1 2021-10-04 Marek Polacek <polacek@redhat.com>
4 * doc/invoke.texi: Document -Warray-compare.
6 2021-10-04 Richard Biener <rguenther@suse.de>
8 * gimplify.c (is_var_need_auto_init): DECL_HARD_REGISTER
9 variables are not to be initialized.
11 2021-10-04 Richard Biener <rguenther@suse.de>
13 * expr.h (non_mem_decl_p): Declare.
14 (mem_ref_refers_to_non_mem_p): Likewise.
15 * expr.c (non_mem_decl_p): Export.
16 (mem_ref_refers_to_non_mem_p): Likewise.
17 * internal-fn.c (expand_DEFERRED_INIT): Do not expand the LHS
18 but check the base with mem_ref_refers_to_non_mem_p
21 2021-10-04 Richard Biener <rguenther@suse.de>
23 PR tree-optimization/102570
24 * tree-ssa-sccvn.h (vn_reference_op_struct): Document
25 we are using clique for the internal function code.
26 * tree-ssa-sccvn.c (vn_reference_op_eq): Compare the
27 internal function code.
28 (print_vn_reference_ops): Print the internal function code.
29 (vn_reference_op_compute_hash): Hash it.
30 (copy_reference_ops_from_call): Record it.
31 (visit_stmt): Remove the restriction around internal function
33 (fully_constant_vn_reference_p): Use fold_const_call and handle
35 (vn_reference_eq): Compare call return types.
36 * tree-ssa-pre.c (create_expression_by_pieces): Handle
37 generating calls to internal functions.
38 (compute_avail): Remove the restriction around internal function
41 2021-10-04 Aldy Hernandez <aldyh@redhat.com>
43 PR tree-optimization/102560
44 * gimple-ssa-warn-alloca.c (alloca_call_type): Remove static
45 marker for invalid_range.
47 2021-10-04 Richard Biener <rguenther@suse.de>
50 * internal-fn.c (expand_DEFERRED_INIT): Guard register
51 initialization path an avoid initializing VLA registers
54 2021-10-04 Eric Botcazou <ebotcazou@adacore.com>
56 * config/rs6000/vxworks.h (TARGET_INIT_LIBFUNCS): Delete.
58 2021-10-03 Martin Liska <mliska@suse.cz>
60 * toplev.c (toplev::main): Check opt_index if it is a part
63 2021-10-02 Aldy Hernandez <aldyh@redhat.com>
65 PR tree-optimization/102563
66 * range-op.cc (operator_lshift::op1_range): Do not clobber
69 2021-10-02 Martin Liska <mliska@suse.cz>
71 * toplev.c (toplev::main): save_decoded_options[0] is program
72 name and so it should be skipped.
74 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
76 PR tree-optimization/102546
77 * range-op.cc (operator_lshift::op1_range): Teach range-ops that
78 X << Y is non-zero implies X is also non-zero.
80 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
82 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
84 * config/aarch64/aarch64-tune.md: Regenerate.
85 * doc/invoke.texi: Update docs.
87 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
89 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
91 * config/aarch64/aarch64-tune.md: Regenerate.
92 * doc/invoke.texi: Update docs.
94 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
96 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
98 * config/aarch64/aarch64-tune.md: Regenerate.
99 * doc/invoke.texi: Update docs.
101 2021-10-01 Martin Sebor <msebor@redhat.com>
104 * doc/invoke.texi (-Waddress): Update.
105 * gengtype.c (write_types): Avoid -Waddress.
106 * poly-int.h (POLY_SET_COEFF): Avoid using null.
108 2021-10-01 John David Anglin <danglin@gcc.gnu.org>
111 * config/pa/pa.c (pa_option_override): Default to dwarf version 4
114 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
116 * config/aarch64/aarch64.h (AARCH64_FL_V9): Update value.
118 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
120 * gimple-range-path.cc (path_range_query::compute_ranges): Use
122 * gimple-range-path.h (class path_range_query): Remove shadowed
124 (path_range_query::get_path_oracle): New.
126 2021-10-01 Jakub Jelinek <jakub@redhat.com>
127 Richard Biener <rguenther@suse.de>
130 * doc/invoke.texi (-fsanitize=integer-divide-by-zero): Remove
131 INT_MIN / -1 division detection from here ...
132 (-fsanitize=signed-integer-overflow): ... and add it here.
134 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
136 * config/aarch64/aarch64-arches.def (AARCH64_ARCH): Added
138 * config/aarch64/aarch64.h (AARCH64_FL_V9): New.
139 (AARCH64_FL_FOR_ARCH9): New flags for Armv9-A.
140 (AARCH64_ISA_V9): New ISA flag.
141 * doc/invoke.texi: Update docs.
143 2021-10-01 Martin Liska <mliska@suse.cz>
145 * toplev.c (toplev::main): Save decoded optimization options.
146 * toplev.h (save_opt_decoded_options): New.
147 * doc/extend.texi: Be more clear about optimize and target
150 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
152 * explow.c: Include langhooks.h.
153 (set_stack_check_libfunc): Build a proper function type.
155 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
158 * config/i386/i386.c (legitimate_pic_address_disp_p): For PE-COFF do
159 not return true for external weak function symbols in medium model.
161 2021-10-01 Jakub Jelinek <jakub@redhat.com>
163 * tree.h (OMP_CLAUSE_ORDER_REPRODUCIBLE): Define.
164 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_ORDER>: Print
165 reproducible: for OMP_CLAUSE_ORDER_REPRODUCIBLE.
166 * omp-general.c (omp_extract_for_data): If OMP_CLAUSE_ORDER is seen
167 without OMP_CLAUSE_ORDER_UNCONSTRAINED, overwrite sched_kind to
168 OMP_CLAUSE_SCHEDULE_STATIC.
170 2021-10-01 Richard Biener <rguenther@suse.de>
173 * tree-inline.c (setup_one_parameter): Avoid substituting
174 an invariant into contexts where a GIMPLE register is not valid.
176 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
178 * config/arm/arm-cpus.in: Add Cortex-R52+ CPU.
179 * config/arm/arm-tables.opt: Regenerate.
180 * config/arm/arm-tune.md: Regenerate.
181 * doc/invoke.texi: Update docs.
183 2021-09-30 Uroš Bizjak <ubizjak@gmail.com>
186 * config/i386/i386.md
187 (sign_extend:WIDE (any_logic:NARROW (memory, immediate)) splitters):
190 2021-09-30 Tobias Burnus <tobias@codesourcery.com>
192 * omp-low.c (omp_runtime_api_call): Add omp_aligned_{,c}alloc and
193 omp_{c,re}alloc, fix omp_alloc/omp_free.
195 2021-09-30 Martin Liska <mliska@suse.cz>
197 * defaults.h (ASM_OUTPUT_ASCII): Do not hide global variable
198 asm_out_file and stream directly to MYFILE.
200 2021-09-30 Richard Biener <rguenther@suse.de>
202 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
203 Restore and fix condition under which we apply npeel to
204 the DRs misalignment value.
206 2021-09-30 Richard Biener <rguenther@suse.de>
208 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
209 Fix npeel check for variable amount of peeling.
211 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
213 * lto-wrapper.c (run_gcc): Plug snprintf overflow.
215 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
217 * gimple-range.cc (gimple_ranger::debug): New.
218 * gimple-range.h (class gimple_ranger): Add debug.
220 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
223 * tree-vrp.c (hybrid_threader::~hybrid_threader): Free m_query.
225 2021-09-29 Indu Bhagat <indu.bhagat@oracle.com>
228 * btfout.c (GTY): Add GTY (()) albeit for cosmetic only purpose.
229 (btf_finalize): Empty the hash_map btf_var_ids.
231 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
233 * tree-vrp.c (thread_through_all_blocks): Return bool.
234 (execute_vrp_threader): Return TODO_* flags.
235 (pass_data_vrp_threader): Set todo_flags_finish to 0.
237 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
239 * timevar.def (TV_TREE_VRP_THREADER): New.
240 * tree-vrp.c: Use TV_TREE_VRP_THREADER for VRP threader pass.
242 2021-09-29 David Faust <david.faust@oracle.com>
244 * config.gcc (bpf-*-*): Do not overwrite extra_headers.
246 2021-09-29 Jonathan Wright <jonathan.wright@arm.com>
248 * config/aarch64/aarch64-builtins.c (TYPES_BINOP_PPU): Define
249 new type qualifier enum.
250 (TYPES_TERNOP_SSSU): Likewise.
251 (TYPES_TERNOP_PPPU): Likewise.
252 * config/aarch64/aarch64-simd-builtins.def: Define PPU, SSU,
253 PPPU and SSSU builtin generator macros for qtbl1 and qtbx1
255 * config/aarch64/arm_neon.h (vqtbl1_p8): Use type-qualified
256 builtin and remove casts.
257 (vqtbl1_s8): Likewise.
258 (vqtbl1q_p8): Likewise.
259 (vqtbl1q_s8): Likewise.
260 (vqtbx1_s8): Likewise.
261 (vqtbx1_p8): Likewise.
262 (vqtbx1q_s8): Likewise.
263 (vqtbx1q_p8): Likewise.
264 (vtbl1_p8): Likewise.
265 (vtbl2_p8): Likewise.
266 (vtbx2_p8): Likewise.
268 2021-09-29 Richard Biener <rguenther@suse.de>
270 * tree-vect-data-refs.c (vect_dr_misalign_for_aligned_access):
272 (vect_update_misalignment_for_peel): Use it to update
273 misaligned to the value necessary for an aligned access.
274 (vect_get_peeling_costs_all_drs): Likewise.
275 (vect_enhance_data_refs_alignment): Likewise.
277 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
279 * config/aarch64/aarch64.c (aarch64_expand_cpymem): Count number of
280 emitted operations and adjust heuristic for code size.
282 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
284 * config/aarch64/aarch64.c (aarch64_expand_setmem): Count number of
285 emitted operations and adjust heuristic for code size.
287 2021-09-29 Jakub Jelinek <jakub@redhat.com>
290 * gimplify.c (gimplify_scan_omp_clauses): Use omp_check_private even
291 in OMP_SCOPE clauses, not just on worksharing construct clauses.
293 2021-09-28 Geng Qi <gengqi@linux.alibaba.com>
295 * config/riscv/riscv.md (mulv<mode>4): Call gen_smul<mode>3_highpart.
296 (<u>mulditi3): Call <su>muldi3_highpart.
297 (<u>muldi3_highpart): Rename to <su>muldi3_highpart.
298 (<u>mulsidi3): Call <su>mulsi3_highpart.
299 (<u>mulsi3_highpart): Rename to <su>mulsi3_highpart.
301 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
303 * config/darwin.h (DSYMUTIL_SPEC): Recognize D sources.
305 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
307 * config/rs6000/darwin.h (FIXED_R13): Add for PPC64.
308 (FIRST_SAVED_GP_REGNO): Save from R13 even when it is one
311 2021-09-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
313 * config/aarch64/aarch64.h (AARCH64_FL_LS64): Define
314 (AARCH64_FL_V8_7): Likewise.
315 (AARCH64_FL_FOR_ARCH8_7): Likewise.
316 * config/aarch64/aarch64-arches.def (armv8.7-a): Define.
317 * config/aarch64/aarch64-option-extensions.def (ls64): Define.
318 * doc/invoke.texi: Document the above.
320 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
322 * dbgcnt.c (dbg_cnt_counter): New.
323 * dbgcnt.h (dbg_cnt_counter): New.
324 * dumpfile.c (dump_options): Add entry for TDF_THREADING.
325 * dumpfile.h (enum dump_flag): Add TDF_THREADING.
326 * gimple-range-path.cc (DEBUG_SOLVER): Use TDF_THREADING.
327 * tree-ssa-threadupdate.c (dump_jump_thread_path): Dump out
330 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
332 * cfgcleanup.c (pass_jump::execute): Check
333 flag_expensive_optimizations.
334 (pass_jump_after_combine::gate): Same.
335 * doc/invoke.texi (-fthread-jumps): Enable for -O1.
336 * opts.c (default_options_table): Enable -fthread-jumps at -O1.
337 * tree-ssa-threadupdate.c
338 (fwd_jt_path_registry::remove_jump_threads_including): Bail unless
341 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
343 * tree-ssa-reassoc.c (biased_names): New global.
344 (propagate_bias_p): New function.
345 (loop_carried_phi): Remove.
346 (propagate_rank): Propagate bias along single uses.
347 (get_rank): Update biased_names when needed.
349 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
351 * passes.def (pass_reassoc): Rename parameter to early_p.
352 * tree-ssa-reassoc.c (reassoc_bias_loop_carried_phi_ranks_p):
354 (phi_rank): Don't bias loop-carried phi ranks
355 before vectorization pass.
356 (execute_reassoc): Add bias_loop_carried_phi_ranks_p parameter.
357 (pass_reassoc::pass_reassoc): Add bias_loop_carried_phi_ranks_p
359 (pass_reassoc::set_param): Set bias_loop_carried_phi_ranks_p
361 (pass_reassoc::execute): Pass bias_loop_carried_phi_ranks_p to
363 (pass_reassoc::bias_loop_carried_phi_ranks_p): New member.
365 2021-09-28 Jakub Jelinek <jakub@redhat.com>
368 * config/i386/i386.c (standard_80387_constant_p): Don't recognize
369 special 80387 instruction XFmode constants if flag_rounding_math.
371 2021-09-28 Richard Biener <rguenther@suse.de>
373 PR tree-optimization/100112
374 * tree-ssa-sccvn.c (visit_reference_op_load): Record the
375 referece into the hashtable twice in case last_vuse is
376 different from the original vuse on the stmt.
378 2021-09-28 Jakub Jelinek <jakub@redhat.com>
381 * gimplify.c (gimplify_adjust_omp_clauses_1): Don't call the
382 omp_finish_clause langhook on implicitly added OMP_CLAUSE_PRIVATE
383 clauses on SIMD constructs.
385 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
387 PR tree-optimization/102511
388 * gimple-range-path.cc (path_range_query::range_on_path_entry):
389 Return VARYING when nothing found.
391 2021-09-28 Hongyu Wang <hongyu.wang@intel.com>
394 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): Add
396 (VALID_SSE2_REG_VHF_MODE): Add V4HFmode and V2HFmode.
397 (VALID_MMX_REG_MODE): Add V4HFmode.
398 (SSE_REG_MODE_P): Replace VALID_AVX512FP16_REG_MODE with
399 vector mode condition.
400 * config/i386/i386.c (classify_argument): Parse V4HF/V2HF
402 (function_arg_32): Add V4HFmode.
403 (function_arg_advance_32): Likewise.
404 * config/i386/i386.md (mode): Add V4HF/V2HF.
405 (MODE_SIZE): Likewise.
406 * config/i386/mmx.md (MMXMODE): Add V4HF mode.
407 (V_32): Add V2HF mode.
408 (VHF_32_64): New mode iterator.
409 (*mov<mode>_internal): Adjust sse alternatives to support
411 (*mov<mode>_internal): Adjust sse alternatives to support
413 (<insn><mode>3): New define_insn for add/sub/mul/div.
415 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
417 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Check
419 (pass_early_thread_jumps::gate): Same.
420 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
421 Return if !flag_thread_jumps.
422 * tree-ssa-threadupdate.c
423 (jt_path_registry::register_jump_thread): Assert that
424 flag_thread_jumps is true.
426 2021-09-28 liuhongt <hongtao.liu@intel.com>
429 (simplify_context::simplify_binary_operation_1): Relax
430 condition of simplifying (vec_concat:M (vec_select op0
431 index0)(vec_select op1 index1)) to allow different modes
432 between op0 and M, but have same inner mode.
434 2021-09-28 liuhongt <hongtao.liu@intel.com>
436 * config/i386/i386-expand.c (emit_reduc_half): Handle
437 V8HF/V16HF/V32HFmode.
438 * config/i386/sse.md (REDUC_SSE_PLUS_MODE): Add V8HF.
439 (REDUC_SSE_SMINMAX_MODE): Ditto.
440 (REDUC_PLUS_MODE): Add V16HF and V32HF.
441 (REDUC_SMINMAX_MODE): Ditto.
443 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
445 * gimple-range-path.cc
446 (path_range_query::precompute_ranges_in_block): Rename to...
447 (path_range_query::compute_ranges_in_block): ...this.
448 (path_range_query::precompute_ranges): Rename to...
449 (path_range_query::compute_ranges): ...this.
450 (path_range_query::precompute_relations): Rename to...
451 (path_range_query::compute_relations): ...this.
452 (path_range_query::precompute_phi_relations): Rename to...
453 (path_range_query::compute_phi_relations): ...this.
454 * gimple-range-path.h: Rename precompute* to compute*.
455 * tree-ssa-threadbackward.c
456 (back_threader::find_taken_edge_switch): Same.
457 (back_threader::find_taken_edge_cond): Same.
458 * tree-ssa-threadedge.c
459 (hybrid_jt_simplifier::compute_ranges_from_state): Same.
460 (hybrid_jt_state::register_equivs_stmt): Inline...
461 * tree-ssa-threadedge.h: ...here.
463 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
465 * tree-vrp.c (lhs_of_dominating_assert): Remove.
466 (class vrp_jt_state): Remove.
467 (class vrp_jt_simplifier): Remove.
468 (vrp_jt_simplifier::simplify): Remove.
469 (class vrp_jump_threader): Remove.
470 (vrp_jump_threader::vrp_jump_threader): Remove.
471 (vrp_jump_threader::~vrp_jump_threader): Remove.
472 (vrp_jump_threader::before_dom_children): Remove.
473 (vrp_jump_threader::after_dom_children): Remove.
475 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
477 * passes.def (pass_vrp_threader): New.
478 * tree-pass.h (make_pass_vrp_threader): Add make_pass_vrp_threader.
479 * tree-ssa-threadedge.c (hybrid_jt_state::register_equivs_stmt): New.
480 (hybrid_jt_simplifier::hybrid_jt_simplifier): New.
481 (hybrid_jt_simplifier::simplify): New.
482 (hybrid_jt_simplifier::compute_ranges_from_state): New.
483 * tree-ssa-threadedge.h (class hybrid_jt_state): New.
484 (class hybrid_jt_simplifier): New.
485 * tree-vrp.c (execute_vrp): Remove ASSERT_EXPR based jump
487 (class hybrid_threader): New.
488 (hybrid_threader::hybrid_threader): New.
489 (hybrid_threader::~hybrid_threader): New.
490 (hybrid_threader::before_dom_children): New.
491 (hybrid_threader::after_dom_children): New.
492 (execute_vrp_threader): New.
493 (class pass_vrp_threader): New.
494 (make_pass_vrp_threader): New.
496 2021-09-27 Martin Liska <mliska@suse.cz>
498 * output.h (enum section_flag): New.
499 (SECTION_FORGET): Remove.
500 (SECTION_ENTSIZE): Make it (1UL << 8) - 1.
501 (SECTION_STYLE_MASK): Define it based on other enum
503 * varasm.c (switch_to_section): Remove unused handling of
506 2021-09-27 Martin Liska <mliska@suse.cz>
508 * common.opt: Add new variable flag_default_complex_method.
509 * opts.c (finish_options): Handle flags related to
510 x_flag_complex_method.
511 * toplev.c (process_options): Remove option handling related
512 to flag_complex_method.
514 2021-09-27 Richard Biener <rguenther@suse.de>
517 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid using
518 type_for_size, instead use int_mode_for_size.
520 2021-09-27 Andrew Pinski <apinski@marvell.com>
523 * gimplify.c (gimplify_save_expr): Return early
524 if the type of val is error_mark_node.
526 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
528 * tree-ssanames.c (ssa_name_has_boolean_range): Use
531 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
533 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
535 * tree-ssa-dom.c (cprop_operand): Convert to range_query API.
536 (cprop_into_stmt): Same.
537 (dom_opt_dom_walker::optimize_stmt): Same.
539 2021-09-27 Richard Biener <rguenther@suse.de>
541 PR tree-optimization/97351
542 PR tree-optimization/97352
543 PR tree-optimization/82426
544 * tree-vectorizer.h (dr_misalignment): Add vector type
546 (aligned_access_p): Likewise.
547 (known_alignment_for_access_p): Likewise.
548 (vect_supportable_dr_alignment): Likewise.
549 (vect_known_alignment_in_bytes): Likewise. Refactor.
550 (DR_MISALIGNMENT): Remove.
551 (vect_update_shared_vectype): Likewise.
552 * tree-vect-data-refs.c (dr_misalignment): Refactor, handle
553 a vector type with larger alignment requirement and apply
554 the negative step adjustment here.
555 (vect_calculate_target_alignment): Remove.
556 (vect_compute_data_ref_alignment): Get explicit vector type
557 argument, do not apply a negative step alignment adjustment
559 (vect_slp_analyze_node_alignment): Re-analyze alignment
560 when we re-visit the DR with a bigger desired alignment but
561 keep more precise results from smaller alignments.
562 * tree-vect-slp.c (vect_update_shared_vectype): Remove.
563 (vect_slp_analyze_node_operations_1): Do not update the
564 shared vector type on stmts.
565 * tree-vect-stmts.c (vect_analyze_stmt): Push/pop the
566 vector type of an SLP node to the representative stmt-info.
567 (vect_transform_stmt): Likewise.
569 2021-09-27 liuhongt <hongtao.liu@intel.com>
572 2021-09-09 liuhongt <hongtao.liu@intel.com>
575 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
576 (reduc_plus_scal_v4sf): .. this, New define_expand.
577 (reduc_plus_scal_v2df): .. and this, New define_expand.
579 2021-09-26 liuhongt <hongtao.liu@intel.com>
581 * doc/extend.texi (Half-Precision): Remove storage only
582 description for _Float16 w/o avx512fp16.
584 2021-09-25 Dimitar Dimitrov <dimitar@dinux.eu>
586 * config/pru/constraints.md (Rrio): New constraint.
587 * config/pru/predicates.md (regio_operand): New predicate.
588 * config/pru/pru-pragma.c (pru_register_pragmas): Register
589 the __regio_symbol address space.
590 * config/pru/pru-protos.h (pru_symref2ioregno): Declaration.
591 * config/pru/pru.c (pru_symref2ioregno): New helper function.
592 (pru_legitimate_address_p): Remove.
593 (pru_addr_space_legitimate_address_p): Use the address space
595 (pru_nongeneric_pointer_addrspace): New helper function.
596 (pru_insert_attributes): New function to validate __regio_symbol
598 (TARGET_INSERT_ATTRIBUTES): New macro.
599 (TARGET_LEGITIMATE_ADDRESS_P): Remove.
600 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): New macro.
601 * config/pru/pru.h (enum reg_class): Add REGIO_REGS class.
602 * config/pru/pru.md (*regio_readsi): New pattern to read I/O
604 (*regio_nozext_writesi): New pattern to write to I/O registers.
605 (*regio_zext_write_r30<EQS0:mode>): Ditto.
606 * doc/extend.texi: Document the new PRU Named Address Space.
608 2021-09-24 Patrick Palka <ppalka@redhat.com>
612 * real.c (encode_ieee_double): Avoid unwanted sign extension.
613 (encode_ieee_quad): Likewise.
615 2021-09-24 Vladimir Makarov <vmakarov@redhat.com>
617 PR rtl-optimization/102147
618 * ira-build.c (ira_conflict_vector_profitable_p): Make
619 profitability calculation independent of host compiler pointer and
622 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
624 * gimple-range-path.cc (path_range_query::path_range_query):
625 Move debugging header...
626 (path_range_query::precompute_ranges): ...here.
627 (path_range_query::internal_range_of_expr): Do not call
628 range_on_path_entry if NAME is defined in the current block.
630 2021-09-24 Richard Biener <rguenther@suse.de>
632 * cfghooks.c (verify_flow_info): Verify unallocated BB and
633 edge flags are not set.
635 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
637 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
639 (jt_path_registry::register_jump_thread): Call
640 cancel_invalid_paths.
641 * tree-ssa-threadupdate.h (class jt_path_registry): Add
642 cancel_invalid_paths.
644 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
646 PR tree-optimization/102400
647 * tree-ssa-sccvn.c (vn_reference_insert_pieces): Initialize
648 result_vdef to zero value.
650 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
652 PR tree-optimization/102451
653 * tree-ssa-dse.c (delete_dead_or_redundant_call): Record bb of stmt
656 2021-09-24 Hongyu Wang <hongyu.wang@intel.com>
658 * config/i386/sse.md (cond_<insn><mode>): Extend to support
660 (cond_mul<mode>): Likewise.
661 (cond_div<mode>): Likewise.
662 (cond_<code><mode>): Likewise.
663 (cond_fma<mode>): Likewise.
664 (cond_fms<mode>): Likewise.
665 (cond_fnma<mode>): Likewise.
666 (cond_fnms<mode>): Likewise.
668 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
670 PR tree-optimization/102463
671 * gimple-range-fold.cc (fold_using_range::relation_fold_and_or): If
672 there is no range-ops handler, don't look for a relation.
674 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
676 * gimple-range-cache.cc (ranger_cache::ranger_cache): Take
677 non-executable_edge flag as parameter.
678 * gimple-range-cache.h (ranger_cache): Adjust prototype.
679 * gimple-range-gori.cc (gori_compute::gori_compute): Take
680 non-executable_edge flag as parameter.
681 (gori_compute::outgoing_edge_range_p): Check new flag.
682 * gimple-range-gori.h (gori_compute): Adjust prototype.
683 * gimple-range.cc (gimple_ranger::gimple_ranger): Create new flag.
684 (gimple_ranger::range_on_edge): Check new flag.
685 * gimple-range.h (gimple_ranger::non_executable_edge_flag): New.
686 * gimple-ssa-evrp.c (rvrp_folder): Pass ranger flag to simplifer.
687 (hybrid_folder::hybrid_folder): Set ranger non-executable flag value.
688 (hybrid_folder::fold_stmt): Set flag value in the simplifer.
689 * vr-values.c (simplify_using_ranges::set_and_propagate_unexecutable):
690 Use not_executable flag if provided inmstead of EDGE_EXECUTABLE.
691 (simplify_using_ranges::simplify_switch_using_ranges): Clear
692 EDGE_EXECUTABLE like it originally did.
693 (simplify_using_ranges::cleanup_edges_and_switches): Clear any
694 NON_EXECUTABLE flags.
695 (simplify_using_ranges::simplify_using_ranges): Adjust.
696 * vr-values.h (class simplify_using_ranges): Adjust.
697 (simplify_using_ranges::set_range_query): Add non-executable flag param.
699 2021-09-23 Bill Schmidt <wschmidt@linux.ibm.com>
702 * config/rs6000/rs6000-call.c (rs6000_aggregate_candidate): Detect
703 zero-width bit fields and return indicator.
704 (rs6000_discover_homogeneous_aggregate): Diagnose when the
705 presence of a zero-width bit field changes parameter passing in
708 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
710 * gimple-range-fold.cc (fold_using_range::range_of_phi):
711 Remove dominator check.
713 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
715 * gimple-range-path.cc (path_range_query::precompute_relations):
716 Hoist edge calculations before using EDGE_SUCC.
718 2021-09-23 Jonathan Wakely <jwakely@redhat.com>
720 * configure.ac: Fix --with-multilib-list description.
721 * configure: Regenerate.
723 2021-09-23 Richard Biener <rguenther@suse.de>
725 PR tree-optimization/102448
726 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
727 Clear alignment info copied from DR_PTR_INFO.
729 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
731 * config/i386/i386-expand.c (ix86_use_mask_cmp_p): Enable
733 * config/i386/sse.md (sseintvecmodelower): Add HF vector modes.
734 (<avx512>_store<mode>_mask): Extend to support HF vector modes.
735 (vec_cmp<mode><avx512fmaskmodelower>): Likewise.
736 (vcond_mask_<mode><avx512fmaskmodelower>): Likewise.
737 (vcond<mode><mode>): New expander.
738 (vcond<mode><sseintvecmodelower>): Likewise.
739 (vcond<sseintvecmodelower><mode>): Likewise.
740 (vcondu<mode><sseintvecmodelower>): Likewise.
742 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
744 * config/i386/sse.md (extend<ssePHmodelower><mode>2):
746 (extendv4hf<mode>2): Likewise.
747 (extendv2hfv2df2): Likewise.
748 (trunc<mode><ssePHmodelower>2): Likewise.
749 (avx512fp16_vcvt<castmode>2ph_<mode>): Rename to ...
750 (trunc<mode>v4hf2): ... this, and drop constraints.
751 (avx512fp16_vcvtpd2ph_v2df): Rename to ...
752 (truncv2dfv2hf2): ... this, and likewise.
754 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
756 * config/i386/sse.md (float<floatunssuffix><mode><ssePHmodelower>2):
758 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>):
760 (float<floatunssuffix><mode>v4hf2): ... this, and drop constraints.
761 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Rename to ...
762 (float<floatunssuffix>v2div2hf2): ... this, and likewise.
764 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
766 * config/i386/i386.md (fix<fixunssuffix>_trunchf<mode>2): New expander.
767 (fixuns_trunchfhi2): Likewise.
768 (*fixuns_trunchfsi2zext): New define_insn.
769 * config/i386/sse.md (ssePHmodelower): New mode_attr.
770 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
771 New expander for same element vector fix_truncate.
772 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
773 Likewise for V4HF to V4SI/V4DI fix_truncate.
774 (fix<fixunssuffix>_truncv2hfv2di2):
775 Likeise for V2HF to V2DI fix_truncate.
777 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
779 * config/i386/i386.md (<code>hf3): New expander.
781 2021-09-23 liuhongt <hongtao.liu@intel.com>
783 * config/i386/sse.md (FMAMODEM): extend to handle FP16.
784 (VFH_SF_AVX512VL): Extend to handle HFmode.
785 (VF_SF_AVX512VL): Deleted.
787 2021-09-23 liuhongt <hongtao.liu@intel.com>
789 * config/i386/i386.md (rinthf2): New expander.
790 (nearbyinthf2): New expander.
792 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
794 * tree-ssa-dom.c (class dom_jump_threader_simplifier): Rename...
795 (class dom_jt_state): ...this and provide virtual overrides.
796 (dom_jt_state::register_equiv): New.
797 (class dom_jt_simplifier): Rename from
798 dom_jump_threader_simplifier.
799 (dom_jump_threader_simplifier::simplify): Rename...
800 (dom_jt_simplifier::simplify): ...to this.
801 (pass_dominator::execute): Use dom_jt_simplifier and
803 * tree-ssa-threadedge.c (jump_threader::jump_threader):
805 (jt_state::register_equivs_stmt): Abstract out...
806 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
808 (jump_threader::thread_around_empty_blocks): Update state.
809 (jump_threader::thread_through_normal_block): Same.
810 (jt_state::jt_state): Remove.
811 (jt_state::push): Remove pass specific bits. Keep block vector
813 (jt_state::append_path): New.
814 (jt_state::pop): Remove pass specific bits.
815 (jt_state::register_equiv): Same.
816 (jt_state::record_ranges_from_stmt): Same.
817 (jt_state::register_equivs_on_edge): Same. Rename...
818 (jt_state::register_equivs_edge): ...to this.
819 (jt_state::dump): New.
820 (jt_state::debug): New.
821 (jump_threader_simplifier::simplify): Remove.
822 (jt_state::get_path): New.
823 * tree-ssa-threadedge.h (class jt_simplifier): Make into a base
824 class. Expose common functionality as virtual methods.
825 (class jump_threader_simplifier): Same. Rename...
826 (class jt_simplifier): ...to this.
827 * tree-vrp.c (class vrp_jump_threader_simplifier): Rename...
828 (class vrp_jt_simplifier): ...to this. Provide pass specific
830 (class vrp_jt_state): New.
831 (vrp_jump_threader_simplifier::simplify): Rename...
832 (vrp_jt_simplifier::simplify): ...to this. Inline code from
833 what used to be the base class.
834 (vrp_jump_threader::vrp_jump_threader): Use vrp_jt_state and
837 2021-09-22 Tobias Burnus <tobias@codesourcery.com>
840 * doc/invoke.texi (-Wno-missing-include-dirs.): Document Fortran
843 2021-09-22 Roger Sayle <roger@nextmovesoftware.com>
844 Richard Biener <rguenther@suse.de>
846 * match.pd (negation simplifications): Implement some negation
847 folding transformations from fold-const.c's fold_negate_expr.
848 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Add a SIMPLIFY
849 argument, to control whether the op should be simplified prior
850 to looking up/assigning a value number.
851 (vn_nary_build_or_lookup): Update call to vn_nary_build_or_lookup_1.
852 (vn_nary_simplify): Likewise.
853 (visit_nary_op): Likewise, but when constructing a NEGATE_EXPR
854 now call vn_nary_build_or_lookup_1 disabling simplification.
856 2021-09-22 Jiufu Guo <guojiufu@linux.ibm.com>
858 PR tree-optimization/102087
859 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
860 Update bound/cmp/control for niter.
862 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
864 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
865 Move check for non-empty BB here.
866 (fur_source::register_outgoing_edges): ...from here.
868 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
870 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
871 Remove call to improve_range_with_equivs.
872 (path_range_query::improve_range_with_equivs): Remove
873 * gimple-range-path.h: Remove improve_range_with_equivs.
875 2021-09-22 dianhong xu <dianhong.xu@intel.com>
877 * config/i386/avx512fp16intrin.h:
878 (_mm512_mask_blend_ph): New intrinsic.
879 (_mm512_permutex2var_ph): Ditto.
880 (_mm512_permutexvar_ph): Ditto.
881 * config/i386/avx512fp16vlintrin.h:
882 (_mm256_mask_blend_ph): New intrinsic.
883 (_mm256_permutex2var_ph): Ditto.
884 (_mm256_permutexvar_ph): Ditto.
885 (_mm_mask_blend_ph): Ditto.
886 (_mm_permutex2var_ph): Ditto.
887 (_mm_permutexvar_ph): Ditto.
889 2021-09-22 dianhong xu <dianhong.xu@intel.com>
891 * config/i386/avx512fp16intrin.h: Add new intrinsics.
892 (_mm512_conj_pch): New intrinsic.
893 (_mm512_mask_conj_pch): Ditto.
894 (_mm512_maskz_conj_pch): Ditto.
895 * config/i386/avx512fp16vlintrin.h: Add new intrinsics.
896 (_mm256_conj_pch): New intrinsic.
897 (_mm256_mask_conj_pch): Ditto.
898 (_mm256_maskz_conj_pch): Ditto.
899 (_mm_conj_pch): Ditto.
900 (_mm_mask_conj_pch): Ditto.
901 (_mm_maskz_conj_pch): Ditto.
903 2021-09-22 dianhong xu <dianhong.xu@intel.com>
905 * config/i386/avx512fp16intrin.h (_MM512_REDUCE_OP): New macro
906 (_mm512_reduce_add_ph): New intrinsic.
907 (_mm512_reduce_mul_ph): Ditto.
908 (_mm512_reduce_min_ph): Ditto.
909 (_mm512_reduce_max_ph): Ditto.
910 * config/i386/avx512fp16vlintrin.h
911 (_MM256_REDUCE_OP/_MM_REDUCE_OP): New macro.
912 (_mm256_reduce_add_ph): New intrinsic.
913 (_mm256_reduce_mul_ph): Ditto.
914 (_mm256_reduce_min_ph): Ditto.
915 (_mm256_reduce_max_ph): Ditto.
916 (_mm_reduce_add_ph): Ditto.
917 (_mm_reduce_mul_ph): Ditto.
918 (_mm_reduce_min_ph): Ditto.
919 (_mm_reduce_max_ph): Ditto.
921 2021-09-22 dianhong xu <dianhong.xu@intel.com>
923 * config/i386/avx512fp16intrin.h (__m512h_u, __m256h_u,
924 __m128h_u): New typedef.
925 (_mm512_load_ph): New intrinsic.
926 (_mm256_load_ph): Ditto.
927 (_mm_load_ph): Ditto.
928 (_mm512_loadu_ph): Ditto.
929 (_mm256_loadu_ph): Ditto.
930 (_mm_loadu_ph): Ditto.
931 (_mm512_store_ph): Ditto.
932 (_mm256_store_ph): Ditto.
933 (_mm_store_ph): Ditto.
934 (_mm512_storeu_ph): Ditto.
935 (_mm256_storeu_ph): Ditto.
936 (_mm_storeu_ph): Ditto.
937 (_mm512_abs_ph): Ditto.
938 * config/i386/avx512fp16vlintrin.h
940 (_mm256_abs_ph): Ditto.
942 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
944 * config/s390/tpf.md (prologue_tpf, epilogue_tpf): Add cc clobber.
946 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
949 * config/s390/s390.c (s390_expand_insv): Emit a normal move if it
950 is actually a full copy of the source operand into the target.
951 Don't emit a strict low part move if source and target mode match.
953 2021-09-22 Jakub Jelinek <jakub@redhat.com>
956 * omp-expand.c (expand_omp_single): If region->exit is NULL,
957 assert region->entry is GIMPLE_OMP_SCOPE region and return.
959 2021-09-22 Jakub Jelinek <jakub@redhat.com>
961 * tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
962 * tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
963 arguments from 2 to 3.
964 * tree-pretty-print.c (dump_omp_clause): Print allocator() around
965 allocate clause allocator and print align if present.
966 * omp-low.c (scan_sharing_clauses): Force allocate_map entry even
967 for omp_default_mem_alloc if align modifier is present. If align
968 modifier is present, use TREE_LIST to encode both allocator and
970 (lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
971 Handle align modifier on allocator clause if present.
973 2021-09-22 liuhongt <hongtao.liu@intel.com>
975 * config/i386/i386.md (define_attr "isa"): Add
977 (define_attr "enabled"): Correspond fma_or_avx512vl to
978 TARGET_FMA || TARGET_AVX512VL.
979 * config/i386/mmx.md (fmav2sf4): Extend to AVX512 fma.
984 2021-09-22 liuhongt <hongtao.liu@intel.com>
986 * config/i386/i386.md (cstorehf3): New define_expand.
988 2021-09-22 liuhongt <hongtao.liu@intel.com>
990 * config/i386/i386.md (<rounding_insn>hf2): New expander.
991 (sse4_1_round<mode>2): Extend from MODEF to MODEFH.
992 * config/i386/sse.md (*sse4_1_round<ssescalarmodesuffix>):
993 Extend from VF_128 to VFH_128.
995 2021-09-22 liuhongt <hongtao.liu@intel.com>
997 * config/i386/i386-features.c (i386-features.c): Handle
999 * config/i386/i386.md (sqrthf2): New expander.
1000 (*sqrthf2): New define_insn.
1001 * config/i386/sse.md
1002 (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
1005 2021-09-22 liuhongt <hongtao.liu@intel.com>
1007 * config/i386/avx512fp16intrin.h (_mm_mask_fcmadd_sch):
1009 (_mm_mask3_fcmadd_sch): Likewise.
1010 (_mm_maskz_fcmadd_sch): Likewise.
1011 (_mm_fcmadd_sch): Likewise.
1012 (_mm_mask_fmadd_sch): Likewise.
1013 (_mm_mask3_fmadd_sch): Likewise.
1014 (_mm_maskz_fmadd_sch): Likewise.
1015 (_mm_fmadd_sch): Likewise.
1016 (_mm_mask_fcmadd_round_sch): Likewise.
1017 (_mm_mask3_fcmadd_round_sch): Likewise.
1018 (_mm_maskz_fcmadd_round_sch): Likewise.
1019 (_mm_fcmadd_round_sch): Likewise.
1020 (_mm_mask_fmadd_round_sch): Likewise.
1021 (_mm_mask3_fmadd_round_sch): Likewise.
1022 (_mm_maskz_fmadd_round_sch): Likewise.
1023 (_mm_fmadd_round_sch): Likewise.
1024 (_mm_fcmul_sch): Likewise.
1025 (_mm_mask_fcmul_sch): Likewise.
1026 (_mm_maskz_fcmul_sch): Likewise.
1027 (_mm_fmul_sch): Likewise.
1028 (_mm_mask_fmul_sch): Likewise.
1029 (_mm_maskz_fmul_sch): Likewise.
1030 (_mm_fcmul_round_sch): Likewise.
1031 (_mm_mask_fcmul_round_sch): Likewise.
1032 (_mm_maskz_fcmul_round_sch): Likewise.
1033 (_mm_fmul_round_sch): Likewise.
1034 (_mm_mask_fmul_round_sch): Likewise.
1035 (_mm_maskz_fmul_round_sch): Likewise.
1036 * config/i386/i386-builtin.def: Add corresponding new builtins.
1037 * config/i386/sse.md
1038 (avx512fp16_fmaddcsh_v8hf_maskz<round_expand_name>): New expander.
1039 (avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>): Ditto.
1040 (avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>):
1042 (avx512fp16_<complexopname>sh_v8hf_mask<round_name>): Ditto.
1043 (avx512fp16_<complexopname>sh_v8hf<mask_scalarc_name><round_scalarcz_name>):
1045 * config/i386/subst.md (mask_scalarcz_name): New.
1046 (mask_scalarc_name): Ditto.
1047 (mask_scalarc_operand3): Ditto.
1048 (mask_scalarcz_operand4): Ditto.
1049 (round_scalarcz_name): Ditto.
1050 (round_scalarc_mask_operand3): Ditto.
1051 (round_scalarcz_mask_operand4): Ditto.
1052 (round_scalarc_mask_op3): Ditto.
1053 (round_scalarcz_mask_op4): Ditto.
1054 (round_scalarcz_constraint): Ditto.
1055 (round_scalarcz_nimm_predicate): Ditto.
1056 (mask_scalarcz): Ditto.
1057 (mask_scalarc): Ditto.
1058 (round_scalarcz): Ditto.
1060 2021-09-22 liuhongt <hongtao.liu@intel.com>
1062 * config/i386/avx512fp16intrin.h (_mm512_fcmadd_pch):
1064 (_mm512_mask_fcmadd_pch): Likewise.
1065 (_mm512_mask3_fcmadd_pch): Likewise.
1066 (_mm512_maskz_fcmadd_pch): Likewise.
1067 (_mm512_fmadd_pch): Likewise.
1068 (_mm512_mask_fmadd_pch): Likewise.
1069 (_mm512_mask3_fmadd_pch): Likewise.
1070 (_mm512_maskz_fmadd_pch): Likewise.
1071 (_mm512_fcmadd_round_pch): Likewise.
1072 (_mm512_mask_fcmadd_round_pch): Likewise.
1073 (_mm512_mask3_fcmadd_round_pch): Likewise.
1074 (_mm512_maskz_fcmadd_round_pch): Likewise.
1075 (_mm512_fmadd_round_pch): Likewise.
1076 (_mm512_mask_fmadd_round_pch): Likewise.
1077 (_mm512_mask3_fmadd_round_pch): Likewise.
1078 (_mm512_maskz_fmadd_round_pch): Likewise.
1079 (_mm512_fcmul_pch): Likewise.
1080 (_mm512_mask_fcmul_pch): Likewise.
1081 (_mm512_maskz_fcmul_pch): Likewise.
1082 (_mm512_fmul_pch): Likewise.
1083 (_mm512_mask_fmul_pch): Likewise.
1084 (_mm512_maskz_fmul_pch): Likewise.
1085 (_mm512_fcmul_round_pch): Likewise.
1086 (_mm512_mask_fcmul_round_pch): Likewise.
1087 (_mm512_maskz_fcmul_round_pch): Likewise.
1088 (_mm512_fmul_round_pch): Likewise.
1089 (_mm512_mask_fmul_round_pch): Likewise.
1090 (_mm512_maskz_fmul_round_pch): Likewise.
1091 * config/i386/avx512fp16vlintrin.h (_mm_fmadd_pch):
1093 (_mm_mask_fmadd_pch): Likewise.
1094 (_mm_mask3_fmadd_pch): Likewise.
1095 (_mm_maskz_fmadd_pch): Likewise.
1096 (_mm256_fmadd_pch): Likewise.
1097 (_mm256_mask_fmadd_pch): Likewise.
1098 (_mm256_mask3_fmadd_pch): Likewise.
1099 (_mm256_maskz_fmadd_pch): Likewise.
1100 (_mm_fcmadd_pch): Likewise.
1101 (_mm_mask_fcmadd_pch): Likewise.
1102 (_mm_mask3_fcmadd_pch): Likewise.
1103 (_mm_maskz_fcmadd_pch): Likewise.
1104 (_mm256_fcmadd_pch): Likewise.
1105 (_mm256_mask_fcmadd_pch): Likewise.
1106 (_mm256_mask3_fcmadd_pch): Likewise.
1107 (_mm256_maskz_fcmadd_pch): Likewise.
1108 (_mm_fmul_pch): Likewise.
1109 (_mm_mask_fmul_pch): Likewise.
1110 (_mm_maskz_fmul_pch): Likewise.
1111 (_mm256_fmul_pch): Likewise.
1112 (_mm256_mask_fmul_pch): Likewise.
1113 (_mm256_maskz_fmul_pch): Likewise.
1114 (_mm_fcmul_pch): Likewise.
1115 (_mm_mask_fcmul_pch): Likewise.
1116 (_mm_maskz_fcmul_pch): Likewise.
1117 (_mm256_fcmul_pch): Likewise.
1118 (_mm256_mask_fcmul_pch): Likewise.
1119 (_mm256_maskz_fcmul_pch): Likewise.
1120 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF_V8HF_V8HF,
1121 V8HF_FTYPE_V16HF_V16HF_V16HF, V16HF_FTYPE_V16HF_V16HF_V16HF_UQI,
1122 V32HF_FTYPE_V32HF_V32HF_V32HF_INT,
1123 V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT): Add new builtin types.
1124 * config/i386/i386-builtin.def: Add new builtins.
1125 * config/i386/i386-expand.c: Handle new builtin types.
1126 * config/i386/subst.md (SUBST_CV): New.
1127 (maskc_name): Ditto.
1128 (maskc_operand3): Ditto.
1130 (sdc_maskz_name): Ditto.
1131 (sdc_mask_op4): Ditto.
1132 (sdc_mask_op5): Ditto.
1133 (sdc_mask_mode512bit_condition): Ditto.
1135 (round_maskc_operand3): Ditto.
1136 (round_sdc_mask_operand4): Ditto.
1137 (round_maskc_op3): Ditto.
1138 (round_sdc_mask_op4): Ditto.
1139 (round_saeonly_sdc_mask_operand5): Ditto.
1140 * config/i386/sse.md (unspec): Add complex fma unspecs.
1141 (avx512fmaskcmode): New.
1142 (UNSPEC_COMPLEX_F_C_MA): Ditto.
1143 (UNSPEC_COMPLEX_F_C_MUL): Ditto.
1144 (complexopname): Ditto.
1145 (<avx512>_fmaddc_<mode>_maskz<round_expand_name>): New expander.
1146 (<avx512>_fcmaddc_<mode>_maskz<round_expand_name>): Ditto.
1147 (fma_<complexopname>_<mode><sdc_maskz_name><round_name>): New
1149 (<avx512>_<complexopname>_<mode>_mask<round_name>): Ditto.
1150 (<avx512>_<complexopname>_<mode><maskc_name><round_name>): Ditto.
1152 2021-09-22 Kewen Lin <linkw@linux.ibm.com>
1154 * config/rs6000/rs6000.opt (rs6000-density-pct-threshold,
1155 rs6000-density-size-threshold, rs6000-density-penalty,
1156 rs6000-density-load-pct-threshold,
1157 rs6000-density-load-num-threshold): New parameter.
1158 * config/rs6000/rs6000.c (rs6000_density_test): Adjust with
1159 corresponding parameters.
1161 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1163 * gimple-range-path.cc (path_range_query::defined_outside_path):
1165 (path_range_query::range_on_path_entry): New.
1166 (path_range_query::internal_range_of_expr): Resolve unknowns
1168 (path_range_query::improve_range_with_equivs): New.
1169 (path_range_query::ssa_range_in_phi): Resolve unknowns with
1171 * gimple-range-path.h (class path_range_query): Add
1172 defined_outside_path, range_on_path_entry, and
1173 improve_range_with_equivs.
1175 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1177 * gimple-range-path.cc (path_range_query::add_to_imports): New.
1178 (path_range_query::add_copies_to_imports): New.
1179 (path_range_query::precompute_ranges): Call
1180 add_copies_to_imports.
1181 * gimple-range-path.h (class path_range_query): Add prototypes
1182 for add_copies_to_imports and add_to_imports.
1184 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1186 * gimple-range-path.cc (path_range_query::range_defined_in_block):
1187 Remove useless code.
1189 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1191 * gimple-range-fold.h (class fur_source): Make oracle protected.
1192 * gimple-range-path.cc (path_range_query::path_range_query): Add
1193 resolve argument. Initialize oracle.
1194 (path_range_query::~path_range_query): Delete oracle.
1195 (path_range_query::range_of_stmt): Adapt to use relations.
1196 (path_range_query::precompute_ranges): Pre-compute relations.
1197 (class jt_fur_source): New
1198 (jt_fur_source::jt_fur_source): New.
1199 (jt_fur_source::register_relation): New.
1200 (jt_fur_source::query_relation): New.
1201 (path_range_query::precompute_relations): New.
1202 (path_range_query::precompute_phi_relations): New.
1203 * gimple-range-path.h (path_range_query): Add resolve argument.
1204 Add oracle, precompute_relations, precompute_phi_relations.
1205 * tree-ssa-threadbackward.c (back_threader::back_threader): Pass
1206 resolve argument to solver.
1208 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1210 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
1211 Rename postfold_gcond_edges to register_outgoing_edges and
1213 (fold_using_range::postfold_gcond_edges): Rename...
1214 (fur_source::register_outgoing_edges): ...to this.
1215 * gimple-range-fold.h (postfold_gcond_edges): Rename to
1216 register_outgoing_edges and move to fur_source.
1218 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1220 * gimple-range-fold.cc (fold_using_range::range_of_phi): Check
1221 dom_info_available_p.
1223 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1225 * gimple-range-cache.cc (non_null_ref::non_null_ref): Use create
1226 and quick_grow_cleared instead of safe_grow_cleared.
1228 2021-09-21 Thomas Schwinge <thomas@codesourcery.com>
1231 * omp-oacc-neuter-broadcast.cc (oacc_do_neutering): Evaluate
1234 2021-09-21 Richard Earnshaw <rearnsha@arm.com>
1236 * configure.ac: Detect when the assembler supports new-style
1237 architecture extensions.
1238 * common/config/arm/arm-common.c (arm_rewrite_mcpu): Return
1239 the full CPU string if the assembler can grok it.
1240 (arm_rewrite_march): Likewise but for the architecture.
1241 * config.in: Regenerate.
1242 * configure: Regenerate.
1244 2021-09-21 Richard Biener <rguenther@suse.de>
1246 PR tree-optimization/102421
1247 * tree-vect-loop.c (vect_dissolve_slp_only_groups): Copy and
1248 adjust alignment info.
1250 2021-09-21 Kewen Lin <linkw@linux.ibm.com>
1252 * ipa-fnsummary.c (ipa_fn_summary_write): Remove inconsistent
1253 bitfield stream out.
1255 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
1257 * gimple-range-fold.cc (fold_using_range::range_of_phi): Ignore
1258 undefined edges, apply an equivalence if appropriate.
1259 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Return
1260 UNDEFINED if EDGE_EXECUTABLE is not set.
1261 * gimple-range.cc (gimple_ranger::gimple_ranger): Set all edges
1262 as EXECUTABLE upon startup.
1263 (gimple_ranger::range_on_edge): Return UNDEFINED for edges without
1264 EDGE_EXECUTABLE set.
1265 * vr-values.c (set_and_propagate_unexecutable): New.
1266 (simplify_using_ranges::fold_cond): Call set_and_propagate.
1267 (simplify_using_ranges::simplify_switch_using_ranges): Ditto.
1268 * vr-values.h: Add prototype.
1270 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
1272 * value-relation.cc (equiv_oracle::register_initial_def): New.
1273 (equiv_oracle::register_relation): Call register_initial_def.
1274 (equiv_oracle::add_equiv_to_block): New. Split register_relation.
1275 (relation_oracle::register_stmt): Check def block of PHI arguments.
1276 * value-relation.h (equiv_oracle): Add new prototypes.
1278 2021-09-20 Matthias Kretz <m.kretz@gsi.de>
1280 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
1281 Define __RECIPROCAL_MATH__, __NO_SIGNED_ZEROS__,
1282 __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__, and
1283 __ROUNDING_MATH__ according to their corresponding flags.
1284 * doc/cpp.texi: Document __RECIPROCAL_MATH__,
1285 __NO_SIGNED_ZEROS__, __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__,
1286 and __ROUNDING_MATH__.
1288 2021-09-20 Richard Biener <rguenther@suse.de>
1290 * tree-vect-stmts.c (vectorizable_load): Use the vectype
1293 2021-09-20 Richard Biener <rguenther@suse.de>
1295 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
1296 Do not compute alignment of the vectorized access here.
1298 2021-09-20 Richard Biener <rguenther@suse.de>
1300 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
1301 Store -1 for runtime alias peeling iterations.
1303 2021-09-20 Richard Biener <rguenther@suse.de>
1305 * config.gcc: Obsolete hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11*.
1307 2021-09-20 Thomas Schwinge <thomas@codesourcery.com>
1309 * input.c (string_concat_db::record_string_concatenation)
1310 (string_concat_db::get_string_concatenation): Skip for
1311 'RESERVED_LOCATION_P'.
1313 2021-09-20 Richard Biener <rguenther@suse.de>
1315 PR tree-optimization/65206
1316 * tree-data-ref.h (struct data_reference): Add alt_indices,
1318 * tree-data-ref.c (free_data_ref): Release alt_indices.
1319 (dr_analyze_indices): Work on struct indices and get DR_REF as tree.
1320 (create_data_ref): Adjust.
1321 (initialize_data_dependence_relation): Split into head
1322 and tail. When the base objects fail to match up try
1323 again with pointer-based analysis of indices.
1324 * tree-vectorizer.c (vec_info_shared::check_datarefs): Do
1325 not compare the lazily computed alternate set of indices.
1327 2021-09-20 Iain Sandoe <iain@sandoe.co.uk>
1329 * gcc.c: Test for execute OK when we find the
1330 programs for assembler linker and dsymutil and those
1331 were specified at configure-time.
1333 2021-09-19 Martin Sebor <msebor@redhat.com>
1335 PR middle-end/102403
1336 * gimple-predicate-analysis.cc (predicate::init_from_control_deps):
1337 Correct a function pre/postcondition.
1339 2021-09-19 Martin Sebor <msebor@redhat.com>
1341 PR middle-end/102243
1342 * tree-ssa-strlen.c (get_range): Handle null cfun.
1344 2021-09-19 Iain Sandoe <iain@sandoe.co.uk>
1346 * config/darwin.h (LINK_COMMAND_SPEC_A): Use Darwin10
1347 unwinder shim as a convenience library.
1349 2021-09-19 Andrew Pinski <apinski@marvell.com>
1351 * doc/install.texi: Add note about
1352 binutils 2.35 is required for LTO usage.
1354 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
1356 * tree-ssa-threadbackward.c
1357 (back_threader_registry::register_path): Use push_edge.
1358 * tree-ssa-threadedge.c
1359 (jump_threader::thread_around_empty_blocks): Same.
1360 (jump_threader::thread_through_normal_block): Same.
1361 (jump_threader::thread_across_edge): Same. Also, use auto_bitmap.
1363 * tree-ssa-threadupdate.c
1364 (jt_path_registry::allocate_thread_edge): Remove.
1365 (jt_path_registry::push_edge): New.
1366 (dump_jump_thread_path): Make static.
1367 * tree-ssa-threadupdate.h (allocate_thread_edge): Remove.
1370 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
1372 * gimple-range-path.cc (path_range_query::path_range_query): Add
1374 (path_range_query::dump): Remove extern declaration of dump_ranger.
1375 * gimple-range-trace.cc (dump_ranger): Add DEBUG_FUNCTION marker.
1376 * gimple-range-trace.h (dump_ranger): Add prototype.
1378 2021-09-19 John Ericson <git@JohnEricson.me>
1380 * gcc.c (find_a_program): New function, factored out of...
1381 (find_a_file): Here.
1382 (execute): Use find_a_program when looking for programs rather
1385 2021-09-19 Matwey V. Kornilov <matwey.kornilov@gmail.com>
1387 * config/avr/avr-mcus.def: Add atmega324pb.
1388 * doc/avr-mmcu.texi: Corresponding changes.
1390 2021-09-19 Roger Sayle <roger@nextmovesoftware.com>
1393 * match.pd (cmp @0 REAL_CST@1): When @0 is also REAL_CST, apply
1394 the same transformations as to @1. For comparisons against NaN,
1395 don't check HONOR_SNANS but confirm that neither operand is a
1398 2021-09-19 Benjamin Peterson <benjamin@locrian.net>
1400 * attribs.c (make_unique_name): Delete.
1401 * attribs.h (make_unique_name): Delete.
1403 2021-09-19 Andrew Pinski <apinski@marvell.com>
1405 * lra-constraints.c (check_and_process_move): Assert
1406 that dclass and sclass are greater than or equal to NO_REGS.
1408 2021-09-18 Jakub Jelinek <jakub@redhat.com>
1410 * tree.h (OMP_CLAUSE_ORDER_UNCONSTRAINED): Define.
1411 * tree-pretty-print.c (dump_omp_clause): Print unconstrained:
1412 for OMP_CLAUSE_ORDER_UNCONSTRAINED.
1414 2021-09-18 liuhongt <hongtao.liu@intel.com>
1416 * config/i386/i386-features.c (remove_partial_avx_dependency):
1417 Restrict TARGET_USE_VECTOR_FP_CONVERTS and
1418 TARGET_USE_VECTOR_CONVERTS to conversion instructions only.
1420 2021-09-18 Jakub Jelinek <jakub@redhat.com>
1422 * gimplify.c (omp_default_clause): For C/C++ default({,first}private),
1423 if file/namespace scope variable doesn't have predetermined sharing,
1424 treat it as if there was default(none).
1426 2021-09-18 liuhongt <hongtao.liu@intel.com>
1428 * config/i386/avx512fp16intrin.h (_mm_fmadd_sh):
1430 (_mm_mask_fmadd_sh): Likewise.
1431 (_mm_mask3_fmadd_sh): Likewise.
1432 (_mm_maskz_fmadd_sh): Likewise.
1433 (_mm_fmadd_round_sh): Likewise.
1434 (_mm_mask_fmadd_round_sh): Likewise.
1435 (_mm_mask3_fmadd_round_sh): Likewise.
1436 (_mm_maskz_fmadd_round_sh): Likewise.
1437 (_mm_fnmadd_sh): Likewise.
1438 (_mm_mask_fnmadd_sh): Likewise.
1439 (_mm_mask3_fnmadd_sh): Likewise.
1440 (_mm_maskz_fnmadd_sh): Likewise.
1441 (_mm_fnmadd_round_sh): Likewise.
1442 (_mm_mask_fnmadd_round_sh): Likewise.
1443 (_mm_mask3_fnmadd_round_sh): Likewise.
1444 (_mm_maskz_fnmadd_round_sh): Likewise.
1445 (_mm_fmsub_sh): Likewise.
1446 (_mm_mask_fmsub_sh): Likewise.
1447 (_mm_mask3_fmsub_sh): Likewise.
1448 (_mm_maskz_fmsub_sh): Likewise.
1449 (_mm_fmsub_round_sh): Likewise.
1450 (_mm_mask_fmsub_round_sh): Likewise.
1451 (_mm_mask3_fmsub_round_sh): Likewise.
1452 (_mm_maskz_fmsub_round_sh): Likewise.
1453 (_mm_fnmsub_sh): Likewise.
1454 (_mm_mask_fnmsub_sh): Likewise.
1455 (_mm_mask3_fnmsub_sh): Likewise.
1456 (_mm_maskz_fnmsub_sh): Likewise.
1457 (_mm_fnmsub_round_sh): Likewise.
1458 (_mm_mask_fnmsub_round_sh): Likewise.
1459 (_mm_mask3_fnmsub_round_sh): Likewise.
1460 (_mm_maskz_fnmsub_round_sh): Likewise.
1461 * config/i386/i386-builtin-types.def
1462 (V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT): New builtin type.
1463 * config/i386/i386-builtin.def: Add new builtins.
1464 * config/i386/i386-expand.c: Handle new builtin type.
1465 * config/i386/sse.md (fmai_vmfmadd_<mode><round_name>):
1466 Ajdust to support FP16.
1467 (fmai_vmfmsub_<mode><round_name>): Ditto.
1468 (fmai_vmfnmadd_<mode><round_name>): Ditto.
1469 (fmai_vmfnmsub_<mode><round_name>): Ditto.
1470 (*fmai_fmadd_<mode>): Ditto.
1471 (*fmai_fmsub_<mode>): Ditto.
1472 (*fmai_fnmadd_<mode><round_name>): Ditto.
1473 (*fmai_fnmsub_<mode><round_name>): Ditto.
1474 (avx512f_vmfmadd_<mode>_mask<round_name>): Ditto.
1475 (avx512f_vmfmadd_<mode>_mask3<round_name>): Ditto.
1476 (avx512f_vmfmadd_<mode>_maskz<round_expand_name>): Ditto.
1477 (avx512f_vmfmadd_<mode>_maskz_1<round_name>): Ditto.
1478 (*avx512f_vmfmsub_<mode>_mask<round_name>): Ditto.
1479 (avx512f_vmfmsub_<mode>_mask3<round_name>): Ditto.
1480 (*avx512f_vmfmsub_<mode>_maskz_1<round_name>): Ditto.
1481 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
1482 (*avx512f_vmfnmsub_<mode>_mask3<round_name>): Ditto.
1483 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
1484 (*avx512f_vmfnmadd_<mode>_mask<round_name>): Renamed to ...
1485 (avx512f_vmfnmadd_<mode>_mask<round_name>) ... this, and
1486 adjust to support FP16.
1487 (avx512f_vmfnmadd_<mode>_mask3<round_name>): Ditto.
1488 (avx512f_vmfnmadd_<mode>_maskz_1<round_name>): Ditto.
1489 (avx512f_vmfnmadd_<mode>_maskz<round_expand_name>): New
1492 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
1494 * config/i386/sse.md (avx512fmaskmodelower): Extend to support
1496 (maskload<mode><avx512fmaskmodelower>): Ditto.
1497 (maskstore<mode><avx512fmaskmodelower>): Ditto.
1499 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
1501 * config/i386/i386-expand.c (ix86_expand_fp_absneg_operator):
1503 (ix86_expand_copysign): Ditto.
1504 (ix86_expand_xorsign): Ditto.
1505 * config/i386/i386.c (ix86_build_const_vector): Handle HF vector
1507 (ix86_build_signbit_mask): Ditto.
1508 (ix86_can_change_mode_class): Ditto.
1509 * config/i386/i386.md
1510 (SSEMODEF): Add HFmode.
1511 (ssevecmodef): Ditto.
1512 (<code>hf2): New define_expand.
1513 (*<code>hf2_1): New define_insn_and_split.
1514 (copysign<mode>): Extend to support HFmode under AVX512FP16.
1515 (xorsign<mode>): Ditto.
1516 * config/i386/sse.md (VFB): New mode iterator.
1517 (VFB_128_256): Ditto.
1519 (sseintvecmode2): Support HF vector mode.
1520 (<code><mode>2): Use new mode iterator.
1521 (*<code><mode>2): Ditto.
1522 (copysign<mode>3): Ditto.
1523 (xorsign<mode>3): Ditto.
1524 (<code><mode>3<mask_name>): Ditto.
1525 (<code><mode>3<mask_name>): Ditto.
1526 (<sse>_andnot<mode>3<mask_name>): Adjust for HF vector mode.
1527 (<sse>_andnot<mode>3<mask_name>): Ditto.
1528 (*<code><mode>3<mask_name>): Ditto.
1529 (*<code><mode>3<mask_name>): Ditto.
1531 2021-09-18 liuhongt <hongtao.liu@intel.com>
1533 * config/i386/avx512fp16intrin.h (_mm512_mask_fmadd_ph):
1535 (_mm512_mask3_fmadd_ph): Likewise.
1536 (_mm512_maskz_fmadd_ph): Likewise.
1537 (_mm512_fmadd_round_ph): Likewise.
1538 (_mm512_mask_fmadd_round_ph): Likewise.
1539 (_mm512_mask3_fmadd_round_ph): Likewise.
1540 (_mm512_maskz_fmadd_round_ph): Likewise.
1541 (_mm512_fnmadd_ph): Likewise.
1542 (_mm512_mask_fnmadd_ph): Likewise.
1543 (_mm512_mask3_fnmadd_ph): Likewise.
1544 (_mm512_maskz_fnmadd_ph): Likewise.
1545 (_mm512_fnmadd_round_ph): Likewise.
1546 (_mm512_mask_fnmadd_round_ph): Likewise.
1547 (_mm512_mask3_fnmadd_round_ph): Likewise.
1548 (_mm512_maskz_fnmadd_round_ph): Likewise.
1549 (_mm512_fmsub_ph): Likewise.
1550 (_mm512_mask_fmsub_ph): Likewise.
1551 (_mm512_mask3_fmsub_ph): Likewise.
1552 (_mm512_maskz_fmsub_ph): Likewise.
1553 (_mm512_fmsub_round_ph): Likewise.
1554 (_mm512_mask_fmsub_round_ph): Likewise.
1555 (_mm512_mask3_fmsub_round_ph): Likewise.
1556 (_mm512_maskz_fmsub_round_ph): Likewise.
1557 (_mm512_fnmsub_ph): Likewise.
1558 (_mm512_mask_fnmsub_ph): Likewise.
1559 (_mm512_mask3_fnmsub_ph): Likewise.
1560 (_mm512_maskz_fnmsub_ph): Likewise.
1561 (_mm512_fnmsub_round_ph): Likewise.
1562 (_mm512_mask_fnmsub_round_ph): Likewise.
1563 (_mm512_mask3_fnmsub_round_ph): Likewise.
1564 (_mm512_maskz_fnmsub_round_ph): Likewise.
1565 * config/i386/avx512fp16vlintrin.h (_mm256_fmadd_ph):
1567 (_mm256_mask_fmadd_ph): Likewise.
1568 (_mm256_mask3_fmadd_ph): Likewise.
1569 (_mm256_maskz_fmadd_ph): Likewise.
1570 (_mm_fmadd_ph): Likewise.
1571 (_mm_mask_fmadd_ph): Likewise.
1572 (_mm_mask3_fmadd_ph): Likewise.
1573 (_mm_maskz_fmadd_ph): Likewise.
1574 (_mm256_fnmadd_ph): Likewise.
1575 (_mm256_mask_fnmadd_ph): Likewise.
1576 (_mm256_mask3_fnmadd_ph): Likewise.
1577 (_mm256_maskz_fnmadd_ph): Likewise.
1578 (_mm_fnmadd_ph): Likewise.
1579 (_mm_mask_fnmadd_ph): Likewise.
1580 (_mm_mask3_fnmadd_ph): Likewise.
1581 (_mm_maskz_fnmadd_ph): Likewise.
1582 (_mm256_fmsub_ph): Likewise.
1583 (_mm256_mask_fmsub_ph): Likewise.
1584 (_mm256_mask3_fmsub_ph): Likewise.
1585 (_mm256_maskz_fmsub_ph): Likewise.
1586 (_mm_fmsub_ph): Likewise.
1587 (_mm_mask_fmsub_ph): Likewise.
1588 (_mm_mask3_fmsub_ph): Likewise.
1589 (_mm_maskz_fmsub_ph): Likewise.
1590 (_mm256_fnmsub_ph): Likewise.
1591 (_mm256_mask_fnmsub_ph): Likewise.
1592 (_mm256_mask3_fnmsub_ph): Likewise.
1593 (_mm256_maskz_fnmsub_ph): Likewise.
1594 (_mm_fnmsub_ph): Likewise.
1595 (_mm_mask_fnmsub_ph): Likewise.
1596 (_mm_mask3_fnmsub_ph): Likewise.
1597 (_mm_maskz_fnmsub_ph): Likewise.
1598 * config/i386/i386-builtin.def: Add corresponding new builtins.
1599 * config/i386/sse.md
1600 (<avx512>_fmadd_<mode>_maskz<round_expand_name>): Adjust to
1601 support HF vector modes.
1602 (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
1604 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
1605 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
1606 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
1607 (<avx512>_fmadd_<mode>_mask<round_name>): Ditto.
1608 (<avx512>_fmadd_<mode>_mask3<round_name>): Ditto.
1609 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Ditto.
1610 (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>):
1612 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
1613 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
1614 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
1615 (<avx512>_fmsub_<mode>_mask<round_name>): Ditto.
1616 (<avx512>_fmsub_<mode>_mask3<round_name>): Ditto.
1617 (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>):
1619 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
1620 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
1621 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
1622 (<avx512>_fnmadd_<mode>_mask<round_name>): Ditto.
1623 (<avx512>_fnmadd_<mode>_mask3<round_name>): Ditto.
1624 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Ditto.
1625 (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>):
1627 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
1628 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
1629 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
1630 (<avx512>_fnmsub_<mode>_mask<round_name>): Ditto.
1631 (<avx512>_fnmsub_<mode>_mask3<round_name>): Ditto.
1633 2021-09-18 liuhongt <hongtao.liu@intel.com>
1635 * config/i386/avx512fp16intrin.h (_mm512_fmaddsub_ph):
1637 (_mm512_mask_fmaddsub_ph): Likewise.
1638 (_mm512_mask3_fmaddsub_ph): Likewise.
1639 (_mm512_maskz_fmaddsub_ph): Likewise.
1640 (_mm512_fmaddsub_round_ph): Likewise.
1641 (_mm512_mask_fmaddsub_round_ph): Likewise.
1642 (_mm512_mask3_fmaddsub_round_ph): Likewise.
1643 (_mm512_maskz_fmaddsub_round_ph): Likewise.
1644 (_mm512_mask_fmsubadd_ph): Likewise.
1645 (_mm512_mask3_fmsubadd_ph): Likewise.
1646 (_mm512_maskz_fmsubadd_ph): Likewise.
1647 (_mm512_fmsubadd_round_ph): Likewise.
1648 (_mm512_mask_fmsubadd_round_ph): Likewise.
1649 (_mm512_mask3_fmsubadd_round_ph): Likewise.
1650 (_mm512_maskz_fmsubadd_round_ph): Likewise.
1651 * config/i386/avx512fp16vlintrin.h (_mm256_fmaddsub_ph):
1653 (_mm256_mask_fmaddsub_ph): Likewise.
1654 (_mm256_mask3_fmaddsub_ph): Likewise.
1655 (_mm256_maskz_fmaddsub_ph): Likewise.
1656 (_mm_fmaddsub_ph): Likewise.
1657 (_mm_mask_fmaddsub_ph): Likewise.
1658 (_mm_mask3_fmaddsub_ph): Likewise.
1659 (_mm_maskz_fmaddsub_ph): Likewise.
1660 (_mm256_fmsubadd_ph): Likewise.
1661 (_mm256_mask_fmsubadd_ph): Likewise.
1662 (_mm256_mask3_fmsubadd_ph): Likewise.
1663 (_mm256_maskz_fmsubadd_ph): Likewise.
1664 (_mm_fmsubadd_ph): Likewise.
1665 (_mm_mask_fmsubadd_ph): Likewise.
1666 (_mm_mask3_fmsubadd_ph): Likewise.
1667 (_mm_maskz_fmsubadd_ph): Likewise.
1668 * config/i386/i386-builtin.def: Add corresponding new builtins.
1669 * config/i386/sse.md (VFH_SF_AVX512VL): New mode iterator.
1670 * (<avx512>_fmsubadd_<mode>_maskz<round_expand_name>): New expander.
1671 * (<avx512>_fmaddsub_<mode>_maskz<round_expand_name>): Use
1673 * (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
1675 * (<avx512>_fmaddsub_<mode>_mask<round_name>): Ditto.
1676 * (<avx512>_fmaddsub_<mode>_mask3<round_name>): Ditto.
1677 * (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
1679 * (<avx512>_fmsubadd_<mode>_mask<round_name>): Ditto.
1680 * (<avx512>_fmsubadd_<mode>_mask3<round_name>): Ditto.
1682 2021-09-18 liuhongt <hongtao.liu@intel.com>
1685 * config/i386/i386.c (ix86_print_operand): Handle
1686 V8HF/V16HF/V32HFmode.
1687 * config/i386/i386.h (VALID_BCST_MODE_P): Add HFmode.
1688 * config/i386/sse.md (avx512bcst): Remove.
1690 2021-09-17 Martin Sebor <msebor@redhat.com>
1692 * Makefile.in (OBJS): Add gimple-predicate-analysis.o.
1693 * tree-ssa-uninit.c (max_phi_args): Move to gimple-predicate-analysis.
1694 (MASK_SET_BIT, MASK_TEST_BIT, MASK_EMPTY): Same.
1695 (check_defs): Add comment.
1696 (can_skip_redundant_opnd): Update comment.
1697 (compute_uninit_opnds_pos): Adjust to namespace change.
1698 (find_pdom): Move to gimple-predicate-analysis.cc.
1700 (struct uninit_undef_val_t): New.
1701 (is_non_loop_exit_postdominating): Move to gimple-predicate-analysis.cc.
1702 (find_control_equiv_block): Same.
1703 (MAX_NUM_CHAINS, MAX_CHAIN_LEN, MAX_POSTDOM_CHECK): Same.
1704 (MAX_SWITCH_CASES): Same.
1705 (compute_control_dep_chain): Same.
1706 (find_uninit_use): Use predicate analyzer.
1707 (struct pred_info): Move to gimple-predicate-analysis.
1708 (convert_control_dep_chain_into_preds): Same.
1709 (find_predicates): Same.
1710 (collect_phi_def_edges): Same.
1711 (warn_uninitialized_phi): Use predicate analyzer.
1712 (find_def_preds): Move to gimple-predicate-analysis.
1713 (dump_pred_info): Same.
1714 (dump_pred_chain): Same.
1715 (dump_predicates): Same.
1716 (destroy_predicate_vecs): Remove.
1717 (execute_late_warn_uninitialized): New.
1718 (get_cmp_code): Move to gimple-predicate-analysis.
1719 (is_value_included_in): Same.
1720 (value_sat_pred_p): Same.
1721 (find_matching_predicate_in_rest_chains): Same.
1722 (is_use_properly_guarded): Same.
1723 (prune_uninit_phi_opnds): Same.
1724 (find_var_cmp_const): Same.
1725 (use_pred_not_overlap_with_undef_path_pred): Same.
1726 (pred_equal_p): Same.
1727 (is_neq_relop_p): Same.
1728 (is_neq_zero_form_p): Same.
1729 (pred_expr_equal_p): Same.
1730 (is_pred_expr_subset_of): Same.
1731 (is_pred_chain_subset_of): Same.
1732 (is_included_in): Same.
1733 (is_superset_of): Same.
1735 (simplify_pred): Same.
1736 (simplify_preds_2): Same.
1737 (simplify_preds_3): Same.
1738 (simplify_preds_4): Same.
1739 (simplify_preds): Same.
1741 (push_to_worklist): Same.
1742 (get_pred_info_from_cmp): Same.
1743 (is_degenerated_phi): Same.
1744 (normalize_one_pred_1): Same.
1745 (normalize_one_pred): Same.
1746 (normalize_one_pred_chain): Same.
1747 (normalize_preds): Same.
1748 (can_one_predicate_be_invalidated_p): Same.
1749 (can_chain_union_be_invalidated_p): Same.
1750 (uninit_uses_cannot_happen): Same.
1751 (pass_late_warn_uninitialized::execute): Define.
1752 * gimple-predicate-analysis.cc: New file.
1753 * gimple-predicate-analysis.h: New file.
1755 2021-09-17 Julian Brown <julian@codesourcery.com>
1757 * config/gcn/gcn.c (gimple.h): Include.
1758 (gcn_fork_join): Emit barrier for worker-level joins.
1759 * omp-oacc-neuter-broadcast.cc (find_local_vars_to_propagate): Add
1760 writes_gang_private bitmap parameter. Set bit for blocks
1761 containing gang-private variable writes.
1762 (worker_single_simple): Don't emit barrier after predicated block.
1763 (worker_single_copy): Don't emit barrier if we're not broadcasting
1764 anything and the block contains no gang-private writes.
1765 (neuter_worker_single): Don't predicate blocks that only contain
1766 NOPs or internal marker functions. Pass has_gang_private_write
1767 argument to worker_single_copy.
1768 (oacc_do_neutering): Add writes_gang_private bitmap handling.
1770 2021-09-17 Julian Brown <julian@codesourcery.com>
1772 * config/gcn/gcn-protos.h
1773 (gcn_goacc_create_worker_broadcast_record): Update prototype.
1774 * config/gcn/gcn-tree.c (gcn_goacc_get_worker_red_decl): Use
1775 preallocated block of LDS memory. Do not cache/share decls for
1776 reduction temporaries between invocations.
1777 (gcn_goacc_reduction_teardown): Unshare VAR on second use.
1778 (gcn_goacc_create_worker_broadcast_record): Add OFFSET parameter
1779 and return temporary LDS space at that offset. Return pointer in
1781 * config/gcn/gcn.c (acc_lds_size, gang_private_hwm, lds_allocs):
1783 (ACC_LDS_SIZE): Define as acc_lds_size.
1784 (gcn_init_machine_status): Don't initialise lds_allocated,
1785 lds_allocs, reduc_decls fields of machine function struct.
1786 (gcn_option_override): Handle default size for gang-private
1787 variables and -mgang-private-size option.
1788 (gcn_expand_prologue): Use LDS_SIZE instead of LDS_SIZE-1 when
1789 initialising M0_REG.
1790 (gcn_shared_mem_layout): New function.
1791 (gcn_print_lds_decl): Update comment. Use global lds_allocs map and
1792 gang_private_hwm variable.
1793 (TARGET_GOACC_SHARED_MEM_LAYOUT): Define target hook.
1794 * config/gcn/gcn.h (machine_function): Remove lds_allocated,
1795 lds_allocs, reduc_decls. Add reduction_base, reduction_limit.
1796 * config/gcn/gcn.opt (gang_private_size_opt): New global.
1797 (mgang-private-size=): New option.
1798 * doc/tm.texi.in (TARGET_GOACC_SHARED_MEM_LAYOUT): Place
1800 * doc/tm.texi: Regenerate.
1801 * omp-oacc-neuter-broadcast.cc (targhooks.h, diagnostic-core.h):
1803 (build_sender_ref): Handle sender_decl being pointer.
1804 (worker_single_copy): Add PLACEMENT and ISOLATE_BROADCASTS
1805 parameters. Pass placement argument to
1806 create_worker_broadcast_record hook invocations. Handle
1807 sender_decl being pointer and isolate_broadcasts inserting extra
1809 (blk_offset_map_t): Add typedef.
1810 (neuter_worker_single): Add BLK_OFFSET_MAP parameter. Pass
1811 preallocated range to worker_single_copy call.
1812 (dfs_broadcast_reachable_1): New function.
1813 (idx_decl_pair_t, used_range_vec_t): New typedefs.
1814 (sort_size_descending): New function.
1815 (addr_range): New class.
1816 (splay_tree_compare_addr_range, splay_tree_free_key)
1817 (first_fit_range, merge_ranges_1, merge_ranges): New functions.
1818 (execute_omp_oacc_neuter_broadcast): Rename to...
1819 (oacc_do_neutering): ... this. Add BOUNDS_LO, BOUNDS_HI
1820 parameters. Arrange layout of shared memory for broadcast
1822 (execute_omp_oacc_neuter_broadcast): New function.
1823 (pass_omp_oacc_neuter_broadcast::gate): Remove num_workers==1
1824 handling from here. Enable pass for all OpenACC routines in order
1825 to call shared memory-layout hook.
1826 * target.def (create_worker_broadcast_record): Add OFFSET
1828 (shared_mem_layout): New hook.
1830 2021-09-17 Julian Brown <julian@codesourcery.com>
1831 Thomas Schwinge <thomas@codesourcery.com>
1833 * omp-oacc-neuter-broadcast.cc
1834 (pass_omp_oacc_neuter_broadcast::gate): Disable if num_workers is
1836 (execute_omp_oacc_neuter_broadcast): Adjust.
1838 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
1840 * value-relation.cc (class equiv_chain): Move to header file.
1841 (path_oracle::path_oracle): New.
1842 (path_oracle::~path_oracle): New.
1843 (path_oracle::register_relation): New.
1844 (path_oracle::query_relation): New.
1845 (path_oracle::reset_path): New.
1846 (path_oracle::dump): New.
1847 * value-relation.h (class equiv_chain): Move to here.
1848 (class path_oracle): New.
1850 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
1852 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a DOM
1854 * gimple-range-fold.cc (fur_depend::register_relation): Use
1855 register_stmt/edge routines.
1856 * value-relation.cc (equiv_chain::find): Relocate from equiv_oracle.
1857 (equiv_oracle::equiv_oracle): Create self equivalence cache.
1858 (equiv_oracle::~equiv_oracle): Release same.
1859 (equiv_oracle::equiv_set): Return entry from self equiv cache if there
1860 are no equivalences.
1861 (equiv_oracle::find_equiv_block): Move list find to equiv_chain.
1862 (equiv_oracle::register_relation): Rename from register_equiv.
1863 (relation_chain_head::find_relation): Relocate from dom_oracle.
1864 (relation_oracle::register_stmt): New.
1865 (relation_oracle::register_edge): New.
1866 (dom_oracle::*): Rename from relation_oracle.
1867 (dom_oracle::register_relation): Adjust to call equiv_oracle.
1868 (dom_oracle::set_one_relation): Split from register_relation.
1869 (dom_oracle::register_transitives): Consolidate 2 methods.
1870 (dom_oracle::find_relation_block): Move core to relation_chain.
1871 (dom_oracle::query_relation): Rename from find_relation_dom and adjust.
1872 * value-relation.h (class relation_oracle): New pure virtual base.
1873 (class equiv_oracle): Inherit from relation_oracle and adjust.
1874 (class dom_oracle): Rename from old relation_oracle and adjust.
1876 2021-09-17 Martin Sebor <msebor@redhat.com>
1878 PR middle-end/102200
1879 * pointer-query.cc (access_ref::inform_access): Handle MIN/MAX_EXPR.
1880 (handle_min_max_size): Change argument. Store original SSA_NAME for
1881 operands to potentially distinct (sub)objects.
1882 (compute_objsize_r): Adjust call to the above.
1884 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1886 * config/rs6000/rs6000.c (rs6000-builtins.h): New include.
1887 (rs6000_new_builtin_vectorized_function): New function.
1888 (rs6000_new_builtin_md_vectorized_function): Likewise.
1889 (rs6000_builtin_vectorized_function): Call
1890 rs6000_new_builtin_vectorized_function.
1891 (rs6000_builtin_md_vectorized_function): Call
1892 rs6000_new_builtin_md_vectorized_function.
1894 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1896 * config/rs6000/rs6000-builtin-new.def (ASSEMBLE_ACC): Add mmaint flag.
1897 (ASSEMBLE_PAIR): Likewise.
1898 (BUILD_ACC): Likewise.
1899 (DISASSEMBLE_ACC): Likewise.
1900 (DISASSEMBLE_PAIR): Likewise.
1901 (PMXVBF16GER2): Likewise.
1902 (PMXVBF16GER2NN): Likewise.
1903 (PMXVBF16GER2NP): Likewise.
1904 (PMXVBF16GER2PN): Likewise.
1905 (PMXVBF16GER2PP): Likewise.
1906 (PMXVF16GER2): Likewise.
1907 (PMXVF16GER2NN): Likewise.
1908 (PMXVF16GER2NP): Likewise.
1909 (PMXVF16GER2PN): Likewise.
1910 (PMXVF16GER2PP): Likewise.
1911 (PMXVF32GER): Likewise.
1912 (PMXVF32GERNN): Likewise.
1913 (PMXVF32GERNP): Likewise.
1914 (PMXVF32GERPN): Likewise.
1915 (PMXVF32GERPP): Likewise.
1916 (PMXVF64GER): Likewise.
1917 (PMXVF64GERNN): Likewise.
1918 (PMXVF64GERNP): Likewise.
1919 (PMXVF64GERPN): Likewise.
1920 (PMXVF64GERPP): Likewise.
1921 (PMXVI16GER2): Likewise.
1922 (PMXVI16GER2PP): Likewise.
1923 (PMXVI16GER2S): Likewise.
1924 (PMXVI16GER2SPP): Likewise.
1925 (PMXVI4GER8): Likewise.
1926 (PMXVI4GER8PP): Likewise.
1927 (PMXVI8GER4): Likewise.
1928 (PMXVI8GER4PP): Likewise.
1929 (PMXVI8GER4SPP): Likewise.
1930 (XVBF16GER2): Likewise.
1931 (XVBF16GER2NN): Likewise.
1932 (XVBF16GER2NP): Likewise.
1933 (XVBF16GER2PN): Likewise.
1934 (XVBF16GER2PP): Likewise.
1935 (XVF16GER2): Likewise.
1936 (XVF16GER2NN): Likewise.
1937 (XVF16GER2NP): Likewise.
1938 (XVF16GER2PN): Likewise.
1939 (XVF16GER2PP): Likewise.
1940 (XVF32GER): Likewise.
1941 (XVF32GERNN): Likewise.
1942 (XVF32GERNP): Likewise.
1943 (XVF32GERPN): Likewise.
1944 (XVF32GERPP): Likewise.
1945 (XVF64GER): Likewise.
1946 (XVF64GERNN): Likewise.
1947 (XVF64GERNP): Likewise.
1948 (XVF64GERPN): Likewise.
1949 (XVF64GERPP): Likewise.
1950 (XVI16GER2): Likewise.
1951 (XVI16GER2PP): Likewise.
1952 (XVI16GER2S): Likewise.
1953 (XVI16GER2SPP): Likewise.
1954 (XVI4GER8): Likewise.
1955 (XVI4GER8PP): Likewise.
1956 (XVI8GER4): Likewise.
1957 (XVI8GER4PP): Likewise.
1958 (XVI8GER4SPP): Likewise.
1959 (XXMFACC): Likewise.
1960 (XXMTACC): Likewise.
1961 (XXSETACCZ): Likewise.
1962 (ASSEMBLE_PAIR_V): Likewise.
1963 (BUILD_PAIR): Likewise.
1964 (DISASSEMBLE_PAIR_V): Likewise.
1967 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_mma_builtin):
1968 Handle RS6000_BIF_LXVP and RS6000_BIF_STXVP.
1969 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add ismmaint.
1970 (parse_bif_attrs): Handle ismmaint.
1971 (write_decls): Add bif_mmaint_bit and bif_is_mmaint.
1972 (write_bif_static_init): Handle ismmaint.
1974 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1976 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin): New
1978 (rs6000_gimple_fold_builtin): Call rs6000_gimple_fold_new_builtin.
1979 (rs6000_new_builtin_valid_without_lhs): New function.
1980 (rs6000_gimple_fold_new_mma_builtin): Likewise.
1981 (rs6000_gimple_fold_new_builtin): Likewise.
1983 2021-09-17 Thomas Schwinge <thomas@codesourcery.com>
1985 * hash-table.h (hash_table<Descriptor, Lazy, Allocator>::expand):
1986 Destruct stale Value objects.
1987 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
1990 2021-09-17 Roger Sayle <roger@nextmovesoftware.com>
1993 * match.pd (shift optimizations): Disable recent sign-changing
1994 optimization for shifts by zero, these will be folded later.
1996 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1998 * config/rs6000/rs6000-builtin-new.def (__builtin_mffsl): Move from
1999 [power9] to [always].
2001 2021-09-17 Richard Biener <rguenther@suse.de>
2003 * tree-vect-stmts.c (vectorizable_load): Do not frob
2006 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
2008 * config/i386/i386-features.c (remove_partial_avx_dependency):
2009 Also check TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY and
2010 and TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY before generating
2012 * config/i386/i386.h (TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY):
2014 (TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
2015 * config/i386/i386.md (SSE FP to FP splitters): Replace
2016 TARGET_SSE_PARTIAL_REG_DEPENDENCY with
2017 TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY.
2018 (SSE INT to FP splitter): Replace TARGET_SSE_PARTIAL_REG_DEPENDENCY
2019 with TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY.
2020 * config/i386/x86-tune.def
2021 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): New.
2022 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
2024 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
2027 * config/i386/i386-features.c (remove_partial_avx_dependency):
2028 Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
2029 before generating vxorps.
2031 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
2033 * config/i386/i386-options.c (processor_cost_table): Use
2034 tremont_cost for Tremont.
2035 * config/i386/x86-tune-costs.h (tremont_memcpy): New.
2036 (tremont_memset): Likewise.
2037 (tremont_cost): Likewise.
2038 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
2041 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
2043 * common/config/i386/i386-common.c: Use Haswell scheduling model
2045 * config/i386/i386.c (ix86_sched_init_global): Prepare for Tremont
2047 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Tremont
2049 (ix86_adjust_cost): Handle Tremont.
2050 * config/i386/x86-tune.def (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY):
2052 (X86_TUNE_USE_LEAVE): Likewise.
2053 (X86_TUNE_PUSH_MEMORY): Likewise.
2054 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
2055 (X86_TUNE_USE_CLTD): Likewise.
2056 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
2057 (X86_TUNE_AVOID_MFENCE): Likewise.
2058 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
2059 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
2060 (X86_TUNE_ACCUMULATE_OUTGOING_ARGS): Disable for Tremont.
2061 (X86_TUNE_FOUR_JUMP_LIMIT): Likewise.
2062 (X86_TUNE_OPT_AGU): Likewise.
2063 (X86_TUNE_AVOID_LEA_FOR_ADDR): Likewise.
2064 (X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE): Likewise.
2065 (X86_TUNE_EXPAND_ABS): Likewise.
2066 (X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS): Likewise.
2067 (X86_TUNE_SLOW_PSHUFB): Likewise.
2069 2021-09-17 Eric Botcazou <ebotcazou@adacore.com>
2071 PR rtl-optimization/102306
2072 * combine.c (try_combine): Abort the combination if we are about to
2073 duplicate volatile references.
2075 2021-09-17 liuhongt <hongtao.liu@intel.com>
2077 * config/i386/avx512fp16intrin.h (_mm_undefined_ph):
2079 (_mm256_undefined_ph): Likewise.
2080 (_mm512_undefined_ph): Likewise.
2081 (_mm_cvtsh_h): Likewise.
2082 (_mm256_cvtsh_h): Likewise.
2083 (_mm512_cvtsh_h): Likewise.
2084 (_mm512_castph_ps): Likewise.
2085 (_mm512_castph_pd): Likewise.
2086 (_mm512_castph_si512): Likewise.
2087 (_mm512_castph512_ph128): Likewise.
2088 (_mm512_castph512_ph256): Likewise.
2089 (_mm512_castph128_ph512): Likewise.
2090 (_mm512_castph256_ph512): Likewise.
2091 (_mm512_zextph128_ph512): Likewise.
2092 (_mm512_zextph256_ph512): Likewise.
2093 (_mm512_castps_ph): Likewise.
2094 (_mm512_castpd_ph): Likewise.
2095 (_mm512_castsi512_ph): Likewise.
2096 * config/i386/avx512fp16vlintrin.h (_mm_castph_ps):
2098 (_mm256_castph_ps): Likewise.
2099 (_mm_castph_pd): Likewise.
2100 (_mm256_castph_pd): Likewise.
2101 (_mm_castph_si128): Likewise.
2102 (_mm256_castph_si256): Likewise.
2103 (_mm_castps_ph): Likewise.
2104 (_mm256_castps_ph): Likewise.
2105 (_mm_castpd_ph): Likewise.
2106 (_mm256_castpd_ph): Likewise.
2107 (_mm_castsi128_ph): Likewise.
2108 (_mm256_castsi256_ph): Likewise.
2109 (_mm256_castph256_ph128): Likewise.
2110 (_mm256_castph128_ph256): Likewise.
2111 (_mm256_zextph128_ph256): Likewise.
2113 2021-09-17 liuhongt <hongtao.liu@intel.com>
2115 * config/i386/avx512fp16intrin.h (_mm_cvtsh_ss):
2117 (_mm_mask_cvtsh_ss): Likewise.
2118 (_mm_maskz_cvtsh_ss): Likewise.
2119 (_mm_cvtsh_sd): Likewise.
2120 (_mm_mask_cvtsh_sd): Likewise.
2121 (_mm_maskz_cvtsh_sd): Likewise.
2122 (_mm_cvt_roundsh_ss): Likewise.
2123 (_mm_mask_cvt_roundsh_ss): Likewise.
2124 (_mm_maskz_cvt_roundsh_ss): Likewise.
2125 (_mm_cvt_roundsh_sd): Likewise.
2126 (_mm_mask_cvt_roundsh_sd): Likewise.
2127 (_mm_maskz_cvt_roundsh_sd): Likewise.
2128 (_mm_cvtss_sh): Likewise.
2129 (_mm_mask_cvtss_sh): Likewise.
2130 (_mm_maskz_cvtss_sh): Likewise.
2131 (_mm_cvtsd_sh): Likewise.
2132 (_mm_mask_cvtsd_sh): Likewise.
2133 (_mm_maskz_cvtsd_sh): Likewise.
2134 (_mm_cvt_roundss_sh): Likewise.
2135 (_mm_mask_cvt_roundss_sh): Likewise.
2136 (_mm_maskz_cvt_roundss_sh): Likewise.
2137 (_mm_cvt_roundsd_sh): Likewise.
2138 (_mm_mask_cvt_roundsd_sh): Likewise.
2139 (_mm_maskz_cvt_roundsd_sh): Likewise.
2140 * config/i386/i386-builtin-types.def
2141 (V8HF_FTYPE_V2DF_V8HF_V8HF_UQI_INT,
2142 V8HF_FTYPE_V4SF_V8HF_V8HF_UQI_INT,
2143 V2DF_FTYPE_V8HF_V2DF_V2DF_UQI_INT,
2144 V4SF_FTYPE_V8HF_V4SF_V4SF_UQI_INT): Add new builtin types.
2145 * config/i386/i386-builtin.def: Add corrresponding new builtins.
2146 * config/i386/i386-expand.c: Handle new builtin types.
2147 * config/i386/sse.md (VF48_128): New mode iterator.
2148 (avx512fp16_vcvtsh2<ssescalarmodesuffix><mask_scalar_name><round_saeonly_scalar_name>):
2150 (avx512fp16_vcvt<ssescalarmodesuffix>2sh<mask_scalar_name><round_scalar_name>):
2153 2021-09-17 liuhongt <hongtao.liu@intel.com>
2155 * config/i386/avx512fp16intrin.h (_mm512_cvtph_pd):
2157 (_mm512_mask_cvtph_pd): Likewise.
2158 (_mm512_maskz_cvtph_pd): Likewise.
2159 (_mm512_cvt_roundph_pd): Likewise.
2160 (_mm512_mask_cvt_roundph_pd): Likewise.
2161 (_mm512_maskz_cvt_roundph_pd): Likewise.
2162 (_mm512_cvtxph_ps): Likewise.
2163 (_mm512_mask_cvtxph_ps): Likewise.
2164 (_mm512_maskz_cvtxph_ps): Likewise.
2165 (_mm512_cvtx_roundph_ps): Likewise.
2166 (_mm512_mask_cvtx_roundph_ps): Likewise.
2167 (_mm512_maskz_cvtx_roundph_ps): Likewise.
2168 (_mm512_cvtxps_ph): Likewise.
2169 (_mm512_mask_cvtxps_ph): Likewise.
2170 (_mm512_maskz_cvtxps_ph): Likewise.
2171 (_mm512_cvtx_roundps_ph): Likewise.
2172 (_mm512_mask_cvtx_roundps_ph): Likewise.
2173 (_mm512_maskz_cvtx_roundps_ph): Likewise.
2174 (_mm512_cvtpd_ph): Likewise.
2175 (_mm512_mask_cvtpd_ph): Likewise.
2176 (_mm512_maskz_cvtpd_ph): Likewise.
2177 (_mm512_cvt_roundpd_ph): Likewise.
2178 (_mm512_mask_cvt_roundpd_ph): Likewise.
2179 (_mm512_maskz_cvt_roundpd_ph): Likewise.
2180 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_pd):
2182 (_mm_mask_cvtph_pd): Likewise.
2183 (_mm_maskz_cvtph_pd): Likewise.
2184 (_mm256_cvtph_pd): Likewise.
2185 (_mm256_mask_cvtph_pd): Likewise.
2186 (_mm256_maskz_cvtph_pd): Likewise.
2187 (_mm_cvtxph_ps): Likewise.
2188 (_mm_mask_cvtxph_ps): Likewise.
2189 (_mm_maskz_cvtxph_ps): Likewise.
2190 (_mm256_cvtxph_ps): Likewise.
2191 (_mm256_mask_cvtxph_ps): Likewise.
2192 (_mm256_maskz_cvtxph_ps): Likewise.
2193 (_mm_cvtxps_ph): Likewise.
2194 (_mm_mask_cvtxps_ph): Likewise.
2195 (_mm_maskz_cvtxps_ph): Likewise.
2196 (_mm256_cvtxps_ph): Likewise.
2197 (_mm256_mask_cvtxps_ph): Likewise.
2198 (_mm256_maskz_cvtxps_ph): Likewise.
2199 (_mm_cvtpd_ph): Likewise.
2200 (_mm_mask_cvtpd_ph): Likewise.
2201 (_mm_maskz_cvtpd_ph): Likewise.
2202 (_mm256_cvtpd_ph): Likewise.
2203 (_mm256_mask_cvtpd_ph): Likewise.
2204 (_mm256_maskz_cvtpd_ph): Likewise.
2205 * config/i386/i386-builtin.def: Add corresponding new builtins.
2206 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2207 * config/i386/i386-expand.c: Handle new builtin types.
2208 * config/i386/sse.md
2209 (VF4_128_8_256): New.
2210 (VF48H_AVX512VL): Ditto.
2211 (ssePHmode): Add HF vector modes.
2212 (castmode): Add new convertable modes.
2215 (avx512fp16_vcvt<castmode>2ph_<mode><mask_name><round_name>): Ditto.
2216 (avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
2217 (*avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
2218 (avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
2219 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
2220 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask_1): Ditto.
2221 (avx512fp16_float_extend_ph<mode>2<mask_name><round_saeonly_name>):
2223 (avx512fp16_float_extend_ph<mode>2<mask_name>): Ditto.
2224 (*avx512fp16_float_extend_ph<mode>2_load<mask_name>): Ditto.
2225 (avx512fp16_float_extend_phv2df2<mask_name>): Ditto.
2226 (*avx512fp16_float_extend_phv2df2_load<mask_name>): Ditto.
2228 2021-09-17 liuhongt <hongtao.liu@intel.com>
2230 * config/i386/avx512fp16intrin.h (_mm_cvttsh_i32):
2232 (_mm_cvttsh_u32): Likewise.
2233 (_mm_cvtt_roundsh_i32): Likewise.
2234 (_mm_cvtt_roundsh_u32): Likewise.
2235 (_mm_cvttsh_i64): Likewise.
2236 (_mm_cvttsh_u64): Likewise.
2237 (_mm_cvtt_roundsh_i64): Likewise.
2238 (_mm_cvtt_roundsh_u64): Likewise.
2239 * config/i386/i386-builtin.def: Add corresponding new builtins.
2240 * config/i386/sse.md
2241 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<round_saeonly_name>):
2244 2021-09-17 liuhongt <hongtao.liu@intel.com>
2246 * config/i386/avx512fp16intrin.h (_mm512_cvttph_epi32):
2248 (_mm512_mask_cvttph_epi32): Likewise.
2249 (_mm512_maskz_cvttph_epi32): Likewise.
2250 (_mm512_cvtt_roundph_epi32): Likewise.
2251 (_mm512_mask_cvtt_roundph_epi32): Likewise.
2252 (_mm512_maskz_cvtt_roundph_epi32): Likewise.
2253 (_mm512_cvttph_epu32): Likewise.
2254 (_mm512_mask_cvttph_epu32): Likewise.
2255 (_mm512_maskz_cvttph_epu32): Likewise.
2256 (_mm512_cvtt_roundph_epu32): Likewise.
2257 (_mm512_mask_cvtt_roundph_epu32): Likewise.
2258 (_mm512_maskz_cvtt_roundph_epu32): Likewise.
2259 (_mm512_cvttph_epi64): Likewise.
2260 (_mm512_mask_cvttph_epi64): Likewise.
2261 (_mm512_maskz_cvttph_epi64): Likewise.
2262 (_mm512_cvtt_roundph_epi64): Likewise.
2263 (_mm512_mask_cvtt_roundph_epi64): Likewise.
2264 (_mm512_maskz_cvtt_roundph_epi64): Likewise.
2265 (_mm512_cvttph_epu64): Likewise.
2266 (_mm512_mask_cvttph_epu64): Likewise.
2267 (_mm512_maskz_cvttph_epu64): Likewise.
2268 (_mm512_cvtt_roundph_epu64): Likewise.
2269 (_mm512_mask_cvtt_roundph_epu64): Likewise.
2270 (_mm512_maskz_cvtt_roundph_epu64): Likewise.
2271 (_mm512_cvttph_epi16): Likewise.
2272 (_mm512_mask_cvttph_epi16): Likewise.
2273 (_mm512_maskz_cvttph_epi16): Likewise.
2274 (_mm512_cvtt_roundph_epi16): Likewise.
2275 (_mm512_mask_cvtt_roundph_epi16): Likewise.
2276 (_mm512_maskz_cvtt_roundph_epi16): Likewise.
2277 (_mm512_cvttph_epu16): Likewise.
2278 (_mm512_mask_cvttph_epu16): Likewise.
2279 (_mm512_maskz_cvttph_epu16): Likewise.
2280 (_mm512_cvtt_roundph_epu16): Likewise.
2281 (_mm512_mask_cvtt_roundph_epu16): Likewise.
2282 (_mm512_maskz_cvtt_roundph_epu16): Likewise.
2283 * config/i386/avx512fp16vlintrin.h (_mm_cvttph_epi32):
2285 (_mm_mask_cvttph_epi32): Likewise.
2286 (_mm_maskz_cvttph_epi32): Likewise.
2287 (_mm256_cvttph_epi32): Likewise.
2288 (_mm256_mask_cvttph_epi32): Likewise.
2289 (_mm256_maskz_cvttph_epi32): Likewise.
2290 (_mm_cvttph_epu32): Likewise.
2291 (_mm_mask_cvttph_epu32): Likewise.
2292 (_mm_maskz_cvttph_epu32): Likewise.
2293 (_mm256_cvttph_epu32): Likewise.
2294 (_mm256_mask_cvttph_epu32): Likewise.
2295 (_mm256_maskz_cvttph_epu32): Likewise.
2296 (_mm_cvttph_epi64): Likewise.
2297 (_mm_mask_cvttph_epi64): Likewise.
2298 (_mm_maskz_cvttph_epi64): Likewise.
2299 (_mm256_cvttph_epi64): Likewise.
2300 (_mm256_mask_cvttph_epi64): Likewise.
2301 (_mm256_maskz_cvttph_epi64): Likewise.
2302 (_mm_cvttph_epu64): Likewise.
2303 (_mm_mask_cvttph_epu64): Likewise.
2304 (_mm_maskz_cvttph_epu64): Likewise.
2305 (_mm256_cvttph_epu64): Likewise.
2306 (_mm256_mask_cvttph_epu64): Likewise.
2307 (_mm256_maskz_cvttph_epu64): Likewise.
2308 (_mm_cvttph_epi16): Likewise.
2309 (_mm_mask_cvttph_epi16): Likewise.
2310 (_mm_maskz_cvttph_epi16): Likewise.
2311 (_mm256_cvttph_epi16): Likewise.
2312 (_mm256_mask_cvttph_epi16): Likewise.
2313 (_mm256_maskz_cvttph_epi16): Likewise.
2314 (_mm_cvttph_epu16): Likewise.
2315 (_mm_mask_cvttph_epu16): Likewise.
2316 (_mm_maskz_cvttph_epu16): Likewise.
2317 (_mm256_cvttph_epu16): Likewise.
2318 (_mm256_mask_cvttph_epu16): Likewise.
2319 (_mm256_maskz_cvttph_epu16): Likewise.
2320 * config/i386/i386-builtin.def: Add new builtins.
2321 * config/i386/sse.md
2322 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name><round_saeonly_name>):
2324 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name>): Ditto.
2325 (*avx512fp16_fix<fixunssuffix>_trunc<mode>2_load<mask_name>): Ditto.
2326 (avx512fp16_fix<fixunssuffix>_truncv2di2<mask_name>): Ditto.
2327 (avx512fp16_fix<fixunssuffix>_truncv2di2_load<mask_name>): Ditto.
2329 2021-09-17 liuhongt <hongtao.liu@intel.com>
2331 * config/i386/avx512fp16intrin.h (_mm_cvtsh_i32): New intrinsic.
2332 (_mm_cvtsh_u32): Likewise.
2333 (_mm_cvt_roundsh_i32): Likewise.
2334 (_mm_cvt_roundsh_u32): Likewise.
2335 (_mm_cvtsh_i64): Likewise.
2336 (_mm_cvtsh_u64): Likewise.
2337 (_mm_cvt_roundsh_i64): Likewise.
2338 (_mm_cvt_roundsh_u64): Likewise.
2339 (_mm_cvti32_sh): Likewise.
2340 (_mm_cvtu32_sh): Likewise.
2341 (_mm_cvt_roundi32_sh): Likewise.
2342 (_mm_cvt_roundu32_sh): Likewise.
2343 (_mm_cvti64_sh): Likewise.
2344 (_mm_cvtu64_sh): Likewise.
2345 (_mm_cvt_roundi64_sh): Likewise.
2346 (_mm_cvt_roundu64_sh): Likewise.
2347 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2348 * config/i386/i386-builtin.def: Add corresponding new builtins.
2349 * config/i386/i386-expand.c (ix86_expand_round_builtin):
2350 Handle new builtin types.
2351 * config/i386/sse.md
2352 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix><round_name>):
2354 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix>_2): Likewise.
2355 (avx512fp16_vcvt<floatsuffix>si2sh<rex64namesuffix><round_name>): Likewise.
2357 2021-09-16 Bill Schmidt <wschmidt@linux.ibm.com>
2359 * config/rs6000/rs6000-c.c (rs6000-builtins.h): New include.
2360 (altivec_resolve_new_overloaded_builtin): New forward decl.
2361 (rs6000_new_builtin_type_compatible): New function.
2362 (altivec_resolve_overloaded_builtin): Call
2363 altivec_resolve_new_overloaded_builtin.
2364 (altivec_build_new_resolved_builtin): New function.
2365 (altivec_resolve_new_overloaded_builtin): Likewise.
2366 * config/rs6000/rs6000-call.c (rs6000_new_builtin_is_supported):
2368 * config/rs6000/rs6000-gen-builtins.c (write_decls): Remove _p from
2369 name of rs6000_new_builtin_is_supported.
2371 2021-09-16 Uroš Bizjak <ubizjak@gmail.com>
2373 * config/i386/i386-protos.h (ix86_decompose_address):
2374 Change return type to bool.
2375 * config/i386/i386.c (ix86_decompose_address): Ditto.
2377 2021-09-16 Tobias Burnus <tobias@codesourcery.com>
2380 * config/rs6000/t-rs6000 (build/rs6000-gen-builtins.o, build/rbtree.o):
2381 Added 'build/' to target, use build/%.o rule.
2382 (build/rs6000-gen-builtins$(build_exeext)): Add 'build/' and
2383 '$(build_exeext)' to target and 'build/' for the *.o files.
2384 (rs6000-builtins.c): Update for those changes; run rs6000-gen-builtins
2387 2021-09-16 Martin Jambor <mjambor@suse.cz>
2389 * cgraph.c (cgraph_node::dump): Do not check caller count sums if
2390 the body has been removed. Remove trailing whitespace.
2392 2021-09-16 Richard Biener <rguenther@suse.de>
2394 PR middle-end/102360
2395 * internal-fn.c (expand_DEFERRED_INIT): Make pattern-init
2396 of non-memory more robust.
2398 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2400 * config/sparc/sparc-opts.h (enum sparc_processor_type): Add LEON5
2401 * config/sparc/sparc.c (struct processor_costs): Add LEON5 costs
2402 (leon5_adjust_cost): Increase cost of store with data dependency
2403 on ALU instruction and FPU anti-dependencies.
2404 (sparc_option_override): Add LEON5 costs
2405 (sparc_adjust_cost): Add LEON5 cost adjustments
2406 * config/sparc/sparc.h: Add LEON5
2407 * config/sparc/sparc.md: Include LEON5 scheduling information
2408 * config/sparc/sparc.opt: Add LEON5
2409 * doc/invoke.texi: Add LEON5
2410 * config/sparc/leon5.md: New file.
2412 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2414 * config/sparc/sparc.md (stack_protect_set32): Add NOP to prevent
2415 sensitive sequence for B2BST errata workaround.
2417 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2419 * config/sparc/sparc.c (sparc_do_work_around_errata): Do not begin
2420 functions with atomic instruction in the UT700 errata workaround.
2422 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2424 * config/sparc/sparc.c (next_active_non_empty_insn): New function
2425 that returns next active non empty assembly instruction.
2426 (sparc_do_work_around_errata): Use new function.
2428 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2430 * config/sparc/sparc.c (store_insn_p): Add predicate for store
2432 (load_insn_p): Add predicate for load attributes.
2433 (sparc_do_work_around_errata): Use new predicates.
2435 2021-09-16 Andreas Larsson <andreas@gaisler.com>
2437 * config/sparc/sparc.c (dump_target_flag_bits): Print bit names for
2440 2021-09-16 Martin Liska <mliska@suse.cz>
2442 * config/mips/netbsd.h: Fix typo in name of a macro.
2444 2021-09-16 liuhongt <hongtao.liu@intel.com>
2446 PR middle-end/102080
2447 * match.pd: Check mask type when doing cond_op related gimple
2449 * tree.c (is_truth_type_for): New function.
2450 * tree.h (is_truth_type_for): New declaration.
2452 2021-09-16 liuhongt <hongtao.liu@intel.com>
2454 * config/i386/avx512fp16intrin.h (_mm512_cvtepi32_ph): New
2456 (_mm512_mask_cvtepi32_ph): Likewise.
2457 (_mm512_maskz_cvtepi32_ph): Likewise.
2458 (_mm512_cvt_roundepi32_ph): Likewise.
2459 (_mm512_mask_cvt_roundepi32_ph): Likewise.
2460 (_mm512_maskz_cvt_roundepi32_ph): Likewise.
2461 (_mm512_cvtepu32_ph): Likewise.
2462 (_mm512_mask_cvtepu32_ph): Likewise.
2463 (_mm512_maskz_cvtepu32_ph): Likewise.
2464 (_mm512_cvt_roundepu32_ph): Likewise.
2465 (_mm512_mask_cvt_roundepu32_ph): Likewise.
2466 (_mm512_maskz_cvt_roundepu32_ph): Likewise.
2467 (_mm512_cvtepi64_ph): Likewise.
2468 (_mm512_mask_cvtepi64_ph): Likewise.
2469 (_mm512_maskz_cvtepi64_ph): Likewise.
2470 (_mm512_cvt_roundepi64_ph): Likewise.
2471 (_mm512_mask_cvt_roundepi64_ph): Likewise.
2472 (_mm512_maskz_cvt_roundepi64_ph): Likewise.
2473 (_mm512_cvtepu64_ph): Likewise.
2474 (_mm512_mask_cvtepu64_ph): Likewise.
2475 (_mm512_maskz_cvtepu64_ph): Likewise.
2476 (_mm512_cvt_roundepu64_ph): Likewise.
2477 (_mm512_mask_cvt_roundepu64_ph): Likewise.
2478 (_mm512_maskz_cvt_roundepu64_ph): Likewise.
2479 (_mm512_cvtepi16_ph): Likewise.
2480 (_mm512_mask_cvtepi16_ph): Likewise.
2481 (_mm512_maskz_cvtepi16_ph): Likewise.
2482 (_mm512_cvt_roundepi16_ph): Likewise.
2483 (_mm512_mask_cvt_roundepi16_ph): Likewise.
2484 (_mm512_maskz_cvt_roundepi16_ph): Likewise.
2485 (_mm512_cvtepu16_ph): Likewise.
2486 (_mm512_mask_cvtepu16_ph): Likewise.
2487 (_mm512_maskz_cvtepu16_ph): Likewise.
2488 (_mm512_cvt_roundepu16_ph): Likewise.
2489 (_mm512_mask_cvt_roundepu16_ph): Likewise.
2490 (_mm512_maskz_cvt_roundepu16_ph): Likewise.
2491 * config/i386/avx512fp16vlintrin.h (_mm_cvtepi32_ph): New
2493 (_mm_mask_cvtepi32_ph): Likewise.
2494 (_mm_maskz_cvtepi32_ph): Likewise.
2495 (_mm256_cvtepi32_ph): Likewise.
2496 (_mm256_mask_cvtepi32_ph): Likewise.
2497 (_mm256_maskz_cvtepi32_ph): Likewise.
2498 (_mm_cvtepu32_ph): Likewise.
2499 (_mm_mask_cvtepu32_ph): Likewise.
2500 (_mm_maskz_cvtepu32_ph): Likewise.
2501 (_mm256_cvtepu32_ph): Likewise.
2502 (_mm256_mask_cvtepu32_ph): Likewise.
2503 (_mm256_maskz_cvtepu32_ph): Likewise.
2504 (_mm_cvtepi64_ph): Likewise.
2505 (_mm_mask_cvtepi64_ph): Likewise.
2506 (_mm_maskz_cvtepi64_ph): Likewise.
2507 (_mm256_cvtepi64_ph): Likewise.
2508 (_mm256_mask_cvtepi64_ph): Likewise.
2509 (_mm256_maskz_cvtepi64_ph): Likewise.
2510 (_mm_cvtepu64_ph): Likewise.
2511 (_mm_mask_cvtepu64_ph): Likewise.
2512 (_mm_maskz_cvtepu64_ph): Likewise.
2513 (_mm256_cvtepu64_ph): Likewise.
2514 (_mm256_mask_cvtepu64_ph): Likewise.
2515 (_mm256_maskz_cvtepu64_ph): Likewise.
2516 (_mm_cvtepi16_ph): Likewise.
2517 (_mm_mask_cvtepi16_ph): Likewise.
2518 (_mm_maskz_cvtepi16_ph): Likewise.
2519 (_mm256_cvtepi16_ph): Likewise.
2520 (_mm256_mask_cvtepi16_ph): Likewise.
2521 (_mm256_maskz_cvtepi16_ph): Likewise.
2522 (_mm_cvtepu16_ph): Likewise.
2523 (_mm_mask_cvtepu16_ph): Likewise.
2524 (_mm_maskz_cvtepu16_ph): Likewise.
2525 (_mm256_cvtepu16_ph): Likewise.
2526 (_mm256_mask_cvtepu16_ph): Likewise.
2527 (_mm256_maskz_cvtepu16_ph): Likewise.
2528 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2529 * config/i386/i386-builtin.def: Add corresponding new builtins.
2530 * config/i386/i386-expand.c
2531 (ix86_expand_args_builtin): Handle new builtin types.
2532 (ix86_expand_round_builtin): Ditto.
2533 * config/i386/i386-modes.def: Declare V2HF and V6HF.
2534 * config/i386/sse.md (VI2H_AVX512VL): New.
2536 (sseintvecmode): Add HF vector modes.
2537 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode><mask_name><round_name>):
2539 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
2540 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
2541 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
2542 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
2543 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask_1): Ditto.
2544 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
2545 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
2546 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
2547 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
2548 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask_1): Ditto.
2549 * config/i386/subst.md (round_qq2phsuff): New subst_attr.
2551 2021-09-16 liuhongt <hongtao.liu@intel.com>
2553 * config/i386/avx512fp16intrin.h (_mm512_cvtph_epi32):
2555 (_mm512_mask_cvtph_epi32): Likewise.
2556 (_mm512_maskz_cvtph_epi32): Likewise.
2557 (_mm512_cvt_roundph_epi32): Likewise.
2558 (_mm512_mask_cvt_roundph_epi32): Likewise.
2559 (_mm512_maskz_cvt_roundph_epi32): Likewise.
2560 (_mm512_cvtph_epu32): Likewise.
2561 (_mm512_mask_cvtph_epu32): Likewise.
2562 (_mm512_maskz_cvtph_epu32): Likewise.
2563 (_mm512_cvt_roundph_epu32): Likewise.
2564 (_mm512_mask_cvt_roundph_epu32): Likewise.
2565 (_mm512_maskz_cvt_roundph_epu32): Likewise.
2566 (_mm512_cvtph_epi64): Likewise.
2567 (_mm512_mask_cvtph_epi64): Likewise.
2568 (_mm512_maskz_cvtph_epi64): Likewise.
2569 (_mm512_cvt_roundph_epi64): Likewise.
2570 (_mm512_mask_cvt_roundph_epi64): Likewise.
2571 (_mm512_maskz_cvt_roundph_epi64): Likewise.
2572 (_mm512_cvtph_epu64): Likewise.
2573 (_mm512_mask_cvtph_epu64): Likewise.
2574 (_mm512_maskz_cvtph_epu64): Likewise.
2575 (_mm512_cvt_roundph_epu64): Likewise.
2576 (_mm512_mask_cvt_roundph_epu64): Likewise.
2577 (_mm512_maskz_cvt_roundph_epu64): Likewise.
2578 (_mm512_cvtph_epi16): Likewise.
2579 (_mm512_mask_cvtph_epi16): Likewise.
2580 (_mm512_maskz_cvtph_epi16): Likewise.
2581 (_mm512_cvt_roundph_epi16): Likewise.
2582 (_mm512_mask_cvt_roundph_epi16): Likewise.
2583 (_mm512_maskz_cvt_roundph_epi16): Likewise.
2584 (_mm512_cvtph_epu16): Likewise.
2585 (_mm512_mask_cvtph_epu16): Likewise.
2586 (_mm512_maskz_cvtph_epu16): Likewise.
2587 (_mm512_cvt_roundph_epu16): Likewise.
2588 (_mm512_mask_cvt_roundph_epu16): Likewise.
2589 (_mm512_maskz_cvt_roundph_epu16): Likewise.
2590 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_epi32):
2592 (_mm_mask_cvtph_epi32): Likewise.
2593 (_mm_maskz_cvtph_epi32): Likewise.
2594 (_mm256_cvtph_epi32): Likewise.
2595 (_mm256_mask_cvtph_epi32): Likewise.
2596 (_mm256_maskz_cvtph_epi32): Likewise.
2597 (_mm_cvtph_epu32): Likewise.
2598 (_mm_mask_cvtph_epu32): Likewise.
2599 (_mm_maskz_cvtph_epu32): Likewise.
2600 (_mm256_cvtph_epu32): Likewise.
2601 (_mm256_mask_cvtph_epu32): Likewise.
2602 (_mm256_maskz_cvtph_epu32): Likewise.
2603 (_mm_cvtph_epi64): Likewise.
2604 (_mm_mask_cvtph_epi64): Likewise.
2605 (_mm_maskz_cvtph_epi64): Likewise.
2606 (_mm256_cvtph_epi64): Likewise.
2607 (_mm256_mask_cvtph_epi64): Likewise.
2608 (_mm256_maskz_cvtph_epi64): Likewise.
2609 (_mm_cvtph_epu64): Likewise.
2610 (_mm_mask_cvtph_epu64): Likewise.
2611 (_mm_maskz_cvtph_epu64): Likewise.
2612 (_mm256_cvtph_epu64): Likewise.
2613 (_mm256_mask_cvtph_epu64): Likewise.
2614 (_mm256_maskz_cvtph_epu64): Likewise.
2615 (_mm_cvtph_epi16): Likewise.
2616 (_mm_mask_cvtph_epi16): Likewise.
2617 (_mm_maskz_cvtph_epi16): Likewise.
2618 (_mm256_cvtph_epi16): Likewise.
2619 (_mm256_mask_cvtph_epi16): Likewise.
2620 (_mm256_maskz_cvtph_epi16): Likewise.
2621 (_mm_cvtph_epu16): Likewise.
2622 (_mm_mask_cvtph_epu16): Likewise.
2623 (_mm_maskz_cvtph_epu16): Likewise.
2624 (_mm256_cvtph_epu16): Likewise.
2625 (_mm256_mask_cvtph_epu16): Likewise.
2626 (_mm256_maskz_cvtph_epu16): Likewise.
2627 * config/i386/i386-builtin-types.def: Add new builtin types.
2628 * config/i386/i386-builtin.def: Add new builtins.
2629 * config/i386/i386-expand.c
2630 (ix86_expand_args_builtin): Handle new builtin types.
2631 (ix86_expand_round_builtin): Ditto.
2632 * config/i386/sse.md (sseintconvert): New.
2634 (UNSPEC_US_FIX_NOTRUNC): Ditto.
2635 (sseintconvertsignprefix): Ditto.
2636 (avx512fp16_vcvtph2<sseintconvertsignprefix><sseintconvert>_<mode><mask_name><round_name>):
2639 2021-09-16 liuhongt <hongtao.liu@intel.com>
2641 * config/i386/avx512fp16intrin.h: (_mm_cvtsi16_si128):
2643 (_mm_cvtsi128_si16): Likewise.
2644 (_mm_mask_load_sh): Likewise.
2645 (_mm_maskz_load_sh): Likewise.
2646 (_mm_mask_store_sh): Likewise.
2647 (_mm_move_sh): Likewise.
2648 (_mm_mask_move_sh): Likewise.
2649 (_mm_maskz_move_sh): Likewise.
2650 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2651 * config/i386/i386-builtin.def: Add corresponding new builtins.
2652 * config/i386/i386-expand.c
2653 (ix86_expand_special_args_builtin): Handle new builtin types.
2654 (ix86_expand_vector_init_one_nonzero): Adjust for FP16 target.
2655 * config/i386/sse.md (VI2F): New mode iterator.
2656 (vec_set<mode>_0): Use new mode iterator.
2657 (avx512f_mov<ssescalarmodelower>_mask): Adjust for HF vector mode.
2658 (avx512f_store<mode>_mask): Ditto.
2660 2021-09-16 Kewen Lin <linkw@linux.ibm.com>
2662 * config/rs6000/rs6000.opt (-mtoc-fusion): Remove.
2664 2021-09-15 David Edelsohn <dje.gcc@gmail.com>
2666 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
2667 Proceed if no symbol summary or the symbol alias flag is false.
2669 2021-09-15 Jakub Jelinek <jakub@redhat.com>
2673 * varasm.c (output_constructor_regular_field): Instead of assertion
2674 that array_size_for_constructor result is equal to size of
2675 TREE_TYPE (local->val) in bytes, assert that the type size is greater
2676 or equal to array_size_for_constructor result and use type size as
2679 2021-09-15 Martin Liska <mliska@suse.cz>
2682 * config/i386/vxworks.h: Use new macro TARGET_CPU_P.
2684 2021-09-15 Martin Liska <mliska@suse.cz>
2687 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
2688 Check that we have a symbol summary for a symbol.
2690 2021-09-15 Richard Biener <rguenther@suse.de>
2693 * config/rs6000/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
2694 to inherit from elfos.h
2696 2021-09-15 liuhongt <hongtao.liu@intel.com>
2699 * config/i386/i386-expand.c
2700 (ix86_expand_vector_init_interleave): Use puncklwd to pack 2
2702 (ix86_expand_vector_set): Use blendw instead of pinsrw.
2703 * config/i386/i386.c (ix86_can_change_mode_class): Adjust for
2704 AVX512FP16 which supports 16bit vector load.
2705 * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>):
2707 (avx512bw_interleave_high<mode><mask_name>): .. this, and
2708 extend to V32HFmode.
2709 (avx2_interleave_highv16hi<mask_name>): Rename to ..
2710 (avx2_interleave_high<mode><mask_name>): .. this, and extend
2712 (vec_interleave_highv8hi<mask_name>): Rename to ..
2713 (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode.
2714 (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>):
2716 (<mask_codefor>avx512bw_interleave_low<mode><mask_name>):
2717 this, and extend to V32HFmode.
2718 (avx2_interleave_lowv16hi<mask_name>): Rename to ..
2719 (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode.
2720 (vec_interleave_lowv8hi<mask_name>): Rename to ..
2721 (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode.
2722 (sse4_1_pblendw): Rename to ..
2723 (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode.
2724 (avx2_pblendph): New define_expand.
2725 (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use
2726 sseintmodesuffix instead of ssemodesuffix.
2727 (blendsuf): New mode attr.
2729 2021-09-15 Richard Biener <rguenther@suse.de>
2731 * tree-vectorizer.h (dr_misalignment): Move out of line.
2732 (dr_target_alignment): New.
2733 (DR_TARGET_ALIGNMENT): Wrap dr_target_alignment.
2734 (set_dr_target_alignment): New.
2735 (SET_DR_TARGET_ALIGNMENT): Wrap set_dr_target_alignment.
2736 * tree-vect-data-refs.c (dr_misalignment): Compute and
2737 return the group members misalignment.
2738 (vect_compute_data_ref_alignment): Use SET_DR_TARGET_ALIGNMENT.
2739 (vect_analyze_data_refs_alignment): Compute alignment only
2740 for the first element of a DR group.
2741 (vect_slp_analyze_node_alignment): Likewise.
2743 2021-09-15 Hongyu Wang <hongyu.wang@intel.com>
2745 * config/i386/avx512fp16intrin.h: Adjust all builtin calls.
2746 * config/i386/avx512fp16vlintrin.h: Likewise.
2747 * config/i386/i386-builtin.def: Adjust builtin name and
2748 enumeration to match AVX512F style.
2750 2021-09-15 Richard Biener <rguenther@suse.de>
2752 PR tree-optimization/102318
2753 * tree-vect-loop.c (vect_transform_cycle_phi): Revert
2754 previous change and do the mode conversion separately from
2755 the sign conversion.
2757 2021-09-15 Hongtao Liu <hongtao.liu@intel.com>
2758 Peter Cordes <peter@cordes.ca>
2761 * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
2762 (*vec_extract<mode><ssescalarmodelower>_valign): Output
2763 vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
2766 2021-09-15 Richard Biener <rguenther@suse.de>
2768 * config.gcc: Remove vax-*-openbsd* configuration.
2770 2021-09-15 Richard Biener <rguenther@suse.de>
2772 * config.gcc: Remove m68k-openbsd.
2774 2021-09-15 Max Filippov <jcmvbkbc@gmail.com>
2777 * config/xtensa/t-xtensa (TM_H): Add include/xtensa-config.h.
2779 2021-09-14 Peter Bergner <bergner@linux.ibm.com>
2781 * config/rs6000/mma.md (unspec): Delete UNSPEC_MMA_XXSETACCZ.
2782 (unspecv): Add UNSPECV_MMA_XXSETACCZ.
2783 (*mma_xxsetaccz): Delete.
2784 (mma_xxsetaccz): Change to define_insn. Remove operand 1.
2785 Use UNSPECV_MMA_XXSETACCZ. Update comment.
2786 * config/rs6000/rs6000.c (rs6000_rtx_costs): Use UNSPECV_MMA_XXSETACCZ.
2788 2021-09-14 Iain Sandoe <iain@sandoe.co.uk>
2790 * Makefile.in: Remove variables related to applying no-PIE
2791 to the exes on $build.
2792 * configure: Regenerate.
2793 * configure.ac: Remove configuration related to applying
2794 no-PIE to the exes on $build.
2796 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
2798 * config/arc/arc.md (doloop_end): Add missing mode.
2799 (loop_end): Likewise.
2801 2021-09-14 Jakub Jelinek <jakub@redhat.com>
2803 * gimplify.c (goa_stabilize_expr): Add depth argument, propagate
2804 it to recursive calls, for depth above 7 just gimplify or return.
2805 Perform a test even for MODIFY_EXPR, ADDR_EXPR, COMPOUND_EXPR with
2806 __builtin_clear_padding and TARGET_EXPR.
2807 (gimplify_omp_atomic): Adjust goa_stabilize_expr callers.
2809 2021-09-14 liuhongt <hongtao.liu@intel.com>
2811 * config/i386/avx512fp16intrin.h (_mm_fpclass_sh_mask):
2813 (_mm_mask_fpclass_sh_mask): Likewise.
2814 (_mm512_mask_fpclass_ph_mask): Likewise.
2815 (_mm512_fpclass_ph_mask): Likewise.
2816 (_mm_getexp_sh): Likewise.
2817 (_mm_mask_getexp_sh): Likewise.
2818 (_mm_maskz_getexp_sh): Likewise.
2819 (_mm512_getexp_ph): Likewise.
2820 (_mm512_mask_getexp_ph): Likewise.
2821 (_mm512_maskz_getexp_ph): Likewise.
2822 (_mm_getexp_round_sh): Likewise.
2823 (_mm_mask_getexp_round_sh): Likewise.
2824 (_mm_maskz_getexp_round_sh): Likewise.
2825 (_mm512_getexp_round_ph): Likewise.
2826 (_mm512_mask_getexp_round_ph): Likewise.
2827 (_mm512_maskz_getexp_round_ph): Likewise.
2828 (_mm_getmant_sh): Likewise.
2829 (_mm_mask_getmant_sh): Likewise.
2830 (_mm_maskz_getmant_sh): Likewise.
2831 (_mm512_getmant_ph): Likewise.
2832 (_mm512_mask_getmant_ph): Likewise.
2833 (_mm512_maskz_getmant_ph): Likewise.
2834 (_mm_getmant_round_sh): Likewise.
2835 (_mm_mask_getmant_round_sh): Likewise.
2836 (_mm_maskz_getmant_round_sh): Likewise.
2837 (_mm512_getmant_round_ph): Likewise.
2838 (_mm512_mask_getmant_round_ph): Likewise.
2839 (_mm512_maskz_getmant_round_ph): Likewise.
2840 * config/i386/avx512fp16vlintrin.h (_mm_mask_fpclass_ph_mask):
2842 (_mm_fpclass_ph_mask): Likewise.
2843 (_mm256_mask_fpclass_ph_mask): Likewise.
2844 (_mm256_fpclass_ph_mask): Likewise.
2845 (_mm256_getexp_ph): Likewise.
2846 (_mm256_mask_getexp_ph): Likewise.
2847 (_mm256_maskz_getexp_ph): Likewise.
2848 (_mm_getexp_ph): Likewise.
2849 (_mm_mask_getexp_ph): Likewise.
2850 (_mm_maskz_getexp_ph): Likewise.
2851 (_mm256_getmant_ph): Likewise.
2852 (_mm256_mask_getmant_ph): Likewise.
2853 (_mm256_maskz_getmant_ph): Likewise.
2854 (_mm_getmant_ph): Likewise.
2855 (_mm_mask_getmant_ph): Likewise.
2856 (_mm_maskz_getmant_ph): Likewise.
2857 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2858 * config/i386/i386-builtin.def: Add corresponding new builtins.
2859 * config/i386/i386-expand.c
2860 (ix86_expand_args_builtin): Handle new builtin types.
2861 (ix86_expand_round_builtin): Ditto.
2862 * config/i386/sse.md (vecmemsuffix): Add HF vector modes.
2863 (<avx512>_getexp<mode><mask_name><round_saeonly_name>): Adjust
2864 to support HF vector modes.
2865 (avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name):
2867 (avx512dq_fpclass<mode><mask_scalar_merge_name>): Ditto.
2868 (avx512dq_vmfpclass<mode><mask_scalar_merge_name>): Ditto.
2869 (<avx512>_getmant<mode><mask_name><round_saeonly_name>): Ditto.
2870 (avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2873 2021-09-14 liuhongt <hongtao.liu@intel.com>
2875 * config/i386/avx512fp16intrin.h (_mm512_reduce_ph):
2877 (_mm512_mask_reduce_ph): Likewise.
2878 (_mm512_maskz_reduce_ph): Likewise.
2879 (_mm512_reduce_round_ph): Likewise.
2880 (_mm512_mask_reduce_round_ph): Likewise.
2881 (_mm512_maskz_reduce_round_ph): Likewise.
2882 (_mm_reduce_sh): Likewise.
2883 (_mm_mask_reduce_sh): Likewise.
2884 (_mm_maskz_reduce_sh): Likewise.
2885 (_mm_reduce_round_sh): Likewise.
2886 (_mm_mask_reduce_round_sh): Likewise.
2887 (_mm_maskz_reduce_round_sh): Likewise.
2888 (_mm512_roundscale_ph): Likewise.
2889 (_mm512_mask_roundscale_ph): Likewise.
2890 (_mm512_maskz_roundscale_ph): Likewise.
2891 (_mm512_roundscale_round_ph): Likewise.
2892 (_mm512_mask_roundscale_round_ph): Likewise.
2893 (_mm512_maskz_roundscale_round_ph): Likewise.
2894 (_mm_roundscale_sh): Likewise.
2895 (_mm_mask_roundscale_sh): Likewise.
2896 (_mm_maskz_roundscale_sh): Likewise.
2897 (_mm_roundscale_round_sh): Likewise.
2898 (_mm_mask_roundscale_round_sh): Likewise.
2899 (_mm_maskz_roundscale_round_sh): Likewise.
2900 * config/i386/avx512fp16vlintrin.h: (_mm_reduce_ph):
2902 (_mm_mask_reduce_ph): Likewise.
2903 (_mm_maskz_reduce_ph): Likewise.
2904 (_mm256_reduce_ph): Likewise.
2905 (_mm256_mask_reduce_ph): Likewise.
2906 (_mm256_maskz_reduce_ph): Likewise.
2907 (_mm_roundscale_ph): Likewise.
2908 (_mm_mask_roundscale_ph): Likewise.
2909 (_mm_maskz_roundscale_ph): Likewise.
2910 (_mm256_roundscale_ph): Likewise.
2911 (_mm256_mask_roundscale_ph): Likewise.
2912 (_mm256_maskz_roundscale_ph): Likewise.
2913 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2914 * config/i386/i386-builtin.def: Add corresponding new builtins.
2915 * config/i386/i386-expand.c
2916 (ix86_expand_args_builtin): Handle new builtin types.
2917 (ix86_expand_round_builtin): Ditto.
2918 * config/i386/sse.md (<mask_codefor>reducep<mode><mask_name>):
2920 (<mask_codefor>reducep<mode><mask_name><round_saeonly_name>):
2921 ... this, and adjust for round operands.
2922 (reduces<mode><mask_scalar_name>): Likewise, with ...
2923 (reduces<mode><mask_scalar_name><round_saeonly_scalar_name):
2925 (<avx512>_rndscale<mode><mask_name><round_saeonly_name>):
2926 Adjust for HF vector modes.
2927 (avx512f_rndscale<mode><mask_scalar_name><round_saeonly_scalar_name>):
2929 (*avx512f_rndscale<mode><round_saeonly_name>): Ditto.
2931 2021-09-14 liuhongt <hongtao.liu@intel.com>
2933 * config/i386/avx512fp16intrin.h: (_mm512_rcp_ph):
2935 (_mm512_mask_rcp_ph): Likewise.
2936 (_mm512_maskz_rcp_ph): Likewise.
2937 (_mm_rcp_sh): Likewise.
2938 (_mm_mask_rcp_sh): Likewise.
2939 (_mm_maskz_rcp_sh): Likewise.
2940 (_mm512_scalef_ph): Likewise.
2941 (_mm512_mask_scalef_ph): Likewise.
2942 (_mm512_maskz_scalef_ph): Likewise.
2943 (_mm512_scalef_round_ph): Likewise.
2944 (_mm512_mask_scalef_round_ph): Likewise.
2945 (_mm512_maskz_scalef_round_ph): Likewise.
2946 (_mm_scalef_sh): Likewise.
2947 (_mm_mask_scalef_sh): Likewise.
2948 (_mm_maskz_scalef_sh): Likewise.
2949 (_mm_scalef_round_sh): Likewise.
2950 (_mm_mask_scalef_round_sh): Likewise.
2951 (_mm_maskz_scalef_round_sh): Likewise.
2952 * config/i386/avx512fp16vlintrin.h (_mm_rcp_ph):
2954 (_mm256_rcp_ph): Likewise.
2955 (_mm_mask_rcp_ph): Likewise.
2956 (_mm256_mask_rcp_ph): Likewise.
2957 (_mm_maskz_rcp_ph): Likewise.
2958 (_mm256_maskz_rcp_ph): Likewise.
2959 (_mm_scalef_ph): Likewise.
2960 (_mm256_scalef_ph): Likewise.
2961 (_mm_mask_scalef_ph): Likewise.
2962 (_mm256_mask_scalef_ph): Likewise.
2963 (_mm_maskz_scalef_ph): Likewise.
2964 (_mm256_maskz_scalef_ph): Likewise.
2965 * config/i386/i386-builtin.def: Add new builtins.
