1 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
3 * gimple-range-cache.cc (non_null_ref::adjust_range): Call new
5 * gimple-range-fold.cc (adjust_pointer_diff_expr): Ditto.
6 (adjust_imagpart_expr): Ditto.
7 * value-range.cc (irange::irange_intersect): Call new routine if
9 (irange::intersect): New wide_int version.
10 * value-range.h (class irange): New prototype.
12 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
14 * gimple-range-edge.cc (gimple_outgoing_range::gimple_outgoing_range):
15 Add parameter to limit size when recognizing switches.
16 (gimple_outgoing_range::edge_range_p): Check size limit.
17 * gimple-range-edge.h (gimple_outgoing_range): Add size field.
18 * gimple-range-gori.cc (gori_map::calculate_gori): Ignore switches
19 that exceed the size limit.
20 (gori_compute::gori_compute): Add initializer.
21 * params.opt (evrp-switch-limit): New.
22 * doc/invoke.texi: Update docs.
24 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
26 * value-range.h (irange::set_varying): Use TYPE_MIN_VALUE and
27 TYPE_MAX_VALUE instead of creating new trees when possible.
29 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
31 * gimple-range-cache.cc (non_null_ref::adjust_range): Check for
32 zero and non-zero more efficently.
34 2021-10-06 Richard Biener <rguenther@suse.de>
37 * dumpfile.h (TDF_GIMPLE_VAL): New.
38 (dump_flag): Re-order and adjust TDF_* flags. Make
39 the enum uint32_t. Use std::underlying_type in the
41 (optgroup_flag): Likewise for the operator overloads.
42 * tree-pretty-print.c (dump_generic_node): Wrap ADDR_EXPR
43 in _Literal if TDF_GIMPLE_VAL.
44 * gimple-pretty-print.c (dump_gimple_assign): Add
45 TDF_GIMPLE_VAL to flags when dumping operands where only
46 is_gimple_val are allowed.
47 (dump_gimple_cond): Likewise.
49 2021-10-06 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
51 * gimple-isel.cc (gimple_expand_vec_cond_expr): Remove redundant if
54 2021-10-05 qing zhao <qing.zhao@oracle.com>
57 * gimplify.c (gimplify_decl_expr): Not add initialization for an
58 auto variable when it has been initialized by frontend.
60 2021-10-05 Aldy Hernandez <aldyh@redhat.com>
62 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
65 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
67 * common/config/avr/avr-common.c (avr_handle_option): Mark
68 argument as ATTRIBUTE_UNUSED.
70 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
72 * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC):
73 Undefine before redefinition.
75 2021-10-05 Richard Biener <rguenther@suse.de>
77 * toplev.c (no_backend): Remove global var.
78 (process_options): Pass in no_backend, move post_options
79 langhook call to toplev::main.
80 (do_compile): Pass in no_backend, move process_options call
82 (toplev::run_self_tests): Check no_backend at the caller.
83 (toplev::main): Call post_options and process_options
84 split out from do_compile, do self-tests only if
85 no_backend is initialized.
87 2021-10-05 Richard Biener <rguenther@suse.de>
89 * tree-cfg.c (dump_function_to_file): Dump the UID of the
90 function as part of the name when requested.
91 * tree-pretty-print.c (dump_function_name): Dump the UID when
92 requested and the langhook produced the actual name.
94 2021-10-05 Richard Biener <rguenther@suse.de>
98 * internal-fn.c (expand_DEFERRED_INIT): Fall back to
99 zero-initialization as last resort, use the constant
100 size as given by the DEFERRED_INIT argument to build
103 2021-10-04 Marek Polacek <polacek@redhat.com>
106 * doc/invoke.texi: Document -Warray-compare.
108 2021-10-04 Richard Biener <rguenther@suse.de>
110 * gimplify.c (is_var_need_auto_init): DECL_HARD_REGISTER
111 variables are not to be initialized.
113 2021-10-04 Richard Biener <rguenther@suse.de>
115 * expr.h (non_mem_decl_p): Declare.
116 (mem_ref_refers_to_non_mem_p): Likewise.
117 * expr.c (non_mem_decl_p): Export.
118 (mem_ref_refers_to_non_mem_p): Likewise.
119 * internal-fn.c (expand_DEFERRED_INIT): Do not expand the LHS
120 but check the base with mem_ref_refers_to_non_mem_p
123 2021-10-04 Richard Biener <rguenther@suse.de>
125 PR tree-optimization/102570
126 * tree-ssa-sccvn.h (vn_reference_op_struct): Document
127 we are using clique for the internal function code.
128 * tree-ssa-sccvn.c (vn_reference_op_eq): Compare the
129 internal function code.
130 (print_vn_reference_ops): Print the internal function code.
131 (vn_reference_op_compute_hash): Hash it.
132 (copy_reference_ops_from_call): Record it.
133 (visit_stmt): Remove the restriction around internal function
135 (fully_constant_vn_reference_p): Use fold_const_call and handle
137 (vn_reference_eq): Compare call return types.
138 * tree-ssa-pre.c (create_expression_by_pieces): Handle
139 generating calls to internal functions.
140 (compute_avail): Remove the restriction around internal function
143 2021-10-04 Aldy Hernandez <aldyh@redhat.com>
145 PR tree-optimization/102560
146 * gimple-ssa-warn-alloca.c (alloca_call_type): Remove static
147 marker for invalid_range.
149 2021-10-04 Richard Biener <rguenther@suse.de>
152 * internal-fn.c (expand_DEFERRED_INIT): Guard register
153 initialization path an avoid initializing VLA registers
156 2021-10-04 Eric Botcazou <ebotcazou@adacore.com>
158 * config/rs6000/vxworks.h (TARGET_INIT_LIBFUNCS): Delete.
160 2021-10-03 Martin Liska <mliska@suse.cz>
162 * toplev.c (toplev::main): Check opt_index if it is a part
165 2021-10-02 Aldy Hernandez <aldyh@redhat.com>
167 PR tree-optimization/102563
168 * range-op.cc (operator_lshift::op1_range): Do not clobber
171 2021-10-02 Martin Liska <mliska@suse.cz>
173 * toplev.c (toplev::main): save_decoded_options[0] is program
174 name and so it should be skipped.
176 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
178 PR tree-optimization/102546
179 * range-op.cc (operator_lshift::op1_range): Teach range-ops that
180 X << Y is non-zero implies X is also non-zero.
182 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
184 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
186 * config/aarch64/aarch64-tune.md: Regenerate.
187 * doc/invoke.texi: Update docs.
189 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
191 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
193 * config/aarch64/aarch64-tune.md: Regenerate.
194 * doc/invoke.texi: Update docs.
196 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
198 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
200 * config/aarch64/aarch64-tune.md: Regenerate.
201 * doc/invoke.texi: Update docs.
203 2021-10-01 Martin Sebor <msebor@redhat.com>
206 * doc/invoke.texi (-Waddress): Update.
207 * gengtype.c (write_types): Avoid -Waddress.
208 * poly-int.h (POLY_SET_COEFF): Avoid using null.
210 2021-10-01 John David Anglin <danglin@gcc.gnu.org>
213 * config/pa/pa.c (pa_option_override): Default to dwarf version 4
216 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
218 * config/aarch64/aarch64.h (AARCH64_FL_V9): Update value.
220 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
222 * gimple-range-path.cc (path_range_query::compute_ranges): Use
224 * gimple-range-path.h (class path_range_query): Remove shadowed
226 (path_range_query::get_path_oracle): New.
228 2021-10-01 Jakub Jelinek <jakub@redhat.com>
229 Richard Biener <rguenther@suse.de>
232 * doc/invoke.texi (-fsanitize=integer-divide-by-zero): Remove
233 INT_MIN / -1 division detection from here ...
234 (-fsanitize=signed-integer-overflow): ... and add it here.
236 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
238 * config/aarch64/aarch64-arches.def (AARCH64_ARCH): Added
240 * config/aarch64/aarch64.h (AARCH64_FL_V9): New.
241 (AARCH64_FL_FOR_ARCH9): New flags for Armv9-A.
242 (AARCH64_ISA_V9): New ISA flag.
243 * doc/invoke.texi: Update docs.
245 2021-10-01 Martin Liska <mliska@suse.cz>
247 * toplev.c (toplev::main): Save decoded optimization options.
248 * toplev.h (save_opt_decoded_options): New.
249 * doc/extend.texi: Be more clear about optimize and target
252 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
254 * explow.c: Include langhooks.h.
255 (set_stack_check_libfunc): Build a proper function type.
257 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
260 * config/i386/i386.c (legitimate_pic_address_disp_p): For PE-COFF do
261 not return true for external weak function symbols in medium model.
263 2021-10-01 Jakub Jelinek <jakub@redhat.com>
265 * tree.h (OMP_CLAUSE_ORDER_REPRODUCIBLE): Define.
266 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_ORDER>: Print
267 reproducible: for OMP_CLAUSE_ORDER_REPRODUCIBLE.
268 * omp-general.c (omp_extract_for_data): If OMP_CLAUSE_ORDER is seen
269 without OMP_CLAUSE_ORDER_UNCONSTRAINED, overwrite sched_kind to
270 OMP_CLAUSE_SCHEDULE_STATIC.
272 2021-10-01 Richard Biener <rguenther@suse.de>
275 * tree-inline.c (setup_one_parameter): Avoid substituting
276 an invariant into contexts where a GIMPLE register is not valid.
278 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
280 * config/arm/arm-cpus.in: Add Cortex-R52+ CPU.
281 * config/arm/arm-tables.opt: Regenerate.
282 * config/arm/arm-tune.md: Regenerate.
283 * doc/invoke.texi: Update docs.
285 2021-09-30 Uroš Bizjak <ubizjak@gmail.com>
288 * config/i386/i386.md
289 (sign_extend:WIDE (any_logic:NARROW (memory, immediate)) splitters):
292 2021-09-30 Tobias Burnus <tobias@codesourcery.com>
294 * omp-low.c (omp_runtime_api_call): Add omp_aligned_{,c}alloc and
295 omp_{c,re}alloc, fix omp_alloc/omp_free.
297 2021-09-30 Martin Liska <mliska@suse.cz>
299 * defaults.h (ASM_OUTPUT_ASCII): Do not hide global variable
300 asm_out_file and stream directly to MYFILE.
302 2021-09-30 Richard Biener <rguenther@suse.de>
304 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
305 Restore and fix condition under which we apply npeel to
306 the DRs misalignment value.
308 2021-09-30 Richard Biener <rguenther@suse.de>
310 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
311 Fix npeel check for variable amount of peeling.
313 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
315 * lto-wrapper.c (run_gcc): Plug snprintf overflow.
317 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
319 * gimple-range.cc (gimple_ranger::debug): New.
320 * gimple-range.h (class gimple_ranger): Add debug.
322 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
325 * tree-vrp.c (hybrid_threader::~hybrid_threader): Free m_query.
327 2021-09-29 Indu Bhagat <indu.bhagat@oracle.com>
330 * btfout.c (GTY): Add GTY (()) albeit for cosmetic only purpose.
331 (btf_finalize): Empty the hash_map btf_var_ids.
333 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
335 * tree-vrp.c (thread_through_all_blocks): Return bool.
336 (execute_vrp_threader): Return TODO_* flags.
337 (pass_data_vrp_threader): Set todo_flags_finish to 0.
339 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
341 * timevar.def (TV_TREE_VRP_THREADER): New.
342 * tree-vrp.c: Use TV_TREE_VRP_THREADER for VRP threader pass.
344 2021-09-29 David Faust <david.faust@oracle.com>
346 * config.gcc (bpf-*-*): Do not overwrite extra_headers.
348 2021-09-29 Jonathan Wright <jonathan.wright@arm.com>
350 * config/aarch64/aarch64-builtins.c (TYPES_BINOP_PPU): Define
351 new type qualifier enum.
352 (TYPES_TERNOP_SSSU): Likewise.
353 (TYPES_TERNOP_PPPU): Likewise.
354 * config/aarch64/aarch64-simd-builtins.def: Define PPU, SSU,
355 PPPU and SSSU builtin generator macros for qtbl1 and qtbx1
357 * config/aarch64/arm_neon.h (vqtbl1_p8): Use type-qualified
358 builtin and remove casts.
359 (vqtbl1_s8): Likewise.
360 (vqtbl1q_p8): Likewise.
361 (vqtbl1q_s8): Likewise.
362 (vqtbx1_s8): Likewise.
363 (vqtbx1_p8): Likewise.
364 (vqtbx1q_s8): Likewise.
365 (vqtbx1q_p8): Likewise.
366 (vtbl1_p8): Likewise.
367 (vtbl2_p8): Likewise.
368 (vtbx2_p8): Likewise.
370 2021-09-29 Richard Biener <rguenther@suse.de>
372 * tree-vect-data-refs.c (vect_dr_misalign_for_aligned_access):
374 (vect_update_misalignment_for_peel): Use it to update
375 misaligned to the value necessary for an aligned access.
376 (vect_get_peeling_costs_all_drs): Likewise.
377 (vect_enhance_data_refs_alignment): Likewise.
379 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
381 * config/aarch64/aarch64.c (aarch64_expand_cpymem): Count number of
382 emitted operations and adjust heuristic for code size.
384 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
386 * config/aarch64/aarch64.c (aarch64_expand_setmem): Count number of
387 emitted operations and adjust heuristic for code size.
389 2021-09-29 Jakub Jelinek <jakub@redhat.com>
392 * gimplify.c (gimplify_scan_omp_clauses): Use omp_check_private even
393 in OMP_SCOPE clauses, not just on worksharing construct clauses.
395 2021-09-28 Geng Qi <gengqi@linux.alibaba.com>
397 * config/riscv/riscv.md (mulv<mode>4): Call gen_smul<mode>3_highpart.
398 (<u>mulditi3): Call <su>muldi3_highpart.
399 (<u>muldi3_highpart): Rename to <su>muldi3_highpart.
400 (<u>mulsidi3): Call <su>mulsi3_highpart.
401 (<u>mulsi3_highpart): Rename to <su>mulsi3_highpart.
403 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
405 * config/darwin.h (DSYMUTIL_SPEC): Recognize D sources.
407 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
409 * config/rs6000/darwin.h (FIXED_R13): Add for PPC64.
410 (FIRST_SAVED_GP_REGNO): Save from R13 even when it is one
413 2021-09-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
415 * config/aarch64/aarch64.h (AARCH64_FL_LS64): Define
416 (AARCH64_FL_V8_7): Likewise.
417 (AARCH64_FL_FOR_ARCH8_7): Likewise.
418 * config/aarch64/aarch64-arches.def (armv8.7-a): Define.
419 * config/aarch64/aarch64-option-extensions.def (ls64): Define.
420 * doc/invoke.texi: Document the above.
422 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
424 * dbgcnt.c (dbg_cnt_counter): New.
425 * dbgcnt.h (dbg_cnt_counter): New.
426 * dumpfile.c (dump_options): Add entry for TDF_THREADING.
427 * dumpfile.h (enum dump_flag): Add TDF_THREADING.
428 * gimple-range-path.cc (DEBUG_SOLVER): Use TDF_THREADING.
429 * tree-ssa-threadupdate.c (dump_jump_thread_path): Dump out
432 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
434 * cfgcleanup.c (pass_jump::execute): Check
435 flag_expensive_optimizations.
436 (pass_jump_after_combine::gate): Same.
437 * doc/invoke.texi (-fthread-jumps): Enable for -O1.
438 * opts.c (default_options_table): Enable -fthread-jumps at -O1.
439 * tree-ssa-threadupdate.c
440 (fwd_jt_path_registry::remove_jump_threads_including): Bail unless
443 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
445 * tree-ssa-reassoc.c (biased_names): New global.
446 (propagate_bias_p): New function.
447 (loop_carried_phi): Remove.
448 (propagate_rank): Propagate bias along single uses.
449 (get_rank): Update biased_names when needed.
451 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
453 * passes.def (pass_reassoc): Rename parameter to early_p.
454 * tree-ssa-reassoc.c (reassoc_bias_loop_carried_phi_ranks_p):
456 (phi_rank): Don't bias loop-carried phi ranks
457 before vectorization pass.
458 (execute_reassoc): Add bias_loop_carried_phi_ranks_p parameter.
459 (pass_reassoc::pass_reassoc): Add bias_loop_carried_phi_ranks_p
461 (pass_reassoc::set_param): Set bias_loop_carried_phi_ranks_p
463 (pass_reassoc::execute): Pass bias_loop_carried_phi_ranks_p to
465 (pass_reassoc::bias_loop_carried_phi_ranks_p): New member.
467 2021-09-28 Jakub Jelinek <jakub@redhat.com>
470 * config/i386/i386.c (standard_80387_constant_p): Don't recognize
471 special 80387 instruction XFmode constants if flag_rounding_math.
473 2021-09-28 Richard Biener <rguenther@suse.de>
475 PR tree-optimization/100112
476 * tree-ssa-sccvn.c (visit_reference_op_load): Record the
477 referece into the hashtable twice in case last_vuse is
478 different from the original vuse on the stmt.
480 2021-09-28 Jakub Jelinek <jakub@redhat.com>
483 * gimplify.c (gimplify_adjust_omp_clauses_1): Don't call the
484 omp_finish_clause langhook on implicitly added OMP_CLAUSE_PRIVATE
485 clauses on SIMD constructs.
487 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
489 PR tree-optimization/102511
490 * gimple-range-path.cc (path_range_query::range_on_path_entry):
491 Return VARYING when nothing found.
493 2021-09-28 Hongyu Wang <hongyu.wang@intel.com>
496 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): Add
498 (VALID_SSE2_REG_VHF_MODE): Add V4HFmode and V2HFmode.
499 (VALID_MMX_REG_MODE): Add V4HFmode.
500 (SSE_REG_MODE_P): Replace VALID_AVX512FP16_REG_MODE with
501 vector mode condition.
502 * config/i386/i386.c (classify_argument): Parse V4HF/V2HF
504 (function_arg_32): Add V4HFmode.
505 (function_arg_advance_32): Likewise.
506 * config/i386/i386.md (mode): Add V4HF/V2HF.
507 (MODE_SIZE): Likewise.
508 * config/i386/mmx.md (MMXMODE): Add V4HF mode.
509 (V_32): Add V2HF mode.
510 (VHF_32_64): New mode iterator.
511 (*mov<mode>_internal): Adjust sse alternatives to support
513 (*mov<mode>_internal): Adjust sse alternatives to support
515 (<insn><mode>3): New define_insn for add/sub/mul/div.
517 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
519 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Check
521 (pass_early_thread_jumps::gate): Same.
522 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
523 Return if !flag_thread_jumps.
524 * tree-ssa-threadupdate.c
525 (jt_path_registry::register_jump_thread): Assert that
526 flag_thread_jumps is true.
528 2021-09-28 liuhongt <hongtao.liu@intel.com>
531 (simplify_context::simplify_binary_operation_1): Relax
532 condition of simplifying (vec_concat:M (vec_select op0
533 index0)(vec_select op1 index1)) to allow different modes
534 between op0 and M, but have same inner mode.
536 2021-09-28 liuhongt <hongtao.liu@intel.com>
538 * config/i386/i386-expand.c (emit_reduc_half): Handle
539 V8HF/V16HF/V32HFmode.
540 * config/i386/sse.md (REDUC_SSE_PLUS_MODE): Add V8HF.
541 (REDUC_SSE_SMINMAX_MODE): Ditto.
542 (REDUC_PLUS_MODE): Add V16HF and V32HF.
543 (REDUC_SMINMAX_MODE): Ditto.
545 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
547 * gimple-range-path.cc
548 (path_range_query::precompute_ranges_in_block): Rename to...
549 (path_range_query::compute_ranges_in_block): ...this.
550 (path_range_query::precompute_ranges): Rename to...
551 (path_range_query::compute_ranges): ...this.
552 (path_range_query::precompute_relations): Rename to...
553 (path_range_query::compute_relations): ...this.
554 (path_range_query::precompute_phi_relations): Rename to...
555 (path_range_query::compute_phi_relations): ...this.
556 * gimple-range-path.h: Rename precompute* to compute*.
557 * tree-ssa-threadbackward.c
558 (back_threader::find_taken_edge_switch): Same.
559 (back_threader::find_taken_edge_cond): Same.
560 * tree-ssa-threadedge.c
561 (hybrid_jt_simplifier::compute_ranges_from_state): Same.
562 (hybrid_jt_state::register_equivs_stmt): Inline...
563 * tree-ssa-threadedge.h: ...here.
565 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
567 * tree-vrp.c (lhs_of_dominating_assert): Remove.
568 (class vrp_jt_state): Remove.
569 (class vrp_jt_simplifier): Remove.
570 (vrp_jt_simplifier::simplify): Remove.
571 (class vrp_jump_threader): Remove.
572 (vrp_jump_threader::vrp_jump_threader): Remove.
573 (vrp_jump_threader::~vrp_jump_threader): Remove.
574 (vrp_jump_threader::before_dom_children): Remove.
575 (vrp_jump_threader::after_dom_children): Remove.
577 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
579 * passes.def (pass_vrp_threader): New.
580 * tree-pass.h (make_pass_vrp_threader): Add make_pass_vrp_threader.
581 * tree-ssa-threadedge.c (hybrid_jt_state::register_equivs_stmt): New.
582 (hybrid_jt_simplifier::hybrid_jt_simplifier): New.
583 (hybrid_jt_simplifier::simplify): New.
584 (hybrid_jt_simplifier::compute_ranges_from_state): New.
585 * tree-ssa-threadedge.h (class hybrid_jt_state): New.
586 (class hybrid_jt_simplifier): New.
587 * tree-vrp.c (execute_vrp): Remove ASSERT_EXPR based jump
589 (class hybrid_threader): New.
590 (hybrid_threader::hybrid_threader): New.
591 (hybrid_threader::~hybrid_threader): New.
592 (hybrid_threader::before_dom_children): New.
593 (hybrid_threader::after_dom_children): New.
594 (execute_vrp_threader): New.
595 (class pass_vrp_threader): New.
596 (make_pass_vrp_threader): New.
598 2021-09-27 Martin Liska <mliska@suse.cz>
600 * output.h (enum section_flag): New.
601 (SECTION_FORGET): Remove.
602 (SECTION_ENTSIZE): Make it (1UL << 8) - 1.
603 (SECTION_STYLE_MASK): Define it based on other enum
605 * varasm.c (switch_to_section): Remove unused handling of
608 2021-09-27 Martin Liska <mliska@suse.cz>
610 * common.opt: Add new variable flag_default_complex_method.
611 * opts.c (finish_options): Handle flags related to
612 x_flag_complex_method.
613 * toplev.c (process_options): Remove option handling related
614 to flag_complex_method.
616 2021-09-27 Richard Biener <rguenther@suse.de>
619 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid using
620 type_for_size, instead use int_mode_for_size.
622 2021-09-27 Andrew Pinski <apinski@marvell.com>
625 * gimplify.c (gimplify_save_expr): Return early
626 if the type of val is error_mark_node.
628 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
630 * tree-ssanames.c (ssa_name_has_boolean_range): Use
633 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
635 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
637 * tree-ssa-dom.c (cprop_operand): Convert to range_query API.
638 (cprop_into_stmt): Same.
639 (dom_opt_dom_walker::optimize_stmt): Same.
641 2021-09-27 Richard Biener <rguenther@suse.de>
643 PR tree-optimization/97351
644 PR tree-optimization/97352
645 PR tree-optimization/82426
646 * tree-vectorizer.h (dr_misalignment): Add vector type
648 (aligned_access_p): Likewise.
649 (known_alignment_for_access_p): Likewise.
650 (vect_supportable_dr_alignment): Likewise.
651 (vect_known_alignment_in_bytes): Likewise. Refactor.
652 (DR_MISALIGNMENT): Remove.
653 (vect_update_shared_vectype): Likewise.
654 * tree-vect-data-refs.c (dr_misalignment): Refactor, handle
655 a vector type with larger alignment requirement and apply
656 the negative step adjustment here.
657 (vect_calculate_target_alignment): Remove.
658 (vect_compute_data_ref_alignment): Get explicit vector type
659 argument, do not apply a negative step alignment adjustment
661 (vect_slp_analyze_node_alignment): Re-analyze alignment
662 when we re-visit the DR with a bigger desired alignment but
663 keep more precise results from smaller alignments.
664 * tree-vect-slp.c (vect_update_shared_vectype): Remove.
665 (vect_slp_analyze_node_operations_1): Do not update the
666 shared vector type on stmts.
667 * tree-vect-stmts.c (vect_analyze_stmt): Push/pop the
668 vector type of an SLP node to the representative stmt-info.
669 (vect_transform_stmt): Likewise.
671 2021-09-27 liuhongt <hongtao.liu@intel.com>
674 2021-09-09 liuhongt <hongtao.liu@intel.com>
677 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
678 (reduc_plus_scal_v4sf): .. this, New define_expand.
679 (reduc_plus_scal_v2df): .. and this, New define_expand.
681 2021-09-26 liuhongt <hongtao.liu@intel.com>
683 * doc/extend.texi (Half-Precision): Remove storage only
684 description for _Float16 w/o avx512fp16.
686 2021-09-25 Dimitar Dimitrov <dimitar@dinux.eu>
688 * config/pru/constraints.md (Rrio): New constraint.
689 * config/pru/predicates.md (regio_operand): New predicate.
690 * config/pru/pru-pragma.c (pru_register_pragmas): Register
691 the __regio_symbol address space.
692 * config/pru/pru-protos.h (pru_symref2ioregno): Declaration.
693 * config/pru/pru.c (pru_symref2ioregno): New helper function.
694 (pru_legitimate_address_p): Remove.
695 (pru_addr_space_legitimate_address_p): Use the address space
697 (pru_nongeneric_pointer_addrspace): New helper function.
698 (pru_insert_attributes): New function to validate __regio_symbol
700 (TARGET_INSERT_ATTRIBUTES): New macro.
701 (TARGET_LEGITIMATE_ADDRESS_P): Remove.
702 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): New macro.
703 * config/pru/pru.h (enum reg_class): Add REGIO_REGS class.
704 * config/pru/pru.md (*regio_readsi): New pattern to read I/O
706 (*regio_nozext_writesi): New pattern to write to I/O registers.
707 (*regio_zext_write_r30<EQS0:mode>): Ditto.
708 * doc/extend.texi: Document the new PRU Named Address Space.
710 2021-09-24 Patrick Palka <ppalka@redhat.com>
714 * real.c (encode_ieee_double): Avoid unwanted sign extension.
715 (encode_ieee_quad): Likewise.
717 2021-09-24 Vladimir Makarov <vmakarov@redhat.com>
719 PR rtl-optimization/102147
720 * ira-build.c (ira_conflict_vector_profitable_p): Make
721 profitability calculation independent of host compiler pointer and
724 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
726 * gimple-range-path.cc (path_range_query::path_range_query):
727 Move debugging header...
728 (path_range_query::precompute_ranges): ...here.
729 (path_range_query::internal_range_of_expr): Do not call
730 range_on_path_entry if NAME is defined in the current block.
732 2021-09-24 Richard Biener <rguenther@suse.de>
734 * cfghooks.c (verify_flow_info): Verify unallocated BB and
735 edge flags are not set.
737 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
739 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
741 (jt_path_registry::register_jump_thread): Call
742 cancel_invalid_paths.
743 * tree-ssa-threadupdate.h (class jt_path_registry): Add
744 cancel_invalid_paths.
746 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
748 PR tree-optimization/102400
749 * tree-ssa-sccvn.c (vn_reference_insert_pieces): Initialize
750 result_vdef to zero value.
752 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
754 PR tree-optimization/102451
755 * tree-ssa-dse.c (delete_dead_or_redundant_call): Record bb of stmt
758 2021-09-24 Hongyu Wang <hongyu.wang@intel.com>
760 * config/i386/sse.md (cond_<insn><mode>): Extend to support
762 (cond_mul<mode>): Likewise.
763 (cond_div<mode>): Likewise.
764 (cond_<code><mode>): Likewise.
765 (cond_fma<mode>): Likewise.
766 (cond_fms<mode>): Likewise.
767 (cond_fnma<mode>): Likewise.
768 (cond_fnms<mode>): Likewise.
770 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
772 PR tree-optimization/102463
773 * gimple-range-fold.cc (fold_using_range::relation_fold_and_or): If
774 there is no range-ops handler, don't look for a relation.
776 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
778 * gimple-range-cache.cc (ranger_cache::ranger_cache): Take
779 non-executable_edge flag as parameter.
780 * gimple-range-cache.h (ranger_cache): Adjust prototype.
781 * gimple-range-gori.cc (gori_compute::gori_compute): Take
782 non-executable_edge flag as parameter.
783 (gori_compute::outgoing_edge_range_p): Check new flag.
784 * gimple-range-gori.h (gori_compute): Adjust prototype.
785 * gimple-range.cc (gimple_ranger::gimple_ranger): Create new flag.
786 (gimple_ranger::range_on_edge): Check new flag.
787 * gimple-range.h (gimple_ranger::non_executable_edge_flag): New.
788 * gimple-ssa-evrp.c (rvrp_folder): Pass ranger flag to simplifer.
789 (hybrid_folder::hybrid_folder): Set ranger non-executable flag value.
790 (hybrid_folder::fold_stmt): Set flag value in the simplifer.
791 * vr-values.c (simplify_using_ranges::set_and_propagate_unexecutable):
792 Use not_executable flag if provided inmstead of EDGE_EXECUTABLE.
793 (simplify_using_ranges::simplify_switch_using_ranges): Clear
794 EDGE_EXECUTABLE like it originally did.
795 (simplify_using_ranges::cleanup_edges_and_switches): Clear any
796 NON_EXECUTABLE flags.
797 (simplify_using_ranges::simplify_using_ranges): Adjust.
798 * vr-values.h (class simplify_using_ranges): Adjust.
799 (simplify_using_ranges::set_range_query): Add non-executable flag param.
801 2021-09-23 Bill Schmidt <wschmidt@linux.ibm.com>
804 * config/rs6000/rs6000-call.c (rs6000_aggregate_candidate): Detect
805 zero-width bit fields and return indicator.
806 (rs6000_discover_homogeneous_aggregate): Diagnose when the
807 presence of a zero-width bit field changes parameter passing in
810 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
812 * gimple-range-fold.cc (fold_using_range::range_of_phi):
813 Remove dominator check.
815 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
817 * gimple-range-path.cc (path_range_query::precompute_relations):
818 Hoist edge calculations before using EDGE_SUCC.
820 2021-09-23 Jonathan Wakely <jwakely@redhat.com>
822 * configure.ac: Fix --with-multilib-list description.
823 * configure: Regenerate.
825 2021-09-23 Richard Biener <rguenther@suse.de>
827 PR tree-optimization/102448
828 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
829 Clear alignment info copied from DR_PTR_INFO.
831 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
833 * config/i386/i386-expand.c (ix86_use_mask_cmp_p): Enable
835 * config/i386/sse.md (sseintvecmodelower): Add HF vector modes.
836 (<avx512>_store<mode>_mask): Extend to support HF vector modes.
837 (vec_cmp<mode><avx512fmaskmodelower>): Likewise.
838 (vcond_mask_<mode><avx512fmaskmodelower>): Likewise.
839 (vcond<mode><mode>): New expander.
840 (vcond<mode><sseintvecmodelower>): Likewise.
841 (vcond<sseintvecmodelower><mode>): Likewise.
842 (vcondu<mode><sseintvecmodelower>): Likewise.
844 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
846 * config/i386/sse.md (extend<ssePHmodelower><mode>2):
848 (extendv4hf<mode>2): Likewise.
849 (extendv2hfv2df2): Likewise.
850 (trunc<mode><ssePHmodelower>2): Likewise.
851 (avx512fp16_vcvt<castmode>2ph_<mode>): Rename to ...
852 (trunc<mode>v4hf2): ... this, and drop constraints.
853 (avx512fp16_vcvtpd2ph_v2df): Rename to ...
854 (truncv2dfv2hf2): ... this, and likewise.
856 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
858 * config/i386/sse.md (float<floatunssuffix><mode><ssePHmodelower>2):
860 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>):
862 (float<floatunssuffix><mode>v4hf2): ... this, and drop constraints.
863 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Rename to ...
864 (float<floatunssuffix>v2div2hf2): ... this, and likewise.
866 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
868 * config/i386/i386.md (fix<fixunssuffix>_trunchf<mode>2): New expander.
869 (fixuns_trunchfhi2): Likewise.
870 (*fixuns_trunchfsi2zext): New define_insn.
871 * config/i386/sse.md (ssePHmodelower): New mode_attr.
872 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
873 New expander for same element vector fix_truncate.
874 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
875 Likewise for V4HF to V4SI/V4DI fix_truncate.
876 (fix<fixunssuffix>_truncv2hfv2di2):
877 Likeise for V2HF to V2DI fix_truncate.
879 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
881 * config/i386/i386.md (<code>hf3): New expander.
883 2021-09-23 liuhongt <hongtao.liu@intel.com>
885 * config/i386/sse.md (FMAMODEM): extend to handle FP16.
886 (VFH_SF_AVX512VL): Extend to handle HFmode.
887 (VF_SF_AVX512VL): Deleted.
889 2021-09-23 liuhongt <hongtao.liu@intel.com>
891 * config/i386/i386.md (rinthf2): New expander.
892 (nearbyinthf2): New expander.
894 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
896 * tree-ssa-dom.c (class dom_jump_threader_simplifier): Rename...
897 (class dom_jt_state): ...this and provide virtual overrides.
898 (dom_jt_state::register_equiv): New.
899 (class dom_jt_simplifier): Rename from
900 dom_jump_threader_simplifier.
901 (dom_jump_threader_simplifier::simplify): Rename...
902 (dom_jt_simplifier::simplify): ...to this.
903 (pass_dominator::execute): Use dom_jt_simplifier and
905 * tree-ssa-threadedge.c (jump_threader::jump_threader):
907 (jt_state::register_equivs_stmt): Abstract out...
908 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
910 (jump_threader::thread_around_empty_blocks): Update state.
911 (jump_threader::thread_through_normal_block): Same.
912 (jt_state::jt_state): Remove.
913 (jt_state::push): Remove pass specific bits. Keep block vector
915 (jt_state::append_path): New.
916 (jt_state::pop): Remove pass specific bits.
917 (jt_state::register_equiv): Same.
918 (jt_state::record_ranges_from_stmt): Same.
919 (jt_state::register_equivs_on_edge): Same. Rename...
920 (jt_state::register_equivs_edge): ...to this.
921 (jt_state::dump): New.
922 (jt_state::debug): New.
923 (jump_threader_simplifier::simplify): Remove.
924 (jt_state::get_path): New.
925 * tree-ssa-threadedge.h (class jt_simplifier): Make into a base
926 class. Expose common functionality as virtual methods.
927 (class jump_threader_simplifier): Same. Rename...
928 (class jt_simplifier): ...to this.
929 * tree-vrp.c (class vrp_jump_threader_simplifier): Rename...
930 (class vrp_jt_simplifier): ...to this. Provide pass specific
932 (class vrp_jt_state): New.
933 (vrp_jump_threader_simplifier::simplify): Rename...
934 (vrp_jt_simplifier::simplify): ...to this. Inline code from
935 what used to be the base class.
936 (vrp_jump_threader::vrp_jump_threader): Use vrp_jt_state and
939 2021-09-22 Tobias Burnus <tobias@codesourcery.com>
942 * doc/invoke.texi (-Wno-missing-include-dirs.): Document Fortran
945 2021-09-22 Roger Sayle <roger@nextmovesoftware.com>
946 Richard Biener <rguenther@suse.de>
948 * match.pd (negation simplifications): Implement some negation
949 folding transformations from fold-const.c's fold_negate_expr.
950 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Add a SIMPLIFY
951 argument, to control whether the op should be simplified prior
952 to looking up/assigning a value number.
953 (vn_nary_build_or_lookup): Update call to vn_nary_build_or_lookup_1.
954 (vn_nary_simplify): Likewise.
955 (visit_nary_op): Likewise, but when constructing a NEGATE_EXPR
956 now call vn_nary_build_or_lookup_1 disabling simplification.
958 2021-09-22 Jiufu Guo <guojiufu@linux.ibm.com>
960 PR tree-optimization/102087
961 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
962 Update bound/cmp/control for niter.
964 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
966 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
967 Move check for non-empty BB here.
968 (fur_source::register_outgoing_edges): ...from here.
970 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
972 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
973 Remove call to improve_range_with_equivs.
974 (path_range_query::improve_range_with_equivs): Remove
975 * gimple-range-path.h: Remove improve_range_with_equivs.
977 2021-09-22 dianhong xu <dianhong.xu@intel.com>
979 * config/i386/avx512fp16intrin.h:
980 (_mm512_mask_blend_ph): New intrinsic.
981 (_mm512_permutex2var_ph): Ditto.
982 (_mm512_permutexvar_ph): Ditto.
983 * config/i386/avx512fp16vlintrin.h:
984 (_mm256_mask_blend_ph): New intrinsic.
985 (_mm256_permutex2var_ph): Ditto.
986 (_mm256_permutexvar_ph): Ditto.
987 (_mm_mask_blend_ph): Ditto.
988 (_mm_permutex2var_ph): Ditto.
989 (_mm_permutexvar_ph): Ditto.
991 2021-09-22 dianhong xu <dianhong.xu@intel.com>
993 * config/i386/avx512fp16intrin.h: Add new intrinsics.
994 (_mm512_conj_pch): New intrinsic.
995 (_mm512_mask_conj_pch): Ditto.
996 (_mm512_maskz_conj_pch): Ditto.
997 * config/i386/avx512fp16vlintrin.h: Add new intrinsics.
998 (_mm256_conj_pch): New intrinsic.
999 (_mm256_mask_conj_pch): Ditto.
1000 (_mm256_maskz_conj_pch): Ditto.
1001 (_mm_conj_pch): Ditto.
1002 (_mm_mask_conj_pch): Ditto.
1003 (_mm_maskz_conj_pch): Ditto.
1005 2021-09-22 dianhong xu <dianhong.xu@intel.com>
1007 * config/i386/avx512fp16intrin.h (_MM512_REDUCE_OP): New macro
1008 (_mm512_reduce_add_ph): New intrinsic.
1009 (_mm512_reduce_mul_ph): Ditto.
1010 (_mm512_reduce_min_ph): Ditto.
1011 (_mm512_reduce_max_ph): Ditto.
1012 * config/i386/avx512fp16vlintrin.h
1013 (_MM256_REDUCE_OP/_MM_REDUCE_OP): New macro.
1014 (_mm256_reduce_add_ph): New intrinsic.
1015 (_mm256_reduce_mul_ph): Ditto.
1016 (_mm256_reduce_min_ph): Ditto.
1017 (_mm256_reduce_max_ph): Ditto.
1018 (_mm_reduce_add_ph): Ditto.
1019 (_mm_reduce_mul_ph): Ditto.
1020 (_mm_reduce_min_ph): Ditto.
1021 (_mm_reduce_max_ph): Ditto.
1023 2021-09-22 dianhong xu <dianhong.xu@intel.com>
1025 * config/i386/avx512fp16intrin.h (__m512h_u, __m256h_u,
1026 __m128h_u): New typedef.
1027 (_mm512_load_ph): New intrinsic.
1028 (_mm256_load_ph): Ditto.
1029 (_mm_load_ph): Ditto.
1030 (_mm512_loadu_ph): Ditto.
1031 (_mm256_loadu_ph): Ditto.
1032 (_mm_loadu_ph): Ditto.
1033 (_mm512_store_ph): Ditto.
1034 (_mm256_store_ph): Ditto.
1035 (_mm_store_ph): Ditto.
1036 (_mm512_storeu_ph): Ditto.
1037 (_mm256_storeu_ph): Ditto.
1038 (_mm_storeu_ph): Ditto.
1039 (_mm512_abs_ph): Ditto.
1040 * config/i386/avx512fp16vlintrin.h
1041 (_mm_abs_ph): Ditto.
1042 (_mm256_abs_ph): Ditto.
1044 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
1046 * config/s390/tpf.md (prologue_tpf, epilogue_tpf): Add cc clobber.
1048 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
1051 * config/s390/s390.c (s390_expand_insv): Emit a normal move if it
1052 is actually a full copy of the source operand into the target.
1053 Don't emit a strict low part move if source and target mode match.
1055 2021-09-22 Jakub Jelinek <jakub@redhat.com>
1057 PR middle-end/102415
1058 * omp-expand.c (expand_omp_single): If region->exit is NULL,
1059 assert region->entry is GIMPLE_OMP_SCOPE region and return.
1061 2021-09-22 Jakub Jelinek <jakub@redhat.com>
1063 * tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
1064 * tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
1065 arguments from 2 to 3.
1066 * tree-pretty-print.c (dump_omp_clause): Print allocator() around
1067 allocate clause allocator and print align if present.
1068 * omp-low.c (scan_sharing_clauses): Force allocate_map entry even
1069 for omp_default_mem_alloc if align modifier is present. If align
1070 modifier is present, use TREE_LIST to encode both allocator and
1072 (lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
1073 Handle align modifier on allocator clause if present.
1075 2021-09-22 liuhongt <hongtao.liu@intel.com>
1077 * config/i386/i386.md (define_attr "isa"): Add
1079 (define_attr "enabled"): Correspond fma_or_avx512vl to
1080 TARGET_FMA || TARGET_AVX512VL.
1081 * config/i386/mmx.md (fmav2sf4): Extend to AVX512 fma.
1086 2021-09-22 liuhongt <hongtao.liu@intel.com>
1088 * config/i386/i386.md (cstorehf3): New define_expand.
1090 2021-09-22 liuhongt <hongtao.liu@intel.com>
1092 * config/i386/i386.md (<rounding_insn>hf2): New expander.
1093 (sse4_1_round<mode>2): Extend from MODEF to MODEFH.
1094 * config/i386/sse.md (*sse4_1_round<ssescalarmodesuffix>):
1095 Extend from VF_128 to VFH_128.
1097 2021-09-22 liuhongt <hongtao.liu@intel.com>
1099 * config/i386/i386-features.c (i386-features.c): Handle
1101 * config/i386/i386.md (sqrthf2): New expander.
1102 (*sqrthf2): New define_insn.
1103 * config/i386/sse.md
1104 (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
1107 2021-09-22 liuhongt <hongtao.liu@intel.com>
1109 * config/i386/avx512fp16intrin.h (_mm_mask_fcmadd_sch):
1111 (_mm_mask3_fcmadd_sch): Likewise.
1112 (_mm_maskz_fcmadd_sch): Likewise.
1113 (_mm_fcmadd_sch): Likewise.
1114 (_mm_mask_fmadd_sch): Likewise.
1115 (_mm_mask3_fmadd_sch): Likewise.
1116 (_mm_maskz_fmadd_sch): Likewise.
1117 (_mm_fmadd_sch): Likewise.
1118 (_mm_mask_fcmadd_round_sch): Likewise.
1119 (_mm_mask3_fcmadd_round_sch): Likewise.
1120 (_mm_maskz_fcmadd_round_sch): Likewise.
1121 (_mm_fcmadd_round_sch): Likewise.
1122 (_mm_mask_fmadd_round_sch): Likewise.
1123 (_mm_mask3_fmadd_round_sch): Likewise.
1124 (_mm_maskz_fmadd_round_sch): Likewise.
1125 (_mm_fmadd_round_sch): Likewise.
1126 (_mm_fcmul_sch): Likewise.
1127 (_mm_mask_fcmul_sch): Likewise.
1128 (_mm_maskz_fcmul_sch): Likewise.
1129 (_mm_fmul_sch): Likewise.
1130 (_mm_mask_fmul_sch): Likewise.
1131 (_mm_maskz_fmul_sch): Likewise.
1132 (_mm_fcmul_round_sch): Likewise.
1133 (_mm_mask_fcmul_round_sch): Likewise.
1134 (_mm_maskz_fcmul_round_sch): Likewise.
1135 (_mm_fmul_round_sch): Likewise.
1136 (_mm_mask_fmul_round_sch): Likewise.
1137 (_mm_maskz_fmul_round_sch): Likewise.
1138 * config/i386/i386-builtin.def: Add corresponding new builtins.
1139 * config/i386/sse.md
1140 (avx512fp16_fmaddcsh_v8hf_maskz<round_expand_name>): New expander.
1141 (avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>): Ditto.
1142 (avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>):
1144 (avx512fp16_<complexopname>sh_v8hf_mask<round_name>): Ditto.
1145 (avx512fp16_<complexopname>sh_v8hf<mask_scalarc_name><round_scalarcz_name>):
1147 * config/i386/subst.md (mask_scalarcz_name): New.
1148 (mask_scalarc_name): Ditto.
1149 (mask_scalarc_operand3): Ditto.
1150 (mask_scalarcz_operand4): Ditto.
1151 (round_scalarcz_name): Ditto.
1152 (round_scalarc_mask_operand3): Ditto.
1153 (round_scalarcz_mask_operand4): Ditto.
1154 (round_scalarc_mask_op3): Ditto.
1155 (round_scalarcz_mask_op4): Ditto.
1156 (round_scalarcz_constraint): Ditto.
1157 (round_scalarcz_nimm_predicate): Ditto.
1158 (mask_scalarcz): Ditto.
1159 (mask_scalarc): Ditto.
1160 (round_scalarcz): Ditto.
1162 2021-09-22 liuhongt <hongtao.liu@intel.com>
1164 * config/i386/avx512fp16intrin.h (_mm512_fcmadd_pch):
1166 (_mm512_mask_fcmadd_pch): Likewise.
1167 (_mm512_mask3_fcmadd_pch): Likewise.
1168 (_mm512_maskz_fcmadd_pch): Likewise.
1169 (_mm512_fmadd_pch): Likewise.
1170 (_mm512_mask_fmadd_pch): Likewise.
1171 (_mm512_mask3_fmadd_pch): Likewise.
1172 (_mm512_maskz_fmadd_pch): Likewise.
1173 (_mm512_fcmadd_round_pch): Likewise.
1174 (_mm512_mask_fcmadd_round_pch): Likewise.
1175 (_mm512_mask3_fcmadd_round_pch): Likewise.
1176 (_mm512_maskz_fcmadd_round_pch): Likewise.
1177 (_mm512_fmadd_round_pch): Likewise.
1178 (_mm512_mask_fmadd_round_pch): Likewise.
1179 (_mm512_mask3_fmadd_round_pch): Likewise.
1180 (_mm512_maskz_fmadd_round_pch): Likewise.
1181 (_mm512_fcmul_pch): Likewise.
1182 (_mm512_mask_fcmul_pch): Likewise.
1183 (_mm512_maskz_fcmul_pch): Likewise.
1184 (_mm512_fmul_pch): Likewise.
1185 (_mm512_mask_fmul_pch): Likewise.
1186 (_mm512_maskz_fmul_pch): Likewise.
1187 (_mm512_fcmul_round_pch): Likewise.
1188 (_mm512_mask_fcmul_round_pch): Likewise.
1189 (_mm512_maskz_fcmul_round_pch): Likewise.
1190 (_mm512_fmul_round_pch): Likewise.
1191 (_mm512_mask_fmul_round_pch): Likewise.
1192 (_mm512_maskz_fmul_round_pch): Likewise.
1193 * config/i386/avx512fp16vlintrin.h (_mm_fmadd_pch):
1195 (_mm_mask_fmadd_pch): Likewise.
1196 (_mm_mask3_fmadd_pch): Likewise.
1197 (_mm_maskz_fmadd_pch): Likewise.
1198 (_mm256_fmadd_pch): Likewise.
1199 (_mm256_mask_fmadd_pch): Likewise.
1200 (_mm256_mask3_fmadd_pch): Likewise.
1201 (_mm256_maskz_fmadd_pch): Likewise.
1202 (_mm_fcmadd_pch): Likewise.
1203 (_mm_mask_fcmadd_pch): Likewise.
1204 (_mm_mask3_fcmadd_pch): Likewise.
1205 (_mm_maskz_fcmadd_pch): Likewise.
1206 (_mm256_fcmadd_pch): Likewise.
1207 (_mm256_mask_fcmadd_pch): Likewise.
1208 (_mm256_mask3_fcmadd_pch): Likewise.
1209 (_mm256_maskz_fcmadd_pch): Likewise.
1210 (_mm_fmul_pch): Likewise.
1211 (_mm_mask_fmul_pch): Likewise.
1212 (_mm_maskz_fmul_pch): Likewise.
1213 (_mm256_fmul_pch): Likewise.
1214 (_mm256_mask_fmul_pch): Likewise.
1215 (_mm256_maskz_fmul_pch): Likewise.
1216 (_mm_fcmul_pch): Likewise.
1217 (_mm_mask_fcmul_pch): Likewise.
1218 (_mm_maskz_fcmul_pch): Likewise.
1219 (_mm256_fcmul_pch): Likewise.
1220 (_mm256_mask_fcmul_pch): Likewise.
1221 (_mm256_maskz_fcmul_pch): Likewise.
1222 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF_V8HF_V8HF,
1223 V8HF_FTYPE_V16HF_V16HF_V16HF, V16HF_FTYPE_V16HF_V16HF_V16HF_UQI,
1224 V32HF_FTYPE_V32HF_V32HF_V32HF_INT,
1225 V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT): Add new builtin types.
1226 * config/i386/i386-builtin.def: Add new builtins.
1227 * config/i386/i386-expand.c: Handle new builtin types.
1228 * config/i386/subst.md (SUBST_CV): New.
1229 (maskc_name): Ditto.
1230 (maskc_operand3): Ditto.
1232 (sdc_maskz_name): Ditto.
1233 (sdc_mask_op4): Ditto.
1234 (sdc_mask_op5): Ditto.
1235 (sdc_mask_mode512bit_condition): Ditto.
1237 (round_maskc_operand3): Ditto.
1238 (round_sdc_mask_operand4): Ditto.
1239 (round_maskc_op3): Ditto.
1240 (round_sdc_mask_op4): Ditto.
1241 (round_saeonly_sdc_mask_operand5): Ditto.
1242 * config/i386/sse.md (unspec): Add complex fma unspecs.
1243 (avx512fmaskcmode): New.
1244 (UNSPEC_COMPLEX_F_C_MA): Ditto.
1245 (UNSPEC_COMPLEX_F_C_MUL): Ditto.
1246 (complexopname): Ditto.
1247 (<avx512>_fmaddc_<mode>_maskz<round_expand_name>): New expander.
1248 (<avx512>_fcmaddc_<mode>_maskz<round_expand_name>): Ditto.
1249 (fma_<complexopname>_<mode><sdc_maskz_name><round_name>): New
1251 (<avx512>_<complexopname>_<mode>_mask<round_name>): Ditto.
1252 (<avx512>_<complexopname>_<mode><maskc_name><round_name>): Ditto.
1254 2021-09-22 Kewen Lin <linkw@linux.ibm.com>
1256 * config/rs6000/rs6000.opt (rs6000-density-pct-threshold,
1257 rs6000-density-size-threshold, rs6000-density-penalty,
1258 rs6000-density-load-pct-threshold,
1259 rs6000-density-load-num-threshold): New parameter.
1260 * config/rs6000/rs6000.c (rs6000_density_test): Adjust with
1261 corresponding parameters.
1263 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1265 * gimple-range-path.cc (path_range_query::defined_outside_path):
1267 (path_range_query::range_on_path_entry): New.
1268 (path_range_query::internal_range_of_expr): Resolve unknowns
1270 (path_range_query::improve_range_with_equivs): New.
1271 (path_range_query::ssa_range_in_phi): Resolve unknowns with
1273 * gimple-range-path.h (class path_range_query): Add
1274 defined_outside_path, range_on_path_entry, and
1275 improve_range_with_equivs.
1277 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1279 * gimple-range-path.cc (path_range_query::add_to_imports): New.
1280 (path_range_query::add_copies_to_imports): New.
1281 (path_range_query::precompute_ranges): Call
1282 add_copies_to_imports.
1283 * gimple-range-path.h (class path_range_query): Add prototypes
1284 for add_copies_to_imports and add_to_imports.
1286 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1288 * gimple-range-path.cc (path_range_query::range_defined_in_block):
1289 Remove useless code.
1291 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1293 * gimple-range-fold.h (class fur_source): Make oracle protected.
1294 * gimple-range-path.cc (path_range_query::path_range_query): Add
1295 resolve argument. Initialize oracle.
1296 (path_range_query::~path_range_query): Delete oracle.
1297 (path_range_query::range_of_stmt): Adapt to use relations.
1298 (path_range_query::precompute_ranges): Pre-compute relations.
1299 (class jt_fur_source): New
1300 (jt_fur_source::jt_fur_source): New.
1301 (jt_fur_source::register_relation): New.
1302 (jt_fur_source::query_relation): New.
1303 (path_range_query::precompute_relations): New.
1304 (path_range_query::precompute_phi_relations): New.
1305 * gimple-range-path.h (path_range_query): Add resolve argument.
1306 Add oracle, precompute_relations, precompute_phi_relations.
1307 * tree-ssa-threadbackward.c (back_threader::back_threader): Pass
1308 resolve argument to solver.
1310 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1312 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
1313 Rename postfold_gcond_edges to register_outgoing_edges and
1315 (fold_using_range::postfold_gcond_edges): Rename...
1316 (fur_source::register_outgoing_edges): ...to this.
1317 * gimple-range-fold.h (postfold_gcond_edges): Rename to
1318 register_outgoing_edges and move to fur_source.
1320 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1322 * gimple-range-fold.cc (fold_using_range::range_of_phi): Check
1323 dom_info_available_p.
1325 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
1327 * gimple-range-cache.cc (non_null_ref::non_null_ref): Use create
1328 and quick_grow_cleared instead of safe_grow_cleared.
1330 2021-09-21 Thomas Schwinge <thomas@codesourcery.com>
1333 * omp-oacc-neuter-broadcast.cc (oacc_do_neutering): Evaluate
1336 2021-09-21 Richard Earnshaw <rearnsha@arm.com>
1338 * configure.ac: Detect when the assembler supports new-style
1339 architecture extensions.
1340 * common/config/arm/arm-common.c (arm_rewrite_mcpu): Return
1341 the full CPU string if the assembler can grok it.
1342 (arm_rewrite_march): Likewise but for the architecture.
1343 * config.in: Regenerate.
1344 * configure: Regenerate.
1346 2021-09-21 Richard Biener <rguenther@suse.de>
1348 PR tree-optimization/102421
1349 * tree-vect-loop.c (vect_dissolve_slp_only_groups): Copy and
1350 adjust alignment info.
1352 2021-09-21 Kewen Lin <linkw@linux.ibm.com>
1354 * ipa-fnsummary.c (ipa_fn_summary_write): Remove inconsistent
1355 bitfield stream out.
1357 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
1359 * gimple-range-fold.cc (fold_using_range::range_of_phi): Ignore
1360 undefined edges, apply an equivalence if appropriate.
1361 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Return
1362 UNDEFINED if EDGE_EXECUTABLE is not set.
1363 * gimple-range.cc (gimple_ranger::gimple_ranger): Set all edges
1364 as EXECUTABLE upon startup.
1365 (gimple_ranger::range_on_edge): Return UNDEFINED for edges without
1366 EDGE_EXECUTABLE set.
1367 * vr-values.c (set_and_propagate_unexecutable): New.
1368 (simplify_using_ranges::fold_cond): Call set_and_propagate.
1369 (simplify_using_ranges::simplify_switch_using_ranges): Ditto.
1370 * vr-values.h: Add prototype.
1372 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
1374 * value-relation.cc (equiv_oracle::register_initial_def): New.
1375 (equiv_oracle::register_relation): Call register_initial_def.
1376 (equiv_oracle::add_equiv_to_block): New. Split register_relation.
1377 (relation_oracle::register_stmt): Check def block of PHI arguments.
1378 * value-relation.h (equiv_oracle): Add new prototypes.
1380 2021-09-20 Matthias Kretz <m.kretz@gsi.de>
1382 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
1383 Define __RECIPROCAL_MATH__, __NO_SIGNED_ZEROS__,
1384 __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__, and
1385 __ROUNDING_MATH__ according to their corresponding flags.
1386 * doc/cpp.texi: Document __RECIPROCAL_MATH__,
1387 __NO_SIGNED_ZEROS__, __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__,
1388 and __ROUNDING_MATH__.
1390 2021-09-20 Richard Biener <rguenther@suse.de>
1392 * tree-vect-stmts.c (vectorizable_load): Use the vectype
1395 2021-09-20 Richard Biener <rguenther@suse.de>
1397 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
1398 Do not compute alignment of the vectorized access here.
1400 2021-09-20 Richard Biener <rguenther@suse.de>
1402 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
1403 Store -1 for runtime alias peeling iterations.
1405 2021-09-20 Richard Biener <rguenther@suse.de>
1407 * config.gcc: Obsolete hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11*.
1409 2021-09-20 Thomas Schwinge <thomas@codesourcery.com>
1411 * input.c (string_concat_db::record_string_concatenation)
1412 (string_concat_db::get_string_concatenation): Skip for
1413 'RESERVED_LOCATION_P'.
1415 2021-09-20 Richard Biener <rguenther@suse.de>
1417 PR tree-optimization/65206
1418 * tree-data-ref.h (struct data_reference): Add alt_indices,
1420 * tree-data-ref.c (free_data_ref): Release alt_indices.
1421 (dr_analyze_indices): Work on struct indices and get DR_REF as tree.
1422 (create_data_ref): Adjust.
1423 (initialize_data_dependence_relation): Split into head
1424 and tail. When the base objects fail to match up try
1425 again with pointer-based analysis of indices.
1426 * tree-vectorizer.c (vec_info_shared::check_datarefs): Do
1427 not compare the lazily computed alternate set of indices.
1429 2021-09-20 Iain Sandoe <iain@sandoe.co.uk>
1431 * gcc.c: Test for execute OK when we find the
1432 programs for assembler linker and dsymutil and those
1433 were specified at configure-time.
1435 2021-09-19 Martin Sebor <msebor@redhat.com>
1437 PR middle-end/102403
1438 * gimple-predicate-analysis.cc (predicate::init_from_control_deps):
1439 Correct a function pre/postcondition.
1441 2021-09-19 Martin Sebor <msebor@redhat.com>
1443 PR middle-end/102243
1444 * tree-ssa-strlen.c (get_range): Handle null cfun.
1446 2021-09-19 Iain Sandoe <iain@sandoe.co.uk>
1448 * config/darwin.h (LINK_COMMAND_SPEC_A): Use Darwin10
1449 unwinder shim as a convenience library.
1451 2021-09-19 Andrew Pinski <apinski@marvell.com>
1453 * doc/install.texi: Add note about
1454 binutils 2.35 is required for LTO usage.
1456 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
1458 * tree-ssa-threadbackward.c
1459 (back_threader_registry::register_path): Use push_edge.
1460 * tree-ssa-threadedge.c
1461 (jump_threader::thread_around_empty_blocks): Same.
1462 (jump_threader::thread_through_normal_block): Same.
1463 (jump_threader::thread_across_edge): Same. Also, use auto_bitmap.
1465 * tree-ssa-threadupdate.c
1466 (jt_path_registry::allocate_thread_edge): Remove.
1467 (jt_path_registry::push_edge): New.
1468 (dump_jump_thread_path): Make static.
1469 * tree-ssa-threadupdate.h (allocate_thread_edge): Remove.
1472 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
1474 * gimple-range-path.cc (path_range_query::path_range_query): Add
1476 (path_range_query::dump): Remove extern declaration of dump_ranger.
1477 * gimple-range-trace.cc (dump_ranger): Add DEBUG_FUNCTION marker.
1478 * gimple-range-trace.h (dump_ranger): Add prototype.
1480 2021-09-19 John Ericson <git@JohnEricson.me>
1482 * gcc.c (find_a_program): New function, factored out of...
1483 (find_a_file): Here.
1484 (execute): Use find_a_program when looking for programs rather
1487 2021-09-19 Matwey V. Kornilov <matwey.kornilov@gmail.com>
1489 * config/avr/avr-mcus.def: Add atmega324pb.
1490 * doc/avr-mmcu.texi: Corresponding changes.
1492 2021-09-19 Roger Sayle <roger@nextmovesoftware.com>
1495 * match.pd (cmp @0 REAL_CST@1): When @0 is also REAL_CST, apply
1496 the same transformations as to @1. For comparisons against NaN,
1497 don't check HONOR_SNANS but confirm that neither operand is a
1500 2021-09-19 Benjamin Peterson <benjamin@locrian.net>
1502 * attribs.c (make_unique_name): Delete.
1503 * attribs.h (make_unique_name): Delete.
1505 2021-09-19 Andrew Pinski <apinski@marvell.com>
1507 * lra-constraints.c (check_and_process_move): Assert
1508 that dclass and sclass are greater than or equal to NO_REGS.
1510 2021-09-18 Jakub Jelinek <jakub@redhat.com>
1512 * tree.h (OMP_CLAUSE_ORDER_UNCONSTRAINED): Define.
1513 * tree-pretty-print.c (dump_omp_clause): Print unconstrained:
1514 for OMP_CLAUSE_ORDER_UNCONSTRAINED.
1516 2021-09-18 liuhongt <hongtao.liu@intel.com>
1518 * config/i386/i386-features.c (remove_partial_avx_dependency):
1519 Restrict TARGET_USE_VECTOR_FP_CONVERTS and
1520 TARGET_USE_VECTOR_CONVERTS to conversion instructions only.
1522 2021-09-18 Jakub Jelinek <jakub@redhat.com>
1524 * gimplify.c (omp_default_clause): For C/C++ default({,first}private),
1525 if file/namespace scope variable doesn't have predetermined sharing,
1526 treat it as if there was default(none).
1528 2021-09-18 liuhongt <hongtao.liu@intel.com>
1530 * config/i386/avx512fp16intrin.h (_mm_fmadd_sh):
1532 (_mm_mask_fmadd_sh): Likewise.
1533 (_mm_mask3_fmadd_sh): Likewise.
1534 (_mm_maskz_fmadd_sh): Likewise.
1535 (_mm_fmadd_round_sh): Likewise.
1536 (_mm_mask_fmadd_round_sh): Likewise.
1537 (_mm_mask3_fmadd_round_sh): Likewise.
1538 (_mm_maskz_fmadd_round_sh): Likewise.
1539 (_mm_fnmadd_sh): Likewise.
1540 (_mm_mask_fnmadd_sh): Likewise.
1541 (_mm_mask3_fnmadd_sh): Likewise.
1542 (_mm_maskz_fnmadd_sh): Likewise.
1543 (_mm_fnmadd_round_sh): Likewise.
1544 (_mm_mask_fnmadd_round_sh): Likewise.
1545 (_mm_mask3_fnmadd_round_sh): Likewise.
1546 (_mm_maskz_fnmadd_round_sh): Likewise.
1547 (_mm_fmsub_sh): Likewise.
1548 (_mm_mask_fmsub_sh): Likewise.
1549 (_mm_mask3_fmsub_sh): Likewise.
1550 (_mm_maskz_fmsub_sh): Likewise.
1551 (_mm_fmsub_round_sh): Likewise.
1552 (_mm_mask_fmsub_round_sh): Likewise.
1553 (_mm_mask3_fmsub_round_sh): Likewise.
1554 (_mm_maskz_fmsub_round_sh): Likewise.
1555 (_mm_fnmsub_sh): Likewise.
1556 (_mm_mask_fnmsub_sh): Likewise.
1557 (_mm_mask3_fnmsub_sh): Likewise.
1558 (_mm_maskz_fnmsub_sh): Likewise.
1559 (_mm_fnmsub_round_sh): Likewise.
1560 (_mm_mask_fnmsub_round_sh): Likewise.
1561 (_mm_mask3_fnmsub_round_sh): Likewise.
1562 (_mm_maskz_fnmsub_round_sh): Likewise.
1563 * config/i386/i386-builtin-types.def
1564 (V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT): New builtin type.
1565 * config/i386/i386-builtin.def: Add new builtins.
1566 * config/i386/i386-expand.c: Handle new builtin type.
1567 * config/i386/sse.md (fmai_vmfmadd_<mode><round_name>):
1568 Ajdust to support FP16.
1569 (fmai_vmfmsub_<mode><round_name>): Ditto.
1570 (fmai_vmfnmadd_<mode><round_name>): Ditto.
1571 (fmai_vmfnmsub_<mode><round_name>): Ditto.
1572 (*fmai_fmadd_<mode>): Ditto.
1573 (*fmai_fmsub_<mode>): Ditto.
1574 (*fmai_fnmadd_<mode><round_name>): Ditto.
1575 (*fmai_fnmsub_<mode><round_name>): Ditto.
1576 (avx512f_vmfmadd_<mode>_mask<round_name>): Ditto.
1577 (avx512f_vmfmadd_<mode>_mask3<round_name>): Ditto.
1578 (avx512f_vmfmadd_<mode>_maskz<round_expand_name>): Ditto.
1579 (avx512f_vmfmadd_<mode>_maskz_1<round_name>): Ditto.
1580 (*avx512f_vmfmsub_<mode>_mask<round_name>): Ditto.
1581 (avx512f_vmfmsub_<mode>_mask3<round_name>): Ditto.
1582 (*avx512f_vmfmsub_<mode>_maskz_1<round_name>): Ditto.
1583 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
1584 (*avx512f_vmfnmsub_<mode>_mask3<round_name>): Ditto.
