 Samuel Thibault <samuel.thibault@ens-lyon.org>
24464 * configure: Re-generate.
24466 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24468 * doc/invoke.texi (-std=c++20): Adjust for the publication of
24469 ISO 14882:2020 standard.
24470 * doc/standards.texi: Likewise.
24472 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24474 PR tree-optimization/94802
24475 * expr.h (maybe_optimize_sub_cmp_0): Declare.
24476 * expr.c: Include tree-pretty-print.h and flags.h.
24477 (maybe_optimize_sub_cmp_0): New function.
24478 (do_store_flag): Use it.
24479 * cfgexpand.c (expand_gimple_cond): Likewise.
24481 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24483 * mux-utils.h (pointer_mux::m_ptr): Tweak description of contents.
24484 * rtlanal.c (simple_regno_set): Tweak description to clarify the
24487 2021-01-05 Richard Biener <rguenther@suse.de>
24489 PR tree-optimization/98516
24490 * tree-vect-slp.c (vect_optimize_slp): Permute the incoming
24491 lanes when materializing on a VEC_PERM node.
24492 (vectorizable_slp_permutation): Dump the permute properly.
24494 2021-01-05 Richard Biener <rguenther@suse.de>
24496 * tree-vect-slp.c (vect_slp_region): Move debug counter
24497 to cover individual subgraphs.
24499 2021-01-05 Richard Biener <rguenther@suse.de>
24501 PR tree-optimization/98428
24502 * tree-vect-slp.c (vect_build_slp_tree_1): Properly reject
24503 vector lane extracts for loop vectorization.
24505 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24507 PR tree-optimization/98514
24508 * tree-ssa-reassoc.c (bb_rank): Change type from long * to
24510 (operand_rank): Change type from hash_map<tree, long> to
24511 hash_map<tree, int64_t>.
24512 (phi_rank): Change return type from long to int64_t.
24513 (loop_carried_phi): Change block_rank variable type from long to
24515 (propagate_rank): Change return type, rank parameter type and
24516 op_rank variable type from long to int64_t.
24517 (find_operand_rank): Change return type from long to int64_t
24518 and change slot variable type from long * to int64_t *.
24519 (insert_operand_rank): Change rank parameter type from long to
24521 (get_rank): Change return type and rank variable type from long to
24522 int64_t. Use PRId64 instead of ld to print the rank.
24523 (init_reassoc): Change rank variable type from long to int64_t
24524 and adjust correspondingly bb_rank and operand_rank initialization.
24526 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24528 PR tree-optimization/96928
24529 * tree-ssa-phiopt.c (xor_replacement): New function.
24530 (tree_ssa_phiopt_worker): Call it.
24532 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24534 PR tree-optimization/96930
24535 * match.pd ((A / (1 << B)) -> (A >> B)): If A is extended
24536 from narrower value which has the same type as 1 << B, perform
24537 the right shift on the narrower value followed by extension.
24539 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24541 PR tree-optimization/96239
24542 * gimple-ssa-store-merging.c (maybe_optimize_vector_constructor): New
24544 (get_status_for_store_merging): Don't return BB_INVALID for blocks
24545 with potential bswap optimizable CONSTRUCTORs.
24546 (pass_store_merging::execute): Optimize vector CONSTRUCTORs with bswap
24549 2021-01-05 Richard Biener <rguenther@suse.de>
24551 PR tree-optimization/98381
24552 * tree.c (vector_element_bits): Properly compute bool vector
24554 * tree-vect-loop.c (vectorizable_live_operation): Properly
24555 compute the last lane bit offset.
24557 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
24560 * config/i386/sse.md (sse_cvtps2pi): Redefine as define_insn_and_split.
24561 Clear the top 64 bytes of the input XMM register.
24562 (sse_cvttps2pi): Ditto.
24564 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
24567 * config/i386/xopintrin.h (_mm256_cmov_si256): New.
24569 2021-01-05 H.J. Lu <hjl.tools@gmail.com>
24572 * config/i386/xmmintrin.h (_mm_extract_pi16): Cast to unsigned
24575 2021-01-05 Claudiu Zissulescu <claziss@synopsys.com>
24577 * config/arc/arc.md (maddsidi4_split): Use ACC_REG_FIRST.