2966 * config/i386/sse.md (VFH_AVX512VL): New.
2967 (avx512fp16_rcp<mode>2<mask_name>): Ditto.
2968 (avx512fp16_vmrcpv8hf2<mask_scalar_name>): Ditto.
2969 (avx512f_vmscalef<mode><mask_scalar_name><round_scalar_name>):
2970 Adjust to support HF vector modes.
2971 (<avx512>_scalef<mode><mask_name><round_name>): Ditto.
2973 2021-09-14 liuhongt <hongtao.liu@intel.com>
2975 * config/i386/avx512fp16intrin.h: (_mm512_sqrt_ph):
2977 (_mm512_mask_sqrt_ph): Likewise.
2978 (_mm512_maskz_sqrt_ph): Likewise.
2979 (_mm512_sqrt_round_ph): Likewise.
2980 (_mm512_mask_sqrt_round_ph): Likewise.
2981 (_mm512_maskz_sqrt_round_ph): Likewise.
2982 (_mm512_rsqrt_ph): Likewise.
2983 (_mm512_mask_rsqrt_ph): Likewise.
2984 (_mm512_maskz_rsqrt_ph): Likewise.
2985 (_mm_rsqrt_sh): Likewise.
2986 (_mm_mask_rsqrt_sh): Likewise.
2987 (_mm_maskz_rsqrt_sh): Likewise.
2988 (_mm_sqrt_sh): Likewise.
2989 (_mm_mask_sqrt_sh): Likewise.
2990 (_mm_maskz_sqrt_sh): Likewise.
2991 (_mm_sqrt_round_sh): Likewise.
2992 (_mm_mask_sqrt_round_sh): Likewise.
2993 (_mm_maskz_sqrt_round_sh): Likewise.
2994 * config/i386/avx512fp16vlintrin.h (_mm_sqrt_ph): New intrinsic.
2995 (_mm256_sqrt_ph): Likewise.
2996 (_mm_mask_sqrt_ph): Likewise.
2997 (_mm256_mask_sqrt_ph): Likewise.
2998 (_mm_maskz_sqrt_ph): Likewise.
2999 (_mm256_maskz_sqrt_ph): Likewise.
3000 (_mm_rsqrt_ph): Likewise.
3001 (_mm256_rsqrt_ph): Likewise.
3002 (_mm_mask_rsqrt_ph): Likewise.
3003 (_mm256_mask_rsqrt_ph): Likewise.
3004 (_mm_maskz_rsqrt_ph): Likewise.
3005 (_mm256_maskz_rsqrt_ph): Likewise.
3006 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3007 * config/i386/i386-builtin.def: Add corresponding new builtins.
3008 * config/i386/i386-expand.c
3009 (ix86_expand_args_builtin): Handle new builtins.
3010 (ix86_expand_round_builtin): Ditto.
3011 * config/i386/sse.md (VF_AVX512FP16VL): New.
3012 (sqrt<mode>2): Adjust for HF vector modes.
3013 (<sse>_sqrt<mode>2<mask_name><round_name>): Likewise.
3014 (<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
3016 (<sse>_rsqrt<mode>2<mask_name>): New.
3017 (avx512fp16_vmrsqrtv8hf2<mask_scalar_name>): Likewise.
3019 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
3022 * diagnostic-spec.c (warning_suppressed_at, copy_warning): Handle
3023 'RESERVED_LOCATION_P' locations.
3024 * warning-control.cc (get_nowarn_spec, suppress_warning)
3025 (copy_warning): Likewise.
3027 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
3029 * diagnostic-spec.h (typedef xint_hash_t): Use 'location_t' instead of...
3030 (typedef key_type_t): ... this. Remove.
3031 (nowarn_map): Document.
3032 * diagnostic-spec.c (nowarn_map): Likewise.
3033 * warning-control.cc (convert_to_key): Evolve functions into...
3034 (get_location): ... these. Adjust all users.
3036 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
3038 * warning-control.cc (copy_warning): Remove 'nowarn_map' setup.
3040 2021-09-13 Jason Merrill <jason@redhat.com>
3042 * params.opt: Add destructive-interference-size and
3043 constructive-interference-size.
3044 * doc/invoke.texi: Document them.
3045 * config/aarch64/aarch64.c (aarch64_override_options_internal):
3047 * config/arm/arm.c (arm_option_override): Set them.
3048 * config/i386/i386-options.c (ix86_option_override_internal):
3051 2021-09-13 Martin Liska <mliska@suse.cz>
3052 H.J. Lu <hjl.tools@gmail.com>
3055 * common/config/i386/cpuinfo.h (cpu_indicator_init): Add support
3056 for x86-64 micro levels for __builtin_cpu_supports.
3057 * common/config/i386/i386-cpuinfo.h (enum feature_priority):
3058 Add priorities for the micro-arch levels.
3059 (enum processor_features): Add new features.
3060 * common/config/i386/i386-isas.h: Add micro-arch features.
3061 * config/i386/i386-builtins.c (get_builtin_code_for_version):
3062 Support the micro-arch levels by callsing
3063 __builtin_cpu_supports.
3064 * doc/extend.texi: Document that the levels are support by
3065 __builtin_cpu_supports.
3067 2021-09-13 Andrew Pinski <apinski@marvell.com>
3070 * config/aarch64/aarch64-builtins.c (aarch64_fold_builtin_lane_check):
3072 (aarch64_general_fold_builtin): Handle AARCH64_SIMD_BUILTIN_LANE_CHECK.
3073 (aarch64_general_gimple_fold_builtin): Likewise.
3075 2021-09-13 Andrew Pinski <apinski@marvell.com>
3077 * config.gcc: Add m32r-*-linux* and m32rle-*-linux*
3078 to the Unsupported targets list.
3079 Remove support for m32r-*-linux* and m32rle-*-linux*.
3080 * config/m32r/linux.h: Removed.
3081 * config/m32r/t-linux: Removed.
3083 2021-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3086 * config/aarch64/aarch64.c (aarch64_classify_address): Don't allow
3087 register index for SVE predicate modes.
3089 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
3091 * tree-ssa-threadbackward.c
3092 (back_threader_profitability::profitable_path_p): Remove FSM
3094 (back_threader_registry::register_path): Same.
3095 * tree-ssa-threadedge.c
3096 (jump_threader::simplify_control_stmt_condition): Same.
3097 * tree-ssa-threadupdate.c (jt_path_registry::jt_path_registry):
3098 Add backedge_threads argument.
3099 (fwd_jt_path_registry::fwd_jt_path_registry): Pass
3100 backedge_threads argument.
3101 (back_jt_path_registry::back_jt_path_registry): Same.
3102 (dump_jump_thread_path): Adjust for FSM removal.
3103 (back_jt_path_registry::rewire_first_differing_edge): Same.
3104 (back_jt_path_registry::adjust_paths_after_duplication): Same.
3105 (back_jt_path_registry::update_cfg): Same.
3106 (jt_path_registry::register_jump_thread): Same.
3107 * tree-ssa-threadupdate.h (enum jump_thread_edge_type): Remove
3109 (class back_jt_path_registry): Add backedge_threads to
3112 2021-09-13 Martin Liska <mliska@suse.cz>
3115 * asan.h (sanitize_coverage_p): Handle when fn == NULL.
3117 2021-09-13 H.J. Lu <hjl.tools@gmail.com>
3120 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): New.
3121 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
3122 (MOVE_MAX): Check TARGET_AVX256_MOVE_BY_PIECES and
3123 TARGET_AVX256_STORE_BY_PIECES instead of
3124 TARGET_AVX256_SPLIT_UNALIGNED_LOAD and
3125 TARGET_AVX256_SPLIT_UNALIGNED_STORE.
3126 (STORE_MAX_PIECES): Check TARGET_AVX256_STORE_BY_PIECES instead
3127 of TARGET_AVX256_SPLIT_UNALIGNED_STORE.
3128 * config/i386/x86-tune.def (X86_TUNE_AVX256_MOVE_BY_PIECES): New.
3129 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
3131 2021-09-13 liuhongt <hongtao.liu@intel.com>
3134 * expmed.c (extract_bit_field_using_extv): Use
3135 gen_lowpart_if_possible instead of gen_lowpart to avoid ICE.
3137 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
3139 * Makefile.in (OBJS): Add value-pointer-equiv.o.
3140 * gimple-ssa-evrp.c (class ssa_equiv_stack): Move to
3141 value-pointer-equiv.*.
3142 (ssa_equiv_stack::ssa_equiv_stack): Same.
3143 (ssa_equiv_stack::enter): Same.
3144 (ssa_equiv_stack::leave): Same.
3145 (ssa_equiv_stack::push_replacement): Same.
3146 (ssa_equiv_stack::get_replacement): Same.
3147 (is_pointer_ssa): Same.
3148 (class pointer_equiv_analyzer): Same.
3149 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
3150 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Same.
3151 (pointer_equiv_analyzer::set_global_equiv): Same.
3152 (pointer_equiv_analyzer::set_cond_equiv): Same.
3153 (pointer_equiv_analyzer::get_equiv): Same.
3154 (pointer_equiv_analyzer::enter): Same.
3155 (pointer_equiv_analyzer::leave): Same.
3156 (pointer_equiv_analyzer::get_equiv_expr): Same.
3157 (pta_valueize): Same.
3158 (pointer_equiv_analyzer::visit_stmt): Same.
3159 (pointer_equiv_analyzer::visit_edge): Same.
3160 (hybrid_folder::value_of_expr): Same.
3161 (hybrid_folder::value_on_edge): Same.
3162 * value-pointer-equiv.cc: New file.
3163 * value-pointer-equiv.h: New file.
3165 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
3168 * gimple-fold.c (gimple_fold_builtin_memory_op): Allow folding
3169 memcpy if the size is not more than MOVE_MAX * MOVE_RATIO.
3171 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
3174 * config/arm/arm.md (movmisaligndi): New define_expand.
3175 * config/arm/vec-common.md (movmisalign<mode>): Iterate over VDQ mode.
3177 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
3180 * emit-rtl.c (gen_highpart): Use adjust_address to handle
3181 MEM rather than calling simplify_gen_subreg.
3183 2021-09-13 Jan-Benedict Glaw <jbglaw@ług-owl.de>
3185 * config/alpha/vms.h (INIT_CUMULATIVE_ARGS): Wrap multi-statment
3186 define into a block.
3188 2021-09-13 Richard Biener <rguenther@suse.de>
3190 * config/darwin.h (DARWIN_PREFER_DWARF): Do not define.
3191 * config/i386/darwin.h (PREFERRED_DEBUGGING_TYPE): Do not
3192 change based on DARWIN_PREFER_DWARF not being defined.
3194 2021-09-13 Richard Biener <rguenther@suse.de>
3196 * config/i386/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
3197 to inherit from elfos.h
3199 2021-09-13 Richard Biener <rguenther@suse.de>
3201 * config.gcc: Add cr16-*-* to the list of obsoleted targets.
3203 2021-09-13 Richard Biener <rguenther@suse.de>
3205 * config/avr/elf.h (PREFERRED_DEBUGGING_TYPE): Remove
3206 override, pick up DWARF2_DEBUG define from elfos.h
3208 2021-09-13 Richard Biener <rguenther@suse.de>
3210 * config/rx/rx.h (PREFERRED_DEBUGGING_TYPE): Always define to
3213 2021-09-13 Richard Biener <rguenther@suse.de>
3215 * config/alpha/vms.h (PREFERRED_DEBUGGING_TYPE): Define to
3218 2021-09-13 Richard Biener <rguenther@suse.de>
3220 * config/i386/cygming.h: Always default to DWARF2 debugging.
3221 Do not define DBX_DEBUGGING_INFO, that's done via dbxcoff.h
3223 * doc/install.texi: Document binutils 2.16 as minimum
3224 requirement for mingw.
3226 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
3228 * config/rs6000/rs6000.c (struct rs6000_cost_data): New members
3229 nstmts, nloads and extra_ctor_cost.
3230 (rs6000_density_test): Add load density related heuristics. Do
3231 extra costing on vector construction statements if need.
3232 (rs6000_init_cost): Init new members.
3233 (rs6000_update_target_cost_per_stmt): New function.
3234 (rs6000_add_stmt_cost): Factor vect_nonmem hunk out to function
3235 rs6000_update_target_cost_per_stmt and call it.
3237 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
3239 * config/rs6000/rs6000.c (struct rs6000_cost_data): Remove typedef.
3240 (rs6000_init_cost): Adjust.
3242 2021-09-13 liuhongt <hongtao.liu@intel.com>
3244 * config/i386/i386.md: (UNSPEC_COPYSIGN): Remove.
3245 (UNSPEC_XORSIGN): Ditto.
3247 2021-09-12 Roger Sayle <roger@nextmovesoftware.com>
3249 * expr.c (convert_move): Preserve SUBREG_PROMOTED_VAR_P when
3250 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
3253 2021-09-11 Aldy Hernandez <aldyh@redhat.com>
3255 * tree-ssa-threadbackward.c (class back_threader_registry): Use
3256 back_jt_path_registry.
3257 * tree-ssa-threadedge.c (jump_threader::jump_threader): Use
3258 fwd_jt_path_registry.
3259 * tree-ssa-threadedge.h (class jump_threader): Same..
3260 * tree-ssa-threadupdate.c
3261 (jump_thread_path_registry::jump_thread_path_registry): Rename...
3262 (jt_path_registry::jt_path_registry): ...to this.
3263 (jump_thread_path_registry::~jump_thread_path_registry): Rename...
3264 (jt_path_registry::~jt_path_registry): ...this.
3265 (fwd_jt_path_registry::fwd_jt_path_registry): New.
3266 (fwd_jt_path_registry::~fwd_jt_path_registry): New.
3267 (jump_thread_path_registry::allocate_thread_edge): Rename...
3268 (jt_path_registry::allocate_thread_edge): ...to this.
3269 (jump_thread_path_registry::allocate_thread_path): Rename...
3270 (jt_path_registry::allocate_thread_path): ...to this.
3271 (jump_thread_path_registry::lookup_redirection_data): Rename...
3272 (fwd_jt_path_registry::lookup_redirection_data): ...to this.
3273 (jump_thread_path_registry::thread_block_1): Rename...
3274 (fwd_jt_path_registry::thread_block_1): ...to this.
3275 (jump_thread_path_registry::thread_block): Rename...
3276 (fwd_jt_path_registry::thread_block): ...to this.
3277 (jt_path_registry::thread_through_loop_header): Rename...
3278 (fwd_jt_path_registry::thread_through_loop_header): ...to this.
3279 (jump_thread_path_registry::mark_threaded_blocks): Rename...
3280 (fwd_jt_path_registry::mark_threaded_blocks): ...to this.
3281 (jump_thread_path_registry::debug_path): Rename...
3282 (jt_path_registry::debug_path): ...to this.
3283 (jump_thread_path_registry::dump): Rename...
3284 (jt_path_registry::debug): ...to this.
3285 (jump_thread_path_registry::rewire_first_differing_edge): Rename...
3286 (back_jt_path_registry::rewire_first_differing_edge): ...to this.
3287 (jump_thread_path_registry::adjust_paths_after_duplication): Rename...
3288 (back_jt_path_registry::adjust_paths_after_duplication): ...to this.
3289 (jump_thread_path_registry::duplicate_thread_path): Rename...
3290 (back_jt_path_registry::duplicate_thread_path): ...to this. Also,
3291 drop ill-formed candidates.
3292 (jump_thread_path_registry::remove_jump_threads_including): Rename...
3293 (fwd_jt_path_registry::remove_jump_threads_including): ...to this.
3294 (jt_path_registry::thread_through_all_blocks): New.
3295 (back_jt_path_registry::update_cfg): New.
3296 (fwd_jt_path_registry::update_cfg): New.
3297 (jump_thread_path_registry::register_jump_thread): Rename...
3298 (jt_path_registry::register_jump_thread): ...to this.
3299 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
3301 (class jt_path_registry): ...here.
3302 (class fwd_jt_path_registry): New.
3303 (class back_jt_path_registry): New.
3305 2021-09-10 liuhongt <hongtao.liu@intel.com>
3308 2021-09-01 liuhongt <hongtao.liu@intel.com>
3310 * emit-rtl.c (validate_subreg): Get rid of all float-int
3313 2021-09-10 Jakub Jelinek <jakub@redhat.com>
3315 * tree-core.h (enum omp_memory_order): Add OMP_MEMORY_ORDER_MASK,
3316 OMP_FAIL_MEMORY_ORDER_UNSPECIFIED, OMP_FAIL_MEMORY_ORDER_RELAXED,
3317 OMP_FAIL_MEMORY_ORDER_ACQUIRE, OMP_FAIL_MEMORY_ORDER_RELEASE,
3318 OMP_FAIL_MEMORY_ORDER_ACQ_REL, OMP_FAIL_MEMORY_ORDER_SEQ_CST and
3319 OMP_FAIL_MEMORY_ORDER_MASK enumerators.
3320 (OMP_FAIL_MEMORY_ORDER_SHIFT): Define.
3321 * gimple-pretty-print.c (dump_gimple_omp_atomic_load,
3322 dump_gimple_omp_atomic_store): Print [weak] for weak atomic
3324 * gimple.h (enum gf_mask): Change GF_OMP_ATOMIC_MEMORY_ORDER
3325 to 6-bit mask, adjust GF_OMP_ATOMIC_NEED_VALUE value and add
3327 (gimple_omp_atomic_weak_p, gimple_omp_atomic_set_weak): New inline
3329 * tree.h (OMP_ATOMIC_WEAK): Define.
3330 * tree-pretty-print.c (dump_omp_atomic_memory_order): Adjust for
3331 fail memory order being encoded in the same enum and also print
3332 fail clause if present.
3333 (dump_generic_node): Print weak clause if OMP_ATOMIC_WEAK.
3334 * gimplify.c (goa_stabilize_expr): Add target_expr and rhs arguments,
3335 handle pre_p == NULL case as a test mode that only returns value
3336 but doesn't change gimplify nor change anything otherwise, adjust
3337 recursive calls, add MODIFY_EXPR, ADDR_EXPR, COND_EXPR, TARGET_EXPR
3338 and CALL_EXPR handling, adjust COMPOUND_EXPR handling for
3339 __builtin_clear_padding calls, for !rhs gimplify as lvalue rather
3341 (gimplify_omp_atomic): Adjust goa_stabilize_expr caller. Handle
3342 COND_EXPR rhs. Set weak flag on gimple load/store for
3344 * omp-expand.c (omp_memory_order_to_fail_memmodel): New function.
3345 (omp_memory_order_to_memmodel): Adjust for fail clause encoded
3347 (expand_omp_atomic_cas): New function.
3348 (expand_omp_atomic_pipeline): Use omp_memory_order_to_fail_memmodel
3350 (expand_omp_atomic): Attempt to optimize atomic compare and exchange
3351 using expand_omp_atomic_cas.
3353 2021-09-10 Aldy Hernandez <aldyh@redhat.com>
3354 Michael Matz <matz@suse.de>
3356 * tree-pass.h (PROP_loop_opts_done): New.
3357 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
3358 Intersect with global range.
3359 * tree-ssa-loop.c (tree_ssa_loop_done): Set PROP_loop_opts_done.
3360 * tree-ssa-threadbackward.c
3361 (back_threader_profitability::profitable_path_p): Disable
3362 threading through latches until after loop optimizations have run.
3364 2021-09-10 David Faust <david.faust@oracle.com>
3366 * doc/invoke.texi: Document BPF -mcpu, -mjmpext, -mjmp32 and -malu32
3369 2021-09-10 David Faust <david.faust@oracle.com>
3371 * config/bpf/bpf-opts.h (bpf_isa_version): New enum.
3372 * config/bpf/bpf-protos.h (bpf_expand_cbranch): New.
3373 * config/bpf/bpf.c (bpf_option_override): Handle -mcpu option.
3374 (bpf_expand_cbranch): New function.
3375 * config/bpf/bpf.md (AM mode iterator): Conditionalize support for SI
3377 (zero_extendsidi2): Only use mov32 instruction if it is available.
3378 (SIM mode iterator): Conditionalize support for SI mode.
3379 (JM mode iterator): New.
3380 (cbranchdi4): Update name, use new JM iterator. Use bpf_expand_cbranch.
3381 (*branch_on_di): Update name, use new JM iterator.
3382 * config/bpf/bpf.opt: (mjmpext): New option.
3386 (bpf_isa): New enum.
3388 2021-09-10 David Faust <david.faust@oracle.com>
3390 * config/bpf/bpf.md (zero_extendhidi2): Add new output template
3391 for register-to-register extensions.
3392 (zero_extendqidi2): Likewise.
3394 2021-09-10 Richard Biener <rguenther@suse.de>
3396 PR middle-end/102273
3397 * internal-fn.c (expand_DEFERRED_INIT): Always expand non-SSA vars.
3399 2021-09-10 Richard Biener <rguenther@suse.de>
3401 PR middle-end/102269
3402 * gimplify.c (is_var_need_auto_init): Empty types do not need
3405 2021-09-10 Richard Biener <rguenther@suse.de>
3407 * configure.ac (--with-stabs): Remove.
3408 * configure: Regenerate.
3409 * doc/install.texi: Remove --with-stabs documentation.
3411 2021-09-10 liuhongt <hongtao.liu@intel.com>
3413 * config/i386/avx512fp16intrin.h: (_mm512_cmp_ph_mask):
3415 (_mm512_mask_cmp_ph_mask): Likewise.
3416 (_mm512_cmp_round_ph_mask): Likewise.
3417 (_mm512_mask_cmp_round_ph_mask): Likewise.
3418 (_mm_cmp_sh_mask): Likewise.
3419 (_mm_mask_cmp_sh_mask): Likewise.
3420 (_mm_cmp_round_sh_mask): Likewise.
3421 (_mm_mask_cmp_round_sh_mask): Likewise.
3422 (_mm_comieq_sh): Likewise.
3423 (_mm_comilt_sh): Likewise.
3424 (_mm_comile_sh): Likewise.
3425 (_mm_comigt_sh): Likewise.
3426 (_mm_comige_sh): Likewise.
3427 (_mm_comineq_sh): Likewise.
3428 (_mm_ucomieq_sh): Likewise.
3429 (_mm_ucomilt_sh): Likewise.
3430 (_mm_ucomile_sh): Likewise.
3431 (_mm_ucomigt_sh): Likewise.
3432 (_mm_ucomige_sh): Likewise.
3433 (_mm_ucomineq_sh): Likewise.
3434 (_mm_comi_round_sh): Likewise.
3435 (_mm_comi_sh): Likewise.
3436 * config/i386/avx512fp16vlintrin.h (_mm_cmp_ph_mask): New intrinsic.
3437 (_mm_mask_cmp_ph_mask): Likewise.
3438 (_mm256_cmp_ph_mask): Likewise.
3439 (_mm256_mask_cmp_ph_mask): Likewise.
3440 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3441 * config/i386/i386-builtin.def: Add corresponding new builtins.
3442 * config/i386/i386-expand.c
3443 (ix86_expand_args_builtin): Handle new builtin types.
3444 (ix86_expand_round_builtin): Ditto.
3445 * config/i386/i386.md (ssevecmode): Add HF mode.
3446 (MODEFH): New mode iterator.
3447 * config/i386/sse.md
3448 (V48H_AVX512VL): New mode iterator to support HF vector modes.
3449 Ajdust corresponding description.
3450 (ssecmpintprefix): New.
3451 (VI12_AVX512VL): Adjust to support HF vector modes.
3452 (cmp_imm_predicate): Likewise.
3453 (<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
3455 (avx512f_vmcmp<mode>3<round_saeonly_name>): Likewise.
3456 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Likewise.
3457 (<sse>_<unord>comi<round_saeonly_name>): Likewise.
3459 2021-09-10 liuhongt <hongtao.liu@intel.com>
3461 * config/i386/avx512fp16intrin.h: (_mm512_max_ph): New intrinsic.
3462 (_mm512_mask_max_ph): Likewise.
3463 (_mm512_maskz_max_ph): Likewise.
3464 (_mm512_min_ph): Likewise.
3465 (_mm512_mask_min_ph): Likewise.
3466 (_mm512_maskz_min_ph): Likewise.
3467 (_mm512_max_round_ph): Likewise.
3468 (_mm512_mask_max_round_ph): Likewise.
3469 (_mm512_maskz_max_round_ph): Likewise.
3470 (_mm512_min_round_ph): Likewise.
3471 (_mm512_mask_min_round_ph): Likewise.
3472 (_mm512_maskz_min_round_ph): Likewise.
3473 (_mm_max_sh): Likewise.
3474 (_mm_mask_max_sh): Likewise.
3475 (_mm_maskz_max_sh): Likewise.
3476 (_mm_min_sh): Likewise.
3477 (_mm_mask_min_sh): Likewise.
3478 (_mm_maskz_min_sh): Likewise.
3479 (_mm_max_round_sh): Likewise.
3480 (_mm_mask_max_round_sh): Likewise.
3481 (_mm_maskz_max_round_sh): Likewise.
3482 (_mm_min_round_sh): Likewise.
3483 (_mm_mask_min_round_sh): Likewise.
3484 (_mm_maskz_min_round_sh): Likewise.
3485 * config/i386/avx512fp16vlintrin.h (_mm_max_ph): New intrinsic.
3486 (_mm256_max_ph): Likewise.
3487 (_mm_mask_max_ph): Likewise.
3488 (_mm256_mask_max_ph): Likewise.
3489 (_mm_maskz_max_ph): Likewise.
3490 (_mm256_maskz_max_ph): Likewise.
3491 (_mm_min_ph): Likewise.
3492 (_mm256_min_ph): Likewise.
3493 (_mm_mask_min_ph): Likewise.
3494 (_mm256_mask_min_ph): Likewise.
3495 (_mm_maskz_min_ph): Likewise.
3496 (_mm256_maskz_min_ph): Likewise.
3497 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3498 * config/i386/i386-builtin.def: Add corresponding new builtins.
3499 * config/i386/i386-expand.c
3500 (ix86_expand_args_builtin): Handle new builtin types.
3501 * config/i386/sse.md
3502 (<code><mode>3<mask_name><round_saeonly_name>): Adjust to
3503 support HF vector modes.
3504 (*<code><mode>3<mask_name><round_saeonly_name>): Likewise.
3505 (ieee_<ieee_maxmin><mode>3<mask_name><round_saeonly_name>):
3507 (<sse>_vm<code><mode>3<mask_scalar_name><round_saeonly_scalar_name>):
3509 * config/i386/subst.md (round_saeonly_mode512bit_condition):
3510 Adjust for HF vector modes.
3512 2021-09-10 Liu, Hongtao <hongtao.liu@intel.com>
3514 * config/i386/avx512fp16intrin.h (_mm_add_sh): New intrinsic.
3515 (_mm_mask_add_sh): Likewise.
3516 (_mm_maskz_add_sh): Likewise.
3517 (_mm_sub_sh): Likewise.
3518 (_mm_mask_sub_sh): Likewise.
3519 (_mm_maskz_sub_sh): Likewise.
3520 (_mm_mul_sh): Likewise.
3521 (_mm_mask_mul_sh): Likewise.
3522 (_mm_maskz_mul_sh): Likewise.
3523 (_mm_div_sh): Likewise.
3524 (_mm_mask_div_sh): Likewise.
3525 (_mm_maskz_div_sh): Likewise.
3526 (_mm_add_round_sh): Likewise.
3527 (_mm_mask_add_round_sh): Likewise.
3528 (_mm_maskz_add_round_sh): Likewise.
3529 (_mm_sub_round_sh): Likewise.
3530 (_mm_mask_sub_round_sh): Likewise.
3531 (_mm_maskz_sub_round_sh): Likewise.
3532 (_mm_mul_round_sh): Likewise.
3533 (_mm_mask_mul_round_sh): Likewise.
3534 (_mm_maskz_mul_round_sh): Likewise.
3535 (_mm_div_round_sh): Likewise.
3536 (_mm_mask_div_round_sh): Likewise.
3537 (_mm_maskz_div_round_sh): Likewise.
3538 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3539 * config/i386/i386-builtin.def: Add corresponding new builtins.
3540 * config/i386/i386-expand.c
3541 (ix86_expand_round_builtin): Handle new builtins.
3542 * config/i386/sse.md (VF_128): Change description.
3543 (<sse>_vm<plusminus_insn><mode>3<mask_scalar_name><round_scalar_name>):
3544 Adjust to support HF vector modes.
3545 (<sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name><round_scalar_name>):
3548 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
3550 * config/i386/i386-expand.c
3551 (ix86_avx256_split_vector_move_misalign): Handle V16HF mode.
3552 * config/i386/i386.c
3553 (ix86_preferred_simd_mode): Handle HF mode.
3554 * config/i386/sse.md (V_256H): New mode iterator.
3555 (avx_vextractf128<mode>): Use it.
3556 (VEC_INIT_MODE): Align vector HFmode condition to vector
3557 HImodes since there're no real HF instruction used.
3558 (VEC_INIT_HALF_MODE): Ditto.
3560 (VIHF_AVX512BW): Ditto.
3561 (*vec_extracthf): Ditto.
3562 (VEC_EXTRACT_MODE): Ditto.
3564 2021-09-10 Richard Biener <rguenther@suse.de>
3567 * config/dbx.h: Remove.
3568 * config/dbxcoff.h: Do not define PREFERRED_DEBUGGING_TYPE.
3569 * config/lynx.h: Likewise.
3571 2021-09-10 liuhongt <hongtao.liu@intel.com>
3573 * config/i386/i386-expand.c (ix86_expand_copysign): Expand
3574 right into ANDNOT + AND + IOR, using paradoxical subregs.
3575 (ix86_split_copysign_const): Remove.
3576 (ix86_split_copysign_var): Ditto.
3577 * config/i386/i386-protos.h (ix86_split_copysign_const): Dotto.
3578 (ix86_split_copysign_var): Ditto.
3579 * config/i386/i386.md (@copysign<mode>3_const): Ditto.
3580 (@copysign<mode>3_var): Ditto.
3582 2021-09-09 qing zhao <qing.zhao@oracle.com>
3584 * builtins.c (expand_builtin_memset): Make external visible.
3585 * builtins.h (expand_builtin_memset): Declare extern.
3586 * common.opt (ftrivial-auto-var-init=): New option.
3587 * doc/extend.texi: Document the uninitialized attribute.
3588 * doc/invoke.texi: Document -ftrivial-auto-var-init.
3589 * flag-types.h (enum auto_init_type): New enumerated type
3591 * gimple-fold.c (clear_padding_type): Add one new parameter.
3592 (clear_padding_union): Likewise.
3593 (clear_padding_emit_loop): Likewise.
3594 (clear_type_padding_in_mask): Likewise.
3595 (gimple_fold_builtin_clear_padding): Handle this new parameter.
3596 * gimplify.c (gimple_add_init_for_auto_var): New function.
3597 (gimple_add_padding_init_for_auto_var): New function.
3598 (is_var_need_auto_init): New function.
3599 (gimplify_decl_expr): Add initialization to automatic variables per
3601 (gimplify_call_expr): Add one new parameter for call to
3602 __builtin_clear_padding.
3603 (gimplify_init_constructor): Add padding initialization in the end.
3604 * internal-fn.c (INIT_PATTERN_VALUE): New macro.
3605 (expand_DEFERRED_INIT): New function.
3606 * internal-fn.def (DEFERRED_INIT): New internal function.
3607 * tree-cfg.c (verify_gimple_call): Verify calls to .DEFERRED_INIT.
3608 * tree-sra.c (generate_subtree_deferred_init): New function.
3609 (scan_function): Avoid setting cannot_scalarize_away_bitmap for
3610 calls to .DEFERRED_INIT.
3611 (sra_modify_deferred_init): New function.
3612 (sra_modify_function_body): Handle calls to DEFERRED_INIT specially.
3613 * tree-ssa-structalias.c (find_func_aliases_for_call): Likewise.
3614 * tree-ssa-uninit.c (warn_uninit): Handle calls to DEFERRED_INIT
3616 (check_defs): Likewise.
3617 (warn_uninitialized_vars): Likewise.
3618 * tree-ssa.c (ssa_undefined_value_p): Likewise.
3619 * tree.c (build_common_builtin_nodes): Build tree node for
3620 BUILT_IN_CLEAR_PADDING when needed.
3622 2021-09-09 Richard Biener <rguenther@suse.de>
3624 * tree-ssa-loop-im.c (fill_always_executed_in_1): Walk
3627 2021-09-09 Richard Biener <rguenther@suse.de>
3629 * tree-ssa-loop-im.c (fill_always_executed_in_1): Integrate
3630 DOM walk from get_loop_body_in_dom_order using a worklist
3633 2021-09-09 liuhongt <hongtao.liu@intel.com>
3635 * config.gcc: Add avx512fp16vlintrin.h.
3636 * config/i386/avx512fp16intrin.h: (_mm512_add_ph): New intrinsic.
3637 (_mm512_mask_add_ph): Likewise.
3638 (_mm512_maskz_add_ph): Likewise.
3639 (_mm512_sub_ph): Likewise.
3640 (_mm512_mask_sub_ph): Likewise.
3641 (_mm512_maskz_sub_ph): Likewise.
3642 (_mm512_mul_ph): Likewise.
3643 (_mm512_mask_mul_ph): Likewise.
3644 (_mm512_maskz_mul_ph): Likewise.
3645 (_mm512_div_ph): Likewise.
3646 (_mm512_mask_div_ph): Likewise.
3647 (_mm512_maskz_div_ph): Likewise.
3648 (_mm512_add_round_ph): Likewise.
3649 (_mm512_mask_add_round_ph): Likewise.
3650 (_mm512_maskz_add_round_ph): Likewise.
3651 (_mm512_sub_round_ph): Likewise.
3652 (_mm512_mask_sub_round_ph): Likewise.
3653 (_mm512_maskz_sub_round_ph): Likewise.
3654 (_mm512_mul_round_ph): Likewise.
3655 (_mm512_mask_mul_round_ph): Likewise.
3656 (_mm512_maskz_mul_round_ph): Likewise.
3657 (_mm512_div_round_ph): Likewise.
3658 (_mm512_mask_div_round_ph): Likewise.
3659 (_mm512_maskz_div_round_ph): Likewise.
3660 * config/i386/avx512fp16vlintrin.h: New header.
3661 * config/i386/i386-builtin-types.def (V16HF, V8HF, V32HF):
3662 Add new builtin types.
3663 * config/i386/i386-builtin.def: Add corresponding builtins.
3664 * config/i386/i386-expand.c
3665 (ix86_expand_args_builtin): Handle new builtin types.
3666 (ix86_expand_round_builtin): Likewise.
3667 * config/i386/immintrin.h: Include avx512fp16vlintrin.h
3668 * config/i386/sse.md (VFH): New mode_iterator.
3670 (avx512fmaskmode): Add HF vector modes.
3671 (avx512fmaskhalfmode): Likewise.
3672 (<plusminus_insn><mode>3<mask_name><round_name>): Adjust to for
3674 (*<plusminus_insn><mode>3<mask_name><round_name>): Likewise.
3675 (mul<mode>3<mask_name><round_name>): Likewise.
3676 (*mul<mode>3<mask_name><round_name>): Likewise.
3677 (div<mode>3): Likewise.
3678 (<sse>_div<mode>3<mask_name><round_name>): Likewise.
3679 * config/i386/subst.md (SUBST_V): Add HF vector modes.
3680 (SUBST_A): Likewise.
3681 (round_mode512bit_condition): Adjust for V32HFmode.
3683 2021-09-09 liuhongt <hongtao.liu@intel.com>
3686 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
3687 (reduc_plus_scal_v4sf): .. this, New define_expand.
3688 (reduc_plus_scal_v2df): .. and this, New define_expand.
3690 2021-09-09 liuhongt <hongtao.liu@intel.com>
3693 * config/i386/sse.md (*vec_extract<mode><ssescalarmodelower>_valign):
3696 2021-09-08 Jonathan Wakely <jwakely@redhat.com>
3699 * doc/trouble.texi (Copy Assignment): Fix description of
3700 behaviour and fix code in example.
3702 2021-09-08 Segher Boessenkool <segher@kernel.crashing.org>
3705 * config/rs6000/rs6000-logue.c (rs6000_emit_epilogue): For ELFv2 use
3706 r11 instead of r12 for restoring CR.
3708 2021-09-08 Jakub Jelinek <jakub@redhat.com>
3709 liuhongt <hongtao.liu@intel.com>
3712 * config/i386/i386.md (@xorsign<mode>3_1): Remove.
3713 * config/i386/i386-expand.c (ix86_expand_xorsign): Expand right away
3714 into AND with mask and XOR, using paradoxical subregs.
3715 (ix86_split_xorsign): Remove.
3716 * config/i386/i386-protos.h (ix86_split_xorsign): Remove.
3718 2021-09-08 Di Zhao <dizhao@os.amperecomputing.com>
3720 * tree-ssa-sccvn.c (vn_nary_op_insert_into): fix result compare
3722 2021-09-08 Jakub Jelinek <jakub@redhat.com>
3725 * config/i386/i386.md (xorsign<mode>3): If operands[1] is equal to
3726 operands[2], emit abs<mode>2 instead.
3727 (@xorsign<mode>3_1): Add early-clobbers for output operand, enable
3728 first alternative even for avx, add another alternative with
3729 =&Yv <- 0, Yv, Yvm constraints.
3730 * config/i386/i386-expand.c (ix86_split_xorsign): If op0 is equal
3731 to op1, emit vpandn instead.
3733 2021-09-08 liuhongt <hongtao.liu@intel.com>
3735 * config/i386/avx512fp16intrin.h (_mm_set_ph): New intrinsic.
3736 (_mm256_set_ph): Likewise.
3737 (_mm512_set_ph): Likewise.
3738 (_mm_setr_ph): Likewise.
3739 (_mm256_setr_ph): Likewise.
3740 (_mm512_setr_ph): Likewise.
3741 (_mm_set1_ph): Likewise.
3742 (_mm256_set1_ph): Likewise.
3743 (_mm512_set1_ph): Likewise.
3744 (_mm_setzero_ph): Likewise.
3745 (_mm256_setzero_ph): Likewise.
3746 (_mm512_setzero_ph): Likewise.
3747 (_mm_set_sh): Likewise.
3748 (_mm_load_sh): Likewise.
3749 (_mm_store_sh): Likewise.
3750 * config/i386/i386-builtin-types.def (V8HF): New type.
3751 (DEF_FUNCTION_TYPE (V8HF, V8HI)): New builtin function type
3752 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
3753 Support vector HFmodes.
3754 (ix86_expand_vector_init_one_nonzero): Likewise.
3755 (ix86_expand_vector_init_one_var): Likewise.
3756 (ix86_expand_vector_init_interleave): Likewise.
3757 (ix86_expand_vector_init_general): Likewise.
3758 (ix86_expand_vector_set): Likewise.
3759 (ix86_expand_vector_extract): Likewise.
3760 (ix86_expand_vector_init_concat): Likewise.
3761 (ix86_expand_sse_movcc): Handle vector HFmodes.
3762 (ix86_expand_vector_set_var): Ditto.
3763 * config/i386/i386-modes.def: Add HF vector modes in comment.
3764 * config/i386/i386.c (classify_argument): Add HF vector modes.
3765 (ix86_hard_regno_mode_ok): Allow HF vector modes for AVX512FP16.
3766 (ix86_vector_mode_supported_p): Likewise.
3767 (ix86_set_reg_reg_cost): Handle vector HFmode.
3768 (ix86_get_ssemov): Handle vector HFmode.
3769 (function_arg_advance_64): Pass unamed V16HFmode and V32HFmode
3771 (function_arg_advance_32): Pass V8HF/V16HF/V32HF by sse reg for 32bit
3773 (function_arg_advance_32): Ditto.
3774 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): New.
3775 (VALID_AVX256_REG_OR_OI_MODE): Rename to ..
3776 (VALID_AVX256_REG_OR_OI_VHF_MODE): .. this, and add V16HF.
3777 (VALID_SSE2_REG_VHF_MODE): New.
3778 (VALID_AVX512VL_128_REG_MODE): Add V8HF and TImode.
3779 (SSE_REG_MODE_P): Add vector HFmode.
3780 * config/i386/i386.md (mode): Add HF vector modes.
3781 (MODE_SIZE): Likewise.
3782 (ssemodesuffix): Add ph suffix for HF vector modes.
3783 * config/i386/sse.md (VFH_128): New mode iterator.
3784 (VMOVE): Adjust for HF vector modes.
3786 (V_256_512): Likewise.
3788 (avx512fmaskmode): Likewise.
3789 (shuffletype): Likewise.
3790 (sseinsnmode): Likewise.
3791 (ssedoublevecmode): Likewise.
3792 (ssehalfvecmode): Likewise.
3793 (ssehalfvecmodelower): Likewise.
3794 (ssePScmode): Likewise.
3795 (ssescalarmode): Likewise.
3796 (ssescalarmodelower): Likewise.
3797 (sseintprefix): Likewise.
3799 (bcstscalarsuff): Likewise.
3800 (xtg_mode): Likewise.
3801 (VI12HF_AVX512VL): New mode_iterator.
3802 (VF_AVX512FP16): Likewise.
3804 (VIHF_256): Likewise.
3805 (VIHF_AVX512BW): Likewise.
3806 (V16_256): Likewise.
3807 (V32_512): Likewise.
3808 (sseintmodesuffix): New mode_attr.
3809 (sse): Add scalar and vector HFmodes.
3810 (ssescalarmode): Add vector HFmode mapping.
3811 (ssescalarmodesuffix): Add sh suffix for HFmode.
3812 (*<sse>_vm<insn><mode>3): Use VFH_128.
3813 (*<sse>_vm<multdiv_mnemonic><mode>3): Likewise.
3814 (*ieee_<ieee_maxmin><mode>3): Likewise.
3815 (<avx512>_blendm<mode>): New define_insn.
3816 (vec_setv8hf): New define_expand.
3817 (vec_set<mode>_0): New define_insn for HF vector set.
3818 (*avx512fp16_movsh): Likewise.
3819 (avx512fp16_movsh): Likewise.
3820 (vec_extract_lo_v32hi): Rename to ...
3821 (vec_extract_lo_<mode>): ... this, and adjust to allow HF
3823 (vec_extract_hi_v32hi): Likewise.
3824 (vec_extract_hi_<mode>): Likewise.
3825 (vec_extract_lo_v16hi): Likewise.
3826 (vec_extract_lo_<mode>): Likewise.
3827 (vec_extract_hi_v16hi): Likewise.
3828 (vec_extract_hi_<mode>): Likewise.
3829 (vec_set_hi_v16hi): Likewise.
3830 (vec_set_hi_<mode>): Likewise.
3831 (vec_set_lo_v16hi): Likewise.
3832 (vec_set_lo_<mode>): Likewise.
3833 (*vec_extract<mode>_0): New define_insn_and_split for HF
3835 (*vec_extracthf): New define_insn.
3836 (VEC_EXTRACT_MODE): Add HF vector modes.
3837 (PINSR_MODE): Add V8HF.
3838 (sse2p4_1): Likewise.
3839 (pinsr_evex_isa): Likewise.
3840 (<sse2p4_1>_pinsr<ssemodesuffix>): Adjust to support
3841 insert for V8HFmode.
3842 (pbroadcast_evex_isa): Add HF vector modes.
3843 (AVX2_VEC_DUP_MODE): Likewise.
3844 (VEC_INIT_MODE): Likewise.
3845 (VEC_INIT_HALF_MODE): Likewise.
3846 (avx2_pbroadcast<mode>): Adjust to support HF vector mode
3848 (avx2_pbroadcast<mode>_1): Likewise.
3849 (<avx512>_vec_dup<mode>_1): Likewise.
3850 (<avx512>_vec_dup<mode><mask_name>): Likewise.
3851 (<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>):
3854 2021-09-08 Guo, Xuepeng <xuepeng.guo@intel.com>
3855 H.J. Lu <hongjiu.lu@intel.com>
3856 Liu Hongtao <hongtao.liu@intel.com>
3857 Wang Hongyu <hongyu.wang@intel.com>
3858 Xu Dianhong <dianhong.xu@intel.com>
3860 * common/config/i386/cpuinfo.h (get_available_features):
3861 Detect FEATURE_AVX512FP16.
3862 * common/config/i386/i386-common.c
3863 (OPTION_MASK_ISA_AVX512FP16_SET,
3864 OPTION_MASK_ISA_AVX512FP16_UNSET,
3865 OPTION_MASK_ISA2_AVX512FP16_SET,
3866 OPTION_MASK_ISA2_AVX512FP16_UNSET): New.
3867 (OPTION_MASK_ISA2_AVX512BW_UNSET,
3868 OPTION_MASK_ISA2_AVX512BF16_UNSET): Add AVX512FP16.
3869 (ix86_handle_option): Handle -mavx512fp16.
3870 * common/config/i386/i386-cpuinfo.h (enum processor_features):
3871 Add FEATURE_AVX512FP16.
3872 * common/config/i386/i386-isas.h: Add entry for AVX512FP16.
3873 * config.gcc: Add avx512fp16intrin.h.
3874 * config/i386/avx512fp16intrin.h: New intrinsic header.
3875 * config/i386/cpuid.h: Add bit_AVX512FP16.
3876 * config/i386/i386-builtin-types.def: (FLOAT16): New primitive type.
3877 * config/i386/i386-builtins.c: Support _Float16 type for i386
3879 (ix86_register_float16_builtin_type): New function.
3880 (ix86_float16_type_node): New.
3881 * config/i386/i386-c.c (ix86_target_macros_internal): Define
3883 * config/i386/i386-expand.c (ix86_expand_branch): Support
3885 (ix86_prepare_fp_compare_args): Adjust TARGET_SSE_MATH &&
3886 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
3887 (ix86_expand_fp_movcc): Ditto.
3888 * config/i386/i386-isa.def: Add PTA define for AVX512FP16.
3889 * config/i386/i386-options.c (isa2_opts): Add -mavx512fp16.
3890 (ix86_valid_target_attribute_inner_p): Add avx512fp16 attribute.
3891 * config/i386/i386.c (ix86_get_ssemov): Use
3892 vmovdqu16/vmovw/vmovsh for HFmode/HImode scalar or vector.
3893 (ix86_get_excess_precision): Use
3894 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when TARGET_AVX512FP16
3896 (sse_store_index): Use SFmode cost for HFmode cost.
3897 (inline_memory_move_cost): Add HFmode, and perfer SSE cost over
3898 GPR cost for HFmode.
3899 (ix86_hard_regno_mode_ok): Allow HImode in sse register.
3900 (ix86_mangle_type): Add manlging for _Float16 type.
3901 (inline_secondary_memory_needed): No memory is needed for
3902 16bit movement between gpr and sse reg under
3904 (ix86_multiplication_cost): Adjust TARGET_SSE_MATH &&
3905 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
3906 (ix86_division_cost): Ditto.
3907 (ix86_rtx_costs): Ditto.
3908 (ix86_add_stmt_cost): Ditto.
3909 (ix86_optab_supported_p): Ditto.
3910 * config/i386/i386.h (VALID_AVX512F_SCALAR_MODE): Add HFmode.
3911 (SSE_FLOAT_MODE_SSEMATH_OR_HF_P): Add HFmode.
3912 (PTA_SAPPHIRERAPIDS): Add PTA_AVX512FP16.
3913 * config/i386/i386.md (mode): Add HFmode.
3914 (MODE_SIZE): Add HFmode.
3915 (isa): Add avx512fp16.
3916 (enabled): Handle avx512fp16.
3917 (ssemodesuffix): Add sh suffix for HFmode.
3918 (comm): Add mult, div.
3919 (plusminusmultdiv): New code iterator.
3920 (insn): Add mult, div.
3921 (*movhf_internal): Adjust for avx512fp16 instruction.
3922 (*movhi_internal): Ditto.
3923 (*cmpi<unord>hf): New define_insn for HFmode.
3924 (*ieee_s<ieee_maxmin>hf3): Likewise.
3925 (extendhf<mode>2): Likewise.
3926 (trunc<mode>hf2): Likewise.
3927 (float<floatunssuffix><mode>hf2): Likewise.
3928 (*<insn>hf): Likewise.
3929 (cbranchhf4): New expander.
3930 (movhfcc): Likewise.
3931 (<insn>hf3): Likewise.
3934 * config/i386/i386.opt: Add mavx512fp16.
3935 * config/i386/immintrin.h: Include avx512fp16intrin.h.
3936 * doc/invoke.texi: Add mavx512fp16.
3937 * doc/extend.texi: Add avx512fp16 Usage Notes.
3939 2021-09-08 liuhongt <hongtao.liu@intel.com>
3941 * common.opt: Support -fexcess-precision=16.
3942 * config/aarch64/aarch64.c (aarch64_excess_precision): Return
3943 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when
3944 EXCESS_PRECISION_TYPE_FLOAT16.
3945 * config/arm/arm.c (arm_excess_precision): Ditto.
3946 * config/i386/i386.c (ix86_get_excess_precision): Ditto.
3947 * config/m68k/m68k.c (m68k_excess_precision): Issue an error
3948 when EXCESS_PRECISION_TYPE_FLOAT16.
3949 * config/s390/s390.c (s390_excess_precision): Ditto.
3950 * coretypes.h (enum excess_precision_type): Add
3951 EXCESS_PRECISION_TYPE_FLOAT16.
3952 * doc/tm.texi (TARGET_C_EXCESS_PRECISION): Update documents.
3953 * doc/tm.texi.in (TARGET_C_EXCESS_PRECISION): Ditto.
3954 * doc/extend.texi (Half-Precision): Document
3955 -fexcess-precision=16.
3956 * flag-types.h (enum excess_precision): Add
3957 EXCESS_PRECISION_FLOAT16.
3958 * target.def (excess_precision): Update document.
3959 * tree.c (excess_precision_type): Set excess_precision_type to
3960 EXCESS_PRECISION_FLOAT16 when -fexcess-precision=16.
3962 2021-09-08 liuhongt <hongtao.liu@intel.com>
3964 * doc/extend.texi: (@node Floating Types): Adjust the wording.
3965 (@node Half-Precision): Ditto.
3967 2021-09-07 Takayuki 'January June' Suwa <jjsuwa_sys3175@yahoo.co.jp>
3970 * config/xtensa/xtensa.c (xtensa_emit_move_sequence): Add
3971 'CONST_INT_P (src)' to the condition of the block that tries to
3972 eliminate literal when loading integer contant.
3974 2021-09-07 David Faust <david.faust@oracle.com>
3976 * doc/extend.texi (BPF Type Attributes) New node.
3977 Document new preserve_access_index attribute.
3978 Document new preserve_access_index builtin.
3979 * doc/invoke.texi: Document -mco-re and -mno-co-re options.
3981 2021-09-07 David Faust <david.faust@oracle.com>
3983 * config/bpf/bpf.c: Adjust includes.
3984 (bpf_handle_preserve_access_index_attribute): New function.
3985 (bpf_attribute_table): Use it here.
3986 (bpf_builtins): Add BPF_BUILTIN_PRESERVE_ACCESS_INDEX.
3987 (bpf_option_override): Handle "-mco-re" option.
3988 (bpf_asm_init_sections): New.
3989 (TARGET_ASM_INIT_SECTIONS): Redefine.
3990 (bpf_file_end): New.
3991 (TARGET_ASM_FILE_END): Redefine.
3992 (bpf_init_builtins): Add "__builtin_preserve_access_index".
3993 (bpf_core_compute, bpf_core_get_index): New.
3994 (is_attr_preserve_access): New.
3995 (bpf_expand_builtin): Handle new builtins.
3996 (bpf_core_newdecl, bpf_core_is_maybe_aggregate_access): New.
3997 (bpf_core_walk): New.
3998 (bpf_resolve_overloaded_builtin): New.
3999 (TARGET_RESOLVE_OVERLOADED_BUILTIN): Redefine.
4001 (pass_bpf_core_attr): New RTL pass.
4002 * config/bpf/bpf-passes.def: New file.
4003 * config/bpf/bpf-protos.h (make_pass_bpf_core_attr): New.
4004 * config/bpf/coreout.c: New file.
4005 * config/bpf/coreout.h: Likewise.
4006 * config/bpf/t-bpf (TM_H): Add $(srcdir)/config/bpf/coreout.h.
4007 (coreout.o): New rule.
4008 (PASSES_EXTRA): Add $(srcdir)/config/bpf/bpf-passes.def.
4009 * config.gcc (bpf): Add coreout.h to extra_headers.
4010 Add coreout.o to extra_objs.
4011 Add $(srcdir)/config/bpf/coreout.c to target_gtfiles.
4013 2021-09-07 David Faust <david.faust@oracle.com>
4015 * btfout.c (get_btf_id): Function is no longer static.
4016 * ctfc.h: Expose it here.
4018 2021-09-07 David Faust <david.faust@oracle.com>
4020 * ctfc.c (ctf_lookup_tree_type): New function.
4023 2021-09-07 David Faust <david.faust@oracle.com>
4025 * ctfc.c (ctf_dtd_lookup): Function is no longer static.
4026 * ctfc.h: Analogous change.
4028 2021-09-07 David Faust <david.faust@oracle.com>
4030 * dwarf2out.c (lookup_type_die): Function is no longer static.
4031 * dwarf2out.h: Expose it here.
4033 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
4035 * dwarf2ctf.c (ctf_debug_finalize): Make it static.
4036 (ctf_debug_early_finish): New definition.
4037 (ctf_debug_finish): Likewise.
4038 * dwarf2ctf.h (ctf_debug_finalize): Remove declaration.
4039 (ctf_debug_early_finish): New declaration.
4040 (ctf_debug_finish): Likewise.
4041 * dwarf2out.c (dwarf2out_finish): Invoke ctf_debug_finish.
4042 (dwarf2out_early_finish): Invoke ctf_debug_early_finish.
4044 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
4046 * config/bpf/bpf.c (bpf_option_override): For BPF backend, disable LTO
4047 support when compiling for CO-RE.
4048 * config/bpf/bpf.opt: Add new command line option -mco-re.
4050 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
4052 * flag-types.h (enum debug_info_type): Add new enum
4053 DINFO_TYPE_BTF_WITH_CORE.
4054 (BTF_WITH_CORE_DEBUG): New bitmask.
4055 * flags.h (btf_with_core_debuginfo_p): New declaration.
4056 * opts.c (btf_with_core_debuginfo_p): New definition.
4058 2021-09-07 Jason Merrill <jason@redhat.com>
4060 * tree.h (error_operand_p): Change to inline function.
4062 2021-09-07 Aldy Hernandez <aldyh@redhat.com>
4064 * tree-ssa-threadedge.c (forwarder_block_p): Rename to...
4065 (empty_block_with_phis_p): ...this.
4066 (potentially_threadable_block): Same.
4067 (jump_threader::thread_through_normal_block): Same.
4069 2021-09-07 Eric Botcazou <ebotcazou@adacore.com>
4072 * dwarf2out.c (mark_base_types): New overloaded function.
4073 (dwarf2out_early_finish): Invoke it on the COMDAT type list as well
4074 as the compilation unit, and call move_marked_base_types afterward.
4076 2021-09-07 H.J. Lu <hjl.tools@gmail.com>
4079 * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
4081 (ix86_expand_vector_convert_uns_vsivsf): Likewise.
4083 2021-09-07 Richard Biener <rguenther@suse.de>
4085 PR tree-optimization/102226
4086 * tree-vect-loop.c (vect_transform_cycle_phi): Record
4087 the converted value for the epilogue PHI use.
4089 2021-09-07 Martin Liska <mliska@suse.cz>
4091 PR gcov-profile/80223
4092 * ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
4093 options, do not inline when no_profile_instrument_function
4094 attributes are different in early inliner. It's fine to inline
4095 it after PGO instrumentation.
4097 2021-09-07 Richard Biener <rguenther@suse.de>
4099 PR tree-optimization/101555
4100 * tree-ssa-pre.c (translate_vuse_through_block): Do not
4101 perform an alias walk to determine the validity of the
4102 mem at the start of the block which is already guaranteed
4103 by means of prune_clobbered_mems.
4104 (phi_translate_1): Pass edge to translate_vuse_through_block.
4106 2021-09-07 Xionghu Luo <luoxhu@linux.ibm.com>
4109 * config/rs6000/rs6000.md (fmod<mode>3): New define_expand.
4110 (remainder<mode>3): Likewise.
4112 2021-09-07 YunQiang Su <yunqiang.su@cipunited.com>
4114 * config/mips/mips.c (mips_file_start): add .module for
4117 2021-09-06 Roger Sayle <roger@nextmovesoftware.com>
4119 * wide-int.cc (wi::clz): Reorder tests to ensure the result
4120 is zero for all negative values.
4122 2021-09-06 Tobias Burnus <tobias@codesourcery.com>
4124 * doc/invoke.texi (-foffload-options): Fix @opindex.
4126 2021-09-06 H.J. Lu <hjl.tools@gmail.com>
4129 * config/i386/i386-expand.c (ix86_split_xorsign): Use operands[2].
4130 * config/i386/i386.md (@xorsign<mode>3_1): Add non-destructive
4131 source alternative for AVX.
4133 2021-09-06 liuhongt <hongtao.liu@intel.com>
4135 PR middle-end/102182
4136 * optabs.c (expand_fix): Add from1 to avoid from being
4139 2021-09-06 Eric Botcazou <ebotcazou@adacore.com>
4141 * dwarf2out.c (modified_type_die): Deal with all array types earlier
4142 and use local variable consistently throughout the function.
4144 2021-09-06 Jakub Jelinek <jakub@redhat.com>
4146 PR tree-optimization/102207
4147 * match.pd: Don't demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
4148 were promoted from signed to wider unsigned type.
4150 2021-09-06 Andrew Pinski <apinski@marvell.com>
4152 PR tree-optimization/63184
4153 * match.pd: Add simplification of pointer_diff of two pointer_plus
4154 with addr_expr in the first operand of each pointer_plus.
4155 Add simplificatoin of ne/eq of two pointer_plus with addr_expr
4156 in the first operand of each pointer_plus.
4158 2021-09-06 Richard Biener <rguenther@suse.de>
4160 PR tree-optimization/102176
4161 * tree-vect-slp.c (vect_slp_gather_vectorized_scalar_stmts):
4163 (vect_bb_slp_scalar_cost): Use the computed set of
4164 vectorized scalar stmts instead of relying on the out-of-date
4165 and not accurate PURE_SLP_STMT.
4166 (vect_bb_vectorization_profitable_p): Compute the set
4167 of vectorized scalar stmts.
4169 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
4171 * gimple-range-path.cc (path_range_query::range_of_stmt): Remove
4172 GIMPLE_COND special casing.
4173 (path_range_query::range_defined_in_block): Use range_of_stmt
4174 instead of calling fold_range directly.
4176 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
4178 * gimple-range-path.cc (path_range_query::range_of_expr): Set
4179 m_undefined_path when appropriate.
4180 (path_range_query::internal_range_of_expr): Copy from range_of_expr.
4181 (path_range_query::unreachable_path_p): New.
4182 (path_range_query::precompute_ranges): Set m_undefined_path.
4183 * gimple-range-path.h (path_range_query::unreachable_path_p): New.
4184 (path_range_query::internal_range_of_expr): New.
4185 * tree-ssa-threadbackward.c (back_threader::find_taken_edge_cond):
4186 Use unreachable_path_p.
4188 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
4190 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
4191 Remove argument and call find_taken_edge.
4192 (back_threader::resolve_phi): Do not calculate taken edge before
4193 calling maybe_register_path.
4194 (back_threader::find_paths_to_names): Same.
4196 2021-09-05 Jeff Law <jlaw@localhost.localdomain>
4198 * config/h8300/h8300.md (QHSI2 mode iterator): New mode iterator.
4199 * config/h8300/testcompare.md (store_c): Update name, use new
4201 (store_neg_c, store_shifted_c): New patterns.
4203 2021-09-03 Segher Boessenkool <segher@kernel.crashing.org>
4206 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): On ELFv2 use r11
4207 instead of r12 for CR save, in all cases.
4209 2021-09-03 Andrew Pinski <apinski@marvell.com>
4211 * config/aarch64/aarch64-sve-builtins.cc (register_vector_type):
4212 Handle error_mark_node as the type of the type_decl.
4214 2021-09-03 Andrew Pinski <apinski@marvell.com>
4216 * config/aarch64/aarch64-builtins.c (struct aarch64_simd_type_info):
4218 (aarch64_simd_types): Likewise.
4219 (aarch64_simd_intOI_type_node): Likewise.
4220 (aarch64_simd_intCI_type_node): Likewise.
4221 (aarch64_simd_intXI_type_node): Likewise.
4222 * config/aarch64/aarch64.h (aarch64_fp16_type_node): Likewise.
4223 (aarch64_fp16_ptr_type_node): Likewise.
4224 (aarch64_bf16_type_node): Likewise.
4225 (aarch64_bf16_ptr_type_node): Likewise.
4227 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4229 * range-op.cc (operator_minus::op1_op2_relation_effect): Abstract
4231 (minus_op1_op2_relation_effect): ...here.
4232 (class operator_pointer_diff): New.
4233 (operator_pointer_diff::op1_op2_relation_effect): Call
4234 minus_op1_op2_relation_effect.
4235 (integral_table::integral_table): Add entry for POINTER_DIFF_EXPR.
4237 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4239 * tree-ssa-threadbackward.c (back_threader::thread_through_all_blocks):
4240 Add may_peel_loop_headers.
4241 (back_threader_registry::thread_through_all_blocks): Same.
4242 (try_thread_blocks): Pass may_peel_loop_headers argument.
4243 (pass_early_thread_jumps::execute): Same.
4245 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4247 * tree-ssa-threadedge.c (has_phis_p): New.
4248 (forwarder_block_p): New.
4249 (potentially_threadable_block): Call forwarder_block_p.
4250 (jump_threader::thread_around_empty_blocks): Call has_phis_p.
4251 (jump_threader::thread_through_normal_block): Call
4254 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4256 * tree-ssa-threadbackward.c (back_threader::dump): New.
4257 (back_threader::debug): New.
4258 (back_threader_profitability::profitable_path_p): Dump blocks
4259 even if we are bailing early.
4261 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4263 * tree-ssa-threadupdate.c (cancel_thread): New.
4264 (jump_thread_path_registry::thread_block_1): Use cancel_thread.
4265 (jump_thread_path_registry::mark_threaded_blocks): Same.
4266 (jump_thread_path_registry::register_jump_thread): Same.
4268 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4270 * tree-ssa-threadedge.c (jt_state::push): Only call methods for
4271 which objects are available.
4272 (jt_state::pop): Same.
4273 (jt_state::register_equiv): Same.
4274 (jt_state::register_equivs_on_edge): Same.
4276 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4278 * tree-ssa-threadedge.c (jump_threader::thread_across_edge):
4279 Move pop until after a thread is registered.
4281 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4283 * tree-ssa-threadupdate.c (debug): New.
4285 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4287 * gimple-range-trace.cc (push_dump_file::push_dump_file): New.
4288 (push_dump_file::~push_dump_file): New.
4289 (dump_ranger): Change dump_file temporarily while dumping
4291 * gimple-range-trace.h (class push_dump_file): New.
4293 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4295 * gimple-range-trace.cc (debug_seed_ranger): Remove static.
4296 (dump_ranger): Dump function name.
4298 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4300 * gimple-range-path.cc (path_range_query::range_defined_in_block):
4301 Adjust for non-null.
4302 (path_range_query::adjust_for_non_null_uses): New.
4303 (path_range_query::precompute_ranges): Call
4304 adjust_for_non_null_uses.
4305 * gimple-range-path.h: Add m_non_null and
4306 adjust_for_non_null_uses.
4308 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4310 * gimple-range-path.cc (path_range_query::dump): Dump path
4312 (path_range_query::precompute_ranges): Dump entire path.
4314 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4316 * value-relation.cc (relation_oracle::debug): New.
4317 * value-relation.h (relation_oracle::debug): New.
4319 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4321 * tree-ssa-loop-ch.c: Remove unnecessary include file.
4323 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4325 * gimple-range-fold.cc (fold_using_range::postfold_gcond_edges):
4326 Skip statements with no defining BB.
4327 * gimple-range-path.cc (path_range_query::range_defined_in_block):
4328 Do not get confused by statements with no defining BB.
4330 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4332 * gimple-range-fold.cc (adjust_imagpart_expr): Move from
4333 gimple_range_adjustment. Add support for constants.
4334 (adjust_realpart_expr): New.
4335 (gimple_range_adjustment): Move IMAGPART_EXPR code to
4336 adjust_imagpart_expr.
4337 * range-op.cc (integral_table::integral_table): Add entry for
4340 2021-09-03 Jakub Jelinek <jakub@redhat.com>
4342 * omp-expand.c (expand_omp_atomic_pipeline): Use
4343 IFN_ATOMIC_COMPARE_EXCHANGE instead of
4344 BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_? so that memory order
4347 2021-09-03 Jakub Jelinek <jakub@redhat.com>
4350 * tree.h (DECL_FIELD_ABI_IGNORED): Changed into rvalue only macro
4351 that is false if DECL_BIT_FIELD.
4352 (SET_DECL_FIELD_ABI_IGNORED, DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD,
4353 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD): Define.
4354 * tree-streamer-out.c (pack_ts_decl_common_value_fields): For
4355 DECL_BIT_FIELD stream DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead
4356 of DECL_FIELD_ABI_IGNORED.
4357 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
4358 SET_DECL_FIELD_ABI_IGNORED instead of writing to
4359 DECL_FIELD_ABI_IGNORED and for DECL_BIT_FIELD use
4360 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead.
4361 * lto-streamer-out.c (hash_tree): For DECL_BIT_FIELD hash
4362 DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead of DECL_FIELD_ABI_IGNORED.
4364 2021-09-03 liuhongt <hongtao.liu@intel.com>
4367 * config/i386/amxbf16intrin.h : Remove macro check for __AMX_BF16__.
4368 * config/i386/amxint8intrin.h : Remove macro check for __AMX_INT8__.
4369 * config/i386/amxtileintrin.h : Remove macro check for __AMX_TILE__.
4371 2021-09-02 Martin Sebor <msebor@redhat.com>
4373 PR tree-optimization/17506
4375 * tree-ssa-uninit.c (warn_uninit): Remove conditional guarding note.
4377 2021-09-02 Richard Biener <rguenther@suse.de>
4379 * tree-ssa-loop-im.c (fill_always_executed_in_1): Refine
4380 fix for PR78185 and continue processing when leaving
4383 2021-09-02 Jakub Jelinek <jakub@redhat.com>
4385 PR tree-optimization/99591
4386 * match.pd: Demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
4389 2021-09-02 Richard Biener <rguenther@suse.de>
4392 2021-09-02 Richard Biener <rguenther@suse.de>
4394 PR tree-optimization/102155
4395 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
4396 over a part of the RPO array and do not recurse here.
4397 Dump blocks marked as always executed.
4398 (fill_always_executed_in): Walk over the RPO array and
4399 process loops whose header we run into.
4400 (loop_invariant_motion_in_fun): Compute the first RPO
4401 using rev_post_order_and_mark_dfs_back_seme in iteration
4402 order and pass that to fill_always_executed_in.
4404 2021-09-02 liuhongt <hongtao.liu@intel.com>
4406 * config/i386/i386-modes.def (FLOAT_MODE): Define ieee HFmode.
4407 * config/i386/i386.c (enum x86_64_reg_class): Add
4409 (merge_classes): Handle X86_64_SSEHF_CLASS.
4410 (examine_argument): Ditto.
4411 (construct_container): Ditto.
4412 (classify_argument): Ditto, and set HFmode/HCmode to
4414 (function_value_32): Return _FLoat16/Complex Float16 by
4416 (function_value_64): Return _Float16/Complex Float16 by SSE
4418 (ix86_print_operand): Handle CONST_DOUBLE HFmode.
4419 (ix86_secondary_reload): Require gpr as intermediate register
4420 to store _Float16 from sse register when sse4 is not
4422 (ix86_libgcc_floating_mode_supported_p): Enable _FLoat16 under
4424 (ix86_scalar_mode_supported_p): Ditto.
4425 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Defined.
4426 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add HFmode.
4427 (VALID_INT_MODE_P): Add HFmode and HCmode.
4428 * config/i386/i386.md (*pushhf_rex64): New define_insn.
4430 (*movhf_internal): Ditto.
4431 * doc/extend.texi (Half-Precision Floating Point): Documemt
4434 2021-09-02 Richard Biener <rguenther@suse.de>
4436 PR tree-optimization/102155
4437 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
4438 over a part of the RPO array and do not recurse here.
4439 Dump blocks marked as always executed.
4440 (fill_always_executed_in): Walk over the RPO array and
4441 process loops whose header we run into.
4442 (loop_invariant_motion_in_fun): Compute the first RPO
4443 using rev_post_order_and_mark_dfs_back_seme in iteration
4444 order and pass that to fill_always_executed_in.
4446 2021-09-02 YunQiang Su <syq@debian.org>
4449 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
4451 * config/mips/mips.c (mips_module_isa_name): New.
4452 mips_file_start: add .module mipsREV to all asm output
4454 2021-09-01 Jeff Law <jlaw@localhost.localdomain>
4456 PR tree-optimization/102152
4457 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Reduce a vector
4458 comparison to a scalar comparison before calling
4459 update_stmt_if_modified.
4461 2021-09-01 Andrew Pinski <apinski@marvell.com>
4464 * config/aarch64/aarch64.c (aarch64_expand_setmem):
4465 Check STRICT_ALIGNMENT before creating an overlapping
4468 2021-09-01 Martin Sebor <msebor@redhat.com>
4470 * gimple-ssa-warn-access.cc (get_size_range): Add argument.
4471 (check_access): Pass additional argument.
4472 (check_memop_access): Remove template and make a member function.
4473 (maybe_check_dealloc_call): Make a pass_waccess member function.
4474 (class pass_waccess): Add, rename, and remove members.
4475 (pass_waccess::pass_waccess): Adjust to name change.
4476 (pass_waccess::~pass_waccess): Same.
4477 (check_alloca): Make a member function.
4478 (check_alloc_size_call): Same.
4479 (check_strcat): Same.
4480 (check_strncat): Same.
4481 (check_stxcpy): Same.
4482 (check_stxncpy): Same.
4483 (check_strncmp): Same.
4484 (maybe_warn_rdwr_sizes): Rename...
4485 (pass_waccess::maybe_check_access_sizes): ...to this.
4486 (pass_waccess::check_call): Adjust to name changes.
4487 (pass_waccess::maybe_check_dealloc_call): Make a pass_waccess member
4489 (pass_waccess::execute): Adjust to name changes.
4490 * gimple-ssa-warn-access.h (check_memop_access): Remove.
4491 * pointer-query.cc (access_ref::phi): Handle null pointer.
4492 (access_ref::inform_access): Same.
4493 (pointer_query::put_ref): Modify a cached value, not a copy of it.
4494 (pointer_query::dump): New function.
4495 (compute_objsize_r): Avoid overwriting access_ref::bndrng. Cache
4497 * pointer-query.h (pointer_query::dump): Declare.
4498 * tree-ssa-strlen.c (get_range): Simplify. Use function query.
4499 (dump_strlen_info): Use function query.
4500 (printf_strlen_execute): Factor code out into pointer_query::put_ref.
4502 2021-09-01 Thomas Schwinge <thomas@codesourcery.com>
4504 * tree.c (walk_tree_1) <OMP_CLAUSE>: Simplify.
4506 2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
4508 * doc/extend.texi: Document unavailable attribute.
4509 * print-tree.c (print_node): Handle unavailable attribute.
4510 * tree-core.h (struct tree_base): Add a bit to carry unavailability.
4511 * tree.c (error_unavailable_use): New.
4512 * tree.h (TREE_UNAVAILABLE): New.
4513 (error_unavailable_use): New.
4515 2021-09-01 Jakub Jelinek <jakub@redhat.com>
4517 PR tree-optimization/102124
4518 * tree-vect-patterns.c (vect_recog_widen_op_pattern): For ORIG_CODE
4519 MINUS_EXPR, if itype is unsigned with smaller precision than type,
4520 add an extra cast to signed variant of itype to ensure sign-extension.
4522 2021-09-01 Martin Liska <mliska@suse.cz>
4524 * graph.c (draw_cfg_node_succ_edges): Do not color fallthru
4525 edges and rather use colors for TRUE and FALSE edges.
4527 2021-09-01 Richard Biener <rguenther@suse.de>
4529 PR tree-optimization/93491
4530 * tree-ssa-pre.c (compute_avail): Set BB_MAY_NOTRETURN
4531 after processing the stmt itself. Do not consider
4532 pure functions possibly not returning. Properly avoid
4533 adding possibly trapping calls to EXP_GEN when there's
4534 a preceeding possibly not returning call.
4535 * tree-ssa-sccvn.c (vn_reference_may_trap): Conservatively
4538 2021-09-01 Richard Biener <rguenther@suse.de>
4540 PR tree-optimization/102139
4541 * tree-vectorizer.h (vec_base_alignments): Adjust hash-map
4542 type to record a std::pair of the stmt-info and the innermost
4544 (dr_vec_info::group): New member.
4545 * tree-vect-data-refs.c (vect_record_base_alignment): Adjust.
4546 (vect_compute_data_ref_alignment): Verify the recorded
4547 base alignment can be used.
4548 (data_ref_pair): Remove.
4549 (dr_group_sort_cmp): Adjust.
4550 (vect_analyze_data_ref_accesses): Store the group-ID in the
4551 dr_vec_info and operate on a vector of dr_vec_infos.
4553 2021-09-01 YunQiang Su <yunqiang.su@cipunited.com>
4555 * read-md.c (md_reader::handle_enum): support value assignation.
4556 * doc/md.texi: record define_c_enum value assignation support.
4558 2021-09-01 Jakub Jelinek <jakub@redhat.com>
4560 PR tree-optimization/102141
4561 * gimple-ssa-store-merging.c (bswap_view_convert): Add BEFORE
4562 argument. If false, emit stmts after gsi instead of before, and
4564 (bswap_replace): Adjust callers. When converting output of bswap,
4565 emit VIEW_CONVERT prepratation stmts after a copy of gsi instead
4568 2021-09-01 liuhongt <hongtao.liu@intel.com>
4570 * emit-rtl.c (validate_subreg): Get rid of all float-int
4573 2021-09-01 liuhongt <hongtao.liu@intel.com>
4576 2021-08-30 liuhongt <hongtao.liu@intel.com>
4578 * expmed.c (extract_bit_field_1): Make sure we're playing with
4579 integral modes before call extract_integral_bit_field.
4580 (extract_integral_bit_field): Add a parameter of type
4581 scalar_int_mode which corresponds to of tmode.
4582 And call extract_and_convert_fixed_bit_field instead of
4583 extract_fixed_bit_field and convert_extracted_bit_field.
4584 (extract_and_convert_fixed_bit_field): New function, it's a
4585 combination of extract_fixed_bit_field and
4586 convert_extracted_bit_field.
4588 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
4590 * tree.c (walk_tree_1) <OMP_CLAUSE_TILE>: Handle three operands.
4592 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
4594 * omp-general.h (omp_is_reference): Rename to...
4595 (omp_privatize_by_reference): ... this. Adjust all users...
4596 * omp-general.c: ... here, ...
4597 * gimplify.c: ... here, ...
4598 * omp-expand.c: ... here, ...
4599 * omp-low.c: ... here.
4601 2021-08-31 Martin Sebor <msebor@redhat.com>
4603 * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Test
4604 pointer element for equality to zero, not that of the cotaining
4607 2021-08-31 Martin Sebor <msebor@redhat.com>
4609 * gcc-rich-location.h (gcc_rich_location): Make ctor explicit.
4611 2021-08-31 Martin Sebor <msebor@redhat.com>
4613 * function.h (function): Add comments.
4614 (get_range_query): Same. Add attribute returns nonnull.
4616 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
4618 * expr.c (convert_modes): Don't use subreg_promoted_mode on a
4619 SUBREG if it can't be guaranteed to a SUBREG_PROMOTED_VAR_P set.
4620 Instead use the standard (safer) is_a <scalar_int_mode> idiom.
4622 2021-08-31 Jeff Law <jlaw@localhost.localdomain>
4624 * config.gcc (cris-*-elf, cris-*-none): Remove dbxelf.h from
4626 (m32r-*-elf, m32rle-*-elf, m32r-*-linux): Likewise.
4627 (mn10300-*-*, am33_2.0-*-linux*): Likewise.
4628 (xtensa*-*-elf, xtensa*-*-linux, xtensa*-*-uclinux): Likewise.
4629 (m32c-*-elf*, m32c-*-rtems*): Likewise.
4630 * config/cris/cris.h (DBX_NO_XREFS): Remove.
4631 (DBX_CONTIN_LENGTH, DBX_CONTIN_CHAR): Likewise.
4632 * config/m32r/m32r.h (DBXOUT_SOURCE_LINE): Likewise.
4633 (DBX_DEBUGGING_INFO, DBX_CONTIN_LENGTH): Likewise.
4634 * config/mn10300/mn10300.h (DEFAULT_GDB_EXTENSIONS): Likewise.
4635 * config/mn10300/linux.h (DBX_REGISTER_NAMES): Likewise.
4637 2021-08-31 Marcel Vollweiler <marcel@codesourcery.com>
4639 * gimplify.c (gimplify_scan_omp_clauses): Error handling. 'ancestor' only
4640 allowed on target constructs and only with particular other clauses.
4641 * omp-expand.c (expand_omp_target): Output of 'sorry, not supported' if
4643 * omp-low.c (check_omp_nesting_restrictions): Error handling. No nested OpenMP
4644 structs when 'ancestor' is used.
4645 (scan_omp_1_stmt): No usage of OpenMP runtime routines in a target region when
4647 * tree-pretty-print.c (dump_omp_clause): Append 'ancestor'.
4648 * tree.h (OMP_CLAUSE_DEVICE_ANCESTOR): Define macro.
4650 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
4652 * expr.c (convert_modes): Preserve SUBREG_PROMOTED_VAR_P when
4653 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
4655 * simplify-rtx.c (simplify_unary_operation_1) [SIGN_EXTEND]:
4656 Likewise, preserve SUBREG_PROMOTED_VAR_P when creating a (wider)
4657 partial subreg from a SUBREG_PROMOTED_VAR_P subreg. Generate
4658 SIGN_EXTEND of the SUBREG_REG when a subreg would be paradoxical.
4659 [ZERO_EXTEND]: Likewise, preserve SUBREG_PROMOTED_VAR_P when
4660 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
4661 subreg. Generate ZERO_EXTEND of the SUBREG_REG when a subreg
4662 would be paradoxical.
4664 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
4666 * combine.c (combine_simplify_rtx): Avoid converting an explicit
4667 TRUNCATE into a lowpart SUBREG on !TRULY_NOOP_TRUNCATION targets.
4668 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
4670 2021-08-31 Richard Biener <rguenther@suse.de>
4672 PR tree-optimization/102142
4673 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Fix
4674 condition under which to unset the visited flag.
4676 2021-08-31 Richard Biener <rguenther@suse.de>
4678 PR middle-end/102129
4679 * tree-ssa-ter.c (find_replaceable_in_bb): Do not move
4680 possibly trapping expressions across calls.
4682 2021-08-31 Jakub Jelinek <jakub@redhat.com>
4684 PR tree-optimization/102134
4685 * tree-ssa-ccp.c (bit_value_binop) <case RSHIFT_EXPR>: If sgn is
4686 UNSIGNED and r1val | r1mask has MSB set, ensure lzcount doesn't
4689 2021-08-31 Andrew Pinski <apinski@marvell.com>
4692 * collect-utils.c (setup_signals): New declaration.
4693 * collect-utils.h (setup_signals): New function.
4694 * collect2.c (handler): Delete.
4695 (main): Instead of manually setting up the signals,
4696 just call setup_signals.
4697 * lto-wrapper.c (main): Likewise.
4699 2021-08-31 Andrew Pinski <apinski@marvell.com>
4702 * config/i386/i386-protos.h (x86_output_aligned_bss):
4703 Change align argument to unsigned type.
4704 (x86_elf_aligned_decl_common): Likewise.
4705 * config/i386/i386.c (x86_elf_aligned_decl_common): Likewise.
4706 (x86_output_aligned_bss): Likewise.
4708 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
4710 * config/mips/mips.c (mips_module_isa_name): New.
4711 mips_file_start: add .module mipsREV to all asm output
4713 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
4715 * config/mips/mips.h (struct mips_cpu_info): define enum mips_isa;
4716 use enum instead of int for 'isa' member.
4717 * config.gcc, config/mips/mips.c, config/mips/mips-cpus.def,
4718 config/mips/netbsd.h: replace hardcoded numbers with enum.
4720 2021-08-31 liuhongt <hongtao.liu@intel.com>
4722 * config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
4723 define_split to define_insn_and_split.
4724 (*avx2_eq<mode>3): Removed.
4725 (<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
4726 (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
4727 (*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
4729 (*avx2_gt<mode>3): Removed.
4730 (<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
4731 define_insn to define_expand, and adjust pattern.
4732 (UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.
4734 2021-08-30 David Malcolm <dmalcolm@redhat.com>
4737 * Makefile.in (ANALYZER_OBJS): Add analyzer/call-info.o.
4739 2021-08-30 Jason Merrill <jason@redhat.com>
4741 * doc/invoke.texi: Document -Wmissing-requires.
4743 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
4745 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Remove
4746 TARGET_EXTRA_BUILTINS guard.
4748 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
4750 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Change
4751 initialization of V2DI_type_node and unsigned_V2DI_type_node.
4753 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
4755 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Use the new
4756 decl when new_builtins_are_live.
4757 * config/rs6000/rs6000-builtin-new.def (__builtin_cfstring): New
4760 2021-08-30 Pat Haugen <pthaugen@linux.ibm.com>
4762 * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add
4763 OPTION_MASK_P10_FUSION_2STORE.
4764 (POWERPC_MASKS): Likewise.
4765 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
4766 store fusion for Power10.
4767 (is_fusable_store): New.
4768 (power10_sched_reorder): Likewise.
4769 (rs6000_sched_reorder): Do Power10 specific reordering.
4770 (rs6000_sched_reorder2): Likewise.
4771 * config/rs6000/rs6000.opt: Add new option.
4773 2021-08-30 Richard Biener <rguenther@suse.de>
4775 PR tree-optimization/102128
4776 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
4777 Move scanning for if-converted scalar code to the caller
4778 and instead delay clearing the visited flag for profitable
4780 (vect_slp_region): Cost all subgraphs before scheduling.
4781 For if-converted BB vectorization scan for scalar COND_EXPRs
4782 and do not vectorize if any found and the cost model is
4785 2021-08-30 Richard Biener <rguenther@suse.de>
4787 * common.opt (fexceptions): Mark
4788 EnabledBy(fnon-call-exceptions).
4789 * doc/invoke.texi (fnon-call-exceptions): Document this
4790 enables -fexceptions.
4792 2021-08-30 Sebastian Huber <sebastian.huber@embedded-brains.de>
4794 * tsystem.h (abort): Define abort() if inhibit_libc is defined and it
4795 is not already defined.
4797 2021-08-30 liuhongt <hongtao.liu@intel.com>
4799 * expmed.c (extract_bit_field_1): Make sure we're playing with
4800 integral modes before call extract_integral_bit_field.
4801 (extract_integral_bit_field): Add a parameter of type
4802 scalar_int_mode which corresponds to of tmode.
4803 And call extract_and_convert_fixed_bit_field instead of
4804 extract_fixed_bit_field and convert_extracted_bit_field.
4805 (extract_and_convert_fixed_bit_field): New function, it's a
4806 combination of extract_fixed_bit_field and
4807 convert_extracted_bit_field.
4809 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
4811 * config/darwin.c (darwin_libc_has_function): Do not run
4812 the checks for x86 or modern Darwin. Make sure that there
4813 is a value set for darwin_macosx_version_min before testing.
4815 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
4817 * config/i386/darwin.h (CLEAR_INSN_CACHE): New.
4819 2021-08-28 Jan Hubicka <hubicka@ucw.cz>
4821 * ipa-modref-tree.h (modref_access_node::merge): Break out
4822 logic combining offsets and logic merging ranges to ...
4823 (modref_access_node::combined_offsets): ... here
4824 (modref_access_node::update2): ... here
4825 (modref_access_node::closer_pair_p): New member function.
4826 (modref_access_node::forced_merge): New member function.
4827 (modre_ref_node::insert): Do merging when table is full.
4829 2021-08-28 YunQiang Su <yunqiang.su@cipunited.com>
4832 * config.gcc: MIPS: use N64 ABI by default if the triple end
4833 with -gnuabi64, which is used by Debian since 2013.
4835 2021-08-28 Alexandre Oliva <oliva@adacore.com>
4837 * ipa-modref.c (analyze_function): Skip debug stmts.
4838 * tree-inline.c (estimate_num_insn): Consider builtins even
4839 without a cgraph_node.
4841 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
4843 * config/h8300/bitfield.md (cstore<mode>4): Remove expander.
4844 * config/h8300/h8300.c (h8300_expand_branch): Remove function.
4845 * config/h8300/h8300-protos.h (h8300_expadn_branch): Remove prototype.
4846 * config/h8300/h8300.md (eqne): New code iterator.
4847 (geultu, geultu_to_c): Similarly.
4848 * config/h8300/testcompare.md (cstore<mode>4): Dummy expander.
4849 (store_c_<mode>, store_c_i_<mode>): New define_insn_and_splits
4850 (cmp<mode>_c): New pattern
4852 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
4854 * tree-ssa-dom.c (reduce_vector_comparison_to_scalar_comparison): New
4856 (dom_opt_dom_walker::optimize_stmt): Use it.
4858 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
4860 * config/darwin.c (finalize_ctors): Add a section-start linker-
4862 (finalize_dtors): Likewise.
4863 * config/darwin.h (MIN_LD64_INIT_TERM_START_LABELS): New.
4865 2021-08-27 Bill Schmidt <wschmidt@linux.ibm.com>
4867 * config/rs6000/rs6000-call.c (rs6000-builtins.h): New #include.
4868 (rs6000_init_builtins): Call rs6000_init_generated_builtins. Skip the
4869 old initialization logic when new builtins are enabled.
4870 * config/rs6000/rs6000-gen-builtins.c (write_decls): Rename
4871 rs6000_autoinit_builtins to rs6000_init_generated_builtins.
4872 (write_init_file): Likewise.
4874 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
4876 * configure.ac (darwin2[[0-9]]* | darwin19*): Alter use of
4877 gcc_GAS_CHECK_FEATURE to remove an extraneous parameter.
4878 (amdgcn-* | gcn-*) Likewise.
4880 2021-08-27 Anthony Sharp <anthonysharp15@gmail.com>
4882 * symbol-summary.h: Added missing template keyword.
4884 2021-08-27 Richard Biener <rguenther@suse.de>
4886 PR tree-optimization/45178
4887 * tree-ssa-dce.c (find_obviously_necessary_stmts): For
4888 infinite loops without exit do not mark control dependent
4889 edges of the latch necessary.
4891 2021-08-27 konglin1 <lingling.kong@intel.com>
4894 * config/i386/sse.md: (<avx512>scattersi<mode>): Add mask operand to
4896 (<avx512>scattersi<mode>): Likewise.
4897 (*avx512f_scattersi<VI48F:mode>): Merge mask operand to set_dest.
4898 (*avx512f_scatterdi<VI48F:mode>): Likewise
4900 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
4902 * config/rs6000/rs6000.c (rs6000_builtin_md_vectorized_function): Add
4903 support for built-in functions MISC_BUILTIN_DIVWE, MISC_BUILTIN_DIVWEU,
4904 MISC_BUILTIN_DIVDE, MISC_BUILTIN_DIVDEU, P10_BUILTIN_CFUGED,
4905 P10_BUILTIN_CNTLZDM, P10_BUILTIN_CNTTZDM, P10_BUILTIN_PDEPD and
4906 P10_BUILTIN_PEXTD on Power10.
4908 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
4910 * config/rs6000/rs6000-call.c (builtin_function_type): Add unsigned
4911 signedness for some Power10 bifs.
4913 2021-08-27 David Edelsohn <dje.gcc@gmail.com>
4916 * config/rs6000/rs6000.c (rs6000_adjust_field_align): Use
4917 computed alignment if the entire struct has attribute packed.
4919 2021-08-27 liuhongt <hongtao.liu@intel.com>
4923 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
4924 IX86_BUILTIN_SHUFPD512, IX86_BUILTIN_SHUFPS512,
4925 IX86_BUILTIN_SHUFPD256, IX86_BUILTIN_SHUFPS,
4926 IX86_BUILTIN_SHUFPS256.
4927 (ix86_masked_all_ones): New function.
4929 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
4931 * config/i386/i386.md (*btr<mode>_1): Call force_reg unconditionally.
4932 (conditional moves with memory inputs splitters): Ditto.
4933 * config/i386/sse.md (one_cmpl<mode>2): Simplify.
4935 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
4937 * ipa-modref-tree.h (modref_access_node::try_merge_with): Restart
4938 search after merging.
4940 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
4942 * config/rs6000/rs6000-overload.def: Add remaining overloads.
4944 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
4946 * config/rs6000/rs6000-builtin-new.def: Add cell stanza.
4948 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
4950 * config/rs6000/rs6000-builtin-new.def: Add ieee128-hw, dfp,
4951 crypto, and htm stanzas.
4953 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
4955 * config/rs6000/rs6000-builtin-new.def: Add mma stanza.
4957 2021-08-26 Martin Sebor <msebor@redhat.com>
4959 * tree-ssa-uninit.c (warn_uninit): Refactor and simplify.
4960 (warn_uninit_phi_uses): Remove argument from calls to warn_uninit.
4961 (warn_uninitialized_vars): Same. Reduce visibility of locals.
4962 (warn_uninitialized_phi): Same.
4964 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
4966 * tree-ssa-ccp.c (get_individual_bits): Helper function to
4967 extract the individual bits from a widest_int constant (mask).
4968 (gray_code_bit_flips): New read-only table for effiently
4969 enumerating permutations/combinations of bits.
4970 (bit_value_binop) [LROTATE_EXPR, RROTATE_EXPR]: Handle rotates
4971 by unknown counts that are guaranteed less than the target
4972 precision and four or fewer unknown bits by enumeration.
4973 [LSHIFT_EXPR, RSHIFT_EXPR]: Likewise, also handle shifts by
4974 enumeration under the same conditions. Handle remaining
4975 shifts as a mask based upon the minimum possible shift value.
4977 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
4978 Richard Biener <rguenther@suse.de>
4980 * match.pd (shift transformations): Remove a redundant
4981 !POINTER_TYPE_P check.
4983 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
4986 * config/i386/i386.md (cmove reg-reg move elimination peephole2s):
4987 Set all_regs to true in the call to replace_rtx.
4989 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
4991 * ipa-modref-tree.c (test_insert_search_collapse): Update test.
4992 * ipa-modref-tree.h (modref_base_node::insert): Be smarter when
4993 hiting --param modref-max-refs limit.
4994 (modref_tree:insert_base): Be smarter when hitting
4995 --param modref-max-bases limit. Add new parameter REF.
4996 (modref_tree:insert): Update.
4997 (modref_tree:merge): Update.
4998 * ipa-modref.c (read_modref_records): Update.
5000 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
5002 * params.opt: (modref-max-adjustments): Add full stop.
5004 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
5006 * ipa-modref-tree.h (modref_ref_node::verify): New member
5008 (modref_ref_node::insert): Use it.
5009 (modref_ref_node::try_mere_with): Fix off by one error.
5011 2021-08-26 Martin Liska <mliska@suse.cz>
5012 Stefan Kneifel <stefan.kneifel@bluewin.ch>
5014 * cgraph.h (create_version_clone_with_body): Add new parameter.
5015 * cgraphclones.c: Likewise.
5016 * multiple_target.c (create_dispatcher_calls): Do not use
5018 (create_target_clone): Likewise here.
5020 2021-08-26 Jonathan Yong <10walls@gmail.com>
5022 * doc/extend.texi: Add note about reserved priorities
5023 to the constructor attribute.
5025 2021-08-25 Martin Sebor <msebor@redhat.com>
5027 * gimple-range-cache.cc (ssa_global_cache::dump): Avoid printing
5028 range table header alone.
5029 * gimple-range.cc (gimple_ranger::export_global_ranges): Same.
5031 2021-08-25 Jan Hubicka <hubicka@ucw.cz>
5033 * doc/invoke.texi: Document --param modref-max-adjustments.
5034 * ipa-modref-tree.c (test_insert_search_collapse): Update.
5035 (test_merge): Update.
5036 * ipa-modref-tree.h (struct modref_access_node): Add adjustments;
5037 (modref_access_node::operator==): Fix handling of access ranges.
5038 (modref_access_node::contains): Constify parameter; handle also
5039 mismatched parm offsets.
5040 (modref_access_node::update): New function.
5041 (modref_access_node::merge): New function.
5042 (unspecified_modref_access_node): Update constructor.
5043 (modref_ref_node::insert_access): Add record_adjustments parameter;
5045 (modref_ref_node::try_merge_with): New private function.
5046 (modref_tree::insert): New record_adjustments parameter.
5047 (modref_tree::merge): New record_adjustments parameter.
5048 (modref_tree::copy_from): Update.
5049 * ipa-modref.c (dump_access): Dump adjustments field.
5050 (get_access): Update constructor.
5051 (record_access): Update call of insert.
5052 (record_access_lto): Update call of insert.
5053 (merge_call_side_effects): Add record_adjustments parameter.
5054 (get_access_for_fnspec): Update.
5055 (process_fnspec): Update.
5056 (analyze_call): Update.
5057 (analyze_function): Update.
5058 (read_modref_records): Update.
5059 (ipa_merge_modref_summary_after_inlining): Update.
5060 (propagate_unknown_call): Update.
5061 (modref_propagate_in_scc): Update.
5062 * params.opt (param-max-modref-adjustments=): New.
5064 2021-08-25 Michael Meissner <meissner@linux.ibm.com>
5066 * config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
5068 (xxspltiw_v4si): Use vecperm type attribute.
5069 (xxspltiw_v4si_inst): Use vecperm type attribute.
5070 (xxspltiw_v4sf_inst): Likewise.
5071 (xxspltidp_v2df): Use vecperm type attribute. Use
5072 UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
5073 (xxspltidp_v2df_inst): Likewise.
5074 (xxsplti32dx_v4si): Use vecperm type attribute.
5075 (xxsplti32dx_v4si_inst): Likewise.
5076 (xxsplti32dx_v4sf_inst): Likewise.
5077 (xxblend_<mode>): Likewise.
5078 (xxpermx): Likewise.
5079 (xxpermx_inst): Likewise.
5082 2021-08-25 Lewis Hyatt <lhyatt@gmail.com>
5085 * coretypes.h (typedef diagnostic_input_charset_callback): Declare.
5086 * diagnostic.c (diagnostic_initialize_input_context): New function.
5087 * diagnostic.h (diagnostic_initialize_input_context): Declare.
5088 * input.c (default_charset_callback): New function.
5089 (file_cache::initialize_input_context): New function.
5090 (file_cache_slot::create): Added ability to convert the input
5091 according to the input context.
5092 (file_cache::file_cache): Initialize the new input context.
5093 (class file_cache_slot): Added new m_alloc_offset member.
5094 (file_cache_slot::file_cache_slot): Initialize the new member.
5095 (file_cache_slot::~file_cache_slot): Handle potentially offset buffer.
5096 (file_cache_slot::maybe_grow): Likewise.
5097 (file_cache_slot::needs_read_p): Handle NULL fp, which is now possible.
5098 (file_cache_slot::get_next_line): Likewise.
5099 * input.h (class file_cache): Added input context member.
5101 2021-08-25 Richard Biener <rguenther@suse.de>
5103 PR tree-optimization/102046
5104 * tree-vect-slp.c (vect_build_slp_tree_2): Conservatively
5105 update ->any_pattern when swapping operands.
5107 2021-08-25 Hongyu Wang <hongyu.wang@intel.com>
5110 * config/i386/i386.c (ix86_live_on_entry): Adjust comment.
5111 (ix86_decompose_address): Remove retval check for ASHIFT,
5112 allow non-canonical zero extend if AND mask covers ASHIFT
5114 (ix86_legitimate_address_p): Adjust condition for decompose.
5115 (ix86_rtx_costs): Adjust cost for lea with non-canonical
5117 Co-Authored by: Uros Bizjak <ubizjak@gmail.com>
5119 2021-08-25 Jiufu Guo <guojiufu@linux.ibm.com>
5121 PR tree-optimization/101145
5122 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
5124 (number_of_iterations_lt): Invoke above function.
5125 (adjust_cond_for_loop_until_wrap):
5126 Merge to number_of_iterations_until_wrap.
5127 (number_of_iterations_cond): Update invokes for
5128 adjust_cond_for_loop_until_wrap and number_of_iterations_lt.
5130 2021-08-25 konglin1 <lingling.kong@intel.com>
5133 * config/i386/avx512dqintrin.h (_mm512_fpclass_ps_mask): Fix
5135 (_mm512_mask_fpclass_ps_mask): Ditto.
5137 2021-08-25 Kewen Lin <linkw@linux.ibm.com>
5139 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Remove.
5140 (vec_unpacku_hi_v8hi): Likewise.
5141 (vec_unpacku_lo_v16qi): Likewise.
5142 (vec_unpacku_lo_v8hi): Likewise.
5143 (vec_unpacku_hi_<VP_small_lc>): New define_expand.
5144 (vec_unpacku_lo_<VP_small_lc>): Likewise.
5146 2021-08-24 David Edelsohn <dje.gcc@gmail.com>
5148 * config/rs6000/aix.h (SYSTEM_IMPLICIT_EXTERN_C): Delete.
5149 * config/rs6000/aix71.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
5150 * config/rs6000/aix72.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
5151 * config/rs6000/aix73.h (TARGET_AIX_VERSION): Increase to 73.
5153 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5155 PR middle-end/102031
5156 * simplify-rtx.c (simplify_truncation): When comparing precisions
5157 use "subreg_prec" variable, not "subreg_mode".
5159 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5161 * config/rs6000/rs6000-builtin-new.def: Add power10 and power10-64
5164 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5166 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Initialize
5167 various pointer type nodes.
5168 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add enum
5169 values for various pointer types.
5170 (ptr_V16QI_type_node): New macro.
5171 (ptr_V1TI_type_node): New macro.
5172 (ptr_V2DI_type_node): New macro.
5173 (ptr_V2DF_type_node): New macro.
5174 (ptr_V4SI_type_node): New macro.
5175 (ptr_V4SF_type_node): New macro.
5176 (ptr_V8HI_type_node): New macro.
5177 (ptr_unsigned_V16QI_type_node): New macro.
5178 (ptr_unsigned_V1TI_type_node): New macro.
5179 (ptr_unsigned_V8HI_type_node): New macro.
5180 (ptr_unsigned_V4SI_type_node): New macro.
5181 (ptr_unsigned_V2DI_type_node): New macro.
5182 (ptr_bool_V16QI_type_node): New macro.
5183 (ptr_bool_V8HI_type_node): New macro.
5184 (ptr_bool_V4SI_type_node): New macro.
5185 (ptr_bool_V2DI_type_node): New macro.
5186 (ptr_bool_V1TI_type_node): New macro.
5187 (ptr_pixel_type_node): New macro.
5188 (ptr_intQI_type_node): New macro.
5189 (ptr_uintQI_type_node): New macro.
5190 (ptr_intHI_type_node): New macro.
5191 (ptr_uintHI_type_node): New macro.
5192 (ptr_intSI_type_node): New macro.
5193 (ptr_uintSI_type_node): New macro.
5194 (ptr_intDI_type_node): New macro.
5195 (ptr_uintDI_type_node): New macro.
5196 (ptr_intTI_type_node): New macro.
5197 (ptr_uintTI_type_node): New macro.
5198 (ptr_long_integer_type_node): New macro.
5199 (ptr_long_unsigned_type_node): New macro.
5200 (ptr_float_type_node): New macro.
5201 (ptr_double_type_node): New macro.
5202 (ptr_long_double_type_node): New macro.
5203 (ptr_dfloat64_type_node): New macro.
5204 (ptr_dfloat128_type_node): New macro.
5205 (ptr_ieee128_type_node): New macro.
5206 (ptr_ibm128_type_node): New macro.
5207 (ptr_vector_pair_type_node): New macro.
5208 (ptr_vector_quad_type_node): New macro.
5209 (ptr_long_long_integer_type_node): New macro.
5210 (ptr_long_long_unsigned_type_node): New macro.
5212 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5214 * config/rs6000/rs6000-builtin-new.def: Add power9-vector, power9,
5215 and power9-64 stanzas.
5217 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5218 Tom de Vries <tdevries@suse.de>
5220 * config.gcc (nvptx-*-*): Define {c,c++}_target_objs.
5221 * config/nvptx/nvptx-protos.h (nvptx_cpu_cpp_builtins): Prototype.
5222 * config/nvptx/nvptx.h (TARGET_CPU_CPP_BUILTINS): Implement with
5223 a call to the new nvptx_cpu_cpp_builtins function in nvptx-c.c.
5224 * config/nvptx/t-nvptx (nvptx-c.o): New rule.
5225 * config/nvptx/nvptx-c.c: New source file.
5226 (nvptx_cpu_cpp_builtins): Move implementation here.
5228 2021-08-24 Martin Sebor <msebor@redhat.com>
5230 PR middle-end/101600
5231 PR middle-end/101977
5232 * gimple-ssa-warn-access.cc (maybe_warn_for_bound): Tighten up
5233 the phrasing of a warning.
5234 (check_access): Use the remaining size after subtracting any offset
5235 rather than the whole object size.
5236 * pointer-query.cc (access_ref::get_ref): Clear BASE0 flag if it's
5237 clear for any nonnull PHI argument.
5238 (compute_objsize): Clear argument.
5240 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5242 * config/rs6000/rs6000-builtin-new.def: Add power8-vector stanza.
5244 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5246 * config/rs6000/rs6000-builtin-new.def: Add power7 and power7-64
5249 2021-08-24 Andrew MacLeod <amacleod@redhat.com>
5251 * value-relation.cc (rr_transitive_table): New.
5252 (relation_transitive): New.
5253 (value_relation::swap): Remove.
5254 (value_relation::apply_transitive): New.
5255 (relation_oracle::relation_oracle): Allocate a new tmp bitmap.
5256 (relation_oracle::register_relation): Call register_transitives.
5257 (relation_oracle::register_transitives): New.
5258 * value-relation.h (relation_oracle): Add new temporary bitmap and
5261 2021-08-24 H.J. Lu <hjl.tools@gmail.com>
5264 * config/i386/i386-expand.c (ix86_expand_vector_move): Broadcast
5265 from integer to a pseudo vector register.
5267 2021-08-24 Richard Biener <rguenther@suse.de>
5269 PR tree-optimization/100089
5270 * tree-vectorizer.h (vect_slp_bb): Rename to ...
5271 (vect_slp_if_converted_bb): ... this and get the original
5272 loop as new argument.
5273 * tree-vectorizer.c (try_vectorize_loop_1): Revert previous fix,
5274 pass original loop to vect_slp_if_converted_bb.
5275 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
5276 If orig_loop was passed scan the not vectorized stmts
5277 for COND_EXPRs and force not profitable if found.
5278 (vect_slp_region): Pass down all SLP instances to costing
5279 if orig_loop was specified.
5280 (vect_slp_bbs): Pass through orig_loop.
5281 (vect_slp_bb): Rename to ...
5282 (vect_slp_if_converted_bb): ... this and get the original
5283 loop as new argument.
5284 (vect_slp_function): Adjust.
5286 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
5289 * config/arm/arm.md (attribute arch): Add fix_vlldm.
5290 (arch_enabled): Use it.
5291 * config/arm/vfp.md (lazy_store_multiple_insn): Add alternative to
5292 use when erratum mitigation is needed.
5294 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
5297 * config/arm/arm.opt (mfix-cmse-cve-2021-35465): New option.
5298 * doc/invoke.texi (Arm Options): Document it.
5299 * config/arm/arm-cpus.in (quirk_vlldm): New feature bit.
5300 (ALL_QUIRKS): Add quirk_vlldm.
5301 (cortex-m33): Add quirk_vlldm.
5302 (cortex-m35p, cortex-m55): Likewise.
5303 * config/arm/arm.c (arm_option_override): Enable fix_vlldm if
5304 targetting an affected CPU and not explicitly controlled on
5307 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
5309 * config/arm/vfp.md (lazy_store_multiple_insn): Rewrite as valid RTL.
5310 (lazy_load_multiple_insn): Likewise.
5312 2021-08-24 liuhongt <hongtao.liu@intel.com>
5315 * config/i386/sse.md (<avx512>_vternlog<mode><sd_maskz_name>):
5316 Enable avx512 embedded broadcast.
5317 (*<avx512>_vternlog<mode>_all): Ditto.
5318 (<avx512>_vternlog<mode>_mask): Ditto.
5320 2021-08-24 liuhongt <hongtao.liu@intel.com>
5323 * config/i386/i386.c (ix86_rtx_costs): Define cost for
5325 * config/i386/i386.h (STRIP_UNARY): New macro.
5326 * config/i386/predicates.md (reg_or_notreg_operand): New
5328 * config/i386/sse.md (*<avx512>_vternlog<mode>_all): New define_insn.
5329 (*<avx512>_vternlog<mode>_1): New pre_reload
5330 define_insn_and_split.
5331 (*<avx512>_vternlog<mode>_2): Ditto.
5332 (*<avx512>_vternlog<mode>_3): Ditto.
5333 (any_logic1,any_logic2): New code iterator.
5334 (logic_op): New code attribute.
5335 (ternlogsuffix): Extend to VNxDF and VNxSF.
5337 2021-08-24 Richard Biener <rguenther@suse.de>
5339 * doc/invoke.texi (vect-inner-loop-cost-factor): Adjust.
5340 * params.opt (--param vect-inner-loop-cost-factor): Adjust
5342 * tree-vect-loop.c (vect_analyze_loop_form): Initialize
5343 inner_loop_cost_factor to the minimum of the estimated number
5344 of iterations of the inner loop and vect-inner-loop-cost-factor.
5346 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5347 Richard Biener <rguenther@suse.de>
5349 * config/i386/i386-features.c (compute_convert_gain): Provide
5350 more accurate values for CONST_INT, when optimizing for size.
5351 * config/i386/i386.c (COSTS_N_BYTES): Move definition from here...
5352 * config/i386/i386.h (COSTS_N_BYTES): to here.
5354 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5355 Jakub Jelinek <jakub@redhat.com>
5357 PR middle-end/102029
5358 * match.pd (shift transformations): Add an additional check for
5359 !POINTER_TYPE_P in the recently added left shift transformation.
5361 2021-08-24 liuhongt <hongtao.liu@intel.com>
5363 PR tree-optimization/100089
5364 * tree-vectorizer.c (try_vectorize_loop_1): Disable slp in
5365 loop vectorizer when cost model is very-cheap.
5367 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5369 * config/rs6000/rs6000-gen-builtins.c (parse_bif_entry): Don't call
5370 asprintf, which is not available on AIX.
5372 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5374 * config.gcc (target_gtfiles): Add ./rs6000-builtins.h.
5375 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Set.
5377 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5379 * config.gcc (powerpc*-*-*): Add rs6000-builtins.o to extra_objs.
5380 * config/rs6000/rs6000-gen-builtins.c (main): Close init_file
5382 * config/rs6000/t-rs6000 (rs6000-gen-builtins.o): New target.
5383 (rbtree.o): Likewise.
5384 (rs6000-gen-builtins): Likewise.
5385 (rs6000-builtins.c): Likewise.
5386 (rs6000-builtins.h): Likewise.
5387 (rs6000.o): Add dependency.
5388 (EXTRA_HEADERS): Add rs6000-vecdefines.h.
5389 (rs6000-vecdefines.h): New target.
5390 (rs6000-builtins.o): Likewise.
5391 (rs6000-call.o): Add rs6000-builtins.h as a dependency.
5392 (rs6000-c.o): Likewise.
5394 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5397 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace):
5398 Diagnose buffer overrun.
5399 (safe_inc_pos): Fix overrun detection.
5400 (match_identifier): Diagnose buffer overrun.
5401 (match_integer): Likewise.
5402 (match_to_right_bracket): Likewise.
5404 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
5406 * ipa-modref-tree.h (modref_access_node::range_info_useful_p):
5407 Improve range compare.
5408 (modref_access_node::contains): New member function.
5409 (modref_access_node::search): Remove.
5410 (modref_access_node::insert): Be smarter about subaccesses.
5412 2021-08-23 Thomas Schwinge <thomas@codesourcery.com>
5414 * config/i386/i386-options.c (ix86_omp_device_kind_arch_isa)
5415 <omp_device_arch> [ACCEL_COMPILER]: Match "intel_mic".
5416 * config/i386/t-omp-device (omp-device-properties-i386) <arch>:
5419 2021-08-23 Jeff Law <jlaw@localhost.localdomain>
5421 * config/h8300/h8300-protos.h (h8300_expand_epilogue): Add new
5423 * config/h8300/jumpcall.md (call, call_value): Restrict to
5424 !SIBLING_CALL_P cases.
5425 (subcall, sibcall_value): New patterns & expanders.
5426 * config/h8300/proepi.md (epilogue): Pass new argument to
5427 h8300_expand_epilogue.
5428 (sibcall_epilogue): New expander.
5429 * config/h8300/h8300.c (h8300_expand_epilogue): Handle sibcall
5431 (h8300_ok_for_sibcall_p): New function.
5432 (TARGET_FUNCTION_OK_FOR_SIBCALL): define.
5434 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5436 * simplify-rtx.c (simplify_unary_operation_1): [TRUNCATE]:
5437 Handle case where the operand is already the desired mode.
5439 2021-08-23 Richard Biener <rguenther@suse.de>
5442 * tree-ssa-structalias.c (ipa_pta_execute): Check in_other_partition
5443 in addition to has_gimple_body.
5445 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
5447 PR middle-end/101949
5448 * ipa-modref.c (analyze_ssa_name_flags): Fix merging of
5451 2021-08-23 Martin Liska <mliska@suse.cz>
5453 * doc/invoke.texi: Put the option out of -mxl-mode-app-model
5456 2021-08-23 Richard Biener <rguenther@suse.de>
5458 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5459 Properly scale the inner loop cost only once.
5461 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5463 * tree-ssa-ccp.c (bit_value_binop) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]:
5464 Provide bounds for unsigned (and signed with non-negative operands)
5465 division and modulus.
5467 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5469 * simplify-rtx.c (simplify_truncation): Generalize simplification
5470 of (truncate:A (subreg:B X)).
5471 (simplify_unary_operation_1) [FLOAT_TRUNCATE, FLOAT_EXTEND,
5472 SIGN_EXTEND, ZERO_EXTEND]: Handle cases where the operand
5473 already has the desired machine mode.
5474 (test_scalar_int_ops): Add tests that useless extensions and
5475 truncations are optimized away.
5476 (test_scalar_int_ext_ops): New self-test function to confirm
5477 that truncations of extensions are correctly simplified.
5478 (test_scalar_int_ext_ops2): New self-test function to check
5479 truncations of truncations, extensions of extensions, and
5480 truncations of extensions.
5481 (test_scalar_ops): Call the above two functions with a
5482 representative sampling of integer machine modes.
5484 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5486 * match.pd (shift transformations): Change the sign of an
5487 LSHIFT_EXPR if it reduces the number of explicit conversions.
5489 2021-08-23 Jakub Jelinek <jakub@redhat.com>
5491 PR tree-optimization/86723
5492 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
5493 cast64_to_32 argument, set *cast64_to_32 to false, unless n is
5494 non-memory permutation of 64-bit src which only has bytes of
5495 0 or [5..8] and n->range is 4.
5496 (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
5497 find_bswap_or_nop_finalize caller, support bswap with some bytes
5498 zeroed, as long as at least two bytes are not zeroed.
5499 (bswap_replace): Add mask argument and handle masking of bswap
5501 (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
5502 caller, punt if cast64_to_32 or mask is not all ones.
5503 (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
5504 caller, for now punt if cast64_to_32.
5506 2021-08-23 Richard Biener <rguenther@suse.de>
5508 PR tree-optimization/79334
5509 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Record
5510 a type also for COMPONENT_REFs.
5511 (vn_reference_may_trap): Check ARRAY_REF with constant index
5512 against the array domain.
5514 2021-08-23 liuhongt <hongtao.liu@intel.com>
5517 * config/i386/sse.md (*avx512f_pshufb_truncv8hiv8qi_1): Add
5518 TARGET_AVX512BW to condition.
5520 2021-08-23 Jakub Jelinek <jakub@redhat.com>
5523 * dwarf2out.c (gen_variable_die): Add DW_AT_location for global
5524 register variables already during early_dwarf if possible.
5526 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
5528 * config/arm/arm_mve.h: Fix __arm_vctp16q return type.
5530 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
5533 * config/arm/arm.opt: Fix typo.
5534 * config/arm/t-rmprofile: Fix typo.
5536 2021-08-23 Jakub Jelinek <jakub@redhat.com>
5538 * tree.h (OMP_CLAUSE_GRAINSIZE_STRICT): Define.
5539 (OMP_CLAUSE_NUM_TASKS_STRICT): Define.
5540 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_GRAINSIZE,
5541 case OMP_CLAUSE_NUM_TASKS>: Print strict: modifier.
5542 * omp-expand.c (expand_task_call): Use GOMP_TASK_FLAG_STRICT in iflags
5543 if either grainsize or num_tasks clause has the strict modifier.
5545 2021-08-23 Martin Liska <mliska@suse.cz>
5547 * dbgcnt.def (DEBUG_COUNTER): New counter.
5548 * gimple.c (gimple_call_arg_flags): Use it in IPA PTA.
5550 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
5552 * ipa-modref.c (analyze_ssa_name_flags): Improve handling of return slot.
5554 2021-08-23 Xi Ruoyao <xry111@mengyan1223.wang>
5557 * config/mips/mips-protos.h (mips_msa_output_shift_immediate):
5559 * config/mips/mips.c (mips_msa_output_shift_immediate): New
5561 * config/mips/mips-msa.md (vashl<mode>3, vashr<mode>3,
5562 vlshr<mode>3): Call it.
5564 2021-08-22 Jan Hubicka <hubicka@ucw.cz>
5565 Martin Liska <mliska@suse.cz>
5567 PR middle-end/101949
5568 * ipa-modref.c (analyze_ssa_name_flags): Indirect call implies
5571 2021-08-21 Dragan Mladjenovic <OT_Dragan.Mladjenovic@mediatek.com>
5573 * config/mips/mips.c (mips_function_rodata_section,
5574 TARGET_ASM_FUNCTION_RODATA_SECTION): Removed.
5576 2021-08-21 John David Anglin <danglin@gcc.gnu.org>
5578 * config/pa/pa.c (pa_asm_output_aligned_common): Remove warning.
5580 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5582 * configure.ac (thread-local storage support): Remove tls_first_major
5583 and tls_first_minor. Use "$conftest_s" to check support.
5584 * configure: Regenerate.
5586 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5588 * configure.ac: Fixup formatting.
5590 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5592 * acinclude.m4 (gcc_GAS_CHECK_FEATURE): Remove third argument and ...
5593 * configure.ac: ... update all callers.
5595 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5598 * acinclude.m4 (_gcc_COMPUTE_GAS_VERSION, _gcc_GAS_VERSION_GTE_IFELSE)
5599 (gcc_GAS_VERSION_GTE_IFELSE): Remove.
5600 (gcc_GAS_CHECK_FEATURE): Do not handle in-tree case specially.
5601 * configure.ac: Remove gcc_cv_gas_major_version, gcc_cv_gas_minor_version.
5602 Remove remaining checks for in-tree assembler.
5603 * configure: Regenerate.
5605 2021-08-20 Jeff Law <jlaw@localhost.localdomain>
5607 * config/h8300/h8300.c (shift_alg_hi): Improve arithmetic shift right
5608 by 15 bits for H8/300H and H8/S. Improve logical shifts by 12
5610 (shift_alg_si): Improve arithmetic right shift by 28-30 bits for
5611 H8/300H. Improve arithmetic shift right by 15 bits for H8/S.
5612 Improve logical shifts by 27 bits for H8/S.
5613 (get_shift_alg): Corresponding changes.
5614 (h8300_option_override): Revert to loops for -Os when profitable.
5616 2021-08-20 Richard Biener <rguenther@suse.de>
5618 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
5620 (vect_analyze_data_ref_accesses): Likewise. Assign the BB
5621 index as group_id when dataref_groups were not computed.
5622 * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
5623 we advace to the next BB.
5625 2021-08-20 Jakub Jelinek <jakub@redhat.com>
5627 * omp-builtins.def (BUILT_IN_GOMP_WARNING, BUILT_IN_GOMP_ERROR): New
5630 2021-08-20 Martin Liska <mliska@suse.cz>
5632 PR gcov-profile/89961
5633 * gcov.c (make_gcov_file_name): Rewrite using std::string.
5634 (mangle_name): Simplify, do not used the second argument.
5635 (strip_extention): New function.
5636 (get_md5sum): Likewise.
5637 (get_gcov_intermediate_filename): Handle properly -p and -x
5639 (output_gcov_file): Use string type.
5640 (generate_results): Likewise.
5641 (md5sum_to_hex): Remove.
5643 2021-08-20 Michael Meissner <meissner@linux.ibm.com>
5645 * config/rs6000/altivec.md (UNSPEC_XXEVAL): Move to vsx.md.
5646 (UNSPEC_XXSPLTIW): Move to vsx.md.
5647 (UNSPEC_XXSPLTID): Move to vsx.md.
5648 (UNSPEC_XXSPLTI32DX): Move to vsx.md.
5649 (UNSPEC_XXBLEND): Move to vsx.md.
5650 (UNSPEC_XXPERMX): Move to vsx.md.
5651 (VM3): Move to vsx.md.
5652 (VM3_char): Move to vsx.md.
5653 (xxspltiw_v4si): Move to vsx.md.
5654 (xxspltiw_v4sf): Move to vsx.md.
5655 (xxspltiw_v4sf_inst): Move to vsx.md.
5656 (xxspltidp_v2df): Move to vsx.md.
5657 (xxspltidp_v2df_inst): Move to vsx.md.
5658 (xxsplti32dx_v4si_inst): Move to vsx.md.
5659 (xxsplti32dx_v4sf): Move to vsx.md.
5660 (xxsplti32dx_v4sf_inst): Move to vsx.md.
5661 (xxblend_<mode>): Move to vsx.md.
5662 (xxpermx): Move to vsx.md.
5663 (xxpermx_inst): Move to vsx.md.
5664 * config/rs6000/vsx.md (UNSPEC_XXEVAL): Move from altivec.md.
5665 (UNSPEC_XXSPLTIW): Move from altivec.md.
5666 (UNSPEC_XXSPLTID): Move from altivec.md.
5667 (UNSPEC_XXSPLTI32DX): Move from altivec.md.
5668 (UNSPEC_XXBLEND): Move from altivec.md.
5669 (UNSPEC_XXPERMX): Move from altivec.md.
5670 (VM3): Move from altivec.md.
5671 (VM3_char): Move from altivec.md.
5672 (xxspltiw_v4si): Move from altivec.md.
5673 (xxspltiw_v4sf): Move from altivec.md.
5674 (xxspltiw_v4sf_inst): Move from altivec.md.
5675 (xxspltidp_v2df): Move from altivec.md.
5676 (xxspltidp_v2df_inst): Move from altivec.md.
5677 (xxsplti32dx_v4si_inst): Move from altivec.md.
5678 (xxsplti32dx_v4sf): Move from altivec.md.
5679 (xxsplti32dx_v4sf_inst): Move from altivec.md.
5680 (xxblend_<mode>): Move from altivec.md.
5681 (xxpermx): Move from altivec.md.
5682 (xxpermx_inst): Move from altivec.md.
5684 2021-08-19 Roger Sayle <roger@nextmovesoftware.com>
5686 * tree-vect-generic.c (expand_vector_operations_1): Use either
5687 gimplify_build1 or gimplify_build2 instead of gimple_build_assign
5688 when constructing scalar splat expressions.
5690 2021-08-19 Peter Bergner <bergner@linux.ibm.com>
5693 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Cast
5694 pointer to __vector_pair *.
5696 2021-08-19 Martin Sebor <msebor@redhat.com>
5698 * gimple-range.cc: Add comments.
5699 * gimple-range.h: Same.
5701 2021-08-19 Martin Sebor <msebor@redhat.com>
5703 PR middle-end/101984
5704 * gimple-ssa-warn-access.cc (pass_waccess::execute): Also call
5707 2021-08-19 Jeff Law <jlaw@localhost.localdomain>
5709 * config.gcc (h8300-*-elf*): Do not include dbxelf.h.
5710 (h8300-*-linux*, v850-*-rtems*, v850*-elf*): Likewise.
5711 * config/v850/v850.h (DEFAULT_GDB_EXTENSIONS): Remove.
5713 2021-08-19 Jakub Jelinek <jakub@redhat.com>
5715 PR middle-end/101950
5716 * optabs.c (expand_clrsb_using_clz): New function.
5717 (expand_unop): Use it as another clrsb expansion fallback.
5719 2021-08-19 liuhongt <hongtao.liu@intel.com>
5722 2021-07-28 liuhongt <hongtao.liu@intel.com>
5725 * config/i386/i386.h (processor_costs): Add new member
5727 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
5728 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
5729 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
5730 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
5731 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
5732 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
5733 generic_cost, core_cost): Initialize integer_to_sse same value
5735 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
5736 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5737 Use integer_to_sse instead of sse_op to calculate the cost of
5740 2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
5742 * config.gcc: Include rpath.opt for Darwin.
5743 * config/darwin.h (DRIVER_SELF_SPECS): Handle -rpath.
5745 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
5748 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
5751 2021-08-18 Jonathan Wright <jonathan.wright@arm.com>
5753 * config/aarch64/arm_neon.h (vld3_lane_f64): Use float RTL
5754 pattern and type cast.
5755 (vld4_lane_f32): Use float RTL pattern.
5756 (vld4q_lane_f64): Use float type cast.
5758 2021-08-18 Jan Hubicka <hubicka@ucw.cz>
5760 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Check also
5763 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
5765 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor): Extend.
5766 (test_map_of_type_with_ctor_and_dtor_expand): Add function.
5767 (hash_map_tests_c_tests): Call it.
5769 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
5771 * ggc.h (enum ggc_collect): New.
5772 (ggc_collect): Use it.
5773 * ggc-page.c: Adjust.
5774 * ggc-common.c: Likewise.
5775 * ggc-tests.c: Likewise.
5776 * read-rtl-function.c: Likewise.
5777 * selftest-run-tests.c: Likewise.
5778 * doc/gty.texi (Invoking the garbage collector): Likewise.
5780 2021-08-18 liuhongt <hongtao.liu@intel.com>
5783 * config/i386/i386.h (TARGET_V2DF_REDUCTION_PREFER_HADDPD):
5785 * config/i386/sse.md (*sse3_haddv2df3_low): Add
5786 TARGET_V2DF_REDUCTION_PREFER_HADDPD.
5787 (*sse3_hsubv2df3_low): Ditto.
5788 * config/i386/x86-tune.def
5789 (X86_TUNE_V2DF_REDUCTION_PREFER_HADDPD): New tune.
5791 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
5793 * gimple-range-gori.cc (gori_compute::gori_compute): Enable tracing.
5794 (gori_compute::compute_operand_range): Add tracing.
5795 (gori_compute::logical_combine): Ditto.
5796 (gori_compute::compute_logical_operands): Ditto.
5797 (gori_compute::compute_operand1_range): Ditto.
5798 (gori_compute::compute_operand2_range): Ditto.
5799 (gori_compute::outgoing_edge_range_p): Ditto.
5800 * gimple-range-gori.h (class gori_compute): Add range_tracer.
5802 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
5804 * flag-types.h (enum evrp_mode): Adjust evrp-mode values.
5805 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Relocate from.
5806 * gimple-range-trace.h (DEBUG_RANGE_CACHE): Here.
5807 * params.opt (--param=evrp-mode): Adjust options.
5809 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
5811 * Makefile.in (OBJS): Add gimple-range-trace.o.
5812 * gimple-range-cache.h (enable_new_values): Remove unused prototype.
5813 * gimple-range-fold.cc: Adjust headers.
5814 * gimple-range-trace.cc: New.
5815 * gimple-range-trace.h: New.
5816 * gimple-range.cc (gimple_ranger::gimple_ranger): Enable tracer.
5817 (gimple_ranger::range_of_expr): Add tracing.
5818 (gimple_ranger::range_on_entry): Ditto.
5819 (gimple_ranger::range_on_exit): Ditto.
5820 (gimple_ranger::range_on_edge): Ditto.
5821 (gimple_ranger::fold_range_internal): Ditto.
5822 (gimple_ranger::dump_bb): Do not calculate edge range twice.
5823 (trace_ranger::*): Remove.
5824 (enable_ranger): Never create a trace_ranger.
5825 (debug_seed_ranger): Move to gimple-range-trace.cc.
5826 (dump_ranger): Ditto.
5827 (debug_ranger): Ditto.
5828 * gimple-range.h: Include gimple-range-trace.h.
5829 (range_on_entry, range_on_exit): No longer virtual.
5830 (class trace_ranger): Remove.
5831 (DEBUG_RANGE_CACHE): Move to gimple-range-trace.h.
5833 2021-08-17 Martin Sebor <msebor@redhat.com>
5835 PR middle-end/101854
5836 * builtins.c (expand_builtin_alloca): Move warning code to check_alloca
5837 in gimple-ssa-warn-access.cc.
5838 * calls.c (alloc_max_size): Move code to check_alloca.
5839 (get_size_range): Move to pointer-query.cc.
5840 (maybe_warn_alloc_args_overflow): Move to gimple-ssa-warn-access.cc.
5841 (get_attr_nonstring_decl): Move to tree.c.
5842 (fntype_argno_type): Move to gimple-ssa-warn-access.cc.
5843 (append_attrname): Same.
5844 (maybe_warn_rdwr_sizes): Same.
5845 (initialize_argument_information): Move code to
5846 gimple-ssa-warn-access.cc.
5847 * calls.h (maybe_warn_alloc_args_overflow): Move to
5848 gimple-ssa-warn-access.h.
5849 (get_attr_nonstring_decl): Move to tree.h.
5850 (maybe_warn_nonstring_arg): Move to gimple-ssa-warn-access.h.
5851 (enum size_range_flags): Move to pointer-query.h.
5852 (get_size_range): Same.
5853 * gimple-ssa-warn-access.cc (has_location): Remove unused overload
5854 to avoid Clang -Wunused-function.
5855 (get_size_range): Declare static.
5856 (maybe_emit_free_warning): Rename...
5857 (maybe_check_dealloc_call): ...to this for consistency.
5858 (class pass_waccess): Add members.
5859 (pass_waccess::~pass_waccess): Defined.
5860 (alloc_max_size): Move here from calls.c.
5861 (maybe_warn_alloc_args_overflow): Same.
5862 (check_alloca): New function.
5863 (check_alloc_size_call): New function.
5864 (check_strncat): Handle another warning flag.
5865 (pass_waccess::check_builtin): Handle alloca.
5866 (fntype_argno_type): Move here from calls.c.
5867 (append_attrname): Same.
5868 (maybe_warn_rdwr_sizes): Same.
5869 (pass_waccess::check_call): Define.
5870 (check_nonstring_args): New function.
5871 (pass_waccess::check): Call new member functions.
5872 (pass_waccess::execute): Enable ranger.
5873 * gimple-ssa-warn-access.h (get_size_range): Move here from calls.h.
5874 (maybe_warn_nonstring_arg): Same.
5875 * gimple-ssa-warn-restrict.c: Remove #include.
5876 * pointer-query.cc (get_size_range): Move here from calls.c.
5877 * pointer-query.h (enum size_range_flags): Same.
5878 (get_size_range): Same.
5879 * tree.c (get_attr_nonstring_decl): Move here from calls.c.
5880 * tree.h (get_attr_nonstring_decl): Move here from calls.h.
5882 2021-08-17 Thomas Schwinge <thomas@codesourcery.com>
5884 * ggc.h (ggc_collect): Add 'force_collect' parameter.
5885 * ggc-page.c (ggc_collect): Use that one instead of global
5886 'ggc_force_collect'. Adjust all users.
5887 * doc/gty.texi (Invoking the garbage collector): Update.
5888 * ggc-internal.h (ggc_force_collect): Remove.
5889 * ggc-common.c (ggc_force_collect): Likewise.
5890 * selftest.h (forcibly_ggc_collect): Remove.
5891 * ggc-tests.c (selftest::forcibly_ggc_collect): Likewise.
5892 * read-rtl-function.c (test_loading_labels): Adjust.
5893 * selftest-run-tests.c (run_tests): Likewise.
5895 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
5897 * config/darwin.c (darwin_file_end): Reset and reclaim the
5898 section names table at the end of compile.
5900 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
5903 * config.in: Regenerate.
5904 * config/i386/darwin.h (EXTRA_ASM_OPTS): New
5905 (ASM_SPEC): Pass options to disable branch shortening where
5907 * configure: Regenerate.
5908 * configure.ac: Detect versions of 'as' that support the
5909 optimisation which has the bug.
5911 2021-08-17 Richard Biener <rguenther@suse.de>
5913 * optabs-query.c (supports_vec_gather_load_p): Also check
5915 (supports_vec_scatter_store_p): Likewise.
5916 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Fall
5917 back to masked variants if non-masked are not supported.
5918 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
5919 When we need to use masked gather/scatter but do not have
5920 a mask set up a constant true one.
5921 * tree-vect-stmts.c (vect_check_scalar_mask): Also allow
5924 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
5926 * tree-ssa-ccp.c (bit_value_binop) [MINUS_EXPR]: Use same
5927 algorithm as PLUS_EXPR to improve subtraction bit bounds.
5928 [POINTER_DIFF_EXPR]: Treat as synonymous with MINUS_EXPR.
5930 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
5932 * tree-ssa-ccp.c (bit_value_mult_const): New helper function to
5933 calculate the mask-value pair result of a multiplication by an
5935 (bit_value_binop) [MULT_EXPR]: Call it from here for
5936 multiplications by (sparse) non-negative constants.
5938 2021-08-17 Christophe Lyon <christophe.lyon@foss.st.com>
5941 * config.gcc (gcc_cv_initfini_array): Leave undefined for
5942 uclinuxfdpiceabi targets.
5944 2021-08-17 Alexandre Oliva <oliva@adacore.com>
5946 * tree-inline.c (maybe_move_debug_stmts_to_successors): Don't
5947 reverse debug stmts.
5949 2021-08-17 Alexandre Oliva <oliva@adacore.com>
5951 * tree-cfg.c (dump_function_to_file): Use fun, not cfun.
5953 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
5955 * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete.
5956 (__LD4Q_LANE_FUNC): Likewise.
5957 (vld4_lane_u8): Define without macro.
5958 (vld4_lane_u16): Likewise.
5959 (vld4_lane_u32): Likewise.
5960 (vld4_lane_u64): Likewise.
5961 (vld4_lane_s8): Likewise.
5962 (vld4_lane_s16): Likewise.
5963 (vld4_lane_s32): Likewise.
5964 (vld4_lane_s64): Likewise.
5965 (vld4_lane_f16): Likewise.
5966 (vld4_lane_f32): Likewise.
5967 (vld4_lane_f64): Likewise.
5968 (vld4_lane_p8): Likewise.
5969 (vld4_lane_p16): Likewise.
5970 (vld4_lane_p64): Likewise.
5971 (vld4q_lane_u8): Likewise.
5972 (vld4q_lane_u16): Likewise.
5973 (vld4q_lane_u32): Likewise.
5974 (vld4q_lane_u64): Likewise.
5975 (vld4q_lane_s8): Likewise.
5976 (vld4q_lane_s16): Likewise.
5977 (vld4q_lane_s32): Likewise.
5978 (vld4q_lane_s64): Likewise.
5979 (vld4q_lane_f16): Likewise.
5980 (vld4q_lane_f32): Likewise.
5981 (vld4q_lane_f64): Likewise.
5982 (vld4q_lane_p8): Likewise.
5983 (vld4q_lane_p16): Likewise.
5984 (vld4q_lane_p64): Likewise.
5985 (vld4_lane_bf16): Likewise.
5986 (vld4q_lane_bf16): Likewise.
5988 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
5990 * config/aarch64/arm_neon.h (__LD3_LANE_FUNC): Delete.
5991 (__LD3Q_LANE_FUNC): Delete.
5992 (vld3_lane_u8): Define without macro.
5993 (vld3_lane_u16): Likewise.
5994 (vld3_lane_u32): Likewise.
5995 (vld3_lane_u64): Likewise.
5996 (vld3_lane_s8): Likewise.
5997 (vld3_lane_s16): Likewise.
5998 (vld3_lane_s32): Likewise.
5999 (vld3_lane_s64): Likewise.
6000 (vld3_lane_f16): Likewise.
6001 (vld3_lane_f32): Likewise.
6002 (vld3_lane_f64): Likewise.
6003 (vld3_lane_p8): Likewise.
6004 (vld3_lane_p16): Likewise.
6005 (vld3_lane_p64): Likewise.
6006 (vld3q_lane_u8): Likewise.
6007 (vld3q_lane_u16): Likewise.
6008 (vld3q_lane_u32): Likewise.
6009 (vld3q_lane_u64): Likewise.
6010 (vld3q_lane_s8): Likewise.
6011 (vld3q_lane_s16): Likewise.
6012 (vld3q_lane_s32): Likewise.
6013 (vld3q_lane_s64): Likewise.
6014 (vld3q_lane_f16): Likewise.
6015 (vld3q_lane_f32): Likewise.
6016 (vld3q_lane_f64): Likewise.
6017 (vld3q_lane_p8): Likewise.
6018 (vld3q_lane_p16): Likewise.
6019 (vld3q_lane_p64): Likewise.
6020 (vld3_lane_bf16): Likewise.
6021 (vld3q_lane_bf16): Likewise.
6023 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
6025 * config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Delete.
6026 (__LD2Q_LANE_FUNC): Likewise.
6027 (vld2_lane_u8): Define without macro.
6028 (vld2_lane_u16): Likewise.
6029 (vld2_lane_u32): Likewise.
6030 (vld2_lane_u64): Likewise.
6031 (vld2_lane_s8): Likewise.
6032 (vld2_lane_s16): Likewise.
6033 (vld2_lane_s32): Likewise.
6034 (vld2_lane_s64): Likewise.
6035 (vld2_lane_f16): Likewise.
6036 (vld2_lane_f32): Likewise.
6037 (vld2_lane_f64): Likewise.
6038 (vld2_lane_p8): Likewise.
6039 (vld2_lane_p16): Likewise.
6040 (vld2_lane_p64): Likewise.
6041 (vld2q_lane_u8): Likewise.
6042 (vld2q_lane_u16): Likewise.
6043 (vld2q_lane_u32): Likewise.
6044 (vld2q_lane_u64): Likewise.
6045 (vld2q_lane_s8): Likewise.
6046 (vld2q_lane_s16): Likewise.
6047 (vld2q_lane_s32): Likewise.
6048 (vld2q_lane_s64): Likewise.
6049 (vld2q_lane_f16): Likewise.
6050 (vld2q_lane_f32): Likewise.
6051 (vld2q_lane_f64): Likewise.
6052 (vld2q_lane_p8): Likewise.
6053 (vld2q_lane_p16): Likewise.
6054 (vld2q_lane_p64): Likewise.
6055 (vld2_lane_bf16): Likewise.
6056 (vld2q_lane_bf16): Likewise.
6058 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
6060 * haifa-sched.c (advance_one_cycle): Output more context-synchronization
6063 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
6065 * haifa-sched.c (enum rfs_decision, rfs_str): Add RFS_AUTOPREF.
6066 (rank_for_schedule): Use it.
6068 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
6070 PR rtl-optimization/91598
6071 * haifa-sched.c (autopref_rank_for_schedule): Prioritize "irrelevant"
6072 insns after memory reads and before memory writes.
6074 2021-08-17 Alistair_Lee <alistair.lee@arm.com>
6076 * rtl.h (CONST_VECTOR_P): New macro.
6077 * config/aarch64/aarch64.c (aarch64_get_sve_pred_bits): Use RTL
6078 code testing macros.
6079 (aarch64_ptrue_all_mode): Likewise.
6080 (aarch64_expand_mov_immediate): Likewise.
6081 (aarch64_const_vec_all_in_range_p): Likewise.
6082 (aarch64_rtx_costs): Likewise.
6083 (aarch64_legitimate_constant_p): Likewise.
6084 (aarch64_simd_valid_immediate): Likewise.