1585 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
1586 (*avx512f_vmfnmadd_<mode>_mask<round_name>): Renamed to ...
1587 (avx512f_vmfnmadd_<mode>_mask<round_name>) ... this, and
1588 adjust to support FP16.
1589 (avx512f_vmfnmadd_<mode>_mask3<round_name>): Ditto.
1590 (avx512f_vmfnmadd_<mode>_maskz_1<round_name>): Ditto.
1591 (avx512f_vmfnmadd_<mode>_maskz<round_expand_name>): New
1594 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
1596 * config/i386/sse.md (avx512fmaskmodelower): Extend to support
1598 (maskload<mode><avx512fmaskmodelower>): Ditto.
1599 (maskstore<mode><avx512fmaskmodelower>): Ditto.
1601 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
1603 * config/i386/i386-expand.c (ix86_expand_fp_absneg_operator):
1605 (ix86_expand_copysign): Ditto.
1606 (ix86_expand_xorsign): Ditto.
1607 * config/i386/i386.c (ix86_build_const_vector): Handle HF vector
1609 (ix86_build_signbit_mask): Ditto.
1610 (ix86_can_change_mode_class): Ditto.
1611 * config/i386/i386.md
1612 (SSEMODEF): Add HFmode.
1613 (ssevecmodef): Ditto.
1614 (<code>hf2): New define_expand.
1615 (*<code>hf2_1): New define_insn_and_split.
1616 (copysign<mode>): Extend to support HFmode under AVX512FP16.
1617 (xorsign<mode>): Ditto.
1618 * config/i386/sse.md (VFB): New mode iterator.
1619 (VFB_128_256): Ditto.
1621 (sseintvecmode2): Support HF vector mode.
1622 (<code><mode>2): Use new mode iterator.
1623 (*<code><mode>2): Ditto.
1624 (copysign<mode>3): Ditto.
1625 (xorsign<mode>3): Ditto.
1626 (<code><mode>3<mask_name>): Ditto.
1627 (<code><mode>3<mask_name>): Ditto.
1628 (<sse>_andnot<mode>3<mask_name>): Adjust for HF vector mode.
1629 (<sse>_andnot<mode>3<mask_name>): Ditto.
1630 (*<code><mode>3<mask_name>): Ditto.
1631 (*<code><mode>3<mask_name>): Ditto.
1633 2021-09-18 liuhongt <hongtao.liu@intel.com>
1635 * config/i386/avx512fp16intrin.h (_mm512_mask_fmadd_ph):
1637 (_mm512_mask3_fmadd_ph): Likewise.
1638 (_mm512_maskz_fmadd_ph): Likewise.
1639 (_mm512_fmadd_round_ph): Likewise.
1640 (_mm512_mask_fmadd_round_ph): Likewise.
1641 (_mm512_mask3_fmadd_round_ph): Likewise.
1642 (_mm512_maskz_fmadd_round_ph): Likewise.
1643 (_mm512_fnmadd_ph): Likewise.
1644 (_mm512_mask_fnmadd_ph): Likewise.
1645 (_mm512_mask3_fnmadd_ph): Likewise.
1646 (_mm512_maskz_fnmadd_ph): Likewise.
1647 (_mm512_fnmadd_round_ph): Likewise.
1648 (_mm512_mask_fnmadd_round_ph): Likewise.
1649 (_mm512_mask3_fnmadd_round_ph): Likewise.
1650 (_mm512_maskz_fnmadd_round_ph): Likewise.
1651 (_mm512_fmsub_ph): Likewise.
1652 (_mm512_mask_fmsub_ph): Likewise.
1653 (_mm512_mask3_fmsub_ph): Likewise.
1654 (_mm512_maskz_fmsub_ph): Likewise.
1655 (_mm512_fmsub_round_ph): Likewise.
1656 (_mm512_mask_fmsub_round_ph): Likewise.
1657 (_mm512_mask3_fmsub_round_ph): Likewise.
1658 (_mm512_maskz_fmsub_round_ph): Likewise.
1659 (_mm512_fnmsub_ph): Likewise.
1660 (_mm512_mask_fnmsub_ph): Likewise.
1661 (_mm512_mask3_fnmsub_ph): Likewise.
1662 (_mm512_maskz_fnmsub_ph): Likewise.
1663 (_mm512_fnmsub_round_ph): Likewise.
1664 (_mm512_mask_fnmsub_round_ph): Likewise.
1665 (_mm512_mask3_fnmsub_round_ph): Likewise.
1666 (_mm512_maskz_fnmsub_round_ph): Likewise.
1667 * config/i386/avx512fp16vlintrin.h (_mm256_fmadd_ph):
1669 (_mm256_mask_fmadd_ph): Likewise.
1670 (_mm256_mask3_fmadd_ph): Likewise.
1671 (_mm256_maskz_fmadd_ph): Likewise.
1672 (_mm_fmadd_ph): Likewise.
1673 (_mm_mask_fmadd_ph): Likewise.
1674 (_mm_mask3_fmadd_ph): Likewise.
1675 (_mm_maskz_fmadd_ph): Likewise.
1676 (_mm256_fnmadd_ph): Likewise.
1677 (_mm256_mask_fnmadd_ph): Likewise.
1678 (_mm256_mask3_fnmadd_ph): Likewise.
1679 (_mm256_maskz_fnmadd_ph): Likewise.
1680 (_mm_fnmadd_ph): Likewise.
1681 (_mm_mask_fnmadd_ph): Likewise.
1682 (_mm_mask3_fnmadd_ph): Likewise.
1683 (_mm_maskz_fnmadd_ph): Likewise.
1684 (_mm256_fmsub_ph): Likewise.
1685 (_mm256_mask_fmsub_ph): Likewise.
1686 (_mm256_mask3_fmsub_ph): Likewise.
1687 (_mm256_maskz_fmsub_ph): Likewise.
1688 (_mm_fmsub_ph): Likewise.
1689 (_mm_mask_fmsub_ph): Likewise.
1690 (_mm_mask3_fmsub_ph): Likewise.
1691 (_mm_maskz_fmsub_ph): Likewise.
1692 (_mm256_fnmsub_ph): Likewise.
1693 (_mm256_mask_fnmsub_ph): Likewise.
1694 (_mm256_mask3_fnmsub_ph): Likewise.
1695 (_mm256_maskz_fnmsub_ph): Likewise.
1696 (_mm_fnmsub_ph): Likewise.
1697 (_mm_mask_fnmsub_ph): Likewise.
1698 (_mm_mask3_fnmsub_ph): Likewise.
1699 (_mm_maskz_fnmsub_ph): Likewise.
1700 * config/i386/i386-builtin.def: Add corresponding new builtins.
1701 * config/i386/sse.md
1702 (<avx512>_fmadd_<mode>_maskz<round_expand_name>): Adjust to
1703 support HF vector modes.
1704 (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
1706 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
1707 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
1708 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
1709 (<avx512>_fmadd_<mode>_mask<round_name>): Ditto.
1710 (<avx512>_fmadd_<mode>_mask3<round_name>): Ditto.
1711 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Ditto.
1712 (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>):
1714 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
1715 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
1716 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
1717 (<avx512>_fmsub_<mode>_mask<round_name>): Ditto.
1718 (<avx512>_fmsub_<mode>_mask3<round_name>): Ditto.
1719 (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>):
1721 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
1722 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
1723 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
1724 (<avx512>_fnmadd_<mode>_mask<round_name>): Ditto.
1725 (<avx512>_fnmadd_<mode>_mask3<round_name>): Ditto.
1726 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Ditto.
1727 (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>):
1729 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
1730 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
1731 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
1732 (<avx512>_fnmsub_<mode>_mask<round_name>): Ditto.
1733 (<avx512>_fnmsub_<mode>_mask3<round_name>): Ditto.
1735 2021-09-18 liuhongt <hongtao.liu@intel.com>
1737 * config/i386/avx512fp16intrin.h (_mm512_fmaddsub_ph):
1739 (_mm512_mask_fmaddsub_ph): Likewise.
1740 (_mm512_mask3_fmaddsub_ph): Likewise.
1741 (_mm512_maskz_fmaddsub_ph): Likewise.
1742 (_mm512_fmaddsub_round_ph): Likewise.
1743 (_mm512_mask_fmaddsub_round_ph): Likewise.
1744 (_mm512_mask3_fmaddsub_round_ph): Likewise.
1745 (_mm512_maskz_fmaddsub_round_ph): Likewise.
1746 (_mm512_mask_fmsubadd_ph): Likewise.
1747 (_mm512_mask3_fmsubadd_ph): Likewise.
1748 (_mm512_maskz_fmsubadd_ph): Likewise.
1749 (_mm512_fmsubadd_round_ph): Likewise.
1750 (_mm512_mask_fmsubadd_round_ph): Likewise.
1751 (_mm512_mask3_fmsubadd_round_ph): Likewise.
1752 (_mm512_maskz_fmsubadd_round_ph): Likewise.
1753 * config/i386/avx512fp16vlintrin.h (_mm256_fmaddsub_ph):
1755 (_mm256_mask_fmaddsub_ph): Likewise.
1756 (_mm256_mask3_fmaddsub_ph): Likewise.
1757 (_mm256_maskz_fmaddsub_ph): Likewise.
1758 (_mm_fmaddsub_ph): Likewise.
1759 (_mm_mask_fmaddsub_ph): Likewise.
1760 (_mm_mask3_fmaddsub_ph): Likewise.
1761 (_mm_maskz_fmaddsub_ph): Likewise.
1762 (_mm256_fmsubadd_ph): Likewise.
1763 (_mm256_mask_fmsubadd_ph): Likewise.
1764 (_mm256_mask3_fmsubadd_ph): Likewise.
1765 (_mm256_maskz_fmsubadd_ph): Likewise.
1766 (_mm_fmsubadd_ph): Likewise.
1767 (_mm_mask_fmsubadd_ph): Likewise.
1768 (_mm_mask3_fmsubadd_ph): Likewise.
1769 (_mm_maskz_fmsubadd_ph): Likewise.
1770 * config/i386/i386-builtin.def: Add corresponding new builtins.
1771 * config/i386/sse.md (VFH_SF_AVX512VL): New mode iterator.
1772 * (<avx512>_fmsubadd_<mode>_maskz<round_expand_name>): New expander.
1773 * (<avx512>_fmaddsub_<mode>_maskz<round_expand_name>): Use
1775 * (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
1777 * (<avx512>_fmaddsub_<mode>_mask<round_name>): Ditto.
1778 * (<avx512>_fmaddsub_<mode>_mask3<round_name>): Ditto.
1779 * (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
1781 * (<avx512>_fmsubadd_<mode>_mask<round_name>): Ditto.
1782 * (<avx512>_fmsubadd_<mode>_mask3<round_name>): Ditto.
1784 2021-09-18 liuhongt <hongtao.liu@intel.com>
1787 * config/i386/i386.c (ix86_print_operand): Handle
1788 V8HF/V16HF/V32HFmode.
1789 * config/i386/i386.h (VALID_BCST_MODE_P): Add HFmode.
1790 * config/i386/sse.md (avx512bcst): Remove.
1792 2021-09-17 Martin Sebor <msebor@redhat.com>
1794 * Makefile.in (OBJS): Add gimple-predicate-analysis.o.
1795 * tree-ssa-uninit.c (max_phi_args): Move to gimple-predicate-analysis.
1796 (MASK_SET_BIT, MASK_TEST_BIT, MASK_EMPTY): Same.
1797 (check_defs): Add comment.
1798 (can_skip_redundant_opnd): Update comment.
1799 (compute_uninit_opnds_pos): Adjust to namespace change.
1800 (find_pdom): Move to gimple-predicate-analysis.cc.
1802 (struct uninit_undef_val_t): New.
1803 (is_non_loop_exit_postdominating): Move to gimple-predicate-analysis.cc.
1804 (find_control_equiv_block): Same.
1805 (MAX_NUM_CHAINS, MAX_CHAIN_LEN, MAX_POSTDOM_CHECK): Same.
1806 (MAX_SWITCH_CASES): Same.
1807 (compute_control_dep_chain): Same.
1808 (find_uninit_use): Use predicate analyzer.
1809 (struct pred_info): Move to gimple-predicate-analysis.
1810 (convert_control_dep_chain_into_preds): Same.
1811 (find_predicates): Same.
1812 (collect_phi_def_edges): Same.
1813 (warn_uninitialized_phi): Use predicate analyzer.
1814 (find_def_preds): Move to gimple-predicate-analysis.
1815 (dump_pred_info): Same.
1816 (dump_pred_chain): Same.
1817 (dump_predicates): Same.
1818 (destroy_predicate_vecs): Remove.
1819 (execute_late_warn_uninitialized): New.
1820 (get_cmp_code): Move to gimple-predicate-analysis.
1821 (is_value_included_in): Same.
1822 (value_sat_pred_p): Same.
1823 (find_matching_predicate_in_rest_chains): Same.
1824 (is_use_properly_guarded): Same.
1825 (prune_uninit_phi_opnds): Same.
1826 (find_var_cmp_const): Same.
1827 (use_pred_not_overlap_with_undef_path_pred): Same.
1828 (pred_equal_p): Same.
1829 (is_neq_relop_p): Same.
1830 (is_neq_zero_form_p): Same.
1831 (pred_expr_equal_p): Same.
1832 (is_pred_expr_subset_of): Same.
1833 (is_pred_chain_subset_of): Same.
1834 (is_included_in): Same.
1835 (is_superset_of): Same.
1837 (simplify_pred): Same.
1838 (simplify_preds_2): Same.
1839 (simplify_preds_3): Same.
1840 (simplify_preds_4): Same.
1841 (simplify_preds): Same.
1843 (push_to_worklist): Same.
1844 (get_pred_info_from_cmp): Same.
1845 (is_degenerated_phi): Same.
1846 (normalize_one_pred_1): Same.
1847 (normalize_one_pred): Same.
1848 (normalize_one_pred_chain): Same.
1849 (normalize_preds): Same.
1850 (can_one_predicate_be_invalidated_p): Same.
1851 (can_chain_union_be_invalidated_p): Same.
1852 (uninit_uses_cannot_happen): Same.
1853 (pass_late_warn_uninitialized::execute): Define.
1854 * gimple-predicate-analysis.cc: New file.
1855 * gimple-predicate-analysis.h: New file.
1857 2021-09-17 Julian Brown <julian@codesourcery.com>
1859 * config/gcn/gcn.c (gimple.h): Include.
1860 (gcn_fork_join): Emit barrier for worker-level joins.
1861 * omp-oacc-neuter-broadcast.cc (find_local_vars_to_propagate): Add
1862 writes_gang_private bitmap parameter. Set bit for blocks
1863 containing gang-private variable writes.
1864 (worker_single_simple): Don't emit barrier after predicated block.
1865 (worker_single_copy): Don't emit barrier if we're not broadcasting
1866 anything and the block contains no gang-private writes.
1867 (neuter_worker_single): Don't predicate blocks that only contain
1868 NOPs or internal marker functions. Pass has_gang_private_write
1869 argument to worker_single_copy.
1870 (oacc_do_neutering): Add writes_gang_private bitmap handling.
1872 2021-09-17 Julian Brown <julian@codesourcery.com>
1874 * config/gcn/gcn-protos.h
1875 (gcn_goacc_create_worker_broadcast_record): Update prototype.
1876 * config/gcn/gcn-tree.c (gcn_goacc_get_worker_red_decl): Use
1877 preallocated block of LDS memory. Do not cache/share decls for
1878 reduction temporaries between invocations.
1879 (gcn_goacc_reduction_teardown): Unshare VAR on second use.
1880 (gcn_goacc_create_worker_broadcast_record): Add OFFSET parameter
1881 and return temporary LDS space at that offset. Return pointer in
1883 * config/gcn/gcn.c (acc_lds_size, gang_private_hwm, lds_allocs):
1885 (ACC_LDS_SIZE): Define as acc_lds_size.
1886 (gcn_init_machine_status): Don't initialise lds_allocated,
1887 lds_allocs, reduc_decls fields of machine function struct.
1888 (gcn_option_override): Handle default size for gang-private
1889 variables and -mgang-private-size option.
1890 (gcn_expand_prologue): Use LDS_SIZE instead of LDS_SIZE-1 when
1891 initialising M0_REG.
1892 (gcn_shared_mem_layout): New function.
1893 (gcn_print_lds_decl): Update comment. Use global lds_allocs map and
1894 gang_private_hwm variable.
1895 (TARGET_GOACC_SHARED_MEM_LAYOUT): Define target hook.
1896 * config/gcn/gcn.h (machine_function): Remove lds_allocated,
1897 lds_allocs, reduc_decls. Add reduction_base, reduction_limit.
1898 * config/gcn/gcn.opt (gang_private_size_opt): New global.
1899 (mgang-private-size=): New option.
1900 * doc/tm.texi.in (TARGET_GOACC_SHARED_MEM_LAYOUT): Place
1902 * doc/tm.texi: Regenerate.
1903 * omp-oacc-neuter-broadcast.cc (targhooks.h, diagnostic-core.h):
1905 (build_sender_ref): Handle sender_decl being pointer.
1906 (worker_single_copy): Add PLACEMENT and ISOLATE_BROADCASTS
1907 parameters. Pass placement argument to
1908 create_worker_broadcast_record hook invocations. Handle
1909 sender_decl being pointer and isolate_broadcasts inserting extra
1911 (blk_offset_map_t): Add typedef.
1912 (neuter_worker_single): Add BLK_OFFSET_MAP parameter. Pass
1913 preallocated range to worker_single_copy call.
1914 (dfs_broadcast_reachable_1): New function.
1915 (idx_decl_pair_t, used_range_vec_t): New typedefs.
1916 (sort_size_descending): New function.
1917 (addr_range): New class.
1918 (splay_tree_compare_addr_range, splay_tree_free_key)
1919 (first_fit_range, merge_ranges_1, merge_ranges): New functions.
1920 (execute_omp_oacc_neuter_broadcast): Rename to...
1921 (oacc_do_neutering): ... this. Add BOUNDS_LO, BOUNDS_HI
1922 parameters. Arrange layout of shared memory for broadcast
1924 (execute_omp_oacc_neuter_broadcast): New function.
1925 (pass_omp_oacc_neuter_broadcast::gate): Remove num_workers==1
1926 handling from here. Enable pass for all OpenACC routines in order
1927 to call shared memory-layout hook.
1928 * target.def (create_worker_broadcast_record): Add OFFSET
1930 (shared_mem_layout): New hook.
1932 2021-09-17 Julian Brown <julian@codesourcery.com>
1933 Thomas Schwinge <thomas@codesourcery.com>
1935 * omp-oacc-neuter-broadcast.cc
1936 (pass_omp_oacc_neuter_broadcast::gate): Disable if num_workers is
1938 (execute_omp_oacc_neuter_broadcast): Adjust.
1940 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
1942 * value-relation.cc (class equiv_chain): Move to header file.
1943 (path_oracle::path_oracle): New.
1944 (path_oracle::~path_oracle): New.
1945 (path_oracle::register_relation): New.
1946 (path_oracle::query_relation): New.
1947 (path_oracle::reset_path): New.
1948 (path_oracle::dump): New.
1949 * value-relation.h (class equiv_chain): Move to here.
1950 (class path_oracle): New.
1952 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
1954 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a DOM
1956 * gimple-range-fold.cc (fur_depend::register_relation): Use
1957 register_stmt/edge routines.
1958 * value-relation.cc (equiv_chain::find): Relocate from equiv_oracle.
1959 (equiv_oracle::equiv_oracle): Create self equivalence cache.
1960 (equiv_oracle::~equiv_oracle): Release same.
1961 (equiv_oracle::equiv_set): Return entry from self equiv cache if there
1962 are no equivalences.
1963 (equiv_oracle::find_equiv_block): Move list find to equiv_chain.
1964 (equiv_oracle::register_relation): Rename from register_equiv.
1965 (relation_chain_head::find_relation): Relocate from dom_oracle.
1966 (relation_oracle::register_stmt): New.
1967 (relation_oracle::register_edge): New.
1968 (dom_oracle::*): Rename from relation_oracle.
1969 (dom_oracle::register_relation): Adjust to call equiv_oracle.
1970 (dom_oracle::set_one_relation): Split from register_relation.
1971 (dom_oracle::register_transitives): Consolidate 2 methods.
1972 (dom_oracle::find_relation_block): Move core to relation_chain.
1973 (dom_oracle::query_relation): Rename from find_relation_dom and adjust.
1974 * value-relation.h (class relation_oracle): New pure virtual base.
1975 (class equiv_oracle): Inherit from relation_oracle and adjust.
1976 (class dom_oracle): Rename from old relation_oracle and adjust.
1978 2021-09-17 Martin Sebor <msebor@redhat.com>
1980 PR middle-end/102200
1981 * pointer-query.cc (access_ref::inform_access): Handle MIN/MAX_EXPR.
1982 (handle_min_max_size): Change argument. Store original SSA_NAME for
1983 operands to potentially distinct (sub)objects.
1984 (compute_objsize_r): Adjust call to the above.
1986 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1988 * config/rs6000/rs6000.c (rs6000-builtins.h): New include.
1989 (rs6000_new_builtin_vectorized_function): New function.
1990 (rs6000_new_builtin_md_vectorized_function): Likewise.
1991 (rs6000_builtin_vectorized_function): Call
1992 rs6000_new_builtin_vectorized_function.
1993 (rs6000_builtin_md_vectorized_function): Call
1994 rs6000_new_builtin_md_vectorized_function.
1996 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
1998 * config/rs6000/rs6000-builtin-new.def (ASSEMBLE_ACC): Add mmaint flag.
1999 (ASSEMBLE_PAIR): Likewise.
2000 (BUILD_ACC): Likewise.
2001 (DISASSEMBLE_ACC): Likewise.
2002 (DISASSEMBLE_PAIR): Likewise.
2003 (PMXVBF16GER2): Likewise.
2004 (PMXVBF16GER2NN): Likewise.
2005 (PMXVBF16GER2NP): Likewise.
2006 (PMXVBF16GER2PN): Likewise.
2007 (PMXVBF16GER2PP): Likewise.
2008 (PMXVF16GER2): Likewise.
2009 (PMXVF16GER2NN): Likewise.
2010 (PMXVF16GER2NP): Likewise.
2011 (PMXVF16GER2PN): Likewise.
2012 (PMXVF16GER2PP): Likewise.
2013 (PMXVF32GER): Likewise.
2014 (PMXVF32GERNN): Likewise.
2015 (PMXVF32GERNP): Likewise.
2016 (PMXVF32GERPN): Likewise.
2017 (PMXVF32GERPP): Likewise.
2018 (PMXVF64GER): Likewise.
2019 (PMXVF64GERNN): Likewise.
2020 (PMXVF64GERNP): Likewise.
2021 (PMXVF64GERPN): Likewise.
2022 (PMXVF64GERPP): Likewise.
2023 (PMXVI16GER2): Likewise.
2024 (PMXVI16GER2PP): Likewise.
2025 (PMXVI16GER2S): Likewise.
2026 (PMXVI16GER2SPP): Likewise.
2027 (PMXVI4GER8): Likewise.
2028 (PMXVI4GER8PP): Likewise.
2029 (PMXVI8GER4): Likewise.
2030 (PMXVI8GER4PP): Likewise.
2031 (PMXVI8GER4SPP): Likewise.
2032 (XVBF16GER2): Likewise.
2033 (XVBF16GER2NN): Likewise.
2034 (XVBF16GER2NP): Likewise.
2035 (XVBF16GER2PN): Likewise.
2036 (XVBF16GER2PP): Likewise.
2037 (XVF16GER2): Likewise.
2038 (XVF16GER2NN): Likewise.
2039 (XVF16GER2NP): Likewise.
2040 (XVF16GER2PN): Likewise.
2041 (XVF16GER2PP): Likewise.
2042 (XVF32GER): Likewise.
2043 (XVF32GERNN): Likewise.
2044 (XVF32GERNP): Likewise.
2045 (XVF32GERPN): Likewise.
2046 (XVF32GERPP): Likewise.
2047 (XVF64GER): Likewise.
2048 (XVF64GERNN): Likewise.
2049 (XVF64GERNP): Likewise.
2050 (XVF64GERPN): Likewise.
2051 (XVF64GERPP): Likewise.
2052 (XVI16GER2): Likewise.
2053 (XVI16GER2PP): Likewise.
2054 (XVI16GER2S): Likewise.
2055 (XVI16GER2SPP): Likewise.
2056 (XVI4GER8): Likewise.
2057 (XVI4GER8PP): Likewise.
2058 (XVI8GER4): Likewise.
2059 (XVI8GER4PP): Likewise.
2060 (XVI8GER4SPP): Likewise.
2061 (XXMFACC): Likewise.
2062 (XXMTACC): Likewise.
2063 (XXSETACCZ): Likewise.
2064 (ASSEMBLE_PAIR_V): Likewise.
2065 (BUILD_PAIR): Likewise.
2066 (DISASSEMBLE_PAIR_V): Likewise.
2069 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_mma_builtin):
2070 Handle RS6000_BIF_LXVP and RS6000_BIF_STXVP.
2071 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add ismmaint.
2072 (parse_bif_attrs): Handle ismmaint.
2073 (write_decls): Add bif_mmaint_bit and bif_is_mmaint.
2074 (write_bif_static_init): Handle ismmaint.
2076 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
2078 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin): New
2080 (rs6000_gimple_fold_builtin): Call rs6000_gimple_fold_new_builtin.
2081 (rs6000_new_builtin_valid_without_lhs): New function.
2082 (rs6000_gimple_fold_new_mma_builtin): Likewise.
2083 (rs6000_gimple_fold_new_builtin): Likewise.
2085 2021-09-17 Thomas Schwinge <thomas@codesourcery.com>
2087 * hash-table.h (hash_table<Descriptor, Lazy, Allocator>::expand):
2088 Destruct stale Value objects.
2089 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
2092 2021-09-17 Roger Sayle <roger@nextmovesoftware.com>
2095 * match.pd (shift optimizations): Disable recent sign-changing
2096 optimization for shifts by zero, these will be folded later.
2098 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
2100 * config/rs6000/rs6000-builtin-new.def (__builtin_mffsl): Move from
2101 [power9] to [always].
2103 2021-09-17 Richard Biener <rguenther@suse.de>
2105 * tree-vect-stmts.c (vectorizable_load): Do not frob
2108 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
2110 * config/i386/i386-features.c (remove_partial_avx_dependency):
2111 Also check TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY and
2112 and TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY before generating
2114 * config/i386/i386.h (TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY):
2116 (TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
2117 * config/i386/i386.md (SSE FP to FP splitters): Replace
2118 TARGET_SSE_PARTIAL_REG_DEPENDENCY with
2119 TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY.
2120 (SSE INT to FP splitter): Replace TARGET_SSE_PARTIAL_REG_DEPENDENCY
2121 with TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY.
2122 * config/i386/x86-tune.def
2123 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): New.
2124 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
2126 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
2129 * config/i386/i386-features.c (remove_partial_avx_dependency):
2130 Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
2131 before generating vxorps.
2133 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
2135 * config/i386/i386-options.c (processor_cost_table): Use
2136 tremont_cost for Tremont.
2137 * config/i386/x86-tune-costs.h (tremont_memcpy): New.
2138 (tremont_memset): Likewise.
2139 (tremont_cost): Likewise.
2140 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
2143 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
2145 * common/config/i386/i386-common.c: Use Haswell scheduling model
2147 * config/i386/i386.c (ix86_sched_init_global): Prepare for Tremont
2149 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Tremont
2151 (ix86_adjust_cost): Handle Tremont.
2152 * config/i386/x86-tune.def (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY):
2154 (X86_TUNE_USE_LEAVE): Likewise.
2155 (X86_TUNE_PUSH_MEMORY): Likewise.
2156 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
2157 (X86_TUNE_USE_CLTD): Likewise.
2158 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
2159 (X86_TUNE_AVOID_MFENCE): Likewise.
2160 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
2161 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
2162 (X86_TUNE_ACCUMULATE_OUTGOING_ARGS): Disable for Tremont.
2163 (X86_TUNE_FOUR_JUMP_LIMIT): Likewise.
2164 (X86_TUNE_OPT_AGU): Likewise.
2165 (X86_TUNE_AVOID_LEA_FOR_ADDR): Likewise.
2166 (X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE): Likewise.
2167 (X86_TUNE_EXPAND_ABS): Likewise.
2168 (X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS): Likewise.
2169 (X86_TUNE_SLOW_PSHUFB): Likewise.
2171 2021-09-17 Eric Botcazou <ebotcazou@adacore.com>
2173 PR rtl-optimization/102306
2174 * combine.c (try_combine): Abort the combination if we are about to
2175 duplicate volatile references.
2177 2021-09-17 liuhongt <hongtao.liu@intel.com>
2179 * config/i386/avx512fp16intrin.h (_mm_undefined_ph):
2181 (_mm256_undefined_ph): Likewise.
2182 (_mm512_undefined_ph): Likewise.
2183 (_mm_cvtsh_h): Likewise.
2184 (_mm256_cvtsh_h): Likewise.
2185 (_mm512_cvtsh_h): Likewise.
2186 (_mm512_castph_ps): Likewise.
2187 (_mm512_castph_pd): Likewise.
2188 (_mm512_castph_si512): Likewise.
2189 (_mm512_castph512_ph128): Likewise.
2190 (_mm512_castph512_ph256): Likewise.
2191 (_mm512_castph128_ph512): Likewise.
2192 (_mm512_castph256_ph512): Likewise.
2193 (_mm512_zextph128_ph512): Likewise.
2194 (_mm512_zextph256_ph512): Likewise.
2195 (_mm512_castps_ph): Likewise.
2196 (_mm512_castpd_ph): Likewise.
2197 (_mm512_castsi512_ph): Likewise.
2198 * config/i386/avx512fp16vlintrin.h (_mm_castph_ps):
2200 (_mm256_castph_ps): Likewise.
2201 (_mm_castph_pd): Likewise.
2202 (_mm256_castph_pd): Likewise.
2203 (_mm_castph_si128): Likewise.
2204 (_mm256_castph_si256): Likewise.
2205 (_mm_castps_ph): Likewise.
2206 (_mm256_castps_ph): Likewise.
2207 (_mm_castpd_ph): Likewise.
2208 (_mm256_castpd_ph): Likewise.
2209 (_mm_castsi128_ph): Likewise.
2210 (_mm256_castsi256_ph): Likewise.
2211 (_mm256_castph256_ph128): Likewise.
2212 (_mm256_castph128_ph256): Likewise.
2213 (_mm256_zextph128_ph256): Likewise.
2215 2021-09-17 liuhongt <hongtao.liu@intel.com>
2217 * config/i386/avx512fp16intrin.h (_mm_cvtsh_ss):
2219 (_mm_mask_cvtsh_ss): Likewise.
2220 (_mm_maskz_cvtsh_ss): Likewise.
2221 (_mm_cvtsh_sd): Likewise.
2222 (_mm_mask_cvtsh_sd): Likewise.
2223 (_mm_maskz_cvtsh_sd): Likewise.
2224 (_mm_cvt_roundsh_ss): Likewise.
2225 (_mm_mask_cvt_roundsh_ss): Likewise.
2226 (_mm_maskz_cvt_roundsh_ss): Likewise.
2227 (_mm_cvt_roundsh_sd): Likewise.
2228 (_mm_mask_cvt_roundsh_sd): Likewise.
2229 (_mm_maskz_cvt_roundsh_sd): Likewise.
2230 (_mm_cvtss_sh): Likewise.
2231 (_mm_mask_cvtss_sh): Likewise.
2232 (_mm_maskz_cvtss_sh): Likewise.
2233 (_mm_cvtsd_sh): Likewise.
2234 (_mm_mask_cvtsd_sh): Likewise.
2235 (_mm_maskz_cvtsd_sh): Likewise.
2236 (_mm_cvt_roundss_sh): Likewise.
2237 (_mm_mask_cvt_roundss_sh): Likewise.
2238 (_mm_maskz_cvt_roundss_sh): Likewise.
2239 (_mm_cvt_roundsd_sh): Likewise.
2240 (_mm_mask_cvt_roundsd_sh): Likewise.
2241 (_mm_maskz_cvt_roundsd_sh): Likewise.
2242 * config/i386/i386-builtin-types.def
2243 (V8HF_FTYPE_V2DF_V8HF_V8HF_UQI_INT,
2244 V8HF_FTYPE_V4SF_V8HF_V8HF_UQI_INT,
2245 V2DF_FTYPE_V8HF_V2DF_V2DF_UQI_INT,
2246 V4SF_FTYPE_V8HF_V4SF_V4SF_UQI_INT): Add new builtin types.
2247 * config/i386/i386-builtin.def: Add corrresponding new builtins.
2248 * config/i386/i386-expand.c: Handle new builtin types.
2249 * config/i386/sse.md (VF48_128): New mode iterator.
2250 (avx512fp16_vcvtsh2<ssescalarmodesuffix><mask_scalar_name><round_saeonly_scalar_name>):
2252 (avx512fp16_vcvt<ssescalarmodesuffix>2sh<mask_scalar_name><round_scalar_name>):
2255 2021-09-17 liuhongt <hongtao.liu@intel.com>
2257 * config/i386/avx512fp16intrin.h (_mm512_cvtph_pd):
2259 (_mm512_mask_cvtph_pd): Likewise.
2260 (_mm512_maskz_cvtph_pd): Likewise.
2261 (_mm512_cvt_roundph_pd): Likewise.
2262 (_mm512_mask_cvt_roundph_pd): Likewise.
2263 (_mm512_maskz_cvt_roundph_pd): Likewise.
2264 (_mm512_cvtxph_ps): Likewise.
2265 (_mm512_mask_cvtxph_ps): Likewise.
2266 (_mm512_maskz_cvtxph_ps): Likewise.
2267 (_mm512_cvtx_roundph_ps): Likewise.
2268 (_mm512_mask_cvtx_roundph_ps): Likewise.
2269 (_mm512_maskz_cvtx_roundph_ps): Likewise.
2270 (_mm512_cvtxps_ph): Likewise.
2271 (_mm512_mask_cvtxps_ph): Likewise.
2272 (_mm512_maskz_cvtxps_ph): Likewise.
2273 (_mm512_cvtx_roundps_ph): Likewise.
2274 (_mm512_mask_cvtx_roundps_ph): Likewise.
2275 (_mm512_maskz_cvtx_roundps_ph): Likewise.
2276 (_mm512_cvtpd_ph): Likewise.
2277 (_mm512_mask_cvtpd_ph): Likewise.
2278 (_mm512_maskz_cvtpd_ph): Likewise.
2279 (_mm512_cvt_roundpd_ph): Likewise.
2280 (_mm512_mask_cvt_roundpd_ph): Likewise.
2281 (_mm512_maskz_cvt_roundpd_ph): Likewise.
2282 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_pd):
2284 (_mm_mask_cvtph_pd): Likewise.
2285 (_mm_maskz_cvtph_pd): Likewise.
2286 (_mm256_cvtph_pd): Likewise.
2287 (_mm256_mask_cvtph_pd): Likewise.
2288 (_mm256_maskz_cvtph_pd): Likewise.
2289 (_mm_cvtxph_ps): Likewise.
2290 (_mm_mask_cvtxph_ps): Likewise.
2291 (_mm_maskz_cvtxph_ps): Likewise.
2292 (_mm256_cvtxph_ps): Likewise.
2293 (_mm256_mask_cvtxph_ps): Likewise.
2294 (_mm256_maskz_cvtxph_ps): Likewise.
2295 (_mm_cvtxps_ph): Likewise.
2296 (_mm_mask_cvtxps_ph): Likewise.
2297 (_mm_maskz_cvtxps_ph): Likewise.
2298 (_mm256_cvtxps_ph): Likewise.
2299 (_mm256_mask_cvtxps_ph): Likewise.
2300 (_mm256_maskz_cvtxps_ph): Likewise.
2301 (_mm_cvtpd_ph): Likewise.
2302 (_mm_mask_cvtpd_ph): Likewise.
2303 (_mm_maskz_cvtpd_ph): Likewise.
2304 (_mm256_cvtpd_ph): Likewise.
2305 (_mm256_mask_cvtpd_ph): Likewise.
2306 (_mm256_maskz_cvtpd_ph): Likewise.
2307 * config/i386/i386-builtin.def: Add corresponding new builtins.
2308 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2309 * config/i386/i386-expand.c: Handle new builtin types.
2310 * config/i386/sse.md
2311 (VF4_128_8_256): New.
2312 (VF48H_AVX512VL): Ditto.
2313 (ssePHmode): Add HF vector modes.
2314 (castmode): Add new convertable modes.
2317 (avx512fp16_vcvt<castmode>2ph_<mode><mask_name><round_name>): Ditto.
2318 (avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
2319 (*avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
2320 (avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
2321 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
2322 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask_1): Ditto.
2323 (avx512fp16_float_extend_ph<mode>2<mask_name><round_saeonly_name>):
2325 (avx512fp16_float_extend_ph<mode>2<mask_name>): Ditto.
2326 (*avx512fp16_float_extend_ph<mode>2_load<mask_name>): Ditto.
2327 (avx512fp16_float_extend_phv2df2<mask_name>): Ditto.
2328 (*avx512fp16_float_extend_phv2df2_load<mask_name>): Ditto.
2330 2021-09-17 liuhongt <hongtao.liu@intel.com>
2332 * config/i386/avx512fp16intrin.h (_mm_cvttsh_i32):
2334 (_mm_cvttsh_u32): Likewise.
2335 (_mm_cvtt_roundsh_i32): Likewise.
2336 (_mm_cvtt_roundsh_u32): Likewise.
2337 (_mm_cvttsh_i64): Likewise.
2338 (_mm_cvttsh_u64): Likewise.
2339 (_mm_cvtt_roundsh_i64): Likewise.
2340 (_mm_cvtt_roundsh_u64): Likewise.
2341 * config/i386/i386-builtin.def: Add corresponding new builtins.
2342 * config/i386/sse.md
2343 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<round_saeonly_name>):
2346 2021-09-17 liuhongt <hongtao.liu@intel.com>
2348 * config/i386/avx512fp16intrin.h (_mm512_cvttph_epi32):
2350 (_mm512_mask_cvttph_epi32): Likewise.
2351 (_mm512_maskz_cvttph_epi32): Likewise.
2352 (_mm512_cvtt_roundph_epi32): Likewise.
2353 (_mm512_mask_cvtt_roundph_epi32): Likewise.
2354 (_mm512_maskz_cvtt_roundph_epi32): Likewise.
2355 (_mm512_cvttph_epu32): Likewise.
2356 (_mm512_mask_cvttph_epu32): Likewise.
2357 (_mm512_maskz_cvttph_epu32): Likewise.
2358 (_mm512_cvtt_roundph_epu32): Likewise.
2359 (_mm512_mask_cvtt_roundph_epu32): Likewise.
2360 (_mm512_maskz_cvtt_roundph_epu32): Likewise.
2361 (_mm512_cvttph_epi64): Likewise.
2362 (_mm512_mask_cvttph_epi64): Likewise.
2363 (_mm512_maskz_cvttph_epi64): Likewise.
2364 (_mm512_cvtt_roundph_epi64): Likewise.
2365 (_mm512_mask_cvtt_roundph_epi64): Likewise.
2366 (_mm512_maskz_cvtt_roundph_epi64): Likewise.
2367 (_mm512_cvttph_epu64): Likewise.
2368 (_mm512_mask_cvttph_epu64): Likewise.
2369 (_mm512_maskz_cvttph_epu64): Likewise.
2370 (_mm512_cvtt_roundph_epu64): Likewise.
2371 (_mm512_mask_cvtt_roundph_epu64): Likewise.
2372 (_mm512_maskz_cvtt_roundph_epu64): Likewise.
2373 (_mm512_cvttph_epi16): Likewise.
2374 (_mm512_mask_cvttph_epi16): Likewise.
2375 (_mm512_maskz_cvttph_epi16): Likewise.
2376 (_mm512_cvtt_roundph_epi16): Likewise.
2377 (_mm512_mask_cvtt_roundph_epi16): Likewise.
2378 (_mm512_maskz_cvtt_roundph_epi16): Likewise.
2379 (_mm512_cvttph_epu16): Likewise.
2380 (_mm512_mask_cvttph_epu16): Likewise.
2381 (_mm512_maskz_cvttph_epu16): Likewise.
2382 (_mm512_cvtt_roundph_epu16): Likewise.
2383 (_mm512_mask_cvtt_roundph_epu16): Likewise.
2384 (_mm512_maskz_cvtt_roundph_epu16): Likewise.
2385 * config/i386/avx512fp16vlintrin.h (_mm_cvttph_epi32):
2387 (_mm_mask_cvttph_epi32): Likewise.
2388 (_mm_maskz_cvttph_epi32): Likewise.
2389 (_mm256_cvttph_epi32): Likewise.
2390 (_mm256_mask_cvttph_epi32): Likewise.
2391 (_mm256_maskz_cvttph_epi32): Likewise.
2392 (_mm_cvttph_epu32): Likewise.
2393 (_mm_mask_cvttph_epu32): Likewise.
2394 (_mm_maskz_cvttph_epu32): Likewise.
2395 (_mm256_cvttph_epu32): Likewise.
2396 (_mm256_mask_cvttph_epu32): Likewise.
2397 (_mm256_maskz_cvttph_epu32): Likewise.
2398 (_mm_cvttph_epi64): Likewise.
2399 (_mm_mask_cvttph_epi64): Likewise.
2400 (_mm_maskz_cvttph_epi64): Likewise.
2401 (_mm256_cvttph_epi64): Likewise.
2402 (_mm256_mask_cvttph_epi64): Likewise.
2403 (_mm256_maskz_cvttph_epi64): Likewise.
2404 (_mm_cvttph_epu64): Likewise.
2405 (_mm_mask_cvttph_epu64): Likewise.
2406 (_mm_maskz_cvttph_epu64): Likewise.
2407 (_mm256_cvttph_epu64): Likewise.
2408 (_mm256_mask_cvttph_epu64): Likewise.
2409 (_mm256_maskz_cvttph_epu64): Likewise.
2410 (_mm_cvttph_epi16): Likewise.
2411 (_mm_mask_cvttph_epi16): Likewise.
2412 (_mm_maskz_cvttph_epi16): Likewise.
2413 (_mm256_cvttph_epi16): Likewise.
2414 (_mm256_mask_cvttph_epi16): Likewise.
2415 (_mm256_maskz_cvttph_epi16): Likewise.
2416 (_mm_cvttph_epu16): Likewise.
2417 (_mm_mask_cvttph_epu16): Likewise.
2418 (_mm_maskz_cvttph_epu16): Likewise.
2419 (_mm256_cvttph_epu16): Likewise.
2420 (_mm256_mask_cvttph_epu16): Likewise.
2421 (_mm256_maskz_cvttph_epu16): Likewise.
2422 * config/i386/i386-builtin.def: Add new builtins.
2423 * config/i386/sse.md
2424 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name><round_saeonly_name>):
2426 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name>): Ditto.
2427 (*avx512fp16_fix<fixunssuffix>_trunc<mode>2_load<mask_name>): Ditto.
2428 (avx512fp16_fix<fixunssuffix>_truncv2di2<mask_name>): Ditto.
2429 (avx512fp16_fix<fixunssuffix>_truncv2di2_load<mask_name>): Ditto.
2431 2021-09-17 liuhongt <hongtao.liu@intel.com>
2433 * config/i386/avx512fp16intrin.h (_mm_cvtsh_i32): New intrinsic.
2434 (_mm_cvtsh_u32): Likewise.
2435 (_mm_cvt_roundsh_i32): Likewise.
2436 (_mm_cvt_roundsh_u32): Likewise.
2437 (_mm_cvtsh_i64): Likewise.
2438 (_mm_cvtsh_u64): Likewise.
2439 (_mm_cvt_roundsh_i64): Likewise.
2440 (_mm_cvt_roundsh_u64): Likewise.
2441 (_mm_cvti32_sh): Likewise.
2442 (_mm_cvtu32_sh): Likewise.
2443 (_mm_cvt_roundi32_sh): Likewise.
2444 (_mm_cvt_roundu32_sh): Likewise.
2445 (_mm_cvti64_sh): Likewise.
2446 (_mm_cvtu64_sh): Likewise.
2447 (_mm_cvt_roundi64_sh): Likewise.
2448 (_mm_cvt_roundu64_sh): Likewise.
2449 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2450 * config/i386/i386-builtin.def: Add corresponding new builtins.
2451 * config/i386/i386-expand.c (ix86_expand_round_builtin):
2452 Handle new builtin types.
2453 * config/i386/sse.md
2454 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix><round_name>):
2456 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix>_2): Likewise.
2457 (avx512fp16_vcvt<floatsuffix>si2sh<rex64namesuffix><round_name>): Likewise.
2459 2021-09-16 Bill Schmidt <wschmidt@linux.ibm.com>
2461 * config/rs6000/rs6000-c.c (rs6000-builtins.h): New include.
2462 (altivec_resolve_new_overloaded_builtin): New forward decl.
2463 (rs6000_new_builtin_type_compatible): New function.
2464 (altivec_resolve_overloaded_builtin): Call
2465 altivec_resolve_new_overloaded_builtin.
2466 (altivec_build_new_resolved_builtin): New function.
2467 (altivec_resolve_new_overloaded_builtin): Likewise.
2468 * config/rs6000/rs6000-call.c (rs6000_new_builtin_is_supported):
2470 * config/rs6000/rs6000-gen-builtins.c (write_decls): Remove _p from
2471 name of rs6000_new_builtin_is_supported.
2473 2021-09-16 Uroš Bizjak <ubizjak@gmail.com>
2475 * config/i386/i386-protos.h (ix86_decompose_address):
2476 Change return type to bool.
2477 * config/i386/i386.c (ix86_decompose_address): Ditto.
2479 2021-09-16 Tobias Burnus <tobias@codesourcery.com>
2482 * config/rs6000/t-rs6000 (build/rs6000-gen-builtins.o, build/rbtree.o):
2483 Added 'build/' to target, use build/%.o rule.
2484 (build/rs6000-gen-builtins$(build_exeext)): Add 'build/' and
2485 '$(build_exeext)' to target and 'build/' for the *.o files.
2486 (rs6000-builtins.c): Update for those changes; run rs6000-gen-builtins
2489 2021-09-16 Martin Jambor <mjambor@suse.cz>
2491 * cgraph.c (cgraph_node::dump): Do not check caller count sums if
2492 the body has been removed. Remove trailing whitespace.
2494 2021-09-16 Richard Biener <rguenther@suse.de>
2496 PR middle-end/102360
2497 * internal-fn.c (expand_DEFERRED_INIT): Make pattern-init
2498 of non-memory more robust.
2500 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2502 * config/sparc/sparc-opts.h (enum sparc_processor_type): Add LEON5
2503 * config/sparc/sparc.c (struct processor_costs): Add LEON5 costs
2504 (leon5_adjust_cost): Increase cost of store with data dependency
2505 on ALU instruction and FPU anti-dependencies.
2506 (sparc_option_override): Add LEON5 costs
2507 (sparc_adjust_cost): Add LEON5 cost adjustments
2508 * config/sparc/sparc.h: Add LEON5
2509 * config/sparc/sparc.md: Include LEON5 scheduling information
2510 * config/sparc/sparc.opt: Add LEON5
2511 * doc/invoke.texi: Add LEON5
2512 * config/sparc/leon5.md: New file.
2514 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2516 * config/sparc/sparc.md (stack_protect_set32): Add NOP to prevent
2517 sensitive sequence for B2BST errata workaround.
2519 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2521 * config/sparc/sparc.c (sparc_do_work_around_errata): Do not begin
2522 functions with atomic instruction in the UT700 errata workaround.
2524 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2526 * config/sparc/sparc.c (next_active_non_empty_insn): New function
2527 that returns next active non empty assembly instruction.
2528 (sparc_do_work_around_errata): Use new function.
2530 2021-09-16 Daniel Cederman <cederman@gaisler.com>
2532 * config/sparc/sparc.c (store_insn_p): Add predicate for store
2534 (load_insn_p): Add predicate for load attributes.
2535 (sparc_do_work_around_errata): Use new predicates.
2537 2021-09-16 Andreas Larsson <andreas@gaisler.com>
2539 * config/sparc/sparc.c (dump_target_flag_bits): Print bit names for
2542 2021-09-16 Martin Liska <mliska@suse.cz>
2544 * config/mips/netbsd.h: Fix typo in name of a macro.
2546 2021-09-16 liuhongt <hongtao.liu@intel.com>
2548 PR middle-end/102080
2549 * match.pd: Check mask type when doing cond_op related gimple
2551 * tree.c (is_truth_type_for): New function.
2552 * tree.h (is_truth_type_for): New declaration.
2554 2021-09-16 liuhongt <hongtao.liu@intel.com>
2556 * config/i386/avx512fp16intrin.h (_mm512_cvtepi32_ph): New
2558 (_mm512_mask_cvtepi32_ph): Likewise.
2559 (_mm512_maskz_cvtepi32_ph): Likewise.
2560 (_mm512_cvt_roundepi32_ph): Likewise.
2561 (_mm512_mask_cvt_roundepi32_ph): Likewise.
2562 (_mm512_maskz_cvt_roundepi32_ph): Likewise.
2563 (_mm512_cvtepu32_ph): Likewise.
2564 (_mm512_mask_cvtepu32_ph): Likewise.
2565 (_mm512_maskz_cvtepu32_ph): Likewise.
2566 (_mm512_cvt_roundepu32_ph): Likewise.
2567 (_mm512_mask_cvt_roundepu32_ph): Likewise.
2568 (_mm512_maskz_cvt_roundepu32_ph): Likewise.
2569 (_mm512_cvtepi64_ph): Likewise.
2570 (_mm512_mask_cvtepi64_ph): Likewise.
2571 (_mm512_maskz_cvtepi64_ph): Likewise.
2572 (_mm512_cvt_roundepi64_ph): Likewise.
2573 (_mm512_mask_cvt_roundepi64_ph): Likewise.
2574 (_mm512_maskz_cvt_roundepi64_ph): Likewise.
2575 (_mm512_cvtepu64_ph): Likewise.
2576 (_mm512_mask_cvtepu64_ph): Likewise.
2577 (_mm512_maskz_cvtepu64_ph): Likewise.
2578 (_mm512_cvt_roundepu64_ph): Likewise.
2579 (_mm512_mask_cvt_roundepu64_ph): Likewise.
2580 (_mm512_maskz_cvt_roundepu64_ph): Likewise.
2581 (_mm512_cvtepi16_ph): Likewise.
2582 (_mm512_mask_cvtepi16_ph): Likewise.
2583 (_mm512_maskz_cvtepi16_ph): Likewise.
2584 (_mm512_cvt_roundepi16_ph): Likewise.
2585 (_mm512_mask_cvt_roundepi16_ph): Likewise.
2586 (_mm512_maskz_cvt_roundepi16_ph): Likewise.
2587 (_mm512_cvtepu16_ph): Likewise.
2588 (_mm512_mask_cvtepu16_ph): Likewise.
2589 (_mm512_maskz_cvtepu16_ph): Likewise.
2590 (_mm512_cvt_roundepu16_ph): Likewise.
2591 (_mm512_mask_cvt_roundepu16_ph): Likewise.
2592 (_mm512_maskz_cvt_roundepu16_ph): Likewise.
2593 * config/i386/avx512fp16vlintrin.h (_mm_cvtepi32_ph): New
2595 (_mm_mask_cvtepi32_ph): Likewise.
2596 (_mm_maskz_cvtepi32_ph): Likewise.
2597 (_mm256_cvtepi32_ph): Likewise.
2598 (_mm256_mask_cvtepi32_ph): Likewise.
2599 (_mm256_maskz_cvtepi32_ph): Likewise.
2600 (_mm_cvtepu32_ph): Likewise.
2601 (_mm_mask_cvtepu32_ph): Likewise.
2602 (_mm_maskz_cvtepu32_ph): Likewise.
2603 (_mm256_cvtepu32_ph): Likewise.
2604 (_mm256_mask_cvtepu32_ph): Likewise.
2605 (_mm256_maskz_cvtepu32_ph): Likewise.
2606 (_mm_cvtepi64_ph): Likewise.
2607 (_mm_mask_cvtepi64_ph): Likewise.
2608 (_mm_maskz_cvtepi64_ph): Likewise.
2609 (_mm256_cvtepi64_ph): Likewise.
2610 (_mm256_mask_cvtepi64_ph): Likewise.
2611 (_mm256_maskz_cvtepi64_ph): Likewise.
2612 (_mm_cvtepu64_ph): Likewise.
2613 (_mm_mask_cvtepu64_ph): Likewise.
2614 (_mm_maskz_cvtepu64_ph): Likewise.
2615 (_mm256_cvtepu64_ph): Likewise.
2616 (_mm256_mask_cvtepu64_ph): Likewise.
2617 (_mm256_maskz_cvtepu64_ph): Likewise.
2618 (_mm_cvtepi16_ph): Likewise.
2619 (_mm_mask_cvtepi16_ph): Likewise.
2620 (_mm_maskz_cvtepi16_ph): Likewise.
2621 (_mm256_cvtepi16_ph): Likewise.
2622 (_mm256_mask_cvtepi16_ph): Likewise.
2623 (_mm256_maskz_cvtepi16_ph): Likewise.
2624 (_mm_cvtepu16_ph): Likewise.
2625 (_mm_mask_cvtepu16_ph): Likewise.
2626 (_mm_maskz_cvtepu16_ph): Likewise.
2627 (_mm256_cvtepu16_ph): Likewise.
2628 (_mm256_mask_cvtepu16_ph): Likewise.
2629 (_mm256_maskz_cvtepu16_ph): Likewise.
2630 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2631 * config/i386/i386-builtin.def: Add corresponding new builtins.
2632 * config/i386/i386-expand.c
2633 (ix86_expand_args_builtin): Handle new builtin types.
2634 (ix86_expand_round_builtin): Ditto.
2635 * config/i386/i386-modes.def: Declare V2HF and V6HF.
2636 * config/i386/sse.md (VI2H_AVX512VL): New.
2638 (sseintvecmode): Add HF vector modes.
2639 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode><mask_name><round_name>):
2641 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
2642 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
2643 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
2644 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
2645 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask_1): Ditto.
2646 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
2647 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
2648 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
2649 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
2650 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask_1): Ditto.
2651 * config/i386/subst.md (round_qq2phsuff): New subst_attr.
2653 2021-09-16 liuhongt <hongtao.liu@intel.com>
2655 * config/i386/avx512fp16intrin.h (_mm512_cvtph_epi32):
2657 (_mm512_mask_cvtph_epi32): Likewise.
2658 (_mm512_maskz_cvtph_epi32): Likewise.
2659 (_mm512_cvt_roundph_epi32): Likewise.
2660 (_mm512_mask_cvt_roundph_epi32): Likewise.
2661 (_mm512_maskz_cvt_roundph_epi32): Likewise.
2662 (_mm512_cvtph_epu32): Likewise.
2663 (_mm512_mask_cvtph_epu32): Likewise.
2664 (_mm512_maskz_cvtph_epu32): Likewise.
2665 (_mm512_cvt_roundph_epu32): Likewise.
2666 (_mm512_mask_cvt_roundph_epu32): Likewise.
2667 (_mm512_maskz_cvt_roundph_epu32): Likewise.
2668 (_mm512_cvtph_epi64): Likewise.
2669 (_mm512_mask_cvtph_epi64): Likewise.
2670 (_mm512_maskz_cvtph_epi64): Likewise.
2671 (_mm512_cvt_roundph_epi64): Likewise.
2672 (_mm512_mask_cvt_roundph_epi64): Likewise.
2673 (_mm512_maskz_cvt_roundph_epi64): Likewise.
2674 (_mm512_cvtph_epu64): Likewise.
2675 (_mm512_mask_cvtph_epu64): Likewise.
2676 (_mm512_maskz_cvtph_epu64): Likewise.
2677 (_mm512_cvt_roundph_epu64): Likewise.
2678 (_mm512_mask_cvt_roundph_epu64): Likewise.
2679 (_mm512_maskz_cvt_roundph_epu64): Likewise.
2680 (_mm512_cvtph_epi16): Likewise.
2681 (_mm512_mask_cvtph_epi16): Likewise.
2682 (_mm512_maskz_cvtph_epi16): Likewise.
2683 (_mm512_cvt_roundph_epi16): Likewise.
2684 (_mm512_mask_cvt_roundph_epi16): Likewise.
2685 (_mm512_maskz_cvt_roundph_epi16): Likewise.
2686 (_mm512_cvtph_epu16): Likewise.
2687 (_mm512_mask_cvtph_epu16): Likewise.
2688 (_mm512_maskz_cvtph_epu16): Likewise.
2689 (_mm512_cvt_roundph_epu16): Likewise.
2690 (_mm512_mask_cvt_roundph_epu16): Likewise.
2691 (_mm512_maskz_cvt_roundph_epu16): Likewise.
2692 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_epi32):
2694 (_mm_mask_cvtph_epi32): Likewise.
2695 (_mm_maskz_cvtph_epi32): Likewise.
2696 (_mm256_cvtph_epi32): Likewise.
2697 (_mm256_mask_cvtph_epi32): Likewise.
2698 (_mm256_maskz_cvtph_epi32): Likewise.
2699 (_mm_cvtph_epu32): Likewise.
2700 (_mm_mask_cvtph_epu32): Likewise.
2701 (_mm_maskz_cvtph_epu32): Likewise.
2702 (_mm256_cvtph_epu32): Likewise.
2703 (_mm256_mask_cvtph_epu32): Likewise.
2704 (_mm256_maskz_cvtph_epu32): Likewise.
2705 (_mm_cvtph_epi64): Likewise.
2706 (_mm_mask_cvtph_epi64): Likewise.
2707 (_mm_maskz_cvtph_epi64): Likewise.
2708 (_mm256_cvtph_epi64): Likewise.
2709 (_mm256_mask_cvtph_epi64): Likewise.
2710 (_mm256_maskz_cvtph_epi64): Likewise.
2711 (_mm_cvtph_epu64): Likewise.
2712 (_mm_mask_cvtph_epu64): Likewise.
2713 (_mm_maskz_cvtph_epu64): Likewise.
2714 (_mm256_cvtph_epu64): Likewise.
2715 (_mm256_mask_cvtph_epu64): Likewise.
2716 (_mm256_maskz_cvtph_epu64): Likewise.
2717 (_mm_cvtph_epi16): Likewise.
2718 (_mm_mask_cvtph_epi16): Likewise.
2719 (_mm_maskz_cvtph_epi16): Likewise.
2720 (_mm256_cvtph_epi16): Likewise.
2721 (_mm256_mask_cvtph_epi16): Likewise.
2722 (_mm256_maskz_cvtph_epi16): Likewise.
2723 (_mm_cvtph_epu16): Likewise.
2724 (_mm_mask_cvtph_epu16): Likewise.
2725 (_mm_maskz_cvtph_epu16): Likewise.
2726 (_mm256_cvtph_epu16): Likewise.
2727 (_mm256_mask_cvtph_epu16): Likewise.
2728 (_mm256_maskz_cvtph_epu16): Likewise.
2729 * config/i386/i386-builtin-types.def: Add new builtin types.
2730 * config/i386/i386-builtin.def: Add new builtins.
2731 * config/i386/i386-expand.c
2732 (ix86_expand_args_builtin): Handle new builtin types.
2733 (ix86_expand_round_builtin): Ditto.
2734 * config/i386/sse.md (sseintconvert): New.
2736 (UNSPEC_US_FIX_NOTRUNC): Ditto.
2737 (sseintconvertsignprefix): Ditto.
2738 (avx512fp16_vcvtph2<sseintconvertsignprefix><sseintconvert>_<mode><mask_name><round_name>):
2741 2021-09-16 liuhongt <hongtao.liu@intel.com>
2743 * config/i386/avx512fp16intrin.h: (_mm_cvtsi16_si128):
2745 (_mm_cvtsi128_si16): Likewise.
2746 (_mm_mask_load_sh): Likewise.
2747 (_mm_maskz_load_sh): Likewise.
2748 (_mm_mask_store_sh): Likewise.
2749 (_mm_move_sh): Likewise.
2750 (_mm_mask_move_sh): Likewise.
2751 (_mm_maskz_move_sh): Likewise.
2752 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2753 * config/i386/i386-builtin.def: Add corresponding new builtins.
2754 * config/i386/i386-expand.c
2755 (ix86_expand_special_args_builtin): Handle new builtin types.
2756 (ix86_expand_vector_init_one_nonzero): Adjust for FP16 target.
2757 * config/i386/sse.md (VI2F): New mode iterator.
2758 (vec_set<mode>_0): Use new mode iterator.
2759 (avx512f_mov<ssescalarmodelower>_mask): Adjust for HF vector mode.
2760 (avx512f_store<mode>_mask): Ditto.
2762 2021-09-16 Kewen Lin <linkw@linux.ibm.com>
2764 * config/rs6000/rs6000.opt (-mtoc-fusion): Remove.
2766 2021-09-15 David Edelsohn <dje.gcc@gmail.com>
2768 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
2769 Proceed if no symbol summary or the symbol alias flag is false.
2771 2021-09-15 Jakub Jelinek <jakub@redhat.com>
2775 * varasm.c (output_constructor_regular_field): Instead of assertion
2776 that array_size_for_constructor result is equal to size of
2777 TREE_TYPE (local->val) in bytes, assert that the type size is greater
2778 or equal to array_size_for_constructor result and use type size as
2781 2021-09-15 Martin Liska <mliska@suse.cz>
2784 * config/i386/vxworks.h: Use new macro TARGET_CPU_P.
2786 2021-09-15 Martin Liska <mliska@suse.cz>
2789 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
2790 Check that we have a symbol summary for a symbol.
2792 2021-09-15 Richard Biener <rguenther@suse.de>
2795 * config/rs6000/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
2796 to inherit from elfos.h
2798 2021-09-15 liuhongt <hongtao.liu@intel.com>
2801 * config/i386/i386-expand.c
2802 (ix86_expand_vector_init_interleave): Use puncklwd to pack 2
2804 (ix86_expand_vector_set): Use blendw instead of pinsrw.
2805 * config/i386/i386.c (ix86_can_change_mode_class): Adjust for
2806 AVX512FP16 which supports 16bit vector load.
2807 * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>):
2809 (avx512bw_interleave_high<mode><mask_name>): .. this, and
2810 extend to V32HFmode.
2811 (avx2_interleave_highv16hi<mask_name>): Rename to ..
2812 (avx2_interleave_high<mode><mask_name>): .. this, and extend
2814 (vec_interleave_highv8hi<mask_name>): Rename to ..
2815 (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode.
2816 (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>):
2818 (<mask_codefor>avx512bw_interleave_low<mode><mask_name>):
2819 this, and extend to V32HFmode.
2820 (avx2_interleave_lowv16hi<mask_name>): Rename to ..
2821 (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode.
2822 (vec_interleave_lowv8hi<mask_name>): Rename to ..
2823 (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode.
2824 (sse4_1_pblendw): Rename to ..
2825 (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode.
2826 (avx2_pblendph): New define_expand.
2827 (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use
2828 sseintmodesuffix instead of ssemodesuffix.
2829 (blendsuf): New mode attr.
2831 2021-09-15 Richard Biener <rguenther@suse.de>
2833 * tree-vectorizer.h (dr_misalignment): Move out of line.
2834 (dr_target_alignment): New.
2835 (DR_TARGET_ALIGNMENT): Wrap dr_target_alignment.
2836 (set_dr_target_alignment): New.
2837 (SET_DR_TARGET_ALIGNMENT): Wrap set_dr_target_alignment.
2838 * tree-vect-data-refs.c (dr_misalignment): Compute and
2839 return the group members misalignment.
2840 (vect_compute_data_ref_alignment): Use SET_DR_TARGET_ALIGNMENT.
2841 (vect_analyze_data_refs_alignment): Compute alignment only
2842 for the first element of a DR group.
2843 (vect_slp_analyze_node_alignment): Likewise.
2845 2021-09-15 Hongyu Wang <hongyu.wang@intel.com>
2847 * config/i386/avx512fp16intrin.h: Adjust all builtin calls.
2848 * config/i386/avx512fp16vlintrin.h: Likewise.
2849 * config/i386/i386-builtin.def: Adjust builtin name and
2850 enumeration to match AVX512F style.
2852 2021-09-15 Richard Biener <rguenther@suse.de>
2854 PR tree-optimization/102318
2855 * tree-vect-loop.c (vect_transform_cycle_phi): Revert
2856 previous change and do the mode conversion separately from
2857 the sign conversion.
2859 2021-09-15 Hongtao Liu <hongtao.liu@intel.com>
2860 Peter Cordes <peter@cordes.ca>
2863 * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
2864 (*vec_extract<mode><ssescalarmodelower>_valign): Output
2865 vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
2868 2021-09-15 Richard Biener <rguenther@suse.de>
2870 * config.gcc: Remove vax-*-openbsd* configuration.
2872 2021-09-15 Richard Biener <rguenther@suse.de>
2874 * config.gcc: Remove m68k-openbsd.
2876 2021-09-15 Max Filippov <jcmvbkbc@gmail.com>
2879 * config/xtensa/t-xtensa (TM_H): Add include/xtensa-config.h.