24578 (umaddsidi4_split): Likewise.
24580 2021-01-05 liuhongt <hongtao.liu@intel.com>
24583 * config/i386/sse.md (*sse2_pmovskb_zexthisi): New
24584 define_insn_and_split for zero_extend of subreg HI of pmovskb
24586 (*sse2_pmovskb_zexthisi): Add new combine splitters for
24587 zero_extend of not of subreg HI of pmovskb result.
24589 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24592 * explow.c (convert_memory_address_addr_space_1): Handle UNSPECs
24594 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate): Use
24595 convert_memory_address to convert symbolic immediates to ptr_mode
24596 before forcing them to memory.
24598 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24600 PR rtl-optimization/97144
24601 * recog.c (constrain_operands): Initialize matching_operand
24602 for each alternative, rather than only doing it once.
24604 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24606 PR rtl-optimization/98403
24607 * rtl-ssa/changes.cc (function_info::finalize_new_accesses): Explain
24608 why we don't remove call clobbers.
24609 (function_info::apply_changes_to_insn): Don't attempt to add
24610 call clobbers here.
24612 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24614 PR tree-optimization/98371
24615 * tree-vect-loop.c (vect_reanalyze_as_main_loop): New function.
24616 (vect_analyze_loop): If an epilogue loop appears to be cheaper
24617 than the main loop, re-analyze it as a main loop before adopting
24620 2021-01-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
24623 * configure.ac (NETLIBS): Determine using AX_LIB_SOCKET_NSL.
24624 * aclocal.m4, configure: Regenerate.
24625 * Makefile.in (NETLIBS): Define.
24626 (BACKEND): Remove $(CODYLIB).
24628 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24630 PR rtl-optimization/98334
24631 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
24632 Optimize (X - 1) * Y + Y to X * Y or (X + 1) * Y - Y to X * Y.
24634 2021-01-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
24636 * tree-inline.c (expand_call_inline): Restore input_location.
24637 Return result from recursive call.
24639 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
24641 PR tree-optimization/95401
24642 * config/aarch64/aarch64-sve-builtins.cc
24643 (gimple_folder::load_store_cookie): Use bits rather than bytes
24644 for the alignment argument to IFN_MASK_LOAD and IFN_MASK_STORE.
24645 * gimple-fold.c (gimple_fold_mask_load_store_mem_ref): Likewise.
24646 * tree-vect-stmts.c (vectorizable_store): Likewise.
24647 (vectorizable_load): Likewise.
24649 2021-01-04 Richard Biener <rguenther@suse.de>
24651 PR tree-optimization/98308
24652 * tree-vect-stmts.c (vectorizable_load): Set invariant mask
24655 2021-01-04 Jakub Jelinek <jakub@redhat.com>
24657 PR tree-optimization/95771
24658 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Handle types
24659 with precision smaller than int's precision and types with precision
24660 twice as large as long long. Formatting fixes.
24662 2021-01-04 Richard Biener <rguenther@suse.de>
24664 PR tree-optimization/98464
24665 * tree-ssa-sccvn.c (vn_valueize_for_srt): Rename from ...
24666 (vn_valueize_wrapper): ... this. Temporarily adjust vn_context_bb.
24667 (process_bb): Adjust.
24669 2021-01-04 Matthew Malcomson <matthew.malcomson@arm.com>
24672 * doc/invoke.texi (-fsanitize=address): Fix wording describing
24673 clash with -fsanitize=hwaddress.
24675 2021-01-04 Richard Biener <rguenther@suse.de>
24677 PR tree-optimization/98282
24678 * tree-ssa-sccvn.c (vn_get_stmt_kind): Classify tcc_reference on
24679 invariants as VN_NARY.
24681 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
24684 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Accept
24685 aarch64_simd_reg_or_zero for operand 2. Use the combinez patterns
24686 to handle zero operands.
24688 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
24690 * config/aarch64/aarch64.c (offset_6bit_signed_scaled_p): New function.
24691 (offset_6bit_unsigned_scaled_p): Fix typo in comment.