6085 (aarch64_simd_make_constant): Likewise.
6086 (aarch64_convert_mult_to_shift): Likewise.
6087 (aarch64_expand_sve_vec_perm): Likewise.
6088 (aarch64_vec_fpconst_pow_of_2): Likewise.
6090 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
6092 PR tree-optimization/101938
6093 * range-op.cc (operator_abs::op1_range): Special case
6094 -TYPE_MIN_VALUE for flag_wrapv.
6096 2021-08-17 Kewen Lin <linkw@linux.ibm.com>
6098 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Add the cost for
6101 2021-08-17 Jakub Jelinek <jakub@redhat.com>
6103 * tree.def (OMP_SCOPE): New tree code.
6104 * tree.h (OMP_SCOPE_BODY, OMP_SCOPE_CLAUSES): Define.
6105 * tree-nested.c (convert_nonlocal_reference_stmt,
6106 convert_local_reference_stmt, convert_gimple_call): Handle
6108 * tree-pretty-print.c (dump_generic_node): Handle OMP_SCOPE.
6109 * gimple.def (GIMPLE_OMP_SCOPE): New gimple code.
6110 * gimple.c (gimple_build_omp_scope): New function.
6111 (gimple_copy): Handle GIMPLE_OMP_SCOPE.
6112 * gimple.h (gimple_build_omp_scope): Declare.
6113 (gimple_has_substatements): Handle GIMPLE_OMP_SCOPE.
6114 (gimple_omp_scope_clauses, gimple_omp_scope_clauses_ptr,
6115 gimple_omp_scope_set_clauses): New inline functions.
6116 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_SCOPE.
6117 * gimple-pretty-print.c (dump_gimple_omp_scope): New function.
6118 (pp_gimple_stmt_1): Handle GIMPLE_OMP_SCOPE.
6119 * gimple-walk.c (walk_gimple_stmt): Likewise.
6120 * gimple-low.c (lower_stmt): Likewise.
6121 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
6122 (gimplify_scan_omp_clauses): For task reductions, handle OMP_SCOPE
6123 like ORT_WORKSHARE constructs. Adjust diagnostics for %<scope%>
6124 allowing task reductions. Reject inscan reductions on scope.
6125 (omp_find_stores_stmt): Handle GIMPLE_OMP_SCOPE.
6126 (gimplify_omp_workshare, gimplify_expr): Handle OMP_SCOPE.
6127 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_SCOPE.
6128 (estimate_num_insns): Likewise.
6129 * omp-low.c (build_outer_var_ref): Look through GIMPLE_OMP_SCOPE
6130 contexts if var isn't privatized there.
6131 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_SCOPE.
6132 (scan_omp_1_stmt): Likewise.
6133 (maybe_add_implicit_barrier_cancel): Look through outer
6135 (lower_omp_scope): New function.
6136 (lower_omp_task_reductions): Handle OMP_SCOPE.
6137 (lower_omp_1): Handle GIMPLE_OMP_SCOPE.
6138 (diagnose_sb_1, diagnose_sb_2): Likewise.
6139 * omp-expand.c (expand_omp_single): Support also GIMPLE_OMP_SCOPE.
6140 (expand_omp): Handle GIMPLE_OMP_SCOPE.
6141 (omp_make_gimple_edges): Likewise.
6142 * omp-builtins.def (BUILT_IN_GOMP_SCOPE_START): New built-in.
6144 2021-08-17 Richard Biener <rguenther@suse.de>
6146 PR tree-optimization/101925
6147 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Set
6148 reverse on COMPONENT_REF and ARRAY_REF according to
6149 what reverse_storage_order_for_component_p does.
6150 (vn_reference_eq): Compare reversed on reference ops.
6151 (reverse_storage_order_for_component_p): New overload.
6152 (vn_reference_lookup_3): Check reverse_storage_order_for_component_p
6153 on the reference looked up.
6155 2021-08-17 Jeff Law <jlaw@localhost.localdomain>
6157 * config/h8300/h8300.c (shift_alg_si): Avoid loops for most SImode
6159 (h8300_option_override): Use loops on H8/S more often when optimizing
6161 (get_shift_alg): Handle new "special" cases on H8/S. Simplify
6162 accordingly. Handle various arithmetic right shifts with special
6163 sequences that we couldn't handle before.
6165 2021-08-16 Jeff Law <jlaw@localhost.localdomain>
6167 * config.gcc (rl78-*-elf*): Do not include dbxelf.h.
6169 2021-08-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
6171 * config/sparc/rtemself.h (SPARC_GCOV_TYPE_SIZE): Define.
6172 * config/sparc/sparc.c (sparc_gcov_type_size): New.
6173 (TARGET_GCOV_TYPE_SIZE): Redefine if SPARC_GCOV_TYPE_SIZE is defined.
6174 * coverage.c (get_gcov_type): Use targetm.gcov_type_size().
6175 * doc/tm.texi (TARGET_GCOV_TYPE_SIZE): Add hook under "Misc".
6176 * doc/tm.texi.in: Regenerate.
6177 * target.def (gcov_type_size): New target hook.
6178 * targhooks.c (default_gcov_type_size): New.
6179 * targhooks.h (default_gcov_type_size): Declare.
6180 * tree-profile.c (gimple_gen_edge_profiler): Use precision of
6182 (gimple_gen_time_profiler): Likewise.
6184 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
6186 * dwarf2out.c (add_scalar_info): Deal with DW_AT_data_bit_offset.
6188 2021-08-16 Tobias Burnus <tobias@codesourcery.com>
6190 PR middle-end/101931
6191 * omp-low.c (omp_runtime_api_call): Update for routines
6192 added in the meanwhile.
6194 2021-08-16 Martin Liska <mliska@suse.cz>
6196 PR tree-optimization/100393
6197 * tree-switch-conversion.c (group_cluster::dump): Use
6198 get_comparison_count.
6199 (jump_table_cluster::find_jump_tables): Pre-compute number of
6200 comparisons and then decrement it. Cache also max_ratio.
6201 (jump_table_cluster::can_be_handled): Change signature.
6202 * tree-switch-conversion.h (get_comparison_count): New.
6204 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
6206 * dwarf2out.c (add_data_member_location_attribute): Use GNAT
6207 encodings only when -fgnat-encodings=all is specified.
6208 (add_bound_info): Likewise.
6209 (add_byte_size_attribute): Likewise.
6210 (gen_member_die): Likewise.
6212 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
6214 * omp-oacc-neuter-broadcast.cc
6215 (execute_omp_oacc_neuter_broadcast): Plug 'par' memory leak.
6217 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
6219 * omp-oacc-neuter-broadcast.cc
6220 (execute_omp_oacc_neuter_broadcast): Clarify memory management for
6223 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
6225 * omp-oacc-neuter-broadcast.cc (field_map): Move variable into...
6226 (execute_omp_oacc_neuter_broadcast): ... here.
6227 (install_var_field, build_receiver_ref, build_sender_ref): Take
6228 'field_map_t *' parameter. Adjust all users.
6229 (worker_single_copy, neuter_worker_single): Take a
6230 'record_field_map_t *' parameter. Adjust all users.
6232 2021-08-16 liuhongt <hongtao.liu@intel.com>
6235 * config/i386/i386.md (ldexp<mode>3): Force operands[1] to
6238 2021-08-16 Martin Liska <mliska@suse.cz>
6241 * multiple_target.c (create_dispatcher_calls): Make default
6242 function local only if it is a definition.
6244 2021-08-16 Martin Liska <mliska@suse.cz>
6247 * ipa-icf-gimple.c (func_checker::compare_ssa_name): Do not
6248 consider equal SSA_NAMEs when one is a param.
6250 2021-08-16 liuhongt <hongtao.liu@intel.com>
6253 * config/i386/i386-expand.c (ix86_expand_vec_perm_vpermt2):
6254 Support vpermi2b for V32QI/V16QImode.
6255 (ix86_extract_perm_from_pool_constant): New function.
6256 (ix86_expand_vec_one_operand_perm_avx512): Support
6257 vpermw/vpermb under TARGET_AVX512BW/TARGET_AVX512VBMI.
6258 (expand_vec_perm_1): Adjust comments for upper.
6259 * config/i386/i386-protos.h (ix86_extract_perm_from_pool_constant):
6261 * config/i386/predicates.md (permvar_truncate_operand): New predicate.
6262 (pshufb_truncv4siv4hi_operand): Ditto.
6263 (pshufb_truncv8hiv8qi_operand): Ditto.
6264 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1):
6265 New pre_reload define_insn_and_split.
6266 (*avx512f_permvar_truncv8siv8hi_1): Ditto.
6267 (*avx512f_vpermvar_truncv8div8si_1): Ditto.
6268 (*avx512f_permvar_truncv32hiv32qi_1): Ditto.
6269 (*avx512f_permvar_truncv16hiv16qi_1): Ditto.
6270 (*avx512f_permvar_truncv4div4si_1): Ditto.
6271 (*avx512f_pshufb_truncv8hiv8qi_1): Ditto.
6272 (*avx512f_pshufb_truncv4siv4hi_1): Ditto.
6273 (*avx512f_pshufd_truncv2div2si_1): Ditto.
6275 2021-08-16 Kito Cheng <kito.cheng@sifive.com>
6277 * config/riscv/multilib-generator: Support code model option for
6279 * doc/install.texi: Add document of new option for
6280 --with-multilib-generator.
6282 2021-08-15 Clément Chigot <clement.chigot@atos.net>
6284 * config/rs6000/rs6000.c (xcoff_tls_exec_model_detected): New.
6285 (rs6000_legitimize_tls_address_aix): Use it.
6286 (rs6000_xcoff_file_end): Add ".ref __tls_get_addr" when
6287 xcoff_tls_exec_model_detected is true.
6289 2021-08-15 Jeff Law <jlaw@localhost.localdomain>
6291 * config/h8300/h8300.c (shift_alg_si): Retune H8/300H shifts
6292 to allow a bit more code growth, saving many dozens of cycles.
6293 (h8300_option_override): Adjus shift_alg_si if optimizing for
6295 (get_shift_alg): Use special + inline shifts for residuals
6298 2021-08-14 Stafford Horne <shorne@gmail.com>
6301 * config/or1k/or1k-opts.h: New file.
6302 * config/or1k/or1k.c (or1k_legitimize_address_1, print_reloc):
6303 Support generating gotha relocations if -mcmodel=large is
6305 * config/or1k/or1k.h (TARGET_CMODEL_SMALL, TARGET_CMODEL_LARGE):
6307 * config/or1k/or1k.opt (mcmodel=): New option.
6308 * doc/invoke.texi (OpenRISC Options): Document mcmodel.
6310 2021-08-14 Martin Sebor <msebor@redhat.com>
6312 PR middle-end/101791
6313 * gimple-ssa-warn-access.cc (new_delete_mismatch_p): Use new argument
6314 to valid_new_delete_pair_p.
6315 * tree.c (valid_new_delete_pair_p): Add argument.
6316 * tree.h (valid_new_delete_pair_p): Same.
6318 2021-08-14 Jakub Jelinek <jakub@redhat.com>
6321 * config/i386/i386-expand.c (expand_vec_perm_broadcast_1)
6322 <case E_V64QImode>: For this mode assert
6323 !TARGET_AVX512BW || d->perm[0] rather than !TARGET_AVX2 || d->perm[0].
6325 2021-08-13 Michael Meissner <meissner@linux.ibm.com>
6328 * config/rs6000/altivec.md (xxeval): Use register_predicate
6329 instead of altivec_register_predicate.
6331 2021-08-13 Martin Sebor <msebor@redhat.com>
6333 PR middle-end/101734
6334 * tree-ssa-uninit.c (maybe_warn_read_write_only): New function.
6335 (maybe_warn_operand): Call it.
6337 2021-08-13 Martin Liska <mliska@suse.cz>
6340 * attribs.c (decl_attributes): Make naked functions "noipa"
6343 2021-08-13 Martin Liska <mliska@suse.cz>
6346 * symtab.c (symtab_node::noninterposable_alias): Do not create
6347 local aliases for target_clone functions as the clonning pass
6350 2021-08-13 Martin Liska <mliska@suse.cz>
6352 * opts.c (LIVE_PATCHING_OPTION): Define.
6353 (control_options_for_live_patching): Use it in error messages.
6355 2021-08-13 Jan Hubicka <hubicka@ucw.cz>
6357 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOREAD.
6358 (implicit_const_eaf_flags, implicit_pure_eaf_flags,
6359 ignore_stores_eaf_flags): New constants.
6360 (remove_useless_eaf_flags): New function.
6361 (eaf_flags_useful_p): Use it.
6362 (deref_flags): Add EAF_NOT_RETURNED if flag is unused;
6364 (modref_lattice::init): Add EAF_NOREAD.
6365 (modref_lattice::add_escape_point): Do not reacord escape point if
6367 (modref_lattice::merge): EAF_NOESCAPE implies EAF_NODIRECTESCAPE;
6368 use remove_useless_eaf_flags.
6369 (modref_lattice::merge_deref): Use ignore_stores_eaf_flags.
6370 (modref_lattice::merge_direct_load): Add EAF_NOREAD
6371 (analyze_ssa_name_flags): Fix handling EAF_NOT_RETURNED
6372 (analyze_parms): Use remove_useless_eaf_flags.
6373 (ipa_merge_modref_summary_after_inlining): Use ignore_stores_eaf_flags.
6374 (modref_merge_call_site_flags): Add caller and ecf_flags parameter;
6375 use remove_useless_eaf_flags.
6376 (modref_propagate_flags_in_scc): Update.
6377 * ipa-modref.h: Turn eaf_flags_t back to char.
6378 * tree-core.h (EAF_NOT_RETURNED): Fix.
6379 (EAF_NOREAD): New constant
6380 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1): Check for
6382 * tree-ssa-structalias.c (handle_rhs_call): Handle new flags.
6383 (handle_pure_call): Likewise.
6385 2021-08-12 Jakub Jelinek <jakub@redhat.com>
6387 * tree.def (OMP_MASKED): New tree code.
6388 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_FILTER.
6389 * tree.h (OMP_MASKED_BODY, OMP_MASKED_CLAUSES, OMP_MASKED_COMBINED,
6390 OMP_CLAUSE_FILTER_EXPR): Define.
6391 * tree.c (omp_clause_num_ops): Add OMP_CLAUSE_FILTER entry.
6392 (omp_clause_code_name): Likewise.
6393 (walk_tree_1): Handle OMP_CLAUSE_FILTER.
6394 * tree-nested.c (convert_nonlocal_omp_clauses,
6395 convert_local_omp_clauses): Handle OMP_CLAUSE_FILTER.
6396 (convert_nonlocal_reference_stmt, convert_local_reference_stmt,
6397 convert_gimple_call): Handle GIMPLE_OMP_MASTER.
6398 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_FILTER.
6399 (dump_generic_node): Handle OMP_MASTER.
6400 * gimple.def (GIMPLE_OMP_MASKED): New gimple code.
6401 * gimple.c (gimple_build_omp_masked): New function.
6402 (gimple_copy): Handle GIMPLE_OMP_MASKED.
6403 * gimple.h (gimple_build_omp_masked): Declare.
6404 (gimple_has_substatements): Handle GIMPLE_OMP_MASKED.
6405 (gimple_omp_masked_clauses, gimple_omp_masked_clauses_ptr,
6406 gimple_omp_masked_set_clauses): New inline functions.
6407 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_MASKED.
6408 * gimple-pretty-print.c (dump_gimple_omp_masked): New function.
6409 (pp_gimple_stmt_1): Handle GIMPLE_OMP_MASKED.
6410 * gimple-walk.c (walk_gimple_stmt): Likewise.
6411 * gimple-low.c (lower_stmt): Likewise.
6412 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
6413 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_FILTER. For clauses
6414 that take one expression rather than decl or constant, force
6415 gimplification of that into a SSA_NAME or temporary unless min
6417 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_FILTER.
6418 (gimplify_expr): Handle OMP_MASKED.
6419 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_MASKED.
6420 (estimate_num_insns): Likewise.
6421 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_FILTER.
6422 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_MASKED. Adjust
6423 diagnostics for existence of masked construct.
6424 (scan_omp_1_stmt, lower_omp_master, lower_omp_1, diagnose_sb_1,
6425 diagnose_sb_2): Handle GIMPLE_OMP_MASKED.
6426 * omp-expand.c (expand_omp_synch, expand_omp, omp_make_gimple_edges):
6429 2021-08-12 Uroš Bizjak <ubizjak@gmail.com>
6432 * config/i386/i386.md (avx512f_scalef<mode>2): New insn pattern.
6433 (ldexp<mode>3): Use avx512f_scalef<mode>2.
6434 (UNSPEC_SCALEF): Move from ...
6435 * config/i386/sse.md (UNSPEC_SCALEF): ... here.
6437 2021-08-12 Jan Hubicka <hubicka@ucw.cz>
6439 * ipa-split.c (consider_split): Fix condition testing void functions.
6441 2021-08-12 Aldy Hernandez <aldyh@redhat.com>
6443 * doc/invoke.texi: Remove docs for threader-mode param.
6444 * flag-types.h (enum threader_mode): Remove.
6445 * params.opt: Remove threader-mode param.
6446 * tree-ssa-threadbackward.c (class back_threader): Remove
6447 path_is_unreachable_p.
6448 Make find_paths private.
6449 Add maybe_thread and thread_through_all_blocks.
6450 Remove reference marker for m_registry.
6451 Remove reference marker for m_profit.
6452 (back_threader::back_threader): Adjust for registry and profit not
6454 (dump_path): Move down.
6456 (class thread_jumps): Remove.
6457 (class back_threader_registry): Remove m_all_paths.
6459 (thread_jumps::thread_through_all_blocks): Move to back_threader
6461 (fsm_find_thread_path): Remove
6462 (back_threader::maybe_thread): New.
6463 (back_threader::thread_through_all_blocks): Move from
6465 (back_threader_registry::back_threader_registry): Remove
6467 (back_threader_registry::~back_threader_registry): Remove.
6468 (thread_jumps::find_taken_edge): Remove.
6469 (thread_jumps::check_subpath_and_update_thread_path): Remove.
6470 (thread_jumps::maybe_register_path): Remove.
6471 (thread_jumps::handle_phi): Remove.
6472 (handle_assignment_p): Remove.
6473 (thread_jumps::handle_assignment): Remove.
6474 (thread_jumps::fsm_find_control_statement_thread_paths): Remove.
6475 (thread_jumps::find_jump_threads_backwards): Remove.
6476 (thread_jumps::find_jump_threads_backwards_with_ranger): Remove.
6477 (try_thread_blocks): Rename find_jump_threads_backwards to
6479 (pass_early_thread_jumps::execute): Same.
6481 2021-08-12 Tobias Burnus <tobias@codesourcery.com>
6483 * tree-core.h (omp_clause_proc_bind_kind): Add
6484 OMP_CLAUSE_PROC_BIND_PRIMARY.
6485 * tree-pretty-print.c (dump_omp_clause): Add TODO comment to
6486 change 'master' to 'primary' in proc_bind for OpenMP 5.1.
6488 2021-08-12 Claudiu Zissulescu <claziss@synopsys.com>
6490 * common/config/arc/arc-common.c (arc_option_init_struct): Remove
6491 fno-common reference.
6492 * config/arc/arc.c (arc_override_options): Remove overriding of
6495 2021-08-12 Jakub Jelinek <jakub@redhat.com>
6498 * config/i386/i386-expand.c (ix86_expand_vec_one_operand_perm_avx512):
6499 If d->testing_p, return true after performing checks instead of
6500 actually expanding the insn.
6501 (expand_vec_perm_broadcast_1): Handle V32HImode - assert
6502 !TARGET_AVX512BW and return false.
6504 2021-08-12 Eric Botcazou <ebotcazou@gcc.gnu.org>
6506 * configure.ac (PE linker --disable-dynamicbase support): New check.
6507 * configure: Regenerate.
6508 * config.in: Likewise.
6509 * config/i386/mingw32.h (LINK_SPEC_DISABLE_DYNAMICBASE): New define.
6510 (LINK_SPEC): Use it.
6511 * config/i386/mingw-w64.h (LINK_SPEC_DISABLE_DYNAMICBASE): Likewise.
6512 (LINK_SPEC): Likewise.
6514 2021-08-12 liuhongt <hongtao.liu@intel.com>
6517 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_2): New
6518 post_reload define_insn_and_split.
6519 (*avx512bw_zero_extendv32qiv32hi2_2): Ditto.
6520 (*sse4_1_zero_extendv8qiv8hi2_4): Ditto.
6521 (*avx512f_zero_extendv16hiv16si2_2): Ditto.
6522 (*avx2_zero_extendv8hiv8si2_2): Ditto.
6523 (*sse4_1_zero_extendv4hiv4si2_4): Ditto.
6524 (*avx512f_zero_extendv8siv8di2_2): Ditto.
6525 (*avx2_zero_extendv4siv4di2_2): Ditto.
6526 (*sse4_1_zero_extendv2siv2di2_4): Ditto.
6527 (VI248_256, VI248_512, VI148_512, VI148_256, VI148_128): New
6530 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
6532 * config/rs6000/rs6000-builtin-new.def: Add always, power5, and
6535 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
6537 * config/rs6000/rs6000-builtin-new.def: Add vsx stanza.
6539 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
6541 * config/rs6000/rs6000-builtin-new.def: Finish altivec stanza.
6542 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
6543 initialization of pcvoid_type_node here...
6544 (altivec_init_builtins): ...from here.
6545 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add
6546 RS6000_BTI_const_ptr_void.
6547 (pcvoid_type_node): New macro.
6549 2021-08-11 Richard Biener <rguenther@suse.de>
6552 * tree-ssa-forwprop.c (pass_forwprop::execute): Do not decompose
6553 hard-register accesses.
6555 2021-08-11 Richard Biener <rguenther@suse.de>
6557 * tree-ssa-operands.c (operands_scanner::get_expr_operands):
6558 Do not look at COMPONENT_REF FIELD_DECLs TREE_THIS_VOLATILE
6559 to determine has_volatile_ops.
6561 2021-08-11 Eric Botcazou <ebotcazou@gcc.gnu.org>
6563 * cfgexpand.c (expand_used_vars): Reuse attribs local variable.
6565 2021-08-11 Jan Hubicka <hubicka@ucw.cz>
6566 Alexandre Oliva <oliva@adacore.com>
6568 * ipa-modref.c (modref_lattice::dump): Fix escape_point's min_flags
6570 (modref_lattice::merge_deref): Fix handling of indirect scape points.
6571 (update_escape_summary_1): Likewise.
6572 (update_escape_summary): Likewise.
6573 (ipa_merge_modref_summary_after_inlining): Likewise.
6575 2021-08-11 Richard Biener <rguenther@suse.de>
6577 PR middle-end/101858
6578 * fold-const.c (fold_binary_loc): Guard simplification
6579 of X < (cast) (1 << Y) to integer types.
6581 2021-08-11 Richard Biener <rguenther@suse.de>
6583 PR tree-optimization/101861
6584 * tree-vect-stmts.c (vectorizable_load): Fix error in
6585 previous change with regard to gather vectorization.
6587 2021-08-11 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
6590 * config/arm/arm_neon.h (vdup_n_s8): Replace call to builtin
6592 (vdup_n_s16): Likewise.
6593 (vdup_n_s32): Likewise.
6594 (vdup_n_s64): Likewise.
6595 (vdup_n_u8): Likewise.
6596 (vdup_n_u16): Likewise.
6597 (vdup_n_u32): Likewise.
6598 (vdup_n_u64): Likewise.
6599 (vdup_n_p8): Likewise.
6600 (vdup_n_p16): Likewise.
6601 (vdup_n_p64): Likewise.
6602 (vdup_n_f16): Likewise.
6603 (vdup_n_f32): Likewise.
6604 (vdupq_n_s8): Likewise.
6605 (vdupq_n_s16): Likewise.
6606 (vdupq_n_s32): Likewise.
6607 (vdupq_n_s64): Likewise.
6608 (vdupq_n_u8): Likewise.
6609 (vdupq_n_u16): Likewise.
6610 (vdupq_n_u32): Likewise.
6611 (vdupq_n_u64): Likewise.
6612 (vdupq_n_p8): Likewise.
6613 (vdupq_n_p16): Likewise.
6614 (vdupq_n_p64): Likewise.
6615 (vdupq_n_f16): Likewise.
6616 (vdupq_n_f32): Likewise.
6617 (vmov_n_s8): Replace call to builtin with call to corresponding
6619 (vmov_n_s16): Likewise.
6620 (vmov_n_s32): Likewise.
6621 (vmov_n_s64): Likewise.
6622 (vmov_n_u8): Likewise.
6623 (vmov_n_u16): Likewise.
6624 (vmov_n_u32): Likewise.
6625 (vmov_n_u64): Likewise.
6626 (vmov_n_p8): Likewise.
6627 (vmov_n_p16): Likewise.
6628 (vmov_n_f16): Likewise.
6629 (vmov_n_f32): Likewise.
6630 (vmovq_n_s8): Likewise.
6631 (vmovq_n_s16): Likewise.
6632 (vmovq_n_s32): Likewise.
6633 (vmovq_n_s64): Likewise.
6634 (vmovq_n_u8): Likewise.
6635 (vmovq_n_u16): Likewise.
6636 (vmovq_n_u32): Likewise.
6637 (vmovq_n_u64): Likewise.
6638 (vmovq_n_p8): Likewise.
6639 (vmovq_n_p16): Likewise.
6640 (vmovq_n_f16): Likewise.
6641 (vmovq_n_f32): Likewise.
6642 * config/arm/arm_neon_builtins.def: Remove entries for vdup_n.
6644 2021-08-11 liuhongt <hongtao.liu@intel.com>
6647 * config/i386/i386.md (ldexp<mode>3): Extend to vscalefs[sd]
6648 when TARGET_AVX512F and TARGET_SSE_MATH.
6650 2021-08-10 Jakub Jelinek <jakub@redhat.com>
6653 * config/i386/i386-expand.c (expand_vec_perm_even_odd): Return false
6654 for V32HImode if !TARGET_AVX512BW.
6655 (ix86_vectorize_vec_perm_const) <case E_V32HImode, case E_V64QImode>:
6656 If !TARGET_AVX512BW and TARGET_AVX512F and d.testing_p, don't fail
6657 early, but actually check the permutation.
6659 2021-08-10 Richard Biener <rguenther@suse.de>
6661 PR tree-optimization/101809
6662 * tree-vect-stmts.c (get_load_store_type): Allow emulated
6663 gathers with offset vector nunits being a constant multiple
6664 of the data vector nunits.
6665 (vect_get_gather_scatter_ops): Use the appropriate nunits
6666 for the offset vector defs.
6667 (vectorizable_store): Adjust call to
6668 vect_get_gather_scatter_ops.
6669 (vectorizable_load): Likewise. Handle the case of less
6670 offset vectors than data vectors.
6672 2021-08-10 Jakub Jelinek <jakub@redhat.com>
6675 * config/i386/sse.md (*avx512f_shuf_<shuffletype>64x2_1<mask_name>_1,
6676 *avx512f_shuf_<shuffletype>32x4_1<mask_name>_1): New define_insn
6679 2021-08-10 Richard Biener <rguenther@suse.de>
6681 PR tree-optimization/101801
6682 PR tree-optimization/101819
6683 * tree-vectorizer.h (vect_emulated_vector_p): Declare.
6684 * tree-vect-loop.c (vect_emulated_vector_p): New function.
6685 (vectorizable_reduction): Re-instantiate a check for emulated
6687 * tree-vect-stmts.c (vectorizable_shift): Likewise.
6688 (vectorizable_operation): Likewise. Cost emulated vector
6689 operations according to the scalar sequence synthesized by
6692 2021-08-10 Richard Biener <rguenther@suse.de>
6694 PR middle-end/101824
6695 * tree-nested.c (get_frame_field): Mark the COMPONENT_REF as
6696 volatile in case the variable was.
6698 2021-08-10 H.J. Lu <hjl.tools@gmail.com>
6701 * config/i386/constraints.md (BC): Document for integer SSE
6702 constant all bits set operand.
6703 (BF): New constraint for const floating-point all bits set
6705 * config/i386/i386.c (standard_sse_constant_p): Likewise.
6706 (standard_sse_constant_opcode): Likewise.
6707 * config/i386/sse.md (sseconstm1): New mode attribute.
6708 (mov<mode>_internal): Replace BC with <sseconstm1>.
6710 2021-08-10 liuhongt <hongtao.liu@intel.com>
6712 * config/i386/sse.md (cond_<insn><mode>): New expander.
6713 (VI248_AVX512VLBW): New mode iterator.
6714 * config/i386/predicates.md
6715 (nonimmediate_or_const_vec_dup_operand): New predicate.
6717 2021-08-09 Andrew MacLeod <amacleod@redhat.com>
6719 PR tree-optimization/101741
6720 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Check
6721 type of parameter for toupper/tolower.
6723 2021-08-09 Martin Jambor <mjambor@suse.cz>
6726 * ipa-prop.c (propagate_controlled_uses): Removed a spurious space.
6728 2021-08-09 Pat Haugen <pthaugen@linux.ibm.com>
6730 * config/rs6000/rs6000.c (is_load_insn1): Verify destination is a
6732 (is_store_insn1): Verify source is a register.
6734 2021-08-09 Uroš Bizjak <ubizjak@gmail.com>
6737 * config/i386/mmx.md (<any_logic:code>v2sf3):
6738 Rename from *mmx_<any_logic:code>v2sf3
6740 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6742 * config/nvptx/nvptx.c: Cross-reference parts adapted in
6743 'gcc/omp-oacc-neuter-broadcast.cc'.
6744 * omp-low.c: Likewise.
6745 * omp-oacc-neuter-broadcast.cc: Cross-reference parts adapted from
6748 2021-08-09 Julian Brown <julian@codesourcery.com>
6749 Kwok Cheung Yeung <kcy@codesourcery.com>
6750 Thomas Schwinge <thomas@codesourcery.com>
6752 * config/gcn/gcn.c (gcn_init_builtins): Override decls for
6753 BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START,
6754 BUILT_IN_GOACC_SINGLE_COPY_END and BUILT_IN_GOACC_BARRIER.
6755 (gcn_goacc_validate_dims): Turn on worker partitioning unconditionally.
6756 (gcn_fork_join): Update comment.
6757 * config/gcn/gcn.opt (flag_worker_partitioning): Remove.
6758 (macc_experimental_workers): Remove unused option.
6760 2021-08-09 Julian Brown <julian@codesourcery.com>
6761 Nathan Sidwell <nathan@codesourcery.com> (via 'gcc/config/nvptx/nvptx.c' master)
6762 Kwok Cheung Yeung <kcy@codesourcery.com>
6763 Thomas Schwinge <thomas@codesourcery.com>
6765 * Makefile.in (OBJS): Add omp-oacc-neuter-broadcast.o.
6766 * doc/tm.texi.in (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD):
6767 Add documentation hook.
6768 * doc/tm.texi: Regenerate.
6769 * omp-oacc-neuter-broadcast.cc: New file.
6770 * omp-builtins.def (BUILT_IN_GOACC_BARRIER)
6771 (BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START)
6772 (BUILT_IN_GOACC_SINGLE_COPY_END): New builtins.
6773 * passes.def (pass_omp_oacc_neuter_broadcast): Add pass.
6774 * target.def (goacc.create_worker_broadcast_record): Add target
6776 * tree-pass.h (make_pass_omp_oacc_neuter_broadcast): Add
6778 * config/gcn/gcn-protos.h (gcn_goacc_adjust_propagation_record):
6779 Rename prototype to...
6780 (gcn_goacc_create_worker_broadcast_record): ... this.
6781 * config/gcn/gcn-tree.c (gcn_goacc_adjust_propagation_record): Rename
6783 (gcn_goacc_create_worker_broadcast_record): ... this.
6784 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_PROPAGATION_RECORD):
6786 (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD): ... this.
6788 2021-08-09 Tejas Belagod <tejas.belagod@arm.com>
6791 * config/aarch64/aarch64-simd.md (vlshr<mode>3, vashr<mode>3): Use
6794 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6796 * Makefile.in (GTFILES): Remove '$(srcdir)/omp-offload.c'.
6798 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6800 * builtins.def (DEF_GOACC_BUILTIN, DEF_GOMP_BUILTIN): Don't
6801 consider '-foffload-abi'.
6802 * common.opt (-foffload-abi): Remove 'Var', 'Init'.
6803 * opts.c (common_handle_option) <-foffload-abi> [ACCEL_COMPILER]:
6806 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6808 * optc-gen.awk: Sanity check that 'Init' doesn't appear without
6811 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6813 * omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE): Remove.
6815 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6817 * doc/gty.texi (Files): Update.
6819 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6821 * doc/gty.texi (Files): Fix GTY header file example.
6823 2021-08-09 Roger Sayle <roger@nextmovesoftware.com>
6825 * tree-ssa-ccp.c (value_mask_to_min_max): Helper function to
6826 determine the upper and lower bounds from a mask-value pair.
6827 (bit_value_unop) [ABS_EXPR, ABSU_EXPR]: Add support for
6828 absolute value and unsigned absolute value expressions.
6829 (bit_value_binop): Initialize *VAL's precision.
6830 [LT_EXPR, LE_EXPR]: Use value_mask_to_min_max to determine
6831 upper and lower bounds of operands. Add LE_EXPR/GE_EXPR
6832 support when the operands are unknown but potentially equal.
6833 [MIN_EXPR, MAX_EXPR]: Support minimum/maximum expressions.
6835 2021-08-09 Bin Cheng <bin.cheng@linux.alibaba.com>
6837 * config/aarch64/aarch64.md
6838 (*extend<SHORT:mode><GPI:mode>2_aarch64): Use %<GPI:w>0.
6840 2021-08-08 Sergei Trofimovich <siarheit@google.com>
6842 * lra-constraints.c: Fix s/otput/output/ typo.
6844 2021-08-06 Martin Sebor <msebor@redhat.com>
6846 * builtins.c (expand_builtin_memchr): Move to gimple-ssa-warn-access.cc.
6847 (expand_builtin_strcat): Same.
6848 (expand_builtin_stpncpy): Same.
6849 (expand_builtin_strncat): Same.
6850 (check_read_access): Same.
6851 (check_memop_access): Same.
6852 (expand_builtin_strlen): Move checks to gimple-ssa-warn-access.cc.
6853 (expand_builtin_strnlen): Same.
6854 (expand_builtin_memcpy): Same.
6855 (expand_builtin_memmove): Same.
6856 (expand_builtin_mempcpy): Same.
6857 (expand_builtin_strcpy): Same.
6858 (expand_builtin_strcpy_args): Same.
6859 (expand_builtin_stpcpy_1): Same.
6860 (expand_builtin_strncpy): Same.
6861 (expand_builtin_memset): Same.
6862 (expand_builtin_bzero): Same.
6863 (expand_builtin_strcmp): Same.
6864 (expand_builtin_strncmp): Same.
6865 (expand_builtin): Remove handlers.
6866 (fold_builtin_strlen): Add a comment.
6867 * builtins.h (check_access): Move to gimple-ssa-warn-access.cc.
6868 * calls.c (maybe_warn_nonstring_arg): Same.
6869 * diagnostic-spec.c (nowarn_spec_t::nowarn_spec_t): Add warning option.
6870 * gimple-fold.c (gimple_fold_builtin_strcpy): Pass argument to callee.
6871 (gimple_fold_builtin_stpcpy): Same.
6872 * gimple-ssa-warn-access.cc (has_location): New function.
6873 (get_location): Same.
6874 (get_callee_fndecl): Same.
6877 (warn_string_no_nul): Define.
6878 (unterminated_array): Same.
6879 (check_nul_terminated_array): Same.
6880 (maybe_warn_nonstring_arg): Same.
6881 (maybe_warn_for_bound): Same.
6882 (warn_for_access): Same.
6883 (check_access): Same.
6884 (check_memop_access): Same.
6885 (check_read_access): Same.
6886 (warn_dealloc_offset): Use helper functions.
6887 (maybe_emit_free_warning): Same.
6888 (class pass_waccess): Add members.
6889 (check_strcat): New function.
6890 (check_strncat): New function.
6891 (check_stxcpy): New function.
6892 (check_stxncpy): New function.
6893 (check_strncmp): New function.
6894 (pass_waccess::check_builtin): New function.
6895 (pass_waccess::check): Call it.
6896 * gimple-ssa-warn-access.h (warn_string_no_nul): Move here from
6898 (maybe_warn_for_bound): Same.
6899 (check_access): Same.
6900 (check_memop_access): Same.
6901 (check_read_access): Same.
6902 * pointer-query.h (struct access_data): Define a ctor overload.
6904 2021-08-06 Richard Biener <rguenther@suse.de>
6906 PR tree-optimization/101801
6907 * tree-vectorizer.h (vect_worthwhile_without_simd_p): Rename...
6908 (vect_can_vectorize_without_simd_p): ... to this.
6909 * tree-vect-loop.c (vect_worthwhile_without_simd_p): Rename...
6910 (vect_can_vectorize_without_simd_p): ... to this and fold
6911 in vect_min_worthwhile_factor.
6912 (vect_min_worthwhile_factor): Remove.
6913 (vectorizable_reduction): Adjust and remove the cost part.
6914 * tree-vect-stmts.c (vectorizable_shift): Likewise.
6915 (vectorizable_operation): Likewise.
6917 2021-08-06 Uroš Bizjak <ubizjak@gmail.com>
6920 * config/i386/i386.md (cmove reg-to-reg move elimination peephole2s):
6921 Add general_gr_operand predicate to operand 3.
6923 2021-08-06 Roger Sayle <roger@nextmovesoftware.com>
6925 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern): Use
6926 CFN_BUILT_IN_CLRSB* instead of BUILT_IN_CLRSB* for consistency.
6928 2021-08-06 Tamar Christina <tamar.christina@arm.com>
6930 * config/aarch64/aarch64-sve-builtins.cc (register_svpattern,
6931 register_svprfop): Pass vec<> by pointer.
6932 * langhooks-def.h (lhd_simulate_enum_decl): Likewise.
6933 * langhooks.c (lhd_simulate_enum_decl): Likewise.
6934 * langhooks.h (struct lang_hooks_for_types): Likewise.
6936 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
6938 * config/aarch64/arm_neon.h (vst1_bf16_x2): Use
6939 __builtin_memcpy instead of constructing an additional
6940 __builtin_aarch64_simd_oi one vector at a time.
6941 (vst1q_bf16_x2): Likewise.
6942 (vst1_bf16_x3): Use __builtin_memcpy instead of constructing
6943 an additional __builtin_aarch64_simd_ci one vector at a time.
6944 (vst1q_bf16_x3): Likewise.
6945 (vst1_bf16_x4): Use __builtin_memcpy instead of a union.
6946 (vst1q_bf16_x4): Likewise.
6947 (vst2_bf16): Use __builtin_memcpy instead of constructing an
6948 additional __builtin_aarch64_simd_oi one vector at a time.
6949 (vst2q_bf16): Likewise.
6950 (vst3_bf16): Use __builtin_memcpy instead of constructing an
6951 additional __builtin_aarch64_simd_ci mode one vector at a
6953 (vst3q_bf16): Likewise.
6954 (vst4_bf16): Use __builtin_memcpy instead of constructing an
6955 additional __builtin_aarch64_simd_xi one vector at a time.
6956 (vst4q_bf16): Likewise.
6958 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
6960 * config/aarch64/arm_neon.h (__ST2_LANE_FUNC): Delete.
6961 (__ST2Q_LANE_FUNC): Delete.
6962 (vst2_lane_f16): Use __builtin_memcpy to copy vector
6963 structure instead of constructing __builtin_aarch64_simd_oi
6964 one vector at a time.
6965 (vst2_lane_f32): Likewise.
6966 (vst2_lane_f64): Likewise.
6967 (vst2_lane_p8): Likewise.
6968 (vst2_lane_p16): Likewise.
6969 (vst2_lane_p64): Likewise.
6970 (vst2_lane_s8): Likewise.
6971 (vst2_lane_s16): Likewise.
6972 (vst2_lane_s32): Likewise.
6973 (vst2_lane_s64): Likewise.
6974 (vst2_lane_u8): Likewise.
6975 (vst2_lane_u16): Likewise.
6976 (vst2_lane_u32): Likewise.
6977 (vst2_lane_u64): Likewise.
6978 (vst2_lane_bf16): Likewise.
6979 (vst2q_lane_f16): Use __builtin_memcpy to copy vector
6980 structure instead of using a union.
6981 (vst2q_lane_f32): Likewise.
6982 (vst2q_lane_f64): Likewise.
6983 (vst2q_lane_p8): Likewise.
6984 (vst2q_lane_p16): Likewise.
6985 (vst2q_lane_p64): Likewise.
6986 (vst2q_lane_s8): Likewise.
6987 (vst2q_lane_s16): Likewise.
6988 (vst2q_lane_s32): Likewise.
6989 (vst2q_lane_s64): Likewise.
6990 (vst2q_lane_u8): Likewise.
6991 (vst2q_lane_u16): Likewise.
6992 (vst2q_lane_u32): Likewise.
6993 (vst2q_lane_u64): Likewise.
6994 (vst2q_lane_bf16): Likewise.
6996 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
6998 * config/aarch64/arm_neon.h (__ST3_LANE_FUNC): Delete.
6999 (__ST3Q_LANE_FUNC): Delete.
7000 (vst3_lane_f16): Use __builtin_memcpy to copy vector
7001 structure instead of constructing __builtin_aarch64_simd_ci
7002 one vector at a time.
7003 (vst3_lane_f32): Likewise.
7004 (vst3_lane_f64): Likewise.
7005 (vst3_lane_p8): Likewise.
7006 (vst3_lane_p16): Likewise.
7007 (vst3_lane_p64): Likewise.
7008 (vst3_lane_s8): Likewise.
7009 (vst3_lane_s16): Likewise.
7010 (vst3_lane_s32): Likewise.
7011 (vst3_lane_s64): Likewise.
7012 (vst3_lane_u8): Likewise.
7013 (vst3_lane_u16): Likewise.
7014 (vst3_lane_u32): Likewise.
7015 (vst3_lane_u64): Likewise.
7016 (vst3_lane_bf16): Likewise.
7017 (vst3q_lane_f16): Use __builtin_memcpy to copy vector
7018 structure instead of using a union.
7019 (vst3q_lane_f32): Likewise.
7020 (vst3q_lane_f64): Likewise.
7021 (vst3q_lane_p8): Likewise.
7022 (vst3q_lane_p16): Likewise.
7023 (vst3q_lane_p64): Likewise.
7024 (vst3q_lane_s8): Likewise.
7025 (vst3q_lane_s16): Likewise.
7026 (vst3q_lane_s32): Likewise.
7027 (vst3q_lane_s64): Likewise.
7028 (vst3q_lane_u8): Likewise.
7029 (vst3q_lane_u16): Likewise.
7030 (vst3q_lane_u32): Likewise.
7031 (vst3q_lane_u64): Likewise.
7032 (vst3q_lane_bf16): Likewise.
7034 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
7036 * config/aarch64/arm_neon.h (__ST4_LANE_FUNC): Delete.
7037 (__ST4Q_LANE_FUNC): Delete.
7038 (vst4_lane_f16): Use __builtin_memcpy to copy vector
7039 structure instead of constructing __builtin_aarch64_simd_xi
7040 one vector at a time.
7041 (vst4_lane_f32): Likewise.
7042 (vst4_lane_f64): Likewise.
7043 (vst4_lane_p8): Likewise.
7044 (vst4_lane_p16): Likewise.
7045 (vst4_lane_p64): Likewise.
7046 (vst4_lane_s8): Likewise.
7047 (vst4_lane_s16): Likewise.
7048 (vst4_lane_s32): Likewise.
7049 (vst4_lane_s64): Likewise.
7050 (vst4_lane_u8): Likewise.
7051 (vst4_lane_u16): Likewise.
7052 (vst4_lane_u32): Likewise.
7053 (vst4_lane_u64): Likewise.
7054 (vst4_lane_bf16): Likewise.
7055 (vst4q_lane_f16): Use __builtin_memcpy to copy vector
7056 structure instead of using a union.
7057 (vst4q_lane_f32): Likewise.
7058 (vst4q_lane_f64): Likewise.
7059 (vst4q_lane_p8): Likewise.
7060 (vst4q_lane_p16): Likewise.
7061 (vst4q_lane_p64): Likewise.
7062 (vst4q_lane_s8): Likewise.
7063 (vst4q_lane_s16): Likewise.
7064 (vst4q_lane_s32): Likewise.
7065 (vst4q_lane_s64): Likewise.
7066 (vst4q_lane_u8): Likewise.
7067 (vst4q_lane_u16): Likewise.
7068 (vst4q_lane_u32): Likewise.
7069 (vst4q_lane_u64): Likewise.
7070 (vst4q_lane_bf16): Likewise.
7072 2021-08-06 Martin Liska <mliska@suse.cz>
7074 * config/rs6000/rs6000.c (rs6000_option_override_internal): When
7075 a target option is restored, it can have
7076 rs6000_long_double_type_size set to FLOAT_PRECISION_TFmode
7077 and error should not be emitted.
7079 2021-08-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
7081 * gcov-io.h (gcov_write): Declare.
7082 * gcov-io.c (gcov_write): New.
7083 (gcov_write_counter): Remove.
7084 (gcov_write_tag_length): Likewise.
7085 (gcov_write_summary): Replace gcov_write_tag_length() with calls to
7086 gcov_write_unsigned().
7087 * doc/invoke.texi (fprofile-info-section): Mention
7088 __gcov_info_to_gdca().
7090 2021-08-06 Martin Sebor <msebor@redhat.com>
7092 * dominance.c (prune_bbs_to_update_dominators): Adjust by-value vec
7093 arguments to by-reference.
7094 (iterate_fix_dominators): Same.
7095 * dominance.h (iterate_fix_dominators): Same.
7096 * ipa-prop.h: Call auto_vec::to_vec_legacy.
7097 * tree-data-ref.c (dump_data_dependence_relation): Adjust by-value vec
7098 arguments to by-reference.
7099 (debug_data_dependence_relation): Same.
7100 (dump_data_dependence_relations): Same.
7101 * tree-data-ref.h (debug_data_dependence_relation): Same.
7102 (dump_data_dependence_relations): Same.
7103 * tree-predcom.c (dump_chains): Same.
7104 (initialize_root_vars_lm): Same.
7105 (determine_unroll_factor): Same.
7106 (replace_phis_by_defined_names): Same.
7107 (insert_init_seqs): Same.
7108 (pcom_worker::tree_predictive_commoning_loop): Call
7109 auto_vec::to_vec_legacy.
7110 * tree-ssa-pre.c (insert_into_preds_of_block): Adjust by-value vec
7111 arguments to by-reference.
7112 * tree-ssa-threadbackward.c (populate_worklist): Same.
7113 (back_threader::resolve_def): Same.
7114 * tree-vect-data-refs.c (vect_check_nonzero_value): Same.
7115 (vect_enhance_data_refs_alignment): Same.
7116 (vect_check_lower_bound): Same.
7117 (vect_prune_runtime_alias_test_list): Same.
7118 (vect_permute_store_chain): Same.
7119 * tree-vect-slp-patterns.c (vect_normalize_conj_loc): Same.
7120 * tree-vect-stmts.c (vect_create_vectorized_demotion_stmts): Same.
7121 * tree-vectorizer.h (vect_permute_store_chain): Same.
7122 * vec.c (test_init): New function.
7123 (vec_c_tests): Call new function.
7124 * vec.h (vec): Declare ctors, dtor, and assignment.
7125 (auto_vec::vec_to_legacy): New function.
7126 (vec::copy): Adjust initialization.
7128 2021-08-05 H.J. Lu <hjl.tools@gmail.com>
7131 * config/i386/i386.c (ix86_can_inline_p): Ignore MASK_80387 if
7132 callee only uses GPRs.
7133 * config/i386/ia32intrin.h: Revert commit 5463cee2770.
7134 * config/i386/serializeintrin.h: Revert commit 71958f740f1.
7135 * config/i386/x86gprintrin.h: Add
7136 #pragma GCC target("general-regs-only") and #pragma GCC pop_options
7137 to disable non-GPR ISAs.
7139 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
7141 PR middle-end/101787
7142 * doc/md.texi (cond_ashl, cond_ashr, cond_lshr): Document.
7144 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
7146 * tree-vectorizer.h (vect_is_store_elt_extraction, vect_is_reduction)
7147 (vect_reduc_type, vect_embedded_comparison_type, vect_comparison_type)
7148 (vect_is_extending_load, vect_is_integer_truncation): New functions,
7149 moved from aarch64.c but given different names.
7150 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction)
7151 (aarch64_is_reduction, aarch64_reduc_type)
7152 (aarch64_embedded_comparison_type, aarch64_comparison_type)
7153 (aarch64_extending_load_p, aarch64_integer_truncation_p): Delete
7154 in favor of the above. Update callers accordingly.
7156 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
7159 * config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
7160 writing .cpu directive in asm output.
7161 * config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
7162 (arm_last_printed_arch_string): Delete.
7163 (arm_last-printed_fpu_string): Delete.
7164 (arm_configure_build_target): If use of floating-point/SIMD is
7165 disabled, remove all fp/simd related features from the target ISA.
7166 (last_arm_targ_options): New variable.
7167 (arm_print_asm_arch_directives): Add new parameters. Change order
7168 of emitted directives and handle all cases here.
7169 (arm_file_start): Always call arm_print_asm_arch_directives, move
7170 all generation of .arch/.arch_extension here.
7171 (arm_file_end): Call arm_print_asm_arch.
7172 (arm_declare_function_name): Call arm_print_asm_arch_directives
7173 instead of printing .arch/.fpu directives directly.
7175 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
7177 * config/arm/arm.c (arm_configure_build_target): Don't call
7178 arm_option_reconfigure_globals.
7179 (arm_option_restore): Call arm_option_reconfigure_globals after
7180 reconfiguring the target.
7181 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
7183 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
7185 * config/arm/arm.c (arm_configure_build_target): Ensure the target's
7186 arch_name is always set.
7188 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
7190 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
7191 of vec_select high-half from being added into Neon subtract
7194 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
7196 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
7197 of vec_select high-half from being added into Neon add cost.
7199 2021-08-05 Kewen Lin <linkw@linux.ibm.com>
7201 * cfgloop.h (loops_list::loops_list): Add one optional argument
7202 root and adjust accordingly, update loop tree walking and factor
7204 * cfgloop.c (loops_list::walk_loop_tree): ... this. New function.
7206 2021-08-05 Eric Botcazou <ebotcazou@gcc.gnu.org>
7208 PR tree-optimization/101626
7209 * tree-sra.c (propagate_subaccesses_from_rhs): Do not set the
7210 reverse scalar storage order on a pointer or vector component.
7212 2021-08-05 liuhongt <hongtao.liu@intel.com>
7214 * config/i386/sse.md (cond_<code><mode>): New expander.
7216 2021-08-05 liuhongt <hongtao.liu@intel.com>
7218 * config/i386/sse.md (cond_<code><mode>): New expander.
7220 2021-08-05 liuhongt <hongtao.liu@intel.com>
7222 * config/i386/sse.md (cond_<code><mode>): New expander.
7224 2021-08-04 David Malcolm <dmalcolm@redhat.com>
7227 * Makefile.in (ANALYZER_OBJS): Add analyzer/region-model-asm.o.
7229 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
7232 * config/i386/i386.h (STORE_MAX_PIECES): Allow 16/32/64 bytes
7233 only if TARGET_INTER_UNIT_MOVES_TO_VEC is true.
7235 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
7238 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
7239 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
7240 data with SSE register from one memory location to another.
7242 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7244 * config/s390/s390.c (expand_perm_with_vpdi): New function.
7245 (vectorize_vec_perm_const_1): Call expand_perm_with_vpdi.
7246 * config/s390/vector.md (*vpdi1<mode>, @vpdi1<mode>): Enable a
7247 parameterized expander.
7248 (*vpdi4<mode>, @vpdi4<mode>): Likewise.
7250 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7252 * config/s390/s390.c (MAX_VECT_LEN): Define macro.
7253 (struct expand_vec_perm_d): Define struct.
7254 (expand_perm_with_merge): New function.
7255 (vectorize_vec_perm_const_1): New function.
7256 (s390_vectorize_vec_perm_const): New function.
7257 (TARGET_VECTORIZE_VEC_PERM_CONST): Define target macro.
7259 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7261 * config/s390/vector.md (V_HW_64): Remove mode iterator.
7262 (*vec_load_pair<mode>): Use V_HW_2 instead of V_HW_64.
7263 * config/s390/vx-builtins.md
7264 (vec_scatter_element<V_HW_2:mode>_SI): Use V_HW_2 instead of
7267 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7269 * config/s390/s390.md (UNSPEC_VEC_PERMI): Remove constant
7271 * config/s390/vector.md (*vpdi1<mode>, *vpdi4<mode>): New pattern
7273 * config/s390/vx-builtins.md (*vec_permi<mode>): Emit generic rtx
7274 instead of an unspec.
7276 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7278 * config/s390/s390-modes.def: Add more vector modes to support
7279 concatenation of two vectors.
7280 * config/s390/s390-protos.h (s390_expand_merge_perm_const): Add
7282 (s390_expand_merge): Likewise.
7283 * config/s390/s390.c (s390_expand_merge_perm_const): New function.
7284 (s390_expand_merge): New function.
7285 * config/s390/s390.md (UNSPEC_VEC_MERGEH, UNSPEC_VEC_MERGEL):
7286 Remove constant definitions.
7287 * config/s390/vector.md (V_HW_2): Add mode iterators.
7288 (VI_HW_4, V_HW_4): Rename VI_HW_4 to V_HW_4.
7289 (vec_2x_nelts, vec_2x_wide): New mode attributes.
7290 (*vmrhb, *vmrlb, *vmrhh, *vmrlh, *vmrhf, *vmrlf, *vmrhg, *vmrlg):
7291 New pattern definitions.
7292 (vec_widen_umult_lo_<mode>, vec_widen_umult_hi_<mode>)
7293 (vec_widen_smult_lo_<mode>, vec_widen_smult_hi_<mode>)
7294 (vec_unpacks_lo_v4sf, vec_unpacks_hi_v4sf, vec_unpacks_lo_v2df)
7295 (vec_unpacks_hi_v2df): Adjust expanders to emit non-unspec RTX for
7297 * config/s390/vx-builtins.md (V_HW_4): Remove mode iterator. Now
7299 (vec_mergeh<mode>, vec_mergel<mode>): Use s390_expand_merge to
7300 emit vec merge pattern.
7302 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
7304 * config/aarch64/aarch64.c (aarch64_strip_extend_vec_half):
7306 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent cost of
7307 vec_select high-half from being added into Neon multiply
7309 * rtlanal.c (vec_series_highpart_p): Define.
7310 * rtlanal.h (vec_series_highpart_p): Declare.
7312 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
7314 * config/aarch64/aarch64.c (aarch64_strip_duplicate_vec_elt):
7316 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent
7317 vec_select cost from being added into Neon multiply cost.
7319 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
7321 * tree-vect-loop.c (vect_better_loop_vinfo_p): Detect cases in
7322 which old_loop_vinfo is an epilogue loop that handles a constant
7323 number of iterations.
7325 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
7327 * tree-vect-loop.c (vect_analyze_loop): Print a dump message
7328 when a reanalyzed loop fails to be cheaper than the current
7331 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
7333 * config/aarch64/aarch64.c: Fix a typo.
7335 2021-08-04 Vincent Lefèvre <vincent-gcc@vinc17.net>
7337 PR gcov-profile/101773
7338 * gcov-io.c (gcov_close): Check return code of a fclose.
7340 2021-08-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
7343 * dwarf2out.c (dwarf2out_assembly_start): Emit a dummy
7344 .file statement when needed.
7346 2021-08-04 Richard Biener <rguenther@suse.de>
7348 * tree-vect-data-refs.c (vect_check_gather_scatter):
7349 Include widening conversions only when the result is
7350 still handed by native gather or the current offset
7351 size not already matches the data size.
7352 Also succeed analysis in case there's no native support,
7353 noted by a IFN_LAST ifn and a NULL decl.
7354 (vect_analyze_data_refs): Always consider gathers.
7355 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
7356 Test for no IFN gather rather than decl gather.
7357 * tree-vect-stmts.c (vect_model_load_cost): Pass in the
7358 gather-scatter info and cost emulated gathers accordingly.
7359 (vect_truncate_gather_scatter_offset): Properly test for
7361 (vect_use_strided_gather_scatters_p): Likewise.
7362 (get_load_store_type): Handle emulated gathers and its
7364 (vectorizable_load): Likewise. Emulate them by extracting
7365 scalar offsets, doing scalar loads and a vector construct.
7367 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
7370 * expr.c (op_by_pieces_d::op_by_pieces_d): Add a max_pieces
7371 argument to set m_max_size.
7372 (move_by_pieces_d): Pass MOVE_MAX_PIECES to op_by_pieces_d.
7373 (store_by_pieces_d): Pass STORE_MAX_PIECES to op_by_pieces_d.
7374 (compare_by_pieces_d): Pass COMPARE_MAX_PIECES to op_by_pieces_d.
7376 2021-08-04 Roger Sayle <roger@nextmovesoftware.com>
7377 Marc Glisse <marc.glisse@inria.fr>
7379 * match.pd (bit_ior, bit_xor): Canonicalize (X*C1)|(X*C2) and
7380 (X*C1)^(X*C2) as X*(C1+C2), and related variants, using
7381 tree_nonzero_bits to ensure that operands are bit-wise disjoint.
7383 2021-08-04 Richard Biener <rguenther@suse.de>
7385 * tree-ssa-forwprop.c (pass_forwprop::execute): Split
7386 out code to decompose vector loads ...
7387 (optimize_vector_load): ... here. Generalize it to
7388 handle intermediate widening and TARGET_MEM_REF loads
7389 and apply it to loads with a supported vector mode as well.
7391 2021-08-04 Richard Biener <rguenther@suse.de>
7393 PR tree-optimization/101756
7394 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Make sure
7395 the result of the reduction epilogue is compatible to the original
7398 2021-08-04 liuhongt <hongtao.liu@intel.com>
7401 * config/i386/i386.md (peephole2): Refine predicate from
7402 register_operand to general_reg_operand.
7404 2021-08-04 Aldy Hernandez <aldyh@redhat.com>
7406 * gimple-range-path.h (path_range_query::dump): Mark override.
7408 2021-08-04 Richard Biener <rguenther@suse.de>
7410 PR tree-optimization/101769
7411 * tree-tailcall.c (eliminate_tail_call): Add the created loop
7412 for the first recursion and return it via the new output parameter.
7413 (optimize_tail_call): Pass through new output param.
7414 (tree_optimize_tail_calls_1): After creating all latches,
7415 add the created loop to the loop tree. Do not mark loops for fixup.
7417 2021-08-04 Martin Liska <mliska@suse.cz>
7419 * doc/invoke.texi: Document threader-mode param.
7421 2021-08-04 liuhongt <hongtao.liu@intel.com>
7423 * config/i386/sse.md (cond_fma<mode>): New expander.
7424 (cond_fms<mode>): Ditto.
7425 (cond_fnma<mode>): Ditto.
7426 (cond_fnms<mode>): Ditto.
7428 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
7430 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode>): Use && instead of &.
7432 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
7434 * config/rs6000/constraints.md: Remove "e" from the list of available
7435 constraint characters.
7437 2021-08-03 Eugene Rozenfeld <erozen@microsoft.com>
7439 PR gcov-profile/71672
7440 * auto-profile.c (afdo_indirect_call): Fix setup of the historgram value for indirect calls.
7442 2021-08-03 Paul A. Clarke <pc@us.ibm.com>
7444 * config/rs6000/smmintrin.h (_mm_minpos_epu16): New.
7446 2021-08-03 H.J. Lu <hjl.tools@gmail.com>
7448 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): In 64-bit mode,
7449 try XMM31 to avoid vzeroupper.
7451 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7453 * doc/invoke.texi: Document -mtune=neoverse-512tvb and
7454 -mcpu=neoverse-512tvb.
7455 * config/aarch64/aarch64-cores.def (neoverse-512tvb): New entry.
7456 * config/aarch64/aarch64-tune.md: Regenerate.
7457 * config/aarch64/aarch64.c (neoverse512tvb_sve_vector_cost)
7458 (neoverse512tvb_sve_issue_info, neoverse512tvb_vec_issue_info)
7459 (neoverse512tvb_vector_cost, neoverse512tvb_tunings): New structures.
7460 (aarch64_adjust_body_cost_sve): Handle -mtune=neoverse-512tvb.
7461 (aarch64_adjust_body_cost): Likewise.
7463 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7465 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Only
7466 record issue information for operations that occur in the
7469 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7471 * config/aarch64/aarch64.c (aarch64_multiply_add_p): Add a vec_flags
7472 parameter. Detect cases in which an Advanced SIMD MLA would almost
7473 certainly require a MOV.
7474 (aarch64_count_ops): Update accordingly.
7476 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7478 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction): New
7479 function, split out from...
7480 (aarch64_detect_vector_stmt_subtype): ...here.
7481 (aarch64_add_stmt_cost): Treat extracting element 0 as free.
7483 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7485 * config/aarch64/aarch64-protos.h (sve_vec_cost):
7486 Add gather_load_x32_cost and gather_load_x64_cost.
7487 * config/aarch64/aarch64.c (generic_sve_vector_cost)
7488 (a64fx_sve_vector_cost, neoversev1_sve_vector_cost): Update
7489 accordingly, using the values given by the scalar_load * number
7490 of elements calculation that we used previously.
7491 (aarch64_detect_vector_stmt_subtype): Use the new fields.
7493 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7495 * config/aarch64/aarch64.c (aarch64_adjust_body_cost_sve): New
7496 function, split out from...
7497 (aarch64_adjust_body_cost): ...here.
7499 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7501 * config/aarch64/fractional-cost.h: New file.
7502 * config/aarch64/aarch64.c: Include <algorithm> (indirectly)
7503 and cost_fraction.h.
7504 (vec_cost_fraction): New typedef.
7505 (aarch64_detect_scalar_stmt_subtype): Use it for statement costs.
7506 (aarch64_detect_vector_stmt_subtype): Likewise.
7507 (aarch64_sve_adjust_stmt_cost, aarch64_adjust_stmt_cost): Likewise.
7508 (aarch64_estimate_min_cycles_per_iter): Use vec_cost_fraction
7510 (aarch64_adjust_body_cost): Likewise.
7511 (aarch64_test_cost_fraction): New function.
7512 (aarch64_run_selftests): Call it.
7514 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7516 * config/aarch64/aarch64-protos.h (tune_params::sve_width): Turn
7518 * config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Update
7520 (aarch64_estimated_poly_value): Likewise. Use the least significant
7521 set bit for the minimum and likely values. Use the most significant
7522 set bit for the maximum value.
7524 2021-08-03 liuhongt <hongtao.liu@intel.com>
7526 * config/i386/sse.md (cond_<insn><mode>): New expander.
7527 (cond_mul<mode>): Ditto.
7529 2021-08-03 Kewen Lin <linkw@linux.ibm.com>
7531 * tree-cfg.c (move_sese_region_to_fn): Fix typos on dloop.
7533 2021-08-03 liuhongt <hongtao.liu@intel.com>
7535 * config/i386/sse.md (cond_<insn><mode>):New expander.
7536 (cond_mul<mode>): Ditto.
7537 (cond_div<mode>): Ditto.
7539 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
7541 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Also
7542 check stack_realign_needed for stack realignment.
7543 (ix86_legitimate_constant_p): Always allow CONST_WIDE_INT smaller
7544 than the largest integer supported by vector register.
7545 * config/i386/i386.h (MAX_MOVE_MAX): New. Set to 64.
7546 (MOVE_MAX): Set to bytes of the largest integer supported by
7548 (STORE_MAX_PIECES): New.
7550 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
7552 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
7553 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
7554 data from one memory location to another.
7556 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
7559 * config/i386/i386.c (TARGET_GEN_MEMSET_SCRATCH_RTX): New.
7561 2021-08-02 Aldy Hernandez <aldyh@redhat.com>
7563 PR tree-optimization/101724
7564 * params.opt: Remove --param=threader-iterative.
7565 * tree-ssa-threadbackward.c (pass_thread_jumps::execute): Remove
7568 2021-08-02 Tom de Vries <tdevries@suse.de>
7570 PR middle-end/101665
7571 * doc/extend.texi (nonnull attribute): Improve documentation.
7573 2021-08-02 Andrew Pinski <apinski@marvell.com>
7575 PR rtl-optimization/101683
7576 * rtlanal.c (may_trap_p_1): Handle UNSIGNED_FIX.
7578 2021-08-02 Roger Sayle <roger@nextmovesoftware.com>
7580 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern):
7581 Renamed from cond_removal_in_popcount_clz_ctz_pattern.
7582 Add support for BSWAP, FFS, PARITY and CLRSB builtins.
7583 (tree_ssa_phiop_worker): Update call to function above.
7585 2021-08-01 H.J. Lu <hjl.tools@gmail.com>
7588 * config/i386/i386.md (bsr_rex64_1_zext): New.
7589 (combine splitter for constant - clzll): Replace gen_bsr_rex64_1
7590 with gen_bsr_rex64_1_zext.
7592 2021-07-31 Jakub Jelinek <jakub@redhat.com>
7595 * config/i386/i386.md (bsr_rex64_1, bsr_1, bsr_zext_1): New
7596 define_insn patterns.
7597 (*bsr_rex64_2, *bsr_2): New define_insn_and_split patterns.
7598 Add combine splitters for constant - clz.
7599 (clz<mode>2): Use a temporary pseudo for bsr result.
7601 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
7603 * config/rs6000/smmintrin.h (_mm_floor_pd, _mm_floor_ps,
7604 _mm_floor_sd, _mm_floor_ss): New.
7606 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
7608 * config/rs6000/smmintrin.h (_mm_ceil_pd, _mm_ceil_ps,
7609 _mm_ceil_sd, _mm_ceil_ss): New.
7611 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
7613 * config/rs6000/smmintrin.h (_mm_blend_pd, _mm_blendv_pd,
7614 _mm_blend_ps, _mm_blendv_ps): New.
7616 2021-07-30 Roger Sayle <roger@nextmovesoftware.com>
7617 Uroš Bizjak <ubizjak@gmail.com>
7619 * config/i386/i386.md (*dec_cmov<mode>): New define_insn_and_split
7620 to generate a conditional move using the carry flag after sub $1.
7621 (peephole2): Eliminate a register-to-register move by inverting
7622 the condition of a conditional move.
7624 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
7626 * config/mmix/mmix.md ("call", "call_value", "*call_real")
7627 ("*call_value_real"): Don't generate rtx mentioning the generic
7628 operands 1 and 2 to "call", and similarly for "call_value".
7629 * config/mmix/mmix.c (mmix_print_operand_punct_valid_p)
7630 (mmix_print_operand): Use '!' instead of 'p'.
7632 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
7634 * doc/md.texi (call): Correct information about operand 2.
7635 * config/mmix/mmix.md ("call", "call_value"): Remove fixed FIXMEs.
7637 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
7639 * range-op.cc (operator_trunc_mod::wi_fold): Fold constants.
7641 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
7643 * range-op.cc (operator_div::wi_fold): Return UNDEFINED for [0, 0] divisor.
7645 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
7647 * gimple-range-cache.cc (*::set_bb_range): Change const basic_block to
7649 (*::get_bb_range): Ditto.
7650 (*::bb_range_p): Ditto.
7651 * gimple-range-cache.h: Change prototypes.
7653 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
7656 * builtins.c (builtin_memcpy_read_str): Change the mode argument
7657 from scalar_int_mode to fixed_size_mode.
7658 (builtin_strncpy_read_str): Likewise.
7659 (gen_memset_value_from_prev): New function.
7660 (builtin_memset_read_str): Change the mode argument from
7661 scalar_int_mode to fixed_size_mode. Use gen_memset_value_from_prev
7662 and support CONST_VECTOR.
7663 (builtin_memset_gen_str): Likewise.
7664 (try_store_by_multiple_pieces): Use by_pieces_constfn to declare
7666 * builtins.h (builtin_strncpy_read_str): Replace scalar_int_mode
7667 with fixed_size_mode.
7668 (builtin_memset_read_str): Likewise.
7669 * expr.c (widest_int_mode_for_size): Renamed to ...
7670 (widest_fixed_size_mode_for_size): Add a bool argument to
7671 indicate if QI vector mode can be used.
7672 (by_pieces_ninsns): Call widest_fixed_size_mode_for_size
7673 instead of widest_int_mode_for_size.
7674 (pieces_addr::adjust): Change the mode argument from
7675 scalar_int_mode to fixed_size_mode.
7676 (op_by_pieces_d): Make m_len read-only. Add a bool member,
7677 m_qi_vector_mode, to indicate that QI vector mode can be used.
7678 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
7679 initialize m_qi_vector_mode. Call widest_fixed_size_mode_for_size
7680 instead of widest_int_mode_for_size.
7681 (op_by_pieces_d::get_usable_mode): Change the mode argument from
7682 scalar_int_mode to fixed_size_mode. Call
7683 widest_fixed_size_mode_for_size instead of
7684 widest_int_mode_for_size.
7685 (op_by_pieces_d::smallest_fixed_size_mode_for_size): New member
7686 function to return the smallest integer or QI vector mode.
7687 (op_by_pieces_d::run): Call widest_fixed_size_mode_for_size
7688 instead of widest_int_mode_for_size. Call
7689 smallest_fixed_size_mode_for_size instead of
7690 smallest_int_mode_for_size.
7691 (store_by_pieces_d::store_by_pieces_d): Add a bool argument to
7692 indicate that QI vector mode can be used and pass it to
7693 op_by_pieces_d::op_by_pieces_d.
7694 (can_store_by_pieces): Call widest_fixed_size_mode_for_size
7695 instead of widest_int_mode_for_size. Pass memsetp to
7696 widest_fixed_size_mode_for_size to support QI vector mode.
7697 Allow all CONST_VECTORs for memset if vec_duplicate is supported.
7698 (store_by_pieces): Pass memsetp to
7699 store_by_pieces_d::store_by_pieces_d.
7700 (clear_by_pieces_1): Removed.
7701 (clear_by_pieces): Replace clear_by_pieces_1 with
7702 builtin_memset_read_str and pass true to store_by_pieces_d to
7703 support vector mode broadcast.
7704 (string_cst_read_str): Change the mode argument from
7705 scalar_int_mode to fixed_size_mode.
7706 * expr.h (by_pieces_constfn): Change scalar_int_mode to
7708 (by_pieces_prev): Likewise.
7709 * rtl.h (lowpart_subreg_regno): New.
7710 * rtlanal.c (lowpart_subreg_regno): New. A wrapper around
7711 simplify_subreg_regno.
7712 * target.def (gen_memset_scratch_rtx): New hook.
7713 * doc/tm.texi.in: Add TARGET_GEN_MEMSET_SCRATCH_RTX.
7714 * doc/tm.texi: Regenerated.
7716 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
7719 * config/mips/mips.c (mips_atomic_assign_expand_fenv): Use
7720 TARGET_EXPR instead of MODIFY_EXPR.
7722 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
7725 * config/mips/mips-protos.h (mips_expand_vec_cmp_expr): Declare.
7726 * config/mips/mips.c (mips_expand_vec_cmp_expr): New function.
7727 * config/mips/mips-msa.md (vec_cmp<MSA:mode><mode_i>): New
7729 (vec_cmpu<IMSA:mode><mode_i>): New expander.
7731 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
7734 * config/i386/i386-options.c (ix86_option_override_internal):
7735 Don't enable LZCNT/POPCNT if they have been disabled explicitly.
7737 2021-07-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
7740 * config/arm/arm_neon.h (vld1_p64): Replace call to builtin by
7741 explicitly dereferencing __a.
7742 (vld1_s64): Likewise.
7743 (vld1_u64): Likewise.
7744 * config/arm/arm_neon_builtins.def (vld1): Remove entry for di
7745 and change to VAR13.
7747 2021-07-30 Aldy Hernandez <aldyh@redhat.com>
7749 * gimple-loop-versioning.cc (lv_dom_walker::lv_dom_walker): Remove
7750 use of m_range_analyzer.
7751 (loop_versioning::lv_dom_walker::before_dom_children): Same.
7752 (loop_versioning::lv_dom_walker::after_dom_children): Remove.
7753 (loop_versioning::prune_loop_conditions): Replace vr_values use
7754 with range_query interface.
7755 (pass_loop_versioning::execute): Use ranger.
7757 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
7760 * ipa-devirt.c (ipa_odr_read_section): Compare the precision of
7761 enum values, and emit a warning if they mismatch.
7763 2021-07-30 Kewen Lin <linkw@linux.ibm.com>
7765 * cfgloop.h (as_const): New function.
7766 (class loop_iterator): Rename to ...
7767 (class loops_list): ... this.
7768 (loop_iterator::next): Rename to ...
7769 (loops_list::Iter::fill_curr_loop): ... this and adjust.
7770 (loop_iterator::loop_iterator): Rename to ...
7771 (loops_list::loops_list): ... this and adjust.
7772 (loops_list::Iter): New class.
7773 (loops_list::iterator): New type.
7774 (loops_list::const_iterator): New type.
7775 (loops_list::begin): New function.
7776 (loops_list::end): Likewise.
7777 (loops_list::begin const): Likewise.
7778 (loops_list::end const): Likewise.
7779 (FOR_EACH_LOOP): Remove.
7780 (FOR_EACH_LOOP_FN): Remove.
7781 * cfgloop.c (flow_loops_dump): Adjust FOR_EACH_LOOP* with range-based
7782 for loop with loops_list instance.
7783 (sort_sibling_loops): Likewise.
7784 (disambiguate_loops_with_multiple_latches): Likewise.
7785 (verify_loop_structure): Likewise.
7786 * cfgloopmanip.c (create_preheaders): Likewise.
7787 (force_single_succ_latches): Likewise.
7788 * config/aarch64/falkor-tag-collision-avoidance.c
7789 (execute_tag_collision_avoidance): Likewise.
7790 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Likewise.
7791 * config/s390/s390.c (s390_adjust_loops): Likewise.
7792 * doc/loop.texi: Likewise.
7793 * gimple-loop-interchange.cc (pass_linterchange::execute): Likewise.
7794 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
7795 * gimple-loop-versioning.cc (loop_versioning::analyze_blocks): Likewise.
7796 (loop_versioning::make_versioning_decisions): Likewise.
7797 * gimple-ssa-split-paths.c (split_paths): Likewise.
7798 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
7799 * graphite.c (canonicalize_loop_form): Likewise.
7800 (graphite_transform_loops): Likewise.
7801 * ipa-fnsummary.c (analyze_function_body): Likewise.
7802 * ipa-pure-const.c (analyze_function): Likewise.
7803 * loop-doloop.c (doloop_optimize_loops): Likewise.
7804 * loop-init.c (loop_optimizer_finalize): Likewise.
7805 (fix_loop_structure): Likewise.
7806 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
7807 (move_loop_invariants): Likewise.
7808 * loop-unroll.c (decide_unrolling): Likewise.
7809 (unroll_loops): Likewise.
7810 * modulo-sched.c (sms_schedule): Likewise.
7811 * predict.c (predict_loops): Likewise.
7812 (pass_profile::execute): Likewise.
7813 * profile.c (branch_prob): Likewise.
7814 * sel-sched-ir.c (sel_finish_pipelining): Likewise.
7815 (sel_find_rgns): Likewise.
7816 * tree-cfg.c (replace_loop_annotate): Likewise.
7817 (replace_uses_by): Likewise.
7818 (move_sese_region_to_fn): Likewise.
7819 * tree-if-conv.c (pass_if_conversion::execute): Likewise.
7820 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
7821 * tree-parloops.c (parallelize_loops): Likewise.
7822 * tree-predcom.c (tree_predictive_commoning): Likewise.
7823 * tree-scalar-evolution.c (scev_initialize): Likewise.
7824 (scev_reset): Likewise.
7825 * tree-ssa-dce.c (find_obviously_necessary_stmts): Likewise.
7826 * tree-ssa-live.c (remove_unused_locals): Likewise.
7827 * tree-ssa-loop-ch.c (ch_base::copy_headers): Likewise.
7828 * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
7829 (tree_ssa_lim_initialize): Likewise.
7830 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Likewise.
7831 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize): Likewise.
7832 * tree-ssa-loop-manip.c (get_loops_exits): Likewise.
7833 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations): Likewise.
7834 (free_numbers_of_iterations_estimates): Likewise.
7835 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Likewise.
7836 * tree-ssa-loop-split.c (tree_ssa_split_loops): Likewise.
7837 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Likewise.
7838 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
7839 (pass_scev_cprop::execute): Likewise.
7840 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
7841 * tree-ssa-sccvn.c (do_rpo_vn): Likewise.
7842 * tree-ssa-threadupdate.c
7843 (jump_thread_path_registry::thread_through_all_blocks): Likewise.
7844 * tree-vectorizer.c (vectorize_loops): Likewise.
7845 * tree-vrp.c (vrp_asserts::find_assert_locations): Likewise.
7847 2021-07-29 Hans-Peter Nilsson <hp@bitrange.com>
7849 * config/mmix/mmix.c (mmix_function_arg_1): Avoid
7850 generating a VOIDmode register for e.g the
7851 function_arg_info::end_marker.
7853 2021-07-29 Jeff Law <jeffreyalaw@gmail.com>
7855 * config/h8300/h8300-modes.def: Add CCZ, CCV and CCC, drop CCZNV.
7856 * config/h8300/h8300.md (H8cc mode iterator): Add CCZ.
7857 (cc mode_attr): Similarly.
7858 (ccz subst_attr): Similarly.
7859 * config/h8300/jumpcall.md: Add new patterns for branch-on-bit.
7860 * config/h8300/testcompare.md: Remove various cc0 based patterns
7861 that had been commented out. Add pattern to set CCZ from a bit
7864 2021-07-29 Thomas Schwinge <thomas@codesourcery.com>
7865 Julian Brown <julian@codesourcery.com>
7866 Kwok Cheung Yeung <kcy@codesourcery.com>
7868 * omp-offload.c (oacc_loop_xform_head_tail, oacc_loop_process):
7869 'update_stmt' after modification.
7870 (pass_oacc_loop_designation): New function, extracted out of...
7871 (pass_oacc_device_lower): ... this.
7872 (pass_data_oacc_loop_designation, pass_oacc_loop_designation)
7873 (make_pass_oacc_loop_designation): New
7874 * passes.def: Add it.
7875 * tree-parloops.c (create_parallel_loop): Adjust.
7876 * tree-pass.h (make_pass_oacc_loop_designation): New.
7878 2021-07-29 Aldy Hernandez <aldyh@redhat.com>
7880 * flag-types.h (enum threader_mode): New.
7881 * params.opt: Add entry for --param=threader-mode.
7882 * tree-ssa-threadbackward.c (THREADER_ITERATIVE_MODE): New.
7883 (class back_threader): New.
7884 (back_threader::back_threader): New.
7885 (back_threader::~back_threader): New.
7886 (back_threader::maybe_register_path): New.
7887 (back_threader::find_taken_edge): New.
7888 (back_threader::find_taken_edge_switch): New.
7889 (back_threader::find_taken_edge_cond): New.
7890 (back_threader::resolve_def): New.
7891 (back_threader::resolve_phi): New.
7892 (back_threader::find_paths_to_names): New.
7893 (back_threader::find_paths): New.
7896 (thread_jumps::find_jump_threads_backwards): Call ranger threader.
7897 (thread_jumps::find_jump_threads_backwards_with_ranger): New.
7898 (pass_thread_jumps::execute): Abstract out code...
7899 (try_thread_blocks): ...here.
7900 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
7901 Abstract out threading candidate code to...
7902 (single_succ_to_potentially_threadable_block): ...here.
7903 * tree-ssa-threadedge.h (single_succ_to_potentially_threadable_block):
7905 * tree-ssa-threadupdate.c (register_jump_thread): Return boolean.
7906 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
7907 Return bool from register_jump_thread.
7909 2021-07-29 Andreas Krebbel <krebbel@linux.ibm.com>
7911 * target.def: in0 and in1 do not need to be registers.
7912 * doc/tm.texi: Regenerate.
7914 2021-07-29 liuhongt <hongtao.liu@intel.com>
7917 * config/i386/i386.c (ix86_widen_mult_cost): New function.
7918 (ix86_add_stmt_cost): Use ix86_widen_mult_cost for
7921 2021-07-29 Jiufu Guo <guojiufu@linux.ibm.com>
7924 * config/rs6000/rs6000.c (TARGET_PREFERRED_DOLOOP_MODE): New hook.
7925 (rs6000_preferred_doloop_mode): New hook.
7926 * doc/tm.texi: Regenerate.
7927 * doc/tm.texi.in: Add hook preferred_doloop_mode.
7928 * target.def (preferred_doloop_mode): New hook.
7929 * targhooks.c (default_preferred_doloop_mode): New hook.
7930 * targhooks.h (default_preferred_doloop_mode): New hook.
7931 * tree-ssa-loop-ivopts.c (compute_doloop_base_on_mode): New function.
7932 (add_iv_candidate_for_doloop): Call targetm.preferred_doloop_mode
7933 and compute_doloop_base_on_mode.
7935 2021-07-28 Martin Sebor <msebor@redhat.com>
7937 PR middle-end/101494
7938 * tree-ssa-uninit.c (maybe_warn_operand): Correct object offset
7939 and size computation.
7941 2021-07-28 Martin Sebor <msebor@redhat.com>
7943 PR middle-end/101601
7944 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Remove
7946 Handle pointers to functions.
7948 2021-07-28 Martin Sebor <msebor@redhat.com>
7950 * Makefile.in (OBJS): Add gimple-ssa-warn-access.o and pointer-query.o.
7951 * attribs.h (fndecl_dealloc_argno): Move fndecl_dealloc_argno to tree.h.
7952 * builtins.c (compute_objsize_r): Move to pointer-query.cc.
7953 (access_ref::access_ref): Same.
7954 (access_ref::phi): Same.
7955 (access_ref::get_ref): Same.
7956 (access_ref::size_remaining): Same.
7957 (access_ref::offset_in_range): Same.
7958 (access_ref::add_offset): Same.
7959 (access_ref::inform_access): Same.
7960 (ssa_name_limit_t::visit_phi): Same.
7961 (ssa_name_limit_t::leave_phi): Same.
7962 (ssa_name_limit_t::next): Same.
7963 (ssa_name_limit_t::next_phi): Same.
7964 (ssa_name_limit_t::~ssa_name_limit_t): Same.
7965 (pointer_query::pointer_query): Same.
7966 (pointer_query::get_ref): Same.
7967 (pointer_query::put_ref): Same.
7968 (pointer_query::flush_cache): Same.
7969 (warn_string_no_nul): Move to gimple-ssa-warn-access.cc.
7970 (check_nul_terminated_array): Same.
7971 (unterminated_array): Same.
7972 (maybe_warn_for_bound): Same.
7973 (check_read_access): Same.
7974 (warn_for_access): Same.
7975 (get_size_range): Same.
7976 (check_access): Same.
7977 (gimple_call_alloc_size): Move to tree.c.
7978 (gimple_parm_array_size): Move to pointer-query.cc.
7979 (get_offset_range): Same.
7980 (gimple_call_return_array): Same.
7981 (handle_min_max_size): Same.
7982 (handle_array_ref): Same.
7983 (handle_mem_ref): Same.
7984 (compute_objsize): Same.
7985 (gimple_call_alloc_p): Move to gimple-ssa-warn-access.cc.
7986 (call_dealloc_argno): Same.
7987 (fndecl_dealloc_argno): Same.
7988 (new_delete_mismatch_p): Same.
7989 (matching_alloc_calls_p): Same.
7990 (warn_dealloc_offset): Same.
7991 (maybe_emit_free_warning): Same.
7992 * builtins.h (check_nul_terminated_array): Move to
7993 gimple-ssa-warn-access.h.
7994 (check_nul_terminated_array): Same.
7995 (warn_string_no_nul): Same.
7996 (unterminated_array): Same.
7997 (class ssa_name_limit_t): Same.
7998 (class pointer_query): Same.
7999 (struct access_ref): Same.
8000 (class range_query): Same.
8001 (struct access_data): Same.
8002 (gimple_call_alloc_size): Same.
8003 (gimple_parm_array_size): Same.
8004 (compute_objsize): Same.
8005 (class access_data): Same.
8006 (maybe_emit_free_warning): Same.
8007 * calls.c (initialize_argument_information): Remove call to
8008 maybe_emit_free_warning.
8009 * gimple-array-bounds.cc: Include new header..
8010 * gimple-fold.c: Same.
8011 * gimple-ssa-sprintf.c: Same.
8012 * gimple-ssa-warn-restrict.c: Same.
8013 * passes.def: Add pass_warn_access.
8014 * tree-pass.h (make_pass_warn_access): Declare.
8015 * tree-ssa-strlen.c: Include new headers.
8016 * tree.c (fndecl_dealloc_argno): Move here from builtins.c.
8017 * tree.h (fndecl_dealloc_argno): Move here from attribs.h.
8018 * gimple-ssa-warn-access.cc: New file.
8019 * gimple-ssa-warn-access.h: New file.
8020 * pointer-query.cc: New file.
8021 * pointer-query.h: New file.
8023 2021-07-28 Jakub Jelinek <jakub@redhat.com>
8025 PR middle-end/101624
8026 * ubsan.c (maybe_instrument_pointer_overflow,
8027 instrument_object_size): Only test DECL_REGISTER on VAR_DECLs,
8028 PARM_DECLs or RESULT_DECLs.
8029 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
8031 2021-07-28 Jakub Jelinek <jakub@redhat.com>
8033 PR middle-end/101642
8034 * match.pd (bswap16 (x) == bswap16 (y)): Cast both operands
8035 to type of bswap16 for comparison.
8036 (bswap16 (x) == cst): Cast bswap16 operand to type of cst.
8038 2021-07-28 Richard Biener <rguenther@suse.de>
8040 PR tree-optimization/101615
8041 * tree-vect-slp.c (vect_optimize_slp): Materialize permutes
8042 at CTOR SLP graph entries.
8044 2021-07-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8046 * config/aarch64/aarch64.md (*extend<SHORT:mode><GPI:mode>2_aarch64):
8047 Add "r,w" alternative.
8049 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
8052 * config/i386/i386.c (ix86_avx_u128_mode_needed): Don't set
8053 AVX_U128_DIRTY when all bits are zero.
8055 2021-07-28 Richard Biener <rguenther@suse.de>
8057 PR tree-optimization/101615
8058 * tree-vect-slp.c (vect_optimize_slp): Pre-existing vector
8059 external nodes cannot be permuted so make them perm_out 0.
8061 2021-07-28 Andrew Stubbs <ams@codesourcery.com>
8064 * config.in: Regenerate.
8065 * config/gcn/gcn-hsa.h (A_FIJI): New define.
8066 (A_900): New define.
8067 (A_906): New define.
8068 (A_908): New define.
8069 (ASM_SPEC): Use A_FIJI, A_900, A_906 and A_908.
8070 * config/gcn/gcn.c (output_file_start): Adjust attributes according
8071 to the assembler capabilities.
8072 * config/gcn/mkoffload.c (main): Likewise.
8073 * configure: Regenerate.
8074 * configure.ac: Add tests for LLVM assembler attribute features.
8076 2021-07-28 Andrew MacLeod <amacleod@redhat.com>
8078 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Check for
8079 cond_false and cond_true on branches.
8081 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
8083 * config/aarch64/aarch64.c (aarch64_gen_adjusted_ldpstp): use
8086 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
8088 * alias.c (init_alias_analysis): Don't skip prologue/epilogue.
8090 2021-07-28 Jakub Jelinek <jakub@redhat.com>
8093 * config/i386/sse.md (vashr<mode>3): Split into vashrv8di3 expander
8094 and vashrv4di3 expander, where the latter requires just TARGET_AVX2
8095 and has special !TARGET_AVX512VL expansion.
8096 (vashrv2di3<mask_name>): Rename to ...
8097 (vashrv2di3): ... this. Change condition to TARGET_XOP || TARGET_AVX2
8098 and add special !TARGET_XOP && !TARGET_AVX512VL expansion.
8100 2021-07-28 Martin Uecker <muecker@gwdg.de>
8102 * calls.c (maybe_warn_rdwr_sizes): Correct argument
8103 numbers in warning that were switched.
8105 2021-07-28 Kewen Lin <linkw@linux.ibm.com>
8107 PR tree-optimization/101596
8108 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Fix wrong check
8109 by using new_type's precision instead.
8111 2021-07-28 liuhongt <hongtao.liu@intel.com>
8114 * config/i386/i386.h (processor_costs): Add new member
8116 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
8117 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
8118 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
8119 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
8120 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
8121 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
8122 generic_cost, core_cost): Initialize integer_to_sse same value
8124 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
8125 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8126 Use integer_to_sse instead of sse_op to calculate the cost of
8129 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
8131 * config/rs6000/rs6000-gen-builtins.c (write_ovld_static_init): New
8133 (write_init_file): Call write_ovld_static_init.
8135 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
8137 * config/rs6000/rs6000-gen-builtins.c (write_bif_static_init): New
8139 (write_init_file): Call write_bif_static_init.
8141 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
8143 * config/rs6000/rs6000-gen-builtins.c (typemap): New struct.
8144 (TYPE_MAP_SIZE): New macro.
8145 (type_map): New initialized variable.
8146 (typemap_cmp): New function.
8147 (write_type_node): Likewise.
8148 (write_fntype_init): Implement.
8150 2021-07-27 Martin Sebor <msebor@redhat.com>
8152 PR tree-optimization/101584
8153 * tree-ssa-uninit.c (builtin_call_nomodifying_p): New function.
8154 (check_defs): Call it.
8156 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
8158 * tree-ssa-dom.c (dom_jump_threader_simplifier):
8159 Put avail_exprs_stack in the class, instead of passing it to
8160 jump_threader_simplifier.
8161 (dom_jump_threader_simplifier::simplify): Add state argument.
8162 (dom_opt_dom_walker): Add state.
8163 (pass_dominator::execute): Pass state to threader.
8164 (dom_opt_dom_walker::before_dom_children): Use state.
8165 * tree-ssa-threadedge.c (jump_threader::jump_threader): Replace
8167 (jump_threader::record_temporary_equivalences_from_phis):
8168 Register equivalences through the state variable.
8169 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
8170 Record ranges in a statement through the state variable.
8171 (jump_threader::simplify_control_stmt_condition): Pass state to
8173 (jump_threader::simplify_control_stmt_condition_1): Same.
8174 (jump_threader::thread_around_empty_blocks): Remove obsolete
8176 (jump_threader::thread_through_normal_block): Record equivalences
8177 on edge through the state variable.
8178 (jump_threader::thread_across_edge): Abstract state pushing.
8179 (jt_state::jt_state): New.
8180 (jt_state::push): New.
8181 (jt_state::pop): New.
8182 (jt_state::register_equiv): New.
8183 (jt_state::record_ranges_from_stmt): New.
8184 (jt_state::register_equivs_on_edge): New.
8185 (jump_threader_simplifier::jump_threader_simplifier): Move from
8187 (jump_threader_simplifier::simplify): Add state argument.
8188 * tree-ssa-threadedge.h (class jt_state): New.
8189 (class jump_threader): Add state to constructor.
8190 (class jump_threader_simplifier): Add state to simplify. Remove
8191 avail_exprs_stack from class.
8192 * tree-vrp.c (vrp_jump_threader_simplifier::simplify): Add state
8194 (vrp_jump_threader::vrp_jump_threader): Add state.
8195 (vrp_jump_threader::~vrp_jump_threader): Cleanup state.
8197 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
8199 * Makefile.in (OBJS): Add gimple-range-path.o.
8200 * gimple-range-path.cc: New file.
8201 * gimple-range-path.h: New file.
8203 2021-07-27 Jonathan Wright <jonathan.wright@arm.com>
8205 * config/aarch64/aarch64-simd.md: Push sign/zero-extension
8206 inside vec_duplicate for all patterns.
8207 * simplify-rtx.c (simplify_context::simplify_unary_operation_1):
8208 Push sign/zero-extension inside vec_duplicate.
8210 2021-07-27 Richard Biener <rguenther@suse.de>
8212 PR tree-optimization/101573
8213 * tree-ssa-uninit.c (warn_uninit_phi_uses): New function
8214 looking at uninitialized PHI arg defs in some constrained cases.
8215 (warn_uninitialized_vars): Call it.
8216 (execute_early_warn_uninitialized): Calculate dominators.
8218 2021-07-27 Richard Biener <rguenther@suse.de>
8220 PR tree-optimization/39821
8221 * tree-vect-stmts.c (vect_model_promotion_demotion_cost): Use
8222 vector_stmt for widening arithmetic.
8223 (vectorizable_conversion): Adjust.
8225 2021-07-27 Martin Jambor <mjambor@suse.cz>
8227 * cgraph.h (ipa_replace_map): New field force_load_ref.
8228 * ipa-prop.h (ipa_param_descriptor): Reduce precision of move_cost,
8229 aded new flag load_dereferenced, adjusted comments.
8230 (ipa_get_param_dereferenced): New function.
8231 (ipa_set_param_dereferenced): Likewise.
8232 * cgraphclones.c (cgraph_node::create_virtual_clone): Follow it.
8233 * ipa-cp.c: Include gimple.h.
8234 (ipcp_discover_new_direct_edges): Take into account dereferenced flag.
8235 (get_replacement_map): New parameter force_load_ref, set the
8236 appropriate flag in ipa_replace_map if set.
8237 (struct symbol_and_index_together): New type.
8238 (adjust_refs_in_act_callers): New function.
8239 (adjust_references_in_caller): Likewise.
8240 (create_specialized_node): When appropriate, call
8241 adjust_references_in_caller and force only load references.
8242 * ipa-prop.c (load_from_dereferenced_name): New function.
8243 (ipa_analyze_controlled_uses): Also detect loads from a
8244 dereference, harden testing of call statements.
8245 (ipa_write_node_info): Stream the dereferenced flag.
8246 (ipa_read_node_info): Likewise.
8247 (ipa_set_jf_constant): Also create refdesc when jump function
8248 references a variable.
8249 (cgraph_node_for_jfunc): Rename to symtab_node_for_jfunc, work
8250 also on references of variables and return a symtab_node. Adjust
8252 (propagate_controlled_uses): Also remove references to VAR_DECLs.
8254 2021-07-27 Jakub Jelinek <jakub@redhat.com>
8256 PR middle-end/101586
8257 * gimple-fold.c (clear_padding_type): Ignore FIELD_DECLs with byte
8258 positions above or equal to sz except for diagnostics of flexible
8261 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
8263 PR tree-optimization/78888
8264 * gimple-range-fold.cc (get_letter_range): New.
8265 (fold_using_range::range_of_builtin_call): Call get_letter_range.
8267 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
8269 PR tree-optimization/78888
8270 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Add cases
8271 for CFN_BUILT_IN_TOUPPER and CFN_BUILT_IN_TOLOWER.
8273 2021-07-26 Roger Sayle <roger@nextmovesoftware.com>
8274 Marc Glisse <marc.glisse@inria.fr>
8276 * match.pd (rotate): Simplify equality/inequality of rotations.
8277 (bswap): Simplify equality/inequality tests of byte swapping.
8279 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8281 * range-op.cc (operator_bitwise_xor::op1_op2_relation_effect):
8284 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8286 * range-op.cc (operator_lshift::fold_range): Pass rel to
8287 base class fold_range.
8288 (operator_rshift::fold_range): Same.
8290 2021-07-26 Ashimida <ashimida@linux.alibaba.com>
8293 * toplev.h (min_align_loops_log): Remove declaration.
8294 (min_align_jumps_log, min_align_labels_log): Likewise.
8295 (min_align_functions_log): Likewise.
8297 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8299 * tree-vrp.c (vrp_simplify_cond_using_ranges): Rename vr_values
8301 (execute_vrp): Abstract out simplification of conditionals...
8302 (simplify_casted_conds): ...here.
8304 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8306 * gimple-array-bounds.cc (array_bounds_checker::get_value_range):
8307 Add gimple argument.
8308 (array_bounds_checker::check_array_ref): Same.
8309 (array_bounds_checker::check_addr_expr): Same.
8310 (array_bounds_checker::check_array_bounds): Pass statement to
8311 check_array_bounds and check_addr_expr.
8312 * gimple-array-bounds.h (check_array_bounds): Add gimple argument.
8313 (check_addr_expr): Same.
8314 (get_value_range): Same.
8316 2021-07-26 Tamar Christina <tamar.christina@arm.com>
8318 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): Rename to..
8319 (sdot_prod, udot_prod): ... This.
8320 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): Merged
8322 (<sur>dot_prod<vsi2qi>): ... this.
8323 (aarch64_<sur>dot_lane<vsi2qi>, aarch64_<sur>dot_laneq<vsi2qi>):
8324 Change operands order.
8325 (<sur>sadv16qi): Use new operands order.
8326 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32,
8327 vdotq_s32): Use new RTL ordering.
8329 2021-07-26 Tamar Christina <tamar.christina@arm.com>
8331 * config/aarch64/aarch64-builtins.c (TYPES_TERNOP_SUSS,
8332 aarch64_types_ternop_suss_qualifiers): New.
8333 * config/aarch64/aarch64-simd-builtins.def (usdot_prod): Use it.
8334 * config/aarch64/aarch64-simd.md (usdot_prod<vsi2qi>): Re-organize RTL.
8335 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Use it.
8337 2021-07-23 Jakub Jelinek <jakub@redhat.com>
8339 PR rtl-optimization/101562
8340 * expmed.c (store_integral_bit_field): Only use movstrict_optab
8341 if the operand isn't paradoxical.
8343 2021-07-23 Aldy Hernandez <aldyh@redhat.com>
8345 * gimple-array-bounds.h (class array_bounds_checker): Change
8346 ranges type to range_query.
8348 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8350 * config/aarch64/arm_neon.h (vst1_s64_x2): Use
8351 __builtin_memcpy instead of constructing
8352 __builtin_aarch64_simd_oi one vector at a time.
8353 (vst1_u64_x2): Likewise.
8354 (vst1_f64_x2): Likewise.
8355 (vst1_s8_x2): Likewise.
8356 (vst1_p8_x2): Likewise.
8357 (vst1_s16_x2): Likewise.
8358 (vst1_p16_x2): Likewise.
8359 (vst1_s32_x2): Likewise.
8360 (vst1_u8_x2): Likewise.
8361 (vst1_u16_x2): Likewise.
8362 (vst1_u32_x2): Likewise.
8363 (vst1_f16_x2): Likewise.
8364 (vst1_f32_x2): Likewise.
8365 (vst1_p64_x2): Likewise.
8366 (vst1q_s8_x2): Likewise.
8367 (vst1q_p8_x2): Likewise.
8368 (vst1q_s16_x2): Likewise.
8369 (vst1q_p16_x2): Likewise.
8370 (vst1q_s32_x2): Likewise.
8371 (vst1q_s64_x2): Likewise.
8372 (vst1q_u8_x2): Likewise.
8373 (vst1q_u16_x2): Likewise.
8374 (vst1q_u32_x2): Likewise.
8375 (vst1q_u64_x2): Likewise.
8376 (vst1q_f16_x2): Likewise.
8377 (vst1q_f32_x2): Likewise.
8378 (vst1q_f64_x2): Likewise.
8379 (vst1q_p64_x2): Likewise.
8381 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8383 * config/aarch64/arm_neon.h (vst1_s64_x3): Use
8384 __builtin_memcpy instead of constructing
8385 __builtin_aarch64_simd_ci one vector at a time.
8386 (vst1_u64_x3): Likewise.
8387 (vst1_f64_x3): Likewise.
8388 (vst1_s8_x3): Likewise.
8389 (vst1_p8_x3): Likewise.
8390 (vst1_s16_x3): Likewise.
8391 (vst1_p16_x3): Likewise.
8392 (vst1_s32_x3): Likewise.
8393 (vst1_u8_x3): Likewise.
8394 (vst1_u16_x3): Likewise.
8395 (vst1_u32_x3): Likewise.
8396 (vst1_f16_x3): Likewise.
8397 (vst1_f32_x3): Likewise.
8398 (vst1_p64_x3): Likewise.
8399 (vst1q_s8_x3): Likewise.
8400 (vst1q_p8_x3): Likewise.
8401 (vst1q_s16_x3): Likewise.
8402 (vst1q_p16_x3): Likewise.
8403 (vst1q_s32_x3): Likewise.
8404 (vst1q_s64_x3): Likewise.
8405 (vst1q_u8_x3): Likewise.
8406 (vst1q_u16_x3): Likewise.
8407 (vst1q_u32_x3): Likewise.
8408 (vst1q_u64_x3): Likewise.
8409 (vst1q_f16_x3): Likewise.
8410 (vst1q_f32_x3): Likewise.
8411 (vst1q_f64_x3): Likewise.
8412 (vst1q_p64_x3): Likewise.
8414 2021-07-23 H.J. Lu <hjl.tools@gmail.com>
8417 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): Don't return
8418 hard register when LRA is in progress.
8420 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8422 * config/aarch64/arm_neon.h (vst1_s8_x4): Use
8423 __builtin_memcpy instead of using a union.
8424 (vst1q_s8_x4): Likewise.
8425 (vst1_s16_x4): Likewise.
8426 (vst1q_s16_x4): Likewise.
8427 (vst1_s32_x4): Likewise.
8428 (vst1q_s32_x4): Likewise.
8429 (vst1_u8_x4): Likewise.
8430 (vst1q_u8_x4): Likewise.
8431 (vst1_u16_x4): Likewise.
8432 (vst1q_u16_x4): Likewise.
8433 (vst1_u32_x4): Likewise.
8434 (vst1q_u32_x4): Likewise.
8435 (vst1_f16_x4): Likewise.
8436 (vst1q_f16_x4): Likewise.
8437 (vst1_f32_x4): Likewise.
8438 (vst1q_f32_x4): Likewise.
8439 (vst1_p8_x4): Likewise.
8440 (vst1q_p8_x4): Likewise.
8441 (vst1_p16_x4): Likewise.
8442 (vst1q_p16_x4): Likewise.
8443 (vst1_s64_x4): Likewise.
8444 (vst1_u64_x4): Likewise.
8445 (vst1_p64_x4): Likewise.
8446 (vst1q_s64_x4): Likewise.
8447 (vst1q_u64_x4): Likewise.
8448 (vst1q_p64_x4): Likewise.
8449 (vst1_f64_x4): Likewise.
8450 (vst1q_f64_x4): Likewise.
8452 2021-07-23 Jonathan Wrightt <jonathan.wright@arm.com>
8454 * config/aarch64/arm_neon.h (vst2_s64): Use __builtin_memcpy
8455 instead of constructing __builtin_aarch64_simd_oi one vector
8457 (vst2_u64): Likewise.
8458 (vst2_f64): Likewise.
8459 (vst2_s8): Likewise.
8460 (vst2_p8): Likewise.
8461 (vst2_s16): Likewise.
8462 (vst2_p16): Likewise.
8463 (vst2_s32): Likewise.
8464 (vst2_u8): Likewise.
8465 (vst2_u16): Likewise.
8466 (vst2_u32): Likewise.
8467 (vst2_f16): Likewise.
8468 (vst2_f32): Likewise.
8469 (vst2_p64): Likewise.
8470 (vst2q_s8): Likewise.
8471 (vst2q_p8): Likewise.
8472 (vst2q_s16): Likewise.
8473 (vst2q_p16): Likewise.
8474 (vst2q_s32): Likewise.
8475 (vst2q_s64): Likewise.
8476 (vst2q_u8): Likewise.
8477 (vst2q_u16): Likewise.
8478 (vst2q_u32): Likewise.
8479 (vst2q_u64): Likewise.
8480 (vst2q_f16): Likewise.
8481 (vst2q_f32): Likewise.
8482 (vst2q_f64): Likewise.
8483 (vst2q_p64): Likewise.
8485 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8487 * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy
8488 instead of constructing __builtin_aarch64_simd_ci one vector
8490 (vst3_u64): Likewise.
8491 (vst3_f64): Likewise.
8492 (vst3_s8): Likewise.
8493 (vst3_p8): Likewise.
8494 (vst3_s16): Likewise.
8495 (vst3_p16): Likewise.
8496 (vst3_s32): Likewise.
8497 (vst3_u8): Likewise.
8498 (vst3_u16): Likewise.
8499 (vst3_u32): Likewise.
8500 (vst3_f16): Likewise.
8501 (vst3_f32): Likewise.
8502 (vst3_p64): Likewise.
8503 (vst3q_s8): Likewise.
8504 (vst3q_p8): Likewise.
8505 (vst3q_s16): Likewise.
8506 (vst3q_p16): Likewise.
8507 (vst3q_s32): Likewise.
8508 (vst3q_s64): Likewise.
8509 (vst3q_u8): Likewise.
8510 (vst3q_u16): Likewise.
8511 (vst3q_u32): Likewise.
8512 (vst3q_u64): Likewise.
8513 (vst3q_f16): Likewise.
8514 (vst3q_f32): Likewise.
8515 (vst3q_f64): Likewise.
8516 (vst3q_p64): Likewise.
8518 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8520 * config/aarch64/arm_neon.h (vst4_s64): Use __builtin_memcpy
8521 instead of constructing __builtin_aarch64_simd_xi one vector
8523 (vst4_u64): Likewise.
8524 (vst4_f64): Likewise.
8525 (vst4_s8): Likewise.
8526 (vst4_p8): Likewise.
8527 (vst4_s16): Likewise.
8528 (vst4_p16): Likewise.
8529 (vst4_s32): Likewise.
8530 (vst4_u8): Likewise.
8531 (vst4_u16): Likewise.
8532 (vst4_u32): Likewise.
8533 (vst4_f16): Likewise.
8534 (vst4_f32): Likewise.
8535 (vst4_p64): Likewise.
8536 (vst4q_s8): Likewise.
8537 (vst4q_p8): Likewise.
8538 (vst4q_s16): Likewise.
8539 (vst4q_p16): Likewise.
8540 (vst4q_s32): Likewise.
8541 (vst4q_s64): Likewise.
8542 (vst4q_u8): Likewise.
8543 (vst4q_u16): Likewise.
8544 (vst4q_u32): Likewise.
8545 (vst4q_u64): Likewise.
8546 (vst4q_f16): Likewise.
8547 (vst4q_f32): Likewise.
8548 (vst4q_f64): Likewise.
8549 (vst4q_p64): Likewise.
8551 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8553 * config/aarch64/arm_neon.h (vtbx4_s8): Use __builtin_memcpy
8554 instead of constructing __builtin_aarch64_simd_oi one vector
8556 (vtbx4_u8): Likewise.
8557 (vtbx4_p8): Likewise.
8559 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8561 * config/aarch64/arm_neon.h (vtbl3_s8): Use __builtin_memcpy
8562 instead of constructing __builtin_aarch64_simd_oi one vector
8564 (vtbl3_u8): Likewise.
8565 (vtbl3_p8): Likewise.
8566 (vtbl4_s8): Likewise.
8567 (vtbl4_u8): Likewise.
8568 (vtbl4_p8): Likewise.
8570 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8572 * config/aarch64/arm_neon.h (vqtbx2_s8): Use __builtin_memcpy
8573 instead of constructing __builtin_aarch64_simd_oi one vector
8575 (vqtbx2_u8): Likewise.
8576 (vqtbx2_p8): Likewise.
8577 (vqtbx2q_s8): Likewise.
8578 (vqtbx2q_u8): Likewise.
8579 (vqtbx2q_p8): Likewise.
8580 (vqtbx3_s8): Use __builtin_memcpy instead of constructing
8581 __builtin_aarch64_simd_ci one vector at a time.
8582 (vqtbx3_u8): Likewise.
8583 (vqtbx3_p8): Likewise.
8584 (vqtbx3q_s8): Likewise.
8585 (vqtbx3q_u8): Likewise.
8586 (vqtbx3q_p8): Likewise.
8587 (vqtbx4_s8): Use __builtin_memcpy instead of constructing
8588 __builtin_aarch64_simd_xi one vector at a time.
8589 (vqtbx4_u8): Likewise.
8590 (vqtbx4_p8): Likewise.
8591 (vqtbx4q_s8): Likewise.
8592 (vqtbx4q_u8): Likewise.
8593 (vqtbx4q_p8): Likewise.
8595 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8597 * config/aarch64/arm_neon.h (vqtbl2_s8): Use __builtin_memcpy
8598 instead of constructing __builtin_aarch64_simd_oi one vector
8600 (vqtbl2_u8): Likewise.
8601 (vqtbl2_p8): Likewise.
8602 (vqtbl2q_s8): Likewise.
8603 (vqtbl2q_u8): Likewise.
8604 (vqtbl2q_p8): Likewise.
8605 (vqtbl3_s8): Use __builtin_memcpy instead of constructing
8606 __builtin_aarch64_simd_ci one vector at a time.
8607 (vqtbl3_u8): Likewise.
8608 (vqtbl3_p8): Likewise.
8609 (vqtbl3q_s8): Likewise.
8610 (vqtbl3q_u8): Likewise.
8611 (vqtbl3q_p8): Likewise.
8612 (vqtbl4_s8): Use __builtin_memcpy instead of constructing
8613 __builtin_aarch64_simd_xi one vector at a time.
8614 (vqtbl4_u8): Likewise.
8615 (vqtbl4_p8): Likewise.
8616 (vqtbl4q_s8): Likewise.
8617 (vqtbl4q_u8): Likewise.
8618 (vqtbl4q_p8): Likewise.
8620 2021-07-23 Haochen Gui <guihaoc@gcc.gnu.org>
8623 * config/rs6000/rs6000.md (cstore<mode>4): Fix wrong fall through.
8625 2021-07-22 Andrew Pinski <apinski@marvell.com>
8627 PR tree-optimization/10153
8628 * tree-tailcall.c (create_tailcall_accumulator):
8629 Don't call fold_convert as the type should be correct already.
8630 (tree_optimize_tail_calls_1): Use build_{one,zero}_cst instead
8631 of integer_{one,zero}_node for the call of create_tailcall_accumulator.
8633 2021-07-22 Aldy Hernandez <aldyh@redhat.com>
8635 * gimple-range-cache.cc (non_null_ref::adjust_range): Replace
8636 varying_p check for null/non-null check.
8638 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
8640 PR tree-optimization/101511
8641 * value-relation.cc (relation_oracle::query_relation): Check if ssa1
8642 is in ssa2's equiv set, and don't trap if so.
8644 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
8646 PR tree-optimization/101497
8647 * gimple-range-fold.cc (fold_using_range::range_of_cond_expr): Check
8650 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
8652 PR tree-optimization/101496
8653 * vr-values.c (simplify_using_ranges::fold_cond): Call range_of_stmt
8654 first, then vrp_visit_cond_Stmt.
8656 2021-07-22 liuhongt <hongtao.liu@intel.com>
8658 * config/i386/i386-expand.c
8659 (ix86_broadcast_from_integer_constant): Rename to ..
8660 (ix86_broadcast_from_constant): .. this, and extend it to
8662 (ix86_expand_vector_move): Extend to float mode.
8663 * config/i386/i386-features.c
8664 (replace_constant_pool_with_broadcast): Remove.
8665 (remove_partial_avx_dependency_gate): Ditto.
8666 (constant_pool_broadcast): Ditto.
8667 (class pass_constant_pool_broadcast): Ditto.
8668 (make_pass_constant_pool_broadcast): Ditto.
8669 (remove_partial_avx_dependency): Adjust gate.
8670 * config/i386/i386-passes.def: Remove pass_constant_pool_broadcast.
8671 * config/i386/i386-protos.h
8672 (make_pass_constant_pool_broadcast): Remove.
8674 2021-07-22 liuhongt <hongtao.liu@intel.com>
8676 * config/i386/constraints.md (Wb): New constraint.
8678 * config/i386/i386.md (*ashlhi3_1): Extend to avx512 mask
8680 (*ashlqi3_1): Ditto.
8681 (*<insn><mode>3_1): Split to ..
8682 (*ashr<mode>3_1): this, ...
8683 (*lshr<mode>3_1): and this, also extend this pattern to avx512
8685 (*<insn><mode>3_1): Split to ..
8686 (*ashr<mode>3_1): this, ...
8687 (*lshrqi3_1): and this, also extend this pattern to avx512
8689 (*lshrhi3_1): And this, also extend this pattern to avx512
8691 * config/i386/sse.md (k<code><mode>): New define_split after
8692 it to convert generic shift pattern to mask shift ones.
8694 2021-07-21 Thomas Schwinge <thomas@codesourcery.com>
8695 Joseph Myers <joseph@codesourcery.com>
8696 Cesar Philippidis <cesar@codesourcery.com>
8698 * tree-core.h (omp_clause_code): Add 'OMP_CLAUSE_NOHOST'.
8699 * tree.c (omp_clause_num_ops, omp_clause_code_name, walk_tree_1):
8701 * tree-pretty-print.c (dump_omp_clause): Likewise.
8702 * omp-general.c (oacc_verify_routine_clauses): Likewise.
8703 * gimplify.c (gimplify_scan_omp_clauses)
8704 (gimplify_adjust_omp_clauses): Likewise.
8705 * tree-nested.c (convert_nonlocal_omp_clauses)
8706 (convert_local_omp_clauses): Likewise.
8707 * omp-low.c (scan_sharing_clauses): Likewise.
8708 * omp-offload.c (execute_oacc_device_lower): Update.
8710 2021-07-21 Martin Sebor <msebor@redhat.com>
8712 * tree-ssa-alias.c (walk_aliased_vdefs_1): Fix typos in a comment.
8714 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8716 * config/rs6000/rs6000-gen-builtins.c (write_init_bif_table):
8719 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8721 * config/rs6000/rs6000-gen-builtins.c (write_fntype): New
8723 (write_fntype_init): New stub function.
8724 (write_init_bif_table): Likewise.
8725 (write_init_ovld_table): New function.
8726 (write_init_file): Implement.
8728 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8730 * config/rs6000/rs6000-gen-builtins.c
8731 (write_autogenerated_header): New function.
8732 (write_decls): Likewise.
8733 (write_extern_fntype): New callback function.
8734 (write_header_file): Implement.
8736 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8738 * config/rs6000/rs6000-gen-builtins.c (write_defines_file):
8741 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8743 * config/rs6000/rs6000-gen-builtins.c (complete_vector_type): New
8745 (complete_base_type): Likewise.
8746 (construct_fntype_id): Likewise.
8747 (parse_bif_entry): Call contruct_fntype_id.
8748 (parse_ovld_entry): Likewise.
8750 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8752 * config/rs6000/rs6000-gen-builtins.c (ovld_stanza): New struct.
8753 (MAXOVLDSTANZAS): New macro.
8754 (ovld_stanzas): New variable.
8755 (curr_ovld_stanza): Likewise.
8756 (MAXOVLDS): New macro.
8757 (ovlddata): New struct.
8758 (ovlds): New variable.
8759 (curr_ovld): Likewise.
8760 (max_ovld_args): Likewise.
8761 (parse_ovld_entry): New function.
8762 (parse_ovld_stanza): Likewise.
8763 (parse_ovld): Implement.
8765 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8767 * config/rs6000/rs6000-gen-builtins.c (parse_bif_attrs):
8770 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8772 * config/rs6000/rs6000-gen-builtins.c (parse_args): New function.
8773 (parse_prototype): Implement.
8775 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8777 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): New enum.
8778 (curr_bif_stanza): New variable.
8779 (stanza_entry): New struct.
8780 (stanza_map): New initialized variable.
8781 (enable_string): Likewise.
8782 (fnkinds): New enum.
8783 (typelist): New struct.
8784 (attrinfo): Likewise.
8785 (MAXRESTROPNDS): New macro.
8786 (prototype): New struct.
8787 (MAXBIFS): New macro.
8788 (bifdata): New struct.
8789 (bifs): New variable.
8790 (curr_bif): Likewise.
8791 (bif_order): Likewise.
8792 (bif_index): Likewise.
8793 (fatal): New function.
8794 (stanza_name_to_stanza): Likewise.
8795 (parse_bif_attrs): New stub function.
8796 (parse_prototype): Likewise.
8797 (parse_bif_entry): New function.
8798 (parse_bif_stanza): Likewise.
8799 (parse_bif): Implement.
8800 (set_bif_order): New function.
8801 (create_bif_order): Implement.
8803 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8805 * config/rs6000/rs6000-gen-builtins.c (rbtree.h): New #include.
8806 (num_bifs): New variable.
8807 (num_ovld_stanzas): Likewise.
8808 (num_ovlds): Likewise.
8809 (parse_codes): New enum.
8810 (bif_rbt): New variable.
8811 (ovld_rbt): Likewise.
8812 (fntype_rbt): Likewise.
8813 (bifo_rbt): Likewise.
8814 (parse_bif): New stub function.
8815 (create_bif_order): Likewise.
8816 (parse_ovld): Likewise.
8817 (write_header_file): Likewise.
8818 (write_init_file): Likewise.
8819 (write_defines_file): Likewise.
8820 (delete_output_files): New function.
8823 2021-07-21 H.J. Lu <hjl.tools@gmail.com>
8826 * config/i386/i386-builtin.def: Remove OPTION_MASK_ISA_SSE4_2
8827 from CRC32 _builtin functions.
8829 2021-07-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
8831 * coverage.c (build_gcov_info_var_registration): Mark the object placed
8832 in the linker set as referenced so that it does not get optimized away.
8834 2021-07-21 Kito Cheng <kito.cheng@sifive.com>
8837 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
8839 * config.gcc (riscv*-*-*): Detect which python is available.
8841 2021-07-21 Jakub Jelinek <jakub@redhat.com>
8843 PR middle-end/101535
8844 * gimplify.c (omp_check_private): Properly skip ORT_TARGET_DATA
8845 contexts in which decl isn't privatized and for ORT_TARGET return
8846 false if decl is mapped.
8848 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
8850 * gimple-loop-jam.c: Include tree-ssa-sccvn.h.
8851 (tree_loop_unroll_and_jam): Run value-numbering on a loop that
8852 has been successfully unrolled.
8854 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
8856 * tree-ssa-loop-manip.c (determine_exit_conditions): Return a null
8857 exit condition if no tail loop is needed, and if the original exit
8858 condition should therefore be kept as-is.
8859 (tree_transform_and_unroll_loop): Handle that case here too.
8861 2021-07-21 Kewen Lin <linkw@linux.ibm.com>
8863 * tree-data-ref.c (free_dependence_relations): Adjust to pass vec
8865 (free_data_refs): Likewise.
8866 * tree-data-ref.h (free_dependence_relations): Likewise.
8867 (free_data_refs): Likewise.
8868 * tree-predcom.c (struct chain): Use auto_vec instead of vec for
8870 (struct component): Likewise.
8871 (pcom_worker::pcom_worker): Adjust for auto_vec and renaming changes.
8872 (pcom_worker::~pcom_worker): Likewise.
8873 (pcom_worker::release_chain): Adjust as auto_vec changes.
8874 (pcom_worker::loop): Rename to ...
8875 (pcom_worker::m_loop): ... this.
8876 (pcom_worker::datarefs): Rename to ...
8877 (pcom_worker::m_datarefs): ... this. Use auto_vec instead of vec.
8878 (pcom_worker::dependences): Rename to ...
8879 (pcom_worker::m_dependences): ... this. Use auto_vec instead of vec.
8880 (pcom_worker::chains): Rename to ...
8881 (pcom_worker::m_chains): ... this. Use auto_vec instead of vec.
8882 (pcom_worker::looparound_phis): Rename to ...
8883 (pcom_worker::m_looparound_phis): ... this. Use auto_vec instead of
8885 (pcom_worker::cache): Rename to ...
8886 (pcom_worker::m_cache): ... this. Use auto_vec instead of vec.
8887 (pcom_worker::release_chain): Adjust for auto_vec changes.
8888 (pcom_worker::release_chains): Adjust for auto_vec and renaming
8890 (release_component): Remove.
8891 (release_components): Adjust for release_component removal.
8892 (component_of): Adjust to use vec.
8893 (merge_comps): Likewise.
8894 (pcom_worker::aff_combination_dr_offset): Adjust for renaming changes.
8895 (pcom_worker::determine_offset): Likewise.
8896 (class comp_ptrs): Remove.
8897 (pcom_worker::split_data_refs_to_components): Adjust for renaming
8898 changes, for comp_ptrs removal with auto_vec.
8899 (pcom_worker::suitable_component_p): Adjust for renaming changes.
8900 (pcom_worker::filter_suitable_components): Adjust for release_component
8902 (pcom_worker::valid_initializer_p): Adjust for renaming changes.
8903 (pcom_worker::find_looparound_phi): Likewise.
8904 (pcom_worker::add_looparound_copies): Likewise.
8905 (pcom_worker::determine_roots_comp): Likewise.
8906 (pcom_worker::single_nonlooparound_use): Likewise.
8907 (pcom_worker::execute_pred_commoning_chain): Likewise.
8908 (pcom_worker::execute_pred_commoning): Likewise.
8909 (pcom_worker::try_combine_chains): Likewise.
8910 (pcom_worker::prepare_initializers_chain): Likewise.
8911 (pcom_worker::prepare_initializers): Likewise.
8912 (pcom_worker::prepare_finalizers_chain): Likewise.
8913 (pcom_worker::prepare_finalizers): Likewise.
8914 (pcom_worker::tree_predictive_commoning_loop): Likewise.
8916 2021-07-20 Martin Sebor <msebor@redhat.com>
8918 PR middle-end/101397
8919 * builtins.c (gimple_call_return_array): Add argument. Correct
8920 offsets for memchr, mempcpy, stpcpy, and stpncpy.
8921 (compute_objsize_r): Adjust offset computation for argument returning
8924 2021-07-20 Martin Sebor <msebor@redhat.com>
8926 PR middle-end/101300
8927 * tree-ssa-uninit.c (check_defs): Handle UBSAN built-ins.
8929 2021-07-20 Jeff Law <jlaw@localhost.localdomain>
8931 * function.c (assign_parm_setup_block): Use adjust_address instead
8932 of change_address to preserve MEM_EXPR and friends.
8934 2021-07-20 Martin Sebor <msebor@redhat.com>
8936 * cfgloop.h (single_likely_exit): Adjust by-value argument to
8938 * cfgloopanal.c (single_likely_exit): Same.
8939 * cgraph.h (struct cgraph_node): Same.
8940 * cgraphclones.c (cgraph_node::create_virtual_clone): Same.
8941 * genautomata.c (merge_states): Same.
8942 * genextract.c (VEC_char_to_string): Same.
8943 * genmatch.c (dt_node::gen_kids_1): Same.
8944 (walk_captures): Adjust by-value argument to by-reference.
8945 * gimple-ssa-store-merging.c (check_no_overlap): Adjust by-value argument
8946 to by-const-reference.
8947 * gimple.c (gimple_build_call_vec): Same.
8948 (gimple_build_call_internal_vec): Same.
8949 (gimple_build_switch): Same.
8950 (sort_case_labels): Same.
8951 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
8953 * gimple.h (gimple_build_call_vec): Adjust by-value argument to
8955 (gimple_build_call_internal_vec): Same.
8956 (gimple_build_switch): Same.
8957 (sort_case_labels): Same.
8958 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
8960 * haifa-sched.c (calc_priorities): Adjust by-value argument to
8962 (sched_init_luids): Same.
8963 (haifa_init_h_i_d): Same.
8964 * ipa-cp.c (ipa_get_indirect_edge_target_1): Same.
8965 (adjust_callers_for_value_intersection): Adjust by-value argument to
8967 (find_more_scalar_values_for_callers_subset): Adjust by-value argument to
8969 (find_more_contexts_for_caller_subset): Same.
8970 (find_aggregate_values_for_callers_subset): Same.
8971 (copy_useful_known_contexts): Same.
8972 * ipa-fnsummary.c (remap_edge_summaries): Same.
8973 (remap_freqcounting_predicate): Same.
8974 * ipa-inline.c (add_new_edges_to_heap): Adjust by-value argument to
8976 * ipa-predicate.c (predicate::remap_after_inlining): Adjust by-value argument
8977 to by-const-reference.
8978 * ipa-predicate.h (predicate::remap_after_inlining): Same.
8979 * ipa-prop.c (ipa_find_agg_cst_for_param): Same.
8980 * ipa-prop.h (ipa_find_agg_cst_for_param): Same.
8981 * ira-build.c (ira_loop_tree_body_rev_postorder): Same.
8982 * read-rtl.c (add_overload_instance): Same.
8983 * rtl.h (native_decode_rtx): Same.
8984 (native_decode_vector_rtx): Same.
8985 * sched-int.h (sched_init_luids): Same.
8986 (haifa_init_h_i_d): Same.
8987 * simplify-rtx.c (native_decode_vector_rtx): Same.
8988 (native_decode_rtx): Same.
8989 * tree-call-cdce.c (gen_shrink_wrap_conditions): Same.
8990 (shrink_wrap_one_built_in_call_with_conds): Same.
8991 (shrink_wrap_conditional_dead_built_in_calls): Same.
8992 * tree-data-ref.c (create_runtime_alias_checks): Same.
8993 (compute_all_dependences): Same.
8994 * tree-data-ref.h (compute_all_dependences): Same.
8995 (create_runtime_alias_checks): Same.
8996 (index_in_loop_nest): Same.
8997 * tree-if-conv.c (mask_exists): Same.
8998 * tree-loop-distribution.c (class loop_distribution): Same.
8999 (loop_distribution::create_rdg_vertices): Same.
9000 (dump_rdg_partitions): Same.
9001 (debug_rdg_partitions): Same.
9002 (partition_contains_all_rw): Same.
9003 (loop_distribution::distribute_loop): Same.
9004 * tree-parloops.c (oacc_entry_exit_ok_1): Same.
9005 (oacc_entry_exit_single_gang): Same.
9006 * tree-ssa-loop-im.c (hoist_memory_references): Same.
9007 (loop_suitable_for_sm): Same.
9008 * tree-ssa-loop-niter.c (bound_index): Same.
9009 * tree-ssa-reassoc.c (update_ops): Same.
9010 (swap_ops_for_binary_stmt): Same.
9011 (rewrite_expr_tree): Same.
9012 (rewrite_expr_tree_parallel): Same.
9013 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Same.
9014 * tree-ssa-sccvn.h (ao_ref_init_from_vn_reference): Same.
9015 * tree-ssa-structalias.c (process_all_all_constraints): Same.
9016 (make_constraints_to): Same.
9017 (handle_lhs_call): Same.
9018 (find_func_aliases_for_builtin_call): Same.
9019 (sort_fieldstack): Same.
9020 (check_for_overlaps): Same.
9021 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Same.
9022 (vect_create_cond_for_unequal_addrs): Same.
9023 (vect_create_cond_for_lower_bounds): Same.
9024 (vect_create_cond_for_alias_checks): Same.
9025 * tree-vect-slp-patterns.c (vect_validate_multiplication): Same.
9026 * tree-vect-slp.c (vect_analyze_slp_instance): Same.
9027 (vect_make_slp_decision): Same.
9028 (vect_slp_bbs): Same.
9029 (duplicate_and_interleave): Same.
9030 (vect_transform_slp_perm_load): Same.
9031 (vect_schedule_slp): Same.
9032 * tree-vectorizer.h (vect_transform_slp_perm_load): Same.
9033 (vect_schedule_slp): Same.
9034 (duplicate_and_interleave): Same.
9035 * tree.c (build_vector_from_ctor): Same.
9036 (build_vector): Same.
9037 (check_vector_cst): Same.
9038 (check_vector_cst_duplicate): Same.
9039 (check_vector_cst_fill): Same.
9040 (check_vector_cst_stepped): Same.
9041 * tree.h (build_vector_from_ctor): Same.
9043 2021-07-20 Jakub Jelinek <jakub@redhat.com>
9046 * config/rs6000/rs6000-protos.h (easy_altivec_constant): Change return
9047 type from bool to int.
9048 * config/rs6000/rs6000.c (vspltis_constant): Fix up handling the
9049 EASY_VECTOR_MSB case if either step or copies is not 1.
9050 (vspltis_shifted): Fix comment typo.
9051 (easy_altivec_constant): Change return type from bool to int, instead
9052 of returning true return byte size of the element mode that should be
9053 used to synthetize the constant.
9054 * config/rs6000/predicates.md (easy_vector_constant_msb): Require
9055 that vspltis_shifted is 0, handle the case where easy_altivec_constant
9056 assumes using different vector mode from CONST_VECTOR's mode.
9057 * config/rs6000/altivec.md (easy_vector_constant_msb splitter): Use
9058 easy_altivec_constant to determine mode in which -1 >> -1 should be
9059 performed, use rs6000_expand_vector_init instead of gen_vec_initv4sisi.
9061 2021-07-20 Richard Biener <rguenther@suse.de>
9064 * dwarf2out.h (dwarf_file_data): Add key member.
9065 * dwarf2out.c (dwarf_file_hasher::equal): Compare key.
9066 (dwarf_file_hasher::hash): Hash key.
9067 (lookup_filename): Remap the filename and store it in the
9068 filename member of dwarf_file_data when creating a new
9070 (file_name_acquire): Do not remap the filename again.
9071 (maybe_emit_file): Likewise.
9073 2021-07-20 Jonathan Wright <jonathan.wright@arm.com>
9075 * config/aarch64/aarch64-simd-builtins.def: Use two variant
9076 generators for all TBL/TBX intrinsics and rename to
9077 consistent forms: qtbl[1234] or qtbx[1234].
9078 * config/aarch64/aarch64-simd.md (aarch64_tbl1<mode>):
9080 (aarch64_qtbl1<mode>): This.
9081 (aarch64_tbx1<mode>): Rename to...
9082 (aarch64_qtbx1<mode>): This.
9083 (aarch64_tbl2v16qi): Delete.
9084 (aarch64_tbl3<mode>): Rename to...
9085 (aarch64_qtbl2<mode>): This.
9086 (aarch64_tbx4<mode>): Rename to...
9087 (aarch64_qtbx2<mode>): This.
9088 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_1): Use
9089 renamed qtbl1 and qtbl2 RTL patterns.
9090 * config/aarch64/arm_neon.h (vqtbl1_p8): Use renamed qtbl1
9092 (vqtbl1_s8): Likewise.
9093 (vqtbl1_u8): Likewise.
9094 (vqtbl1q_p8): Likewise.
9095 (vqtbl1q_s8): Likewise.
9096 (vqtbl1q_u8): Likewise.
9097 (vqtbx1_s8): Use renamed qtbx1 RTL pattern.
9098 (vqtbx1_u8): Likewise.
9099 (vqtbx1_p8): Likewise.
9100 (vqtbx1q_s8): Likewise.
9101 (vqtbx1q_u8): Likewise.
9102 (vqtbx1q_p8): Likewise.
9103 (vtbl1_s8): Use renamed qtbl1 RTL pattern.
9104 (vtbl1_u8): Likewise.
9105 (vtbl1_p8): Likewise.
9106 (vtbl2_s8): Likewise
9107 (vtbl2_u8): Likewise.
9108 (vtbl2_p8): Likewise.
9109 (vtbl3_s8): Use renamed qtbl2 RTL pattern.
9110 (vtbl3_u8): Likewise.
9111 (vtbl3_p8): Likewise.
9112 (vtbl4_s8): Likewise.
9113 (vtbl4_u8): Likewise.
9114 (vtbl4_p8): Likewise.
9115 (vtbx2_s8): Use renamed qtbx2 RTL pattern.
9116 (vtbx2_u8): Likewise.
9117 (vtbx2_p8): Likewise.
9118 (vqtbl2_s8): Use renamed qtbl2 RTL pattern.
9119 (vqtbl2_u8): Likewise.
9120 (vqtbl2_p8): Likewise.
9121 (vqtbl2q_s8): Likewise.
9122 (vqtbl2q_u8): Likewise.
9123 (vqtbl2q_p8): Likewise.
9124 (vqtbx2_s8): Use renamed qtbx2 RTL pattern.
9125 (vqtbx2_u8): Likewise.
9126 (vqtbx2_p8): Likewise.
9127 (vqtbx2q_s8): Likewise.
9128 (vqtbx2q_u8): Likewise.
9129 (vqtbx2q_p8): Likewise.
9130 (vtbx4_s8): Likewise.
9131 (vtbx4_u8): Likewise.
9132 (vtbx4_p8): Likewise.
9134 2021-07-20 Uroš Bizjak <ubizjak@gmail.com>
9137 * config/i386/sync.md (define_peephole2 atomic_storedi_fpu):
9139 (define_peephole2 atomic_loaddi_fpu): Ditto.
9141 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
9143 * config.gcc (riscv*-*-*): Detect which python is available.
9145 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
9147 * config/rs6000/vsx.md (mulhs_<mode>): Rename to...
9148 (smul<mode>3_highpart): ... this.
9149 (mulhu_<mode>): Rename to...
9150 (umul<mode>3_highpart): ... this.
9151 * config/rs6000/rs6000-builtin.def (MULHS_V2DI, MULHS_V4SI,
9152 MULHU_V2DI, MULHU_V4SI): Adjust.
9154 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
9156 PR tree-optimization/100696
9157 * internal-fn.c (first_commutative_argument): Add info for IFN_MULH.
9158 * internal-fn.def (IFN_MULH): New internal function.
9159 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Add support to
9160 recog normal multiply highpart as IFN_MULH.
9161 * config/i386/i386.c (ix86_add_stmt_cost): Adjust for combined
9164 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
9166 * config/elfos.h (CTF_DEBUGGING_INFO): New definition.
9167 (BTF_DEBUGGING_INFO): Likewise.
9168 * doc/tm.texi.in: Document the new macros.
9169 * doc/tm.texi: Regenerated.
9170 * toplev.c: Guard initialization of debug hooks.
9172 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
9174 * flags.h (ctf_debuginfo_p): New function declaration.
9175 * opts.c (ctf_debuginfo_p): New function definition.
9177 2021-07-19 Andrew Stubbs <ams@codesourcery.com>
9180 * config/gcn/gcn-hsa.h (DRIVER_SELF_SPECS): New.
9181 (ASM_SPEC): Set -mattr for xnack and sram-ecc.
9182 * config/gcn/gcn-opts.h (enum sram_ecc_type): New.
9183 * config/gcn/gcn-valu.md: Add a warning comment.
9184 * config/gcn/gcn.c (gcn_option_override): Add "sorry" for -mxnack.
9185 (output_file_start): Add xnack and sram-ecc state to ".amdgcn_target".
9186 * config/gcn/gcn.md: Add a warning comment.
9187 * config/gcn/gcn.opt: Add -mxnack and -msram-ecc.
9188 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): Remove
9190 (EF_AMDGPU_XNACK): New.
9191 (EF_AMDGPU_SRAM_ECC): New.
9193 (copy_early_debug_info): Use elf_flags.
9194 (main): Handle -mxnack and -msram-ecc options.
9195 * doc/invoke.texi: Document -mxnack and -msram-ecc.
9197 2021-07-19 Andrew Pinski <apinski@marvell.com>
9200 * config/aarch64/aarch64.md (csneg3_uxtw_insn): Rename to ...
9201 (*cs<neg_not_cs>3_uxtw_insn4): and extend to NEG_NOT.
9203 2021-07-19 Richard Biener <rguenther@suse.de>
9205 PR tree-optimization/101505
9206 * tree-vect-patterns.c (vect_determine_precisions): Walk
9207 PHIs also for loop vectorization.
9209 2021-07-19 Richard Biener <rguenther@suse.de>
9211 * gimple.h (gimple_expr_type): Remove.
9212 * doc/gimple.texi: Remove gimple_expr_type documentation.
9214 2021-07-19 Richard Biener <rguenther@suse.de>
9216 * tree-ssa-sccvn.c (vn_reference_eq): Handle NULL vr->type.
9217 (ao_ref_init_from_vn_reference): Likewise.
9218 (fully_constant_reference): Likewise.
9219 (vn_reference_lookup_call): Do not set vr->type to random
9221 * tree-ssa-pre.c (compute_avail): Do not try to PRE calls
9223 * tree-vect-generic.c (expand_vector_piecewise): Pass in
9224 whether we expanded parallel.
9225 (expand_vector_parallel): Adjust.
9226 (expand_vector_addition): Likewise.
9227 (expand_vector_comparison): Likewise.
9228 (expand_vector_operation): Likewise.
9229 (expand_vector_scalar_condition): Likewise.
9230 (expand_vector_conversion): Likewise.
9232 2021-07-19 Richard Biener <rguenther@suse.de>
9234 * tree-vrp.c (register_edge_assert_for_2): Use the
9236 (vrp_folder::fold_predicate_in): Likewise.