2881 2021-09-14 Peter Bergner <bergner@linux.ibm.com>
2883 * config/rs6000/mma.md (unspec): Delete UNSPEC_MMA_XXSETACCZ.
2884 (unspecv): Add UNSPECV_MMA_XXSETACCZ.
2885 (*mma_xxsetaccz): Delete.
2886 (mma_xxsetaccz): Change to define_insn. Remove operand 1.
2887 Use UNSPECV_MMA_XXSETACCZ. Update comment.
2888 * config/rs6000/rs6000.c (rs6000_rtx_costs): Use UNSPECV_MMA_XXSETACCZ.
2890 2021-09-14 Iain Sandoe <iain@sandoe.co.uk>
2892 * Makefile.in: Remove variables related to applying no-PIE
2893 to the exes on $build.
2894 * configure: Regenerate.
2895 * configure.ac: Remove configuration related to applying
2896 no-PIE to the exes on $build.
2898 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
2900 * config/arc/arc.md (doloop_end): Add missing mode.
2901 (loop_end): Likewise.
2903 2021-09-14 Jakub Jelinek <jakub@redhat.com>
2905 * gimplify.c (goa_stabilize_expr): Add depth argument, propagate
2906 it to recursive calls, for depth above 7 just gimplify or return.
2907 Perform a test even for MODIFY_EXPR, ADDR_EXPR, COMPOUND_EXPR with
2908 __builtin_clear_padding and TARGET_EXPR.
2909 (gimplify_omp_atomic): Adjust goa_stabilize_expr callers.
2911 2021-09-14 liuhongt <hongtao.liu@intel.com>
2913 * config/i386/avx512fp16intrin.h (_mm_fpclass_sh_mask):
2915 (_mm_mask_fpclass_sh_mask): Likewise.
2916 (_mm512_mask_fpclass_ph_mask): Likewise.
2917 (_mm512_fpclass_ph_mask): Likewise.
2918 (_mm_getexp_sh): Likewise.
2919 (_mm_mask_getexp_sh): Likewise.
2920 (_mm_maskz_getexp_sh): Likewise.
2921 (_mm512_getexp_ph): Likewise.
2922 (_mm512_mask_getexp_ph): Likewise.
2923 (_mm512_maskz_getexp_ph): Likewise.
2924 (_mm_getexp_round_sh): Likewise.
2925 (_mm_mask_getexp_round_sh): Likewise.
2926 (_mm_maskz_getexp_round_sh): Likewise.
2927 (_mm512_getexp_round_ph): Likewise.
2928 (_mm512_mask_getexp_round_ph): Likewise.
2929 (_mm512_maskz_getexp_round_ph): Likewise.
2930 (_mm_getmant_sh): Likewise.
2931 (_mm_mask_getmant_sh): Likewise.
2932 (_mm_maskz_getmant_sh): Likewise.
2933 (_mm512_getmant_ph): Likewise.
2934 (_mm512_mask_getmant_ph): Likewise.
2935 (_mm512_maskz_getmant_ph): Likewise.
2936 (_mm_getmant_round_sh): Likewise.
2937 (_mm_mask_getmant_round_sh): Likewise.
2938 (_mm_maskz_getmant_round_sh): Likewise.
2939 (_mm512_getmant_round_ph): Likewise.
2940 (_mm512_mask_getmant_round_ph): Likewise.
2941 (_mm512_maskz_getmant_round_ph): Likewise.
2942 * config/i386/avx512fp16vlintrin.h (_mm_mask_fpclass_ph_mask):
2944 (_mm_fpclass_ph_mask): Likewise.
2945 (_mm256_mask_fpclass_ph_mask): Likewise.
2946 (_mm256_fpclass_ph_mask): Likewise.
2947 (_mm256_getexp_ph): Likewise.
2948 (_mm256_mask_getexp_ph): Likewise.
2949 (_mm256_maskz_getexp_ph): Likewise.
2950 (_mm_getexp_ph): Likewise.
2951 (_mm_mask_getexp_ph): Likewise.
2952 (_mm_maskz_getexp_ph): Likewise.
2953 (_mm256_getmant_ph): Likewise.
2954 (_mm256_mask_getmant_ph): Likewise.
2955 (_mm256_maskz_getmant_ph): Likewise.
2956 (_mm_getmant_ph): Likewise.
2957 (_mm_mask_getmant_ph): Likewise.
2958 (_mm_maskz_getmant_ph): Likewise.
2959 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
2960 * config/i386/i386-builtin.def: Add corresponding new builtins.
2961 * config/i386/i386-expand.c
2962 (ix86_expand_args_builtin): Handle new builtin types.
2963 (ix86_expand_round_builtin): Ditto.
2964 * config/i386/sse.md (vecmemsuffix): Add HF vector modes.
2965 (<avx512>_getexp<mode><mask_name><round_saeonly_name>): Adjust
2966 to support HF vector modes.
2967 (avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name):
2969 (avx512dq_fpclass<mode><mask_scalar_merge_name>): Ditto.
2970 (avx512dq_vmfpclass<mode><mask_scalar_merge_name>): Ditto.
2971 (<avx512>_getmant<mode><mask_name><round_saeonly_name>): Ditto.
2972 (avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
2975 2021-09-14 liuhongt <hongtao.liu@intel.com>
2977 * config/i386/avx512fp16intrin.h (_mm512_reduce_ph):
2979 (_mm512_mask_reduce_ph): Likewise.
2980 (_mm512_maskz_reduce_ph): Likewise.
2981 (_mm512_reduce_round_ph): Likewise.
2982 (_mm512_mask_reduce_round_ph): Likewise.
2983 (_mm512_maskz_reduce_round_ph): Likewise.
2984 (_mm_reduce_sh): Likewise.
2985 (_mm_mask_reduce_sh): Likewise.
2986 (_mm_maskz_reduce_sh): Likewise.
2987 (_mm_reduce_round_sh): Likewise.
2988 (_mm_mask_reduce_round_sh): Likewise.
2989 (_mm_maskz_reduce_round_sh): Likewise.
2990 (_mm512_roundscale_ph): Likewise.
2991 (_mm512_mask_roundscale_ph): Likewise.
2992 (_mm512_maskz_roundscale_ph): Likewise.
2993 (_mm512_roundscale_round_ph): Likewise.
2994 (_mm512_mask_roundscale_round_ph): Likewise.
2995 (_mm512_maskz_roundscale_round_ph): Likewise.
2996 (_mm_roundscale_sh): Likewise.
2997 (_mm_mask_roundscale_sh): Likewise.
2998 (_mm_maskz_roundscale_sh): Likewise.
2999 (_mm_roundscale_round_sh): Likewise.
3000 (_mm_mask_roundscale_round_sh): Likewise.
3001 (_mm_maskz_roundscale_round_sh): Likewise.
3002 * config/i386/avx512fp16vlintrin.h: (_mm_reduce_ph):
3004 (_mm_mask_reduce_ph): Likewise.
3005 (_mm_maskz_reduce_ph): Likewise.
3006 (_mm256_reduce_ph): Likewise.
3007 (_mm256_mask_reduce_ph): Likewise.
3008 (_mm256_maskz_reduce_ph): Likewise.
3009 (_mm_roundscale_ph): Likewise.
3010 (_mm_mask_roundscale_ph): Likewise.
3011 (_mm_maskz_roundscale_ph): Likewise.
3012 (_mm256_roundscale_ph): Likewise.
3013 (_mm256_mask_roundscale_ph): Likewise.
3014 (_mm256_maskz_roundscale_ph): Likewise.
3015 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3016 * config/i386/i386-builtin.def: Add corresponding new builtins.
3017 * config/i386/i386-expand.c
3018 (ix86_expand_args_builtin): Handle new builtin types.
3019 (ix86_expand_round_builtin): Ditto.
3020 * config/i386/sse.md (<mask_codefor>reducep<mode><mask_name>):
3022 (<mask_codefor>reducep<mode><mask_name><round_saeonly_name>):
3023 ... this, and adjust for round operands.
3024 (reduces<mode><mask_scalar_name>): Likewise, with ...
3025 (reduces<mode><mask_scalar_name><round_saeonly_scalar_name):
3027 (<avx512>_rndscale<mode><mask_name><round_saeonly_name>):
3028 Adjust for HF vector modes.
3029 (avx512f_rndscale<mode><mask_scalar_name><round_saeonly_scalar_name>):
3031 (*avx512f_rndscale<mode><round_saeonly_name>): Ditto.
3033 2021-09-14 liuhongt <hongtao.liu@intel.com>
3035 * config/i386/avx512fp16intrin.h: (_mm512_rcp_ph):
3037 (_mm512_mask_rcp_ph): Likewise.
3038 (_mm512_maskz_rcp_ph): Likewise.
3039 (_mm_rcp_sh): Likewise.
3040 (_mm_mask_rcp_sh): Likewise.
3041 (_mm_maskz_rcp_sh): Likewise.
3042 (_mm512_scalef_ph): Likewise.
3043 (_mm512_mask_scalef_ph): Likewise.
3044 (_mm512_maskz_scalef_ph): Likewise.
3045 (_mm512_scalef_round_ph): Likewise.
3046 (_mm512_mask_scalef_round_ph): Likewise.
3047 (_mm512_maskz_scalef_round_ph): Likewise.
3048 (_mm_scalef_sh): Likewise.
3049 (_mm_mask_scalef_sh): Likewise.
3050 (_mm_maskz_scalef_sh): Likewise.
3051 (_mm_scalef_round_sh): Likewise.
3052 (_mm_mask_scalef_round_sh): Likewise.
3053 (_mm_maskz_scalef_round_sh): Likewise.
3054 * config/i386/avx512fp16vlintrin.h (_mm_rcp_ph):
3056 (_mm256_rcp_ph): Likewise.
3057 (_mm_mask_rcp_ph): Likewise.
3058 (_mm256_mask_rcp_ph): Likewise.
3059 (_mm_maskz_rcp_ph): Likewise.
3060 (_mm256_maskz_rcp_ph): Likewise.
3061 (_mm_scalef_ph): Likewise.
3062 (_mm256_scalef_ph): Likewise.
3063 (_mm_mask_scalef_ph): Likewise.
3064 (_mm256_mask_scalef_ph): Likewise.
3065 (_mm_maskz_scalef_ph): Likewise.
3066 (_mm256_maskz_scalef_ph): Likewise.
3067 * config/i386/i386-builtin.def: Add new builtins.
3068 * config/i386/sse.md (VFH_AVX512VL): New.
3069 (avx512fp16_rcp<mode>2<mask_name>): Ditto.
3070 (avx512fp16_vmrcpv8hf2<mask_scalar_name>): Ditto.
3071 (avx512f_vmscalef<mode><mask_scalar_name><round_scalar_name>):
3072 Adjust to support HF vector modes.
3073 (<avx512>_scalef<mode><mask_name><round_name>): Ditto.
3075 2021-09-14 liuhongt <hongtao.liu@intel.com>
3077 * config/i386/avx512fp16intrin.h: (_mm512_sqrt_ph):
3079 (_mm512_mask_sqrt_ph): Likewise.
3080 (_mm512_maskz_sqrt_ph): Likewise.
3081 (_mm512_sqrt_round_ph): Likewise.
3082 (_mm512_mask_sqrt_round_ph): Likewise.
3083 (_mm512_maskz_sqrt_round_ph): Likewise.
3084 (_mm512_rsqrt_ph): Likewise.
3085 (_mm512_mask_rsqrt_ph): Likewise.
3086 (_mm512_maskz_rsqrt_ph): Likewise.
3087 (_mm_rsqrt_sh): Likewise.
3088 (_mm_mask_rsqrt_sh): Likewise.
3089 (_mm_maskz_rsqrt_sh): Likewise.
3090 (_mm_sqrt_sh): Likewise.
3091 (_mm_mask_sqrt_sh): Likewise.
3092 (_mm_maskz_sqrt_sh): Likewise.
3093 (_mm_sqrt_round_sh): Likewise.
3094 (_mm_mask_sqrt_round_sh): Likewise.
3095 (_mm_maskz_sqrt_round_sh): Likewise.
3096 * config/i386/avx512fp16vlintrin.h (_mm_sqrt_ph): New intrinsic.
3097 (_mm256_sqrt_ph): Likewise.
3098 (_mm_mask_sqrt_ph): Likewise.
3099 (_mm256_mask_sqrt_ph): Likewise.
3100 (_mm_maskz_sqrt_ph): Likewise.
3101 (_mm256_maskz_sqrt_ph): Likewise.
3102 (_mm_rsqrt_ph): Likewise.
3103 (_mm256_rsqrt_ph): Likewise.
3104 (_mm_mask_rsqrt_ph): Likewise.
3105 (_mm256_mask_rsqrt_ph): Likewise.
3106 (_mm_maskz_rsqrt_ph): Likewise.
3107 (_mm256_maskz_rsqrt_ph): Likewise.
3108 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3109 * config/i386/i386-builtin.def: Add corresponding new builtins.
3110 * config/i386/i386-expand.c
3111 (ix86_expand_args_builtin): Handle new builtins.
3112 (ix86_expand_round_builtin): Ditto.
3113 * config/i386/sse.md (VF_AVX512FP16VL): New.
3114 (sqrt<mode>2): Adjust for HF vector modes.
3115 (<sse>_sqrt<mode>2<mask_name><round_name>): Likewise.
3116 (<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
3118 (<sse>_rsqrt<mode>2<mask_name>): New.
3119 (avx512fp16_vmrsqrtv8hf2<mask_scalar_name>): Likewise.
3121 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
3124 * diagnostic-spec.c (warning_suppressed_at, copy_warning): Handle
3125 'RESERVED_LOCATION_P' locations.
3126 * warning-control.cc (get_nowarn_spec, suppress_warning)
3127 (copy_warning): Likewise.
3129 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
3131 * diagnostic-spec.h (typedef xint_hash_t): Use 'location_t' instead of...
3132 (typedef key_type_t): ... this. Remove.
3133 (nowarn_map): Document.
3134 * diagnostic-spec.c (nowarn_map): Likewise.
3135 * warning-control.cc (convert_to_key): Evolve functions into...
3136 (get_location): ... these. Adjust all users.
3138 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
3140 * warning-control.cc (copy_warning): Remove 'nowarn_map' setup.
3142 2021-09-13 Jason Merrill <jason@redhat.com>
3144 * params.opt: Add destructive-interference-size and
3145 constructive-interference-size.
3146 * doc/invoke.texi: Document them.
3147 * config/aarch64/aarch64.c (aarch64_override_options_internal):
3149 * config/arm/arm.c (arm_option_override): Set them.
3150 * config/i386/i386-options.c (ix86_option_override_internal):
3153 2021-09-13 Martin Liska <mliska@suse.cz>
3154 H.J. Lu <hjl.tools@gmail.com>
3157 * common/config/i386/cpuinfo.h (cpu_indicator_init): Add support
3158 for x86-64 micro levels for __builtin_cpu_supports.
3159 * common/config/i386/i386-cpuinfo.h (enum feature_priority):
3160 Add priorities for the micro-arch levels.
3161 (enum processor_features): Add new features.
3162 * common/config/i386/i386-isas.h: Add micro-arch features.
3163 * config/i386/i386-builtins.c (get_builtin_code_for_version):
3164 Support the micro-arch levels by callsing
3165 __builtin_cpu_supports.
3166 * doc/extend.texi: Document that the levels are support by
3167 __builtin_cpu_supports.
3169 2021-09-13 Andrew Pinski <apinski@marvell.com>
3172 * config/aarch64/aarch64-builtins.c (aarch64_fold_builtin_lane_check):
3174 (aarch64_general_fold_builtin): Handle AARCH64_SIMD_BUILTIN_LANE_CHECK.
3175 (aarch64_general_gimple_fold_builtin): Likewise.
3177 2021-09-13 Andrew Pinski <apinski@marvell.com>
3179 * config.gcc: Add m32r-*-linux* and m32rle-*-linux*
3180 to the Unsupported targets list.
3181 Remove support for m32r-*-linux* and m32rle-*-linux*.
3182 * config/m32r/linux.h: Removed.
3183 * config/m32r/t-linux: Removed.
3185 2021-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3188 * config/aarch64/aarch64.c (aarch64_classify_address): Don't allow
3189 register index for SVE predicate modes.
3191 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
3193 * tree-ssa-threadbackward.c
3194 (back_threader_profitability::profitable_path_p): Remove FSM
3196 (back_threader_registry::register_path): Same.
3197 * tree-ssa-threadedge.c
3198 (jump_threader::simplify_control_stmt_condition): Same.
3199 * tree-ssa-threadupdate.c (jt_path_registry::jt_path_registry):
3200 Add backedge_threads argument.
3201 (fwd_jt_path_registry::fwd_jt_path_registry): Pass
3202 backedge_threads argument.
3203 (back_jt_path_registry::back_jt_path_registry): Same.
3204 (dump_jump_thread_path): Adjust for FSM removal.
3205 (back_jt_path_registry::rewire_first_differing_edge): Same.
3206 (back_jt_path_registry::adjust_paths_after_duplication): Same.
3207 (back_jt_path_registry::update_cfg): Same.
3208 (jt_path_registry::register_jump_thread): Same.
3209 * tree-ssa-threadupdate.h (enum jump_thread_edge_type): Remove
3211 (class back_jt_path_registry): Add backedge_threads to
3214 2021-09-13 Martin Liska <mliska@suse.cz>
3217 * asan.h (sanitize_coverage_p): Handle when fn == NULL.
3219 2021-09-13 H.J. Lu <hjl.tools@gmail.com>
3222 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): New.
3223 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
3224 (MOVE_MAX): Check TARGET_AVX256_MOVE_BY_PIECES and
3225 TARGET_AVX256_STORE_BY_PIECES instead of
3226 TARGET_AVX256_SPLIT_UNALIGNED_LOAD and
3227 TARGET_AVX256_SPLIT_UNALIGNED_STORE.
3228 (STORE_MAX_PIECES): Check TARGET_AVX256_STORE_BY_PIECES instead
3229 of TARGET_AVX256_SPLIT_UNALIGNED_STORE.
3230 * config/i386/x86-tune.def (X86_TUNE_AVX256_MOVE_BY_PIECES): New.
3231 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
3233 2021-09-13 liuhongt <hongtao.liu@intel.com>
3236 * expmed.c (extract_bit_field_using_extv): Use
3237 gen_lowpart_if_possible instead of gen_lowpart to avoid ICE.
3239 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
3241 * Makefile.in (OBJS): Add value-pointer-equiv.o.
3242 * gimple-ssa-evrp.c (class ssa_equiv_stack): Move to
3243 value-pointer-equiv.*.
3244 (ssa_equiv_stack::ssa_equiv_stack): Same.
3245 (ssa_equiv_stack::enter): Same.
3246 (ssa_equiv_stack::leave): Same.
3247 (ssa_equiv_stack::push_replacement): Same.
3248 (ssa_equiv_stack::get_replacement): Same.
3249 (is_pointer_ssa): Same.
3250 (class pointer_equiv_analyzer): Same.
3251 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
3252 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Same.
3253 (pointer_equiv_analyzer::set_global_equiv): Same.
3254 (pointer_equiv_analyzer::set_cond_equiv): Same.
3255 (pointer_equiv_analyzer::get_equiv): Same.
3256 (pointer_equiv_analyzer::enter): Same.
3257 (pointer_equiv_analyzer::leave): Same.
3258 (pointer_equiv_analyzer::get_equiv_expr): Same.
3259 (pta_valueize): Same.
3260 (pointer_equiv_analyzer::visit_stmt): Same.
3261 (pointer_equiv_analyzer::visit_edge): Same.
3262 (hybrid_folder::value_of_expr): Same.
3263 (hybrid_folder::value_on_edge): Same.
3264 * value-pointer-equiv.cc: New file.
3265 * value-pointer-equiv.h: New file.
3267 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
3270 * gimple-fold.c (gimple_fold_builtin_memory_op): Allow folding
3271 memcpy if the size is not more than MOVE_MAX * MOVE_RATIO.
3273 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
3276 * config/arm/arm.md (movmisaligndi): New define_expand.
3277 * config/arm/vec-common.md (movmisalign<mode>): Iterate over VDQ mode.
3279 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
3282 * emit-rtl.c (gen_highpart): Use adjust_address to handle
3283 MEM rather than calling simplify_gen_subreg.
3285 2021-09-13 Jan-Benedict Glaw <jbglaw@ług-owl.de>
3287 * config/alpha/vms.h (INIT_CUMULATIVE_ARGS): Wrap multi-statment
3288 define into a block.
3290 2021-09-13 Richard Biener <rguenther@suse.de>
3292 * config/darwin.h (DARWIN_PREFER_DWARF): Do not define.
3293 * config/i386/darwin.h (PREFERRED_DEBUGGING_TYPE): Do not
3294 change based on DARWIN_PREFER_DWARF not being defined.
3296 2021-09-13 Richard Biener <rguenther@suse.de>
3298 * config/i386/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
3299 to inherit from elfos.h
3301 2021-09-13 Richard Biener <rguenther@suse.de>
3303 * config.gcc: Add cr16-*-* to the list of obsoleted targets.
3305 2021-09-13 Richard Biener <rguenther@suse.de>
3307 * config/avr/elf.h (PREFERRED_DEBUGGING_TYPE): Remove
3308 override, pick up DWARF2_DEBUG define from elfos.h
3310 2021-09-13 Richard Biener <rguenther@suse.de>
3312 * config/rx/rx.h (PREFERRED_DEBUGGING_TYPE): Always define to
3315 2021-09-13 Richard Biener <rguenther@suse.de>
3317 * config/alpha/vms.h (PREFERRED_DEBUGGING_TYPE): Define to
3320 2021-09-13 Richard Biener <rguenther@suse.de>
3322 * config/i386/cygming.h: Always default to DWARF2 debugging.
3323 Do not define DBX_DEBUGGING_INFO, that's done via dbxcoff.h
3325 * doc/install.texi: Document binutils 2.16 as minimum
3326 requirement for mingw.
3328 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
3330 * config/rs6000/rs6000.c (struct rs6000_cost_data): New members
3331 nstmts, nloads and extra_ctor_cost.
3332 (rs6000_density_test): Add load density related heuristics. Do
3333 extra costing on vector construction statements if need.
3334 (rs6000_init_cost): Init new members.
3335 (rs6000_update_target_cost_per_stmt): New function.
3336 (rs6000_add_stmt_cost): Factor vect_nonmem hunk out to function
3337 rs6000_update_target_cost_per_stmt and call it.
3339 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
3341 * config/rs6000/rs6000.c (struct rs6000_cost_data): Remove typedef.
3342 (rs6000_init_cost): Adjust.
3344 2021-09-13 liuhongt <hongtao.liu@intel.com>
3346 * config/i386/i386.md: (UNSPEC_COPYSIGN): Remove.
3347 (UNSPEC_XORSIGN): Ditto.
3349 2021-09-12 Roger Sayle <roger@nextmovesoftware.com>
3351 * expr.c (convert_move): Preserve SUBREG_PROMOTED_VAR_P when
3352 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
3355 2021-09-11 Aldy Hernandez <aldyh@redhat.com>
3357 * tree-ssa-threadbackward.c (class back_threader_registry): Use
3358 back_jt_path_registry.
3359 * tree-ssa-threadedge.c (jump_threader::jump_threader): Use
3360 fwd_jt_path_registry.
3361 * tree-ssa-threadedge.h (class jump_threader): Same..
3362 * tree-ssa-threadupdate.c
3363 (jump_thread_path_registry::jump_thread_path_registry): Rename...
3364 (jt_path_registry::jt_path_registry): ...to this.
3365 (jump_thread_path_registry::~jump_thread_path_registry): Rename...
3366 (jt_path_registry::~jt_path_registry): ...this.
3367 (fwd_jt_path_registry::fwd_jt_path_registry): New.
3368 (fwd_jt_path_registry::~fwd_jt_path_registry): New.
3369 (jump_thread_path_registry::allocate_thread_edge): Rename...
3370 (jt_path_registry::allocate_thread_edge): ...to this.
3371 (jump_thread_path_registry::allocate_thread_path): Rename...
3372 (jt_path_registry::allocate_thread_path): ...to this.
3373 (jump_thread_path_registry::lookup_redirection_data): Rename...
3374 (fwd_jt_path_registry::lookup_redirection_data): ...to this.
3375 (jump_thread_path_registry::thread_block_1): Rename...
3376 (fwd_jt_path_registry::thread_block_1): ...to this.
3377 (jump_thread_path_registry::thread_block): Rename...
3378 (fwd_jt_path_registry::thread_block): ...to this.
3379 (jt_path_registry::thread_through_loop_header): Rename...
3380 (fwd_jt_path_registry::thread_through_loop_header): ...to this.
3381 (jump_thread_path_registry::mark_threaded_blocks): Rename...
3382 (fwd_jt_path_registry::mark_threaded_blocks): ...to this.
3383 (jump_thread_path_registry::debug_path): Rename...
3384 (jt_path_registry::debug_path): ...to this.
3385 (jump_thread_path_registry::dump): Rename...
3386 (jt_path_registry::debug): ...to this.
3387 (jump_thread_path_registry::rewire_first_differing_edge): Rename...
3388 (back_jt_path_registry::rewire_first_differing_edge): ...to this.
3389 (jump_thread_path_registry::adjust_paths_after_duplication): Rename...
3390 (back_jt_path_registry::adjust_paths_after_duplication): ...to this.
3391 (jump_thread_path_registry::duplicate_thread_path): Rename...
3392 (back_jt_path_registry::duplicate_thread_path): ...to this. Also,
3393 drop ill-formed candidates.
3394 (jump_thread_path_registry::remove_jump_threads_including): Rename...
3395 (fwd_jt_path_registry::remove_jump_threads_including): ...to this.
3396 (jt_path_registry::thread_through_all_blocks): New.
3397 (back_jt_path_registry::update_cfg): New.
3398 (fwd_jt_path_registry::update_cfg): New.
3399 (jump_thread_path_registry::register_jump_thread): Rename...
3400 (jt_path_registry::register_jump_thread): ...to this.
3401 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
3403 (class jt_path_registry): ...here.
3404 (class fwd_jt_path_registry): New.
3405 (class back_jt_path_registry): New.
3407 2021-09-10 liuhongt <hongtao.liu@intel.com>
3410 2021-09-01 liuhongt <hongtao.liu@intel.com>
3412 * emit-rtl.c (validate_subreg): Get rid of all float-int
3415 2021-09-10 Jakub Jelinek <jakub@redhat.com>
3417 * tree-core.h (enum omp_memory_order): Add OMP_MEMORY_ORDER_MASK,
3418 OMP_FAIL_MEMORY_ORDER_UNSPECIFIED, OMP_FAIL_MEMORY_ORDER_RELAXED,
3419 OMP_FAIL_MEMORY_ORDER_ACQUIRE, OMP_FAIL_MEMORY_ORDER_RELEASE,
3420 OMP_FAIL_MEMORY_ORDER_ACQ_REL, OMP_FAIL_MEMORY_ORDER_SEQ_CST and
3421 OMP_FAIL_MEMORY_ORDER_MASK enumerators.
3422 (OMP_FAIL_MEMORY_ORDER_SHIFT): Define.
3423 * gimple-pretty-print.c (dump_gimple_omp_atomic_load,
3424 dump_gimple_omp_atomic_store): Print [weak] for weak atomic
3426 * gimple.h (enum gf_mask): Change GF_OMP_ATOMIC_MEMORY_ORDER
3427 to 6-bit mask, adjust GF_OMP_ATOMIC_NEED_VALUE value and add
3429 (gimple_omp_atomic_weak_p, gimple_omp_atomic_set_weak): New inline
3431 * tree.h (OMP_ATOMIC_WEAK): Define.
3432 * tree-pretty-print.c (dump_omp_atomic_memory_order): Adjust for
3433 fail memory order being encoded in the same enum and also print
3434 fail clause if present.
3435 (dump_generic_node): Print weak clause if OMP_ATOMIC_WEAK.
3436 * gimplify.c (goa_stabilize_expr): Add target_expr and rhs arguments,
3437 handle pre_p == NULL case as a test mode that only returns value
3438 but doesn't change gimplify nor change anything otherwise, adjust
3439 recursive calls, add MODIFY_EXPR, ADDR_EXPR, COND_EXPR, TARGET_EXPR
3440 and CALL_EXPR handling, adjust COMPOUND_EXPR handling for
3441 __builtin_clear_padding calls, for !rhs gimplify as lvalue rather
3443 (gimplify_omp_atomic): Adjust goa_stabilize_expr caller. Handle
3444 COND_EXPR rhs. Set weak flag on gimple load/store for
3446 * omp-expand.c (omp_memory_order_to_fail_memmodel): New function.
3447 (omp_memory_order_to_memmodel): Adjust for fail clause encoded
3449 (expand_omp_atomic_cas): New function.
3450 (expand_omp_atomic_pipeline): Use omp_memory_order_to_fail_memmodel
3452 (expand_omp_atomic): Attempt to optimize atomic compare and exchange
3453 using expand_omp_atomic_cas.
3455 2021-09-10 Aldy Hernandez <aldyh@redhat.com>
3456 Michael Matz <matz@suse.de>
3458 * tree-pass.h (PROP_loop_opts_done): New.
3459 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
3460 Intersect with global range.
3461 * tree-ssa-loop.c (tree_ssa_loop_done): Set PROP_loop_opts_done.
3462 * tree-ssa-threadbackward.c
3463 (back_threader_profitability::profitable_path_p): Disable
3464 threading through latches until after loop optimizations have run.
3466 2021-09-10 David Faust <david.faust@oracle.com>
3468 * doc/invoke.texi: Document BPF -mcpu, -mjmpext, -mjmp32 and -malu32
3471 2021-09-10 David Faust <david.faust@oracle.com>
3473 * config/bpf/bpf-opts.h (bpf_isa_version): New enum.
3474 * config/bpf/bpf-protos.h (bpf_expand_cbranch): New.
3475 * config/bpf/bpf.c (bpf_option_override): Handle -mcpu option.
3476 (bpf_expand_cbranch): New function.
3477 * config/bpf/bpf.md (AM mode iterator): Conditionalize support for SI
3479 (zero_extendsidi2): Only use mov32 instruction if it is available.
3480 (SIM mode iterator): Conditionalize support for SI mode.
3481 (JM mode iterator): New.
3482 (cbranchdi4): Update name, use new JM iterator. Use bpf_expand_cbranch.
3483 (*branch_on_di): Update name, use new JM iterator.
3484 * config/bpf/bpf.opt: (mjmpext): New option.
3488 (bpf_isa): New enum.
3490 2021-09-10 David Faust <david.faust@oracle.com>
3492 * config/bpf/bpf.md (zero_extendhidi2): Add new output template
3493 for register-to-register extensions.
3494 (zero_extendqidi2): Likewise.
3496 2021-09-10 Richard Biener <rguenther@suse.de>
3498 PR middle-end/102273
3499 * internal-fn.c (expand_DEFERRED_INIT): Always expand non-SSA vars.
3501 2021-09-10 Richard Biener <rguenther@suse.de>
3503 PR middle-end/102269
3504 * gimplify.c (is_var_need_auto_init): Empty types do not need
3507 2021-09-10 Richard Biener <rguenther@suse.de>
3509 * configure.ac (--with-stabs): Remove.
3510 * configure: Regenerate.
3511 * doc/install.texi: Remove --with-stabs documentation.
3513 2021-09-10 liuhongt <hongtao.liu@intel.com>
3515 * config/i386/avx512fp16intrin.h: (_mm512_cmp_ph_mask):
3517 (_mm512_mask_cmp_ph_mask): Likewise.
3518 (_mm512_cmp_round_ph_mask): Likewise.
3519 (_mm512_mask_cmp_round_ph_mask): Likewise.
3520 (_mm_cmp_sh_mask): Likewise.
3521 (_mm_mask_cmp_sh_mask): Likewise.
3522 (_mm_cmp_round_sh_mask): Likewise.
3523 (_mm_mask_cmp_round_sh_mask): Likewise.
3524 (_mm_comieq_sh): Likewise.
3525 (_mm_comilt_sh): Likewise.
3526 (_mm_comile_sh): Likewise.
3527 (_mm_comigt_sh): Likewise.
3528 (_mm_comige_sh): Likewise.
3529 (_mm_comineq_sh): Likewise.
3530 (_mm_ucomieq_sh): Likewise.
3531 (_mm_ucomilt_sh): Likewise.
3532 (_mm_ucomile_sh): Likewise.
3533 (_mm_ucomigt_sh): Likewise.
3534 (_mm_ucomige_sh): Likewise.
3535 (_mm_ucomineq_sh): Likewise.
3536 (_mm_comi_round_sh): Likewise.
3537 (_mm_comi_sh): Likewise.
3538 * config/i386/avx512fp16vlintrin.h (_mm_cmp_ph_mask): New intrinsic.
3539 (_mm_mask_cmp_ph_mask): Likewise.
3540 (_mm256_cmp_ph_mask): Likewise.
3541 (_mm256_mask_cmp_ph_mask): Likewise.
3542 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3543 * config/i386/i386-builtin.def: Add corresponding new builtins.
3544 * config/i386/i386-expand.c
3545 (ix86_expand_args_builtin): Handle new builtin types.
3546 (ix86_expand_round_builtin): Ditto.
3547 * config/i386/i386.md (ssevecmode): Add HF mode.
3548 (MODEFH): New mode iterator.
3549 * config/i386/sse.md
3550 (V48H_AVX512VL): New mode iterator to support HF vector modes.
3551 Ajdust corresponding description.
3552 (ssecmpintprefix): New.
3553 (VI12_AVX512VL): Adjust to support HF vector modes.
3554 (cmp_imm_predicate): Likewise.
3555 (<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
3557 (avx512f_vmcmp<mode>3<round_saeonly_name>): Likewise.
3558 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Likewise.
3559 (<sse>_<unord>comi<round_saeonly_name>): Likewise.
3561 2021-09-10 liuhongt <hongtao.liu@intel.com>
3563 * config/i386/avx512fp16intrin.h: (_mm512_max_ph): New intrinsic.
3564 (_mm512_mask_max_ph): Likewise.
3565 (_mm512_maskz_max_ph): Likewise.
3566 (_mm512_min_ph): Likewise.
3567 (_mm512_mask_min_ph): Likewise.
3568 (_mm512_maskz_min_ph): Likewise.
3569 (_mm512_max_round_ph): Likewise.
3570 (_mm512_mask_max_round_ph): Likewise.
3571 (_mm512_maskz_max_round_ph): Likewise.
3572 (_mm512_min_round_ph): Likewise.
3573 (_mm512_mask_min_round_ph): Likewise.
3574 (_mm512_maskz_min_round_ph): Likewise.
3575 (_mm_max_sh): Likewise.
3576 (_mm_mask_max_sh): Likewise.
3577 (_mm_maskz_max_sh): Likewise.
3578 (_mm_min_sh): Likewise.
3579 (_mm_mask_min_sh): Likewise.
3580 (_mm_maskz_min_sh): Likewise.
3581 (_mm_max_round_sh): Likewise.
3582 (_mm_mask_max_round_sh): Likewise.
3583 (_mm_maskz_max_round_sh): Likewise.
3584 (_mm_min_round_sh): Likewise.
3585 (_mm_mask_min_round_sh): Likewise.
3586 (_mm_maskz_min_round_sh): Likewise.
3587 * config/i386/avx512fp16vlintrin.h (_mm_max_ph): New intrinsic.
3588 (_mm256_max_ph): Likewise.
3589 (_mm_mask_max_ph): Likewise.
3590 (_mm256_mask_max_ph): Likewise.
3591 (_mm_maskz_max_ph): Likewise.
3592 (_mm256_maskz_max_ph): Likewise.
3593 (_mm_min_ph): Likewise.
3594 (_mm256_min_ph): Likewise.
3595 (_mm_mask_min_ph): Likewise.
3596 (_mm256_mask_min_ph): Likewise.
3597 (_mm_maskz_min_ph): Likewise.
3598 (_mm256_maskz_min_ph): Likewise.
3599 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3600 * config/i386/i386-builtin.def: Add corresponding new builtins.
3601 * config/i386/i386-expand.c
3602 (ix86_expand_args_builtin): Handle new builtin types.
3603 * config/i386/sse.md
3604 (<code><mode>3<mask_name><round_saeonly_name>): Adjust to
3605 support HF vector modes.
3606 (*<code><mode>3<mask_name><round_saeonly_name>): Likewise.
3607 (ieee_<ieee_maxmin><mode>3<mask_name><round_saeonly_name>):
3609 (<sse>_vm<code><mode>3<mask_scalar_name><round_saeonly_scalar_name>):
3611 * config/i386/subst.md (round_saeonly_mode512bit_condition):
3612 Adjust for HF vector modes.
3614 2021-09-10 Liu, Hongtao <hongtao.liu@intel.com>
3616 * config/i386/avx512fp16intrin.h (_mm_add_sh): New intrinsic.
3617 (_mm_mask_add_sh): Likewise.
3618 (_mm_maskz_add_sh): Likewise.
3619 (_mm_sub_sh): Likewise.
3620 (_mm_mask_sub_sh): Likewise.
3621 (_mm_maskz_sub_sh): Likewise.
3622 (_mm_mul_sh): Likewise.
3623 (_mm_mask_mul_sh): Likewise.
3624 (_mm_maskz_mul_sh): Likewise.
3625 (_mm_div_sh): Likewise.
3626 (_mm_mask_div_sh): Likewise.
3627 (_mm_maskz_div_sh): Likewise.
3628 (_mm_add_round_sh): Likewise.
3629 (_mm_mask_add_round_sh): Likewise.
3630 (_mm_maskz_add_round_sh): Likewise.
3631 (_mm_sub_round_sh): Likewise.
3632 (_mm_mask_sub_round_sh): Likewise.
3633 (_mm_maskz_sub_round_sh): Likewise.
3634 (_mm_mul_round_sh): Likewise.
3635 (_mm_mask_mul_round_sh): Likewise.
3636 (_mm_maskz_mul_round_sh): Likewise.
3637 (_mm_div_round_sh): Likewise.
3638 (_mm_mask_div_round_sh): Likewise.
3639 (_mm_maskz_div_round_sh): Likewise.
3640 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
3641 * config/i386/i386-builtin.def: Add corresponding new builtins.
3642 * config/i386/i386-expand.c
3643 (ix86_expand_round_builtin): Handle new builtins.
3644 * config/i386/sse.md (VF_128): Change description.
3645 (<sse>_vm<plusminus_insn><mode>3<mask_scalar_name><round_scalar_name>):
3646 Adjust to support HF vector modes.
3647 (<sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name><round_scalar_name>):
3650 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
3652 * config/i386/i386-expand.c
3653 (ix86_avx256_split_vector_move_misalign): Handle V16HF mode.
3654 * config/i386/i386.c
3655 (ix86_preferred_simd_mode): Handle HF mode.
3656 * config/i386/sse.md (V_256H): New mode iterator.
3657 (avx_vextractf128<mode>): Use it.
3658 (VEC_INIT_MODE): Align vector HFmode condition to vector
3659 HImodes since there're no real HF instruction used.
3660 (VEC_INIT_HALF_MODE): Ditto.
3662 (VIHF_AVX512BW): Ditto.
3663 (*vec_extracthf): Ditto.
3664 (VEC_EXTRACT_MODE): Ditto.
3666 2021-09-10 Richard Biener <rguenther@suse.de>
3669 * config/dbx.h: Remove.
3670 * config/dbxcoff.h: Do not define PREFERRED_DEBUGGING_TYPE.
3671 * config/lynx.h: Likewise.
3673 2021-09-10 liuhongt <hongtao.liu@intel.com>
3675 * config/i386/i386-expand.c (ix86_expand_copysign): Expand
3676 right into ANDNOT + AND + IOR, using paradoxical subregs.
3677 (ix86_split_copysign_const): Remove.
3678 (ix86_split_copysign_var): Ditto.
3679 * config/i386/i386-protos.h (ix86_split_copysign_const): Dotto.
3680 (ix86_split_copysign_var): Ditto.
3681 * config/i386/i386.md (@copysign<mode>3_const): Ditto.
3682 (@copysign<mode>3_var): Ditto.
3684 2021-09-09 qing zhao <qing.zhao@oracle.com>
3686 * builtins.c (expand_builtin_memset): Make external visible.
3687 * builtins.h (expand_builtin_memset): Declare extern.
3688 * common.opt (ftrivial-auto-var-init=): New option.
3689 * doc/extend.texi: Document the uninitialized attribute.
3690 * doc/invoke.texi: Document -ftrivial-auto-var-init.
3691 * flag-types.h (enum auto_init_type): New enumerated type
3693 * gimple-fold.c (clear_padding_type): Add one new parameter.
3694 (clear_padding_union): Likewise.
3695 (clear_padding_emit_loop): Likewise.
3696 (clear_type_padding_in_mask): Likewise.
3697 (gimple_fold_builtin_clear_padding): Handle this new parameter.
3698 * gimplify.c (gimple_add_init_for_auto_var): New function.
3699 (gimple_add_padding_init_for_auto_var): New function.
3700 (is_var_need_auto_init): New function.
3701 (gimplify_decl_expr): Add initialization to automatic variables per
3703 (gimplify_call_expr): Add one new parameter for call to
3704 __builtin_clear_padding.
3705 (gimplify_init_constructor): Add padding initialization in the end.
3706 * internal-fn.c (INIT_PATTERN_VALUE): New macro.
3707 (expand_DEFERRED_INIT): New function.
3708 * internal-fn.def (DEFERRED_INIT): New internal function.
3709 * tree-cfg.c (verify_gimple_call): Verify calls to .DEFERRED_INIT.
3710 * tree-sra.c (generate_subtree_deferred_init): New function.
3711 (scan_function): Avoid setting cannot_scalarize_away_bitmap for
3712 calls to .DEFERRED_INIT.
3713 (sra_modify_deferred_init): New function.
3714 (sra_modify_function_body): Handle calls to DEFERRED_INIT specially.
3715 * tree-ssa-structalias.c (find_func_aliases_for_call): Likewise.
3716 * tree-ssa-uninit.c (warn_uninit): Handle calls to DEFERRED_INIT
3718 (check_defs): Likewise.
3719 (warn_uninitialized_vars): Likewise.
3720 * tree-ssa.c (ssa_undefined_value_p): Likewise.
3721 * tree.c (build_common_builtin_nodes): Build tree node for
3722 BUILT_IN_CLEAR_PADDING when needed.
3724 2021-09-09 Richard Biener <rguenther@suse.de>
3726 * tree-ssa-loop-im.c (fill_always_executed_in_1): Walk
3729 2021-09-09 Richard Biener <rguenther@suse.de>
3731 * tree-ssa-loop-im.c (fill_always_executed_in_1): Integrate
3732 DOM walk from get_loop_body_in_dom_order using a worklist
3735 2021-09-09 liuhongt <hongtao.liu@intel.com>
3737 * config.gcc: Add avx512fp16vlintrin.h.
3738 * config/i386/avx512fp16intrin.h: (_mm512_add_ph): New intrinsic.
3739 (_mm512_mask_add_ph): Likewise.
3740 (_mm512_maskz_add_ph): Likewise.
3741 (_mm512_sub_ph): Likewise.
3742 (_mm512_mask_sub_ph): Likewise.
3743 (_mm512_maskz_sub_ph): Likewise.
3744 (_mm512_mul_ph): Likewise.
3745 (_mm512_mask_mul_ph): Likewise.
3746 (_mm512_maskz_mul_ph): Likewise.
3747 (_mm512_div_ph): Likewise.
3748 (_mm512_mask_div_ph): Likewise.
3749 (_mm512_maskz_div_ph): Likewise.
3750 (_mm512_add_round_ph): Likewise.
3751 (_mm512_mask_add_round_ph): Likewise.
3752 (_mm512_maskz_add_round_ph): Likewise.
3753 (_mm512_sub_round_ph): Likewise.
3754 (_mm512_mask_sub_round_ph): Likewise.
3755 (_mm512_maskz_sub_round_ph): Likewise.
3756 (_mm512_mul_round_ph): Likewise.
3757 (_mm512_mask_mul_round_ph): Likewise.
3758 (_mm512_maskz_mul_round_ph): Likewise.
3759 (_mm512_div_round_ph): Likewise.
3760 (_mm512_mask_div_round_ph): Likewise.
3761 (_mm512_maskz_div_round_ph): Likewise.
3762 * config/i386/avx512fp16vlintrin.h: New header.
3763 * config/i386/i386-builtin-types.def (V16HF, V8HF, V32HF):
3764 Add new builtin types.
3765 * config/i386/i386-builtin.def: Add corresponding builtins.
3766 * config/i386/i386-expand.c
3767 (ix86_expand_args_builtin): Handle new builtin types.
3768 (ix86_expand_round_builtin): Likewise.
3769 * config/i386/immintrin.h: Include avx512fp16vlintrin.h
3770 * config/i386/sse.md (VFH): New mode_iterator.
3772 (avx512fmaskmode): Add HF vector modes.
3773 (avx512fmaskhalfmode): Likewise.
3774 (<plusminus_insn><mode>3<mask_name><round_name>): Adjust to for
3776 (*<plusminus_insn><mode>3<mask_name><round_name>): Likewise.
3777 (mul<mode>3<mask_name><round_name>): Likewise.
3778 (*mul<mode>3<mask_name><round_name>): Likewise.
3779 (div<mode>3): Likewise.
3780 (<sse>_div<mode>3<mask_name><round_name>): Likewise.
3781 * config/i386/subst.md (SUBST_V): Add HF vector modes.
3782 (SUBST_A): Likewise.
3783 (round_mode512bit_condition): Adjust for V32HFmode.
3785 2021-09-09 liuhongt <hongtao.liu@intel.com>
3788 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
3789 (reduc_plus_scal_v4sf): .. this, New define_expand.
3790 (reduc_plus_scal_v2df): .. and this, New define_expand.
3792 2021-09-09 liuhongt <hongtao.liu@intel.com>
3795 * config/i386/sse.md (*vec_extract<mode><ssescalarmodelower>_valign):
3798 2021-09-08 Jonathan Wakely <jwakely@redhat.com>
3801 * doc/trouble.texi (Copy Assignment): Fix description of
3802 behaviour and fix code in example.
3804 2021-09-08 Segher Boessenkool <segher@kernel.crashing.org>
3807 * config/rs6000/rs6000-logue.c (rs6000_emit_epilogue): For ELFv2 use
3808 r11 instead of r12 for restoring CR.
3810 2021-09-08 Jakub Jelinek <jakub@redhat.com>
3811 liuhongt <hongtao.liu@intel.com>
3814 * config/i386/i386.md (@xorsign<mode>3_1): Remove.
3815 * config/i386/i386-expand.c (ix86_expand_xorsign): Expand right away
3816 into AND with mask and XOR, using paradoxical subregs.
3817 (ix86_split_xorsign): Remove.
3818 * config/i386/i386-protos.h (ix86_split_xorsign): Remove.
3820 2021-09-08 Di Zhao <dizhao@os.amperecomputing.com>
3822 * tree-ssa-sccvn.c (vn_nary_op_insert_into): fix result compare
3824 2021-09-08 Jakub Jelinek <jakub@redhat.com>
3827 * config/i386/i386.md (xorsign<mode>3): If operands[1] is equal to
3828 operands[2], emit abs<mode>2 instead.
3829 (@xorsign<mode>3_1): Add early-clobbers for output operand, enable
3830 first alternative even for avx, add another alternative with
3831 =&Yv <- 0, Yv, Yvm constraints.
3832 * config/i386/i386-expand.c (ix86_split_xorsign): If op0 is equal
3833 to op1, emit vpandn instead.
3835 2021-09-08 liuhongt <hongtao.liu@intel.com>
3837 * config/i386/avx512fp16intrin.h (_mm_set_ph): New intrinsic.
3838 (_mm256_set_ph): Likewise.
3839 (_mm512_set_ph): Likewise.
3840 (_mm_setr_ph): Likewise.
3841 (_mm256_setr_ph): Likewise.
3842 (_mm512_setr_ph): Likewise.
3843 (_mm_set1_ph): Likewise.
3844 (_mm256_set1_ph): Likewise.
3845 (_mm512_set1_ph): Likewise.
3846 (_mm_setzero_ph): Likewise.
3847 (_mm256_setzero_ph): Likewise.
3848 (_mm512_setzero_ph): Likewise.
3849 (_mm_set_sh): Likewise.
3850 (_mm_load_sh): Likewise.
3851 (_mm_store_sh): Likewise.
3852 * config/i386/i386-builtin-types.def (V8HF): New type.
3853 (DEF_FUNCTION_TYPE (V8HF, V8HI)): New builtin function type
3854 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
3855 Support vector HFmodes.
3856 (ix86_expand_vector_init_one_nonzero): Likewise.
3857 (ix86_expand_vector_init_one_var): Likewise.
3858 (ix86_expand_vector_init_interleave): Likewise.
3859 (ix86_expand_vector_init_general): Likewise.
3860 (ix86_expand_vector_set): Likewise.
3861 (ix86_expand_vector_extract): Likewise.
3862 (ix86_expand_vector_init_concat): Likewise.
3863 (ix86_expand_sse_movcc): Handle vector HFmodes.
3864 (ix86_expand_vector_set_var): Ditto.
3865 * config/i386/i386-modes.def: Add HF vector modes in comment.
3866 * config/i386/i386.c (classify_argument): Add HF vector modes.
3867 (ix86_hard_regno_mode_ok): Allow HF vector modes for AVX512FP16.
3868 (ix86_vector_mode_supported_p): Likewise.
3869 (ix86_set_reg_reg_cost): Handle vector HFmode.
3870 (ix86_get_ssemov): Handle vector HFmode.
3871 (function_arg_advance_64): Pass unamed V16HFmode and V32HFmode
3873 (function_arg_advance_32): Pass V8HF/V16HF/V32HF by sse reg for 32bit
3875 (function_arg_advance_32): Ditto.
3876 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): New.
3877 (VALID_AVX256_REG_OR_OI_MODE): Rename to ..
3878 (VALID_AVX256_REG_OR_OI_VHF_MODE): .. this, and add V16HF.
3879 (VALID_SSE2_REG_VHF_MODE): New.
3880 (VALID_AVX512VL_128_REG_MODE): Add V8HF and TImode.
3881 (SSE_REG_MODE_P): Add vector HFmode.
3882 * config/i386/i386.md (mode): Add HF vector modes.
3883 (MODE_SIZE): Likewise.
3884 (ssemodesuffix): Add ph suffix for HF vector modes.
3885 * config/i386/sse.md (VFH_128): New mode iterator.
3886 (VMOVE): Adjust for HF vector modes.
3888 (V_256_512): Likewise.
3890 (avx512fmaskmode): Likewise.
3891 (shuffletype): Likewise.
3892 (sseinsnmode): Likewise.
3893 (ssedoublevecmode): Likewise.
3894 (ssehalfvecmode): Likewise.
3895 (ssehalfvecmodelower): Likewise.
3896 (ssePScmode): Likewise.
3897 (ssescalarmode): Likewise.
3898 (ssescalarmodelower): Likewise.
3899 (sseintprefix): Likewise.
3901 (bcstscalarsuff): Likewise.
3902 (xtg_mode): Likewise.
3903 (VI12HF_AVX512VL): New mode_iterator.
3904 (VF_AVX512FP16): Likewise.
3906 (VIHF_256): Likewise.
3907 (VIHF_AVX512BW): Likewise.
3908 (V16_256): Likewise.
3909 (V32_512): Likewise.
3910 (sseintmodesuffix): New mode_attr.
3911 (sse): Add scalar and vector HFmodes.
3912 (ssescalarmode): Add vector HFmode mapping.
3913 (ssescalarmodesuffix): Add sh suffix for HFmode.
3914 (*<sse>_vm<insn><mode>3): Use VFH_128.
3915 (*<sse>_vm<multdiv_mnemonic><mode>3): Likewise.
3916 (*ieee_<ieee_maxmin><mode>3): Likewise.
3917 (<avx512>_blendm<mode>): New define_insn.
3918 (vec_setv8hf): New define_expand.
3919 (vec_set<mode>_0): New define_insn for HF vector set.
3920 (*avx512fp16_movsh): Likewise.
3921 (avx512fp16_movsh): Likewise.
3922 (vec_extract_lo_v32hi): Rename to ...
3923 (vec_extract_lo_<mode>): ... this, and adjust to allow HF
3925 (vec_extract_hi_v32hi): Likewise.
3926 (vec_extract_hi_<mode>): Likewise.
3927 (vec_extract_lo_v16hi): Likewise.
3928 (vec_extract_lo_<mode>): Likewise.
3929 (vec_extract_hi_v16hi): Likewise.
3930 (vec_extract_hi_<mode>): Likewise.
3931 (vec_set_hi_v16hi): Likewise.
3932 (vec_set_hi_<mode>): Likewise.
3933 (vec_set_lo_v16hi): Likewise.
3934 (vec_set_lo_<mode>): Likewise.
3935 (*vec_extract<mode>_0): New define_insn_and_split for HF
3937 (*vec_extracthf): New define_insn.
3938 (VEC_EXTRACT_MODE): Add HF vector modes.
3939 (PINSR_MODE): Add V8HF.
3940 (sse2p4_1): Likewise.
3941 (pinsr_evex_isa): Likewise.
3942 (<sse2p4_1>_pinsr<ssemodesuffix>): Adjust to support
3943 insert for V8HFmode.
3944 (pbroadcast_evex_isa): Add HF vector modes.
3945 (AVX2_VEC_DUP_MODE): Likewise.
3946 (VEC_INIT_MODE): Likewise.
3947 (VEC_INIT_HALF_MODE): Likewise.
3948 (avx2_pbroadcast<mode>): Adjust to support HF vector mode
3950 (avx2_pbroadcast<mode>_1): Likewise.
3951 (<avx512>_vec_dup<mode>_1): Likewise.
3952 (<avx512>_vec_dup<mode><mask_name>): Likewise.
3953 (<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>):
3956 2021-09-08 Guo, Xuepeng <xuepeng.guo@intel.com>
3957 H.J. Lu <hongjiu.lu@intel.com>
3958 Liu Hongtao <hongtao.liu@intel.com>
3959 Wang Hongyu <hongyu.wang@intel.com>
3960 Xu Dianhong <dianhong.xu@intel.com>
3962 * common/config/i386/cpuinfo.h (get_available_features):
3963 Detect FEATURE_AVX512FP16.
3964 * common/config/i386/i386-common.c
3965 (OPTION_MASK_ISA_AVX512FP16_SET,
3966 OPTION_MASK_ISA_AVX512FP16_UNSET,
3967 OPTION_MASK_ISA2_AVX512FP16_SET,
3968 OPTION_MASK_ISA2_AVX512FP16_UNSET): New.
3969 (OPTION_MASK_ISA2_AVX512BW_UNSET,
3970 OPTION_MASK_ISA2_AVX512BF16_UNSET): Add AVX512FP16.
3971 (ix86_handle_option): Handle -mavx512fp16.
3972 * common/config/i386/i386-cpuinfo.h (enum processor_features):
3973 Add FEATURE_AVX512FP16.
3974 * common/config/i386/i386-isas.h: Add entry for AVX512FP16.
3975 * config.gcc: Add avx512fp16intrin.h.
3976 * config/i386/avx512fp16intrin.h: New intrinsic header.
3977 * config/i386/cpuid.h: Add bit_AVX512FP16.
3978 * config/i386/i386-builtin-types.def: (FLOAT16): New primitive type.
3979 * config/i386/i386-builtins.c: Support _Float16 type for i386
3981 (ix86_register_float16_builtin_type): New function.
3982 (ix86_float16_type_node): New.
3983 * config/i386/i386-c.c (ix86_target_macros_internal): Define
3985 * config/i386/i386-expand.c (ix86_expand_branch): Support
3987 (ix86_prepare_fp_compare_args): Adjust TARGET_SSE_MATH &&
3988 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
3989 (ix86_expand_fp_movcc): Ditto.
3990 * config/i386/i386-isa.def: Add PTA define for AVX512FP16.
3991 * config/i386/i386-options.c (isa2_opts): Add -mavx512fp16.
3992 (ix86_valid_target_attribute_inner_p): Add avx512fp16 attribute.
3993 * config/i386/i386.c (ix86_get_ssemov): Use
3994 vmovdqu16/vmovw/vmovsh for HFmode/HImode scalar or vector.
3995 (ix86_get_excess_precision): Use
3996 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when TARGET_AVX512FP16
3998 (sse_store_index): Use SFmode cost for HFmode cost.
3999 (inline_memory_move_cost): Add HFmode, and perfer SSE cost over
4000 GPR cost for HFmode.
4001 (ix86_hard_regno_mode_ok): Allow HImode in sse register.
4002 (ix86_mangle_type): Add manlging for _Float16 type.
4003 (inline_secondary_memory_needed): No memory is needed for
4004 16bit movement between gpr and sse reg under
4006 (ix86_multiplication_cost): Adjust TARGET_SSE_MATH &&
4007 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
4008 (ix86_division_cost): Ditto.
4009 (ix86_rtx_costs): Ditto.
4010 (ix86_add_stmt_cost): Ditto.
4011 (ix86_optab_supported_p): Ditto.
4012 * config/i386/i386.h (VALID_AVX512F_SCALAR_MODE): Add HFmode.
4013 (SSE_FLOAT_MODE_SSEMATH_OR_HF_P): Add HFmode.
4014 (PTA_SAPPHIRERAPIDS): Add PTA_AVX512FP16.
4015 * config/i386/i386.md (mode): Add HFmode.
4016 (MODE_SIZE): Add HFmode.
4017 (isa): Add avx512fp16.
4018 (enabled): Handle avx512fp16.
4019 (ssemodesuffix): Add sh suffix for HFmode.
4020 (comm): Add mult, div.
4021 (plusminusmultdiv): New code iterator.
4022 (insn): Add mult, div.
4023 (*movhf_internal): Adjust for avx512fp16 instruction.
4024 (*movhi_internal): Ditto.
4025 (*cmpi<unord>hf): New define_insn for HFmode.
4026 (*ieee_s<ieee_maxmin>hf3): Likewise.
4027 (extendhf<mode>2): Likewise.
4028 (trunc<mode>hf2): Likewise.
4029 (float<floatunssuffix><mode>hf2): Likewise.
4030 (*<insn>hf): Likewise.
4031 (cbranchhf4): New expander.
4032 (movhfcc): Likewise.
4033 (<insn>hf3): Likewise.
4036 * config/i386/i386.opt: Add mavx512fp16.
4037 * config/i386/immintrin.h: Include avx512fp16intrin.h.
4038 * doc/invoke.texi: Add mavx512fp16.
4039 * doc/extend.texi: Add avx512fp16 Usage Notes.
4041 2021-09-08 liuhongt <hongtao.liu@intel.com>
4043 * common.opt: Support -fexcess-precision=16.
4044 * config/aarch64/aarch64.c (aarch64_excess_precision): Return
4045 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when
4046 EXCESS_PRECISION_TYPE_FLOAT16.
4047 * config/arm/arm.c (arm_excess_precision): Ditto.
4048 * config/i386/i386.c (ix86_get_excess_precision): Ditto.
4049 * config/m68k/m68k.c (m68k_excess_precision): Issue an error
4050 when EXCESS_PRECISION_TYPE_FLOAT16.
4051 * config/s390/s390.c (s390_excess_precision): Ditto.
4052 * coretypes.h (enum excess_precision_type): Add
4053 EXCESS_PRECISION_TYPE_FLOAT16.
4054 * doc/tm.texi (TARGET_C_EXCESS_PRECISION): Update documents.
4055 * doc/tm.texi.in (TARGET_C_EXCESS_PRECISION): Ditto.
4056 * doc/extend.texi (Half-Precision): Document
4057 -fexcess-precision=16.
4058 * flag-types.h (enum excess_precision): Add
4059 EXCESS_PRECISION_FLOAT16.
4060 * target.def (excess_precision): Update document.
4061 * tree.c (excess_precision_type): Set excess_precision_type to
4062 EXCESS_PRECISION_FLOAT16 when -fexcess-precision=16.
4064 2021-09-08 liuhongt <hongtao.liu@intel.com>
4066 * doc/extend.texi: (@node Floating Types): Adjust the wording.
4067 (@node Half-Precision): Ditto.
4069 2021-09-07 Takayuki 'January June' Suwa <jjsuwa_sys3175@yahoo.co.jp>
4072 * config/xtensa/xtensa.c (xtensa_emit_move_sequence): Add
4073 'CONST_INT_P (src)' to the condition of the block that tries to
4074 eliminate literal when loading integer contant.
4076 2021-09-07 David Faust <david.faust@oracle.com>
4078 * doc/extend.texi (BPF Type Attributes) New node.
4079 Document new preserve_access_index attribute.
4080 Document new preserve_access_index builtin.
4081 * doc/invoke.texi: Document -mco-re and -mno-co-re options.
4083 2021-09-07 David Faust <david.faust@oracle.com>
4085 * config/bpf/bpf.c: Adjust includes.
4086 (bpf_handle_preserve_access_index_attribute): New function.
4087 (bpf_attribute_table): Use it here.
4088 (bpf_builtins): Add BPF_BUILTIN_PRESERVE_ACCESS_INDEX.
4089 (bpf_option_override): Handle "-mco-re" option.
4090 (bpf_asm_init_sections): New.
4091 (TARGET_ASM_INIT_SECTIONS): Redefine.
4092 (bpf_file_end): New.
4093 (TARGET_ASM_FILE_END): Redefine.
4094 (bpf_init_builtins): Add "__builtin_preserve_access_index".
4095 (bpf_core_compute, bpf_core_get_index): New.
4096 (is_attr_preserve_access): New.
4097 (bpf_expand_builtin): Handle new builtins.
4098 (bpf_core_newdecl, bpf_core_is_maybe_aggregate_access): New.
4099 (bpf_core_walk): New.
4100 (bpf_resolve_overloaded_builtin): New.
4101 (TARGET_RESOLVE_OVERLOADED_BUILTIN): Redefine.
4103 (pass_bpf_core_attr): New RTL pass.
4104 * config/bpf/bpf-passes.def: New file.
4105 * config/bpf/bpf-protos.h (make_pass_bpf_core_attr): New.
4106 * config/bpf/coreout.c: New file.
4107 * config/bpf/coreout.h: Likewise.
4108 * config/bpf/t-bpf (TM_H): Add $(srcdir)/config/bpf/coreout.h.
4109 (coreout.o): New rule.
4110 (PASSES_EXTRA): Add $(srcdir)/config/bpf/bpf-passes.def.
4111 * config.gcc (bpf): Add coreout.h to extra_headers.
4112 Add coreout.o to extra_objs.
4113 Add $(srcdir)/config/bpf/coreout.c to target_gtfiles.
4115 2021-09-07 David Faust <david.faust@oracle.com>
4117 * btfout.c (get_btf_id): Function is no longer static.
4118 * ctfc.h: Expose it here.
4120 2021-09-07 David Faust <david.faust@oracle.com>
4122 * ctfc.c (ctf_lookup_tree_type): New function.
4125 2021-09-07 David Faust <david.faust@oracle.com>
4127 * ctfc.c (ctf_dtd_lookup): Function is no longer static.
4128 * ctfc.h: Analogous change.
4130 2021-09-07 David Faust <david.faust@oracle.com>
4132 * dwarf2out.c (lookup_type_die): Function is no longer static.
4133 * dwarf2out.h: Expose it here.
4135 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
4137 * dwarf2ctf.c (ctf_debug_finalize): Make it static.
4138 (ctf_debug_early_finish): New definition.
4139 (ctf_debug_finish): Likewise.
4140 * dwarf2ctf.h (ctf_debug_finalize): Remove declaration.
4141 (ctf_debug_early_finish): New declaration.
4142 (ctf_debug_finish): Likewise.
4143 * dwarf2out.c (dwarf2out_finish): Invoke ctf_debug_finish.
4144 (dwarf2out_early_finish): Invoke ctf_debug_early_finish.
4146 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
4148 * config/bpf/bpf.c (bpf_option_override): For BPF backend, disable LTO
4149 support when compiling for CO-RE.
4150 * config/bpf/bpf.opt: Add new command line option -mco-re.
4152 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
4154 * flag-types.h (enum debug_info_type): Add new enum
4155 DINFO_TYPE_BTF_WITH_CORE.
4156 (BTF_WITH_CORE_DEBUG): New bitmask.
4157 * flags.h (btf_with_core_debuginfo_p): New declaration.
4158 * opts.c (btf_with_core_debuginfo_p): New definition.
4160 2021-09-07 Jason Merrill <jason@redhat.com>
4162 * tree.h (error_operand_p): Change to inline function.
4164 2021-09-07 Aldy Hernandez <aldyh@redhat.com>
4166 * tree-ssa-threadedge.c (forwarder_block_p): Rename to...
4167 (empty_block_with_phis_p): ...this.
4168 (potentially_threadable_block): Same.
4169 (jump_threader::thread_through_normal_block): Same.
4171 2021-09-07 Eric Botcazou <ebotcazou@adacore.com>
4174 * dwarf2out.c (mark_base_types): New overloaded function.
4175 (dwarf2out_early_finish): Invoke it on the COMDAT type list as well
4176 as the compilation unit, and call move_marked_base_types afterward.