24692 (aarch64_sve_prefetch_operand_p): Accept MUL VLs in the range
24695 2021-01-04 Richard Biener <rguenther@suse.de>
24697 PR tree-optimization/98393
24698 * tree-vect-slp.c (vect_build_slp_tree): Properly zero matches
24699 when hitting the limit.
24701 2021-01-04 Richard Biener <rguenther@suse.de>
24703 PR tree-optimization/98291
24704 * tree-vect-loop.c (vectorizable_reduction): Bypass
24705 associativity check for SLP reductions with VF 1.
24707 2021-01-04 Jakub Jelinek <jakub@redhat.com>
24709 PR tree-optimization/96782
24710 * match.pd (x == ~x -> false, x != ~x -> true): New simplifications.
24712 2021-01-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
24714 * collect-utils.c (collect_execute): Check dumppfx.
24715 * collect2.c (maybe_run_lto_and_relink, do_link): Pass atsuffix
24716 to collect_execute.
24717 (do_link): Add new parameter atsuffix.
24718 (main): Handle -dumpdir option. Skip one argument for
24719 -o, -isystem and -B options.
24720 * gcc.c (make_at_file): New helper function.
24721 (close_at_file): Use it.
24723 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24725 * config/darwin.h (MIN_LD64_NO_COAL_SECTS): Adjust.
24726 Amend handling for LD64_VERSION fallback defaults.
24728 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24730 * config.gcc: Compute default version information
24731 from the configured target. Likewise defaults for
24733 * config/darwin10.h: Removed.
24734 * config/darwin12.h: Removed.
24735 * config/darwin9.h: Removed.
24736 * config/rs6000/darwin8.h: Removed.
24738 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24740 * config/darwin9.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
24742 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24744 * config/darwin9.h (STACK_CHECK_STATIC_BUILTIN): Move from here..
24745 * config/darwin.h (STACK_CHECK_STATIC_BUILTIN): .. to here.
24747 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24749 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move from
24751 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): ... to here.
24753 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24755 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move the spec
24756 for the Darwin10 unwinder stub from here ...
24757 * config/darwin.h (LINK_COMMAND_SPEC_A): ... to here.
24759 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
24761 * config/darwin.h (DSYMUTIL_SPEC): Default to DWARF
24762 (ASM_DEBUG_SPEC):Only define if the assembler supports
24764 (PREFERRED_DEBUGGING_TYPE): Default to DWARF.
24765 (DARWIN_PREFER_DWARF): Define.
24766 * config/darwin9.h (PREFERRED_DEBUGGING_TYPE): Remove.
24767 (DARWIN_PREFER_DWARF): Likewise
24768 (DSYMUTIL_SPEC): Likewise.
24769 (COLLECT_RUN_DSYMUTIL): Likewise.
24770 (ASM_DEBUG_SPEC): Likewise.
24771 (ASM_DEBUG_OPTION_SPEC): Likewise.
24773 2021-01-02 Jan Hubicka <jh@suse.cz>
24775 * cfg.c (free_block): ggc_free bb.
24777 2021-01-01 Jakub Jelinek <jakub@redhat.com>
24779 * gcc.c (process_command): Update copyright notice dates.
24780 * gcov-dump.c (print_version): Ditto.
24781 * gcov.c (print_version): Ditto.
24782 * gcov-tool.c (print_version): Ditto.
24783 * gengtype.c (create_file): Ditto.
24784 * doc/cpp.texi: Bump @copying's copyright year.
24785 * doc/cppinternals.texi: Ditto.
24786 * doc/gcc.texi: Ditto.
24787 * doc/gccint.texi: Ditto.
24788 * doc/gcov.texi: Ditto.
24789 * doc/install.texi: Ditto.
24790 * doc/invoke.texi: Ditto.
24792 2021-01-01 Jakub Jelinek <jakub@redhat.com>
24794 * ChangeLog-2020: Rotate ChangeLog. New file.
24797 Copyright (C) 2021 Free Software Foundation, Inc.
24799 Copying and distribution of this file, with or without modification,
24800 are permitted in any medium without royalty provided the copyright
24801 notice and this notice are preserved.