9237 * vr-values.c (gimple_assign_nonzero_p): Likewise.
9238 (vr_values::extract_range_from_comparison): Likewise.
9239 (vr_values::extract_range_from_ubsan_builtin): Use the
9240 type of the first operand.
9241 (vr_values::extract_range_basic): Push down type
9242 computation, use the appropriate LHS.
9243 (vr_values::extract_range_from_assignment): Use the
9246 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
9249 * common/config/i386/i386-common.c (ix86_handle_option): For
9250 -mgeneral-regs-only, enable the GPR only instructions which are
9251 enabled implicitly by SSE ISAs unless they have been disabled
9254 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
9257 * config/i386/i386.c (ix86_check_avx_upper_stores): Moved before
9258 ix86_avx_u128_mode_needed.
9259 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY if callee
9260 returns AVX register.
9262 2021-07-17 Jan Hubicka <hubicka@ucw.cz>
9264 * tree-ssa-structalias.c (handle_rhs_call): Support EAF_NOT_RETURNED.
9265 (handle_const_call): Liekise
9266 (handle_pure_call): Liekise
9268 2021-07-17 Andrew MacLeod <amacleod@redhat.com>
9270 PR tree-optimization/96542
9271 * range-op.cc (range_operator::wi_fold_in_parts): New.
9272 (range_operator::fold_range): Call wi_fold_in_parts.
9273 (operator_lshift::wi_fold): Fix broken lshift by [0,0].
9274 * range-op.h (wi_fold_in_parts): Add prototype.
9276 2021-07-16 David Malcolm <dmalcolm@redhat.com>
9278 * doc/analyzer.texi: Add __analyzer_dump_state.
9280 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9282 * config/rs6000/rbtree.c: New file.
9283 * config/rs6000/rbtree.h: New file.
9285 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9287 * config/rs6000/rs6000-gen-builtins.c (restriction): New enum.
9288 (typeinfo): Add restr field.
9289 (match_bracketed_pair): New function.
9290 (match_const_restriction): Implement.
9292 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9294 * config/rs6000/rs6000-gen-builtins.c (match_basetype): Implement.
9296 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9298 * config/rs6000/rs6000-gen-builtins.c (void_status): New enum.
9299 (basetype): Likewise.
9300 (typeinfo): Likewise.
9301 (handle_pointer): New function.
9302 (match_basetype): New stub function.
9303 (match_const_restriction): Likewise.
9304 (match_type): New function.
9306 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9308 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace): New
9310 (advance_line): Likewise.
9311 (safe_inc_pos): Likewise.
9312 (match_identifier): Likewise.
9313 (match_integer): Likewise.
9314 (match_to_right_bracket): Likewise.
9316 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9318 * config/rs6000/rs6000-gen-builtins.c (bif_file): New variable.
9319 (ovld_file): Likewise.
9320 (header_file): Likewise.
9321 (init_file): Likewise.
9322 (defines_file): Likewise.
9323 (pgm_path): Likewise.
9324 (bif_path): Likewise.
9325 (ovld_path): Likewise.
9326 (header_path): Likewise.
9327 (init_path): Likewise.
9328 (defines_path): Likewise.
9329 (LINELEN): New macro.
9330 (linebuf): New variable.
9334 (bif_diag): New function.
9335 (ovld_diag): Likewise.
9337 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9339 * config/rs6000/rs6000-builtin-new.def: New.
9340 * config/rs6000/rs6000-overload.def: New.
9342 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9344 * config/rs6000/rs6000-gen-builtins.c: New.
9346 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9348 * Makefile.in (EXTRA_GTYPE_DEPS): New variable.
9349 (s-gtype): Depend on EXTRA_GTYPE_DEPS.
9350 * gengtype-state.c (state_writer::write_state_file_list): Add a
9351 parameter to the fileslist expression for the number of build
9353 (read_state_files_list): Detect build headers and strip the
9354 initial "./" or ".\" from their names.
9355 * gengtype.c (build_headers): New global variable.
9356 (num_build_headers): Likewise.
9357 (open_base_files): Emit #include for each build header.
9358 (main): Detect and count build headers.
9359 * gengtype.h (build_headers): New extern variable.
9360 (num_build_headers): Likewise.
9362 2021-07-16 Richard Biener <rguenther@suse.de>
9364 * gimple-ssa-store-merging.c (verify_symbolic_number_p): Use
9365 the type of the LHS.
9366 (find_bswap_or_nop_1): Likewise.
9367 (find_bswap_or_nop): Likewise.
9368 * tree-vectorizer.h (vect_get_smallest_scalar_type): Adjust
9370 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
9371 Remove unused parameters, pass in the scalar type. Fix
9372 internal store function handling.
9373 * tree-vect-stmts.c (vect_analyze_stmt): Remove assert.
9374 (vect_get_vector_types_for_stmt): Move down check for
9375 existing vector stmt after we've determined a scalar type.
9376 Pass down the used scalar type to vect_get_smallest_scalar_type.
9377 * tree-vect-generic.c (expand_vector_condition): Use
9378 the type of the LHS.
9379 (expand_vector_scalar_condition): Likewise.
9380 (expand_vector_operations_1): Likewise.
9381 * tree-vect-patterns.c (vect_widened_op_tree): Likewise.
9382 (vect_recog_dot_prod_pattern): Likewise.
9383 (vect_recog_sad_pattern): Likewise.
9384 (vect_recog_widen_op_pattern): Likewise.
9385 (vect_recog_widen_sum_pattern): Likewise.
9386 (vect_recog_mixed_size_cond_pattern): Likewise.
9388 2021-07-16 Jan Hubicka <hubicka@ucw.cz>
9390 * ipa-modref.c (struct escape_entry): Use eaf_fleags_t.
9391 (dump_eaf_flags): Dump EAF_NOT_RETURNED
9392 (eaf_flags_useful_p): Use eaf_fleags_t; handle const functions
9393 and EAF_NOT_RETURNED.
9394 (modref_summary::useful_p): Likewise.
9395 (modref_summary_lto::useful_p): Likewise.
9396 (struct) modref_summary_lto: Use eaf_fleags_t.
9397 (deref_flags): Handle EAF_NOT_RETURNED.
9398 (struct escape_point): Use min_flags.
9399 (modref_lattice::init): Add EAF_NOT_RETURNED.
9400 (merge_call_lhs_flags): Ignore EAF_NOT_RETURNED functions
9401 (analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
9403 (analyze_parms): Also analyze const functions; update conition on
9405 (modref_write): Update streaming.
9406 (read_section): Update streaming.
9407 (remap_arg_flags): Use eaf_flags_t.
9408 (modref_merge_call_site_flags): Hanlde EAF_NOT_RETURNED.
9409 * ipa-modref.h: (eaf_flags_t): New typedef.
9410 (struct modref_summary): Use eaf_flags_t.
9411 * tree-core.h (EAF_NOT_RETURNED): New constant.
9413 2021-07-16 Richard Biener <rguenther@suse.de>
9415 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Use
9416 the type of the LHS.
9417 (gimple_assign_nonnegative_warnv_p): Likewise.
9418 (gimple_call_nonnegative_warnv_p): Likewise. Return false
9419 if the call has no LHS.
9420 * gimple.c (gimple_could_trap_p_1): Use the type of the LHS.
9421 * tree-eh.c (stmt_could_throw_1_p): Likewise.
9422 * tree-inline.c (insert_init_stmt): Likewise.
9423 * tree-ssa-loop-niter.c (get_val_for): Likewise.
9424 * tree-outof-ssa.c (ssa_is_replaceable_p): Use the type of
9426 * tree-ssa-sccvn.c (init_vn_nary_op_from_stmt): Take a
9427 gassign *. Use the type of the lhs.
9428 (vn_nary_op_lookup_stmt): Adjust.
9429 (vn_nary_op_insert_stmt): Likewise.
9431 2021-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
9433 * config/s390/predicates.md (bras_sym_operand): Accept all
9434 functions in 64-bit mode, use UNSPEC_PLT31.
9435 (larl_operand): Use UNSPEC_PLT31.
9436 * config/s390/s390.c (s390_loadrelative_operand_p): Likewise.
9437 (legitimize_pic_address): Likewise.
9438 (s390_emit_tls_call_insn): Mark __tls_get_offset as function,
9440 (s390_delegitimize_address): Use UNSPEC_PLT31.
9441 (s390_output_addr_const_extra): Likewise.
9442 (print_operand): Add @PLT to TLS calls, handle %K.
9443 (s390_function_profiler): Mark __fentry__/_mcount as function,
9444 use %K, use UNSPEC_PLT31.
9445 (s390_output_mi_thunk): Use only UNSPEC_GOT, use %K.
9446 (s390_emit_call): Use UNSPEC_PLT31.
9447 (s390_emit_tpf_eh_return): Mark __tpf_eh_return as function.
9448 * config/s390/s390.md (UNSPEC_PLT31): Rename from UNSPEC_PLT.
9449 (*movdi_64): Use %K.
9450 (reload_base_64): Likewise.
9451 (*sibcall_brc): Likewise.
9452 (*sibcall_brcl): Likewise.
9453 (*sibcall_value_brc): Likewise.
9454 (*sibcall_value_brcl): Likewise.
9457 (*bras_r): Likewise.
9458 (*brasl_r): Likewise.
9459 (*bras_tls): Likewise.
9460 (*brasl_tls): Likewise.
9461 (main_base_64): Likewise.
9462 (reload_base_64): Likewise.
9463 (@split_stack_call<mode>): Likewise.
9465 2021-07-16 Richard Biener <rguenther@suse.de>
9467 PR tree-optimization/101467
9468 * tree-vect-stmts.c (vect_gen_while): Properly guard
9469 make_temp_ssa_name usage.
9471 2021-07-16 Cooper Qu <cooper.qu@linux.alibaba.com>
9473 * config.gcc: Don't use forked print-sysroot-suffix.sh and
9474 t-sysroot-suffix for C-SKY.
9475 * config/csky/print-sysroot-suffix.sh: Delete.
9476 * config/csky/t-csky-linux: Delete.
9477 * config/csky/t-sysroot-suffix: Define MULTILIB_DIRNAMES
9478 instead of CSKY_MULTILIB_DIRNAMES.
9480 2021-07-16 Richard Biener <rguenther@suse.de>
9482 * tree-vect-loop.c (vect_transform_cycle_phi): Correct sign
9483 conversion issues with the partial reduction of the reused
9486 2021-07-16 Richard Biener <rguenther@suse.de>
9488 * config/i386/i386-options.c (ix86_option_override_internal): Set
9489 param_vect_partial_vector_usage to zero if not set.
9491 2021-07-15 Uroš Bizjak <ubizjak@gmail.com>
9494 * config/i386/i386.h (VALID_SSE_REG_MODE): Add TDmode.
9495 (VALID_INT_MODE_P): Add SDmode and DDmode.
9496 Add TDmode for TARGET_64BIT.
9497 (VALID_DFP_MODE_P): Remove.
9498 * config/i386/i386.c (ix86_hard_regno_mode_ok):
9499 Do not use VALID_DFP_MODE_P.
9501 2021-07-15 Andrew MacLeod <amacleod@redhat.com>
9503 * gimple-range-fold.cc (adjust_pointer_diff_expr): Use
9505 (fold_using_range::fold_stmt): Ditto.
9506 (fold_using_range::range_of_range_op): Ditto.
9507 (fold_using_range::range_of_phi): Ditto.
9508 (fold_using_range::range_of_call): Ditto.
9509 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
9510 (fold_using_range::range_of_builtin_call): Ditto.
9511 (fold_using_range::range_of_cond_expr): Ditto.
9512 * gimple-range-fold.h (gimple_range_type): New.
9514 2021-07-15 Martin Sebor <msebor@redhat.com>
9517 * tree-ssa-strlen.c (handle_assign): New function.
9518 (maybe_warn_overflow): Add argument.
9519 (nonzero_bytes_for_type): New function.
9520 (count_nonzero_bytes): Handle more tree types. Call
9521 nonzero_bytes_for_tye.
9522 (count_nonzero_bytes): Handle types.
9523 (handle_store): Handle stores from function calls.
9524 (strlen_check_and_optimize_call): Move code to handle_assign. Call
9525 it for assignments from function calls.
9527 2021-07-15 David Malcolm <dmalcolm@redhat.com>
9532 * doc/invoke.texi: Add -Wanalyzer-use-of-uninitialized-value.
9534 2021-07-15 David Malcolm <dmalcolm@redhat.com>
9536 * doc/invoke.texi (-fdump-analyzer-exploded-paths): New.
9538 2021-07-15 Martin Sebor <msebor@redhat.com>
9542 * fold-const.c (operand_compare::operand_equal_p): Handle OEP_DECL_NAME.
9543 (operand_compare::verify_hash_value): Same.
9544 * tree-core.h (OEP_DECL_NAME): New.
9546 2021-07-15 Martin Jambor <mjambor@suse.cz>
9548 * profile-count.h (profile_count::value): Change the return type to
9550 * gimple-pretty-print.c (dump_gimple_bb_header): Adjust print
9552 * tree-cfg.c (dump_function_to_file): Likewise.
9554 2021-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
9557 * config/rs6000/rs6000-p8swap.c (has_part_mult): New.
9558 (rs6000_analyze_swaps): Insns containing a subreg of a mult are
9561 2021-07-15 Richard Biener <rguenther@suse.de>
9563 * tree-vectorizer.h (vect_gen_while): Match up with
9565 * tree-vect-stmts.c (vect_gen_while): Adjust API to that
9566 of vect_gen_while_not.
9567 (vect_gen_while_not): Adjust.
9568 * tree-vect-loop-manip.c (vect_set_loop_controls_directly): Likewise.
9570 2021-07-15 Aldy Hernandez <aldyh@redhat.com>
9572 * gimple-range-cache.cc (non_null_ref::adjust_range): New.
9573 (ranger_cache::range_of_def): Call adjust_range.
9574 (ranger_cache::entry_range): Same.
9575 * gimple-range-cache.h (non_null_ref::adjust_range): New.
9576 * gimple-range.cc (gimple_ranger::range_of_expr): Call
9578 (gimple_ranger::range_on_entry): Same.
9580 2021-07-15 Tamar Christina <tamar.christina@arm.com>
9583 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9585 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
9587 2021-07-15 Tamar Christina <tamar.christina@arm.com>
9590 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9592 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
9593 (sdot_prod, udot_prod): ...These.
9594 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
9595 (aarch64_<sur>dot<vsi2qi>): Rename to...
9596 (<sur>dot_prod<vsi2qi>): ...This.
9597 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
9600 2021-07-15 Jakub Jelinek <jakub@redhat.com>
9602 PR middle-end/101437
9603 * gimplify.c (gimplify_expr): Throw away volatile reads from empty
9604 types even if they have non-BLKmode TYPE_MODE.
9606 2021-07-15 Richard Biener <rguenther@suse.de>
9609 * gcc.c (process_command): Process -gtoggle like process_options
9610 would after parsing options.
9612 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
9614 * cfgexpand.c (expand_asm_loc): Adjust.
9615 (expand_asm_stmt): Likewise.
9616 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Likewise.
9617 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
9618 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
9619 * config/avr/avr.c (avr_md_asm_adjust): Likewise.
9620 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
9621 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
9622 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
9623 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
9624 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
9625 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
9626 * config/s390/s390.c (s390_md_asm_adjust): Likewise.
9627 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
9628 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
9629 * doc/tm.texi: Regenerate.
9630 * target.def: Add location argument to md_asm_adjust.
9632 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
9634 * tree-diagnostic.c (diagnostic_report_current_function): Use the
9635 diagnostic's location, not input_location.
9637 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
9639 * cfgexpand.c (tree_conflicts_with_clobbers_p): Pass location to
9641 (expand_asm_stmt): Likewise.
9643 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
9645 * config/rs6000/rs6000.c (adjacent_mem_locations): Return the lower
9646 addressed memory rtx, if any.
9647 (rs6000_split_multireg_move): Fix code formatting.
9648 Handle MMA build built-ins with operands in adjacent memory locations.
9650 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
9652 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Move to later
9655 2021-07-14 Jason Merrill <jason@redhat.com>
9657 * sel-sched-ir.h (get_all_loop_exits): Use auto_vec.
9659 2021-07-14 Jason Merrill <jason@redhat.com>
9661 * doc/invoke.texi: -fdelete-dead-exceptions is on by default for
9664 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9666 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
9667 Remove erroneous line.
9669 2021-07-14 Andrew MacLeod <amacleod@redhat.com>
9671 * params.opt (param_evrp_mode): Change default.
9673 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9675 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
9676 (sdot_prod, udot_prod): ...These.
9677 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
9678 (aarch64_<sur>dot<vsi2qi>): Rename to...
9679 (<sur>dot_prod<vsi2qi>): ...This.
9680 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
9683 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9685 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
9687 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9689 * doc/sourcebuild.texi (arm_v8_2a_i8mm_neon_hw): Document.
9691 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9693 * config/arm/neon.md (usdot_prod<vsi2qi>): New.
9695 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9697 * config/aarch64/aarch64-simd.md (aarch64_usdot<vsi2qi>): Rename to...
9698 (usdot_prod<vsi2qi>): ... This.
9699 * config/aarch64/aarch64-simd-builtins.def (usdot): Rename to...
9700 (usdot_prod): ...This.
9701 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Likewise.
9702 * config/aarch64/aarch64-sve.md (@aarch64_<sur>dot_prod<vsi2qi>):
9704 (@<sur>dot_prod<vsi2qi>): ...This.
9705 * config/aarch64/aarch64-sve-builtins-base.cc
9706 (svusdot_impl::expand): Use it.
9708 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9710 * optabs.def (usdot_prod_optab): New.
9711 * doc/md.texi: Document it and clarify other dot prod optabs.
9712 * optabs-tree.h (enum optab_subtype): Add optab_vector_mixed_sign.
9713 * optabs-tree.c (optab_for_tree_code): Support usdot_prod_optab.
9714 * optabs.c (expand_widen_pattern_expr): Likewise.
9715 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
9716 * tree-vect-loop.c (vectorizable_reduction): Query dot-product kind.
9717 * tree-vect-patterns.c (vect_supportable_direct_optab_p): Take optional
9719 (vect_widened_op_tree): Optionally ignore
9721 (vect_recog_dot_prod_pattern): Support usdot_prod_optab.
9723 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
9726 * config/i386/driver-i386.c (host_detect_local_cpu): Check
9727 "arch [32|64]" and "tune [32|64]" for 32-bit and 64-bit codegen.
9728 Enable UINTR only for 64-bit codegen.
9729 * config/i386/i386-options.c
9730 (ix86_option_override_internal::DEF_PTA): Skip PTA_UINTR if not
9732 * config/i386/i386.h (ARCH_ARG): New.
9733 (CC1_CPU_SPEC): Pass "[arch|tune] 32" for 32-bit codegen and
9734 "[arch|tune] 64" for 64-bit codegen.
9736 2021-07-14 Richard Biener <rguenther@suse.de>
9738 PR tree-optimization/101445
9739 * tree-vect-stmts.c (vectorizable_load): Do the gap adjustment
9740 of the IV in the correct direction for negative stride
9743 2021-07-14 Jakub Jelinek <jakub@redhat.com>
9746 * godump.c (godump_str_hash): New type.
9747 (godump_container::pot_dummy_types): Use string_hash instead of
9748 ptr_hash in the hash_set.
9750 2021-07-14 Richard Biener <rguenther@suse.de>
9752 * tree-vect-loop.c (vect_find_reusable_accumulator): Handle
9753 vector types where the old vector type has a multiple of
9754 the new vector type elements.
9755 (vect_create_partial_epilog): New function, split out from...
9756 (vect_create_epilog_for_reduction): ... here.
9757 (vect_transform_cycle_phi): Reduce the re-used accumulator
9758 to the new vector type.
9760 2021-07-14 Alexandre Oliva <oliva@adacore.com>
9762 * tree-ssa-alias.c (attr_fnspec::verify): Fix index in
9763 non-'t'-sized arg check.
9765 2021-07-14 Alexandre Oliva <oliva@adacore.com>
9767 * tree-cfg.c (cleanup_dead_labels_eh): Update
9768 post_landing_pad label upon change of landing pad block's
9770 (cleanup_dead_labels): Check that a removed label is not that
9773 2021-07-13 Jonathan Wright <jonathan.wright@arm.com>
9775 * combine.c (combine_simplify_rtx): Add vec_select -> subreg
9777 * config/aarch64/aarch64.md (*zero_extend<SHORT:mode><GPI:mode>2_aarch64):
9778 Add Neon to general purpose register case for zero-extend
9780 * config/arm/vfp.md (*arm_movsi_vfp): Remove "*" from *t -> r
9781 case to prevent some cases opting to go through memory.
9782 * cse.c (fold_rtx): Add vec_select -> subreg simplification.
9783 * rtl.c (rtvec_series_p): Define predicate to determine
9784 whether a vector contains a linear series of integers.
9785 * rtl.h (rtvec_series_p): Define.
9786 * rtlanal.c (vec_series_lowpart_p): Define predicate to
9787 determine if a vector selection is equivalent to the low part
9789 * rtlanal.h (vec_series_lowpart_p): Define.
9790 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
9791 Add vec_select -> subreg simplification.
9793 2021-07-13 Paul A. Clarke <pc@us.ibm.com>
9795 * config/rs6000/smmintrin.h (_mm_testz_si128, _mm_testc_si128,
9796 _mm_testnzc_si128, _mm_test_all_ones, _mm_test_all_zeros,
9797 _mm_test_mix_ones_zeros): New.
9799 2021-07-13 Roger Sayle <roger@nextmovesoftware.com>
9800 Richard Biener <rguenther@suse.de>
9802 * gimple.c (gimple_could_trap_p_1): Make S argument a
9803 "const gimple*". Preserve constness in call to
9804 gimple_asm_volatile_p.
9805 (gimple_could_trap_p): Make S argument a "const gimple*".
9806 * gimple.h (gimple_could_trap_p_1, gimple_could_trap_p):
9807 Update function prototypes.
9809 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9811 * tree-vectorizer.h (vect_reusable_accumulator): New structure.
9812 (_loop_vec_info::main_loop_edge): New field.
9813 (_loop_vec_info::skip_main_loop_edge): Likewise.
9814 (_loop_vec_info::skip_this_loop_edge): Likewise.
9815 (_loop_vec_info::reusable_accumulators): Likewise.
9816 (_stmt_vec_info::reduc_scalar_results): Likewise.
9817 (_stmt_vec_info::reused_accumulator): Likewise.
9818 (vect_get_main_loop_result): Declare.
9819 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize
9820 reduc_scalar_inputs.
9821 (vec_info::free_stmt_vec_info): Free reduc_scalar_inputs.
9822 * tree-vect-loop-manip.c (vect_get_main_loop_result): New function.
9823 (vect_do_peeling): Fill an epilogue loop's main_loop_edge,
9824 skip_main_loop_edge and skip_this_loop_edge fields.
9825 * tree-vect-loop.c (INCLUDE_ALGORITHM): Define.
9826 (vect_emit_reduction_init_stmts): New function.
9827 (get_initial_def_for_reduction): Use it.
9828 (get_initial_defs_for_reduction): Likewise. Change the vinfo
9829 parameter to a loop_vec_info.
9830 (vect_create_epilog_for_reduction): Store the scalar results
9831 in the reduc_info. If an epilogue loop is reusing an accumulator
9832 from the main loop, and if the epilogue loop can also be skipped,
9833 try to place the reduction code in the join block. Record
9834 accumulators that could potentially be reused by epilogue loops.
9835 (vect_transform_cycle_phi): When vectorizing epilogue loops,
9836 try to reuse accumulators from the main loop. Record the initial
9837 value in reduc_info for non-SLP reductions too.
9839 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9841 * tree-vect-loop.c (get_initial_def_for_reduction): Remove
9842 adjustment handling. Take the neutral value as an argument,
9843 in place of the code argument.
9844 (vect_transform_cycle_phi): Update accordingly. Handle the
9845 initial values of cond reductions separately from code reductions.
9846 Choose the adjustment here rather than in
9847 get_initial_def_for_reduction. Sink the splat of vec_initial_def.
9849 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9851 * tree-vect-loop.c (neutral_op_for_slp_reduction): Replace with...
9852 (neutral_op_for_reduction): ...this, providing a more general
9854 (vect_create_epilog_for_reduction): Update accordingly.
9855 (vectorizable_reduction): Likewise.
9856 (vect_transform_cycle_phi): Likewise.
9858 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9860 * tree-vect-loop.c (get_initial_def_for_reduction): Take the
9861 reduc_info instead of the original stmt_vec_info.
9862 (vect_transform_cycle_phi): Update accordingly.
9864 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9866 * tree-vect-loop.c (get_initial_defs_for_reduction): Take the
9867 reduc_info as an additional parameter.
9868 (vect_transform_cycle_phi): Update accordingly.
9870 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9872 * tree-vectorizer.h: Include tree-ssa-operands.h.
9873 (vect_phi_initial_value): New function.
9874 * tree-vect-loop.c (neutral_op_for_slp_reduction): Use it.
9875 (get_initial_defs_for_reduction, info_for_reduction): Likewise.
9876 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
9877 (vect_transform_cycle_phi, vectorizable_induction): Likewise.
9879 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9881 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
9882 the phi results to vectype after creating them. Remove later
9883 conversion code that thus becomes redundant.
9885 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9887 * tree-vect-loop.c (vect_create_epilog_for_reduction): Replace
9888 the new_phis vector with a reduc_inputs vector. Combine handling
9889 of reduction chains and ncopies > 1.
9891 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9893 * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
9894 scalar_results to group_size elements after reducing down from
9895 N*group_size elements. Construct an array_slice of the live-out
9896 stmts and assert that there is one stmt per scalar result.
9898 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9900 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
9901 nested_in_vect_loop and use double_reduc everywhere. Remove dead
9902 assignment to "loop".
9904 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9906 * internal-fn.c (vectorized_internal_fn_supported_p): Handle
9907 vector types first. For scalar types, consider both the preferred
9908 vector mode and the alternative vector modes.
9909 * optabs-query.c (can_vec_mask_load_store_p): Use the same
9910 structure as above, in particular using related_vector_mode
9911 for modes provided by autovectorize_vector_modes.
9913 2021-07-13 Jakub Jelinek <jakub@redhat.com>
9914 Richard Biener <rguenther@suse.de>
9916 PR tree-optimization/101419
9917 * tree-pass.h (PROP_objsz): Define.
9918 (make_pass_early_object_sizes): Declare.
9919 * passes.def (pass_all_early_optimizations): Rename pass_object_sizes
9920 there to pass_early_object_sizes, drop parameter.
9921 (pass_all_optimizations): Move pass_object_sizes right after pass_ccp,
9922 drop parameter, move pass_post_ipa_warn right after that.
9923 * tree-object-size.c (pass_object_sizes::execute): Rename to...
9924 (object_sizes_execute): ... this. Add insert_min_max_p argument.
9925 (pass_data_object_sizes): Move after object_sizes_execute.
9926 (pass_object_sizes): Likewise. In execute method call
9927 object_sizes_execute, drop set_pass_param method and insert_min_max_p
9928 non-static data member and its initializer in the ctor.
9929 (pass_data_early_object_sizes, pass_early_object_sizes,
9930 make_pass_early_object_sizes): New.
9931 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Use
9932 (cfun->curr_properties & PROP_objsz) instead of cfun->after_inlining.
9934 2021-07-13 Kito Cheng <kito.cheng@sifive.com>
9937 * config/riscv/constraints.md ("S"): Update description and remove
9939 * doc/md.texi (Machine Constraints): Document the 'S' constraints
9942 2021-07-13 Richard Biener <rguenther@suse.de>
9945 2021-07-12 Richard Biener <rguenther@suse.de>
9947 * tree-vect-slp.c (vect_slp_region): Show the number of
9948 SLP graph entries in the optimization message.
9950 2021-07-13 Michael Meissner <meissner@linux.ibm.com>
9952 * config/rs6000/altivec.md (xxspltiw_v4sf): Change local variable
9954 * config/rs6000/rs6000-protos.h (rs6000_const_f32_to_i32): Change
9955 return type to long.
9956 * config/rs6000/rs6000.c (rs6000_const_f32_to_i32): Change return
9959 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
9961 * gimple-range-fold.cc (fold_using_range::range_of_builtin_ubsan_call):
9962 Query relation between the 2 operands and use it.
9964 2021-07-12 Sergei Trofimovich <siarheit@google.com>
9966 * doc/cfg.texi: Fix s/ei_safe_safe/ei_safe_edge/ typo.
9968 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
9971 * config/i386/predicates.md (vec_setm_sse41_operand):
9972 Rename from vec_setm_operand.
9973 (vec_setm_avx2_operand): New predicate.
9974 * config/i386/sse.md (vec_set<V_128:mode>): Use V_128 mode iterator.
9975 Use vec_setm_sse41_operand as operand 2 predicate.
9976 (vec_set<V_256_512:mode): New expander.
9977 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_sse41_operand
9978 as operand 2 predicate.
9980 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
9982 PR tree-optimization/101335
9983 * range-op.cc (operator_cast::lhs_op1_relation): Delete.
9985 2021-07-12 Andrew Pinski <apinski@marvell.com>
9987 * tree-ssa-phiopt.c (match_simplify_replacement): Move
9988 insert of the sequence before the movement of the
9989 statement. Check if to see if the statement is used
9990 outside of the original phi to see if we should move it.
9992 2021-07-12 Richard Biener <rguenther@suse.de>
9994 * dump-context.h (debug_dump_context::debug_dump_context):
9995 Add FILE * parameter defaulted to stderr.
9996 * dumpfile.c (debug_dump_context::debug_dump_context): Adjust.
9997 * tree-vect-slp.c (dot_slp_tree): New functions.
9999 2021-07-12 Richard Biener <rguenther@suse.de>
10001 PR tree-optimization/101373
10002 * tree-ssa-pre.c (prune_clobbered_mems): Also prune trapping
10003 references when the BB may not return.
10004 (compute_avail): Pass in the function we're working on and
10005 replace cfun references with it. Externally throwing
10006 const calls also possibly terminate the function.
10007 (pass_pre::execute): Pass down the function we're working on.
10008 * gcse.c (compute_hash_table_work): Externally throwing
10009 const/pure calls also need record_last_mem_set_info.
10010 * postreload-gcse.c (record_opr_changes): Looping or externally
10011 throwing const/pure calls also need record_last_mem_set_info.
10013 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
10015 * recog.c (memory_address_addr_space_p): Change the type to bool.
10016 Return true/false instead of 1/0.
10017 (offsettable_memref_p): Ditto.
10018 (offsettable_nonstrict_memref_p): Ditto.
10019 (offsettable_address_addr_space_p): Ditto.
10020 Change the type of addressp indirect function to bool.
10021 * recog.h (memory_address_addr_space_p): Change the type to bool.
10022 (strict_memory_address_addr_space_p): Ditto.
10023 (offsettable_memref_p): Ditto.
10024 (offsettable_nonstrict_memref_p): Ditto.
10025 (offsettable_address_addr_space_p): Ditto.
10026 * reload.c (maybe_memory_address_addr_space_p): Ditto.
10027 (strict_memory_address_addr_space_p): Change the type to bool.
10028 Return true/false instead of 1/0.
10029 (maybe_memory_address_addr_space_p): Change the type to bool.
10031 2021-07-12 Richard Biener <rguenther@suse.de>
10033 * tree-vect-slp.c (vect_slp_region): Show the number of
10034 SLP graph entries in the optimization message.
10036 2021-07-12 Richard Biener <rguenther@suse.de>
10038 PR tree-optimization/101394
10039 * tree-ssa-pre.c (do_pre_regular_insertion): Avoid inserting
10040 copies from abnormals for a full redundancy.
10042 2021-07-12 Richard Biener <rguenther@suse.de>
10044 PR middle-end/101423
10045 * gimple.c (gimple_could_trap_p_1): Internal function calls
10047 * tree-eh.c (tree_could_trap_p): Likewise.
10049 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
10052 * config/arm/arm_neon.h (vmul_n_u32): Replace call to builtin with
10054 (vmulq_n_u32): Likewise.
10055 (vmul_n_f32): Gate __a * __b on __FAST_MATH__.
10056 (vmulq_n_f32): Likewise.
10057 (vmul_n_f16): Likewise.
10058 (vmulq_n_f16): Likewise.
10060 2021-07-12 Martin Liska <mliska@suse.cz>
10062 PR sanitizer/101425
10063 * gcc.c (check_offload_target_name): Call
10064 candidates_list_and_hint only if we have a candidate.
10066 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
10069 * config/arm/neon.md (vec_init): Move to ...
10070 * config/arm/vec-common.md (vec_init): ... here.
10071 Change the pattern's mode to VDQX and gate it on VALID_MVE_MODE.
10073 2021-07-12 Roger Sayle <roger@nextmovesoftware.com>
10075 PR tree-optimization/101403
10076 * match.pd ((T)bswap(X)>>C): Correctly handle cases where
10077 signedness of the shift is not the same as the signedness of
10078 the type extension.
10080 2021-07-09 Roger Sayle <roger@nextmovesoftware.com>
10081 Uroš Bizjak <ubizjak@gmail.com>
10083 * config/i386/i386.md (*divmodsi4_const): Optimize SImode
10084 divmod of a constant numerator with new define_insn_and_split.
10086 2021-07-09 Iain Sandoe <iain@sandoe.co.uk>
10089 * config/i386/i386-expand.c (ix86_expand_call): If a call is
10090 to a non-local-binding, or local but to a public symbol, then
10091 assume that it might be indirected via the lazy symbol binder.
10092 Mark R10 and R10 as clobbered in that case.
10094 2021-07-09 Eric Botcazou <ebotcazou@adacore.com>
10097 * gcc.c (ASM_DEBUG_DWARF_OPTION): Set again to --gdwarf2 in
10098 the case where HAVE_AS_WORKING_DWARF_N_FLAG is not defined
10099 and HAVE_LD_BROKEN_PE_DWARF5 is defined.
10101 2021-07-09 Uroš Bizjak <ubizjak@gmail.com>
10103 * config/i386/i386.md (*udivmodsi4_pow2_zext_1): Limit the
10104 log2 range of operands[3] to [1,31].
10105 (*udivmodsi4_pow2_zext_2): Ditto. Correct insn RTX pattern.
10107 2021-07-09 Sergei Trofimovich <siarheit@google.com>
10109 * doc/md.texi: Don't split @smallexample in multiple @groups.
10111 2021-07-09 Sergei Trofimovich <siarheit@google.com>
10113 * doc/md.texi: Add missing 'see' word.
10115 2021-07-09 Andrew Pinski <apinski@marvell.com>
10117 * tree-ssa-phiopt.c (phiopt_early_allow): Change arguments
10118 to take sequence and gimple_match_op. Accept the case where
10119 op is a SSA_NAME and one statement in the sequence.
10120 Also allow constants.
10121 (gimple_simplify_phiopt): Always pass a sequence to resimplify.
10122 Update call to phiopt_early_allow. Discard the sequence if not
10125 2021-07-09 Xi Ruoyao <xry111@mengyan1223.wang>
10130 * config/mips/mips.c (mips_const_insns): Use MSA_SUPPORTED_MODE_P
10131 instead of ISA_HAS_MSA.
10132 (mips_expand_vec_unpack): Likewise.
10133 (mips_expand_vector_init): Likewise.
10135 2021-07-09 Kewen Lin <linkw@linux.ibm.com>
10137 * config/rs6000/vsx.md (mods_<mode>): Rename to...
10138 (mod<mode>3): ... this.
10139 (modu_<mode>): Rename to...
10140 (umod<mode>3): ... this.
10141 * config/rs6000/rs6000-builtin.def (MODS_V2DI, MODS_V4SI, MODU_V2DI,
10142 MODU_V4SI): Adjust.
10144 2021-07-08 Jeff Law <jeffreyalaw@gmail.com>
10146 * config/h8300/shiftrotate.md (variable shifts): Expose condition
10147 code handling for the test before the loop.
10149 2021-07-08 Martin Jambor <mjambor@suse.cz>
10152 * ipa-sra.c (class isra_call_summary): New member
10153 m_before_any_store, initialize it in the constructor.
10154 (isra_call_summary::dump): Dump the new field.
10155 (ipa_sra_call_summaries::duplicate): Copy it.
10156 (process_scan_results): Set it.
10157 (isra_write_edge_summary): Stream it.
10158 (isra_read_edge_summary): Likewise.
10159 (param_splitting_across_edge): Only override
10160 safe_to_import_accesses if m_before_any_store is set.
10162 2021-07-08 Martin Sebor <msebor@redhat.com>
10164 PR bootstrap/101374
10165 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref):
10166 Use Object Size Type 0 instead of 1.
10168 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
10170 * tree-vect-loop.c (vectorizable_reduction): Remove always-true
10173 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
10175 * match.pd: Simplify an extend-operate-truncate sequence involving
10178 2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
10179 Richard Biener <rguenther@suse.de>
10181 PR tree-optimization/40210
10182 * match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
10183 (x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
10184 when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
10186 2021-07-08 Uroš Bizjak <ubizjak@gmail.com>
10189 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
10191 * config/i386/mmx.md (V_32): New mode iterator.
10192 (mov<V_32:mode>): Use V_32 mode iterator.
10193 (*mov<V_32:mode>_internal): Ditto.
10194 (*push<V_32:mode>2_rex64): Ditto.
10195 (*push<V_32:mode>2): Ditto.
10196 (movmisalign<V_32:mode>): Ditto.
10197 (mmx_<any_shiftrt:insn>v1si3): New insn pattern.
10198 (sse4_1_<any_extend:code>v2qiv2hi2): Ditto.
10199 (vec_unpacks_lo_v4qi): New expander.
10200 (vec_unpacks_hi_v4qi): Ditto.
10201 (vec_unpacku_lo_v4qi): Ditto.
10202 (vec_unpacku_hi_v4qi): Ditto.
10203 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add V1SImode.
10204 (VALID_INT_MODE_P): Ditto.
10206 2021-07-08 Michael Meissner <meissner@linux.ibm.com>
10209 * config/rs6000/rs6000.md (udivti3): New insn.
10210 (divti3): New insn.
10211 (umodti3): New insn.
10212 (modti3): New insn.
10214 2021-07-07 Martin Sebor <msebor@redhat.com>
10216 PR tree-optimization/100137
10217 PR tree-optimization/99121
10218 PR tree-optimization/97027
10219 * builtins.c (access_ref::access_ref): Also set offmax.
10220 (access_ref::offset_in_range): Define new function.
10221 (access_ref::add_offset): Set offmax.
10222 (access_ref::inform_access): Handle access_none.
10223 (handle_mem_ref): Clear ostype.
10224 (compute_objsize_r): Handle ASSERT_EXPR.
10225 * builtins.h (struct access_ref): Add offmax member.
10226 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Use
10227 compute_objsize() and simplify.
10229 2021-07-07 Peter Bergner <bergner@linux.ibm.com>
10231 * config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
10232 and VSX_BUILTIN_STXVP.
10234 2021-07-07 Martin Sebor <msebor@redhat.com>
10237 * config/aarch64/aarch64.c (aarch64_simd_lane_bounds): Remove
10238 a stray %K from error_at() missed in r12-2088.
10240 2021-07-07 Richard Biener <rguenther@suse.de>
10242 PR tree-optimization/99728
10243 * tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
10245 (mem_refs_may_alias_p): Add assert we handled aggregate
10247 (sm_seq_valid_bb): Give up when running into aggregate copies.
10248 (ref_indep_loop_p): Handle aggregate copies as never
10249 being invariant themselves but allow other refs to be
10250 disambiguated against them.
10251 (can_sm_ref_p): Do not try to apply store-motion to aggregate
10254 2021-07-06 Indu Bhagat <indu.bhagat@oracle.com>
10257 * dwarf2ctf.c (ctf_get_AT_data_member_location): Multiply by 8 to get
10260 2021-07-06 Martin Sebor <msebor@redhat.com>
10262 * gimple-pretty-print.c (percent_G_format): Remove.
10263 * tree-diagnostic.c (default_tree_printer): Remove calls.
10264 * tree-pretty-print.c (percent_K_format): Remove.
10265 * tree-pretty-print.h (percent_K_format): Remove.
10267 2021-07-06 Martin Sebor <msebor@redhat.com>
10269 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_builtin):
10270 Remove %K and use error_at.
10271 (aarch64_expand_fcmla_builtin): Same.
10272 (aarch64_expand_builtin_tme): Same.
10273 (aarch64_expand_builtin_memtag): Same.
10274 * config/arm/arm-builtins.c (arm_expand_acle_builtin): Same.
10275 (arm_expand_builtin): Same.
10276 * config/arm/arm.c (bounds_check): Same.
10278 2021-07-06 Martin Sebor <msebor@redhat.com>
10280 * builtins.c (warn_string_no_nul): Remove %G.
10281 (maybe_warn_for_bound): Same.
10282 (warn_for_access): Same.
10283 (check_access): Same.
10284 (check_strncat_sizes): Same.
10285 (expand_builtin_strncat): Same.
10286 (expand_builtin_strncmp): Same.
10287 (expand_builtin): Same.
10288 (expand_builtin_object_size): Same.
10289 (warn_dealloc_offset): Same.
10290 (maybe_emit_free_warning): Same.
10291 * calls.c (maybe_warn_alloc_args_overflow): Same.
10292 (maybe_warn_nonstring_arg): Same.
10293 (maybe_warn_rdwr_sizes): Same.
10294 * expr.c (expand_expr_real_1): Remove %K.
10295 * gimple-fold.c (gimple_fold_builtin_strncpy): Remove %G.
10296 (gimple_fold_builtin_strncat): Same.
10297 * gimple-ssa-sprintf.c (format_directive): Same.
10298 (handle_printf_call): Same.
10299 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Same.
10300 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
10301 (maybe_diag_access_bounds): Same. Call gimple_location.
10302 (check_bounds_or_overlap): Same.
10303 * trans-mem.c (ipa_tm_scan_irr_block): Remove %K. Simplify.
10304 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Remove %G.
10305 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
10306 (maybe_diag_stxncpy_trunc): Same.
10307 (handle_builtin_stxncpy_strncat): Same.
10308 (maybe_warn_pointless_strcmp): Same.
10309 * tree-ssa-uninit.c (maybe_warn_operand): Same.
10311 2021-07-06 Uroš Bizjak <ubizjak@gmail.com>
10314 * config/i386/predicates.md (vec_setm_operand): Enable
10315 register_operand for TARGET_SSE4_1.
10316 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_operand
10317 as operand 2 predicate. Call ix86_expand_vector_set_var
10318 for non-constant index operand.
10319 (vec_setv4qi): Use vec_setm_mmx_operand as operand 2 predicate.
10320 Call ix86_expand_vector_set_var for non-constant index operand.
10322 2021-07-06 Jeff Law <jeffreyalaw@gmail.com>
10324 * config/h8300/jumpcall.md (*branch): When possible, generate
10325 the comparison in CCZN mode.
10326 * config/h8300/predicates.md (simple_memory_operand): Reject all
10327 auto-increment addressing modes.
10329 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
10331 PR bootstrap/100246
10332 * config/i386/i386.h (struct stringop_algs): Define a CTOR for
10335 2021-07-06 Richard Biener <rguenther@suse.de>
10337 * doc/md.texi (vec_fmaddsub<mode>4): Document.
10338 (vec_fmsubadd<mode>4): Likewise.
10339 * optabs.def (vec_fmaddsub$a4): Add.
10340 (vec_fmsubadd$a4): Likewise.
10341 * internal-fn.def (IFN_VEC_FMADDSUB): Add.
10342 (IFN_VEC_FMSUBADD): Likewise.
10343 * tree-vect-slp-patterns.c (addsub_pattern::recognize):
10344 Refactor to handle IFN_VEC_FMADDSUB and IFN_VEC_FMSUBADD.
10345 (addsub_pattern::build): Likewise.
10346 * tree-vect-slp.c (vect_optimize_slp): CFN_VEC_FMADDSUB
10347 and CFN_VEC_FMSUBADD are not transparent for permutes.
10348 * config/i386/sse.md (vec_fmaddsub<mode>4): New expander.
10349 (vec_fmsubadd<mode>4): Likewise.
10351 2021-07-06 Richard Biener <rguenther@suse.de>
10353 * doc/invoke.texi (fmove-loop-stores): Document.
10354 * common.opt (fmove-loop-stores): New option.
10355 * opts.c (default_options_table): Enable -fmove-loop-stores
10356 at -O1 but not -Og.
10357 * tree-ssa-loop-im.c (pass_lim::execute): Pass
10358 flag_move_loop_stores instead of true to
10359 loop_invariant_motion_in_fun.
10361 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
10363 * doc/install.texi: Document --with-dsymutil.
10365 2021-07-06 Andrew Pinski <apinski@marvell.com>
10367 PR tree-optimization/101256
10368 * dbgcnt.def (phiopt_edge_range): New counter.
10369 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
10370 Check to make sure the new name is defined in the same
10371 bb as the conditional before duplicating range info.
10372 Also add debug counter.
10374 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
10376 PR rtl-optimization/100328
10377 * config/i386/i386-options.c (ix86_option_override_internal):
10378 Set param_ira_consider_dup_in_all_alts to 0.
10380 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
10382 PR rtl-optimization/100328
10383 * doc/invoke.texi (ira-consider-dup-in-all-alts): Document new
10385 * ira.c (ira_get_dup_out_num): Adjust as parameter
10386 param_ira_consider_dup_in_all_alts.
10387 * params.opt (ira-consider-dup-in-all-alts): New.
10388 * ira-conflicts.c (process_regs_for_copy): Add one parameter
10389 single_input_op_has_cstr_p.
10390 (get_freq_for_shuffle_copy): New function.
10391 (add_insn_allocno_copies): Adjust as single_input_op_has_cstr_p.
10392 * ira-int.h (ira_get_dup_out_num): Add one bool parameter.
10394 2021-07-05 Jeff Law <jeffreyalaw@gmail.com>
10396 * config/h8300/shiftrotate.md (shift-by-variable patterns): Update to
10397 generate condition code aware RTL directly.
10399 2021-07-05 Andrew Pinski <apinski@marvell.com>
10401 PR tree-optimization/101039
10402 * match.pd (A CMP 0 ? A : -A): New patterns.
10403 * tree-ssa-phiopt.c (abs_replacement): Delete function.
10404 (tree_ssa_phiopt_worker): Don't call abs_replacement.
10405 Update comment about abs_replacement.
10407 2021-07-05 Andrew Pinski <apinski@marvell.com>
10409 * tree-ssa-phiopt.c (gimple_simplify_phiopt):
10410 If "A ? B : C" fails to simplify, try "(!A) ? C : B".
10412 2021-07-05 Andrew Pinski <apinski@marvell.com>
10414 * tree-ssa-phiopt.c (match_simplify_replacement):
10415 Add early_p argument. Call gimple_simplify_phiopt
10416 instead of gimple_simplify.
10417 (tree_ssa_phiopt_worker): Update call to
10418 match_simplify_replacement and allow unconditionally.
10419 (phiopt_early_allow): New function.
10420 (gimple_simplify_phiopt): New function.
10422 2021-07-05 Andrew Pinski <apinski@marvell.com>
10424 PR middle-end/101237
10425 * fold-const.c (negate_expr_p): Remove call to element_mode
10426 and TREE_MODE/TREE_TYPE when calling HONOR_SIGNED_ZEROS,
10427 HONOR_SIGN_DEPENDENT_ROUNDING, and HONOR_SNANS.
10428 (fold_negate_expr_1): Likewise.
10429 (const_unop): Likewise.
10430 (fold_cond_expr_with_comparison): Likewise.
10431 (fold_binary_loc): Likewise.
10432 (fold_ternary_loc): Likewise.
10433 (tree_call_nonnegative_warnv_p): Likewise.
10434 * match.pd (-(A + B) -> (-B) - A): Likewise.
10436 2021-07-05 Iain Sandoe <iain@sandoe.co.uk>
10438 * configure.ac: Handle --with-dsymutil in the same way as we
10439 do for the assembler and linker. (DEFAULT_DSYMUTIL): New.
10440 Extract the type and version for the dsymutil configured or
10441 found by the default searches.
10442 * config.in: Regenerated.
10443 * configure: Regenerated.
10444 * collect2.c (do_dsymutil): Handle locating dsymutil in the
10445 same way as for the assembler and linker.
10446 * config/darwin.h (DSYMUTIL): Delete.
10447 * gcc.c: Report a configured dsymutil correctly.
10448 * exec-tool.in: Allow for dsymutil.
10450 2021-07-05 Uroš Bizjak <ubizjak@gmail.com>
10452 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
10453 Handle V4QI and V2HI modes.
10454 (expand_vec_perm_blend): Allow 4-byte vector modes with TARGET_SSE4_1.
10455 Handle V4QI mode. Emit mmx_pblendvb32 for 4-byte modes.
10456 (expand_vec_perm_pshufb): Rewrite to use switch statemets.
10457 Handle 4-byte dual operands with TARGET_XOP and single operands
10458 with TARGET_SSSE3. Emit mmx_ppermv32 for TARGET_XOP and
10459 mmx_pshufbv4qi3 for TARGET_SSSE3.
10460 (expand_vec_perm_pblendv): Allow 4-byte vector modes with TARGET_SSE4_1.
10461 (expand_vec_perm_interleave2): Allow 4-byte vector modes.
10462 (expand_vec_perm_pshufb2): Allow 4-byte vector modes with TARGET_SSSE3.
10463 (expand_vec_perm_even_odd_1): Handle V4QI mode.
10464 (expand_vec_perm_broadcast_1): Handle V4QI mode.
10465 (ix86_vectorize_vec_perm_const): Handle V4QI mode.
10466 * config/i386/mmx.md (mmx_ppermv32): New insn pattern.
10467 (mmx_pshufbv4qi3): Ditto.
10468 (*mmx_pblendw32): Ditto.
10469 (*mmx_pblendw64): Rename from *mmx_pblendw.
10470 (mmx_punpckhbw_low): New insn_and_split pattern.
10471 (mmx_punpcklbw_low): Ditto.
10473 2021-07-05 Richard Biener <rguenther@suse.de>
10475 * tree-vect-loop-manip.c (vect_loop_versioning): Do not
10476 set LOOP_C_INFINITE on the vectorized loop.
10478 2021-07-05 Richard Biener <rguenther@suse.de>
10480 PR middle-end/101291
10481 * cfgloopmanip.c (loop_version): Set the loop copy of the
10482 versioned loop to the new loop.
10484 2021-07-04 Iain Sandoe <iain@sandoe.co.uk>
10487 * config.gcc: Ensure that Darwin biarch definitions are
10488 added before i386.h.
10489 * config/i386/darwin.h (TARGET_64BIT): Remove.
10490 (PR80556_WORKAROUND): New.
10491 (REAL_LIBGCC_SPEC): Amend to use PR80556_WORKAROUND.
10492 (DARWIN_SUBARCH_SPEC): New.
10493 * config/i386/darwin32-biarch.h (TARGET_64BIT_DEFAULT,
10494 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
10495 (REAL_LIBGCC_SPEC): Remove.
10496 * config/i386/darwin64-biarch.h (TARGET_64BIT_DEFAULT,
10497 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
10498 (REAL_LIBGCC_SPEC): Remove.
10500 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
10502 PR middle-end/101294
10503 * expr.c (store_constructor): Don't use vec_duplicate on vector.
10505 2021-07-02 Martin Sebor <msebor@redhat.com>
10507 PR middle-end/98871
10508 PR middle-end/98512
10509 * diagnostic.c (get_any_inlining_info): New.
10510 (update_effective_level_from_pragmas): Handle inlining context.
10511 (diagnostic_enabled): Same.
10512 (diagnostic_report_diagnostic): Same.
10513 * diagnostic.h (struct diagnostic_info): Add ctor.
10514 (struct diagnostic_context): Add new member.
10515 * tree-diagnostic.c (set_inlining_locations): New.
10516 (tree_diagnostics_defaults): Set new callback pointer.
10518 2021-07-02 Peter Bergner <bergner@linux.ibm.com>
10520 * config/rs6000/rs6000-builtin.def (BU_MMA_PAIR_LD, BU_MMA_PAIR_ST):
10522 (__builtin_vsx_lxvp, __builtin_vsx_stxvp): New built-ins.
10523 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Expand
10524 lxvp and stxvp built-ins.
10525 (mma_init_builtins): Handle lxvp and stxvp built-ins.
10526 (builtin_function_type): Likewise.
10527 * doc/extend.texi (__builtin_vsx_lxvp, __builtin_mma_stxvp): Document.
10529 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10531 * config/h8300/h8300-protos.h (compute_a_shift_cc): Accept
10532 additional argument for the code.
10533 * config/h8300/h8300.c (compute_a_shift_cc): Accept additional
10534 argument for the code. Just return if the ZN bits are useful or
10535 not rather than the old style CC_* enums.
10536 * config/h8300/shiftrotate.md (shiftqi_noscratch): Move before
10537 more generic shiftqi patterns.
10538 (shifthi_noscratch, shiftsi_noscratch): Similarly.
10539 (shiftqi_noscratch_set_flags): New pattern.
10540 (shifthi_noscratch_set_flags, shiftsi_noscratch_set_flags): Likewise.
10542 2021-07-02 Andrew MacLeod <amacleod@redhat.com>
10544 PR tree-optimization/101223
10545 * range-op.cc (build_lt): Add -1 for signed values.
10546 (built_gt): Subtract -1 for signed values.
10548 2021-07-02 David Faust <david.faust@oracle.com>
10550 * btfout.c (get_btf_kind): Support BTF_KIND_FLOAT.
10551 (btf_asm_type): Likewise.
10553 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10555 * config/h8300/h8300-protos.h (output_a_shift): Make first argument
10556 an array of rtx rather than a pointer to rtx. Add code argument.
10557 (compute_a_shift_length): Similarly.
10558 * config/h8300/h8300.c (h8300_shift_costs): Adjust now that the
10559 shift itself isn't an operand. Create dummy operand[0] to carry
10560 a mode and pass a suitable rtx code to compute_a_shift_length.
10561 (get_shift_alg): Adjust operand number of clobber in output templates.
10562 (output_a_shift): Make first argument an array of rtx rather than
10563 a pointer to rtx. Add code argument for the type of shift.
10564 Adjust now that the shift itself is no longer an operand.
10565 (compute_a_shift_length): Similarly.
10566 * config/h8300/shiftrotate.md (shiftqi, shifthi, shiftsi): Use an
10567 iterator rather than nshift_operator.
10568 (shiftqi_noscratch, shifthi_noscratch, shiftsi_noscratch): Likewise.
10569 (shiftqi_clobber_flags): Adjust to API changes in output_a_shift
10570 and compute_a_shift_length.
10571 (shiftqi_noscratch_clobber_flags): Likewise.
10572 (shifthi_noscratch_clobber_flags): Likewise.
10573 (shiftsi_noscratch_clobber_flags): Likewise.
10575 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
10578 * config/darwin.h (DSYMUTIL_SPEC): Do not try to run
10579 dsymutil for BTF/CTF.
10581 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
10584 * config/darwin.h (CTF_INFO_SECTION_NAME): Update the
10585 segment to include BTF.
10586 (BTF_INFO_SECTION_NAME): New.
10588 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10590 * config/m32r/m32r-protos.h (call_operand): Adjust return type.
10591 (small_data_operand, memreg_operand, small_insn_p): Likewise.
10592 * config/m32r/m32r.c (call_operand): Adjust return type.
10593 (small_data_operand, memreg_operand): Likewise.
10595 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10597 * config/frv/frv-protos.h (integer_register_operand): Adjust return
10599 (frv_load_operand, gpr_or_fpr_operand, gpr_no_subreg_operand): Likewise.
10600 (fpr_or_int6_operand, gpr_or_int_operand); Likewise.
10601 (gpr_or_int12_operand, gpr_or_int10_operand); Likewise.
10602 (move_source_operand, move_destination_operand): Likewise.
10603 (condexec_source_operand, condexec_dest_operand): Likewise.
10604 (lr_operand, gpr_or_memory_operand, fpr_or_memory_operand): Likewise.
10605 (reg_or_0_operand, fcc_operand, icc_operand, cc_operand): Likewise.
10606 (fcr_operand, icr_operand, cr_operand, call_operand): Likewise.
10607 (fpr_operand, even_reg_operand, odd_reg_operand): Likewise.
10608 (even_gpr_operand, odd_gpr_operand, quad_fpr_operand): Likewise.
10609 (even_fpr_operand, odd_fpr_operand): Likewise.
10610 (dbl_memory_one_insn_operand, dbl_memory_two_insn_operand): Likewise.
10611 (int12_operand, int6_operand, int5_operand, uint5_operand): Likewise.
10612 (uint4_operand, uint1_operand, int_2word_operand): Likewise
10613 (upper_int16_operand, uint16_operand, symbolic_operand): Likewise.
10614 (relational_operator, float_relational_operator): Likewise.
10615 (ccr_eqne_operator, minmax_operator): Likewise.
10616 (condexec_si_binary_operator, condexec_si_media_operator): Likewise.
10617 (condexec_si_divide_operator, condexec_si_unary_operator): Likewise.
10618 (condexec_sf_conv_operator, condexec_sf_add_operator): Likewise.
10619 (intop_compare_operator, acc_operand, even_acc_operand): Likewise.
10620 (quad_acc_operand, accg_operand): Likewise.
10622 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10624 * config/stormy16/stormy16-protos.h (xstormy16_below_100_symbol): Change
10625 return type to a bool.
10626 (nonimmediate_nonstack_operand): Likewise.
10627 (xstormy16_splittable_below100_operand): Likewise.
10628 * config/stormy16/stormy16.c (xstormy16_below_100_symbol): Fix
10630 (xstormy16_splittable_below100_operand): Likewise.
10632 2021-07-02 Richard Biener <rguenther@suse.de>
10634 PR tree-optimization/101293
10635 * tree-ssa-loop-im.c (mem_ref_hasher::equal): Compare MEM_REF bases
10636 with combined offsets.
10637 (gather_mem_refs_stmt): Hash MEM_REFs as if their offset were
10638 combined with the rest of the offset.
10640 2021-07-02 Eric Botcazou <ebotcazou@adacore.com>
10642 * config/i386/i386.c (asm_preferred_eh_data_format): Always use the
10643 PIC encodings for PE-COFF targets.
10645 2021-07-02 Jakub Jelinek <jakub@redhat.com>
10648 * config/i386/i386-expand.c (ix86_broadcast_from_integer_constant):
10649 Return nullptr for TImode inner mode.
10651 2021-07-02 Richard Biener <rguenther@suse.de>
10653 PR tree-optimization/101280
10654 PR tree-optimization/101173
10655 * gimple-loop-interchange.cc
10656 (tree_loop_interchange::valid_data_dependences): Properly
10657 guard all dependence checks with DDR_REVERSED_P or its
10660 2021-07-02 Hongyu Wang <hongyu.wang@intel.com>
10662 * config/i386/i386-expand.c (ix86_expand_builtin):
10663 Add branch to clear odata when ZF is set for asedecenc_expand
10664 and wideaesdecenc_expand.
10666 2021-07-02 Eugene Rozenfeld <erozen@microsoft.com>
10668 * config/i386/gcc-auto-profile: regenerate
10670 2021-07-02 liuhongt <hongtao.liu@intel.com>
10672 * config/i386/sse.md (trunc<mode><pmov_dst_4>2): Refined to ..
10673 (trunc<mode><pmov_dst_4_lower>2): this.
10675 2021-07-01 David Malcolm <dmalcolm@redhat.com>
10677 * diagnostic.h (diagnostic_context::m_file_cache): New field.
10678 * input.c (class fcache): Rename to...
10679 (class file_cache_slot): ...this, making most members private and
10680 prefixing fields with "m_".
10681 (file_cache_slot::get_file_path): New accessor.
10682 (file_cache_slot::get_use_count): New accessor.
10683 (file_cache_slot::missing_trailing_newline_p): New accessor.
10684 (file_cache_slot::inc_use_count): New.
10685 (fcache_buffer_size): Move to...
10686 (file_cache_slot::buffer_size): ...here.
10687 (fcache_line_record_size): Move to...
10688 (file_cache_slot::line_record_size): ...here.
10689 (fcache_tab): Delete, in favor of global_dc->m_file_cache.
10690 (fcache_tab_size): Move to file_cache::num_file_slots.
10691 (diagnostic_file_cache_init): Update for move of fcache_tab
10692 to global_dc->m_file_cache.
10693 (diagnostic_file_cache_fini): Likewise.
10694 (lookup_file_in_cache_tab): Convert to...
10695 (file_cache::lookup_file): ...this.
10696 (diagnostics_file_cache_forcibly_evict_file): Update for move of
10697 fcache_tab to global_dc->m_file_cache, moving most of
10698 implementation to...
10699 (file_cache::forcibly_evict_file): ...this new function and...
10700 (file_cache_slot::evict): ...this new function.
10701 (evicted_cache_tab_entry): Convert to...
10702 (file_cache::evicted_cache_tab_entry): ...this.
10703 (add_file_to_cache_tab): Convert to...
10704 (file_cache::add_file): ...this, moving bulk of implementation
10706 (file_cache_slot::create): ..this new function.
10707 (file_cache::file_cache): New.
10708 (file_cache::~file_cache): New.
10709 (lookup_or_add_file_to_cache_tab): Convert to...
10710 (file_cache::lookup_or_add_file): ..this new function.
10711 (fcache::fcache): Rename to...
10712 (file_cache_slot::file_cache_slot): ...this, adding "m_" prefixes
10714 (fcache::~fcache): Rename to...
10715 (file_cache_slot::~file_cache_slot): ...this, adding "m_" prefixes
10717 (needs_read): Convert to...
10718 (file_cache_slot::needs_read_p): ...this.
10719 (needs_grow): Convert to...
10720 (file_cache_slot::needs_grow_p): ...this.
10721 (maybe_grow): Convert to...
10722 (file_cache_slot::maybe_grow): ...this.
10723 (read_data): Convert to...
10724 (file_cache_slot::read_data): ...this.
10725 (maybe_read_data): Convert to...
10726 (file_cache_slot::maybe_read_data): ...this.
10727 (get_next_line): Convert to...
10728 (file_cache_slot::get_next_line): ...this.
10729 (goto_next_line): Convert to...
10730 (file_cache_slot::goto_next_line): ...this.
10731 (read_line_num): Convert to...
10732 (file_cache_slot::read_line_num): ...this.
10733 (location_get_source_line): Update for moving of globals to
10734 global_dc->m_file_cache.
10735 (location_missing_trailing_newline): Likewise.
10736 * input.h (class file_cache_slot): New forward decl.
10737 (class file_cache): New.
10739 2021-07-01 Michael Meissner <meissner@linux.ibm.com>
10741 * config/rs6000/rs6000.c (rs6000_maybe_emit_fp_cmove): Add IEEE
10742 128-bit floating point conditional move support.
10743 (have_compare_and_set_mask): Add IEEE 128-bit floating point
10745 * config/rs6000/rs6000.md (mov<mode>cc, IEEE128 iterator): New insn.
10746 (mov<mode>cc_p10, IEEE128 iterator): New insn.
10747 (mov<mode>cc_invert_p10, IEEE128 iterator): New insn.
10748 (fpmask<mode>, IEEE128 iterator): New insn.
10749 (xxsel<mode>, IEEE128 iterator): New insn.
10751 2021-07-01 Iain Sandoe <iain@sandoe.co.uk>
10754 * config/darwin.h (CTF_INFO_SECTION_NAME): New.
10756 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
10758 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
10760 * config/i386/i386-protos.h (ix86_expand_vector_init_duplicate):
10762 * config/i386/sse.md (INT_BROADCAST_MODE): New mode iterator.
10763 (vec_duplicate<mode>): New expander.
10765 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
10768 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
10770 (ix86_byte_broadcast): New function.
10771 (ix86_convert_const_wide_int_to_broadcast): Likewise.
10772 (ix86_expand_move): Convert CONST_WIDE_INT to broadcast if mode
10773 size is 16 bytes or bigger.
10774 (ix86_broadcast_from_integer_constant): New function.
10775 (ix86_expand_vector_move): Convert CONST_WIDE_INT and CONST_VECTOR
10776 to broadcast if mode size is 16 bytes or bigger.
10777 * config/i386/i386-protos.h (ix86_gen_scratch_sse_rtx): New
10779 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): New function.
10781 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10783 * config/i386/predicates.md (ix86_endbr_immediate_operand):
10784 Return true/false instead of 1/0.
10785 (movq_parallel): Ditto.
10787 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10789 * recog.c (general_operand): Return true/false instead of 1/0.
10790 (register_operand): Ditto.
10791 (immediate_operand): Ditto.
10792 (const_int_operand): Ditto.
10793 (const_scalar_int_operand): Ditto.
10794 (const_double_operand): Ditto.
10795 (push_operand): Ditto.
10796 (pop_operand): Ditto.
10797 (memory_operand): Ditto.
10798 (indirect_operand): Ditto.
10800 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10802 * genpreds.c (write_predicate_subfunction):
10803 Change the type of written subfunction to bool.
10804 (write_one_predicate_function):
10805 Change the type of written function to bool.
10806 (write_tm_preds_h): Ditto.
10807 * recog.h (*insn_operand_predicate_fn): Change the type to bool.
10808 * recog.c (general_operand): Change the type to bool.
10809 (address_operand): Ditto.
10810 (register_operand): Ditto.
10811 (pmode_register_operand): Ditto.
10812 (scratch_operand): Ditto.
10813 (immediate_operand): Ditto.
10814 (const_int_operand): Ditto.
10815 (const_scalar_int_operand): Ditto.
10816 (const_double_operand): Ditto.
10817 (nonimmediate_operand): Ditto.
10818 (nonmemory_operand): Ditto.
10819 (push_operand): Ditto.
10820 (pop_operand): Ditto.
10821 (memory_operand): Ditto.
10822 (indirect_operand): Ditto.
10823 (ordered_comparison_operator): Ditto.
10824 (comparison_operator): Ditto.
10825 * config/i386/i386-expand.c (ix86_expand_sse_cmp):
10826 Change the type of indirect predicate function to bool.
10827 * config/rs6000/rs6000.c (easy_vector_constant):
10828 Change the type to bool.
10829 * config/mips/mips-protos.h (m16_based_address_p):
10830 Change the type of operand 3 to bool.
10832 2021-07-01 Richard Biener <rguenther@suse.de>
10834 PR tree-optimization/101280
10835 PR tree-optimization/101173
10836 * gimple-loop-interchange.cc
10837 (tree_loop_interchange::valid_data_dependences): Revert
10838 previous change and instead correctly handle DDR_REVERSED_P
10841 2021-07-01 Richard Biener <rguenther@suse.de>
10843 PR tree-optimization/101278
10844 * tree-ssa-dse.c (dse_classify_store): First check for
10845 uses, then ignore stmt for chaining purposes.
10847 2021-07-01 Richard Biener <rguenther@suse.de>
10849 PR tree-optimization/100778
10850 * tree-vect-slp.c (vect_schedule_slp_node): Do not place trapping
10851 vectorized ops ahead of their scalar BB.
10853 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10856 * config/i386/i386.md (*nabs<dwi>2_doubleword):
10857 New insn_and_split pattern.
10858 (*nabs<dwi>2_1): Ditto.
10859 * config/i386/i386-features.c
10860 (general_scalar_chain::compute_convert_gain):
10861 Handle (NEG (ABS (...))) RTX. Rewrite src code
10862 scanner as switch statement.
10863 (general_scalar_chain::convert_insn):
10864 Handle (NEG (ABS (...))) RTX.
10865 (general_scalar_to_vector_candidate_p):
10866 Detect (NEG (ABS (...))) RTX. Reorder case statements
10867 for (AND (NOT (...) ...)) fallthrough.
10869 2021-07-01 Richard Biener <rguenther@suse.de>
10871 PR tree-optimization/101178
10872 * tree-vect-slp.c (slpg_vertex::materialize): Remove.
10873 (slpg::perm_in): Add.
10874 (slpg::get_perm_in): Remove.
10875 (slpg::get_perm_materialized): Add.
10876 (vect_optimize_slp): Handle VEC_PERM nodes more optimally
10877 during permute propagation and materialization.
10879 2021-07-01 Jakub Jelinek <jakub@redhat.com>
10882 * dwarf2out.c (loc_list_from_tree_1): Handle COMPOUND_LITERAL_EXPR.
10884 2021-07-01 Jakub Jelinek <jakub@redhat.com>
10886 PR middle-end/94366
10887 * omp-low.c (lower_rec_input_clauses): Rename is_fp_and_or to
10888 is_truth_op, set it for TRUTH_*IF_EXPR regardless of new_var's type,
10889 use boolean_type_node instead of integer_type_node as NE_EXPR type.
10890 (lower_reduction_clauses): Likewise.
10892 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
10894 * config/gcn/gcn.c: Include dwarf2.h.
10895 (gcn_addr_space_debug): New function.
10896 (TARGET_ADDR_SPACE_DEBUG): New hook.
10898 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
10900 * common/config/gcn/gcn-common.c
10901 (gcn_option_optimization_table): Change OPT_fomit_frame_pointer to -O3.
10902 * config/gcn/gcn.c (gcn_expand_prologue): Prefer the frame pointer
10904 (gcn_expand_prologue): Prefer the frame pointer when emitting CFI.
10905 (gcn_frame_pointer_rqd): New function.
10906 (TARGET_FRAME_POINTER_REQUIRED): New hook.
10908 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
10910 * config/gcn/gcn.c (move_callee_saved_registers): Emit CFI notes for
10911 prologue register saves.
10912 (gcn_debug_unwind_info): Use UI_DWARF2.
10913 (gcn_dwarf_register_number): Map DWARF_LINK_REGISTER to DWARF PC.
10914 (gcn_dwarf_register_span): DWARF_LINK_REGISTER doesn't span.
10915 * config/gcn/gcn.h: (DWARF_FRAME_RETURN_COLUMN): New define.
10916 (DWARF_LINK_REGISTER): New define.
10917 (FIRST_PSEUDO_REGISTER): Increment.
10918 (FIXED_REGISTERS): Add entry for DWARF_LINK_REGISTER.
10919 (CALL_USED_REGISTERS): Likewise.
10920 (REGISTER_NAMES): Likewise.
10922 2021-06-30 Richard Biener <rguenther@suse.de>
10924 PR tree-optimization/101267
10925 * tree-vect-stmts.c (vect_check_scalar_mask): Adjust
10926 API and use SLP compatible interface of vect_is_simple_use.
10927 Reject not vectorized SLP defs for callers that do not support
10929 (vect_check_store_rhs): Handle masked stores and pass down
10930 the appropriate operator index.
10931 (vectorizable_call): Adjust.
10932 (vectorizable_store): Likewise.
10933 (vectorizable_load): Likewise. Handle SLP pecularity of
10935 (vect_is_simple_use): Remove special-casing of masked stores.
10937 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
10939 * common.opt (foffload): Remove help as Driver only.
10940 * gcc.c (display_help): Add -foffload.
10942 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
10944 * gcc.c (close_at_file, execute): Replace alloca by XALLOCAVEC.
10945 (check_offload_target_name): Fix splitting OFFLOAD_TARGETS into
10946 a candidate list; better inform no offload target is configured
10947 and fix hint extraction when passed target is not '\0' at [len].
10948 * common.opt (foffload): Add tailing '.'.
10949 (foffload-options): Likewise; fix flag name in the help string.
10951 2021-06-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
10954 * config/arm/arm_neon.h: Move vabs intrinsics before vcage_f32.
10955 (vcage_f32): Gate comparison on __FAST_MATH__.
10956 (vcageq_f32): Likewise.
10957 (vcale_f32): Likewise.
10958 (vcaleq_f32): Likewise.
10959 (vcagt_f32): Likewise.
10960 (vcagtq_f32): Likewise.
10961 (vcalt_f32): Likewise.
10962 (vcaltq_f32): Likewise.
10963 (vcage_f16): Likewise.
10964 (vcageq_f16): Likewise.
10965 (vcale_f16): Likewise.
10966 (vcaleq_f16): Likewise.
10967 (vcagt_f16): Likewise.
10968 (vcagtq_f16): Likewise.
10969 (vcalt_f16): Likewise.
10970 (vcaltq_f16): Likewise.
10972 2021-06-30 Richard Biener <rguenther@suse.de>
10974 PR tree-optimization/101264
10975 * tree-vect-slp.c (vect_optimize_slp): Propagate the
10976 computed perm_in to all "any" permute successors
10977 we cannot de-duplicate immediately.
10979 2021-06-30 liuhongt <hongtao.liu@intel.com>
10982 * config/i386/sse.md
10983 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>):
10985 (avx512f_sfixupimm<mode><maskz_scalar_name><round_saeonly_name>):
10987 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>"): Refined.
10988 * config/i386/subst.md (maskz_scalar): New define_subst.
10989 (maskz_scalar_name): New subst_attr.
10990 (maskz_scalar_op5): Ditto.
10991 (round_saeonly_maskz_scalar_op5): Ditto.
10992 (round_saeonly_maskz_scalar_operand5): Ditto.
10994 2021-06-30 David Edelsohn <dje.gcc@gmail.com>
10996 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
10997 Increase code CSECT alignment to at least 32 bytes.
10998 * config/rs6000/xcoff.h (TEXT_SECTION_ASM_OP): Add 32 byte
10999 alignment designation.
11001 2021-06-29 Sergei Trofimovich <siarheit@google.com>
11003 * doc/generic.texi: Fix s/net yet/not yet/ typo.
11005 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
11007 PR tree-optimization/101254
11008 * range-op.cc (operator_minus::op1_op2_relation_effect): Check for
11009 wrapping/non-wrapping when setting the result range.
11011 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
11013 * value-query.cc (gimple_range_global): Allow phis.
11015 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
11017 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): Use stmt.
11018 (simplify_using_ranges::op_with_boolean_value_range_p): Add a
11019 statement for location context.
11020 (check_for_binary_op_overflow): Ditto.
11021 (simplify_using_ranges::get_vr_for_comparison): Ditto.
11022 (simplify_using_ranges::compare_name_with_value): Ditto.
11023 (simplify_using_ranges::compare_names): Ditto.
11024 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Ditto.
11025 (simplify_using_ranges::simplify_truth_ops_using_ranges): Ditto.
11026 (simplify_using_ranges::simplify_min_or_max_using_ranges): Ditto.
11027 (simplify_using_ranges::simplify_internal_call_using_ranges): Ditto.
11028 (simplify_using_ranges::two_valued_val_range_p): Ditto.
11029 (simplify_using_ranges::simplify): Ditto.
11030 * vr-values.h: Adjust prototypes.
11032 2021-06-29 Uroš Bizjak <ubizjak@gmail.com>
11035 * config/i386/mmx.md (vec_addsubv2sf3): New insn pattern.
11037 2021-06-29 Julian Brown <julian@codesourcery.com>
11039 * config/gcn/gcn.c (gcn_init_libfuncs): New function.
11040 (TARGET_INIT_LIBFUNCS): Define target hook using above function.
11041 * config/gcn/gcn.h (UNITS_PER_WORD): Define to 8 for IN_LIBGCC2, 4
11043 (LIBGCC2_UNITS_PER_WORD, BITS_PER_WORD): Remove definitions.
11044 (MAX_FIXED_MODE_SIZE): Change to 128.
11046 2021-06-29 Julian Brown <julian@codesourcery.com>
11048 * config/gcn/gcn.md (UNSPEC_FLBIT_INT): New unspec constant.
11049 (s_mnemonic): Add clrsb.
11050 (gcn_flbit<mode>_int): Add insn pattern for SImode/DImode.
11051 (clrsb<mode>2): Add expander for SImode/DImode.
11053 2021-06-29 Julian Brown <julian@codesourcery.com>
11055 * config/gcn/gcn.md (<su>mulsidi3, <su>mulsidi3_reg, <su>mulsidi3_imm,
11056 muldi3): Add patterns.
11058 2021-06-29 Julian Brown <julian@codesourcery.com>
11060 * config/gcn/gcn.md (<su>mulsi3_highpart): Change to expander.
11061 (<su>mulsi3_highpart_reg, <su>mulsi3_highpart_imm): New patterns.
11063 2021-06-29 Julian Brown <julian@codesourcery.com>
11065 * config/gcn/gcn.md (mulsi3): Make s_mulk_i32 variant clobber SCC.
11067 2021-06-29 Joseph Myers <joseph@codesourcery.com>
11069 * btfout.c, ctfout.c: Include "memmodel.h".
11071 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
11073 * gcc.c (check_offload_target_name): Cast len argument to
11074 %q.*s to 'int'; avoid -Wstringop-truncation warning.
11076 2021-06-29 Richard Biener <rguenther@suse.de>
11078 * tree-vect-slp.c (vect_optimize_slp): Forward propagate
11079 to "any" permute nodes and relax "any" permute proapgation
11080 during iterative backward propagation.
11082 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
11085 * common.opt (-foffload=): Update description.
11086 (-foffload-options=): New.
11087 * doc/invoke.texi (C Language Options): Document
11088 -foffload and -foffload-options.
11089 * gcc.c (check_offload_target_name): New, split off from
11090 handle_foffload_option.
11091 (check_foffload_target_names): New.
11092 (handle_foffload_option): Handle -foffload=default.
11093 (driver_handle_option): Update for -foffload-options.
11094 * lto-opts.c (lto_write_options): Use -foffload-options
11095 instead of -foffload.
11096 * lto-wrapper.c (merge_and_complain, append_offload_options):
11098 * opts.c (common_handle_option): Likewise.
11100 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
11102 * doc/invoke.texi (C Language Options): Sort options
11103 alphabetically in optlist and also the description itself.
11104 Remove leftover -fallow-single-precision from and add missing
11105 -fgnu-tm to the optlist.
11107 2021-06-29 Richard Biener <rguenther@suse.de>
11109 * tree-vect-slp.c (slpg_vertex::visited): Remove.
11110 (vect_slp_perms_eq): Handle -1 permutes.
11111 (vect_optimize_slp): Rewrite permute propagation.
11113 2021-06-29 Jakub Jelinek <jakub@redhat.com>
11116 * match.pd ((intptr_t)x eq/ne CST to x eq/ne (typeof x) CST): Don't
11117 perform the optimization in GENERIC when sanitizing and x has a
11120 2021-06-29 Richard Biener <rguenther@suse.de>
11122 PR tree-optimization/101242
11123 * tree-vect-slp.c (vect_slp_build_vertices): Force-add
11124 PHIs with not represented initial values as leafs.
11126 2021-06-29 Jan-Benedict Glaw <jbglaw@getslash.de>
11128 * config/pdp11/pdp11.h (ASM_OUTPUT_SKIP): Fix signedness warning.
11129 * config/pdp11/pdp11.c (pdp11_asm_print_operand_punct_valid_p): Remove
11130 "register" keyword.
11131 (pdp11_initial_elimination_offset) Remove unused variable.
11132 (pdp11_cmp_length) Ditto.
11133 (pdp11_insn_cost): Ditto, and fix signedness warning.
11135 2021-06-29 David Edelsohn <dje.gcc@gmail.com>
11137 * btfout.c: Include tm_p.h.
11140 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
11142 * config/bpf/bpf.c (bpf_expand_prologue): Do not mark insns as
11144 (bpf_expand_epilogue): Likewise.
11145 * config/bpf/bpf.h (DWARF2_FRAME_INFO): Define to 0.
11146 Do not define DBX_DEBUGGING_INFO.
11148 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
11150 * doc/invoke.texi: Document the CTF and BTF debug info options.
11152 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
11153 David Faust <david.faust@oracle.com>
11154 Jose E. Marchesi <jose.marchesi@oracle.com>
11155 Weimin Pan <weimin.pan@oracle.com>
11157 * Makefile.in: Add ctfc.*, ctfout.c and btfout.c files to
11158 GTFILES. Add new object files.
11159 * common.opt: Add CTF and BTF debug info options.
11160 * btfout.c: New file.
11161 * ctfc.c: Likewise.
11162 * ctfc.h: Likewise.
11163 * ctfout.c: Likewise.
11164 * dwarf2ctf.c: Likewise.
11165 * dwarf2ctf.h: Likewise.
11166 * dwarf2cfi.c (dwarf2out_do_frame): Acknowledge CTF_DEBUG and
11168 * dwarf2out.c (dwarf2out_source_line): Likewise.
11169 (dwarf2out_finish): Skip emitting DWARF if CTF or BTF are to
11171 (debug_format_do_cu): New function.
11172 (dwarf2out_early_finish): Traverse DIEs and emit CTF/BTF for
11174 Include dwarf2ctf.c.
11175 * final.c (dwarf2_debug_info_emitted_p): Acknowledge DWARF-based debug
11177 * flag-types.h (enum debug_info_type): Add CTF_DEBUG and BTF_DEBUG.
11178 (CTF_DEBUG): New bitmask.
11179 (BTF_DEBUG): Likewise.
11180 (enum ctf_debug_info_levels): New enum.
11181 * gengtype.c (open_base_files): Handle ctfc.h.
11182 (main): Handle uint32_t type.
11183 * flags.h (btf_debuginfo_p): New definition.
11184 (dwarf_based_debuginfo_p): Likewise.
11185 * opts.c (debug_type_names): Add entries for CTF and BTF.
11186 (btf_debuginfo_p): New function.
11187 (dwarf_based_debuginfo_p): Likewise.
11188 (common_handle_option): Handle -gctfN and -gbtf options.
11189 (set_debug_level): Set CTF_DEBUG, BTF_DEBUG whenever appropriate.
11190 * toplev.c (process_options): Inform the user and ignore -gctfLEVEL if
11193 2021-06-28 Jose E. Marchesi <jose.marchesi@oracle.com>
11195 * dwarf2out.c (AT_class): Function is no longer static.
11196 (AT_int): Likewise.
11197 (AT_unsigned): Likewise.
11198 (AT_loc): Likewise.
11199 (get_AT): Likewise.
11200 (get_AT_string): Likewise.
11201 (get_AT_flag): Likewise.
11202 (get_AT_unsigned): Likewise.
11203 (get_AT_ref): Likewise.
11204 (new_die_raw): Likewise.
11205 (lookup_decl_die): Likewise.
11206 (base_type_die): Likewise.
11207 (add_name_attribute): Likewise.
11208 (add_AT_int): Likewise.
11209 (add_AT_unsigned): Likewise.
11210 (add_AT_loc): Likewise.
11211 (dw_get_die_tag): New function.
11212 (dw_get_die_child): Likewise.
11213 (dw_get_die_sib): Likewise.
11214 (struct dwarf_file_data): Move from here to dwarf2out.h
11215 (struct dw_attr_struct): Likewise.
11216 * dwarf2out.h: Analogous changes.
11218 2021-06-28 Martin Jambor <mjambor@suse.cz>
11221 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
11222 members m_dead_stmts and m_dead_ssas.
11223 * ipa-param-manipulation.c
11224 (ipa_param_body_adjustments::mark_dead_statements): New function.
11225 (ipa_param_body_adjustments::common_initialization): Call it on
11226 all removed but not split parameters.
11227 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
11229 (ipa_param_body_adjustments::modify_call_stmt): Remove arguments that
11231 * tree-inline.c (remap_gimple_stmt): Do not copy dead statements, reset
11232 dead debug statements.
11233 (copy_phis_for_bb): Do not copy dead PHI nodes.
11235 2021-06-28 Martin Jambor <mjambor@suse.cz>
11238 * symtab-clones.h (clone_info): Removed member param_adjustments.
11239 * ipa-param-manipulation.h: Adjust initial comment to reflect how we
11240 deal with pass-through splits now.
11241 (ipa_param_performed_split): Removed.
11242 (ipa_param_adjustments::modify_call): Adjusted parameters.
11243 (class ipa_param_body_adjustments): Adjusted parameters of
11244 register_replacement, modify_gimple_stmt and modify_call_stmt.
11245 (ipa_verify_edge_has_no_modifications): Declare.
11246 (ipa_edge_modifications_finalize): Declare.
11247 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Remove
11248 performed_splits processing, pas only edge to padjs->modify_call,
11249 check that call arguments were not modified if they should not have
11251 * cgraphclones.c (cgraph_node::create_clone): Do not copy performed
11253 * ipa-param-manipulation.c (struct pass_through_split_map): New type.
11254 (ipa_edge_modification_info): Likewise.
11255 (ipa_edge_modification_sum): Likewise.
11256 (ipa_edge_modifications): New edge summary.
11257 (ipa_verify_edge_has_no_modifications): New function.
11258 (transitive_split_p): Removed.
11259 (transitive_split_map): Likewise.
11260 (init_transitive_splits): Likewise.
11261 (ipa_param_adjustments::modify_call): Adjusted to use the new edge
11262 summary instead of performed_splits.
11263 (ipa_param_body_adjustments::register_replacement): Drop dummy
11264 parameter, set base_index of the created ipa_param_body_replacement.
11265 (phi_arg_will_live_p): New function.
11266 (ipa_param_body_adjustments::common_initialization): Do not create
11267 IPA_SRA dummy decls.
11268 (simple_tree_swap_info): Removed.
11269 (remap_split_decl_to_dummy): Likewise.
11270 (record_argument_state_1): New function.
11271 (record_argument_state): Likewise.
11272 (ipa_param_body_adjustments::modify_call_stmt): New parameter
11273 orig_stmt. Do not work with dummy decls, save necessary info about
11274 changes to ipa_edge_modifications.
11275 (ipa_param_body_adjustments::modify_gimple_stmt): New parameter
11276 orig_stmt, pass it to modify_call_stmt.
11277 (ipa_param_body_adjustments::modify_cfun_body): Adjust call to
11278 modify_gimple_stmt.
11279 (ipa_edge_modifications_finalize): New function.
11280 * tree-inline.c (remap_gimple_stmt): Pass original statement to
11281 modify_gimple_stmt.
11282 (copy_phis_for_bb): Do not copy dead PHI nodes.
11283 (expand_call_inline): Do not remap performed_splits.
11284 (update_clone_info): Likewise.
11285 * toplev.c: Include ipa-param-manipulation.h.
11286 (toplev::finalize): Call ipa_edge_modifications_finalize.
11288 2021-06-28 Andrew Pinski <apinski@marvell.com>
11290 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Duplicate range
11291 info if we're the only things setting the target PHI.
11292 (value_replacement): Don't duplicate range here.
11293 (minmax_replacement): Likewise.
11295 2021-06-28 Richard Biener <rguenther@suse.de>
11297 PR tree-optimization/101229
11298 * gimple-walk.c (gimple_walk_op): Handle PHIs.
11300 2021-06-28 Martin Liska <mliska@suse.cz>
11302 * config/v850/v850.c (construct_dispose_instruction): Allocate
11304 (construct_prepare_instruction): Likewise.
11306 2021-06-28 Martin Liska <mliska@suse.cz>
11308 * config/v850/v850.c (v850_option_override): Build default
11310 (v850_can_inline_p): New. Allow MASK_PROLOG_FUNCTION to be
11311 ignored for inlining.
11312 (TARGET_CAN_INLINE_P): New.
11314 2021-06-28 Richard Biener <rguenther@suse.de>
11316 PR tree-optimization/101207
11317 * tree-vect-slp.c (vect_optimize_slp): Do BB reduction
11318 permute eliding for load permutations properly.
11320 2021-06-28 Richard Biener <rguenther@suse.de>
11322 PR tree-optimization/101173
11323 * gimple-loop-interchange.cc
11324 (tree_loop_interchange::valid_data_dependences): Disallow outer
11325 loop dependence distance of zero.
11327 2021-06-28 liuhongt <hongtao.liu@intel.com>
11330 * config/i386/sse.md (*avx_cmp<mode>3_lt): New
11331 define_insn_and_split.
11332 (*avx_cmp<mode>3_ltint): Ditto.
11333 (*avx2_pcmp<mode>3_3): Ditto.
11334 (*avx2_pcmp<mode>3_4): Ditto.
11335 (*avx2_pcmp<mode>3_5): Ditto.
11337 2021-06-28 liuhongt <hongtao.liu@intel.com>
11339 * config/i386/i386-builtin.def (IX86_BUILTIN_BLENDVPD256,
11340 IX86_BUILTIN_BLENDVPS256, IX86_BUILTIN_PBLENDVB256,
11341 IX86_BUILTIN_BLENDVPD, IX86_BUILTIN_BLENDVPS,
11342 IX86_BUILTIN_PBLENDVB128): Replace icode with
11344 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold blendv
11346 * config/i386/sse.md (*<sse4_1_avx2>_pblendvb_lt_subreg_not):
11347 New pre_reload splitter.
11349 2021-06-27 Andrew Pinski <apinski@marvell.com>
11351 PR middle-end/101230
11352 * fold-const.c (fold_ternary_loc): Check
11353 the return value of invert_tree_comparison.
11355 2021-06-27 David Edelsohn <dje.gcc@gmail.com>
11357 * config.gcc: Add SPDX License Identifier.
11358 (powerpc-ibm-aix789): Default to aix73.h.
11359 (powerpc-ibm-aix7.2.*.*): New stanza.
11360 * config/rs6000/aix72.h: Add SPDX License Identifier.
11361 * config/rs6000/aix73.h: New file.
11363 2021-06-26 Jason Merrill <jason@redhat.com>
11365 * except.c: #include "dwarf2.h" instead of "dwarf2out.h".
11367 2021-06-26 Andrew Pinski <apinski@marvell.com>
11369 * genmatch.c (lower_cond): Copy for_subst_vec
11370 for the simplify also.
11371 (lower): Swap the order for lower_for and lower_cond.
11373 2021-06-26 Andrew Pinski <apinski@marvell.com>
11375 * tree-ssa-phiopt.c (match_simplify_replacement): Reset
11376 flow senatitive info on the moved ssa set.
11378 2021-06-26 Andrew Pinski <apinski@marvell.com>
11380 * fold-const.c (fold_cond_expr_with_comparison):
11381 Exand arg0 into comp_code, arg00, and arg01.
11382 (fold_ternary_loc): Use invert_tree_comparison
11383 instead of fold_invert_truthvalue for the case
11384 where we have A CMP B ? C : A.
11386 2021-06-25 Martin Sebor <msebor@redhat.com>
11388 PR middle-end/101216
11389 * calls.c (maybe_warn_rdwr_sizes): Use the no_warning constant.
11391 2021-06-25 Jeff Law <jeffreyalaw@gmail.com>
11393 * config/h8300/h8300.c (select_cc_mode): Handle ASHIFTRT and LSHIFTRT.
11395 2021-06-25 Richard Biener <rguenther@suse.de>
11397 PR tree-optimization/101202
11398 * tree-vect-slp.c (vect_optimize_slp): Explicitely handle
11401 2021-06-25 Richard Biener <rguenther@suse.de>
11403 * tree-vect-slp-patterns.c (addsub_pattern::build): Copy
11404 STMT_VINFO_REDUC_DEF from the original representative.
11406 2021-06-25 Martin Sebor <msebor@redhat.com>
11408 * builtins.c (warn_string_no_nul): Replace uses of TREE_NO_WARNING,
11409 gimple_no_warning_p and gimple_set_no_warning with
11410 warning_suppressed_p, and suppress_warning.
11412 (maybe_warn_for_bound): Same.
11413 (warn_for_access): Same.
11414 (check_access): Same.
11415 (expand_builtin_strncmp): Same.
11416 (fold_builtin_varargs): Same.
11417 * calls.c (maybe_warn_nonstring_arg): Same.
11418 (maybe_warn_rdwr_sizes): Same.
11419 * cfgexpand.c (expand_call_stmt): Same.
11420 * cgraphunit.c (check_global_declaration): Same.
11421 * fold-const.c (fold_undefer_overflow_warnings): Same.
11422 (fold_truth_not_expr): Same.
11423 (fold_unary_loc): Same.
11424 (fold_checksum_tree): Same.
11425 * gimple-array-bounds.cc (array_bounds_checker::check_array_ref): Same.
11426 (array_bounds_checker::check_mem_ref): Same.
11427 (array_bounds_checker::check_addr_expr): Same.
11428 (array_bounds_checker::check_array_bounds): Same.
11429 * gimple-expr.c (copy_var_decl): Same.
11430 * gimple-fold.c (gimple_fold_builtin_strcpy): Same.
11431 (gimple_fold_builtin_strncat): Same.
11432 (gimple_fold_builtin_stxcpy_chk): Same.
11433 (gimple_fold_builtin_stpcpy): Same.
11434 (gimple_fold_builtin_sprintf): Same.
11435 (fold_stmt_1): Same.
11436 * gimple-ssa-isolate-paths.c (diag_returned_locals): Same.
11437 * gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Same.
11438 * gimple-ssa-sprintf.c (handle_printf_call): Same.
11439 * gimple-ssa-store-merging.c (imm_store_chain_info::output_merged_store): Same.
11440 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
11441 * gimple-ssa-warn-restrict.h: Adjust declarations.
11442 (maybe_diag_access_bounds): Replace uses of TREE_NO_WARNING,
11443 gimple_no_warning_p and gimple_set_no_warning with
11444 warning_suppressed_p, and suppress_warning.
11445 (check_call): Same.
11446 (check_bounds_or_overlap): Same.
11447 * gimple.c (gimple_build_call_from_tree): Same.
11448 * gimplify.c (gimplify_return_expr): Same.
11449 (gimplify_cond_expr): Same.
11450 (gimplify_modify_expr_complex_part): Same.
11451 (gimplify_modify_expr): Same.
11452 (gimple_push_cleanup): Same.
11453 (gimplify_expr): Same.
11454 * omp-expand.c (expand_omp_for_generic): Same.
11455 (expand_omp_taskloop_for_outer): Same.
11456 * omp-low.c (lower_rec_input_clauses): Same.
11457 (lower_lastprivate_clauses): Same.
11458 (lower_send_clauses): Same.
11459 (lower_omp_target): Same.
11460 * tree-cfg.c (pass_warn_function_return::execute): Same.
11461 * tree-complex.c (create_one_component_var): Same.
11462 * tree-inline.c (remap_gimple_op_r): Same.
11463 (copy_tree_body_r): Same.
11464 (declare_return_variable): Same.
11465 (expand_call_inline): Same.
11466 * tree-nested.c (lookup_field_for_decl): Same.
11467 * tree-sra.c (create_access_replacement): Same.
11468 (generate_subtree_copies): Same.
11469 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Same.
11470 * tree-ssa-forwprop.c (combine_cond_expr_cond): Same.
11471 * tree-ssa-loop-ch.c (ch_base::copy_headers): Same.
11472 * tree-ssa-loop-im.c (execute_sm): Same.
11473 * tree-ssa-phiopt.c (cond_store_replacement): Same.
11474 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
11475 (handle_builtin_strcpy): Same.
11476 (maybe_diag_stxncpy_trunc): Same.
11477 (handle_builtin_stxncpy_strncat): Same.
11478 (handle_builtin_strcat): Same.
11479 * tree-ssa-uninit.c (get_no_uninit_warning): Same.
11480 (set_no_uninit_warning): Same.
11481 (uninit_undefined_value_p): Same.
11482 (warn_uninit): Same.
11483 (maybe_warn_operand): Same.
11484 * tree-vrp.c (compare_values_warnv): Same.
11485 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr): Same.
11486 (test_for_singularity): Same.
11487 * gimple.h (warning_suppressed_p): New function.
11488 (suppress_warning): Same.
11489 (copy_no_warning): Same.
11490 (gimple_set_block): Call gimple_set_location.
11491 (gimple_set_location): Call copy_warning.
11493 2021-06-25 Martin Sebor <msebor@redhat.com>
11495 * tree.h (warning_suppressed_at, copy_warning,
11496 warning_suppressed_p, suppress_warning): New functions.
11498 2021-06-25 Martin Sebor <msebor@redhat.com>
11500 * Makefile.in (OBJS-libcommon): Add diagnostic-spec.o.
11501 * gengtype.c (open_base_files): Add diagnostic-spec.h.
11502 * diagnostic-spec.c: New file.
11503 * diagnostic-spec.h: New file.
11504 * tree.h (no_warning, all_warnings, suppress_warning_at): New
11506 * warning-control.cc: New file.
11508 2021-06-25 liuhongt <hongtao.liu@intel.com>
11511 * config/i386/i386.c (x86_order_regs_for_local_alloc):
11514 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11516 PR tree-optimization/101189
11517 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Pass
11518 LHS range of condition to postfold routine.
11519 (fold_using_range::postfold_gcond_edges): Only process the TRUE or
11520 FALSE edge if the LHS range supports it being taken.
11521 * gimple-range-fold.h (postfold_gcond_edges): Add range parameter.
11523 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11525 * value-relation.cc (equiv_oracle::dump): Do not dump NULL blocks.
11526 (relation_oracle::find_relation_block): Check correct bitmap.
11527 (relation_oracle::dump): Do not dump NULL blocks.
11529 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11531 * gimple-range-cache.cc (ranger_cache::propagate_cache): Call
11532 range_on_edge instead of manually calculating.
11534 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11536 * range-op.cc: Fix comment.
11538 2021-06-24 Uroš Bizjak <ubizjak@gmail.com>
11541 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
11542 Handle V8QI and V4HI modes.
11543 * config/i386/mmx.md (sse4_1_<any_extend:code>v4qiv4hi2):
11545 (sse4_1_<any_extend:code>v4qiv4hi2): Ditto.
11546 (mmxpackmode): New mode attribute.
11547 (vec_pack_trunc_<mmxpackmode:mode>): New expander.
11548 (mmxunpackmode): New mode attribute.
11549 (vec_unpacks_lo_<mmxunpackmode:mode>): New expander.
11550 (vec_unpacks_hi_<mmxunpackmode:mode>): Ditto.
11551 (vec_unpacku_lo_<mmxunpackmode:mode>): Ditto.
11552 (vec_unpacku_hi_<mmxunpackmode:mode>): Ditto.
11553 * config/i386/i386.md (extsuffix): Move from ...
11554 * config/i386/sse.md: ... here.
11556 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
11558 * dwarf2out.c (dwarf2out_assembly_start): Emit .file 0 marker here..
11559 (dwarf2out_finish): ...instead of here.
11561 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
11563 * configure.ac (--gdwarf-5 option): Use objdump instead of readelf.
11564 (working --gdwarf-4/--gdwarf-5 for all sources): Likewise.
11565 (--gdwarf-4 not refusing generated .debug_line): Adjust for Windows.
11566 * configure: Regenerate.
11568 2021-06-24 Richard Biener <rguenther@suse.de>
11570 * config/i386/sse.md (vec_addsubv4df3, vec_addsubv2df3,
11571 vec_addsubv8sf3, vec_addsubv4sf3): Merge into ...
11572 (vec_addsub<mode>3): ... using a new addsub_cst mode attribute.
11574 2021-06-24 Richard Biener <rguenther@suse.de>
11576 * config/i386/sse.md (avx_addsubv4df3): Rename to
11578 (avx_addsubv8sf3): Rename to vec_addsubv8sf3.
11579 (sse3_addsubv2df3): Rename to vec_addsubv2df3.
11580 (sse3_addsubv4sf3): Rename to vec_addsubv4sf3.
11581 * config/i386/i386-builtin.def: Adjust.
11582 * internal-fn.def (VEC_ADDSUB): New internal optab fn.
11583 * optabs.def (vec_addsub_optab): New optab.
11584 * tree-vect-slp-patterns.c (class addsub_pattern): New.
11585 (slp_patterns): Add addsub_pattern.
11586 * tree-vect-slp.c (vect_optimize_slp): Disable propagation
11587 across CFN_VEC_ADDSUB.
11588 * tree-vectorizer.h (vect_pattern::vect_pattern): Make
11590 * doc/md.texi (vec_addsub<mode>3): Document.
11592 2021-06-24 Jakub Jelinek <jakub@redhat.com>
11594 PR middle-end/101170
11595 * df-scan.c (df_ref_record): For paradoxical big-endian SUBREGs
11596 where regno + subreg_regno_offset wraps around use 0 as starting
11599 2021-06-24 Jakub Jelinek <jakub@redhat.com>
11601 PR middle-end/101172
11602 * stor-layout.c (finish_bitfield_representative): If nextf has
11603 error_mark_node type, set repr type to error_mark_node too.
11605 2021-06-24 Ilya Leoshkevich <iii@linux.ibm.com>
11607 * config/s390/s390.c (s390_function_profiler): Ignore labelno
11609 * config/s390/s390.h (NO_PROFILE_COUNTERS): Define.
11611 2021-06-24 Richard Biener <rguenther@suse.de>
11613 * tree-vect-slp.c (vect_optimize_slp): Do not propagate
11614 across operations that have different semantics on different
11617 2021-06-24 Jakub Jelinek <jakub@redhat.com>
11619 * tree.h (OMP_CLAUSE_MAP_IN_REDUCTION): Document meaning for OpenMP.
11620 * gimplify.c (gimplify_scan_omp_clauses): For OpenMP map clauses
11621 with OMP_CLAUSE_MAP_IN_REDUCTION flag partially defer gimplification
11622 of non-decl OMP_CLAUSE_DECL. For OMP_CLAUSE_IN_REDUCTION on
11623 OMP_TARGET user outer_ctx instead of ctx for placeholders and
11624 initializer/combiner gimplification.
11625 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_MAP_IN_REDUCTION
11626 on target constructs.
11627 (lower_rec_input_clauses): Likewise.
11628 (lower_omp_target): Likewise.
11629 * omp-expand.c (expand_omp_target): Temporarily ignore nowait clause
11630 on target if in_reduction is present.
11632 2021-06-24 Kewen Lin <linkw@linux.ibm.com>
11634 * tree-predcom.c (class pcom_worker): New class.
11635 (release_chain): Renamed to...
11636 (pcom_worker::release_chain): ...this.
11637 (release_chains): Renamed to...
11638 (pcom_worker::release_chains): ...this.
11639 (aff_combination_dr_offset): Renamed to...
11640 (pcom_worker::aff_combination_dr_offset): ...this.
11641 (determine_offset): Renamed to...
11642 (pcom_worker::determine_offset): ...this.
11643 (class comp_ptrs): New class.
11644 (split_data_refs_to_components): Renamed to...
11645 (pcom_worker::split_data_refs_to_components): ...this,
11646 and update with class comp_ptrs.
11647 (suitable_component_p): Renamed to...
11648 (pcom_worker::suitable_component_p): ...this.
11649 (filter_suitable_components): Renamed to...
11650 (pcom_worker::filter_suitable_components): ...this.
11651 (valid_initializer_p): Renamed to...
11652 (pcom_worker::valid_initializer_p): ...this.
11653 (find_looparound_phi): Renamed to...
11654 (pcom_worker::find_looparound_phi): ...this.
11655 (add_looparound_copies): Renamed to...
11656 (pcom_worker::add_looparound_copies): ...this.
11657 (determine_roots_comp): Renamed to...
11658 (pcom_worker::determine_roots_comp): ...this.
11659 (determine_roots): Renamed to...
11660 (pcom_worker::determine_roots): ...this.
11661 (single_nonlooparound_use): Renamed to...
11662 (pcom_worker::single_nonlooparound_use): ...this.
11663 (remove_stmt): Renamed to...
11664 (pcom_worker::remove_stmt): ...this.
11665 (execute_pred_commoning_chain): Renamed to...
11666 (pcom_worker::execute_pred_commoning_chain): ...this.
11667 (execute_pred_commoning): Renamed to...
11668 (pcom_worker::execute_pred_commoning): ...this.
11669 (struct epcc_data): New member worker.
11670 (execute_pred_commoning_cbck): Call execute_pred_commoning
11671 with pcom_worker pointer.
11672 (find_use_stmt): Renamed to...
11673 (pcom_worker::find_use_stmt): ...this.
11674 (find_associative_operation_root): Renamed to...
11675 (pcom_worker::find_associative_operation_root): ...this.
11676 (find_common_use_stmt): Renamed to...
11677 (pcom_worker::find_common_use_stmt): ...this.
11678 (combinable_refs_p): Renamed to...
11679 (pcom_worker::combinable_refs_p): ...this.
11680 (reassociate_to_the_same_stmt): Renamed to...
11681 (pcom_worker::reassociate_to_the_same_stmt): ...this.
11682 (stmt_combining_refs): Renamed to...
11683 (pcom_worker::stmt_combining_refs): ...this.
11684 (combine_chains): Renamed to...
11685 (pcom_worker::combine_chains): ...this.
11686 (try_combine_chains): Renamed to...
11687 (pcom_worker::try_combine_chains): ...this.
11688 (prepare_initializers_chain): Renamed to...
11689 (pcom_worker::prepare_initializers_chain): ...this.
11690 (prepare_initializers): Renamed to...
11691 (pcom_worker::prepare_initializers): ...this.
11692 (prepare_finalizers_chain): Renamed to...
11693 (pcom_worker::prepare_finalizers_chain): ...this.
11694 (prepare_finalizers): Renamed to...
11695 (pcom_worker::prepare_finalizers): ...this.
11696 (tree_predictive_commoning_loop): Renamed to...
11697 (pcom_worker::tree_predictive_commoning_loop): ...this, adjust
11698 some calls and remove some cleanup code.
11699 (tree_predictive_commoning): Adjusted to use pcom_worker instance.
11700 (static variable looparound_phis): Remove.
11701 (static variable name_expansions): Remove.
11703 2021-06-24 Richard Biener <rguenther@suse.de>
11705 * tree-vect-slp.c (slpg_vertex): New struct.
11706 (vect_slp_build_vertices): Adjust.
11707 (vect_optimize_slp): Likewise. Maintain an outgoing permute
11708 and a materialized one.
11710 2021-06-24 Richard Biener <rguenther@suse.de>
11712 PR tree-optimization/101105
11713 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
11714 Only ignore steps when they are equal or scalar order is preserved.
11716 2021-06-24 liuhongt <hongtao.liu@intel.com>
11719 * config/i386/i386-expand.c (ix86_expand_vec_interleave):
11720 Adjust comments for ix86_expand_vecop_qihi2.
11721 (ix86_expand_vecmul_qihi): Renamed to ..
11722 (ix86_expand_vecop_qihi2): Adjust function prototype to
11723 support shift operation, add static to definition.
11724 (ix86_expand_vec_shift_qihi_constant): Add static to definition.
11725 (ix86_expand_vecop_qihi): Call ix86_expand_vecop_qihi2 and
11726 ix86_expand_vec_shift_qihi_constant.
11727 * config/i386/i386-protos.h (ix86_expand_vecmul_qihi): Deleted.
11728 (ix86_expand_vec_shift_qihi_constant): Deleted.
11729 * config/i386/sse.md (VI12_256_512_AVX512VL): New mode
11731 (mulv8qi3): Call ix86_expand_vecop_qihi directly, add
11732 condition TARGET_64BIT.
11733 (mul<mode>3): Ditto.
11734 (<insn><mode>3): Ditto.
11735 (vlshr<mode>3): Extend to support avx512 vlshr.
11736 (v<insn><mode>3): New expander for
11738 (v<insn>v8qi3): Ditto.
11739 (vashrv8hi3<mask_name>): Renamed to ..
11740 (vashr<mode>3): And extend to support V16QImode for avx512.
11741 (vashrv16qi3): Deleted.
11742 (vashrv2di3<mask_name>): Extend expander to support avx512
11745 2021-06-23 Dimitar Dimitrov <dimitar@dinux.eu>
11747 * doc/lto.texi (Design Overview): Update that slim objects are
11750 2021-06-23 Aaron Sawdey <acsawdey@linux.ibm.com>
11752 * config/rs6000/rs6000-cpus.def: Take OPTION_MASK_PCREL_OPT out
11753 of OTHER_POWER10_MASKS so it will not be enabled by default.
11755 2021-06-23 Richard Biener <rguenther@suse.de>
11756 Martin Jambor <mjambor@suse.cz>
11758 * tree-inline.c (setup_one_parameter): Set TREE_READONLY of the
11759 param replacement unconditionally. Adjust comment.
11761 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11763 * Makefile.in (OBJS): Add gimple-range-fold.o
11764 * gimple-range-fold.cc: New.
11765 * gimple-range-fold.h: New.
11766 * gimple-range-gori.cc (gimple_range_calc_op1): Move to here.
11767 (gimple_range_calc_op2): Ditto.
11768 * gimple-range-gori.h: Move prototypes to here.
11769 * gimple-range.cc: Adjust include files.
11770 (fur_source:fur_source): Relocate to gimple-range-fold.cc.
11771 (fur_source::get_operand): Ditto.
11772 (fur_source::get_phi_operand): Ditto.
11773 (fur_source::query_relation): Ditto.
11774 (fur_source::register_relation): Ditto.
11775 (class fur_edge): Ditto.
11776 (fur_edge::fur_edge): Ditto.
11777 (fur_edge::get_operand): Ditto.
11778 (fur_edge::get_phi_operand): Ditto.
11779 (fur_stmt::fur_stmt): Ditto.
11780 (fur_stmt::get_operand): Ditto.
11781 (fur_stmt::get_phi_operand): Ditto.
11782 (fur_stmt::query_relation): Ditto.
11783 (class fur_depend): Relocate to gimple-range-fold.h.
11784 (fur_depend::fur_depend): Relocate to gimple-range-fold.cc.
11785 (fur_depend::register_relation): Ditto.
11786 (fur_depend::register_relation): Ditto.
11787 (class fur_list): Ditto.
11788 (fur_list::fur_list): Ditto.
11789 (fur_list::get_operand): Ditto.
11790 (fur_list::get_phi_operand): Ditto.
11791 (fold_range): Ditto.
11792 (adjust_pointer_diff_expr): Ditto.
11793 (gimple_range_adjustment): Ditto.
11794 (gimple_range_base_of_assignment): Ditto.
11795 (gimple_range_operand1): Ditto.
11796 (gimple_range_operand2): Ditto.
11797 (gimple_range_calc_op1): Relocate to gimple-range-gori.cc.
11798 (gimple_range_calc_op2): Ditto.
11799 (fold_using_range::fold_stmt): Relocate to gimple-range-fold.cc.
11800 (fold_using_range::range_of_range_op): Ditto.
11801 (fold_using_range::range_of_address): Ditto.
11802 (fold_using_range::range_of_phi): Ditto.
11803 (fold_using_range::range_of_call): Ditto.
11804 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
11805 (fold_using_range::range_of_builtin_call): Ditto.
11806 (fold_using_range::range_of_cond_expr): Ditto.
11807 (fold_using_range::range_of_ssa_name_with_loop_info): Ditto.
11808 (fold_using_range::relation_fold_and_or): Ditto.
11809 (fold_using_range::postfold_gcond_edges): Ditto.
11810 * gimple-range.h: Add gimple-range-fold.h to include files. Change
11811 GIMPLE_RANGE_STMT_H to GIMPLE_RANGE_H.
11812 (gimple_range_handler): Relocate to gimple-range-fold.h.
11813 (gimple_range_ssa_p): Ditto.
11814 (range_compatible_p): Ditto.
11815 (class fur_source): Ditto.
11816 (class fur_stmt): Ditto.
11817 (class fold_using_range): Ditto.
11818 (gimple_range_calc_op1): Relocate to gimple-range-gori.h
11819 (gimple_range_calc_op2): Ditto.
11821 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11823 PR tree-optimization/101148
11824 PR tree-optimization/101014
11825 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
11826 (ranger_cache::~ranger_cache): Adjust.
11827 (ranger_cache::block_range): Check if propagation disallowed.
11828 (ranger_cache::propagate_cache): Disallow propagation if new value
11829 can't be stored properly.
11830 * gimple-range-cache.h (ranger_cache::m_propfail): New member.
11832 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11834 * gimple-range-cache.cc (class ssa_block_ranges): Adjust prototype.
11835 (sbr_vector::set_bb_range): Return true.
11836 (class sbr_sparse_bitmap): Adjust.
11837 (sbr_sparse_bitmap::set_bb_range): Return value.
11838 (block_range_cache::set_bb_range): Return value.
11839 (ranger_cache::propagate_cache): Use return value to print msg.
11840 * gimple-range-cache.h (class block_range_cache): Adjust.
11842 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11844 * gimple-range.cc (dump_bb): Use range_on_edge from the cache.
11846 2021-06-23 Jeff Law <jeffreyalaw@gmail.com>
11848 * config/h8300/logical.md (<code><mode>3<ccnz>): Use <cczn>
11849 so this pattern can be used for test/compare removal. Pass
11850 current insn to compute_logical_op_length and output_logical_op.
11851 * config/h8300/h8300.c (compute_logical_op_cc): Remove.
11852 (h8300_and_costs): Add argument to compute_logical_op_length.
11853 (output_logical_op): Add new argument. Use it to determine if the
11854 condition codes are used and adjust the output accordingly.
11855 (compute_logical_op_length): Add new argument and update length
11856 computations when condition codes are used.
11857 * config/h8300/h8300-protos.h (compute_logical_op_length): Update
11859 (output_logical_op): Likewise.
11861 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
11864 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
11865 Handle 64bit modes for TARGET_XOP. Use indirect gen_* functions.
11866 * config/i386/mmx.md (mmx_ppermv64): New insn pattern.
11867 * config/i386/i386.md (unspec): Move UNSPEC_XOP_PERMUTE from ...
11868 * config/i386/sse.md (unspec): ... here.
11870 2021-06-23 Martin Liska <mliska@suse.cz>
11873 * optc-save-gen.awk: Put back arm_fp16_format to
11876 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
11879 * config/i386/i386.md (bsr_rex64): Add zero-flag setting RTX.
11882 (clz<mode>2): Update RTX pattern for additions.
11884 2021-06-23 Jakub Jelinek <jakub@redhat.com>
11886 PR middle-end/101167
11887 * omp-low.c (lower_omp_regimplify_p): Regimplify also PARM_DECLs
11888 and RESULT_DECLs that have DECL_HAS_VALUE_EXPR_P set.
11890 2021-06-22 Sergei Trofimovich <siarheit@google.com>
11892 * doc/rtl.texi: drop unbalanced parenthesis.
11894 2021-06-22 Richard Biener <rguenther@suse.de>
11896 PR middle-end/101156
11897 * gimplify.c (gimplify_expr): Remove premature incorrect
11900 2021-06-22 Jakub Jelinek <jakub@redhat.com>
11902 PR tree-optimization/101159
11903 * tree-vect-patterns.c (vect_recog_popcount_pattern): Fix some
11906 2021-06-22 Jakub Jelinek <jakub@redhat.com>
11908 PR middle-end/101160
11909 * function.c (assign_parms): For decl_result with TYPE_EMPTY_P type
11910 clear crtl->return_rtx instead of keeping it referencing a pseudo.
11912 2021-06-22 Jakub Jelinek <jakub@redhat.com>
11913 Andrew Pinski <apinski@marvell.com>
11915 PR tree-optimization/101162
11916 * fold-const.c (range_check_type): Handle OFFSET_TYPE like pointer
11919 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11921 * range-op.cc (range_relational_tests): New.
11922 (range_op_tests): Call range_relational_tests.
11924 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11926 * range-op.cc (operator_cast::lhs_op1_relation): New.
11927 (operator_identity::lhs_op1_relation): Mew.
11929 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11931 * range-op.cc (operator_minus::op1_op2_relation_effect): New.
11933 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11935 * range-op.cc (operator_plus::lhs_op1_relation): New.
11936 (operator_plus::lhs_op2_relation): New.
11938 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11940 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a
11941 relation_oracle if dominators exist.
11942 (ranger_cache::~ranger_cache): Dispose of oracle.
11943 (ranger_cache::dump_bb): Dump oracle.
11944 * gimple-range.cc (fur_source::fur_source): New.
11945 (fur_source::get_operand): Use mmeber query.
11946 (fur_source::get_phi_operand): Use member_query.
11947 (fur_source::query_relation): New.
11948 (fur_source::register_dependency): Delete.
11949 (fur_source::register_relation): New.
11950 (fur_edge::fur_edge): Adjust.
11951 (fur_edge::get_phi_operand): Fix comment.
11952 (fur_edge::query): Delete.
11953 (fur_stmt::fur_stmt): Adjust.
11954 (fur_stmt::query): Delete.
11955 (fur_depend::fur_depend): Adjust.
11956 (fur_depend::register_relation): New.
11957 (fur_depend::register_relation): New.
11958 (fur_list::fur_list): Adjust.
11959 (fur_list::get_operand): Use member query.
11960 (fold_using_range::range_of_range_op): Process and query relations.
11961 (fold_using_range::range_of_address): Adjust dependency call.
11962 (fold_using_range::range_of_phi): Ditto.
11963 (gimple_ranger::gimple_ranger): New. Use ranger_ache oracle.
11964 (fold_using_range::relation_fold_and_or): New.
11965 (fold_using_range::postfold_gcond_edges): New.
11966 * gimple-range.h (class gimple_ranger): Adjust.
11967 (class fur_source): Adjust members.
11968 (class fur_stmt): Ditto.
11969 (class fold_using_range): Ditto.
11971 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11973 * range-op.cc (range_operator::wi_fold): Apply relation effect.
11974 (range_operator::fold_range): Adjust and apply relation effect.
11975 (*::fold_range): Add relation parameters.
11976 (*::op1_range): Ditto.
11977 (*::op2_range): Ditto.
11978 (range_operator::lhs_op1_relation): New.
11979 (range_operator::lhs_op2_relation): New.
11980 (range_operator::op1_op2_relation): New.
11981 (range_operator::op1_op2_relation_effect): New.
11982 (relop_early_resolve): New.
11983 (operator_equal::op1_op2_relation): New.
11984 (operator_equal::fold_range): Call relop_early_resolve.
11985 (operator_not_equal::op1_op2_relation): New.
11986 (operator_not_equal::fold_range): Call relop_early_resolve.
11987 (operator_lt::op1_op2_relation): New.
11988 (operator_lt::fold_range): Call relop_early_resolve.
11989 (operator_le::op1_op2_relation): New.
11990 (operator_le::fold_range): Call relop_early_resolve.
11991 (operator_gt::op1_op2_relation): New.
11992 (operator_gt::fold_range): Call relop_early_resolve.
11993 (operator_ge::op1_op2_relation): New.
11994 (operator_ge::fold_range): Call relop_early_resolve.
11995 * range-op.h (class range_operator): Adjust parameters and methods.
11997 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
11999 * Makefile.in (OBJS): Add value-relation.o.
12000 * gimple-range.h: Adjust include files.
12001 * tree-data-ref.c: Adjust include file order.
12002 * value-query.cc (range_query::get_value_range): Default to no oracle.
12003 (range_query::query_relation): New.
12004 (range_query::query_relation): New.
12005 * value-query.h (class range_query): Adjust.
12006 * value-relation.cc: New.
12007 * value-relation.h: New.
12009 2021-06-22 Richard Biener <rguenther@suse.de>
12011 PR tree-optimization/101151
12012 * tree-ssa-sink.c (statement_sink_location): Expand irreducible
12015 2021-06-22 Jojo R <rjiejie@linux.alibaba.com>
12017 * config/riscv/riscv.c (thead_c906_tune_info): New.
12018 (riscv_tune_info_table): Use new tune.
12020 2021-06-22 Richard Biener <rguenther@suse.de>
12022 PR tree-optimization/101158
12023 * tree-vect-slp.c (vect_build_slp_tree_1): Move same operand
12024 checking after checking for matching operation.
12026 2021-06-22 Richard Biener <rguenther@suse.de>
12028 PR tree-optimization/101159
12029 * tree-vect-patterns.c (vect_recog_popcount_pattern): Add
12030 missing NULL vectype check.
12032 2021-06-22 Richard Biener <rguenther@suse.de>
12034 PR tree-optimization/101154
12035 * tree-vect-slp.c (vect_build_slp_tree_2): Fix out-of-bound access.
12037 2021-06-22 Jakub Jelinek <jakub@redhat.com>
12040 * config/i386/i386-protos.h (ix86_last_zero_store_uid): Declare.
12041 * config/i386/i386-expand.c (ix86_last_zero_store_uid): New variable.
12042 * config/i386/i386.c (ix86_expand_prologue): Clear it.
12043 * config/i386/i386.md (peephole2s for 1/2/4 stores of const0_rtx):
12044 Remove "" from match_operand. Emit new insns using emit_move_insn and
12045 set ix86_last_zero_store_uid to INSN_UID of the last store.
12046 Add peephole2s for 1/2/4 stores of const0_rtx following previous
12049 2021-06-22 Martin Liska <mliska@suse.cz>
12051 * auto-profile.c (AUTO_PROFILE_VERSION): Bump as string format
12054 2021-06-22 Martin Liska <mliska@suse.cz>
12056 * gcov-io.h: Remove padding entries.
12058 2021-06-22 liuhongt <hongtao.liu@intel.com>
12060 PR tree-optimization/97770
12061 * tree-vect-patterns.c (vect_recog_popcount_pattern):
12063 (vect_recog_func vect_vect_recog_func_ptrs): Add new pattern.
12065 2021-06-22 liuhongt <hongtao.liu@intel.com>
12068 * config/i386/i386-builtin.def (BDESC): Adjust builtin name.
12069 * config/i386/sse.md (<avx512>_expand<mode>_mask): Rename to ..
12070 (expand<mode>_mask): this ..
12071 (*expand<mode>_mask): New pre_reload splitter to transform
12072 v{,p}expand* to vmov* when mask is zero, all ones, or has all
12073 ones in it's lower part, otherwise still generate
12076 2021-06-22 liuhongt <hongtao.liu@intel.com>
12079 * config/i386/i386-expand.c
12080 (ix86_expand_special_args_builtin): Keep constm1_operand only
12081 if it satisfies insn's operand predicate.
12083 2021-06-21 Jason Merrill <jason@redhat.com>
12086 * df-scan.c (df_ref_record): Check that regno < endregno.
12087 * function.c (assign_parms, expand_function_end): Do nothing with a
12088 TYPE_EMPTY_P result.
12090 2021-06-21 Richard Biener <rguenther@suse.de>
12092 PR tree-optimization/101120
12093 * tree-vect-data-refs.c (bump_vector_ptr): Fold the
12095 * tree-vect-slp.c (vect_transform_slp_perm_load): Add
12096 DR chain DCE capability.
12097 * tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
12098 * tree-vect-stmts.c (vectorizable_load): Remove unused
12099 loads in the DR chain for SLP.
12101 2021-06-21 Jakub Jelinek <jakub@redhat.com>
12103 PR inline-asm/100785
12104 * gimplify.c (gimplify_asm_expr): Don't diagnose errors if
12105 output or input operands were already error_mark_node.
12106 * cfgexpand.c (expand_asm_stmt): If errors are emitted,
12107 remove all inputs, outputs and clobbers from the asm and
12108 set template to "".
12110 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
12112 * config/arm/arm_neon.h (vceq_s8): Replace builtin with __a == __b.
12113 (vceq_s16): Likewise.
12114 (vceq_s32): Likewise.
12115 (vceq_u8): Likewise.
12116 (vceq_u16): Likewise.
12117 (vceq_u32): Likewise.
12118 (vceq_p8): Likewise.
12119 (vceqq_s8): Likewise.
12120 (vceqq_s16): Likewise.
12121 (vceqq_s32): Likewise.
12122 (vceqq_u8): Likewise.
12123 (vceqq_u16): Likewise.
12124 (vceqq_u32): Likewise.
12125 (vceqq_p8): Likewise.
12126 (vceq_f32): Gate __a == __b on __FAST_MATH__.
12127 (vceqq_f32): Likewise.
12128 (vceq_f16): Likewise.
12129 (vceqq_f16): Likewise.
12131 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
12134 * config/arm/iterators.md (NEON_VACMP): Remove.
12135 * config/arm/neon.md (neon_vca<cmp_op><mode>): Use GLTE instead of GTGE
12137 (neon_vca<cmp_op><mode>_insn): Likewise.
12138 (neon_vca<cmp_op_unsp><mode>_insn_unspec): Use NEON_VAGLTE instead of
12141 2021-06-21 Richard Biener <rguenther@suse.de>
12143 PR tree-optimization/101121
12144 * tree-vect-slp.c (vect_build_slp_tree_2): To not fail fatally
12145 when we just lack a stmt with the desired op when doing permutation.
12146 (vect_build_slp_tree): When caching a failed SLP build attempt
12147 assert that at least one lane is marked as not matching.
12149 2021-06-21 liuhongt <hongtao.liu@intel.com>
12152 * config/i386/i386.md: (*anddi_1): Disparage slightly the mask
12153 register alternative.
12154 (*and<mode>_1): Ditto.
12156 (*andn<mode>_1): Ditto.
12157 (*<code><mode>_1): Ditto.
12158 (*<code>qi_1): Ditto.
12159 (*one_cmpl<mode>2_1): Ditto.
12160 (*one_cmplsi2_1_zext): Ditto.
12161 (*one_cmplqi2_1): Ditto.
12162 * config/i386/i386.c (x86_order_regs_for_local_alloc): Change
12163 the order of mask registers to be before general registers.
12165 2021-06-21 Roger Sayle <roger@nextmovesoftware.com>
12168 * config/i386/i386.md: New define_peephole2s to shrink writing
12169 1, 2 or 4 consecutive zeros to memory when optimizing for size.
12171 2021-06-18 Jeff Law <jeffreyalaw@gmail.com>
12173 * config/h8300/h8300.c (h8300_select_cc_mode): Handle SYMBOL_REF.
12174 * config/h8300/logical.md (<code><mode>3 logcial expander): Generate
12175 more efficient code when the source can be trivially simplified.
12177 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
12179 * gimple-range-cache.cc (ranger_cache::range_of_def): Calculate
12180 a range if global is not available.
12181 (ranger_cache::entry_range): Fallback to range_of_def.
12182 * gimple-range-cache.h (range_of_def): Adjust prototype.
12184 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
12186 PR tree-optimization/101014
12187 * gimple-range-cache.cc (ranger_cache::ranger_cache): Remove poor
12189 (ranger_cache::~ranger_cache): Ditto.
12190 (ranger_cache::enable_new_values): Delete.
12191 (ranger_cache::push_poor_value): Delete.
12192 (ranger_cache::range_of_def): Remove poor value processing.
12193 (ranger_cache::entry_range): Ditto.
12194 (ranger_cache::fill_block_cache): Ditto.
12195 * gimple-range-cache.h (class ranger_cache): Remove poor value members.
12196 * gimple-range.cc (gimple_ranger::range_of_expr): Remove call.
12197 * gimple-range.h (class gimple_ranger): Adjust.
12199 2021-06-18 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
12202 * common/config/arm/arm-common.c (arm_canon_arch_option_1): New function
12203 derived from arm_canon_arch.
12204 (arm_canon_arch_option): Call it.
12205 (arm_canon_arch_multilib_option): New function.
12206 * config/arm/arm-cpus.in (IGNORE_FOR_MULTILIB): New fgroup.
12207 * config/arm/arm.h (arm_canon_arch_multilib_option): New prototype.
12208 (CANON_ARCH_MULTILIB_SPEC_FUNCTION): New macro.
12209 (MULTILIB_ARCH_CANONICAL_SPECS): New macro.
12210 (DRIVER_SELF_SPECS): Add MULTILIB_ARCH_CANONICAL_SPECS.
12211 * config/arm/arm.opt (mlibarch): New option.
12212 * config/arm/t-rmprofile (MULTILIB_MATCHES): For armv8*-m, replace use
12213 of march on RHS with mlibarch.
12215 2021-06-18 Marcel Vollweiler <marcel@codesourcery.com>
12217 * config.in: Regenerate.
12218 * config/gcn/gcn.c (print_operand_address): Fix for global_load assembler
12220 * configure: Regenerate.
12221 * configure.ac: Fix for global_load assembler functions.
12223 2021-06-18 Richard Biener <rguenther@suse.de>
12225 PR tree-optimization/101112
12226 * tree-vect-slp.c (vect_slp_linearize_chain): Fix condition
12227 to lookup a pattern stmt def.
12229 2021-06-18 Jakub Jelinek <jakub@redhat.com>
12231 PR middle-end/101062
12232 * stor-layout.c (finish_bitfield_layout): Don't add bitfield
12233 representatives in QUAL_UNION_TYPE.
12235 2021-06-18 Andrew Pinski <apinski@marvell.com>
12237 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
12238 Add counting of how many times it is done.
12239 (factor_out_conditional_conversion): Likewise.
12240 (match_simplify_replacement): Likewise.
12241 (value_replacement): Likewise.
12242 (spaceship_replacement): Likewise.
12243 (cond_store_replacement): Likewise.
12244 (cond_if_else_store_replacement_1): Likewise.
12245 (hoist_adjacent_loads): Likewise.
12247 2021-06-18 Andrew Pinski <apinski@marvell.com>
12249 * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
12250 types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and CNONJ_EXPR.
12251 (verify_gimple_assign_binary): Reject point and offset types on
12252 MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
12253 FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
12254 FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.
12256 2021-06-18 Michael Meissner <meissner@linux.ibm.com>
12258 * config/rs6000/rs6000.c (rs6000_emit_minmax): Add support for ISA
12259 3.1 IEEE 128-bit floating point xsmaxcqp/xsmincqp instructions.
12260 * config/rs6000/rs6000.md (s<minmax><mode>3, IEEE128 iterator):
12263 2021-06-17 Aaron Sawdey <acsawdey@linux.ibm.com>
12265 * config/rs6000/genfusion.pl (gen_logical_addsubf): Add
12266 earlyclobber to alts 0/1.
12267 (gen_addadd): Add earlyclobber to alts 0/1.
12268 * config/rs6000/fusion.md: Regenerate file.
12270 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12272 * cfgloopanal.c (get_loop_hot_path): Make path an auto_vec.
12274 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
12276 * gimple-range-cache.cc: Comment cleanups.
12277 * gimple-range-gori.cc: Comment cleanups.
12278 * gimple-range.cc: Comment/spacing cleanups
12279 * value-range.h: Comment cleanups.
12281 2021-06-17 H.J. Lu <hjl.tools@gmail.com>
12284 * calls.c (expand_call): Replace PUSH_ARGS with
12285 targetm.calls.push_argument (0).
12286 (emit_library_call_value_1): Likewise.
12287 * defaults.h (PUSH_ARGS): Removed.
12288 (PUSH_ARGS_REVERSED): Replace PUSH_ARGS with
12289 targetm.calls.push_argument (0).
12290 * expr.c (block_move_libcall_safe_for_call_parm): Likewise.
12291 (emit_push_insn): Pass the number bytes to push to
12292 targetm.calls.push_argument and pass 0 if ARGS_ADDR is 0.
12293 * hooks.c (hook_bool_uint_true): New.
12294 * hooks.h (hook_bool_uint_true): Likewise.
12295 * rtlanal.c (nonzero_bits1): Replace PUSH_ARGS with
12296 targetm.calls.push_argument (0).
12297 * target.def (push_argument): Add a targetm.calls hook.
12298 * targhooks.c (default_push_argument): New.
12299 * targhooks.h (default_push_argument): Likewise.
12300 * config/bpf/bpf.h (PUSH_ARGS): Removed.
12301 * config/cr16/cr16.c (TARGET_PUSH_ARGUMENT): New.
12302 * config/cr16/cr16.h (PUSH_ARGS): Removed.
12303 * config/i386/i386.c (ix86_push_argument): New.
12304 (TARGET_PUSH_ARGUMENT): Likewise.
12305 * config/i386/i386.h (PUSH_ARGS): Removed.
12306 * config/m32c/m32c.c (TARGET_PUSH_ARGUMENT): New.
12307 * config/m32c/m32c.h (PUSH_ARGS): Removed.
12308 * config/nios2/nios2.h (PUSH_ARGS): Likewise.
12309 * config/pru/pru.h (PUSH_ARGS): Likewise.
12310 * doc/tm.texi.in: Remove PUSH_ARGS documentation. Add
12311 TARGET_PUSH_ARGUMENT hook.
12312 * doc/tm.texi: Regenerated.
12314 2021-06-17 Uroš Bizjak <ubizjak@gmail.com>
12317 * config/i386/i386-expand.c (expand_vector_set_var):
12318 Handle V2FS mode remapping. Pass TARGET_MMX_WITH_SSE to
12319 ix86_expand_vector_init_duplicate.
12320 (ix86_expand_vector_init_duplicate): Emit insv_1 for
12321 QImode for !TARGET_PARTIAL_REG_STALL.
12322 * config/i386/predicates.md (vec_setm_mmx_operand): New predicate.
12323 * config/i386/mmx.md (vec_setv2sf): Use vec_setm_mmx_operand
12324 as operand 2 predicate. Call ix86_expand_vector_set_var
12325 for non-constant index operand.
12326 (vec_setv2si): Ditto.
12327 (vec_setv4hi): Ditto.
12328 (vec_setv8qi): ditto.
12330 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
12332 PR tree-optimization/100790
12333 * gimple-range.cc (range_of_builtin_call): Cleanup clz and ctz
12336 2021-06-17 Martin Liska <mliska@suse.cz>
12338 * doc/invoke.texi: Use consistently -O1 instead of -O.
12340 2021-06-17 Martin Liska <mliska@suse.cz>
12342 * gcov-io.h: Update documentation entry about string format.
12344 2021-06-17 Marius Hillenbrand <mhillen@linux.ibm.com>
12347 * config/s390/vecintrin.h (vec_doublee): Fix to use
12348 __builtin_s390_vflls.
12349 (vec_floate): Fix to use __builtin_s390_vflrd.
12351 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12353 * dominance.c (get_dominated_to_depth): Return auto_vec<basic_block>.
12354 * dominance.h (get_dominated_to_depth): Likewise.
12355 (get_all_dominated_blocks): Likewise.
12356 * cfgcleanup.c (delete_unreachable_blocks): Adjust.
12357 * gcse.c (hoist_code): Likewise.
12358 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
12359 * tree-parloops.c (oacc_entry_exit_ok): Likewise.
12360 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Likewise.
12361 * tree-ssa-phiprop.c (pass_phiprop::execute): Likewise.
12363 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12365 * dominance.c (get_dominated_by_region): Return auto_vec<basic_block>.
12366 * dominance.h (get_dominated_by_region): Likewise.
12367 * tree-cfg.c (gimple_duplicate_sese_region): Adjust.
12368 (gimple_duplicate_sese_tail): Likewise.
12369 (move_sese_region_to_fn): Likewise.
12371 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12373 * dominance.c (get_dominated_by): Return auto_vec<basic_block>.
12374 * dominance.h (get_dominated_by): Likewise.
12375 * auto-profile.c (afdo_find_equiv_class): Adjust.
12376 * cfgloopmanip.c (duplicate_loop_to_header_edge): Likewise.
12377 * loop-unroll.c (unroll_loop_runtime_iterations): Likewise.
12378 * tree-cfg.c (test_linear_chain): Likewise.
12379 (test_diamond): Likewise.
12381 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12383 * cfgloop.h (get_loop_hot_path): Return auto_vec<basic_block>.
12384 * cfgloopanal.c (get_loop_hot_path): Likewise.
12385 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
12387 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12389 * cgraph.c (cgraph_node::collect_callers): Return
12390 auto_vec<cgraph_edge *>.
12391 * cgraph.h (cgraph_node::collect_callers): Likewise.
12392 * ipa-cp.c (create_specialized_node): Adjust.
12393 (decide_about_value): Likewise.
12394 (decide_whether_version_node): Likewise.
12395 * ipa-sra.c (process_isra_node_results): Likewise.
12397 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12399 * vec.h (vl_ptr>::using_auto_storage): Handle null m_vec.
12400 (auto_vec<T, 0>::auto_vec): Define move constructor, and delete copy
12402 (auto_vec<T, 0>::operator=): Define move assignment and delete copy
12405 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
12407 * gimple-range.cc (debug_seed_ranger): New.
12408 (dump_ranger): New.
12409 (debug_ranger): New.
12411 2021-06-17 Richard Biener <rguenther@suse.de>
12413 PR tree-optimization/54400
12414 * tree-vectorizer.h (enum slp_instance_kind): Add
12415 slp_inst_kind_bb_reduc.
12416 (reduction_fn_for_scalar_code): Declare.
12417 * tree-vect-data-refs.c (vect_slp_analyze_instance_dependence):
12418 Check SLP_INSTANCE_KIND instead of looking at the
12420 (vect_slp_analyze_instance_alignment): Likewise.
12421 * tree-vect-loop.c (reduction_fn_for_scalar_code): Export.
12422 * tree-vect-slp.c (vect_slp_linearize_chain): Split out
12423 chain linearization from vect_build_slp_tree_2 and generalize
12424 for the use of BB reduction vectorization.
12425 (vect_build_slp_tree_2): Adjust accordingly.
12426 (vect_optimize_slp): Elide permutes at the root of BB reduction
12428 (vectorizable_bb_reduc_epilogue): New function.
12429 (vect_slp_prune_covered_roots): Likewise.
12430 (vect_slp_analyze_operations): Use them.