4178 2021-09-07 H.J. Lu <hjl.tools@gmail.com>
4181 * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
4183 (ix86_expand_vector_convert_uns_vsivsf): Likewise.
4185 2021-09-07 Richard Biener <rguenther@suse.de>
4187 PR tree-optimization/102226
4188 * tree-vect-loop.c (vect_transform_cycle_phi): Record
4189 the converted value for the epilogue PHI use.
4191 2021-09-07 Martin Liska <mliska@suse.cz>
4193 PR gcov-profile/80223
4194 * ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
4195 options, do not inline when no_profile_instrument_function
4196 attributes are different in early inliner. It's fine to inline
4197 it after PGO instrumentation.
4199 2021-09-07 Richard Biener <rguenther@suse.de>
4201 PR tree-optimization/101555
4202 * tree-ssa-pre.c (translate_vuse_through_block): Do not
4203 perform an alias walk to determine the validity of the
4204 mem at the start of the block which is already guaranteed
4205 by means of prune_clobbered_mems.
4206 (phi_translate_1): Pass edge to translate_vuse_through_block.
4208 2021-09-07 Xionghu Luo <luoxhu@linux.ibm.com>
4211 * config/rs6000/rs6000.md (fmod<mode>3): New define_expand.
4212 (remainder<mode>3): Likewise.
4214 2021-09-07 YunQiang Su <yunqiang.su@cipunited.com>
4216 * config/mips/mips.c (mips_file_start): add .module for
4219 2021-09-06 Roger Sayle <roger@nextmovesoftware.com>
4221 * wide-int.cc (wi::clz): Reorder tests to ensure the result
4222 is zero for all negative values.
4224 2021-09-06 Tobias Burnus <tobias@codesourcery.com>
4226 * doc/invoke.texi (-foffload-options): Fix @opindex.
4228 2021-09-06 H.J. Lu <hjl.tools@gmail.com>
4231 * config/i386/i386-expand.c (ix86_split_xorsign): Use operands[2].
4232 * config/i386/i386.md (@xorsign<mode>3_1): Add non-destructive
4233 source alternative for AVX.
4235 2021-09-06 liuhongt <hongtao.liu@intel.com>
4237 PR middle-end/102182
4238 * optabs.c (expand_fix): Add from1 to avoid from being
4241 2021-09-06 Eric Botcazou <ebotcazou@adacore.com>
4243 * dwarf2out.c (modified_type_die): Deal with all array types earlier
4244 and use local variable consistently throughout the function.
4246 2021-09-06 Jakub Jelinek <jakub@redhat.com>
4248 PR tree-optimization/102207
4249 * match.pd: Don't demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
4250 were promoted from signed to wider unsigned type.
4252 2021-09-06 Andrew Pinski <apinski@marvell.com>
4254 PR tree-optimization/63184
4255 * match.pd: Add simplification of pointer_diff of two pointer_plus
4256 with addr_expr in the first operand of each pointer_plus.
4257 Add simplificatoin of ne/eq of two pointer_plus with addr_expr
4258 in the first operand of each pointer_plus.
4260 2021-09-06 Richard Biener <rguenther@suse.de>
4262 PR tree-optimization/102176
4263 * tree-vect-slp.c (vect_slp_gather_vectorized_scalar_stmts):
4265 (vect_bb_slp_scalar_cost): Use the computed set of
4266 vectorized scalar stmts instead of relying on the out-of-date
4267 and not accurate PURE_SLP_STMT.
4268 (vect_bb_vectorization_profitable_p): Compute the set
4269 of vectorized scalar stmts.
4271 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
4273 * gimple-range-path.cc (path_range_query::range_of_stmt): Remove
4274 GIMPLE_COND special casing.
4275 (path_range_query::range_defined_in_block): Use range_of_stmt
4276 instead of calling fold_range directly.
4278 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
4280 * gimple-range-path.cc (path_range_query::range_of_expr): Set
4281 m_undefined_path when appropriate.
4282 (path_range_query::internal_range_of_expr): Copy from range_of_expr.
4283 (path_range_query::unreachable_path_p): New.
4284 (path_range_query::precompute_ranges): Set m_undefined_path.
4285 * gimple-range-path.h (path_range_query::unreachable_path_p): New.
4286 (path_range_query::internal_range_of_expr): New.
4287 * tree-ssa-threadbackward.c (back_threader::find_taken_edge_cond):
4288 Use unreachable_path_p.
4290 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
4292 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
4293 Remove argument and call find_taken_edge.
4294 (back_threader::resolve_phi): Do not calculate taken edge before
4295 calling maybe_register_path.
4296 (back_threader::find_paths_to_names): Same.
4298 2021-09-05 Jeff Law <jlaw@localhost.localdomain>
4300 * config/h8300/h8300.md (QHSI2 mode iterator): New mode iterator.
4301 * config/h8300/testcompare.md (store_c): Update name, use new
4303 (store_neg_c, store_shifted_c): New patterns.
4305 2021-09-03 Segher Boessenkool <segher@kernel.crashing.org>
4308 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): On ELFv2 use r11
4309 instead of r12 for CR save, in all cases.
4311 2021-09-03 Andrew Pinski <apinski@marvell.com>
4313 * config/aarch64/aarch64-sve-builtins.cc (register_vector_type):
4314 Handle error_mark_node as the type of the type_decl.
4316 2021-09-03 Andrew Pinski <apinski@marvell.com>
4318 * config/aarch64/aarch64-builtins.c (struct aarch64_simd_type_info):
4320 (aarch64_simd_types): Likewise.
4321 (aarch64_simd_intOI_type_node): Likewise.
4322 (aarch64_simd_intCI_type_node): Likewise.
4323 (aarch64_simd_intXI_type_node): Likewise.
4324 * config/aarch64/aarch64.h (aarch64_fp16_type_node): Likewise.
4325 (aarch64_fp16_ptr_type_node): Likewise.
4326 (aarch64_bf16_type_node): Likewise.
4327 (aarch64_bf16_ptr_type_node): Likewise.
4329 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4331 * range-op.cc (operator_minus::op1_op2_relation_effect): Abstract
4333 (minus_op1_op2_relation_effect): ...here.
4334 (class operator_pointer_diff): New.
4335 (operator_pointer_diff::op1_op2_relation_effect): Call
4336 minus_op1_op2_relation_effect.
4337 (integral_table::integral_table): Add entry for POINTER_DIFF_EXPR.
4339 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4341 * tree-ssa-threadbackward.c (back_threader::thread_through_all_blocks):
4342 Add may_peel_loop_headers.
4343 (back_threader_registry::thread_through_all_blocks): Same.
4344 (try_thread_blocks): Pass may_peel_loop_headers argument.
4345 (pass_early_thread_jumps::execute): Same.
4347 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4349 * tree-ssa-threadedge.c (has_phis_p): New.
4350 (forwarder_block_p): New.
4351 (potentially_threadable_block): Call forwarder_block_p.
4352 (jump_threader::thread_around_empty_blocks): Call has_phis_p.
4353 (jump_threader::thread_through_normal_block): Call
4356 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4358 * tree-ssa-threadbackward.c (back_threader::dump): New.
4359 (back_threader::debug): New.
4360 (back_threader_profitability::profitable_path_p): Dump blocks
4361 even if we are bailing early.
4363 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4365 * tree-ssa-threadupdate.c (cancel_thread): New.
4366 (jump_thread_path_registry::thread_block_1): Use cancel_thread.
4367 (jump_thread_path_registry::mark_threaded_blocks): Same.
4368 (jump_thread_path_registry::register_jump_thread): Same.
4370 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4372 * tree-ssa-threadedge.c (jt_state::push): Only call methods for
4373 which objects are available.
4374 (jt_state::pop): Same.
4375 (jt_state::register_equiv): Same.
4376 (jt_state::register_equivs_on_edge): Same.
4378 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4380 * tree-ssa-threadedge.c (jump_threader::thread_across_edge):
4381 Move pop until after a thread is registered.
4383 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4385 * tree-ssa-threadupdate.c (debug): New.
4387 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4389 * gimple-range-trace.cc (push_dump_file::push_dump_file): New.
4390 (push_dump_file::~push_dump_file): New.
4391 (dump_ranger): Change dump_file temporarily while dumping
4393 * gimple-range-trace.h (class push_dump_file): New.
4395 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4397 * gimple-range-trace.cc (debug_seed_ranger): Remove static.
4398 (dump_ranger): Dump function name.
4400 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4402 * gimple-range-path.cc (path_range_query::range_defined_in_block):
4403 Adjust for non-null.
4404 (path_range_query::adjust_for_non_null_uses): New.
4405 (path_range_query::precompute_ranges): Call
4406 adjust_for_non_null_uses.
4407 * gimple-range-path.h: Add m_non_null and
4408 adjust_for_non_null_uses.
4410 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4412 * gimple-range-path.cc (path_range_query::dump): Dump path
4414 (path_range_query::precompute_ranges): Dump entire path.
4416 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4418 * value-relation.cc (relation_oracle::debug): New.
4419 * value-relation.h (relation_oracle::debug): New.
4421 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4423 * tree-ssa-loop-ch.c: Remove unnecessary include file.
4425 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4427 * gimple-range-fold.cc (fold_using_range::postfold_gcond_edges):
4428 Skip statements with no defining BB.
4429 * gimple-range-path.cc (path_range_query::range_defined_in_block):
4430 Do not get confused by statements with no defining BB.
4432 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
4434 * gimple-range-fold.cc (adjust_imagpart_expr): Move from
4435 gimple_range_adjustment. Add support for constants.
4436 (adjust_realpart_expr): New.
4437 (gimple_range_adjustment): Move IMAGPART_EXPR code to
4438 adjust_imagpart_expr.
4439 * range-op.cc (integral_table::integral_table): Add entry for
4442 2021-09-03 Jakub Jelinek <jakub@redhat.com>
4444 * omp-expand.c (expand_omp_atomic_pipeline): Use
4445 IFN_ATOMIC_COMPARE_EXCHANGE instead of
4446 BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_? so that memory order
4449 2021-09-03 Jakub Jelinek <jakub@redhat.com>
4452 * tree.h (DECL_FIELD_ABI_IGNORED): Changed into rvalue only macro
4453 that is false if DECL_BIT_FIELD.
4454 (SET_DECL_FIELD_ABI_IGNORED, DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD,
4455 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD): Define.
4456 * tree-streamer-out.c (pack_ts_decl_common_value_fields): For
4457 DECL_BIT_FIELD stream DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead
4458 of DECL_FIELD_ABI_IGNORED.
4459 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
4460 SET_DECL_FIELD_ABI_IGNORED instead of writing to
4461 DECL_FIELD_ABI_IGNORED and for DECL_BIT_FIELD use
4462 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead.
4463 * lto-streamer-out.c (hash_tree): For DECL_BIT_FIELD hash
4464 DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead of DECL_FIELD_ABI_IGNORED.
4466 2021-09-03 liuhongt <hongtao.liu@intel.com>
4469 * config/i386/amxbf16intrin.h : Remove macro check for __AMX_BF16__.
4470 * config/i386/amxint8intrin.h : Remove macro check for __AMX_INT8__.
4471 * config/i386/amxtileintrin.h : Remove macro check for __AMX_TILE__.
4473 2021-09-02 Martin Sebor <msebor@redhat.com>
4475 PR tree-optimization/17506
4477 * tree-ssa-uninit.c (warn_uninit): Remove conditional guarding note.
4479 2021-09-02 Richard Biener <rguenther@suse.de>
4481 * tree-ssa-loop-im.c (fill_always_executed_in_1): Refine
4482 fix for PR78185 and continue processing when leaving
4485 2021-09-02 Jakub Jelinek <jakub@redhat.com>
4487 PR tree-optimization/99591
4488 * match.pd: Demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
4491 2021-09-02 Richard Biener <rguenther@suse.de>
4494 2021-09-02 Richard Biener <rguenther@suse.de>
4496 PR tree-optimization/102155
4497 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
4498 over a part of the RPO array and do not recurse here.
4499 Dump blocks marked as always executed.
4500 (fill_always_executed_in): Walk over the RPO array and
4501 process loops whose header we run into.
4502 (loop_invariant_motion_in_fun): Compute the first RPO
4503 using rev_post_order_and_mark_dfs_back_seme in iteration
4504 order and pass that to fill_always_executed_in.
4506 2021-09-02 liuhongt <hongtao.liu@intel.com>
4508 * config/i386/i386-modes.def (FLOAT_MODE): Define ieee HFmode.
4509 * config/i386/i386.c (enum x86_64_reg_class): Add
4511 (merge_classes): Handle X86_64_SSEHF_CLASS.
4512 (examine_argument): Ditto.
4513 (construct_container): Ditto.
4514 (classify_argument): Ditto, and set HFmode/HCmode to
4516 (function_value_32): Return _FLoat16/Complex Float16 by
4518 (function_value_64): Return _Float16/Complex Float16 by SSE
4520 (ix86_print_operand): Handle CONST_DOUBLE HFmode.
4521 (ix86_secondary_reload): Require gpr as intermediate register
4522 to store _Float16 from sse register when sse4 is not
4524 (ix86_libgcc_floating_mode_supported_p): Enable _FLoat16 under
4526 (ix86_scalar_mode_supported_p): Ditto.
4527 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Defined.
4528 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add HFmode.
4529 (VALID_INT_MODE_P): Add HFmode and HCmode.
4530 * config/i386/i386.md (*pushhf_rex64): New define_insn.
4532 (*movhf_internal): Ditto.
4533 * doc/extend.texi (Half-Precision Floating Point): Documemt
4536 2021-09-02 Richard Biener <rguenther@suse.de>
4538 PR tree-optimization/102155
4539 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
4540 over a part of the RPO array and do not recurse here.
4541 Dump blocks marked as always executed.
4542 (fill_always_executed_in): Walk over the RPO array and
4543 process loops whose header we run into.
4544 (loop_invariant_motion_in_fun): Compute the first RPO
4545 using rev_post_order_and_mark_dfs_back_seme in iteration
4546 order and pass that to fill_always_executed_in.
4548 2021-09-02 YunQiang Su <syq@debian.org>
4551 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
4553 * config/mips/mips.c (mips_module_isa_name): New.
4554 mips_file_start: add .module mipsREV to all asm output
4556 2021-09-01 Jeff Law <jlaw@localhost.localdomain>
4558 PR tree-optimization/102152
4559 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Reduce a vector
4560 comparison to a scalar comparison before calling
4561 update_stmt_if_modified.
4563 2021-09-01 Andrew Pinski <apinski@marvell.com>
4566 * config/aarch64/aarch64.c (aarch64_expand_setmem):
4567 Check STRICT_ALIGNMENT before creating an overlapping
4570 2021-09-01 Martin Sebor <msebor@redhat.com>
4572 * gimple-ssa-warn-access.cc (get_size_range): Add argument.
4573 (check_access): Pass additional argument.
4574 (check_memop_access): Remove template and make a member function.
4575 (maybe_check_dealloc_call): Make a pass_waccess member function.
4576 (class pass_waccess): Add, rename, and remove members.
4577 (pass_waccess::pass_waccess): Adjust to name change.
4578 (pass_waccess::~pass_waccess): Same.
4579 (check_alloca): Make a member function.
4580 (check_alloc_size_call): Same.
4581 (check_strcat): Same.
4582 (check_strncat): Same.
4583 (check_stxcpy): Same.
4584 (check_stxncpy): Same.
4585 (check_strncmp): Same.
4586 (maybe_warn_rdwr_sizes): Rename...
4587 (pass_waccess::maybe_check_access_sizes): ...to this.
4588 (pass_waccess::check_call): Adjust to name changes.
4589 (pass_waccess::maybe_check_dealloc_call): Make a pass_waccess member
4591 (pass_waccess::execute): Adjust to name changes.
4592 * gimple-ssa-warn-access.h (check_memop_access): Remove.
4593 * pointer-query.cc (access_ref::phi): Handle null pointer.
4594 (access_ref::inform_access): Same.
4595 (pointer_query::put_ref): Modify a cached value, not a copy of it.
4596 (pointer_query::dump): New function.
4597 (compute_objsize_r): Avoid overwriting access_ref::bndrng. Cache
4599 * pointer-query.h (pointer_query::dump): Declare.
4600 * tree-ssa-strlen.c (get_range): Simplify. Use function query.
4601 (dump_strlen_info): Use function query.
4602 (printf_strlen_execute): Factor code out into pointer_query::put_ref.
4604 2021-09-01 Thomas Schwinge <thomas@codesourcery.com>
4606 * tree.c (walk_tree_1) <OMP_CLAUSE>: Simplify.
4608 2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
4610 * doc/extend.texi: Document unavailable attribute.
4611 * print-tree.c (print_node): Handle unavailable attribute.
4612 * tree-core.h (struct tree_base): Add a bit to carry unavailability.
4613 * tree.c (error_unavailable_use): New.
4614 * tree.h (TREE_UNAVAILABLE): New.
4615 (error_unavailable_use): New.
4617 2021-09-01 Jakub Jelinek <jakub@redhat.com>
4619 PR tree-optimization/102124
4620 * tree-vect-patterns.c (vect_recog_widen_op_pattern): For ORIG_CODE
4621 MINUS_EXPR, if itype is unsigned with smaller precision than type,
4622 add an extra cast to signed variant of itype to ensure sign-extension.
4624 2021-09-01 Martin Liska <mliska@suse.cz>
4626 * graph.c (draw_cfg_node_succ_edges): Do not color fallthru
4627 edges and rather use colors for TRUE and FALSE edges.
4629 2021-09-01 Richard Biener <rguenther@suse.de>
4631 PR tree-optimization/93491
4632 * tree-ssa-pre.c (compute_avail): Set BB_MAY_NOTRETURN
4633 after processing the stmt itself. Do not consider
4634 pure functions possibly not returning. Properly avoid
4635 adding possibly trapping calls to EXP_GEN when there's
4636 a preceeding possibly not returning call.
4637 * tree-ssa-sccvn.c (vn_reference_may_trap): Conservatively
4640 2021-09-01 Richard Biener <rguenther@suse.de>
4642 PR tree-optimization/102139
4643 * tree-vectorizer.h (vec_base_alignments): Adjust hash-map
4644 type to record a std::pair of the stmt-info and the innermost
4646 (dr_vec_info::group): New member.
4647 * tree-vect-data-refs.c (vect_record_base_alignment): Adjust.
4648 (vect_compute_data_ref_alignment): Verify the recorded
4649 base alignment can be used.
4650 (data_ref_pair): Remove.
4651 (dr_group_sort_cmp): Adjust.
4652 (vect_analyze_data_ref_accesses): Store the group-ID in the
4653 dr_vec_info and operate on a vector of dr_vec_infos.
4655 2021-09-01 YunQiang Su <yunqiang.su@cipunited.com>
4657 * read-md.c (md_reader::handle_enum): support value assignation.
4658 * doc/md.texi: record define_c_enum value assignation support.
4660 2021-09-01 Jakub Jelinek <jakub@redhat.com>
4662 PR tree-optimization/102141
4663 * gimple-ssa-store-merging.c (bswap_view_convert): Add BEFORE
4664 argument. If false, emit stmts after gsi instead of before, and
4666 (bswap_replace): Adjust callers. When converting output of bswap,
4667 emit VIEW_CONVERT prepratation stmts after a copy of gsi instead
4670 2021-09-01 liuhongt <hongtao.liu@intel.com>
4672 * emit-rtl.c (validate_subreg): Get rid of all float-int
4675 2021-09-01 liuhongt <hongtao.liu@intel.com>
4678 2021-08-30 liuhongt <hongtao.liu@intel.com>
4680 * expmed.c (extract_bit_field_1): Make sure we're playing with
4681 integral modes before call extract_integral_bit_field.
4682 (extract_integral_bit_field): Add a parameter of type
4683 scalar_int_mode which corresponds to of tmode.
4684 And call extract_and_convert_fixed_bit_field instead of
4685 extract_fixed_bit_field and convert_extracted_bit_field.
4686 (extract_and_convert_fixed_bit_field): New function, it's a
4687 combination of extract_fixed_bit_field and
4688 convert_extracted_bit_field.
4690 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
4692 * tree.c (walk_tree_1) <OMP_CLAUSE_TILE>: Handle three operands.
4694 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
4696 * omp-general.h (omp_is_reference): Rename to...
4697 (omp_privatize_by_reference): ... this. Adjust all users...
4698 * omp-general.c: ... here, ...
4699 * gimplify.c: ... here, ...
4700 * omp-expand.c: ... here, ...
4701 * omp-low.c: ... here.
4703 2021-08-31 Martin Sebor <msebor@redhat.com>
4705 * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Test
4706 pointer element for equality to zero, not that of the cotaining
4709 2021-08-31 Martin Sebor <msebor@redhat.com>
4711 * gcc-rich-location.h (gcc_rich_location): Make ctor explicit.
4713 2021-08-31 Martin Sebor <msebor@redhat.com>
4715 * function.h (function): Add comments.
4716 (get_range_query): Same. Add attribute returns nonnull.
4718 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
4720 * expr.c (convert_modes): Don't use subreg_promoted_mode on a
4721 SUBREG if it can't be guaranteed to a SUBREG_PROMOTED_VAR_P set.
4722 Instead use the standard (safer) is_a <scalar_int_mode> idiom.
4724 2021-08-31 Jeff Law <jlaw@localhost.localdomain>
4726 * config.gcc (cris-*-elf, cris-*-none): Remove dbxelf.h from
4728 (m32r-*-elf, m32rle-*-elf, m32r-*-linux): Likewise.
4729 (mn10300-*-*, am33_2.0-*-linux*): Likewise.
4730 (xtensa*-*-elf, xtensa*-*-linux, xtensa*-*-uclinux): Likewise.
4731 (m32c-*-elf*, m32c-*-rtems*): Likewise.
4732 * config/cris/cris.h (DBX_NO_XREFS): Remove.
4733 (DBX_CONTIN_LENGTH, DBX_CONTIN_CHAR): Likewise.
4734 * config/m32r/m32r.h (DBXOUT_SOURCE_LINE): Likewise.
4735 (DBX_DEBUGGING_INFO, DBX_CONTIN_LENGTH): Likewise.
4736 * config/mn10300/mn10300.h (DEFAULT_GDB_EXTENSIONS): Likewise.
4737 * config/mn10300/linux.h (DBX_REGISTER_NAMES): Likewise.
4739 2021-08-31 Marcel Vollweiler <marcel@codesourcery.com>
4741 * gimplify.c (gimplify_scan_omp_clauses): Error handling. 'ancestor' only
4742 allowed on target constructs and only with particular other clauses.
4743 * omp-expand.c (expand_omp_target): Output of 'sorry, not supported' if
4745 * omp-low.c (check_omp_nesting_restrictions): Error handling. No nested OpenMP
4746 structs when 'ancestor' is used.
4747 (scan_omp_1_stmt): No usage of OpenMP runtime routines in a target region when
4749 * tree-pretty-print.c (dump_omp_clause): Append 'ancestor'.
4750 * tree.h (OMP_CLAUSE_DEVICE_ANCESTOR): Define macro.
4752 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
4754 * expr.c (convert_modes): Preserve SUBREG_PROMOTED_VAR_P when
4755 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
4757 * simplify-rtx.c (simplify_unary_operation_1) [SIGN_EXTEND]:
4758 Likewise, preserve SUBREG_PROMOTED_VAR_P when creating a (wider)
4759 partial subreg from a SUBREG_PROMOTED_VAR_P subreg. Generate
4760 SIGN_EXTEND of the SUBREG_REG when a subreg would be paradoxical.
4761 [ZERO_EXTEND]: Likewise, preserve SUBREG_PROMOTED_VAR_P when
4762 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
4763 subreg. Generate ZERO_EXTEND of the SUBREG_REG when a subreg
4764 would be paradoxical.
4766 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
4768 * combine.c (combine_simplify_rtx): Avoid converting an explicit
4769 TRUNCATE into a lowpart SUBREG on !TRULY_NOOP_TRUNCATION targets.
4770 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
4772 2021-08-31 Richard Biener <rguenther@suse.de>
4774 PR tree-optimization/102142
4775 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Fix
4776 condition under which to unset the visited flag.
4778 2021-08-31 Richard Biener <rguenther@suse.de>
4780 PR middle-end/102129
4781 * tree-ssa-ter.c (find_replaceable_in_bb): Do not move
4782 possibly trapping expressions across calls.
4784 2021-08-31 Jakub Jelinek <jakub@redhat.com>
4786 PR tree-optimization/102134
4787 * tree-ssa-ccp.c (bit_value_binop) <case RSHIFT_EXPR>: If sgn is
4788 UNSIGNED and r1val | r1mask has MSB set, ensure lzcount doesn't
4791 2021-08-31 Andrew Pinski <apinski@marvell.com>
4794 * collect-utils.c (setup_signals): New declaration.
4795 * collect-utils.h (setup_signals): New function.
4796 * collect2.c (handler): Delete.
4797 (main): Instead of manually setting up the signals,
4798 just call setup_signals.
4799 * lto-wrapper.c (main): Likewise.
4801 2021-08-31 Andrew Pinski <apinski@marvell.com>
4804 * config/i386/i386-protos.h (x86_output_aligned_bss):
4805 Change align argument to unsigned type.
4806 (x86_elf_aligned_decl_common): Likewise.
4807 * config/i386/i386.c (x86_elf_aligned_decl_common): Likewise.
4808 (x86_output_aligned_bss): Likewise.
4810 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
4812 * config/mips/mips.c (mips_module_isa_name): New.
4813 mips_file_start: add .module mipsREV to all asm output
4815 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
4817 * config/mips/mips.h (struct mips_cpu_info): define enum mips_isa;
4818 use enum instead of int for 'isa' member.
4819 * config.gcc, config/mips/mips.c, config/mips/mips-cpus.def,
4820 config/mips/netbsd.h: replace hardcoded numbers with enum.
4822 2021-08-31 liuhongt <hongtao.liu@intel.com>
4824 * config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
4825 define_split to define_insn_and_split.
4826 (*avx2_eq<mode>3): Removed.
4827 (<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
4828 (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
4829 (*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
4831 (*avx2_gt<mode>3): Removed.
4832 (<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
4833 define_insn to define_expand, and adjust pattern.
4834 (UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.
4836 2021-08-30 David Malcolm <dmalcolm@redhat.com>
4839 * Makefile.in (ANALYZER_OBJS): Add analyzer/call-info.o.
4841 2021-08-30 Jason Merrill <jason@redhat.com>
4843 * doc/invoke.texi: Document -Wmissing-requires.
4845 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
4847 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Remove
4848 TARGET_EXTRA_BUILTINS guard.
4850 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
4852 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Change
4853 initialization of V2DI_type_node and unsigned_V2DI_type_node.
4855 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
4857 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Use the new
4858 decl when new_builtins_are_live.
4859 * config/rs6000/rs6000-builtin-new.def (__builtin_cfstring): New
4862 2021-08-30 Pat Haugen <pthaugen@linux.ibm.com>
4864 * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add
4865 OPTION_MASK_P10_FUSION_2STORE.
4866 (POWERPC_MASKS): Likewise.
4867 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
4868 store fusion for Power10.
4869 (is_fusable_store): New.
4870 (power10_sched_reorder): Likewise.
4871 (rs6000_sched_reorder): Do Power10 specific reordering.
4872 (rs6000_sched_reorder2): Likewise.
4873 * config/rs6000/rs6000.opt: Add new option.
4875 2021-08-30 Richard Biener <rguenther@suse.de>
4877 PR tree-optimization/102128
4878 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
4879 Move scanning for if-converted scalar code to the caller
4880 and instead delay clearing the visited flag for profitable
4882 (vect_slp_region): Cost all subgraphs before scheduling.
4883 For if-converted BB vectorization scan for scalar COND_EXPRs
4884 and do not vectorize if any found and the cost model is
4887 2021-08-30 Richard Biener <rguenther@suse.de>
4889 * common.opt (fexceptions): Mark
4890 EnabledBy(fnon-call-exceptions).
4891 * doc/invoke.texi (fnon-call-exceptions): Document this
4892 enables -fexceptions.
4894 2021-08-30 Sebastian Huber <sebastian.huber@embedded-brains.de>
4896 * tsystem.h (abort): Define abort() if inhibit_libc is defined and it
4897 is not already defined.
4899 2021-08-30 liuhongt <hongtao.liu@intel.com>
4901 * expmed.c (extract_bit_field_1): Make sure we're playing with
4902 integral modes before call extract_integral_bit_field.
4903 (extract_integral_bit_field): Add a parameter of type
4904 scalar_int_mode which corresponds to of tmode.
4905 And call extract_and_convert_fixed_bit_field instead of
4906 extract_fixed_bit_field and convert_extracted_bit_field.
4907 (extract_and_convert_fixed_bit_field): New function, it's a
4908 combination of extract_fixed_bit_field and
4909 convert_extracted_bit_field.
4911 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
4913 * config/darwin.c (darwin_libc_has_function): Do not run
4914 the checks for x86 or modern Darwin. Make sure that there
4915 is a value set for darwin_macosx_version_min before testing.
4917 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
4919 * config/i386/darwin.h (CLEAR_INSN_CACHE): New.
4921 2021-08-28 Jan Hubicka <hubicka@ucw.cz>
4923 * ipa-modref-tree.h (modref_access_node::merge): Break out
4924 logic combining offsets and logic merging ranges to ...
4925 (modref_access_node::combined_offsets): ... here
4926 (modref_access_node::update2): ... here
4927 (modref_access_node::closer_pair_p): New member function.
4928 (modref_access_node::forced_merge): New member function.
4929 (modre_ref_node::insert): Do merging when table is full.
4931 2021-08-28 YunQiang Su <yunqiang.su@cipunited.com>
4934 * config.gcc: MIPS: use N64 ABI by default if the triple end
4935 with -gnuabi64, which is used by Debian since 2013.
4937 2021-08-28 Alexandre Oliva <oliva@adacore.com>
4939 * ipa-modref.c (analyze_function): Skip debug stmts.
4940 * tree-inline.c (estimate_num_insn): Consider builtins even
4941 without a cgraph_node.
4943 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
4945 * config/h8300/bitfield.md (cstore<mode>4): Remove expander.
4946 * config/h8300/h8300.c (h8300_expand_branch): Remove function.
4947 * config/h8300/h8300-protos.h (h8300_expadn_branch): Remove prototype.
4948 * config/h8300/h8300.md (eqne): New code iterator.
4949 (geultu, geultu_to_c): Similarly.
4950 * config/h8300/testcompare.md (cstore<mode>4): Dummy expander.
4951 (store_c_<mode>, store_c_i_<mode>): New define_insn_and_splits
4952 (cmp<mode>_c): New pattern
4954 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
4956 * tree-ssa-dom.c (reduce_vector_comparison_to_scalar_comparison): New
4958 (dom_opt_dom_walker::optimize_stmt): Use it.
4960 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
4962 * config/darwin.c (finalize_ctors): Add a section-start linker-
4964 (finalize_dtors): Likewise.
4965 * config/darwin.h (MIN_LD64_INIT_TERM_START_LABELS): New.
4967 2021-08-27 Bill Schmidt <wschmidt@linux.ibm.com>
4969 * config/rs6000/rs6000-call.c (rs6000-builtins.h): New #include.
4970 (rs6000_init_builtins): Call rs6000_init_generated_builtins. Skip the
4971 old initialization logic when new builtins are enabled.
4972 * config/rs6000/rs6000-gen-builtins.c (write_decls): Rename
4973 rs6000_autoinit_builtins to rs6000_init_generated_builtins.
4974 (write_init_file): Likewise.
4976 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
4978 * configure.ac (darwin2[[0-9]]* | darwin19*): Alter use of
4979 gcc_GAS_CHECK_FEATURE to remove an extraneous parameter.
4980 (amdgcn-* | gcn-*) Likewise.
4982 2021-08-27 Anthony Sharp <anthonysharp15@gmail.com>
4984 * symbol-summary.h: Added missing template keyword.
4986 2021-08-27 Richard Biener <rguenther@suse.de>
4988 PR tree-optimization/45178
4989 * tree-ssa-dce.c (find_obviously_necessary_stmts): For
4990 infinite loops without exit do not mark control dependent
4991 edges of the latch necessary.
4993 2021-08-27 konglin1 <lingling.kong@intel.com>
4996 * config/i386/sse.md: (<avx512>scattersi<mode>): Add mask operand to
4998 (<avx512>scattersi<mode>): Likewise.
4999 (*avx512f_scattersi<VI48F:mode>): Merge mask operand to set_dest.
5000 (*avx512f_scatterdi<VI48F:mode>): Likewise
5002 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
5004 * config/rs6000/rs6000.c (rs6000_builtin_md_vectorized_function): Add
5005 support for built-in functions MISC_BUILTIN_DIVWE, MISC_BUILTIN_DIVWEU,
5006 MISC_BUILTIN_DIVDE, MISC_BUILTIN_DIVDEU, P10_BUILTIN_CFUGED,
5007 P10_BUILTIN_CNTLZDM, P10_BUILTIN_CNTTZDM, P10_BUILTIN_PDEPD and
5008 P10_BUILTIN_PEXTD on Power10.
5010 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
5012 * config/rs6000/rs6000-call.c (builtin_function_type): Add unsigned
5013 signedness for some Power10 bifs.
5015 2021-08-27 David Edelsohn <dje.gcc@gmail.com>
5018 * config/rs6000/rs6000.c (rs6000_adjust_field_align): Use
5019 computed alignment if the entire struct has attribute packed.
5021 2021-08-27 liuhongt <hongtao.liu@intel.com>
5025 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
5026 IX86_BUILTIN_SHUFPD512, IX86_BUILTIN_SHUFPS512,
5027 IX86_BUILTIN_SHUFPD256, IX86_BUILTIN_SHUFPS,
5028 IX86_BUILTIN_SHUFPS256.
5029 (ix86_masked_all_ones): New function.
5031 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
5033 * config/i386/i386.md (*btr<mode>_1): Call force_reg unconditionally.
5034 (conditional moves with memory inputs splitters): Ditto.
5035 * config/i386/sse.md (one_cmpl<mode>2): Simplify.
5037 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
5039 * ipa-modref-tree.h (modref_access_node::try_merge_with): Restart
5040 search after merging.
5042 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
5044 * config/rs6000/rs6000-overload.def: Add remaining overloads.
5046 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
5048 * config/rs6000/rs6000-builtin-new.def: Add cell stanza.
5050 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
5052 * config/rs6000/rs6000-builtin-new.def: Add ieee128-hw, dfp,
5053 crypto, and htm stanzas.
5055 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
5057 * config/rs6000/rs6000-builtin-new.def: Add mma stanza.
5059 2021-08-26 Martin Sebor <msebor@redhat.com>
5061 * tree-ssa-uninit.c (warn_uninit): Refactor and simplify.
5062 (warn_uninit_phi_uses): Remove argument from calls to warn_uninit.
5063 (warn_uninitialized_vars): Same. Reduce visibility of locals.
5064 (warn_uninitialized_phi): Same.
5066 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
5068 * tree-ssa-ccp.c (get_individual_bits): Helper function to
5069 extract the individual bits from a widest_int constant (mask).
5070 (gray_code_bit_flips): New read-only table for effiently
5071 enumerating permutations/combinations of bits.
5072 (bit_value_binop) [LROTATE_EXPR, RROTATE_EXPR]: Handle rotates
5073 by unknown counts that are guaranteed less than the target
5074 precision and four or fewer unknown bits by enumeration.
5075 [LSHIFT_EXPR, RSHIFT_EXPR]: Likewise, also handle shifts by
5076 enumeration under the same conditions. Handle remaining
5077 shifts as a mask based upon the minimum possible shift value.
5079 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
5080 Richard Biener <rguenther@suse.de>
5082 * match.pd (shift transformations): Remove a redundant
5083 !POINTER_TYPE_P check.
5085 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
5088 * config/i386/i386.md (cmove reg-reg move elimination peephole2s):
5089 Set all_regs to true in the call to replace_rtx.
5091 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
5093 * ipa-modref-tree.c (test_insert_search_collapse): Update test.
5094 * ipa-modref-tree.h (modref_base_node::insert): Be smarter when
5095 hiting --param modref-max-refs limit.
5096 (modref_tree:insert_base): Be smarter when hitting
5097 --param modref-max-bases limit. Add new parameter REF.
5098 (modref_tree:insert): Update.
5099 (modref_tree:merge): Update.
5100 * ipa-modref.c (read_modref_records): Update.
5102 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
5104 * params.opt: (modref-max-adjustments): Add full stop.
5106 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
5108 * ipa-modref-tree.h (modref_ref_node::verify): New member
5110 (modref_ref_node::insert): Use it.
5111 (modref_ref_node::try_mere_with): Fix off by one error.
5113 2021-08-26 Martin Liska <mliska@suse.cz>
5114 Stefan Kneifel <stefan.kneifel@bluewin.ch>
5116 * cgraph.h (create_version_clone_with_body): Add new parameter.
5117 * cgraphclones.c: Likewise.
5118 * multiple_target.c (create_dispatcher_calls): Do not use
5120 (create_target_clone): Likewise here.
5122 2021-08-26 Jonathan Yong <10walls@gmail.com>
5124 * doc/extend.texi: Add note about reserved priorities
5125 to the constructor attribute.
5127 2021-08-25 Martin Sebor <msebor@redhat.com>
5129 * gimple-range-cache.cc (ssa_global_cache::dump): Avoid printing
5130 range table header alone.
5131 * gimple-range.cc (gimple_ranger::export_global_ranges): Same.
5133 2021-08-25 Jan Hubicka <hubicka@ucw.cz>
5135 * doc/invoke.texi: Document --param modref-max-adjustments.
5136 * ipa-modref-tree.c (test_insert_search_collapse): Update.
5137 (test_merge): Update.
5138 * ipa-modref-tree.h (struct modref_access_node): Add adjustments;
5139 (modref_access_node::operator==): Fix handling of access ranges.
5140 (modref_access_node::contains): Constify parameter; handle also
5141 mismatched parm offsets.
5142 (modref_access_node::update): New function.
5143 (modref_access_node::merge): New function.
5144 (unspecified_modref_access_node): Update constructor.
5145 (modref_ref_node::insert_access): Add record_adjustments parameter;
5147 (modref_ref_node::try_merge_with): New private function.
5148 (modref_tree::insert): New record_adjustments parameter.
5149 (modref_tree::merge): New record_adjustments parameter.
5150 (modref_tree::copy_from): Update.
5151 * ipa-modref.c (dump_access): Dump adjustments field.
5152 (get_access): Update constructor.
5153 (record_access): Update call of insert.
5154 (record_access_lto): Update call of insert.
5155 (merge_call_side_effects): Add record_adjustments parameter.
5156 (get_access_for_fnspec): Update.
5157 (process_fnspec): Update.
5158 (analyze_call): Update.
5159 (analyze_function): Update.
5160 (read_modref_records): Update.
5161 (ipa_merge_modref_summary_after_inlining): Update.
5162 (propagate_unknown_call): Update.
5163 (modref_propagate_in_scc): Update.
5164 * params.opt (param-max-modref-adjustments=): New.
5166 2021-08-25 Michael Meissner <meissner@linux.ibm.com>
5168 * config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
5170 (xxspltiw_v4si): Use vecperm type attribute.
5171 (xxspltiw_v4si_inst): Use vecperm type attribute.
5172 (xxspltiw_v4sf_inst): Likewise.
5173 (xxspltidp_v2df): Use vecperm type attribute. Use
5174 UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
5175 (xxspltidp_v2df_inst): Likewise.
5176 (xxsplti32dx_v4si): Use vecperm type attribute.
5177 (xxsplti32dx_v4si_inst): Likewise.
5178 (xxsplti32dx_v4sf_inst): Likewise.
5179 (xxblend_<mode>): Likewise.
5180 (xxpermx): Likewise.
5181 (xxpermx_inst): Likewise.
5184 2021-08-25 Lewis Hyatt <lhyatt@gmail.com>
5187 * coretypes.h (typedef diagnostic_input_charset_callback): Declare.
5188 * diagnostic.c (diagnostic_initialize_input_context): New function.
5189 * diagnostic.h (diagnostic_initialize_input_context): Declare.
5190 * input.c (default_charset_callback): New function.
5191 (file_cache::initialize_input_context): New function.
5192 (file_cache_slot::create): Added ability to convert the input
5193 according to the input context.
5194 (file_cache::file_cache): Initialize the new input context.
5195 (class file_cache_slot): Added new m_alloc_offset member.
5196 (file_cache_slot::file_cache_slot): Initialize the new member.
5197 (file_cache_slot::~file_cache_slot): Handle potentially offset buffer.
5198 (file_cache_slot::maybe_grow): Likewise.
5199 (file_cache_slot::needs_read_p): Handle NULL fp, which is now possible.
5200 (file_cache_slot::get_next_line): Likewise.
5201 * input.h (class file_cache): Added input context member.
5203 2021-08-25 Richard Biener <rguenther@suse.de>
5205 PR tree-optimization/102046
5206 * tree-vect-slp.c (vect_build_slp_tree_2): Conservatively
5207 update ->any_pattern when swapping operands.
5209 2021-08-25 Hongyu Wang <hongyu.wang@intel.com>
5212 * config/i386/i386.c (ix86_live_on_entry): Adjust comment.
5213 (ix86_decompose_address): Remove retval check for ASHIFT,
5214 allow non-canonical zero extend if AND mask covers ASHIFT
5216 (ix86_legitimate_address_p): Adjust condition for decompose.
5217 (ix86_rtx_costs): Adjust cost for lea with non-canonical
5219 Co-Authored by: Uros Bizjak <ubizjak@gmail.com>
5221 2021-08-25 Jiufu Guo <guojiufu@linux.ibm.com>
5223 PR tree-optimization/101145
5224 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
5226 (number_of_iterations_lt): Invoke above function.
5227 (adjust_cond_for_loop_until_wrap):
5228 Merge to number_of_iterations_until_wrap.
5229 (number_of_iterations_cond): Update invokes for
5230 adjust_cond_for_loop_until_wrap and number_of_iterations_lt.
5232 2021-08-25 konglin1 <lingling.kong@intel.com>
5235 * config/i386/avx512dqintrin.h (_mm512_fpclass_ps_mask): Fix
5237 (_mm512_mask_fpclass_ps_mask): Ditto.
5239 2021-08-25 Kewen Lin <linkw@linux.ibm.com>
5241 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Remove.
5242 (vec_unpacku_hi_v8hi): Likewise.
5243 (vec_unpacku_lo_v16qi): Likewise.
5244 (vec_unpacku_lo_v8hi): Likewise.
5245 (vec_unpacku_hi_<VP_small_lc>): New define_expand.
5246 (vec_unpacku_lo_<VP_small_lc>): Likewise.
5248 2021-08-24 David Edelsohn <dje.gcc@gmail.com>
5250 * config/rs6000/aix.h (SYSTEM_IMPLICIT_EXTERN_C): Delete.
5251 * config/rs6000/aix71.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
5252 * config/rs6000/aix72.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
5253 * config/rs6000/aix73.h (TARGET_AIX_VERSION): Increase to 73.
5255 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5257 PR middle-end/102031
5258 * simplify-rtx.c (simplify_truncation): When comparing precisions
5259 use "subreg_prec" variable, not "subreg_mode".
5261 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5263 * config/rs6000/rs6000-builtin-new.def: Add power10 and power10-64
5266 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5268 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Initialize
5269 various pointer type nodes.
5270 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add enum
5271 values for various pointer types.
5272 (ptr_V16QI_type_node): New macro.
5273 (ptr_V1TI_type_node): New macro.
5274 (ptr_V2DI_type_node): New macro.
5275 (ptr_V2DF_type_node): New macro.
5276 (ptr_V4SI_type_node): New macro.
5277 (ptr_V4SF_type_node): New macro.
5278 (ptr_V8HI_type_node): New macro.
5279 (ptr_unsigned_V16QI_type_node): New macro.
5280 (ptr_unsigned_V1TI_type_node): New macro.
5281 (ptr_unsigned_V8HI_type_node): New macro.
5282 (ptr_unsigned_V4SI_type_node): New macro.
5283 (ptr_unsigned_V2DI_type_node): New macro.
5284 (ptr_bool_V16QI_type_node): New macro.
5285 (ptr_bool_V8HI_type_node): New macro.
5286 (ptr_bool_V4SI_type_node): New macro.
5287 (ptr_bool_V2DI_type_node): New macro.
5288 (ptr_bool_V1TI_type_node): New macro.
5289 (ptr_pixel_type_node): New macro.
5290 (ptr_intQI_type_node): New macro.
5291 (ptr_uintQI_type_node): New macro.
5292 (ptr_intHI_type_node): New macro.
5293 (ptr_uintHI_type_node): New macro.
5294 (ptr_intSI_type_node): New macro.
5295 (ptr_uintSI_type_node): New macro.
5296 (ptr_intDI_type_node): New macro.
5297 (ptr_uintDI_type_node): New macro.
5298 (ptr_intTI_type_node): New macro.
5299 (ptr_uintTI_type_node): New macro.
5300 (ptr_long_integer_type_node): New macro.
5301 (ptr_long_unsigned_type_node): New macro.
5302 (ptr_float_type_node): New macro.
5303 (ptr_double_type_node): New macro.
5304 (ptr_long_double_type_node): New macro.
5305 (ptr_dfloat64_type_node): New macro.
5306 (ptr_dfloat128_type_node): New macro.
5307 (ptr_ieee128_type_node): New macro.
5308 (ptr_ibm128_type_node): New macro.
5309 (ptr_vector_pair_type_node): New macro.
5310 (ptr_vector_quad_type_node): New macro.
5311 (ptr_long_long_integer_type_node): New macro.
5312 (ptr_long_long_unsigned_type_node): New macro.
5314 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5316 * config/rs6000/rs6000-builtin-new.def: Add power9-vector, power9,
5317 and power9-64 stanzas.
5319 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5320 Tom de Vries <tdevries@suse.de>
5322 * config.gcc (nvptx-*-*): Define {c,c++}_target_objs.
5323 * config/nvptx/nvptx-protos.h (nvptx_cpu_cpp_builtins): Prototype.
5324 * config/nvptx/nvptx.h (TARGET_CPU_CPP_BUILTINS): Implement with
5325 a call to the new nvptx_cpu_cpp_builtins function in nvptx-c.c.
5326 * config/nvptx/t-nvptx (nvptx-c.o): New rule.
5327 * config/nvptx/nvptx-c.c: New source file.
5328 (nvptx_cpu_cpp_builtins): Move implementation here.
5330 2021-08-24 Martin Sebor <msebor@redhat.com>
5332 PR middle-end/101600
5333 PR middle-end/101977
5334 * gimple-ssa-warn-access.cc (maybe_warn_for_bound): Tighten up
5335 the phrasing of a warning.
5336 (check_access): Use the remaining size after subtracting any offset
5337 rather than the whole object size.
5338 * pointer-query.cc (access_ref::get_ref): Clear BASE0 flag if it's
5339 clear for any nonnull PHI argument.
5340 (compute_objsize): Clear argument.
5342 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5344 * config/rs6000/rs6000-builtin-new.def: Add power8-vector stanza.
5346 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
5348 * config/rs6000/rs6000-builtin-new.def: Add power7 and power7-64
5351 2021-08-24 Andrew MacLeod <amacleod@redhat.com>
5353 * value-relation.cc (rr_transitive_table): New.
5354 (relation_transitive): New.
5355 (value_relation::swap): Remove.
5356 (value_relation::apply_transitive): New.
5357 (relation_oracle::relation_oracle): Allocate a new tmp bitmap.
5358 (relation_oracle::register_relation): Call register_transitives.
5359 (relation_oracle::register_transitives): New.
5360 * value-relation.h (relation_oracle): Add new temporary bitmap and
5363 2021-08-24 H.J. Lu <hjl.tools@gmail.com>
5366 * config/i386/i386-expand.c (ix86_expand_vector_move): Broadcast
5367 from integer to a pseudo vector register.
5369 2021-08-24 Richard Biener <rguenther@suse.de>
5371 PR tree-optimization/100089
5372 * tree-vectorizer.h (vect_slp_bb): Rename to ...
5373 (vect_slp_if_converted_bb): ... this and get the original
5374 loop as new argument.
5375 * tree-vectorizer.c (try_vectorize_loop_1): Revert previous fix,
5376 pass original loop to vect_slp_if_converted_bb.
5377 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
5378 If orig_loop was passed scan the not vectorized stmts
5379 for COND_EXPRs and force not profitable if found.
5380 (vect_slp_region): Pass down all SLP instances to costing
5381 if orig_loop was specified.
5382 (vect_slp_bbs): Pass through orig_loop.
5383 (vect_slp_bb): Rename to ...
5384 (vect_slp_if_converted_bb): ... this and get the original
5385 loop as new argument.
5386 (vect_slp_function): Adjust.
5388 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
5391 * config/arm/arm.md (attribute arch): Add fix_vlldm.
5392 (arch_enabled): Use it.
5393 * config/arm/vfp.md (lazy_store_multiple_insn): Add alternative to
5394 use when erratum mitigation is needed.
5396 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
5399 * config/arm/arm.opt (mfix-cmse-cve-2021-35465): New option.
5400 * doc/invoke.texi (Arm Options): Document it.
5401 * config/arm/arm-cpus.in (quirk_vlldm): New feature bit.
5402 (ALL_QUIRKS): Add quirk_vlldm.
5403 (cortex-m33): Add quirk_vlldm.
5404 (cortex-m35p, cortex-m55): Likewise.
5405 * config/arm/arm.c (arm_option_override): Enable fix_vlldm if
5406 targetting an affected CPU and not explicitly controlled on
5409 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
5411 * config/arm/vfp.md (lazy_store_multiple_insn): Rewrite as valid RTL.
5412 (lazy_load_multiple_insn): Likewise.
5414 2021-08-24 liuhongt <hongtao.liu@intel.com>
5417 * config/i386/sse.md (<avx512>_vternlog<mode><sd_maskz_name>):
5418 Enable avx512 embedded broadcast.
5419 (*<avx512>_vternlog<mode>_all): Ditto.
5420 (<avx512>_vternlog<mode>_mask): Ditto.
5422 2021-08-24 liuhongt <hongtao.liu@intel.com>
5425 * config/i386/i386.c (ix86_rtx_costs): Define cost for
5427 * config/i386/i386.h (STRIP_UNARY): New macro.
5428 * config/i386/predicates.md (reg_or_notreg_operand): New
5430 * config/i386/sse.md (*<avx512>_vternlog<mode>_all): New define_insn.
5431 (*<avx512>_vternlog<mode>_1): New pre_reload
5432 define_insn_and_split.
5433 (*<avx512>_vternlog<mode>_2): Ditto.
5434 (*<avx512>_vternlog<mode>_3): Ditto.
5435 (any_logic1,any_logic2): New code iterator.
5436 (logic_op): New code attribute.
5437 (ternlogsuffix): Extend to VNxDF and VNxSF.
5439 2021-08-24 Richard Biener <rguenther@suse.de>
5441 * doc/invoke.texi (vect-inner-loop-cost-factor): Adjust.
5442 * params.opt (--param vect-inner-loop-cost-factor): Adjust
5444 * tree-vect-loop.c (vect_analyze_loop_form): Initialize
5445 inner_loop_cost_factor to the minimum of the estimated number
5446 of iterations of the inner loop and vect-inner-loop-cost-factor.
5448 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5449 Richard Biener <rguenther@suse.de>
5451 * config/i386/i386-features.c (compute_convert_gain): Provide
5452 more accurate values for CONST_INT, when optimizing for size.
5453 * config/i386/i386.c (COSTS_N_BYTES): Move definition from here...
5454 * config/i386/i386.h (COSTS_N_BYTES): to here.
5456 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
5457 Jakub Jelinek <jakub@redhat.com>
5459 PR middle-end/102029
5460 * match.pd (shift transformations): Add an additional check for
5461 !POINTER_TYPE_P in the recently added left shift transformation.
5463 2021-08-24 liuhongt <hongtao.liu@intel.com>
5465 PR tree-optimization/100089
5466 * tree-vectorizer.c (try_vectorize_loop_1): Disable slp in
5467 loop vectorizer when cost model is very-cheap.
5469 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5471 * config/rs6000/rs6000-gen-builtins.c (parse_bif_entry): Don't call
5472 asprintf, which is not available on AIX.
5474 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5476 * config.gcc (target_gtfiles): Add ./rs6000-builtins.h.
5477 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Set.
5479 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5481 * config.gcc (powerpc*-*-*): Add rs6000-builtins.o to extra_objs.
5482 * config/rs6000/rs6000-gen-builtins.c (main): Close init_file
5484 * config/rs6000/t-rs6000 (rs6000-gen-builtins.o): New target.
5485 (rbtree.o): Likewise.
5486 (rs6000-gen-builtins): Likewise.
5487 (rs6000-builtins.c): Likewise.
5488 (rs6000-builtins.h): Likewise.
5489 (rs6000.o): Add dependency.
5490 (EXTRA_HEADERS): Add rs6000-vecdefines.h.
5491 (rs6000-vecdefines.h): New target.
5492 (rs6000-builtins.o): Likewise.
5493 (rs6000-call.o): Add rs6000-builtins.h as a dependency.
5494 (rs6000-c.o): Likewise.
5496 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
5499 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace):
5500 Diagnose buffer overrun.
5501 (safe_inc_pos): Fix overrun detection.
5502 (match_identifier): Diagnose buffer overrun.
5503 (match_integer): Likewise.
5504 (match_to_right_bracket): Likewise.
5506 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
5508 * ipa-modref-tree.h (modref_access_node::range_info_useful_p):
5509 Improve range compare.
5510 (modref_access_node::contains): New member function.
5511 (modref_access_node::search): Remove.
5512 (modref_access_node::insert): Be smarter about subaccesses.
5514 2021-08-23 Thomas Schwinge <thomas@codesourcery.com>
5516 * config/i386/i386-options.c (ix86_omp_device_kind_arch_isa)
5517 <omp_device_arch> [ACCEL_COMPILER]: Match "intel_mic".
5518 * config/i386/t-omp-device (omp-device-properties-i386) <arch>:
5521 2021-08-23 Jeff Law <jlaw@localhost.localdomain>
5523 * config/h8300/h8300-protos.h (h8300_expand_epilogue): Add new
5525 * config/h8300/jumpcall.md (call, call_value): Restrict to
5526 !SIBLING_CALL_P cases.
5527 (subcall, sibcall_value): New patterns & expanders.
5528 * config/h8300/proepi.md (epilogue): Pass new argument to
5529 h8300_expand_epilogue.
5530 (sibcall_epilogue): New expander.
5531 * config/h8300/h8300.c (h8300_expand_epilogue): Handle sibcall
5533 (h8300_ok_for_sibcall_p): New function.
5534 (TARGET_FUNCTION_OK_FOR_SIBCALL): define.
5536 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5538 * simplify-rtx.c (simplify_unary_operation_1): [TRUNCATE]:
5539 Handle case where the operand is already the desired mode.
5541 2021-08-23 Richard Biener <rguenther@suse.de>
5544 * tree-ssa-structalias.c (ipa_pta_execute): Check in_other_partition
5545 in addition to has_gimple_body.
5547 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
5549 PR middle-end/101949
5550 * ipa-modref.c (analyze_ssa_name_flags): Fix merging of
5553 2021-08-23 Martin Liska <mliska@suse.cz>
5555 * doc/invoke.texi: Put the option out of -mxl-mode-app-model
5558 2021-08-23 Richard Biener <rguenther@suse.de>
5560 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
5561 Properly scale the inner loop cost only once.
5563 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5565 * tree-ssa-ccp.c (bit_value_binop) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]:
5566 Provide bounds for unsigned (and signed with non-negative operands)
5567 division and modulus.
5569 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5571 * simplify-rtx.c (simplify_truncation): Generalize simplification
5572 of (truncate:A (subreg:B X)).
5573 (simplify_unary_operation_1) [FLOAT_TRUNCATE, FLOAT_EXTEND,
5574 SIGN_EXTEND, ZERO_EXTEND]: Handle cases where the operand
5575 already has the desired machine mode.
5576 (test_scalar_int_ops): Add tests that useless extensions and
5577 truncations are optimized away.
5578 (test_scalar_int_ext_ops): New self-test function to confirm
5579 that truncations of extensions are correctly simplified.
5580 (test_scalar_int_ext_ops2): New self-test function to check
5581 truncations of truncations, extensions of extensions, and
5582 truncations of extensions.
5583 (test_scalar_ops): Call the above two functions with a
5584 representative sampling of integer machine modes.
5586 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
5588 * match.pd (shift transformations): Change the sign of an
5589 LSHIFT_EXPR if it reduces the number of explicit conversions.
5591 2021-08-23 Jakub Jelinek <jakub@redhat.com>
5593 PR tree-optimization/86723
5594 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
5595 cast64_to_32 argument, set *cast64_to_32 to false, unless n is
5596 non-memory permutation of 64-bit src which only has bytes of
5597 0 or [5..8] and n->range is 4.
5598 (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
5599 find_bswap_or_nop_finalize caller, support bswap with some bytes
5600 zeroed, as long as at least two bytes are not zeroed.
5601 (bswap_replace): Add mask argument and handle masking of bswap
5603 (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
5604 caller, punt if cast64_to_32 or mask is not all ones.
5605 (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
5606 caller, for now punt if cast64_to_32.
5608 2021-08-23 Richard Biener <rguenther@suse.de>
5610 PR tree-optimization/79334
5611 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Record
5612 a type also for COMPONENT_REFs.
5613 (vn_reference_may_trap): Check ARRAY_REF with constant index
5614 against the array domain.
5616 2021-08-23 liuhongt <hongtao.liu@intel.com>
5619 * config/i386/sse.md (*avx512f_pshufb_truncv8hiv8qi_1): Add
5620 TARGET_AVX512BW to condition.
5622 2021-08-23 Jakub Jelinek <jakub@redhat.com>
5625 * dwarf2out.c (gen_variable_die): Add DW_AT_location for global
5626 register variables already during early_dwarf if possible.
5628 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
5630 * config/arm/arm_mve.h: Fix __arm_vctp16q return type.
5632 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
5635 * config/arm/arm.opt: Fix typo.
5636 * config/arm/t-rmprofile: Fix typo.
5638 2021-08-23 Jakub Jelinek <jakub@redhat.com>
5640 * tree.h (OMP_CLAUSE_GRAINSIZE_STRICT): Define.
5641 (OMP_CLAUSE_NUM_TASKS_STRICT): Define.
5642 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_GRAINSIZE,
5643 case OMP_CLAUSE_NUM_TASKS>: Print strict: modifier.
5644 * omp-expand.c (expand_task_call): Use GOMP_TASK_FLAG_STRICT in iflags
5645 if either grainsize or num_tasks clause has the strict modifier.
5647 2021-08-23 Martin Liska <mliska@suse.cz>
5649 * dbgcnt.def (DEBUG_COUNTER): New counter.
5650 * gimple.c (gimple_call_arg_flags): Use it in IPA PTA.
5652 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
5654 * ipa-modref.c (analyze_ssa_name_flags): Improve handling of return slot.
5656 2021-08-23 Xi Ruoyao <xry111@mengyan1223.wang>
5659 * config/mips/mips-protos.h (mips_msa_output_shift_immediate):
5661 * config/mips/mips.c (mips_msa_output_shift_immediate): New
5663 * config/mips/mips-msa.md (vashl<mode>3, vashr<mode>3,
5664 vlshr<mode>3): Call it.
5666 2021-08-22 Jan Hubicka <hubicka@ucw.cz>
5667 Martin Liska <mliska@suse.cz>
5669 PR middle-end/101949
5670 * ipa-modref.c (analyze_ssa_name_flags): Indirect call implies
5673 2021-08-21 Dragan Mladjenovic <OT_Dragan.Mladjenovic@mediatek.com>
5675 * config/mips/mips.c (mips_function_rodata_section,
5676 TARGET_ASM_FUNCTION_RODATA_SECTION): Removed.
5678 2021-08-21 John David Anglin <danglin@gcc.gnu.org>
5680 * config/pa/pa.c (pa_asm_output_aligned_common): Remove warning.
5682 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5684 * configure.ac (thread-local storage support): Remove tls_first_major
5685 and tls_first_minor. Use "$conftest_s" to check support.
5686 * configure: Regenerate.
5688 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5690 * configure.ac: Fixup formatting.
5692 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5694 * acinclude.m4 (gcc_GAS_CHECK_FEATURE): Remove third argument and ...
5695 * configure.ac: ... update all callers.
5697 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
5700 * acinclude.m4 (_gcc_COMPUTE_GAS_VERSION, _gcc_GAS_VERSION_GTE_IFELSE)
5701 (gcc_GAS_VERSION_GTE_IFELSE): Remove.
5702 (gcc_GAS_CHECK_FEATURE): Do not handle in-tree case specially.
5703 * configure.ac: Remove gcc_cv_gas_major_version, gcc_cv_gas_minor_version.
5704 Remove remaining checks for in-tree assembler.
5705 * configure: Regenerate.
5707 2021-08-20 Jeff Law <jlaw@localhost.localdomain>
5709 * config/h8300/h8300.c (shift_alg_hi): Improve arithmetic shift right
5710 by 15 bits for H8/300H and H8/S. Improve logical shifts by 12
5712 (shift_alg_si): Improve arithmetic right shift by 28-30 bits for
5713 H8/300H. Improve arithmetic shift right by 15 bits for H8/S.
5714 Improve logical shifts by 27 bits for H8/S.
5715 (get_shift_alg): Corresponding changes.
5716 (h8300_option_override): Revert to loops for -Os when profitable.
5718 2021-08-20 Richard Biener <rguenther@suse.de>
5720 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
5722 (vect_analyze_data_ref_accesses): Likewise. Assign the BB
5723 index as group_id when dataref_groups were not computed.
5724 * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
5725 we advace to the next BB.
5727 2021-08-20 Jakub Jelinek <jakub@redhat.com>
5729 * omp-builtins.def (BUILT_IN_GOMP_WARNING, BUILT_IN_GOMP_ERROR): New
5732 2021-08-20 Martin Liska <mliska@suse.cz>
5734 PR gcov-profile/89961
5735 * gcov.c (make_gcov_file_name): Rewrite using std::string.
5736 (mangle_name): Simplify, do not used the second argument.
5737 (strip_extention): New function.
5738 (get_md5sum): Likewise.
5739 (get_gcov_intermediate_filename): Handle properly -p and -x
5741 (output_gcov_file): Use string type.
5742 (generate_results): Likewise.
5743 (md5sum_to_hex): Remove.
5745 2021-08-20 Michael Meissner <meissner@linux.ibm.com>
5747 * config/rs6000/altivec.md (UNSPEC_XXEVAL): Move to vsx.md.
5748 (UNSPEC_XXSPLTIW): Move to vsx.md.
5749 (UNSPEC_XXSPLTID): Move to vsx.md.
5750 (UNSPEC_XXSPLTI32DX): Move to vsx.md.
5751 (UNSPEC_XXBLEND): Move to vsx.md.
5752 (UNSPEC_XXPERMX): Move to vsx.md.
5753 (VM3): Move to vsx.md.
5754 (VM3_char): Move to vsx.md.
5755 (xxspltiw_v4si): Move to vsx.md.
5756 (xxspltiw_v4sf): Move to vsx.md.
5757 (xxspltiw_v4sf_inst): Move to vsx.md.
5758 (xxspltidp_v2df): Move to vsx.md.
5759 (xxspltidp_v2df_inst): Move to vsx.md.
5760 (xxsplti32dx_v4si_inst): Move to vsx.md.
5761 (xxsplti32dx_v4sf): Move to vsx.md.
5762 (xxsplti32dx_v4sf_inst): Move to vsx.md.
5763 (xxblend_<mode>): Move to vsx.md.
5764 (xxpermx): Move to vsx.md.
5765 (xxpermx_inst): Move to vsx.md.
5766 * config/rs6000/vsx.md (UNSPEC_XXEVAL): Move from altivec.md.
5767 (UNSPEC_XXSPLTIW): Move from altivec.md.
5768 (UNSPEC_XXSPLTID): Move from altivec.md.
5769 (UNSPEC_XXSPLTI32DX): Move from altivec.md.
5770 (UNSPEC_XXBLEND): Move from altivec.md.
5771 (UNSPEC_XXPERMX): Move from altivec.md.
5772 (VM3): Move from altivec.md.
5773 (VM3_char): Move from altivec.md.
5774 (xxspltiw_v4si): Move from altivec.md.
5775 (xxspltiw_v4sf): Move from altivec.md.
5776 (xxspltiw_v4sf_inst): Move from altivec.md.
5777 (xxspltidp_v2df): Move from altivec.md.
5778 (xxspltidp_v2df_inst): Move from altivec.md.
5779 (xxsplti32dx_v4si_inst): Move from altivec.md.
5780 (xxsplti32dx_v4sf): Move from altivec.md.
5781 (xxsplti32dx_v4sf_inst): Move from altivec.md.
5782 (xxblend_<mode>): Move from altivec.md.
5783 (xxpermx): Move from altivec.md.
5784 (xxpermx_inst): Move from altivec.md.
5786 2021-08-19 Roger Sayle <roger@nextmovesoftware.com>
5788 * tree-vect-generic.c (expand_vector_operations_1): Use either
5789 gimplify_build1 or gimplify_build2 instead of gimple_build_assign
5790 when constructing scalar splat expressions.
5792 2021-08-19 Peter Bergner <bergner@linux.ibm.com>
5795 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Cast
5796 pointer to __vector_pair *.
5798 2021-08-19 Martin Sebor <msebor@redhat.com>
5800 * gimple-range.cc: Add comments.
5801 * gimple-range.h: Same.
5803 2021-08-19 Martin Sebor <msebor@redhat.com>
5805 PR middle-end/101984
5806 * gimple-ssa-warn-access.cc (pass_waccess::execute): Also call
5809 2021-08-19 Jeff Law <jlaw@localhost.localdomain>
5811 * config.gcc (h8300-*-elf*): Do not include dbxelf.h.
5812 (h8300-*-linux*, v850-*-rtems*, v850*-elf*): Likewise.
5813 * config/v850/v850.h (DEFAULT_GDB_EXTENSIONS): Remove.
5815 2021-08-19 Jakub Jelinek <jakub@redhat.com>
5817 PR middle-end/101950
5818 * optabs.c (expand_clrsb_using_clz): New function.
5819 (expand_unop): Use it as another clrsb expansion fallback.
5821 2021-08-19 liuhongt <hongtao.liu@intel.com>
5824 2021-07-28 liuhongt <hongtao.liu@intel.com>
5827 * config/i386/i386.h (processor_costs): Add new member
5829 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
5830 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
5831 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
5832 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
5833 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
5834 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
5835 generic_cost, core_cost): Initialize integer_to_sse same value
5837 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
5838 * config/i386/i386.c (ix86_builtin_vectorization_cost):
5839 Use integer_to_sse instead of sse_op to calculate the cost of
5842 2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
5844 * config.gcc: Include rpath.opt for Darwin.
5845 * config/darwin.h (DRIVER_SELF_SPECS): Handle -rpath.
5847 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
5850 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
5853 2021-08-18 Jonathan Wright <jonathan.wright@arm.com>
5855 * config/aarch64/arm_neon.h (vld3_lane_f64): Use float RTL
5856 pattern and type cast.
5857 (vld4_lane_f32): Use float RTL pattern.
5858 (vld4q_lane_f64): Use float type cast.
5860 2021-08-18 Jan Hubicka <hubicka@ucw.cz>
5862 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Check also
5865 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
5867 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor): Extend.
5868 (test_map_of_type_with_ctor_and_dtor_expand): Add function.
5869 (hash_map_tests_c_tests): Call it.
5871 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
5873 * ggc.h (enum ggc_collect): New.
5874 (ggc_collect): Use it.
5875 * ggc-page.c: Adjust.
5876 * ggc-common.c: Likewise.
5877 * ggc-tests.c: Likewise.
5878 * read-rtl-function.c: Likewise.
5879 * selftest-run-tests.c: Likewise.
5880 * doc/gty.texi (Invoking the garbage collector): Likewise.
5882 2021-08-18 liuhongt <hongtao.liu@intel.com>
5885 * config/i386/i386.h (TARGET_V2DF_REDUCTION_PREFER_HADDPD):
5887 * config/i386/sse.md (*sse3_haddv2df3_low): Add
5888 TARGET_V2DF_REDUCTION_PREFER_HADDPD.
5889 (*sse3_hsubv2df3_low): Ditto.
5890 * config/i386/x86-tune.def
5891 (X86_TUNE_V2DF_REDUCTION_PREFER_HADDPD): New tune.
5893 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
5895 * gimple-range-gori.cc (gori_compute::gori_compute): Enable tracing.
5896 (gori_compute::compute_operand_range): Add tracing.
5897 (gori_compute::logical_combine): Ditto.
5898 (gori_compute::compute_logical_operands): Ditto.
5899 (gori_compute::compute_operand1_range): Ditto.
5900 (gori_compute::compute_operand2_range): Ditto.
5901 (gori_compute::outgoing_edge_range_p): Ditto.
5902 * gimple-range-gori.h (class gori_compute): Add range_tracer.
5904 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
5906 * flag-types.h (enum evrp_mode): Adjust evrp-mode values.
5907 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Relocate from.
5908 * gimple-range-trace.h (DEBUG_RANGE_CACHE): Here.
5909 * params.opt (--param=evrp-mode): Adjust options.
5911 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
5913 * Makefile.in (OBJS): Add gimple-range-trace.o.
5914 * gimple-range-cache.h (enable_new_values): Remove unused prototype.
5915 * gimple-range-fold.cc: Adjust headers.
5916 * gimple-range-trace.cc: New.
5917 * gimple-range-trace.h: New.
5918 * gimple-range.cc (gimple_ranger::gimple_ranger): Enable tracer.
5919 (gimple_ranger::range_of_expr): Add tracing.
5920 (gimple_ranger::range_on_entry): Ditto.
5921 (gimple_ranger::range_on_exit): Ditto.
5922 (gimple_ranger::range_on_edge): Ditto.
5923 (gimple_ranger::fold_range_internal): Ditto.
5924 (gimple_ranger::dump_bb): Do not calculate edge range twice.
5925 (trace_ranger::*): Remove.
5926 (enable_ranger): Never create a trace_ranger.
5927 (debug_seed_ranger): Move to gimple-range-trace.cc.
5928 (dump_ranger): Ditto.
5929 (debug_ranger): Ditto.
5930 * gimple-range.h: Include gimple-range-trace.h.
5931 (range_on_entry, range_on_exit): No longer virtual.
5932 (class trace_ranger): Remove.
5933 (DEBUG_RANGE_CACHE): Move to gimple-range-trace.h.
5935 2021-08-17 Martin Sebor <msebor@redhat.com>
5937 PR middle-end/101854
5938 * builtins.c (expand_builtin_alloca): Move warning code to check_alloca
5939 in gimple-ssa-warn-access.cc.
5940 * calls.c (alloc_max_size): Move code to check_alloca.
5941 (get_size_range): Move to pointer-query.cc.
5942 (maybe_warn_alloc_args_overflow): Move to gimple-ssa-warn-access.cc.
5943 (get_attr_nonstring_decl): Move to tree.c.
5944 (fntype_argno_type): Move to gimple-ssa-warn-access.cc.
5945 (append_attrname): Same.
5946 (maybe_warn_rdwr_sizes): Same.
5947 (initialize_argument_information): Move code to
5948 gimple-ssa-warn-access.cc.
5949 * calls.h (maybe_warn_alloc_args_overflow): Move to
5950 gimple-ssa-warn-access.h.
5951 (get_attr_nonstring_decl): Move to tree.h.
5952 (maybe_warn_nonstring_arg): Move to gimple-ssa-warn-access.h.
5953 (enum size_range_flags): Move to pointer-query.h.
5954 (get_size_range): Same.
5955 * gimple-ssa-warn-access.cc (has_location): Remove unused overload
5956 to avoid Clang -Wunused-function.
5957 (get_size_range): Declare static.
5958 (maybe_emit_free_warning): Rename...
5959 (maybe_check_dealloc_call): ...to this for consistency.
5960 (class pass_waccess): Add members.
5961 (pass_waccess::~pass_waccess): Defined.
5962 (alloc_max_size): Move here from calls.c.
5963 (maybe_warn_alloc_args_overflow): Same.
5964 (check_alloca): New function.
5965 (check_alloc_size_call): New function.
5966 (check_strncat): Handle another warning flag.
5967 (pass_waccess::check_builtin): Handle alloca.
5968 (fntype_argno_type): Move here from calls.c.
5969 (append_attrname): Same.
5970 (maybe_warn_rdwr_sizes): Same.
5971 (pass_waccess::check_call): Define.
5972 (check_nonstring_args): New function.
5973 (pass_waccess::check): Call new member functions.
5974 (pass_waccess::execute): Enable ranger.
5975 * gimple-ssa-warn-access.h (get_size_range): Move here from calls.h.
5976 (maybe_warn_nonstring_arg): Same.
5977 * gimple-ssa-warn-restrict.c: Remove #include.
5978 * pointer-query.cc (get_size_range): Move here from calls.c.
5979 * pointer-query.h (enum size_range_flags): Same.
5980 (get_size_range): Same.
5981 * tree.c (get_attr_nonstring_decl): Move here from calls.c.
5982 * tree.h (get_attr_nonstring_decl): Move here from calls.h.
5984 2021-08-17 Thomas Schwinge <thomas@codesourcery.com>
5986 * ggc.h (ggc_collect): Add 'force_collect' parameter.
5987 * ggc-page.c (ggc_collect): Use that one instead of global
5988 'ggc_force_collect'. Adjust all users.
5989 * doc/gty.texi (Invoking the garbage collector): Update.
5990 * ggc-internal.h (ggc_force_collect): Remove.
5991 * ggc-common.c (ggc_force_collect): Likewise.
5992 * selftest.h (forcibly_ggc_collect): Remove.
5993 * ggc-tests.c (selftest::forcibly_ggc_collect): Likewise.
5994 * read-rtl-function.c (test_loading_labels): Adjust.
5995 * selftest-run-tests.c (run_tests): Likewise.
5997 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
5999 * config/darwin.c (darwin_file_end): Reset and reclaim the
6000 section names table at the end of compile.
6002 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
6005 * config.in: Regenerate.
6006 * config/i386/darwin.h (EXTRA_ASM_OPTS): New
6007 (ASM_SPEC): Pass options to disable branch shortening where
6009 * configure: Regenerate.
6010 * configure.ac: Detect versions of 'as' that support the
6011 optimisation which has the bug.
6013 2021-08-17 Richard Biener <rguenther@suse.de>
6015 * optabs-query.c (supports_vec_gather_load_p): Also check
6017 (supports_vec_scatter_store_p): Likewise.
6018 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Fall
6019 back to masked variants if non-masked are not supported.
6020 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
6021 When we need to use masked gather/scatter but do not have
6022 a mask set up a constant true one.
6023 * tree-vect-stmts.c (vect_check_scalar_mask): Also allow
6026 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
6028 * tree-ssa-ccp.c (bit_value_binop) [MINUS_EXPR]: Use same
6029 algorithm as PLUS_EXPR to improve subtraction bit bounds.
6030 [POINTER_DIFF_EXPR]: Treat as synonymous with MINUS_EXPR.
6032 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
6034 * tree-ssa-ccp.c (bit_value_mult_const): New helper function to
6035 calculate the mask-value pair result of a multiplication by an
6037 (bit_value_binop) [MULT_EXPR]: Call it from here for
6038 multiplications by (sparse) non-negative constants.
6040 2021-08-17 Christophe Lyon <christophe.lyon@foss.st.com>
6043 * config.gcc (gcc_cv_initfini_array): Leave undefined for
6044 uclinuxfdpiceabi targets.
6046 2021-08-17 Alexandre Oliva <oliva@adacore.com>
6048 * tree-inline.c (maybe_move_debug_stmts_to_successors): Don't
6049 reverse debug stmts.
6051 2021-08-17 Alexandre Oliva <oliva@adacore.com>
6053 * tree-cfg.c (dump_function_to_file): Use fun, not cfun.
6055 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
6057 * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete.
6058 (__LD4Q_LANE_FUNC): Likewise.
6059 (vld4_lane_u8): Define without macro.
6060 (vld4_lane_u16): Likewise.
6061 (vld4_lane_u32): Likewise.
6062 (vld4_lane_u64): Likewise.
6063 (vld4_lane_s8): Likewise.
6064 (vld4_lane_s16): Likewise.
6065 (vld4_lane_s32): Likewise.
6066 (vld4_lane_s64): Likewise.
6067 (vld4_lane_f16): Likewise.
6068 (vld4_lane_f32): Likewise.
6069 (vld4_lane_f64): Likewise.
6070 (vld4_lane_p8): Likewise.
6071 (vld4_lane_p16): Likewise.
6072 (vld4_lane_p64): Likewise.
6073 (vld4q_lane_u8): Likewise.
6074 (vld4q_lane_u16): Likewise.
6075 (vld4q_lane_u32): Likewise.
6076 (vld4q_lane_u64): Likewise.
6077 (vld4q_lane_s8): Likewise.
6078 (vld4q_lane_s16): Likewise.
6079 (vld4q_lane_s32): Likewise.
6080 (vld4q_lane_s64): Likewise.
6081 (vld4q_lane_f16): Likewise.
6082 (vld4q_lane_f32): Likewise.
6083 (vld4q_lane_f64): Likewise.
6084 (vld4q_lane_p8): Likewise.
6085 (vld4q_lane_p16): Likewise.
6086 (vld4q_lane_p64): Likewise.
6087 (vld4_lane_bf16): Likewise.
6088 (vld4q_lane_bf16): Likewise.
6090 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
6092 * config/aarch64/arm_neon.h (__LD3_LANE_FUNC): Delete.
6093 (__LD3Q_LANE_FUNC): Delete.
6094 (vld3_lane_u8): Define without macro.
6095 (vld3_lane_u16): Likewise.
6096 (vld3_lane_u32): Likewise.
6097 (vld3_lane_u64): Likewise.
6098 (vld3_lane_s8): Likewise.
6099 (vld3_lane_s16): Likewise.
6100 (vld3_lane_s32): Likewise.
6101 (vld3_lane_s64): Likewise.
6102 (vld3_lane_f16): Likewise.
6103 (vld3_lane_f32): Likewise.
6104 (vld3_lane_f64): Likewise.
6105 (vld3_lane_p8): Likewise.
6106 (vld3_lane_p16): Likewise.
6107 (vld3_lane_p64): Likewise.
6108 (vld3q_lane_u8): Likewise.
6109 (vld3q_lane_u16): Likewise.
6110 (vld3q_lane_u32): Likewise.
6111 (vld3q_lane_u64): Likewise.
6112 (vld3q_lane_s8): Likewise.
6113 (vld3q_lane_s16): Likewise.
6114 (vld3q_lane_s32): Likewise.
6115 (vld3q_lane_s64): Likewise.
6116 (vld3q_lane_f16): Likewise.
6117 (vld3q_lane_f32): Likewise.
6118 (vld3q_lane_f64): Likewise.
6119 (vld3q_lane_p8): Likewise.
6120 (vld3q_lane_p16): Likewise.
6121 (vld3q_lane_p64): Likewise.
6122 (vld3_lane_bf16): Likewise.
6123 (vld3q_lane_bf16): Likewise.
6125 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
6127 * config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Delete.
6128 (__LD2Q_LANE_FUNC): Likewise.
6129 (vld2_lane_u8): Define without macro.
6130 (vld2_lane_u16): Likewise.
6131 (vld2_lane_u32): Likewise.
6132 (vld2_lane_u64): Likewise.
6133 (vld2_lane_s8): Likewise.
6134 (vld2_lane_s16): Likewise.
6135 (vld2_lane_s32): Likewise.
6136 (vld2_lane_s64): Likewise.
6137 (vld2_lane_f16): Likewise.
6138 (vld2_lane_f32): Likewise.
6139 (vld2_lane_f64): Likewise.
6140 (vld2_lane_p8): Likewise.
6141 (vld2_lane_p16): Likewise.
6142 (vld2_lane_p64): Likewise.
6143 (vld2q_lane_u8): Likewise.
6144 (vld2q_lane_u16): Likewise.
6145 (vld2q_lane_u32): Likewise.
6146 (vld2q_lane_u64): Likewise.
6147 (vld2q_lane_s8): Likewise.
6148 (vld2q_lane_s16): Likewise.
6149 (vld2q_lane_s32): Likewise.
6150 (vld2q_lane_s64): Likewise.
6151 (vld2q_lane_f16): Likewise.
6152 (vld2q_lane_f32): Likewise.
6153 (vld2q_lane_f64): Likewise.
6154 (vld2q_lane_p8): Likewise.
6155 (vld2q_lane_p16): Likewise.
6156 (vld2q_lane_p64): Likewise.
6157 (vld2_lane_bf16): Likewise.
6158 (vld2q_lane_bf16): Likewise.
6160 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
6162 * haifa-sched.c (advance_one_cycle): Output more context-synchronization
6165 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
6167 * haifa-sched.c (enum rfs_decision, rfs_str): Add RFS_AUTOPREF.
6168 (rank_for_schedule): Use it.
6170 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
6172 PR rtl-optimization/91598
6173 * haifa-sched.c (autopref_rank_for_schedule): Prioritize "irrelevant"
6174 insns after memory reads and before memory writes.
6176 2021-08-17 Alistair_Lee <alistair.lee@arm.com>
6178 * rtl.h (CONST_VECTOR_P): New macro.
6179 * config/aarch64/aarch64.c (aarch64_get_sve_pred_bits): Use RTL
6180 code testing macros.
6181 (aarch64_ptrue_all_mode): Likewise.
6182 (aarch64_expand_mov_immediate): Likewise.
6183 (aarch64_const_vec_all_in_range_p): Likewise.
6184 (aarch64_rtx_costs): Likewise.
6185 (aarch64_legitimate_constant_p): Likewise.
6186 (aarch64_simd_valid_immediate): Likewise.
6187 (aarch64_simd_make_constant): Likewise.
6188 (aarch64_convert_mult_to_shift): Likewise.
6189 (aarch64_expand_sve_vec_perm): Likewise.
6190 (aarch64_vec_fpconst_pow_of_2): Likewise.
6192 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
6194 PR tree-optimization/101938
6195 * range-op.cc (operator_abs::op1_range): Special case
6196 -TYPE_MIN_VALUE for flag_wrapv.
6198 2021-08-17 Kewen Lin <linkw@linux.ibm.com>
6200 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Add the cost for
6203 2021-08-17 Jakub Jelinek <jakub@redhat.com>
6205 * tree.def (OMP_SCOPE): New tree code.
6206 * tree.h (OMP_SCOPE_BODY, OMP_SCOPE_CLAUSES): Define.
6207 * tree-nested.c (convert_nonlocal_reference_stmt,
6208 convert_local_reference_stmt, convert_gimple_call): Handle
6210 * tree-pretty-print.c (dump_generic_node): Handle OMP_SCOPE.
6211 * gimple.def (GIMPLE_OMP_SCOPE): New gimple code.
6212 * gimple.c (gimple_build_omp_scope): New function.
6213 (gimple_copy): Handle GIMPLE_OMP_SCOPE.
6214 * gimple.h (gimple_build_omp_scope): Declare.
6215 (gimple_has_substatements): Handle GIMPLE_OMP_SCOPE.
6216 (gimple_omp_scope_clauses, gimple_omp_scope_clauses_ptr,
6217 gimple_omp_scope_set_clauses): New inline functions.
6218 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_SCOPE.
6219 * gimple-pretty-print.c (dump_gimple_omp_scope): New function.
6220 (pp_gimple_stmt_1): Handle GIMPLE_OMP_SCOPE.
6221 * gimple-walk.c (walk_gimple_stmt): Likewise.
6222 * gimple-low.c (lower_stmt): Likewise.
6223 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
6224 (gimplify_scan_omp_clauses): For task reductions, handle OMP_SCOPE
6225 like ORT_WORKSHARE constructs. Adjust diagnostics for %<scope%>
6226 allowing task reductions. Reject inscan reductions on scope.
6227 (omp_find_stores_stmt): Handle GIMPLE_OMP_SCOPE.
6228 (gimplify_omp_workshare, gimplify_expr): Handle OMP_SCOPE.
6229 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_SCOPE.
6230 (estimate_num_insns): Likewise.
6231 * omp-low.c (build_outer_var_ref): Look through GIMPLE_OMP_SCOPE
6232 contexts if var isn't privatized there.
6233 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_SCOPE.
6234 (scan_omp_1_stmt): Likewise.
6235 (maybe_add_implicit_barrier_cancel): Look through outer
6237 (lower_omp_scope): New function.
6238 (lower_omp_task_reductions): Handle OMP_SCOPE.
6239 (lower_omp_1): Handle GIMPLE_OMP_SCOPE.
6240 (diagnose_sb_1, diagnose_sb_2): Likewise.
6241 * omp-expand.c (expand_omp_single): Support also GIMPLE_OMP_SCOPE.
6242 (expand_omp): Handle GIMPLE_OMP_SCOPE.
6243 (omp_make_gimple_edges): Likewise.
6244 * omp-builtins.def (BUILT_IN_GOMP_SCOPE_START): New built-in.
6246 2021-08-17 Richard Biener <rguenther@suse.de>
6248 PR tree-optimization/101925
6249 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Set
6250 reverse on COMPONENT_REF and ARRAY_REF according to
6251 what reverse_storage_order_for_component_p does.
6252 (vn_reference_eq): Compare reversed on reference ops.
6253 (reverse_storage_order_for_component_p): New overload.
6254 (vn_reference_lookup_3): Check reverse_storage_order_for_component_p
6255 on the reference looked up.
6257 2021-08-17 Jeff Law <jlaw@localhost.localdomain>
6259 * config/h8300/h8300.c (shift_alg_si): Avoid loops for most SImode
6261 (h8300_option_override): Use loops on H8/S more often when optimizing
6263 (get_shift_alg): Handle new "special" cases on H8/S. Simplify
6264 accordingly. Handle various arithmetic right shifts with special
6265 sequences that we couldn't handle before.
6267 2021-08-16 Jeff Law <jlaw@localhost.localdomain>
6269 * config.gcc (rl78-*-elf*): Do not include dbxelf.h.
6271 2021-08-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
6273 * config/sparc/rtemself.h (SPARC_GCOV_TYPE_SIZE): Define.
6274 * config/sparc/sparc.c (sparc_gcov_type_size): New.
6275 (TARGET_GCOV_TYPE_SIZE): Redefine if SPARC_GCOV_TYPE_SIZE is defined.
6276 * coverage.c (get_gcov_type): Use targetm.gcov_type_size().
6277 * doc/tm.texi (TARGET_GCOV_TYPE_SIZE): Add hook under "Misc".
6278 * doc/tm.texi.in: Regenerate.
6279 * target.def (gcov_type_size): New target hook.
6280 * targhooks.c (default_gcov_type_size): New.
6281 * targhooks.h (default_gcov_type_size): Declare.
6282 * tree-profile.c (gimple_gen_edge_profiler): Use precision of
6284 (gimple_gen_time_profiler): Likewise.
6286 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
6288 * dwarf2out.c (add_scalar_info): Deal with DW_AT_data_bit_offset.
6290 2021-08-16 Tobias Burnus <tobias@codesourcery.com>
6292 PR middle-end/101931
6293 * omp-low.c (omp_runtime_api_call): Update for routines
6294 added in the meanwhile.
6296 2021-08-16 Martin Liska <mliska@suse.cz>
6298 PR tree-optimization/100393
6299 * tree-switch-conversion.c (group_cluster::dump): Use
6300 get_comparison_count.
6301 (jump_table_cluster::find_jump_tables): Pre-compute number of
6302 comparisons and then decrement it. Cache also max_ratio.
6303 (jump_table_cluster::can_be_handled): Change signature.
6304 * tree-switch-conversion.h (get_comparison_count): New.
6306 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
6308 * dwarf2out.c (add_data_member_location_attribute): Use GNAT
6309 encodings only when -fgnat-encodings=all is specified.
6310 (add_bound_info): Likewise.
6311 (add_byte_size_attribute): Likewise.
6312 (gen_member_die): Likewise.
6314 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
6316 * omp-oacc-neuter-broadcast.cc
6317 (execute_omp_oacc_neuter_broadcast): Plug 'par' memory leak.
6319 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
6321 * omp-oacc-neuter-broadcast.cc
6322 (execute_omp_oacc_neuter_broadcast): Clarify memory management for
6325 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
6327 * omp-oacc-neuter-broadcast.cc (field_map): Move variable into...
6328 (execute_omp_oacc_neuter_broadcast): ... here.
6329 (install_var_field, build_receiver_ref, build_sender_ref): Take
6330 'field_map_t *' parameter. Adjust all users.
6331 (worker_single_copy, neuter_worker_single): Take a
6332 'record_field_map_t *' parameter. Adjust all users.
6334 2021-08-16 liuhongt <hongtao.liu@intel.com>
6337 * config/i386/i386.md (ldexp<mode>3): Force operands[1] to
6340 2021-08-16 Martin Liska <mliska@suse.cz>
6343 * multiple_target.c (create_dispatcher_calls): Make default
6344 function local only if it is a definition.
6346 2021-08-16 Martin Liska <mliska@suse.cz>
6349 * ipa-icf-gimple.c (func_checker::compare_ssa_name): Do not
6350 consider equal SSA_NAMEs when one is a param.
6352 2021-08-16 liuhongt <hongtao.liu@intel.com>
6355 * config/i386/i386-expand.c (ix86_expand_vec_perm_vpermt2):
6356 Support vpermi2b for V32QI/V16QImode.
6357 (ix86_extract_perm_from_pool_constant): New function.
6358 (ix86_expand_vec_one_operand_perm_avx512): Support
6359 vpermw/vpermb under TARGET_AVX512BW/TARGET_AVX512VBMI.
6360 (expand_vec_perm_1): Adjust comments for upper.
6361 * config/i386/i386-protos.h (ix86_extract_perm_from_pool_constant):
6363 * config/i386/predicates.md (permvar_truncate_operand): New predicate.
6364 (pshufb_truncv4siv4hi_operand): Ditto.
6365 (pshufb_truncv8hiv8qi_operand): Ditto.
6366 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1):
6367 New pre_reload define_insn_and_split.
6368 (*avx512f_permvar_truncv8siv8hi_1): Ditto.
6369 (*avx512f_vpermvar_truncv8div8si_1): Ditto.
6370 (*avx512f_permvar_truncv32hiv32qi_1): Ditto.
6371 (*avx512f_permvar_truncv16hiv16qi_1): Ditto.
6372 (*avx512f_permvar_truncv4div4si_1): Ditto.
6373 (*avx512f_pshufb_truncv8hiv8qi_1): Ditto.
6374 (*avx512f_pshufb_truncv4siv4hi_1): Ditto.
6375 (*avx512f_pshufd_truncv2div2si_1): Ditto.
6377 2021-08-16 Kito Cheng <kito.cheng@sifive.com>
6379 * config/riscv/multilib-generator: Support code model option for
6381 * doc/install.texi: Add document of new option for
6382 --with-multilib-generator.
6384 2021-08-15 Clément Chigot <clement.chigot@atos.net>
6386 * config/rs6000/rs6000.c (xcoff_tls_exec_model_detected): New.
6387 (rs6000_legitimize_tls_address_aix): Use it.
6388 (rs6000_xcoff_file_end): Add ".ref __tls_get_addr" when
6389 xcoff_tls_exec_model_detected is true.
6391 2021-08-15 Jeff Law <jlaw@localhost.localdomain>
6393 * config/h8300/h8300.c (shift_alg_si): Retune H8/300H shifts
6394 to allow a bit more code growth, saving many dozens of cycles.
6395 (h8300_option_override): Adjus shift_alg_si if optimizing for
6397 (get_shift_alg): Use special + inline shifts for residuals
6400 2021-08-14 Stafford Horne <shorne@gmail.com>
6403 * config/or1k/or1k-opts.h: New file.
6404 * config/or1k/or1k.c (or1k_legitimize_address_1, print_reloc):
6405 Support generating gotha relocations if -mcmodel=large is
6407 * config/or1k/or1k.h (TARGET_CMODEL_SMALL, TARGET_CMODEL_LARGE):
6409 * config/or1k/or1k.opt (mcmodel=): New option.
6410 * doc/invoke.texi (OpenRISC Options): Document mcmodel.
6412 2021-08-14 Martin Sebor <msebor@redhat.com>
6414 PR middle-end/101791
6415 * gimple-ssa-warn-access.cc (new_delete_mismatch_p): Use new argument
6416 to valid_new_delete_pair_p.
6417 * tree.c (valid_new_delete_pair_p): Add argument.
6418 * tree.h (valid_new_delete_pair_p): Same.
6420 2021-08-14 Jakub Jelinek <jakub@redhat.com>
6423 * config/i386/i386-expand.c (expand_vec_perm_broadcast_1)
6424 <case E_V64QImode>: For this mode assert
6425 !TARGET_AVX512BW || d->perm[0] rather than !TARGET_AVX2 || d->perm[0].
6427 2021-08-13 Michael Meissner <meissner@linux.ibm.com>
6430 * config/rs6000/altivec.md (xxeval): Use register_predicate
6431 instead of altivec_register_predicate.
6433 2021-08-13 Martin Sebor <msebor@redhat.com>
6435 PR middle-end/101734
6436 * tree-ssa-uninit.c (maybe_warn_read_write_only): New function.
6437 (maybe_warn_operand): Call it.
6439 2021-08-13 Martin Liska <mliska@suse.cz>
6442 * attribs.c (decl_attributes): Make naked functions "noipa"
6445 2021-08-13 Martin Liska <mliska@suse.cz>
6448 * symtab.c (symtab_node::noninterposable_alias): Do not create
6449 local aliases for target_clone functions as the clonning pass
6452 2021-08-13 Martin Liska <mliska@suse.cz>
6454 * opts.c (LIVE_PATCHING_OPTION): Define.
6455 (control_options_for_live_patching): Use it in error messages.
6457 2021-08-13 Jan Hubicka <hubicka@ucw.cz>
6459 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOREAD.
6460 (implicit_const_eaf_flags, implicit_pure_eaf_flags,
6461 ignore_stores_eaf_flags): New constants.
6462 (remove_useless_eaf_flags): New function.
6463 (eaf_flags_useful_p): Use it.
6464 (deref_flags): Add EAF_NOT_RETURNED if flag is unused;
6466 (modref_lattice::init): Add EAF_NOREAD.
6467 (modref_lattice::add_escape_point): Do not reacord escape point if
6469 (modref_lattice::merge): EAF_NOESCAPE implies EAF_NODIRECTESCAPE;
6470 use remove_useless_eaf_flags.
6471 (modref_lattice::merge_deref): Use ignore_stores_eaf_flags.
6472 (modref_lattice::merge_direct_load): Add EAF_NOREAD
6473 (analyze_ssa_name_flags): Fix handling EAF_NOT_RETURNED
6474 (analyze_parms): Use remove_useless_eaf_flags.
6475 (ipa_merge_modref_summary_after_inlining): Use ignore_stores_eaf_flags.
6476 (modref_merge_call_site_flags): Add caller and ecf_flags parameter;
6477 use remove_useless_eaf_flags.
6478 (modref_propagate_flags_in_scc): Update.
6479 * ipa-modref.h: Turn eaf_flags_t back to char.
6480 * tree-core.h (EAF_NOT_RETURNED): Fix.
6481 (EAF_NOREAD): New constant
6482 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1): Check for
6484 * tree-ssa-structalias.c (handle_rhs_call): Handle new flags.
6485 (handle_pure_call): Likewise.
6487 2021-08-12 Jakub Jelinek <jakub@redhat.com>
6489 * tree.def (OMP_MASKED): New tree code.
6490 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_FILTER.
6491 * tree.h (OMP_MASKED_BODY, OMP_MASKED_CLAUSES, OMP_MASKED_COMBINED,
6492 OMP_CLAUSE_FILTER_EXPR): Define.
6493 * tree.c (omp_clause_num_ops): Add OMP_CLAUSE_FILTER entry.
6494 (omp_clause_code_name): Likewise.
6495 (walk_tree_1): Handle OMP_CLAUSE_FILTER.
6496 * tree-nested.c (convert_nonlocal_omp_clauses,
6497 convert_local_omp_clauses): Handle OMP_CLAUSE_FILTER.
6498 (convert_nonlocal_reference_stmt, convert_local_reference_stmt,
6499 convert_gimple_call): Handle GIMPLE_OMP_MASTER.
6500 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_FILTER.
6501 (dump_generic_node): Handle OMP_MASTER.
6502 * gimple.def (GIMPLE_OMP_MASKED): New gimple code.
6503 * gimple.c (gimple_build_omp_masked): New function.
6504 (gimple_copy): Handle GIMPLE_OMP_MASKED.
6505 * gimple.h (gimple_build_omp_masked): Declare.
6506 (gimple_has_substatements): Handle GIMPLE_OMP_MASKED.
6507 (gimple_omp_masked_clauses, gimple_omp_masked_clauses_ptr,
6508 gimple_omp_masked_set_clauses): New inline functions.
6509 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_MASKED.
6510 * gimple-pretty-print.c (dump_gimple_omp_masked): New function.
6511 (pp_gimple_stmt_1): Handle GIMPLE_OMP_MASKED.
6512 * gimple-walk.c (walk_gimple_stmt): Likewise.
6513 * gimple-low.c (lower_stmt): Likewise.
6514 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
6515 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_FILTER. For clauses
6516 that take one expression rather than decl or constant, force
6517 gimplification of that into a SSA_NAME or temporary unless min
6519 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_FILTER.
6520 (gimplify_expr): Handle OMP_MASKED.
6521 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_MASKED.
6522 (estimate_num_insns): Likewise.
6523 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_FILTER.
6524 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_MASKED. Adjust
6525 diagnostics for existence of masked construct.
6526 (scan_omp_1_stmt, lower_omp_master, lower_omp_1, diagnose_sb_1,
6527 diagnose_sb_2): Handle GIMPLE_OMP_MASKED.
6528 * omp-expand.c (expand_omp_synch, expand_omp, omp_make_gimple_edges):
6531 2021-08-12 Uroš Bizjak <ubizjak@gmail.com>
6534 * config/i386/i386.md (avx512f_scalef<mode>2): New insn pattern.
6535 (ldexp<mode>3): Use avx512f_scalef<mode>2.
6536 (UNSPEC_SCALEF): Move from ...
6537 * config/i386/sse.md (UNSPEC_SCALEF): ... here.
6539 2021-08-12 Jan Hubicka <hubicka@ucw.cz>
6541 * ipa-split.c (consider_split): Fix condition testing void functions.
6543 2021-08-12 Aldy Hernandez <aldyh@redhat.com>
6545 * doc/invoke.texi: Remove docs for threader-mode param.
6546 * flag-types.h (enum threader_mode): Remove.
6547 * params.opt: Remove threader-mode param.
6548 * tree-ssa-threadbackward.c (class back_threader): Remove
6549 path_is_unreachable_p.
6550 Make find_paths private.
6551 Add maybe_thread and thread_through_all_blocks.
6552 Remove reference marker for m_registry.
6553 Remove reference marker for m_profit.
6554 (back_threader::back_threader): Adjust for registry and profit not
6556 (dump_path): Move down.
6558 (class thread_jumps): Remove.
6559 (class back_threader_registry): Remove m_all_paths.
6561 (thread_jumps::thread_through_all_blocks): Move to back_threader
6563 (fsm_find_thread_path): Remove
6564 (back_threader::maybe_thread): New.
6565 (back_threader::thread_through_all_blocks): Move from
6567 (back_threader_registry::back_threader_registry): Remove
6569 (back_threader_registry::~back_threader_registry): Remove.
6570 (thread_jumps::find_taken_edge): Remove.
6571 (thread_jumps::check_subpath_and_update_thread_path): Remove.
6572 (thread_jumps::maybe_register_path): Remove.
6573 (thread_jumps::handle_phi): Remove.
6574 (handle_assignment_p): Remove.
6575 (thread_jumps::handle_assignment): Remove.
6576 (thread_jumps::fsm_find_control_statement_thread_paths): Remove.
6577 (thread_jumps::find_jump_threads_backwards): Remove.
6578 (thread_jumps::find_jump_threads_backwards_with_ranger): Remove.
6579 (try_thread_blocks): Rename find_jump_threads_backwards to
6581 (pass_early_thread_jumps::execute): Same.
6583 2021-08-12 Tobias Burnus <tobias@codesourcery.com>
6585 * tree-core.h (omp_clause_proc_bind_kind): Add
6586 OMP_CLAUSE_PROC_BIND_PRIMARY.
6587 * tree-pretty-print.c (dump_omp_clause): Add TODO comment to
6588 change 'master' to 'primary' in proc_bind for OpenMP 5.1.
6590 2021-08-12 Claudiu Zissulescu <claziss@synopsys.com>
6592 * common/config/arc/arc-common.c (arc_option_init_struct): Remove
6593 fno-common reference.
6594 * config/arc/arc.c (arc_override_options): Remove overriding of
6597 2021-08-12 Jakub Jelinek <jakub@redhat.com>
6600 * config/i386/i386-expand.c (ix86_expand_vec_one_operand_perm_avx512):
6601 If d->testing_p, return true after performing checks instead of
6602 actually expanding the insn.
6603 (expand_vec_perm_broadcast_1): Handle V32HImode - assert
6604 !TARGET_AVX512BW and return false.
6606 2021-08-12 Eric Botcazou <ebotcazou@gcc.gnu.org>
6608 * configure.ac (PE linker --disable-dynamicbase support): New check.
6609 * configure: Regenerate.
6610 * config.in: Likewise.
6611 * config/i386/mingw32.h (LINK_SPEC_DISABLE_DYNAMICBASE): New define.
6612 (LINK_SPEC): Use it.
6613 * config/i386/mingw-w64.h (LINK_SPEC_DISABLE_DYNAMICBASE): Likewise.
6614 (LINK_SPEC): Likewise.
6616 2021-08-12 liuhongt <hongtao.liu@intel.com>
6619 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_2): New
6620 post_reload define_insn_and_split.
6621 (*avx512bw_zero_extendv32qiv32hi2_2): Ditto.
6622 (*sse4_1_zero_extendv8qiv8hi2_4): Ditto.
6623 (*avx512f_zero_extendv16hiv16si2_2): Ditto.
6624 (*avx2_zero_extendv8hiv8si2_2): Ditto.
6625 (*sse4_1_zero_extendv4hiv4si2_4): Ditto.
6626 (*avx512f_zero_extendv8siv8di2_2): Ditto.
6627 (*avx2_zero_extendv4siv4di2_2): Ditto.
6628 (*sse4_1_zero_extendv2siv2di2_4): Ditto.
6629 (VI248_256, VI248_512, VI148_512, VI148_256, VI148_128): New
6632 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
6634 * config/rs6000/rs6000-builtin-new.def: Add always, power5, and
6637 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
6639 * config/rs6000/rs6000-builtin-new.def: Add vsx stanza.
6641 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
6643 * config/rs6000/rs6000-builtin-new.def: Finish altivec stanza.
6644 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
6645 initialization of pcvoid_type_node here...
6646 (altivec_init_builtins): ...from here.
6647 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add
6648 RS6000_BTI_const_ptr_void.
6649 (pcvoid_type_node): New macro.
6651 2021-08-11 Richard Biener <rguenther@suse.de>
6654 * tree-ssa-forwprop.c (pass_forwprop::execute): Do not decompose
6655 hard-register accesses.
6657 2021-08-11 Richard Biener <rguenther@suse.de>
6659 * tree-ssa-operands.c (operands_scanner::get_expr_operands):
6660 Do not look at COMPONENT_REF FIELD_DECLs TREE_THIS_VOLATILE
6661 to determine has_volatile_ops.
6663 2021-08-11 Eric Botcazou <ebotcazou@gcc.gnu.org>
6665 * cfgexpand.c (expand_used_vars): Reuse attribs local variable.
6667 2021-08-11 Jan Hubicka <hubicka@ucw.cz>
6668 Alexandre Oliva <oliva@adacore.com>
6670 * ipa-modref.c (modref_lattice::dump): Fix escape_point's min_flags
6672 (modref_lattice::merge_deref): Fix handling of indirect scape points.
6673 (update_escape_summary_1): Likewise.
6674 (update_escape_summary): Likewise.
6675 (ipa_merge_modref_summary_after_inlining): Likewise.
6677 2021-08-11 Richard Biener <rguenther@suse.de>
6679 PR middle-end/101858
6680 * fold-const.c (fold_binary_loc): Guard simplification
6681 of X < (cast) (1 << Y) to integer types.
6683 2021-08-11 Richard Biener <rguenther@suse.de>
6685 PR tree-optimization/101861
6686 * tree-vect-stmts.c (vectorizable_load): Fix error in
6687 previous change with regard to gather vectorization.
6689 2021-08-11 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
6692 * config/arm/arm_neon.h (vdup_n_s8): Replace call to builtin
6694 (vdup_n_s16): Likewise.
6695 (vdup_n_s32): Likewise.
6696 (vdup_n_s64): Likewise.
6697 (vdup_n_u8): Likewise.
6698 (vdup_n_u16): Likewise.
6699 (vdup_n_u32): Likewise.
6700 (vdup_n_u64): Likewise.
6701 (vdup_n_p8): Likewise.
6702 (vdup_n_p16): Likewise.
6703 (vdup_n_p64): Likewise.
6704 (vdup_n_f16): Likewise.
6705 (vdup_n_f32): Likewise.
6706 (vdupq_n_s8): Likewise.
6707 (vdupq_n_s16): Likewise.
6708 (vdupq_n_s32): Likewise.
6709 (vdupq_n_s64): Likewise.
6710 (vdupq_n_u8): Likewise.
6711 (vdupq_n_u16): Likewise.
6712 (vdupq_n_u32): Likewise.
6713 (vdupq_n_u64): Likewise.
6714 (vdupq_n_p8): Likewise.
6715 (vdupq_n_p16): Likewise.
6716 (vdupq_n_p64): Likewise.
6717 (vdupq_n_f16): Likewise.
6718 (vdupq_n_f32): Likewise.
6719 (vmov_n_s8): Replace call to builtin with call to corresponding
6721 (vmov_n_s16): Likewise.
6722 (vmov_n_s32): Likewise.
6723 (vmov_n_s64): Likewise.
6724 (vmov_n_u8): Likewise.
6725 (vmov_n_u16): Likewise.
6726 (vmov_n_u32): Likewise.
6727 (vmov_n_u64): Likewise.
6728 (vmov_n_p8): Likewise.
6729 (vmov_n_p16): Likewise.
6730 (vmov_n_f16): Likewise.
6731 (vmov_n_f32): Likewise.
6732 (vmovq_n_s8): Likewise.
6733 (vmovq_n_s16): Likewise.
6734 (vmovq_n_s32): Likewise.
6735 (vmovq_n_s64): Likewise.
6736 (vmovq_n_u8): Likewise.
6737 (vmovq_n_u16): Likewise.
6738 (vmovq_n_u32): Likewise.
6739 (vmovq_n_u64): Likewise.
6740 (vmovq_n_p8): Likewise.
6741 (vmovq_n_p16): Likewise.
6742 (vmovq_n_f16): Likewise.
6743 (vmovq_n_f32): Likewise.
6744 * config/arm/arm_neon_builtins.def: Remove entries for vdup_n.
6746 2021-08-11 liuhongt <hongtao.liu@intel.com>
6749 * config/i386/i386.md (ldexp<mode>3): Extend to vscalefs[sd]
6750 when TARGET_AVX512F and TARGET_SSE_MATH.
6752 2021-08-10 Jakub Jelinek <jakub@redhat.com>
6755 * config/i386/i386-expand.c (expand_vec_perm_even_odd): Return false
6756 for V32HImode if !TARGET_AVX512BW.
6757 (ix86_vectorize_vec_perm_const) <case E_V32HImode, case E_V64QImode>:
6758 If !TARGET_AVX512BW and TARGET_AVX512F and d.testing_p, don't fail
6759 early, but actually check the permutation.
6761 2021-08-10 Richard Biener <rguenther@suse.de>
6763 PR tree-optimization/101809
6764 * tree-vect-stmts.c (get_load_store_type): Allow emulated
6765 gathers with offset vector nunits being a constant multiple
6766 of the data vector nunits.
6767 (vect_get_gather_scatter_ops): Use the appropriate nunits
6768 for the offset vector defs.
6769 (vectorizable_store): Adjust call to
6770 vect_get_gather_scatter_ops.
6771 (vectorizable_load): Likewise. Handle the case of less
6772 offset vectors than data vectors.
6774 2021-08-10 Jakub Jelinek <jakub@redhat.com>
6777 * config/i386/sse.md (*avx512f_shuf_<shuffletype>64x2_1<mask_name>_1,
6778 *avx512f_shuf_<shuffletype>32x4_1<mask_name>_1): New define_insn
6781 2021-08-10 Richard Biener <rguenther@suse.de>
6783 PR tree-optimization/101801
6784 PR tree-optimization/101819
6785 * tree-vectorizer.h (vect_emulated_vector_p): Declare.
6786 * tree-vect-loop.c (vect_emulated_vector_p): New function.
6787 (vectorizable_reduction): Re-instantiate a check for emulated
6789 * tree-vect-stmts.c (vectorizable_shift): Likewise.
6790 (vectorizable_operation): Likewise. Cost emulated vector
6791 operations according to the scalar sequence synthesized by
6794 2021-08-10 Richard Biener <rguenther@suse.de>
6796 PR middle-end/101824
6797 * tree-nested.c (get_frame_field): Mark the COMPONENT_REF as
6798 volatile in case the variable was.
6800 2021-08-10 H.J. Lu <hjl.tools@gmail.com>
6803 * config/i386/constraints.md (BC): Document for integer SSE
6804 constant all bits set operand.
6805 (BF): New constraint for const floating-point all bits set
6807 * config/i386/i386.c (standard_sse_constant_p): Likewise.
6808 (standard_sse_constant_opcode): Likewise.
6809 * config/i386/sse.md (sseconstm1): New mode attribute.
6810 (mov<mode>_internal): Replace BC with <sseconstm1>.
6812 2021-08-10 liuhongt <hongtao.liu@intel.com>
6814 * config/i386/sse.md (cond_<insn><mode>): New expander.
6815 (VI248_AVX512VLBW): New mode iterator.
6816 * config/i386/predicates.md
6817 (nonimmediate_or_const_vec_dup_operand): New predicate.
6819 2021-08-09 Andrew MacLeod <amacleod@redhat.com>
6821 PR tree-optimization/101741
6822 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Check
6823 type of parameter for toupper/tolower.
6825 2021-08-09 Martin Jambor <mjambor@suse.cz>
6828 * ipa-prop.c (propagate_controlled_uses): Removed a spurious space.
6830 2021-08-09 Pat Haugen <pthaugen@linux.ibm.com>
6832 * config/rs6000/rs6000.c (is_load_insn1): Verify destination is a
6834 (is_store_insn1): Verify source is a register.
6836 2021-08-09 Uroš Bizjak <ubizjak@gmail.com>
6839 * config/i386/mmx.md (<any_logic:code>v2sf3):
6840 Rename from *mmx_<any_logic:code>v2sf3
6842 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6844 * config/nvptx/nvptx.c: Cross-reference parts adapted in
6845 'gcc/omp-oacc-neuter-broadcast.cc'.
6846 * omp-low.c: Likewise.
6847 * omp-oacc-neuter-broadcast.cc: Cross-reference parts adapted from
6850 2021-08-09 Julian Brown <julian@codesourcery.com>
6851 Kwok Cheung Yeung <kcy@codesourcery.com>
6852 Thomas Schwinge <thomas@codesourcery.com>
6854 * config/gcn/gcn.c (gcn_init_builtins): Override decls for
6855 BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START,
6856 BUILT_IN_GOACC_SINGLE_COPY_END and BUILT_IN_GOACC_BARRIER.
6857 (gcn_goacc_validate_dims): Turn on worker partitioning unconditionally.
6858 (gcn_fork_join): Update comment.
6859 * config/gcn/gcn.opt (flag_worker_partitioning): Remove.
6860 (macc_experimental_workers): Remove unused option.
6862 2021-08-09 Julian Brown <julian@codesourcery.com>
6863 Nathan Sidwell <nathan@codesourcery.com> (via 'gcc/config/nvptx/nvptx.c' master)
6864 Kwok Cheung Yeung <kcy@codesourcery.com>
6865 Thomas Schwinge <thomas@codesourcery.com>
6867 * Makefile.in (OBJS): Add omp-oacc-neuter-broadcast.o.
6868 * doc/tm.texi.in (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD):
6869 Add documentation hook.
6870 * doc/tm.texi: Regenerate.
6871 * omp-oacc-neuter-broadcast.cc: New file.
6872 * omp-builtins.def (BUILT_IN_GOACC_BARRIER)
6873 (BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START)
6874 (BUILT_IN_GOACC_SINGLE_COPY_END): New builtins.
6875 * passes.def (pass_omp_oacc_neuter_broadcast): Add pass.
6876 * target.def (goacc.create_worker_broadcast_record): Add target
6878 * tree-pass.h (make_pass_omp_oacc_neuter_broadcast): Add
6880 * config/gcn/gcn-protos.h (gcn_goacc_adjust_propagation_record):
6881 Rename prototype to...
6882 (gcn_goacc_create_worker_broadcast_record): ... this.
6883 * config/gcn/gcn-tree.c (gcn_goacc_adjust_propagation_record): Rename
6885 (gcn_goacc_create_worker_broadcast_record): ... this.
6886 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_PROPAGATION_RECORD):
6888 (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD): ... this.
6890 2021-08-09 Tejas Belagod <tejas.belagod@arm.com>
6893 * config/aarch64/aarch64-simd.md (vlshr<mode>3, vashr<mode>3): Use
6896 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6898 * Makefile.in (GTFILES): Remove '$(srcdir)/omp-offload.c'.
6900 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6902 * builtins.def (DEF_GOACC_BUILTIN, DEF_GOMP_BUILTIN): Don't
6903 consider '-foffload-abi'.
6904 * common.opt (-foffload-abi): Remove 'Var', 'Init'.
6905 * opts.c (common_handle_option) <-foffload-abi> [ACCEL_COMPILER]:
6908 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6910 * optc-gen.awk: Sanity check that 'Init' doesn't appear without
6913 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6915 * omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE): Remove.
6917 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6919 * doc/gty.texi (Files): Update.
6921 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
6923 * doc/gty.texi (Files): Fix GTY header file example.
6925 2021-08-09 Roger Sayle <roger@nextmovesoftware.com>
6927 * tree-ssa-ccp.c (value_mask_to_min_max): Helper function to
6928 determine the upper and lower bounds from a mask-value pair.
6929 (bit_value_unop) [ABS_EXPR, ABSU_EXPR]: Add support for
6930 absolute value and unsigned absolute value expressions.
6931 (bit_value_binop): Initialize *VAL's precision.
6932 [LT_EXPR, LE_EXPR]: Use value_mask_to_min_max to determine
6933 upper and lower bounds of operands. Add LE_EXPR/GE_EXPR
6934 support when the operands are unknown but potentially equal.
6935 [MIN_EXPR, MAX_EXPR]: Support minimum/maximum expressions.
6937 2021-08-09 Bin Cheng <bin.cheng@linux.alibaba.com>
6939 * config/aarch64/aarch64.md
6940 (*extend<SHORT:mode><GPI:mode>2_aarch64): Use %<GPI:w>0.
6942 2021-08-08 Sergei Trofimovich <siarheit@google.com>
6944 * lra-constraints.c: Fix s/otput/output/ typo.
6946 2021-08-06 Martin Sebor <msebor@redhat.com>
6948 * builtins.c (expand_builtin_memchr): Move to gimple-ssa-warn-access.cc.
6949 (expand_builtin_strcat): Same.
6950 (expand_builtin_stpncpy): Same.
6951 (expand_builtin_strncat): Same.
6952 (check_read_access): Same.
6953 (check_memop_access): Same.
6954 (expand_builtin_strlen): Move checks to gimple-ssa-warn-access.cc.
6955 (expand_builtin_strnlen): Same.
6956 (expand_builtin_memcpy): Same.
6957 (expand_builtin_memmove): Same.
6958 (expand_builtin_mempcpy): Same.
6959 (expand_builtin_strcpy): Same.
6960 (expand_builtin_strcpy_args): Same.
6961 (expand_builtin_stpcpy_1): Same.
6962 (expand_builtin_strncpy): Same.
6963 (expand_builtin_memset): Same.
6964 (expand_builtin_bzero): Same.
6965 (expand_builtin_strcmp): Same.
6966 (expand_builtin_strncmp): Same.
6967 (expand_builtin): Remove handlers.
6968 (fold_builtin_strlen): Add a comment.
6969 * builtins.h (check_access): Move to gimple-ssa-warn-access.cc.
6970 * calls.c (maybe_warn_nonstring_arg): Same.
6971 * diagnostic-spec.c (nowarn_spec_t::nowarn_spec_t): Add warning option.
6972 * gimple-fold.c (gimple_fold_builtin_strcpy): Pass argument to callee.
6973 (gimple_fold_builtin_stpcpy): Same.
6974 * gimple-ssa-warn-access.cc (has_location): New function.
6975 (get_location): Same.
6976 (get_callee_fndecl): Same.
6979 (warn_string_no_nul): Define.
6980 (unterminated_array): Same.
6981 (check_nul_terminated_array): Same.
6982 (maybe_warn_nonstring_arg): Same.
6983 (maybe_warn_for_bound): Same.
6984 (warn_for_access): Same.
6985 (check_access): Same.
6986 (check_memop_access): Same.
6987 (check_read_access): Same.
6988 (warn_dealloc_offset): Use helper functions.
6989 (maybe_emit_free_warning): Same.
6990 (class pass_waccess): Add members.
6991 (check_strcat): New function.
6992 (check_strncat): New function.
6993 (check_stxcpy): New function.
6994 (check_stxncpy): New function.
6995 (check_strncmp): New function.
6996 (pass_waccess::check_builtin): New function.
6997 (pass_waccess::check): Call it.
6998 * gimple-ssa-warn-access.h (warn_string_no_nul): Move here from
7000 (maybe_warn_for_bound): Same.
7001 (check_access): Same.
7002 (check_memop_access): Same.
7003 (check_read_access): Same.
7004 * pointer-query.h (struct access_data): Define a ctor overload.
7006 2021-08-06 Richard Biener <rguenther@suse.de>
7008 PR tree-optimization/101801
7009 * tree-vectorizer.h (vect_worthwhile_without_simd_p): Rename...
7010 (vect_can_vectorize_without_simd_p): ... to this.
7011 * tree-vect-loop.c (vect_worthwhile_without_simd_p): Rename...
7012 (vect_can_vectorize_without_simd_p): ... to this and fold
7013 in vect_min_worthwhile_factor.
7014 (vect_min_worthwhile_factor): Remove.
7015 (vectorizable_reduction): Adjust and remove the cost part.
7016 * tree-vect-stmts.c (vectorizable_shift): Likewise.
7017 (vectorizable_operation): Likewise.
7019 2021-08-06 Uroš Bizjak <ubizjak@gmail.com>
7022 * config/i386/i386.md (cmove reg-to-reg move elimination peephole2s):
7023 Add general_gr_operand predicate to operand 3.
7025 2021-08-06 Roger Sayle <roger@nextmovesoftware.com>
7027 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern): Use
7028 CFN_BUILT_IN_CLRSB* instead of BUILT_IN_CLRSB* for consistency.
7030 2021-08-06 Tamar Christina <tamar.christina@arm.com>
7032 * config/aarch64/aarch64-sve-builtins.cc (register_svpattern,
7033 register_svprfop): Pass vec<> by pointer.
7034 * langhooks-def.h (lhd_simulate_enum_decl): Likewise.
7035 * langhooks.c (lhd_simulate_enum_decl): Likewise.
7036 * langhooks.h (struct lang_hooks_for_types): Likewise.
7038 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
7040 * config/aarch64/arm_neon.h (vst1_bf16_x2): Use
7041 __builtin_memcpy instead of constructing an additional
7042 __builtin_aarch64_simd_oi one vector at a time.
7043 (vst1q_bf16_x2): Likewise.
7044 (vst1_bf16_x3): Use __builtin_memcpy instead of constructing
7045 an additional __builtin_aarch64_simd_ci one vector at a time.
7046 (vst1q_bf16_x3): Likewise.
7047 (vst1_bf16_x4): Use __builtin_memcpy instead of a union.
7048 (vst1q_bf16_x4): Likewise.
7049 (vst2_bf16): Use __builtin_memcpy instead of constructing an
7050 additional __builtin_aarch64_simd_oi one vector at a time.
7051 (vst2q_bf16): Likewise.
7052 (vst3_bf16): Use __builtin_memcpy instead of constructing an
7053 additional __builtin_aarch64_simd_ci mode one vector at a
7055 (vst3q_bf16): Likewise.
7056 (vst4_bf16): Use __builtin_memcpy instead of constructing an
7057 additional __builtin_aarch64_simd_xi one vector at a time.
7058 (vst4q_bf16): Likewise.
7060 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
7062 * config/aarch64/arm_neon.h (__ST2_LANE_FUNC): Delete.
7063 (__ST2Q_LANE_FUNC): Delete.
7064 (vst2_lane_f16): Use __builtin_memcpy to copy vector
7065 structure instead of constructing __builtin_aarch64_simd_oi
7066 one vector at a time.
7067 (vst2_lane_f32): Likewise.
7068 (vst2_lane_f64): Likewise.
7069 (vst2_lane_p8): Likewise.
7070 (vst2_lane_p16): Likewise.
7071 (vst2_lane_p64): Likewise.
7072 (vst2_lane_s8): Likewise.
7073 (vst2_lane_s16): Likewise.
7074 (vst2_lane_s32): Likewise.
7075 (vst2_lane_s64): Likewise.
7076 (vst2_lane_u8): Likewise.
7077 (vst2_lane_u16): Likewise.
7078 (vst2_lane_u32): Likewise.
7079 (vst2_lane_u64): Likewise.
7080 (vst2_lane_bf16): Likewise.
7081 (vst2q_lane_f16): Use __builtin_memcpy to copy vector
7082 structure instead of using a union.
7083 (vst2q_lane_f32): Likewise.
7084 (vst2q_lane_f64): Likewise.
7085 (vst2q_lane_p8): Likewise.
7086 (vst2q_lane_p16): Likewise.
7087 (vst2q_lane_p64): Likewise.
7088 (vst2q_lane_s8): Likewise.
7089 (vst2q_lane_s16): Likewise.
7090 (vst2q_lane_s32): Likewise.
7091 (vst2q_lane_s64): Likewise.
7092 (vst2q_lane_u8): Likewise.
7093 (vst2q_lane_u16): Likewise.
7094 (vst2q_lane_u32): Likewise.
7095 (vst2q_lane_u64): Likewise.
7096 (vst2q_lane_bf16): Likewise.
7098 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
7100 * config/aarch64/arm_neon.h (__ST3_LANE_FUNC): Delete.
7101 (__ST3Q_LANE_FUNC): Delete.
7102 (vst3_lane_f16): Use __builtin_memcpy to copy vector
7103 structure instead of constructing __builtin_aarch64_simd_ci
7104 one vector at a time.
7105 (vst3_lane_f32): Likewise.
7106 (vst3_lane_f64): Likewise.
7107 (vst3_lane_p8): Likewise.
7108 (vst3_lane_p16): Likewise.
7109 (vst3_lane_p64): Likewise.
7110 (vst3_lane_s8): Likewise.
7111 (vst3_lane_s16): Likewise.
7112 (vst3_lane_s32): Likewise.
7113 (vst3_lane_s64): Likewise.
7114 (vst3_lane_u8): Likewise.
7115 (vst3_lane_u16): Likewise.
7116 (vst3_lane_u32): Likewise.
7117 (vst3_lane_u64): Likewise.
7118 (vst3_lane_bf16): Likewise.
7119 (vst3q_lane_f16): Use __builtin_memcpy to copy vector
7120 structure instead of using a union.
7121 (vst3q_lane_f32): Likewise.
7122 (vst3q_lane_f64): Likewise.
7123 (vst3q_lane_p8): Likewise.
7124 (vst3q_lane_p16): Likewise.
7125 (vst3q_lane_p64): Likewise.
7126 (vst3q_lane_s8): Likewise.
7127 (vst3q_lane_s16): Likewise.
7128 (vst3q_lane_s32): Likewise.
7129 (vst3q_lane_s64): Likewise.
7130 (vst3q_lane_u8): Likewise.
7131 (vst3q_lane_u16): Likewise.
7132 (vst3q_lane_u32): Likewise.
7133 (vst3q_lane_u64): Likewise.
7134 (vst3q_lane_bf16): Likewise.
7136 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
7138 * config/aarch64/arm_neon.h (__ST4_LANE_FUNC): Delete.
7139 (__ST4Q_LANE_FUNC): Delete.
7140 (vst4_lane_f16): Use __builtin_memcpy to copy vector
7141 structure instead of constructing __builtin_aarch64_simd_xi
7142 one vector at a time.
7143 (vst4_lane_f32): Likewise.
7144 (vst4_lane_f64): Likewise.
7145 (vst4_lane_p8): Likewise.
7146 (vst4_lane_p16): Likewise.
7147 (vst4_lane_p64): Likewise.
7148 (vst4_lane_s8): Likewise.
7149 (vst4_lane_s16): Likewise.
7150 (vst4_lane_s32): Likewise.
7151 (vst4_lane_s64): Likewise.
7152 (vst4_lane_u8): Likewise.
7153 (vst4_lane_u16): Likewise.
7154 (vst4_lane_u32): Likewise.
7155 (vst4_lane_u64): Likewise.
7156 (vst4_lane_bf16): Likewise.
7157 (vst4q_lane_f16): Use __builtin_memcpy to copy vector
7158 structure instead of using a union.
7159 (vst4q_lane_f32): Likewise.
7160 (vst4q_lane_f64): Likewise.
7161 (vst4q_lane_p8): Likewise.
7162 (vst4q_lane_p16): Likewise.
7163 (vst4q_lane_p64): Likewise.
7164 (vst4q_lane_s8): Likewise.
7165 (vst4q_lane_s16): Likewise.
7166 (vst4q_lane_s32): Likewise.
7167 (vst4q_lane_s64): Likewise.
7168 (vst4q_lane_u8): Likewise.
7169 (vst4q_lane_u16): Likewise.
7170 (vst4q_lane_u32): Likewise.
7171 (vst4q_lane_u64): Likewise.
7172 (vst4q_lane_bf16): Likewise.
7174 2021-08-06 Martin Liska <mliska@suse.cz>
7176 * config/rs6000/rs6000.c (rs6000_option_override_internal): When
7177 a target option is restored, it can have
7178 rs6000_long_double_type_size set to FLOAT_PRECISION_TFmode
7179 and error should not be emitted.
7181 2021-08-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
7183 * gcov-io.h (gcov_write): Declare.
7184 * gcov-io.c (gcov_write): New.
7185 (gcov_write_counter): Remove.
7186 (gcov_write_tag_length): Likewise.
7187 (gcov_write_summary): Replace gcov_write_tag_length() with calls to
7188 gcov_write_unsigned().
7189 * doc/invoke.texi (fprofile-info-section): Mention
7190 __gcov_info_to_gdca().
7192 2021-08-06 Martin Sebor <msebor@redhat.com>
7194 * dominance.c (prune_bbs_to_update_dominators): Adjust by-value vec
7195 arguments to by-reference.
7196 (iterate_fix_dominators): Same.
7197 * dominance.h (iterate_fix_dominators): Same.
7198 * ipa-prop.h: Call auto_vec::to_vec_legacy.
7199 * tree-data-ref.c (dump_data_dependence_relation): Adjust by-value vec
7200 arguments to by-reference.
7201 (debug_data_dependence_relation): Same.
7202 (dump_data_dependence_relations): Same.
7203 * tree-data-ref.h (debug_data_dependence_relation): Same.
7204 (dump_data_dependence_relations): Same.
7205 * tree-predcom.c (dump_chains): Same.
7206 (initialize_root_vars_lm): Same.
7207 (determine_unroll_factor): Same.
7208 (replace_phis_by_defined_names): Same.
7209 (insert_init_seqs): Same.
7210 (pcom_worker::tree_predictive_commoning_loop): Call
7211 auto_vec::to_vec_legacy.
7212 * tree-ssa-pre.c (insert_into_preds_of_block): Adjust by-value vec
7213 arguments to by-reference.
7214 * tree-ssa-threadbackward.c (populate_worklist): Same.
7215 (back_threader::resolve_def): Same.
7216 * tree-vect-data-refs.c (vect_check_nonzero_value): Same.
7217 (vect_enhance_data_refs_alignment): Same.
7218 (vect_check_lower_bound): Same.
7219 (vect_prune_runtime_alias_test_list): Same.
7220 (vect_permute_store_chain): Same.
7221 * tree-vect-slp-patterns.c (vect_normalize_conj_loc): Same.
7222 * tree-vect-stmts.c (vect_create_vectorized_demotion_stmts): Same.
7223 * tree-vectorizer.h (vect_permute_store_chain): Same.
7224 * vec.c (test_init): New function.
7225 (vec_c_tests): Call new function.
7226 * vec.h (vec): Declare ctors, dtor, and assignment.
7227 (auto_vec::vec_to_legacy): New function.
7228 (vec::copy): Adjust initialization.
7230 2021-08-05 H.J. Lu <hjl.tools@gmail.com>
7233 * config/i386/i386.c (ix86_can_inline_p): Ignore MASK_80387 if
7234 callee only uses GPRs.
7235 * config/i386/ia32intrin.h: Revert commit 5463cee2770.
7236 * config/i386/serializeintrin.h: Revert commit 71958f740f1.
7237 * config/i386/x86gprintrin.h: Add
7238 #pragma GCC target("general-regs-only") and #pragma GCC pop_options
7239 to disable non-GPR ISAs.
7241 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
7243 PR middle-end/101787
7244 * doc/md.texi (cond_ashl, cond_ashr, cond_lshr): Document.
7246 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
7248 * tree-vectorizer.h (vect_is_store_elt_extraction, vect_is_reduction)
7249 (vect_reduc_type, vect_embedded_comparison_type, vect_comparison_type)
7250 (vect_is_extending_load, vect_is_integer_truncation): New functions,
7251 moved from aarch64.c but given different names.