12431 (vect_slp_check_for_constructors): Recognize associatable
12432 chains for BB reduction vectorization.
12433 (vectorize_slp_instance_root_stmt): Generate code for the
12434 BB reduction epilogue.
12436 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
12438 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Check with
12440 (gori_compute::may_recompute_p): New.
12441 (gori_compute::outgoing_edge_range_p): Perform recomputations.
12442 * gimple-range-gori.h (class gori_compute): Add prototype.
12444 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
12446 * gimple-range-cache.cc (ranger_cache::range_on_edge): Always return
12447 true when a range can be calculated.
12448 * gimple-range.cc (gimple_ranger::dump_bb): Check has_edge_range_p.
12450 2021-06-16 Martin Sebor <msebor@redhat.com>
12452 * doc/invoke.texi (-Wmismatched-dealloc, -Wmismatched-new-delete):
12453 Correct documented defaults.
12455 2021-06-16 Andrew MacLeod <amacleod@redhat.com>
12457 * gimple-range-cache.cc (ranger_cache::ranger_cache): Initialize
12458 m_new_value_p directly.
12460 2021-06-16 Uroš Bizjak <ubizjak@gmail.com>
12463 * config/i386/i386-expand.c (expand_vec_perm_2perm_pblendv):
12464 Handle 64bit modes for TARGET_SSE4_1.
12465 (expand_vec_perm_pshufb2): Handle 64bit modes for TARGET_SSSE3.
12466 (expand_vec_perm_even_odd_pack): Handle V4HI mode.
12467 (expand_vec_perm_even_odd_1) <case E_V4HImode>: Expand via
12468 expand_vec_perm_pshufb2 for TARGET_SSSE3 and via
12469 expand_vec_perm_even_odd_pack for TARGET_SSE4_1.
12470 * config/i386/mmx.md (mmx_packusdw): New insn pattern.
12472 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12474 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn<mode>):
12475 Change to an expander that emits the correct instruction
12476 depending on endianness.
12477 (aarch64_<sur><addsub>hn<mode>_insn_le): Define.
12478 (aarch64_<sur><addsub>hn<mode>_insn_be): Define.
12480 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12482 * config/aarch64/aarch64-simd-builtins.def: Split generator
12483 for aarch64_<su>qmovn builtins into scalar and vector
12485 * config/aarch64/aarch64-simd.md (aarch64_<su>qmovn<mode>_insn_le):
12487 (aarch64_<su>qmovn<mode>_insn_be): Define.
12488 (aarch64_<su>qmovn<mode>): Split into scalar and vector
12489 variants. Change vector variant to an expander that emits the
12490 correct instruction depending on endianness.
12492 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12494 * config/aarch64/aarch64-simd-builtins.def: Split generator
12495 for aarch64_sqmovun builtins into scalar and vector variants.
12496 * config/aarch64/aarch64-simd.md (aarch64_sqmovun<mode>):
12497 Split into scalar and vector variants. Change vector variant
12498 to an expander that emits the correct instruction depending
12500 (aarch64_sqmovun<mode>_insn_le): Define.
12501 (aarch64_sqmovun<mode>_insn_be): Define.
12503 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12505 * config/aarch64/aarch64-simd.md (aarch64_xtn<mode>_insn_le):
12506 Define - modeling zero-high-half semantics.
12507 (aarch64_xtn<mode>): Change to an expander that emits the
12508 appropriate instruction depending on endianness.
12509 (aarch64_xtn<mode>_insn_be): Define - modeling zero-high-half
12511 (aarch64_xtn2<mode>_le): Rename to...
12512 (aarch64_xtn2<mode>_insn_le): This.
12513 (aarch64_xtn2<mode>_be): Rename to...
12514 (aarch64_xtn2<mode>_insn_be): This.
12515 (vec_pack_trunc_<mode>): Emit truncation instruction instead
12517 * config/aarch64/iterators.md (Vnarrowd): Add Vnarrowd mode
12518 attribute iterator.
12520 2021-06-16 Martin Jambor <mjambor@suse.cz>
12522 PR tree-optimization/100453
12523 * tree-sra.c (create_access): Disqualify any const candidates
12524 which are written to.
12525 (sra_modify_expr): Do not store sub-replacements back to a const base.
12526 (handle_unscalarized_data_in_subtree): Likewise.
12527 (sra_modify_assign): Likewise. Earlier, use TREE_READONLy test
12528 instead of constant_decl_p.
12530 2021-06-16 Jakub Jelinek <jakub@redhat.com>
12532 PR middle-end/101062
12533 * stor-layout.c (finish_bitfield_representative): For fields in unions
12534 assume nextf is always NULL.
12535 (finish_bitfield_layout): Compute bit field representatives also in
12536 unions, but handle it as if each bitfield was the only field in the
12539 2021-06-16 Richard Biener <rguenther@suse.de>
12541 PR tree-optimization/101088
12542 * tree-ssa-loop-im.c (sm_seq_valid_bb): Only look for
12543 supported refs on edges. Do not assert same ref but
12544 different kind stores are unsuported but mark them so.
12545 (hoist_memory_references): Only look for supported refs
12548 2021-06-16 Roger Sayle <roger@nextmovesoftware.com>
12550 PR rtl-optimization/46235
12551 * config/i386/i386.md: New define_split for bt followed by cmov.
12552 (*bt<mode>_setcqi): New define_insn_and_split for bt followed by setc.
12553 (*bt<mode>_setncqi): New define_insn_and_split for bt then setnc.
12554 (*bt<mode>_setnc<mode>): New define_insn_and_split for bt followed
12555 by setnc with zero extension.
12557 2021-06-16 Richard Biener <rguenther@suse.de>
12559 PR tree-optimization/101083
12560 * tree-vect-slp.c (vect_slp_build_two_operator_nodes): Get
12561 vectype as argument.
12562 (vect_build_slp_tree_2): Adjust.
12564 2021-06-15 Martin Sebor <msebor@redhat.com>
12566 PR middle-end/100876
12567 * builtins.c: (gimple_call_return_array): Account for size_t
12568 mangling as either unsigned int or unsigned long
12570 2021-06-15 Jeff Law <jeffreyalaw@gmail.com>
12572 * compare-elim.c (try_eliminate_compare): Run DCE to clean things
12573 up before eliminating comparisons.
12575 2021-06-15 Aldy Hernandez <aldyh@redhat.com>
12577 * range-op.cc (operator_bitwise_or::wi_fold): Make sure
12578 nonzero|X is nonzero.
12579 (range_op_bitwise_and_tests): Add tests for above.
12581 2021-06-15 Carl Love <cel@us.ibm.com>
12584 * config/rs6000/rs6000-builtin.def (VCMPEQUT): Fix the ICODE for the
12586 (VRLQ, VSLQ, VSRQ, VSRAQ): Remove unused BU_P10_OVERLOAD_2
12589 2021-06-15 Tobias Burnus <tobias@codesourcery.com>
12592 * gimplify.c (enum gimplify_defaultmap_kind): Add GDMK_SCALAR_TARGET.
12593 (struct gimplify_omp_ctx): Extend defaultmap array by one.
12594 (new_omp_context): Init defaultmap[GDMK_SCALAR_TARGET].
12595 (omp_notice_variable): Update type classification for Fortran.
12596 (gimplify_scan_omp_clauses): Update calls for new argument; handle
12597 GDMK_SCALAR_TARGET; for Fortran, GDMK_POINTER avoid GOVD_MAP_0LEN_ARRAY.
12598 * langhooks-def.h (lhd_omp_scalar_p): Add 'ptr_ok' argument.
12599 * langhooks.c (lhd_omp_scalar_p): Likewise.
12600 (LANG_HOOKS_OMP_ALLOCATABLE_P, LANG_HOOKS_OMP_SCALAR_TARGET_P): New.
12601 (LANG_HOOKS_DECLS): Add them.
12602 * langhooks.h (struct lang_hooks_for_decls): Add new hooks, update
12603 omp_scalar_p pointer type to include the new bool argument.
12605 2021-06-15 David Malcolm <dmalcolm@redhat.com>
12607 * doc/analyzer.texi
12608 (Special Functions for Debugging the Analyzer): Add
12609 __analyzer_dump_capacity.
12611 2021-06-15 Jakub Jelinek <jakub@redhat.com>
12614 * expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
12615 case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.
12617 2021-06-15 Richard Biener <rguenther@suse.de>
12619 * cfgloopanal.c (mark_irreducible_loops): Use a dominance
12620 check to identify loop latches.
12621 * cfgloop.c (verify_loop_structure): Likewise.
12622 * loop-init.c (apply_loop_flags): Allow marked irreducible
12623 regions even with multiple latches.
12624 * predict.c (rebuild_frequencies): Simplify.
12626 2021-06-15 Richard Biener <rguenther@suse.de>
12628 * tree-ssa-threadupdate.c
12629 (jump_thread_path_registry::mark_threaded_blocks): Assert we
12630 have marked irreducible regions.
12632 2021-06-14 Martin Sebor <msebor@redhat.com>
12635 * builtins.c (gimple_call_return_array): Check for attribute fn spec.
12636 Handle calls to placement new.
12637 (ndecl_dealloc_argno): Avoid placement delete.
12639 2021-06-14 Peter Bergner <bergner@linux.ibm.com>
12642 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
12643 create_tmp_reg_or_ssa_name().
12645 2021-06-14 Andrew MacLeod <amacleod@redhat.com>
12647 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
12648 (ranger_cache::enable_new_values): Set to specified value and
12649 return the old value.
12650 (ranger_cache::disable_new_values): Delete.
12651 (ranger_cache::fill_block_cache): Disable non 1st order derived
12653 * gimple-range-cache.h (ranger_cache): Adjust prototypes.
12654 * gimple-range.cc (gimple_ranger::range_of_expr): Adjust.
12656 2021-06-14 Uroš Bizjak <ubizjak@gmail.com>
12659 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
12660 Return true early when testing with V2HImode.
12661 * config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.
12663 2021-06-14 Christophe Lyon <christophe.lyon@linaro.org>
12665 * config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
12666 (mve_vec_unpack<US>_hi_<mode>): New pattern.
12667 (@mve_vec_pack_trunc_lo_<mode>): New pattern.
12668 (mve_vmovntq_<supf><mode>): Prefix with '@'.
12669 * config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
12671 (vec_unpack<US>_lo_<mode>): Likewise.
12672 (vec_pack_trunc_<mode>): Rename to
12673 neon_quad_vec_pack_trunc_<mode>.
12674 * config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
12676 (vec_unpack<US>_lo_<mode>): New.
12677 (vec_pack_trunc_<mode>): New.
12679 2021-06-14 Richard Biener <rguenther@suse.de>
12681 PR tree-optimization/100934
12682 * tree-ssa-dom.c (pass_dominator::execute): Properly
12683 mark irreducible regions.
12685 2021-06-14 Martin Liska <mliska@suse.cz>
12687 * doc/invoke.texi: Put r{...} on the same line as @item.
12689 2021-06-14 Martin Liska <mliska@suse.cz>
12691 * doc/invoke.texi: Add missing newline.
12693 2021-06-14 Martin Liska <mliska@suse.cz>
12695 * doc/invoke.texi: Remove '+' charasters.
12697 2021-06-14 Claudiu Zissulescu <claziss@synopsys.com>
12699 * config.gcc (arc): Add support for with_cpu option.
12700 * config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.
12702 2021-06-14 Richard Biener <rguenther@suse.de>
12704 PR tree-optimization/101031
12705 * tree-ssa-strlen.c (maybe_invalidate): Increment max_size
12706 instead of size when accounting for a possibly string
12709 2021-06-14 Martin Liska <mliska@suse.cz>
12711 * gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].
12713 2021-06-14 Aldy Hernandez <aldyh@redhat.com>
12715 * value-query.cc (gimple_range_global): Call get_range_global
12716 if called after inlining.
12718 2021-06-13 Uroš Bizjak <ubizjak@gmail.com>
12721 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
12722 Emit constant permutation insn directly from here.
12724 2021-06-13 Trevor Saunders <tbsaunde@tbsaunde.org>
12726 * attribs.c (find_attribute_namespace): Iterate over vec<> with
12728 * auto-profile.c (afdo_find_equiv_class): Likewise.
12729 * gcc.c (do_specs_vec): Likewise.
12730 (do_spec_1): Likewise.
12731 (driver::set_up_specs): Likewise.
12732 * gimple-loop-jam.c (any_access_function_variant_p): Likewise.
12733 * gimple-ssa-store-merging.c (compatible_load_p): Likewise.
12734 (imm_store_chain_info::try_coalesce_bswap): Likewise.
12735 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
12736 (get_location_for_stmts): Likewise.
12737 * graphite-poly.c (print_iteration_domains): Likewise.
12738 (free_poly_bb): Likewise.
12739 (remove_gbbs_in_scop): Likewise.
12740 (free_scop): Likewise.
12741 (dump_gbb_cases): Likewise.
12742 (dump_gbb_conditions): Likewise.
12743 (print_pdrs): Likewise.
12744 (print_scop): Likewise.
12745 * ifcvt.c (cond_move_process_if_block): Likewise.
12746 * lower-subreg.c (decompose_multiword_subregs): Likewise.
12747 * regcprop.c (pass_cprop_hardreg::execute): Likewise.
12748 * sanopt.c (sanitize_rewrite_addressable_params): Likewise.
12749 * sel-sched-dump.c (dump_insn_vector): Likewise.
12750 * store-motion.c (store_ops_ok): Likewise.
12751 (store_killed_in_insn): Likewise.
12752 * timevar.c (timer::named_items::print): Likewise.
12753 * tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
12754 (cleanup_tree_cfg_noloop): Likewise.
12755 * tree-data-ref.c (dump_data_references): Likewise.
12756 (print_dir_vectors): Likewise.
12757 (print_dist_vectors): Likewise.
12758 (dump_data_dependence_relations): Likewise.
12759 (dump_dist_dir_vectors): Likewise.
12760 (dump_ddrs): Likewise.
12761 (create_runtime_alias_checks): Likewise.
12762 (free_subscripts): Likewise.
12763 (save_dist_v): Likewise.
12764 (save_dir_v): Likewise.
12765 (invariant_access_functions): Likewise.
12766 (same_access_functions): Likewise.
12767 (access_functions_are_affine_or_constant_p): Likewise.
12768 (find_data_references_in_stmt): Likewise.
12769 (graphite_find_data_references_in_stmt): Likewise.
12770 (free_dependence_relations): Likewise.
12771 (free_data_refs): Likewise.
12772 * tree-inline.c (copy_debug_stmts): Likewise.
12773 * tree-into-ssa.c (dump_currdefs): Likewise.
12774 (rewrite_update_phi_arguments): Likewise.
12775 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
12776 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
12778 (vect_slp_analyze_node_dependences): Likewise.
12779 (vect_slp_analyze_instance_dependence): Likewise.
12780 (vect_record_base_alignments): Likewise.
12781 (vect_get_peeling_costs_all_drs): Likewise.
12782 (vect_peeling_supportable): Likewise.
12783 * tree-vectorizer.c (vec_info::~vec_info): Likewise.
12784 (vec_info::free_stmt_vec_infos): Likewise.
12786 2021-06-13 Jeff Law <jeffreyalaw@gmail.com>
12788 * config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
12789 (andqi3_1<cczn>): Removed.
12790 (<ors>qi3_1): Do not split for IOR/XOR a single bit.
12791 (H8/SX bit logicals): Split out from other patterns.
12792 * config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
12793 mulqihi3_const_clobber_flags.
12794 (mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly
12796 2021-06-13 H.J. Lu <hjl.tools@gmail.com>
12799 * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
12800 to true if red zone is used.
12801 (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
12802 ix86_red_zone_used.
12803 * config/i386/i386.h (machine_function): Add red_zone_used.
12804 (ix86_red_zone_size): Removed.
12805 (ix86_red_zone_used): New.
12806 * config/i386/i386.md (peephole2 patterns): Replace
12807 ix86_red_zone_size with ix86_red_zone_used.
12809 2021-06-12 Jason Merrill <jason@redhat.com>
12811 * doc/extend.texi (unused variable attribute): Applies to
12812 structure fields as well.
12814 2021-06-12 Eugene Rozenfeld <erozen@microsoft.com>
12816 * auto-profile.c (read_profile): fix a typo in an error string
12818 2021-06-11 Thomas Schwinge <thomas@codesourcery.com>
12820 * tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
12822 * tree-pretty-print.c (dump_omp_clauses): Update.
12823 (dump_generic_node) <OMP_CLAUSE>: Use it.
12825 2021-06-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
12828 * config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
12829 int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
12830 the polymorphic variants matching code.
12831 (__arm_vld1q_z): Likewise.
12832 (__arm_vld2q): Likewise.
12833 (__arm_vld4q): Likewise.
12834 (__arm_vldrbq_gather_offset): Likewise.
12835 (__arm_vldrbq_gather_offset_z): Likewise.
12837 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
12839 PR tree-optimization/96392
12840 * fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.
12842 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
12844 PR tree-optimization/96392
12845 * fold-const.c (fold_real_zero_addition_p): Take both arguments
12846 of the addition or subtraction, not just the zero. Use this
12847 other argument in tests for signaling NaNs and signed zeros.
12848 (tree_expr_maybe_real_minus_zero_p): New predicate.
12849 * fold-const.h (fold_real_zero_addition_p): Update prototype.
12850 (tree_expr_maybe_real_minus_zero_p): New function prototype.
12851 * match.pd: Update calls to fold_real_zero_addition_p.
12852 Replace HONOR_NANS with tree_expr_maybe_nan_p.
12853 Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
12854 Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
12855 * tree-ssa-reassoc.c (eliminate_using_constants): Update
12856 call to fold_real_zero_addition_p.
12858 2021-06-11 Richard Biener <rguenther@suse.de>
12860 PR tree-optimization/101025
12861 * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
12862 all refs that require dependence checking.
12864 2021-06-11 Richard Biener <rguenther@suse.de>
12866 PR tree-optimization/101028
12867 * tree-vect-slp.c (vect_build_slp_tree_2): When SLP
12868 reassoc discovery fails fatally, mark appropriate lanes
12871 2021-06-11 Richard Biener <rguenther@suse.de>
12873 PR tree-optimization/101026
12874 * tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
12875 have a representative for the associated chain nodes.
12877 2021-06-11 Jakub Jelinek <jakub@redhat.com>
12879 PR rtl-optimization/101008
12880 * simplify-rtx.c (relational_result): New function.
12881 (simplify_logical_relational_operation,
12882 simplify_relational_operation): Use it.
12884 2021-06-11 Jakub Jelinek <jakub@redhat.com>
12887 * config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.
12889 2021-06-11 Uroš Bizjak <ubizjak@gmail.com>
12892 * config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
12893 false if the permutation can be implemented with constant
12894 permutation instruction in wider mode.
12895 (canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
12896 Handle V8QImode and V4HImode.
12898 2021-06-11 Martin Liska <mliska@suse.cz>
12900 PR gcov-profile/100788
12901 * common.opt: Add new option.
12902 * coverage.c (coverage_begin_function): Emit warning instead on
12903 the internal compiler error.
12904 * doc/invoke.texi: Document the option.
12905 * toplev.c (process_options): Enable it by default.
12907 2021-06-11 Richard Biener <rguenther@suse.de>
12909 PR middle-end/101009
12910 * tree-data-ref.c (build_classic_dist_vector_1): Make sure
12911 to set *init_b to true when we encounter a constant equal
12913 (compute_affine_dependence): Also dump the actual DR_REF.
12915 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
12917 PR tree-optimization/100984
12918 * gimple-ssa-evrp.c (ssa_equiv_stack): Use auto_vec for
12919 replacements table.
12920 (ssa_equiv_stack::~ssa_equiv_stack): Remove.
12922 2021-06-11 Kewen Lin <linkw@linux.ibm.com>
12924 * config/rs6000/rs6000.md
12925 (floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
12926 define_insn_and_split.
12928 2021-06-11 Richard Biener <rguenther@suse.de>
12930 * tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
12931 to sort operands of the associative chain.
12933 2021-06-11 Richard Biener <rguenther@suse.de>
12935 * system.h (gcc_stablesort_r): Declare.
12936 * sort.cc (gcc_sort_r): Support stable sort.
12937 (gcc_stablesort_r): Define.
12938 * vec.h (vec<>::stablesort): Add.
12940 2021-06-10 Uroš Bizjak <ubizjak@gmail.com>
12943 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
12944 Handle V2SF mode. Emit SHUFPS to fixup unpack-high for V2SF mode.
12945 (expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
12946 (expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
12947 (expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
12948 (expand_vec_perm_interleave2): Handle 64bit modes.
12949 (expand_vec_perm_even_odd_pack): Handle V8QI mode.
12950 (expand_vec_perm_even_odd_1): Ditto.
12951 (ix86_vectorize_vec_perm_const): Ditto.
12952 * config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
12953 * config/i386/sse.md: ... here.
12954 * config/i386/mmx.md (*vec_interleave_lowv2sf):
12955 New insn_and_split pattern.
12956 (*vec_interleave_highv2sf): Ditto.
12957 (mmx_pshufbv8qi3): New insn pattern.
12958 (*mmx_pblendw): Ditto.
12960 2021-06-10 Peter Bergner <bergner@linux.ibm.com>
12962 * config/rs6000/rs6000-builtin.def (build_pair): New built-in.
12963 (build_acc): Likewise.
12964 * config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
12965 source operands in little-endian mode.
12966 (rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
12967 (mma_init_builtins): Likewise.
12968 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
12969 ordering for the MMA assemble and build source operands.
12970 * doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
12972 (__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
12975 2021-06-10 Jeff Law <jeffreyalaw@gmail.com>
12977 * config/h8300/h8300.c (select_cc_mode): Handle MEM. Use
12979 * config/h8300/extensions.md: Replace _clobber_flags patterns
12982 2021-06-10 Robin Dapp <rdapp@linux.ibm.com>
12984 * config/s390/vector.md (vcond_mask_<mode><mode>): Change to
12985 (vcond_mask_<mode><tointvec>): this.
12987 2021-06-10 Andrew Stubbs <ams@codesourcery.com>
12988 Thomas Schwinge <thomas@codesourcery.com>
12990 * omp-builtins.def (BUILT_IN_GOACC_ENTER_EXIT_DATA): Split into...
12991 (BUILT_IN_GOACC_ENTER_DATA, BUILT_IN_GOACC_EXIT_DATA): ... these.
12992 * gimple.h (enum gf_mask): Split
12993 'GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA' into
12994 'GF_OMP_TARGET_KIND_OACC_ENTER_DATA' and
12995 'GF_OMP_TARGET_KIND_OACC_EXIT_DATA'.
12996 (is_gimple_omp_oacc): Update.
12997 * gimple-pretty-print.c (dump_gimple_omp_target): Likewise.
12998 * gimplify.c (gimplify_omp_target_update): Likewise.
12999 * omp-expand.c (expand_omp_target, build_omp_regions_1)
13000 (omp_make_gimple_edges): Likewise.
13001 * omp-low.c (check_omp_nesting_restrictions, lower_omp_target):
13004 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
13006 * value-query.cc (value_query::value_on_edge): Rename name to
13008 (range_query::range_on_edge): Same.
13009 (range_query::value_of_expr): Same.
13010 (range_query::value_on_edge): Same.
13011 * value-query.h (class value_query): Same.
13012 (class range_query): Same.
13014 2021-06-10 Richard Biener <rguenther@suse.de>
13016 PR tree-optimization/101003
13017 * tree-vect-slp.c (vect_build_slp_tree_2): Appropriately
13018 use the pattern stmt defs when linearizing a chain.
13020 2021-06-10 Jakub Jelinek <jakub@redhat.com>
13023 * ifcvt.c (noce_get_alt_condition, noce_try_abs): Use
13024 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
13026 2021-06-10 Clement Chigot <clement.chigot@atos.net>
13028 * config/rs6000/aix71.h (ASM_CPU_SPEC): Add Power10 directive.
13029 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
13031 2021-06-09 Andrew Pinski <apinski@marvell.com>
13033 PR tree-optimization/100925
13034 * match.pd (a ? CST1 : CST2): Limit transformations
13035 that would produce a negative to integeral types only.
13036 Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.
13038 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
13041 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
13043 * doc/tm.texi: Correctly update.
13045 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
13047 * doc/tm.texi: Correctly update.
13049 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
13052 * doc/tm.texi.in (Trampolines): Add a missing blank line.
13054 2021-06-09 Paul Eggert <eggert@cs.ucla.edu>
13057 * doc/invoke.texi (Code Gen Options); Document that -fno-trampolines
13058 and -ftrampolines work only with Ada.
13059 * doc/tm.texi.in (Trampolines): Likewise.
13060 * doc/tm.texi: Regenerated.
13062 2021-06-09 Carl Love <cel@us.ibm.com>
13064 * config/rs6000/altivec.h (vec_signextll, vec_signexti, vec_signextq):
13065 Add define for new builtins.
13066 * config/rs6000/altivec.md(altivec_vreveti2): Add define_expand.
13067 * config/rs6000/rs6000-builtin.def (VSIGNEXTI, VSIGNEXTLL): Add
13068 overloaded builtin definitions.
13069 (VSIGNEXTSB2W, VSIGNEXTSH2W, VSIGNEXTSB2D, VSIGNEXTSH2D,VSIGNEXTSW2D,
13070 VSIGNEXTSD2Q): Add builtin expansions.
13071 (SIGNEXT): Add P10 overload definition.
13072 * config/rs6000/rs6000-call.c (P9V_BUILTIN_VEC_VSIGNEXTI, P9V_BUILTIN_VEC_VSIGNEXTLL,
13073 P10_BUILTIN_VEC_SIGNEXT): Add overloaded argument definitions.
13074 * config/rs6000/vsx.md (vsx_sign_extend_v2di_v1ti): Add define_insn.
13075 (vsignextend_v2di_v1ti, vsignextend_qi_<mode>, vsignextend_hi_<mode>,
13076 vsignextend_si_v2di)[VIlong]: Add define_expand.
13077 Make define_insn vsx_sign_extend_si_v2di visible.
13078 * doc/extend.texi: Add documentation for the vec_signexti,
13079 vec_signextll builtins and vec_signextq.
13081 2021-06-09 Carl Love <cel@us.ibm.com>
13083 * config/rs6000/rs6000.c (__fixkfti, __fixunskfti, __floattikf,
13084 __floatuntikf): Names changed to __fixkfti_sw, __fixunskfti_sw,
13085 __floattikf_sw, __floatuntikf_sw respectively.
13086 * config/rs6000/rs6000.md (floatti<mode>2, floatunsti<mode>2,
13087 fix_trunc<mode>ti2, fixuns_trunc<mode>ti2): Add
13088 define_insn for mode IEEE 128.
13090 2021-06-09 Carl Love <cel@us.ibm.com>
13092 * config/rs6000/altivec.md (altivec_vslq, altivec_vsrq):
13093 Rename to altivec_vslq_<mode>, altivec_vsrq_<mode>, mode VEC_TI.
13094 * config/rs6000/vector.md (VEC_TI): Was named VSX_TI in vsx.md.
13095 (vashlv1ti3): Change to vashl<mode>3, mode VEC_TI.
13096 (vlshrv1ti3): Change to vlshr<mode>3, mode VEC_TI.
13097 * config/rs6000/vsx.md (VSX_TI): Remove define_mode_iterator. Update
13098 uses of VSX_TI to VEC_TI.
13100 2021-06-09 Carl Love <cel@us.ibm.com>
13102 * config/rs6000/dfp.md (floattitd2, fixtdti2): New define_insns.
13104 2021-06-09 Carl Love <cel@us.ibm.com>
13106 * config/rs6000/altivec.h (vec_dive, vec_mod): Add define for new
13108 * config/rs6000/altivec.md (UNSPEC_VMULEUD, UNSPEC_VMULESD,
13109 UNSPEC_VMULOUD, UNSPEC_VMULOSD): New unspecs.
13110 (altivec_eqv1ti, altivec_gtv1ti, altivec_gtuv1ti, altivec_vmuleud,
13111 altivec_vmuloud, altivec_vmulesd, altivec_vmulosd, altivec_vrlq,
13112 altivec_vrlqmi, altivec_vrlqmi_inst, altivec_vrlqnm,
13113 altivec_vrlqnm_inst, altivec_vslq, altivec_vsrq, altivec_vsraq,
13114 altivec_vcmpequt_p, altivec_vcmpgtst_p, altivec_vcmpgtut_p): New
13116 (vec_widen_umult_even_v2di, vec_widen_smult_even_v2di,
13117 vec_widen_umult_odd_v2di, vec_widen_smult_odd_v2di, altivec_vrlqmi,
13118 altivec_vrlqnm): New define_expands.
13119 * config/rs6000/rs6000-builtin.def (VCMPEQUT_P, VCMPGTST_P,
13120 VCMPGTUT_P): Add macro expansions.
13121 (BU_P10V_AV_P): Add builtin predicate definition.
13122 (VCMPGTUT, VCMPGTST, VCMPEQUT, CMPNET, CMPGE_1TI,
13123 CMPGE_U1TI, CMPLE_1TI, CMPLE_U1TI, VNOR_V1TI_UNS, VNOR_V1TI, VCMPNET_P,
13124 VCMPAET_P, VMULEUD, VMULESD, VMULOUD, VMULOSD, VRLQ,
13125 VSLQ, VSRQ, VSRAQ, VRLQNM, DIV_V1TI, UDIV_V1TI, DIVES_V1TI, DIVEU_V1TI,
13126 MODS_V1TI, MODU_V1TI, VRLQMI): New macro expansions.
13127 (VRLQ, VSLQ, VSRQ, VSRAQ, DIVE, MOD): New overload expansions.
13128 * config/rs6000/rs6000-call.c (P10_BUILTIN_VCMPEQUT,
13129 P10V_BUILTIN_CMPGE_1TI, P10V_BUILTIN_CMPGE_U1TI,
13130 P10V_BUILTIN_VCMPGTUT, P10V_BUILTIN_VCMPGTST,
13131 P10V_BUILTIN_CMPLE_1TI, P10V_BUILTIN_VCMPLE_U1TI,
13132 P10V_BUILTIN_DIV_V1TI, P10V_BUILTIN_UDIV_V1TI,
13133 P10V_BUILTIN_VMULESD, P10V_BUILTIN_VMULEUD,
13134 P10V_BUILTIN_VMULOSD, P10V_BUILTIN_VMULOUD,
13135 P10V_BUILTIN_VNOR_V1TI, P10V_BUILTIN_VNOR_V1TI_UNS,
13136 P10V_BUILTIN_VRLQ, P10V_BUILTIN_VRLQMI,
13137 P10V_BUILTIN_VRLQNM, P10V_BUILTIN_VSLQ,
13138 P10V_BUILTIN_VSRQ, P10V_BUILTIN_VSRAQ,
13139 P10V_BUILTIN_VCMPGTUT_P, P10V_BUILTIN_VCMPGTST_P,
13140 P10V_BUILTIN_VCMPEQUT_P, P10V_BUILTIN_VCMPGTUT_P,
13141 P10V_BUILTIN_VCMPGTST_P, P10V_BUILTIN_CMPNET,
13142 P10V_BUILTIN_VCMPNET_P, P10V_BUILTIN_VCMPAET_P,
13143 P10V_BUILTIN_DIVES_V1TI, P10V_BUILTIN_MODS_V1TI,
13144 P10V_BUILTIN_MODU_V1TI):
13145 New overloaded definitions.
13146 (rs6000_gimple_fold_builtin) [P10V_BUILTIN_VCMPEQUT,
13147 P10V_BUILTIN_CMPNET, P10V_BUILTIN_CMPGE_1TI,
13148 P10V_BUILTIN_CMPGE_U1TI, P10V_BUILTIN_VCMPGTUT,
13149 P10V_BUILTIN_VCMPGTST, P10V_BUILTIN_CMPLE_1TI,
13150 P10V_BUILTIN_CMPLE_U1TI]: New case statements.
13151 (rs6000_init_builtins) [bool_V1TI_type_node, int_ftype_int_v1ti_v1ti]:
13153 (altivec_init_builtins): New E_V1TImode case statement.
13154 (builtin_function_type)[P10_BUILTIN_128BIT_VMULEUD,
13155 P10_BUILTIN_128BIT_VMULOUD, P10_BUILTIN_128BIT_DIVEU_V1TI,
13156 P10_BUILTIN_128BIT_MODU_V1TI, P10_BUILTIN_CMPGE_U1TI,
13157 P10_BUILTIN_VCMPGTUT, P10_BUILTIN_VCMPEQUT]: New case statements.
13158 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute) [E_TImode,
13159 E_V1TImode]: New case statements.
13160 * config/rs6000/rs6000.h (rs6000_builtin_type_index): New enum
13161 value RS6000_BTI_bool_V1TI.
13162 * config/rs6000/vector.md (vector_gtv1ti,vector_nltv1ti,
13163 vector_gtuv1ti, vector_nltuv1ti, vector_ngtv1ti, vector_ngtuv1ti,
13164 vector_eq_v1ti_p, vector_ne_v1ti_p, vector_ae_v1ti_p,
13165 vector_gt_v1ti_p, vector_gtu_v1ti_p, vrotlv1ti3, vashlv1ti3,
13166 vlshrv1ti3, vashrv1ti3): New define_expands.
13167 * config/rs6000/vsx.md (UNSPEC_VSX_DIVSQ, UNSPEC_VSX_DIVUQ,
13168 UNSPEC_VSX_DIVESQ, UNSPEC_VSX_DIVEUQ, UNSPEC_VSX_MODSQ,
13169 UNSPEC_VSX_MODUQ): New unspecs.
13170 (mulv2di3, vsx_div_v1ti, vsx_udiv_v1ti, vsx_dives_v1ti,
13171 vsx_diveu_v1ti, vsx_mods_v1ti, vsx_modu_v1ti, xxswapd_v1ti): New
13173 (vcmpnet): New define_expand.
13174 * doc/extend.texi: Add documentation for the new builtins vec_rl,
13175 vec_rlmi, vec_rlnm, vec_sl, vec_sr, vec_sra, vec_mule, vec_mulo,
13176 vec_div, vec_dive, vec_mod, vec_cmpeq, vec_cmpne, vec_cmpgt, vec_cmplt,
13177 vec_cmpge, vec_cmple, vec_all_eq, vec_all_ne, vec_all_gt, vec_all_lt,
13178 vec_all_ge, vec_all_le, vec_any_eq, vec_any_ne, vec_any_gt, vec_any_lt,
13179 vec_any_ge, vec_any_le.
13181 2021-06-09 Carl Love <cel@us.ibm.com>
13183 * config/rs6000/altivec.md (altivec_vrl<VI_char>mi): Fix
13184 bug in argument generation.
13186 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
13188 * config/arm/iterators.md (<supf>): Remove VCLZQ_U, VCLZQ_S.
13190 * config/arm/mve.md (mve_vclzq_<supf><mode>): Add '@' prefix,
13191 remove <supf> iterator.
13192 (mve_vclzq_u<mode>): New.
13193 * config/arm/neon.md (clz<mode>2): Rename to neon_vclz<mode>.
13194 (neon_vclz<mode): Move to ...
13195 * config/arm/unspecs.md (VCLZQ_U, VCLZQ_S): Remove.
13196 * config/arm/vec-common.md: ... here. Add support for MVE.
13198 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
13200 * config/arm/mve.md (mve_vhaddq_<supf><mode>): Prefix with '@'.
13201 (@mve_vrhaddq_<supf><mode): Likewise.
13202 * config/arm/neon.md (neon_v<r>hadd<sup><mode>): Likewise.
13203 * config/arm/vec-common.md (avg<mode>3_floor, uavg<mode>3_floor)
13204 (avg<mode>3_ceil", uavg<mode>3_ceil): New patterns.
13206 2021-06-09 imba-tjd <109224573@qq.com>
13208 * doc/invoke.texi: Fix typo.
13210 2021-06-09 Roger Sayle <roger@nextmovesoftware.com>
13212 PR middle-end/53267
13213 * fold-const-call.c (fold_const_call_sss) [CASE_CFN_FMOD]:
13214 Support evaluation of fmod/fmodf/fmodl at compile-time.
13216 2021-06-09 Richard Biener <rguenther@suse.de>
13218 PR tree-optimization/100981
13219 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
13220 gimple_get_lhs to also handle calls.
13221 * tree-vect-slp-patterns.c (complex_pattern::build): Transfer
13224 2021-06-09 Richard Biener <rguenther@suse.de>
13226 PR tree-optimization/97832
13227 * tree-vectorizer.h (_slp_tree::failed): New.
13228 * tree-vect-slp.c (_slp_tree::_slp_tree): Initialize
13230 (_slp_tree::~_slp_tree): Free failed.
13231 (vect_build_slp_tree): Retain failed nodes and record
13232 matches in them, copying that back out when running
13233 into a cached fail. Dump start and end of discovery.
13234 (dt_sort_cmp): New.
13235 (vect_build_slp_tree_2): Handle associatable chains
13236 together doing more aggressive operand swapping.
13238 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
13241 * config.gcc (gcc_cv_initfini_array): Set to yes for Linux and
13243 * doc/install.texi: Require glibc 2.1 and binutils 2.12 for
13244 Linux and GNU targets.
13246 2021-06-09 Richard Biener <rguenther@suse.de>
13248 * tree-vect-stmts.c (vect_is_simple_use): Always get dt
13251 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
13253 * config/arc/arc.md (loop_end): Change it to
13254 define_insn_and_split.
13256 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
13258 * config/arc/arc.md (maddhisi4): Use VMAC2H instruction.
13259 (machi): New pattern.
13260 (umaddhisi4): Use VMAC2HU instruction.
13261 (umachi): New pattern.
13263 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
13265 * config/arc/arc-protos.h (arc_split_move_p): New prototype.
13266 * config/arc/arc.c (arc_split_move_p): New function.
13267 (arc_split_move): Clean up.
13268 * config/arc/arc.md (movdi_insn): Clean up, use arc_split_move_p.
13269 (movdf_insn): Likewise.
13270 * config/arc/simdext.md (mov<VWH>_insn): Likewise.
13272 2021-06-09 Uroš Bizjak <ubizjak@gmail.com>
13275 * config/i386/i386.c (print_operand_address_as): Rename "no_rip"
13276 argument to "raw". Do not emit segment overrides when "raw" is true.
13278 2021-06-09 Martin Liska <mliska@suse.cz>
13280 * doc/gcov.texi: Create a proper JSON files.
13281 * doc/invoke.texi: Remove dots in order to make it a valid
13284 2021-06-09 Xionghu Luo <luoxhu@linux.ibm.com>
13286 * config/rs6000/rs6000-p8swap.c (pattern_is_rotate64): New.
13287 (insn_is_load_p): Use pattern_is_rotate64.
13288 (insn_is_swap_p): Likewise.
13289 (quad_aligned_load_p): Likewise.
13290 (const_load_sequence_p): Likewise.
13291 (replace_swapped_aligned_load): Likewise.
13292 (recombine_lvx_pattern): Likewise.
13293 (recombine_stvx_pattern): Likewise.
13295 2021-06-09 Andrew MacLeod <amacleod@redhat.com>
13297 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Use a
13298 fur_stmt source record.
13299 * gimple-range.cc (fur_source::get_operand): Generic range query.
13300 (fur_source::get_phi_operand): New.
13301 (fur_source::register_dependency): New.
13302 (fur_source::query): New.
13303 (class fur_edge): New. Edge source for operands.
13304 (fur_edge::fur_edge): New.
13305 (fur_edge::get_operand): New.
13306 (fur_edge::get_phi_operand): New.
13307 (fur_edge::query): New.
13308 (fur_stmt::fur_stmt): New.
13309 (fur_stmt::get_operand): New.
13310 (fur_stmt::get_phi_operand): New.
13311 (fur_stmt::query): New.
13312 (class fur_depend): New. Statement source and process dependencies.
13313 (fur_depend::fur_depend): New.
13314 (fur_depend::register_dependency): New.
13315 (class fur_list): New. List source for operands.
13316 (fur_list::fur_list): New.
13317 (fur_list::get_operand): New.
13318 (fur_list::get_phi_operand): New.
13319 (fold_range): New. Instantiate appropriate fur_source class and fold.
13320 (fold_using_range::range_of_range_op): Use new API.
13321 (fold_using_range::range_of_address): Ditto.
13322 (fold_using_range::range_of_phi): Ditto.
13323 (imple_ranger::fold_range_internal): Use fur_depend class.
13324 (fold_using_range::range_of_ssa_name_with_loop_info): Use new API.
13325 * gimple-range.h (class fur_source): Now a base class.
13326 (class fur_stmt): New.
13327 (fold_range): New prototypes.
13328 (fur_source::fur_source): Delete.
13330 2021-06-08 Andrew Pinski <apinski@marvell.com>
13332 PR tree-optimization/25290
13333 * tree-ssa-phiopt.c (xor_replacement): Delete.
13334 (tree_ssa_phiopt_worker): Delete use of xor_replacement.
13335 (match_simplify_replacement): Allow one cheap preparation
13336 statement that can be moved to before the if.
13338 2021-06-08 Pat Haugen <pthaugen@linux.ibm.com>
13340 * config/rs6000/power10.md (power10-fused-load, power10-fused-store,
13341 power10-fused_alu, power10-fused-vec, power10-fused-branch): New.
13343 2021-06-08 Jeff Law <jeffreyalaw@gmail.com>
13345 * config/h8300/logical.md (andqi3_1): Move BCLR case into define_insn_and_split.
13346 Create length attribute on define_insn_and_split. Only split for cases which we
13348 (andqi3_1<cczn>): Renamed from andqi3_1_clobber_flags. Only handle AND here and
13349 fix length computation.
13350 (b<code><mode>msx): Combine QImode and HImode H8/SX patterns using iterator.
13352 2021-06-08 Richard Biener <rguenther@suse.de>
13354 PR tree-optimization/100923
13355 * tree-ssa-sccvn.c (valueize_refs_1): Take a pointer to
13356 the operand vector to be valueized.
13357 (valueize_refs): Likewise.
13358 (valueize_shared_reference_ops_from_ref): Adjust.
13359 (valueize_shared_reference_ops_from_call): Likewise.
13360 (vn_reference_lookup_3): Likewise.
13361 (vn_reference_lookup_pieces): Likewise. Re-valueize
13362 with honoring availability when we are about to create
13363 the ao_ref and valueized before.
13364 (vn_reference_lookup): Likewise.
13365 (vn_reference_insert_pieces): Adjust.
13367 2021-06-08 Richard Biener <rguenther@suse.de>
13369 * tree-vectorizer.h (_slp_instance::root_stmt): Change to...
13370 (_slp_instance::root_stmts): ... a vector.
13371 (SLP_INSTANCE_ROOT_STMT): Rename to ...
13372 (SLP_INSTANCE_ROOT_STMTS): ... this.
13373 (slp_root::root): Change to...
13374 (slp_root::roots): ... a vector.
13375 (slp_root::slp_root): Adjust.
13376 * tree-vect-slp.c (_slp_instance::location): Adjust.
13377 (vect_free_slp_instance): Release the root stmt vector.
13378 (vect_build_slp_instance): Adjust.
13379 (vect_analyze_slp): Likewise.
13380 (_bb_vec_info::~_bb_vec_info): Likewise.
13381 (vect_slp_analyze_operations): Likewise.
13382 (vect_bb_vectorization_profitable_p): Likewise. Adjust
13383 costs for the root stmt.
13384 (vect_slp_check_for_constructors): Gather all BIT_INSERT_EXPRs
13386 (vect_slp_analyze_bb_1): Simplify by marking all root stmts
13388 (vectorize_slp_instance_root_stmt): Adjust.
13389 (vect_schedule_slp): Likewise.
13391 2021-06-08 Aldy Hernandez <aldyh@redhat.com>
13393 * gimple-ssa-evrp.c (class ssa_equiv_stack): New.
13394 (ssa_equiv_stack::ssa_equiv_stack): New.
13395 (ssa_equiv_stack::~ssa_equiv_stack): New.
13396 (ssa_equiv_stack::enter): New.
13397 (ssa_equiv_stack::leave): New.
13398 (ssa_equiv_stack::push_replacement): New.
13399 (ssa_equiv_stack::get_replacement): New.
13400 (is_pointer_ssa): New.
13401 (class pointer_equiv_analyzer): New.
13402 (pointer_equiv_analyzer::pointer_equiv_analyzer): New.
13403 (pointer_equiv_analyzer::~pointer_equiv_analyzer): New.
13404 (pointer_equiv_analyzer::set_global_equiv): New.
13405 (pointer_equiv_analyzer::set_cond_equiv): New.
13406 (pointer_equiv_analyzer::get_equiv): New.
13407 (pointer_equiv_analyzer::enter): New.
13408 (pointer_equiv_analyzer::leave): New.
13409 (pointer_equiv_analyzer::get_equiv_expr): New.
13410 (pta_valueize): New.
13411 (pointer_equiv_analyzer::visit_stmt): New.
13412 (pointer_equiv_analyzer::visit_edge): New.
13413 (hybrid_folder::value_of_expr): Call PTA.
13414 (hybrid_folder::value_on_edge): Same.
13415 (hybrid_folder::pre_fold_bb): New.
13416 (hybrid_folder::post_fold_bb): New.
13417 (hybrid_folder::pre_fold_stmt): New.
13418 (rvrp_folder::pre_fold_bb): New.
13419 (rvrp_folder::post_fold_bb): New.
13420 (rvrp_folder::pre_fold_stmt): New.
13421 (rvrp_folder::value_of_expr): Call PTA.
13422 (rvrp_folder::value_on_edge): Same.
13424 2021-06-08 Jakub Jelinek <jakub@redhat.com>
13427 * tree-inline.c (copy_tree_body_r): For OMP_CLAUSE_DEPEND don't
13428 check TREE_CODE if OMP_CLAUSE_DECL is NULL.
13430 2021-06-08 Richard Biener <rguenther@suse.de>
13432 PR middle-end/100951
13433 * tree-vect-generic.c (expand_vector_piecewise): Build a
13434 VECTOR_CST if all elements are constant.
13435 (expand_vector_condition): Likewise.
13436 (lower_vec_perm): Likewise.
13437 (expand_vector_conversion): Likewise.
13439 2021-06-08 Martin Liska <mliska@suse.cz>
13441 * doc/invoke.texi: Document new param evrp-sparse-threshold.
13443 2021-06-08 Martin Liska <mliska@suse.cz>
13445 * genautomata.c (create_automata): Fix typo.
13447 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
13449 PR tree-optimization/100794
13450 * tree-predcom.c (tree_predictive_commoning_loop): Add parameter
13451 allow_unroll_p and only allow unrolling when it's true.
13452 (tree_predictive_commoning): Add parameter allow_unroll_p and
13454 (run_tree_predictive_commoning): Likewise.
13455 (pass_predcom::gate): Check flag_tree_loop_vectorize and
13456 global_options_set.x_flag_predictive_commoning.
13457 (pass_predcom::execute): Adjust for allow_unroll_p.
13459 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
13461 * tree-predcom.c (execute_pred_commoning): Remove update_ssa call.
13462 (tree_predictive_commoning_loop): Factor some cleanup stuffs into
13463 lambda function cleanup, remove scev_reset call, and adjust return
13465 (tree_predictive_commoning): Adjust for different changed values,
13466 only set flag TODO_update_ssa_only_virtuals if changed.
13467 (pass_data pass_data_predcom): Remove TODO_update_ssa_only_virtuals
13468 from todo_flags_finish.
13470 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
13472 * gimple-range-cache.cc (class sbr_sparse_bitmap): New.
13473 (sbr_sparse_bitmap::sbr_sparse_bitmap): New.
13474 (sbr_sparse_bitmap::bitmap_set_quad): New.
13475 (sbr_sparse_bitmap::bitmap_get_quad): New.
13476 (sbr_sparse_bitmap::set_bb_range): New.
13477 (sbr_sparse_bitmap::get_bb_range): New.
13478 (sbr_sparse_bitmap::bb_range_p): New.
13479 (block_range_cache::block_range_cache): initialize bitmap obstack.
13480 (block_range_cache::~block_range_cache): Destruct obstack.
13481 (block_range_cache::set_bb_range): Decide when to utilze the
13482 sparse on entry cache.
13483 * gimple-range-cache.h (block_range_cache): Add bitmap obstack.
13484 * params.opt (-param=evrp-sparse-threshold): New.
13486 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
13488 * bitmap.c (bitmap_set_aligned_chunk): New.
13489 (bitmap_get_aligned_chunk): New.
13490 (test_aligned_chunk): New.
13491 (bitmap_c_tests): Call test_aligned_chunk.
13492 * bitmap.h (bitmap_set_aligned_chunk, bitmap_get_aligned_chunk): New.
13494 2021-06-07 Uroš Bizjak <ubizjak@gmail.com>
13497 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
13499 (ix86_expand_vector_init_one_nonzero): Ditto.
13500 (ix86_expand_vector_init_one_var): Ditto.
13501 (ix86_expand_vector_init_general): Ditto.
13502 * config/i386/mmx.md (vec_initv4qiqi): New expander.
13504 2021-06-07 Jeff Law <jeffreyalaw@gmail.com>
13506 * config/h8300/movepush.md: Change most _clobber_flags
13507 patterns to instead use <cczn> subst.
13508 (movsi_cczn): New pattern with usable CC cases split out.
13509 (movsi_h8sx_cczn): Likewise.
13511 2021-06-07 Martin Liska <mliska@suse.cz>
13513 * common/common-target.def: Split long lines and replace them
13515 * target.def: Likewise.
13516 * doc/tm.texi: Re-generated.
13518 2021-06-07 Jakub Jelinek <jakub@redhat.com>
13521 * fold-const.c (fold_read_from_vector): Return NULL if trying to
13522 read from a CONSTRUCTOR with vector type elements.
13524 2021-06-07 Jakub Jelinek <jakub@redhat.com>
13526 PR middle-end/100898
13527 * tree-inline.c (copy_bb): Only use gimple_call_arg_ptr if memcpy
13528 should copy any arguments. Don't call gimple_call_num_args
13529 on id->call_stmt or call_stmt more than once.
13531 2021-06-07 liuhongt <hongtao.liu@intel.com>
13534 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): Refine
13536 (<insn>v4siv4di2): Delete constraints for define_expand.
13538 2021-06-07 liuhongt <hongtao.liu@intel.com>
13541 * config/i386/i386-expand.c (ix86_expand_builtin): Remove
13542 assignment of cfun->machine->has_explicit_vzeroupper.
13543 * config/i386/i386-features.c
13544 (ix86_add_reg_usage_to_vzerouppers): Delete.
13545 (ix86_add_reg_usage_to_vzeroupper): Ditto.
13546 (rest_of_handle_insert_vzeroupper): Remove
13547 ix86_add_reg_usage_to_vzerouppers, add df_analyze at the end
13549 (gate): Remove cfun->machine->has_explicit_vzeroupper.
13550 * config/i386/i386-protos.h (ix86_expand_avx_vzeroupper):
13552 * config/i386/i386.c (ix86_insn_callee_abi): New function.
13553 (ix86_initialize_callee_abi): Ditto.
13554 (ix86_expand_avx_vzeroupper): Ditto.
13555 (ix86_hard_regno_call_part_clobbered): Adjust for vzeroupper
13557 (TARGET_INSN_CALLEE_ABI): Define as ix86_insn_callee_abi.
13558 (ix86_emit_mode_set): Call ix86_expand_avx_vzeroupper
13560 * config/i386/i386.h (struct GTY(()) machine_function): Delete
13561 has_explicit_vzeroupper.
13562 * config/i386/i386.md (enum unspec): New member
13564 (ABI_DEFAULT,ABI_VZEROUPPER,ABI_UNKNOWN): New
13565 define_constants for insn callee abi index.
13566 * config/i386/predicates.md (vzeroupper_pattern): Adjust.
13567 * config/i386/sse.md (UNSPECV_VZEROUPPER): Deleted.
13568 (avx_vzeroupper): Call ix86_expand_avx_vzeroupper.
13569 (*avx_vzeroupper): Rename to ..
13570 (avx_vzeroupper_callee_abi): .. this, and adjust pattern as
13571 call_insn which has a special vzeroupper ABI.
13572 (*avx_vzeroupper_1): Deleted.
13574 2021-06-07 liuhongt <hongtao.liu@intel.com>
13577 * df-scan.c (df_get_call_refs): When call_insn is a fake call,
13578 it won't use stack pointer reg.
13579 * final.c (leaf_function_p): When call_insn is a fake call, it
13580 won't affect caller as a leaf function.
13581 * reg-stack.c (callee_clobbers_any_stack_reg): New.
13582 (subst_stack_regs): When call_insn doesn't clobber any stack
13583 reg, don't clear the arguments.
13584 * rtl.c (shallow_copy_rtx): Don't clear flag used when orig is
13586 * shrink-wrap.c (requires_stack_frame_p): No need for stack
13587 frame for a fake call.
13588 * rtl.h (FAKE_CALL_P): New macro.
13590 2021-06-06 Eric Botcazou <ebotcazou@adacore.com>
13592 * config/sparc/sparc-protos.h (order_regs_for_local_alloc): Rename
13594 (sparc_order_regs_for_local_alloc): ...this.
13595 (sparc_leaf_reg_remap): Declare.
13596 * config/sparc/sparc.h (ADJUST_REG_ALLOC_ORDER): Adjust.
13597 (LEAF_REG_REMAP): Reimplement as call to sparc_leaf_reg_remap.
13598 * config/sparc/sparc.c (leaf_reg_remap): Delete.
13599 (order_regs_for_local_alloc): Rename to...
13600 (sparc_order_regs_for_local_alloc): ...this.
13601 (sparc_leaf_reg_remap): New function.
13602 (sparc_conditional_register_usage): Do not modify leaf_reg_remap.
13604 2021-06-06 David Edelsohn <dje.gcc@gmail.com>
13606 * config/rs6000/rs6000.c (rs6000_xcoff_asm_output_aligned_decl_common):
13607 Use assemble_name to output BSS section name.
13609 2021-06-06 Uroš Bizjak <ubizjak@gmail.com>
13611 * config/i386/constraints.md (Bs):
13612 Remove boolean operators from match_test RTX.
13615 (M): Use "mode" variable instead of GET_MODE (op) in match_test RTX.
13618 2021-06-06 Martin Liska <mliska@suse.cz>
13620 * doc/extend.texi: Add missing @headitem.
13621 * doc/invoke.texi: Likewise.
13622 * doc/objc.texi: Likewise.
13624 2021-06-06 Martin Liska <mliska@suse.cz>
13626 * genhooks.c (emit_findices): Remove unused function.
13627 (emit_documentation): Do not call emit_findices
13628 and do not search for @Fcode directives.
13630 2021-06-06 Martin Liska <mliska@suse.cz>
13632 * doc/invoke.texi: Remove extra character.
13634 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13636 * config/sh/sh.md (doloop_end_split): Fix empty split condition.
13638 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13640 * config/sparc/sparc.md (*snedi<W:mode>_zero_vis3,
13641 *neg_snedi<W:mode>_zero_subxc, *plus_snedi<W:mode>_zero,
13642 *plus_plus_snedi<W:mode>_zero, *minus_snedi<W:mode>_zero,
13643 *minus_minus_snedi<W:mode>_zero): Fix empty split condition.
13645 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13647 * config/or1k/or1k.md (*movdi): Fix empty split condition.
13649 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13651 * config/mips/mips.md (<anonymous>, bswapsi2, bswapdi2): Fix empty
13654 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13656 * config/m68k/m68k.md (*zero_extend_inc, *zero_extend_dec,
13657 *zero_extendsidi2): Fix empty split condition.
13659 2021-06-05 Jeff Law <jeffreyalaw@gmail.com>
13661 * config/h8300/addsub.md: Fix split condition in define_insn_and_split
13663 * config/h8300/bitfield.md: Likewise.
13664 * config/h8300/combiner.md: Likewise.
13665 * config/h8300/divmod.md: Likewise.
13666 * config/h8300/extensions.md: Likewise.
13667 * config/h8300/jumpcall.md: Likewise.
13668 * config/h8300/movepush.md: Likewise.
13669 * config/h8300/multiply.md: Likewise.
13670 * config/h8300/other.md: Likewise.
13671 * config/h8300/shiftrotate.md: Likewise.
13672 * config/h8300/logical.md: Likewise. Fix split pattern to use
13673 code iterator that somehow slipped through.
13675 2021-06-04 Tobias Burnus <tobias@codesourcery.com>
13677 PR middle-end/100905
13678 * tree-nested.c (convert_nonlocal_omp_clauses,
13679 convert_local_omp_clauses): Handle OMP_CLAUSE_BIND.
13681 2021-06-04 Martin Sebor <msebor@redhat.com>
13683 PR middle-end/100732
13684 * gimple-fold.c (gimple_fold_builtin_sprintf): Avoid folding calls
13685 with either source or destination argument of invalid type.
13686 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Avoid checking
13687 calls with arguments of invalid type.
13689 2021-06-04 Martin Sebor <msebor@redhat.com>
13691 * attribs.c (init_attr_rdwr_indices): Use VLA bounds in the expected
13693 (attr_access::vla_bounds): Also handle VLA bounds.
13695 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
13697 * config/i386/predicates.md (GOT_memory_operand):
13698 Implement using match_code RTXes.
13699 (GOT32_symbol_operand): Ditto.
13701 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
13704 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
13706 (ix86_expand_vector_init_general): Ditto.
13707 Use SImode instead of word_mode for logic operations
13708 when GET_MODE_SIZE (mode) < UNITS_PER_WORD.
13709 (expand_vec_perm_even_odd_1): Assert that V2HI mode should be
13710 implemented by expand_vec_perm_1.
13711 (expand_vec_perm_broadcast_1): Assert that V2HI and V4HI modes
13712 should be implemented using standard shuffle patterns.
13713 (ix86_vectorize_vec_perm_const): Handle V2HImode. Add V4HI and
13714 V2HI modes to modes, implementable with shuffle for one operand.
13715 * config/i386/mmx.md (*punpckwd): New insn_and_split pattern.
13716 (*pshufw_1): New insn pattern.
13717 (*vec_dupv2hi): Ditto.
13718 (vec_initv2hihi): New expander.
13720 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
13722 * config/arm/vfp.md (no_literal_pool_df_immediate,
13723 no_literal_pool_sf_immediate): Fix empty split condition.
13725 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
13727 * config/i386/i386.md (*load_tp_x32_zext, *add_tp_x32_zext,
13728 *tls_dynamic_gnu2_combine_32): Fix empty split condition.
13729 * config/i386/sse.md (*<sse2_avx2>_pmovmskb_lt,
13730 *<sse2_avx2>_pmovmskb_zext_lt, *sse2_pmovmskb_ext_lt,
13731 *<sse4_1_avx2>_pblendvb_lt): Likewise.
13733 2021-06-04 Jakub Jelinek <jakub@redhat.com>
13736 * config/i386/i386-expand.c (ix86_expand_vector_init): Handle
13737 concatenation from half-sized modes with TImode elements.
13739 2021-06-04 Claudiu Zissulescu <claziss@synopsys.com>
13741 * config/arc/arc.c (arc_override_options): Disable millicode
13742 thunks when RF16 is on.
13744 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
13746 * config/rs6000/rs6000.h (PROMOTE_MODE): Remove.
13748 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
13750 * config/rs6000/rs6000-call.c (rs6000_promote_function_mode):
13751 Replace PROMOTE_MODE marco with its content.
13753 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
13755 * config/cris/cris.md (*addi_reload): Fix empty split condition.
13757 2021-06-03 Jim Wilson <jimw@sifive.com>
13759 * config.gcc (riscv*-*-*): If --with-riscv-attribute not used,
13760 turn it on for all riscv targets.
13762 2021-06-03 Uroš Bizjak <ubizjak@gmail.com>
13765 * config/i386/i386-expand.c (ix86_expand_vector_set):
13766 Handle V2HI and V4QI modes.
13767 (ix86_expand_vector_extract): Ditto.
13768 * config/i386/mmx.md (*pinsrw): New insn pattern.
13771 (*pextrw_zext): Ditto.
13773 (*pextrb_zext): Ditto.
13774 (vec_setv2hi): New expander.
13775 (vec_extractv2hihi): Ditto.
13776 (vec_setv4qi): Ditto.
13777 (vec_extractv4qiqi): Ditto.
13778 (vec_setv8qi): Enable only for TARGET_SSE4_1.
13779 (vec_extractv8qiqi): Ditto.
13781 2021-06-03 Aaron Sawdey <acsawdey@linux.ibm.com>
13783 * config/rs6000/genfusion.pl (gen_logical_addsubf): Fix input
13784 order to subf instruction.
13785 * config/rs6000/fusion.md: Regenerate.
13787 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
13789 * calls.c (get_size_range): Use range_of_expr instead of
13790 determine_value_range.
13791 * tree-affine.c (expr_to_aff_combination): Same.
13792 * tree-data-ref.c (split_constant_offset): Same.
13793 * tree-vrp.c (determine_value_range_1): Remove.
13794 (determine_value_range): Remove.
13795 * tree-vrp.h (determine_value_range): Remove.
13797 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
13799 * function-tests.c (test_ranges): Call gimple_range_tests.
13800 * gimple-range-cache.cc (ranger_cache::range_of_expr): Pass stmt
13802 * gimple-range.cc (fur_source::get_operand): Do not call
13803 get_tree_range or gimple_range_global.
13805 (get_tree_range): Move to value-query.cc.
13806 Call get_arith_expr_range.
13807 (gimple_ranger::range_of_expr): Add argument to get_tree_range.
13808 Include gimple-range-tests.cc.
13809 * gimple-range.h (fold_range): Add argument.
13810 (get_tree_range): Remove.
13811 * selftest.h (gimple_range_tests): New.
13812 * value-query.cc (global_range_query::range_of_expr): Add
13814 (range_query::get_tree_range): Move from gimple-range.cc.
13815 * value-query.h (class range_query): Add get_tree_range and
13816 get_arith_expr_range. Make fur_source a friend.
13817 * vr-values.c (vr_values::range_of_expr): Pass stmt to
13819 * gimple-range-tests.cc: New file.
13821 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
13823 * gimple-range.cc (gimple_ranger::export_global_ranges): Call
13824 update_global_range.
13825 * value-query.cc (update_global_range): New.
13826 * value-query.h (update_global_range): New.
13828 2021-06-03 David Malcolm <dmalcolm@redhat.com>
13830 * diagnostic-show-locus.c (diagnostic_show_locus): Don't reject
13831 printing the same location twice if there are fix-it hints,
13832 multiple locations, or a label.
13834 2021-06-03 Andre Vieira <andre.simoesdiasvieira@arm.com>
13836 * tree-vect-loop.c (vect_transform_loop): Use main loop's various'
13837 thresholds to narrow the upper bound on epilogue iterations.
13839 2021-06-03 Christophe Lyon <christophe.lyon@linaro.org>
13841 * config/arm/mve.md (mve_vabsq_f<mode>): Use 'abs' instead of unspec.
13842 (mve_vabsq_s<mode>): Likewise.
13843 * config/arm/neon.md (abs<mode>2): Rename to neon_abs<mode>2.
13844 * config/arm/unspecs.md (VABSQ_F, VABSQ_S): Delete.
13845 * config/arm/vec-common.md (neg<mode>2): Rename to
13846 <absneg_str><mode>2.
13848 2021-06-03 Claudiu Zissulescu <claziss@synopsys.com>
13850 * common/config/arc/arc-common.c (arc_option_optimization_table):
13851 Remove malign-call.
13852 * config/arc/arc.c (arc_unalign_branch_p): Remove unused function.
13853 * config/arc/arc.h (TARGET_MIXED_CODE): Remove macro.
13854 (INDEX_REG_CLASS): Only refer to GENERAL_REGS.
13855 * config/arc/arc.md (abssi2_mixed): Remove pattern.
13856 * config/arc/arc.opt (munalign-prob-threshold): Mark it obsolete.
13857 (malign-call): Likewise.
13858 (mmixed-code): Likewise.
13859 * doc/invoke.texi (ARC): Update doc.
13861 2021-06-03 Martin Liska <mliska@suse.cz>
13863 * common.opt: Use proper Enum values.
13864 * opts.c (COVERAGE_SANITIZER_OPT): Remove.
13865 (parse_sanitizer_options): Handle only sanitizer_opts.
13866 (common_handle_option): Just assign value.
13868 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13871 * tree-inline.c (inline_forbidden_p): Remove test on return type.
13873 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13875 * dwarf2out.c (loc_list_from_tree_1) <FUNCTION_DECL>: Also generate
13876 DW_OP_GNU_variable_value referencing an existing DIE at file scope.
13877 (type_byte_size): Inline into...
13878 (add_byte_size_attribute): ...this and call add_scalar_info.
13880 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13882 * dwarf2out.c (mem_loc_descriptor) <UDIV>: Fix typo.
13883 (typed_binop_from_tree): New function.
13884 (loc_list_from_tree_1) <EXACT_DIV_EXPR>: For an unsigned type,
13885 turn a divide by a power of 2 into a shift.
13886 <CEIL_DIV_EXPR>: For an unsigned type, use a signed divide if the
13887 size of the mode is lower than DWARF2_ADDR_SIZE; otherwise, do a
13888 typed divide by calling typed_binop_from_tree.
13890 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13892 * dwarf2out.c (scompare_loc_descriptor): Fix head comment.
13893 (is_handled_procedure_type): Likewise.
13894 (struct loc_descr_context): Add strict_signedness field.
13895 (resolve_args_picking_1): Deal with DW_OP_[GNU_]deref_type,
13896 DW_OP_[GNU_]convert and DW_OP_[GNU_]reinterpret.
13897 (resolve_args_picking): Minor tweak.
13898 (function_to_dwarf_procedure): Initialize strict_signedness field.
13899 (type_byte_size): Likewise.
13900 (field_byte_offset): Likewise.
13901 (gen_descr_array_type_die): Likewise.
13902 (gen_variant_part): Likewise.
13903 (loc_list_from_tree_1) <CALL_EXPR>: Tidy up and set strict_signedness
13904 to true when a context is present before evaluating the arguments.
13905 <COND_EXPR>: Do not generate a useless comparison with zero.
13906 When dereferencing an address, if strict_signedness is true and the
13907 type is small and signed, use DW_OP_deref_type to do the dereference
13908 and then DW_OP_convert to convert back to the generic type.
13910 2021-06-03 Jakub Jelinek <jakub@redhat.com>
13913 * tree-inline.c (copy_tree_body_r): Handle iterators on
13914 OMP_CLAUSE_AFFINITY or OMP_CLAUSE_DEPEND.
13916 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
13918 * config/arc/arc.md (*bbit_di): Remove.
13920 2021-06-02 Christoph Muellner <cmuellner@gcc.gnu.org>
13922 PR rtl-optimization/100264
13923 * ree.c (get_sub_rtx): Ignore SET expressions without register
13924 destinations and remove assertion, as it is not valid anymore
13925 with this new behaviour.
13926 (merge_def_and_ext): Eliminate destination check for register
13927 as such SET expressions can't occur anymore.
13928 (combine_reaching_defs): Likewise.
13930 2021-06-02 Jakub Jelinek <jakub@redhat.com>
13933 * config/xtensa/xtensa.h (LEAF_REG_REMAP): Cast REGNO to int to avoid
13934 -Wtype-limits warnings.
13935 (DWARF_FRAME_REGISTER): Rewrite into ternary operator with addition
13936 in operands to avoid -Wsign-compare warnings.
13938 2021-06-02 Pat Haugen <pthaugen@linux.ibm.com>
13940 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): Use
13943 2021-06-02 Vineet Gupta <vgupta@synopsys.com>
13945 * config/arc/arc.h (TARGET_CPU_DEFAULT): Change to hs38_linux.
13947 2021-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
13949 * config/s390/s390.md(*ashrdi3_31<setcc><cconly>): Use a single
13951 * config/s390/subst.md(cconly_subst): Use a single constraint
13952 in (match_scratch).
13954 2021-06-02 Martin Liska <mliska@suse.cz>
13956 * ipa-icf.h: Use auto_vec for memory_access_types.
13958 2021-06-02 Jeff Law <jeffreyalaw@gmail.com>
13960 * config/h8300/h8300-protos.h (compute_a_shift_length): Drop unused
13961 argument from prototype.
13962 (output_logical_op): Add rtx_code argument.
13963 (compute_logical_op_length): Likewise.
13964 * config/h8300/h8300.c (h8300_and_costs): Pass additional argument
13965 to compute_a_shift_length.
13966 (output_logical_op); New argument with the rtx code rather than
13967 extracting it from an operand. Handle QImode too.
13968 (compute_logical_op_length): Similary.
13969 (compute_a_shift_length): Drop unused argument.
13970 * config/h8300/h8300.md (logicals): New code iterator.
13971 * config/h8300/logical.md (<code><mode>3 expander): Combine
13972 the "and" expander with the "ior"/"xor" expander.
13973 (bclr<mode>msx): Combine the QI/HI mode patterns.
13974 (<logical><mode>3 insns): Use code iterator rather than match_operator.
13975 Handle QImode as well. Update call to output_logical_op and
13976 compute_logical_op_length to pass in rtx_code
13977 Fix split condition on all define_insn_and_split patterns.
13978 (one_cmpl<mode>2<cczn>): Use <cczn> to support both clobbering
13979 the flags and setting ZN via existing define_subst.
13980 * config/h8300/shiftrotate.md: Drop unused argument from
13981 calls to compute_a_shift_length.
13982 Signed-off-by: Jeff Law <jeffreyalaw@gmail.com>
13984 2021-06-01 Andrew Pinski <apinski@marvell.com>
13986 PR tree-optimization/25290
13987 * tree-ssa-phiopt.c (match_simplify_replacement):
13989 (tree_ssa_phiopt_worker): Use match_simplify_replacement.
13990 (two_value_replacement): Change the comment about
13991 conditional_replacement.
13992 (conditional_replacement): Delete.
13994 2021-06-01 Andrew Pinski <apinski@marvell.com>
13996 PR tree-optimization/95481
13997 * tree-tailcall.c (find_tail_calls): Handle empty typed
14000 2021-06-01 Andrew Pinski <apinski@marvell.com>
14002 * gimplify.c (zero_sized_field_decl): Delete
14003 (zero_sized_type): Delete
14004 (gimplify_init_ctor_eval): Use is_empty_type instead
14005 of zero_sized_field_decl.
14006 (gimplify_modify_expr): Use is_empty_type instead of
14009 2021-06-01 Jason Merrill <jason@redhat.com>
14012 * tree.h (CALL_FROM_NEW_OR_DELETE_P): Adjust comment.
14014 2021-06-01 Jason Merrill <jason@redhat.com>
14017 * diagnostic.h (warning_enabled_at): Declare.
14018 * diagnostic.c (diagnostic_enabled): Factor out from...
14019 (diagnostic_report_diagnostic): ...here.
14020 (warning_enabled_at): New.
14022 2021-06-01 Aldy Hernandez <aldyh@redhat.com>
14024 * gimple-ssa-evrp.c: Enable exporting of global ranges.
14026 2021-06-01 Martin Liska <mliska@suse.cz>
14029 * doc/invoke.texi: Mention that -fgcse-after-reload
14030 is enabled with -O3.
14032 2021-06-01 liuhongt <hongtao.liu@intel.com>
14034 PR tree-optimization/98365
14035 * tree-if-conv.c (strip_nop_cond_scalar_reduction): New function.
14036 (is_cond_scalar_reduction): Handle nop_expr in cond scalar reduction.
14037 (convert_scalar_cond_reduction): Ditto.
14038 (predicate_scalar_phi): Ditto.
14040 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
14042 PR tree-optimization/100781
14043 * gimple-range-cache.cc (ranger_cache::ranger_cache): Enable new
14044 value calculation by default.
14045 (ranger_cache::enable_new_values): New.
14046 (ranger_cache::disable_new_values): New.
14047 (ranger_cache::push_poor_value): Check if new values are allowed.
14048 * gimple-range-cache.h (class ranger_cache): New member/methods.
14049 * gimple-range.cc (gimple_ranger::range_of_expr): Check for debug
14050 statement, and disable/renable new value calculation.
14052 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
14054 * gimple-range-cache.cc (ranger_cache::ssa_range_in_bb): Delete.
14055 (ranger_cache::range_of_def): New.
14056 (ranger_cache::entry_range): New.
14057 (ranger_cache::exit_range): New.
14058 (ranger_cache::range_of_expr): Adjust.
14059 (ranger_cache::range_on_edge): Adjust.
14060 (ranger_cache::propagate_cache): Call exit_range directly.
14061 * gimple-range-cache.h (class ranger_cache): Adjust.
14063 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
14065 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust for
14066 gori_compute being a member rather than base class.
14067 dervied call to member call.
14068 (ranger_cache::dump): No longer dump gori_map.
14069 (ranger_cache::dump_bb): New.
14070 (ranger_cache::get_non_stale_global_range): Adjust for gori_compute
14071 being a member rather than base class.
14072 (ranger_cache::set_global_range): Ditto.
14073 (ranger_cache::ssa_range_in_bb): Ditto.
14074 (ranger_cache::range_of_expr): New.
14075 (ranger_cache::range_on_edge): New.
14076 (ranger_cache::block_range): Adjust for gori_computes. Debug changes.
14077 (ranger_cache::propagate_cache): Adjust debugging output.
14078 (ranger_cache::fill_block_cache): Adjust for gori_computes. Debug
14080 * gimple-range-cache.h (class ranger_cache): Make gori_compute a
14081 member, and inherit from range_query instead.
14082 (ranger_cache::dump_bb): New. split from dump.
14083 * gimple-range-gori.cc (gori_compute::ssa_range_in_bb): Delete.
14084 (gori_compute::expr_range_at_stmt): Delete.
14085 (gori_compute::compute_name_range_op): Delete.
14086 (gori_compute::compute_operand_range_switch): Add fur_source.
14087 (gori_compute::compute_operand_range): Add fur_source param, inline
14088 old compute_name_range_op and optimize_logical_operands.
14089 (struct tf_range): Delete.
14090 (gori_compute::logical_combine): Adjust
14091 (gori_compute::optimize_logical_operands): Delete.
14092 (gori_compute::compute_logical_operands_in_chain): Delete.
14093 (gori_compute::compute_logical_operands): Adjust.
14094 (gori_compute::compute_operand1_range): Adjust to fur_source.
14095 (gori_compute::compute_operand2_range): Ditto.
14096 (gori_compute::compute_operand1_and_operand2_range): Ditto.
14097 (gori_compute::outgoing_edge_range_p): Add range_query parameter,
14098 and adjust to fur_source.
14099 * gimple-range-gori.h (class gori_compute): Simplify and adjust to
14100 range_query and fur_source.
14101 * gimple-range.cc (gimple_ranger::range_on_edge): Query range_on_edge
14102 from the ranger_cache..
14103 (gimple_ranger::fold_range_internal): Adjust to base class change of
14105 (gimple_ranger::dump_bb): Adjust dump.
14106 * gimple-range.h (gimple_ranger):export gori computes object.
14108 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
14110 PR tree-optimization/100774
14111 * gimple-range-cache.cc (ranger_cache::get_non_stale_global_range):
14112 Constant values are also not stale.
14113 (ranger_cache::set_global_range): Range invariant values should also
14114 have the correct timestamp.
14116 2021-05-31 Martin Liska <mliska@suse.cz>
14118 * tree-streamer-in.c (unpack_ts_function_decl_value_fields):
14119 Unpack FUNCTION_DECL_DECL_TYPE.
14120 * tree-streamer-out.c (pack_ts_function_decl_value_fields):
14121 Stream FUNCTION_DECL_DECL_TYPE instead of
14122 DECL_IS_OPERATOR_NEW_P.
14123 * tree.h (set_function_decl_type): Use FUNCTION_DECL_DECL_TYPE
14125 (DECL_IS_OPERATOR_NEW_P): Likewise.
14126 (DECL_IS_OPERATOR_DELETE_P): Likewise.
14127 (DECL_LAMBDA_FUNCTION_P): Likewise.
14129 2021-05-31 Richard Biener <rguenther@suse.de>
14132 * internal-fn.c (expand_SHUFFLEVECTOR): Define.
14133 * internal-fn.def (SHUFFLEVECTOR): New.
14134 * internal-fn.h (expand_SHUFFLEVECTOR): Declare.
14135 * doc/extend.texi: Document __builtin_shufflevector.
14137 2021-05-31 Peter Bergner <bergner@linux.ibm.com>
14140 * config/rs6000/predicates.md(mma_assemble_input_operand): Allow
14141 indexed form addresses.
14143 2021-05-29 Jeff Law <jlaw@tachyum.com>
14145 * config/h8300/h8300.c (h8300_emit_stack_adjustment): Drop unused
14146 parameter. Call callers fixed.
14148 (output_plussi): Add FALLTHRU markers.
14149 (h8300_shift_needs_scratch_p): Add gcc_unreachable marker.
14151 2021-05-29 Jakub Jelinek <jakub@redhat.com>
14153 PR middle-end/99928
14154 * gimplify.c (gimplify_scan_omp_clauses): For taskloop simd
14155 combined with parallel, make sure to add shared clause to
14156 parallel for explicit linear clause.
14158 2021-05-29 Aldy Hernandez <aldyh@redhat.com>
14160 PR tree-optimization/100787
14161 * gimple-ssa-evrp.c: Disable exporting of global ranges.
14163 2021-05-28 Jason Merrill <jason@redhat.com>
14165 * tree-iterator.h (struct tree_stmt_iterator): Add operator++,
14166 operator--, operator*, operator==, and operator!=.
14167 (class tsi_range): New.
14169 2021-05-28 Richard Biener <rguenther@suse.de>
14171 PR tree-optimization/100778
14172 * tree-vect-slp.c (vect_build_slp_tree_1): Prevent possibly
14173 trapping ops in different BBs.
14175 2021-05-28 Richard Biener <rguenther@suse.de>
14178 * tree-inline.c (copy_bb): When processing __builtin_va_arg_pack
14179 copy fntype from original call.
14181 2021-05-28 Martin Liska <mliska@suse.cz>
14183 PR gcov-profile/100751
14184 * doc/gcov.texi: Revert partially a hunk that was wrong.
14186 2021-05-28 Cooper Qu <cooper.qu@linux.alibaba.com>
14188 * config/csky/csky-linux-elf.h (HAVE_sync_compare_and_swapqi):
14190 (HAVE_sync_compare_and_swaphi): Likewise.
14191 (HAVE_sync_compare_and_swapsi): Likewise.
14193 2021-05-28 Jakub Jelinek <jakub@redhat.com>
14195 PR middle-end/99928
14196 * tree.h (OMP_CLAUSE_MAP_IMPLICIT): Define.
14198 2021-05-28 Tobias Burnus <tobias@codesourcery.com>
14200 * gimplify.c (gimplify_omp_affinity): New.
14201 (gimplify_scan_omp_clauses): Call it; remove affinity clause afterwards.
14202 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_AFFINITY.
14203 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_AFFINITY.
14204 * tree.c (omp_clause_num_ops, omp_clause_code_name): Add clause.
14205 (walk_tree_1): Handle OMP_CLAUSE_AFFINITY.
14207 2021-05-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
14208 Richard Biener <rguenther@suse.de>
14210 * match.pd <popcount & / + pattern matching>:
14211 When generating popcount directly fails, try doing it in two halves.
14213 2021-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
14215 * Makefile.in (generated_files): Add gimple-match.c and
14218 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
14220 * gensupport.c (alter_predicate_for_insn): Handle MATCH_DUP.
14222 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
14224 * gensupport.c (alter_constraints): Add MATCH_SCRATCH case.
14226 2021-05-28 Kewen Lin <linkw@linux.ibm.com>
14228 PR tree-optimization/99398
14229 * tree-ssa-forwprop.c (simplify_permutation): Optimize some cases
14230 where the fed operands are CTOR/CST and propagated through
14231 VIEW_CONVERT_EXPR. Call vec_perm_indices::new_shrunk_vector.
14232 * vec-perm-indices.c (vec_perm_indices::new_shrunk_vector): New
14234 * vec-perm-indices.h (vec_perm_indices::new_shrunk_vector): New
14237 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
14239 * config/i386/mmx.md (addv2sf3): Do not call
14240 ix86_fixup_binary_operands_no_copy.
14243 (<smaxmin:code>v2sf3): Ditto.
14244 (<plusminus:insn><MMXMODEI:mode>3): Ditto.
14245 (<plusminus:insn><VI_32:mode>3): Remove expander.
14246 (<plusminus:insn><VI_32:mode>3): Rename from
14247 "*<plusminus:insn><VI_32:mode>3".
14248 (mulv4hi): Do not call ix86_fixup_binary_operands_no_copy.
14249 (mulv2hi3): Remove expander.
14250 (mulv2hi3): Rename from *mulv2hi3.
14251 (<s>mulv2hi3_highpart): Remove expander.
14252 (<s>mulv2hi3_highpart): Rename from *<s>mulv2hi3_highpart.
14253 (<smaxmin:code><MMXMODE14:mode>3): Rename from
14254 "*mmx_<smaxmin:code><MMXMODE14:mode>3".
14255 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Remove expander.
14256 (SMAXMIN_MMXMODEI): Remove mode iterator.
14257 (<smaxmin:code>v4hi3): New expander.
14258 (<smaxmin:code>v4qi3): Rename from *<smaxmin:code>v4qi3.
14259 (<smaxmin:code>v2hi3): Rename from *<smaxmin:code>v2hi3.
14260 (<smaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
14261 (SMAXMIN_VI_32): Remove mode iterator.
14262 (<umaxmin:code><MMXMODE24:mode>3): Rename from
14263 "*mmx_<umaxmin:code><MMXMODE24:mode>3".
14264 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Remove expander.
14265 (UMAXMIN_MMXMODEI): Remove mode iterator.
14266 (<umaxmin:code>v8qi3): New expander.
14267 (<umaxmin:code>v4qi3): Rename from *<umaxmin:code>v4qi3.
14268 (<umaxmin:code>v2hi3): Rename from *<umaxmin:code>v2hi3.
14269 (<umaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
14270 (UMAXMIN_VI_32): Remove mode iterator.
14271 (<any_shift:insn>v2hi3): Remove expander.
14272 (<any_shift:insn>v2hi3): Rename from *<any_shift:insn>v2hi3.
14273 (<any_logic:code><MMXMODEI:mode>3): Do not call
14274 ix86_fixup_binary_operands_no_copy.
14275 (<any_logic:code><VI_32:mode>3): Remove expander.
14276 (<any_logic:code><VI_32:mode>3): Rename from
14277 "*<any_logic:code><VI_32:mode>3".
14278 (uavg<mode>3_ceil): Do not call ix86_fixup_binary_operands_no_copy.
14279 * config/i386/sse.md (div<VF2:mode>3): Do not call
14280 ix86_fixup_binary_operands_no_copy.
14281 (div<VF1:mode>3): Ditto.
14282 (<maxmin:code><VI8_AVX2_AVX512F:mode>3): Ditto.
14283 (smulhrsv4hi3): Ditto.
14284 (smulhrsv2hi3): Ditto.
14286 2021-05-27 Martin Sebor <msebor@redhat.com>
14288 * ggc.h (gt_ggc_mx): Add overloads for all integers.
14290 * hash-map.h (class hash_map): Add pch_nx_helper overloads for all
14292 (hash_map::operator==): New function.
14294 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
14297 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
14298 For TARGET_XOP bypass SSE comparisons for all supported vector modes.
14299 * config/i386/mmx.md (*xop_maskcmp<MMXMODEI:mode>3): New insn pattern.
14300 (*xop_maskcmp<VI_32:mode>3): Ditto.
14301 (*xop_maskcmp_uns<MMXMODEI:mode>3): Ditto.
14302 (*xop_maskcmp_uns<VI_32:mode>3): Ditto.
14304 2021-05-27 Richard Earnshaw <rearnsha@arm.com>
14307 * config/arm/arm.c (arm_configure_build_target): Remove parameter
14308 opts_set, directly check opts parameters for being non-null.
14309 (arm_option_restore): Update call to arm_configure_build_target.
14310 (arm_option_override): Likewise.
14311 (arm_can_inline_p): Likewise.
14312 (arm_valid_target_attribute_tree): Likewise.
14313 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
14314 * config/arm/arm-protos.h (arm_configure_build_target): Adjust
14317 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
14319 * vr-values.c (simplify_conversion_using_ranges): Use
14320 get_range_query instead of get_global_range_query.
14322 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
14324 * gimple-range.cc (get_range_global): Move to value-query.cc.
14325 (gimple_range_global): Same.
14326 (get_global_range_query): Same.
14327 (global_range_query::range_of_expr): Same.
14328 * gimple-range.h (class global_range_query): Move to
14330 (gimple_range_global): Same.
14331 * tree-ssanames.c (get_range_info): Move to value-query.cc.
14332 (get_ptr_nonnull): Same.
14333 * tree-ssanames.h (get_range_info): Remove.
14334 (get_ptr_nonnull): Remove.
14335 * value-query.cc (get_ssa_name_range_info): Move from
14337 (get_ssa_name_ptr_info_nonnull): Same.
14338 (get_range_global): Move from gimple-range.cc.
14339 (gimple_range_global): Same.
14340 (get_global_range_query): Same.
14341 (global_range_query::range_of_expr): Same.
14342 * value-query.h (class global_range_query): Move from
14344 (gimple_range_global): Same.
14346 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
14349 * config/i386/mmx.md (uavgv4qi3_ceil): New insn pattern.
14350 (uavgv2hi3_ceil): Ditto.
14352 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
14355 * doc/extend.texi (scalar_storage_order): Rephrase slightly.
14357 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14359 * tree-ssanames.c (get_range_info): Merge both copies of
14360 get_range_info into one that works with irange.
14361 * tree-ssanames.h (get_range_info): Remove version that works on
14364 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14366 * builtins.c (check_nul_terminated_array): Convert to get_range_query.
14367 (expand_builtin_strnlen): Same.
14368 (determine_block_size): Same.
14369 * fold-const.c (expr_not_equal_to): Same.
14370 * gimple-fold.c (size_must_be_zero_p): Same.
14371 * gimple-match-head.c: Include gimple-range.h.
14372 * gimple-pretty-print.c (dump_ssaname_info): Convert to get_range_query.
14373 * gimple-ssa-warn-restrict.c
14374 (builtin_memref::extend_offset_range): Same.
14375 * graphite-sese-to-poly.c (add_param_constraints): Same.
14376 * internal-fn.c (get_min_precision): Same.
14377 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): Same.
14378 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Same.
14380 * tree-data-ref.c (split_constant_offset): Same.
14381 (dr_step_indicator): Same.
14382 * tree-dfa.c (get_ref_base_and_extent): Same.
14383 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
14384 * tree-ssa-loop-niter.c (refine_value_range_using_guard): Same.
14385 (determine_value_range): Same.
14386 (record_nonwrapping_iv): Same.
14387 (infer_loop_bounds_from_signedness): Same.
14388 (scev_var_range_cant_overflow): Same.
14389 * tree-ssa-phiopt.c (two_value_replacement): Same.
14390 * tree-ssa-pre.c (insert_into_preds_of_block): Same.
14391 * tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Same.
14392 * tree-ssa-strlen.c (handle_builtin_stxncpy_strncat): Same.
14394 (dump_strlen_info): Same.
14395 (set_strlen_range): Same.
14396 (maybe_diag_stxncpy_trunc): Same.
14397 (get_len_or_size): Same.
14398 (handle_integral_assign): Same.
14399 * tree-ssa-structalias.c (find_what_p_points_to): Same.
14400 * tree-ssa-uninit.c (find_var_cmp_const): Same.
14401 * tree-switch-conversion.c (bit_test_cluster::emit): Same.
14402 * tree-vect-patterns.c (vect_get_range_info): Same.
14403 (vect_recog_divmod_pattern): Same.
14404 * tree-vrp.c (intersect_range_with_nonzero_bits): Same.
14405 (register_edge_assert_for_2): Same.
14406 (determine_value_range_1): Same.
14407 * tree.c (get_range_pos_neg): Same.
14408 * vr-values.c (vr_values::get_lattice_entry): Same.
14409 (vr_values::update_value_range): Same.
14410 (simplify_conversion_using_ranges): Same.
14412 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14414 * gimple-ssa-warn-alloca.c (alloca_call_type): Use
14415 get_range_query instead of query argument.
14416 (pass_walloca::execute): Enable and disable global ranger.
14418 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14420 * gimple-ssa-evrp.c (rvrp_folder::rvrp_folder): Call
14422 (rvrp_folder::~rvrp_folder): Call disable_ranger.
14423 (hybrid_folder::hybrid_folder): Call enable_ranger.
14424 (hybrid_folder::~hybrid_folder): Call disable_ranger.
14426 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14428 * function.c (allocate_struct_function): Set cfun->x_range_query.
14429 * function.h (struct function): Declare x_range_query.
14430 (get_range_query): New.
14431 (get_global_range_query): New.
14432 * gimple-range-cache.cc (ssa_global_cache::ssa_global_cache):
14433 Remove call to safe_grow_cleared.
14434 * gimple-range.cc (get_range_global): New.
14435 (gimple_range_global): Move from gimple-range.h.
14436 (get_global_range_query): New.
14437 (global_range_query::range_of_expr): New.
14438 (enable_ranger): New.
14439 (disable_ranger): New.
14440 * gimple-range.h (gimple_range_global): Move to gimple-range.cc.
14441 (class global_range_query): New.
14442 (enable_ranger): New.
14443 (disable_ranger): New.
14444 * gimple-ssa-evrp.c (evrp_folder::~evrp_folder): Rename
14445 dump_all_value_ranges to dump.
14446 * tree-vrp.c (vrp_prop::finalize): Same.
14447 * value-query.cc (range_query::dump): New.
14448 * value-query.h (range_query::dump): New.
14449 * vr-values.c (vr_values::dump_all_value_ranges): Rename to...
14450 (vr_values::dump): ...this.
14451 * vr-values.h (class vr_values): Rename dump_all_value_ranges to
14452 dump and make virtual.
14454 2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
14456 * config/i386/i386.c (ix86_autovectorize_vector_modes):
14457 Add V4QImode and V16QImode for TARGET_SSE2.
14458 * doc/sourcebuild.texi (Vector-specific attributes):
14459 Add vect64 and vect32 description.
14461 2021-05-26 Bernd Edlinger <bernd.edlinger@hotmail.de>
14463 * gimple-range-gori.cc (range_def_chain::register_dependency):
14464 Resize m_def_chain when needed.
14466 2021-05-26 Christophe Lyon <christophe.lyon@linaro.org>
14468 * config/arm/mve.md (mve_vaddvq_<supf><mode>): Prefix with '@'.
14469 * config/arm/neon.md (reduc_plus_scal_<mode>): Move to ..
14470 * config/arm/vec-common.md: .. here. Add support for MVE.
14472 2021-05-26 Jakub Jelinek <jakub@redhat.com>
14474 * config/epiphany/epiphany.c (epiphany_print_operand_address): Remove
14476 * config/microblaze/microblaze.c (microblaze_legitimize_address,
14478 microblaze_option_override, print_operand): Likewise.
14479 * config/microblaze/microblaze.md (call_internal_plt,
14480 call_value_intern_plt, call_value_intern): Likewise.
14481 * config/arm/aout.h (ASM_OUTPUT_ALIGN): Likewise.
14482 * config/iq2000/iq2000.md (call_internal1, call_value_internal1,
14483 call_value_multiple_internal1): Likewise.
14484 * config/bfin/bfin.c (symbolic_reference_mentioned_p): Likewise.
14486 2021-05-26 Jan-Benedict Glaw <jbglaw@lug-owl.de>
14488 * config/arc/arc.c (arc_address_cost, arc_print_operand_address,
14489 arc_ccfsm_advance, symbolic_reference_mentioned_p,
14490 arc_raw_symbolic_reference_mentioned_p): Remove register
14493 2021-05-26 Jakub Jelinek <jakub@redhat.com>
14496 * omp-low.c: Include omp-offload.h.
14497 (create_omp_child_function): If current_function_decl has
14498 "omp declare target" attribute and is_gimple_omp_offloaded,
14499 remove that attribute from the copy of attribute list and
14500 add "omp target entrypoint" attribute instead.
14501 (lower_omp_target): Mark .omp_data_sizes.* and .omp_data_kinds.*
14502 variables for offloading if in omp_maybe_offloaded_ctx.
14503 * omp-offload.c (pass_omp_target_link::execute): Nullify second
14504 argument to GOMP_target_data_ext in offloaded code.
14506 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14508 * config/csky/csky.c (csky_can_change_mode_class): Delete.
14509 For csky, HF/SF mode use the low bits of VREGS.
14511 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
14513 * gimplify.c (gimplify_decl_expr): Do not clear TREE_READONLY on a
14514 DECL which is a reference for OMP.
14516 2021-05-26 Martin Liska <mliska@suse.cz>
14518 PR gcov-profile/100751
14519 * doc/gcov.texi: Document that __gcov_dump can be called just
14520 once and that __gcov_reset resets run-time counters.
14522 2021-05-26 Martin Liska <mliska@suse.cz>
14524 * doc/install.texi: Port relevant part from install-old.texi
14525 and re-generate list of CPUs and systems.
14527 2021-05-26 Martin Liska <mliska@suse.cz>
14529 * Makefile.in: Remove it.
14530 * doc/include/fdl.texi: Update next/previous chapters.
14531 * doc/install.texi: Likewise.
14532 * doc/install-old.texi: Removed.
14534 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14536 * config/csky/csky.c (ck810_legitimate_index_p): Support
14537 "base + index" with DF mode.
14538 * config/csky/constraints.md ("Y"): New constraint for memory operands
14539 without index register.
14540 * config/csky/csky_insn_fpuv2.md (fpuv3_movdf): Use "Y" instead of "m"
14541 when mov between memory and general registers, and lower their priority.
14542 * config/csky/csky_insn_fpuv3.md (fpuv2_movdf): Likewise.
14544 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14546 * config/csky/csky.c (TARGET_PROMOTE_PROTOTYPES): Delete.
14548 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14550 * config/csky/csky.md (untyped_call): Emit clobber for return
14551 registers to mark them used.
14553 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14555 * config/csky/csky.md (cskyv2_sextend_ldbs): New.
14557 2021-05-26 Andrew Pinski <apinski@marvell.com>
14559 * match.pd (x < 0 ? ~y : y): New patterns.
14561 2021-05-26 Andrew Pinski <apinski@marvell.com>
14563 * match.pd (A?CST1:CST2): Add simplifcations for A?0:+-1, A?+-1:0,
14564 A?POW2:0 and A?0:POW2.
14566 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14568 * gimple-range-gori.cc (class logical_stmt_cache): Delete
14569 (logical_stmt_cache::logical_stmt_cache ): Delete.
14570 (logical_stmt_cache::~logical_stmt_cache): Delete.
14571 (logical_stmt_cache::cache_entry::dump): Delete.
14572 (logical_stmt_cache::get_range): Delete.
14573 (logical_stmt_cache::cached_name ): Delete.
14574 (logical_stmt_cache::same_cached_name): Delete.
14575 (logical_stmt_cache::cacheable_p): Delete.
14576 (logical_stmt_cache::slot_diagnostics ): Delete.
14577 (logical_stmt_cache::dump): Delete.
14578 (gori_compute_cache::gori_compute_cache): Delete.
14579 (gori_compute_cache::~gori_compute_cache): Delete.
14580 (gori_compute_cache::compute_operand_range): Delete.
14581 (gori_compute_cache::cache_stmt): Delete.
14582 * gimple-range-gori.h (gori_compute::compute_operand_range): Remove
14584 (class gori_compute_cache): Delete.
14586 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14588 * gimple-range.cc (fold_using_range::range_of_range_op): Use m_gori
14590 (fold_using_range::range_of_address): Adjust.
14591 (fold_using_range::range_of_phi): Adjust.
14592 * gimple-range.h (class fur_source): Adjust.
14593 (fur_source::fur_source): Adjust.
14595 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14597 * gimple-range-gori.cc (gori_compute::expr_range_at_stmt): Rename
14598 from expr_range_in_bb and adjust.
14599 (gori_compute::compute_name_range_op): Adjust.
14600 (gori_compute::optimize_logical_operands): Adjust.
14601 (gori_compute::compute_logical_operands_in_chain): Adjust.
14602 (gori_compute::compute_operand1_range): Adjust.
14603 (gori_compute::compute_operand2_range): Adjust.
14604 (ori_compute_cache::cache_stmt): Adjust.
14605 * gimple-range-gori.h (gori_compute): Rename prototype.
14607 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14609 * gimple-range.cc (gimple_ranger::range_of_expr): Non-null should be
14610 checked only after range_of_stmt, not range_on_entry.
14611 (gimple_ranger::range_on_entry): Check for non-null in any
14612 predecessor block, if it is not already non-null.
14613 (gimple_ranger::range_on_exit): DOnt check for non-null after
14614 range on entry call.
14615 (gimple_ranger::dump_bb): New. Split from dump.
14616 (gimple_ranger::dump): Adjust.
14617 * gimple-range.h (class gimple_ranger): Adjust.
14619 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14621 * gimple-range-cache.cc (struct range_timestamp): Delete.
14622 (class temporal_cache): Adjust.
14623 (temporal_cache::get_timestamp): Delete.
14624 (temporal_cache::set_dependency): Delete.
14625 (temporal_cache::temporal_value): Adjust.
14626 (temporal_cache::current_p): Take dependencies as params.
14627 (temporal_cache::set_timestamp): Adjust.
14628 (temporal_cache::set_always_current): Adjust.
14629 (ranger_cache::get_non_stale_global_range): Adjust.
14630 (ranger_cache::register_dependency): Delete.
14631 * gimple-range-cache.h (class range_cache): Adjust.
14633 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14635 * gimple-range-gori.cc (range_def_chain::range_def_chain): init
14637 (range_def_chain::~range_def_chain): Dispose of obstack rather than
14638 each individual bitmap.
14639 (range_def_chain::set_import): New.
14640 (range_def_chain::get_imports): New.
14641 (range_def_chain::chain_import_p): New.
14642 (range_def_chain::register_dependency): Rename from build_def_chain
14644 (range_def_chain::def_chain_in_bitmap_p): New.
14645 (range_def_chain::add_def_chain_to_bitmap): New.
14646 (range_def_chain::has_def_chain): Just check first depenedence.
14647 (range_def_chain::get_def_chain): Process imports, use generic
14648 register_dependency routine.
14649 (range_def_chain::dump): New.
14650 (gori_map::gori_map): Allocate import list.
14651 (gori_map::~gori_map): Release imports.
14652 (gori_map::exports): Check for past allocated block size.
14653 (gori_map::imports): New.
14654 (gori_map::def_chain_in_export_p): Delete.
14655 (gori_map::is_import_p): New.
14656 (gori_map::maybe_add_gori): Handle imports.
14657 (gori_map::dump): Adjust output, add imports.
14658 (gori_compute::has_edge_range_p): Remove def_chain_in_export call.
14659 (gori_export_iterator::gori_export_iterator): New.
14660 (gori_export_iterator::next): New.
14661 (gori_export_iterator::get_name): New.
14662 * gimple-range-gori.h (range_def_chain): Add imports and direct
14663 dependecies via struct rdc.
14664 (range_def_chain::depend1): New.
14665 (range_def_chain::depend2): New.
14666 (class gori_map): Adjust.
14667 (FOR_EACH_GORI_IMPORT_NAME): New.
14668 (FOR_EACH_GORI_EXPORT_NAME): New.
14669 (class gori_export_iterator): New.
14671 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14673 * gimple-range-cache.cc (ranger_cache::ranger_cache): Move initial
14674 export cache filling to here.
14675 * gimple-range-gori.cc (gori_compute::gori_compute) : From Here.
14677 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14679 * gimple-range-gori.cc (range_def_chain): Move to gimple-range-gori.h.
14680 (gori_map): Move to gimple-range-gori.h.
14681 (gori_compute::gori_compute): Adjust.
14682 (gori_compute::~gori_compute): Delete.
14683 (gori_compute::compute_operand_range_switch): Adjust.
14684 (gori_compute::compute_operand_range): Adjust.
14685 (gori_compute::compute_logical_operands): Adjust.
14686 (gori_compute::has_edge_range_p ): Adjust.
14687 (gori_compute::set_range_invariant): Delete.
14688 (gori_compute::dump): Adjust.
14689 (gori_compute::outgoing_edge_range_p): Adjust.
14690 * gimple-range-gori.h (class range_def_chain): Relocate here.
14691 (class gori_map): Relocate here.
14692 (class gori_compute): Inherit from gori_map, and adjust.
14694 2021-05-25 Aldy Hernandez <aldyh@redhat.com>
14696 * value-range.cc (range_tests_legacy): Use
14697 build_nonstandard_integer_type instead of int and short.
14699 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
14701 * gimplify.c (gimplify_decl_expr): Clear TREE_READONLY on the DECL
14702 when really creating an initialization statement for it.
14704 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
14706 * tree-inline.c (setup_one_parameter): Fix thinko in new condition.
14708 2021-05-25 Kito Cheng <kito.cheng@sifive.com>
14710 * config/riscv/riscv.h (ASM_SPEC): Pass -mno-relax.
14712 2021-05-25 Martin Liska <mliska@suse.cz>
14714 PR tree-optimization/92860
14716 * optc-save-gen.awk: Remove exceptions.
14718 2021-05-25 Martin Liska <mliska@suse.cz>
14720 * asan.h (sanitize_coverage_p): New function.
14721 * doc/extend.texi: Document it.
14722 * fold-const.c (fold_range_test): Use sanitize_flags_p
14723 instead of flag_sanitize_coverage.
14724 (fold_truth_andor): Likewise.
14725 * sancov.c: Likewise.
14726 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
14727 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Handle
14728 -fsanitize-coverage when inlining.
14730 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
14732 * config/csky/csky-modes.def : Fix copyright.
14734 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
14736 * config/csky/csky-modes.def : Amend copyright.
14737 * config/csky/csky_insn_fpuv2.md : Likewise.
14738 * config/csky/csky_insn_fpuv3.md : Likewise.
14740 2021-05-25 Richard Biener <rguenther@suse.de>
14742 PR middle-end/100727
14743 * calls.c (initialize_argument_information): Explicitely test
14744 for WITH_SIZE_EXPR.
14745 * gimple-expr.c (mark_addressable): Skip outer WITH_SIZE_EXPR.
14747 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14749 * config/csky/csky.h (FRAME_POINTER_REGNUM): Use
14750 HARD_FRAME_POINTER_REGNUM and FRAME_POINTER_REGNUM instead of
14751 the signle definition. The signle definition may not work well
14752 at simplify_subreg_regno().
14753 (HARD_FRAME_POINTER_REGNUM): New.
14754 (ELIMINABLE_REGS): Add for HARD_FRAME_POINTER_REGNUM.
14755 * config/csky/csky.c (get_csky_live_regs, csky_can_eliminate,
14756 csky_initial_elimination_offset, csky_expand_prologue,
14757 csky_expand_epilogue): Add for HARD_FRAME_POINTER_REGNUM.
14759 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14761 * config/csky/csky.c (csky_option_override):
14762 Init csky_arch_isa_features[] in advance, so TARGET_DSP
14763 and TARGET_DIV can be set well.
14765 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14767 * config/csky/constraints.md ("l", "h"): Delete.
14768 * config/csky/csky.h (reg_class, REG_CLASS_NAMES,
14769 REG_CLASS_CONTENTS): Delete LO_REGS and HI_REGS.
14770 * config/csky/csky.c (regno_reg_classm,
14771 csky_secondary_reload, csky_register_move_cost):
14772 Use HILO_REGS instead of LO_REGS and HI_REGS.
14774 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14776 * config/csky/constraints.md ("W"): New constriant for mem operand
14777 with base reg, index register.
14778 ("Q"): Renamed and modified "csky_valid_fpuv2_mem_operand" to
14779 "csky_valid_mem_constraint_operand" to deal with both "Q" and "W"
14781 ("Dv"): New constraint for const double value that can be used at
14783 * config/csky/csky-modes.def (HFmode): New mode.
14784 * config/csky/csky-protos.h (csky_valid_fpuv2_mem_operand): Rename
14785 to "csky_valid_mem_constraint_operand" and support new constraint
14787 (csky_get_movedouble_length): New.
14788 (fpuv3_output_move): New.
14789 (fpuv3_const_double): New.
14790 * config/csky/csky.c (csky_option_override): New arch CK860 with fpv3.
14791 (decompose_csky_address): Refine.
14792 (csky_print_operand): New "CONST_DOUBLE" operand.
14793 (csky_output_move): Support fpv3 instructions.
14794 (csky_get_movedouble_length): New.
14795 (fpuv3_output_move): New.
14796 (fpuv3_const_double): New.
14797 (csky_emit_compare): Cover float comparsion.
14798 (csky_emit_compare_float): Refine.
14799 (csky_vaild_fpuv2_mem_operand): Rename to
14800 "csky_valid_mem_constraint_operand" and support new constraint "W".
14801 (ck860_rtx_costs): New.
14802 (csky_rtx_costs): Add the cost calculation of CK860.
14803 (regno_reg_class): New vregs for fpuv3.
14804 (csky_dbx_regno): Likewise.
14805 (csky_cpu_cpp_builtins): New builtin macro for fpuv3.
14806 (csky_conditional_register_usage): Suporrot fpuv3.
14807 (csky_dwarf_register_span): Suporrot fpuv3.
14808 (csky_init_builtins, csky_mangle_type): Support "__fp16" type.
14809 (ck810_legitimate_index_p): Support fp16.
14810 * config/csky/csky.h (TARGET_TLS): ADD CK860.
14811 (CSKY_VREG_P, CSKY_VREG_LO_P, CSKY_VREG_HI_P): Support fpuv3.
14812 (TARGET_SINGLE_FPU): Support fpuv3.
14813 (TARGET_SUPPORT_FPV3): New.
14814 (FIRST_PSEUDO_REGISTER): Change to 202 to hold the new fpuv3 registers.
14815 (FIXED_REGISTERS, CALL_REALLY_USED_REGISTERS, REGISTER_NAMES,
14816 REG_CLASS_CONTENTS): Support fpuv3.
14817 * config/csky/csky.md (movsf): Move to cksy_insn_fpu.md and refine.
14818 (csky_movsf_fpv2): Likewise.
14819 (ck801_movsf): Likewise.
14820 (csky_movsf): Likewise.
14822 (csky_movdf_fpv2): Likewise.
14823 (ck801_movdf): Likewise.
14824 (csky_movdf): Likewise.
14825 (movsicc): Refine. Use "comparison_operatior" instead of
14826 "ordered_comparison_operatior".
14827 (addsicc): Likewise.
14828 (CSKY_FIRST_VFP3_REGNUM, CSKY_LAST_VFP3_REGNUM): New constant.
14829 (call_value_internal_vh): New.
14830 * config/csky/csky_cores.def (CK860): New arch and cpu.
14835 * config/csky/csky_insn_fpu.md: Refactor. Separate all float patterns
14836 into emit-patterns and match-patterns, remain the emit-patterns here,
14837 and move the match-patterns to csky_insn_fpuv2.md or
14838 csky_insn_fpuv3.md.
14839 * config/csky/csky_insn_fpuv2.md: New file for fpuv2 instructions.
14840 * config/csky/csky_insn_fpuv3.md: New file and new patterns for fpuv3
14842 * config/csky/csky_isa.def (fcr): New.
14847 (CK860): New definition for ck860.
14848 * config/csky/csky_tables.opt (ck860): New processors ck860,
14849 ck860f. And new arch ck860.
14854 * config/csky/predicates.md (csky_float_comparsion_operator): Delete
14855 "geu", "gtu", "leu", "ltu", which will never appear at float comparison.
14856 * config/csky/t-csky-elf: Support 860.
14857 * config/csky/t-csky-linux: Likewise.
14858 * doc/md.texi: Add "Q" and "W" constraints for C-SKY.
14860 2021-05-24 Aaron Sawdey <acsawdey@linux.ibm.com>
14862 * config/rs6000/genfusion.pl (gen_logical_addsubf): Refactor to
14863 add generation of logical-add and add-logical fusion pairs.
14864 * config/rs6000/rs6000-cpus.def: Add new fusion to ISA 3.1 mask
14866 * config/rs6000/rs6000.c (rs6000_option_override_internal): Turn on
14867 logical-add and add-logical fusion by default.
14868 * config/rs6000/rs6000.opt: Add -mpower10-fusion-logical-add and
14869 -mpower10-fusion-add-logical options.
14870 * config/rs6000/fusion.md: Regenerate file.
14872 2021-05-24 Aldy Hernandez <aldyh@redhat.com>
14874 * value-range.cc (irange::legacy_equal_p): Check type when
14875 comparing VR_VARYING types.
14876 (range_tests_legacy): Test comparing VARYING ranges of different
14879 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
14881 * config/aarch64/aarch64.c (neoversen1_tunings):
14882 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
14884 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
14886 * config/aarch64/aarch64.c (aarch64_classify_symbol): Use GOT for
14887 extern weak symbols. Limit symbol offsets for non-GOT symbols with
14890 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
14892 * config/arm/neon.md (vec_load_lanesxi<mode>)
14893 (vec_store_lanexoi<mode>): Move ...
14894 * config/arm/vec-common.md: here.
14896 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
14898 * config/arm/neon.md (vec_load_lanesoi<mode>)
14899 (vec_store_lanesoi<mode>): Move ...
14900 * config/arm/vec-common.md: here.
14902 2021-05-24 liuhongt <hongtao.liu@intel.com>
14905 * config/i386/i386.c (ix86_gimple_fold_builtin): Replacing
14906 stmt with GIMPLE_NOP when lhs doesn't exist.
14908 2021-05-23 Uroš Bizjak <ubizjak@gmail.com>
14911 * config/i386/mmx.md (*push<VI_32:mode>2_rex64):
14912 New instruction pattern.
14913 (*push<VI_32:mode>2): Ditto.
14914 (push splitter for SSE registers): New splitter.
14916 2021-05-23 Andrew Pinski <apinski@marvell.com>
14918 * match.pd ((A & C) != 0 ? D : 0): Limit to non pointer types.
14920 2021-05-22 Aaron Sawdey <acsawdey@linux.ibm.com>
14922 * config/rs6000/genfusion.pl (gen_addadd): Fix incorrect attr types.
14923 * config/rs6000/fusion.md: Regenerate file.
14925 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
14927 * config/rs6000/genfusion.pl (gen_addadd): New function.
14928 * config/rs6000/fusion.md: Regenerate file.
14929 * config/rs6000/rs6000-cpus.def: Add
14930 OPTION_MASK_P10_FUSION_2ADD to masks.
14931 * config/rs6000/rs6000.c (rs6000_option_override_internal):
14932 Handle default value of OPTION_MASK_P10_FUSION_2ADD.
14933 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2add.
14935 2021-05-21 Jakub Jelinek <jakub@redhat.com>
14937 PR middle-end/99928
14938 * tree.h (OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET): Define.
14939 * gimplify.c (enum gimplify_omp_var_data): Fix up
14940 GOVD_MAP_HAS_ATTACHMENTS value, add GOVD_FIRSTPRIVATE_IMPLICIT.
14941 (omp_lastprivate_for_combined_outer_constructs): If combined target
14942 has GOVD_FIRSTPRIVATE_IMPLICIT set for the decl, change it to
14943 GOVD_MAP | GOVD_SEEN.
14944 (gimplify_scan_omp_clauses): Set GOVD_FIRSTPRIVATE_IMPLICIT for
14945 firstprivate clauses with OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT.
14946 (gimplify_adjust_omp_clauses): For firstprivate clauses with
14947 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT either clear that bit and
14948 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET too, or remove it and
14949 let it be replaced by implicit map clause.
14951 2021-05-21 Jakub Jelinek <jakub@redhat.com>
14953 PR middle-end/99928
14954 * gimplify.c (omp_lastprivate_for_combined_outer_constructs): New
14956 (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LASTPRIVATE>: Use it.
14957 (gimplify_omp_for): Likewise.
14959 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14961 PR middle-end/90115
14962 * omp-low.c (oacc_privatization_candidate_p): Reject 'static',
14963 'external' in blocks.
14965 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14967 PR middle-end/90115
14968 * flag-types.h (enum openacc_privatization): New.
14969 * params.opt (-param=openacc-privatization): New.
14970 * doc/invoke.texi (openacc-privatization): Document it.
14971 * omp-general.h (get_openacc_privatization_dump_flags): New
14973 * omp-low.c (oacc_privatization_candidate_p): Add diagnostics.
14974 * omp-offload.c (execute_oacc_device_lower)
14975 <IFN_UNIQUE_OACC_PRIVATE>: Re-work diagnostics.
14976 * target.def (goacc.adjust_private_decl): Add 'location_t'
14978 * doc/tm.texi: Regenerate.
14979 * config/gcn/gcn-protos.h (gcn_goacc_adjust_private_decl): Adjust.
14980 * config/gcn/gcn-tree.c (gcn_goacc_adjust_private_decl): Likewise.
14981 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl):
14982 Likewise. Preserve it for...
14983 (nvptx_goacc_expand_var_decl): ... use here.
14985 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14987 * doc/sourcebuild.texi (Other attributes): Document '__OPTIMIZE__'
14990 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
14992 PR middle-end/90115
14993 * omp-low.c (oacc_privatization_candidate_p): New function.
14994 (oacc_privatization_scan_clause_chain)
14995 (oacc_privatization_scan_decl_chain): Use it. Also
14996 'gcc_checking_assert' that we're not seeing duplicates.
14998 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15000 PR middle-end/90115
15001 * omp-offload.c (execute_oacc_device_lower): Skip processing if no
15004 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15006 PR middle-end/90115
15007 * omp-offload.c (execute_oacc_device_lower): Explain.
15009 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15011 PR middle-end/90115
15012 * omp-offload.c (execute_oacc_device_lower)
15013 <IFN_UNIQUE_OACC_PRIVATE>: Diagnose and handle for 'level == -1'
15015 * internal-fn.c (expand_UNIQUE): Don't expect
15016 'IFN_UNIQUE_OACC_PRIVATE'.
15018 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15020 PR middle-end/90115
15021 * omp-low.c (lower_omp_for): Don't evaluate OpenMP 'for' clauses.
15023 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15025 PR middle-end/90115
15026 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl)
15027 (nvptx_goacc_expand_var_decl): Tighten.
15029 2021-05-21 Julian Brown <julian@codesourcery.com>
15030 Chung-Lin Tang <cltang@codesourcery.com>
15031 Thomas Schwinge <thomas@codesourcery.com>
15033 PR middle-end/90115
15034 * doc/tm.texi.in (TARGET_GOACC_EXPAND_VAR_DECL)
15035 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Add documentation hooks.
15036 * doc/tm.texi: Regenerate.
15037 * expr.c (expand_expr_real_1): Expand decls using the
15038 expand_var_decl OpenACC hook if defined.
15039 * internal-fn.c (expand_UNIQUE): Handle IFN_UNIQUE_OACC_PRIVATE.
15040 * internal-fn.h (IFN_UNIQUE_CODES): Add OACC_PRIVATE.
15041 * omp-low.c (omp_context): Add oacc_privatization_candidates
15043 (lower_oacc_reductions): Add PRIVATE_MARKER parameter. Insert
15045 (lower_oacc_head_tail): Add PRIVATE_MARKER parameter. Modify
15046 private marker's gimple call arguments, and pass it to
15047 lower_oacc_reductions.
15048 (oacc_privatization_scan_clause_chain)
15049 (oacc_privatization_scan_decl_chain, lower_oacc_private_marker):
15051 (lower_omp_for, lower_omp_target, lower_omp_1): Use these.
15052 * omp-offload.c (convert.h): Include.
15053 (oacc_loop_xform_head_tail): Treat private-variable markers like
15054 fork/join when transforming head/tail sequences.
15055 (struct var_decl_rewrite_info): Add struct.
15056 (oacc_rewrite_var_decl, is_sync_builtin_call): New functions.
15057 (execute_oacc_device_lower): Support rewriting gang-private
15058 variables using target hook, and fix up addr_expr and var_decl
15060 * target.def (adjust_private_decl, expand_var_decl): New hooks.
15061 * config/gcn/gcn-protos.h (gcn_goacc_adjust_gangprivate_decl):
15063 (gcn_goacc_adjust_private_decl): ...this.
15064 * config/gcn/gcn-tree.c (gcn_goacc_adjust_gangprivate_decl):
15066 (gcn_goacc_adjust_private_decl): ...this. Add LEVEL parameter.
15067 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_GANGPRIVATE_DECL): Rename
15068 definition using gcn_goacc_adjust_gangprivate_decl...
15069 (TARGET_GOACC_ADJUST_PRIVATE_DECL): ...to this, using
15070 gcn_goacc_adjust_private_decl.
15071 * config/nvptx/nvptx.c (tree-pretty-print.h): Include.
15072 (gang_private_shared_size): New global variable.
15073 (gang_private_shared_align): Likewise.
15074 (gang_private_shared_sym): Likewise.
15075 (gang_private_shared_hmap): Likewise.
15076 (nvptx_option_override): Initialize these.
15077 (nvptx_file_end): Output gang_private_shared_sym.
15078 (nvptx_goacc_adjust_private_decl, nvptx_goacc_expand_var_decl):
15080 (nvptx_set_current_function): Clear gang_private_shared_hmap.
15081 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Define hook.
15082 (TARGET_GOACC_EXPAND_VAR_DECL): Likewise.
15084 2021-05-21 H.J. Lu <hjl.tools@gmail.com>
15086 * config/i386/i386-modes.def (MAX_BITSIZE_MODE_ANY_INT): Removed.
15088 2021-05-21 Richard Biener <rguenther@suse.de>
15089 H.J. Lu <hjl.tools@gmail.com>
15091 PR middle-end/90773
15092 * expr.c (expand_constructor): Elide expand_constructor if
15093 move by pieces is preferred.
15095 2021-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15097 * config/aarch64/aarch64-builtins.c (aarch64_call_properties):
15098 Take a flag and mode value as arguments.
15099 (aarch64_modifies_global_state_p): Likewise.
15100 (aarch64_reads_global_state_p): Likewise.
15101 (aarch64_could_trap_p): Likewise.
15102 (aarch64_get_attributes): Likewise.
15103 (aarch64_init_simd_builtins): Adjust callsite of above.
15104 (aarch64_init_fcmla_laneq_builtins): Use aarch64_get_attributes to get
15105 function attributes to apply to builtins.
15106 (aarch64_init_crc32_builtins): Likewise.
15107 (aarch64_init_builtin_rsqrt): Likewise.
15109 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
15111 * config/rs6000/rs6000.md (define_attr "type"): Add types for fusion.
15112 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10): Use new fusion types.
15113 (gen_2logical): Use new fusion types.
15114 * config/rs6000/fusion.md: Regenerate.
15116 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
15119 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
15120 Handle V4QI and V2HI modes.
15121 (ix86_expand_sse_movcc): Ditto.
15122 * config/i386/mmx.md (*<sat_plusminus:insn><VI_32:mode>3):
15123 New instruction pattern.
15124 (*eq<VI_32:mode>3): Ditto.
15125 (*gt<VI_32:mode>3): Ditto.
15126 (*xop_pcmov_<VI_32:mode>): Ditto.
15127 (mmx_pblendvb32): Ditto.
15128 (mmx_pblendvb64): Rename from mmx_pblendvb.
15129 (vec_cmp<VI_32:mode><VI_32:mode>): New expander.
15130 (vec_cmpu<VI_32:mode><VI_32:mode>): Ditto.
15131 (vcond<VI_32:mode><VI_32:mode>): Ditto.
15132 (vcondu<VI_32:mode><VI_32:mode>): Ditto.
15133 (vcond_mask_<VI_32:mode><VI_32:mode>): Ditto.
15135 2021-05-21 Jakub Jelinek <jakub@redhat.com>
15137 PR tree-optimization/94589
15138 * tree-ssa-phiopt.c (spaceship_replacement): For integral rhs1 and
15139 rhs2, treat x <= 4 equivalently to x < 5 etc. In cmp1 and cmp2 (if
15140 not the same as cmp3) treat <= the same as < and >= the same as >.
15141 Don't require that cond2_phi_edge is true edge, instead take
15142 false/true edges into account based on cmp1/cmp2 comparison kinds.
15144 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
15147 * config/i386/mmx.md (SMAXMIN_MMXMODEI): New mode iterator.
15148 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Macroize expander
15149 from <smaxmin:code>v4hi3> and <smaxmin:code><MMXMODE14:mode>3
15150 using SMAXMIN_MMXMODEI mode iterator.
15151 (*<smaxmin:code>v4qi3): New insn pattern.
15152 (*<smaxmin:code>v2hi3): Ditto.
15153 (SMAXMIN_VI_32): New mode iterator.
15154 (<smaxmin:code><SMAXMIN_VI_32>mode3): New expander.
15155 (UMAXMIN_MMXMODEI): New mode iterator.
15156 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Macroize expander
15157 from <umaxmin:code>v8qi3> and <umaxmin:code><MMXMODE24:mode>3
15158 using UMAXMIN_MMXMODEI mode iterator.
15159 (*<umaxmin:code>v4qi3): New insn pattern.
15160 (*<umaxmin:code>v2hi3): Ditto.
15161 (UMAXMIN_VI_32): New mode iterator.
15162 (<umaxmin:code><UMAXMIN_VI_32>mode3): New expander.
15163 (abs<VI_32:mode>2): New insn pattern.
15164 (ssse3_abs<MMXMODEI:mode>2, abs<MMXMODEI:mode>2): Move from ...
15165 * config/i386/sse.md: ... here.
15167 2021-05-20 Clement Chigot <clement.chigot@atos.net>
15168 David Edelsohn <dje.gcc@gmail.com>
15170 * collect2.c (scan_prog_file): Issue non-fatal warning for
15173 2021-05-20 Jonathan Wakely <jwakely@redhat.com>
15175 * doc/invoke.texi (-Wno-c++11-extensions)
15176 (-Wno-c++14-extensions, -Wno-c++17-extensions)
15177 (-Wno-c++20-extensions, -Wno-c++23-extensions): Document
15180 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
15182 * config/c6x/c6x.c (c6x_output_file_unwind): Use dwarf_debuginfo_p.
15183 * config/darwin.c (darwin_override_options): Likewise.
15184 * config/i386/cygming.h (DBX_REGISTER_NUMBER): Likewise.
15185 * config/i386/darwin.h (DBX_REGISTER_NUMBER): Likewise.
15186 (DWARF2_FRAME_REG_OUT): Likewise.
15187 * config/mips/mips.c (mips_output_filename): Likewise.
15188 * config/rs6000/rs6000.c (rs6000_xcoff_declare_function_name):
15190 (rs6000_dbx_register_number): Likewise.
15191 * dbxout.c: Include flags.h.
15192 * dwarf2cfi.c (cfi_label_required_p): Likewise.
15193 (dwarf2out_do_frame): Likewise.
15194 * except.c: Include flags.h.
15195 * final.c (dwarf2_debug_info_emitted_p): Likewise.
15196 (final_scan_insn_1): Likewise.
15197 * flags.h (dwarf_debuginfo_p): New function declaration.
15198 * opts.c (dwarf_debuginfo_p): New function definition.
15199 * targhooks.c (default_debug_unwind_info): Use dwarf_debuginfo_p.
15200 * toplev.c (process_options): Likewise.
15202 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
15204 * common.opt: Change type to support bitmasks.
15205 * flag-types.h (enum debug_info_type): Rename enumerator constants.
15206 (NO_DEBUG): New bitmask.
15207 (DBX_DEBUG): Likewise.
15208 (DWARF2_DEBUG): Likewise.
15209 (XCOFF_DEBUG): Likewise.
15210 (VMS_DEBUG): Likewise.
15211 (VMS_AND_DWARF2_DEBUG): Likewise.
15212 * flags.h (debug_set_to_format): New function declaration.
15213 (debug_set_count): Likewise.
15214 (debug_set_names): Likewise.
15215 * opts.c (debug_type_masks): Array of bitmasks for debug formats.
15216 (debug_set_to_format): New function definition.
15217 (debug_set_count): Likewise.
15218 (debug_set_names): Likewise.
15219 (set_debug_level): Update access to debug_type_names.
15220 * toplev.c: Likewise.
15222 2021-05-20 Martin Sebor <msebor@redhat.com>
15224 PR middle-end/100684
15225 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Handle C++ lambda.
15227 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
15230 * config/i386/i386.md (isa): Remove x64_bmi.
15231 (enabled): Remove x64_bmi.
15232 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
15233 Remove general register alternative.
15234 (*andnot<VI_32:mode>3): Ditto.
15235 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
15236 (*<any_logic:code><VI_32:mode>3): Ditto.
15238 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
15240 * config/arm/arm.c: Include head files tree-vectorizer.h and
15243 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
15246 * config/i386/mmx.md (Yv_Yw): Revert adding V4QI and V2HI modes.
15247 (*<plusminus:insn><VI_32:mode>3): Use Yw instad of <Yv_Yw> constrint.
15248 (<s>mulv4hi3_highpart): New expander.
15249 (*<s>mulv2hi3_highpart): New insn pattern.
15250 (<s>mulv2hi3_higpart): New expander.
15251 (*<any_shift:insn>v2hi3): New insn pattern.
15252 (<any_shift:insn>v2hi3): New expander.
15253 * config/i386/sse.md (smulhrsv2hi3): New expander.
15254 (*smulhrsv2hi3): New insn pattern.
15256 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
15258 * doc/invoke.texi (vect-inner-loop-cost-factor): Document new
15260 * params.opt (vect-inner-loop-cost-factor): New.
15261 * targhooks.c (default_add_stmt_cost): Replace hardcoded factor
15262 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR, include head file
15263 tree-vectorizer.h and its required ones.
15264 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Replace
15265 hardcoded factor 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR.
15266 * config/arm/arm.c (arm_add_stmt_cost): Likewise.
15267 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
15268 * config/rs6000/rs6000.c (rs6000_add_stmt_cost): Likewise.
15269 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
15271 (_loop_vec_info::_loop_vec_info): Init inner_loop_cost_factor.
15272 * tree-vectorizer.h (_loop_vec_info): Add inner_loop_cost_factor.
15273 (LOOP_VINFO_INNER_LOOP_COST_FACTOR): New macro.
15275 2021-05-20 Christophe Lyon <christophe.lyon@linaro.org>
15276 Torbjörn Svensson <torbjorn.svensson@st.com>
15279 * doc/cpp.texi (Common Predefined Macros): Document __FILE_NAME__.
15281 2021-05-20 Jakub Jelinek <jakub@redhat.com>
15283 PR middle-end/99928
15284 * gimplify.c (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LINEAR>: For
15285 explicit linear clause when combined with target, make it map(tofrom:)
15286 instead of no clause or firstprivate.
15288 2021-05-20 Jakub Jelinek <jakub@redhat.com>
15290 PR tree-optimization/94589
15291 * match.pd ((X & Y) == X -> (X & ~Y) == 0): Simplify even in presence
15292 of integral conversions.
15294 2021-05-19 Andrew MacLeod <amacleod@redhat.com>
15296 * gimple-range.cc (fur_source::get_operand): New.
15297 (gimple_range_fold): Delete.
15298 (fold_using_range::fold_stmt): Move from gimple_ranger::calc_stmt.
15299 (fold_using_range::range_of_range_op): Move from gimple_ranger.
15300 (fold_using_range::range_of_address): Ditto.
15301 (fold_using_range::range_of_phi): Ditto.
15302 (fold_using_range::range_of_call): Ditto.
15303 (fold_using_range::range_of_builtin_ubsan_call): Move from
15304 range_of_builtin_ubsan_call.
15305 (fold_using_range::range_of_builtin_call): Move from
15306 range_of_builtin_call.
15307 (gimple_ranger::range_of_builtin_call): Delete.
15308 (fold_using_range::range_of_cond_expr): Move from gimple_ranger.
15309 (gimple_ranger::fold_range_internal): New.
15310 (gimple_ranger::range_of_stmt): Use new fold_using_range API.
15311 (fold_using_range::range_of_ssa_name_with_loop_info): Move from
15312 gimple_ranger. Improve ranges of SSA_NAMES when possible.
15313 * gimple-range.h (gimple_ranger): Remove various range_of routines.
15314 (class fur_source): New.
15315 (class fold_using_range): New.
15316 (fur_source::fur_source): New.
15318 * vr-values.c (vr_values::extract_range_basic): Use fold_using_range
15319 instead of range_of_builtin_call.
15321 2021-05-19 Jonathan Wakely <jwakely@redhat.com>
15323 * doc/cpp.texi (Common Predefined Macros): Update documentation
15324 for the __GXX_EXPERIMENTAL_CXX0X__ macro.
15326 2021-05-19 Alex Coplan <alex.coplan@arm.com>
15329 * config/arm/arm.md (nonsecure_call_internal): Always ensure
15330 callee's address is in a register.
15332 2021-05-19 Geng Qi <gengqi@linux.alibaba.com>
15334 * common/config/riscv/riscv-common.c
15335 (riscv_subset_list::parsing_subset_version): Properly parse the letter
15337 (riscv_subset_list::parse_std_ext,
15338 riscv_subset_list::parse_multiletter_ext): To handle errors generated
15339 in riscv_subset_list::parsing_subset_version.
15341 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15343 * config/aarch64/aarch64-simd.md: Use "neon_move_narrow_q"
15344 type attribute in patterns generating XTN(2).
15346 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15348 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_pack_trunc_<mode>):
15349 Remove as duplicate of...
15350 (aarch64_xtn<mode>): This.
15351 (aarch64_xtn2<mode>_le): Move position in file.
15352 (aarch64_xtn2<mode>_be): Move position in file.
15353 (aarch64_xtn2<mode>): Move position in file.
15354 (vec_pack_trunc_<mode>): Define as an expander.
15356 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15358 * config/aarch64/aarch64-simd-builtins.def: Split builtin
15359 generation for aarch64_<sur>q<r>shr<u>n_n<mode> pattern into
15360 separate scalar and vector generators.
15361 * config/aarch64/aarch64-simd.md
15362 (aarch64_<sur>q<r>shr<u>n_n<mode>): Define as an expander and
15364 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_le): This and...
15365 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_be): This.
15366 * config/aarch64/iterators.md: Define SD_HSDI iterator.
15368 2021-05-19 Jonathn Wright <jonathan.wright@arm.com>
15370 * config/aarch64/aarch64-simd.md: Use UNSPEC_SQXTUN instead
15372 * config/aarch64/iterators.md: Remove UNSPEC_SQXTUN2.
15374 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15376 * config/aarch64/aarch64-simd.md (aarch64_<sur>q<r>shr<u>n2_n<mode>):
15377 Implement as an expand emitting a big/little endian
15378 instruction pattern.
15379 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_le): Define.
15380 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_be): Define.
15382 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15384 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn2<mode>):
15385 Implement as an expand emitting a big/little endian
15386 instruction pattern.
15387 (aarch64_<sur><addsub>hn2<mode>_insn_le): Define.
15388 (aarch64_<sur><addsub>hn2<mode>_insn_be): Define.
15389 * config/aarch64/iterators.md: Remove UNSPEC_[R]ADDHN2 and
15390 UNSPEC_[R]SUBHN2 unspecs and ADDSUBHN2 iterator.
15392 2021-05-19 Richard Biener <rguenther@suse.de>
15394 PR middle-end/100672
15395 * fold-const.c (fold_negate_expr_1): Use element_precision.
15396 (negate_expr_p): Likewise.
15398 2021-05-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
15400 * config/aarch64/iterators.md (SVE_PRED_LOAD): New iterator.
15401 (pred_load): New int attribute.
15402 * config/aarch64/aarch64-sve.md
15403 (aarch64_load_<ANY_EXTEND:optab><SVE_HSDI:mode><SVE_PARTIAL_I:mode>): Use
15404 SVE_PRED_LOAD enum iterator and corresponding pred_load attribute.
15405 * config/aarch64/aarch64-sve-builtins-base.cc (expand): Update call to
15406 code_for_aarch64_load.
15408 2021-05-19 Richard Biener <rguenther@suse.de>
15410 * cfgexpand.c (discover_nonconstant_array_refs_r): Make
15411 sure TARGET_MEM_REF bases are expanded as memory.
15412 * tree-ssa-operands.c (operands_scanner::get_tmr_operands):
15413 Do not mark TARGET_MEM_REF bases addressable.
15414 * tree-ssa.c (non_rewritable_mem_ref_base): Handle
15415 TARGET_MEM_REF bases as never rewritable.
15416 * gimple-walk.c (walk_stmt_load_store_addr_ops): Do not
15417 walk TARGET_MEM_REF bases as address-takens.
15418 * tree-ssa-dce.c (ref_may_be_aliased): Handle TARGET_MEM_REF.
15420 2021-05-19 Richard Biener <rguenther@suse.de>
15422 * builtins.c (get_object_alignment_1): Strip outer
15424 * tree-dfa.c (get_ref_base_and_extent): Handle outer
15425 WITH_SIZE_EXPR for size processing and process the
15427 * tree-ssa-alias.c (ao_ref_base_alias_set): Strip
15428 outer WITH_SIZE_EXPR.
15429 (ao_ref_base_alias_ptr_type): Likewise.
15430 (refs_may_alias_p_2): Allow WITH_SIZE_EXPR in ref->ref
15431 and handle that accordingly, stripping it for the
15432 core alias workers.
15433 * tree.c (get_base_address): Handle WITH_SIZE_EXPR by
15434 looking through it instead of returning NULL.
15436 2021-05-19 Jakub Jelinek <jakub@redhat.com>
15438 PR middle-end/100576
15439 * builtins.c (check_read_access): Convert bound to size_type_node if
15442 2021-05-19 Richard Biener <rguenther@suse.de>
15444 * tree-cfg.c (verify_types_in_gimple_min_lval): Inline...
15445 (verify_types_in_gimple_reference): ... here. Sanitize.
15446 (verify_gimple_call): Verify references in LHS and arguments.
15447 (verify_gimple_assign_single): Reject WITH_SIZE_EXPR.
15449 2021-05-19 Uroš Bizjak <ubizjak@gmail.com>
15451 * config/i386/i386.h (VALID_INT_MODE_P):
15452 Add V8QI, V4HI and V2SI modes for TARGET_64BIT.
15453 * config/i386/i386.md (isa): Add x64_bmi.
15454 (enabled): Handle x64_bmi.
15455 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
15456 Add alternative using 64bit general registers.
15457 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
15459 2021-05-19 Jakub Jelinek <jakub@redhat.com>
15461 PR middle-end/99928
15462 * tree.h (OMP_MASTER_COMBINED): Define.
15463 * gimplify.c (gimplify_scan_omp_clauses): Rewrite lastprivate
15464 handling for outer combined/composite constructs to a loop.
15465 Handle lastprivate on combined target.
15466 (gimplify_expr): Formatting fix.
15468 2021-05-19 Xionghu Luo <luoxhu@linux.ibm.com>
15470 * passes.def: Add sink_code pass before store_merging.
15471 * tree-ssa-sink.c (pass_sink_code:clone): New.
15473 2021-05-18 Bill Schmidt <wschmidt@linux.ibm.com>
15475 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Remove call to
15476 rs6000_special_adjust_field_align_p.
15477 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
15478 * config/rs6000/rs6000-call.c (rs6000_function_arg_boundary):
15479 Remove ABI warning.
15480 (rs6000_function_arg): Likewise.
15481 * config/rs6000/rs6000-protos.h
15482 (rs6000_special_adjust_field_align_p): Remove prototype.
15483 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
15485 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Remove call to
15486 rs6000_special_adjust_field_align_p.
15488 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
15491 * config/i386/i386.h (VALID_SSE2_REG_MODE):
15492 Add V4QI and V2HI modes.
15493 (VALID_INT_MODE_P): Ditto.
15494 * config/i386/mmx.md (VI_32): New mode iterator.
15495 (mmxvecsize): Handle V4QI and V2HI.
15497 (mov<VI_32:mode>): New expander.
15498 (*mov<mode>_internal): New insn pattern.
15499 (movmisalign<VI_32:mode>): New expander.
15500 (neg<VI_32:mode>): New expander.
15501 (<plusminus:insn><VI_32:mode>3): New expander.
15502 (*<plusminus:insn><VI_32:mode>3): New insn pattern.
15503 (mulv2hi3): New expander.
15504 (*mulv2hi3): New insn pattern.
15505 (one_cmpl<VI_32:mode>2): New expander.
15506 (*andnot<VI_32:mode>3): New insn pattern.
15507 (<any_logic:code><VI_32:mode>3): New expander.
15508 (*<any_logic:code><VI_32:mode>3): New insn pattern.