7252 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction)
7253 (aarch64_is_reduction, aarch64_reduc_type)
7254 (aarch64_embedded_comparison_type, aarch64_comparison_type)
7255 (aarch64_extending_load_p, aarch64_integer_truncation_p): Delete
7256 in favor of the above. Update callers accordingly.
7258 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
7261 * config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
7262 writing .cpu directive in asm output.
7263 * config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
7264 (arm_last_printed_arch_string): Delete.
7265 (arm_last-printed_fpu_string): Delete.
7266 (arm_configure_build_target): If use of floating-point/SIMD is
7267 disabled, remove all fp/simd related features from the target ISA.
7268 (last_arm_targ_options): New variable.
7269 (arm_print_asm_arch_directives): Add new parameters. Change order
7270 of emitted directives and handle all cases here.
7271 (arm_file_start): Always call arm_print_asm_arch_directives, move
7272 all generation of .arch/.arch_extension here.
7273 (arm_file_end): Call arm_print_asm_arch.
7274 (arm_declare_function_name): Call arm_print_asm_arch_directives
7275 instead of printing .arch/.fpu directives directly.
7277 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
7279 * config/arm/arm.c (arm_configure_build_target): Don't call
7280 arm_option_reconfigure_globals.
7281 (arm_option_restore): Call arm_option_reconfigure_globals after
7282 reconfiguring the target.
7283 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
7285 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
7287 * config/arm/arm.c (arm_configure_build_target): Ensure the target's
7288 arch_name is always set.
7290 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
7292 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
7293 of vec_select high-half from being added into Neon subtract
7296 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
7298 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
7299 of vec_select high-half from being added into Neon add cost.
7301 2021-08-05 Kewen Lin <linkw@linux.ibm.com>
7303 * cfgloop.h (loops_list::loops_list): Add one optional argument
7304 root and adjust accordingly, update loop tree walking and factor
7306 * cfgloop.c (loops_list::walk_loop_tree): ... this. New function.
7308 2021-08-05 Eric Botcazou <ebotcazou@gcc.gnu.org>
7310 PR tree-optimization/101626
7311 * tree-sra.c (propagate_subaccesses_from_rhs): Do not set the
7312 reverse scalar storage order on a pointer or vector component.
7314 2021-08-05 liuhongt <hongtao.liu@intel.com>
7316 * config/i386/sse.md (cond_<code><mode>): New expander.
7318 2021-08-05 liuhongt <hongtao.liu@intel.com>
7320 * config/i386/sse.md (cond_<code><mode>): New expander.
7322 2021-08-05 liuhongt <hongtao.liu@intel.com>
7324 * config/i386/sse.md (cond_<code><mode>): New expander.
7326 2021-08-04 David Malcolm <dmalcolm@redhat.com>
7329 * Makefile.in (ANALYZER_OBJS): Add analyzer/region-model-asm.o.
7331 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
7334 * config/i386/i386.h (STORE_MAX_PIECES): Allow 16/32/64 bytes
7335 only if TARGET_INTER_UNIT_MOVES_TO_VEC is true.
7337 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
7340 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
7341 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
7342 data with SSE register from one memory location to another.
7344 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7346 * config/s390/s390.c (expand_perm_with_vpdi): New function.
7347 (vectorize_vec_perm_const_1): Call expand_perm_with_vpdi.
7348 * config/s390/vector.md (*vpdi1<mode>, @vpdi1<mode>): Enable a
7349 parameterized expander.
7350 (*vpdi4<mode>, @vpdi4<mode>): Likewise.
7352 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7354 * config/s390/s390.c (MAX_VECT_LEN): Define macro.
7355 (struct expand_vec_perm_d): Define struct.
7356 (expand_perm_with_merge): New function.
7357 (vectorize_vec_perm_const_1): New function.
7358 (s390_vectorize_vec_perm_const): New function.
7359 (TARGET_VECTORIZE_VEC_PERM_CONST): Define target macro.
7361 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7363 * config/s390/vector.md (V_HW_64): Remove mode iterator.
7364 (*vec_load_pair<mode>): Use V_HW_2 instead of V_HW_64.
7365 * config/s390/vx-builtins.md
7366 (vec_scatter_element<V_HW_2:mode>_SI): Use V_HW_2 instead of
7369 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7371 * config/s390/s390.md (UNSPEC_VEC_PERMI): Remove constant
7373 * config/s390/vector.md (*vpdi1<mode>, *vpdi4<mode>): New pattern
7375 * config/s390/vx-builtins.md (*vec_permi<mode>): Emit generic rtx
7376 instead of an unspec.
7378 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
7380 * config/s390/s390-modes.def: Add more vector modes to support
7381 concatenation of two vectors.
7382 * config/s390/s390-protos.h (s390_expand_merge_perm_const): Add
7384 (s390_expand_merge): Likewise.
7385 * config/s390/s390.c (s390_expand_merge_perm_const): New function.
7386 (s390_expand_merge): New function.
7387 * config/s390/s390.md (UNSPEC_VEC_MERGEH, UNSPEC_VEC_MERGEL):
7388 Remove constant definitions.
7389 * config/s390/vector.md (V_HW_2): Add mode iterators.
7390 (VI_HW_4, V_HW_4): Rename VI_HW_4 to V_HW_4.
7391 (vec_2x_nelts, vec_2x_wide): New mode attributes.
7392 (*vmrhb, *vmrlb, *vmrhh, *vmrlh, *vmrhf, *vmrlf, *vmrhg, *vmrlg):
7393 New pattern definitions.
7394 (vec_widen_umult_lo_<mode>, vec_widen_umult_hi_<mode>)
7395 (vec_widen_smult_lo_<mode>, vec_widen_smult_hi_<mode>)
7396 (vec_unpacks_lo_v4sf, vec_unpacks_hi_v4sf, vec_unpacks_lo_v2df)
7397 (vec_unpacks_hi_v2df): Adjust expanders to emit non-unspec RTX for
7399 * config/s390/vx-builtins.md (V_HW_4): Remove mode iterator. Now
7401 (vec_mergeh<mode>, vec_mergel<mode>): Use s390_expand_merge to
7402 emit vec merge pattern.
7404 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
7406 * config/aarch64/aarch64.c (aarch64_strip_extend_vec_half):
7408 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent cost of
7409 vec_select high-half from being added into Neon multiply
7411 * rtlanal.c (vec_series_highpart_p): Define.
7412 * rtlanal.h (vec_series_highpart_p): Declare.
7414 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
7416 * config/aarch64/aarch64.c (aarch64_strip_duplicate_vec_elt):
7418 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent
7419 vec_select cost from being added into Neon multiply cost.
7421 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
7423 * tree-vect-loop.c (vect_better_loop_vinfo_p): Detect cases in
7424 which old_loop_vinfo is an epilogue loop that handles a constant
7425 number of iterations.
7427 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
7429 * tree-vect-loop.c (vect_analyze_loop): Print a dump message
7430 when a reanalyzed loop fails to be cheaper than the current
7433 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
7435 * config/aarch64/aarch64.c: Fix a typo.
7437 2021-08-04 Vincent Lefèvre <vincent-gcc@vinc17.net>
7439 PR gcov-profile/101773
7440 * gcov-io.c (gcov_close): Check return code of a fclose.
7442 2021-08-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
7445 * dwarf2out.c (dwarf2out_assembly_start): Emit a dummy
7446 .file statement when needed.
7448 2021-08-04 Richard Biener <rguenther@suse.de>
7450 * tree-vect-data-refs.c (vect_check_gather_scatter):
7451 Include widening conversions only when the result is
7452 still handed by native gather or the current offset
7453 size not already matches the data size.
7454 Also succeed analysis in case there's no native support,
7455 noted by a IFN_LAST ifn and a NULL decl.
7456 (vect_analyze_data_refs): Always consider gathers.
7457 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
7458 Test for no IFN gather rather than decl gather.
7459 * tree-vect-stmts.c (vect_model_load_cost): Pass in the
7460 gather-scatter info and cost emulated gathers accordingly.
7461 (vect_truncate_gather_scatter_offset): Properly test for
7463 (vect_use_strided_gather_scatters_p): Likewise.
7464 (get_load_store_type): Handle emulated gathers and its
7466 (vectorizable_load): Likewise. Emulate them by extracting
7467 scalar offsets, doing scalar loads and a vector construct.
7469 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
7472 * expr.c (op_by_pieces_d::op_by_pieces_d): Add a max_pieces
7473 argument to set m_max_size.
7474 (move_by_pieces_d): Pass MOVE_MAX_PIECES to op_by_pieces_d.
7475 (store_by_pieces_d): Pass STORE_MAX_PIECES to op_by_pieces_d.
7476 (compare_by_pieces_d): Pass COMPARE_MAX_PIECES to op_by_pieces_d.
7478 2021-08-04 Roger Sayle <roger@nextmovesoftware.com>
7479 Marc Glisse <marc.glisse@inria.fr>
7481 * match.pd (bit_ior, bit_xor): Canonicalize (X*C1)|(X*C2) and
7482 (X*C1)^(X*C2) as X*(C1+C2), and related variants, using
7483 tree_nonzero_bits to ensure that operands are bit-wise disjoint.
7485 2021-08-04 Richard Biener <rguenther@suse.de>
7487 * tree-ssa-forwprop.c (pass_forwprop::execute): Split
7488 out code to decompose vector loads ...
7489 (optimize_vector_load): ... here. Generalize it to
7490 handle intermediate widening and TARGET_MEM_REF loads
7491 and apply it to loads with a supported vector mode as well.
7493 2021-08-04 Richard Biener <rguenther@suse.de>
7495 PR tree-optimization/101756
7496 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Make sure
7497 the result of the reduction epilogue is compatible to the original
7500 2021-08-04 liuhongt <hongtao.liu@intel.com>
7503 * config/i386/i386.md (peephole2): Refine predicate from
7504 register_operand to general_reg_operand.
7506 2021-08-04 Aldy Hernandez <aldyh@redhat.com>
7508 * gimple-range-path.h (path_range_query::dump): Mark override.
7510 2021-08-04 Richard Biener <rguenther@suse.de>
7512 PR tree-optimization/101769
7513 * tree-tailcall.c (eliminate_tail_call): Add the created loop
7514 for the first recursion and return it via the new output parameter.
7515 (optimize_tail_call): Pass through new output param.
7516 (tree_optimize_tail_calls_1): After creating all latches,
7517 add the created loop to the loop tree. Do not mark loops for fixup.
7519 2021-08-04 Martin Liska <mliska@suse.cz>
7521 * doc/invoke.texi: Document threader-mode param.
7523 2021-08-04 liuhongt <hongtao.liu@intel.com>
7525 * config/i386/sse.md (cond_fma<mode>): New expander.
7526 (cond_fms<mode>): Ditto.
7527 (cond_fnma<mode>): Ditto.
7528 (cond_fnms<mode>): Ditto.
7530 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
7532 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode>): Use && instead of &.
7534 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
7536 * config/rs6000/constraints.md: Remove "e" from the list of available
7537 constraint characters.
7539 2021-08-03 Eugene Rozenfeld <erozen@microsoft.com>
7541 PR gcov-profile/71672
7542 * auto-profile.c (afdo_indirect_call): Fix setup of the historgram value for indirect calls.
7544 2021-08-03 Paul A. Clarke <pc@us.ibm.com>
7546 * config/rs6000/smmintrin.h (_mm_minpos_epu16): New.
7548 2021-08-03 H.J. Lu <hjl.tools@gmail.com>
7550 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): In 64-bit mode,
7551 try XMM31 to avoid vzeroupper.
7553 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7555 * doc/invoke.texi: Document -mtune=neoverse-512tvb and
7556 -mcpu=neoverse-512tvb.
7557 * config/aarch64/aarch64-cores.def (neoverse-512tvb): New entry.
7558 * config/aarch64/aarch64-tune.md: Regenerate.
7559 * config/aarch64/aarch64.c (neoverse512tvb_sve_vector_cost)
7560 (neoverse512tvb_sve_issue_info, neoverse512tvb_vec_issue_info)
7561 (neoverse512tvb_vector_cost, neoverse512tvb_tunings): New structures.
7562 (aarch64_adjust_body_cost_sve): Handle -mtune=neoverse-512tvb.
7563 (aarch64_adjust_body_cost): Likewise.
7565 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7567 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Only
7568 record issue information for operations that occur in the
7571 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7573 * config/aarch64/aarch64.c (aarch64_multiply_add_p): Add a vec_flags
7574 parameter. Detect cases in which an Advanced SIMD MLA would almost
7575 certainly require a MOV.
7576 (aarch64_count_ops): Update accordingly.
7578 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7580 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction): New
7581 function, split out from...
7582 (aarch64_detect_vector_stmt_subtype): ...here.
7583 (aarch64_add_stmt_cost): Treat extracting element 0 as free.
7585 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7587 * config/aarch64/aarch64-protos.h (sve_vec_cost):
7588 Add gather_load_x32_cost and gather_load_x64_cost.
7589 * config/aarch64/aarch64.c (generic_sve_vector_cost)
7590 (a64fx_sve_vector_cost, neoversev1_sve_vector_cost): Update
7591 accordingly, using the values given by the scalar_load * number
7592 of elements calculation that we used previously.
7593 (aarch64_detect_vector_stmt_subtype): Use the new fields.
7595 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7597 * config/aarch64/aarch64.c (aarch64_adjust_body_cost_sve): New
7598 function, split out from...
7599 (aarch64_adjust_body_cost): ...here.
7601 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7603 * config/aarch64/fractional-cost.h: New file.
7604 * config/aarch64/aarch64.c: Include <algorithm> (indirectly)
7605 and cost_fraction.h.
7606 (vec_cost_fraction): New typedef.
7607 (aarch64_detect_scalar_stmt_subtype): Use it for statement costs.
7608 (aarch64_detect_vector_stmt_subtype): Likewise.
7609 (aarch64_sve_adjust_stmt_cost, aarch64_adjust_stmt_cost): Likewise.
7610 (aarch64_estimate_min_cycles_per_iter): Use vec_cost_fraction
7612 (aarch64_adjust_body_cost): Likewise.
7613 (aarch64_test_cost_fraction): New function.
7614 (aarch64_run_selftests): Call it.
7616 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
7618 * config/aarch64/aarch64-protos.h (tune_params::sve_width): Turn
7620 * config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Update
7622 (aarch64_estimated_poly_value): Likewise. Use the least significant
7623 set bit for the minimum and likely values. Use the most significant
7624 set bit for the maximum value.
7626 2021-08-03 liuhongt <hongtao.liu@intel.com>
7628 * config/i386/sse.md (cond_<insn><mode>): New expander.
7629 (cond_mul<mode>): Ditto.
7631 2021-08-03 Kewen Lin <linkw@linux.ibm.com>
7633 * tree-cfg.c (move_sese_region_to_fn): Fix typos on dloop.
7635 2021-08-03 liuhongt <hongtao.liu@intel.com>
7637 * config/i386/sse.md (cond_<insn><mode>):New expander.
7638 (cond_mul<mode>): Ditto.
7639 (cond_div<mode>): Ditto.
7641 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
7643 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Also
7644 check stack_realign_needed for stack realignment.
7645 (ix86_legitimate_constant_p): Always allow CONST_WIDE_INT smaller
7646 than the largest integer supported by vector register.
7647 * config/i386/i386.h (MAX_MOVE_MAX): New. Set to 64.
7648 (MOVE_MAX): Set to bytes of the largest integer supported by
7650 (STORE_MAX_PIECES): New.
7652 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
7654 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
7655 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
7656 data from one memory location to another.
7658 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
7661 * config/i386/i386.c (TARGET_GEN_MEMSET_SCRATCH_RTX): New.
7663 2021-08-02 Aldy Hernandez <aldyh@redhat.com>
7665 PR tree-optimization/101724
7666 * params.opt: Remove --param=threader-iterative.
7667 * tree-ssa-threadbackward.c (pass_thread_jumps::execute): Remove
7670 2021-08-02 Tom de Vries <tdevries@suse.de>
7672 PR middle-end/101665
7673 * doc/extend.texi (nonnull attribute): Improve documentation.
7675 2021-08-02 Andrew Pinski <apinski@marvell.com>
7677 PR rtl-optimization/101683
7678 * rtlanal.c (may_trap_p_1): Handle UNSIGNED_FIX.
7680 2021-08-02 Roger Sayle <roger@nextmovesoftware.com>
7682 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern):
7683 Renamed from cond_removal_in_popcount_clz_ctz_pattern.
7684 Add support for BSWAP, FFS, PARITY and CLRSB builtins.
7685 (tree_ssa_phiop_worker): Update call to function above.
7687 2021-08-01 H.J. Lu <hjl.tools@gmail.com>
7690 * config/i386/i386.md (bsr_rex64_1_zext): New.
7691 (combine splitter for constant - clzll): Replace gen_bsr_rex64_1
7692 with gen_bsr_rex64_1_zext.
7694 2021-07-31 Jakub Jelinek <jakub@redhat.com>
7697 * config/i386/i386.md (bsr_rex64_1, bsr_1, bsr_zext_1): New
7698 define_insn patterns.
7699 (*bsr_rex64_2, *bsr_2): New define_insn_and_split patterns.
7700 Add combine splitters for constant - clz.
7701 (clz<mode>2): Use a temporary pseudo for bsr result.
7703 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
7705 * config/rs6000/smmintrin.h (_mm_floor_pd, _mm_floor_ps,
7706 _mm_floor_sd, _mm_floor_ss): New.
7708 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
7710 * config/rs6000/smmintrin.h (_mm_ceil_pd, _mm_ceil_ps,
7711 _mm_ceil_sd, _mm_ceil_ss): New.
7713 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
7715 * config/rs6000/smmintrin.h (_mm_blend_pd, _mm_blendv_pd,
7716 _mm_blend_ps, _mm_blendv_ps): New.
7718 2021-07-30 Roger Sayle <roger@nextmovesoftware.com>
7719 Uroš Bizjak <ubizjak@gmail.com>
7721 * config/i386/i386.md (*dec_cmov<mode>): New define_insn_and_split
7722 to generate a conditional move using the carry flag after sub $1.
7723 (peephole2): Eliminate a register-to-register move by inverting
7724 the condition of a conditional move.
7726 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
7728 * config/mmix/mmix.md ("call", "call_value", "*call_real")
7729 ("*call_value_real"): Don't generate rtx mentioning the generic
7730 operands 1 and 2 to "call", and similarly for "call_value".
7731 * config/mmix/mmix.c (mmix_print_operand_punct_valid_p)
7732 (mmix_print_operand): Use '!' instead of 'p'.
7734 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
7736 * doc/md.texi (call): Correct information about operand 2.
7737 * config/mmix/mmix.md ("call", "call_value"): Remove fixed FIXMEs.
7739 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
7741 * range-op.cc (operator_trunc_mod::wi_fold): Fold constants.
7743 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
7745 * range-op.cc (operator_div::wi_fold): Return UNDEFINED for [0, 0] divisor.
7747 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
7749 * gimple-range-cache.cc (*::set_bb_range): Change const basic_block to
7751 (*::get_bb_range): Ditto.
7752 (*::bb_range_p): Ditto.
7753 * gimple-range-cache.h: Change prototypes.
7755 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
7758 * builtins.c (builtin_memcpy_read_str): Change the mode argument
7759 from scalar_int_mode to fixed_size_mode.
7760 (builtin_strncpy_read_str): Likewise.
7761 (gen_memset_value_from_prev): New function.
7762 (builtin_memset_read_str): Change the mode argument from
7763 scalar_int_mode to fixed_size_mode. Use gen_memset_value_from_prev
7764 and support CONST_VECTOR.
7765 (builtin_memset_gen_str): Likewise.
7766 (try_store_by_multiple_pieces): Use by_pieces_constfn to declare
7768 * builtins.h (builtin_strncpy_read_str): Replace scalar_int_mode
7769 with fixed_size_mode.
7770 (builtin_memset_read_str): Likewise.
7771 * expr.c (widest_int_mode_for_size): Renamed to ...
7772 (widest_fixed_size_mode_for_size): Add a bool argument to
7773 indicate if QI vector mode can be used.
7774 (by_pieces_ninsns): Call widest_fixed_size_mode_for_size
7775 instead of widest_int_mode_for_size.
7776 (pieces_addr::adjust): Change the mode argument from
7777 scalar_int_mode to fixed_size_mode.
7778 (op_by_pieces_d): Make m_len read-only. Add a bool member,
7779 m_qi_vector_mode, to indicate that QI vector mode can be used.
7780 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
7781 initialize m_qi_vector_mode. Call widest_fixed_size_mode_for_size
7782 instead of widest_int_mode_for_size.
7783 (op_by_pieces_d::get_usable_mode): Change the mode argument from
7784 scalar_int_mode to fixed_size_mode. Call
7785 widest_fixed_size_mode_for_size instead of
7786 widest_int_mode_for_size.
7787 (op_by_pieces_d::smallest_fixed_size_mode_for_size): New member
7788 function to return the smallest integer or QI vector mode.
7789 (op_by_pieces_d::run): Call widest_fixed_size_mode_for_size
7790 instead of widest_int_mode_for_size. Call
7791 smallest_fixed_size_mode_for_size instead of
7792 smallest_int_mode_for_size.
7793 (store_by_pieces_d::store_by_pieces_d): Add a bool argument to
7794 indicate that QI vector mode can be used and pass it to
7795 op_by_pieces_d::op_by_pieces_d.
7796 (can_store_by_pieces): Call widest_fixed_size_mode_for_size
7797 instead of widest_int_mode_for_size. Pass memsetp to
7798 widest_fixed_size_mode_for_size to support QI vector mode.
7799 Allow all CONST_VECTORs for memset if vec_duplicate is supported.
7800 (store_by_pieces): Pass memsetp to
7801 store_by_pieces_d::store_by_pieces_d.
7802 (clear_by_pieces_1): Removed.
7803 (clear_by_pieces): Replace clear_by_pieces_1 with
7804 builtin_memset_read_str and pass true to store_by_pieces_d to
7805 support vector mode broadcast.
7806 (string_cst_read_str): Change the mode argument from
7807 scalar_int_mode to fixed_size_mode.
7808 * expr.h (by_pieces_constfn): Change scalar_int_mode to
7810 (by_pieces_prev): Likewise.
7811 * rtl.h (lowpart_subreg_regno): New.
7812 * rtlanal.c (lowpart_subreg_regno): New. A wrapper around
7813 simplify_subreg_regno.
7814 * target.def (gen_memset_scratch_rtx): New hook.
7815 * doc/tm.texi.in: Add TARGET_GEN_MEMSET_SCRATCH_RTX.
7816 * doc/tm.texi: Regenerated.
7818 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
7821 * config/mips/mips.c (mips_atomic_assign_expand_fenv): Use
7822 TARGET_EXPR instead of MODIFY_EXPR.
7824 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
7827 * config/mips/mips-protos.h (mips_expand_vec_cmp_expr): Declare.
7828 * config/mips/mips.c (mips_expand_vec_cmp_expr): New function.
7829 * config/mips/mips-msa.md (vec_cmp<MSA:mode><mode_i>): New
7831 (vec_cmpu<IMSA:mode><mode_i>): New expander.
7833 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
7836 * config/i386/i386-options.c (ix86_option_override_internal):
7837 Don't enable LZCNT/POPCNT if they have been disabled explicitly.
7839 2021-07-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
7842 * config/arm/arm_neon.h (vld1_p64): Replace call to builtin by
7843 explicitly dereferencing __a.
7844 (vld1_s64): Likewise.
7845 (vld1_u64): Likewise.
7846 * config/arm/arm_neon_builtins.def (vld1): Remove entry for di
7847 and change to VAR13.
7849 2021-07-30 Aldy Hernandez <aldyh@redhat.com>
7851 * gimple-loop-versioning.cc (lv_dom_walker::lv_dom_walker): Remove
7852 use of m_range_analyzer.
7853 (loop_versioning::lv_dom_walker::before_dom_children): Same.
7854 (loop_versioning::lv_dom_walker::after_dom_children): Remove.
7855 (loop_versioning::prune_loop_conditions): Replace vr_values use
7856 with range_query interface.
7857 (pass_loop_versioning::execute): Use ranger.
7859 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
7862 * ipa-devirt.c (ipa_odr_read_section): Compare the precision of
7863 enum values, and emit a warning if they mismatch.
7865 2021-07-30 Kewen Lin <linkw@linux.ibm.com>
7867 * cfgloop.h (as_const): New function.
7868 (class loop_iterator): Rename to ...
7869 (class loops_list): ... this.
7870 (loop_iterator::next): Rename to ...
7871 (loops_list::Iter::fill_curr_loop): ... this and adjust.
7872 (loop_iterator::loop_iterator): Rename to ...
7873 (loops_list::loops_list): ... this and adjust.
7874 (loops_list::Iter): New class.
7875 (loops_list::iterator): New type.
7876 (loops_list::const_iterator): New type.
7877 (loops_list::begin): New function.
7878 (loops_list::end): Likewise.
7879 (loops_list::begin const): Likewise.
7880 (loops_list::end const): Likewise.
7881 (FOR_EACH_LOOP): Remove.
7882 (FOR_EACH_LOOP_FN): Remove.
7883 * cfgloop.c (flow_loops_dump): Adjust FOR_EACH_LOOP* with range-based
7884 for loop with loops_list instance.
7885 (sort_sibling_loops): Likewise.
7886 (disambiguate_loops_with_multiple_latches): Likewise.
7887 (verify_loop_structure): Likewise.
7888 * cfgloopmanip.c (create_preheaders): Likewise.
7889 (force_single_succ_latches): Likewise.
7890 * config/aarch64/falkor-tag-collision-avoidance.c
7891 (execute_tag_collision_avoidance): Likewise.
7892 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Likewise.
7893 * config/s390/s390.c (s390_adjust_loops): Likewise.
7894 * doc/loop.texi: Likewise.
7895 * gimple-loop-interchange.cc (pass_linterchange::execute): Likewise.
7896 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
7897 * gimple-loop-versioning.cc (loop_versioning::analyze_blocks): Likewise.
7898 (loop_versioning::make_versioning_decisions): Likewise.
7899 * gimple-ssa-split-paths.c (split_paths): Likewise.
7900 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
7901 * graphite.c (canonicalize_loop_form): Likewise.
7902 (graphite_transform_loops): Likewise.
7903 * ipa-fnsummary.c (analyze_function_body): Likewise.
7904 * ipa-pure-const.c (analyze_function): Likewise.
7905 * loop-doloop.c (doloop_optimize_loops): Likewise.
7906 * loop-init.c (loop_optimizer_finalize): Likewise.
7907 (fix_loop_structure): Likewise.
7908 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
7909 (move_loop_invariants): Likewise.
7910 * loop-unroll.c (decide_unrolling): Likewise.
7911 (unroll_loops): Likewise.
7912 * modulo-sched.c (sms_schedule): Likewise.
7913 * predict.c (predict_loops): Likewise.
7914 (pass_profile::execute): Likewise.
7915 * profile.c (branch_prob): Likewise.
7916 * sel-sched-ir.c (sel_finish_pipelining): Likewise.
7917 (sel_find_rgns): Likewise.
7918 * tree-cfg.c (replace_loop_annotate): Likewise.
7919 (replace_uses_by): Likewise.
7920 (move_sese_region_to_fn): Likewise.
7921 * tree-if-conv.c (pass_if_conversion::execute): Likewise.
7922 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
7923 * tree-parloops.c (parallelize_loops): Likewise.
7924 * tree-predcom.c (tree_predictive_commoning): Likewise.
7925 * tree-scalar-evolution.c (scev_initialize): Likewise.
7926 (scev_reset): Likewise.
7927 * tree-ssa-dce.c (find_obviously_necessary_stmts): Likewise.
7928 * tree-ssa-live.c (remove_unused_locals): Likewise.
7929 * tree-ssa-loop-ch.c (ch_base::copy_headers): Likewise.
7930 * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
7931 (tree_ssa_lim_initialize): Likewise.
7932 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Likewise.
7933 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize): Likewise.
7934 * tree-ssa-loop-manip.c (get_loops_exits): Likewise.
7935 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations): Likewise.
7936 (free_numbers_of_iterations_estimates): Likewise.
7937 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Likewise.
7938 * tree-ssa-loop-split.c (tree_ssa_split_loops): Likewise.
7939 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Likewise.
7940 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
7941 (pass_scev_cprop::execute): Likewise.
7942 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
7943 * tree-ssa-sccvn.c (do_rpo_vn): Likewise.
7944 * tree-ssa-threadupdate.c
7945 (jump_thread_path_registry::thread_through_all_blocks): Likewise.
7946 * tree-vectorizer.c (vectorize_loops): Likewise.
7947 * tree-vrp.c (vrp_asserts::find_assert_locations): Likewise.
7949 2021-07-29 Hans-Peter Nilsson <hp@bitrange.com>
7951 * config/mmix/mmix.c (mmix_function_arg_1): Avoid
7952 generating a VOIDmode register for e.g the
7953 function_arg_info::end_marker.
7955 2021-07-29 Jeff Law <jeffreyalaw@gmail.com>
7957 * config/h8300/h8300-modes.def: Add CCZ, CCV and CCC, drop CCZNV.
7958 * config/h8300/h8300.md (H8cc mode iterator): Add CCZ.
7959 (cc mode_attr): Similarly.
7960 (ccz subst_attr): Similarly.
7961 * config/h8300/jumpcall.md: Add new patterns for branch-on-bit.
7962 * config/h8300/testcompare.md: Remove various cc0 based patterns
7963 that had been commented out. Add pattern to set CCZ from a bit
7966 2021-07-29 Thomas Schwinge <thomas@codesourcery.com>
7967 Julian Brown <julian@codesourcery.com>
7968 Kwok Cheung Yeung <kcy@codesourcery.com>
7970 * omp-offload.c (oacc_loop_xform_head_tail, oacc_loop_process):
7971 'update_stmt' after modification.
7972 (pass_oacc_loop_designation): New function, extracted out of...
7973 (pass_oacc_device_lower): ... this.
7974 (pass_data_oacc_loop_designation, pass_oacc_loop_designation)
7975 (make_pass_oacc_loop_designation): New
7976 * passes.def: Add it.
7977 * tree-parloops.c (create_parallel_loop): Adjust.
7978 * tree-pass.h (make_pass_oacc_loop_designation): New.
7980 2021-07-29 Aldy Hernandez <aldyh@redhat.com>
7982 * flag-types.h (enum threader_mode): New.
7983 * params.opt: Add entry for --param=threader-mode.
7984 * tree-ssa-threadbackward.c (THREADER_ITERATIVE_MODE): New.
7985 (class back_threader): New.
7986 (back_threader::back_threader): New.
7987 (back_threader::~back_threader): New.
7988 (back_threader::maybe_register_path): New.
7989 (back_threader::find_taken_edge): New.
7990 (back_threader::find_taken_edge_switch): New.
7991 (back_threader::find_taken_edge_cond): New.
7992 (back_threader::resolve_def): New.
7993 (back_threader::resolve_phi): New.
7994 (back_threader::find_paths_to_names): New.
7995 (back_threader::find_paths): New.
7998 (thread_jumps::find_jump_threads_backwards): Call ranger threader.
7999 (thread_jumps::find_jump_threads_backwards_with_ranger): New.
8000 (pass_thread_jumps::execute): Abstract out code...
8001 (try_thread_blocks): ...here.
8002 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
8003 Abstract out threading candidate code to...
8004 (single_succ_to_potentially_threadable_block): ...here.
8005 * tree-ssa-threadedge.h (single_succ_to_potentially_threadable_block):
8007 * tree-ssa-threadupdate.c (register_jump_thread): Return boolean.
8008 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
8009 Return bool from register_jump_thread.
8011 2021-07-29 Andreas Krebbel <krebbel@linux.ibm.com>
8013 * target.def: in0 and in1 do not need to be registers.
8014 * doc/tm.texi: Regenerate.
8016 2021-07-29 liuhongt <hongtao.liu@intel.com>
8019 * config/i386/i386.c (ix86_widen_mult_cost): New function.
8020 (ix86_add_stmt_cost): Use ix86_widen_mult_cost for
8023 2021-07-29 Jiufu Guo <guojiufu@linux.ibm.com>
8026 * config/rs6000/rs6000.c (TARGET_PREFERRED_DOLOOP_MODE): New hook.
8027 (rs6000_preferred_doloop_mode): New hook.
8028 * doc/tm.texi: Regenerate.
8029 * doc/tm.texi.in: Add hook preferred_doloop_mode.
8030 * target.def (preferred_doloop_mode): New hook.
8031 * targhooks.c (default_preferred_doloop_mode): New hook.
8032 * targhooks.h (default_preferred_doloop_mode): New hook.
8033 * tree-ssa-loop-ivopts.c (compute_doloop_base_on_mode): New function.
8034 (add_iv_candidate_for_doloop): Call targetm.preferred_doloop_mode
8035 and compute_doloop_base_on_mode.
8037 2021-07-28 Martin Sebor <msebor@redhat.com>
8039 PR middle-end/101494
8040 * tree-ssa-uninit.c (maybe_warn_operand): Correct object offset
8041 and size computation.
8043 2021-07-28 Martin Sebor <msebor@redhat.com>
8045 PR middle-end/101601
8046 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Remove
8048 Handle pointers to functions.
8050 2021-07-28 Martin Sebor <msebor@redhat.com>
8052 * Makefile.in (OBJS): Add gimple-ssa-warn-access.o and pointer-query.o.
8053 * attribs.h (fndecl_dealloc_argno): Move fndecl_dealloc_argno to tree.h.
8054 * builtins.c (compute_objsize_r): Move to pointer-query.cc.
8055 (access_ref::access_ref): Same.
8056 (access_ref::phi): Same.
8057 (access_ref::get_ref): Same.
8058 (access_ref::size_remaining): Same.
8059 (access_ref::offset_in_range): Same.
8060 (access_ref::add_offset): Same.
8061 (access_ref::inform_access): Same.
8062 (ssa_name_limit_t::visit_phi): Same.
8063 (ssa_name_limit_t::leave_phi): Same.
8064 (ssa_name_limit_t::next): Same.
8065 (ssa_name_limit_t::next_phi): Same.
8066 (ssa_name_limit_t::~ssa_name_limit_t): Same.
8067 (pointer_query::pointer_query): Same.
8068 (pointer_query::get_ref): Same.
8069 (pointer_query::put_ref): Same.
8070 (pointer_query::flush_cache): Same.
8071 (warn_string_no_nul): Move to gimple-ssa-warn-access.cc.
8072 (check_nul_terminated_array): Same.
8073 (unterminated_array): Same.
8074 (maybe_warn_for_bound): Same.
8075 (check_read_access): Same.
8076 (warn_for_access): Same.
8077 (get_size_range): Same.
8078 (check_access): Same.
8079 (gimple_call_alloc_size): Move to tree.c.
8080 (gimple_parm_array_size): Move to pointer-query.cc.
8081 (get_offset_range): Same.
8082 (gimple_call_return_array): Same.
8083 (handle_min_max_size): Same.
8084 (handle_array_ref): Same.
8085 (handle_mem_ref): Same.
8086 (compute_objsize): Same.
8087 (gimple_call_alloc_p): Move to gimple-ssa-warn-access.cc.
8088 (call_dealloc_argno): Same.
8089 (fndecl_dealloc_argno): Same.
8090 (new_delete_mismatch_p): Same.
8091 (matching_alloc_calls_p): Same.
8092 (warn_dealloc_offset): Same.
8093 (maybe_emit_free_warning): Same.
8094 * builtins.h (check_nul_terminated_array): Move to
8095 gimple-ssa-warn-access.h.
8096 (check_nul_terminated_array): Same.
8097 (warn_string_no_nul): Same.
8098 (unterminated_array): Same.
8099 (class ssa_name_limit_t): Same.
8100 (class pointer_query): Same.
8101 (struct access_ref): Same.
8102 (class range_query): Same.
8103 (struct access_data): Same.
8104 (gimple_call_alloc_size): Same.
8105 (gimple_parm_array_size): Same.
8106 (compute_objsize): Same.
8107 (class access_data): Same.
8108 (maybe_emit_free_warning): Same.
8109 * calls.c (initialize_argument_information): Remove call to
8110 maybe_emit_free_warning.
8111 * gimple-array-bounds.cc: Include new header..
8112 * gimple-fold.c: Same.
8113 * gimple-ssa-sprintf.c: Same.
8114 * gimple-ssa-warn-restrict.c: Same.
8115 * passes.def: Add pass_warn_access.
8116 * tree-pass.h (make_pass_warn_access): Declare.
8117 * tree-ssa-strlen.c: Include new headers.
8118 * tree.c (fndecl_dealloc_argno): Move here from builtins.c.
8119 * tree.h (fndecl_dealloc_argno): Move here from attribs.h.
8120 * gimple-ssa-warn-access.cc: New file.
8121 * gimple-ssa-warn-access.h: New file.
8122 * pointer-query.cc: New file.
8123 * pointer-query.h: New file.
8125 2021-07-28 Jakub Jelinek <jakub@redhat.com>
8127 PR middle-end/101624
8128 * ubsan.c (maybe_instrument_pointer_overflow,
8129 instrument_object_size): Only test DECL_REGISTER on VAR_DECLs,
8130 PARM_DECLs or RESULT_DECLs.
8131 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
8133 2021-07-28 Jakub Jelinek <jakub@redhat.com>
8135 PR middle-end/101642
8136 * match.pd (bswap16 (x) == bswap16 (y)): Cast both operands
8137 to type of bswap16 for comparison.
8138 (bswap16 (x) == cst): Cast bswap16 operand to type of cst.
8140 2021-07-28 Richard Biener <rguenther@suse.de>
8142 PR tree-optimization/101615
8143 * tree-vect-slp.c (vect_optimize_slp): Materialize permutes
8144 at CTOR SLP graph entries.
8146 2021-07-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8148 * config/aarch64/aarch64.md (*extend<SHORT:mode><GPI:mode>2_aarch64):
8149 Add "r,w" alternative.
8151 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
8154 * config/i386/i386.c (ix86_avx_u128_mode_needed): Don't set
8155 AVX_U128_DIRTY when all bits are zero.
8157 2021-07-28 Richard Biener <rguenther@suse.de>
8159 PR tree-optimization/101615
8160 * tree-vect-slp.c (vect_optimize_slp): Pre-existing vector
8161 external nodes cannot be permuted so make them perm_out 0.
8163 2021-07-28 Andrew Stubbs <ams@codesourcery.com>
8166 * config.in: Regenerate.
8167 * config/gcn/gcn-hsa.h (A_FIJI): New define.
8168 (A_900): New define.
8169 (A_906): New define.
8170 (A_908): New define.
8171 (ASM_SPEC): Use A_FIJI, A_900, A_906 and A_908.
8172 * config/gcn/gcn.c (output_file_start): Adjust attributes according
8173 to the assembler capabilities.
8174 * config/gcn/mkoffload.c (main): Likewise.
8175 * configure: Regenerate.
8176 * configure.ac: Add tests for LLVM assembler attribute features.
8178 2021-07-28 Andrew MacLeod <amacleod@redhat.com>
8180 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Check for
8181 cond_false and cond_true on branches.
8183 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
8185 * config/aarch64/aarch64.c (aarch64_gen_adjusted_ldpstp): use
8188 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
8190 * alias.c (init_alias_analysis): Don't skip prologue/epilogue.
8192 2021-07-28 Jakub Jelinek <jakub@redhat.com>
8195 * config/i386/sse.md (vashr<mode>3): Split into vashrv8di3 expander
8196 and vashrv4di3 expander, where the latter requires just TARGET_AVX2
8197 and has special !TARGET_AVX512VL expansion.
8198 (vashrv2di3<mask_name>): Rename to ...
8199 (vashrv2di3): ... this. Change condition to TARGET_XOP || TARGET_AVX2
8200 and add special !TARGET_XOP && !TARGET_AVX512VL expansion.
8202 2021-07-28 Martin Uecker <muecker@gwdg.de>
8204 * calls.c (maybe_warn_rdwr_sizes): Correct argument
8205 numbers in warning that were switched.
8207 2021-07-28 Kewen Lin <linkw@linux.ibm.com>
8209 PR tree-optimization/101596
8210 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Fix wrong check
8211 by using new_type's precision instead.
8213 2021-07-28 liuhongt <hongtao.liu@intel.com>
8216 * config/i386/i386.h (processor_costs): Add new member
8218 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
8219 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
8220 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
8221 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
8222 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
8223 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
8224 generic_cost, core_cost): Initialize integer_to_sse same value
8226 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
8227 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8228 Use integer_to_sse instead of sse_op to calculate the cost of
8231 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
8233 * config/rs6000/rs6000-gen-builtins.c (write_ovld_static_init): New
8235 (write_init_file): Call write_ovld_static_init.
8237 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
8239 * config/rs6000/rs6000-gen-builtins.c (write_bif_static_init): New
8241 (write_init_file): Call write_bif_static_init.
8243 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
8245 * config/rs6000/rs6000-gen-builtins.c (typemap): New struct.
8246 (TYPE_MAP_SIZE): New macro.
8247 (type_map): New initialized variable.
8248 (typemap_cmp): New function.
8249 (write_type_node): Likewise.
8250 (write_fntype_init): Implement.
8252 2021-07-27 Martin Sebor <msebor@redhat.com>
8254 PR tree-optimization/101584
8255 * tree-ssa-uninit.c (builtin_call_nomodifying_p): New function.
8256 (check_defs): Call it.
8258 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
8260 * tree-ssa-dom.c (dom_jump_threader_simplifier):
8261 Put avail_exprs_stack in the class, instead of passing it to
8262 jump_threader_simplifier.
8263 (dom_jump_threader_simplifier::simplify): Add state argument.
8264 (dom_opt_dom_walker): Add state.
8265 (pass_dominator::execute): Pass state to threader.
8266 (dom_opt_dom_walker::before_dom_children): Use state.
8267 * tree-ssa-threadedge.c (jump_threader::jump_threader): Replace
8269 (jump_threader::record_temporary_equivalences_from_phis):
8270 Register equivalences through the state variable.
8271 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
8272 Record ranges in a statement through the state variable.
8273 (jump_threader::simplify_control_stmt_condition): Pass state to
8275 (jump_threader::simplify_control_stmt_condition_1): Same.
8276 (jump_threader::thread_around_empty_blocks): Remove obsolete
8278 (jump_threader::thread_through_normal_block): Record equivalences
8279 on edge through the state variable.
8280 (jump_threader::thread_across_edge): Abstract state pushing.
8281 (jt_state::jt_state): New.
8282 (jt_state::push): New.
8283 (jt_state::pop): New.
8284 (jt_state::register_equiv): New.
8285 (jt_state::record_ranges_from_stmt): New.
8286 (jt_state::register_equivs_on_edge): New.
8287 (jump_threader_simplifier::jump_threader_simplifier): Move from
8289 (jump_threader_simplifier::simplify): Add state argument.
8290 * tree-ssa-threadedge.h (class jt_state): New.
8291 (class jump_threader): Add state to constructor.
8292 (class jump_threader_simplifier): Add state to simplify. Remove
8293 avail_exprs_stack from class.
8294 * tree-vrp.c (vrp_jump_threader_simplifier::simplify): Add state
8296 (vrp_jump_threader::vrp_jump_threader): Add state.
8297 (vrp_jump_threader::~vrp_jump_threader): Cleanup state.
8299 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
8301 * Makefile.in (OBJS): Add gimple-range-path.o.
8302 * gimple-range-path.cc: New file.
8303 * gimple-range-path.h: New file.
8305 2021-07-27 Jonathan Wright <jonathan.wright@arm.com>
8307 * config/aarch64/aarch64-simd.md: Push sign/zero-extension
8308 inside vec_duplicate for all patterns.
8309 * simplify-rtx.c (simplify_context::simplify_unary_operation_1):
8310 Push sign/zero-extension inside vec_duplicate.
8312 2021-07-27 Richard Biener <rguenther@suse.de>
8314 PR tree-optimization/101573
8315 * tree-ssa-uninit.c (warn_uninit_phi_uses): New function
8316 looking at uninitialized PHI arg defs in some constrained cases.
8317 (warn_uninitialized_vars): Call it.
8318 (execute_early_warn_uninitialized): Calculate dominators.
8320 2021-07-27 Richard Biener <rguenther@suse.de>
8322 PR tree-optimization/39821
8323 * tree-vect-stmts.c (vect_model_promotion_demotion_cost): Use
8324 vector_stmt for widening arithmetic.
8325 (vectorizable_conversion): Adjust.
8327 2021-07-27 Martin Jambor <mjambor@suse.cz>
8329 * cgraph.h (ipa_replace_map): New field force_load_ref.
8330 * ipa-prop.h (ipa_param_descriptor): Reduce precision of move_cost,
8331 aded new flag load_dereferenced, adjusted comments.
8332 (ipa_get_param_dereferenced): New function.
8333 (ipa_set_param_dereferenced): Likewise.
8334 * cgraphclones.c (cgraph_node::create_virtual_clone): Follow it.
8335 * ipa-cp.c: Include gimple.h.
8336 (ipcp_discover_new_direct_edges): Take into account dereferenced flag.
8337 (get_replacement_map): New parameter force_load_ref, set the
8338 appropriate flag in ipa_replace_map if set.
8339 (struct symbol_and_index_together): New type.
8340 (adjust_refs_in_act_callers): New function.
8341 (adjust_references_in_caller): Likewise.
8342 (create_specialized_node): When appropriate, call
8343 adjust_references_in_caller and force only load references.
8344 * ipa-prop.c (load_from_dereferenced_name): New function.
8345 (ipa_analyze_controlled_uses): Also detect loads from a
8346 dereference, harden testing of call statements.
8347 (ipa_write_node_info): Stream the dereferenced flag.
8348 (ipa_read_node_info): Likewise.
8349 (ipa_set_jf_constant): Also create refdesc when jump function
8350 references a variable.
8351 (cgraph_node_for_jfunc): Rename to symtab_node_for_jfunc, work
8352 also on references of variables and return a symtab_node. Adjust
8354 (propagate_controlled_uses): Also remove references to VAR_DECLs.
8356 2021-07-27 Jakub Jelinek <jakub@redhat.com>
8358 PR middle-end/101586
8359 * gimple-fold.c (clear_padding_type): Ignore FIELD_DECLs with byte
8360 positions above or equal to sz except for diagnostics of flexible
8363 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
8365 PR tree-optimization/78888
8366 * gimple-range-fold.cc (get_letter_range): New.
8367 (fold_using_range::range_of_builtin_call): Call get_letter_range.
8369 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
8371 PR tree-optimization/78888
8372 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Add cases
8373 for CFN_BUILT_IN_TOUPPER and CFN_BUILT_IN_TOLOWER.
8375 2021-07-26 Roger Sayle <roger@nextmovesoftware.com>
8376 Marc Glisse <marc.glisse@inria.fr>
8378 * match.pd (rotate): Simplify equality/inequality of rotations.
8379 (bswap): Simplify equality/inequality tests of byte swapping.
8381 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8383 * range-op.cc (operator_bitwise_xor::op1_op2_relation_effect):
8386 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8388 * range-op.cc (operator_lshift::fold_range): Pass rel to
8389 base class fold_range.
8390 (operator_rshift::fold_range): Same.
8392 2021-07-26 Ashimida <ashimida@linux.alibaba.com>
8395 * toplev.h (min_align_loops_log): Remove declaration.
8396 (min_align_jumps_log, min_align_labels_log): Likewise.
8397 (min_align_functions_log): Likewise.
8399 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8401 * tree-vrp.c (vrp_simplify_cond_using_ranges): Rename vr_values
8403 (execute_vrp): Abstract out simplification of conditionals...
8404 (simplify_casted_conds): ...here.
8406 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
8408 * gimple-array-bounds.cc (array_bounds_checker::get_value_range):
8409 Add gimple argument.
8410 (array_bounds_checker::check_array_ref): Same.
8411 (array_bounds_checker::check_addr_expr): Same.
8412 (array_bounds_checker::check_array_bounds): Pass statement to
8413 check_array_bounds and check_addr_expr.
8414 * gimple-array-bounds.h (check_array_bounds): Add gimple argument.
8415 (check_addr_expr): Same.
8416 (get_value_range): Same.
8418 2021-07-26 Tamar Christina <tamar.christina@arm.com>
8420 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): Rename to..
8421 (sdot_prod, udot_prod): ... This.
8422 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): Merged
8424 (<sur>dot_prod<vsi2qi>): ... this.
8425 (aarch64_<sur>dot_lane<vsi2qi>, aarch64_<sur>dot_laneq<vsi2qi>):
8426 Change operands order.
8427 (<sur>sadv16qi): Use new operands order.
8428 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32,
8429 vdotq_s32): Use new RTL ordering.
8431 2021-07-26 Tamar Christina <tamar.christina@arm.com>
8433 * config/aarch64/aarch64-builtins.c (TYPES_TERNOP_SUSS,
8434 aarch64_types_ternop_suss_qualifiers): New.
8435 * config/aarch64/aarch64-simd-builtins.def (usdot_prod): Use it.
8436 * config/aarch64/aarch64-simd.md (usdot_prod<vsi2qi>): Re-organize RTL.
8437 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Use it.
8439 2021-07-23 Jakub Jelinek <jakub@redhat.com>
8441 PR rtl-optimization/101562
8442 * expmed.c (store_integral_bit_field): Only use movstrict_optab
8443 if the operand isn't paradoxical.
8445 2021-07-23 Aldy Hernandez <aldyh@redhat.com>
8447 * gimple-array-bounds.h (class array_bounds_checker): Change
8448 ranges type to range_query.
8450 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8452 * config/aarch64/arm_neon.h (vst1_s64_x2): Use
8453 __builtin_memcpy instead of constructing
8454 __builtin_aarch64_simd_oi one vector at a time.
8455 (vst1_u64_x2): Likewise.
8456 (vst1_f64_x2): Likewise.
8457 (vst1_s8_x2): Likewise.
8458 (vst1_p8_x2): Likewise.
8459 (vst1_s16_x2): Likewise.
8460 (vst1_p16_x2): Likewise.
8461 (vst1_s32_x2): Likewise.
8462 (vst1_u8_x2): Likewise.
8463 (vst1_u16_x2): Likewise.
8464 (vst1_u32_x2): Likewise.
8465 (vst1_f16_x2): Likewise.
8466 (vst1_f32_x2): Likewise.
8467 (vst1_p64_x2): Likewise.
8468 (vst1q_s8_x2): Likewise.
8469 (vst1q_p8_x2): Likewise.
8470 (vst1q_s16_x2): Likewise.
8471 (vst1q_p16_x2): Likewise.
8472 (vst1q_s32_x2): Likewise.
8473 (vst1q_s64_x2): Likewise.
8474 (vst1q_u8_x2): Likewise.
8475 (vst1q_u16_x2): Likewise.
8476 (vst1q_u32_x2): Likewise.
8477 (vst1q_u64_x2): Likewise.
8478 (vst1q_f16_x2): Likewise.
8479 (vst1q_f32_x2): Likewise.
8480 (vst1q_f64_x2): Likewise.
8481 (vst1q_p64_x2): Likewise.
8483 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8485 * config/aarch64/arm_neon.h (vst1_s64_x3): Use
8486 __builtin_memcpy instead of constructing
8487 __builtin_aarch64_simd_ci one vector at a time.
8488 (vst1_u64_x3): Likewise.
8489 (vst1_f64_x3): Likewise.
8490 (vst1_s8_x3): Likewise.
8491 (vst1_p8_x3): Likewise.
8492 (vst1_s16_x3): Likewise.
8493 (vst1_p16_x3): Likewise.
8494 (vst1_s32_x3): Likewise.
8495 (vst1_u8_x3): Likewise.
8496 (vst1_u16_x3): Likewise.
8497 (vst1_u32_x3): Likewise.
8498 (vst1_f16_x3): Likewise.
8499 (vst1_f32_x3): Likewise.
8500 (vst1_p64_x3): Likewise.
8501 (vst1q_s8_x3): Likewise.
8502 (vst1q_p8_x3): Likewise.
8503 (vst1q_s16_x3): Likewise.
8504 (vst1q_p16_x3): Likewise.
8505 (vst1q_s32_x3): Likewise.
8506 (vst1q_s64_x3): Likewise.
8507 (vst1q_u8_x3): Likewise.
8508 (vst1q_u16_x3): Likewise.
8509 (vst1q_u32_x3): Likewise.
8510 (vst1q_u64_x3): Likewise.
8511 (vst1q_f16_x3): Likewise.
8512 (vst1q_f32_x3): Likewise.
8513 (vst1q_f64_x3): Likewise.
8514 (vst1q_p64_x3): Likewise.
8516 2021-07-23 H.J. Lu <hjl.tools@gmail.com>
8519 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): Don't return
8520 hard register when LRA is in progress.
8522 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8524 * config/aarch64/arm_neon.h (vst1_s8_x4): Use
8525 __builtin_memcpy instead of using a union.
8526 (vst1q_s8_x4): Likewise.
8527 (vst1_s16_x4): Likewise.
8528 (vst1q_s16_x4): Likewise.
8529 (vst1_s32_x4): Likewise.
8530 (vst1q_s32_x4): Likewise.
8531 (vst1_u8_x4): Likewise.
8532 (vst1q_u8_x4): Likewise.
8533 (vst1_u16_x4): Likewise.
8534 (vst1q_u16_x4): Likewise.
8535 (vst1_u32_x4): Likewise.
8536 (vst1q_u32_x4): Likewise.
8537 (vst1_f16_x4): Likewise.
8538 (vst1q_f16_x4): Likewise.
8539 (vst1_f32_x4): Likewise.
8540 (vst1q_f32_x4): Likewise.
8541 (vst1_p8_x4): Likewise.
8542 (vst1q_p8_x4): Likewise.
8543 (vst1_p16_x4): Likewise.
8544 (vst1q_p16_x4): Likewise.
8545 (vst1_s64_x4): Likewise.
8546 (vst1_u64_x4): Likewise.
8547 (vst1_p64_x4): Likewise.
8548 (vst1q_s64_x4): Likewise.
8549 (vst1q_u64_x4): Likewise.
8550 (vst1q_p64_x4): Likewise.
8551 (vst1_f64_x4): Likewise.
8552 (vst1q_f64_x4): Likewise.
8554 2021-07-23 Jonathan Wrightt <jonathan.wright@arm.com>
8556 * config/aarch64/arm_neon.h (vst2_s64): Use __builtin_memcpy
8557 instead of constructing __builtin_aarch64_simd_oi one vector
8559 (vst2_u64): Likewise.
8560 (vst2_f64): Likewise.
8561 (vst2_s8): Likewise.
8562 (vst2_p8): Likewise.
8563 (vst2_s16): Likewise.
8564 (vst2_p16): Likewise.
8565 (vst2_s32): Likewise.
8566 (vst2_u8): Likewise.
8567 (vst2_u16): Likewise.
8568 (vst2_u32): Likewise.
8569 (vst2_f16): Likewise.
8570 (vst2_f32): Likewise.
8571 (vst2_p64): Likewise.
8572 (vst2q_s8): Likewise.
8573 (vst2q_p8): Likewise.
8574 (vst2q_s16): Likewise.
8575 (vst2q_p16): Likewise.
8576 (vst2q_s32): Likewise.
8577 (vst2q_s64): Likewise.
8578 (vst2q_u8): Likewise.
8579 (vst2q_u16): Likewise.
8580 (vst2q_u32): Likewise.
8581 (vst2q_u64): Likewise.
8582 (vst2q_f16): Likewise.
8583 (vst2q_f32): Likewise.
8584 (vst2q_f64): Likewise.
8585 (vst2q_p64): Likewise.
8587 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8589 * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy
8590 instead of constructing __builtin_aarch64_simd_ci one vector
8592 (vst3_u64): Likewise.
8593 (vst3_f64): Likewise.
8594 (vst3_s8): Likewise.
8595 (vst3_p8): Likewise.
8596 (vst3_s16): Likewise.
8597 (vst3_p16): Likewise.
8598 (vst3_s32): Likewise.
8599 (vst3_u8): Likewise.
8600 (vst3_u16): Likewise.
8601 (vst3_u32): Likewise.
8602 (vst3_f16): Likewise.
8603 (vst3_f32): Likewise.
8604 (vst3_p64): Likewise.
8605 (vst3q_s8): Likewise.
8606 (vst3q_p8): Likewise.
8607 (vst3q_s16): Likewise.
8608 (vst3q_p16): Likewise.
8609 (vst3q_s32): Likewise.
8610 (vst3q_s64): Likewise.
8611 (vst3q_u8): Likewise.
8612 (vst3q_u16): Likewise.
8613 (vst3q_u32): Likewise.
8614 (vst3q_u64): Likewise.
8615 (vst3q_f16): Likewise.
8616 (vst3q_f32): Likewise.
8617 (vst3q_f64): Likewise.
8618 (vst3q_p64): Likewise.
8620 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8622 * config/aarch64/arm_neon.h (vst4_s64): Use __builtin_memcpy
8623 instead of constructing __builtin_aarch64_simd_xi one vector
8625 (vst4_u64): Likewise.
8626 (vst4_f64): Likewise.
8627 (vst4_s8): Likewise.
8628 (vst4_p8): Likewise.
8629 (vst4_s16): Likewise.
8630 (vst4_p16): Likewise.
8631 (vst4_s32): Likewise.
8632 (vst4_u8): Likewise.
8633 (vst4_u16): Likewise.
8634 (vst4_u32): Likewise.
8635 (vst4_f16): Likewise.
8636 (vst4_f32): Likewise.
8637 (vst4_p64): Likewise.
8638 (vst4q_s8): Likewise.
8639 (vst4q_p8): Likewise.
8640 (vst4q_s16): Likewise.
8641 (vst4q_p16): Likewise.
8642 (vst4q_s32): Likewise.
8643 (vst4q_s64): Likewise.
8644 (vst4q_u8): Likewise.
8645 (vst4q_u16): Likewise.
8646 (vst4q_u32): Likewise.
8647 (vst4q_u64): Likewise.
8648 (vst4q_f16): Likewise.
8649 (vst4q_f32): Likewise.
8650 (vst4q_f64): Likewise.
8651 (vst4q_p64): Likewise.
8653 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8655 * config/aarch64/arm_neon.h (vtbx4_s8): Use __builtin_memcpy
8656 instead of constructing __builtin_aarch64_simd_oi one vector
8658 (vtbx4_u8): Likewise.
8659 (vtbx4_p8): Likewise.
8661 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8663 * config/aarch64/arm_neon.h (vtbl3_s8): Use __builtin_memcpy
8664 instead of constructing __builtin_aarch64_simd_oi one vector
8666 (vtbl3_u8): Likewise.
8667 (vtbl3_p8): Likewise.
8668 (vtbl4_s8): Likewise.
8669 (vtbl4_u8): Likewise.
8670 (vtbl4_p8): Likewise.
8672 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8674 * config/aarch64/arm_neon.h (vqtbx2_s8): Use __builtin_memcpy
8675 instead of constructing __builtin_aarch64_simd_oi one vector
8677 (vqtbx2_u8): Likewise.
8678 (vqtbx2_p8): Likewise.
8679 (vqtbx2q_s8): Likewise.
8680 (vqtbx2q_u8): Likewise.
8681 (vqtbx2q_p8): Likewise.
8682 (vqtbx3_s8): Use __builtin_memcpy instead of constructing
8683 __builtin_aarch64_simd_ci one vector at a time.
8684 (vqtbx3_u8): Likewise.
8685 (vqtbx3_p8): Likewise.
8686 (vqtbx3q_s8): Likewise.
8687 (vqtbx3q_u8): Likewise.
8688 (vqtbx3q_p8): Likewise.
8689 (vqtbx4_s8): Use __builtin_memcpy instead of constructing
8690 __builtin_aarch64_simd_xi one vector at a time.
8691 (vqtbx4_u8): Likewise.
8692 (vqtbx4_p8): Likewise.
8693 (vqtbx4q_s8): Likewise.
8694 (vqtbx4q_u8): Likewise.
8695 (vqtbx4q_p8): Likewise.
8697 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
8699 * config/aarch64/arm_neon.h (vqtbl2_s8): Use __builtin_memcpy
8700 instead of constructing __builtin_aarch64_simd_oi one vector
8702 (vqtbl2_u8): Likewise.
8703 (vqtbl2_p8): Likewise.
8704 (vqtbl2q_s8): Likewise.
8705 (vqtbl2q_u8): Likewise.
8706 (vqtbl2q_p8): Likewise.
8707 (vqtbl3_s8): Use __builtin_memcpy instead of constructing
8708 __builtin_aarch64_simd_ci one vector at a time.
8709 (vqtbl3_u8): Likewise.
8710 (vqtbl3_p8): Likewise.
8711 (vqtbl3q_s8): Likewise.
8712 (vqtbl3q_u8): Likewise.
8713 (vqtbl3q_p8): Likewise.
8714 (vqtbl4_s8): Use __builtin_memcpy instead of constructing
8715 __builtin_aarch64_simd_xi one vector at a time.
8716 (vqtbl4_u8): Likewise.
8717 (vqtbl4_p8): Likewise.
8718 (vqtbl4q_s8): Likewise.
8719 (vqtbl4q_u8): Likewise.
8720 (vqtbl4q_p8): Likewise.
8722 2021-07-23 Haochen Gui <guihaoc@gcc.gnu.org>
8725 * config/rs6000/rs6000.md (cstore<mode>4): Fix wrong fall through.
8727 2021-07-22 Andrew Pinski <apinski@marvell.com>
8729 PR tree-optimization/10153
8730 * tree-tailcall.c (create_tailcall_accumulator):
8731 Don't call fold_convert as the type should be correct already.
8732 (tree_optimize_tail_calls_1): Use build_{one,zero}_cst instead
8733 of integer_{one,zero}_node for the call of create_tailcall_accumulator.
8735 2021-07-22 Aldy Hernandez <aldyh@redhat.com>
8737 * gimple-range-cache.cc (non_null_ref::adjust_range): Replace
8738 varying_p check for null/non-null check.
8740 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
8742 PR tree-optimization/101511
8743 * value-relation.cc (relation_oracle::query_relation): Check if ssa1
8744 is in ssa2's equiv set, and don't trap if so.
8746 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
8748 PR tree-optimization/101497
8749 * gimple-range-fold.cc (fold_using_range::range_of_cond_expr): Check
8752 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
8754 PR tree-optimization/101496
8755 * vr-values.c (simplify_using_ranges::fold_cond): Call range_of_stmt
8756 first, then vrp_visit_cond_Stmt.
8758 2021-07-22 liuhongt <hongtao.liu@intel.com>
8760 * config/i386/i386-expand.c
8761 (ix86_broadcast_from_integer_constant): Rename to ..
8762 (ix86_broadcast_from_constant): .. this, and extend it to
8764 (ix86_expand_vector_move): Extend to float mode.
8765 * config/i386/i386-features.c
8766 (replace_constant_pool_with_broadcast): Remove.
8767 (remove_partial_avx_dependency_gate): Ditto.
8768 (constant_pool_broadcast): Ditto.
8769 (class pass_constant_pool_broadcast): Ditto.
8770 (make_pass_constant_pool_broadcast): Ditto.
8771 (remove_partial_avx_dependency): Adjust gate.
8772 * config/i386/i386-passes.def: Remove pass_constant_pool_broadcast.
8773 * config/i386/i386-protos.h
8774 (make_pass_constant_pool_broadcast): Remove.
8776 2021-07-22 liuhongt <hongtao.liu@intel.com>
8778 * config/i386/constraints.md (Wb): New constraint.
8780 * config/i386/i386.md (*ashlhi3_1): Extend to avx512 mask
8782 (*ashlqi3_1): Ditto.
8783 (*<insn><mode>3_1): Split to ..
8784 (*ashr<mode>3_1): this, ...
8785 (*lshr<mode>3_1): and this, also extend this pattern to avx512
8787 (*<insn><mode>3_1): Split to ..
8788 (*ashr<mode>3_1): this, ...
8789 (*lshrqi3_1): and this, also extend this pattern to avx512
8791 (*lshrhi3_1): And this, also extend this pattern to avx512
8793 * config/i386/sse.md (k<code><mode>): New define_split after
8794 it to convert generic shift pattern to mask shift ones.
8796 2021-07-21 Thomas Schwinge <thomas@codesourcery.com>
8797 Joseph Myers <joseph@codesourcery.com>
8798 Cesar Philippidis <cesar@codesourcery.com>
8800 * tree-core.h (omp_clause_code): Add 'OMP_CLAUSE_NOHOST'.
8801 * tree.c (omp_clause_num_ops, omp_clause_code_name, walk_tree_1):
8803 * tree-pretty-print.c (dump_omp_clause): Likewise.
8804 * omp-general.c (oacc_verify_routine_clauses): Likewise.
8805 * gimplify.c (gimplify_scan_omp_clauses)
8806 (gimplify_adjust_omp_clauses): Likewise.
8807 * tree-nested.c (convert_nonlocal_omp_clauses)
8808 (convert_local_omp_clauses): Likewise.
8809 * omp-low.c (scan_sharing_clauses): Likewise.
8810 * omp-offload.c (execute_oacc_device_lower): Update.
8812 2021-07-21 Martin Sebor <msebor@redhat.com>
8814 * tree-ssa-alias.c (walk_aliased_vdefs_1): Fix typos in a comment.