15510 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
15512 * config/i386/sse.md (<any_extend:insn>v4qiv4di2):
15513 Fix a mode mismatch with operand 1.
15515 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
15518 * config/i386/i386-expand.c (split_double_mode): Return
15519 temporary register when simplify_gen_subreg fails with
15520 the high half od the paradoxical subreg.
15522 2021-05-18 Richard Biener <rguenther@suse.de>
15524 * cfgexpand.c (expand_one_var): Pass in forced_stack_var
15525 and honor it when expanding.
15526 (expand_used_vars_for_block): Pass through forced_stack_var.
15527 (expand_used_vars): Likewise.
15528 (discover_nonconstant_array_refs_r): Set bits in
15529 forced_stack_vars instead of marking vars TREE_ADDRESSABLE.
15530 (avoid_type_punning_on_regs): Likewise.
15531 (discover_nonconstant_array_refs): Likewise.
15532 (pass_expand::execute): Create and pass down forced_stack_var
15533 bitmap. For parameters and returns temporarily set
15534 TREE_ADDRESSABLE when expand_function_start.
15536 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
15538 * doc/sourcebuild.texi: Document 'dg-note'.
15540 2021-05-18 Tobias Burnus <tobias@codesourcery.com>
15543 * configure: Regenerate.
15544 * configure.ac (BUILD_CFLAG, BUILD_CXXFLAGS): Add $(CFLAGS-$@).
15546 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
15548 * gimple.h (is_gimple_omp_oacc): Tighten.
15549 * omp-low.c (check_omp_nesting_restrictions): Adjust.
15551 2021-05-18 Richard Biener <rguenther@suse.de>
15553 * tree-ssa-operands.c (mark_address_taken): Simplify.
15555 2021-05-18 Martin Liska <mliska@suse.cz>
15557 * config/gcn/mkoffload.c (STR): Redefine.
15558 * config/i386/intelmic-mkoffload.c (STR): Likewise.
15559 * config/nvptx/mkoffload.c (STR): Likewise.
15561 2021-05-18 Martin Liska <mliska@suse.cz>
15563 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
15564 Use startswith function instead of strncmp.
15565 * common/config/bfin/bfin-common.c (bfin_handle_option): Likewise.
15566 * common/config/riscv/riscv-common.c (riscv_subset_list::parse): Likewise.
15567 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): Likewise.
15568 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Likewise.
15569 * config/alpha/alpha.c (alpha_elf_section_type_flags): Likewise.
15570 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
15571 * config/arm/arm.c (arm_file_start): Likewise.
15572 (arm_valid_target_attribute_rec): Likewise.
15573 (thumb1_md_asm_adjust): Likewise.
15574 * config/arm/driver-arm.c (host_detect_local_cpu): Likewise.
15575 * config/avr/avr.c (STR_PREFIX_P): Likewise.
15576 (avr_set_current_function): Likewise.
15577 (avr_handle_addr_attribute): Likewise.
15578 (avr_asm_output_aligned_decl_common): Likewise.
15579 (avr_asm_named_section): Likewise.
15580 (avr_section_type_flags): Likewise.
15581 (avr_asm_select_section): Likewise.
15582 * config/c6x/c6x.c (c6x_in_small_data_p): Likewise.
15583 (c6x_section_type_flags): Likewise.
15584 * config/darwin-c.c (darwin_cfstring_ref_p): Likewise.
15585 (darwin_objc_declare_unresolved_class_reference): Likewise.
15586 (darwin_objc_declare_class_definition): Likewise.
15587 * config/darwin.c (indirect_data): Likewise.
15588 (darwin_encode_section_info): Likewise.
15589 (darwin_objc2_section): Likewise.
15590 (darwin_objc1_section): Likewise.
15591 (machopic_select_section): Likewise.
15592 (darwin_globalize_label): Likewise.
15593 (darwin_label_is_anonymous_local_objc_name): Likewise.
15594 (darwin_asm_named_section): Likewise.
15595 (darwin_asm_output_dwarf_offset): Likewise.
15596 * config/frv/frv.c (frv_string_begins_with): Likewise.
15597 (frv_in_small_data_p): Likewise.
15598 * config/gcn/mkoffload.c (STR): Likewise.
15600 * config/i386/i386-builtins.c (get_builtin_code_for_version): Likewise.
15601 * config/i386/i386-options.c (ix86_option_override_internal): Likewise.
15602 * config/i386/i386.c (x86_64_elf_section_type_flags): Likewise.
15603 (ix86_md_asm_adjust): Likewise.
15604 * config/i386/intelmic-mkoffload.c (STR): Likewise.
15605 * config/i386/winnt.c (i386_pe_asm_named_section): Likewise.
15606 (i386_pe_file_end): Likewise.
15607 * config/ia64/ia64.c (ia64_in_small_data_p): Likewise.
15608 (ia64_section_type_flags): Likewise.
15609 * config/mips/driver-native.c (host_detect_local_cpu): Likewise.
15610 * config/mips/mips.c (mips_handle_interrupt_attr): Likewise.
15611 (mips16_stub_function_p): Likewise.
15612 (mips_function_rodata_section): Likewise.
15613 * config/msp430/msp430.c (msp430_mcu_name): Likewise.
15614 (msp430_function_section): Likewise.
15615 (msp430_section_type_flags): Likewise.
15616 (msp430_expand_helper): Likewise.
15617 * config/nios2/nios2.c (nios2_small_section_name_p): Likewise.
15618 (nios2_valid_target_attribute_rec): Likewise.
15619 * config/nvptx/mkoffload.c (process): Likewise.
15621 * config/pa/som.h: Likewise.
15622 * config/pdp11/pdp11.c (pdp11_output_ident): Likewise.
15623 * config/riscv/riscv.c (riscv_elf_select_rtx_section): Likewise.
15624 * config/rs6000/rs6000.c (VTABLE_NAME_P): Likewise.
15625 (rs6000_inner_target_options): Likewise.
15626 * config/s390/driver-native.c (s390_host_detect_local_cpu): Likewise.
15627 * config/sparc/driver-sparc.c (host_detect_local_cpu): Likewise.
15628 * config/vax/vax.c (vax_output_int_move): Likewise.
15629 * config/vms/vms-ld.c (startswith): Likewise.
15630 (process_args): Likewise.
15632 * config/vms/vms.c: Likewise.
15634 2021-05-18 Jakub Jelinek <jakub@redhat.com>
15636 PR rtl-optimization/100590
15637 * regcprop.c (copyprop_hardreg_forward_1): Only DCE dead sets if
15638 they are NONJUMP_INSN_P.
15640 2021-05-18 Jakub Jelinek <jakub@redhat.com>
15643 * function.c (push_dummy_function): Set DECL_ARTIFICIAL and
15644 DECL_ASSEMBLER_NAME on the fn_decl.
15646 2021-05-18 Jakub Jelinek <jakub@redhat.com>
15648 PR tree-optimization/94589
15649 * tree-ssa-phiopt.c (spaceship_replacement): Pattern match
15650 phi result used in (res & ~1) == 0 comparison as res >= 0 as
15651 res == 2 would be UB with -ffinite-math-only.
15653 2021-05-18 Martin Liska <mliska@suse.cz>
15655 * Makefile.in: genversion.o should depend on DATESTAMP.
15657 2021-05-18 Claudiu Zissulescu <claziss@synopsys.com>
15659 * config/arc/simdext.md (negv2si2): Remove round bracket.
15661 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
15663 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Define
15664 _Bool as macro expanding to _Bool.
15666 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
15669 * tree.c (build_reference_type_for_mode)
15670 (build_pointer_type_for_mode): Pick pointer mode if MODE argument
15672 (build_reference_type, build_pointer_type): Invoke
15673 build_*_type_for_mode with VOIDmode.
15675 2021-05-17 Andrew MacLeod <amacleod@redhat.com>
15677 PR tree-optimization/100512
15678 * gimple-range-cache.cc (ranger_cache::set_global_range): Mark const
15679 and non-zero pointer ranges as invariant.
15680 * gimple-range.cc (gimple_ranger::range_of_stmt): Remove pointer
15681 processing from here.
15683 2021-05-17 Tom de Vries <tdevries@suse.de>
15686 * config/nvptx/nvptx-protos.h (nvptx_output_atomic_insn): Declare
15687 * config/nvptx/nvptx.c (nvptx_output_barrier)
15688 (nvptx_output_atomic_insn): New function.
15689 (nvptx_print_operand): Add support for 'B'.
15690 * config/nvptx/nvptx.md: Use nvptx_output_atomic_insn for atomic
15693 2021-05-17 Aldy Hernandez <aldyh@redhat.com>
15695 PR tree-optimization/100349
15696 * vr-values.c (bounds_of_var_in_loop): Bail if scev returns
15699 2021-05-17 Tamar Christina <tamar.christina@arm.com>
15701 * config/aarch64/driver-aarch64.c (DEFAULT_ARCH): New.
15702 (host_detect_local_cpu): Use it.
15704 2021-05-17 Martin Liska <mliska@suse.cz>
15706 * doc/invoke.texi: Add 2 missing dots.
15708 2021-05-17 Marius Hillenbrand <mhillen@linux.ibm.com>
15710 PR bootstrap/100552
15711 * configure.ac: Replace pattern substitution with call to sed.
15712 * configure: Regenerate.
15714 2021-05-17 Richard Biener <rguenther@suse.de>
15716 PR middle-end/100582
15717 * tree.c (array_at_struct_end_p): Get to the base of the
15718 reference before looking for the underlying decl.
15720 2021-05-17 Joern Rennecke <joern.rennecke@embecosm.com>
15722 * genoutput.c (validate_insn_alternatives) Make "wrong number of
15723 alternatives" message more specific, and remove assumption on where
15726 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
15728 * config/arm/iterators.md (V16): New iterator.
15729 (VH_cvtto): New iterator.
15730 (v_cmp_result): Added V4HF and V8HF support.
15731 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result>): Use VDQWH.
15732 (vcond<mode><mode>): Likewise.
15733 (vcond_mask_<mode><v_cmp_result>): Likewise.
15734 (vcond<VH_cvtto><mode>): New expander.
15736 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
15738 * config/arm/arm-protos.h (arm_expand_vector_compare): Update
15740 * config/arm/arm.c (arm_expand_vector_compare): Add support for
15742 (arm_expand_vcond): Likewise.
15743 * config/arm/iterators.md (supf): Remove VCMPNEQ_S, VCMPEQQ_S,
15744 VCMPEQQ_N_S, VCMPNEQ_N_S.
15745 (VCMPNEQ, VCMPEQQ, VCMPEQQ_N, VCMPNEQ_N): Remove.
15746 * config/arm/mve.md (@mve_vcmp<mve_cmp_op>q_<mode>): Add '@' prefix.
15747 (@mve_vcmp<mve_cmp_op>q_f<mode>): Likewise.
15748 (@mve_vcmp<mve_cmp_op>q_n_f<mode>): Likewise.
15749 (@mve_vpselq_<supf><mode>): Likewise.
15750 (@mve_vpselq_f<mode>"): Likewise.
15751 * config/arm/neon.md (vec_cmp<mode><v_cmp_result): Enable for MVE
15752 and move to vec-common.md.
15753 (vec_cmpu<mode><mode>): Likewise.
15754 (vcond<mode><mode>): Likewise.
15755 (vcond<V_cvtto><mode>): Likewise.
15756 (vcondu<mode><v_cmp_result>): Likewise.
15757 (vcond_mask_<mode><v_cmp_result>): Likewise.
15758 * config/arm/unspecs.md (VCMPNEQ_U, VCMPNEQ_S, VCMPEQQ_S)
15759 (VCMPEQQ_N_S, VCMPNEQ_N_S, VCMPEQQ_U, CMPEQQ_N_U, VCMPNEQ_N_U)
15760 (VCMPGEQ_N_S, VCMPGEQ_S, VCMPGTQ_N_S, VCMPGTQ_S, VCMPLEQ_N_S)
15761 (VCMPLEQ_S, VCMPLTQ_N_S, VCMPLTQ_S, VCMPCSQ_N_U, VCMPCSQ_U)
15762 (VCMPHIQ_N_U, VCMPHIQ_U): Remove.
15763 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result): Moved
15765 (vec_cmpu<mode><mode>): Likewise.
15766 (vcond<mode><mode>): Likewise.
15767 (vcond<V_cvtto><mode>): Likewise.
15768 (vcondu<mode><v_cmp_result>): Likewise.
15769 (vcond_mask_<mode><v_cmp_result>): Likewise. Added unsafe math
15772 2021-05-17 liuhongt <hongtao.liu@intel.com>
15775 * config/i386/i386.c (ix86_gimple_fold_builtin): Use
15776 gsi_insert_seq_before instead.
15778 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
15780 * doc/sourcebuild.texi (arm_qbit_ok): Rename into...
15781 (arm_sat_ok): ...this.
15783 2021-05-17 Martin Liska <mliska@suse.cz>
15785 * lto-wrapper.c (merge_flto_options): Factor out a new function.
15786 (merge_and_complain): Use it.
15787 (run_gcc): Merge also linker command line -flto=foo argument
15790 2021-05-16 Christophe Lyon <christophe.lyon@linaro.org>
15792 * config/arm/arm.h (CPP_SPEC): Remove error message about
15793 -mlittle-endian/-mbig-endian conflict.
15795 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
15797 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
15798 __ROP_PROTECT__ if -mrop-protect is selected.
15800 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
15802 * config/rs6000/rs6000-internal.h (rs6000_stack): Add
15803 rop_hash_save_offset and rop_hash_size.
15804 * config/rs6000/rs6000-logue.c (rs6000_stack_info): Compute
15805 rop_hash_size and rop_hash_save_offset.
15806 (debug_stack_info): Dump rop_hash_save_offset and rop_hash_size.
15807 (rs6000_emit_prologue): Emit hashst[p] in prologue.
15808 (rs6000_emit_epilogue): Emit hashchk[p] in epilogue.
15809 * config/rs6000/rs6000.md (unspec): Add UNSPEC_HASHST and
15811 (hashst): New define_insn.
15812 (hashchk): Likewise.
15814 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
15816 * config/rs6000/rs6000.c (rs6000_option_override_internal):
15817 Disable shrink wrap when inserting ROP-protect instructions.
15818 * config/rs6000/rs6000.opt (mrop-protect): New option.
15819 (mprivileged): Likewise.
15820 * doc/invoke.texi: Document mrop-protect and mprivileged.
15822 2021-05-15 Hans-Peter Nilsson <hp@axis.com>
15824 * reorg.c (fill_slots_from_thread): Reinstate code typoed out in
15827 2021-05-15 Martin Jambor <mjambor@suse.cz>
15830 2021-05-13 Martin Jambor <mjambor@suse.cz>
15832 PR tree-optimization/100453
15833 * tree-sra.c (sra_modify_assign): All const base accesses do not
15834 need refreshing, not just those from decl_pool.
15835 (sra_modify_assign): Do not refresh into a const base decl.
15837 2021-05-15 Jakub Jelinek <jakub@redhat.com>
15839 PR rtl-optimization/100342
15840 * regcprop.c (copy_value): When copying a source reg in a wider
15841 mode than it has recorded for the value, adjust recorded destination
15842 mode too or punt if !REG_CAN_CHANGE_MODE_P.
15844 2021-05-14 Jason Merrill <jason@redhat.com>
15846 * intl.h: Add comments.
15848 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15850 * config/aarch64/aarch64-simd.md
15851 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Split into...
15852 (aarch64_sqdmlsl2_lane<mode>_internal): ... This...
15853 (aarch64_sqdmlal2_lane<mode>_internal): ... And this.
15854 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Split into ...
15855 (aarch64_sqdmlsl2_laneq<mode>_internal): ... This...
15856 (aarch64_sqdmlal2_laneq<mode>_internal): ... And this.
15857 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>_internal): Split into...
15858 (aarch64_sqdmlsl2_n<mode>_internal): ... This...
15859 (aarch64_sqdmlal2_n<mode>_internal): ... And this.
15861 2021-05-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15864 * config/arm/arm_neon.h (vtst_s8): Replace call to vtst builtin with it's
15865 boolean logic equivalent.
15866 (vtst_s16): Likewise.
15867 (vtst_s32): Likewise.
15868 (vtst_u8): Likewise.
15869 (vtst_u16): Likewise.
15870 (vtst_u32): Likewise.
15871 (vtst_p8): Likewise.
15872 (vtst_p16): Likewise.
15873 (vtstq_s8): Likewise.
15874 (vtstq_s16): Likewise.
15875 (vtstq_s32): Likewise.
15876 (vtstq_u8): Likewise.
15877 (vtstq_u16): Likewise.
15878 (vtstq_u32): Likewise.
15879 (vtstq_p8): Likewise.
15880 (vtstq_p16): Likewise.
15881 * config/arm/arm_neon_builtins.def: Remove entry for vtst.
15882 * config/arm/neon.md (neon_vtst<mode>): Remove pattern.
15884 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15886 * config/aarch64/aarch64-simd.md (aarch64_sqdmlal2<mode>): Merge into...
15887 (aarch64_sqdml<SBINQOPS:as>l2<mode>): ... This.
15888 (aarch64_sqdmlsl2<mode>): Delete.
15889 (aarch64_sqdmlal2_lane<mode>): Merge this...
15890 (aarch64_sqdmlsl2_lane<mode>): ... And this...
15891 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>): ... Into this.
15892 (aarch64_sqdmlal2_laneq<mode>): Merge this...
15893 (aarch64_sqdmlsl2_laneq<mode>): ... And this...
15894 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>): ... Into this.
15895 (aarch64_sqdmlal2_n<mode>): Merge this...
15896 (aarch64_sqdmlsl2_n<mode>): ... And this...
15897 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>): ... Into this.
15899 2021-05-13 Martin Sebor <msebor@redhat.com>
15901 PR middle-end/100574
15902 * builtins.c (access_ref::get_ref): Improve detection of PHIs with
15903 all null arguments.
15905 2021-05-13 Martin Sebor <msebor@redhat.com>
15907 PR tree-optimization/93100
15908 PR middle-end/98583
15909 * tree-ssa-uninit.c (check_defs): Exclude intrinsic functions that
15910 don't modify referenced objects.
15912 2021-05-13 Martin Jambor <mjambor@suse.cz>
15914 PR tree-optimization/100453
15915 * tree-sra.c (sra_modify_assign): All const base accesses do not
15916 need refreshing, not just those from decl_pool.
15917 (sra_modify_assign): Do not refresh into a const base decl.
15919 2021-05-13 Martin Liska <mliska@suse.cz>
15921 * tree-ssa-dom.c: Remove m_simplifier.
15923 2021-05-13 Richard Earnshaw <rearnsha@arm.com>
15926 * config/arm/arm.c (arm_canonicalize_comparison): Correctly
15927 canonicalize DImode inequality comparisons against the
15928 maximum integral value.
15930 2021-05-13 Jakub Jelinek <jakub@redhat.com>
15932 PR tree-optimization/98856
15933 * config/i386/i386.c (ix86_shift_rotate_cost): Add CODE argument.
15934 Expect V2DI and V4DI arithmetic right shifts to be emulated.
15935 (ix86_rtx_costs, ix86_add_stmt_cost): Adjust ix86_shift_rotate_cost
15937 * config/i386/i386-expand.c (expand_vec_perm_2perm_interleave,
15938 expand_vec_perm_2perm_pblendv): New functions.
15939 (ix86_expand_vec_perm_const_1): Use them.
15940 * config/i386/sse.md (ashr<mode>3<mask_name>): Rename to ...
15941 (<mask_codefor>ashr<mode>3<mask_name>): ... this.
15942 (ashr<mode>3): New define_expand with VI248_AVX512BW iterator.
15943 (ashrv4di3): New define_expand.
15944 (ashrv2di3): Change condition to TARGET_SSE2, handle !TARGET_XOP
15945 and !TARGET_AVX512VL expansion.
15947 2021-05-13 Uroš Bizjak <ubizjak@gmail.com>
15950 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Force mode
15951 sizes < 16 to a register when constructing vpcmov pattern.
15952 * config/i386/mmx.md (*xop_pcmov_<mode>): Use MMXMODE124 mode.
15954 2021-05-13 Martin Liska <mliska@suse.cz>
15956 * gcov-io.c (gcov_write_block): Remove.
15957 (gcov_write_words): Likewise.
15958 (gcov_read_words): Re-implement using gcov_read_bytes.
15959 (gcov_allocate): Remove.
15960 (GCOV_BLOCK_SIZE): Likewise.
15961 (struct gcov_var): Remove most of the fields.
15962 (gcov_position): Implement with ftell.
15963 (gcov_rewrite): Remove setting of start and offset fields.
15964 (from_file): Re-format.
15965 (gcov_open): Remove setbuf call. It should not be needed.
15966 (gcov_close): Remove internal buffer handling.
15967 (gcov_magic): Use __builtin_bswap32.
15968 (gcov_write_counter): Use directly gcov_write_unsigned.
15969 (gcov_write_string): Use direct fwrite and do not round
15971 (gcov_seek): Use directly fseek.
15972 (gcov_write_tag): Use gcov_write_unsigned directly.
15973 (gcov_write_length): Likewise.
15974 (gcov_write_tag_length): Likewise.
15975 (gcov_read_bytes): Use directly fread.
15976 (gcov_read_unsigned): Use gcov_read_words.
15977 (gcov_read_counter): Likewise.
15978 (gcov_read_string): Use gcov_read_bytes.
15979 * gcov-io.h (GCOV_WORD_SIZE): Adjust to reflect
15980 that size is not in bytes, but words (4B).
15981 (GCOV_TAG_FUNCTION_LENGTH): Likewise.
15982 (GCOV_TAG_ARCS_LENGTH): Likewise.
15983 (GCOV_TAG_ARCS_NUM): Likewise.
15984 (GCOV_TAG_COUNTER_LENGTH): Likewise.
15985 (GCOV_TAG_COUNTER_NUM): Likewise.
15986 (GCOV_TAG_SUMMARY_LENGTH): Likewise.
15988 2021-05-13 liuhongt <hongtao.liu@intel.com>
15991 * config/i386/sse.md (ssedoublevecmode): Add attribute for
15992 V64QI/V32HI/V16SI/V4DI.
15993 (ssehalfvecmode): Add attribute for V2DI/V2DF.
15994 (*vec_concatv4si_0): Extend to VI124_128.
15995 (*vec_concat<mode>_0): New pre-reload splitter.
15996 * config/i386/predicates.md (movq_parallel): New predicate.
15998 2021-05-13 Alexandre Oliva <oliva@adacore.com>
16000 * targhooks.c (default_zero_call_used_regs): Retry using
16001 successfully-zeroed registers as sources.
16003 2021-05-12 Tobias Burnus <tobias@codesourcery.com>
16005 * omp-low.c (finish_taskreg_scan): Use the proper detach decl.
16007 2021-05-12 Aldy Hernandez <aldyh@redhat.com>
16010 * gimple-range.cc (range_of_builtin_call): Skip out on
16011 processing __builtin_clz when varying.
16013 2021-05-12 Tom de Vries <tdevries@suse.de>
16016 * config/nvptx/nvptx-opts.h (enum ptx_version): New enum.
16017 * config/nvptx/nvptx.c (nvptx_file_start): Print .version according
16018 to ptx_version_option.
16019 * config/nvptx/nvptx.h (TARGET_PTX_6_3): Define.
16020 * config/nvptx/nvptx.md (define_insn "nvptx_shuffle<mode>")
16021 (define_insn "nvptx_vote_ballot"): Use sync variant for
16023 * config/nvptx/nvptx.opt (ptx_version): Add enum.
16024 (mptx): Add option.
16025 * doc/invoke.texi (Nvidia PTX Options): Add mptx item.
16027 2021-05-12 Richard Biener <rguenther@suse.de>
16029 PR tree-optimization/100566
16030 * tree-ssa-sccvn.c (dominated_by_p_w_unex): Properly handle
16031 allow_back for all edge queries.
16033 2021-05-12 liuhongt <hongtao.liu@intel.com>
16036 * config/i386/sse.md (<sse4_1_avx2>_pblendvb): Add
16037 splitters for pblendvb of NOT mask register.
16039 2021-05-12 Richard Biener <rguenther@suse.de>
16041 PR tree-optimization/100519
16042 * tree-ssa-reassoc.c (can_associate_p): Split into...
16043 (can_associate_op_p): ... this
16044 (can_associate_type_p): ... and this.
16045 (is_reassociable_op): Call can_associate_op_p.
16046 (break_up_subtract_bb): Call the appropriate predicates.
16047 (reassociate_bb): Likewise.
16049 2021-05-12 Martin Liska <mliska@suse.cz>
16051 * lto-wrapper.c (merge_and_complain): Merge -flto=arg options.
16052 (run_gcc): Use -flto argument detection for merged
16055 2021-05-12 Martin Liska <mliska@suse.cz>
16057 * lto-wrapper.c (print_lto_docs_link): New function.
16058 (run_gcc): Print warning about missing job server detection
16059 after we know NR of partitions. Do the same for -flto{,=1}.
16060 * opts.c (get_option_html_page): Support -flto option.
16062 2021-05-12 Martin Liska <mliska@suse.cz>
16064 * lto-wrapper.c (get_options_from_collect_gcc_options): Change
16066 (append_option): Remove.
16067 (find_option): Rework to use the vector type.
16068 (remove_option): Remove.
16069 (merge_and_complain): Use vectors for cl_decoded_option data
16071 (append_compiler_options): Likewise.
16072 (append_diag_options): Likewise.
16073 (append_linker_options): Likewise.
16074 (append_offload_options): Likewise.
16075 (compile_offload_image): Likewise.
16076 (compile_images_for_offload_targets): Likewise.
16077 (find_and_merge_options): Likewise.
16078 (run_gcc): Likewise.
16080 2021-05-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
16083 * dwarf2out.c (dwarf2out_finish): Set
16084 have_multiple_function_sections with multi-range text_section.
16086 2021-05-12 Martin Liska <mliska@suse.cz>
16088 PR bootstrap/100560
16089 * Makefile.in: Remove version.h from linker command line.
16091 2021-05-12 Richard Biener <rguenther@suse.de>
16093 PR middle-end/100547
16094 * rtl.h (rtvec_alloc): Make argument size_t.
16095 * rtl.c (rtvec_alloc): Verify the count is less than INT_MAX.
16097 2021-05-12 Jakub Jelinek <jakub@redhat.com>
16099 PR middle-end/100508
16100 * cfgexpand.c (expand_debug_expr): For DEBUG_EXPR_DECL with vector
16101 type, don't reuse DECL_RTL if it has different mode, instead force
16102 creation of a new DEBUG_EXPR.
16104 2021-05-12 Jakub Jelinek <jakub@redhat.com>
16105 Marc Glisse <marc.glisse@inria.fr>
16107 PR tree-optimization/94589
16108 * match.pd ((X & Y) == X -> (X & ~Y) == 0,
16109 (X | Y) == Y -> (X & ~Y) == 0): New GIMPLE simplifications.
16111 2021-05-12 Uroš Bizjak <ubizjak@gmail.com>
16114 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Handle V2SF mode.
16115 * config/i386/mmx.md (MMXMODE124): New mode iterator.
16117 (mmxintvecmode): New mode attribute.
16118 (mmxintvecmodelower): Ditto.
16119 (*mmx_maskcmpv2sf3_comm): New insn pattern.
16120 (*mmx_maskcmpv2sf3): Ditto.
16121 (vec_cmpv2sfv2si): New expander.
16122 (vcond<V2FI:mode>v2si): Ditto.
16123 (mmx_vlendvps): New insn pattern.
16124 (vcond<MMXMODE124:mode><MMXMODEI:mode>): Also handle V2SFmode.
16125 (vcondu<MMXMODE124:mode><MMXMODEI:mode>): Ditto.
16126 (vcond_mask_<mode><mmxintvecmodelower>): Ditto.
16128 2021-05-11 Martin Sebor <msebor@redhat.com>
16130 PR middle-end/21433
16131 * expr.c (expand_expr_real_1): Replace unreachable code with an assert.
16133 2021-05-11 Richard Biener <rguenther@suse.de>
16135 * gimple-fold.c (gimple_fold_call): Do not call
16136 maybe_fold_reference on call arguments or the static chain.
16137 (fold_stmt_1): Do not call maybe_fold_reference on GIMPLE_ASM
16140 2021-05-11 Martin Liska <mliska@suse.cz>
16142 * builtins.def (DEF_HSAIL_BUILTIN): Remove.
16143 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
16144 (DEF_HSAIL_SAT_BUILTIN): Likewise.
16145 (DEF_HSAIL_INTR_BUILTIN): Likewise.
16146 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
16147 * doc/frontends.texi: Remove BRIG.
16148 * doc/install.texi: Likewise.
16149 * doc/invoke.texi: Likewise.
16150 * doc/standards.texi: Likewise.
16151 * brig-builtins.def: Removed.
16152 * brig/ChangeLog: Removed.
16153 * brig/Make-lang.in: Removed.
16154 * brig/brig-builtins.h: Removed.
16155 * brig/brig-c.h: Removed.
16156 * brig/brig-lang.c: Removed.
16157 * brig/brigfrontend/brig-arg-block-handler.cc: Removed.
16158 * brig/brigfrontend/brig-atomic-inst-handler.cc: Removed.
16159 * brig/brigfrontend/brig-basic-inst-handler.cc: Removed.
16160 * brig/brigfrontend/brig-branch-inst-handler.cc: Removed.
16161 * brig/brigfrontend/brig-cmp-inst-handler.cc: Removed.
16162 * brig/brigfrontend/brig-code-entry-handler.cc: Removed.
16163 * brig/brigfrontend/brig-code-entry-handler.h: Removed.
16164 * brig/brigfrontend/brig-comment-handler.cc: Removed.
16165 * brig/brigfrontend/brig-control-handler.cc: Removed.
16166 * brig/brigfrontend/brig-copy-move-inst-handler.cc: Removed.
16167 * brig/brigfrontend/brig-cvt-inst-handler.cc: Removed.
16168 * brig/brigfrontend/brig-fbarrier-handler.cc: Removed.
16169 * brig/brigfrontend/brig-function-handler.cc: Removed.
16170 * brig/brigfrontend/brig-function.cc: Removed.
16171 * brig/brigfrontend/brig-function.h: Removed.
16172 * brig/brigfrontend/brig-inst-mod-handler.cc: Removed.
16173 * brig/brigfrontend/brig-label-handler.cc: Removed.
16174 * brig/brigfrontend/brig-lane-inst-handler.cc: Removed.
16175 * brig/brigfrontend/brig-machine.c: Removed.
16176 * brig/brigfrontend/brig-machine.h: Removed.
16177 * brig/brigfrontend/brig-mem-inst-handler.cc: Removed.
16178 * brig/brigfrontend/brig-module-handler.cc: Removed.
16179 * brig/brigfrontend/brig-queue-inst-handler.cc: Removed.
16180 * brig/brigfrontend/brig-seg-inst-handler.cc: Removed.
16181 * brig/brigfrontend/brig-signal-inst-handler.cc: Removed.
16182 * brig/brigfrontend/brig-to-generic.cc: Removed.
16183 * brig/brigfrontend/brig-to-generic.h: Removed.
16184 * brig/brigfrontend/brig-util.cc: Removed.
16185 * brig/brigfrontend/brig-util.h: Removed.
16186 * brig/brigfrontend/brig-variable-handler.cc: Removed.
16187 * brig/brigfrontend/hsa-brig-format.h: Removed.
16188 * brig/brigfrontend/phsa.h: Removed.
16189 * brig/brigspec.c: Removed.
16190 * brig/config-lang.in: Removed.
16191 * brig/gccbrig.texi: Removed.
16192 * brig/lang-specs.h: Removed.
16193 * brig/lang.opt: Removed.
16195 2021-05-11 Richard Biener <rguenther@suse.de>
16198 * ipa-param-manipulation.c
16199 (ipa_param_body_adjustments::modify_call_stmt): Avoid
16200 altering SSA_NAME_DEF_STMT by adjusting the calls LHS
16201 via gimple_call_lhs_ptr.
16203 2021-05-11 Alex Coplan <alex.coplan@arm.com>
16206 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear):
16207 Avoid emitting CFA adjusts on the sp if we have the fp.
16209 2021-05-11 Richard Sandiford <richard.sandiford@arm.com>
16211 * config/aarch64/iterators.md (VMUL_CHANGE_NLANES): Delete.
16212 (VMULD): New iterator.
16213 (VCOND): Handle V4HF and V8HF.
16214 (VCONQ): Fix entry for V2SF.
16215 * config/aarch64/aarch64-simd.md (mul_lane<mode>3): Use VMULD
16216 instead of VMUL. Use a 64-bit vector mode for the indexed operand.
16217 (*aarch64_mul3_elt_<vswap_width_name><mode>): Merge with...
16218 (mul_laneq<mode>3): ...this define_insn. Use VMUL instead of VDQSF.
16219 Use a 128-bit vector mode for the indexed operand. Use stype for
16220 the scheduling type.
16222 2021-05-11 Richard Biener <rguenther@suse.de>
16224 * gimple-fold.c (maybe_fold_reference): Only return
16225 is_gimple_min_invariant values.
16227 2021-05-11 Richard Biener <rguenther@suse.de>
16229 PR middle-end/100509
16230 * gimple-fold.c (fold_gimple_assign): Only call
16231 get_symbol_constant_value on register type symbols.
16233 2021-05-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
16234 Joe Ramsay <joe.ramsay@arm.com>
16237 * config/arm/arm_mve.h (__arm_vstrwq_scatter_offset): Fix wrong arguments.
16238 (__arm_vcmpneq): Remove duplicate definition.
16239 (__arm_vstrwq_scatter_offset_p): Likewise.
16240 (__arm_vmaxq_x): Likewise.
16241 (__arm_vmlsdavaq): Likewise.
16242 (__arm_vmlsdavaxq): Likewise.
16243 (__arm_vmlsdavq_p): Likewise.
16244 (__arm_vmlsdavxq_p): Likewise.
16245 (__arm_vrmlaldavhaq): Likewise.
16246 (__arm_vstrbq_p): Likewise.
16247 (__arm_vstrbq_scatter_offset): Likewise.
16248 (__arm_vstrbq_scatter_offset_p): Likewise.
16249 (__arm_vstrdq_scatter_offset): Likewise.
16250 (__arm_vstrdq_scatter_offset_p): Likewise.
16251 (__arm_vstrdq_scatter_shifted_offset): Likewise.
16252 (__arm_vstrdq_scatter_shifted_offset_p): Likewise.
16254 2021-05-11 Jakub Jelinek <jakub@redhat.com>
16256 PR middle-end/100471
16257 * omp-low.c (lower_omp_task_reductions): For OMP_TASKLOOP, if data
16258 is 0, bypass the reduction loop including
16259 GOMP_taskgroup_reduction_unregister call.
16261 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
16263 * config/rs6000/rs6000.c (struct rs6000_cost_data): New member
16264 costing_for_scalar.
16265 (rs6000_density_test): Early return if costing_for_scalar is true.
16266 (rs6000_init_cost): Init costing_for_scalar of rs6000_cost_data.
16268 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
16270 * doc/tm.texi: Regenerated.
16271 * target.def (init_cost): Add new parameter costing_for_scalar.
16272 * targhooks.c (default_init_cost): Adjust for new parameter.
16273 * targhooks.h (default_init_cost): Likewise.
16274 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
16275 (vect_compute_single_scalar_iteration_cost): Likewise.
16276 (vect_analyze_loop_2): Likewise.
16277 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
16278 (vect_bb_vectorization_profitable_p): Likewise.
16279 * tree-vectorizer.h (init_cost): Likewise.
16280 * config/aarch64/aarch64.c (aarch64_init_cost): Likewise.
16281 * config/i386/i386.c (ix86_init_cost): Likewise.
16282 * config/rs6000/rs6000.c (rs6000_init_cost): Likewise.
16284 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
16286 * config/rs6000/rs6000.c (rs6000_vect_nonmem): Renamed to
16287 vect_nonmem and moved into...
16288 (struct rs6000_cost_data): ...here.
16289 (rs6000_init_cost): Use vect_nonmem of cost_data instead.
16290 (rs6000_add_stmt_cost): Likewise.
16291 (rs6000_finish_cost): Likewise.
16293 2021-05-10 Eric Botcazou <ebotcazou@adacore.com>
16295 * range-op.cc (get_bool_state): Adjust head comment.
16296 (operator_not_equal::op1_range): Fix comment.
16297 (operator_bitwise_xor::op1_range): Remove call to gcc_unreachable.
16299 2021-05-10 Martin Sebor <msebor@redhat.com>
16301 PR middle-end/100425
16302 PR middle-end/100510
16303 * gimple-ssa-warn-alloca.c (pass_walloca::firast_time_p): Rename...
16304 (pass_walloca::xlimit_certain_p): ...to this.
16305 (pass_walloca::gate): Execute for any kind of handled warning.
16306 (pass_walloca::execute): Avoid issuing "maybe" and "unbounded"
16307 warnings when xlimit_certain_p is set.
16309 2021-05-10 Pat Haugen <pthaugen@linux.ibm.com>
16311 * config/rs6000/rs6000.c (rs6000_ira_change_pseudo_allocno_class):
16312 Return ALTIVEC_REGS if that is best_class.
16313 (rs6000_compute_pressure_classes): Add ALTIVEC_REGS.
16315 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16317 * config/arm/arm.h (CPP_SPEC): Remove error message about
16320 2021-05-10 Martin Jambor <mjambor@suse.cz>
16322 * ipa-prop.h (IPA_NODE_REF): Removed.
16323 (IPA_NODE_REF_GET_CREATE): Likewise.
16324 (IPA_EDGE_REF): Likewise.
16325 (IPA_EDGE_REF_GET_CREATE): Likewise.
16326 (IS_VALID_JUMP_FUNC_INDEX): Likewise.
16327 * ipa-cp.c (print_all_lattices): Replaced IPA_NODE_REF with a direct
16328 use of ipa_node_params_sum.
16329 (ipcp_versionable_function_p): Likewise.
16330 (push_node_to_stack): Likewise.
16331 (pop_node_from_stack): Likewise.
16332 (set_single_call_flag): Replaced two IPA_NODE_REF with one single
16333 direct use of ipa_node_params_sum.
16334 (initialize_node_lattices): Replaced IPA_NODE_REF with a direct use of
16335 ipa_node_params_sum.
16336 (ipa_context_from_jfunc): Replaced IPA_EDGE_REF with a direct use of
16338 (ipcp_verify_propagated_values): Replaced IPA_NODE_REF with a direct
16339 use of ipa_node_params_sum.
16340 (self_recursively_generated_p): Likewise.
16341 (propagate_scalar_across_jump_function): Likewise.
16342 (propagate_context_across_jump_function): Replaced IPA_EDGE_REF with a
16343 direct use of ipa_edge_args_sum, moved the lookup after the early
16344 exit. Replaced IPA_NODE_REF with a direct use of ipa_node_params_sum.
16345 (propagate_bits_across_jump_function): Replaced IPA_NODE_REF with
16346 direct uses of ipa_node_params_sum.
16347 (propagate_vr_across_jump_function): Likewise.
16348 (propagate_aggregate_lattice): Likewise.
16349 (propagate_aggs_across_jump_function): Likewise.
16350 (propagate_constants_across_call): Likewise, also replaced
16351 IPA_EDGE_REF with a direct use of ipa_edge_args_sum.
16352 (good_cloning_opportunity_p): Replaced IPA_NODE_REF with a direct use
16353 of ipa_node_params_sum.
16354 (estimate_local_effects): Likewise.
16355 (add_all_node_vals_to_toposort): Likewise.
16356 (propagate_constants_topo): Likewise.
16357 (ipcp_propagate_stage): Likewise.
16358 (ipcp_discover_new_direct_edges): Likewise.
16359 (calls_same_node_or_its_all_contexts_clone_p): Likewise.
16360 (cgraph_edge_brings_value_p): Likewise (in both overloaded functions).
16361 (get_info_about_necessary_edges): Likewise.
16362 (want_remove_some_param_p): Likewise.
16363 (create_specialized_node): Likewise.
16364 (self_recursive_pass_through_p): Likewise.
16365 (self_recursive_agg_pass_through_p): Likewise.
16366 (find_more_scalar_values_for_callers_subset): Likewise and also
16367 replaced IPA_EDGE_REF with direct uses of ipa_edge_args_sum, in one
16368 case replacing two of those with a single query.
16369 (find_more_contexts_for_caller_subset): Likewise for the
16370 ipa_polymorphic_call_context overload.
16371 (intersect_aggregates_with_edge): Replaced IPA_EDGE_REF with a direct
16372 use of ipa_edge_args_sum. Replaced IPA_NODE_REF with direct uses of
16373 ipa_node_params_sum.
16374 (find_aggregate_values_for_callers_subset): Likewise, also reusing
16375 results of ipa_edge_args_sum->get.
16376 (cgraph_edge_brings_all_scalars_for_node): Replaced IPA_NODE_REF with
16377 direct uses of ipa_node_params_sum, replaced IPA_EDGE_REF with a
16378 direct use of ipa_edge_args_sum.
16379 (cgraph_edge_brings_all_agg_vals_for_node): Likewise, moved node
16380 summary query after the early exit and reused the result later.
16381 (decide_about_value): Replaced IPA_NODE_REF with a direct use of
16382 ipa_node_params_sum.
16383 (decide_whether_version_node): Likewise. Removed re-querying for
16384 summaries after cloning.
16385 (spread_undeadness): Replaced IPA_NODE_REF with a direct use of
16386 ipa_node_params_sum.
16387 (has_undead_caller_from_outside_scc_p): Likewise, reusing results of
16389 (identify_dead_nodes): Likewise.
16390 (ipcp_store_bits_results): Replaced IPA_NODE_REF with direct uses of
16391 ipa_node_params_sum.
16392 (ipcp_store_vr_results): Likewise.
16393 * ipa-fnsummary.c (evaluate_properties_for_edge): Likewise.
16394 (ipa_fn_summary_t::duplicate): Likewise.
16395 (analyze_function_body): Likewise.
16396 (estimate_calls_size_and_time): Likewise.
16397 (ipa_cached_call_context::duplicate_from): Likewise.
16398 (ipa_call_context::equal_to): Likewise.
16399 (remap_edge_params): Likewise.
16400 (ipa_merge_fn_summary_after_inlining): Likewise.
16401 (inline_read_section): Likewise.
16402 * ipa-icf.c (sem_function::param_used_p): Likewise.
16403 * ipa-modref.c (compute_parm_map): Likewise.
16404 (compute_parm_map): Replaced IPA_EDGE_REF with a direct use of
16406 (get_access_for_fnspec): Replaced IPA_NODE_REF with a direct use of
16407 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
16409 * ipa-profile.c (check_argument_count): Likewise.
16410 * ipa-prop.c (ipa_alloc_node_params): Replaced IPA_NODE_REF_GET_CREATE
16411 with a direct use of ipa_node_params_sum.
16412 (ipa_initialize_node_params): Likewise.
16413 (ipa_print_node_jump_functions_for_edge): Replaced IPA_EDGE_REF with a
16414 direct use of ipa_edge_args_sum and reused the query result.
16415 (ipa_compute_jump_functions_for_edge): Replaced IPA_NODE_REF with a
16416 direct use of ipa_node_params_sum and replaced IPA_EDGE_REF with a
16417 direct use of ipa_edge_args_sum.
16418 (ipa_note_param_call): Replaced IPA_NODE_REF with a direct use of
16419 ipa_node_params_sum and reused the result of the query.
16420 (ipa_analyze_node): Likewise.
16421 (ipa_analyze_controlled_uses): Replaced IPA_NODE_REF with a direct use
16422 of ipa_node_params_sum.
16423 (update_jump_functions_after_inlining): Replaced IPA_EDGE_REF with
16424 direct uses of ipa_edge_args_sum.
16425 (update_indirect_edges_after_inlining): Replaced IPA_NODE_REF with
16426 direct uses of ipa_node_params_sum and replaced IPA_EDGE_REF with a
16427 direct use of ipa_edge_args_sum. Removed superficial re-querying the
16429 (propagate_controlled_uses): Replaced IPA_NODE_REF with direct uses of
16430 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
16432 (ipa_propagate_indirect_call_infos): Replaced IPA_EDGE_REF with a
16433 direct use of ipa_edge_args_sum.
16434 (ipa_edge_args_sum_t::duplicate): Replaced IPA_NODE_REF with a direct
16435 use of ipa_node_params_sum.
16436 (ipa_print_node_params): Likewise.
16437 (ipa_write_node_info): Likewise and also replaced IPA_EDGE_REF with
16438 direct uses of ipa_edge_args_sum.
16439 (ipa_read_edge_info): Replaced IPA_EDGE_REF with a direct use of
16441 (ipa_read_node_info): Replaced IPA_NODE_REF with a direct use of
16442 ipa_node_params_sum.
16443 (ipa_prop_write_jump_functions): Likewise. Move variable node to the
16444 scopes where it is used.
16446 2021-05-10 Uroš Bizjak <ubizjak@gmail.com>
16448 * config/i386/i386-expand.c (ix86_expand_sse_movcc)
16449 <case E_V2SImode>: Force op_true to register.
16451 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16453 * config/arm/iterators.md (MVE_FP_COMPARISONS): New.
16454 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_f<mode>)
16455 (mve_vcmp<mve_cmp_op>q_n_f<mode>): New, merge all vcmp_*f*
16457 (mve_vcmpeqq_f<mode>, mve_vcmpeqq_n_f<mode>, mve_vcmpgeq_f<mode>)
16458 (mve_vcmpgeq_n_f<mode>, mve_vcmpgtq_f<mode>)
16459 (mve_vcmpgtq_n_f<mode>, mve_vcmpleq_f<mode>)
16460 (mve_vcmpleq_n_f<mode>, mve_vcmpltq_f<mode>)
16461 (mve_vcmpltq_n_f<mode>, mve_vcmpneq_f<mode>)
16462 (mve_vcmpneq_n_f<mode>): Remove.
16463 * config/arm/unspecs.md (VCMPEQQ_F, VCMPEQQ_N_F, VCMPGEQ_F)
16464 (VCMPGEQ_N_F, VCMPGTQ_F, VCMPGTQ_N_F, VCMPLEQ_F, VCMPLEQ_N_F)
16465 (VCMPLTQ_F, VCMPLTQ_N_F, VCMPNEQ_F, VCMPNEQ_N_F): Remove.
16467 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16469 * config/arm/iterators.md (MVE_COMPARISONS): New.
16471 (mve_cmp_type): New.
16472 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_<mode>): New, merge all
16474 (mve_vcmpneq_<mode>, mve_vcmpcsq_n_<mode>, mve_vcmpcsq_<mode>)
16475 (mve_vcmpeqq_n_<mode>, mve_vcmpeqq_<mode>, mve_vcmpgeq_n_<mode>)
16476 (mve_vcmpgeq_<mode>, mve_vcmpgtq_n_<mode>, mve_vcmpgtq_<mode>)
16477 (mve_vcmphiq_n_<mode>, mve_vcmphiq_<mode>, mve_vcmpleq_n_<mode>)
16478 (mve_vcmpleq_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
16479 (mve_vcmpneq_n_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
16480 (mve_vcmpneq_n_<mode>): Remove.
16482 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16484 * config/arm/arm_mve.h (__arm_vcmp*): Remove 's' suffix.
16485 * config/arm/arm_mve_builtins.def (vcmp*): Remove 's' suffix.
16486 * config/arm/mve.md (mve_vcmp*): Remove 's' suffix in pattern
16489 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16491 * config/arm/arm_mve_builtins.def (vcmpneq_u): Remove.
16492 (vcmpneq_n_u): Likewise.
16493 (vcmpeqq_u,): Likewise.
16494 (vcmpeqq_n_u): Likewise.
16495 * config/arm/iterators.md (supf): Remove VCMPNEQ_U, VCMPEQQ_U,
16496 VCMPEQQ_N_U and VCMPNEQ_N_U.
16497 * config/arm/mve.md (mve_vcmpneq): Remove <supf> iteration.
16498 (mve_vcmpeqq_n): Likewise.
16499 (mve_vcmpeqq): Likewise.
16500 (mve_vcmpneq_n): Likewise.
16502 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16504 * config/arm/arm_mve.h (__arm_vcmpeq*u*, __arm_vcmpne*u*): Call
16505 the 's' version of the builtin.
16507 2021-05-10 Richard Biener <rguenther@suse.de>
16509 PR tree-optimization/100492
16510 * tree-loop-distribution.c (find_seed_stmts_for_distribution):
16511 Find nothing when the loop contains an irreducible region.
16513 2021-05-10 Richard Biener <rguenther@suse.de>
16515 PR middle-end/100464
16517 * gimple-fold.c (canonicalize_constructor_val): Do not set
16520 2021-05-10 Richard Biener <rguenther@suse.de>
16522 PR tree-optimization/100434
16523 * tree-ssa-dse.c (initialize_ao_ref_for_dse): Handle
16525 (dse_optimize_stmt): Handle call LHS by dropping the
16526 LHS or the whole call if it doesn't have other
16528 (pass_dse::execute): Adjust.
16530 2021-05-10 Martin Liska <mliska@suse.cz>
16532 * Makefile.in: Add missing genversion rule.
16534 2021-05-10 Alex Coplan <alex.coplan@arm.com>
16537 * config/arm/mve.md (*mve_mov<mode>): Simplify output code. Use
16538 vldrw.u32 and vstrw.32 for V2D[IF]mode loads and stores.
16540 2021-05-10 Martin Liska <mliska@suse.cz>
16542 * builtins.c (is_builtin_name): Use startswith
16543 function instead of strncmp.
16544 * collect2.c (main): Likewise.
16545 (has_lto_section): Likewise.
16546 (scan_libraries): Likewise.
16547 * coverage.c (coverage_checksum_string): Likewise.
16548 (coverage_init): Likewise.
16549 * dwarf2out.c (is_cxx): Likewise.
16550 (gen_compile_unit_die): Likewise.
16551 * gcc-ar.c (main): Likewise.
16552 * gcc.c (init_spec): Likewise.
16553 (read_specs): Likewise.
16554 (execute): Likewise.
16555 (check_live_switch): Likewise.
16556 * genattrtab.c (write_attr_case): Likewise.
16557 (IS_ATTR_GROUP): Likewise.
16558 * gencfn-macros.c (main): Likewise.
16559 * gengtype.c (type_for_name): Likewise.
16560 (gen_rtx_next): Likewise.
16561 (get_file_langdir): Likewise.
16562 (write_local): Likewise.
16563 * genmatch.c (get_operator): Likewise.
16564 (get_operand_type): Likewise.
16565 (expr::gen_transform): Likewise.
16566 * genoutput.c (validate_optab_operands): Likewise.
16567 * incpath.c (add_sysroot_to_chain): Likewise.
16568 * langhooks.c (lang_GNU_C): Likewise.
16569 (lang_GNU_CXX): Likewise.
16570 (lang_GNU_Fortran): Likewise.
16571 (lang_GNU_OBJC): Likewise.
16572 * lto-wrapper.c (run_gcc): Likewise.
16573 * omp-general.c (omp_max_simt_vf): Likewise.
16574 * omp-low.c (omp_runtime_api_call): Likewise.
16575 * opts-common.c (parse_options_from_collect_gcc_options): Likewise.
16576 * read-rtl-function.c (function_reader::read_rtx_operand_r): Likewise.
16577 * real.c (real_from_string): Likewise.
16578 * selftest.c (assert_str_startswith): Likewise.
16579 * timevar.c (timer::validate_phases): Likewise.
16580 * tree.c (get_file_function_name): Likewise.
16581 * ubsan.c (ubsan_use_new_style_p): Likewise.
16582 * varasm.c (default_function_rodata_section): Likewise.
16583 (incorporeal_function_p): Likewise.
16584 (default_section_type_flags): Likewise.
16585 * system.h (startswith): Define startswith.
16587 2021-05-10 Martin Liska <mliska@suse.cz>
16589 * bitmap.h (class auto_bitmap): Remove
16590 __cplusplus >= 201103.
16591 * config/aarch64/aarch64.c: Likewise.
16592 * gimple-ssa-store-merging.c (store_immediate_info::store_immediate_info):
16594 * sbitmap.h: Likewise.
16596 2021-05-10 Martin Liska <mliska@suse.cz>
16598 * Makefile.in: Rename gcov-iov to genversion and depend
16599 on version.h (instead of gcov-iov.h).
16600 * gcov-io.h: Include version.h instread of gcov-iov.h.
16601 * gengtype-state.c (read_state_version): Likewise.
16602 * gcov-iov.c: Moved to...
16603 * genversion.c: ...here.
16604 * lto-streamer.h (LTO_major_version): Define it with
16606 * version.c: Removed.
16607 * version.h: Removed.
16609 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16611 * config/arc/arc.md (UNSPEC_ARC_DMPYWH): Define.
16612 * config/arc/simdext.md (VCT): Add predicates for iterator
16615 (voptab): Likewise.
16616 (vec_widen_<V_US>mult_hi_v4hi): Change pattern predicate.
16617 (<voptab>v2si3): New patterns.
16619 (reduc_plus_scal_v4hi): Likewise.
16620 (reduc_plus_scal_v2si): Likewise.
16621 (vec_duplicatev2si): Likewise.
16622 (vec_duplicatev4hi): Likewise.
16624 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16626 * config/arc/simdext.md: Format and cleanup file.
16628 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16630 * config/arc/simdext.md (movmisalignv2hi): Allow misaligned access
16631 only when munaligned-access option is on.
16632 (movmisalign<mode>): Likewise.
16634 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16636 * common/config/arc/arc-common.c (arc_handle_option): Remove dot
16638 * config/arc/arc.c (arc_reorg): Remove underscore from string.
16640 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16642 * config/arc/arc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
16643 (CTZ_DEFINED_VALUE_AT_ZERO): Likewise.
16644 * config/arc/arc.md (clrsbsi2): Cleanup pattern.
16645 (norm_f): Likewise.
16648 (clzsi2): Use fls instruction when available.
16649 (arc_clzsi2): Likewise.
16651 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16653 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add r26 and r27.
16655 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16657 * doc/extend.texi (__builtin_arc_sr): Swap arguments.
16659 2021-05-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
16661 PR middle-end/100467
16662 * toplev.c (compile_file): Call insn_locations_init before
16663 targetm.asm_out.code_end.
16665 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
16668 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
16670 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
16672 2021-05-07 Jakub Jelinek <jakub@redhat.com>
16673 Andrew Stubbs <amd@codesourcery.com>
16676 * builtins.c (try_store_by_multiple_pieces): Use force_operand for
16677 emit_move_insn operands.
16679 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
16681 * cfgexpand.c (expand_gimple_basic_block): Do not inherit a current
16682 location for the outgoing edges of an empty block.
16683 * dwarf2out.c (add_subscript_info): Retrieve the bounds and index
16684 type by means of the get_array_descr_info langhook, if it is set and
16685 returns true. Remove obsolete code dealing with unnamed subtypes.
16687 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16689 * gimple-range-cache.cc (ssa_block_ranges): Virtualize.
16690 (sbr_vector): Renamed from ssa_block_cache.
16691 (sbr_vector::sbr_vector): Allocate from obstack abd initialize.
16692 (ssa_block_ranges::~ssa_block_ranges): Remove.
16693 (sbr_vector::set_bb_range): Use varying and undefined cached values.
16694 (ssa_block_ranges::set_bb_varying): Remove.
16695 (sbr_vector::get_bb_range): Adjust assert.
16696 (sbr_vector::bb_range_p): Adjust assert.
16697 (~block_range_cache): No freeing loop required.
16698 (block_range_cache::get_block_ranges): Remove.
16699 (block_range_cache::set_bb_range): Inline get_block_ranges.
16700 (block_range_cache::set_bb_varying): Remove.
16701 * gimple-range-cache.h (set_bb_varying): Remove prototype.
16702 * value-range.h (irange_allocator::get_memory): New.
16704 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16706 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Search
16707 dominator tree is available and requested.
16708 (ranger_cache::ssa_range_in_bb): Don't search dom tree here.
16709 (ranger_cache::fill_block_cache): Don't search dom tree here either.
16710 * gimple-range-cache.h (non_null_deref_p): Add dom_search param.
16712 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16714 * gimple-range.cc (gimple_ranger::range_on_exit): Handle block with
16715 only PHI nodes better.
16717 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16719 * gimple-range-edge.h (gimple_outgoing_range): Rename from
16721 (gcond_edge_range): Export prototype.
16722 * gimple-range-edge.cc (gcond_edge_range): New.
16723 (gimple_outgoing_range::edge_range_p): Use gcond_edge_range.
16724 * gimple-range-gori.h (gori_compute): Use gimple_outgoing_range.
16726 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16728 * gimple-range-edge.cc (outgoing_range::calc_switch_ranges): Compute
16729 default range into a temp and allocate only what is needed.
16731 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16733 * range-op.cc (operator_trunc_mod::wi_fold): x % 0 is UNDEFINED.
16735 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16737 * gimple-range.h (gimple_range_global): Pick up parameter initial
16738 values, and use-before defined locals are UNDEFINED.
16740 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
16742 * doc/extend.texi (scalar_storage_order): Mention effect on pointer
16744 * tree.h (reverse_storage_order_for_component_p): Return false if
16745 the type is a pointer.
16747 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
16749 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
16751 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
16754 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
16755 Handle V8QI, V4HI and V2SI modes.
16756 * config/i386/mmx.md (mmx_pblendvb): New insn pattern.
16757 * config/i386/sse.md (unspec): Move UNSPEC_BLENDV ...
16758 * config/i386/i386.md (unspec): ... here.
16760 2021-05-07 Tobias Burnus <tobias@codesourcery.com>
16761 Tom de Vries <tdevries@suse.de>
16763 * omp-low.c (lower_rec_simd_input_clauses): Set max_vf = 1 if
16764 a truth_value_p reduction variable is nonintegral.
16766 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
16769 * config/i386/i386-expand.c (ix86_use_mask_cmp_p):
16770 Return false for mode sizes < 16.
16772 2021-05-07 Jakub Jelinek <jakub@redhat.com>
16775 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn.
16777 2021-05-06 Martin Jambor <mjambor@suse.cz>
16779 * ipa-sra.c (ipa_sra_dump_all_summaries): Dump edge summaries even
16780 when there is no function summary.
16781 (ipa_sra_summarize_function): produce edge summaries even when
16784 2021-05-06 Tom Tromey <tom@tromey.com>
16786 * godump.c (string_hash_eq): Remove.
16787 (go_finish): Use htab_eq_string.
16789 2021-05-06 Tom Tromey <tom@tromey.com>
16791 * gengtype-state.c (read_state): Use htab_eq_string.
16792 (string_eq): Remove.
16794 2021-05-06 Tom Tromey <tom@tromey.com>
16796 * gensupport.c (htab_eq_string): Remove.
16798 2021-05-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
16801 * debug.h (gcc_debug_hooks): Add set_ignored_loc function pointer.
16802 * dwarf2out.h (dw_fde_node::ignored_debug): New data item.
16803 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Add dummy
16804 set_ignored_loc callbacks.
16805 * debug.c (do_nothing_debug_hooks): Likewise.
16806 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
16807 * dwarf2out.c (text_section_used, cold_text_section_used): Remove.
16808 (in_text_section_p, last_text_label, last_cold_label,
16809 switch_text_ranges, switch_cold_ranges): New data items.
16810 (dwarf2out_note_section_used): Remove.
16811 (dwarf2out_begin_prologue): Set fde->ignored_debug and
16813 (mark_ignored_debug_section): New helper function.
16814 (dwarf2out_end_epilogue, dwarf2out_switch_text_section): Call
16815 mark_ignored_debug_section.
16816 (dwarf2_debug_hooks): Use dwarf2out_set_ignored_loc.
16817 (dwarf2_lineno_debug_hooks): Use dummy for set_ignored_loc.
16818 (size_of_aranges): Adjust formula for multi-part text ranges size.
16819 (output_aranges): Output multi-part text ranges.
16820 (dwarf2out_set_ignored_loc): New callback function.
16821 (dwarf2out_finish): Output multi-part text ranges.
16822 (dwarf2out_c_finalize): Clear new data items.
16823 * final.c (final_start_function_1): Call set_ignored_loc callback.
16824 (final_scan_insn_1): Likewise.
16825 * ggc-page.c (gt_ggc_mx): New helper function.
16826 * stringpool.c (gt_pch_nx): Likewise.
16828 2021-05-06 Richard Biener <rguenther@suse.de>
16830 * timevar.def (TV_TREE_INSERT_PHI_NODES): Remove.
16831 (TV_TREE_SSA_REWRITE_BLOCKS): Likewise.
16832 (TV_TREE_INTO_SSA): New.
16833 * tree-into-ssa.c (insert_phi_nodes): Do not account separately.
16834 (rewrite_blocks): Likewise.
16835 (pass_data_build_ssa): Account to TV_TREE_INTO_SSA.
16837 2021-05-06 Jakub Jelinek <jakub@redhat.com>
16839 * tree-ssa-phiopt.c (value_replacement, minmax_replacement,
16840 abs_replacement, xor_replacement,
16841 cond_removal_in_popcount_clz_ctz_pattern,
16842 replace_phi_edge_with_variable): Change type of phi argument from
16843 gimple * to gphi *.
16845 2021-05-06 Richard Biener <rguenther@suse.de>
16847 * tree-ssa-loop-split.c (split_loop): Delay updating SSA form.
16848 Output an opt-info message.
16849 (do_split_loop_on_cond): Likewise.
16850 (tree_ssa_split_loops): Update SSA form here.
16852 2021-05-06 Richard Biener <rguenther@suse.de>
16854 * tree-inline.c (tree_function_versioning): Fix DECL_BY_REFERENCE
16855 return variable removal.
16857 2021-05-06 Marius Hillenbrand <mhillen@linux.ibm.com>
16859 * config/s390/s390-builtins.def (O_M5, O1_M5, ...): Remove unused macros.
16860 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
16861 (s390_vec_permi_dbl, s390_vpdi): Use the O3_U2 type for the immediate
16863 * config/s390/s390.c (s390_const_operand_ok): Remove unused
16866 2021-05-06 Jakub Jelinek <jakub@redhat.com>
16868 PR tree-optimization/94589
16869 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Call
16870 spaceship_replacement.
16871 (cond_only_block_p, spaceship_replacement): New functions.
16873 2021-05-06 Richard Biener <rguenther@suse.de>
16876 * tree-emutls.c (gen_emutls_addr): Pass in whether we're
16877 dealing with a debug use and only query existing addresses
16879 (lower_emutls_1): Avoid splitting out addresses for debug
16880 stmts, reset the debug stmt when we fail to find existing
16882 (lower_emutls_phi_arg): Set wi.stmt.
16884 2021-05-06 Christoph Muellner <cmuellner@gcc.gnu.org>
16887 * config/riscv/riscv.c (riscv_block_move_loop): Use cbranch helper.
16888 * config/riscv/riscv.md (cbranch<mode>4): Generate helpers.
16889 (stack_protect_test): Use cbranch helper.
16891 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
16894 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
16895 always return the establisher frame for __builtin_frame_address (0).
16897 2021-05-05 Ivan Sorokin <vanyacpp@gmail.com>
16900 * config/i386/i386-builtins.c (ix86_cpu_model_type_node): New.
16901 (ix86_cpu_model_var): Likewise.
16902 (ix86_cpu_features2_type_node): Likewise.
16903 (ix86_cpu_features2_var): Likewise.
16904 (fold_builtin_cpu): Cache __cpu_model and __cpu_features2 with
16907 2021-05-05 Martin Sebor <msebor@redhat.com>
16909 * passes.def (pass_warn_printf): Run after SSA.
16911 2021-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16913 * config/arm/neon.md (neon_vtst_combine<mode>): New pattern.
16914 * config/arm/predicates.md (minus_one_operand): New predicate.
16916 2021-05-05 Jeff Law <jlaw@tachyum.com>
16918 * config/avr/avr.md: Remove references to CC_STATUS_INIT.
16920 2021-05-05 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
16922 PR rtl-optimization/100263
16923 * postreload.c (move2add_valid_value_p): Ensure register can
16926 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
16928 PR rtl-optimization/100411
16929 * cfgcleanup.c (try_crossjump_to_edge): Also skip end of prologue
16930 and beginning of function markers.
16932 2021-05-05 Jeff Law <jlaw@tachyum.com>
16934 * config/cr16/cr16.h (NOTICE_UPDATE_CC): Remove.
16935 * config/cr16/cr16.c (notice_update_cc): Remove.
16936 * config/cr16/cr16-protos.h (notice_update_cc): Remove.
16938 2021-05-05 Uroš Bizjak <ubizjak@gmail.com>
16941 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
16942 Handle V8QI, V4HI and V2SI modes.
16943 * config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
16944 (ix86_build_signbit_mask): Ditto.
16945 * config/i386/mmx.md (MMXMODE14): New mode iterator.
16946 (<smaxmin:code><MMXMODE14:mode>3): New expander.
16947 (*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
16948 (<umaxmin:code><MMXMODE24:mode>3): New expander.
16949 (*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
16950 (vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
16951 (vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
16952 (vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
16953 (vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
16954 (vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
16956 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
16958 * dwarf2out.c (loc_list_from_tree_1) <DECL>: During early DWARF, do
16959 not expand the VALUE_EXPR of variables put in the non-local frame.
16960 * gimplify.c (gimplify_type_sizes) <RECORD_TYPE>: If the type is not
16961 to be ignored for debug info, ensure its variable offsets are not.
16963 2021-05-05 Richard Biener <rguenther@suse.de>
16965 PR tree-optimization/79333
16966 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
16967 Fold stmt following SSA edges.
16969 2021-05-05 Richard Biener <rguenther@suse.de>
16971 PR middle-end/100394
16972 * calls.c (expand_call): Preserve possibly throwing calls.
16973 * cfgexpand.c (expand_call_stmt): When a call can throw signal
16974 RTL expansion there are side-effects.
16975 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Simplify,
16976 mark all possibly throwing stmts necessary unless we can elide
16978 * tree-ssa-dse.c (pass_dse::execute): Preserve exceptions unless
16979 -fdelete-dead-exceptions.
16980 * tree.h (DECL_PURE_P): Add note about exceptions.
16982 2021-05-05 Alexandre Oliva <oliva@adacore.com>
16984 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Make it
16987 2021-05-04 David Edelsohn <dje.gcc@gmail.com>
16989 * config/rs6000/rs6000-call.c (rs6000_output_mi_thunk): Use
16990 get_fnname_from_decl for name of thunk.
16991 * config/rs6000/rs6000.c (rs6000_declare_alias): Use assemble_name
16992 and ASM_OUTPUT_LABEL.
16993 (rs6000_xcoff_declare_function_name): Use assemble_name and
16995 (rs6000_xcoff_declare_object_name): Use ASM_OUTPUT_LABEL.
16996 (rs6000_xcoff_encode_section_info): Don't add mapping class
16997 for aliases. Always add [DS] mapping class to primary
16999 (rs6000_asm_weaken_decl): Don't explicitly add [DS].
17001 2021-05-04 Martin Sebor <msebor@redhat.com>
17003 PR middle-end/100307
17004 * builtins.c (compute_objsize_r): Clear base0 for pointers.
17006 2021-05-04 Jeff Law <jlaw@tachyum.com>
17008 * config/bfin/bfin.h (NOTICE_UPDATE_CC): Remove.
17010 2021-05-04 Segher Boessenkool <segher@kernel.crashing.org>
17012 * caller-save.c: Remove CC0.
17013 * cfgcleanup.c: Remove CC0.
17014 * cfgrtl.c: Remove CC0.
17015 * combine.c: Remove CC0.
17016 * compare-elim.c: Remove CC0.
17017 * conditions.h: Remove CC0.
17018 * config/h8300/h8300.h: Remove CC0.
17019 * config/h8300/h8300-protos.h: Remove CC0.
17020 * config/h8300/peepholes.md: Remove CC0.
17021 * config/i386/x86-tune-sched.c: Remove CC0.
17022 * config/m68k/m68k.c: Remove CC0.
17023 * config/rl78/rl78.c: Remove CC0.
17024 * config/sparc/sparc.c: Remove CC0.
17025 * config/xtensa/xtensa.c: Remove CC0.
17026 (gen_conditional_move): Use pc_rtx instead of cc0_rtx in a piece of
17027 RTL where that is used as a placeholder only.
17028 * cprop.c: Remove CC0.
17029 * cse.c: Remove CC0.
17030 * cselib.c: Remove CC0.
17031 * df-problems.c: Remove CC0.
17032 * df-scan.c: Remove CC0.
17033 * doc/md.texi: Remove CC0. Adjust an example.
17034 * doc/rtl.texi: Remove CC0. Adjust an example.
17035 * doc/tm.texi: Regenerate.
17036 * doc/tm.texi.in: Remove CC0.
17037 * emit-rtl.c: Remove CC0.
17038 * final.c: Remove CC0.
17039 * fwprop.c: Remove CC0.
17040 * gcse-common.c: Remove CC0.
17041 * gcse.c: Remove CC0.
17042 * genattrtab.c: Remove CC0.
17043 * genconfig.c: Remove CC0.
17044 * genemit.c: Remove CC0.
17045 * genextract.c: Remove CC0.
17046 * gengenrtl.c: Remove CC0.
17047 * genrecog.c: Remove CC0.
17048 * haifa-sched.c: Remove CC0.
17049 * ifcvt.c: Remove CC0.
17050 * ira-costs.c: Remove CC0.
17051 * ira.c: Remove CC0.
17052 * jump.c: Remove CC0.
17053 * loop-invariant.c: Remove CC0.
17054 * lra-constraints.c: Remove CC0.
17055 * lra-eliminations.c: Remove CC0.
17056 * optabs.c: Remove CC0.
17057 * postreload-gcse.c: Remove CC0.
17058 * postreload.c: Remove CC0.
17059 * print-rtl.c: Remove CC0.
17060 * read-rtl-function.c: Remove CC0.
17061 * reg-notes.def: Remove CC0.
17062 * reg-stack.c: Remove CC0.
17063 * reginfo.c: Remove CC0.
17064 * regrename.c: Remove CC0.
17065 * reload.c: Remove CC0.
17066 * reload1.c: Remove CC0.
17067 * reorg.c: Remove CC0.
17068 * resource.c: Remove CC0.
17069 * rtl.c: Remove CC0.
17070 * rtl.def: Remove CC0.
17071 * rtl.h: Remove CC0.
17072 * rtlanal.c: Remove CC0.
17073 * sched-deps.c: Remove CC0.
17074 * sched-rgn.c: Remove CC0.
17075 * shrink-wrap.c: Remove CC0.
17076 * simplify-rtx.c: Remove CC0.
17077 * system.h: Remove CC0. Poison NOTICE_UPDATE_CC, CC_STATUS_MDEP_INIT,
17078 CC_STATUS_MDEP, and CC_STATUS.
17079 * target.def: Remove CC0.
17080 * valtrack.c: Remove CC0.
17081 * var-tracking.c: Remove CC0.
17083 2021-05-04 Richard Biener <rguenther@suse.de>
17085 PR tree-optimization/100414
17086 * tree-ssa-phiopt.c (get_non_trapping): Do not compute dominance
17088 (tree_ssa_phiopt_worker): But unconditionally here.
17090 2021-05-04 Tobias Burnus <tobias@codesourcery.com>
17092 * omp-low.c (lower_rec_input_clauses, lower_reduction_clauses): Handle
17093 && and || with floating-point and complex arguments.
17095 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
17097 * tree-inline.c (insert_debug_decl_map): Delete.
17098 (copy_debug_stmt): Minor tweak.
17099 (setup_one_parameter): Do not use a variable if the value is either
17100 a read-only DECL or a non-addressable local variable in the caller.
17101 In this case, insert the debug-only variable in the map manually.
17102 (expand_call_inline): Do not generate a CLOBBER for these values.
17103 * tree-inline.h (debug_map): Minor tweak.
17105 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
17107 * builtins.c (builtin_with_linkage_p): Return true for stp[n]cpy.
17108 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Tidy up.
17110 2021-05-04 Richard Biener <rguenther@suse.de>
17112 PR tree-optimization/100329
17113 * tree-ssa-reassoc.c (can_reassociate_p): Do not reassociate
17115 (insert_stmt_after): Assert we're not running into asm goto.
17117 2021-05-04 Richard Biener <rguenther@suse.de>
17119 PR tree-optimization/100398
17120 * tree-ssa-dse.c (pass_dse::execute): Preserve control
17123 2021-05-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17125 * builtins.c (try_store_by_multiple_pieces): Fix constfun's prototype.
17127 2021-05-04 Alexandre Oliva <oliva@adacore.com>
17129 * builtins.c (try_store_by_multiple_pieces): New.
17130 (expand_builtin_memset_args): Use it. If target_char_cast
17131 fails, proceed as for non-constant val. Pass len's ctz to...
17132 * expr.c (clear_storage_hints): ... this. Try store by
17133 multiple pieces after setmem.
17134 (clear_storage): Adjust.
17135 * expr.h (clear_storage_hints): Likewise.
17136 (try_store_by_multiple_pieces): Declare.
17137 * passes.def: Replace the last copy_prop with ccp.
17139 2021-05-03 Tom de Vries <tdevries@suse.de>
17142 * omp-low.c (lower_rec_input_clauses): Disable SIMT for user-defined
17145 2021-05-03 Richard Biener <rguenther@suse.de>
17147 * tree-ssa-dse.c (dse_classify_store): Track two PHI defs.
17149 2021-05-03 Richard Biener <rguenther@suse.de>
17151 * tree-ssa-dse.c: Do not include domwalk.h but cfganal.h.
17152 (dse_dom_walker): Remove.
17153 (dse_dom_walker::dse_optimize_stmt): Rename...
17154 (dse_optimize_stmt): ... to this, pass in live_bytes sbitmap.
17155 (dse_dom_walker::before_dom_children): Inline ...
17156 (pass_dse::execute): ... here. Perform a reverse program
17159 2021-05-03 H.J. Lu <hjl.tools@gmail.com>
17162 * configure: Regenerated.
17164 2021-05-03 Ilya Leoshkevich <iii@linux.ibm.com>
17167 * config/s390/s390.c (s390_hard_fp_reg_p): New function.
17168 (s390_md_asm_adjust): Handle hard registers.
17170 2021-05-03 Jakub Jelinek <jakub@redhat.com>
17172 PR tree-optimization/100382
17173 * tree-ssa-dse.c: Include tree-eh.h.
17174 (dse_dom_walker::before_dom_children): Don't remove stmts if
17175 stmt_unremovable_because_of_non_call_eh_p is true.
17177 2021-05-02 David Edelsohn <dje.gcc@gmail.com>
17179 * varasm.c (compute_reloc_for_var): Split out from...
17180 (get_variable_section): Use it.
17181 * output.h (compute_reloc_for_var): Declare.
17182 * config/rs6000/rs6000-protos.h
17183 (rs6000_xcoff_asm_output_aligned_decl_common): Change alignment to
17185 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address_aix):
17186 Don't append storage mapping class to symbol.
17187 (rs6000_xcoff_asm_named_section): Add BS and UL mapping classes.
17188 Don't convert TLS BSS to common.
17189 (rs6000_xcoff_unique_section): Don't fall back to select_secton.
17190 (rs6000_xcoff_section_type_flags): Add SECTION_BSS if DECL is
17192 (rs6000_xcoff_asm_globalize_decl_name): Don't strip storage
17194 (rs6000_xcoff_asm_output_aligned_decl_common): Align is unsigned int.
17195 If align is 0 from TLS class, use the same rules as varasm.c
17196 If not common, switch to BSS section manually.
17197 If common, emit appropriate comm or lcomm directive.
17198 (rs6000_xcoff_encode_section_info): Add logic to append all
17199 storage mapping classes.
17200 (rs6000_asm_weaken_decl): Adjust for qualname symbols.
17201 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
17202 rs6000_xcoff_asm_output_aligned_decl_common.
17203 (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
17204 rs6000_xcoff_asm_output_aligned_decl_common.
17205 (ASM_OUTPUT_TLS_COMMON): Use
17206 rs6000_xcoff_asm_output_aligned_decl_common.
17208 2021-05-02 Jakub Jelinek <jakub@redhat.com>
17211 * config/nvptx/nvptx.c (nvptx_sese_pseudo): Use nullptr instead of 0
17212 as first argument of pseudo_node_t constructors.
17214 2021-05-02 Jakub Jelinek <jakub@redhat.com>
17217 * config/i386/t-i386 (TM_H): Add $(srcdir)/config/i386/i386-isa.def.
17219 2021-05-01 Aldy Hernandez <aldyh@redhat.com>
17221 * value-range.cc (DEFINE_INT_RANGE_GC_STUBS): Remove.
17222 (gt_pch_nx (int_range<1> *&)): New.
17223 (gt_ggc_mx (int_range<1> *&)): New.
17224 * value-range.h (class irange): Add GTY support for
17227 2021-05-01 Geng Qi <gengqi@linux.alibaba.com>
17229 * doc/options.texi (Negative): Change either or to both and.
17231 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17233 * config/aarch64/aarch64-simd-builtins.def: Add
17234 float_ml[as][q]_laneq builtin generator macros.
17235 * config/aarch64/aarch64-simd.md (mul_laneq<mode>3): Define.
17236 (aarch64_float_mla_laneq<mode>): Define.
17237 (aarch64_float_mls_laneq<mode>): Define.
17238 * config/aarch64/arm_neon.h (vmla_laneq_f32): Use RTL builtin
17239 instead of GCC vector extensions.
17240 (vmlaq_laneq_f32): Likewise.
17241 (vmls_laneq_f32): Likewise.
17242 (vmlsq_laneq_f32): Likewise.
17244 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17246 * config/aarch64/aarch64-simd-builtins.def: Add
17247 float_ml[as]_lane builtin generator macros.
17248 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt<mode>):
17250 (mul_lane<mode>3): This, and re-order arguments.
17251 (aarch64_float_mla_lane<mode>): Define.
17252 (aarch64_float_mls_lane<mode>): Define.
17253 * config/aarch64/arm_neon.h (vmla_lane_f32): Use RTL builtin
17254 instead of GCC vector extensions.
17255 (vmlaq_lane_f32): Likewise.
17256 (vmls_lane_f32): Likewise.
17257 (vmlsq_lane_f32): Likewise.
17259 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17261 * config/aarch64/aarch64-simd-builtins.def: Add float_ml[as]
17262 builtin generator macros.
17263 * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>):
17265 (aarch64_float_mls<mode>): Define.
17266 * config/aarch64/arm_neon.h (vmla_f32): Use RTL builtin
17267 instead of relying on GCC vector extensions.
17268 (vmla_f64): Likewise.
17269 (vmlaq_f32): Likewise.
17270 (vmlaq_f64): Likewise.
17271 (vmls_f32): Likewise.
17272 (vmls_f64): Likewise.
17273 (vmlsq_f32): Likewise.
17274 (vmlsq_f64): Likewise.
17275 * config/aarch64/iterators.md: Define VDQF_DF mode iterator.
17277 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17279 * config/aarch64/aarch64-simd-builtins.def: Add
17280 float_ml[as]_n_builtin generator macros.
17281 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt_from_dup<mode>):
17283 (mul_n<mode>3): This, and re-order arguments.
17284 (aarch64_float_mla_n<mode>): Define.
17285 (aarch64_float_mls_n<mode>): Define.
17286 * config/aarch64/arm_neon.h (vmla_n_f32): Use RTL builtin
17287 instead of inline asm.
17288 (vmlaq_n_f32): Likewise.
17289 (vmls_n_f32): Likewise.
17290 (vmlsq_n_f32): Likewise.
17292 2021-04-30 Jonathan Wright <joanthan.wright@arm.com>
17294 * config/aarch64/aarch64-simd-builtins.def: Add pmull[2]
17295 builtin generator macros.
17296 * config/aarch64/aarch64-simd.md (aarch64_pmullv8qi): Define.
17297 (aarch64_pmull_hiv16qi_insn): Define.
17298 (aarch64_pmull_hiv16qi): Define.
17299 * config/aarch64/arm_neon.h (vmull_high_p8): Use RTL builtin
17300 instead of inline asm.
17301 (vmull_p8): Likewise.
17303 2021-04-30 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
17305 * config/avr/avr.md: Adjust peepholes to match and
17306 generate parallels with clobber of REG_CC.
17307 (mov<mode>_insn): Rename to mov<mode>_insn_split.
17308 (*mov<mode>_insn): Rename to mov<mode>_insn.
17310 2021-04-30 David Edelsohn <dje.gcc@gmail.com>
17312 * varasm.c (use_blocks_for_decl_p): Don't use section anchors
17313 for VAR_DECLs if -fdata-sections enabled.
17315 2021-04-30 Michael Meissner <meissner@linux.ibm.com>
17317 PR bootstrap/100327
17318 * config/rs6000/rs6000.c
17319 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Define.
17320 (rs6000_libgcc_floating_mode_supported_p): New target hook.
17322 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
17324 * tree-ssa-threadbackward.c (class thread_jumps): Split out code
17326 (class back_threader_registry): ...to here...
17327 (class back_threader_profitability): ...and here...
17328 (thread_jumps::thread_through_all_blocks): Remove argument.
17329 (back_threader_registry::back_threader_registry): New.
17330 (back_threader_registry::~back_threader_registry): New.
17331 (back_threader_registry::thread_through_all_blocks): New.
17332 (thread_jumps::profitable_jump_thread_path): Move from here...
17333 (back_threader_profitability::profitable_path_p): ...to here.
17334 (thread_jumps::find_taken_edge): New.
17335 (thread_jumps::convert_and_register_current_path): Move...
17336 (back_threader_registry::register_path): ...to here.
17337 (thread_jumps::register_jump_thread_path_if_profitable): Move...
17338 (thread_jumps::maybe_register_path): ...to here.
17339 (thread_jumps::handle_phi): Call find_taken_edge and
17340 maybe_register_path.
17341 (thread_jumps::handle_assignment): Same.
17342 (thread_jumps::fsm_find_control_statement_thread_paths): Remove
17343 tree argument to handle_phi and handle_assignment.
17344 (thread_jumps::find_jump_threads_backwards): Set m_name. Remove
17345 set of m_speed_p and m_max_threaded_paths.
17346 (pass_thread_jumps::execute): Remove second argument from
17347 find_jump_threads_backwards.
17348 (pass_early_thread_jumps::execute): Same.
17350 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
17352 * tree-ssa-dom.c (class dom_jump_threader_simplifier): New.
17353 (class dom_opt_dom_walker): Initialize some class variables.
17354 (pass_dominator::execute): Pass evrp_range_analyzer and
17355 dom_jump_threader_simplifier to dom_opt_dom_walker.
17356 Adjust for some functions moving into classes.
17357 (simplify_stmt_for_jump_threading): Adjust and move to...
17358 (jump_threader_simplifier::simplify): ...here.
17359 (dom_opt_dom_walker::before_dom_children): Adjust for
17360 m_evrp_range_analyzer.
17361 (dom_opt_dom_walker::after_dom_children): Remove x_vr_values hack.
17362 (test_for_singularity): Place in dom_opt_dom_walker class.
17363 (dom_opt_dom_walker::optimize_stmt): The argument
17364 evrp_range_analyzer is now a class field.
17365 * tree-ssa-threadbackward.c (class thread_jumps): Add m_registry.
17366 (thread_jumps::thread_through_all_blocks): New.
17367 (thread_jumps::convert_and_register_current_path): Use m_registry.
17368 (pass_thread_jumps::execute): Adjust for thread_through_all_blocks
17369 being in the threader class.
17370 (pass_early_thread_jumps::execute): Same.
17371 * tree-ssa-threadedge.c (threadedge_initialize_values): Move...
17372 (jump_threader::jump_threader): ...here.
17373 (threadedge_finalize_values): Move...
17374 (jump_threader::~jump_threader): ...here.
17375 (jump_threader::remove_jump_threads_including): New.
17376 (jump_threader::thread_through_all_blocks): New.
17377 (record_temporary_equivalences_from_phis): Move...
17378 (jump_threader::record_temporary_equivalences_from_phis): ...here.
17379 (record_temporary_equivalences_from_stmts_at_dest): Move...
17380 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
17382 (simplify_control_stmt_condition_1): Move to jump_threader class.
17383 (simplify_control_stmt_condition): Move...
17384 (jump_threader::simplify_control_stmt_condition): ...here.
17385 (thread_around_empty_blocks): Move...
17386 (jump_threader::thread_around_empty_blocks): ...here.
17387 (thread_through_normal_block): Move...
17388 (jump_threader::thread_through_normal_block): ...here.
17389 (thread_across_edge): Move...
17390 (jump_threader::thread_across_edge): ...here.
17391 (thread_outgoing_edges): Move...
17392 (jump_threader::thread_outgoing_edges): ...here.
17393 * tree-ssa-threadedge.h: Move externally facing functings...
17394 (class jump_threader): ...here...
17395 (class jump_threader_simplifier): ...and here.
17396 * tree-ssa-threadupdate.c (struct redirection_data): Remove comment.
17397 (jump_thread_path_allocator::jump_thread_path_allocator): New.
17398 (jump_thread_path_allocator::~jump_thread_path_allocator): New.
17399 (jump_thread_path_allocator::allocate_thread_edge): New.
17400 (jump_thread_path_allocator::allocate_thread_path): New.
17401 (jump_thread_path_registry::jump_thread_path_registry): New.
17402 (jump_thread_path_registry::~jump_thread_path_registry): New.
17403 (jump_thread_path_registry::allocate_thread_edge): New.
17404 (jump_thread_path_registry::allocate_thread_path): New.
17405 (dump_jump_thread_path): Make extern.
17406 (debug (const vec<jump_thread_edge *> &path)): New.
17407 (struct removed_edges): Move to tree-ssa-threadupdate.h.
17408 (struct thread_stats_d): Remove.
17409 (remove_ctrl_stmt_and_useless_edges): Make static.
17410 (lookup_redirection_data): Move...
17411 (jump_thread_path_registry::lookup_redirection_data): ...here.
17412 (ssa_redirect_edges): Make static.
17413 (thread_block_1): Move...
17414 (jump_thread_path_registry::thread_block_1): ...here.
17415 (thread_block): Move...
17416 (jump_thread_path_registry::thread_block): ...here.
17417 (thread_through_loop_header): Move...
17418 (jump_thread_path_registry::thread_through_loop_header): ...here.
17419 (mark_threaded_blocks): Move...
17420 (jump_thread_path_registry::mark_threaded_blocks): ...here.
17421 (debug_path): Move...
17422 (jump_thread_path_registry::debug_path): ...here.
17423 (debug_all_paths): Move...
17424 (jump_thread_path_registry::dump): ..here.
17425 (rewire_first_differing_edge): Move...
17426 (jump_thread_path_registry::rewire_first_differing_edge): ...here.
17427 (adjust_paths_after_duplication): Move...
17428 (jump_thread_path_registry::adjust_paths_after_duplication): ...here.
17429 (duplicate_thread_path): Move...
17430 (jump_thread_path_registry::duplicate_thread_path): ..here.
17431 (remove_jump_threads_including): Move...
17432 (jump_thread_path_registry::remove_jump_threads_including): ...here.
17433 (thread_through_all_blocks): Move to...
17434 (jump_thread_path_registry::thread_through_all_blocks): ...here.
17435 (delete_jump_thread_path): Remove.
17436 (register_jump_thread): Move...
17437 (jump_thread_path_registry::register_jump_thread): ...here.
17438 * tree-ssa-threadupdate.h: Move externally facing functions...
17439 (class jump_thread_path_allocator): ...here...
17440 (class jump_thread_path_registry): ...and here.
17441 (thread_through_all_blocks): Remove.
17442 (struct removed_edges): New.
17443 (register_jump_thread): Remove.
17444 (remove_jump_threads_including): Remove.
17445 (delete_jump_thread_path): Remove.
17446 (remove_ctrl_stmt_and_useless_edges): Remove.
17447 (free_dom_edge_info): New prototype.
17448 * tree-vrp.c: Remove x_vr_values hack.
17449 (class vrp_jump_threader_simplifier): New.
17450 (vrp_jump_threader_simplifier::simplify): New.
17451 (vrp_jump_threader::vrp_jump_threader): Adjust method signature.
17452 Remove m_dummy_cond.
17453 Instantiate m_simplifier and m_threader.
17454 (vrp_jump_threader::thread_through_all_blocks): New.
17455 (vrp_jump_threader::simplify_stmt): Remove.
17456 (vrp_jump_threader::after_dom_children): Do not set m_dummy_cond.
17457 Remove x_vr_values hack.
17458 (execute_vrp): Adjust for thread_through_all_blocks being in a
17461 2021-04-30 Christophe Lyon <christophe.lyon@linaro.org>
17463 * genflags.c (gen_insn): Print failed expansion string.
17465 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
17467 * expr.c (alignment_for_piecewise_move): Call mode_for_size
17468 without limit to MAX_FIXED_MODE_SIZE.
17470 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
17472 PR middle-end/90773
17473 * builtins.c (builtin_memset_gen_str): Don't use return from
17474 simplify_gen_subreg.
17476 2021-04-30 Uroš Bizjak <ubizjak@gmail.com>
17479 * config/i386/i386.md (*add<mode>3_carry_0r): New insn pattern.
17480 (*addsi3_carry_zext_0r): Ditto.
17481 (*sub<mode>3_carry_0): Ditto.
17482 (*subsi3_carry_zext_0r): Ditto.
17483 * config/i386/predicates.md (ix86_carry_flag_unset_operator):
17485 * config/i386/i386.c (ix86_rtx_costs) <case PLUS, case MINUS>:
17486 Also consider ix86_carry_flag_unset_operator to calculate
17487 the cost of adc/sbb insn.
17489 2021-04-30 Roman Zhuykov <zhroma@ispras.ru>
17491 PR rtl-optimization/100225
17492 PR rtl-optimization/84878
17493 * modulo-sched.c (sms_schedule): Use note_stores to skip loops
17494 where we have an instruction which touches (writes) any hard
17495 register from df->regular_block_artificial_uses set.
17496 Allow not-single-set instruction only right before basic block
17499 2021-04-30 Geng Qi <gengqi@linux.alibaba.com>
17501 * config/riscv/riscv.opt (march=,mabi=): Negative itself.
17503 2021-04-30 LevyHsu <admin@levyhsu.com>
17505 * config/riscv/riscv.c (riscv_min_arithmetic_precision): New.
17506 * config/riscv/riscv.h (TARGET_MIN_ARITHMETIC_PRECISION): New.
17507 * config/riscv/riscv.md (addv<mode>4, uaddv<mode>4): New.
17508 (subv<mode>4, usubv<mode>4, mulv<mode>4, umulv<mode>4): New.
17510 2021-04-29 Alexandre Oliva <oliva@adacore.com>
17512 * config.gcc: Merged x86 and x86_64 cpu_type-setting cases.
17514 2021-04-29 Alexandre Oliva <oliva@adacore.com>
17516 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Rename to...
17517 (ASM_OUTPUT_MAX_SKIP_ALIGN): ... this. Enclose in do/while(0).
17518 * config/i386/i386.c: Adjust.
17519 * config/i386/i386.md: Adjust.
17520 * config/i386/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Drop.
17521 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17522 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17523 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17524 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17525 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17526 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17527 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17528 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17529 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17530 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
17532 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
17534 * config/i386/i386-expand.c (ix86_expand_int_compare):
17535 Swap operands of GTU and LEU comparison to emit carry flag comparison.
17536 * config/i386/i386.md (*add<mode>3_carry_0): Change insn
17537 predicate to allow more combine opportunities with memory operands.
17538 (*sub<mode>3_carry_0): Ditto.
17540 2021-04-29 Richard Sandiford <richard.sandiford@arm.com>
17542 PR rtl-optimization/100303
17543 * rtl-ssa/accesses.cc (function_info::make_use_available): Take a
17544 boolean that indicates whether the use will only be used in
17545 debug instructions. Treat it in the same way that existing
17546 cross-EBB debug references would be handled if so.
17547 (function_info::make_uses_available): Likewise.
17548 * rtl-ssa/functions.h (function_info::make_uses_available): Update
17549 prototype accordingly.
17550 (function_info::make_uses_available): Likewise.
17551 * fwprop.c (try_fwprop_subst): Update call accordingly.
17553 2021-04-29 Jeff Law <jlaw@tachyum.com>
17555 * config/nios2/nios2-protos.h (nios2_fpu_insn_enabled): Move outside
17558 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
17559 Richard Biener <rguenther@suse.de>
17562 * config/i386/i386-builtin.def (IX86_BUILTIN_MASKLOADPD)
17563 (IX86_BUILTIN_MASKLOADPS, IX86_BUILTIN_MASKLOADPD256)
17564 (IX86_BUILTIN_MASKLOADPS256, IX86_BUILTIN_MASKLOADD)
17565 (IX86_BUILTIN_MASKLOADQ, IX86_BUILTIN_MASKLOADD256)
17566 (IX86_BUILTIN_MASKLOADQ256): Move from SPECIAL_ARGS
17567 to PURE_ARGS category.
17568 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
17569 Handle PURE_ARGS category.
17570 * config/i386/i386-expand.c (ix86_expand_builtin): Ditto.
17572 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
17574 * configure.ac: Check for the presence of sys/locking.h header and
17575 for whether _LK_LOCK is supported by _locking.
17576 * configure: Regenerate.
17577 * config.in: Likewise.
17578 * gcov-io.h: Define GCOV_LOCKED_WITH_LOCKING if HOST_HAS_LK_LOCK.
17579 * gcov-io.c (gcov_open): Add support for GCOV_LOCKED_WITH_LOCKING.
17580 * system.h: Include <sys/locking.h> if HAVE_SYS_LOCKING_H.
17582 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
17584 * config/i386/predicates.md (fcmov_comparison_operator):
17585 Do not check for trivial FP comparison operator.
17586 <case GEU, case LTU>: Allow CCGZmode.
17587 <case GTU, case LEU>: Do not allow CCCmode.
17588 (ix86_comparison_operator) <case GTU, case LEU>: Allow only CCmode.
17589 (ix86_carry_flag_operator): Match only LTU and UNLT code.
17590 Do not check for trivial FP comparison operator. Allow CCGZmode.
17592 2021-04-29 Tom de Vries <tdevries@suse.de>
17594 * omp-expand.c (expand_omp_simd): Add step_orig, and replace uses of
17595 fd->loop.step by either step or orig_step.
17597 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
17599 * config/sparc/sparc.c (gen_load_pcrel_sym): Delete.
17600 (load_got_register): Do the PIC dance here.
17601 (sparc_legitimize_tls_address): Simplify.
17602 (sparc_emit_probe_stack_range): Likewise.
17603 (sparc32_initialize_trampoline): Likewise.
17604 (sparc64_initialize_trampoline): Likewise.
17605 * config/sparc/sparc.md (load_pcrel_sym<P:mode>): Add @ marker.
17606 (probe_stack_range<P:mode>): Likewise.
17607 (flush<P:mode>): Likewise.
17608 (tgd_hi22<P:mode>): Likewise.
17609 (tgd_lo10<P:mode>): Likewise.
17610 (tgd_add<P:mode>): Likewise.
17611 (tgd_call<P:mode>): Likewise.
17612 (tldm_hi22<P:mode>): Likewise.
17613 (tldm_lo10<P:mode>): Likewise.
17614 (tldm_add<P:mode>): Likewise.
17615 (tldm_call<P:mode>): Likewise.
17616 (tldo_hix22<P:mode>): Likewise.
17617 (tldo_lox10<P:mode>): Likewise.
17618 (tldo_add<P:mode>): Likewise.
17619 (tie_hi22<P:mode>): Likewise.
17620 (tie_lo10<P:mode>): Likewise.
17621 (tie_add<P:mode>): Likewise.
17622 (tle_hix22<P:mode>): Likewise.
17623 (tle_lox10<P:mode>): Likewise.
17624 (stack_protect_setsi): Rename to...
17625 (stack_protect_set32): ...this.
17626 (stack_protect_setdi): Rename to...
17627 (stack_protect_set64): ...this.
17628 (stack_protect_set): Adjust calls to above.
17629 (stack_protect_testsi): Rename to...
17630 (stack_protect_test32): ...this.
17631 (stack_protect_testdi): Rename to...
17632 (stack_protect_test64): ...this.
17633 (stack_protect_test): Adjust calls to above.
17635 2021-04-29 H.J. Lu <hjl.tools@gmail.com>
17637 PR middle-end/90773
17638 * builtins.c (builtin_memcpy_read_str): Add a dummy argument.
17639 (builtin_strncpy_read_str): Likewise.
17640 (builtin_memset_read_str): Add an argument for the previous RTL
17641 information and generate the new RTL from the previous RTL info.
17642 (builtin_memset_gen_str): Likewise.
17643 * builtins.h (builtin_strncpy_read_str): Update the prototype.
17644 (builtin_memset_read_str): Likewise.
17645 * expr.c (by_pieces_ninsns): If targetm.overlap_op_by_pieces_p()
17646 returns true, round up size and alignment to the widest integer
17647 mode for maximum size.
17648 (pieces_addr::adjust): Add a pointer to by_pieces_prev argument
17649 and pass it to m_constfn.
17650 (op_by_pieces_d): Add m_push and m_overlap_op_by_pieces.
17651 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
17652 initialize m_push. Initialize m_overlap_op_by_pieces with
17653 targetm.overlap_op_by_pieces_p ().
17654 (op_by_pieces_d::run): Pass the previous RTL information to
17655 pieces_addr::adjust and generate overlapping operations if
17656 m_overlap_op_by_pieces is true.
17658 (move_by_pieces_d::move_by_pieces_d): Updated for op_by_pieces_d
17660 (store_by_pieces_d::store_by_pieces_d): Updated for op_by_pieces_d
17662 (can_store_by_pieces): Use by_pieces_constfn on constfun.
17663 (store_by_pieces): Use by_pieces_constfn on constfun. Updated
17664 for op_by_pieces_d change.
17665 (clear_by_pieces_1): Add a dummy argument.
17666 (clear_by_pieces): Updated for op_by_pieces_d change.
17667 (compare_by_pieces_d::compare_by_pieces_d): Likewise.
17668 (string_cst_read_str): Add a dummy argument.
17669 * expr.h (by_pieces_constfn): Add a dummy argument.
17670 (by_pieces_prev): New.
17671 * target.def (overlap_op_by_pieces_p): New target hook.
17672 * config/i386/i386.c (TARGET_OVERLAP_OP_BY_PIECES_P): New.
17673 * doc/tm.texi.in: Add TARGET_OVERLAP_OP_BY_PIECES_P.
17674 * doc/tm.texi: Regenerated.
17676 2021-04-29 Richard Biener <rguenther@suse.de>
17678 PR tree-optimization/100253
17679 * tree-vect-stmts.c (vectorizable_load): Do not assume
17680 element alignment when DR_MISALIGNMENT is -1.
17681 (vectorizable_store): Likewise.
17683 2021-04-29 Jakub Jelinek <jakub@redhat.com>
17686 * config/aarch64/aarch64.c (aarch64_add_offset_1_temporaries): Use
17687 absu_hwi instead of abs_hwi.
17689 2021-04-29 Richard Biener <rguenther@suse.de>
17691 PR middle-end/38474
17692 * tree-ssa-structalias.c (add_graph_edge): Avoid direct
17693 forwarding when indirect forwarding through ESCAPED
17696 2021-04-29 Tom de Vries <tdevries@suse.de>
17699 * internal-fn.c (expand_GOMP_SIMT_ENTER_ALLOC)
17700 (expand_GOMP_SIMT_LAST_LANE, expand_GOMP_SIMT_ORDERED_PRED)
17701 (expand_GOMP_SIMT_VOTE_ANY, expand_GOMP_SIMT_XCHG_BFLY)
17702 (expand_GOMP_SIMT_XCHG_IDX): Ensure target is assigned to.
17704 2021-04-29 Richard Biener <rguenther@suse.de>
17706 PR tree-optimization/99912
17707 * tree-ssa-dse.c (dse_dom_walker::m_need_cfg_cleanup): New.
17708 (dse_dom_walker::todo): Likewise.
17709 (dse_dom_walker::dse_optimize_stmt): Move VDEF check to the
17711 (dse_dom_walker::before_dom_children): Remove trivially
17712 dead SSA defs and schedule CFG cleanup if we removed all
17714 (pass_dse::execute): Get TODO as computed by the DOM walker
17715 and return it. Wipe dominator info earlier.
17717 2021-04-29 Richard Biener <rguenther@suse.de>
17720 * ipa-prop.c (ipcp_modif_dom_walker::before_dom_children):
17721 Track blocks to cleanup EH in new m_need_eh_cleanup.
17722 (ipcp_modif_dom_walker::cleanup_eh): New.
17723 (ipcp_transform_function): Release dominator info before
17726 2021-04-29 Martin Sebor <msebor@redhat.com>
17728 PR middle-end/100250
17729 * attribs.c (attr_access::array_as_string): Avoid dereferencing
17730 a pointer when it's null.
17732 2021-04-29 Martin Sebor <msebor@redhat.com>
17734 * Makefile.in (OBJS): Add ipa-free-lang-data.o.
17735 * ipa-free-lang-data.cc: New file.
17736 * tree.c: Move pass free_lang_data to file above.
17737 (build_array_type_1): Declare extern.
17738 * tree.h (build_array_type_1): Declare.
17740 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17742 * config/aarch64/aarch64-simd-builtins.def: Modify comment to
17743 make consistent with updated RTL pattern.
17744 * config/aarch64/aarch64-simd.md (aarch64_<sur>qmovn<mode>):
17745 Implement using ss_truncate and us_truncate rather than
17747 * config/aarch64/iterators.md: Remove redundant unspecs and
17748 iterator: UNSPEC_[SU]QXTN and SUQMOVN respectively.
17750 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17752 * config/aarch64/arm_acle.h (__attribute__): Make intrinsic
17753 attributes consistent with those defined in arm_neon.h.
17755 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17757 * config/aarch64/arm_fp16.h (__attribute__): Make intrinsic
17758 attributes consistent with those defined in arm_neon.h.
17760 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17762 * config/aarch64/aarch64-simd-builtins.def: Add
17763 float_trunc_rodd builtin generator macros.
17764 * config/aarch64/aarch64-simd.md (aarch64_float_trunc_rodd_df):
17766 (aarch64_float_trunc_rodd_lo_v2sf): Define.
17767 (aarch64_float_trunc_rodd_hi_v4sf_le): Define.
17768 (aarch64_float_trunc_rodd_hi_v4sf_be): Define.
17769 (aarch64_float_trunc_rodd_hi_v4sf): Define.
17770 * config/aarch64/arm_neon.h (vcvtx_f32_f64): Use RTL builtin
17771 instead of inline asm.
17772 (vcvtx_high_f32_f64): Likewise.
17773 (vcvtxd_f32_f64): Likewise.
17774 * config/aarch64/iterators.md: Add FCVTXN unspec.
17776 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17778 * config/aarch64/aarch64-simd-builtins.def: Add tbx1 builtin
17780 * config/aarch64/aarch64-simd.md (aarch64_tbx1<mode>):
17782 * config/aarch64/arm_neon.h (vqtbx1_s8): USE RTL builtin
17783 instead of inline asm.
17784 (vqtbx1_u8): Likewise.
17785 (vqtbx1_p8): Likewise.
17786 (vqtbx1q_s8): Likewise.
17787 (vqtbx1q_u8): Likewise.
17788 (vqtbx1q_p8): Likewise.
17789 (vtbx2_s8): Likewise.
17790 (vtbx2_u8): Likewise.
17791 (vtbx2_p8): Likewise.
17793 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17795 * config/aarch64/aarch64-simd-builtins.def: Add tbl1 builtin
17797 * config/aarch64/arm_neon.h (vqtbl1_p8): Use RTL builtin
17798 instead of inline asm.
17799 (vqtbl1_s8): Likewise.
17800 (vqtbl1_u8): Likewise.
17801 (vqtbl1q_p8): Likewise.
17802 (vqtbl1q_s8): Likewise.
17803 (vqtbl1q_u8): Likewise.
17804 (vtbl1_s8): Likewise.
17805 (vtbl1_u8): Likewise.
17806 (vtbl1_p8): Likewise.
17807 (vtbl2_s8): Likewise.
17808 (vtbl2_u8): Likewise.
17809 (vtbl2_p8): Likewise.
17811 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17813 * config/aarch64/aarch64-simd-builtins.def: Add polynomial
17814 ssri_n buitin generator macro.
17815 * config/aarch64/arm_neon.h (vsri_n_p8): Use RTL builtin
17816 instead of inline asm.
17817 (vsri_n_p16): Likewise.
17818 (vsri_n_p64): Likewise.
17819 (vsriq_n_p8): Likewise.
17820 (vsriq_n_p16): Likewise.
17821 (vsriq_n_p64): Likewise.
17823 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17825 * config/aarch64/aarch64-simd-builtins.def: Use VALLP mode
17826 iterator for polynomial ssli_n builtin generator macro.
17827 * config/aarch64/arm_neon.h (vsli_n_p8): Use RTL builtin
17828 instead of inline asm.
17829 (vsli_n_p16): Likewise.
17830 (vsliq_n_p8): Likewise.
17831 (vsliq_n_p16): Likewise.
17832 * config/aarch64/iterators.md: Define VALLP mode iterator.
17834 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17836 * config/aarch64/aarch64-simd-builtins.def: Use VDQV_L
17837 iterator to generate [su]adalp RTL builtins.
17838 * config/aarch64/aarch64-simd.md: Use VDQV_L iterator in
17839 [su]adalp RTL pattern.
17840 * config/aarch64/arm_neon.h (vpadal_s32): Use RTL builtin
17841 instead of inline asm.
17842 (vpadal_u32): Likewise.
17844 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17846 * config/aarch64/aarch64-simd-builtins.def: Add [su]addlp
17847 builtin generator macros.
17848 * config/aarch64/aarch64-simd.md (aarch64_<su>addlp<mode>):
17850 * config/aarch64/arm_neon.h (vpaddl_s8): Use RTL builtin
17851 instead of inline asm.
17852 (vpaddl_s16): Likewise.
17853 (vpaddl_s32): Likewise.
17854 (vpaddl_u8): Likewise.
17855 (vpaddl_u16): Likewise.
17856 (vpaddl_u32): Likewise.
17857 (vpaddlq_s8): Likewise.
17858 (vpaddlq_s16): Likewise.
17859 (vpaddlq_s32): Likewise.
17860 (vpaddlq_u8): Likewise.
17861 (vpaddlq_u16): Likewise.
17862 (vpaddlq_u32): Liwewise.
17863 * config/aarch64/iterators.md: Define [SU]ADDLP unspecs with
17864 appropriate attributes.
17866 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17868 * config/aarch64/aarch64-simd-builtins.def: Use VDQ_I iterator
17869 for aarch64_addp<mode> builtin macro generator.
17870 * config/aarch64/aarch64-simd.md: Use VDQ_I iterator in
17871 aarch64_addp<mode> RTL pattern.
17872 * config/aarch64/arm_neon.h (vpaddq_s8): Use RTL builtin
17873 instead of inline asm.
17874 (vpaddq_s16): Likewise.
17875 (vpaddq_s32): Likewise.
17876 (vpaddq_s64): Likewise.
17877 (vpaddq_u8): Likewise.
17878 (vpaddq_u16): Likewise.
17879 (vpaddq_u32): Likewise.
17880 (vpaddq_u64): Likewise.
17882 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17884 * config/aarch64/aarch64-simd-builtins.def: Add sq[r]dmulh_n
17885 builtin generator macros.
17886 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_n<mode>):
17888 * config/aarch64/arm_neon.h (vqdmulh_n_s16): Use RTL builtin
17889 instead of inline asm.
17890 (vqdmulh_n_s32): Likewise.
17891 (vqdmulhq_n_s16): Likewise.
17892 (vqdmulhq_n_s32): Likewise.
17893 (vqrdmulh_n_s16): Likewise.
17894 (vqrdmulh_n_s32): Likewise.
17895 (vqrdmulhq_n_s16): Likewise.
17896 (vqrdmulhq_n_s32): Likewise.
17898 2021-04-28 Tobias Burnus <tobias@codesourcery.com>
17900 * doc/install.texi (--enable-offload-defaulted): Document.
17902 2021-04-28 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
17904 * config/avr/avr-dimode.md: Turn existing patterns into
17905 define_insn_and_split style patterns where the splitter
17906 adds a clobber of the condition code register. Drop "cc"
17907 attribute. Add new patterns to match output of
17909 * config/avr/avr-fixed.md: Likewise.
17910 * config/avr/avr.c (cc_reg_rtx): New.
17911 (avr_parallel_insn_from_insns): Adjust insn count
17912 for removal of set of cc0.
17913 (avr_is_casesi_sequence): Likewise.
17914 (avr_casei_sequence_check_operands): Likewise.
17915 (avr_optimize_casesi): Likewise. Also insert
17916 new insns after jump_insn.
17917 (avr_pass_casesi::avr_rest_of_handle_casesi): Adjust
17918 for removal of set of cc0.
17919 (avr_init_expanders): Initialize cc_reg_rtx.
17920 (avr_regno_reg_class): Handle REG_CC.
17921 (cond_string): Remove usage of CC_OVERFLOW_UNUSABLE.
17922 (avr_notice_update_cc): Remove function.
17923 (ret_cond_branch): Remove usage of CC_OVERFLOW_UNUSABLE.
17924 (compare_condition): Adjust for PARALLEL with
17926 (out_shift_with_cnt): Likewise.
17927 (ashlhi3_out): Likewise.
17928 (ashrhi3_out): Likewise.
17929 (lshrhi3_out): Likewise.
17930 (avr_class_max_nregs): Return single reg for REG_CC.
17931 (avr_compare_pattern): Check for REG_CC instead
17933 (avr_reorg_remove_redundant_compare): Likewise.
17934 (avr_reorg):Adjust for PARALLEL with REG_CC clobber.
17935 (avr_hard_regno_nregs): Return single reg for REG_CC.
17936 (avr_hard_regno_mode_ok): Allow only CCmode for REG_CC.
17937 (avr_md_asm_adjust): Clobber REG_CC.
17938 (TARGET_HARD_REGNO_NREGS): Define.
17939 (TARGET_CLASS_MAX_NREGS): Define.
17940 (TARGET_MD_ASM_ADJUST): Define.
17941 * config/avr/avr.h (FIRST_PSEUDO_REGISTER): Adjust
17943 (enum reg_class): Add CC_REG class.
17944 (NOTICE_UPDATE_CC): Remove.
17945 (CC_OVERFLOW_UNUSABLE): Remove.
17946 (CC_NO_CARRY): Remove.
17947 * config/avr/avr.md: Turn existing patterns into
17948 define_insn_and_split style patterns where the splitter
17949 adds a clobber of the condition code register. Drop "cc"
17950 attribute. Add new patterns to match output of
17952 (sez): Remove unused pattern.
17954 2021-04-28 Richard Earnshaw <rearnsha@arm.com>
17957 * config/arm/arm.c (arm_hard_regno_mode_ok): Only allow VPR to be
17960 2021-04-28 Richard Sandiford <richard.sandiford@arm.com>
17963 * config/aarch64/constraints.md (Utq): Require the address to
17964 be valid for both the element mode and for V2DImode.
17966 2021-04-28 Jakub Jelinek <jakub@redhat.com>
17967 Tobias Burnus <tobias@codesourcery.com>
17969 * configure.ac (OFFLOAD_DEFAULTED): AC_DEFINE if offload-defaulted.
17970 * gcc.c (process_command): New variable.
17971 (driver::maybe_putenv_OFFLOAD_TARGETS): If OFFLOAD_DEFAULTED,
17972 set it if -foffload is defaulted.
17973 * lto-wrapper.c (OFFLOAD_TARGET_DEFAULT_ENV): Define.
17974 (compile_offload_image): If OFFLOAD_DEFAULTED and
17975 OFFLOAD_TARGET_DEFAULT is in the environment, don't fail
17976 if corresponding mkoffload can't be found.
17977 (compile_images_for_offload_targets): Likewise. Free and clear
17978 offload_names if no valid offload is found.
17979 * config.in: Regenerate.
17980 * configure: Regenerate.
17982 2021-04-28 Richard Biener <rguenther@suse.de>
17984 PR tree-optimization/100292
17985 * tree-vect-generic.c (expand_vector_condition): Do not fold
17988 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
17990 * config/rs6000/aix.h (SUBTARGET_DRIVER_SELF_SPECS): New.
17991 * config/rs6000/aix64.opt (m64): New.
17994 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
17996 * config/vax/vax.c (print_operand_address, vax_address_cost_1)
17997 (index_term_p): Handle ASHIFT too.
17999 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
18001 * config/vax/builtins.md (jbb<ccss>i<mode>): Remove operand #3.
18002 (sync_lock_test_and_set<mode>): Adjust accordingly.
18003 (sync_lock_release<mode>): Likewise.
18005 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
18007 * config/vax/vax-protos.h (adjacent_operands_p): Remove
18009 * config/vax/vax.c (adjacent_operands_p): Remove.
18011 2021-04-27 Maciej W. Rozycki <macro@linux-mips.org>
18013 * ifcvt.c (dead_or_predicable) [!IFCVT_MODIFY_TESTS]: Fall
18014 through to the non-conditional execution case if getting the
18015 condition for conditional execution has failed.
18017 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
18019 PR middle-end/100284
18020 * gimple.c (gimple_could_trap_p_1): Remove VEC_COND_EXPR test.
18021 * tree-eh.c (operation_could_trap_p): Handle VEC_COND_EXPR rather
18022 than asserting on it.
18024 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
18026 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): Protect
18027 with TARGET_AIX_OS.
18029 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
18032 * calls.c (precompute_register_parameters): Additionally test
18033 targetm.precompute_tls_p to pre-compute argument.
18034 * config/rs6000/aix.h (TARGET_PRECOMPUTE_TLS_P): Define.
18035 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): New.
18036 * target.def (precompute_tls_p): New.
18037 * doc/tm.texi.in (TARGET_PRECOMPUTE_TLS_P): Add hook documentation.
18038 * doc/tm.texi: Regenerated.
18040 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18043 * config/aarch64/aarch64.c (aarch64_print_operand): Cast -UINTVAL
18044 back to HOST_WIDE_INT.
18046 2021-04-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
18049 * simplify-rtx.c (simplify_context::simplify_subreg): Check the
18050 memory alignment for the outer mode.
18052 2021-04-27 H.J. Lu <hjl.tools@gmail.com>
18054 PR middle-end/90773
18055 * expr.c (op_by_pieces_d::get_usable_mode): New member function.
18056 (op_by_pieces_d::run): Cange a while loop to a do-while loop.
18058 2021-04-27 Alex Coplan <alex.coplan@arm.com>
18061 * config/arm/arm.c (arm_split_compare_and_swap): Fix up codegen
18062 with negative immediates: ensure we expand cbranchsi4_scratch
18063 correctly and ensure we satisfy its constraints.
18064 * config/arm/sync.md
18065 (@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): Don't
18066 attempt to tie two output operands together with constraints;
18067 collapse two alternatives.
18068 (@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
18069 * config/arm/thumb1.md (cbranchsi4_neg_late): New.
18071 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18074 * config/aarch64/predicates.md (aarch64_sub_immediate,
18075 aarch64_plus_immediate): Use -UINTVAL instead of -INTVAL.
18076 * config/aarch64/aarch64.md (casesi, rotl<mode>3): Likewise.
18077 * config/aarch64/aarch64.c (aarch64_print_operand,
18078 aarch64_split_atomic_op, aarch64_expand_subvti): Likewise.
18080 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18082 PR tree-optimization/100239
18083 * tree-vect-generic.c (lower_vec_perm): Don't accept constant
18084 permutations with all indices from the first zero element as vec_shl.
18086 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18088 PR rtl-optimization/100254
18089 * cfgcleanup.c (outgoing_edges_match): Check REG_EH_REGION on
18090 last1 and last2 insns rather than BB_END (bb1) and BB_END (bb2) insns.
18092 2021-04-27 Richard Biener <rguenther@suse.de>
18094 PR tree-optimization/99912
18095 * passes.def: Add comment about new TODO_remove_unused_locals.
18096 * tree-stdarg.c (pass_data_stdarg): Run TODO_remove_unused_locals
18099 2021-04-27 Richard Biener <rguenther@suse.de>
18101 PR tree-optimization/99912
18102 * passes.def (pass_all_optimizations): Add pass_dse before
18103 the first pass_dce, move the first pass_dse before the
18104 pass_dce following pass_pre.
18106 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18108 PR tree-optimization/95527
18109 * generic-match-head.c: Include tm.h.
18110 * gimple-match-head.c: Include tm.h.
18111 * match.pd (CLZ == INTEGER_CST): Don't use
18112 #ifdef CLZ_DEFINED_VALUE_AT_ZERO, only test CLZ_DEFINED_VALUE_AT_ZERO
18113 if clz == CFN_CLZ. Add missing val declaration.
18114 (CTZ cmp CST): New simplifications.
18116 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18118 PR tree-optimization/96696
18119 * expr.c (expand_expr_divmod): New function.
18120 (expand_expr_real_2) <case TRUNC_DIV_EXPR>: Use it for truncations and
18121 divisions. Formatting fixes.
18122 <case MULT_EXPR>: Optimize x / y * y as x - x % y if the latter is
18125 2021-04-27 Martin Jambor <mjambor@suse.cz>
18128 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
18129 If removing a call statement LHS SSA name, release it.
18131 2021-04-27 Richard Earnshaw <rearnsha@arm.com>
18134 * config/arm/arm.c (THUMB2_WORK_REGS): Check PIC_OFFSET_TABLE_REGNUM
18135 is valid before including it in the mask.
18137 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
18140 * config/aarch64/aarch64.c (aarch64_comp_type_attributes): Handle
18143 2021-04-27 Richard Biener <rguenther@suse.de>
18145 PR tree-optimization/100051
18146 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Add
18147 disambiguator based on access size vs. decl size.
18149 2021-04-27 Richard Biener <rguenther@suse.de>
18151 PR tree-optimization/100278
18152 * tree-ssa-pre.c (compute_avail): Give up when we cannot
18153 adjust TBAA beacuse of mismatching bases.
18155 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18158 * config/i386/i386.md (*<insn><mode>3_mask, *<insn><mode>3_mask_1):
18159 For any_rotate define_insn_split and following splitters, use
18160 SWI iterator instead of SWI48.
18162 2021-04-27 Richard Biener <rguenther@suse.de>
18164 PR tree-optimization/99776
18165 * match.pd (bit_field_ref (ctor)): Relax element extract
18166 type compatibility checks.
18168 2021-04-27 Cui,Lili <lili.cui@intel.com>
18170 * common/config/i386/i386-common.c (processor_names):
18171 Sync processor_names with processor_type.
18172 * config/i386/i386-options.c (processor_cost_table):
18173 Sync processor_cost_table with processor_type.
18175 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18177 * value-range.cc (irange::irange_set_1bit_anti_range): Add assert.
18178 (irange::set): Call irange_set_1bit_anti_range for handling all
18179 1-bit ranges. Fall through on ~[MIN,MAX].
18181 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18183 * value-range.cc (irange::legacy_num_pairs): Remove.
18184 (irange::invert): Change gcc_assert to gcc_checking_assert.
18185 * value-range.h (irange::num_pairs): Adjust for a cached
18186 num_pairs(). Also, rename all gcc_assert's to
18187 gcc_checking_assert's.
18189 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18191 * value-range.cc (irange::operator=): Set m_kind.
18192 (irange::copy_to_legacy): Handle varying and undefined sources
18193 as a legacy copy since they can be easily copied.
18194 (irange::irange_set): Set m_kind.
18195 (irange::irange_set_anti_range): Same.
18196 (irange::set): Rename normalize_min_max to normalize_kind.
18197 (irange::verify_range): Adjust for multi-ranges having the
18199 (irange::irange_union): Set m_kind.
18200 (irange::irange_intersect): Same.
18201 (irange::invert): Same.
18202 * value-range.h (irange::kind): Always return m_kind.
18203 (irange::varying_p): Rename to...
18204 (irange::varying_comptaible_p): ...this.
18205 (irange::undefined_p): Only look at m_kind.
18206 (irange::irange): Always set VR_UNDEFINED if applicable.
18207 (irange::set_undefined): Always set VR_UNDEFINED.
18208 (irange::set_varying): Always set m_kind to VR_VARYING.
18209 (irange::normalize_min_max): Rename to...
18210 (irange::normalize_kind): ...this.
18212 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18214 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
18215 Adjust for constant_p including varying_p.
18216 * tree-vrp.c (vrp_prop::finalize): Same.
18217 (determine_value_range): Same.
18218 * vr-values.c (vr_values::range_of_expr): Same.
18219 * value-range.cc (irange::symbolic_p): Do not check varying_p.
18220 (irange::constant_p): Same.
18222 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18224 * value-range.cc (irange::legacy_lower_bound): Replace
18225 !undefined_p check with num_ranges > 0.
18226 (irange::legacy_upper_bound): Same.
18227 * value-range.h (irange::type): Same.
18228 (irange::lower_bound): Same.
18229 (irange::upper_bound): Same.
18231 2021-04-26 Richard Biener <rguenther@suse.de>
18233 PR tree-optimization/99956
18234 * gimple-loop-interchange.cc (compute_access_stride):
18235 Try instantiating the access in a shallower loop nest
18236 if instantiating failed.
18237 (compute_access_strides): Pass adjustable loop_nest
18238 to compute_access_stride.
18240 2021-04-26 Christophe Lyon <christophe.lyon@linaro.org>
18242 * doc/sourcebuild.texi (arm_cmse_hw): Document.
18244 2021-04-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18246 * config/aarch64/iterators.md (vwcore): Handle V4BF, V8BF.
18248 2021-04-26 Thomas Schwinge <thomas@codesourcery.com>
18249 Nathan Sidwell <nathan@codesourcery.com>
18250 Tom de Vries <vries@codesourcery.com>
18251 Julian Brown <julian@codesourcery.com>
18252 Kwok Cheung Yeung <kcy@codesourcery.com>
18254 * omp-offload.c (oacc_validate_dims): Implement
18255 '-Wopenacc-parallelism'.
18256 * doc/invoke.texi (-Wopenacc-parallelism): Document.
18258 2021-04-26 Richard Biener <rguenther@suse.de>
18260 * tree-cfg.h (gimplify_build1): Remove.
18261 (gimplify_build2): Likewise.
18262 (gimplify_build3): Likewise.
18263 * tree-cfg.c (gimplify_build1): Move to tree-vect-generic.c.
18264 (gimplify_build2): Likewise.
18265 (gimplify_build3): Likewise.
18266 * tree-vect-generic.c (gimplify_build1): Move from tree-cfg.c.
18268 (gimplify_build2): Likewise.
18269 (gimplify_build3): Likewise.
18270 (tree_vec_extract): Use resimplify with following SSA edges.
18271 (expand_vector_parallel): Avoid passing NULL size/bitpos
18272 to tree_vec_extract.
18273 * expr.c (store_constructor): Deal with zero-element CTORs.
18274 * match.pd (bit_field_ref <vector CTOR>): Make sure to
18275 produce vector constants when possible.
18277 2021-04-26 Richard Biener <rguenther@suse.de>
18279 * tree-complex.c: Include gimple-fold.h.
18280 (expand_complex_addition): Use gimple_build.
18281 (expand_complex_multiplication_components): Likewise.
18282 (expand_complex_multiplication): Likewise.
18283 (expand_complex_div_straight): Likewise.
18284 (expand_complex_div_wide): Likewise.
18285 (expand_complex_division): Likewise.
18286 (expand_complex_conjugate): Likewise.
18287 (expand_complex_comparison): Likewise.
18289 2021-04-26 Richard Biener <rguenther@suse.de>
18291 * tree-ssa-phiopt.c (two_value_replacement): Remove use
18292 of legacy gimplify_buildN API.
18294 2021-04-26 Richard Biener <rguenther@suse.de>
18296 PR tree-optimization/99473
18297 * tree-ssa-phiopt.c (cond_store_replacement): Handle all
18300 2021-04-26 Richard Biener <rguenther@suse.de>
18302 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin):
18303 Use replace_call_with_value.
18305 2021-04-26 Richard Biener <rguenther@suse.de>
18307 * tree-ssa-propagate.h (valid_gimple_rhs_p): Remove.
18308 (update_gimple_call): Likewise.
18309 (update_call_from_tree): Likewise.
18310 * tree-ssa-propagate.c (valid_gimple_rhs_p): Remove.
18311 (valid_gimple_call_p): Likewise.
18312 (move_ssa_defining_stmt_for_defs): Likewise.
18313 (finish_update_gimple_call): Likewise.
18314 (update_gimple_call): Likewise.
18315 (update_call_from_tree): Likewise.
18316 (propagate_tree_value_into_stmt): Use replace_call_with_value.
18317 * gimple-fold.h (update_gimple_call): Declare.
18318 * gimple-fold.c (valid_gimple_rhs_p): Move here from
18319 tree-ssa-propagate.c.
18320 (update_gimple_call): Likewise.
18321 (valid_gimple_call_p): Likewise.
18322 (finish_update_gimple_call): Likewise, and simplify.
18323 (gimplify_and_update_call_from_tree): Implement
18324 update_call_from_tree functionality, avoid excessive
18325 push/pop_gimplify_context.
18326 (gimple_fold_builtin): Use only gimplify_and_update_call_from_tree.
18327 (gimple_fold_call): Likewise.
18328 * gimple-ssa-sprintf.c (try_substitute_return_value): Likewise.
18329 * tree-ssa-ccp.c (ccp_folder::fold_stmt): Likewise.
18330 (pass_fold_builtins::execute): Likewise.
18331 (optimize_stack_restore): Use replace_call_with_value.
18332 * tree-cfg.c (fold_loop_internal_call): Likewise.
18333 * tree-ssa-dce.c (maybe_optimize_arith_overflow): Use
18334 only gimplify_and_update_call_from_tree.
18335 * tree-ssa-strlen.c (handle_builtin_strlen): Likewise.
18336 (handle_builtin_strchr): Likewise.
18337 * tsan.c: Include gimple-fold.h instead of tree-ssa-propagate.h.
18339 2021-04-26 Jakub Jelinek <jakub@redhat.com>
18342 * vmsdbgout.c (ASM_OUTPUT_DEBUG_STRING, vmsdbgout_begin_block,
18343 vmsdbgout_end_block, lookup_filename, vmsdbgout_source_line): Remove
18346 2021-04-25 liuhongt <hongtao.liu@intel.com>
18349 * config/i386/i386-builtin.def (BDESC): Change the icode of
18350 the following builtins to CODE_FOR_nothing.
18351 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
18352 IX86_BUILTIN_PCMPEQB128, IX86_BUILTIN_PCMPEQW128,
18353 IX86_BUILTIN_PCMPEQD128, IX86_BUILTIN_PCMPEQQ,
18354 IX86_BUILTIN_PCMPEQB256, IX86_BUILTIN_PCMPEQW256,
18355 IX86_BUILTIN_PCMPEQD256, IX86_BUILTIN_PCMPEQQ256,
18356 IX86_BUILTIN_PCMPGTB128, IX86_BUILTIN_PCMPGTW128,
18357 IX86_BUILTIN_PCMPGTD128, IX86_BUILTIN_PCMPGTQ,
18358 IX86_BUILTIN_PCMPGTB256, IX86_BUILTIN_PCMPGTW256,
18359 IX86_BUILTIN_PCMPGTD256, IX86_BUILTIN_PCMPGTQ256.
18360 * config/i386/sse.md (avx2_eq<mode>3): Deleted.
18361 (sse2_eq<mode>3): Ditto.
18362 (sse4_1_eqv2di3): Ditto.
18363 (sse2_gt<mode>3): Rename to ..
18364 (*sse2_gt<mode>3): .. this.
18366 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
18369 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
18372 * config/darwin.c (darwin_binds_local_p): Assume that any
18373 public symbol might be interposed for PIC code. Update function
18374 header comment to reflect current Darwin capability.
18376 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
18379 * config/darwin.c (darwin_binds_local_p): Assume that any
18380 public symbol might be interposed for PIC code. Update function
18381 header comment to reflect current Darwin capability.
18383 2021-04-24 Richard Sandiford <richard.sandiford@arm.com>
18385 * doc/sourcebuild.texi: Document no-opts and any-opts target
18388 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
18390 * config/bpf/bpf.h (ASM_OUTPUT_ALIGNED_BSS): Use .type and .lcomm.
18392 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
18394 * config/bpf/bpf.h (FUNCTION_BOUNDARY): Set to 64.
18396 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
18399 * config/i386/i386-options.c (ix86_option_override_internal):
18400 Error out when -m96bit-long-double is used with 64bit targets.
18401 * config/i386/i386.md (*pushxf_rounded): Remove pattern.
18403 2021-04-23 Martin Liska <mliska@suse.cz>
18405 * lto-wrapper.c: Remove FIXME about usage of
18406 hardware_concurrency. The function is not on par with
18409 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
18412 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
18413 Copy operand 3 to operand 4. Use sse_reg_operand
18414 as operand 3 predicate.
18415 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): Ditto.
18416 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
18417 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
18418 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2):
18419 Copy operand 1 to operand 0.
18420 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): Ditto.
18421 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
18422 (LDX_ATOMIC/LDX_ATOMIC FP store peephole2 with mem blockage): Ditto.
18424 2021-04-23 Alex Coplan <alex.coplan@arm.com>
18426 PR rtl-optimization/100230
18427 * early-remat.c (early_remat::sort_candidates): Use delete[]
18428 instead of delete for array allocated with new[].
18430 2021-04-23 Richard Biener <rguenther@suse.de>
18432 * genmatch.c (lower_cond): Remove VEC_COND_EXPR special-casing.
18433 (capture_info::capture_info): Likewise.
18434 (capture_info::walk_match): Likewise.
18435 (expr::gen_transform): Likewise.
18436 (dt_simplify::gen_1): Likewise.
18437 * gimple-match-head.c (maybe_resimplify_conditional_op):
18438 Remove VEC_COND_EXPR special-casing.
18439 (gimple_simplify): Likewise.
18440 * gimple.c (gimple_could_trap_p_1): Adjust.
18441 * tree-ssa-pre.c (compute_avail): Allow VEC_COND_EXPR
18442 to participate in PRE.
18444 2021-04-23 Richard Biener <rguenther@suse.de>
18446 * cfganal.c (connect_infinite_loops_to_exit): First call
18447 add_noreturn_fake_exit_edges.
18448 * ipa-sra.c (process_scan_results): Do not call the now redundant
18449 add_noreturn_fake_exit_edges.
18450 * predict.c (tree_estimate_probability): Likewise.
18451 (rebuild_frequencies): Likewise.
18452 * store-motion.c (one_store_motion_pass): Likewise.
18454 2021-04-23 Richard Biener <rguenther@suse.de>
18456 PR tree-optimization/100222
18457 * predict.c (pass_profile::execute): Remove redundant call to
18458 mark_irreducible_loops.
18459 (report_predictor_hitrates): Likewise.
18461 2021-04-23 Richard Biener <rguenther@suse.de>
18463 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Avoid
18464 valid_gimple_rhs_p by instead gimplifying to one.
18466 2021-04-23 Richard Biener <rguenther@suse.de>
18468 PR tree-optimization/99971
18469 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
18470 Always use TBAA for loads.
18472 2021-04-23 liuhongt <hongtao.liu@intel.com>
18475 * config/i386/i386-options.c (ix86_option_override_internal):
18476 Clear MASK_AVX256_SPLIT_UNALIGNED_LOAD/STORE in x_target_flags
18477 when X86_TUNE_AVX256_UNALIGNED_LOAD/STORE_OPTIMAL is enabled
18478 by target attribute.
18480 2021-04-23 David Edelsohn <dje.gcc@gmail.com>
18482 * config/rs6000/aix71.h (PREFERRED_DEBUGGING_TYPE): Change to
18484 * config/rs6000/aix72.h (PREFERRED_DEBUGGING_TYPE): Same.
18486 2021-04-22 David Edelsohn <dje.gcc@gmail.com>
18488 * config.gcc (powerpc-ibm-aix6.*): Remove.
18489 * config/rs6000/aix61.h: Delete.
18491 2021-04-22 Martin Liska <mliska@suse.cz>
18493 PR testsuite/100159
18494 PR testsuite/100192
18495 * builtins.c (expand_builtin): Fix typos and missing comments.
18496 * dwarf2out.c (gen_subprogram_die): Likewise.
18497 (gen_struct_or_union_type_die): Likewise.
18499 2021-04-22 Uroš Bizjak <ubizjak@gmail.com>
18502 * config/i386/i386-expand.c (ix86_expand_convert_uns_sidf_sse):
18503 Remove the sign with FE_DOWNWARD, where x - x = -0.0.
18505 2021-04-21 Iain Sandoe <iain@sandoe.co.uk>
18507 * config/i386/darwin.h (TARGET_64BIT): Remove definition
18508 based on TARGET_ISA_64BIT.
18509 (TARGET_64BIT_P): Remove definition based on
18510 TARGET_ISA_64BIT_P().
18512 2021-04-21 Martin Liska <mliska@suse.cz>
18515 2021-04-21 Martin Liska <mliska@suse.cz>
18517 * lto-wrapper.c (cpuset_popcount): Remove.
18518 (init_num_threads): Remove and use hardware_concurrency.
18520 2021-04-21 Martin Liska <mliska@suse.cz>
18523 * main.c (main): Call toplev::finalize in CHECKING_P mode.
18524 * ipa-modref.c (ipa_modref_c_finalize): summaries are NULL
18525 when incremental LTO linking happens.
18527 2021-04-21 Martin Liska <mliska@suse.cz>
18529 * lto-wrapper.c (run_gcc): When -flto=jobserver is used, but the
18530 makeserver cannot be detected, then use -flto=N fallback.
18532 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
18534 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): When cross-compiling,
18535 default to yes for aarch64-linux-gnu.
18536 * configure: Regenerate.
18538 2021-04-21 Martin Liska <mliska@suse.cz>
18540 * lto-wrapper.c (cpuset_popcount): Remove.
18541 (init_num_threads): Remove and use hardware_concurrency.
18543 2021-04-21 Martin Liska <mliska@suse.cz>
18545 * config/i386/i386.c: Remove superfluous || TARGET_MACHO
18546 which remains to be '(... || 0)' and clang complains about it.
18547 * dwarf2out.c (AT_vms_delta): Declare conditionally.
18548 (add_AT_vms_delta): Likewise.
18549 * tree.c (fld_simplified_type): Use rather more common pattern
18550 for disabling of something (#if 0).
18551 (get_tree_code_name): Likewise.
18552 (verify_type_variant): Likewise.
18554 2021-04-21 Martin Liska <mliska@suse.cz>
18556 * config/i386/i386-expand.c (decide_alignment): Use newly named
18557 macro TARGET_CPU_P.
18558 * config/i386/i386.c (ix86_decompose_address): Likewise.
18559 (ix86_address_cost): Likewise.
18560 (ix86_lea_outperforms): Likewise.
18561 (ix86_avoid_lea_for_addr): Likewise.
18562 (ix86_add_stmt_cost): Likewise.
18563 * config/i386/i386.h (TARGET_*): Remove.
18564 (TARGET_CPU_P): New macro.
18565 * config/i386/i386.md: Use newly named macro TARGET_CPU_P.
18566 * config/i386/x86-tune-sched-atom.c (do_reorder_for_imul): Likewise.
18567 (swap_top_of_ready_list): Likewise.
18568 (ix86_atom_sched_reorder): Likewise.
18569 * config/i386/x86-tune-sched-bd.c (ix86_bd_has_dispatch): Likewise.
18570 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Likewise.
18572 2021-04-21 Martin Liska <mliska@suse.cz>
18574 * config/i386/i386-options.c (TARGET_EXPLICIT_NO_SAHF_P):
18576 (SET_TARGET_NO_SAHF): Likewise.
18577 (TARGET_EXPLICIT_PREFETCH_SSE_P): Likewise.
18578 (SET_TARGET_PREFETCH_SSE): Likewise.
18579 (TARGET_EXPLICIT_NO_TUNE_P): Likewise.
18580 (SET_TARGET_NO_TUNE): Likewise.
18581 (TARGET_EXPLICIT_NO_80387_P): Likewise.
18582 (SET_TARGET_NO_80387): Likewise.
18584 * config/i386/i386.h (TARGET_*): Remove.
18585 * opth-gen.awk: Generate new used macros.
18587 2021-04-21 Martin Liska <mliska@suse.cz>
18589 * config/i386/i386.h (PTA_*): Remove.
18590 (enum pta_flag): New.
18591 (DEF_PTA): Generate PTA_* values from i386-isa.def.
18592 * config/i386/i386-isa.def: New file.
18594 2021-04-21 Alex Coplan <alex.coplan@arm.com>
18597 * config/aarch64/aarch64-bti-insert.c (aarch64_bti_j_insn_p): New.
18598 (rest_of_insert_bti): Avoid inserting duplicate bti j insns for
18599 jump table targets.
18601 2021-04-21 H.J. Lu <hjl.tools@gmail.com>
18603 * config.gcc: Install mwaitintrin.h for i[34567]86-*-* and
18604 x86_64-*-* targets.
18605 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_MWAIT_SET):
18607 (OPTION_MASK_ISA2_MWAIT_UNSET): Likewise.
18608 (ix86_handle_option): Handle -mmwait.
18609 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
18610 Replace OPTION_MASK_ISA_SSE3 with OPTION_MASK_ISA2_MWAIT on
18611 __builtin_ia32_monitor and __builtin_ia32_mwait.
18612 * config/i386/i386-options.c (isa2_opts): Add -mmwait.
18613 (ix86_valid_target_attribute_inner_p): Likewise.
18614 (ix86_option_override_internal): Enable mwait/monitor
18615 instructions for -msse3.
18616 * config/i386/i386.h (TARGET_MWAIT): New.
18617 (TARGET_MWAIT_P): Likewise.
18618 * config/i386/i386.opt: Add -mmwait.
18619 * config/i386/mwaitintrin.h: New file.
18620 * config/i386/pmmintrin.h: Include <mwaitintrin.h>.
18621 * config/i386/sse.md (sse3_mwait): Replace TARGET_SSE3 with
18623 (@sse3_monitor_<mode>): Likewise.
18624 * config/i386/x86gprintrin.h: Include <mwaitintrin.h>.
18625 * doc/extend.texi: Document mwait target attribute.
18626 * doc/invoke.texi: Document -mmwait.
18628 2021-04-21 Martin Liska <mliska@suse.cz>
18630 * config/i386/i386-options.c (DEF_ENUM): Remove it.
18631 * config/i386/i386-opts.h (DEF_ENUM): Likewise.
18632 * config/i386/stringop.def (DEF_ENUM): Likewise.
18634 2021-04-21 Martin Liska <mliska@suse.cz>
18636 * tree-cfg.c (gimple_verify_flow_info): Use qD instead
18637 of print_generic_expr.
18639 2021-04-21 Jakub Jelinek <jakub@redhat.com>
18641 PR rtl-optimization/100148
18642 * cprop.c (constprop_register): Use next_nondebug_insn instead of
18645 2021-04-21 Martin Liska <mliska@suse.cz>
18648 * cgraphunit.c (cgraph_node::analyze): Remove duplicate
18649 free_dominance_info calls.
18651 2021-04-21 Richard Biener <rguenther@suse.de>
18653 * gimple-fold.c (maybe_fold_reference): Remove is_lhs
18654 parameter (and assume it to be false).
18655 (fold_gimple_assign): Adjust, remove all callers of
18656 maybe_fold_reference calling it with is_lhs true.
18657 (gimple_fold_call): Likewise.
18658 (fold_stmt_1): Likewise.
18660 2021-04-21 Richard Biener <rguenther@suse.de>
18662 * fold-const.c (pedantic_non_lvalue_loc): Remove.
18663 (fold_binary_loc): Adjust.
18664 (fold_ternary_loc): Likewise.
18666 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
18668 PR middle-end/100130
18669 * varasm.c (get_block_for_decl): Make sure that any use of the
18670 retain attribute matches the section's retain flag.
18671 (switch_to_section): Check for retain mismatches even when
18672 changing sections, but do not warn if the given decl is the
18673 section's named.decl.
18674 (output_object_block): Pass the first decl in the block (if any)
18675 to switch_to_section.
18677 2021-04-20 H.J. Lu <hjl.tools@gmail.com>
18679 * config/i386/i386-c.c (ix86_target_macros_internal): Define
18680 __CRC32__ for -mcrc32.
18681 * config/i386/i386-options.c (ix86_option_override_internal):
18682 Enable crc32 instruction for -msse4.2.
18683 * config/i386/i386.md (sse4_2_crc32<mode>): Remove TARGET_SSE4_2
18685 (sse4_2_crc32di): Likewise.
18686 * config/i386/ia32intrin.h: Use crc32 target option for CRC32
18689 2021-04-20 Segher Boessenkool <segher@kernel.crashing.org>
18692 * config/rs6000/rs6000.c (rs6000_machine_from_flags): Do not consider
18695 2021-04-20 Martin Liska <mliska@suse.cz>
18697 * doc/invoke.texi: Fix typo.
18698 * params.opt: Likewise.
18700 2021-04-20 Martin Liska <mliska@suse.cz>
18702 * doc/invoke.texi: Document new param.
18704 2021-04-19 Andrew MacLeod <amacleod@redhat.com>
18706 PR tree-optimization/100081
18707 * gimple-range-cache.h (ranger_cache): Inherit from gori_compute
18708 rather than gori_compute_cache.
18709 * gimple-range-gori.cc (is_gimple_logical_p): Move to top of file.
18710 (range_def_chain::m_logical_depth): New member.
18711 (range_def_chain::range_def_chain): Initialize m_logical_depth.
18712 (range_def_chain::get_def_chain): Don't build defchains through more
18713 than LOGICAL_LIMIT logical expressions.
18714 * params.opt (param_ranger_logical_depth): New.
18716 2021-04-19 Richard Earnshaw <rearnsha@arm.com>
18719 * config/arm/arm.c (arm_configure_build_target): Do not strip
18720 extended FPU/SIMD feature bits from the target ISA when -mfpu
18721 is specified (partial revert of r11-8168).
18723 2021-04-19 Thomas Schwinge <thomas@codesourcery.com>
18725 * params.opt (-param=openacc-kernels=): Add.
18726 * omp-oacc-kernels-decompose.cc
18727 (pass_omp_oacc_kernels_decompose::gate): Use it.
18728 * doc/invoke.texi (-fopenacc-kernels=@var{mode}): Move...
18729 (--param): ... here, 'openacc-kernels'.
18731 2021-04-19 Martin Liska <mliska@suse.cz>
18734 * gengtype.c (finish_root_table): Align function arguments
18735 in between declaration and definition.
18737 2021-04-19 Eric Botcazou <ebotcazou@adacore.com>
18739 * config/i386/winnt.c (i386_pe_seh_cold_init): Properly deal with
18740 frames larger than the SEH maximum frame size.
18742 2021-04-18 Segher Boessenkool <segher@kernel.crashing.org>
18744 PR rtl-optimization/99927
18745 * combine.c (distribute_notes) [REG_UNUSED]: If the register already
18746 is dead, just drop it.
18748 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
18751 * config/i386/winnt-d.c (TARGET_D_TEMPLATES_ALWAYS_COMDAT): Define.
18752 * doc/tm.texi: Regenerate.
18753 * doc/tm.texi.in (D language and ABI): Add @hook for
18754 TARGET_D_TEMPLATES_ALWAYS_COMDAT.
18756 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
18758 * config/darwin-d.c (darwin_d_handle_target_object_format): New
18760 (darwin_d_register_target_info): New function.
18761 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18762 * config/dragonfly-d.c (dragonfly_d_handle_target_object_format): New
18764 (dragonfly_d_register_target_info): New function.
18765 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18766 * config/freebsd-d.c (freebsd_d_handle_target_object_format): New
18768 (freebsd_d_register_target_info): New function.
18769 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18770 * config/glibc-d.c (glibc_d_handle_target_object_format): New
18772 (glibc_d_register_target_info): New function.
18773 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18774 * config/i386/i386-d.c (ix86_d_handle_target_object_format): New
18776 (ix86_d_register_target_info): Add ix86_d_handle_target_object_format
18777 as handler for objectFormat key.
18778 * config/i386/winnt-d.c (winnt_d_handle_target_object_format): New
18780 (winnt_d_register_target_info): New function.
18781 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18782 * config/netbsd-d.c (netbsd_d_handle_target_object_format): New
18784 (netbsd_d_register_target_info): New function.
18785 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18786 * config/openbsd-d.c (openbsd_d_handle_target_object_format): New
18788 (openbsd_d_register_target_info): New function.
18789 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18790 * config/pa/pa-d.c (pa_d_handle_target_object_format): New function.
18791 (pa_d_register_target_info): Add pa_d_handle_target_object_format as
18792 handler for objectFormat key.
18793 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_object_format): New
18795 (rs6000_d_register_target_info): Add
18796 rs6000_d_handle_target_object_format as handler for objectFormat key.
18797 * config/sol2-d.c (solaris_d_handle_target_object_format): New
18799 (solaris_d_register_target_info): New function.
18800 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18802 2021-04-16 Jakub Jelinek <jakub@redhat.com>
18805 * config/aarch64/aarch64.c (aarch64_function_arg_alignment): Change
18806 abi_break argument from bool * to unsigned *, store there the pre-GCC 9
18808 (aarch64_layout_arg, aarch64_gimplify_va_arg_expr): Adjust callers.
18809 (aarch64_function_arg_regno_p): Likewise. Only emit -Wpsabi note if
18810 the old and new alignment after applying MIN/MAX to it is different.
18812 2021-04-16 Tamar Christina <tamar.christina@arm.com>
18815 * config/aarch64/aarch64-sve.md (@aarch64_sve_trn1_conv<mode>): New.
18816 * config/aarch64/aarch64.c (aarch64_expand_sve_const_pred_trn): Use new
18818 * config/aarch64/iterators.md (UNSPEC_TRN1_CONV): New.
18820 2021-04-16 Bill Schmidt <wschmidt@linux.ibm.com>
18822 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Revise
18823 this section and its subsections.
18825 2021-04-16 Jakub Jelinek <jakub@redhat.com>
18828 * config/aarch64/aarch64.md (*neg_asr_si2_extr, *extrsi5_insn_di): New
18829 define_insn patterns.
18831 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
18833 PR rtl-optimization/98689
18834 * reg-notes.def (UNTYPED_CALL): New note.
18835 * combine.c (distribute_notes): Handle it.
18836 * emit-rtl.c (try_split): Likewise.
18837 * rtlanal.c (rtx_properties::try_to_add_insn): Likewise. Assume
18838 that calls with the note implicitly set all return value registers.
18839 * builtins.c (expand_builtin_apply): Add a REG_UNTYPED_CALL
18842 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
18844 PR rtl-optimization/99596
18845 * rtlanal.c (rtx_properties::try_to_add_insn): Don't add global
18846 register accesses for const calls. Assume that pure functions
18847 can only read from global registers. Ignore cases in which
18848 the stack pointer has been marked global.
18850 2021-04-16 Jakub Jelinek <jakub@redhat.com>
18853 * tree-vect-loop.c (vect_transform_loop): Don't remove just
18854 dead scalar .MASK_LOAD calls, but also dead .COND_* calls - replace
18855 them by their last argument.
18857 2021-04-15 Martin Liska <mliska@suse.cz>
18859 * doc/invoke.texi: Other params don't use it, remove it.
18861 2021-04-15 Richard Biener <rguenther@suse.de>
18863 * gimple-builder.h: Add deprecation note.
18865 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
18868 * attribs.h (restrict_type_identity_attributes_to): Declare.
18869 * attribs.c (restrict_type_identity_attributes_to): New function.
18871 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
18874 * attribs.h (affects_type_identity_attributes): Declare.
18875 * attribs.c (remove_attributes_matching): New function.
18876 (affects_type_identity_attributes): Likewise.
18878 2021-04-15 Jakub Jelinek <jakub@redhat.com>
18881 * config/aarch64/aarch64.md (*<LOGICAL:optab>_<SHIFT:optab><mode>3):
18882 Add combine splitters for *<LOGICAL:optab>_ashl<mode>3 with
18883 ZERO_EXTEND, SIGN_EXTEND or AND.
18885 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
18887 PR rtl-optimization/99929
18888 * rtl.h (same_vector_encodings_p): New function.
18889 * cse.c (exp_equiv_p): Check that CONST_VECTORs have the same encoding.
18890 * cselib.c (rtx_equal_for_cselib_1): Likewise.
18891 * jump.c (rtx_renumbered_equal_p): Likewise.
18892 * lra-constraints.c (operands_match_p): Likewise.
18893 * reload.c (operands_match_p): Likewise.
18894 * rtl.c (rtx_equal_p_cb, rtx_equal_p): Likewise.
18896 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
18898 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Print
18899 more information about variable-length CONST_VECTORs.
18901 2021-04-14 Vladimir N. Makarov <vmakarov@redhat.com>
18903 PR rtl-optimization/100066
18904 * lra-constraints.c (split_reg): Check paradoxical_subreg_p for
18905 ordered modes when choosing splitting mode for hard reg.
18907 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
18910 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector_sel):
18912 (aarch64_expand_sve_const_vector): Use it for nelts_per_pattern==2.
18914 2021-04-14 Andreas Krebbel <krebbel@linux.ibm.com>
18916 * config/s390/s390-builtins.def (O_M5, O_M12, ...): Add new macros
18917 for mask operand types.
18918 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
18919 (s390_vec_permi_dbl, s390_vpdi): Use the M5 type for the immediate
18921 (s390_vec_msum_u128, s390_vmslg): Use the M12 type for the
18923 * config/s390/s390.c (s390_const_operand_ok): Check the new
18924 operand types and generate a list of valid values.
18926 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
18928 * doc/tm.texi: Regenerate.
18929 * doc/tm.texi.in (D language and ABI): Add @hook for
18930 TARGET_D_REGISTER_OS_TARGET_INFO.
18932 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
18934 * config/aarch64/aarch64-d.c (aarch64_d_handle_target_float_abi): New
18936 (aarch64_d_register_target_info): New function.
18937 * config/aarch64/aarch64-protos.h (aarch64_d_register_target_info):
18939 * config/aarch64/aarch64.h (TARGET_D_REGISTER_CPU_TARGET_INFO):
18941 * config/arm/arm-d.c (arm_d_handle_target_float_abi): New function.
18942 (arm_d_register_target_info): New function.
18943 * config/arm/arm-protos.h (arm_d_register_target_info): Declare.
18944 * config/arm/arm.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18945 * config/i386/i386-d.c (ix86_d_handle_target_float_abi): New function.
18946 (ix86_d_register_target_info): New function.
18947 * config/i386/i386-protos.h (ix86_d_register_target_info): Declare.
18948 * config/i386/i386.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18949 * config/mips/mips-d.c (mips_d_handle_target_float_abi): New function.
18950 (mips_d_register_target_info): New function.
18951 * config/mips/mips-protos.h (mips_d_register_target_info): Declare.
18952 * config/mips/mips.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18953 * config/pa/pa-d.c (pa_d_handle_target_float_abi): New function.
18954 (pa_d_register_target_info): New function.
18955 * config/pa/pa-protos.h (pa_d_register_target_info): Declare.
18956 * config/pa/pa.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18957 * config/riscv/riscv-d.c (riscv_d_handle_target_float_abi): New
18959 (riscv_d_register_target_info): New function.
18960 * config/riscv/riscv-protos.h (riscv_d_register_target_info): Declare.
18961 * config/riscv/riscv.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18962 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_float_abi): New
18964 (rs6000_d_register_target_info): New function.
18965 * config/rs6000/rs6000-protos.h (rs6000_d_register_target_info):
18967 * config/rs6000/rs6000.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18968 * config/s390/s390-d.c (s390_d_handle_target_float_abi): New function.
18969 (s390_d_register_target_info): New function.
18970 * config/s390/s390-protos.h (s390_d_register_target_info): Declare.
18971 * config/s390/s390.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18972 * config/sparc/sparc-d.c (sparc_d_handle_target_float_abi): New
18974 (sparc_d_register_target_info): New function.
18975 * config/sparc/sparc-protos.h (sparc_d_register_target_info): Declare.
18976 * config/sparc/sparc.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
18977 * doc/tm.texi: Regenerate.
18978 * doc/tm.texi.in (D language and ABI): Add @hook for
18979 TARGET_D_REGISTER_CPU_TARGET_INFO.
18981 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
18983 * config/i386/i386-d.c (ix86_d_has_stdcall_convention): New function.
18984 * config/i386/i386-protos.h (ix86_d_has_stdcall_convention): Declare.
18985 * config/i386/i386.h (TARGET_D_HAS_STDCALL_CONVENTION): Define.
18986 * doc/tm.texi: Regenerate.
18987 * doc/tm.texi.in (D language and ABI): Add @hook for
18988 TARGET_D_HAS_STDCALL_CONVENTION.
18990 2021-04-14 Richard Biener <rguenther@suse.de>
18992 * tree-cfg.c (verify_gimple_assign_ternary): Verify that
18993 VEC_COND_EXPRs have a gimple_val condition.
18994 * tree-ssa-propagate.c (valid_gimple_rhs_p): VEC_COND_EXPR
18995 can no longer have a GENERIC condition.
18997 2021-04-14 Richard Earnshaw <rearnsha@arm.com>
19000 * config/arm/arm.c (arm_configure_build_target): Strip isa_all_fpbits
19001 from the isa_delta when -mfpu has been used.
19002 (arm_options_perform_arch_sanity_checks): It's the architecture that
19003 lacks an FPU not the processor.
19005 2021-04-13 Richard Biener <rguenther@suse.de>
19007 PR tree-optimization/100053
19008 * tree-ssa-sccvn.c (vn_nary_op_get_predicated_value): Do
19009 not use optimistic dominance queries for backedges to validate
19011 (dominated_by_p_w_unex): Add parameter to ignore executable
19012 state on backedges.
19013 (rpo_elim::eliminate_avail): Adjust.
19015 2021-04-13 Jakub Jelinek <jakub@redhat.com>
19018 * config/aarch64/aarch64.md (*aarch64_bfxil<mode>_extr,
19019 *aarch64_bfxilsi_extrdi): New define_insn patterns.
19021 2021-04-13 Jakub Jelinek <jakub@redhat.com>
19024 * simplify-rtx.c (simplify_immed_subreg): For MODE_COMPOSITE_P
19025 outermode, return NULL if the result doesn't encode back to the
19026 original byte sequence.
19027 (simplify_gen_subreg): Don't create SUBREGs from constants to
19028 MODE_COMPOSITE_P outermode.
19030 2021-04-12 Jakub Jelinek <jakub@redhat.com>
19032 PR rtl-optimization/99905
19033 * combine.c (expand_compound_operation): If pos + len > modewidth,
19034 perform the right shift by pos in inner_mode and then convert to mode,
19035 instead of trying to simplify a shift of rtx with inner_mode by pos
19036 as if it was a shift in mode.
19038 2021-04-12 Jakub Jelinek <jakub@redhat.com>
19041 * combine.c (simplify_and_const_int_1): Don't optimize varop
19042 away if it has side-effects.
19044 2021-04-12 Martin Liska <mliska@suse.cz>
19046 * doc/extend.texi: Escape @smallexample content.
19048 2021-04-12 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
19050 * config/s390/s390.md ("*movdi_31", "*movdi_64"): Add
19051 alternative in order to load a DFP zero.
19053 2021-04-12 Martin Liska <mliska@suse.cz>
19055 * doc/extend.texi: Be more precise in documentation
19056 of symver attribute.
19058 2021-04-12 Martin Liska <mliska@suse.cz>
19061 * gimplify.c (gimplify_expr): Right now, we unpoison all
19062 variables before a goto <dest>. We should not do it if we are
19065 2021-04-12 Cui,Lili <lili.cui@intel.com>
19067 * common/config/i386/cpuinfo.h (get_intel_cpu): Handle
19069 * common/config/i386/i386-common.c (processor_names): Add
19071 (processor_alias_table): Add rocketlake.
19072 * common/config/i386/i386-cpuinfo.h (processor_subtypes): Add
19073 INTEL_COREI7_ROCKETLAKE.
19074 * config.gcc: Add -march=rocketlake.
19075 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19077 * config/i386/i386-options.c (m_ROCKETLAKE) : Define.
19078 (processor_cost_table): Add rocketlake cost.
19079 * config/i386/i386.h (ix86_size_cost) : Define
19081 (processor_type) : Add PROCESSOR_ROCKETLAKE.
19082 (PTA_ROCKETLAKE): Ditto.
19083 * doc/extend.texi: Add rocketlake.
19084 * doc/invoke.texi: Add rocketlake.
19086 2021-04-12 Cui,Lili <lili.cui@intel.com>
19088 * config/i386/i386.h (PTA_ALDERLAKE): Change alderlake ISA list.
19089 * config/i386/i386-options.c (m_CORE_AVX2): Add m_ALDERLAKE.
19090 * common/config/i386/cpuinfo.h (get_intel_cpu): Add AlderLake model.
19091 * doc/invoke.texi: Change alderlake ISA list.
19093 2021-04-11 Hafiz Abid Qadeer <abidh@codesourcery.com>
19095 PR middle-end/98088
19096 * omp-expand.c (expand_oacc_collapse_init): Update condition in
19099 2021-04-10 H.J. Lu <hjl.tools@gmail.com>
19102 * config/i386/serializeintrin.h (_serialize): Defined as macro.
19104 2021-04-10 Jakub Jelinek <jakub@redhat.com>
19107 * expr.c (expand_expr_addr_expr_1): Test is_global_var rather than
19108 just TREE_STATIC on COMPOUND_LITERAL_EXPR_DECLs.
19110 2021-04-10 Jakub Jelinek <jakub@redhat.com>
19112 PR middle-end/99989
19113 * gimple-ssa-warn-alloca.c
19114 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit to
19115 0 with integer precision unconditionally.
19117 2021-04-10 Jakub Jelinek <jakub@redhat.com>
19119 PR rtl-optimization/98601
19120 * rtlanal.c (rtx_addr_can_trap_p_1): Allow in assert unknown size
19121 not just for BLKmode, but also for VOIDmode. For STRICT_ALIGNMENT
19122 unaligned_mems handle VOIDmode like BLKmode.
19124 2021-04-10 Jan Hubicka <hubicka@ucw.cz>
19127 * tree.c (free_lang_data_in_decl): Do not release body of
19128 declare_variant_alt.
19130 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
19132 * config/aarch64/aarch64.c (aarch64_option_restore): If the
19133 architecture was specified explicitly and the tuning wasn't,
19134 tune for the architecture rather than the configured default CPU.
19136 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
19138 * config/aarch64/aarch64.md (tlsdesc_small_sve_<mode>): Use X30
19139 as the temporary register.
19141 2021-04-09 Martin Liska <mliska@suse.cz>
19143 * doc/extend.texi: Move non-target attributes on the top level.
19145 2021-04-09 Martin Liska <mliska@suse.cz>
19147 * doc/invoke.texi: Document minimum and maximum value of the
19148 argument for both supported compression algorithms.
19150 2021-04-08 David Edelsohn <dje.gcc@gmail.com>
19152 * config/rs6000/rs6000.c (rs6000_xcoff_select_section): Select
19153 TLS BSS before TLS data.
19154 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .comm.
19156 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
19158 * doc/sourcebuild.texi (stdint_types_mbig_endian): Document.
19160 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
19162 * match.pd: Extend vec_cond folds to handle shifts.
19164 2021-04-08 Maciej W. Rozycki <macro@orcam.me.uk>
19166 * config/vax/vax.md: Fix comment for `*bit<mode>' pattern's
19169 2021-04-08 Alex Coplan <alex.coplan@arm.com>
19172 * config/arm/iterators.md (MVE_vecs): New.
19173 (V_elem): Also handle V2DF.
19174 * config/arm/mve.md (*mve_mov<mode>): Rename to ...
19175 (*mve_vdup<mode>): ... this. Remove second alternative since
19176 vec_duplicate of const_int is not canonical RTL, and we don't
19177 want to match symbol_refs.
19178 (*mve_vec_duplicate<mode>): Delete (pattern is redundant).
19180 2021-04-08 Xionghu Luo <luoxhu@linux.ibm.com>
19182 * fold-const.c (fold_single_bit_test): Fix typo.
19183 * print-rtl.c (print_rtx_insn_vec): Call print_rtl_single
19186 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
19188 PR tree-optimization/97513
19189 * tree-vect-slp.c (vect_add_slp_permutation): New function,
19191 (vectorizable_slp_permutation): ...here. Detect cases in which
19192 all VEC_PERM_EXPRs are guaranteed to have the same stepped
19193 permute vector and only generate one permute vector for that case.
19194 Extend that case to handle variable-length vectors.
19196 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
19198 PR tree-optimization/99873
19199 * tree-vect-slp.c (vect_slp_prefer_store_lanes_p): New function.
19200 (vect_build_slp_instance): Don't split store groups that could
19201 use IFN_STORE_LANES.
19203 2021-04-07 Jakub Jelinek <jakub@redhat.com>
19206 * varasm.c (output_constant_pool_contents): Don't strip name encoding
19207 from XSTR (desc->sym, 0) or from label before passing those to
19210 2021-04-07 Richard Biener <rguenther@suse.de>
19212 PR tree-optimization/99954
19213 * tree-loop-distribution.c: Include tree-affine.h.
19214 (generate_memcpy_builtin): Try using tree-affine to prove
19216 (loop_distribution::classify_builtin_ldst): Always classify
19219 2021-04-07 Richard Biener <rguenther@suse.de>
19221 PR tree-optimization/99947
19222 * tree-vect-loop.c (vectorizable_induction): Pre-allocate
19223 steps vector to avoid pushing elements from the reallocated
19226 2021-04-07 Richard Biener <rguenther@suse.de>
19228 * tree-ssa-sccvn.h (print_vn_reference_ops): Declare.
19229 * tree-ssa-pre.c (print_pre_expr): Factor out VN reference operand
19231 * tree-ssa-sccvn.c (print_vn_reference_ops): ... into this new
19233 (debug_vn_reference_ops): New.
19235 2021-04-07 Bin Cheng <bin.cheng@linux.alibaba.com>
19237 PR tree-optimization/98736
19238 * tree-loop-distribution.c
19239 * (loop_distribution::bb_top_order_init):
19240 Compute RPO with programing order preserved by calling function
19241 rev_post_order_and_mark_dfs_back_seme.
19243 2021-04-06 Vladimir N. Makarov <vmakarov@redhat.com>
19246 * lra-constraints.c (split_reg): Don't check paradoxical_subreg_p.
19247 * lra-lives.c (clear_sparseset_regnos, regnos_in_sparseset_p): New
19249 (process_bb_lives): Don't update biggest mode of hard reg for
19250 implicit in multi-register group. Use the new functions for
19251 updating dead_set and unused_set by register notes.
19253 2021-04-06 Xianmiao Qu <xianmiao_qu@c-sky.com>
19255 * config/csky/csky_pipeline_ck802.md : Use insn reservation name
19258 2021-04-06 H.J. Lu <hjl.tools@gmail.com>
19260 * config/i386/x86-tune-costs.h (skylake_memcpy): Updated.
19261 (skylake_memset): Likewise.
19262 (skylake_cost): Change CLEAR_RATIO to 17.
19263 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
19264 Replace m_CANNONLAKE, m_ICELAKE_CLIENT, m_ICELAKE_SERVER,
19265 m_TIGERLAKE and m_SAPPHIRERAPIDS with m_SKYLAKE and m_CORE_AVX512.
19267 2021-04-06 Richard Biener <rguenther@suse.de>
19269 PR tree-optimization/99880
19270 * tree-vect-loop.c (maybe_set_vectorized_backedge_value): Only
19271 set vectorized defs of relevant PHIs.
19273 2021-04-06 Richard Biener <rguenther@suse.de>
19275 PR tree-optimization/99924
19276 * tree-vect-slp.c (vect_bb_partition_graph_r): Do not mark
19277 nodes w/o scalar stmts as visited.
19279 2021-04-06 Alex Coplan <alex.coplan@arm.com>
19282 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Also use base
19283 PCS for [su]fix_optab.
19285 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
19287 * config/darwin.c (machopic_legitimize_pic_address): Check
19288 that the current pic register is one of the hard reg set
19289 before setting liveness.
19291 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
19293 * config/darwin.c (machopic_legitimize_pic_address): Fix
19294 whitespace, remove unused code.
19296 2021-04-03 Jakub Jelinek <jakub@redhat.com>
19298 PR tree-optimization/99882
19299 * gimple-ssa-store-merging.c (bswap_view_convert): Handle val with
19302 2021-04-03 Jakub Jelinek <jakub@redhat.com>
19304 PR rtl-optimization/99863
19305 * dse.c (replace_read): Drop regs_live argument. Instead of
19306 regs_live, use store_insn->fixed_regs_live if non-NULL,
19307 otherwise punt if insns sequence clobbers or sets any hard
19310 2021-04-03 Jakub Jelinek <jakub@redhat.com>
19313 * targhooks.h (default_print_patchable_function_entry_1): Declare.
19314 * targhooks.c (default_print_patchable_function_entry_1): New function,
19315 copied from default_print_patchable_function_entry with an added flags
19317 (default_print_patchable_function_entry): Rewritten into a small
19318 wrapper around default_print_patchable_function_entry_1.
19319 * config/rs6000/rs6000.c (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
19321 (rs6000_print_patchable_function_entry): New function.
19323 2021-04-02 Eric Botcazou <ebotcazou@adacore.com>
19325 * doc/invoke.texi (fdelete-dead-exceptions): Minor tweak.
19327 2021-04-01 Jason Merrill <jason@redhat.com>
19330 * common.opt: Document v15 and v16.
19332 2021-04-01 Richard Biener <rguenther@suse.de>
19334 PR tree-optimization/99863
19335 * gimplify.c (gimplify_init_constructor): Recompute vector
19338 2021-04-01 Jakub Jelinek <jakub@redhat.com>
19340 * doc/extend.texi (symver attribute): Fix up syntax errors
19343 2021-04-01 Jakub Jelinek <jakub@redhat.com>
19345 PR tree-optimization/96573
19346 * gimple-ssa-store-merging.c (init_symbolic_number): Handle
19347 also pointer types.
19349 2021-04-01 Richard Biener <rguenther@suse.de>
19351 PR tree-optimization/99856
19352 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Promote
19353 precision to vector element precision.
19355 2021-04-01 Martin Jambor <mjambor@suse.cz>
19357 PR tree-optimization/97009
19358 * tree-sra.c (access_or_its_child_written): New function.
19359 (propagate_subaccesses_from_rhs): Use it instead of a simple grp_write
19362 2021-03-31 Jan Hubicka <hubicka@ucw.cz>
19365 * cif-code.def (USES_COMDAT_LOCAL): Make CIF_FINAL_NORMAL.
19367 2021-03-31 Pat Haugen <pthaugen@linux.ibm.com>
19370 * config/rs6000/altivec.md (xxspltiw_v4si, xxspltiw_v4sf_inst,
19371 xxspltidp_v2df_inst, xxsplti32dx_v4si_inst, xxsplti32dx_v4sf_inst,
19372 xxblend_<mode>, xxpermx_inst, xxeval): Mark prefixed.
19373 * config/rs6000/mma.md (mma_<vvi4i4i8>, mma_<avvi4i4i8>,
19374 mma_<vvi4i4i2>, mma_<avvi4i4i2>, mma_<vvi4i4>, mma_<avvi4i4>,
19375 mma_<pvi4i2>, mma_<apvi4i2>, mma_<vvi4i4i4>, mma_<avvi4i4i4>):
19377 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Adjust test.
19378 * config/rs6000/rs6000.md (define_attr "maybe_prefixed"): New.
19379 (define_attr "prefixed"): Update initializer.
19381 2021-03-31 Jakub Jelinek <jakub@redhat.com>
19384 * dwarf2out.c (debug_ranges_dwo_section): New variable.
19385 (DW_RANGES_IDX_SKELETON): Define.
19386 (struct dw_ranges): Add begin_entry and end_entry members.
19387 (DEBUG_DWO_RNGLISTS_SECTION): Define.
19388 (add_ranges_num): Adjust r initializer for addition of *_entry
19390 (add_ranges_by_labels): For -gsplit-dwarf and force_direct,
19391 set idx to DW_RANGES_IDX_SKELETON.
19392 (use_distinct_base_address_for_range): New function.
19393 (index_rnglists): Don't set r->idx if it is equal to
19394 DW_RANGES_IDX_SKELETON. Initialize r->begin_entry and
19395 r->end_entry for -gsplit-dwarf if those will be needed by
19397 (output_rnglists): Add DWO argument. If true, switch to
19398 debug_ranges_dwo_section rather than debug_ranges_section.
19399 Adjust l1/l2 label indexes. Only output the offset table when
19400 dwo is true and don't include in there the skeleton range
19401 entry if present. For -gsplit-dwarf, skip ranges that belong
19402 to the other rnglists section. Change return type from void
19403 to bool and return true if there are any range entries for
19404 the other section. For dwarf_split_debug_info use
19405 DW_RLE_startx_endx, DW_RLE_startx_length and DW_RLE_base_addressx
19406 entries instead of DW_RLE_start_end, DW_RLE_start_length and
19407 DW_RLE_base_address. Use use_distinct_base_address_for_range.
19408 (init_sections_and_labels): Initialize debug_ranges_dwo_section
19409 if -gsplit-dwarf and DWARF >= 5. Adjust ranges_section_label
19410 and range_base_label indexes.
19411 (dwarf2out_finish): Call index_rnglists earlier before finalizing
19412 .debug_addr. Never emit DW_AT_rnglists_base attribute. For
19413 -gsplit-dwarf and DWARF >= 5 call output_rnglists up to twice
19414 with different dwo arguments.
19415 (dwarf2out_c_finalize): Clear debug_ranges_dwo_section.
19417 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19419 PR tree-optimization/98268
19420 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Call
19421 recompute_tree_invariant_for_addr_expr after successfully
19422 folding a TARGET_MEM_REF that occurs inside an ADDR_EXPR.
19424 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19426 PR tree-optimization/99726
19427 * tree-data-ref.c (create_intersect_range_checks_index): Bail
19428 out if there is more than one access function SCEV for the loop
19431 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19433 PR rtl-optimization/97141
19434 PR rtl-optimization/98726
19435 * emit-rtl.c (valid_for_const_vector_p): Return true for
19437 * rtx-vector-builder.h (rtx_vector_builder::step): Return a
19438 poly_wide_int instead of a wide_int.
19439 (rtx_vector_builder::apply_set): Take a poly_wide_int instead
19441 * rtx-vector-builder.c (rtx_vector_builder::apply_set): Likewise.
19442 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Return
19443 false for CONST_VECTORs that cannot be forced to memory.
19444 * config/aarch64/aarch64-simd.md (mov<mode>): If a CONST_VECTOR
19445 is too complex to force to memory, build it up from individual
19448 2021-03-31 Jan Hubicka <jh@suse.cz>
19451 * cgraph.c (cgraph_node::release_body): Fix overactive check.
19453 2021-03-31 Christophe Lyon <christophe.lyon@linaro.org>
19456 * config/arm/vec-common.md (mul<mode>3): Disable on iwMMXT, expect
19459 2021-03-31 H.J. Lu <hjl.tools@gmail.com>
19461 * config/i386/i386-expand.c (expand_set_or_cpymem_via_rep):
19462 For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, don't convert QImode
19464 (decide_alg): For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, use
19465 "rep movsb/stosb" only for known sizes.
19466 * config/i386/i386-options.c (processor_cost_table): Use Ice
19467 Lake cost for Cannon Lake, Ice Lake, Tiger Lake, Sapphire
19468 Rapids and Alder Lake.
19469 * config/i386/i386.h (TARGET_PREFER_KNOWN_REP_MOVSB_STOSB): New.
19470 * config/i386/x86-tune-costs.h (icelake_memcpy): New.
19471 (icelake_memset): Likewise.
19472 (icelake_cost): Likewise.
19473 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
19476 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19479 * config/aarch64/aarch64.c
19480 (aarch64_vectorize_preferred_vector_alignment): Query the size
19481 of the provided SVE vector; do not assume that all SVE vectors
19482 have the same size.
19484 2021-03-31 Jan Hubicka <jh@suse.cz>
19487 * cgraph.c (cgraph_node::release_body): Remove all callers and
19489 * cgraphclones.c (cgraph_node::materialize_clone): Do not do it here.
19490 * cgraphunit.c (cgraph_node::expand): And here.
19492 2021-03-31 Martin Liska <mliska@suse.cz>
19494 * ipa-modref.c (analyze_ssa_name_flags): Fix coding style
19495 and one negated condition.
19497 2021-03-31 Jakub Jelinek <jakub@redhat.com>
19498 Richard Sandiford <richard.sandiford@arm.com>
19501 * config/aarch64/aarch64.md (*add<mode>3_poly_1): Swap Uai and Uav
19502 constraints on operands[2] and similarly 0 and rk constraints
19503 on operands[1] corresponding to that.
19505 2021-03-31 Jakub Jelinek <jakub@redhat.com>
19508 * configure.ac (HAVE_LD_BROKEN_PE_DWARF5): New AC_DEFINE if PECOFF
19509 linker doesn't support DWARF sections new in DWARF5.
19510 * config/i386/i386-options.c (ix86_option_override_internal): Default
19511 to dwarf_version 4 if HAVE_LD_BROKEN_PE_DWARF5 for TARGET_PECOFF
19513 * config.in: Regenerated.
19514 * configure: Regenerated.
19516 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19519 * config/aarch64/aarch64.c (aarch64_analyze_loop_vinfo): Check for
19520 available issue_info before using it.
19522 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19525 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): Do not allow zero
19528 2021-03-30 Xionghu Luo <luoxhu@linux.ibm.com>
19531 * config/rs6000/altivec.md (altivec_lvsl_reg): Change to ...
19532 (altivec_lvsl_reg_<mode>): ... this.
19533 (altivec_lvsr_reg): Change to ...
19534 (altivec_lvsr_reg_<mode>): ... this.
19535 * config/rs6000/predicates.md (vec_set_index_operand): New.
19536 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
19537 Enable 32bit variable vec_insert for all TARGET_VSX.
19538 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
19539 Enable 32bit variable vec_insert for p9 and above.
19540 (rs6000_expand_vector_set_var_p8): Rename to ...
19541 (rs6000_expand_vector_set_var_p7): ... this.
19542 (rs6000_expand_vector_set): Use TARGET_VSX and adjust assert
19544 * config/rs6000/vector.md (vec_set<mode>): Use vec_set_index_operand.
19545 * config/rs6000/vsx.md (xl_len_r): Use gen_altivec_lvsl_reg_di and
19546 gen_altivec_lvsr_reg_di.
19548 2021-03-30 H.J. Lu <hjl.tools@gmail.com>
19551 * config/i386/ia32intrin.h (__rdtsc): Defined as macro.
19552 (__rdtscp): Likewise.
19554 2021-03-30 Tamar Christina <tamar.christina@arm.com>
19556 PR tree-optimization/99825
19557 * tree-vect-slp-patterns.c (vect_check_evenodd_blend):
19558 Reject non-mult 2 lanes.
19560 2021-03-30 Richard Earnshaw <rearnsha@arm.com>
19563 * config/arm/arm.c (arm_file_start): Fix emission of
19564 Tag_ABI_VFP_args attribute.
19566 2021-03-30 Richard Biener <rguenther@suse.de>
19568 PR tree-optimization/99824
19569 * stor-layout.c (set_min_and_max_values_for_integral_type):
19570 Assert the precision is within the bounds of
19571 WIDE_INT_MAX_PRECISION.
19572 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
19573 the outermost component ref only to lower the access size
19574 and initialize that from the access type.
19576 2021-03-30 Richard Sandiford <richard.sandiford@arm.com>
19579 * config/aarch64/aarch64.md (mov<mode>): Pass multi-instruction
19580 CONST_INTs to aarch64_expand_mov_immediate when called after RA.
19582 2021-03-30 Mihailo Stojanovic <mihailo.stojanovic@typhoon-hil.com>
19584 * config/aarch64/aarch64.md
19585 (<optab>_trunc<fcvt_target><GPI:mode>2): Set the "arch"
19586 attribute to disambiguate between SIMD and FP variants of the
19589 2021-03-29 Jan Hubicka <hubicka@ucw.cz>
19591 * ipa-modref.c (merge_call_lhs_flags): Correct handling of deref.
19592 (analyze_ssa_name_flags): Fix typo in comment.
19594 2021-03-29 Alex Coplan <alex.coplan@arm.com>
19597 * config/aarch64/aarch64-sve-builtins.cc
19598 (function_builder::add_function): Add placeholder_p argument, use
19599 placeholder decls if this is set.
19600 (function_builder::add_unique_function): Instead of conditionally adding
19601 direct overloads, unconditionally add either a direct overload or a
19603 (function_builder::add_overloaded_function): Set placeholder_p if we're
19604 using C++ overloads. Use the obstack for string storage instead
19605 of relying on the tree nodes.
19606 (function_builder::add_overloaded_functions): Don't return early for
19607 m_direct_overloads: we need to add placeholders.
19608 * config/aarch64/aarch64-sve-builtins.h
19609 (function_builder::add_function): Add placeholder_p argument.
19611 2021-03-29 Richard Biener <rguenther@suse.de>
19613 PR tree-optimization/99807
19614 * tree-vect-slp.c (vect_slp_analyze_node_operations_1): Move
19615 assert below VEC_PERM handling.
19617 2021-03-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19620 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>): Use
19621 aarch64_simd_or_scalar_imm_zero to match zeroes. Remove pattern
19622 matching const_int 0.
19623 (move_lo_quad_internal_be_<mode>): Likewise.
19624 (move_lo_quad_<mode>): Update for the above.
19625 * config/aarch64/iterators.md (VQ_2E): Delete.
19627 2021-03-29 Jakub Jelinek <jakub@redhat.com>
19629 PR tree-optimization/99777
19630 * fold-const.c (extract_muldiv_1): For conversions, punt on casts from
19631 types other than scalar integral types.
19633 2021-03-28 David Edelsohn <dje.gcc@gmail.com>
19635 * config/rs6000/rs6000.c (rs6000_output_dwarf_dtprel): Do not add
19636 XCOFF TLS reloc decorations.
19638 2021-03-28 Gerald Pfeifer <gerald@pfeifer.com>
19640 * doc/analyzer.texi (Analyzer Internals): Update link to
19641 "A Memory Model for Static Analysis of C Programs".
19643 2021-03-26 David Edelsohn <dje.gcc@gmail.com>
19645 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Call function.
19646 * config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align):
19648 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align): New.
19649 (rs6000_special_round_type_align): Recursively check innermost first
19652 2021-03-26 Jakub Jelinek <jakub@redhat.com>
19655 * dwarf2out.h (struct dw_fde_node): Add rule18 member.
19656 * dwarf2cfi.c (dwarf2out_frame_debug_expr): When handling (set hfp sp)
19657 assignment with drap_reg active, queue reg save for hfp with offset 0
19658 and flush queued reg saves. When handling a push with rule18,
19659 defer queueing reg save for hfp and just assert the offset is 0.
19660 (scan_trace): Assert that fde->rule18 is false.
19662 2021-03-26 Vladimir Makarov <vmakarov@redhat.com>
19665 * ira-costs.c (record_reg_classes): Put case with
19666 CT_RELAXED_MEMORY adjacent to one with CT_MEMORY.
19667 * ira.c (ira_setup_alts): Ditto.
19668 * lra-constraints.c (process_alt_operands): Ditto.
19669 * recog.c (asm_operand_ok): Ditto.
19670 * reload.c (find_reloads): Ditto.
19672 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19674 * config/aarch64/aarch64-protos.h
19675 (cpu_addrcost_table::post_modify_ld3_st3): New member variable.
19676 (cpu_addrcost_table::post_modify_ld4_st4): Likewise.
19677 * config/aarch64/aarch64.c (generic_addrcost_table): Update
19678 accordingly, using the same costs as for post_modify.
19679 (exynosm1_addrcost_table, xgene1_addrcost_table): Likewise.
19680 (thunderx2t99_addrcost_table, thunderx3t110_addrcost_table):
19681 (tsv110_addrcost_table, qdf24xx_addrcost_table): Likewise.
19682 (a64fx_addrcost_table): Likewise.
19683 (neoversev1_addrcost_table): New.
19684 (neoversev1_tunings): Use neoversev1_addrcost_table.
19685 (aarch64_address_cost): Use the new post_modify costs for CImode
19688 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19690 * config/aarch64/aarch64.opt
19691 (-param=aarch64-loop-vect-issue-rate-niters=): New parameter.
19692 * doc/invoke.texi: Document it.
19693 * config/aarch64/aarch64-protos.h (aarch64_base_vec_issue_info)
19694 (aarch64_scalar_vec_issue_info, aarch64_simd_vec_issue_info)
19695 (aarch64_advsimd_vec_issue_info, aarch64_sve_vec_issue_info)
19696 (aarch64_vec_issue_info): New structures.
19697 (cpu_vector_cost): Write comments above the variables rather
19699 (cpu_vector_cost::issue_info): New member variable.
19700 * config/aarch64/aarch64.c: Include gimple-pretty-print.h
19701 and tree-ssa-loop-niter.h.
19702 (generic_vector_cost, a64fx_vector_cost, qdf24xx_vector_cost)
19703 (thunderx_vector_cost, tsv110_vector_cost, cortexa57_vector_cost)
19704 (exynosm1_vector_cost, xgene1_vector_cost, thunderx2t99_vector_cost)
19705 (thunderx3t110_vector_cost): Initialize issue_info to null.
19706 (neoversev1_scalar_issue_info, neoversev1_advsimd_issue_info)
19707 (neoversev1_sve_issue_info, neoversev1_vec_issue_info): New structures.
19708 (neoversev1_vector_cost): Use them.
19709 (aarch64_vec_op_count, aarch64_sve_op_count): New structures.
19710 (aarch64_vector_costs::saw_sve_only_op): New member variable.
19711 (aarch64_vector_costs::num_vector_iterations): Likewise.
19712 (aarch64_vector_costs::scalar_ops): Likewise.
19713 (aarch64_vector_costs::advsimd_ops): Likewise.
19714 (aarch64_vector_costs::sve_ops): Likewise.
19715 (aarch64_vector_costs::seen_loads): Likewise.
19716 (aarch64_simd_vec_costs_for_flags): New function.
19717 (aarch64_analyze_loop_vinfo): Initialize num_vector_iterations.
19718 Count the number of predicate operations required by SVE WHILE
19720 (aarch64_comparison_type, aarch64_multiply_add_p): New functions.
19721 (aarch64_sve_only_stmt_p, aarch64_in_loop_reduction_latency): Likewise.
19722 (aarch64_count_ops): Likewise.
19723 (aarch64_add_stmt_cost): Record whether see an SVE operation
19724 that cannot currently be implementing using Advanced SIMD.
19725 Record issue information about the scalar, Advanced SIMD
19726 and (where relevant) SVE versions of a loop.
19727 (aarch64_vec_op_count::dump): New function.
19728 (aarch64_sve_op_count::dump): Likewise.
19729 (aarch64_estimate_min_cycles_per_iter): Likewise.
19730 (aarch64_adjust_body_cost): If issue information is available,
19731 try to compare the issue rates of the various loop implementations
19732 and increase or decrease the vector body cost accordingly.
19734 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19736 * config/aarch64/aarch64.c (aarch64_detect_vector_stmt_subtype):
19737 Assume a zero cost for induction phis.
19739 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19741 * config/aarch64/aarch64.c (aarch64_embedded_comparison_type): New
19743 (aarch64_adjust_stmt_cost): Add the costs of embedded scalar and
19744 vector comparisons.
19746 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19748 * config/aarch64/aarch64.c (aarch64_detect_scalar_stmt_subtype):
19750 (aarch64_add_stmt_cost): Call it.
19752 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19754 * config/aarch64/aarch64-tuning-flags.def (matched_vector_throughput):
19755 New tuning parameter.
19756 * config/aarch64/aarch64.c (neoversev1_tunings): Use it.
19757 (aarch64_estimated_sve_vq): New function.
19758 (aarch64_vector_costs::analyzed_vinfo): New member variable.
19759 (aarch64_vector_costs::is_loop): Likewise.
19760 (aarch64_vector_costs::unrolled_advsimd_niters): Likewise.
19761 (aarch64_vector_costs::unrolled_advsimd_stmts): Likewise.
19762 (aarch64_record_potential_advsimd_unrolling): New function.
19763 (aarch64_analyze_loop_vinfo, aarch64_analyze_bb_vinfo): Likewise.
19764 (aarch64_add_stmt_cost): Call aarch64_analyze_loop_vinfo or
19765 aarch64_analyze_bb_vinfo on the first use of a costs structure.
19766 Detect whether we're vectorizing a loop for SVE that might be
19767 completely unrolled if it used Advanced SIMD instead.
19768 (aarch64_adjust_body_cost_for_latency): New function.
19769 (aarch64_finish_cost): Call it.
19771 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19773 * config/aarch64/aarch64.c (aarch64_vector_costs): New structure.
19774 (aarch64_init_cost): New function.
19775 (aarch64_add_stmt_cost): Use aarch64_vector_costs instead of
19776 the default unsigned[3].
19777 (aarch64_finish_cost, aarch64_destroy_cost_data): New functions.
19778 (TARGET_VECTORIZE_INIT_COST): Override.
19779 (TARGET_VECTORIZE_FINISH_COST): Likewise.
19780 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
19782 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19784 * config/aarch64/aarch64.c (neoversev1_advsimd_vector_cost)
19785 (neoversev1_sve_vector_cost): New cost structures.
19786 (neoversev1_vector_cost): Likewise.
19787 (neoversev1_tunings): Use them. Enable use_new_vector_costs.
19789 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19791 * config/aarch64/aarch64-protos.h
19792 (sve_vec_cost::scatter_store_elt_cost): New member variable.
19793 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
19794 accordingly, taking the cost from the cost of a scalar_store.
19795 (a64fx_sve_vector_cost): Likewise.
19796 (aarch64_detect_vector_stmt_subtype): Detect scatter stores.
19798 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19800 * config/aarch64/aarch64-protos.h
19801 (simd_vec_cost::store_elt_extra_cost): New member variable.
19802 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
19803 accordingly, using the vec_to_scalar cost for the new field.
19804 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
19805 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
19806 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
19807 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
19808 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
19809 (thunderx3t110_advsimd_vector_cost): Likewise.
19810 (aarch64_detect_vector_stmt_subtype): Detect single-element stores.
19812 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19814 * config/aarch64/aarch64-protos.h (simd_vec_cost::ld2_st2_permute_cost)
19815 (simd_vec_cost::ld3_st3_permute_cost): New member variables.
19816 (simd_vec_cost::ld4_st4_permute_cost): Likewise.
19817 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
19818 accordingly, using zero for the new costs.
19819 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
19820 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
19821 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
19822 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
19823 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
19824 (thunderx3t110_advsimd_vector_cost): Likewise.
19825 (aarch64_ld234_st234_vectors): New function.
19826 (aarch64_adjust_stmt_cost): Likewise.
19827 (aarch64_add_stmt_cost): Call aarch64_adjust_stmt_cost if using
19828 the new vector costs.
19830 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19832 * config/aarch64/aarch64-protos.h (sve_vec_cost): Turn into a
19833 derived class of simd_vec_cost. Add information about CLAST[AB]
19834 and FADDA instructions.
19835 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
19836 accordingly, using the vec_to_scalar costs for the new fields.
19837 (a64fx_sve_vector_cost): Likewise.
19838 (aarch64_reduc_type): New function.
19839 (aarch64_sve_in_loop_reduction_latency): Likewise.
19840 (aarch64_detect_vector_stmt_subtype): Take a vinfo parameter.
19841 Use aarch64_sve_in_loop_reduction_latency to handle SVE reductions
19842 that occur in the loop body.
19843 (aarch64_add_stmt_cost): Update call accordingly.
19845 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19847 * config/aarch64/aarch64-tuning-flags.def (use_new_vector_costs):
19849 * config/aarch64/aarch64-protos.h (simd_vec_cost): Put comments
19850 above the fields rather than to the right.
19851 (simd_vec_cost::reduc_i8_cost): New member variable.
19852 (simd_vec_cost::reduc_i16_cost): Likewise.
19853 (simd_vec_cost::reduc_i32_cost): Likewise.
19854 (simd_vec_cost::reduc_i64_cost): Likewise.
19855 (simd_vec_cost::reduc_f16_cost): Likewise.
19856 (simd_vec_cost::reduc_f32_cost): Likewise.
19857 (simd_vec_cost::reduc_f64_cost): Likewise.
19858 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
19859 accordingly, using the vec_to_scalar_cost for the new fields.
19860 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
19861 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
19862 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
19863 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
19864 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
19865 (thunderx3t110_advsimd_vector_cost): Likewise.
19866 (aarch64_use_new_vector_costs_p): New function.
19867 (aarch64_simd_vec_costs): New function, split out from...
19868 (aarch64_builtin_vectorization_cost): ...here.
19869 (aarch64_is_reduction): New function.
19870 (aarch64_detect_vector_stmt_subtype): Likewise.
19871 (aarch64_add_stmt_cost): Call aarch64_detect_vector_stmt_subtype if
19872 using the new vector costs.
19874 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19877 * tree-emutls.c (get_emutls_init_templ_addr): Mark initializer of weak
19878 TLS declarations as public.
19880 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19882 * config/aarch64/aarch64-d.c (IN_TARGET_CODE): Define.
19883 * config/arm/arm-d.c (IN_TARGET_CODE): Likewise.
19884 * config/i386/i386-d.c (IN_TARGET_CODE): Likewise.
19885 * config/mips/mips-d.c (IN_TARGET_CODE): Likewise.
19886 * config/pa/pa-d.c (IN_TARGET_CODE): Likewise.
19887 * config/riscv/riscv-d.c (IN_TARGET_CODE): Likewise.
19888 * config/rs6000/rs6000-d.c (IN_TARGET_CODE): Likewise.
19889 * config/s390/s390-d.c (IN_TARGET_CODE): Likewise.
19890 * config/sparc/sparc-d.c (IN_TARGET_CODE): Likewise.
19892 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19895 * config.gcc (*-*-cygwin*): Add winnt-d.o
19896 (*-*-mingw*): Likewise.
19897 * config/i386/cygwin.h (EXTRA_TARGET_D_OS_VERSIONS): New macro.
19898 * config/i386/mingw32.h (EXTRA_TARGET_D_OS_VERSIONS): Likewise.
19899 * config/i386/t-cygming: Add winnt-d.o.
19900 * config/i386/winnt-d.c: New file.
19902 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19904 * config/freebsd-d.c: Include memmodel.h.
19906 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19909 * config.gcc (*-*-openbsd*): Add openbsd-d.o.
19910 * config/t-openbsd: Add openbsd-d.o.
19911 * config/openbsd-d.c: New file.
19913 2021-03-25 Stam Markianos-Wright <stam.markianos-wright@arm.com>
19915 PR tree-optimization/96974
19916 * tree-vect-stmts.c (vect_get_vector_types_for_stmt): Replace assert
19917 with graceful exit.
19919 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
19922 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
19926 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
19927 always_inline in system headers.
19929 2021-03-25 Kewen Lin <linkw@linux.ibm.com>
19931 * tree-vect-loop.c (vect_model_reduction_cost): Init inside_cost.
19933 2021-03-25 Jakub Jelinek <jakub@redhat.com>
19936 * tree-core.h (enum operand_equal_flag): Add OEP_ADDRESS_OF_SAME_FIELD.
19937 * fold-const.c (operand_compare::operand_equal_p): Don't compare
19938 field offsets if OEP_ADDRESS_OF_SAME_FIELD.
19940 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
19944 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
19945 always_inline in system headers.
19947 2021-03-25 Richard Biener <rguenther@suse.de>
19949 PR tree-optimization/99746
19950 * tree-vect-slp-patterns.c (complex_pattern::build): Do not mark
19951 the scalar stmt as patterned. Instead set up required things
19954 2021-03-25 Xionghu Luo <luoxhu@linux.ibm.com>
19956 * config/rs6000/rs6000.c (power8_costs): Change l2 cache
19959 2021-03-24 Martin Liska <mliska@suse.cz>
19962 * common/config/i386/i386-common.c (ARRAY_SIZE): Fix off-by-one
19964 * config/i386/i386-options.c (ix86_option_override_internal):
19965 Add run-time assert.
19967 2021-03-24 Martin Jambor <mjambor@suse.cz>
19970 * ipa-cp.c (initialize_node_lattices): Mark as bottom all
19971 parameters with unknown type.
19972 (ipacp_value_safe_for_type): New function.
19973 (propagate_vals_across_arith_jfunc): Verify that the constant type
19974 can be used for a type of the formal parameter.
19975 (propagate_vals_across_ancestor): Likewise.
19976 (propagate_scalar_across_jump_function): Likewise. Pass the type
19977 also to propagate_vals_across_ancestor.
19979 2021-03-24 Christophe Lyon <christophe.lyon@linaro.org>
19982 * config/arm/mve.md (movmisalign<mode>_mve_store): Use Ux
19984 (movmisalign<mode>_mve_load): Likewise.
19986 2021-03-24 Jakub Jelinek <jakub@redhat.com>
19989 * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
19990 movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.
19992 2021-03-24 Alexandre Oliva <oliva@adacore.com>
19994 * doc/sourcebuild.texi (sysconf): New effective target.
19996 2021-03-24 Alexandre Oliva <oliva@adacore.com>
19998 * config/i386/predicates.md (reg_or_const_vec_operand): New.
19999 * config/i386/sse.md (ssse3_pshufbv8qi3): Add an expander for
20000 the now *-prefixed insn_and_split, turn the splitter const vec
20001 into an input for the insn, making it an ignored immediate for
20002 non-split cases, and loaded into the scratch register
20005 2021-03-23 Vladimir N. Makarov <vmakarov@redhat.com>
20008 * config/aarch64/constraints.md (Utq, UOb, UOh, UOw, UOd, UOty):
20009 Use define_relaxed_memory_constraint for them.
20011 2021-03-23 Iain Sandoe <iain@sandoe.co.uk>
20014 * config/host-darwin.c (darwin_gt_pch_use_address): Add a
20015 colon to the diagnostic message.
20017 2021-03-23 Ilya Leoshkevich <iii@linux.ibm.com>
20019 * fwprop.c (fwprop_propagation::fwprop_propagation): Look at
20021 (try_fwprop_subst_note): Use set_info instead of insn_info.
20022 (try_fwprop_subst_pattern): Likewise.
20023 (try_fwprop_subst_notes): Likewise.
20024 (try_fwprop_subst): Likewise.
20025 (forward_propagate_subreg): Likewise.
20026 (forward_propagate_and_simplify): Likewise.
20027 (forward_propagate_into): Likewise.
20028 * rtl-ssa/accesses.h (set_info::single_nondebug_use) New
20030 (set_info::single_nondebug_insn_use): Likewise.
20031 (set_info::single_phi_use): Likewise.
20032 * rtl-ssa/member-fns.inl (set_info::single_nondebug_use) New
20034 (set_info::single_nondebug_insn_use): Likewise.
20035 (set_info::single_phi_use): Likewise.
20037 2021-03-23 Christophe Lyon <christophe.lyon@linaro.org>
20039 * doc/sourcebuild.texi (arm_dsp_ok, arm_dsp): Document.
20041 2021-03-23 Jakub Jelinek <jakub@redhat.com>
20044 * config/aarch64/aarch64.c (aarch64_add_offset): Tell
20045 expand_mult to perform an unsigned rather than a signed
20048 2021-03-23 H.J. Lu <hjl.tools@gmail.com>
20051 * config/i386/cpuid.h (__cpuid): Add __volatile__.
20052 (__cpuid_count): Likewise.
20054 2021-03-23 Richard Biener <rguenther@suse.de>
20056 PR tree-optimization/99721
20057 * tree-vect-slp.c (vect_slp_analyze_node_operations):
20058 Make sure we can schedule the node.
20060 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20062 * config/riscv/riscv.c (riscv_subword): Take endianness into
20063 account when calculating the byte offset.
20065 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20067 * config/riscv/predicates.md (subreg_lowpart_operator): New predicate
20068 * config/riscv/riscv.md (*addsi3_extended2, *subsi3_extended2)
20069 (*negsi2_extended2, *mulsi3_extended2, *<optab>si3_mask)
20070 (*<optab>si3_mask_1, *<optab>di3_mask, *<optab>di3_mask_1)
20071 (*<optab>si3_extend_mask, *<optab>si3_extend_mask_1): Use
20072 new predicate "subreg_lowpart_operator"
20074 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20076 * config/riscv/riscv.c (riscv_swap_instruction): New function
20077 to byteswap an SImode rtx containing an instruction.
20078 (riscv_trampoline_init): Byteswap the generated instructions
20081 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20083 * common/config/riscv/riscv-common.c
20084 (TARGET_DEFAULT_TARGET_FLAGS): Set default endianness.
20085 * config.gcc (riscv32be-*, riscv64be-*): Set
20086 TARGET_BIG_ENDIAN_DEFAULT to 1.
20087 * config/riscv/elf.h (LINK_SPEC): Change -melf* value
20088 depending on default endianness.
20089 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
20090 * config/riscv/linux.h (LINK_SPEC): Likewise.
20091 * config/riscv/riscv.c (TARGET_DEFAULT_TARGET_FLAGS): Set
20092 default endianness.
20093 * config/riscv/riscv.h (DEFAULT_ENDIAN_SPEC): New macro.
20095 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20097 * config/riscv/elf.h (LINK_SPEC): Pass linker endianness flag.
20098 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
20099 * config/riscv/linux.h (LINK_SPEC): Likewise.
20100 * config/riscv/riscv.h (ASM_SPEC): Pass -mbig-endian and
20102 (BYTES_BIG_ENDIAN): Handle big endian.
20103 (WORDS_BIG_ENDIAN): Define to BYTES_BIG_ENDIAN.
20104 * config/riscv/riscv.opt (-mbig-endian, -mlittle-endian): New
20106 * doc/invoke.texi (-mbig-endian, -mlittle-endian): Document.
20108 2021-03-23 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
20110 * regcprop.c (find_oldest_value_reg): Ask target whether
20111 different mode is fine for replacement register.
20113 2021-03-23 Aldy Hernandez <aldyh@redhat.com>
20115 PR tree-optimization/99296
20116 * value-range.cc (irange::irange_set_1bit_anti_range): New.
20117 (irange::irange_set_anti_range): Call irange_set_1bit_anti_range
20118 * value-range.h (irange::irange_set_1bit_anti_range): New.
20120 2021-03-22 Vladimir N. Makarov <vmakarov@redhat.com>
20123 * config/aarch64/constraints.md (UtQ): Use
20124 define_relaxed_memory_constraint for it.
20125 * doc/md.texi (define_relaxed_memory_constraint): Describe it.
20126 * genoutput.c (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
20127 * genpreds.c (constraint_data): Add bitfield is_relaxed_memory.
20128 (have_relaxed_memory_constraints): New static var.
20129 (relaxed_memory_start, relaxed_memory_end): Ditto.
20130 (add_constraint): Add arg is_relaxed_memory. Check name for
20131 relaxed memory. Set up is_relaxed_memory in constraint_data and
20132 have_relaxed_memory_constraints. Adjust calls.
20133 (choose_enum_order): Process relaxed memory.
20134 (write_tm_preds_h): Ditto.
20135 (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
20136 * gensupport.c (process_rtx): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
20137 * ira-costs.c (record_reg_classes): Process CT_RELAXED_MEMORY.
20138 * ira-lives.c (single_reg_class): Use
20139 insn_extra_relaxed_memory_constraint.
20140 * ira.c (ira_setup_alts): CT_RELAXED_MEMORY.
20141 * lra-constraints.c (valid_address_p): Use
20142 insn_extra_relaxed_memory_constraint instead of other memory
20144 (process_alt_operands): Process CT_RELAXED_MEMORY.
20145 (curr_insn_transform): Use insn_extra_relaxed_memory_constraint.
20146 * recog.c (asm_operand_ok, preprocess_constraints): Process
20148 * reload.c (find_reloads): Ditto.
20149 * rtl.def (DEFINE_RELAXED_MEMORY_CONSTRAINT): New.
20150 * stmt.c (parse_input_constraint): Use
20151 insn_extra_relaxed_memory_constraint.
20153 2021-03-22 Segher Boessenkool <segher@kernel.crashing.org>
20156 * ubsan.c (ubsan_instrument_float_cast): Don't test for unordered if
20159 2021-03-22 Alex Coplan <alex.coplan@arm.com>
20162 * config/arm/arm-protos.h (neon_make_constant): Add generate
20163 argument to guard emitting insns, default to true.
20164 * config/arm/arm.c (arm_legitimate_constant_p_1): Reject
20165 CONST_VECTORs which neon_make_constant can't handle.
20166 (neon_vdup_constant): Add generate argument, avoid emitting
20167 insns if it's not set.
20168 (neon_make_constant): Plumb new generate argument through.
20169 * config/arm/constraints.md (Ui): New. Use it...
20170 * config/arm/mve.md (*mve_mov<mode>): ... here.
20171 * config/arm/vec-common.md (movv8hf): Use neon_make_constant to
20172 synthesize constants.
20174 2021-03-22 Richard Biener <rguenther@suse.de>
20176 * debug.h: Add deprecation warning.
20178 2021-03-22 Richard Biener <rguenther@suse.de>
20180 PR tree-optimization/99694
20181 * tree-ssa-sccvn.c (visit_phi): Ignore edges with the
20184 2021-03-22 Kito Cheng <kito.cheng@sifive.com>
20187 * config/riscv/riscv.c (riscv_expand_block_move): Get RTL value
20188 after type checking.
20190 2021-03-22 Jakub Jelinek <jakub@redhat.com>
20194 * dwarf2out.c (get_full_len): Use get_precision rather than
20196 (add_const_value_attribute): Make sure add_AT_wide argument has
20197 precision prec rather than some very wide one.
20199 2021-03-22 Kewen Lin <linkw@linux.ibm.com>
20201 * config/rs6000/rs6000.md (*rotldi3_insert_sf,
20202 *mov<SFDF:mode><SFDF2:mode>cc_p9, floatsi<mode>2_lfiwax,
20203 floatsi<mode>2_lfiwax_mem, floatunssi<mode>2_lfiwzx,
20204 floatunssi<mode>2_lfiwzx_mem, *floatsidf2_internal,
20205 *floatunssidf2_internal, fix_trunc<mode>si2_stfiwx,
20206 fix_trunc<mode>si2_internal, fixuns_trunc<mode>si2_stfiwx,
20207 *round32<mode>2_fprs, *roundu32<mode>2_fprs,
20208 *fix_trunc<mode>si2_internal): Fix empty split condition.
20209 * config/rs6000/vsx.md (*vsx_le_undo_permute_<mode>,
20210 vsx_reduc_<VEC_reduc_name>_v2df, vsx_reduc_<VEC_reduc_name>_v4sf,
20211 *vsx_reduc_<VEC_reduc_name>_v2df_scalar,
20212 *vsx_reduc_<VEC_reduc_name>_v4sf_scalar): Likewise.
20214 2021-03-22 Xionghu Luo <luoxhu@linux.ibm.com>
20217 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
20218 Convert idx to DImode.
20219 (rs6000_expand_vector_set_var_p8): Likewise.
20221 2021-03-21 Jakub Jelinek <jakub@redhat.com>
20224 * dwarf2out.c (insert_float): Change return type from void to
20225 unsigned, handle GET_MODE_SIZE (mode) == 2 and return element size.
20226 (mem_loc_descriptor, loc_descriptor, add_const_value_attribute):
20229 2021-03-20 H.J. Lu <hjl.tools@gmail.com>
20232 * config/i386/i386.c (construct_container): Check cfun != NULL
20233 before accessing silent_p.
20235 2021-03-20 Ahamed Husni <ahamedhusni73@gmail.com>
20237 * asan.c: Fix typos in comments.
20239 2021-03-20 Vladimir N. Makarov <vmakarov@redhat.com>
20241 PR rtl-optimization/99680
20242 * lra-constraints.c (skip_contraint_modifiers): Rename to skip_constraint_modifiers.
20243 (process_address_1): Check empty constraint before using
20246 2021-03-19 Pat Haugen <pthaugen@linux.ibm.com>
20248 * config/rs6000/rs6000.c (power10_cost): New.
20249 (rs6000_option_override_internal): Set Power10 costs.
20250 (rs6000_issue_rate): Set Power10 issue rate.
20251 * config/rs6000/power10.md: Rewrite for Power10.
20253 2021-03-19 Vladimir N. Makarov <vmakarov@redhat.com>
20256 * lra-constraints.c (process_address_1): Don't use unknown
20257 constraint for address constraint.
20259 2021-03-19 Iain Sandoe <iain@sandoe.co.uk>
20262 * config.gcc (powerpc-*-darwin8): Delete the reference to
20263 the now removed darwin8.h.
20265 2021-03-19 Olivier Hainque <hainque@adacore.com>
20268 * config/vxworksae.h (VX_CPU_PREFIX): Define.
20270 2021-03-19 John David Anglin <danglin@gcc.gnu.org>
20272 * config/pa/pa.c (import_milli): Use memcpy instead of strncpy.
20274 2021-03-19 Tamar Christina <tamar.christina@arm.com>
20276 PR tree-optimization/99656
20277 * tree-vect-slp-patterns.c (linear_loads_p,
20278 complex_add_pattern::matches, is_eq_or_top,
20279 vect_validate_multiplication, complex_mul_pattern::matches,
20280 complex_fms_pattern::matches): Remove complex_perm_kinds_t.
20281 * tree-vectorizer.h: (complex_load_perm_t): Removed.
20282 (slp_tree_to_load_perm_map_t): Use complex_perm_kinds_t instead of
20283 complex_load_perm_t.
20285 2021-03-19 H.J. Lu <hjl.tools@gmail.com>
20288 * config/i386/i386-options.c (ix86_init_machine_status): Set
20290 * config/i386/i386.c (init_cumulative_args): Set silent_p to
20292 (construct_container): Return early for return and argument
20293 errors if silent_p is true.
20294 * config/i386/i386.h (machine_function): Add silent_p.
20296 2021-03-19 Jakub Jelinek <jakub@redhat.com>
20299 * config/arm/constraints.md (Ds): New constraint.
20300 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): Use w,Ds
20301 constraint instead of w,Dm.
20303 2021-03-19 Andrew Stubbs <ams@codesourcery.com>
20305 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Fix quotes
20308 2021-03-19 Eric Botcazou <ebotcazou@adacore.com>
20310 PR middle-end/99641
20311 * fold-const.c (native_encode_initializer) <CONSTRUCTOR>: For an
20312 array type, do the computation of the current position in sizetype.
20314 2021-03-18 Vladimir N. Makarov <vmakarov@redhat.com>
20317 * lra-constraints.c (process_address_1): Use lookup_constraint
20318 only for a single constraint.
20320 2021-03-18 Martin Sebor <msebor@redhat.com>
20322 PR middle-end/99502
20323 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): Rename...
20324 (inbounds_memaccess_p): ...to this. Check the ending offset of
20325 the accessed member.
20327 2021-03-18 Andrew Stubbs <ams@codesourcery.com>
20329 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Add %< and
20330 %> quote markers to error messages.
20331 (gcn_goacc_validate_dims): Likewise.
20332 (gcn_conditional_register_usage): Remove exclaimation mark from error
20334 (gcn_vectorize_vec_perm_const): Ensure perm is fully uninitialized.
20336 2021-03-18 Jan Hubicka <hubicka@ucw.cz>
20338 * config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
20341 2021-03-18 Sinan Lin <sinan@isrc.iscas.ac.cn>
20342 Kito Cheng <kito.cheng@sifive.com>
20344 * config/riscv/riscv.c (riscv_block_move_straight): Change type
20345 to unsigned HOST_WIDE_INT for parameter and local variable with
20346 HOST_WIDE_INT type.
20347 (riscv_adjust_block_mem): Ditto.
20348 (riscv_block_move_loop): Ditto.
20349 (riscv_expand_block_move): Ditto.
20351 2021-03-18 Nick Clifton <nickc@redhat.com>
20353 * config/v850/v850.c (construct_restore_jr): Increase static
20355 (construct_save_jarl): Likewise.
20356 * config/v850/v850.h (DWARF2_DEBUGGING_INFO): Define.
20358 2021-03-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20360 * config/aarch64/aarch64.c (aarch64_adjust_generic_arch_tuning): Define.
20361 (aarch64_override_options_internal): Use it.
20362 (generic_tunings): Add AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS to
20365 2021-03-17 Sandra Loosemore <sandra@codesourcery.com>
20367 * config/nios2/nios2.c (nios2_custom_check_insns): Clean up
20368 error message format issues.
20369 (nios2_option_override): Likewise.
20370 (nios2_expand_fpu_builtin): Likewise.
20371 (nios2_init_custom_builtins): Adjust to avoid bogus strncpy
20372 truncation warning.
20373 (nios2_expand_custom_builtin): More error message format fixes.
20374 (nios2_expand_rdwrctl_builtin): Likewise.
20375 (nios2_expand_rdprs_builtin): Likewise.
20376 (nios2_expand_eni_builtin): Likewise.
20377 (nios2_expand_builtin): Likewise.
20378 (nios2_register_custom_code): Likewise.
20379 (nios2_valid_target_attribute_rec): Likewise.
20380 (nios2_add_insn_asm): Fix uninitialized variable warning.
20382 2021-03-17 Jan Hubicka <jh@suse.cz>
20384 * config/i386/x86-tune-costs.h (struct processor_costs): Update costs
20385 of gather to match reality.
20386 * config/i386/x86-tune.def (X86_TUNE_USE_GATHER): Enable for znver3.
20388 2021-03-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20390 * config/aarch64/aarch64-builtins.c (aarch64_expand_rng_builtin): Use EQ
20391 to compare against CC_REG rather than NE.
20393 2021-03-17 H.J. Lu <hjl.tools@gmail.com>
20396 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support
20397 inline assembly statements.
20398 (ix86_print_operand): Update 'P' handling for -fno-plt.
20400 2021-03-17 Tamar Christina <tamar.christina@arm.com>
20403 * config/aarch64/aarch64.c
20404 (aarch64_simd_clone_compute_vecsize_and_simdlen): Remove unused var.
20406 2021-03-16 Segher Boessenkool <segher@kernel.crashing.org>
20409 * config/rs6000/predicates.md (branch_comparison_operator): Allow
20410 ordered and unordered for CCFPmode, if flag_finite_math_only.
20412 2021-03-16 Jakub Jelinek <jakub@redhat.com>
20415 * config/i386/i386-expand.c (ix86_split_lea_for_addr): Emit a MULT
20416 rather than ASHIFT.
20417 * config/i386/i386.md (mult by 1248 into ashift): New splitter.
20419 2021-03-16 Martin Liska <mliska@suse.cz>
20422 * optc-save-gen.awk: Add flag_ipa_ra to exceptions for
20423 cl_optimization_compare function.
20425 2021-03-16 Ilya Leoshkevich <iii@linux.ibm.com>
20427 * config/s390/s390.c (f_constraint_p): Treat "fv" constraints
20430 2021-03-16 Jakub Jelinek <jakub@redhat.com>
20433 * config/i386/i386.h (struct machine_function): Add
20434 has_explicit_vzeroupper bitfield.
20435 * config/i386/i386-expand.c (ix86_expand_builtin): Set
20436 cfun->machine->has_explicit_vzeroupper when expanding
20437 IX86_BUILTIN_VZEROUPPER.
20438 * config/i386/i386-features.c (rest_of_handle_insert_vzeroupper):
20439 Do the mode switching only when TARGET_VZEROUPPER, expensive
20440 optimizations turned on and not optimizing for size.
20441 (pass_insert_vzeroupper::gate): Enable even when
20442 cfun->machine->has_explicit_vzeroupper is set.
20444 2021-03-16 Jakub Jelinek <jakub@redhat.com>
20447 * config/aarch64/aarch64.c
20448 (aarch64_simd_clone_compute_vecsize_and_simdlen): If not a function
20449 definition, walk TYPE_ARG_TYPES list if non-NULL for argument types
20450 instead of DECL_ARGUMENTS. Ignore types for uniform arguments.
20452 2021-03-15 Richard Biener <rguenther@suse.de>
20454 PR tree-optimization/98834
20455 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle missing
20456 subsetting by truncating the access size.
20458 2021-03-15 Jan Hubicka <hubicka@ucw.cz>
20460 * config/i386/i386-options.c (processor_cost_table): Add znver3_cost.
20461 * config/i386/x86-tune-costs.h (znver3_cost): New gobal variable; copy
20464 2021-03-15 Martin Liska <mliska@suse.cz>
20466 * spellcheck.c: Add missing comma in initialization.
20468 2021-03-14 Uroš Bizjak <ubizjak@gmail.com>
20470 * config/i386/sse.md (*vec_extract<mode>): Merge alternative 0 with
20471 alternative 2 and alternative 1 with alternative 3 using
20472 YW register constraint.
20473 (*vec_extract<PEXTR_MODE12:mode>_zext): Merge alternatives
20474 using YW register constraint.
20475 (*vec_extractv16qi_zext): Ditto.
20476 (*vec_extractv4si): Merge alternatives 4 and 5
20477 using Yw register constraint.
20478 (*ssse3_palignr<mode>_perm): Use Yw instead of v for alternative 3.
20480 2021-03-13 Martin Sebor <msebor@redhat.com>
20482 PR tree-optimization/99489
20483 * builtins.c (gimple_call_alloc_size): Fail gracefully when argument
20484 is not a call statement.
20486 2021-03-13 Jakub Jelinek <jakub@redhat.com>
20488 PR tree-optimization/99544
20489 * match.pd (X + (X << C) -> X * (1 + (1 << C))): Don't simplify
20490 if for vector types multiplication can't be done in type's mode.
20492 2021-03-12 Eric Botcazou <ebotcazou@adacore.com>
20495 * config/sparc/constraints.md (w): Rename to...
20496 (W): ... this and ditch previous implementation.
20497 * config/sparc/sparc.md (*movdi_insn_sp64): Replace W with m.
20498 (*movdf_insn_sp64): Likewise.
20499 (*mov<VM64:mode>_insn_sp64): Likewise.
20500 * config/sparc/sync.md (*atomic_compare_and_swap<mode>_1): Replace
20502 (atomic_compare_and_swap_leon3_1): Likewise.
20503 (*atomic_compare_and_swapdi_v8plus): Likewise.
20504 * config/sparc/sparc.c (memory_ok_for_ldd): Remove useless test on
20505 architecture and add missing address validity check during LRA.
20507 2021-03-12 Tobias Burnus <tobias@codesourcery.com>
20510 * gimplify.c (omp_add_variable): Handle NULL_TREE as size
20511 occuring for assumed-size arrays in use_device_{ptr,addr}.
20513 2021-03-12 Jakub Jelinek <jakub@redhat.com>
20516 * config/i386/constraints.md (YW): New internal constraint.
20517 * config/i386/sse.md (v_Yw): Add V4TI, V2TI, V1TI and TI cases.
20518 (*<sse2_avx2>_<insn><mode>3<mask_name>,
20519 *<sse2_avx2>_uavg<mode>3<mask_name>, *abs<mode>2,
20520 *<s>mul<mode>3_highpart<mask_name>): Use <v_Yw> instead of v in
20522 (<sse2_avx2>_psadbw): Use YW instead of v in constraints.
20523 (*avx2_pmaddwd, *sse2_pmaddwd, *<code>v8hi3, *<code>v16qi3,
20524 avx2_pmaddubsw256, ssse3_pmaddubsw128): Merge last two alternatives
20525 into one, use Yw instead of former x,v.
20526 (ashr<mode>3, <insn><mode>3): Use <v_Yw> instead of x in constraints of
20527 the last alternative.
20528 (<sse2_avx2>_packsswb<mask_name>, <sse2_avx2>_packssdw<mask_name>,
20529 <sse2_avx2>_packuswb<mask_name>, <sse4_1_avx2>_packusdw<mask_name>,
20530 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, <ssse3_avx2>_palignr<mode>,
20531 <ssse3_avx2>_pshufb<mode>3<mask_name>): Merge last two alternatives
20532 into one, use <v_Yw> instead of former x,v.
20533 (avx2_interleave_highv32qi<mask_name>,
20534 vec_interleave_highv16qi<mask_name>): Use Yw instead of v in
20535 constraints. Add && <mask_avx512bw_condition> to condition.
20536 (avx2_interleave_lowv32qi<mask_name>,
20537 vec_interleave_lowv16qi<mask_name>,
20538 avx2_interleave_highv16hi<mask_name>,
20539 vec_interleave_highv8hi<mask_name>,
20540 avx2_interleave_lowv16hi<mask_name>, vec_interleave_lowv8hi<mask_name>,
20541 avx2_pshuflw_1<mask_name>, sse2_pshuflw_1<mask_name>,
20542 avx2_pshufhw_1<mask_name>, sse2_pshufhw_1<mask_name>,
20543 avx2_<code>v16qiv16hi2<mask_name>, sse4_1_<code>v8qiv8hi2<mask_name>,
20544 *sse4_1_<code>v8qiv8hi2<mask_name>_1, <sse2_avx2>_<insn><mode>3): Use
20545 Yw instead of v in constraints.
20546 * config/i386/mmx.md (Yv_Yw): New define_mode_attr.
20547 (*mmx_<insn><mode>3, mmx_ashr<mode>3, mmx_<insn><mode>3): Use <Yv_Yw>
20548 instead of Yv in constraints.
20549 (*mmx_<insn><mode>3, *mmx_mulv4hi3, *mmx_smulv4hi3_highpart,
20550 *mmx_umulv4hi3_highpart, *mmx_pmaddwd, *mmx_<code>v4hi3,
20551 *mmx_<code>v8qi3, mmx_pack<s_trunsuffix>swb, mmx_packssdw,
20552 mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
20553 *mmx_uavgv8qi3, *mmx_uavgv4hi3, mmx_psadbw): Use Yw instead of Yv in
20555 (*mmx_pinsrw, *mmx_pinsrb, *mmx_pextrw, *mmx_pextrw_zext, *mmx_pextrb,
20556 *mmx_pextrb_zext): Use YW instead of Yv in constraints.
20557 (*mmx_eq<mode>3, mmx_gt<mode>3): Use x instead of Yv in constraints.
20558 (mmx_andnot<mode>3, *mmx_<code><mode>3): Split last alternative into
20559 two, one with just x, another isa avx512vl with v.
20561 2021-03-12 Martin Liska <mliska@suse.cz>
20563 * doc/invoke.texi: Add missing param documentation.
20565 2021-03-11 David Malcolm <dmalcolm@redhat.com>
20568 * Makefile.in (ANALYZER_OBJS): Add analyzer/feasible-graph.o and
20569 analyzer/trimmed-graph.o.
20570 * doc/analyzer.texi (Analyzer Paths): Rewrite description of
20571 feasibility checking to reflect new implementation.
20572 * doc/invoke.texi (-fdump-analyzer-feasibility): Document new
20574 * shortest-paths.h (shortest_paths::get_shortest_distance): New.
20576 2021-03-11 David Malcolm <dmalcolm@redhat.com>
20578 * digraph.cc (selftest::test_shortest_paths): Update
20579 shortest_paths init for new param. Add test of
20580 SPS_TO_GIVEN_TARGET.
20581 * shortest-paths.h (enum shortest_path_sense): New.
20582 (shortest_paths::shortest_paths): Add "sense" param.
20583 Update for renamings. Generalize to use "sense" param.
20584 (shortest_paths::get_shortest_path): Rename param.
20585 (shortest_paths::m_sense): New field.
20586 (shortest_paths::m_prev): Rename...
20587 (shortest_paths::m_best_edge): ...to this.
20588 (shortest_paths::get_shortest_path): Update for renamings.
20589 Conditionalize flipping of path on sense of traversal.
20591 2021-03-11 David Malcolm <dmalcolm@redhat.com>
20593 * digraph.cc (selftest::test_shortest_paths): Add test coverage
20594 for paths from B and C.
20595 * shortest-paths.h (shortest_paths::shortest_paths): Handle
20596 unreachable nodes, rather than asserting.
20598 2021-03-11 David Edelsohn <dje.gcc@gmail.com>
20601 * config/rs6000/rs6000.c (rs6000_xcoff_file_start): Don't create
20602 xcoff_tbss_section_name.
20603 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .lcomm.
20604 * xcoffout.c (xcoff_tbss_section_name): Delete.
20605 * xcoffout.h (xcoff_tbss_section_name): Delete.
20607 2021-03-11 Richard Biener <rguenther@suse.de>
20609 PR tree-optimization/99523
20610 * tree-cfg.c (dump_function_to_file): Dump SSA names
20611 w/o identifier to the decls section as well, not only those
20612 without a VAR_DECL.
20614 2021-03-11 Jakub Jelinek <jakub@redhat.com>
20617 * ipa-icf-gimple.c (func_checker::compare_gimple_call): For internal
20618 function calls with lhs fail if the lhs don't have compatible types.
20620 2021-03-11 Hans-Peter Nilsson <hp@axis.com>
20622 * config/cris/cris.h (HARD_FRAME_POINTER_REGNUM): Define.
20623 Change FRAME_POINTER_REGNUM to correspond to a new faked
20624 register faked_fp, part of GENNONACR_REGS like faked_ap.
20625 (CRIS_FAKED_REGS_CONTENTS): New helper macro.
20626 (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS, CALL_USED_REGISTERS):
20627 (REG_ALLOC_ORDER, REG_CLASS_CONTENTS, REGNO_OK_FOR_BASE_P)
20628 (ELIMINABLE_REGS, REGISTER_NAMES): Adjust accordingly.
20629 * config/cris/cris.md (CRIS_FP_REGNUM): Renumber to new faked
20631 (CRIS_REAL_FP_REGNUM): New constant.
20632 * config/cris/cris.c (cris_reg_saved_in_regsave_area): Check
20633 for HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM.
20634 (cris_initial_elimination_offset): Handle elimination changes
20635 to HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM
20636 and add one from FRAME_POINTER_REGNUM to
20637 HARD_FRAME_POINTER_REGNUM.
20638 (cris_expand_prologue, cris_expand_epilogue): Emit code for
20639 hard_frame_pointer_rtx instead of frame_pointer_rtx.
20641 2021-03-10 David Edelsohn <dje.gcc@gmail.com>
20644 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Add check for DCmode.
20645 * config/rs6000/rs6000.c (rs6000_special_round_type_align): Same.
20647 2021-03-10 Vladimir N. Makarov <vmakarov@redhat.com>
20650 * lra-constraints.c (process_address_1): Don't check unknown
20651 constraint, use X for empty constraint.
20653 2021-03-10 Alex Coplan <alex.coplan@arm.com>
20655 * config/aarch64/aarch64.c (aarch64_vfp_is_call_or_return_candidate):
20656 Fix typo in comment describing "is_ha" argument.
20658 2021-03-10 John David Anglin <danglin@gcc.gnu.org>
20660 * doc/sourcebuild.texi: Document LRA target selector.
20662 2021-03-10 David Malcolm <dmalcolm@redhat.com>
20664 * doc/ux.texi: Add subsection contrasting interactive versus
20665 batch usage of GCC.
20667 2021-03-10 Joel Hutton <joel.hutton@arm.com>
20670 * tree-vect-stmts.c (vectorizable_store): Fix scatter store mask
20672 (vectorizable_load): Fix gather load mask check condition.
20674 2021-03-10 Richard Biener <rguenther@suse.de>
20676 PR tree-optimization/99510
20677 * tree.c (check_aligned_type): Check that the candidate
20678 has TYPE_USER_ALIGN set instead of matching with the
20681 2021-03-10 Eric Botcazou <ebotcazou@adacore.com>
20683 * config/sparc/sparc.c (sparc_regmode_natural_size): Return 4 for
20684 float and vector integer modes only if the mode is not larger.
20686 2021-03-10 Hans-Peter Nilsson <hp@axis.com>
20688 * config/cris/cris.h (DWARF_FRAME_REGISTERS): Define.
20690 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
20692 * ira.c (ira_setup_alts, ira_get_dup_out_num): Process digital
20694 * ira-lives.c (single_reg_class): Ditto.
20696 2021-03-09 Sebastian Huber <sebastian.huber@embedded-brains.de>
20698 * config.gcc (aarch64-*-rtems*): Include general rtems.h after
20699 the architecture-specific rtems.h.
20700 (aarch64-*-rtems*): Likewise.
20701 (arm*-*-rtems*): Likewise.
20702 (epiphany-*-rtems*): Likewise.
20703 (riscv*-*-rtems*): Likewise.
20705 2021-03-09 Jakub Jelinek <jakub@redhat.com>
20707 PR tree-optimization/99305
20708 * tree-ssa-phiopt.c (conditional_replacement): Test integer_pow2p
20709 before integer_all_onesp instead of vice versa.
20711 2021-03-09 Richard Earnshaw <rearnsha@arm.com>
20713 * common/config/arm/arm-common.c (arm_config_default): Change type
20714 of 'i' to unsigned.
20716 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
20719 * lra-constraints.c (process_address_1): Process constraint 'g'
20720 separately and digital constraints containing more one digit.
20722 2021-03-09 Nick Clifton <nickc@redhat.com>
20724 * config/rx/rx.h (DBX_DEBUGGING_INFO): Define.
20725 (DWARF"_DEBUGGING_INFO): Define.
20727 2021-03-09 Eric Botcazou <ebotcazou@adacore.com>
20730 * calls.c (initialize_argument_information): When the argument
20731 is passed by reference, do not make a copy in a thunk only if
20732 the argument is already in memory. Remove redundant test for
20733 the case of callee copy.
20735 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
20738 * lra-constraints.c (process_address_1): Process 0..9 constraints
20739 in process_address_1.
20741 2021-03-09 Andreas Krebbel <krebbel@linux.ibm.com>
20743 * config/s390/s390.c (struct s390_processor processor_table):
20744 Binutils name string must not be empty.
20746 2021-03-09 Claudiu Zissulescu <claziss@synopsys.com>
20748 * config/arc/arc.c (arc_attr_type): Remove function.
20750 2021-03-09 Martin Liska <mliska@suse.cz>
20753 * config/i386/i386-options.c (ix86_option_override_internal):
20754 Set isa_flags for OPTS argument and not for the global
20757 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
20759 * config/rs6000/predicates.md (ds_form_mem_operand): Check
20762 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
20765 * config/rs6000/predicates.md (ds_form_mem_operand) New
20767 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10) Use
20768 ds_form_mem_operand in ld/lwa patterns.
20769 * config/rs6000/fusion.md: Regenerate file.
20771 2021-03-08 Martin Sebor <msebor@redhat.com>
20773 PR middle-end/98266
20774 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): New function.
20775 (array_bounds_checker::check_array_bounds): Call it.
20777 2021-03-08 Martin Sebor <msebor@redhat.com>
20779 PR middle-end/97631
20780 * tree-ssa-strlen.c (maybe_warn_overflow): Test rawmem.
20781 (handle_builtin_stxncpy_strncat): Rename locals. Determine
20782 destination size from allocation calls. Issue a more appropriate
20784 (handle_builtin_memcpy): Pass true as rawmem to maybe_warn_overflow.
20785 (handle_builtin_memset): Same.
20787 2021-03-08 Peter Bergner <bergner@linux.ibm.com>
20790 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_permute): Add an assert
20791 to ensure we do not have an Altivec style address.
20792 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): Disable if passed
20793 an Altivec style address.
20794 (*vsx_le_perm_store_<mode>): Likewise.
20795 (splitters after *vsx_le_perm_store_<mode>): Likewise.
20796 (vsx_load_<mode>): Disable special expander if passed an Altivec
20798 (vsx_store_<mode>): Likewise.
20800 2021-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20803 * config/aarch64/predicates.md (aarch64_simd_shift_imm_vec_qi): Define.
20804 (aarch64_simd_shift_imm_vec_hi): Likewise.
20805 (aarch64_simd_shift_imm_vec_si): Likewise.
20806 (aarch64_simd_shift_imm_vec_di): Likewise.
20807 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le): Use
20808 predicate from above.
20809 (aarch64_shrn<mode>_insn_be): Likewise.
20810 (aarch64_rshrn<mode>_insn_le): Likewise.
20811 (aarch64_rshrn<mode>_insn_be): Likewise.
20812 (aarch64_shrn2<mode>_insn_le): Likewise.
20813 (aarch64_shrn2<mode>_insn_be): Likewise.
20814 (aarch64_rshrn2<mode>_insn_le): Likewise.
20815 (aarch64_rshrn2<mode>_insn_be): Likewise.
20817 2021-03-08 Vladimir N. Makarov <vmakarov@redhat.com>
20820 * lra-constraints.c (skip_contraint_modifiers): New function.
20821 (process_address_1): Use it before lookup_constraint call.
20823 2021-03-08 Martin Liska <mliska@suse.cz>
20826 * config/i386/i386-options.c (ix86_option_override_internal):
20827 Enable UINTR and HRESET for -march that supports it.
20829 2021-03-08 Ilya Leoshkevich <iii@linux.ibm.com>
20831 * config/s390/s390.c (f_constraint_p): New function.
20832 (s390_md_asm_adjust): Implement TARGET_MD_ASM_ADJUST.
20833 (TARGET_MD_ASM_ADJUST): Likewise.
20835 2021-03-08 Tobias Burnus <tobias@codesourcery.com>
20838 * tree-nested.c (convert_local_reference_stmt): Avoid calling
20839 lookup_field_for_decl for Fortran module (= namespace context).
20841 2021-03-08 Andreas Krebbel <krebbel@linux.ibm.com>
20843 * config/s390/s390.c (s390_expand_vec_compare): Implement <0
20844 comparison with arithmetic right shift.
20845 (s390_expand_vcond): No need for a force_reg anymore.
20846 s390_vec_compare will do it.
20847 * config/s390/vector.md ("vec_cmp<mode><tointvec>"): Accept also
20848 immediate operands.
20850 2021-03-07 Jakub Jelinek <jakub@redhat.com>
20853 * config/i386/constraints.md (Yw): Use SSE_REGS if TARGET_SSE
20854 but TARGET_AVX512BW or TARGET_AVX512VL is not set. Adjust description
20856 * config/i386/sse.md (v_Yw): New define_mode_attr.
20857 (*<insn><mode>3, *mul<mode>3<mask_name>, *avx2_<code><mode>3,
20858 *sse4_1_<code><mode>3<mask_name>): Use <v_Yw> instead of v
20860 * config/i386/mmx.md (mmx_pshufw_1, *vec_dupv4hi): Use Yw instead of
20861 xYw in constraints.
20863 2021-03-06 Julian Brown <julian@codesourcery.com>
20865 * tree-pretty-print.c (dump_generic_node): Emit non-generic
20866 address space info for aggregates.
20868 2021-03-06 Hans-Peter Nilsson <hp@axis.com>
20870 * config/cris/cris.h (MAX_FIXED_MODE_SIZE): Don't define.
20872 2021-03-05 Jakub Jelinek <jakub@redhat.com>
20874 PR middle-end/99322
20875 * tree-cfg.c (bb_to_omp_idx): New variable.
20876 (execute_build_cfg): Release the bb_to_omp_idx vector after
20877 cleanup_tree_cfg returns.
20878 (handle_abnormal_edges): Remove bb_to_omp_idx argument, adjust
20879 for bb_to_omp_idx being a vec<int> instead of pointer to array
20881 (make_edges): Remove bb_to_omp_idx local variable, don't pass
20882 it to handle_abnormal_edges, adjust for bb_to_omp_idx being a
20883 vec<int> instead of pointer to array of ints and don't free/release
20885 (remove_bb): When removing a bb and placing forced label somewhere
20886 else, ensure it is put into the same OpenMP region during cfg
20887 pass if possible or to entry successor as fallback. Unregister
20888 bb from bb_to_omp_idx.
20890 2021-03-05 Vladimir N. Makarov <vmakarov@redhat.com>
20893 * lra-constraints.c (process_address_1): Skip decomposing address
20894 for asm insn operand with unknown constraint.
20896 2021-03-05 Martin Jambor <mjambor@suse.cz>
20899 * cgraph.c (cgraph_edge::set_call_stmt): Do not update all
20900 corresponding speculative edges if we are about to resolve
20901 sepculation. Make edge direct (and so resolve speculations) before
20902 removing it from call_site_hash.
20903 (cgraph_edge::make_direct): Relax the initial assert to allow calling
20904 the function on speculative direct edges.
20906 2021-03-05 Eric Botcazou <ebotcazou@adacore.com>
20908 PR rtl-optimization/99376
20909 * rtlanal.c (nonzero_bits1) <arithmetic operators>: If the number
20910 of low-order zero bits is too large, set the result to 0 directly.
20912 2021-03-04 Jakub Jelinek <jakub@redhat.com>
20914 PR middle-end/93235
20915 * expmed.c (store_bit_field_using_insv): Return false of xop0 is a
20916 SUBREG and a SUBREG to op_mode can't be created.
20918 2021-03-04 Alex Coplan <alex.coplan@arm.com>
20921 * config/aarch64/aarch64-sve-builtins.cc
20922 (function_resolver::require_vector_type): Handle error_mark_node.
20924 2021-03-04 Ilya Leoshkevich <iii@linux.ibm.com>
20926 * cfgexpand.c (expand_asm_loc): Pass new parameter.
20927 (expand_asm_stmt): Likewise.
20928 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Add new
20930 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
20931 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
20932 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
20933 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
20934 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
20935 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
20936 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
20937 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
20938 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
20939 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
20940 * doc/tm.texi (md_asm_adjust): Likewise.
20941 * target.def (md_asm_adjust): Likewise.
20943 2021-03-04 Richard Biener <rguenther@suse.de>
20945 PR middle-end/97855
20946 * tree-pretty-print.c: Poison pp_printf.
20947 (dump_decl_name): Avoid use of pp_printf.
20948 (dump_block_node): Likewise.
20949 (dump_generic_node): Likewise.
20951 2021-03-04 Martin Sebor <msebor@redhat.com>
20953 PR middle-end/96963
20954 PR middle-end/94655
20955 * builtins.c (handle_array_ref): New helper.
20956 (handle_mem_ref): New helper.
20957 (compute_objsize_r): Factor out ARRAY_REF and MEM_REF handling
20958 into new helper functions. Correct a workaround for vectorized
20961 2021-03-03 Pat Haugen <pthaugen@linux.ibm.com>
20963 * config/rs6000/dfp.md (extendddtd2, trunctddd2, *cmp<mode>_internal1,
20964 floatditd2, ftrunc<mode>2, fix<mode>di2, dfp_ddedpd_<mode>,
20965 dfp_denbcd_<mode>, dfp_dxex_<mode>, dfp_diex_<mode>,
20966 *dfp_sgnfcnc_<mode>, dfp_dscli_<mode>, dfp_dscri_<mode>): Update size
20967 attribute for Power10.
20968 * config/rs6000/mma.md (*movoo): Likewise.
20969 * config/rs6000/rs6000.md (define_attr "size"): Add 256.
20970 (define_mode_attr bits): Add DD/TD modes.
20971 * config/rs6000/sync.md (load_quadpti, store_quadpti, load_lockedpti,
20972 store_conditionalpti): Update size attribute for Power10.
20974 2021-03-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
20977 * config/sparc/t-sparc (tree-ssanames.o-warn): Don't error for
20978 -Wuninitialized, -Wmaybe-uninitialized.
20979 (wide-int.o-warn): Likewise.
20981 2021-03-03 Richard Earnshaw <rearnsha@arm.com>
20983 * common/config/arm/arm-common.c: Include configargs.h.
20984 (arm_config_default): New function.
20985 (arm_target_mode): Renamed from arm_target_thumb_only. Handle
20986 processors that do not support Thumb. Take into account the
20987 --with-mode configuration setting for selecting the default.
20988 * config/arm/arm.h (OPTION_DEFAULT_SPECS): Remove entry for 'mode'.
20989 (TARGET_MODE_SPEC_FUNCTIONS): Update for function name change.
20991 2021-03-03 Martin Liska <mliska@suse.cz>
20993 PR gcov-profile/97461
20994 * gcov-io.h (GCOV_PREALLOCATED_KVP): Remove.
20996 2021-03-03 Eric Botcazou <ebotcazou@adacore.com>
20999 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
21000 point back the hard frame pointer to its default location when the
21001 frame is larger than SEH_MAX_FRAME_SIZE.
21003 2021-03-03 Jakub Jelinek <jakub@redhat.com>
21006 * config/i386/predicates.md (logic_operator): New define_predicate.
21007 * config/i386/i386.md (mov + mem using comm arith peephole2):
21008 Punt if operands[1] is EXT_REX_SSE_REGNO_P, AVX512BW is not enabled
21009 and the inner mode is [QH]Imode.
21011 2021-03-03 Jakub Jelinek <jakub@redhat.com>
21014 * dwarf2out.c (dw_loc_list_struct): Add end_entry member.
21015 (new_loc_list): Clear end_entry.
21016 (output_loc_list): Only use DW_LLE_startx_length for -gsplit-dwarf
21017 if HAVE_AS_LEB128, otherwise use DW_LLE_startx_endx. Fix comment
21019 (index_location_lists): For dwarf_version >= 5 without HAVE_AS_LEB128,
21020 initialize also end_entry.
21022 2021-03-03 Jakub Jelinek <jakub@redhat.com>
21025 * cfgrtl.c (fixup_partitions): When changing some bbs from hot to cold
21026 partitions, if in non-layout mode after reorder_blocks also move
21027 affected blocks to ensure a single partition transition.
21029 2021-03-03 Jason Merrill <jason@redhat.com>
21032 * cgraphunit.c (process_function_and_variable_attributes): Don't
21033 warn about flatten on an alias if the target also has it.
21034 * cgraph.h (symtab_node::get_alias_target_tree): New.
21036 2021-03-02 David Edelsohn <dje.gcc@gmail.com>
21038 * config/rs6000/rs6000.md (tls_get_tpointer_internal): Prepend
21039 period to symbol name.
21040 (tls_get_addr_internal<mode>): Same.
21042 2021-03-02 David Malcolm <dmalcolm@redhat.com>
21045 * diagnostic-show-locus.c
21046 (selftest::test_one_liner_many_fixits_2): Fix accidental usage of
21049 2021-03-02 Martin Sebor <msebor@redhat.com>
21051 PR middle-end/99276
21052 * builtins.c (warn_for_access): Remove stray warning text.
21054 2021-03-02 Martin Sebor <msebor@redhat.com>
21056 PR middle-end/99295
21057 * doc/extend.texi (attribute malloc): Reword and clarify nonaliasing
21060 2021-03-02 Jakub Jelinek <jakub@redhat.com>
21063 * dwarf2out.c (output_macinfo_op): Use DW_MACRO_*_str* even with
21064 -gdwarf-5 -gstrict-dwarf. For -gsplit-dwarf -gdwarf-5 use
21065 DW_MACRO_*_strx instead of DW_MACRO_*_strp. Handle
21066 DW_MACRO_define_strx and DW_MACRO_undef_strx.
21067 (save_macinfo_strings): Use DW_MACRO_*_str* even with
21068 -gdwarf-5 -gstrict-dwarf. Handle DW_MACRO_define_strx and
21069 DW_MACRO_undef_strx.
21071 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
21073 * config/s390/s390-builtin-types.def (BT_FN_V4SF_V8HI_UINT): New
21075 (BT_FN_V8HI_V8HI_UINT): Likewise.
21076 (BT_FN_V8HI_V4SF_V4SF_UINT): Likewise.
21077 * config/s390/s390-builtins.def (B_NNPA): New macro definition.
21078 (s390_vclfnhs, s390_vclfnls, s390_vcrnfs, s390_vcfn, s390_vcnf):
21079 New builtin definitions.
21080 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Bump
21081 vector extension version.
21082 * config/s390/s390.c (s390_expand_builtin): Check if builtins are
21083 available with current -march level.
21084 * config/s390/s390.md (UNSPEC_NNPA_VCLFNHS_V8HI)
21085 (UNSPEC_NNPA_VCLFNLS_V8HI, UNSPEC_NNPA_VCRNFS_V8HI)
21086 (UNSPEC_NNPA_VCFN_V8HI, UNSPEC_NNPA_VCNF_V8HI): New constants.
21087 * config/s390/vecintrin.h (vec_extend_to_fp32_hi): New macro.
21088 (vec_extend_to_fp32_lo): Likewise.
21089 (vec_round_from_fp32): Likewise.
21090 (vec_convert_to_fp16): Likewise.
21091 (vec_convert_from_fp16): Likewise.
21092 * config/s390/vx-builtins.md (vclfnhs_v8hi): New insn pattern.
21093 (vclfnls_v8hi): Likewise.
21094 (vcrnfs_v8hi): Likewise.
21095 (vcfn_v8hi): Likewise.
21096 (vcnf_v8hi): Likewise.
21098 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
21100 * common/config/s390/s390-common.c (processor_flags_table): New entry.
21101 * config.gcc: Enable arch14 for --with-arch and --with-tune.
21102 * config/s390/driver-native.c (s390_host_detect_local_cpu): Pick
21103 arch14 for unknown CPU models.
21104 * config/s390/s390-opts.h (enum processor_type): Add PROCESSOR_ARCH14.
21105 * config/s390/s390.c (s390_issue_rate): Add case for PROCESSOR_ARCH14.
21106 (s390_get_sched_attrmask): Likewise.
21107 (s390_get_unit_mask): Likewise.
21108 * config/s390/s390.h (enum processor_flags): Add PF_NNPA and PF_ARCH14.
21109 (TARGET_CPU_ARCH14, TARGET_CPU_ARCH14_P, TARGET_CPU_NNPA)
21110 (TARGET_CPU_NNPA_P, TARGET_ARCH14, TARGET_ARCH14_P, TARGET_NNPA)
21111 (TARGET_NNPA_P): New macro definitions.
21112 * config/s390/s390.md ("cpu_facility", "enabled"): Add arch14 and nnpa.
21113 * config/s390/s390.opt: Add PROCESSOR_ARCH14.
21115 2021-03-02 Jakub Jelinek <jakub@redhat.com>
21117 PR middle-end/95757
21118 * tree-vrp.c (register_edge_assert_for): Remove superfluous ()s around
21119 condition. Call register_edge_assert_for_1 for == 0, != 0, == 1 and
21120 != 1 comparisons if name is lhs of a comparison.
21122 2021-03-01 Iain Sandoe <iain@sandoe.co.uk>
21126 * config/darwin-protos.h (darwin_should_restore_cfa_state): New.
21127 * config/darwin.c (darwin_should_restore_cfa_state): New.
21128 * config/darwin.h (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New.
21129 * doc/tm.texi: Regenerated.
21130 * doc/tm.texi.in: Document TARGET_ASM_SHOULD_RESTORE_CFA_STATE.
21131 * dwarf2cfi.c (connect_traces): If the target requests, restore
21132 the CFA expression after a DW_CFA_restore.
21133 * target.def (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New hook.
21135 2021-03-01 Martin Liska <mliska@suse.cz>
21138 * optc-save-gen.awk: Add 4 more exceptions.
21140 2021-03-01 Nathan Sidwell <nathan@acm.org>
21143 * tree.h (TYPE_ALIGN_RAW): New accessor.
21144 (TYPE_ALIGN): Use it.
21146 2021-03-01 Jan Hubicka <jh@suse.cz>
21149 * ipa-fnsummary.c (compute_fn_summary): Fix sanity check.
21151 2021-03-01 Eric Botcazou <ebotcazou@adacore.com>
21154 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
21155 point the hard frame pointer to the SSE register save area instead
21156 of the general register save area. Perform only minimal adjustment
21157 for small frames if it is initially not correctly aligned.
21158 (ix86_expand_prologue): Remove early saves for a SEH target.
21159 * config/i386/winnt.c (struct seh_frame_state): Document constraint.
21161 2021-02-28 Jakub Jelinek <jakub@redhat.com>
21164 * ipa.c (symbol_table::remove_unreachable_nodes): Fix a comment
21165 typo - referneced -> referenced.
21166 * tree.c (component_ref_size): Fix comment typo -
21167 refernce -> reference.
21168 * tree-ssa-alias.c (access_path_may_continue_p): Fix comment typo -
21169 traling -> trailing.
21170 (aliasing_component_refs_p): Fix comment typos -
21171 refernce -> reference and refernece -> reference and
21172 traling -> trailing.
21173 (nonoverlapping_refs_since_match_p): Fix comment typo -
21174 referneces -> references.
21175 * doc/invoke.texi (--param modref-max-bases): Fix a typo -
21176 referneces -> references.
21178 2021-02-27 Iain Sandoe <iain@sandoe.co.uk>
21180 * config/host-darwin.c (darwin_gt_pch_use_address): Modify
21181 diagnostic message to avoid use of a contraction and format
21184 2021-02-27 Jakub Jelinek <jakub@redhat.com>
21187 * gcse.c (gcse_or_cprop_is_too_expensive): Use %wu instead of
21188 HOST_WIDE_INT_PRINT_UNSIGNED in warning format string.
21189 * ipa-devirt.c (ipa_odr_read_section): Use %wd instead of
21190 HOST_WIDE_INT_PRINT_DEC in inform format string. Fix comment
21193 2021-02-26 Richard Biener <rguenther@suse.de>
21195 PR middle-end/99281
21196 * expr.c (store_field): For calls with return-slot optimization
21197 and addressable return type expand the store directly.
21199 2021-02-26 Richard Biener <rguenther@suse.de>
21202 * builtins.c (warn_string_no_nul): Fix diagnostic formatting.
21204 2021-02-26 Peter Bergner <bergner@linux.ibm.com>
21207 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Replace assert
21210 2021-02-26 Aaron Sawdey <acsawdey@linux.ibm.com>
21212 * config.gcc: Add rs6000-pcrel-opt.o.
21213 * config/rs6000/rs6000-pcrel-opt.c: New file.
21214 * config/rs6000/pcrel-opt.md: New file.
21215 * config/rs6000/predicates.md: Add d_form_memory predicate.
21216 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_PCREL_OPT.
21217 * config/rs6000/rs6000-passes.def: Add pass_pcrel_opt.
21218 * config/rs6000/rs6000-protos.h: Add reg_to_non_prefixed(),
21219 pcrel_opt_valid_mem_p(), output_pcrel_opt_reloc(),
21220 and make_pass_pcrel_opt().
21221 * config/rs6000/rs6000.c (reg_to_non_prefixed): Make global.
21222 (rs6000_option_override_internal): Add pcrel-opt.
21223 (rs6000_delegitimize_address): Support pcrel-opt.
21224 (rs6000_opt_masks): Add pcrel-opt.
21225 (pcrel_opt_valid_mem_p): New function.
21226 (reg_to_non_prefixed): Make global.
21227 (rs6000_asm_output_opcode): Reset prepend_p_to_next_insn.
21228 (output_pcrel_opt_reloc): New function.
21229 * config/rs6000/rs6000.md (loads_extern_addr): New attr.
21230 (pcrel_extern_addr): Set loads_extern_addr.
21231 Add include for pcrel-opt.md.
21232 * config/rs6000/rs6000.opt: Add -mpcrel-opt.
21233 * config/rs6000/t-rs6000: Add rules for pcrel-opt.c and
21236 2021-02-26 YunQiang Su <yunqiang.su@cipunited.com>
21239 * config/mips/mips.c (mips_expand_ext_as_unaligned_load):
21240 If TARGET_64BIT and dest is SUBREG, we check the width, if it
21241 equal to SImode, we use SImode operation, just like what we are
21244 2021-02-26 Marek Polacek <polacek@redhat.com>
21246 * builtins.c (warn_for_access): Fix typos.
21248 2021-02-25 Iain Sandoe <iain@sandoe.co.uk>
21250 * config/aarch64/aarch64.md (<optab>_rol<mode>3): Add a '#'
21251 mark in front of the immediate quantity.
21252 (<optab>_rolsi3_uxtw): Likewise.
21254 2021-02-25 Richard Earnshaw <rearnsha@arm.com>
21257 * config/arm/thumb2.md (nonsecure_call_reg_thumb2_fpcxt): New pattern.
21258 (nonsecure_call_value_reg_thumb2_fpcxt): Likewise.
21259 (nonsecure_call_reg_thumb2): Restrict to using r4 for the callee
21260 address and disable when the FPCXT is not available.
21261 (nonsecure_call_value_reg_thumb2): Likewise.
21263 2021-02-25 Nathan Sidwell <nathan@acm.org>
21266 * doc/invoke.texi (flang-info-module-cmi): Renamed option.
21268 2021-02-25 Tamar Christina <tamar.christina@arm.com>
21270 * tree-vect-slp.c (optimize_load_redistribution_1): Abort on NULL nodes.
21272 2021-02-25 Richard Biener <rguenther@suse.de>
21274 PR tree-optimization/99253
21275 * tree-vect-loop.c (check_reduction_path): First compute
21276 code, then verify out-of-loop uses.
21278 2021-02-25 Jakub Jelinek <jakub@redhat.com>
21281 * match.pd ((T)(A) + CST -> (T)(A + CST)): Add :s to convert.
21283 2021-02-25 Jakub Jelinek <jakub@redhat.com>
21285 PR tree-optimization/80635
21286 * tree-vrp.c (vrp_simplify_cond_using_ranges): Also handle
21287 VIEW_CONVERT_EXPR if modes are the same, innerop is integral and
21288 has mode precision.
21290 2021-02-25 Richard Biener <rguenther@suse.de>
21292 * tree-vect-slp.c (optimize_load_redistribution_1): Delay
21293 load_map population.
21294 (vect_match_slp_patterns_2): Revert part of last change.
21295 (vect_analyze_slp): Do not interleave optimize_load_redistribution
21296 with pattern detection but do it afterwards. Dump the
21297 whole SLP graph after pattern recognition and load
21298 redistribution optimization finished.
21300 2021-02-24 Jakub Jelinek <jakub@redhat.com>
21303 * omp-low.c (struct omp_context): Add teams_nested_p and
21304 nonteams_nested_p members.
21305 (scan_omp_target): Diagnose teams nested inside of target with other
21306 directives strictly nested inside of the same target.
21307 (check_omp_nesting_restrictions): Set ctx->teams_nested_p or
21308 ctx->nonteams_nested_p as needed.
21310 2021-02-24 Vladimir N. Makarov <vmakarov@redhat.com>
21312 PR inline-asm/99123
21313 * lra-constraints.c (uses_hard_regs_p): Don't use decompose_mem_address.
21315 2021-02-24 Hans-Peter Nilsson <hp@axis.com>
21317 * config/cris/cris.c (cris_expand_prologue): Set
21318 current_function_static_stack_size, if flag_stack_usage_info.
21320 2021-02-24 Pat Haugen <pthaugen@linux.ibm.com>
21322 * config/rs6000/rs6000.c (next_insn_prefixed_p): Rename.
21323 (rs6000_final_prescan_insn): Adjust.
21324 (rs6000_asm_output_opcode): Likewise.
21326 2021-02-24 Martin Sebor <msebor@redhat.com>
21328 PR middle-end/97172
21329 * attribs.c (attr_access::free_lang_data): Clear attribute arg spec
21330 from function arguments.
21332 2021-02-24 Tamar Christina <tamar.christina@arm.com>
21334 PR tree-optimization/99220
21335 * tree-vect-slp.c (optimize_load_redistribution_1): Remove
21336 node from cache when it's about to be deleted.
21338 2021-02-24 Jakub Jelinek <jakub@redhat.com>
21340 PR tree-optimization/99225
21341 * fold-const.c (fold_binary_loc) <case NE_EXPR>: In (x & (1 << y)) != 0
21342 to ((x >> y) & 1) != 0 simplifications use build_one_cst instead of
21343 build_int_cst (..., 1). Formatting fixes.
21345 2021-02-24 Tamar Christina <tamar.christina@arm.com>
21347 PR tree-optimization/99149
21348 * tree-vect-slp-patterns.c (vect_detect_pair_op): Don't recreate the
21350 (vect_slp_reset_pattern): Remove.
21351 (complex_fma_pattern::matches): Remove call to vect_slp_reset_pattern.
21352 (complex_mul_pattern::build, complex_fma_pattern::build,
21353 complex_fms_pattern::build): Fix ref counts.
21354 * tree-vect-slp.c (vect_free_slp_tree): Undo SLP only pattern relevancy
21355 when node is being deleted.
21356 (vect_match_slp_patterns_2): Correct result of cache hit on patterns.
21357 (vect_schedule_slp): Invalidate SLP_TREE_REPRESENTATIVE of removed
21359 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize value.
21361 2021-02-24 Matthias Klose <doko@ubuntu.com>
21364 2020-12-07 Matthias Klose <doko@ubuntu.com>
21366 * genextract.c (print_header): Undefine ENABLE_RTL_CHECKING
21367 and ENABLE_RTL_FLAG_CHECKING.
21369 2021-02-24 Richard Biener <rguenther@suse.de>
21372 * builtins.c (fold_builtin_next_arg): Avoid NULL arg.
21374 2021-02-23 Peter Bergner <bergner@linux.ibm.com>
21376 * config/rs6000/mma.md (mma_assemble_pair): Rename from this...
21377 (vsx_assemble_pair): ...to this.
21378 (*mma_assemble_pair): Rename from this...
21379 (*vsx_assemble_pair): ...to this.
21380 (mma_disassemble_pair): Rename from this...
21381 (vsx_disassemble_pair): ...to this.
21382 (*mma_disassemble_pair): Rename from this...
21383 (*vsx_disassemble_pair): ...to this.
21384 * config/rs6000/rs6000-builtin.def (BU_MMA_V2, BU_MMA_V3,
21385 BU_COMPAT): New macros.
21386 (mma_assemble_pair): Rename from this...
21387 (vsx_assemble_pair): ...to this.
21388 (mma_disassemble_pair): Rename from this...
21389 (vsx_disassemble_pair): ...to this.
21390 (mma_assemble_pair): New compatibility built-in.
21391 (mma_disassemble_pair): Likewise.
21392 * config/rs6000/rs6000-call.c (struct builtin_compatibility): New.
21393 (RS6000_BUILTIN_COMPAT): Define.
21394 (bdesc_compat): New.
21395 (mma_expand_builtin): Use VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
21396 (rs6000_gimple_fold_mma_builtin): Use MMA_BUILTIN_DISASSEMBLE_PAIR
21397 and VSX_BUILTIN_ASSEMBLE_PAIR.
21398 (rs6000_init_builtins): Register compatibility built-ins.
21399 (mma_init_builtins): Use VSX_BUILTIN_ASSEMBLE_PAIR,
21400 VSX_BUILTIN_ASSEMBLE_PAIR_INTERNAL, VSX_BUILTIN_DISASSEMBLE_PAIR and
21401 VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
21402 * doc/extend.texi (__builtin_mma_assemble_pair): Rename from this...
21403 (__builtin_vsx_assemble_pair): ...to this.
21404 (__builtin_mma_disassemble_pair): Rename from this...
21405 (__builtin_vsx_disassemble_pair): ...to this.
21407 2021-02-23 Martin Liska <mliska@suse.cz>
21410 * ipa-icf.c (sem_variable::merge): Do not merge 2 variables
21411 with different alignment. That leads to an invalid red zone
21412 size allocated in runtime.
21414 2021-02-23 Jakub Jelinek <jakub@redhat.com>
21416 PR tree-optimization/99204
21417 * fold-const.c (fold_read_from_constant_string): Check that
21418 tree_fits_uhwi_p (index) rather than just that index is INTEGER_CST.
21420 2021-02-23 Segher Boessenkool <segher@kernel.crashing.org>
21421 Kewen Lin <linkw@gcc.gnu.org>
21423 * config/rs6000/rs6000.md (*rotl<mode>3_insert_3): Renamed to...
21424 (rotl<mode>3_insert_3): ...this.
21425 (plus_ior_xor): New code_iterator.
21426 (define_split for GPR rl*imi): New splitter.
21427 * config/rs6000/vsx.md (vsx_init_v4si): Use gen_rotldi3_insert_3
21428 for integer merging.
21430 2021-02-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21432 * config/aarch64/aarch64-tuning-flags.def (cse_sve_vl_constants):
21434 * config/aarch64/aarch64.md (add<mode>3): Force CONST_POLY_INT immediates
21435 into a register when the above is enabled.
21436 * config/aarch64/aarch64.c (neoversev1_tunings):
21437 AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
21438 (aarch64_rtx_costs): Use AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
21440 2021-02-22 Hans-Peter Nilsson <hp@axis.com>
21442 * config/cris/cris.c (cris_print_operand) <'T'>: Change
21443 valid operand from is now an addi mult-value to shift-value.
21444 * config/cris/cris.md (*addi): Change expression of scaled
21445 operand from mult to ashift.
21446 * config/cris/cris.md (*addi_reload): New insn_and_split.
21448 2021-02-22 John David Anglin <danglin@gcc.gnu.org>
21451 * config/pa/pa.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Define as
21452 hook_bool_const_tree_hwi_hwi_const_tree_true.
21453 (pa_asm_output_mi_thunk): Add support for nonzero vcall_offset.
21455 2021-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
21457 PR rtl-optimization/98791
21458 * ira-conflicts.c (process_regs_for_copy): Don't create allocno copies
21459 for unordered modes.
21461 2021-02-22 Martin Liska <mliska@suse.cz>
21463 * tree-inline.c (inline_forbidden_p): Set
21464 inline_forbidden_reason.
21466 2021-02-22 Richard Biener <rguenther@suse.de>
21468 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Dump
21471 2021-02-22 Richard Biener <rguenther@suse.de>
21473 PR tree-optimization/99165
21474 * gimple-ssa-store-merging.c (pass_store_merging::process_store):
21475 Accumulate changed to ret.
21477 2021-02-21 Uros Bizjak <ubizjak@gmail.com>
21480 2020-12-09 Uroš Bizjak <ubizjak@gmail.com>
21482 * config/i386/i386.h (REG_ALLOC_ORDER): Remove
21484 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
21487 * config/s390/vector.md (trunctf<DFP_ALL:mode>2_vr): New
21489 (trunctf<DFP_ALL:mode>2): Likewise.
21490 (trunctdtf2_vr): Likewise.
21491 (trunctdtf2): Likewise.
21492 (extend<DFP_ALL:mode>tf2_vr): Likewise.
21493 (extend<DFP_ALL:mode>tf2): Likewise.
21494 (extendtftd2_vr): Likewise.
21495 (extendtftd2): Likewise.
21497 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
21499 * config/s390/vector.md (*fprx2_to_tf): Rename to fprx2_to_tf,
21500 add memory alternative.
21501 (tf_to_fprx2): New pattern.
21503 2021-02-19 Martin Sebor <msebor@redhat.com>
21506 * attribs.c (init_attr_rdwr_indices): Guard vblist use.
21507 (attr_access::free_lang_data): Remove a spurious test.
21509 2021-02-19 Nathan Sidwell <nathan@acm.org>
21511 * doc/invoke.texi (flang-info-module-read): Document.
21513 2021-02-19 Martin Liska <mliska@suse.cz>
21515 PR translation/99167
21516 * params.opt: Fix typo.
21518 2021-02-19 Richard Biener <rguenther@suse.de>
21520 PR middle-end/99122
21521 * tree-inline.c (inline_forbidden_p): Do not inline functions
21522 with VLA arguments or return value.
21524 2021-02-19 Jakub Jelinek <jakub@redhat.com>
21527 * config/arm/arm.md (*stack_protect_combined_set_insn,
21528 *stack_protect_combined_test_insn): If force_const_mem result
21529 is not valid general operand, force its address into the destination
21532 2021-02-19 Jakub Jelinek <jakub@redhat.com>
21535 * tree-cfg.c (gimple_merge_blocks): If bb a starts with eh landing
21536 pad or non-local label, put FORCED_LABELs from bb b after that label
21537 rather than before it.
21539 2021-02-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
21542 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3): Use
21543 expand_vector_broadcast' to emit the vec_duplicate operand.
21545 2021-02-18 Vladimir N. Makarov <vmakarov@redhat.com>
21547 PR rtl-optimization/96264
21548 * lra-remat.c (reg_overlap_for_remat_p): Check also output insn
21551 2021-02-18 H.J. Lu <hjl.tools@gmail.com>
21554 * varasm.c (get_section): Replace SUPPORTS_SHF_GNU_RETAIN with
21555 looking up the retain attribute.
21556 (resolve_unique_section): Likewise.
21557 (get_variable_section): Likewise.
21558 (switch_to_section): Likewise. Warn when a symbol without the
21559 retain attribute and a symbol with the retain attribute are
21560 placed in the section with the same name, instead of the used
21562 * doc/extend.texi: Document the "retain" attribute.
21564 2021-02-18 Nathan Sidwell <nathan@acm.org>
21567 * doc/invoke.texi (flang-info-include-translate): Document header
21570 2021-02-18 Richard Biener <rguenther@suse.de>
21572 PR middle-end/99122
21573 * ipa-fnsummary.c (analyze_function_body): Set
21574 CIF_FUNCTION_NOT_INLINABLE for VLA parameter calls.
21575 * tree-inline.c (insert_init_debug_bind): Pass NULL for
21576 error_mark_node values.
21577 (force_value_to_type): Do not build V_C_Es for WITH_SIZE_EXPR
21579 (setup_one_parameter): Delay force_value_to_type until when
21582 2021-02-18 Hans-Peter Nilsson <hp@axis.com>
21584 PR tree-optimization/99142
21585 * match.pd (clz cmp 0): Gate replacement on single_use of clz result.
21587 2021-02-18 Jakub Jelinek <jakub@redhat.com>
21589 * wide-int-bitmask.h (wide_int_bitmask::wide_int_bitmask (),
21590 wide_int_bitmask::wide_int_bitmask (uint64_t),
21591 wide_int_bitmask::wide_int_bitmask (uint64_t, uint64_t),
21592 wide_int_bitmask::operator ~ () const,
21593 wide_int_bitmask::operator | (wide_int_bitmask) const,
21594 wide_int_bitmask::operator & (wide_int_bitmask) const): Use constexpr
21596 * config/i386/i386.h (PTA_3DNOW, PTA_3DNOW_A, PTA_64BIT, PTA_ABM,
21597 PTA_AES, PTA_AVX, PTA_BMI, PTA_CX16, PTA_F16C, PTA_FMA, PTA_FMA4,
21598 PTA_FSGSBASE, PTA_LWP, PTA_LZCNT, PTA_MMX, PTA_MOVBE, PTA_NO_SAHF,
21599 PTA_PCLMUL, PTA_POPCNT, PTA_PREFETCH_SSE, PTA_RDRND, PTA_SSE, PTA_SSE2,
21600 PTA_SSE3, PTA_SSE4_1, PTA_SSE4_2, PTA_SSE4A, PTA_SSSE3, PTA_TBM,
21601 PTA_XOP, PTA_AVX2, PTA_BMI2, PTA_RTM, PTA_HLE, PTA_PRFCHW, PTA_RDSEED,
21602 PTA_ADX, PTA_FXSR, PTA_XSAVE, PTA_XSAVEOPT, PTA_AVX512F, PTA_AVX512ER,
21603 PTA_AVX512PF, PTA_AVX512CD, PTA_NO_TUNE, PTA_SHA, PTA_PREFETCHWT1,
21604 PTA_CLFLUSHOPT, PTA_XSAVEC, PTA_XSAVES, PTA_AVX512DQ, PTA_AVX512BW,
21605 PTA_AVX512VL, PTA_AVX512IFMA, PTA_AVX512VBMI, PTA_CLWB, PTA_MWAITX,
21606 PTA_CLZERO, PTA_NO_80387, PTA_PKU, PTA_AVX5124VNNIW, PTA_AVX5124FMAPS,
21607 PTA_AVX512VPOPCNTDQ, PTA_SGX, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES,
21608 PTA_AVX512VBMI2, PTA_VPCLMULQDQ, PTA_AVX512BITALG, PTA_RDPID,
21609 PTA_PCONFIG, PTA_WBNOINVD, PTA_AVX512VP2INTERSECT, PTA_PTWRITE,
21610 PTA_AVX512BF16, PTA_WAITPKG, PTA_MOVDIRI, PTA_MOVDIR64B, PTA_ENQCMD,
21611 PTA_CLDEMOTE, PTA_SERIALIZE, PTA_TSXLDTRK, PTA_AMX_TILE, PTA_AMX_INT8,
21612 PTA_AMX_BF16, PTA_UINTR, PTA_HRESET, PTA_KL, PTA_WIDEKL, PTA_AVXVNNI,
21613 PTA_X86_64_BASELINE, PTA_X86_64_V2, PTA_X86_64_V3, PTA_X86_64_V4,
21614 PTA_CORE2, PTA_NEHALEM, PTA_WESTMERE, PTA_SANDYBRIDGE, PTA_IVYBRIDGE,
21615 PTA_HASWELL, PTA_BROADWELL, PTA_SKYLAKE, PTA_SKYLAKE_AVX512,
21616 PTA_CASCADELAKE, PTA_COOPERLAKE, PTA_CANNONLAKE, PTA_ICELAKE_CLIENT,
21617 PTA_ICELAKE_SERVER, PTA_TIGERLAKE, PTA_SAPPHIRERAPIDS, PTA_ALDERLAKE,
21618 PTA_KNL, PTA_BONNELL, PTA_SILVERMONT, PTA_GOLDMONT, PTA_GOLDMONT_PLUS,
21619 PTA_TREMONT, PTA_KNM): Use constexpr instead of const.
21621 2021-02-18 Jakub Jelinek <jakub@redhat.com>
21623 PR middle-end/99109
21624 * gimple-array-bounds.cc (build_zero_elt_array_type): Rename to ...
21625 (build_printable_array_type): ... this. Add nelts argument. For
21626 overaligned eltype, use TYPE_MAIN_VARIANT (eltype) instead. If
21627 nelts, call build_array_type_nelts.
21628 (array_bounds_checker::check_mem_ref): Use build_printable_array_type
21629 instead of build_zero_elt_array_type and build_array_type_nelts.
21631 2021-02-18 Jakub Jelinek <jakub@redhat.com>
21634 * config/i386/i386.c (distance_non_agu_define): Don't call
21635 extract_insn_cached here.
21636 (ix86_lea_outperforms): Save and restore recog_data around call
21637 to distance_non_agu_define and distance_agu_use.
21638 (ix86_ok_to_clobber_flags): Remove.
21639 (ix86_avoid_lea_for_add): Don't call ix86_ok_to_clobber_flags.
21640 (ix86_avoid_lea_for_addr): Likewise. Adjust function comment.
21641 * config/i386/i386.md (*lea<mode>): Change from define_insn_and_split
21642 into define_insn. Move the splitting to define_peephole2 and
21643 check there using peep2_regno_dead_p if FLAGS_REG is dead.
21645 2021-02-17 Julian Brown <julian@codesourcery.com>
21647 * gimplify.c (gimplify_scan_omp_clauses): Handle ATTACH_DETACH
21650 2021-02-17 Xi Ruoyao <xry111@mengyan1223.wang>
21653 * config/mips/mips.c (mips_symbol_insns): Do not use
21654 MSA_SUPPORTED_MODE_P if mode is MAX_MACHINE_MODE.
21656 2021-02-16 Vladimir N. Makarov <vmakarov@redhat.com>
21658 PR inline-asm/98096
21659 * stmt.c (resolve_operand_name_1): Take inout operands into account
21660 for access to labels by names.
21661 * doc/extend.texi: Describe counting operands for accessing labels.
21663 2021-02-16 Richard Biener <rguenther@suse.de>
21665 PR tree-optimization/38474
21666 * tree-ssa-structalias.c (variable_info::address_taken): New.
21667 (new_var_info): Initialize address_taken.
21668 (process_constraint): Set address_taken.
21669 (solve_constraints): Use the new address_taken flag rather
21670 than is_reg_var for sorting variables.
21671 (dump_constraint): Dump the variable number if the name
21674 2021-02-16 Jakub Jelinek <jakub@redhat.com>
21677 * tree-vect-stmts.c (vectorizable_simd_clone_call): For num_calls != 1
21678 multiply by 4096 and for inbranch by 8192.
21679 * config/i386/i386.c (ix86_simd_clone_usable): For TARGET_AVX512F,
21680 return 3, 2 or 1 for mangle letters 'b', 'c' or 'd'.
21682 2021-02-15 Maya Rashish <coypu@sdf.org>
21684 * config/aarch64/aarch64.c (aarch64_init_builtins):
21685 Call SUBTARGET_INIT_BUILTINS.
21687 2021-02-15 Peter Bergner <bergner@linux.ibm.com>
21689 PR rtl-optimization/98872
21690 * init-regs.c (initialize_uninitialized_regs): Skip initialization
21691 if CONST0_RTX is NULL.
21693 2021-02-15 Richard Sandiford <richard.sandiford@arm.com>
21695 PR rtl-optimization/98863
21696 * rtl-ssa/functions.h (function_info::bb_live_out_info): Delete.
21697 (function_info::build_info): Turn into a declaration, moving the
21698 definition to internals.h.
21699 (function_info::bb_walker): Declare.
21700 (function_info::create_reg_use): Likewise.
21701 (function_info::calculate_potential_phi_regs): Take a build_info
21703 (function_info::place_phis, function_info::create_ebbs): Declare.
21704 (function_info::calculate_ebb_live_in_for_debug): Likewise.
21705 (function_info::populate_backedge_phis): Delete.
21706 (function_info::start_block, function_info::end_block): Declare.
21707 (function_info::populate_phi_inputs): Delete.
21708 (function_info::m_potential_phi_regs): Move information to build_info.
21709 * rtl-ssa/internals.h: New file.
21710 (function_info::bb_phi_info): New class.
21711 (function_info::build_info): Moved from functions.h.
21712 Add a constructor and destructor.
21713 (function_info::build_info::ebb_use): Delete.
21714 (function_info::build_info::ebb_def): Likewise.
21715 (function_info::build_info::bb_live_out): Likewise.
21716 (function_info::build_info::tmp_ebb_live_in_for_debug): New variable.
21717 (function_info::build_info::potential_phi_regs): Likewise.
21718 (function_info::build_info::potential_phi_regs_for_debug): Likewise.
21719 (function_info::build_info::ebb_def_regs): Likewise.
21720 (function_info::build_info::bb_phis): Likewise.
21721 (function_info::build_info::bb_mem_live_out): Likewise.
21722 (function_info::build_info::bb_to_rpo): Likewise.
21723 (function_info::build_info::def_stack): Likewise.
21724 (function_info::build_info::old_def_stack_limit): Likewise.
21725 * rtl-ssa/internals.inl (function_info::build_info::record_reg_def):
21726 Remove the regno argument. Push the previous definition onto the
21727 definition stack where necessary.
21728 * rtl-ssa/accesses.cc: Include internals.h.
21729 * rtl-ssa/changes.cc: Likewise.
21730 * rtl-ssa/blocks.cc: Likewise.
21731 (function_info::build_info::build_info): Define.
21732 (function_info::build_info::~build_info): Likewise.
21733 (function_info::bb_walker): New class.
21734 (function_info::bb_walker::bb_walker): Define.
21735 (function_info::add_live_out_use): Convert a logarithmic-complexity
21736 test into a linear one. Allow the same definition to be passed
21738 (function_info::calculate_potential_phi_regs): Moved from
21739 functions.cc. Take a build_info parameter and store the
21740 information there instead.
21741 (function_info::place_phis): New function.
21742 (function_info::add_entry_block_defs): Update call to record_reg_def.
21743 (function_info::calculate_ebb_live_in_for_debug): New function.
21744 (function_info::add_phi_nodes): Use bb_phis to decide which
21745 registers need phi nodes and initialize ebb_def_regs accordingly.
21746 Do not add degenerate phis here.
21747 (function_info::add_artificial_accesses): Use create_reg_use.
21748 Assert that all definitions are listed in the DF LR sets.
21749 Update call to record_reg_def.
21750 (function_info::record_block_live_out): Record live-out register
21751 values in the phis of successor blocks. Use the live-out set
21752 when processing the last block in an EBB, instead of always
21753 using the live-in sets of successor blocks. AND the live sets
21754 with the set of registers that have been defined in the EBB,
21755 rather than with all potential phi registers. Cope correctly
21756 with branches back to the start of the current EBB.
21757 (function_info::start_block): New function.
21758 (function_info::end_block): Likewise.
21759 (function_info::populate_phi_inputs): Likewise.
21760 (function_info::create_ebbs): Likewise.
21761 (function_info::process_all_blocks): Rewrite into a multi-phase
21763 * rtl-ssa/functions.cc: Include internals.h.
21764 (function_info::calculate_potential_phi_regs): Move to blocks.cc.
21765 (function_info::init_function_data): Remove caller.
21766 * rtl-ssa/insns.cc: Include internals.h
21767 (function_info::create_reg_use): New function. Lazily any
21768 degenerate phis needed by the linear RPO view.
21769 (function_info::record_use): Use create_reg_use. When processing
21770 debug uses, use potential_phi_regs and test it before checking
21771 whether the register is live on entry to the current EBB. Lazily
21772 calculate ebb_live_in_for_debug.
21773 (function_info::record_call_clobbers): Update call to record_reg_def.
21774 (function_info::record_def): Likewise.
21776 2021-02-15 Martin Liska <mliska@suse.cz>
21778 * toplev.c (init_asm_output): Free output of
21779 gen_command_line_string function.