8816 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8818 * config/rs6000/rs6000-gen-builtins.c (write_init_bif_table):
8821 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8823 * config/rs6000/rs6000-gen-builtins.c (write_fntype): New
8825 (write_fntype_init): New stub function.
8826 (write_init_bif_table): Likewise.
8827 (write_init_ovld_table): New function.
8828 (write_init_file): Implement.
8830 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8832 * config/rs6000/rs6000-gen-builtins.c
8833 (write_autogenerated_header): New function.
8834 (write_decls): Likewise.
8835 (write_extern_fntype): New callback function.
8836 (write_header_file): Implement.
8838 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8840 * config/rs6000/rs6000-gen-builtins.c (write_defines_file):
8843 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8845 * config/rs6000/rs6000-gen-builtins.c (complete_vector_type): New
8847 (complete_base_type): Likewise.
8848 (construct_fntype_id): Likewise.
8849 (parse_bif_entry): Call contruct_fntype_id.
8850 (parse_ovld_entry): Likewise.
8852 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8854 * config/rs6000/rs6000-gen-builtins.c (ovld_stanza): New struct.
8855 (MAXOVLDSTANZAS): New macro.
8856 (ovld_stanzas): New variable.
8857 (curr_ovld_stanza): Likewise.
8858 (MAXOVLDS): New macro.
8859 (ovlddata): New struct.
8860 (ovlds): New variable.
8861 (curr_ovld): Likewise.
8862 (max_ovld_args): Likewise.
8863 (parse_ovld_entry): New function.
8864 (parse_ovld_stanza): Likewise.
8865 (parse_ovld): Implement.
8867 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8869 * config/rs6000/rs6000-gen-builtins.c (parse_bif_attrs):
8872 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8874 * config/rs6000/rs6000-gen-builtins.c (parse_args): New function.
8875 (parse_prototype): Implement.
8877 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8879 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): New enum.
8880 (curr_bif_stanza): New variable.
8881 (stanza_entry): New struct.
8882 (stanza_map): New initialized variable.
8883 (enable_string): Likewise.
8884 (fnkinds): New enum.
8885 (typelist): New struct.
8886 (attrinfo): Likewise.
8887 (MAXRESTROPNDS): New macro.
8888 (prototype): New struct.
8889 (MAXBIFS): New macro.
8890 (bifdata): New struct.
8891 (bifs): New variable.
8892 (curr_bif): Likewise.
8893 (bif_order): Likewise.
8894 (bif_index): Likewise.
8895 (fatal): New function.
8896 (stanza_name_to_stanza): Likewise.
8897 (parse_bif_attrs): New stub function.
8898 (parse_prototype): Likewise.
8899 (parse_bif_entry): New function.
8900 (parse_bif_stanza): Likewise.
8901 (parse_bif): Implement.
8902 (set_bif_order): New function.
8903 (create_bif_order): Implement.
8905 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
8907 * config/rs6000/rs6000-gen-builtins.c (rbtree.h): New #include.
8908 (num_bifs): New variable.
8909 (num_ovld_stanzas): Likewise.
8910 (num_ovlds): Likewise.
8911 (parse_codes): New enum.
8912 (bif_rbt): New variable.
8913 (ovld_rbt): Likewise.
8914 (fntype_rbt): Likewise.
8915 (bifo_rbt): Likewise.
8916 (parse_bif): New stub function.
8917 (create_bif_order): Likewise.
8918 (parse_ovld): Likewise.
8919 (write_header_file): Likewise.
8920 (write_init_file): Likewise.
8921 (write_defines_file): Likewise.
8922 (delete_output_files): New function.
8925 2021-07-21 H.J. Lu <hjl.tools@gmail.com>
8928 * config/i386/i386-builtin.def: Remove OPTION_MASK_ISA_SSE4_2
8929 from CRC32 _builtin functions.
8931 2021-07-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
8933 * coverage.c (build_gcov_info_var_registration): Mark the object placed
8934 in the linker set as referenced so that it does not get optimized away.
8936 2021-07-21 Kito Cheng <kito.cheng@sifive.com>
8939 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
8941 * config.gcc (riscv*-*-*): Detect which python is available.
8943 2021-07-21 Jakub Jelinek <jakub@redhat.com>
8945 PR middle-end/101535
8946 * gimplify.c (omp_check_private): Properly skip ORT_TARGET_DATA
8947 contexts in which decl isn't privatized and for ORT_TARGET return
8948 false if decl is mapped.
8950 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
8952 * gimple-loop-jam.c: Include tree-ssa-sccvn.h.
8953 (tree_loop_unroll_and_jam): Run value-numbering on a loop that
8954 has been successfully unrolled.
8956 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
8958 * tree-ssa-loop-manip.c (determine_exit_conditions): Return a null
8959 exit condition if no tail loop is needed, and if the original exit
8960 condition should therefore be kept as-is.
8961 (tree_transform_and_unroll_loop): Handle that case here too.
8963 2021-07-21 Kewen Lin <linkw@linux.ibm.com>
8965 * tree-data-ref.c (free_dependence_relations): Adjust to pass vec
8967 (free_data_refs): Likewise.
8968 * tree-data-ref.h (free_dependence_relations): Likewise.
8969 (free_data_refs): Likewise.
8970 * tree-predcom.c (struct chain): Use auto_vec instead of vec for
8972 (struct component): Likewise.
8973 (pcom_worker::pcom_worker): Adjust for auto_vec and renaming changes.
8974 (pcom_worker::~pcom_worker): Likewise.
8975 (pcom_worker::release_chain): Adjust as auto_vec changes.
8976 (pcom_worker::loop): Rename to ...
8977 (pcom_worker::m_loop): ... this.
8978 (pcom_worker::datarefs): Rename to ...
8979 (pcom_worker::m_datarefs): ... this. Use auto_vec instead of vec.
8980 (pcom_worker::dependences): Rename to ...
8981 (pcom_worker::m_dependences): ... this. Use auto_vec instead of vec.
8982 (pcom_worker::chains): Rename to ...
8983 (pcom_worker::m_chains): ... this. Use auto_vec instead of vec.
8984 (pcom_worker::looparound_phis): Rename to ...
8985 (pcom_worker::m_looparound_phis): ... this. Use auto_vec instead of
8987 (pcom_worker::cache): Rename to ...
8988 (pcom_worker::m_cache): ... this. Use auto_vec instead of vec.
8989 (pcom_worker::release_chain): Adjust for auto_vec changes.
8990 (pcom_worker::release_chains): Adjust for auto_vec and renaming
8992 (release_component): Remove.
8993 (release_components): Adjust for release_component removal.
8994 (component_of): Adjust to use vec.
8995 (merge_comps): Likewise.
8996 (pcom_worker::aff_combination_dr_offset): Adjust for renaming changes.
8997 (pcom_worker::determine_offset): Likewise.
8998 (class comp_ptrs): Remove.
8999 (pcom_worker::split_data_refs_to_components): Adjust for renaming
9000 changes, for comp_ptrs removal with auto_vec.
9001 (pcom_worker::suitable_component_p): Adjust for renaming changes.
9002 (pcom_worker::filter_suitable_components): Adjust for release_component
9004 (pcom_worker::valid_initializer_p): Adjust for renaming changes.
9005 (pcom_worker::find_looparound_phi): Likewise.
9006 (pcom_worker::add_looparound_copies): Likewise.
9007 (pcom_worker::determine_roots_comp): Likewise.
9008 (pcom_worker::single_nonlooparound_use): Likewise.
9009 (pcom_worker::execute_pred_commoning_chain): Likewise.
9010 (pcom_worker::execute_pred_commoning): Likewise.
9011 (pcom_worker::try_combine_chains): Likewise.
9012 (pcom_worker::prepare_initializers_chain): Likewise.
9013 (pcom_worker::prepare_initializers): Likewise.
9014 (pcom_worker::prepare_finalizers_chain): Likewise.
9015 (pcom_worker::prepare_finalizers): Likewise.
9016 (pcom_worker::tree_predictive_commoning_loop): Likewise.
9018 2021-07-20 Martin Sebor <msebor@redhat.com>
9020 PR middle-end/101397
9021 * builtins.c (gimple_call_return_array): Add argument. Correct
9022 offsets for memchr, mempcpy, stpcpy, and stpncpy.
9023 (compute_objsize_r): Adjust offset computation for argument returning
9026 2021-07-20 Martin Sebor <msebor@redhat.com>
9028 PR middle-end/101300
9029 * tree-ssa-uninit.c (check_defs): Handle UBSAN built-ins.
9031 2021-07-20 Jeff Law <jlaw@localhost.localdomain>
9033 * function.c (assign_parm_setup_block): Use adjust_address instead
9034 of change_address to preserve MEM_EXPR and friends.
9036 2021-07-20 Martin Sebor <msebor@redhat.com>
9038 * cfgloop.h (single_likely_exit): Adjust by-value argument to
9040 * cfgloopanal.c (single_likely_exit): Same.
9041 * cgraph.h (struct cgraph_node): Same.
9042 * cgraphclones.c (cgraph_node::create_virtual_clone): Same.
9043 * genautomata.c (merge_states): Same.
9044 * genextract.c (VEC_char_to_string): Same.
9045 * genmatch.c (dt_node::gen_kids_1): Same.
9046 (walk_captures): Adjust by-value argument to by-reference.
9047 * gimple-ssa-store-merging.c (check_no_overlap): Adjust by-value argument
9048 to by-const-reference.
9049 * gimple.c (gimple_build_call_vec): Same.
9050 (gimple_build_call_internal_vec): Same.
9051 (gimple_build_switch): Same.
9052 (sort_case_labels): Same.
9053 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
9055 * gimple.h (gimple_build_call_vec): Adjust by-value argument to
9057 (gimple_build_call_internal_vec): Same.
9058 (gimple_build_switch): Same.
9059 (sort_case_labels): Same.
9060 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
9062 * haifa-sched.c (calc_priorities): Adjust by-value argument to
9064 (sched_init_luids): Same.
9065 (haifa_init_h_i_d): Same.
9066 * ipa-cp.c (ipa_get_indirect_edge_target_1): Same.
9067 (adjust_callers_for_value_intersection): Adjust by-value argument to
9069 (find_more_scalar_values_for_callers_subset): Adjust by-value argument to
9071 (find_more_contexts_for_caller_subset): Same.
9072 (find_aggregate_values_for_callers_subset): Same.
9073 (copy_useful_known_contexts): Same.
9074 * ipa-fnsummary.c (remap_edge_summaries): Same.
9075 (remap_freqcounting_predicate): Same.
9076 * ipa-inline.c (add_new_edges_to_heap): Adjust by-value argument to
9078 * ipa-predicate.c (predicate::remap_after_inlining): Adjust by-value argument
9079 to by-const-reference.
9080 * ipa-predicate.h (predicate::remap_after_inlining): Same.
9081 * ipa-prop.c (ipa_find_agg_cst_for_param): Same.
9082 * ipa-prop.h (ipa_find_agg_cst_for_param): Same.
9083 * ira-build.c (ira_loop_tree_body_rev_postorder): Same.
9084 * read-rtl.c (add_overload_instance): Same.
9085 * rtl.h (native_decode_rtx): Same.
9086 (native_decode_vector_rtx): Same.
9087 * sched-int.h (sched_init_luids): Same.
9088 (haifa_init_h_i_d): Same.
9089 * simplify-rtx.c (native_decode_vector_rtx): Same.
9090 (native_decode_rtx): Same.
9091 * tree-call-cdce.c (gen_shrink_wrap_conditions): Same.
9092 (shrink_wrap_one_built_in_call_with_conds): Same.
9093 (shrink_wrap_conditional_dead_built_in_calls): Same.
9094 * tree-data-ref.c (create_runtime_alias_checks): Same.
9095 (compute_all_dependences): Same.
9096 * tree-data-ref.h (compute_all_dependences): Same.
9097 (create_runtime_alias_checks): Same.
9098 (index_in_loop_nest): Same.
9099 * tree-if-conv.c (mask_exists): Same.
9100 * tree-loop-distribution.c (class loop_distribution): Same.
9101 (loop_distribution::create_rdg_vertices): Same.
9102 (dump_rdg_partitions): Same.
9103 (debug_rdg_partitions): Same.
9104 (partition_contains_all_rw): Same.
9105 (loop_distribution::distribute_loop): Same.
9106 * tree-parloops.c (oacc_entry_exit_ok_1): Same.
9107 (oacc_entry_exit_single_gang): Same.
9108 * tree-ssa-loop-im.c (hoist_memory_references): Same.
9109 (loop_suitable_for_sm): Same.
9110 * tree-ssa-loop-niter.c (bound_index): Same.
9111 * tree-ssa-reassoc.c (update_ops): Same.
9112 (swap_ops_for_binary_stmt): Same.
9113 (rewrite_expr_tree): Same.
9114 (rewrite_expr_tree_parallel): Same.
9115 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Same.
9116 * tree-ssa-sccvn.h (ao_ref_init_from_vn_reference): Same.
9117 * tree-ssa-structalias.c (process_all_all_constraints): Same.
9118 (make_constraints_to): Same.
9119 (handle_lhs_call): Same.
9120 (find_func_aliases_for_builtin_call): Same.
9121 (sort_fieldstack): Same.
9122 (check_for_overlaps): Same.
9123 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Same.
9124 (vect_create_cond_for_unequal_addrs): Same.
9125 (vect_create_cond_for_lower_bounds): Same.
9126 (vect_create_cond_for_alias_checks): Same.
9127 * tree-vect-slp-patterns.c (vect_validate_multiplication): Same.
9128 * tree-vect-slp.c (vect_analyze_slp_instance): Same.
9129 (vect_make_slp_decision): Same.
9130 (vect_slp_bbs): Same.
9131 (duplicate_and_interleave): Same.
9132 (vect_transform_slp_perm_load): Same.
9133 (vect_schedule_slp): Same.
9134 * tree-vectorizer.h (vect_transform_slp_perm_load): Same.
9135 (vect_schedule_slp): Same.
9136 (duplicate_and_interleave): Same.
9137 * tree.c (build_vector_from_ctor): Same.
9138 (build_vector): Same.
9139 (check_vector_cst): Same.
9140 (check_vector_cst_duplicate): Same.
9141 (check_vector_cst_fill): Same.
9142 (check_vector_cst_stepped): Same.
9143 * tree.h (build_vector_from_ctor): Same.
9145 2021-07-20 Jakub Jelinek <jakub@redhat.com>
9148 * config/rs6000/rs6000-protos.h (easy_altivec_constant): Change return
9149 type from bool to int.
9150 * config/rs6000/rs6000.c (vspltis_constant): Fix up handling the
9151 EASY_VECTOR_MSB case if either step or copies is not 1.
9152 (vspltis_shifted): Fix comment typo.
9153 (easy_altivec_constant): Change return type from bool to int, instead
9154 of returning true return byte size of the element mode that should be
9155 used to synthetize the constant.
9156 * config/rs6000/predicates.md (easy_vector_constant_msb): Require
9157 that vspltis_shifted is 0, handle the case where easy_altivec_constant
9158 assumes using different vector mode from CONST_VECTOR's mode.
9159 * config/rs6000/altivec.md (easy_vector_constant_msb splitter): Use
9160 easy_altivec_constant to determine mode in which -1 >> -1 should be
9161 performed, use rs6000_expand_vector_init instead of gen_vec_initv4sisi.
9163 2021-07-20 Richard Biener <rguenther@suse.de>
9166 * dwarf2out.h (dwarf_file_data): Add key member.
9167 * dwarf2out.c (dwarf_file_hasher::equal): Compare key.
9168 (dwarf_file_hasher::hash): Hash key.
9169 (lookup_filename): Remap the filename and store it in the
9170 filename member of dwarf_file_data when creating a new
9172 (file_name_acquire): Do not remap the filename again.
9173 (maybe_emit_file): Likewise.
9175 2021-07-20 Jonathan Wright <jonathan.wright@arm.com>
9177 * config/aarch64/aarch64-simd-builtins.def: Use two variant
9178 generators for all TBL/TBX intrinsics and rename to
9179 consistent forms: qtbl[1234] or qtbx[1234].
9180 * config/aarch64/aarch64-simd.md (aarch64_tbl1<mode>):
9182 (aarch64_qtbl1<mode>): This.
9183 (aarch64_tbx1<mode>): Rename to...
9184 (aarch64_qtbx1<mode>): This.
9185 (aarch64_tbl2v16qi): Delete.
9186 (aarch64_tbl3<mode>): Rename to...
9187 (aarch64_qtbl2<mode>): This.
9188 (aarch64_tbx4<mode>): Rename to...
9189 (aarch64_qtbx2<mode>): This.
9190 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_1): Use
9191 renamed qtbl1 and qtbl2 RTL patterns.
9192 * config/aarch64/arm_neon.h (vqtbl1_p8): Use renamed qtbl1
9194 (vqtbl1_s8): Likewise.
9195 (vqtbl1_u8): Likewise.
9196 (vqtbl1q_p8): Likewise.
9197 (vqtbl1q_s8): Likewise.
9198 (vqtbl1q_u8): Likewise.
9199 (vqtbx1_s8): Use renamed qtbx1 RTL pattern.
9200 (vqtbx1_u8): Likewise.
9201 (vqtbx1_p8): Likewise.
9202 (vqtbx1q_s8): Likewise.
9203 (vqtbx1q_u8): Likewise.
9204 (vqtbx1q_p8): Likewise.
9205 (vtbl1_s8): Use renamed qtbl1 RTL pattern.
9206 (vtbl1_u8): Likewise.
9207 (vtbl1_p8): Likewise.
9208 (vtbl2_s8): Likewise
9209 (vtbl2_u8): Likewise.
9210 (vtbl2_p8): Likewise.
9211 (vtbl3_s8): Use renamed qtbl2 RTL pattern.
9212 (vtbl3_u8): Likewise.
9213 (vtbl3_p8): Likewise.
9214 (vtbl4_s8): Likewise.
9215 (vtbl4_u8): Likewise.
9216 (vtbl4_p8): Likewise.
9217 (vtbx2_s8): Use renamed qtbx2 RTL pattern.
9218 (vtbx2_u8): Likewise.
9219 (vtbx2_p8): Likewise.
9220 (vqtbl2_s8): Use renamed qtbl2 RTL pattern.
9221 (vqtbl2_u8): Likewise.
9222 (vqtbl2_p8): Likewise.
9223 (vqtbl2q_s8): Likewise.
9224 (vqtbl2q_u8): Likewise.
9225 (vqtbl2q_p8): Likewise.
9226 (vqtbx2_s8): Use renamed qtbx2 RTL pattern.
9227 (vqtbx2_u8): Likewise.
9228 (vqtbx2_p8): Likewise.
9229 (vqtbx2q_s8): Likewise.
9230 (vqtbx2q_u8): Likewise.
9231 (vqtbx2q_p8): Likewise.
9232 (vtbx4_s8): Likewise.
9233 (vtbx4_u8): Likewise.
9234 (vtbx4_p8): Likewise.
9236 2021-07-20 Uroš Bizjak <ubizjak@gmail.com>
9239 * config/i386/sync.md (define_peephole2 atomic_storedi_fpu):
9241 (define_peephole2 atomic_loaddi_fpu): Ditto.
9243 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
9245 * config.gcc (riscv*-*-*): Detect which python is available.
9247 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
9249 * config/rs6000/vsx.md (mulhs_<mode>): Rename to...
9250 (smul<mode>3_highpart): ... this.
9251 (mulhu_<mode>): Rename to...
9252 (umul<mode>3_highpart): ... this.
9253 * config/rs6000/rs6000-builtin.def (MULHS_V2DI, MULHS_V4SI,
9254 MULHU_V2DI, MULHU_V4SI): Adjust.
9256 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
9258 PR tree-optimization/100696
9259 * internal-fn.c (first_commutative_argument): Add info for IFN_MULH.
9260 * internal-fn.def (IFN_MULH): New internal function.
9261 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Add support to
9262 recog normal multiply highpart as IFN_MULH.
9263 * config/i386/i386.c (ix86_add_stmt_cost): Adjust for combined
9266 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
9268 * config/elfos.h (CTF_DEBUGGING_INFO): New definition.
9269 (BTF_DEBUGGING_INFO): Likewise.
9270 * doc/tm.texi.in: Document the new macros.
9271 * doc/tm.texi: Regenerated.
9272 * toplev.c: Guard initialization of debug hooks.
9274 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
9276 * flags.h (ctf_debuginfo_p): New function declaration.
9277 * opts.c (ctf_debuginfo_p): New function definition.
9279 2021-07-19 Andrew Stubbs <ams@codesourcery.com>
9282 * config/gcn/gcn-hsa.h (DRIVER_SELF_SPECS): New.
9283 (ASM_SPEC): Set -mattr for xnack and sram-ecc.
9284 * config/gcn/gcn-opts.h (enum sram_ecc_type): New.
9285 * config/gcn/gcn-valu.md: Add a warning comment.
9286 * config/gcn/gcn.c (gcn_option_override): Add "sorry" for -mxnack.
9287 (output_file_start): Add xnack and sram-ecc state to ".amdgcn_target".
9288 * config/gcn/gcn.md: Add a warning comment.
9289 * config/gcn/gcn.opt: Add -mxnack and -msram-ecc.
9290 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): Remove
9292 (EF_AMDGPU_XNACK): New.
9293 (EF_AMDGPU_SRAM_ECC): New.
9295 (copy_early_debug_info): Use elf_flags.
9296 (main): Handle -mxnack and -msram-ecc options.
9297 * doc/invoke.texi: Document -mxnack and -msram-ecc.
9299 2021-07-19 Andrew Pinski <apinski@marvell.com>
9302 * config/aarch64/aarch64.md (csneg3_uxtw_insn): Rename to ...
9303 (*cs<neg_not_cs>3_uxtw_insn4): and extend to NEG_NOT.
9305 2021-07-19 Richard Biener <rguenther@suse.de>
9307 PR tree-optimization/101505
9308 * tree-vect-patterns.c (vect_determine_precisions): Walk
9309 PHIs also for loop vectorization.
9311 2021-07-19 Richard Biener <rguenther@suse.de>
9313 * gimple.h (gimple_expr_type): Remove.
9314 * doc/gimple.texi: Remove gimple_expr_type documentation.
9316 2021-07-19 Richard Biener <rguenther@suse.de>
9318 * tree-ssa-sccvn.c (vn_reference_eq): Handle NULL vr->type.
9319 (ao_ref_init_from_vn_reference): Likewise.
9320 (fully_constant_reference): Likewise.
9321 (vn_reference_lookup_call): Do not set vr->type to random
9323 * tree-ssa-pre.c (compute_avail): Do not try to PRE calls
9325 * tree-vect-generic.c (expand_vector_piecewise): Pass in
9326 whether we expanded parallel.
9327 (expand_vector_parallel): Adjust.
9328 (expand_vector_addition): Likewise.
9329 (expand_vector_comparison): Likewise.
9330 (expand_vector_operation): Likewise.
9331 (expand_vector_scalar_condition): Likewise.
9332 (expand_vector_conversion): Likewise.
9334 2021-07-19 Richard Biener <rguenther@suse.de>
9336 * tree-vrp.c (register_edge_assert_for_2): Use the
9338 (vrp_folder::fold_predicate_in): Likewise.
9339 * vr-values.c (gimple_assign_nonzero_p): Likewise.
9340 (vr_values::extract_range_from_comparison): Likewise.
9341 (vr_values::extract_range_from_ubsan_builtin): Use the
9342 type of the first operand.
9343 (vr_values::extract_range_basic): Push down type
9344 computation, use the appropriate LHS.
9345 (vr_values::extract_range_from_assignment): Use the
9348 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
9351 * common/config/i386/i386-common.c (ix86_handle_option): For
9352 -mgeneral-regs-only, enable the GPR only instructions which are
9353 enabled implicitly by SSE ISAs unless they have been disabled
9356 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
9359 * config/i386/i386.c (ix86_check_avx_upper_stores): Moved before
9360 ix86_avx_u128_mode_needed.
9361 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY if callee
9362 returns AVX register.
9364 2021-07-17 Jan Hubicka <hubicka@ucw.cz>
9366 * tree-ssa-structalias.c (handle_rhs_call): Support EAF_NOT_RETURNED.
9367 (handle_const_call): Liekise
9368 (handle_pure_call): Liekise
9370 2021-07-17 Andrew MacLeod <amacleod@redhat.com>
9372 PR tree-optimization/96542
9373 * range-op.cc (range_operator::wi_fold_in_parts): New.
9374 (range_operator::fold_range): Call wi_fold_in_parts.
9375 (operator_lshift::wi_fold): Fix broken lshift by [0,0].
9376 * range-op.h (wi_fold_in_parts): Add prototype.
9378 2021-07-16 David Malcolm <dmalcolm@redhat.com>
9380 * doc/analyzer.texi: Add __analyzer_dump_state.
9382 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9384 * config/rs6000/rbtree.c: New file.
9385 * config/rs6000/rbtree.h: New file.
9387 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9389 * config/rs6000/rs6000-gen-builtins.c (restriction): New enum.
9390 (typeinfo): Add restr field.
9391 (match_bracketed_pair): New function.
9392 (match_const_restriction): Implement.
9394 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9396 * config/rs6000/rs6000-gen-builtins.c (match_basetype): Implement.
9398 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9400 * config/rs6000/rs6000-gen-builtins.c (void_status): New enum.
9401 (basetype): Likewise.
9402 (typeinfo): Likewise.
9403 (handle_pointer): New function.
9404 (match_basetype): New stub function.
9405 (match_const_restriction): Likewise.
9406 (match_type): New function.
9408 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9410 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace): New
9412 (advance_line): Likewise.
9413 (safe_inc_pos): Likewise.
9414 (match_identifier): Likewise.
9415 (match_integer): Likewise.
9416 (match_to_right_bracket): Likewise.
9418 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9420 * config/rs6000/rs6000-gen-builtins.c (bif_file): New variable.
9421 (ovld_file): Likewise.
9422 (header_file): Likewise.
9423 (init_file): Likewise.
9424 (defines_file): Likewise.
9425 (pgm_path): Likewise.
9426 (bif_path): Likewise.
9427 (ovld_path): Likewise.
9428 (header_path): Likewise.
9429 (init_path): Likewise.
9430 (defines_path): Likewise.
9431 (LINELEN): New macro.
9432 (linebuf): New variable.
9436 (bif_diag): New function.
9437 (ovld_diag): Likewise.
9439 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9441 * config/rs6000/rs6000-builtin-new.def: New.
9442 * config/rs6000/rs6000-overload.def: New.
9444 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9446 * config/rs6000/rs6000-gen-builtins.c: New.
9448 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
9450 * Makefile.in (EXTRA_GTYPE_DEPS): New variable.
9451 (s-gtype): Depend on EXTRA_GTYPE_DEPS.
9452 * gengtype-state.c (state_writer::write_state_file_list): Add a
9453 parameter to the fileslist expression for the number of build
9455 (read_state_files_list): Detect build headers and strip the
9456 initial "./" or ".\" from their names.
9457 * gengtype.c (build_headers): New global variable.
9458 (num_build_headers): Likewise.
9459 (open_base_files): Emit #include for each build header.
9460 (main): Detect and count build headers.
9461 * gengtype.h (build_headers): New extern variable.
9462 (num_build_headers): Likewise.
9464 2021-07-16 Richard Biener <rguenther@suse.de>
9466 * gimple-ssa-store-merging.c (verify_symbolic_number_p): Use
9467 the type of the LHS.
9468 (find_bswap_or_nop_1): Likewise.
9469 (find_bswap_or_nop): Likewise.
9470 * tree-vectorizer.h (vect_get_smallest_scalar_type): Adjust
9472 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
9473 Remove unused parameters, pass in the scalar type. Fix
9474 internal store function handling.
9475 * tree-vect-stmts.c (vect_analyze_stmt): Remove assert.
9476 (vect_get_vector_types_for_stmt): Move down check for
9477 existing vector stmt after we've determined a scalar type.
9478 Pass down the used scalar type to vect_get_smallest_scalar_type.
9479 * tree-vect-generic.c (expand_vector_condition): Use
9480 the type of the LHS.
9481 (expand_vector_scalar_condition): Likewise.
9482 (expand_vector_operations_1): Likewise.
9483 * tree-vect-patterns.c (vect_widened_op_tree): Likewise.
9484 (vect_recog_dot_prod_pattern): Likewise.
9485 (vect_recog_sad_pattern): Likewise.
9486 (vect_recog_widen_op_pattern): Likewise.
9487 (vect_recog_widen_sum_pattern): Likewise.
9488 (vect_recog_mixed_size_cond_pattern): Likewise.
9490 2021-07-16 Jan Hubicka <hubicka@ucw.cz>
9492 * ipa-modref.c (struct escape_entry): Use eaf_fleags_t.
9493 (dump_eaf_flags): Dump EAF_NOT_RETURNED
9494 (eaf_flags_useful_p): Use eaf_fleags_t; handle const functions
9495 and EAF_NOT_RETURNED.
9496 (modref_summary::useful_p): Likewise.
9497 (modref_summary_lto::useful_p): Likewise.
9498 (struct) modref_summary_lto: Use eaf_fleags_t.
9499 (deref_flags): Handle EAF_NOT_RETURNED.
9500 (struct escape_point): Use min_flags.
9501 (modref_lattice::init): Add EAF_NOT_RETURNED.
9502 (merge_call_lhs_flags): Ignore EAF_NOT_RETURNED functions
9503 (analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
9505 (analyze_parms): Also analyze const functions; update conition on
9507 (modref_write): Update streaming.
9508 (read_section): Update streaming.
9509 (remap_arg_flags): Use eaf_flags_t.
9510 (modref_merge_call_site_flags): Hanlde EAF_NOT_RETURNED.
9511 * ipa-modref.h: (eaf_flags_t): New typedef.
9512 (struct modref_summary): Use eaf_flags_t.
9513 * tree-core.h (EAF_NOT_RETURNED): New constant.
9515 2021-07-16 Richard Biener <rguenther@suse.de>
9517 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Use
9518 the type of the LHS.
9519 (gimple_assign_nonnegative_warnv_p): Likewise.
9520 (gimple_call_nonnegative_warnv_p): Likewise. Return false
9521 if the call has no LHS.
9522 * gimple.c (gimple_could_trap_p_1): Use the type of the LHS.
9523 * tree-eh.c (stmt_could_throw_1_p): Likewise.
9524 * tree-inline.c (insert_init_stmt): Likewise.
9525 * tree-ssa-loop-niter.c (get_val_for): Likewise.
9526 * tree-outof-ssa.c (ssa_is_replaceable_p): Use the type of
9528 * tree-ssa-sccvn.c (init_vn_nary_op_from_stmt): Take a
9529 gassign *. Use the type of the lhs.
9530 (vn_nary_op_lookup_stmt): Adjust.
9531 (vn_nary_op_insert_stmt): Likewise.
9533 2021-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
9535 * config/s390/predicates.md (bras_sym_operand): Accept all
9536 functions in 64-bit mode, use UNSPEC_PLT31.
9537 (larl_operand): Use UNSPEC_PLT31.
9538 * config/s390/s390.c (s390_loadrelative_operand_p): Likewise.
9539 (legitimize_pic_address): Likewise.
9540 (s390_emit_tls_call_insn): Mark __tls_get_offset as function,
9542 (s390_delegitimize_address): Use UNSPEC_PLT31.
9543 (s390_output_addr_const_extra): Likewise.
9544 (print_operand): Add @PLT to TLS calls, handle %K.
9545 (s390_function_profiler): Mark __fentry__/_mcount as function,
9546 use %K, use UNSPEC_PLT31.
9547 (s390_output_mi_thunk): Use only UNSPEC_GOT, use %K.
9548 (s390_emit_call): Use UNSPEC_PLT31.
9549 (s390_emit_tpf_eh_return): Mark __tpf_eh_return as function.
9550 * config/s390/s390.md (UNSPEC_PLT31): Rename from UNSPEC_PLT.
9551 (*movdi_64): Use %K.
9552 (reload_base_64): Likewise.
9553 (*sibcall_brc): Likewise.
9554 (*sibcall_brcl): Likewise.
9555 (*sibcall_value_brc): Likewise.
9556 (*sibcall_value_brcl): Likewise.
9559 (*bras_r): Likewise.
9560 (*brasl_r): Likewise.
9561 (*bras_tls): Likewise.
9562 (*brasl_tls): Likewise.
9563 (main_base_64): Likewise.
9564 (reload_base_64): Likewise.
9565 (@split_stack_call<mode>): Likewise.
9567 2021-07-16 Richard Biener <rguenther@suse.de>
9569 PR tree-optimization/101467
9570 * tree-vect-stmts.c (vect_gen_while): Properly guard
9571 make_temp_ssa_name usage.
9573 2021-07-16 Cooper Qu <cooper.qu@linux.alibaba.com>
9575 * config.gcc: Don't use forked print-sysroot-suffix.sh and
9576 t-sysroot-suffix for C-SKY.
9577 * config/csky/print-sysroot-suffix.sh: Delete.
9578 * config/csky/t-csky-linux: Delete.
9579 * config/csky/t-sysroot-suffix: Define MULTILIB_DIRNAMES
9580 instead of CSKY_MULTILIB_DIRNAMES.
9582 2021-07-16 Richard Biener <rguenther@suse.de>
9584 * tree-vect-loop.c (vect_transform_cycle_phi): Correct sign
9585 conversion issues with the partial reduction of the reused
9588 2021-07-16 Richard Biener <rguenther@suse.de>
9590 * config/i386/i386-options.c (ix86_option_override_internal): Set
9591 param_vect_partial_vector_usage to zero if not set.
9593 2021-07-15 Uroš Bizjak <ubizjak@gmail.com>
9596 * config/i386/i386.h (VALID_SSE_REG_MODE): Add TDmode.
9597 (VALID_INT_MODE_P): Add SDmode and DDmode.
9598 Add TDmode for TARGET_64BIT.
9599 (VALID_DFP_MODE_P): Remove.
9600 * config/i386/i386.c (ix86_hard_regno_mode_ok):
9601 Do not use VALID_DFP_MODE_P.
9603 2021-07-15 Andrew MacLeod <amacleod@redhat.com>
9605 * gimple-range-fold.cc (adjust_pointer_diff_expr): Use
9607 (fold_using_range::fold_stmt): Ditto.
9608 (fold_using_range::range_of_range_op): Ditto.
9609 (fold_using_range::range_of_phi): Ditto.
9610 (fold_using_range::range_of_call): Ditto.
9611 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
9612 (fold_using_range::range_of_builtin_call): Ditto.
9613 (fold_using_range::range_of_cond_expr): Ditto.
9614 * gimple-range-fold.h (gimple_range_type): New.
9616 2021-07-15 Martin Sebor <msebor@redhat.com>
9619 * tree-ssa-strlen.c (handle_assign): New function.
9620 (maybe_warn_overflow): Add argument.
9621 (nonzero_bytes_for_type): New function.
9622 (count_nonzero_bytes): Handle more tree types. Call
9623 nonzero_bytes_for_tye.
9624 (count_nonzero_bytes): Handle types.
9625 (handle_store): Handle stores from function calls.
9626 (strlen_check_and_optimize_call): Move code to handle_assign. Call
9627 it for assignments from function calls.
9629 2021-07-15 David Malcolm <dmalcolm@redhat.com>
9634 * doc/invoke.texi: Add -Wanalyzer-use-of-uninitialized-value.
9636 2021-07-15 David Malcolm <dmalcolm@redhat.com>
9638 * doc/invoke.texi (-fdump-analyzer-exploded-paths): New.
9640 2021-07-15 Martin Sebor <msebor@redhat.com>
9644 * fold-const.c (operand_compare::operand_equal_p): Handle OEP_DECL_NAME.
9645 (operand_compare::verify_hash_value): Same.
9646 * tree-core.h (OEP_DECL_NAME): New.
9648 2021-07-15 Martin Jambor <mjambor@suse.cz>
9650 * profile-count.h (profile_count::value): Change the return type to
9652 * gimple-pretty-print.c (dump_gimple_bb_header): Adjust print
9654 * tree-cfg.c (dump_function_to_file): Likewise.
9656 2021-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
9659 * config/rs6000/rs6000-p8swap.c (has_part_mult): New.
9660 (rs6000_analyze_swaps): Insns containing a subreg of a mult are
9663 2021-07-15 Richard Biener <rguenther@suse.de>
9665 * tree-vectorizer.h (vect_gen_while): Match up with
9667 * tree-vect-stmts.c (vect_gen_while): Adjust API to that
9668 of vect_gen_while_not.
9669 (vect_gen_while_not): Adjust.
9670 * tree-vect-loop-manip.c (vect_set_loop_controls_directly): Likewise.
9672 2021-07-15 Aldy Hernandez <aldyh@redhat.com>
9674 * gimple-range-cache.cc (non_null_ref::adjust_range): New.
9675 (ranger_cache::range_of_def): Call adjust_range.
9676 (ranger_cache::entry_range): Same.
9677 * gimple-range-cache.h (non_null_ref::adjust_range): New.
9678 * gimple-range.cc (gimple_ranger::range_of_expr): Call
9680 (gimple_ranger::range_on_entry): Same.
9682 2021-07-15 Tamar Christina <tamar.christina@arm.com>
9685 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9687 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
9689 2021-07-15 Tamar Christina <tamar.christina@arm.com>
9692 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9694 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
9695 (sdot_prod, udot_prod): ...These.
9696 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
9697 (aarch64_<sur>dot<vsi2qi>): Rename to...
9698 (<sur>dot_prod<vsi2qi>): ...This.
9699 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
9702 2021-07-15 Jakub Jelinek <jakub@redhat.com>
9704 PR middle-end/101437
9705 * gimplify.c (gimplify_expr): Throw away volatile reads from empty
9706 types even if they have non-BLKmode TYPE_MODE.
9708 2021-07-15 Richard Biener <rguenther@suse.de>
9711 * gcc.c (process_command): Process -gtoggle like process_options
9712 would after parsing options.
9714 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
9716 * cfgexpand.c (expand_asm_loc): Adjust.
9717 (expand_asm_stmt): Likewise.
9718 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Likewise.
9719 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
9720 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
9721 * config/avr/avr.c (avr_md_asm_adjust): Likewise.
9722 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
9723 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
9724 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
9725 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
9726 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
9727 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
9728 * config/s390/s390.c (s390_md_asm_adjust): Likewise.
9729 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
9730 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
9731 * doc/tm.texi: Regenerate.
9732 * target.def: Add location argument to md_asm_adjust.
9734 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
9736 * tree-diagnostic.c (diagnostic_report_current_function): Use the
9737 diagnostic's location, not input_location.
9739 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
9741 * cfgexpand.c (tree_conflicts_with_clobbers_p): Pass location to
9743 (expand_asm_stmt): Likewise.
9745 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
9747 * config/rs6000/rs6000.c (adjacent_mem_locations): Return the lower
9748 addressed memory rtx, if any.
9749 (rs6000_split_multireg_move): Fix code formatting.
9750 Handle MMA build built-ins with operands in adjacent memory locations.
9752 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
9754 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Move to later
9757 2021-07-14 Jason Merrill <jason@redhat.com>
9759 * sel-sched-ir.h (get_all_loop_exits): Use auto_vec.
9761 2021-07-14 Jason Merrill <jason@redhat.com>
9763 * doc/invoke.texi: -fdelete-dead-exceptions is on by default for
9766 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9768 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
9769 Remove erroneous line.
9771 2021-07-14 Andrew MacLeod <amacleod@redhat.com>
9773 * params.opt (param_evrp_mode): Change default.
9775 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9777 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
9778 (sdot_prod, udot_prod): ...These.
9779 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
9780 (aarch64_<sur>dot<vsi2qi>): Rename to...
9781 (<sur>dot_prod<vsi2qi>): ...This.
9782 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
9785 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9787 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
9789 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9791 * doc/sourcebuild.texi (arm_v8_2a_i8mm_neon_hw): Document.
9793 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9795 * config/arm/neon.md (usdot_prod<vsi2qi>): New.
9797 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9799 * config/aarch64/aarch64-simd.md (aarch64_usdot<vsi2qi>): Rename to...
9800 (usdot_prod<vsi2qi>): ... This.
9801 * config/aarch64/aarch64-simd-builtins.def (usdot): Rename to...
9802 (usdot_prod): ...This.
9803 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Likewise.
9804 * config/aarch64/aarch64-sve.md (@aarch64_<sur>dot_prod<vsi2qi>):
9806 (@<sur>dot_prod<vsi2qi>): ...This.
9807 * config/aarch64/aarch64-sve-builtins-base.cc
9808 (svusdot_impl::expand): Use it.
9810 2021-07-14 Tamar Christina <tamar.christina@arm.com>
9812 * optabs.def (usdot_prod_optab): New.
9813 * doc/md.texi: Document it and clarify other dot prod optabs.
9814 * optabs-tree.h (enum optab_subtype): Add optab_vector_mixed_sign.
9815 * optabs-tree.c (optab_for_tree_code): Support usdot_prod_optab.
9816 * optabs.c (expand_widen_pattern_expr): Likewise.
9817 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
9818 * tree-vect-loop.c (vectorizable_reduction): Query dot-product kind.
9819 * tree-vect-patterns.c (vect_supportable_direct_optab_p): Take optional
9821 (vect_widened_op_tree): Optionally ignore
9823 (vect_recog_dot_prod_pattern): Support usdot_prod_optab.
9825 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
9828 * config/i386/driver-i386.c (host_detect_local_cpu): Check
9829 "arch [32|64]" and "tune [32|64]" for 32-bit and 64-bit codegen.
9830 Enable UINTR only for 64-bit codegen.
9831 * config/i386/i386-options.c
9832 (ix86_option_override_internal::DEF_PTA): Skip PTA_UINTR if not
9834 * config/i386/i386.h (ARCH_ARG): New.
9835 (CC1_CPU_SPEC): Pass "[arch|tune] 32" for 32-bit codegen and
9836 "[arch|tune] 64" for 64-bit codegen.
9838 2021-07-14 Richard Biener <rguenther@suse.de>
9840 PR tree-optimization/101445
9841 * tree-vect-stmts.c (vectorizable_load): Do the gap adjustment
9842 of the IV in the correct direction for negative stride
9845 2021-07-14 Jakub Jelinek <jakub@redhat.com>
9848 * godump.c (godump_str_hash): New type.
9849 (godump_container::pot_dummy_types): Use string_hash instead of
9850 ptr_hash in the hash_set.
9852 2021-07-14 Richard Biener <rguenther@suse.de>
9854 * tree-vect-loop.c (vect_find_reusable_accumulator): Handle
9855 vector types where the old vector type has a multiple of
9856 the new vector type elements.
9857 (vect_create_partial_epilog): New function, split out from...
9858 (vect_create_epilog_for_reduction): ... here.
9859 (vect_transform_cycle_phi): Reduce the re-used accumulator
9860 to the new vector type.
9862 2021-07-14 Alexandre Oliva <oliva@adacore.com>
9864 * tree-ssa-alias.c (attr_fnspec::verify): Fix index in
9865 non-'t'-sized arg check.
9867 2021-07-14 Alexandre Oliva <oliva@adacore.com>
9869 * tree-cfg.c (cleanup_dead_labels_eh): Update
9870 post_landing_pad label upon change of landing pad block's
9872 (cleanup_dead_labels): Check that a removed label is not that
9875 2021-07-13 Jonathan Wright <jonathan.wright@arm.com>
9877 * combine.c (combine_simplify_rtx): Add vec_select -> subreg
9879 * config/aarch64/aarch64.md (*zero_extend<SHORT:mode><GPI:mode>2_aarch64):
9880 Add Neon to general purpose register case for zero-extend
9882 * config/arm/vfp.md (*arm_movsi_vfp): Remove "*" from *t -> r
9883 case to prevent some cases opting to go through memory.
9884 * cse.c (fold_rtx): Add vec_select -> subreg simplification.
9885 * rtl.c (rtvec_series_p): Define predicate to determine
9886 whether a vector contains a linear series of integers.
9887 * rtl.h (rtvec_series_p): Define.
9888 * rtlanal.c (vec_series_lowpart_p): Define predicate to
9889 determine if a vector selection is equivalent to the low part
9891 * rtlanal.h (vec_series_lowpart_p): Define.
9892 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
9893 Add vec_select -> subreg simplification.
9895 2021-07-13 Paul A. Clarke <pc@us.ibm.com>
9897 * config/rs6000/smmintrin.h (_mm_testz_si128, _mm_testc_si128,
9898 _mm_testnzc_si128, _mm_test_all_ones, _mm_test_all_zeros,
9899 _mm_test_mix_ones_zeros): New.
9901 2021-07-13 Roger Sayle <roger@nextmovesoftware.com>
9902 Richard Biener <rguenther@suse.de>
9904 * gimple.c (gimple_could_trap_p_1): Make S argument a
9905 "const gimple*". Preserve constness in call to
9906 gimple_asm_volatile_p.
9907 (gimple_could_trap_p): Make S argument a "const gimple*".
9908 * gimple.h (gimple_could_trap_p_1, gimple_could_trap_p):
9909 Update function prototypes.
9911 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9913 * tree-vectorizer.h (vect_reusable_accumulator): New structure.
9914 (_loop_vec_info::main_loop_edge): New field.
9915 (_loop_vec_info::skip_main_loop_edge): Likewise.
9916 (_loop_vec_info::skip_this_loop_edge): Likewise.
9917 (_loop_vec_info::reusable_accumulators): Likewise.
9918 (_stmt_vec_info::reduc_scalar_results): Likewise.
9919 (_stmt_vec_info::reused_accumulator): Likewise.
9920 (vect_get_main_loop_result): Declare.
9921 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize
9922 reduc_scalar_inputs.
9923 (vec_info::free_stmt_vec_info): Free reduc_scalar_inputs.
9924 * tree-vect-loop-manip.c (vect_get_main_loop_result): New function.
9925 (vect_do_peeling): Fill an epilogue loop's main_loop_edge,
9926 skip_main_loop_edge and skip_this_loop_edge fields.
9927 * tree-vect-loop.c (INCLUDE_ALGORITHM): Define.
9928 (vect_emit_reduction_init_stmts): New function.
9929 (get_initial_def_for_reduction): Use it.
9930 (get_initial_defs_for_reduction): Likewise. Change the vinfo
9931 parameter to a loop_vec_info.
9932 (vect_create_epilog_for_reduction): Store the scalar results
9933 in the reduc_info. If an epilogue loop is reusing an accumulator
9934 from the main loop, and if the epilogue loop can also be skipped,
9935 try to place the reduction code in the join block. Record
9936 accumulators that could potentially be reused by epilogue loops.
9937 (vect_transform_cycle_phi): When vectorizing epilogue loops,
9938 try to reuse accumulators from the main loop. Record the initial
9939 value in reduc_info for non-SLP reductions too.
9941 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9943 * tree-vect-loop.c (get_initial_def_for_reduction): Remove
9944 adjustment handling. Take the neutral value as an argument,
9945 in place of the code argument.
9946 (vect_transform_cycle_phi): Update accordingly. Handle the
9947 initial values of cond reductions separately from code reductions.
9948 Choose the adjustment here rather than in
9949 get_initial_def_for_reduction. Sink the splat of vec_initial_def.
9951 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9953 * tree-vect-loop.c (neutral_op_for_slp_reduction): Replace with...
9954 (neutral_op_for_reduction): ...this, providing a more general
9956 (vect_create_epilog_for_reduction): Update accordingly.
9957 (vectorizable_reduction): Likewise.
9958 (vect_transform_cycle_phi): Likewise.
9960 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9962 * tree-vect-loop.c (get_initial_def_for_reduction): Take the
9963 reduc_info instead of the original stmt_vec_info.
9964 (vect_transform_cycle_phi): Update accordingly.
9966 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9968 * tree-vect-loop.c (get_initial_defs_for_reduction): Take the
9969 reduc_info as an additional parameter.
9970 (vect_transform_cycle_phi): Update accordingly.
9972 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9974 * tree-vectorizer.h: Include tree-ssa-operands.h.
9975 (vect_phi_initial_value): New function.
9976 * tree-vect-loop.c (neutral_op_for_slp_reduction): Use it.
9977 (get_initial_defs_for_reduction, info_for_reduction): Likewise.
9978 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
9979 (vect_transform_cycle_phi, vectorizable_induction): Likewise.
9981 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9983 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
9984 the phi results to vectype after creating them. Remove later
9985 conversion code that thus becomes redundant.
9987 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9989 * tree-vect-loop.c (vect_create_epilog_for_reduction): Replace
9990 the new_phis vector with a reduc_inputs vector. Combine handling
9991 of reduction chains and ncopies > 1.
9993 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
9995 * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
9996 scalar_results to group_size elements after reducing down from
9997 N*group_size elements. Construct an array_slice of the live-out
9998 stmts and assert that there is one stmt per scalar result.
10000 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10002 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
10003 nested_in_vect_loop and use double_reduc everywhere. Remove dead
10004 assignment to "loop".
10006 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
10008 * internal-fn.c (vectorized_internal_fn_supported_p): Handle
10009 vector types first. For scalar types, consider both the preferred
10010 vector mode and the alternative vector modes.
10011 * optabs-query.c (can_vec_mask_load_store_p): Use the same
10012 structure as above, in particular using related_vector_mode
10013 for modes provided by autovectorize_vector_modes.
10015 2021-07-13 Jakub Jelinek <jakub@redhat.com>
10016 Richard Biener <rguenther@suse.de>
10018 PR tree-optimization/101419
10019 * tree-pass.h (PROP_objsz): Define.
10020 (make_pass_early_object_sizes): Declare.
10021 * passes.def (pass_all_early_optimizations): Rename pass_object_sizes
10022 there to pass_early_object_sizes, drop parameter.
10023 (pass_all_optimizations): Move pass_object_sizes right after pass_ccp,
10024 drop parameter, move pass_post_ipa_warn right after that.
10025 * tree-object-size.c (pass_object_sizes::execute): Rename to...
10026 (object_sizes_execute): ... this. Add insert_min_max_p argument.
10027 (pass_data_object_sizes): Move after object_sizes_execute.
10028 (pass_object_sizes): Likewise. In execute method call
10029 object_sizes_execute, drop set_pass_param method and insert_min_max_p
10030 non-static data member and its initializer in the ctor.
10031 (pass_data_early_object_sizes, pass_early_object_sizes,
10032 make_pass_early_object_sizes): New.
10033 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Use
10034 (cfun->curr_properties & PROP_objsz) instead of cfun->after_inlining.
10036 2021-07-13 Kito Cheng <kito.cheng@sifive.com>
10039 * config/riscv/constraints.md ("S"): Update description and remove
10041 * doc/md.texi (Machine Constraints): Document the 'S' constraints
10044 2021-07-13 Richard Biener <rguenther@suse.de>
10047 2021-07-12 Richard Biener <rguenther@suse.de>
10049 * tree-vect-slp.c (vect_slp_region): Show the number of
10050 SLP graph entries in the optimization message.
10052 2021-07-13 Michael Meissner <meissner@linux.ibm.com>
10054 * config/rs6000/altivec.md (xxspltiw_v4sf): Change local variable
10056 * config/rs6000/rs6000-protos.h (rs6000_const_f32_to_i32): Change
10057 return type to long.
10058 * config/rs6000/rs6000.c (rs6000_const_f32_to_i32): Change return
10061 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
10063 * gimple-range-fold.cc (fold_using_range::range_of_builtin_ubsan_call):
10064 Query relation between the 2 operands and use it.
10066 2021-07-12 Sergei Trofimovich <siarheit@google.com>
10068 * doc/cfg.texi: Fix s/ei_safe_safe/ei_safe_edge/ typo.
10070 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
10073 * config/i386/predicates.md (vec_setm_sse41_operand):
10074 Rename from vec_setm_operand.
10075 (vec_setm_avx2_operand): New predicate.
10076 * config/i386/sse.md (vec_set<V_128:mode>): Use V_128 mode iterator.
10077 Use vec_setm_sse41_operand as operand 2 predicate.
10078 (vec_set<V_256_512:mode): New expander.
10079 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_sse41_operand
10080 as operand 2 predicate.
10082 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
10084 PR tree-optimization/101335
10085 * range-op.cc (operator_cast::lhs_op1_relation): Delete.
10087 2021-07-12 Andrew Pinski <apinski@marvell.com>
10089 * tree-ssa-phiopt.c (match_simplify_replacement): Move
10090 insert of the sequence before the movement of the
10091 statement. Check if to see if the statement is used
10092 outside of the original phi to see if we should move it.
10094 2021-07-12 Richard Biener <rguenther@suse.de>
10096 * dump-context.h (debug_dump_context::debug_dump_context):
10097 Add FILE * parameter defaulted to stderr.
10098 * dumpfile.c (debug_dump_context::debug_dump_context): Adjust.
10099 * tree-vect-slp.c (dot_slp_tree): New functions.
10101 2021-07-12 Richard Biener <rguenther@suse.de>
10103 PR tree-optimization/101373
10104 * tree-ssa-pre.c (prune_clobbered_mems): Also prune trapping
10105 references when the BB may not return.
10106 (compute_avail): Pass in the function we're working on and
10107 replace cfun references with it. Externally throwing
10108 const calls also possibly terminate the function.
10109 (pass_pre::execute): Pass down the function we're working on.
10110 * gcse.c (compute_hash_table_work): Externally throwing
10111 const/pure calls also need record_last_mem_set_info.
10112 * postreload-gcse.c (record_opr_changes): Looping or externally
10113 throwing const/pure calls also need record_last_mem_set_info.
10115 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
10117 * recog.c (memory_address_addr_space_p): Change the type to bool.
10118 Return true/false instead of 1/0.
10119 (offsettable_memref_p): Ditto.
10120 (offsettable_nonstrict_memref_p): Ditto.
10121 (offsettable_address_addr_space_p): Ditto.
10122 Change the type of addressp indirect function to bool.
10123 * recog.h (memory_address_addr_space_p): Change the type to bool.
10124 (strict_memory_address_addr_space_p): Ditto.
10125 (offsettable_memref_p): Ditto.
10126 (offsettable_nonstrict_memref_p): Ditto.
10127 (offsettable_address_addr_space_p): Ditto.
10128 * reload.c (maybe_memory_address_addr_space_p): Ditto.
10129 (strict_memory_address_addr_space_p): Change the type to bool.
10130 Return true/false instead of 1/0.
10131 (maybe_memory_address_addr_space_p): Change the type to bool.
10133 2021-07-12 Richard Biener <rguenther@suse.de>
10135 * tree-vect-slp.c (vect_slp_region): Show the number of
10136 SLP graph entries in the optimization message.
10138 2021-07-12 Richard Biener <rguenther@suse.de>
10140 PR tree-optimization/101394
10141 * tree-ssa-pre.c (do_pre_regular_insertion): Avoid inserting
10142 copies from abnormals for a full redundancy.
10144 2021-07-12 Richard Biener <rguenther@suse.de>
10146 PR middle-end/101423
10147 * gimple.c (gimple_could_trap_p_1): Internal function calls
10149 * tree-eh.c (tree_could_trap_p): Likewise.
10151 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
10154 * config/arm/arm_neon.h (vmul_n_u32): Replace call to builtin with
10156 (vmulq_n_u32): Likewise.
10157 (vmul_n_f32): Gate __a * __b on __FAST_MATH__.
10158 (vmulq_n_f32): Likewise.
10159 (vmul_n_f16): Likewise.
10160 (vmulq_n_f16): Likewise.
10162 2021-07-12 Martin Liska <mliska@suse.cz>
10164 PR sanitizer/101425
10165 * gcc.c (check_offload_target_name): Call
10166 candidates_list_and_hint only if we have a candidate.
10168 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
10171 * config/arm/neon.md (vec_init): Move to ...
10172 * config/arm/vec-common.md (vec_init): ... here.
10173 Change the pattern's mode to VDQX and gate it on VALID_MVE_MODE.
10175 2021-07-12 Roger Sayle <roger@nextmovesoftware.com>
10177 PR tree-optimization/101403
10178 * match.pd ((T)bswap(X)>>C): Correctly handle cases where
10179 signedness of the shift is not the same as the signedness of
10180 the type extension.
10182 2021-07-09 Roger Sayle <roger@nextmovesoftware.com>
10183 Uroš Bizjak <ubizjak@gmail.com>
10185 * config/i386/i386.md (*divmodsi4_const): Optimize SImode
10186 divmod of a constant numerator with new define_insn_and_split.
10188 2021-07-09 Iain Sandoe <iain@sandoe.co.uk>
10191 * config/i386/i386-expand.c (ix86_expand_call): If a call is
10192 to a non-local-binding, or local but to a public symbol, then
10193 assume that it might be indirected via the lazy symbol binder.
10194 Mark R10 and R10 as clobbered in that case.
10196 2021-07-09 Eric Botcazou <ebotcazou@adacore.com>
10199 * gcc.c (ASM_DEBUG_DWARF_OPTION): Set again to --gdwarf2 in
10200 the case where HAVE_AS_WORKING_DWARF_N_FLAG is not defined
10201 and HAVE_LD_BROKEN_PE_DWARF5 is defined.
10203 2021-07-09 Uroš Bizjak <ubizjak@gmail.com>
10205 * config/i386/i386.md (*udivmodsi4_pow2_zext_1): Limit the
10206 log2 range of operands[3] to [1,31].
10207 (*udivmodsi4_pow2_zext_2): Ditto. Correct insn RTX pattern.
10209 2021-07-09 Sergei Trofimovich <siarheit@google.com>
10211 * doc/md.texi: Don't split @smallexample in multiple @groups.
10213 2021-07-09 Sergei Trofimovich <siarheit@google.com>
10215 * doc/md.texi: Add missing 'see' word.
10217 2021-07-09 Andrew Pinski <apinski@marvell.com>
10219 * tree-ssa-phiopt.c (phiopt_early_allow): Change arguments
10220 to take sequence and gimple_match_op. Accept the case where
10221 op is a SSA_NAME and one statement in the sequence.
10222 Also allow constants.
10223 (gimple_simplify_phiopt): Always pass a sequence to resimplify.
10224 Update call to phiopt_early_allow. Discard the sequence if not
10227 2021-07-09 Xi Ruoyao <xry111@mengyan1223.wang>
10232 * config/mips/mips.c (mips_const_insns): Use MSA_SUPPORTED_MODE_P
10233 instead of ISA_HAS_MSA.
10234 (mips_expand_vec_unpack): Likewise.
10235 (mips_expand_vector_init): Likewise.
10237 2021-07-09 Kewen Lin <linkw@linux.ibm.com>
10239 * config/rs6000/vsx.md (mods_<mode>): Rename to...
10240 (mod<mode>3): ... this.
10241 (modu_<mode>): Rename to...
10242 (umod<mode>3): ... this.
10243 * config/rs6000/rs6000-builtin.def (MODS_V2DI, MODS_V4SI, MODU_V2DI,
10244 MODU_V4SI): Adjust.
10246 2021-07-08 Jeff Law <jeffreyalaw@gmail.com>
10248 * config/h8300/shiftrotate.md (variable shifts): Expose condition
10249 code handling for the test before the loop.
10251 2021-07-08 Martin Jambor <mjambor@suse.cz>
10254 * ipa-sra.c (class isra_call_summary): New member
10255 m_before_any_store, initialize it in the constructor.
10256 (isra_call_summary::dump): Dump the new field.
10257 (ipa_sra_call_summaries::duplicate): Copy it.
10258 (process_scan_results): Set it.
10259 (isra_write_edge_summary): Stream it.
10260 (isra_read_edge_summary): Likewise.
10261 (param_splitting_across_edge): Only override
10262 safe_to_import_accesses if m_before_any_store is set.
10264 2021-07-08 Martin Sebor <msebor@redhat.com>
10266 PR bootstrap/101374
10267 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref):
10268 Use Object Size Type 0 instead of 1.
10270 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
10272 * tree-vect-loop.c (vectorizable_reduction): Remove always-true
10275 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
10277 * match.pd: Simplify an extend-operate-truncate sequence involving
10280 2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
10281 Richard Biener <rguenther@suse.de>
10283 PR tree-optimization/40210
10284 * match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
10285 (x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
10286 when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
10288 2021-07-08 Uroš Bizjak <ubizjak@gmail.com>
10291 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
10293 * config/i386/mmx.md (V_32): New mode iterator.
10294 (mov<V_32:mode>): Use V_32 mode iterator.
10295 (*mov<V_32:mode>_internal): Ditto.
10296 (*push<V_32:mode>2_rex64): Ditto.
10297 (*push<V_32:mode>2): Ditto.
10298 (movmisalign<V_32:mode>): Ditto.
10299 (mmx_<any_shiftrt:insn>v1si3): New insn pattern.
10300 (sse4_1_<any_extend:code>v2qiv2hi2): Ditto.
10301 (vec_unpacks_lo_v4qi): New expander.
10302 (vec_unpacks_hi_v4qi): Ditto.
10303 (vec_unpacku_lo_v4qi): Ditto.
10304 (vec_unpacku_hi_v4qi): Ditto.
10305 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add V1SImode.
10306 (VALID_INT_MODE_P): Ditto.
10308 2021-07-08 Michael Meissner <meissner@linux.ibm.com>
10311 * config/rs6000/rs6000.md (udivti3): New insn.
10312 (divti3): New insn.
10313 (umodti3): New insn.
10314 (modti3): New insn.
10316 2021-07-07 Martin Sebor <msebor@redhat.com>
10318 PR tree-optimization/100137
10319 PR tree-optimization/99121
10320 PR tree-optimization/97027
10321 * builtins.c (access_ref::access_ref): Also set offmax.
10322 (access_ref::offset_in_range): Define new function.
10323 (access_ref::add_offset): Set offmax.
10324 (access_ref::inform_access): Handle access_none.
10325 (handle_mem_ref): Clear ostype.
10326 (compute_objsize_r): Handle ASSERT_EXPR.
10327 * builtins.h (struct access_ref): Add offmax member.
10328 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Use
10329 compute_objsize() and simplify.
10331 2021-07-07 Peter Bergner <bergner@linux.ibm.com>
10333 * config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
10334 and VSX_BUILTIN_STXVP.
10336 2021-07-07 Martin Sebor <msebor@redhat.com>
10339 * config/aarch64/aarch64.c (aarch64_simd_lane_bounds): Remove
10340 a stray %K from error_at() missed in r12-2088.
10342 2021-07-07 Richard Biener <rguenther@suse.de>
10344 PR tree-optimization/99728
10345 * tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
10347 (mem_refs_may_alias_p): Add assert we handled aggregate
10349 (sm_seq_valid_bb): Give up when running into aggregate copies.
10350 (ref_indep_loop_p): Handle aggregate copies as never
10351 being invariant themselves but allow other refs to be
10352 disambiguated against them.
10353 (can_sm_ref_p): Do not try to apply store-motion to aggregate
10356 2021-07-06 Indu Bhagat <indu.bhagat@oracle.com>
10359 * dwarf2ctf.c (ctf_get_AT_data_member_location): Multiply by 8 to get
10362 2021-07-06 Martin Sebor <msebor@redhat.com>
10364 * gimple-pretty-print.c (percent_G_format): Remove.
10365 * tree-diagnostic.c (default_tree_printer): Remove calls.
10366 * tree-pretty-print.c (percent_K_format): Remove.
10367 * tree-pretty-print.h (percent_K_format): Remove.
10369 2021-07-06 Martin Sebor <msebor@redhat.com>
10371 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_builtin):
10372 Remove %K and use error_at.
10373 (aarch64_expand_fcmla_builtin): Same.
10374 (aarch64_expand_builtin_tme): Same.
10375 (aarch64_expand_builtin_memtag): Same.
10376 * config/arm/arm-builtins.c (arm_expand_acle_builtin): Same.
10377 (arm_expand_builtin): Same.
10378 * config/arm/arm.c (bounds_check): Same.
10380 2021-07-06 Martin Sebor <msebor@redhat.com>
10382 * builtins.c (warn_string_no_nul): Remove %G.
10383 (maybe_warn_for_bound): Same.
10384 (warn_for_access): Same.
10385 (check_access): Same.
10386 (check_strncat_sizes): Same.
10387 (expand_builtin_strncat): Same.
10388 (expand_builtin_strncmp): Same.
10389 (expand_builtin): Same.
10390 (expand_builtin_object_size): Same.
10391 (warn_dealloc_offset): Same.
10392 (maybe_emit_free_warning): Same.
10393 * calls.c (maybe_warn_alloc_args_overflow): Same.
10394 (maybe_warn_nonstring_arg): Same.
10395 (maybe_warn_rdwr_sizes): Same.
10396 * expr.c (expand_expr_real_1): Remove %K.
10397 * gimple-fold.c (gimple_fold_builtin_strncpy): Remove %G.
10398 (gimple_fold_builtin_strncat): Same.
10399 * gimple-ssa-sprintf.c (format_directive): Same.
10400 (handle_printf_call): Same.
10401 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Same.
10402 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
10403 (maybe_diag_access_bounds): Same. Call gimple_location.
10404 (check_bounds_or_overlap): Same.
10405 * trans-mem.c (ipa_tm_scan_irr_block): Remove %K. Simplify.
10406 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Remove %G.
10407 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
10408 (maybe_diag_stxncpy_trunc): Same.