21780 (process_options): Likewise.
21782 2021-02-15 Martin Liska <mliska@suse.cz>
21784 * params.opt: Add 2 missing Param keywords.
21786 2021-02-15 Eric Botcazou <ebotcazou@adacore.com>
21788 * df-core.c (df_worklist_dataflow_doublequeue): Use proper cast.
21790 2021-02-15 Jakub Jelinek <jakub@redhat.com>
21792 PR tree-optimization/99079
21793 * match.pd (A % (pow2pcst << N) -> A & ((pow2pcst << N) - 1)): Remove
21794 useless tree_nop_conversion_p (type, TREE_TYPE (@3)) check. Instead
21795 require both type and TREE_TYPE (@1) to be integral types and either
21796 type having smaller or equal precision, or TREE_TYPE (@1) being
21797 unsigned type, or type being signed type. If TREE_TYPE (@1)
21798 doesn't have wrapping overflow, perform the subtraction of one in
21801 2021-02-14 Jan Hubicka <hubicka@ucw.cz>
21802 Richard Biener <rguether@suse.de>
21805 * ipa-reference.c (ipa_init): Only conditinally initialize
21806 reference_vars_to_consider.
21807 (propagate): Conditionally deninitialize reference_vars_to_consider.
21808 (ipa_reference_write_optimization_summary): Sanity check that
21809 reference_vars_to_consider is not allocated.
21811 2021-02-13 Levy Hsu <admin@levyhsu.com>
21814 * config/riscv/riscv-shorten-memrefs.c (pass_shorten_memrefs): Add
21815 extend parameter to get_si_mem_base_reg declaration.
21816 (get_si_mem_base_reg): Add extend parameter. Set it.
21817 (analyze): Pass extend arg to get_si_mem_base_reg.
21818 (transform): Likewise. Use it when rewriting mems.
21819 * config/riscv/riscv.c (riscv_legitimize_move): Check for subword
21820 loads and emit sign/zero extending load followed by subreg move.
21822 2021-02-13 Jim Wilson <jimw@sifive.com>
21825 * config/riscv/riscv.c (riscv_compressed_lw_address_p): Drop early
21826 exit when !reload_completed. Only perform check for compressed reg
21827 if reload_completed.
21828 (riscv_rtx_costs): In MEM case, when optimizing for size and
21829 shorten memrefs, if not compressible, then increase cost.
21831 2021-02-13 Jakub Jelinek <jakub@redhat.com>
21833 PR rtl-optimization/98439
21834 * recog.c (pass_split_before_regstack::gate): Enable even when
21835 pass_split_before_sched2 is enabled if -fselective-scheduling2 is
21838 2021-02-13 Jakub Jelinek <jakub@redhat.com>
21841 * config/i386/mmx.md (*mmx_pshufd_1): Add a combine splitter for
21842 swap of V2SImode elements in memory into DImode memory rotate by 32.
21844 2021-02-12 Martin Sebor <msebor@redhat.com>
21846 * tree-pretty-print.c (print_generic_expr_to_str): Update comment.
21848 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
21850 * rtl-ssa/accesses.cc (function_info::make_use_available): Use
21851 m_temp_obstack rather than m_obstack to allocate the temporary use.
21853 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
21855 * df-problems.c (df_lr_bb_local_compute): Treat partial definitions
21856 as read-modify operations.
21858 2021-02-12 Richard Biener <rguenther@suse.de>
21860 PR middle-end/38474
21861 * ipa-fnsummary.c (unmodified_parm_1): Only walk when
21862 fbi->aa_walk_budget is bigger than zero. Update
21863 fbi->aa_walk_budget.
21864 (param_change_prob): Likewise.
21865 * ipa-prop.c (detect_type_change_from_memory_writes):
21866 Properly account walk_aliased_vdefs.
21867 (parm_preserved_before_stmt_p): Canonicalize updates.
21868 (parm_ref_data_preserved_p): Likewise.
21869 (parm_ref_data_pass_through_p): Likewise.
21870 (determine_known_aggregate_parts): Account own alias queries.
21872 2021-02-12 Martin Liska <mliska@suse.cz>
21874 * opts-common.c (decode_cmdline_option): Release werror_arg.
21875 * opts.c (gen_producer_string): Release output of
21876 gen_command_line_string.
21878 2021-02-12 Richard Biener <rguenther@suse.de>
21880 PR tree-optimization/38474
21881 * params.opt (-param=max-store-chains-to-track=): New param.
21882 (-param=max-stores-to-track=): Likewise.
21883 * doc/invoke.texi (max-store-chains-to-track): Document.
21884 (max-stores-to-track): Likewise.
21885 * gimple-ssa-store-merging.c (pass_store_merging::m_n_chains):
21887 (pass_store_merging::m_n_stores): Likewise.
21888 (pass_store_merging::terminate_and_process_chain): Update
21889 m_n_stores and m_n_chains.
21890 (pass_store_merging::process_store): Likewise. Terminate
21891 oldest chains if the number of stores or chains get too large.
21892 (imm_store_chain_info::terminate_and_process_chain): Dump
21895 2021-02-11 Eric Botcazou <ebotcazou@adacore.com>
21897 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
21898 the cold section, emit a nop before the directive if the previous
21899 active instruction can throw.
21901 2021-02-11 Peter Bergner <bergner@linux.ibm.com>
21904 * config/rs6000/predicates.md (mma_assemble_input_operand): Restrict
21905 memory addresses that are legal for quad word accesses.
21907 2021-02-11 Andrea Corallo <andrea.corallo@arm.com>
21910 * config/arm/thumb2.md (*doloop_end_internal): Generate
21911 alternative sequence to handle long range branches.
21913 2021-02-11 Joel Hutton <joel.hutton@arm.com>
21915 PR tree-optimization/98772
21916 * optabs-tree.c (supportable_half_widening_operation): New function
21917 to check for supportable V8QI->V8HI widening patterns.
21918 * optabs-tree.h (supportable_half_widening_operation): New function.
21919 * tree-vect-stmts.c (vect_create_half_widening_stmts): New function
21920 to create promotion stmts for V8QI->V8HI widening patterns.
21921 (vectorizable_conversion): Add case for V8QI->V8HI.
21923 2021-02-11 Richard Biener <rguenther@suse.de>
21925 * sparseset.h (SPARSESET_ELT_BITS): Remove.
21926 (SPARSESET_ELT_TYPE): Use unsigned int.
21927 * fwprop.c: Do not include sparseset.h.
21929 2021-02-10 Jakub Jelinek <jakub@redhat.com>
21932 * varasm.c (declare_weak): For -fsyntax-only, allow even
21933 TREE_ASM_WRITTEN function decls.
21935 2021-02-10 Jakub Jelinek <jakub@redhat.com>
21938 * config/i386/sse.md (fix<fixunssuffix>_truncv2sfv2di2,
21939 <insn>v8qiv8hi2, <insn>v8qiv8si2, <insn>v4qiv4si2, <insn>v4hiv4si2,
21940 <insn>v8qiv8di2, <insn>v4qiv4di2, <insn>v2qiv2di2, <insn>v4hiv4di2,
21941 <insn>v2hiv2di2, <insn>v2siv2di2): Force operands[1] into REG before
21942 calling simplify_gen_subreg on it.
21944 2021-02-10 Martin Liska <mliska@suse.cz>
21946 * config/nvptx/nvptx.c (nvptx_option_override): Use
21947 flag_patchable_function_entry instead of the removed
21948 function_entry_patch_area_size.
21950 2021-02-10 Martin Liska <mliska@suse.cz>
21952 PR tree-optimization/99002
21953 PR tree-optimization/99026
21954 * gimple-if-to-switch.cc (if_chain::is_beneficial): Fix memory
21955 leak when adjacent cases are merged.
21956 * tree-switch-conversion.c (switch_decision_tree::analyze_switch_statement): Use
21958 (make_pass_lower_switch): Remove trailing whitespace.
21959 * tree-switch-conversion.h (release_clusters): New.
21961 2021-02-10 Richard Biener <rguenther@suse.de>
21963 PR rtl-optimization/99054
21964 * cfgrtl.c (rtl-optimization/99054): Return an auto_vec.
21965 (fixup_partitions): Adjust.
21966 (rtl_verify_edges): Likewise.
21968 2021-02-10 Jakub Jelinek <jakub@redhat.com>
21970 PR middle-end/99007
21971 * gimplify.c (gimplify_scan_omp_clauses): For MEM_REF on reductions,
21972 temporarily disable gimplify_ctxp->into_ssa around gimplify_expr
21975 2021-02-10 Richard Biener <rguenther@suse.de>
21978 * ipa-pure-const.c (propagate_malloc): Use an auto_vec<>
21981 2021-02-10 Richard Biener <rguenther@suse.de>
21983 PR tree-optimization/99024
21984 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Only
21985 clear loop->aux if it is associated with the destroyed loop_vinfo.
21987 2021-02-10 Martin Liska <mliska@suse.cz>
21989 PR tree-optimization/99002
21990 * gimple-if-to-switch.cc (find_conditions): Fix memory leak
21993 2021-02-10 Martin Liska <mliska@suse.cz>
21996 * ipa-icf.c (sem_item::add_reference): Fix memory leak when
21997 a reference exists.
21999 2021-02-10 Jakub Jelinek <jakub@redhat.com>
22002 * dwarf2out.c (prune_unused_types_walk): Mark DW_TAG_variable DIEs
22003 at class scope for DWARF5+.
22005 2021-02-09 Eric Botcazou <ebotcazou@adacore.com>
22007 PR rtl-optimization/96015
22008 * reorg.c (skip_consecutive_labels): Minor comment tweaks.
22009 (relax_delay_slots): When deleting a jump to the next active
22010 instruction over a barrier, first delete the barrier if the
22011 jump is the only way to reach the target label.
22013 2021-02-09 Andre Vieira <andre.simoesdiasvieira@arm.com>
22015 * config/aarch64/aarch64-cost-tables.h: Add entries for vect.mul.
22016 * config/aarch64/aarch64.c (aarch64_rtx_mult_cost): Use vect.mul for
22017 vector multiplies and vect.alu for SSRA.
22018 * config/arm/aarch-common-protos.h (struct vector_cost_table): Define
22019 vect.mul cost field.
22020 * config/arm/aarch-cost-tables.h: Add entries for vect.mul.
22021 * config/arm/arm.c: Likewise.
22023 2021-02-09 Richard Biener <rguenther@suse.de>
22025 PR tree-optimization/98863
22026 * tree-ssa-sccvn.h (vn_avail::next_undo): Add.
22027 * tree-ssa-sccvn.c (last_pushed_avail): New global.
22028 (rpo_elim::eliminate_push_avail): Chain pushed avails.
22029 (unwind_state::avail_top): Add.
22030 (do_unwind): Rewrite unwinding of avail entries.
22031 (do_rpo_vn): Initialize last_pushed_avail and
22032 avail_top of the undo state.
22034 2021-02-09 Jakub Jelinek <jakub@redhat.com>
22036 PR middle-end/99004
22037 * calls.c (maybe_warn_rdwr_sizes): Change s0 and s1 type from
22038 const char * to char * and free those pointers after use.
22040 2021-02-09 Richard Biener <rguenther@suse.de>
22042 PR tree-optimization/99017
22043 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Allow
22044 zero vector cost entries.
22046 2021-02-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
22048 PR middle-end/98974
22049 * tree-vect-stmts.c (vectorizable_condition): Remove shadow vec_num
22050 parameter in vectorizable_condition.
22052 2021-02-08 Richard Biener <rguenther@suse.de>
22055 * tree.c (walk_tree_1): Walk VECTOR_CST elements.
22057 2021-02-08 Martin Liska <mliska@suse.cz>
22060 * cfgexpand.c (pass_expand::execute): Parse per-function option
22061 flag_patchable_function_entry and use it.
22062 * common.opt: Remove function_entry_patch_area_size and
22063 function_entry_patch_area_start global variables.
22064 * opts.c (parse_and_check_patch_area): New function.
22065 (common_handle_option): Use it.
22066 * opts.h (parse_and_check_patch_area): New function.
22067 * toplev.c (process_options): Parse and use
22068 function_entry_patch_area_size.
22070 2021-02-08 Martin Sebor <msebor@redhat.com>
22072 * doc/extend.texi (attribute malloc): Correct typos.
22074 2021-02-05 Nathan Sidwell <nathan@acm.org>
22077 * gcc.c (driver::maybe_run_linker): Check for input file
22078 accessibility if not linking.
22080 2021-02-05 Richard Biener <rguenther@suse.de>
22082 PR tree-optimization/98855
22083 * tree-vectorizer.h (add_stmt_cost): New overload.
22084 * tree-vect-slp.c (li_cost_vec_cmp): New.
22085 (vect_bb_slp_scalar_cost): Cost individual loop regions
22086 separately. Account for the scalar instance root stmt.
22088 2021-02-05 Tom de Vries <tdevries@suse.de>
22091 * tree-switch-conversion.c (jump_table_cluster::emit): Add loc
22093 (bit_test_cluster::emit): Reuse location_t for newly created
22095 (switch_decision_tree::try_switch_expansion): Preserve
22097 * tree-switch-conversion.h: Change function signatures.
22099 2021-02-05 Jakub Jelinek <jakub@redhat.com>
22102 * config/i386/i386-options.c (m_NONE, m_ALL): Define.
22103 * config/i386/x86-tune.def (X86_TUNE_BRANCH_PREDICTION_HINTS,
22104 X86_TUNE_PROMOTE_QI_REGS): Use m_NONE instead of 0U.
22105 (X86_TUNE_QIMODE_MATH): Use m_ALL instead of ~0U.
22107 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22109 * config/aarch64/aarch64-simd-builtins.def (get_high): Define builtin.
22110 * config/aarch64/aarch64-simd.md (aarch64_get_high<mode>): Define.
22111 * config/aarch64/arm_neon.h (__GET_HIGH): Delete.
22112 (vget_high_f16): Reimplement using new builtin.
22113 (vget_high_f32): Likewise.
22114 (vget_high_f64): Likewise.
22115 (vget_high_p8): Likewise.
22116 (vget_high_p16): Likewise.
22117 (vget_high_p64): Likewise.
22118 (vget_high_s8): Likewise.
22119 (vget_high_s16): Likewise.
22120 (vget_high_s32): Likewise.
22121 (vget_high_s64): Likewise.
22122 (vget_high_u8): Likewise.
22123 (vget_high_u16): Likewise.
22124 (vget_high_u32): Likewise.
22125 (vget_high_u64): Likewise.
22127 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22129 * config/aarch64/aarch64-simd-builtins.def (get_low): Define builtin.
22130 * config/aarch64/aarch64-simd.md (aarch64_get_low<mode>): Define.
22131 * config/aarch64/arm_neon.h (__GET_LOW): Delete.
22132 (vget_low_f16): Reimplement using new builtin.
22133 (vget_low_f32): Likewise.
22134 (vget_low_f64): Likewise.
22135 (vget_low_p8): Likewise.
22136 (vget_low_p16): Likewise.
22137 (vget_low_p64): Likewise.
22138 (vget_low_s8): Likewise.
22139 (vget_low_s16): Likewise.
22140 (vget_low_s32): Likewise.
22141 (vget_low_s64): Likewise.
22142 (vget_low_u8): Likewise.
22143 (vget_low_u16): Likewise.
22144 (vget_low_u32): Likewise.
22145 (vget_low_u64): Likewise.
22147 2021-02-05 Kito Cheng <kito.cheng@sifive.com>
22149 * gcc.c (print_multilib_info): Check all required argument is provided
22152 2021-02-05 liuhongt <hongtao.liu@intel.com>
22155 * config/i386/i386-expand.c (ix86_expand_sse_cmp): Don't
22156 generate integer mask comparison for 128/256-bits vector when
22157 op_true/op_false is NULL_RTX or CONSTM1_RTX/CONST0_RTX. Also
22158 delete redundant !maskcmp condition.
22159 (ix86_expand_int_vec_cmp): Ditto but no redundant deletion
22161 (ix86_expand_sse_movcc): Delete definition of maskcmp, add the
22162 condition directly to if (maskcmp), add extra check for
22163 cmpmode, it should be MODE_INT.
22164 (ix86_expand_fp_vec_cmp): Pass NULL to ix86_expand_sse_cmp's
22165 parameters op_true/op_false.
22166 (ix86_use_mask_cmp_p): New.
22168 2021-02-05 liuhongt <hongtao.liu@intel.com>
22171 * config/i386/x86-tune.def (X86_TUNE_AVX256_UNALIGNED_LOAD_OPTIMAL):
22172 Remove m_GENERIC from ~list.
22173 (X86_TUNE_AVX256_UNALIGNED_STORE_OPTIMAL): Ditto.
22175 2021-02-04 David Malcolm <dmalcolm@redhat.com>
22178 * diagnostic-show-locus.c (compatible_locations_p): Require
22179 locations in the same macro map to be either both from the
22180 macro definition, or both from the macro arguments.
22182 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
22184 * config/aarch64/aarch64-simd-builtins.def: Add
22185 [su]mull_hi_lane[q] builtin generator macros.
22186 * config/aarch64/aarch64-simd.md
22187 (aarch64_<su>mull_hi_lane<mode>_insn): Define.
22188 (aarch64_<su>mull_hi_lane<mode>): Define.
22189 (aarch64_<su>mull_hi_laneq<mode>_insn): Define.
22190 (aarch64_<su>mull_hi_laneq<mode>): Define.
22191 * config/aarch64/arm_neon.h (vmull_high_lane_s16): Use RTL
22192 builtin instead of inline asm.
22193 (vmull_high_lane_s32): Likewise.
22194 (vmull_high_lane_u16): Likewise.
22195 (vmull_high_lane_u32): Likewise.
22196 (vmull_high_laneq_s16): Likewise.
22197 (vmull_high_laneq_s32): Likewise.
22198 (vmull_high_laneq_u16): Likewise.
22199 (vmull_high_laneq_u32): Liekwise.
22201 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
22203 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_hi_n
22204 builtin generator macros.
22205 * config/aarch64/aarch64-simd.md
22206 (aarch64_<su>mull_hi_n<mode>_insn): Define.
22207 (aarch64_<su>mull_hi_n<mode>): Define.
22208 * config/aarch64/arm_neon.h (vmull_high_n_s16): Use RTL builtin
22209 instead of inline asm.
22210 (vmull_high_n_s32): Likewise.
22211 (vmull_high_n_u16): Likewise.
22212 (vmull_high_n_u32): Likewise.
22214 2021-02-04 Richard Biener <rguenther@suse.de>
22216 PR tree-optimization/98855
22217 * tree-vect-loop.c (vectorizable_phi): Do not cost
22218 single-argument PHIs.
22219 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Likewise.
22220 * tree-vect-stmts.c (vectorizable_bswap): Also perform
22221 costing for SLP operation.
22223 2021-02-04 Martin Liska <mliska@suse.cz>
22225 * doc/extend.texi: Mention -mprefer-vector-width in target
22228 2021-02-03 Martin Sebor <msebor@redhat.com>
22230 PR tree-optimization/98937
22231 * tree-ssa-strlen.c (strlen_dom_walker::~strlen_dom_walker): Define.
22232 Flush pointer_query cache.
22234 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
22236 * config/rs6000/genfusion.pl (gen_2logical): Add missing
22237 fixes based on patch review.
22238 * config/rs6000/fusion.md: Regenerate file.
22240 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
22242 * config/rs6000/t-rs6000: Comment out auto generation of
22245 2021-02-03 Andrew Stubbs <ams@codesourcery.com>
22247 * config/gcn/gcn-opts.h (enum processor_type): Add PROCESSOR_GFX908.
22248 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Add gfx908.
22249 (output_file_start): Add gfx908.
22250 * config/gcn/gcn.opt (gpu_type): Add gfx908.
22251 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Add march=gfx908.
22252 (MULTILIB_DIRNAMES): Add gfx908.
22253 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): New define.
22254 (main): Recognize gfx908.
22255 * config/gcn/t-omp-device: Add gfx908.
22257 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22259 * config/aarch64/aarch64-simd-builtins.def: Add
22260 [su]mlsl_hi_lane[q] builtin macro generators.
22261 * config/aarch64/aarch64-simd.md
22262 (aarch64_<su>mlsl_hi_lane<mode>_insn): Define.
22263 (aarch64_<su>mlsl_hi_lane<mode>): Define.
22264 (aarch64_<su>mlsl_hi_laneq<mode>_insn): Define.
22265 (aarch64_<su>mlsl_hi_laneq<mode>): Define.
22266 * config/aarch64/arm_neon.h (vmlsl_high_lane_s16): Use RTL
22267 builtin instead of inline asm.
22268 (vmlsl_high_lane_s32): Likewise.
22269 (vmlsl_high_lane_u16): Likewise.
22270 (vmlsl_high_lane_u32): Likewise.
22271 (vmlsl_high_laneq_s16): Likewise.
22272 (vmlsl_high_laneq_s32): Likewise.
22273 (vmlsl_high_laneq_u16): Likewise.
22274 (vmlsl_high_laneq_u32): Likewise.
22275 (vmlal_high_laneq_u32): Likewise.
22277 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22279 * config/aarch64/aarch64-simd-builtins.def: Add
22280 [su]mlal_hi_lane[q] builtin generator macros.
22281 * config/aarch64/aarch64-simd.md
22282 (aarch64_<su>mlal_hi_lane<mode>_insn): Define.
22283 (aarch64_<su>mlal_hi_lane<mode>): Define.
22284 (aarch64_<su>mlal_hi_laneq<mode>_insn): Define.
22285 (aarch64_<su>mlal_hi_laneq<mode>): Define.
22286 * config/aarch64/arm_neon.h (vmlal_high_lane_s16): Use RTL
22287 builtin instead of inline asm.
22288 (vmlal_high_lane_s32): Likewise.
22289 (vmlal_high_lane_u16): Likewise.
22290 (vmlal_high_lane_u32): Likewise.
22291 (vmlal_high_laneq_s16): Likewise.
22292 (vmlal_high_laneq_s32): Likewise.
22293 (vmlal_high_laneq_u16): Likewise.
22294 (vmlal_high_laneq_u32): Likewise.
22296 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22298 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_hi_n
22299 builtin generator macros.
22300 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_hi_n<mode>_insn):
22302 (aarch64_<su>mlsl_hi_n<mode>): Define.
22303 * config/aarch64/arm_neon.h (vmlsl_high_n_s16): Use RTL builtin
22304 instead of inline asm.
22305 (vmlsl_high_n_s32): Likewise.
22306 (vmlsl_high_n_u16): Likewise.
22307 (vmlsl_high_n_u32): Likewise.
22309 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22311 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_hi_n
22312 builtin generator macros.
22313 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_hi_n<mode>_insn):
22315 (aarch64_<su>mlal_hi_n<mode>): Define.
22316 * config/aarch64/arm_neon.h (vmlal_high_n_s16): Use RTL builtin
22317 instead of inline asm.
22318 (vmlal_high_n_s32): Likewise.
22319 (vmlal_high_n_u16): Likewise.
22320 (vmlal_high_n_u32): Likewise.
22322 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22324 * config/aarch64/aarch64-simd-builtins.def: Add RTL builtin
22326 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal_hi<mode>):
22328 (aarch64_<su>mlal_hi<mode>_insn): This.
22329 (aarch64_<su>mlal_hi<mode>): Define.
22330 * config/aarch64/arm_neon.h (vmlal_high_s8): Use RTL builtin
22331 instead of inline asm.
22332 (vmlal_high_s16): Likewise.
22333 (vmlal_high_s32): Likewise.
22334 (vmlal_high_u8): Likewise.
22335 (vmlal_high_u16): Likewise.
22336 (vmlal_high_u32): Likewise.
22338 2021-02-03 Ilya Leoshkevich <iii@linux.ibm.com>
22340 * lra-spills.c (remove_pseudos): Call lra_update_insn_recog_data()
22341 after calling alter_subreg() on a (mem).
22343 2021-02-03 Martin Liska <mliska@suse.cz>
22346 * lto-streamer-out.c (produce_lto_section): Fill up missing
22348 * lto-streamer.h (struct lto_section): Add _padding field.
22350 2021-02-03 Richard Biener <rguenther@suse.de>
22352 * lto-streamer.c (lto_get_section_name): Free temporary
22354 * tree-loop-distribution.c
22355 (loop_distribution::merge_dep_scc_partitions): Free edge data.
22357 2021-02-03 Jakub Jelinek <jakub@redhat.com>
22359 PR middle-end/97487
22360 * ifcvt.c (noce_can_force_operand): New function.
22361 (noce_emit_move_insn): Use it.
22362 (noce_try_sign_mask): Likewise. Formatting fix.
22364 2021-02-03 Jakub Jelinek <jakub@redhat.com>
22366 PR middle-end/97971
22367 * lra-constraints.c (process_alt_operands): For inline asm, don't call
22368 fatal_insn, but instead return false.
22370 2021-02-03 Jakub Jelinek <jakub@redhat.com>
22372 PR tree-optimization/98287
22373 * config/i386/mmx.md (<insn><mode>3): For shifts don't enable expander
22376 2021-02-03 Tamar Christina <tamar.christina@arm.com>
22378 PR tree-optimization/98928
22379 * tree-vect-loop.c (vect_analyze_loop_2): Change
22380 STMT_VINFO_SLP_VECT_ONLY to STMT_VINFO_SLP_VECT_ONLY_PATTERN.
22381 * tree-vect-slp-patterns.c (complex_pattern::build): Likewise.
22382 * tree-vectorizer.h (STMT_VINFO_SLP_VECT_ONLY_PATTERN): New.
22383 (class _stmt_vec_info): Add slp_vect_pattern_only_p.
22385 2021-02-02 Richard Biener <rguenther@suse.de>
22387 * gimple-loop-interchange.cc (prepare_data_references):
22389 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
22390 * tree-ssa-loop-im.c (hoist_memory_references): Likewise.
22391 * tree-vect-stmts.c (vectorizable_condition): Do not
22393 (vectorizable_comparison): Likewise.
22395 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22397 * config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
22398 * config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
22399 * config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
22400 (vrsqrteq_u32): Likewise.
22402 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22404 * config/aarch64/aarch64-simd-builtins.def (sqxtun2): Define builtin.
22405 * config/aarch64/aarch64-simd.md (aarch64_sqxtun2<mode>_le): Define.
22406 (aarch64_sqxtun2<mode>_be): Likewise.
22407 (aarch64_sqxtun2<mode>): Likewise.
22408 * config/aarch64/arm_neon.h (vqmovun_high_s16): Reimplement using builtin.
22409 (vqmovun_high_s32): Likewise.
22410 (vqmovun_high_s64): Likewise.
22411 * config/aarch64/iterators.md (UNSPEC_SQXTUN2): Define.
22413 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22415 * config/aarch64/aarch64-simd-builtins.def (bfdot_lane, bfdot_laneq): Use
22417 (bfmlalb_lane, bfmlalt_lane, bfmlalb_lane_q, bfmlalt_lane_q): Use FP flags.
22419 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22421 * config/aarch64/aarch64-simd-builtins.def (fcmla_lane0, fcmla_lane90,
22422 fcmla_lane180, fcmla_lane270, fcmlaq_lane0, fcmlaq_lane90, fcmlaq_lane180,
22423 fcmlaq_lane270, scvtf, ucvtf, fcvtzs, fcvtzu, scvtfsi, scvtfdi, ucvtfsi,
22424 ucvtfdi, fcvtzshf, fcvtzuhf, fmlal_lane_low, fmlsl_lane_low,
22425 fmlal_laneq_low, fmlsl_laneq_low, fmlalq_lane_low, fmlslq_lane_low,
22426 fmlalq_laneq_low, fmlslq_laneq_low, fmlal_lane_high, fmlsl_lane_high,
22427 fmlal_laneq_high, fmlsl_laneq_high, fmlalq_lane_high, fmlslq_lane_high,
22428 fmlalq_laneq_high, fmlslq_laneq_high): Use FP flags.
22430 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22432 * config/aarch64/aarch64-builtins.c (FLAG_LOAD): Define.
22433 * config/aarch64/aarch64-simd-builtins.def (ld1x2, ld2, ld3, ld4, ld2r,
22434 ld3r, ld4r, ld1, ld1x3, ld1x4): Use LOAD flags.
22436 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22438 * config/aarch64/aarch64-simd-builtins.def (combine, zip1, zip2,
22439 uzp1, uzp2, trn1, trn2, simd_bsl): Use AUTO_FP flags.
22441 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22443 * config/aarch64/aarch64-simd-builtins.def (clrsb, clz, ctz, popcount,
22444 vec_smult_lane_, vec_smlal_lane_, vec_smult_laneq_, vec_smlal_laneq_,
22445 vec_umult_lane_, vec_umlal_lane_, vec_umult_laneq_, vec_umlal_laneq_,
22446 ashl, sshl, ushl, srshl, urshl, sdot_lane, udot_lane, sdot_laneq,
22447 udot_laneq, usdot_lane, usdot_laneq, sudot_lane, sudot_laneq, ashr,
22448 ashr_simd, lshr, lshr_simd, srshr_n, urshr_n, ssra_n, usra_n, srsra_n,
22449 ursra_n, sshll_n, ushll_n, sshll2_n, ushll2_n, ssri_n, usri_n, ssli_n,
22450 ssli_n, usli_n, bswap, rbit, simd_bsl, eor3q, rax1q, xarq, bcaxq): Use
22451 NONE builtin flags.
22453 2021-02-02 Jakub Jelinek <jakub@redhat.com>
22455 PR tree-optimization/98848
22456 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
22457 STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
22459 2021-02-02 Kito Cheng <kito.cheng@sifive.com>
22462 * expr.c: Check mode before calling store_expr.
22464 2021-02-02 Christophe Lyon <christophe.lyon@linaro.org>
22466 * config/arm/iterators.md (supf): Remove VORNQ_S and VORNQ_U.
22468 * config/arm/mve.md (mve_vornq_s<mode>): New entry for vorn
22469 instruction using expression ior.
22470 (mve_vornq_u<mode>): New expander.
22471 (mve_vornq_f<mode>): Use ior code instead of unspec.
22472 * config/arm/unspecs.md (VORNQ_S, VORNQ_U, VORNQ_F): Remove.
22474 2021-02-02 Alexandre Oliva <oliva@adacore.com>
22476 * tree-nested.c (convert_nonlocal_reference_op): Move
22477 current_function_decl restore after re-gimplification.
22478 (convert_local_reference_op): Likewise.
22480 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22482 * config/aarch64/aarch64-simd-builtins.def (rshrn, rshrn2):
22484 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>_insn_le):
22486 (aarch64_rshrn<mode>_insn_be): Likewise.
22487 (aarch64_rshrn<mode>): Likewise.
22488 (aarch64_rshrn2<mode>_insn_le): Likewise.
22489 (aarch64_rshrn2<mode>_insn_be): Likewise.
22490 (aarch64_rshrn2<mode>): Likewise.
22491 * config/aarch64/aarch64.md (unspec): Add UNSPEC_RSHRN.
22492 * config/aarch64/arm_neon.h (vrshrn_high_n_s16): Reimplement
22494 (vrshrn_high_n_s32): Likewise.
22495 (vrshrn_high_n_s64): Likewise.
22496 (vrshrn_high_n_u16): Likewise.
22497 (vrshrn_high_n_u32): Likewise.
22498 (vrshrn_high_n_u64): Likewise.
22499 (vrshrn_n_s16): Likewise.
22500 (vrshrn_n_s32): Likewise.
22501 (vrshrn_n_s64): Likewise.
22502 (vrshrn_n_u16): Likewise.
22503 (vrshrn_n_u32): Likewise.
22504 (vrshrn_n_u64): Likewise.
22506 2021-02-01 Sergei Trofimovich <siarheit@google.com>
22508 PR tree-optimization/98499
22509 * ipa-modref.c (analyze_ssa_name_flags): treat RVO
22510 conservatively and assume all possible side-effects.
22512 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22514 * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi,
22515 vec_unpacku_hi_): Define builtins.
22516 * config/aarch64/arm_neon.h (vmovl_high_s8): Reimplement using
22518 (vmovl_high_s16): Likewise.
22519 (vmovl_high_s32): Likewise.
22520 (vmovl_high_u8): Likewise.
22521 (vmovl_high_u16): Likewise.
22522 (vmovl_high_u32): Likewise.
22524 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22526 * config/aarch64/aarch64-simd-builtins.def (sabdl, uabdl):
22528 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl<mode>): New
22530 * config/aarch64/aarch64.md (unspec): Define UNSPEC_SABDL,
22532 * config/aarch64/arm_neon.h (vabdl_s8): Reimplemet using
22534 (vabdl_s16): Likewise.
22535 (vabdl_s32): Likewise.
22536 (vabdl_u8): Likewise.
22537 (vabdl_u16): Likewise.
22538 (vabdl_u32): Likewise.
22539 * config/aarch64/iterators.md (ABDL): New int iterator.
22540 (sur): Handle UNSPEC_SABDL, UNSPEC_UABDL.
22542 2021-02-01 Martin Sebor <msebor@redhat.com>
22544 * tree.h (BLOCK_VARS): Add comment.
22545 (BLOCK_SUBBLOCKS): Same.
22546 (BLOCK_SUPERCONTEXT): Same.
22547 (BLOCK_ABSTRACT_ORIGIN): Same.
22548 (inlined_function_outer_scope_p): Same.
22550 2021-02-01 Martin Sebor <msebor@redhat.com>
22552 PR middle-end/97172
22553 * attribs.c (attr_access::free_lang_data): Define new function.
22554 * attribs.h (attr_access::free_lang_data): Declare new function.
22556 2021-02-01 Richard Biener <rguenther@suse.de>
22558 * vec.h (auto_vec::auto_vec): Add memory stat parameters
22560 * bitmap.h (auto_bitmap::auto_bitmap): Likewise.
22562 2021-02-01 Tamar Christina <tamar.christina@arm.com>
22564 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>,
22565 aarch64_<su>mlsl<mode>, aarch64_<su>mlsl_n<mode>): Flip mult operands.
22567 2021-02-01 Richard Biener <rguenther@suse.de>
22569 PR rtl-optimization/98863
22570 * config/i386/i386-features.c (convert_scalars_to_vector):
22571 Set DF_RD_PRUNE_DEAD_DEFS.
22573 2021-01-31 Eric Botcazou <ebotcazou@adacore.com>
22575 * system.h (SIZE_MAX): Define if not already defined.
22577 2021-01-30 Aaron Sawdey <acsawdey@linux.ibm.com>
22579 * config/rs6000/genfusion.pl (gen_2logical): New function to
22580 generate patterns for logical-logical fusion.
22581 * config/rs6000/fusion.md: Regenerated patterns.
22582 * config/rs6000/rs6000-cpus.def: Add
22583 OPTION_MASK_P10_FUSION_2LOGICAL.
22584 * config/rs6000/rs6000.c (rs6000_option_override_internal):
22585 Enable logical-logical fusion for p10.
22586 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2logical.
22588 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
22590 * config/rs6000/rs6000.opt: Add periods to new AIX options.
22592 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
22594 * config/rs6000/rs6000.opt (mabi=vec-extabi): New.
22595 (mabi=vec-default): New.
22596 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
22597 __EXTABI__ for AIX Vector extended ABI.
22598 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print AIX Vector
22600 (conditional_register_usage): If AIX vec_extabi enabled, vs20-vs31
22602 * doc/invoke.texi (PowerPC mabi): Add AIX vec-extabi and vec-default.
22604 2021-01-30 Jakub Jelinek <jakub@redhat.com>
22606 * config/i386/i386-features.c (remove_partial_avx_dependency): Clear
22607 DF_DEFER_INSN_RESCAN after calling df_process_deferred_rescans.
22609 2021-01-29 Vladimir N. Makarov <vmakarov@redhat.com>
22612 * lra-constraints.c (in_class_p): Don't narrow class only for REG
22615 2021-01-29 Will Schmidt <will_schmidt@vnet.ibm.com>
22617 * config/rs6000/rs6000-call.c (rs6000_expand_binup_builtin): Add
22618 clauses for CODE_FOR_vsx_xvcvuxddp_scale and
22619 CODE_FOR_vsx_xvcvsxddp_scale to the parameter checking code.
22621 2021-01-29 Andrew MacLeod <amacleod@redhat.com>
22623 PR tree-optimization/98866
22624 * gimple-range-gori.h (gori_compute:set_range_invariant): New.
22625 * gimple-range-gori.cc (gori_map::set_range_invariant): New.
22626 (gori_map::m_maybe_invariant): Rename from all_outgoing.
22627 (gori_map::gori_map): Rename all_outgoing to m_maybe_invariant.
22628 (gori_map::is_export_p): Ditto.
22629 (gori_map::calculate_gori): Ditto.
22630 (gori_compute::set_range_invariant): New.
22631 * gimple-range.cc (gimple_ranger::range_of_stmt): Set range
22632 invariant for pointers evaluating to [1, +INF].
22634 2021-01-29 Richard Biener <rguenther@suse.de>
22636 PR rtl-optimization/98863
22637 * config/i386/i386-features.c (remove_partial_avx_dependency):
22638 Do not perform DF analysis.
22639 (pass_data_remove_partial_avx_dependency): Remove
22642 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
22644 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_n
22645 builtin generator macros.
22646 * config/aarch64/aarch64-simd.md (aarch64_<su>mull_n<mode>):
22648 * config/aarch64/arm_neon.h (vmull_n_s16): Use RTL builtin
22649 instead of inline asm.
22650 (vmull_n_s32): Likewise.
22651 (vmull_n_u16): Likewise.
22652 (vmull_n_u32): Likewise.
22654 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22656 * config/aarch64/aarch64-simd-builtins.def (sabdl2, uabdl2):
22658 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
22660 (aarch64_<sur>abdl2<mode>): ... This.
22661 (<sur>sadv16qi): Adjust use of above.
22662 * config/aarch64/arm_neon.h (vabdl_high_s8): Reimplement using
22664 (vabdl_high_s16): Likewise.
22665 (vabdl_high_s32): Likewise.
22666 (vabdl_high_u8): Likewise.
22667 (vabdl_high_u16): Likewise.
22668 (vabdl_high_u32): Likewise.
22670 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22672 * config/aarch64/aarch64-simd-builtins.def (sabal2): Define
22674 (uabal2): Likewise.
22675 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal2<mode>): New
22677 * config/aarch64/aarch64.md (unspec): Add UNSPEC_SABAL2 and
22679 * config/aarch64/arm_neon.h (vabal_high_s8): Reimplement using
22681 (vabal_high_s16): Likewise.
22682 (vabal_high_s32): Likewise.
22683 (vabal_high_u8): Likewise.
22684 (vabal_high_u16): Likewise.
22685 (vabal_high_u32): Likewise.
22686 * config/aarch64/iterators.md (ABAL2): New mode iterator.
22687 (sur): Handle UNSPEC_SABAL2, UNSPEC_UABAL2.
22689 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22691 * config/aarch64/aarch64-simd-builtins.def (sabal): Define
22694 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal<mode>_4):
22696 (aarch64_<sur>abal<mode>): ... This
22697 (<sur>sadv16qi): Adust use of the above.
22698 * config/aarch64/arm_neon.h (vabal_s8): Reimplement using
22700 (vabal_s16): Likewise.
22701 (vabal_s32): Likewise.
22702 (vabal_u8): Likewise.
22703 (vabal_u16): Likewise.
22704 (vabal_u32): Likewise.
22706 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22708 * config/aarch64/aarch64-simd-builtins.def (saddlv, uaddlv):
22710 * config/aarch64/aarch64-simd.md (aarch64_<su>addlv<mode>):
22712 * config/aarch64/arm_neon.h (vaddlv_s8): Reimplement using
22714 (vaddlv_s16): Likewise.
22715 (vaddlv_u8): Likewise.
22716 (vaddlv_u16): Likewise.
22717 (vaddlvq_s8): Likewise.
22718 (vaddlvq_s16): Likewise.
22719 (vaddlvq_s32): Likewise.
22720 (vaddlvq_u8): Likewise.
22721 (vaddlvq_u16): Likewise.
22722 (vaddlvq_u32): Likewise.
22723 (vaddlv_s32): Likewise.
22724 (vaddlv_u32): Likewise.
22725 * config/aarch64/iterators.md (VDQV_L): New mode iterator.
22726 (unspec): Add UNSPEC_SADDLV, UNSPEC_UADDLV.
22727 (Vwstype): New mode attribute.
22729 (VWIDE_S): Likewise.
22730 (USADDLV): New int iterator.
22731 (su): Handle UNSPEC_SADDLV, UNSPEC_UADDLV.
22733 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
22735 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_lane[q]
22736 builtin generator macros.
22737 * config/aarch64/aarch64-simd.md (aarch64_vec_<su>mlsl_lane<Qlane>):
22739 * config/aarch64/arm_neon.h (vmlsl_lane_s16): Use RTL builtin
22740 instead of inline asm.
22741 (vmlsl_lane_s32): Likewise.
22742 (vmlsl_lane_u16): Likewise.
22743 (vmlsl_lane_u32): Likewise.
22744 (vmlsl_laneq_s16): Likewise.
22745 (vmlsl_laneq_s32): Likewise.
22746 (vmlsl_laneq_u16): Likewise.
22747 (vmlsl_laneq_u32): Likewise.
22749 2021-01-29 Richard Biener <rguenther@suse.de>
22751 * doc/invoke.texi (--param max-gcse-memory): Document unit
22753 * gcse.c (gcse_or_cprop_is_too_expensive): Adjust.
22754 * params.opt (--param max-gcse-memory): Adjust default and
22755 document unit of size.
22757 2021-01-29 Richard Biener <rguenther@suse.de>
22759 PR rtl-optimization/98863
22760 * gcse.c (gcse_or_cprop_is_too_expensive): Use unsigned
22761 HOST_WIDE_INT for the memory estimate.
22763 2021-01-29 Bin Cheng <bin.cheng@linux.alibaba.com>
22764 Richard Biener <rguenther@suse.de>
22766 PR tree-optimization/97627
22767 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
22768 Do not analyze fake edges.
22770 2021-01-29 Richard Biener <rguenther@suse.de>
22772 PR rtl-optimization/98144
22773 * df.h (df_mir_bb_info): Add con_visited member.
22774 * df-problems.c (df_mir_alloc): Initialize con_visited,
22775 do not fully populate IN and OUT.
22776 (df_mir_reset): Likewise.
22777 (df_mir_confluence_0): Set con_visited.
22778 (df_mir_confluence_n): Properly handle implicitely
22779 fully populated IN and OUT as designated by con_visited
22780 and update con_visited accordingly.
22782 2021-01-29 Jakub Jelinek <jakub@redhat.com>
22785 * config/arm/vec-common.md (mve_vshlq_<supf><mode>,
22786 vashl<mode>3, vashr<mode>3, vlshr<mode>3): Add
22787 && !TARGET_REALLY_IWMMXT to conditions.
22789 2021-01-29 Jakub Jelinek <jakub@redhat.com>
22792 * cfgbuild.c (find_bb_boundaries): Reset debug_insn when seeing
22795 2021-01-28 Marek Polacek <polacek@redhat.com>
22798 * stor-layout.c (finalize_type_size): If we reset TYPE_USER_ALIGN in
22799 the main variant, maybe reset it in its variants too.
22800 * tree.c (check_base_type): Return true only if TYPE_USER_ALIGN match.
22801 (check_aligned_type): Check if TYPE_USER_ALIGN match.
22803 2021-01-28 Christophe Lyon <christophe.lyon@linaro.org>
22806 * config/arm/arm.c (arm_rtx_costs_internal): Adjust cost of vector
22807 of constant zero for comparisons.
22809 2021-01-28 Michael Meissner <meissner@linux.ibm.com>
22811 * config/rs6000/rs6000.c (rs6000_mangle_decl_assembler_name): Add
22812 support for mapping built-in function names for long double
22813 built-in functions if long double is IEEE 128-bit.
22815 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
22817 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_n
22818 builtin generator macros.
22819 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_n<mode>):
22821 * config/aarch64/arm_neon.h (vmlsl_n_s16): Use RTL builtin
22822 instead of inline asm.
22823 (vmlsl_n_s32): Likewise.
22824 (vmlsl_n_u16): Likewise.
22825 (vmlsl_n_u32): Likewise.
22827 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
22829 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_n
22830 builtin generator macros.
22831 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>):
22833 * config/aarch64/arm_neon.h (vmlal_n_s16): Use RTL builtin
22834 instead of inline asm.
22835 (vmlal_n_s32): Likewise.
22836 (vmlal_n_u16): Likewise.
22837 (vmlal_n_u32): Likewise.
22839 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22841 * config/aarch64/aarch64-simd-builtins.def (shrn2): Define
22843 * config/aarch64/aarch64-simd.md (aarch64_shrn2<mode>_insn_le):
22845 (aarch64_shrn2<mode>_insn_be): Likewise.
22846 (aarch64_shrn2<mode>): Likewise.
22847 * config/aarch64/arm_neon.h (vshrn_high_n_s16): Reimlplement
22849 (vshrn_high_n_s32): Likewise.
22850 (vshrn_high_n_s64): Likewise.
22851 (vshrn_high_n_u16): Likewise.
22852 (vshrn_high_n_u32): Likewise.
22853 (vshrn_high_n_u64): Likewise.
22855 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22857 * config/aarch64/aarch64-simd-builtins.def (shrn): Define
22859 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le):
22861 (aarch64_shrn<mode>_insn_be): Likewise.
22862 (aarch64_shrn<mode>): Likewise.
22863 * config/aarch64/arm_neon.h (vshrn_n_s16): Reimplement using
22865 (vshrn_n_s32): Likewise.
22866 (vshrn_n_s64): Likewise.
22867 (vshrn_n_u16): Likewise.
22868 (vshrn_n_u32): Likewise.
22869 (vshrn_n_u64): Likewise.
22870 * config/aarch64/iterators.md (vn_mode): New mode attribute.
22872 2021-01-28 Richard Biener <rguenther@suse.de>
22874 PR rtl-optimization/80960
22875 * dse.c (check_mem_read_rtx): Call get_addr on the
22878 2021-01-28 Xionghu Luo <luoxhu@linux.ibm.com>
22879 David Edelsohn <dje.gcc@gmail.com>
22882 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22883 Don't generate VIEW_CONVERT_EXPR for fcode ALTIVEC_BUILTIN_VEC_INSERT
22885 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
22887 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Remove the
22888 wrapper call rs6000_expand_vector_set_var for cleanup. Call
22889 rs6000_expand_vector_set_var_p9 and rs6000_expand_vector_set_var_p8
22891 (rs6000_expand_vector_set_var): Delete.
22892 (rs6000_expand_vector_set_var_p9): Make static.
22893 (rs6000_expand_vector_set_var_p8): Make static.
22895 2021-01-28 Xing GUO <higuoxing@gmail.com>
22897 * common/config/riscv/riscv-common.c
22898 (riscv_subset_list::parsing_subset_version): Fix -march option parsing
22899 when `p` extension exists.
22901 2021-01-27 Vladimir N. Makarov <vmakarov@redhat.com>
22903 PR rtl-optimization/97684
22904 * ira.c (ira): Call ira_set_pseudo_classes before
22905 update_equiv_regs when it is necessary.
22907 2021-01-27 Jakub Jelinek <jakub@redhat.com>
22910 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Use
22911 %w0, %w1 and %2 instead of %0, %1 and %2.
22913 2021-01-27 Aaron Sawdey <acsawdey@linux.ibm.com>
22915 * config/rs6000/genfusion.pl: New script to generate
22916 define_insn_and_split patterns so combine can arrange fused
22917 instructions next to each other.
22918 * config/rs6000/fusion.md: New file, generated fused instruction
22919 patterns for combine.
22920 * config/rs6000/predicates.md (const_m1_to_1_operand): New predicate.
22921 (non_update_memory_operand): New predicate.
22922 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_P10_FUSION and
22923 OPTION_MASK_P10_FUSION_LD_CMPI to ISA_3_1_MASKS_SERVER and
22925 * config/rs6000/rs6000-protos.h (address_is_non_pfx_d_or_x): Add
22927 * config/rs6000/rs6000.c (rs6000_option_override_internal):
22928 Automatically set OPTION_MASK_P10_FUSION and
22929 OPTION_MASK_P10_FUSION_LD_CMPI if target is power10.
22930 (rs600_opt_masks): Allow -mpower10-fusion
22931 in function attributes.
22932 (address_is_non_pfx_d_or_x): New function.
22933 * config/rs6000/rs6000.h: Add MASK_P10_FUSION.
22934 * config/rs6000/rs6000.md: Include fusion.md.
22935 * config/rs6000/rs6000.opt: Add -mpower10-fusion
22936 and -mpower10-fusion-ld-cmpi.
22937 * config/rs6000/t-rs6000: Add dependencies involving fusion.md.
22939 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
22941 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal
22942 builtin generator macros.
22943 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal<mode>):
22945 (aarch64_<su>mlal<mode>): This.
22946 * config/aarch64/arm_neon.h (vmlal_s8): Use RTL builtin
22947 instead of inline asm.
22948 (vmlal_s16): Likewise.
22949 (vmlal_s32): Likewise.
22950 (vmlal_u8): Likewise.
22951 (vmlal_u16): Likewise.
22952 (vmlal_u32): Likewise.
22954 2021-01-27 Richard Biener <rguenther@suse.de>
22956 PR tree-optimization/98854
22957 * tree-vect-slp.c (vect_build_slp_tree_2): Also build
22958 PHIs from scalars when the number of CTORs matches the
22959 number of children.
22961 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
22963 * config/aarch64/aarch64-simd-builtins.def: Add mls_n builtin
22965 * config/aarch64/aarch64-simd.md (*aarch64_mls_elt_merge<mode>):
22967 (aarch64_mls_n<mode>): This.
22968 * config/aarch64/arm_neon.h (vmls_n_s16): Use RTL builtin
22970 (vmls_n_s32): Likewise.
22971 (vmls_n_u16): Likewise.
22972 (vmls_n_u32): Likewise.
22973 (vmlsq_n_s16): Likewise.
22974 (vmlsq_n_s32): Likewise.
22975 (vmlsq_n_u16): Likewise.
22976 (vmlsq_n_u32): Likewise.
22978 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
22980 * config/aarch64/aarch64-simd-builtins.def: Add mls builtin
22982 * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather
22984 (vmls_s16): Likewise.
22985 (vmls_s32): Likewise.
22986 (vmls_u8): Likewise.
22987 (vmls_u16): Likewise.
22988 (vmls_u32): Likewise.
22989 (vmlsq_s8): Likewise.
22990 (vmlsq_s16): Likewise.
22991 (vmlsq_s32): Likewise.
22992 (vmlsq_u8): Likewise.
22993 (vmlsq_u16): Likewise.
22994 (vmlsq_u32): Likewise.
22996 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
22998 * config/aarch64/aarch64-simd-builtins.def: Add mla_n builtin
23000 * config/aarch64/aarch64-simd.md (*aarch64_mla_elt_merge<mode>):
23002 (aarch64_mla_n<mode>): This.
23003 * config/aarch64/arm_neon.h (vmla_n_s16): Use RTL builtin
23005 (vmla_n_s32): Likewise.
23006 (vmla_n_u16): Likewise.
23007 (vmla_n_u32): Likewise.
23008 (vmlaq_n_s16): Likewise.
23009 (vmlaq_n_s32): Likewise.
23010 (vmlaq_n_u16): Likewise.
23011 (vmlaq_n_u32): Likewise.
23013 2021-01-27 liuhongt <hongtao.liu@intel.com>
23016 * config/i386/sse.md (sse2_gt<mode>3): Drop !TARGET_XOP in condition.
23017 (*sse2_eq<mode>3): Ditto.
23019 2021-01-27 Jakub Jelinek <jakub@redhat.com>
23021 * tree-pass.h (PROP_trees): Rename to ...
23022 (PROP_gimple): ... this.
23023 * cfgexpand.c (pass_data_expand): Replace PROP_trees with PROP_gimple.
23024 * passes.c (execute_function_dump, execute_function_todo,
23025 execute_one_ipa_transform_pass, execute_one_pass): Likewise.
23026 * varpool.c (ctor_for_folding): Likewise.
23028 2021-01-27 Jakub Jelinek <jakub@redhat.com>
23030 PR tree-optimization/97260
23031 * varpool.c: Include tree-pass.h.
23032 (ctor_for_folding): In GENERIC return DECL_INITIAL for TREE_READONLY
23033 non-TREE_SIDE_EFFECTS automatic variables.
23035 2021-01-26 Paul Fee <paul.f.fee@gmail.com>
23037 * doc/cpp.texi (__cplusplus): Document value for -std=c++23
23039 * doc/invoke.texi: Document -std=c++23 and -std=gnu++23.
23040 * dwarf2out.c (highest_c_language): Recognise C++20 and C++23.
23041 (gen_compile_unit_die): Recognise C++23.
23043 2021-01-26 Jakub Jelinek <jakub@redhat.com>
23046 * dwarf2asm.c (dw2_assemble_integer): Cast DWARF2_ADDR_SIZE to int
23049 2021-01-26 Jakub Jelinek <jakub@redhat.com>
23052 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
23053 Use UINTVAL (shft_amnt) and UINTVAL (mask) instead of INTVAL (shft_amnt)
23054 and INTVAL (mask). Add && INTVAL (mask) > 0 condition.
23056 2021-01-26 Richard Biener <rguenther@suse.de>
23058 * gimple-pretty-print.c (dump_binary_rhs): Handle
23059 VEC_WIDEN_{PLUS,MINUS}_{LO,HI}_EXPR.
23061 2021-01-26 Richard Biener <rguenther@suse.de>
23063 PR middle-end/98726
23064 * tree.h (vector_cst_int_elt): Remove.
23065 * tree.c (vector_cst_int_elt): Use poly_wide_int for computations,
23068 2021-01-26 Andrew Stubbs <ams@codesourcery.com>
23070 * config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
23071 for V64DFmode min/max reductions.
23073 2021-01-26 Jakub Jelinek <jakub@redhat.com>
23075 * dwarf2asm.c (dw2_assemble_integer): Handle size twice as large
23076 as DWARF2_ADDR_SIZE if x is not a scalar int by emitting it as
23077 two halves, one with x and the other with const0_rtx, ordered
23078 depending on endianity.
23080 2021-01-26 Alexandre Oliva <oliva@adacore.com>
23082 * gimplify.c (gimplify_decl_expr): Skip asan marking calls for
23083 temporaries not seen in binding block, and not about to be
23084 added as gimple variables.
23086 2021-01-25 Martin Sebor <msebor@redhat.com>
23089 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust warning text.
23091 2021-01-25 Martin Liska <mliska@suse.cz>
23093 * value-prof.c (get_nth_most_common_value): Use %s instead
23096 2021-01-25 Jakub Jelinek <jakub@redhat.com>
23099 * configure.ac (HAVE_AS_GDWARF_5_DEBUG_FLAG): Only define if
23100 readelf -wi is able to read the emitted .debug_info back.
23101 * configure: Regenerated.
23103 2021-01-25 Martin Liska <mliska@suse.cz>
23105 PR gcov-profile/98739
23106 * common.opt: Add missing sign symbol.
23107 * value-prof.c (get_nth_most_common_value): Restore handling
23108 of PROFILE_REPRODUCIBILITY_PARALLEL_RUNS and
23109 PROFILE_REPRODUCIBILITY_MULTITHREADED.
23111 2021-01-25 Richard Biener <rguenther@suse.de>
23113 PR middle-end/98807
23114 * tree.c (vector_element_bits): Always use precision of
23115 the element type for boolean vectors.
23117 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
23119 * config/rtems.h (STARTFILE_SPEC): Remove qnolinkcmds.
23120 (ENDFILE_SPEC): Evaluate qnolinkcmds.
23122 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
23124 * config/rtems.h (STARTFILE_SPEC): Remove nostdlib and
23125 nostartfiles handling since this is already done by
23126 LINK_COMMAND_SPEC. Evaluate qnolinkcmds.
23127 (ENDFILE_SPEC): Remove nostdlib and nostartfiles handling since this
23128 is already done by LINK_COMMAND_SPEC.
23129 (LIB_SPECS): Remove nostdlib and nodefaultlibs handling since
23130 this is already done by LINK_COMMAND_SPEC. Remove qnolinkcmds
23133 2021-01-25 Jakub Jelinek <jakub@redhat.com>
23136 * fold-const-call.c (host_size_t_cst_p): Renamed to ...
23137 (size_t_cst_p): ... this. Check and store unsigned HOST_WIDE_INT
23138 value rather than host size_t.
23139 (fold_const_call): Change type of s2 from size_t to
23140 unsigned HOST_WIDE_INT. Use size_t_cst_p instead of
23141 host_size_t_cst_p. For strncmp calls, pass MIN (s2, SIZE_MAX)
23142 instead of s2 as last argument.
23144 2021-01-25 Tamar Christina <tamar.christina@arm.com>
23146 * config/arm/iterators.md (rotsplit1, rotsplit2, conj_op, fcmac1,
23147 VCMLA_OP, VCMUL_OP): New.
23148 * config/arm/mve.md (mve_vcmlaq<mve_rot><mode>): Support vec_dup 0.
23149 * config/arm/neon.md (cmul<conj_op><mode>3): New.
23150 * config/arm/unspecs.md (UNSPEC_VCMLA_CONJ, UNSPEC_VCMLA180_CONJ,
23151 UNSPEC_VCMUL_CONJ): New.
23152 * config/arm/vec-common.md (cmul<conj_op><mode>3, arm_vcmla<rot><mode>,
23153 cml<fcmac1><conj_op><mode>4): New.
23155 2021-01-23 Jakub Jelinek <jakub@redhat.com>
23158 * config/rs6000/mmintrin.h (__m64): Add __may_alias__ attribute.
23160 2021-01-22 Jonathan Wright <jonathan.wright@arm.com>
23162 * config/aarch64/aarch64-simd-builtins.def: Add mla builtin
23164 * config/aarch64/arm_neon.h (vmla_s8): Use RTL builtin rather
23166 (vmla_s16): Likewise.
23167 (vmla_s32): Likewise.
23168 (vmla_u8): Likewise.
23169 (vmla_u16): Likewise.
23170 (vmla_u32): Likewise.
23171 (vmlaq_s8): Likewise.
23172 (vmlaq_s16): Likewise.
23173 (vmlaq_s32): Likewise.
23174 (vmlaq_u8): Likewise.
23175 (vmlaq_u16): Likewise.
23176 (vmlaq_u32): Likewise.
23178 2021-01-22 David Malcolm <dmalcolm@redhat.com>
23180 * doc/invoke.texi (GCC_EXTRA_DIAGNOSTIC_OUTPUT): Add @findex
23183 2021-01-22 Jakub Jelinek <jakub@redhat.com>
23186 * dwarf2out.c (output_file_names): For -gdwarf-5, if there are no
23187 filenames to emit, still emit the required 0 index directory and
23188 filename entries that match DW_AT_comp_dir and DW_AT_name of the
23191 2021-01-22 Marek Polacek <polacek@redhat.com>
23194 * doc/invoke.texi: Update C++ ABI Version 15 description.
23196 2021-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23198 PR tree-optimization/98766
23199 * tree-ssa-math-opts.c (convert_mult_to_fma): Use maybe_le when
23200 comparing against type size with param_avoid_fma_max_bits.
23202 2021-01-22 Richard Biener <rguenther@suse.de>
23204 PR middle-end/98793
23205 * tree.c (vector_element_bits): Key single-bit bool vector on
23206 integer mode rather than not vector mode.
23208 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
23211 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
23212 Generate ARRAY_REF(VIEW_CONVERT_EXPR) for P8 and later
23214 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): Update
23215 to call different path for P8 and P9.
23216 (rs6000_expand_vector_set_var_p9): New function.
23217 (rs6000_expand_vector_set_var_p8): New function.
23219 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
23223 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
23224 Ajdust variable index vec_insert from address dereference to
23225 ARRAY_REF(VIEW_CONVERT_EXPR) tree expression.
23226 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
23228 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): New function.
23230 2021-01-22 Martin Liska <mliska@suse.cz>
23232 PR gcov-profile/98739
23233 * profile.c (compute_value_histograms): Drop time profile for
23234 -fprofile-reproducible=multithreaded.
23236 2021-01-22 Nathan Sidwell <nathan@acm.org>
23238 * gcc.c (process_command): Don't check OPT_SPECIAL_input_file
23241 2021-01-22 Richard Biener <rguenther@suse.de>
23243 PR middle-end/98773
23244 * tree-data-ref.c (initalize_matrix_A): Revert previous
23245 change, retaining failing on HOST_WIDE_INT_MIN CHREC_RIGHT.
23247 2021-01-22 Jakub Jelinek <jakub@redhat.com>
23249 PR tree-optimization/90248
23250 * match.pd (X cmp 0.0 ? 1.0 : -1.0 -> copysign(1, +-X),
23251 X cmp 0.0 ? -1.0 : +1.0 -> copysign(1, -+X)): Remove
23253 (X * (X cmp 0.0 ? 1.0 : -1.0) -> +-abs(X),
23254 X * (X cmp 0.0 ? -1.0 : 1.0) -> +-abs(X)): New simplifications.
23256 2021-01-22 Jakub Jelinek <jakub@redhat.com>
23258 PR tree-optimization/98255
23259 * tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
23260 extend index - low_bound from sizetype's precision rather than index
23262 (get_addr_base_and_unit_offset_1): Likewise.
23263 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
23264 * gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
23266 2021-01-22 Richard Biener <rguenther@suse.de>
23268 PR tree-optimization/98786
23269 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Avoid
23270 adding new uses of abnormals. Verify we deal with a conditional
23273 2021-01-22 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23276 * optc-save-gen.awk: Add arm_fp16_format to checked_options.
23278 2021-01-22 liuhongt <hongtao.liu@intel.com>
23282 * config/i386/sse.md (VI_128_256): New mode iterator.
23283 (*avx_cmp<mode>3_1, *avx_cmp<mode>3_2, *avx_cmp<mode>3_3,
23284 *avx_cmp<mode>3_4, *avx2_eq<mode>3, *avx2_pcmp<mode>3_1,
23285 *avx2_pcmp<mode>3_2, *avx2_gt<mode>3): New
23286 define_insn_and_split to lower avx512 vector comparison to avx
23287 version when dest is vector.
23288 (*<avx512>_cmp<mode>3,*<avx512>_cmp<mode>3,*<avx512>_ucmp<mode>3):
23289 define_insn_and_split for negating the comparison result.
23290 * config/i386/predicates.md (float_vector_all_ones_operand):
23292 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Use
23293 general NOT operator without UNSPEC_MASKOP.
23295 2021-01-21 Vladimir N. Makarov <vmakarov@redhat.com>
23297 PR rtl-optimization/98777
23298 * lra-int.h (lra_pmode_pseudo): New extern.
23299 * lra.c (lra_pmode_pseudo): New global.
23301 * lra-eliminations.c (eliminate_regs_in_insn): Use it.
23303 2021-01-21 Ilya Leoshkevich <iii@linux.ibm.com>
23305 * fwprop.c (fwprop_propagation::classify_result): Allow
23306 (subreg (mem)) simplifications.
23308 2021-01-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23310 * config/aarch64/aarch64-simd.md (aarch64_sqdml<SBINQOPS:as>l<mode>):
23312 (aarch64_sqdmlal<mode>): ... This...
23313 (aarch64_sqdmlsl<mode>): ... And this.
23314 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Split into...
23315 (aarch64_sqdmlal_lane<mode>): ... This...
23316 (aarch64_sqdmlsl_lane<mode>): ... And this.
23317 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Split into...
23318 (aarch64_sqdmlsl_laneq<mode>): ... This...
23319 (aarch64_sqdmlal_laneq<mode>): ... And this.
23320 (aarch64_sqdml<SBINQOPS:as>l_n<mode>): Split into...
23321 (aarch64_sqdmlsl_n<mode>): ... This...
23322 (aarch64_sqdmlal_n<mode>): ... And this.
23323 (aarch64_sqdml<SBINQOPS:as>l2<mode>_internal): Split into...
23324 (aarch64_sqdmlal2<mode>_internal): ... This...
23325 (aarch64_sqdmlsl2<mode>_internal): ... And this.
23327 2021-01-21 Christophe Lyon <christophe.lyon@linaro.org>
23329 * config/arm/arm_mve.h (__arm_vcmpneq_s8): Fix return type.
23331 2021-01-21 Andrea Corallo <andrea.corallo@arm.com>
23334 * doc/sourcebuild.texi (arm_thumb2_no_arm_v8_1_lob): Document.
23336 2021-01-21 liuhongt <hongtao.liu@intel.com>
23338 PR rtl-optimization/98694
23339 * regcprop.c (copy_value): If SRC had been assigned a mode
23340 narrower than the copy, we can't link DEST into the chain even
23341 they have same hard_regno_nregs(i.e. HImode/SImode in i386
23344 2021-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23346 * config/aarch64/aarch64-simd.md (aarch64_get_lane<mode>):
23347 Convert to define_insn_and_split. Split into simple move when moving
23350 2021-01-20 Segher Boessenkool <segher@kernel.crashing.org>
23352 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Change assert.
23353 Adjust comment. Simplify code.
23355 2021-01-20 Jakub Jelinek <jakub@redhat.com>
23358 * dwarf2out.c (reset_indirect_string): Also reset indirect strings
23359 with DW_FORM_line_strp form.
23360 (prune_unused_types_update_strings): Don't add into debug_str_hash
23361 indirect strings with DW_FORM_line_strp form.
23362 (adjust_name_comp_dir): New function.
23363 (dwarf2out_finish): Call it on CU DIEs after resetting
23364 debug_line_str_hash.
23366 2021-01-20 Vladimir N. Makarov <vmakarov@redhat.com>
23368 PR rtl-optimization/98722
23369 * lra-eliminations.c (eliminate_regs_in_insn): Check that target
23370 has no 3-op add insn to transform insns containing two pluses.
23372 2021-01-20 Richard Biener <rguenther@suse.de>
23374 * hwint.h (add_hwi): New function.
23375 (mul_hwi): Likewise.
23376 * tree-data-ref.c (initialize_matrix_A): Properly translate
23377 tree constants and avoid HOST_WIDE_INT_MIN.
23378 (lambda_matrix_row_add): Avoid undefined integer overflow
23379 and return true on such overflow.
23380 (lambda_matrix_right_hermite): Handle overflow from
23381 lambda_matrix_row_add gracefully. Simplify previous fix.
23382 (analyze_subscript_affine_affine): Likewise.
23384 2021-01-20 Eugene Rozenfeld <erozen@microsoft.com>
23386 PR tree-optimization/96674
23387 * match.pd: New patterns: x < y || y == XXX_MIN --> x <= y - 1
23388 x >= y && y != XXX_MIN --> x > y - 1
23390 2021-01-20 Richard Sandiford <richard.sandiford@arm.com>
23392 PR tree-optimization/98535
23393 * tree-vect-slp.c (duplicate_and_interleave): Use quick_grow_cleared.
23394 If the high and low permutes are the same, remove the high permutes
23395 from the working set and only continue with the low ones.
23397 2021-01-20 Jakub Jelinek <jakub@redhat.com>
23399 PR tree-optimization/98721
23400 * builtins.c (access_ref::inform_access): Don't assume
23401 SSA_NAME_IDENTIFIER must be non-NULL. Print messages about
23402 object whenever allocfn is NULL, rather than only when DECL_P
23403 is true. Use %qE instead of %qD for that. Formatting fixes.
23405 2021-01-20 Richard Biener <rguenther@suse.de>
23407 PR tree-optimization/98758
23408 * tree-data-ref.c (int_divides_p): Use lambda_int arguments.
23409 (lambda_matrix_right_hermite): Avoid undefinedness with
23410 signed integer abs and multiplication.
23411 (analyze_subscript_affine_affine): Use lambda_int.
23413 2021-01-20 David Malcolm <dmalcolm@redhat.com>
23416 * dwarf2out.c (output_line_info): Rename static variable
23417 "generation", moving it out of the function to...
23418 (output_line_info_generation): New.
23419 (init_sections_and_labels): Likewise, renaming the variable to...
23420 (init_sections_and_labels_generation): New.
23421 (dwarf2out_c_finalize): Reset the new variables.
23423 2021-01-19 Martin Sebor <msebor@redhat.com>
23425 PR middle-end/98664
23426 * tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
23427 all functions, even if they're not declared artificial or inline.
23428 * tree.c (tree_inlined_location): Use macro expansion location
23429 only if scope traversal fails to expose one.
23431 2021-01-19 Richard Sandiford <richard.sandiford@arm.com>
23433 PR rtl-optimization/92294
23434 * alias.c (compare_base_symbol_refs): Take an extra parameter
23435 and add the distance between two symbols to it. Enshrine in
23436 comments that -1 means "either 0 or 1, but we can't tell
23437 which at compile time".
23438 (memrefs_conflict_p): Update call accordingly.
23439 (rtx_equal_for_memref_p): Likewise. Take the distance between symbols
23442 2021-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23444 * config/aarch64/aarch64-simd-builtins.def (sqshl, uqshl,
23445 sqrshl, uqrshl, sqadd, uqadd, sqsub, uqsub, suqadd, usqadd, sqmovn,
23446 uqmovn, sqxtn2, uqxtn2, sqabs, sqneg, sqdmlal, sqdmlsl, sqdmlal_lane,
23447 sqdmlsl_lane, sqdmlal_laneq, sqdmlsl_laneq, sqdmlal_n, sqdmlsl_n,
23448 sqdmlal2, sqdmlsl2, sqdmlal2_lane, sqdmlsl2_lane, sqdmlal2_laneq,
23449 sqdmlsl2_laneq, sqdmlal2_n, sqdmlsl2_n, sqdmull, sqdmull_lane,
23450 sqdmull_laneq, sqdmull_n, sqdmull2, sqdmull2_lane, sqdmull2_laneq,
23451 sqdmull2_n, sqdmulh, sqrdmulh, sqdmulh_lane, sqdmulh_laneq,
23452 sqrdmulh_lane, sqrdmulh_laneq, sqshrun_n, sqrshrun_n, sqshrn_n,
23453 uqshrn_n, sqrshrn_n, uqrshrn_n, sqshlu_n, sqshl_n, uqshl_n, sqrdmlah,
23454 sqrdmlsh, sqrdmlah_lane, sqrdmlsh_lane, sqrdmlah_laneq, sqrdmlsh_laneq,
23455 sqmovun): Use NONE flags.
23457 2021-01-19 Richard Biener <rguenther@suse.de>
23460 * ipa-modref.c (analyze_stmt): Only record a summary for a
23463 2021-01-19 Richard Biener <rguenther@suse.de>
23465 PR middle-end/98638
23466 * tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
23468 2021-01-19 Daniel Hellstrom <daniel@gaisler.com>
23470 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
23471 built-in define __FIX_LEON3FT_TN0018.
23473 2021-01-19 Richard Biener <rguenther@suse.de>
23476 * tree-inline.c (tree_function_versioning): Set input_location
23477 to UNKNOWN_LOCATION throughout the function.
23479 2021-01-19 Tobias Burnus <tobias@codesourcery.com>
23482 * omp-low.c (lower_omp_target): Handle nonpointer is_device_ptr.
23484 2021-01-19 Martin Jambor <mjambor@suse.cz>
23487 * ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
23488 whether non-call exceptions allow removal of a statement.
23489 (isra_analyze_call): Pass the appropriate function to
23490 ssa_name_only_returned_p.
23492 2021-01-19 Geng Qi <gengqi@linux.alibaba.com>
23494 * config/riscv/arch-canonicalize (longext_sort): New function for
23495 sorting 'multi-letter'.
23496 * config/riscv/multilib-generator: Adjusting the loop of 'alt' in
23497 'alts'. The 'arch' may not be the first of 'alts'.
23498 (_expand_combination): Add underline for the 'ext' without '*'.
23499 This is because, a single-letter extension can always be treated well
23500 with a '_' prefix, but it cannot be separated out if it is appended
23503 2021-01-18 Vladimir N. Makarov <vmakarov@redhat.com>
23506 * ira.c (ira): Skip abnormal critical edge splitting.
23508 2021-01-18 Jakub Jelinek <jakub@redhat.com>
23510 PR tree-optimization/98727
23511 * tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
23512 second .MUL_OVERFLOW operand for signed multiplication with overflow
23513 checking if the second operand of multiplication is not constant.
23515 2021-01-18 David Edelsohn <dje.gcc@gmail.com>
23517 * doc/invoke.texi (-gdwarf): TPF defaults to version 2 and AIX
23518 defaults to version 4.
23520 2021-01-18 David Malcolm <dmalcolm@redhat.com>
23522 * attribs.h (fndecl_dealloc_argno): New decl.
23523 * builtins.c (call_dealloc_argno): Split out second half of
23525 (fndecl_dealloc_argno): New.
23526 * doc/extend.texi (Common Function Attributes): Document the
23527 interaction between the analyzer and the malloc attribute.
23528 * doc/invoke.texi (Static Analyzer Options): Likewise.
23530 2021-01-17 David Edelsohn <dje.gcc@gmail.com>
23532 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Override
23533 dwarf_version to 4.
23534 * config/rs6000/aix72.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
23536 2021-01-17 Martin Jambor <mjambor@suse.cz>
23539 * cgraph.c (clone_of_p): Check also former_clone_of as we climb
23542 2021-01-17 Mark Wielaard <mark@klomp.org>
23544 * common.opt (gdwarf-): Init(5).
23545 * doc/invoke.texi (-gdwarf): Document default to 5.
23547 2021-01-16 Kwok Cheung Yeung <kcy@codesourcery.com>
23549 * builtin-types.def
23550 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
23552 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
23553 ...this. Add extra argument.
23554 * gimplify.c (omp_default_clause): Ensure that event handle is
23555 firstprivate in a task region.
23556 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_DETACH.
23557 (gimplify_adjust_omp_clauses): Likewise.
23558 * omp-builtins.def (BUILT_IN_GOMP_TASK): Change function type to
23559 BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR.
23560 * omp-expand.c (expand_task_call): Add GOMP_TASK_FLAG_DETACH to flags
23561 if detach clause specified. Add detach argument when generating
23563 * omp-low.c (scan_sharing_clauses): Setup data environment for detach
23565 (finish_taskreg_scan): Move field for variable containing the event
23566 handle to the front of the struct.
23567 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_DETACH. Fix
23569 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
23570 OMP_CLAUSE_DETACH clause.
23571 (convert_local_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
23572 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_DETACH.
23573 * tree.c (omp_clause_num_ops): Add entry for OMP_CLAUSE_DETACH.
23575 (omp_clause_code_name): Add entry for OMP_CLAUSE_DETACH. Fix
23577 (walk_tree_1): Handle OMP_CLAUSE_DETACH.
23579 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
23581 * config/nios2/t-rtems: Reset all MULTILIB_* variables. Shorten
23582 multilib directory names. Use MULTILIB_REQUIRED instead of
23583 MULTILIB_EXCEPTIONS. Add -mhw-mul -mhw-mulx -mhw-div
23584 -mcustom-fpu-cfg=fph2 multilib.
23586 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
23588 * config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
23589 (nios2_init_fpu_configs): Provide register values for new
23590 -mcustom-fpu-cfg=fph2 option variant.
23591 * doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
23594 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
23596 * config/nios2/nios2.c (nios2_custom_check_insns): Remove
23597 custom instruction warnings.
23599 2021-01-16 Jakub Jelinek <jakub@redhat.com>
23601 PR tree-optimization/96669
23602 * match.pd ((CST << x) & 1 -> x == 0): New simplification.
23604 2021-01-16 Jakub Jelinek <jakub@redhat.com>
23606 PR tree-optimization/96271
23607 * passes.def: Pass false argument to first two pass_cd_dce
23608 instances and true to last instance. Add comment that
23609 last instance rewrites no longer addressed locals.
23610 * tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
23612 (pass_cd_dce::set_pass_param): New method.
23613 (pass_cd_dce::execute): Return TODO_update_address_taken from
23614 last cd_dce instance.
23616 2021-01-15 Carl Love <cel@us.ibm.com>
23618 * config/rs6000/altivec.h (vec_mulh, vec_div, vec_dive, vec_mod):
23620 * config/rs6000/altivec.md (VIlong): Move define to file vsx.md.
23621 * config/rs6000/rs6000-builtin.def (DIVES_V4SI, DIVES_V2DI,
23622 DIVEU_V4SI, DIVEU_V2DI, DIVS_V4SI, DIVS_V2DI, DIVU_V4SI,
23623 DIVU_V2DI, MODS_V2DI, MODS_V4SI, MODU_V2DI, MODU_V4SI,
23624 MULHS_V2DI, MULHS_V4SI, MULHU_V2DI, MULHU_V4SI, MULLD_V2DI):
23625 Add builtin define.
23626 (MULH, DIVE, MOD): Add new BU_P10_OVERLOAD_2 definitions.
23627 * config/rs6000/rs6000-call.c (VSX_BUILTIN_VEC_DIV,
23628 VSX_BUILTIN_VEC_DIVE, P10_BUILTIN_VEC_MOD, P10_BUILTIN_VEC_MULH):
23629 New overloaded definitions.
23630 (builtin_function_type) [P10V_BUILTIN_DIVEU_V4SI,
23631 P10V_BUILTIN_DIVEU_V2DI, P10V_BUILTIN_DIVU_V4SI,
23632 P10V_BUILTIN_DIVU_V2DI, P10V_BUILTIN_MODU_V2DI,
23633 P10V_BUILTIN_MODU_V4SI, P10V_BUILTIN_MULHU_V2DI,
23634 P10V_BUILTIN_MULHU_V4SI]: Add case
23635 statement for builtins.
23636 * config/rs6000/rs6000.md (bits): Add new attribute sizes V4SI, V2DI.
23637 * config/rs6000/vsx.md (VIlong): Moved from config/rs6000/altivec.md.
23638 (UNSPEC_VDIVES, UNSPEC_VDIVEU): New unspec definitions.
23639 (vsx_mul_v2di): Add if TARGET_POWER10 statement.
23640 (vsx_udiv_v2di): Add if TARGET_POWER10 statement.
23641 (dives_<mode>, diveu_<mode>, div<mode>3, uvdiv<mode>3,
23642 mods_<mode>, modu_<mode>, mulhs_<mode>, mulhu_<mode>, mulv2di3):
23643 Add define_insn, mode is VIlong.
23644 * doc/extend.texi (vec_mulh, vec_mul, vec_div, vec_dive, vec_mod):
23645 Add builtin descriptions.
23647 2021-01-15 Eric Botcazou <ebotcazou@adacore.com>
23649 * final.c (final_start_function_1): Reset force_source_line.
23651 2021-01-15 Jakub Jelinek <jakub@redhat.com>
23653 PR tree-optimization/96669
23654 * match.pd (((1 << A) & 1) != 0 -> A == 0,
23655 ((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
23656 possibly different power of two constants and to right shift too.
23658 2021-01-15 Jakub Jelinek <jakub@redhat.com>
23660 PR tree-optimization/96681
23661 * match.pd ((x < 0) ^ (y < 0) to (x ^ y) < 0): New simplification.
23662 ((x >= 0) ^ (y >= 0) to (x ^ y) < 0): Likewise.
23663 ((x < 0) ^ (y >= 0) to (x ^ y) >= 0): Likewise.
23664 ((x >= 0) ^ (y < 0) to (x ^ y) >= 0): Likewise.
23666 2021-01-15 Alexandre Oliva <oliva@adacore.com>
23668 * opts.c (gen_command_line_string): Exclude -dumpbase-ext.
23670 2021-01-15 Tamar Christina <tamar.christina@arm.com>
23672 * config/aarch64/aarch64-simd.md (cml<fcmac1><conj_op><mode>4,
23673 cmul<conj_op><mode>3): New.
23674 * config/aarch64/iterators.md (UNSPEC_FCMUL,
23675 UNSPEC_FCMUL180, UNSPEC_FCMLA_CONJ, UNSPEC_FCMLA180_CONJ,
23676 UNSPEC_CMLA_CONJ, UNSPEC_CMLA180_CONJ, UNSPEC_CMUL, UNSPEC_CMUL180,
23677 FCMLA_OP, FCMUL_OP, conj_op, rotsplit1, rotsplit2, fcmac1, sve_rot1,
23678 sve_rot2, SVE2_INT_CMLA_OP, SVE2_INT_CMUL_OP, SVE2_INT_CADD_OP): New.
23679 (rot): Add UNSPEC_FCMUL, UNSPEC_FCMUL180.
23680 (rot_op): Renamed to conj_op.
23681 * config/aarch64/aarch64-sve.md (cml<fcmac1><conj_op><mode>4,
23682 cmul<conj_op><mode>3): New.
23683 * config/aarch64/aarch64-sve2.md (cml<fcmac1><conj_op><mode>4,
23684 cmul<conj_op><mode>3): New.
23686 2021-01-15 David Malcolm <dmalcolm@redhat.com>
23690 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
23691 Escape the tempfile name when constructing the expected output.
23693 2021-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23695 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlsl_hi<mode>):
23697 (aarch64_<su>mlsl_hi<mode>): ... This.
23698 (aarch64_<su>mlsl_hi<mode>): Define.
23699 (*aarch64_<su>mlsl<mode): Rename to...
23700 (aarch64_<su>mlsl<mode): ... This.
23701 * config/aarch64/aarch64-simd-builtins.def (smlsl, umlsl,
23702 smlsl_hi, umlsl_hi): Define builtins.
23703 * config/aarch64/arm_neon.h (vmlsl_high_s8, vmlsl_high_s16,
23704 vmlsl_high_s32, vmlsl_high_u8, vmlsl_high_u16, vmlsl_high_u32,
23705 vmlsl_s8, vmlsl_s16, vmlsl_s32, vmlsl_u8,
23706 vmlsl_u16, vmlsl_u32): Reimplement with builtins.
23708 2021-01-15 Uroš Bizjak <ubizjak@gmail.com>
23710 * config/i386/i386-c.c (ix86_target_macros):
23711 Use cpp_define_formatted for __SIZEOF_FLOAT80__ definition.
23713 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
23716 * config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
23717 * Makefile.in (RTL_SSA_H): New variable.
23718 * config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
23719 * config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
23720 * config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
23722 * config/aarch64/aarch64-cc-fusion.cc: New file.
23724 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
23726 * recog.h (insn_change_watermark::~insn_change_watermark): Avoid
23727 calling cancel_changes for changes that no longer exist.
23729 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
23731 * rtl-ssa/functions.h (function_info::ref_defs): Rename to...
23732 (function_info::reg_defs): ...this.
23733 * rtl-ssa/member-fns.inl (function_info::ref_defs): Rename to...
23734 (function_info::reg_defs): ...this.
23736 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23739 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
23741 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23744 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23747 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
23749 2021-01-15 Richard Biener <rguenther@suse.de>
23751 PR tree-optimization/96376
23752 * tree-vect-stmts.c (get_load_store_type): Disregard alignment
23753 for VMAT_INVARIANT.
23755 2021-01-15 Martin Liska <mliska@suse.cz>
23757 * doc/install.texi: Document that some tests need pytest module.
23758 * doc/sourcebuild.texi: Likewise.
23760 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23763 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
23765 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23767 * config/arm/mve.md (mve_vshrq_n_s<mode>_imm): New entry.
23768 (mve_vshrq_n_u<mode>_imm): Likewise.
23769 * config/arm/neon.md (vashr<mode>3, vlshr<mode>3): Move to ...
23770 * config/arm/vec-common.md: ... here.
23772 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23774 * config/arm/mve.md (mve_vshlq_<supf><mode>): Move to
23776 * config/arm/neon.md (vashl<mode>3): Delete.
23777 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): New.
23778 (vasl<mode>3): New expander.
23780 2021-01-15 Richard Biener <rguenther@suse.de>
23782 PR tree-optimization/98685
23783 * tree-vect-slp.c (vect_schedule_slp_node): Refactor handling
23784 of vector extern defs.
23786 2021-01-14 David Malcolm <dmalcolm@redhat.com>
23789 * diagnostic.c (diagnostic_kind_text): Break out this array
23791 (diagnostic_build_prefix): ...here.
23792 (fancy_abort): Detect when diagnostic_initialize has not yet been
23793 called and fall back to a minimal implementation of printing the
23794 ICE, rather than segfaulting in internal_error.
23796 2021-01-14 David Malcolm <dmalcolm@redhat.com>
23798 * diagnostic.c (diagnostic_initialize): Eliminate
23799 parseable_fixits_p in favor of initializing extra_output_kind from
23800 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
23801 (convert_column_unit): New function, split out from...
23802 (diagnostic_converted_column): ...this.
23803 (print_parseable_fixits): Add "column_unit" and "tabstop" params.
23804 Use them to call convert_column_unit on the column values.
23805 (diagnostic_report_diagnostic): Eliminate conditional on
23806 parseable_fixits_p in favor of a switch statement on
23807 extra_output_kind, passing the appropriate values to the new
23808 params of print_parseable_fixits.
23809 (selftest::test_print_parseable_fixits_none): Update for new
23810 params of print_parseable_fixits.
23811 (selftest::test_print_parseable_fixits_insert): Likewise.
23812 (selftest::test_print_parseable_fixits_remove): Likewise.
23813 (selftest::test_print_parseable_fixits_replace): Likewise.
23814 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
23816 (selftest::diagnostic_c_tests): Call it.
23817 * diagnostic.h (enum diagnostics_extra_output_kind): New.
23818 (diagnostic_context::parseable_fixits_p): Delete field in favor
23820 (diagnostic_context::extra_output_kind): ...this new field.
23821 * doc/invoke.texi (Environment Variables): Add
23822 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
23823 * opts.c (common_handle_option): Update handling of
23824 OPT_fdiagnostics_parseable_fixits for change to diagnostic_context
23827 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23829 * tree-vect-slp-patterns.c (class complex_operations_pattern,
23830 complex_operations_pattern::matches,
23831 complex_operations_pattern::recognize,
23832 complex_operations_pattern::build): New.
23833 (slp_patterns): Use it.
23835 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23837 * internal-fn.def (COMPLEX_FMS, COMPLEX_FMS_CONJ): New.
23838 * optabs.def (cmls_optab, cmls_conj_optab): New.
23839 * doc/md.texi: Document them.
23840 * tree-vect-slp-patterns.c (class complex_fms_pattern,
23841 complex_fms_pattern::matches, complex_fms_pattern::recognize,
23842 complex_fms_pattern::build): New.
23844 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23846 * internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
23847 * optabs.def (cmla_optab, cmla_conj_optab): New.
23848 * doc/md.texi: Document them.
23849 * tree-vect-slp-patterns.c (vect_match_call_p,
23850 class complex_fma_pattern, vect_slp_reset_pattern,
23851 complex_fma_pattern::matches, complex_fma_pattern::recognize,
23852 complex_fma_pattern::build): New.
23854 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23856 * internal-fn.def (COMPLEX_MUL, COMPLEX_MUL_CONJ): New.
23857 * optabs.def (cmul_optab, cmul_conj_optab): New.
23858 * doc/md.texi: Document them.
23859 * tree-vect-slp-patterns.c (vect_match_call_complex_mla,
23860 vect_normalize_conj_loc, is_eq_or_top, vect_validate_multiplication,
23861 vect_build_combine_node, class complex_mul_pattern,
23862 complex_mul_pattern::matches, complex_mul_pattern::recognize,
23863 complex_mul_pattern::build): New.
23865 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23867 * tree-vect-slp.c (optimize_load_redistribution_1): New.
23868 (optimize_load_redistribution, vect_is_slp_load_node): New.
23869 (vect_match_slp_patterns): Use it.
23871 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23873 * tree-vect-slp-patterns.c (complex_add_pattern::build):
23876 2021-01-14 Thomas Schwinge <thomas@codesourcery.com>
23878 * config/gcn/mkoffload.c (main): Create an offload image only in
23879 64-bit configurations.
23881 2021-01-14 H.J. Lu <hjl.tools@gmail.com>
23884 * config/i386/i386-options.c (ix86_option_override_internal):
23885 Issue an error for -fcf-protection with CF_BRANCH when compiling
23886 for 32-bit non-TARGET_CMOV targets.
23888 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
23891 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p):
23892 Remove declaration and initialization of shadow variable "ret".
23893 (ix86_option_override_internal): Remove delcaration of
23894 shadow variable "i". Redeclare shadowed variable to unsigned.
23895 * common/config/i386/i386-common.c (pta_size): Redeclare to unsigned.
23896 * config/i386/i386-builtins.c (get_builtin_code_for_version):
23897 Update for redeclaration.
23898 * config/i386/i386.h (pta_size): Ditto.
23900 2021-01-14 Richard Biener <rguenther@suse.de>
23902 PR tree-optimization/98674
23903 * tree-data-ref.c (base_supports_access_fn_components_p): New.
23904 (initialize_data_dependence_relation): For two bases without
23905 possible access fns resort to type size equality when determining
23906 shape compatibility.
23908 2021-01-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23911 * config/arm/arm_neon.h: Replace calls to __builtin_vcge* by
23912 <=, >= operators in vcle and vcge intrinsics respectively.
23913 * config/arm/arm_neon_builtins.def: Remove entry for
23916 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
23919 * config/i386/i386-options.c (ix86_function_specific_save):
23920 Remove redundant assignment to opts->x_ix86_branch_cost.
23921 * config/i386/i386.c (ix86_prefetch_sse):
23922 Rename from x86_prefetch_sse. Update all uses.
23923 * config/i386/i386.h: Update for rename.
23924 * config/i386/i386-options.h: Ditto.
23926 2021-01-14 Jakub Jelinek <jakub@redhat.com>
23929 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3,
23930 *sse4_1_zero_extendv4hiv4si2_3, *sse4_1_zero_extendv2siv2di2_3):
23931 Use Bm instead of m for non-avx. Add isa attribute.
23933 2021-01-14 Jakub Jelinek <jakub@redhat.com>
23935 PR tree-optimization/96688
23936 * match.pd (~(X >> Y) -> ~X >> Y): New simplification if
23937 ~X can be simplified.
23939 2021-01-14 Richard Sandiford <richard.sandiford@arm.com>
23941 * tree-vect-stmts.c (vect_model_load_cost): Account for unused
23942 IFN_LOAD_LANES results.
23944 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23946 * config/aarch64/aarch64-simd.md (aarch64_<su>xtl<mode>):
23948 (aarch64_xtn<mode>): Likewise.
23949 * config/aarch64/aarch64-simd-builtins.def (sxtl, uxtl, xtn):
23952 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
23954 (vmovl_s16): Likewise.
23955 (vmovl_s32): Likewise.
23956 (vmovl_u8): Likewise.
23957 (vmovl_u16): Likewise.
23958 (vmovl_u32): Likewise.
23959 (vmovn_s16): Likewise.
23960 (vmovn_s32): Likewise.
23961 (vmovn_s64): Likewise.
23962 (vmovn_u16): Likewise.
23963 (vmovn_u32): Likewise.
23964 (vmovn_u64): Likewise.
23966 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23968 * config/aarch64/aarch64-simd.md (aarch64_<su>qxtn2<mode>_le):
23970 (aarch64_<su>qxtn2<mode>_be): Likewise.
23971 (aarch64_<su>qxtn2<mode>): Likewise.
23972 * config/aarch64/aarch64-simd-builtins.def (sqxtn2, uqxtn2):
23974 * config/aarch64/iterators.md (SAT_TRUNC): Define code_iterator.
23975 (su): Handle ss_truncate and us_truncate.
23976 * config/aarch64/arm_neon.h (vqmovn_high_s16): Reimplement using
23978 (vqmovn_high_s32): Likewise.
23979 (vqmovn_high_s64): Likewise.
23980 (vqmovn_high_u16): Likewise.
23981 (vqmovn_high_u32): Likewise.
23982 (vqmovn_high_u64): Likewise.
23984 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23986 * config/aarch64/aarch64-simd.md (aarch64_xtn2<mode>_le):
23988 (aarch64_xtn2<mode>_be): Likewise.
23989 (aarch64_xtn2<mode>): Likewise.
23990 * config/aarch64/aarch64-simd-builtins.def (xtn2): Define
23992 * config/aarch64/arm_neon.h (vmovn_high_s16): Reimplement using
23994 (vmovn_high_s32): Likewise.
23995 (vmovn_high_s64): Likewise.
23996 (vmovn_high_u16): Likewise.
23997 (vmovn_high_u32): Likewise.
23998 (vmovn_high_u64): Likewise.
24000 2021-01-13 Stafford Horne <shorne@gmail.com>
24002 * config/or1k/or1k.h (ASM_PREFERRED_EH_DATA_FORMAT): New macro.
24004 2021-01-13 Stafford Horne <shorne@gmail.com>
24006 * config/or1k/linux.h (TARGET_ASM_FILE_END): Define macro.
24008 2021-01-13 Stafford Horne <shorne@gmail.com>
24010 * config/or1k/or1k.h (TARGET_CPU_CPP_BUILTINS): Add builtin
24011 define for __or1k_hard_float__.
24013 2021-01-13 Stafford Horne <shorne@gmail.com>
24015 * config/or1k/or1k.h (NO_PROFILE_COUNTERS): Define as 1.
24016 (PROFILE_HOOK): Define to call _mcount.
24017 (FUNCTION_PROFILER): Change from abort to no-op.
24019 2021-01-13 Jakub Jelinek <jakub@redhat.com>
24021 PR tree-optimization/96691
24022 * match.pd ((~X | C) ^ D -> (X | C) ^ (~D ^ C),
24023 (~X & C) ^ D -> (X & C) ^ (D ^ C)): New simplifications if
24024 (~D ^ C) or (D ^ C) can be simplified.
24026 2021-01-13 Richard Biener <rguenther@suse.de>
24028 PR tree-optimization/92645
24029 * match.pd (BIT_FIELD_REF to conversion): Delay canonicalization
24030 until after vector lowering.
24032 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
24034 * config/aarch64/aarch64-sve.md (fnma<mode>4): Extend from SVE_FULL_I
24036 (@aarch64_pred_fnma<mode>, cond_fnma<mode>, *cond_fnma<mode>_2)
24037 (*cond_fnma<mode>_4, *cond_fnma<mode>_any): Likewise.
24039 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
24041 * config/aarch64/aarch64-sve.md (fma<mode>4): Extend from SVE_FULL_I
24043 (@aarch64_pred_fma<mode>, cond_fma<mode>, *cond_fma<mode>_2)
24044 (*cond_fma<mode>_4, *cond_fma<mode>_any): Likewise.
24046 2021-01-13 Richard Biener <rguenther@suse.de>
24048 PR tree-optimization/92645
24049 * tree-vect-slp.c (vect_build_slp_tree_1): Relax supported
24050 BIT_FIELD_REF argument.
24051 (vect_build_slp_tree_2): Record the desired vector type
24052 on the external vector def.
24053 (vectorizable_slp_permutation): Handle required punning
24054 of existing vector defs.
24056 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
24058 * rtl-ssa/accesses.h (def_lookup): Fix order of comparison results.
24060 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
24062 * config/sh/sh.md (movsf_ie): Remove operands[2] test.
24064 2021-01-13 Samuel Thibault <samuel.thibault@ens-lyon.org>
24066 * config.gcc [$target == *-*-gnu*]: Enable
24067 'default_gnu_indirect_function'.
24069 2021-01-13 Jakub Jelinek <jakub@redhat.com>
24072 * optabs.c (expand_vec_perm_const): Don't force v0 and v1 into
24073 registers before calling targetm.vectorize.vec_perm_const, only after
24075 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const): Handle
24076 two argument permutation when one operand is zero vector and only
24077 after that force operands into registers.
24078 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_1): New
24079 define_insn_and_split pattern.
24080 (*avx512bw_zero_extendv32qiv32hi2_1): Likewise.
24081 (*avx512f_zero_extendv16hiv16si2_1): Likewise.
24082 (*avx2_zero_extendv8hiv8si2_1): Likewise.
24083 (*avx512f_zero_extendv8siv8di2_1): Likewise.
24084 (*avx2_zero_extendv4siv4di2_1): Likewise.
24085 * config/mips/mips.c (mips_vectorize_vec_perm_const): Force operands
24087 * config/arm/arm.c (arm_vectorize_vec_perm_const): Likewise.
24088 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Likewise.
24089 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const): Likewise.
24090 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const): Likewise.
24091 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const): Likewise.
24092 * config/gcn/gcn.c (gcn_vectorize_vec_perm_const): Likewise. Use std::swap.
24094 2021-01-13 Martin Liska <mliska@suse.cz>
24096 PR tree-optimization/98455
24097 * gimple-if-to-switch.cc (condition_info::record_phi_mapping):
24098 Record also virtual PHIs.
24099 (pass_if_to_switch::execute): Return TODO_cleanup_cfg only
24102 2021-01-13 Jonathan Wakely <jwakely@redhat.com>
24104 * doc/invoke.texi (C++ Modules): Fix typos.
24106 2021-01-13 Richard Biener <rguenther@suse.de>
24108 PR tree-optimization/98640
24109 * tree-ssa-sccvn.c (visit_nary_op): Do not try to
24110 handle plus or minus from a truncated operand to be
24113 2021-01-13 Jakub Jelinek <jakub@redhat.com>
24116 * config/i386/i386.md (*btr<mode>_1, *btr<mode>_2): New
24117 define_insn_and_split patterns.
24118 (splitter after *btr<mode>_2): New splitter.
24120 2021-01-13 Martin Liska <mliska@suse.cz>
24123 * cgraphunit.c (analyze_functions): Remove dead code.
24125 2021-01-13 Qian Jianhua <qianjh@cn.fujitsu.com>
24127 * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
24128 * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
24129 (a64fx_regmove_cost, a64fx_vector_cost): New.
24130 (a64fx_tunings): Use the new added cost tables.
24132 2021-01-13 Jakub Jelinek <jakub@redhat.com>
24135 * config/i386/predicates.md (pmovzx_parallel): New predicate.
24136 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): New
24137 define_insn_and_split pattern.
24138 (*sse4_1_zero_extendv4hiv4si2_3): Likewise.
24139 (*sse4_1_zero_extendv2siv2di2_3): Likewise.
24141 2021-01-13 Julian Brown <julian@codesourcery.com>
24143 * config/gcn/gcn.c (gcn_conditional_register_usage): Remove dead code
24144 to fix v0 register.
24146 2021-01-13 Julian Brown <julian@codesourcery.com>
24148 * config/gcn/gcn.c (gcn_md_reorg): Fix case where EXEC reg is live
24151 2021-01-13 Julian Brown <julian@codesourcery.com>
24153 * config/gcn/gcn-valu.md (recip<mode>2<exec>, recip<mode>2): Use unspec
24154 for reciprocal-approximation instructions.
24155 (div<mode>3): Use fused multiply-accumulate operations for reciprocal
24156 refinement and division result.
24157 * config/gcn/gcn.md (UNSPEC_RCP): New unspec constant.
24159 2021-01-13 Julian Brown <julian@codesourcery.com>
24161 * config/gcn/gcn-valu.md (subdf): Rename to...
24164 2021-01-12 Martin Liska <mliska@suse.cz>
24166 * gcov.c (source_info::debug): Fix printf format for 32-bit hosts.
24168 2021-01-12 Andrea Corallo <andrea.corallo@arm.com>
24170 * function-abi.h: Fix typo.
24172 2021-01-12 Christophe Lyon <christophe.lyon@linaro.org>
24176 * config/arm/arm.h (ARM_HAVE_NEON_V8QI_LDST): New macro.
24177 (ARM_HAVE_NEON_V16QI_LDST, ARM_HAVE_NEON_V4HI_LDST): Likewise.
24178 (ARM_HAVE_NEON_V8HI_LDST, ARM_HAVE_NEON_V2SI_LDST): Likewise.
24179 (ARM_HAVE_NEON_V4SI_LDST, ARM_HAVE_NEON_V4HF_LDST): Likewise.
24180 (ARM_HAVE_NEON_V8HF_LDST, ARM_HAVE_NEON_V4BF_LDST): Likewise.
24181 (ARM_HAVE_NEON_V8BF_LDST, ARM_HAVE_NEON_V2SF_LDST): Likewise.
24182 (ARM_HAVE_NEON_V4SF_LDST, ARM_HAVE_NEON_DI_LDST): Likewise.
24183 (ARM_HAVE_NEON_V2DI_LDST): Likewise.
24184 (ARM_HAVE_V8QI_LDST, ARM_HAVE_V16QI_LDST): Likewise.
24185 (ARM_HAVE_V4HI_LDST, ARM_HAVE_V8HI_LDST): Likewise.
24186 (ARM_HAVE_V2SI_LDST, ARM_HAVE_V4SI_LDST, ARM_HAVE_V4HF_LDST): Likewise.
24187 (ARM_HAVE_V8HF_LDST, ARM_HAVE_V4BF_LDST, ARM_HAVE_V8BF_LDST): Likewise.
24188 (ARM_HAVE_V2SF_LDST, ARM_HAVE_V4SF_LDST, ARM_HAVE_DI_LDST): Likewise.
24189 (ARM_HAVE_V2DI_LDST): Likewise.
24190 * config/arm/mve.md (*movmisalign<mode>_mve_store): New pattern.
24191 (*movmisalign<mode>_mve_load): New pattern.
24192 * config/arm/neon.md (movmisalign<mode>): Move to ...
24193 * config/arm/vec-common.md: ... here.
24195 2021-01-12 Vladimir N. Makarov <vmakarov@redhat.com>
24198 * lra-eliminations.c (eliminate_regs_in_insn): Add transformation
24199 of pattern 'plus (plus (hard reg, const), pseudo)'.
24201 2021-01-12 Richard Biener <rguenther@suse.de>
24203 PR tree-optimization/98550
24204 * tree-vect-slp.c (vect_record_max_nunits): Check whether
24205 the group size is a multiple of the vector element count.
24206 (vect_build_slp_tree_1): When we need to fail because
24207 the vector type choosen causes unrolling do so lazily
24208 without affecting matches only at the end to guide group splitting.
24210 2021-01-12 Martin Liska <mliska@suse.cz>
24213 * optc-save-gen.awk: Compare also n_target_save vars with
24216 2021-01-12 Martin Liska <mliska@suse.cz>
24218 * gcov.c (source_info::debug): New.
24219 (print_usage): Add --debug (-D) option.
24220 (process_args): Likewise.
24221 (generate_results): Call src->debug after
24222 accumulate_line_counts.
24223 (read_graph_file): Properly assign id for EXIT_BLOCK.
24224 * profile.c (branch_prob): Dump function body before it is
24227 2021-01-12 Jakub Jelinek <jakub@redhat.com>
24229 PR tree-optimization/98629
24230 * tree-ssa-math-opts.c (arith_overflow_check_p): Don't update use_stmt
24231 unless returning non-zero.
24233 2021-01-12 Jakub Jelinek <jakub@redhat.com>
24235 PR tree-optimization/95731
24236 * tree-ssa-reassoc.c (optimize_range_tests_cmp_bitwise): Also optimize
24237 x < 0 && y < 0 && z < 0 into (x | y | z) < 0 for signed x, y, z.
24238 (optimize_range_tests): Call optimize_range_tests_cmp_bitwise
24239 only after optimize_range_tests_var_bound.
24241 2021-01-12 Jakub Jelinek <jakub@redhat.com>
24243 * configure.ac: Ensure c/Make-lang.in comes first in @all_lang_makefrags@.
24244 * configure: Regenerated.
24246 2021-01-12 liuhongt <hongtao.liu@intel.com>
24249 * config/i386/i386-builtins.h (BUILTIN_DESC_SWAP_OPERANDS):
24251 * config/i386/i386-expand.c (ix86_expand_sse_comi): Delete
24254 2021-01-12 Alexandre Oliva <oliva@adacore.com>
24256 * ssa-iterators.h (end_imm_use_stmt_traverse): Forward
24258 (auto_end_imm_use_stmt_traverse): New struct.
24259 (FOR_EACH_IMM_USE_STMT): Use it.
24260 (BREAK_FROM_IMM_USE_STMT, RETURN_FROM_IMM_USE_STMT): Remove,
24262 * gimple-ssa-strength-reduction.c: ... here, ...
24263 * graphite-scop-detection.c: ... here, ...
24264 * ipa-modref.c, ipa-pure-const.c, ipa-sra.c: ... here, ...
24265 * tree-predcom.c, tree-ssa-ccp.c: ... here, ...
24266 * tree-ssa-dce.c, tree-ssa-dse.c: ... here, ...
24267 * tree-ssa-loop-ivopts.c, tree-ssa-math-opts.c: ... here, ...
24268 * tree-ssa-phiprop.c, tree-ssa.c: ... here, ...
24269 * tree-vect-slp.c: ... and here, ...
24270 * doc/tree-ssa.texi: ... and the example here.
24272 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24274 * config/aarch64/aarch64-sve.md (sdiv_pow2<mode>3): Extend from
24275 SVE_FULL_I to SVE_I. Generate an UNSPEC_PRED_X.
24276 (*sdiv_pow2<mode>3): New pattern.
24277 (@cond_<sve_int_op><mode>): Extend from SVE_FULL_I to SVE_I.
24278 Wrap the ASRD in an UNSPEC_PRED_X.
24279 (*cond_<sve_int_op><mode>_2): Likewise. Replace the UNSPEC_PRED_X
24280 predicate with a constant PTRUE, if it isn't already.
24281 (*cond_<sve_int_op><mode>_z): Replace with...
24282 (*cond_<sve_int_op><mode>_any): ...this new pattern.
24284 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24286 * config/aarch64/aarch64-sve.md (*cond_bic<mode>_2): Extend from
24287 SVE_FULL_I to SVE_I.
24288 (*cond_bic<mode>_any): Likewise.
24290 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24292 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart)
24293 (@aarch64_pred_<MUL_HIGHPART:optab><mode>): Extend from SVE_FULL_I
24296 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24298 * config/aarch64/aarch64-sve.md (<su>abd<mode>_3): Extend from
24299 SVE_FULL_I to SVE_I.
24300 (*aarch64_cond_<su>abd<mode>_2): Likewise.
24301 (*aarch64_cond_<su>abd<mode>_any): Likewise.
24302 (@aarch64_pred_<su>abd<mode>): Likewise. Use UNSPEC_PRED_X
24303 for the max and min but not for the minus.
24304 (*aarch64_cond_<su>abd<mode>_3): New pattern.
24306 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24308 * config/aarch64/iterators.md (SVE_24I): New iterator.
24309 * config/aarch64/aarch64-sve.md (*aarch64_adr<mode>_shift): Extend from
24310 SVE_FULL_SDI to SVE_24I. Use containers rather than elements.
24312 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24314 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_BINARY:optab><mode>)
24315 (*cond_<SVE_INT_BINARY:optab><mode>_2): Extend from SVE_FULL_I
24317 (*cond_<SVE_INT_BINARY:optab><mode>_3): Likewise.
24318 (*cond_<SVE_INT_BINARY:optab><mode>_any): Likewise.
24319 (*cond_<SVE_INT_BINARY:optab><mode>_2_const): Likewise.
24320 (*cond_<SVE_INT_BINARY:optab><mode>_any_const): Likewise.
24322 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24324 * config/aarch64/aarch64-sve.md (<SVE_INT_BINARY_IMM:optab><mode>3)
24325 (@aarch64_pred_<SVE_INT_BINARY_IMM:optab><mode>)
24326 (*post_ra_<SVE_INT_BINARY_IMM:optab><mode>3): Extend from SVE_FULL_I
24329 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24331 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3)
24332 (v<ASHIFT:optab><mode>3, @aarch64_pred_<optab><mode>)
24333 (*post_ra_v<ASHIFT:optab><mode>3): Extend from SVE_FULL_I to SVE_I.
24335 2021-01-11 Martin Liska <mliska@suse.cz>
24338 * symtab-clones.h (clone_info::release): Release
24339 symtab::m_clones with ggc_delete as it's a GGC memory.
24341 2021-01-11 Matthias Klose <doko@ubuntu.com>
24343 * Makefile.in (LINK_PROGRESS): Show the link target.
24345 2021-01-11 Richard Biener <rguenther@suse.de>
24347 PR tree-optimization/91403
24348 * tree-vect-data-refs.c (vect_analyze_group_access_1): Cap
24349 single-element interleaving group size at 4096 elements.
24351 2021-01-11 Richard Biener <rguenther@suse.de>
24353 PR tree-optimization/98526
24354 * tree-vect-loop.c (vect_model_reduction_cost): Remove costing
24355 of the actual reduction op for the regular case.
24356 (vectorizable_reduction): Cost the stmts
24357 vect_transform_reduction produces here.
24359 2021-01-11 Andreas Krebbel <krebbel@linux.ibm.com>
24361 * tree-ssa-forwprop.c (simplify_vector_constructor): For
24362 big-endian, use UNPACK[_FLOAT]_HI.
24364 2021-01-11 Tamar Christina <tamar.christina@arm.com>
24366 * tree-vect-slp-patterns.c (class complex_pattern,
24367 class complex_add_pattern): Add parameters to matches.
24368 (complex_add_pattern::build): Free memory.
24369 (complex_add_pattern::matches): Move validation end of match.
24370 (complex_add_pattern::recognize): Likewise.
24372 2021-01-11 Tamar Christina <tamar.christina@arm.com>
24374 * tree-vect-slp-patterns.c (linear_loads_p): Fix externals.
24376 2021-01-11 Tamar Christina <tamar.christina@arm.com>
24378 * tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
24380 2021-01-11 Jakub Jelinek <jakub@redhat.com>
24382 PR tree-optimization/95867
24383 * tree-ssa-math-opts.h: New header.
24384 * tree-ssa-math-opts.c: Include tree-ssa-math-opts.h.
24385 (powi_as_mults): No longer static. Use build_one_cst instead of
24386 build_real. Formatting fix.
24387 * tree-ssa-reassoc.c: Include tree-ssa-math-opts.h.
24388 (attempt_builtin_powi): Handle multiplication reassociation without
24389 powi_fndecl using powi_as_mults.
24390 (reassociate_bb): For integral types don't require
24391 -funsafe-math-optimizations to call attempt_builtin_powi.
24393 2021-01-11 Jakub Jelinek <jakub@redhat.com>
24395 PR tree-optimization/95852
24396 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): Change
24397 mul_stmts parameter type to vec<gimple *> &. Before cond_stmt
24398 allow in the bb any of the stmts in that vector, div_stmt and
24399 up to 3 cast stmts.
24400 (arith_cast_equal_p): New function.
24401 (arith_overflow_check_p): Add cast_stmt argument, handle signed
24402 multiply overflow checks.
24403 (match_arith_overflow): Adjust caller. Handle signed multiply
24406 2021-01-11 Jakub Jelinek <jakub@redhat.com>
24408 PR tree-optimization/95852
24409 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): New function.
24410 (uaddsub_overflow_check_p): Renamed to ...
24411 (arith_overflow_check_p): ... this. Handle also multiplication
24412 with overflow check.
24413 (match_uaddsub_overflow): Renamed to ...
24414 (match_arith_overflow): ... this. Add cfg_changed argument. Handle
24415 also multiplication with overflow check. Adjust function comment.
24416 (math_opts_dom_walker::after_dom_children): Adjust callers. Call
24417 match_arith_overflow also for MULT_EXPR.
24419 2021-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24421 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
24422 __builtin_convertvector.
24423 (vmovl_s16): Likewise.
24424 (vmovl_s32): Likewise.
24425 (vmovl_u8): Likewise.
24426 (vmovl_u16): Likewise.
24427 (vmovl_u32): Likewise.
24428 (vmovn_s16): Likewise.
24429 (vmovn_s32): Likewise.
24430 (vmovn_s64): Likewise.
24431 (vmovn_u16): Likewise.
24432 (vmovn_u32): Likewise.
24433 (vmovn_u64): Likewise.
24435 2021-01-11 Martin Liska <mliska@suse.cz>
24437 * gimple-if-to-switch.cc (struct condition_info): Use auto_var.
24438 (if_chain::is_beneficial): Delete clusters
24439 (find_conditions): Make second argument of conditions_in_bbs a
24440 pointer so that we control over it's lifetime.
24441 (pass_if_to_switch::execute): Delete them.
24443 2021-01-11 Kewen Lin <linkw@linux.ibm.com>
24445 * ira.c (move_unallocated_pseudos): Check other_reg and skip if
24448 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24450 * config/vax/vax.md (cc): Remove mode attribute.
24451 (subst_<cc>, subst_f<cc>): Rename to...
24452 (subst_<mode>, subst_f<VAXccnz:mode>): ... these respectively.
24453 (*cbranch<VAXint:mode>4_<VAXcc:mode>): Update for `cc' removal.
24454 (*cbranch<VAXfp:mode>4_<VAXccnz:mode>): Likewise.
24455 (*branch_<mode>, *branch_<mode>_reversed): Likewise.
24457 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24459 * config/vax/vax.md (subst_f<cc>): Add mode to operands and
24460 `const_double_zero'.
24462 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24464 * config/pdp11/pdp11.md (PDPfp): New mode iterator.
24465 (fcc_cc, fcc_ccnz): Use it. Add mode to `const_double_zero' and
24468 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24470 * genemit.c (gen_exp) <CONST_DOUBLE>: Handle `const_double_zero'
24472 * read-rtl.c (rtx_reader::read_rtx_code): Handle machine mode
24473 with `const_double_zero'.
24474 * doc/rtl.texi (Constant Expression Types): Document it.
24476 2021-01-09 Jakub Jelinek <jakub@redhat.com>
24479 * tree-cfg.c (verify_gimple_assign_binary): Allow lhs of
24480 POINTER_DIFF_EXPR to be any integral type.
24482 2021-01-09 Jakub Jelinek <jakub@redhat.com>
24484 PR rtl-optimization/98603
24485 * function.c (instantiate_virtual_regs_in_insn): For asm goto
24486 with impossible constraints, drop all SETs, CLOBBERs, drop PARALLEL
24487 if any, set ASM_OPERANDS mode to VOIDmode and change
24488 ASM_OPERANDS_OUTPUT_CONSTRAINT and ASM_OPERANDS_OUTPUT_IDX.
24490 2021-01-09 Alexandre Oliva <oliva@gnu.org>
24493 * final.c (notice_source_line): Narrow down the condition to
24494 skip a line-0 marker.
24496 2021-01-08 Sergei Trofimovich <siarheit@google.com>
24498 * ipa-modref.c (merge_call_side_effects): Fix
24499 linebreak split by reordering two print calls.
24501 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
24503 * config/s390/vector.md (*tf_to_fprx2_0): Rename from
24504 "*mov_tf_to_fprx2_0" for consistency, fix constraint.
24505 (*tf_to_fprx2_1): Rename from "*mov_tf_to_fprx2_1" for
24506 consistency, fix constraint.
24508 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
24510 * config/s390/s390-c.c (s390_def_or_undef_macro): Accept
24511 callables instead of mask values.
24512 (struct target_flag_set_p): New predicate.
24513 (s390_cpu_cpp_builtins_internal): Define or undefine
24514 __LONG_DOUBLE_VX__ macro.
24516 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
24519 * config/i386/i386.c (x86_function_profiler): Use R10 and R11
24520 to call mcount in large model with PIC for NO_PROFILE_COUNTERS
24523 2021-01-08 Richard Biener <rguenther@suse.de>
24525 * tree-ssa-sccvn.c (pass_fre::execute): Reset the SCEV hash table.
24527 2021-01-08 Richard Biener <rguenther@suse.de>
24529 * tree-vect-slp.c (scalar_stmts_to_slp_tree_map_t): Fix.
24530 (vect_build_slp_tree): On cache hit release the matched
24531 scalar stmts vector.
24532 * tree-vect-stmts.c (vectorizable_store): Properly free
24533 vec_oprnds before possibly gathering them again.
24535 2021-01-08 Richard Biener <rguenther@suse.de>
24537 PR tree-optimization/98544
24538 * tree-vect-slp.c (vect_optimize_slp): Always materialize
24539 permutes at a permute node.
24541 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
24544 * config/i386/i386.c (x86_function_profiler): Use R10 to call
24545 mcount in large model. Sorry for large model with PIC.
24547 2021-01-08 Jakub Jelinek <jakub@redhat.com>
24550 * config/i386/i386.opt (ix86_cmodel, ix86_incoming_stack_boundary_arg,
24551 ix86_pmode, ix86_preferred_stack_boundary_arg, ix86_regparm,
24552 ix86_veclibabi_type): Remove x_ prefix, use TargetVariable instead of
24553 TargetSave and initialize for variables with enum types.
24554 (mfentry, mstack-protector-guard-reg=, mstack-protector-guard-offset=,
24555 mstack-protector-guard-symbol=): Add Save.
24556 * config/i386/i386-options.c (ix86_function_specific_save,
24557 ix86_function_specific_restore): Don't save or restore x_ix86_cmodel,
24558 x_ix86_incoming_stack_boundary_arg, x_ix86_pmode,
24559 x_ix86_preferred_stack_boundary_arg, x_ix86_regparm,
24560 x_ix86_veclibabi_type.
24562 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
24564 * config/aarch64/aarch64-sve.md (*cnot<mode>): Extend from
24565 SVE_FULL_I to SVE_I.
24566 (*cond_cnot<mode>_2, *cond_cnot<mode>_any): Likewise.
24568 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
24570 * config/aarch64/aarch64-sve.md (*cond_uxt<mode>_2): Extend from
24571 SVE_FULL_I to SVE_I.
24572 (*cond_uxt<mode>_any): Likewise.
24574 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24576 * config/aarch64/iterators.md (Vwhalf): New iterator.
24577 * config/aarch64/aarch64-simd.md (aarch64_<sur>adalp<mode>_3):
24579 (aarch64_<sur>adalp<mode>): ... This. Make more
24581 (<sur>sadv16qi): Adjust callsite of the above.
24582 * config/aarch64/aarch64-simd-builtins.def (sadalp, uadalp): New
24584 * config/aarch64/arm_neon.h (vpadal_s8): Reimplement using
24586 (vpadal_s16): Likewise.
24587 (vpadal_u8): Likewise.
24588 (vpadal_u16): Likewise.
24589 (vpadalq_s8): Likewise.
24590 (vpadalq_s16): Likewise.
24591 (vpadalq_s32): Likewise.
24592 (vpadalq_u8): Likewise.
24593 (vpadalq_u16): Likewise.
24594 (vpadalq_u32): Likewise.
24596 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24598 * config/aarch64/aarch64-simd.md (aarch64_<su>abd<mode>_3):
24600 (aarch64_<su>abd<mode>): ... This.
24601 (<sur>sadv16qi): Adjust callsite of the above.
24602 * config/aarch64/aarch64-simd-builtins.def (sabd, uabd): Define
24604 * config/aarch64/arm_neon.h (vabd_s8): Reimplement using
24606 (vabd_s16): Likewise.
24607 (vabd_s32): Likewise.
24608 (vabd_u8): Likewise.
24609 (vabd_u16): Likewise.
24610 (vabd_u32): Likewise.
24611 (vabdq_s8): Likewise.
24612 (vabdq_s16): Likewise.
24613 (vabdq_s32): Likewise.
24614 (vabdq_u8): Likewise.
24615 (vabdq_u16): Likewise.
24616 (vabdq_u32): Likewise.
24618 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24620 * config/aarch64/aarch64-simd-builtins.def (saba, uaba): Define
24622 * config/aarch64/arm_neon.h (vaba_s8): Implement using builtin.
24623 (vaba_s16): Likewise.
24624 (vaba_s32): Likewise.
24625 (vaba_u8): Likewise.
24626 (vaba_u16): Likewise.
24627 (vaba_u32): Likewise.
24628 (vabaq_s8): Likewise.
24629 (vabaq_s16): Likewise.
24630 (vabaq_s32): Likewise.
24631 (vabaq_u8): Likewise.
24632 (vabaq_u16): Likewise.
24633 (vabaq_u32): Likewise.
24635 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24637 * config/aarch64/aarch64-simd.md (aba<mode>_3): Rename to...
24638 (aarch64_<su>aba<mode>): ... This. Handle uaba as well.
24639 Change RTL pattern to match.
24641 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
24643 * common/config/riscv/riscv-common.c (riscv_current_subset_list): New.
24644 * config/riscv/riscv-c.c (riscv-subset.h): New.
24645 (INCLUDE_STRING): Define.
24646 (riscv_cpu_cpp_builtins): Add new style architecture extension
24648 * config/riscv/riscv-subset.h (riscv_subset_list::begin): New.
24649 (riscv_subset_list::end): New.
24650 (riscv_current_subset_list): New.
24652 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
24654 * common/config/riscv/riscv-common.c (RISCV_DONT_CARE_VERSION):
24655 Move to riscv-subset.h.
24656 (struct riscv_subset_t): Ditto.
24657 (class riscv_subset_list): Ditto.
24658 * config/riscv/riscv-subset.h (RISCV_DONT_CARE_VERSION): Move
24659 from riscv-common.c.
24660 (struct riscv_subset_t): Ditto.
24661 (class riscv_subset_list): Ditto.
24662 * config/riscv/t-riscv ($(common_out_file)): Add file
24665 2021-01-07 Jakub Jelinek <jakub@redhat.com>
24668 * config/i386/i386.md (*bmi_blsi_<mode>_cmp, *bmi_blsi_<mode>_ccno):
24669 New define_insn patterns.
24671 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
24673 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_UNARY:optab><mode>)
24674 (*cond_<SVE_INT_UNARY:optab><mode>_2): Extend from SVE_FULL_I to SVE_I.
24675 (*cond_<SVE_INT_UNARY:optab><mode>_any): Likewise.
24677 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
24679 PR tree-optimization/98560
24680 * internal-fn.def (IFN_VCONDU, IFN_VCONDEQ): Use type vec_cond.
24681 * internal-fn.c (vec_cond_mask_direct): Get the data mode from
24683 (vec_cond_direct): Likewise argument 2.
24684 (vec_condu_direct, vec_condeq_direct): Delete.
24685 (expand_vect_cond_optab_fn): Rename to...
24686 (expand_vec_cond_optab_fn): ...this, replacing old macro.
24687 (expand_vec_condu_optab_fn, expand_vec_condeq_optab_fn): Delete.
24688 (expand_vect_cond_mask_optab_fn): Rename to...
24689 (expand_vec_cond_mask_optab_fn): ...this, replacing old macro.
24690 (direct_vec_cond_mask_optab_supported_p): Treat the optab as a
24692 (direct_vec_cond_optab_supported_p): Likewise.
24693 (direct_vec_condu_optab_supported_p): Delete.
24694 (direct_vec_condeq_optab_supported_p): Delete.
24695 * gimple-isel.cc: Include internal-fn.h.
24696 (gimple_expand_vec_cond_expr): Check that IFN_VCONDEQ is supported
24699 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
24701 PR tree-optimization/98560
24702 * gimple-isel.cc (gimple_expand_vec_cond_expr): If we fail to use
24703 IFN_VCOND{,U,EQ}, fall back on IFN_VCOND_MASK.
24705 2021-01-07 Uroš Bizjak <ubizjak@gmail.com>
24707 * config/i386/i386.md (insn): Merge from plusminus_insn, shift_insn,
24708 rotate_insn and optab code attributes.
24709 Update all uses to merged code attribute.
24710 * config/i386/sse.md: Update all uses to merged code attribute.
24711 * config/i386/mmx.md: Update all uses to merged code attribute.
24713 2021-01-07 Jakub Jelinek <jakub@redhat.com>
24715 PR tree-optimization/98568
24716 * gimple-ssa-store-merging.c (bswap_view_convert): New function.
24717 (bswap_replace): Use it.
24719 2021-01-06 Vladimir N. Makarov <vmakarov@redhat.com>
24721 PR rtl-optimization/97978
24722 * lra-int.h (lra_hard_reg_split_p): New external.
24723 * lra.c (lra_hard_reg_split_p): New global.
24724 (lra): Set up lra_hard_reg_split_p after splitting a hard reg.
24725 * lra-assigns.c (lra_assign): Don't check allocation correctness
24726 after hard reg splitting.
24728 2021-01-06 Martin Sebor <msebor@redhat.com>
24731 * builtins.c (new_delete_mismatch_p): New overload.
24732 (new_delete_mismatch_p (tree, tree)): Call it.
24734 2021-01-06 Alexandre Oliva <oliva@adacore.com>
24736 * Makefile.in (T_GLIMITS_H): New.
24737 (stmp-int-hdrs): Depend on it, use it.
24738 * config/t-vxworks (T_GLIMITS_H): Override it.
24739 (vxw-glimits.h): New.
24741 2021-01-06 Richard Biener <rguenther@suse.de>
24743 PR tree-optimization/98513
24744 * value-range.cc (intersect_ranges): Compare the upper bounds
24745 for the expected relation.
24747 2021-01-06 Gerald Pfeifer <gerald@pfeifer.com>
24750 2020-12-28 Gerald Pfeifer <gerald@pfeifer.com>
24752 * doc/standards.texi (HSAIL): Remove section.
24754 2021-01-05 Samuel Thibault <samuel.thibault@ens-lyon.org>
24756 * configure: Re-generate.
24758 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24760 * doc/invoke.texi (-std=c++20): Adjust for the publication of
24761 ISO 14882:2020 standard.
24762 * doc/standards.texi: Likewise.
24764 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24766 PR tree-optimization/94802
24767 * expr.h (maybe_optimize_sub_cmp_0): Declare.
24768 * expr.c: Include tree-pretty-print.h and flags.h.
24769 (maybe_optimize_sub_cmp_0): New function.
24770 (do_store_flag): Use it.
24771 * cfgexpand.c (expand_gimple_cond): Likewise.
24773 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24775 * mux-utils.h (pointer_mux::m_ptr): Tweak description of contents.
24776 * rtlanal.c (simple_regno_set): Tweak description to clarify the
24779 2021-01-05 Richard Biener <rguenther@suse.de>
24781 PR tree-optimization/98516
24782 * tree-vect-slp.c (vect_optimize_slp): Permute the incoming
24783 lanes when materializing on a VEC_PERM node.
24784 (vectorizable_slp_permutation): Dump the permute properly.
24786 2021-01-05 Richard Biener <rguenther@suse.de>
24788 * tree-vect-slp.c (vect_slp_region): Move debug counter
24789 to cover individual subgraphs.
24791 2021-01-05 Richard Biener <rguenther@suse.de>
24793 PR tree-optimization/98428
24794 * tree-vect-slp.c (vect_build_slp_tree_1): Properly reject
24795 vector lane extracts for loop vectorization.
24797 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24799 PR tree-optimization/98514
24800 * tree-ssa-reassoc.c (bb_rank): Change type from long * to
24802 (operand_rank): Change type from hash_map<tree, long> to
24803 hash_map<tree, int64_t>.
24804 (phi_rank): Change return type from long to int64_t.
24805 (loop_carried_phi): Change block_rank variable type from long to
24807 (propagate_rank): Change return type, rank parameter type and
24808 op_rank variable type from long to int64_t.
24809 (find_operand_rank): Change return type from long to int64_t
24810 and change slot variable type from long * to int64_t *.
24811 (insert_operand_rank): Change rank parameter type from long to
24813 (get_rank): Change return type and rank variable type from long to
24814 int64_t. Use PRId64 instead of ld to print the rank.
24815 (init_reassoc): Change rank variable type from long to int64_t
24816 and adjust correspondingly bb_rank and operand_rank initialization.
24818 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24820 PR tree-optimization/96928
24821 * tree-ssa-phiopt.c (xor_replacement): New function.
24822 (tree_ssa_phiopt_worker): Call it.
24824 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24826 PR tree-optimization/96930
24827 * match.pd ((A / (1 << B)) -> (A >> B)): If A is extended
24828 from narrower value which has the same type as 1 << B, perform
24829 the right shift on the narrower value followed by extension.
24831 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24833 PR tree-optimization/96239
24834 * gimple-ssa-store-merging.c (maybe_optimize_vector_constructor): New
24836 (get_status_for_store_merging): Don't return BB_INVALID for blocks
24837 with potential bswap optimizable CONSTRUCTORs.
24838 (pass_store_merging::execute): Optimize vector CONSTRUCTORs with bswap
24841 2021-01-05 Richard Biener <rguenther@suse.de>
24843 PR tree-optimization/98381
24844 * tree.c (vector_element_bits): Properly compute bool vector
24846 * tree-vect-loop.c (vectorizable_live_operation): Properly
24847 compute the last lane bit offset.
24849 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
24852 * config/i386/sse.md (sse_cvtps2pi): Redefine as define_insn_and_split.
24853 Clear the top 64 bytes of the input XMM register.
24854 (sse_cvttps2pi): Ditto.
24856 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
24859 * config/i386/xopintrin.h (_mm256_cmov_si256): New.
24861 2021-01-05 H.J. Lu <hjl.tools@gmail.com>
24864 * config/i386/xmmintrin.h (_mm_extract_pi16): Cast to unsigned
24867 2021-01-05 Claudiu Zissulescu <claziss@synopsys.com>
24869 * config/arc/arc.md (maddsidi4_split): Use ACC_REG_FIRST.
24870 (umaddsidi4_split): Likewise.
24872 2021-01-05 liuhongt <hongtao.liu@intel.com>
24875 * config/i386/sse.md (*sse2_pmovskb_zexthisi): New
24876 define_insn_and_split for zero_extend of subreg HI of pmovskb
24878 (*sse2_pmovskb_zexthisi): Add new combine splitters for
24879 zero_extend of not of subreg HI of pmovskb result.
24881 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24884 * explow.c (convert_memory_address_addr_space_1): Handle UNSPECs
24886 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate): Use
24887 convert_memory_address to convert symbolic immediates to ptr_mode
24888 before forcing them to memory.
24890 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24892 PR rtl-optimization/97144
24893 * recog.c (constrain_operands): Initialize matching_operand
24894 for each alternative, rather than only doing it once.
24896 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24898 PR rtl-optimization/98403
24899 * rtl-ssa/changes.cc (function_info::finalize_new_accesses): Explain
24900 why we don't remove call clobbers.
24901 (function_info::apply_changes_to_insn): Don't attempt to add
24902 call clobbers here.
24904 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24906 PR tree-optimization/98371
24907 * tree-vect-loop.c (vect_reanalyze_as_main_loop): New function.
24908 (vect_analyze_loop): If an epilogue loop appears to be cheaper
24909 than the main loop, re-analyze it as a main loop before adopting
24912 2021-01-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
24915 * configure.ac (NETLIBS): Determine using AX_LIB_SOCKET_NSL.
24916 * aclocal.m4, configure: Regenerate.
24917 * Makefile.in (NETLIBS): Define.
24918 (BACKEND): Remove $(CODYLIB).
24920 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24922 PR rtl-optimization/98334
24923 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
24924 Optimize (X - 1) * Y + Y to X * Y or (X + 1) * Y - Y to X * Y.
24926 2021-01-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
24928 * tree-inline.c (expand_call_inline): Restore input_location.
24929 Return result from recursive call.
24931 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
24933 PR tree-optimization/95401
24934 * config/aarch64/aarch64-sve-builtins.cc
24935 (gimple_folder::load_store_cookie): Use bits rather than bytes
24936 for the alignment argument to IFN_MASK_LOAD and IFN_MASK_STORE.
24937 * gimple-fold.c (gimple_fold_mask_load_store_mem_ref): Likewise.
24938 * tree-vect-stmts.c (vectorizable_store): Likewise.
24939 (vectorizable_load): Likewise.
24941 2021-01-04 Richard Biener <rguenther@suse.de>
24943 PR tree-optimization/98308
24944 * tree-vect-stmts.c (vectorizable_load): Set invariant mask
24947 2021-01-04 Jakub Jelinek <jakub@redhat.com>
24949 PR tree-optimization/95771
24950 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Handle types
24951 with precision smaller than int's precision and types with precision
24952 twice as large as long long. Formatting fixes.
24954 2021-01-04 Richard Biener <rguenther@suse.de>
24956 PR tree-optimization/98464
24957 * tree-ssa-sccvn.c (vn_valueize_for_srt): Rename from ...
24958 (vn_valueize_wrapper): ... this. Temporarily adjust vn_context_bb.
24959 (process_bb): Adjust.
24961 2021-01-04 Matthew Malcomson <matthew.malcomson@arm.com>
24964 * doc/invoke.texi (-fsanitize=address): Fix wording describing
24965 clash with -fsanitize=hwaddress.
24967 2021-01-04 Richard Biener <rguenther@suse.de>
24969 PR tree-optimization/98282
24970 * tree-ssa-sccvn.c (vn_get_stmt_kind): Classify tcc_reference on
24971 invariants as VN_NARY.
24973 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
24976 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Accept
24977 aarch64_simd_reg_or_zero for operand 2. Use the combinez patterns
24978 to handle zero operands.
24980 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
24982 * config/aarch64/aarch64.c (offset_6bit_signed_scaled_p): New function.
24983 (offset_6bit_unsigned_scaled_p): Fix typo in comment.
24984 (aarch64_sve_prefetch_operand_p): Accept MUL VLs in the range
24987 2021-01-04 Richard Biener <rguenther@suse.de>
24989 PR tree-optimization/98393
24990 * tree-vect-slp.c (vect_build_slp_tree): Properly zero matches
24991 when hitting the limit.
24993 2021-01-04 Richard Biener <rguenther@suse.de>
24995 PR tree-optimization/98291
24996 * tree-vect-loop.c (vectorizable_reduction): Bypass
24997 associativity check for SLP reductions with VF 1.
24999 2021-01-04 Jakub Jelinek <jakub@redhat.com>
25001 PR tree-optimization/96782
25002 * match.pd (x == ~x -> false, x != ~x -> true): New simplifications.
25004 2021-01-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
25006 * collect-utils.c (collect_execute): Check dumppfx.
25007 * collect2.c (maybe_run_lto_and_relink, do_link): Pass atsuffix
25008 to collect_execute.
25009 (do_link): Add new parameter atsuffix.
25010 (main): Handle -dumpdir option. Skip one argument for
25011 -o, -isystem and -B options.
25012 * gcc.c (make_at_file): New helper function.
25013 (close_at_file): Use it.
25015 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25017 * config/darwin.h (MIN_LD64_NO_COAL_SECTS): Adjust.
25018 Amend handling for LD64_VERSION fallback defaults.
25020 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25022 * config.gcc: Compute default version information
25023 from the configured target. Likewise defaults for
25025 * config/darwin10.h: Removed.
25026 * config/darwin12.h: Removed.
25027 * config/darwin9.h: Removed.
25028 * config/rs6000/darwin8.h: Removed.
25030 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25032 * config/darwin9.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
25034 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25036 * config/darwin9.h (STACK_CHECK_STATIC_BUILTIN): Move from here..
25037 * config/darwin.h (STACK_CHECK_STATIC_BUILTIN): .. to here.
25039 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25041 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move from
25043 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): ... to here.
25045 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25047 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move the spec
25048 for the Darwin10 unwinder stub from here ...
25049 * config/darwin.h (LINK_COMMAND_SPEC_A): ... to here.
25051 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25053 * config/darwin.h (DSYMUTIL_SPEC): Default to DWARF
25054 (ASM_DEBUG_SPEC):Only define if the assembler supports
25056 (PREFERRED_DEBUGGING_TYPE): Default to DWARF.
25057 (DARWIN_PREFER_DWARF): Define.
25058 * config/darwin9.h (PREFERRED_DEBUGGING_TYPE): Remove.
25059 (DARWIN_PREFER_DWARF): Likewise
25060 (DSYMUTIL_SPEC): Likewise.
25061 (COLLECT_RUN_DSYMUTIL): Likewise.
25062 (ASM_DEBUG_SPEC): Likewise.
25063 (ASM_DEBUG_OPTION_SPEC): Likewise.
25065 2021-01-02 Jan Hubicka <jh@suse.cz>
25067 * cfg.c (free_block): ggc_free bb.
25069 2021-01-01 Jakub Jelinek <jakub@redhat.com>
25071 * gcc.c (process_command): Update copyright notice dates.
25072 * gcov-dump.c (print_version): Ditto.
25073 * gcov.c (print_version): Ditto.
25074 * gcov-tool.c (print_version): Ditto.
25075 * gengtype.c (create_file): Ditto.
25076 * doc/cpp.texi: Bump @copying's copyright year.
25077 * doc/cppinternals.texi: Ditto.
25078 * doc/gcc.texi: Ditto.
25079 * doc/gccint.texi: Ditto.
25080 * doc/gcov.texi: Ditto.
25081 * doc/install.texi: Ditto.
25082 * doc/invoke.texi: Ditto.
25084 2021-01-01 Jakub Jelinek <jakub@redhat.com>
25086 * ChangeLog-2020: Rotate ChangeLog. New file.
25089 Copyright (C) 2021 Free Software Foundation, Inc.
25091 Copying and distribution of this file, with or without modification,
25092 are permitted in any medium without royalty provided the copyright
25093 notice and this notice are preserved.