10409 (handle_builtin_stxncpy_strncat): Same.
10410 (maybe_warn_pointless_strcmp): Same.
10411 * tree-ssa-uninit.c (maybe_warn_operand): Same.
10413 2021-07-06 Uroš Bizjak <ubizjak@gmail.com>
10416 * config/i386/predicates.md (vec_setm_operand): Enable
10417 register_operand for TARGET_SSE4_1.
10418 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_operand
10419 as operand 2 predicate. Call ix86_expand_vector_set_var
10420 for non-constant index operand.
10421 (vec_setv4qi): Use vec_setm_mmx_operand as operand 2 predicate.
10422 Call ix86_expand_vector_set_var for non-constant index operand.
10424 2021-07-06 Jeff Law <jeffreyalaw@gmail.com>
10426 * config/h8300/jumpcall.md (*branch): When possible, generate
10427 the comparison in CCZN mode.
10428 * config/h8300/predicates.md (simple_memory_operand): Reject all
10429 auto-increment addressing modes.
10431 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
10433 PR bootstrap/100246
10434 * config/i386/i386.h (struct stringop_algs): Define a CTOR for
10437 2021-07-06 Richard Biener <rguenther@suse.de>
10439 * doc/md.texi (vec_fmaddsub<mode>4): Document.
10440 (vec_fmsubadd<mode>4): Likewise.
10441 * optabs.def (vec_fmaddsub$a4): Add.
10442 (vec_fmsubadd$a4): Likewise.
10443 * internal-fn.def (IFN_VEC_FMADDSUB): Add.
10444 (IFN_VEC_FMSUBADD): Likewise.
10445 * tree-vect-slp-patterns.c (addsub_pattern::recognize):
10446 Refactor to handle IFN_VEC_FMADDSUB and IFN_VEC_FMSUBADD.
10447 (addsub_pattern::build): Likewise.
10448 * tree-vect-slp.c (vect_optimize_slp): CFN_VEC_FMADDSUB
10449 and CFN_VEC_FMSUBADD are not transparent for permutes.
10450 * config/i386/sse.md (vec_fmaddsub<mode>4): New expander.
10451 (vec_fmsubadd<mode>4): Likewise.
10453 2021-07-06 Richard Biener <rguenther@suse.de>
10455 * doc/invoke.texi (fmove-loop-stores): Document.
10456 * common.opt (fmove-loop-stores): New option.
10457 * opts.c (default_options_table): Enable -fmove-loop-stores
10458 at -O1 but not -Og.
10459 * tree-ssa-loop-im.c (pass_lim::execute): Pass
10460 flag_move_loop_stores instead of true to
10461 loop_invariant_motion_in_fun.
10463 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
10465 * doc/install.texi: Document --with-dsymutil.
10467 2021-07-06 Andrew Pinski <apinski@marvell.com>
10469 PR tree-optimization/101256
10470 * dbgcnt.def (phiopt_edge_range): New counter.
10471 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
10472 Check to make sure the new name is defined in the same
10473 bb as the conditional before duplicating range info.
10474 Also add debug counter.
10476 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
10478 PR rtl-optimization/100328
10479 * config/i386/i386-options.c (ix86_option_override_internal):
10480 Set param_ira_consider_dup_in_all_alts to 0.
10482 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
10484 PR rtl-optimization/100328
10485 * doc/invoke.texi (ira-consider-dup-in-all-alts): Document new
10487 * ira.c (ira_get_dup_out_num): Adjust as parameter
10488 param_ira_consider_dup_in_all_alts.
10489 * params.opt (ira-consider-dup-in-all-alts): New.
10490 * ira-conflicts.c (process_regs_for_copy): Add one parameter
10491 single_input_op_has_cstr_p.
10492 (get_freq_for_shuffle_copy): New function.
10493 (add_insn_allocno_copies): Adjust as single_input_op_has_cstr_p.
10494 * ira-int.h (ira_get_dup_out_num): Add one bool parameter.
10496 2021-07-05 Jeff Law <jeffreyalaw@gmail.com>
10498 * config/h8300/shiftrotate.md (shift-by-variable patterns): Update to
10499 generate condition code aware RTL directly.
10501 2021-07-05 Andrew Pinski <apinski@marvell.com>
10503 PR tree-optimization/101039
10504 * match.pd (A CMP 0 ? A : -A): New patterns.
10505 * tree-ssa-phiopt.c (abs_replacement): Delete function.
10506 (tree_ssa_phiopt_worker): Don't call abs_replacement.
10507 Update comment about abs_replacement.
10509 2021-07-05 Andrew Pinski <apinski@marvell.com>
10511 * tree-ssa-phiopt.c (gimple_simplify_phiopt):
10512 If "A ? B : C" fails to simplify, try "(!A) ? C : B".
10514 2021-07-05 Andrew Pinski <apinski@marvell.com>
10516 * tree-ssa-phiopt.c (match_simplify_replacement):
10517 Add early_p argument. Call gimple_simplify_phiopt
10518 instead of gimple_simplify.
10519 (tree_ssa_phiopt_worker): Update call to
10520 match_simplify_replacement and allow unconditionally.
10521 (phiopt_early_allow): New function.
10522 (gimple_simplify_phiopt): New function.
10524 2021-07-05 Andrew Pinski <apinski@marvell.com>
10526 PR middle-end/101237
10527 * fold-const.c (negate_expr_p): Remove call to element_mode
10528 and TREE_MODE/TREE_TYPE when calling HONOR_SIGNED_ZEROS,
10529 HONOR_SIGN_DEPENDENT_ROUNDING, and HONOR_SNANS.
10530 (fold_negate_expr_1): Likewise.
10531 (const_unop): Likewise.
10532 (fold_cond_expr_with_comparison): Likewise.
10533 (fold_binary_loc): Likewise.
10534 (fold_ternary_loc): Likewise.
10535 (tree_call_nonnegative_warnv_p): Likewise.
10536 * match.pd (-(A + B) -> (-B) - A): Likewise.
10538 2021-07-05 Iain Sandoe <iain@sandoe.co.uk>
10540 * configure.ac: Handle --with-dsymutil in the same way as we
10541 do for the assembler and linker. (DEFAULT_DSYMUTIL): New.
10542 Extract the type and version for the dsymutil configured or
10543 found by the default searches.
10544 * config.in: Regenerated.
10545 * configure: Regenerated.
10546 * collect2.c (do_dsymutil): Handle locating dsymutil in the
10547 same way as for the assembler and linker.
10548 * config/darwin.h (DSYMUTIL): Delete.
10549 * gcc.c: Report a configured dsymutil correctly.
10550 * exec-tool.in: Allow for dsymutil.
10552 2021-07-05 Uroš Bizjak <ubizjak@gmail.com>
10554 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
10555 Handle V4QI and V2HI modes.
10556 (expand_vec_perm_blend): Allow 4-byte vector modes with TARGET_SSE4_1.
10557 Handle V4QI mode. Emit mmx_pblendvb32 for 4-byte modes.
10558 (expand_vec_perm_pshufb): Rewrite to use switch statemets.
10559 Handle 4-byte dual operands with TARGET_XOP and single operands
10560 with TARGET_SSSE3. Emit mmx_ppermv32 for TARGET_XOP and
10561 mmx_pshufbv4qi3 for TARGET_SSSE3.
10562 (expand_vec_perm_pblendv): Allow 4-byte vector modes with TARGET_SSE4_1.
10563 (expand_vec_perm_interleave2): Allow 4-byte vector modes.
10564 (expand_vec_perm_pshufb2): Allow 4-byte vector modes with TARGET_SSSE3.
10565 (expand_vec_perm_even_odd_1): Handle V4QI mode.
10566 (expand_vec_perm_broadcast_1): Handle V4QI mode.
10567 (ix86_vectorize_vec_perm_const): Handle V4QI mode.
10568 * config/i386/mmx.md (mmx_ppermv32): New insn pattern.
10569 (mmx_pshufbv4qi3): Ditto.
10570 (*mmx_pblendw32): Ditto.
10571 (*mmx_pblendw64): Rename from *mmx_pblendw.
10572 (mmx_punpckhbw_low): New insn_and_split pattern.
10573 (mmx_punpcklbw_low): Ditto.
10575 2021-07-05 Richard Biener <rguenther@suse.de>
10577 * tree-vect-loop-manip.c (vect_loop_versioning): Do not
10578 set LOOP_C_INFINITE on the vectorized loop.
10580 2021-07-05 Richard Biener <rguenther@suse.de>
10582 PR middle-end/101291
10583 * cfgloopmanip.c (loop_version): Set the loop copy of the
10584 versioned loop to the new loop.
10586 2021-07-04 Iain Sandoe <iain@sandoe.co.uk>
10589 * config.gcc: Ensure that Darwin biarch definitions are
10590 added before i386.h.
10591 * config/i386/darwin.h (TARGET_64BIT): Remove.
10592 (PR80556_WORKAROUND): New.
10593 (REAL_LIBGCC_SPEC): Amend to use PR80556_WORKAROUND.
10594 (DARWIN_SUBARCH_SPEC): New.
10595 * config/i386/darwin32-biarch.h (TARGET_64BIT_DEFAULT,
10596 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
10597 (REAL_LIBGCC_SPEC): Remove.
10598 * config/i386/darwin64-biarch.h (TARGET_64BIT_DEFAULT,
10599 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
10600 (REAL_LIBGCC_SPEC): Remove.
10602 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
10604 PR middle-end/101294
10605 * expr.c (store_constructor): Don't use vec_duplicate on vector.
10607 2021-07-02 Martin Sebor <msebor@redhat.com>
10609 PR middle-end/98871
10610 PR middle-end/98512
10611 * diagnostic.c (get_any_inlining_info): New.
10612 (update_effective_level_from_pragmas): Handle inlining context.
10613 (diagnostic_enabled): Same.
10614 (diagnostic_report_diagnostic): Same.
10615 * diagnostic.h (struct diagnostic_info): Add ctor.
10616 (struct diagnostic_context): Add new member.
10617 * tree-diagnostic.c (set_inlining_locations): New.
10618 (tree_diagnostics_defaults): Set new callback pointer.
10620 2021-07-02 Peter Bergner <bergner@linux.ibm.com>
10622 * config/rs6000/rs6000-builtin.def (BU_MMA_PAIR_LD, BU_MMA_PAIR_ST):
10624 (__builtin_vsx_lxvp, __builtin_vsx_stxvp): New built-ins.
10625 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Expand
10626 lxvp and stxvp built-ins.
10627 (mma_init_builtins): Handle lxvp and stxvp built-ins.
10628 (builtin_function_type): Likewise.
10629 * doc/extend.texi (__builtin_vsx_lxvp, __builtin_mma_stxvp): Document.
10631 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10633 * config/h8300/h8300-protos.h (compute_a_shift_cc): Accept
10634 additional argument for the code.
10635 * config/h8300/h8300.c (compute_a_shift_cc): Accept additional
10636 argument for the code. Just return if the ZN bits are useful or
10637 not rather than the old style CC_* enums.
10638 * config/h8300/shiftrotate.md (shiftqi_noscratch): Move before
10639 more generic shiftqi patterns.
10640 (shifthi_noscratch, shiftsi_noscratch): Similarly.
10641 (shiftqi_noscratch_set_flags): New pattern.
10642 (shifthi_noscratch_set_flags, shiftsi_noscratch_set_flags): Likewise.
10644 2021-07-02 Andrew MacLeod <amacleod@redhat.com>
10646 PR tree-optimization/101223
10647 * range-op.cc (build_lt): Add -1 for signed values.
10648 (built_gt): Subtract -1 for signed values.
10650 2021-07-02 David Faust <david.faust@oracle.com>
10652 * btfout.c (get_btf_kind): Support BTF_KIND_FLOAT.
10653 (btf_asm_type): Likewise.
10655 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10657 * config/h8300/h8300-protos.h (output_a_shift): Make first argument
10658 an array of rtx rather than a pointer to rtx. Add code argument.
10659 (compute_a_shift_length): Similarly.
10660 * config/h8300/h8300.c (h8300_shift_costs): Adjust now that the
10661 shift itself isn't an operand. Create dummy operand[0] to carry
10662 a mode and pass a suitable rtx code to compute_a_shift_length.
10663 (get_shift_alg): Adjust operand number of clobber in output templates.
10664 (output_a_shift): Make first argument an array of rtx rather than
10665 a pointer to rtx. Add code argument for the type of shift.
10666 Adjust now that the shift itself is no longer an operand.
10667 (compute_a_shift_length): Similarly.
10668 * config/h8300/shiftrotate.md (shiftqi, shifthi, shiftsi): Use an
10669 iterator rather than nshift_operator.
10670 (shiftqi_noscratch, shifthi_noscratch, shiftsi_noscratch): Likewise.
10671 (shiftqi_clobber_flags): Adjust to API changes in output_a_shift
10672 and compute_a_shift_length.
10673 (shiftqi_noscratch_clobber_flags): Likewise.
10674 (shifthi_noscratch_clobber_flags): Likewise.
10675 (shiftsi_noscratch_clobber_flags): Likewise.
10677 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
10680 * config/darwin.h (DSYMUTIL_SPEC): Do not try to run
10681 dsymutil for BTF/CTF.
10683 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
10686 * config/darwin.h (CTF_INFO_SECTION_NAME): Update the
10687 segment to include BTF.
10688 (BTF_INFO_SECTION_NAME): New.
10690 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10692 * config/m32r/m32r-protos.h (call_operand): Adjust return type.
10693 (small_data_operand, memreg_operand, small_insn_p): Likewise.
10694 * config/m32r/m32r.c (call_operand): Adjust return type.
10695 (small_data_operand, memreg_operand): Likewise.
10697 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10699 * config/frv/frv-protos.h (integer_register_operand): Adjust return
10701 (frv_load_operand, gpr_or_fpr_operand, gpr_no_subreg_operand): Likewise.
10702 (fpr_or_int6_operand, gpr_or_int_operand); Likewise.
10703 (gpr_or_int12_operand, gpr_or_int10_operand); Likewise.
10704 (move_source_operand, move_destination_operand): Likewise.
10705 (condexec_source_operand, condexec_dest_operand): Likewise.
10706 (lr_operand, gpr_or_memory_operand, fpr_or_memory_operand): Likewise.
10707 (reg_or_0_operand, fcc_operand, icc_operand, cc_operand): Likewise.
10708 (fcr_operand, icr_operand, cr_operand, call_operand): Likewise.
10709 (fpr_operand, even_reg_operand, odd_reg_operand): Likewise.
10710 (even_gpr_operand, odd_gpr_operand, quad_fpr_operand): Likewise.
10711 (even_fpr_operand, odd_fpr_operand): Likewise.
10712 (dbl_memory_one_insn_operand, dbl_memory_two_insn_operand): Likewise.
10713 (int12_operand, int6_operand, int5_operand, uint5_operand): Likewise.
10714 (uint4_operand, uint1_operand, int_2word_operand): Likewise
10715 (upper_int16_operand, uint16_operand, symbolic_operand): Likewise.
10716 (relational_operator, float_relational_operator): Likewise.
10717 (ccr_eqne_operator, minmax_operator): Likewise.
10718 (condexec_si_binary_operator, condexec_si_media_operator): Likewise.
10719 (condexec_si_divide_operator, condexec_si_unary_operator): Likewise.
10720 (condexec_sf_conv_operator, condexec_sf_add_operator): Likewise.
10721 (intop_compare_operator, acc_operand, even_acc_operand): Likewise.
10722 (quad_acc_operand, accg_operand): Likewise.
10724 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
10726 * config/stormy16/stormy16-protos.h (xstormy16_below_100_symbol): Change
10727 return type to a bool.
10728 (nonimmediate_nonstack_operand): Likewise.
10729 (xstormy16_splittable_below100_operand): Likewise.
10730 * config/stormy16/stormy16.c (xstormy16_below_100_symbol): Fix
10732 (xstormy16_splittable_below100_operand): Likewise.
10734 2021-07-02 Richard Biener <rguenther@suse.de>
10736 PR tree-optimization/101293
10737 * tree-ssa-loop-im.c (mem_ref_hasher::equal): Compare MEM_REF bases
10738 with combined offsets.
10739 (gather_mem_refs_stmt): Hash MEM_REFs as if their offset were
10740 combined with the rest of the offset.
10742 2021-07-02 Eric Botcazou <ebotcazou@adacore.com>
10744 * config/i386/i386.c (asm_preferred_eh_data_format): Always use the
10745 PIC encodings for PE-COFF targets.
10747 2021-07-02 Jakub Jelinek <jakub@redhat.com>
10750 * config/i386/i386-expand.c (ix86_broadcast_from_integer_constant):
10751 Return nullptr for TImode inner mode.
10753 2021-07-02 Richard Biener <rguenther@suse.de>
10755 PR tree-optimization/101280
10756 PR tree-optimization/101173
10757 * gimple-loop-interchange.cc
10758 (tree_loop_interchange::valid_data_dependences): Properly
10759 guard all dependence checks with DDR_REVERSED_P or its
10762 2021-07-02 Hongyu Wang <hongyu.wang@intel.com>
10764 * config/i386/i386-expand.c (ix86_expand_builtin):
10765 Add branch to clear odata when ZF is set for asedecenc_expand
10766 and wideaesdecenc_expand.
10768 2021-07-02 Eugene Rozenfeld <erozen@microsoft.com>
10770 * config/i386/gcc-auto-profile: regenerate
10772 2021-07-02 liuhongt <hongtao.liu@intel.com>
10774 * config/i386/sse.md (trunc<mode><pmov_dst_4>2): Refined to ..
10775 (trunc<mode><pmov_dst_4_lower>2): this.
10777 2021-07-01 David Malcolm <dmalcolm@redhat.com>
10779 * diagnostic.h (diagnostic_context::m_file_cache): New field.
10780 * input.c (class fcache): Rename to...
10781 (class file_cache_slot): ...this, making most members private and
10782 prefixing fields with "m_".
10783 (file_cache_slot::get_file_path): New accessor.
10784 (file_cache_slot::get_use_count): New accessor.
10785 (file_cache_slot::missing_trailing_newline_p): New accessor.
10786 (file_cache_slot::inc_use_count): New.
10787 (fcache_buffer_size): Move to...
10788 (file_cache_slot::buffer_size): ...here.
10789 (fcache_line_record_size): Move to...
10790 (file_cache_slot::line_record_size): ...here.
10791 (fcache_tab): Delete, in favor of global_dc->m_file_cache.
10792 (fcache_tab_size): Move to file_cache::num_file_slots.
10793 (diagnostic_file_cache_init): Update for move of fcache_tab
10794 to global_dc->m_file_cache.
10795 (diagnostic_file_cache_fini): Likewise.
10796 (lookup_file_in_cache_tab): Convert to...
10797 (file_cache::lookup_file): ...this.
10798 (diagnostics_file_cache_forcibly_evict_file): Update for move of
10799 fcache_tab to global_dc->m_file_cache, moving most of
10800 implementation to...
10801 (file_cache::forcibly_evict_file): ...this new function and...
10802 (file_cache_slot::evict): ...this new function.
10803 (evicted_cache_tab_entry): Convert to...
10804 (file_cache::evicted_cache_tab_entry): ...this.
10805 (add_file_to_cache_tab): Convert to...
10806 (file_cache::add_file): ...this, moving bulk of implementation
10808 (file_cache_slot::create): ..this new function.
10809 (file_cache::file_cache): New.
10810 (file_cache::~file_cache): New.
10811 (lookup_or_add_file_to_cache_tab): Convert to...
10812 (file_cache::lookup_or_add_file): ..this new function.
10813 (fcache::fcache): Rename to...
10814 (file_cache_slot::file_cache_slot): ...this, adding "m_" prefixes
10816 (fcache::~fcache): Rename to...
10817 (file_cache_slot::~file_cache_slot): ...this, adding "m_" prefixes
10819 (needs_read): Convert to...
10820 (file_cache_slot::needs_read_p): ...this.
10821 (needs_grow): Convert to...
10822 (file_cache_slot::needs_grow_p): ...this.
10823 (maybe_grow): Convert to...
10824 (file_cache_slot::maybe_grow): ...this.
10825 (read_data): Convert to...
10826 (file_cache_slot::read_data): ...this.
10827 (maybe_read_data): Convert to...
10828 (file_cache_slot::maybe_read_data): ...this.
10829 (get_next_line): Convert to...
10830 (file_cache_slot::get_next_line): ...this.
10831 (goto_next_line): Convert to...
10832 (file_cache_slot::goto_next_line): ...this.
10833 (read_line_num): Convert to...
10834 (file_cache_slot::read_line_num): ...this.
10835 (location_get_source_line): Update for moving of globals to
10836 global_dc->m_file_cache.
10837 (location_missing_trailing_newline): Likewise.
10838 * input.h (class file_cache_slot): New forward decl.
10839 (class file_cache): New.
10841 2021-07-01 Michael Meissner <meissner@linux.ibm.com>
10843 * config/rs6000/rs6000.c (rs6000_maybe_emit_fp_cmove): Add IEEE
10844 128-bit floating point conditional move support.
10845 (have_compare_and_set_mask): Add IEEE 128-bit floating point
10847 * config/rs6000/rs6000.md (mov<mode>cc, IEEE128 iterator): New insn.
10848 (mov<mode>cc_p10, IEEE128 iterator): New insn.
10849 (mov<mode>cc_invert_p10, IEEE128 iterator): New insn.
10850 (fpmask<mode>, IEEE128 iterator): New insn.
10851 (xxsel<mode>, IEEE128 iterator): New insn.
10853 2021-07-01 Iain Sandoe <iain@sandoe.co.uk>
10856 * config/darwin.h (CTF_INFO_SECTION_NAME): New.
10858 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
10860 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
10862 * config/i386/i386-protos.h (ix86_expand_vector_init_duplicate):
10864 * config/i386/sse.md (INT_BROADCAST_MODE): New mode iterator.
10865 (vec_duplicate<mode>): New expander.
10867 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
10870 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
10872 (ix86_byte_broadcast): New function.
10873 (ix86_convert_const_wide_int_to_broadcast): Likewise.
10874 (ix86_expand_move): Convert CONST_WIDE_INT to broadcast if mode
10875 size is 16 bytes or bigger.
10876 (ix86_broadcast_from_integer_constant): New function.
10877 (ix86_expand_vector_move): Convert CONST_WIDE_INT and CONST_VECTOR
10878 to broadcast if mode size is 16 bytes or bigger.
10879 * config/i386/i386-protos.h (ix86_gen_scratch_sse_rtx): New
10881 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): New function.
10883 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10885 * config/i386/predicates.md (ix86_endbr_immediate_operand):
10886 Return true/false instead of 1/0.
10887 (movq_parallel): Ditto.
10889 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10891 * recog.c (general_operand): Return true/false instead of 1/0.
10892 (register_operand): Ditto.
10893 (immediate_operand): Ditto.
10894 (const_int_operand): Ditto.
10895 (const_scalar_int_operand): Ditto.
10896 (const_double_operand): Ditto.
10897 (push_operand): Ditto.
10898 (pop_operand): Ditto.
10899 (memory_operand): Ditto.
10900 (indirect_operand): Ditto.
10902 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10904 * genpreds.c (write_predicate_subfunction):
10905 Change the type of written subfunction to bool.
10906 (write_one_predicate_function):
10907 Change the type of written function to bool.
10908 (write_tm_preds_h): Ditto.
10909 * recog.h (*insn_operand_predicate_fn): Change the type to bool.
10910 * recog.c (general_operand): Change the type to bool.
10911 (address_operand): Ditto.
10912 (register_operand): Ditto.
10913 (pmode_register_operand): Ditto.
10914 (scratch_operand): Ditto.
10915 (immediate_operand): Ditto.
10916 (const_int_operand): Ditto.
10917 (const_scalar_int_operand): Ditto.
10918 (const_double_operand): Ditto.
10919 (nonimmediate_operand): Ditto.
10920 (nonmemory_operand): Ditto.
10921 (push_operand): Ditto.
10922 (pop_operand): Ditto.
10923 (memory_operand): Ditto.
10924 (indirect_operand): Ditto.
10925 (ordered_comparison_operator): Ditto.
10926 (comparison_operator): Ditto.
10927 * config/i386/i386-expand.c (ix86_expand_sse_cmp):
10928 Change the type of indirect predicate function to bool.
10929 * config/rs6000/rs6000.c (easy_vector_constant):
10930 Change the type to bool.
10931 * config/mips/mips-protos.h (m16_based_address_p):
10932 Change the type of operand 3 to bool.
10934 2021-07-01 Richard Biener <rguenther@suse.de>
10936 PR tree-optimization/101280
10937 PR tree-optimization/101173
10938 * gimple-loop-interchange.cc
10939 (tree_loop_interchange::valid_data_dependences): Revert
10940 previous change and instead correctly handle DDR_REVERSED_P
10943 2021-07-01 Richard Biener <rguenther@suse.de>
10945 PR tree-optimization/101278
10946 * tree-ssa-dse.c (dse_classify_store): First check for
10947 uses, then ignore stmt for chaining purposes.
10949 2021-07-01 Richard Biener <rguenther@suse.de>
10951 PR tree-optimization/100778
10952 * tree-vect-slp.c (vect_schedule_slp_node): Do not place trapping
10953 vectorized ops ahead of their scalar BB.
10955 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
10958 * config/i386/i386.md (*nabs<dwi>2_doubleword):
10959 New insn_and_split pattern.
10960 (*nabs<dwi>2_1): Ditto.
10961 * config/i386/i386-features.c
10962 (general_scalar_chain::compute_convert_gain):
10963 Handle (NEG (ABS (...))) RTX. Rewrite src code
10964 scanner as switch statement.
10965 (general_scalar_chain::convert_insn):
10966 Handle (NEG (ABS (...))) RTX.
10967 (general_scalar_to_vector_candidate_p):
10968 Detect (NEG (ABS (...))) RTX. Reorder case statements
10969 for (AND (NOT (...) ...)) fallthrough.
10971 2021-07-01 Richard Biener <rguenther@suse.de>
10973 PR tree-optimization/101178
10974 * tree-vect-slp.c (slpg_vertex::materialize): Remove.
10975 (slpg::perm_in): Add.
10976 (slpg::get_perm_in): Remove.
10977 (slpg::get_perm_materialized): Add.
10978 (vect_optimize_slp): Handle VEC_PERM nodes more optimally
10979 during permute propagation and materialization.
10981 2021-07-01 Jakub Jelinek <jakub@redhat.com>
10984 * dwarf2out.c (loc_list_from_tree_1): Handle COMPOUND_LITERAL_EXPR.
10986 2021-07-01 Jakub Jelinek <jakub@redhat.com>
10988 PR middle-end/94366
10989 * omp-low.c (lower_rec_input_clauses): Rename is_fp_and_or to
10990 is_truth_op, set it for TRUTH_*IF_EXPR regardless of new_var's type,
10991 use boolean_type_node instead of integer_type_node as NE_EXPR type.
10992 (lower_reduction_clauses): Likewise.
10994 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
10996 * config/gcn/gcn.c: Include dwarf2.h.
10997 (gcn_addr_space_debug): New function.
10998 (TARGET_ADDR_SPACE_DEBUG): New hook.
11000 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
11002 * common/config/gcn/gcn-common.c
11003 (gcn_option_optimization_table): Change OPT_fomit_frame_pointer to -O3.
11004 * config/gcn/gcn.c (gcn_expand_prologue): Prefer the frame pointer
11006 (gcn_expand_prologue): Prefer the frame pointer when emitting CFI.
11007 (gcn_frame_pointer_rqd): New function.
11008 (TARGET_FRAME_POINTER_REQUIRED): New hook.
11010 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
11012 * config/gcn/gcn.c (move_callee_saved_registers): Emit CFI notes for
11013 prologue register saves.
11014 (gcn_debug_unwind_info): Use UI_DWARF2.
11015 (gcn_dwarf_register_number): Map DWARF_LINK_REGISTER to DWARF PC.
11016 (gcn_dwarf_register_span): DWARF_LINK_REGISTER doesn't span.
11017 * config/gcn/gcn.h: (DWARF_FRAME_RETURN_COLUMN): New define.
11018 (DWARF_LINK_REGISTER): New define.
11019 (FIRST_PSEUDO_REGISTER): Increment.
11020 (FIXED_REGISTERS): Add entry for DWARF_LINK_REGISTER.
11021 (CALL_USED_REGISTERS): Likewise.
11022 (REGISTER_NAMES): Likewise.
11024 2021-06-30 Richard Biener <rguenther@suse.de>
11026 PR tree-optimization/101267
11027 * tree-vect-stmts.c (vect_check_scalar_mask): Adjust
11028 API and use SLP compatible interface of vect_is_simple_use.
11029 Reject not vectorized SLP defs for callers that do not support
11031 (vect_check_store_rhs): Handle masked stores and pass down
11032 the appropriate operator index.
11033 (vectorizable_call): Adjust.
11034 (vectorizable_store): Likewise.
11035 (vectorizable_load): Likewise. Handle SLP pecularity of
11037 (vect_is_simple_use): Remove special-casing of masked stores.
11039 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
11041 * common.opt (foffload): Remove help as Driver only.
11042 * gcc.c (display_help): Add -foffload.
11044 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
11046 * gcc.c (close_at_file, execute): Replace alloca by XALLOCAVEC.
11047 (check_offload_target_name): Fix splitting OFFLOAD_TARGETS into
11048 a candidate list; better inform no offload target is configured
11049 and fix hint extraction when passed target is not '\0' at [len].
11050 * common.opt (foffload): Add tailing '.'.
11051 (foffload-options): Likewise; fix flag name in the help string.
11053 2021-06-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
11056 * config/arm/arm_neon.h: Move vabs intrinsics before vcage_f32.
11057 (vcage_f32): Gate comparison on __FAST_MATH__.
11058 (vcageq_f32): Likewise.
11059 (vcale_f32): Likewise.
11060 (vcaleq_f32): Likewise.
11061 (vcagt_f32): Likewise.
11062 (vcagtq_f32): Likewise.
11063 (vcalt_f32): Likewise.
11064 (vcaltq_f32): Likewise.
11065 (vcage_f16): Likewise.
11066 (vcageq_f16): Likewise.
11067 (vcale_f16): Likewise.
11068 (vcaleq_f16): Likewise.
11069 (vcagt_f16): Likewise.
11070 (vcagtq_f16): Likewise.
11071 (vcalt_f16): Likewise.
11072 (vcaltq_f16): Likewise.
11074 2021-06-30 Richard Biener <rguenther@suse.de>
11076 PR tree-optimization/101264
11077 * tree-vect-slp.c (vect_optimize_slp): Propagate the
11078 computed perm_in to all "any" permute successors
11079 we cannot de-duplicate immediately.
11081 2021-06-30 liuhongt <hongtao.liu@intel.com>
11084 * config/i386/sse.md
11085 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>):
11087 (avx512f_sfixupimm<mode><maskz_scalar_name><round_saeonly_name>):
11089 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>"): Refined.
11090 * config/i386/subst.md (maskz_scalar): New define_subst.
11091 (maskz_scalar_name): New subst_attr.
11092 (maskz_scalar_op5): Ditto.
11093 (round_saeonly_maskz_scalar_op5): Ditto.
11094 (round_saeonly_maskz_scalar_operand5): Ditto.
11096 2021-06-30 David Edelsohn <dje.gcc@gmail.com>
11098 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
11099 Increase code CSECT alignment to at least 32 bytes.
11100 * config/rs6000/xcoff.h (TEXT_SECTION_ASM_OP): Add 32 byte
11101 alignment designation.
11103 2021-06-29 Sergei Trofimovich <siarheit@google.com>
11105 * doc/generic.texi: Fix s/net yet/not yet/ typo.
11107 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
11109 PR tree-optimization/101254
11110 * range-op.cc (operator_minus::op1_op2_relation_effect): Check for
11111 wrapping/non-wrapping when setting the result range.
11113 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
11115 * value-query.cc (gimple_range_global): Allow phis.
11117 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
11119 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): Use stmt.
11120 (simplify_using_ranges::op_with_boolean_value_range_p): Add a
11121 statement for location context.
11122 (check_for_binary_op_overflow): Ditto.
11123 (simplify_using_ranges::get_vr_for_comparison): Ditto.
11124 (simplify_using_ranges::compare_name_with_value): Ditto.
11125 (simplify_using_ranges::compare_names): Ditto.
11126 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Ditto.
11127 (simplify_using_ranges::simplify_truth_ops_using_ranges): Ditto.
11128 (simplify_using_ranges::simplify_min_or_max_using_ranges): Ditto.
11129 (simplify_using_ranges::simplify_internal_call_using_ranges): Ditto.
11130 (simplify_using_ranges::two_valued_val_range_p): Ditto.
11131 (simplify_using_ranges::simplify): Ditto.
11132 * vr-values.h: Adjust prototypes.
11134 2021-06-29 Uroš Bizjak <ubizjak@gmail.com>
11137 * config/i386/mmx.md (vec_addsubv2sf3): New insn pattern.
11139 2021-06-29 Julian Brown <julian@codesourcery.com>
11141 * config/gcn/gcn.c (gcn_init_libfuncs): New function.
11142 (TARGET_INIT_LIBFUNCS): Define target hook using above function.
11143 * config/gcn/gcn.h (UNITS_PER_WORD): Define to 8 for IN_LIBGCC2, 4
11145 (LIBGCC2_UNITS_PER_WORD, BITS_PER_WORD): Remove definitions.
11146 (MAX_FIXED_MODE_SIZE): Change to 128.
11148 2021-06-29 Julian Brown <julian@codesourcery.com>
11150 * config/gcn/gcn.md (UNSPEC_FLBIT_INT): New unspec constant.
11151 (s_mnemonic): Add clrsb.
11152 (gcn_flbit<mode>_int): Add insn pattern for SImode/DImode.
11153 (clrsb<mode>2): Add expander for SImode/DImode.
11155 2021-06-29 Julian Brown <julian@codesourcery.com>
11157 * config/gcn/gcn.md (<su>mulsidi3, <su>mulsidi3_reg, <su>mulsidi3_imm,
11158 muldi3): Add patterns.
11160 2021-06-29 Julian Brown <julian@codesourcery.com>
11162 * config/gcn/gcn.md (<su>mulsi3_highpart): Change to expander.
11163 (<su>mulsi3_highpart_reg, <su>mulsi3_highpart_imm): New patterns.
11165 2021-06-29 Julian Brown <julian@codesourcery.com>
11167 * config/gcn/gcn.md (mulsi3): Make s_mulk_i32 variant clobber SCC.
11169 2021-06-29 Joseph Myers <joseph@codesourcery.com>
11171 * btfout.c, ctfout.c: Include "memmodel.h".
11173 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
11175 * gcc.c (check_offload_target_name): Cast len argument to
11176 %q.*s to 'int'; avoid -Wstringop-truncation warning.
11178 2021-06-29 Richard Biener <rguenther@suse.de>
11180 * tree-vect-slp.c (vect_optimize_slp): Forward propagate
11181 to "any" permute nodes and relax "any" permute proapgation
11182 during iterative backward propagation.
11184 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
11187 * common.opt (-foffload=): Update description.
11188 (-foffload-options=): New.
11189 * doc/invoke.texi (C Language Options): Document
11190 -foffload and -foffload-options.
11191 * gcc.c (check_offload_target_name): New, split off from
11192 handle_foffload_option.
11193 (check_foffload_target_names): New.
11194 (handle_foffload_option): Handle -foffload=default.
11195 (driver_handle_option): Update for -foffload-options.
11196 * lto-opts.c (lto_write_options): Use -foffload-options
11197 instead of -foffload.
11198 * lto-wrapper.c (merge_and_complain, append_offload_options):
11200 * opts.c (common_handle_option): Likewise.
11202 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
11204 * doc/invoke.texi (C Language Options): Sort options
11205 alphabetically in optlist and also the description itself.
11206 Remove leftover -fallow-single-precision from and add missing
11207 -fgnu-tm to the optlist.
11209 2021-06-29 Richard Biener <rguenther@suse.de>
11211 * tree-vect-slp.c (slpg_vertex::visited): Remove.
11212 (vect_slp_perms_eq): Handle -1 permutes.
11213 (vect_optimize_slp): Rewrite permute propagation.
11215 2021-06-29 Jakub Jelinek <jakub@redhat.com>
11218 * match.pd ((intptr_t)x eq/ne CST to x eq/ne (typeof x) CST): Don't
11219 perform the optimization in GENERIC when sanitizing and x has a
11222 2021-06-29 Richard Biener <rguenther@suse.de>
11224 PR tree-optimization/101242
11225 * tree-vect-slp.c (vect_slp_build_vertices): Force-add
11226 PHIs with not represented initial values as leafs.
11228 2021-06-29 Jan-Benedict Glaw <jbglaw@getslash.de>
11230 * config/pdp11/pdp11.h (ASM_OUTPUT_SKIP): Fix signedness warning.
11231 * config/pdp11/pdp11.c (pdp11_asm_print_operand_punct_valid_p): Remove
11232 "register" keyword.
11233 (pdp11_initial_elimination_offset) Remove unused variable.
11234 (pdp11_cmp_length) Ditto.
11235 (pdp11_insn_cost): Ditto, and fix signedness warning.
11237 2021-06-29 David Edelsohn <dje.gcc@gmail.com>
11239 * btfout.c: Include tm_p.h.
11242 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
11244 * config/bpf/bpf.c (bpf_expand_prologue): Do not mark insns as
11246 (bpf_expand_epilogue): Likewise.
11247 * config/bpf/bpf.h (DWARF2_FRAME_INFO): Define to 0.
11248 Do not define DBX_DEBUGGING_INFO.
11250 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
11252 * doc/invoke.texi: Document the CTF and BTF debug info options.
11254 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
11255 David Faust <david.faust@oracle.com>
11256 Jose E. Marchesi <jose.marchesi@oracle.com>
11257 Weimin Pan <weimin.pan@oracle.com>
11259 * Makefile.in: Add ctfc.*, ctfout.c and btfout.c files to
11260 GTFILES. Add new object files.
11261 * common.opt: Add CTF and BTF debug info options.
11262 * btfout.c: New file.
11263 * ctfc.c: Likewise.
11264 * ctfc.h: Likewise.
11265 * ctfout.c: Likewise.
11266 * dwarf2ctf.c: Likewise.
11267 * dwarf2ctf.h: Likewise.
11268 * dwarf2cfi.c (dwarf2out_do_frame): Acknowledge CTF_DEBUG and
11270 * dwarf2out.c (dwarf2out_source_line): Likewise.
11271 (dwarf2out_finish): Skip emitting DWARF if CTF or BTF are to
11273 (debug_format_do_cu): New function.
11274 (dwarf2out_early_finish): Traverse DIEs and emit CTF/BTF for
11276 Include dwarf2ctf.c.
11277 * final.c (dwarf2_debug_info_emitted_p): Acknowledge DWARF-based debug
11279 * flag-types.h (enum debug_info_type): Add CTF_DEBUG and BTF_DEBUG.
11280 (CTF_DEBUG): New bitmask.
11281 (BTF_DEBUG): Likewise.
11282 (enum ctf_debug_info_levels): New enum.
11283 * gengtype.c (open_base_files): Handle ctfc.h.
11284 (main): Handle uint32_t type.
11285 * flags.h (btf_debuginfo_p): New definition.
11286 (dwarf_based_debuginfo_p): Likewise.
11287 * opts.c (debug_type_names): Add entries for CTF and BTF.
11288 (btf_debuginfo_p): New function.
11289 (dwarf_based_debuginfo_p): Likewise.
11290 (common_handle_option): Handle -gctfN and -gbtf options.
11291 (set_debug_level): Set CTF_DEBUG, BTF_DEBUG whenever appropriate.
11292 * toplev.c (process_options): Inform the user and ignore -gctfLEVEL if
11295 2021-06-28 Jose E. Marchesi <jose.marchesi@oracle.com>
11297 * dwarf2out.c (AT_class): Function is no longer static.
11298 (AT_int): Likewise.
11299 (AT_unsigned): Likewise.
11300 (AT_loc): Likewise.
11301 (get_AT): Likewise.
11302 (get_AT_string): Likewise.
11303 (get_AT_flag): Likewise.
11304 (get_AT_unsigned): Likewise.
11305 (get_AT_ref): Likewise.
11306 (new_die_raw): Likewise.
11307 (lookup_decl_die): Likewise.
11308 (base_type_die): Likewise.
11309 (add_name_attribute): Likewise.
11310 (add_AT_int): Likewise.
11311 (add_AT_unsigned): Likewise.
11312 (add_AT_loc): Likewise.
11313 (dw_get_die_tag): New function.
11314 (dw_get_die_child): Likewise.
11315 (dw_get_die_sib): Likewise.
11316 (struct dwarf_file_data): Move from here to dwarf2out.h
11317 (struct dw_attr_struct): Likewise.
11318 * dwarf2out.h: Analogous changes.
11320 2021-06-28 Martin Jambor <mjambor@suse.cz>
11323 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
11324 members m_dead_stmts and m_dead_ssas.
11325 * ipa-param-manipulation.c
11326 (ipa_param_body_adjustments::mark_dead_statements): New function.
11327 (ipa_param_body_adjustments::common_initialization): Call it on
11328 all removed but not split parameters.
11329 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
11331 (ipa_param_body_adjustments::modify_call_stmt): Remove arguments that
11333 * tree-inline.c (remap_gimple_stmt): Do not copy dead statements, reset
11334 dead debug statements.
11335 (copy_phis_for_bb): Do not copy dead PHI nodes.
11337 2021-06-28 Martin Jambor <mjambor@suse.cz>
11340 * symtab-clones.h (clone_info): Removed member param_adjustments.
11341 * ipa-param-manipulation.h: Adjust initial comment to reflect how we
11342 deal with pass-through splits now.
11343 (ipa_param_performed_split): Removed.
11344 (ipa_param_adjustments::modify_call): Adjusted parameters.
11345 (class ipa_param_body_adjustments): Adjusted parameters of
11346 register_replacement, modify_gimple_stmt and modify_call_stmt.
11347 (ipa_verify_edge_has_no_modifications): Declare.
11348 (ipa_edge_modifications_finalize): Declare.
11349 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Remove
11350 performed_splits processing, pas only edge to padjs->modify_call,
11351 check that call arguments were not modified if they should not have
11353 * cgraphclones.c (cgraph_node::create_clone): Do not copy performed
11355 * ipa-param-manipulation.c (struct pass_through_split_map): New type.
11356 (ipa_edge_modification_info): Likewise.
11357 (ipa_edge_modification_sum): Likewise.
11358 (ipa_edge_modifications): New edge summary.
11359 (ipa_verify_edge_has_no_modifications): New function.
11360 (transitive_split_p): Removed.
11361 (transitive_split_map): Likewise.
11362 (init_transitive_splits): Likewise.
11363 (ipa_param_adjustments::modify_call): Adjusted to use the new edge
11364 summary instead of performed_splits.
11365 (ipa_param_body_adjustments::register_replacement): Drop dummy
11366 parameter, set base_index of the created ipa_param_body_replacement.
11367 (phi_arg_will_live_p): New function.
11368 (ipa_param_body_adjustments::common_initialization): Do not create
11369 IPA_SRA dummy decls.
11370 (simple_tree_swap_info): Removed.
11371 (remap_split_decl_to_dummy): Likewise.
11372 (record_argument_state_1): New function.
11373 (record_argument_state): Likewise.
11374 (ipa_param_body_adjustments::modify_call_stmt): New parameter
11375 orig_stmt. Do not work with dummy decls, save necessary info about
11376 changes to ipa_edge_modifications.
11377 (ipa_param_body_adjustments::modify_gimple_stmt): New parameter
11378 orig_stmt, pass it to modify_call_stmt.
11379 (ipa_param_body_adjustments::modify_cfun_body): Adjust call to
11380 modify_gimple_stmt.
11381 (ipa_edge_modifications_finalize): New function.
11382 * tree-inline.c (remap_gimple_stmt): Pass original statement to
11383 modify_gimple_stmt.
11384 (copy_phis_for_bb): Do not copy dead PHI nodes.
11385 (expand_call_inline): Do not remap performed_splits.
11386 (update_clone_info): Likewise.
11387 * toplev.c: Include ipa-param-manipulation.h.
11388 (toplev::finalize): Call ipa_edge_modifications_finalize.
11390 2021-06-28 Andrew Pinski <apinski@marvell.com>
11392 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Duplicate range
11393 info if we're the only things setting the target PHI.
11394 (value_replacement): Don't duplicate range here.
11395 (minmax_replacement): Likewise.
11397 2021-06-28 Richard Biener <rguenther@suse.de>
11399 PR tree-optimization/101229
11400 * gimple-walk.c (gimple_walk_op): Handle PHIs.
11402 2021-06-28 Martin Liska <mliska@suse.cz>
11404 * config/v850/v850.c (construct_dispose_instruction): Allocate
11406 (construct_prepare_instruction): Likewise.
11408 2021-06-28 Martin Liska <mliska@suse.cz>
11410 * config/v850/v850.c (v850_option_override): Build default
11412 (v850_can_inline_p): New. Allow MASK_PROLOG_FUNCTION to be
11413 ignored for inlining.
11414 (TARGET_CAN_INLINE_P): New.
11416 2021-06-28 Richard Biener <rguenther@suse.de>
11418 PR tree-optimization/101207
11419 * tree-vect-slp.c (vect_optimize_slp): Do BB reduction
11420 permute eliding for load permutations properly.
11422 2021-06-28 Richard Biener <rguenther@suse.de>
11424 PR tree-optimization/101173
11425 * gimple-loop-interchange.cc
11426 (tree_loop_interchange::valid_data_dependences): Disallow outer
11427 loop dependence distance of zero.
11429 2021-06-28 liuhongt <hongtao.liu@intel.com>
11432 * config/i386/sse.md (*avx_cmp<mode>3_lt): New
11433 define_insn_and_split.
11434 (*avx_cmp<mode>3_ltint): Ditto.
11435 (*avx2_pcmp<mode>3_3): Ditto.
11436 (*avx2_pcmp<mode>3_4): Ditto.
11437 (*avx2_pcmp<mode>3_5): Ditto.
11439 2021-06-28 liuhongt <hongtao.liu@intel.com>
11441 * config/i386/i386-builtin.def (IX86_BUILTIN_BLENDVPD256,
11442 IX86_BUILTIN_BLENDVPS256, IX86_BUILTIN_PBLENDVB256,
11443 IX86_BUILTIN_BLENDVPD, IX86_BUILTIN_BLENDVPS,
11444 IX86_BUILTIN_PBLENDVB128): Replace icode with
11446 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold blendv
11448 * config/i386/sse.md (*<sse4_1_avx2>_pblendvb_lt_subreg_not):
11449 New pre_reload splitter.
11451 2021-06-27 Andrew Pinski <apinski@marvell.com>
11453 PR middle-end/101230
11454 * fold-const.c (fold_ternary_loc): Check
11455 the return value of invert_tree_comparison.
11457 2021-06-27 David Edelsohn <dje.gcc@gmail.com>
11459 * config.gcc: Add SPDX License Identifier.
11460 (powerpc-ibm-aix789): Default to aix73.h.
11461 (powerpc-ibm-aix7.2.*.*): New stanza.
11462 * config/rs6000/aix72.h: Add SPDX License Identifier.
11463 * config/rs6000/aix73.h: New file.
11465 2021-06-26 Jason Merrill <jason@redhat.com>
11467 * except.c: #include "dwarf2.h" instead of "dwarf2out.h".
11469 2021-06-26 Andrew Pinski <apinski@marvell.com>
11471 * genmatch.c (lower_cond): Copy for_subst_vec
11472 for the simplify also.
11473 (lower): Swap the order for lower_for and lower_cond.
11475 2021-06-26 Andrew Pinski <apinski@marvell.com>
11477 * tree-ssa-phiopt.c (match_simplify_replacement): Reset
11478 flow senatitive info on the moved ssa set.
11480 2021-06-26 Andrew Pinski <apinski@marvell.com>
11482 * fold-const.c (fold_cond_expr_with_comparison):
11483 Exand arg0 into comp_code, arg00, and arg01.
11484 (fold_ternary_loc): Use invert_tree_comparison
11485 instead of fold_invert_truthvalue for the case
11486 where we have A CMP B ? C : A.
11488 2021-06-25 Martin Sebor <msebor@redhat.com>
11490 PR middle-end/101216
11491 * calls.c (maybe_warn_rdwr_sizes): Use the no_warning constant.
11493 2021-06-25 Jeff Law <jeffreyalaw@gmail.com>
11495 * config/h8300/h8300.c (select_cc_mode): Handle ASHIFTRT and LSHIFTRT.
11497 2021-06-25 Richard Biener <rguenther@suse.de>
11499 PR tree-optimization/101202
11500 * tree-vect-slp.c (vect_optimize_slp): Explicitely handle
11503 2021-06-25 Richard Biener <rguenther@suse.de>
11505 * tree-vect-slp-patterns.c (addsub_pattern::build): Copy
11506 STMT_VINFO_REDUC_DEF from the original representative.
11508 2021-06-25 Martin Sebor <msebor@redhat.com>
11510 * builtins.c (warn_string_no_nul): Replace uses of TREE_NO_WARNING,
11511 gimple_no_warning_p and gimple_set_no_warning with
11512 warning_suppressed_p, and suppress_warning.
11514 (maybe_warn_for_bound): Same.
11515 (warn_for_access): Same.
11516 (check_access): Same.
11517 (expand_builtin_strncmp): Same.
11518 (fold_builtin_varargs): Same.
11519 * calls.c (maybe_warn_nonstring_arg): Same.
11520 (maybe_warn_rdwr_sizes): Same.
11521 * cfgexpand.c (expand_call_stmt): Same.
11522 * cgraphunit.c (check_global_declaration): Same.
11523 * fold-const.c (fold_undefer_overflow_warnings): Same.
11524 (fold_truth_not_expr): Same.
11525 (fold_unary_loc): Same.
11526 (fold_checksum_tree): Same.
11527 * gimple-array-bounds.cc (array_bounds_checker::check_array_ref): Same.
11528 (array_bounds_checker::check_mem_ref): Same.
11529 (array_bounds_checker::check_addr_expr): Same.
11530 (array_bounds_checker::check_array_bounds): Same.
11531 * gimple-expr.c (copy_var_decl): Same.
11532 * gimple-fold.c (gimple_fold_builtin_strcpy): Same.
11533 (gimple_fold_builtin_strncat): Same.
11534 (gimple_fold_builtin_stxcpy_chk): Same.
11535 (gimple_fold_builtin_stpcpy): Same.
11536 (gimple_fold_builtin_sprintf): Same.
11537 (fold_stmt_1): Same.
11538 * gimple-ssa-isolate-paths.c (diag_returned_locals): Same.
11539 * gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Same.
11540 * gimple-ssa-sprintf.c (handle_printf_call): Same.
11541 * gimple-ssa-store-merging.c (imm_store_chain_info::output_merged_store): Same.
11542 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
11543 * gimple-ssa-warn-restrict.h: Adjust declarations.
11544 (maybe_diag_access_bounds): Replace uses of TREE_NO_WARNING,
11545 gimple_no_warning_p and gimple_set_no_warning with
11546 warning_suppressed_p, and suppress_warning.
11547 (check_call): Same.
11548 (check_bounds_or_overlap): Same.
11549 * gimple.c (gimple_build_call_from_tree): Same.
11550 * gimplify.c (gimplify_return_expr): Same.
11551 (gimplify_cond_expr): Same.
11552 (gimplify_modify_expr_complex_part): Same.
11553 (gimplify_modify_expr): Same.
11554 (gimple_push_cleanup): Same.
11555 (gimplify_expr): Same.
11556 * omp-expand.c (expand_omp_for_generic): Same.
11557 (expand_omp_taskloop_for_outer): Same.
11558 * omp-low.c (lower_rec_input_clauses): Same.
11559 (lower_lastprivate_clauses): Same.
11560 (lower_send_clauses): Same.
11561 (lower_omp_target): Same.
11562 * tree-cfg.c (pass_warn_function_return::execute): Same.
11563 * tree-complex.c (create_one_component_var): Same.
11564 * tree-inline.c (remap_gimple_op_r): Same.
11565 (copy_tree_body_r): Same.
11566 (declare_return_variable): Same.
11567 (expand_call_inline): Same.
11568 * tree-nested.c (lookup_field_for_decl): Same.
11569 * tree-sra.c (create_access_replacement): Same.
11570 (generate_subtree_copies): Same.
11571 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Same.
11572 * tree-ssa-forwprop.c (combine_cond_expr_cond): Same.
11573 * tree-ssa-loop-ch.c (ch_base::copy_headers): Same.
11574 * tree-ssa-loop-im.c (execute_sm): Same.
11575 * tree-ssa-phiopt.c (cond_store_replacement): Same.
11576 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
11577 (handle_builtin_strcpy): Same.
11578 (maybe_diag_stxncpy_trunc): Same.
11579 (handle_builtin_stxncpy_strncat): Same.
11580 (handle_builtin_strcat): Same.
11581 * tree-ssa-uninit.c (get_no_uninit_warning): Same.
11582 (set_no_uninit_warning): Same.
11583 (uninit_undefined_value_p): Same.
11584 (warn_uninit): Same.
11585 (maybe_warn_operand): Same.
11586 * tree-vrp.c (compare_values_warnv): Same.
11587 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr): Same.
11588 (test_for_singularity): Same.
11589 * gimple.h (warning_suppressed_p): New function.
11590 (suppress_warning): Same.
11591 (copy_no_warning): Same.
11592 (gimple_set_block): Call gimple_set_location.
11593 (gimple_set_location): Call copy_warning.
11595 2021-06-25 Martin Sebor <msebor@redhat.com>
11597 * tree.h (warning_suppressed_at, copy_warning,
11598 warning_suppressed_p, suppress_warning): New functions.
11600 2021-06-25 Martin Sebor <msebor@redhat.com>
11602 * Makefile.in (OBJS-libcommon): Add diagnostic-spec.o.
11603 * gengtype.c (open_base_files): Add diagnostic-spec.h.
11604 * diagnostic-spec.c: New file.
11605 * diagnostic-spec.h: New file.
11606 * tree.h (no_warning, all_warnings, suppress_warning_at): New
11608 * warning-control.cc: New file.
11610 2021-06-25 liuhongt <hongtao.liu@intel.com>
11613 * config/i386/i386.c (x86_order_regs_for_local_alloc):
11616 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11618 PR tree-optimization/101189
11619 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Pass
11620 LHS range of condition to postfold routine.
11621 (fold_using_range::postfold_gcond_edges): Only process the TRUE or
11622 FALSE edge if the LHS range supports it being taken.
11623 * gimple-range-fold.h (postfold_gcond_edges): Add range parameter.
11625 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11627 * value-relation.cc (equiv_oracle::dump): Do not dump NULL blocks.
11628 (relation_oracle::find_relation_block): Check correct bitmap.
11629 (relation_oracle::dump): Do not dump NULL blocks.
11631 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11633 * gimple-range-cache.cc (ranger_cache::propagate_cache): Call
11634 range_on_edge instead of manually calculating.
11636 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
11638 * range-op.cc: Fix comment.
11640 2021-06-24 Uroš Bizjak <ubizjak@gmail.com>
11643 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
11644 Handle V8QI and V4HI modes.
11645 * config/i386/mmx.md (sse4_1_<any_extend:code>v4qiv4hi2):
11647 (sse4_1_<any_extend:code>v4qiv4hi2): Ditto.
11648 (mmxpackmode): New mode attribute.
11649 (vec_pack_trunc_<mmxpackmode:mode>): New expander.
11650 (mmxunpackmode): New mode attribute.
11651 (vec_unpacks_lo_<mmxunpackmode:mode>): New expander.
11652 (vec_unpacks_hi_<mmxunpackmode:mode>): Ditto.
11653 (vec_unpacku_lo_<mmxunpackmode:mode>): Ditto.
11654 (vec_unpacku_hi_<mmxunpackmode:mode>): Ditto.
11655 * config/i386/i386.md (extsuffix): Move from ...
11656 * config/i386/sse.md: ... here.
11658 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
11660 * dwarf2out.c (dwarf2out_assembly_start): Emit .file 0 marker here..
11661 (dwarf2out_finish): ...instead of here.
11663 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
11665 * configure.ac (--gdwarf-5 option): Use objdump instead of readelf.
11666 (working --gdwarf-4/--gdwarf-5 for all sources): Likewise.
11667 (--gdwarf-4 not refusing generated .debug_line): Adjust for Windows.
11668 * configure: Regenerate.
11670 2021-06-24 Richard Biener <rguenther@suse.de>
11672 * config/i386/sse.md (vec_addsubv4df3, vec_addsubv2df3,
11673 vec_addsubv8sf3, vec_addsubv4sf3): Merge into ...
11674 (vec_addsub<mode>3): ... using a new addsub_cst mode attribute.
11676 2021-06-24 Richard Biener <rguenther@suse.de>
11678 * config/i386/sse.md (avx_addsubv4df3): Rename to
11680 (avx_addsubv8sf3): Rename to vec_addsubv8sf3.
11681 (sse3_addsubv2df3): Rename to vec_addsubv2df3.
11682 (sse3_addsubv4sf3): Rename to vec_addsubv4sf3.
11683 * config/i386/i386-builtin.def: Adjust.
11684 * internal-fn.def (VEC_ADDSUB): New internal optab fn.
11685 * optabs.def (vec_addsub_optab): New optab.
11686 * tree-vect-slp-patterns.c (class addsub_pattern): New.
11687 (slp_patterns): Add addsub_pattern.
11688 * tree-vect-slp.c (vect_optimize_slp): Disable propagation
11689 across CFN_VEC_ADDSUB.
11690 * tree-vectorizer.h (vect_pattern::vect_pattern): Make
11692 * doc/md.texi (vec_addsub<mode>3): Document.
11694 2021-06-24 Jakub Jelinek <jakub@redhat.com>
11696 PR middle-end/101170
11697 * df-scan.c (df_ref_record): For paradoxical big-endian SUBREGs
11698 where regno + subreg_regno_offset wraps around use 0 as starting
11701 2021-06-24 Jakub Jelinek <jakub@redhat.com>
11703 PR middle-end/101172
11704 * stor-layout.c (finish_bitfield_representative): If nextf has
11705 error_mark_node type, set repr type to error_mark_node too.
11707 2021-06-24 Ilya Leoshkevich <iii@linux.ibm.com>
11709 * config/s390/s390.c (s390_function_profiler): Ignore labelno
11711 * config/s390/s390.h (NO_PROFILE_COUNTERS): Define.
11713 2021-06-24 Richard Biener <rguenther@suse.de>
11715 * tree-vect-slp.c (vect_optimize_slp): Do not propagate
11716 across operations that have different semantics on different
11719 2021-06-24 Jakub Jelinek <jakub@redhat.com>
11721 * tree.h (OMP_CLAUSE_MAP_IN_REDUCTION): Document meaning for OpenMP.
11722 * gimplify.c (gimplify_scan_omp_clauses): For OpenMP map clauses
11723 with OMP_CLAUSE_MAP_IN_REDUCTION flag partially defer gimplification
11724 of non-decl OMP_CLAUSE_DECL. For OMP_CLAUSE_IN_REDUCTION on
11725 OMP_TARGET user outer_ctx instead of ctx for placeholders and
11726 initializer/combiner gimplification.
11727 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_MAP_IN_REDUCTION
11728 on target constructs.
11729 (lower_rec_input_clauses): Likewise.
11730 (lower_omp_target): Likewise.
11731 * omp-expand.c (expand_omp_target): Temporarily ignore nowait clause
11732 on target if in_reduction is present.
11734 2021-06-24 Kewen Lin <linkw@linux.ibm.com>
11736 * tree-predcom.c (class pcom_worker): New class.
11737 (release_chain): Renamed to...
11738 (pcom_worker::release_chain): ...this.
11739 (release_chains): Renamed to...
11740 (pcom_worker::release_chains): ...this.
11741 (aff_combination_dr_offset): Renamed to...
11742 (pcom_worker::aff_combination_dr_offset): ...this.
11743 (determine_offset): Renamed to...
11744 (pcom_worker::determine_offset): ...this.
11745 (class comp_ptrs): New class.
11746 (split_data_refs_to_components): Renamed to...
11747 (pcom_worker::split_data_refs_to_components): ...this,
11748 and update with class comp_ptrs.
11749 (suitable_component_p): Renamed to...
11750 (pcom_worker::suitable_component_p): ...this.
11751 (filter_suitable_components): Renamed to...
11752 (pcom_worker::filter_suitable_components): ...this.
11753 (valid_initializer_p): Renamed to...
11754 (pcom_worker::valid_initializer_p): ...this.
11755 (find_looparound_phi): Renamed to...
11756 (pcom_worker::find_looparound_phi): ...this.
11757 (add_looparound_copies): Renamed to...
11758 (pcom_worker::add_looparound_copies): ...this.
11759 (determine_roots_comp): Renamed to...
11760 (pcom_worker::determine_roots_comp): ...this.
11761 (determine_roots): Renamed to...
11762 (pcom_worker::determine_roots): ...this.
11763 (single_nonlooparound_use): Renamed to...
11764 (pcom_worker::single_nonlooparound_use): ...this.
11765 (remove_stmt): Renamed to...
11766 (pcom_worker::remove_stmt): ...this.
11767 (execute_pred_commoning_chain): Renamed to...
11768 (pcom_worker::execute_pred_commoning_chain): ...this.
11769 (execute_pred_commoning): Renamed to...
11770 (pcom_worker::execute_pred_commoning): ...this.
11771 (struct epcc_data): New member worker.
11772 (execute_pred_commoning_cbck): Call execute_pred_commoning
11773 with pcom_worker pointer.
11774 (find_use_stmt): Renamed to...
11775 (pcom_worker::find_use_stmt): ...this.
11776 (find_associative_operation_root): Renamed to...
11777 (pcom_worker::find_associative_operation_root): ...this.
11778 (find_common_use_stmt): Renamed to...
11779 (pcom_worker::find_common_use_stmt): ...this.
11780 (combinable_refs_p): Renamed to...
11781 (pcom_worker::combinable_refs_p): ...this.
11782 (reassociate_to_the_same_stmt): Renamed to...
11783 (pcom_worker::reassociate_to_the_same_stmt): ...this.
11784 (stmt_combining_refs): Renamed to...
11785 (pcom_worker::stmt_combining_refs): ...this.
11786 (combine_chains): Renamed to...
11787 (pcom_worker::combine_chains): ...this.
11788 (try_combine_chains): Renamed to...
11789 (pcom_worker::try_combine_chains): ...this.
11790 (prepare_initializers_chain): Renamed to...
11791 (pcom_worker::prepare_initializers_chain): ...this.
11792 (prepare_initializers): Renamed to...
11793 (pcom_worker::prepare_initializers): ...this.
11794 (prepare_finalizers_chain): Renamed to...
11795 (pcom_worker::prepare_finalizers_chain): ...this.
11796 (prepare_finalizers): Renamed to...
11797 (pcom_worker::prepare_finalizers): ...this.
11798 (tree_predictive_commoning_loop): Renamed to...
11799 (pcom_worker::tree_predictive_commoning_loop): ...this, adjust
11800 some calls and remove some cleanup code.
11801 (tree_predictive_commoning): Adjusted to use pcom_worker instance.
11802 (static variable looparound_phis): Remove.
11803 (static variable name_expansions): Remove.
11805 2021-06-24 Richard Biener <rguenther@suse.de>
11807 * tree-vect-slp.c (slpg_vertex): New struct.
11808 (vect_slp_build_vertices): Adjust.
11809 (vect_optimize_slp): Likewise. Maintain an outgoing permute
11810 and a materialized one.
11812 2021-06-24 Richard Biener <rguenther@suse.de>
11814 PR tree-optimization/101105
11815 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
11816 Only ignore steps when they are equal or scalar order is preserved.
11818 2021-06-24 liuhongt <hongtao.liu@intel.com>
11821 * config/i386/i386-expand.c (ix86_expand_vec_interleave):
11822 Adjust comments for ix86_expand_vecop_qihi2.
11823 (ix86_expand_vecmul_qihi): Renamed to ..
11824 (ix86_expand_vecop_qihi2): Adjust function prototype to
11825 support shift operation, add static to definition.
11826 (ix86_expand_vec_shift_qihi_constant): Add static to definition.
11827 (ix86_expand_vecop_qihi): Call ix86_expand_vecop_qihi2 and
11828 ix86_expand_vec_shift_qihi_constant.
11829 * config/i386/i386-protos.h (ix86_expand_vecmul_qihi): Deleted.
11830 (ix86_expand_vec_shift_qihi_constant): Deleted.
11831 * config/i386/sse.md (VI12_256_512_AVX512VL): New mode
11833 (mulv8qi3): Call ix86_expand_vecop_qihi directly, add
11834 condition TARGET_64BIT.
11835 (mul<mode>3): Ditto.
11836 (<insn><mode>3): Ditto.
11837 (vlshr<mode>3): Extend to support avx512 vlshr.
11838 (v<insn><mode>3): New expander for
11840 (v<insn>v8qi3): Ditto.
11841 (vashrv8hi3<mask_name>): Renamed to ..
11842 (vashr<mode>3): And extend to support V16QImode for avx512.
11843 (vashrv16qi3): Deleted.
11844 (vashrv2di3<mask_name>): Extend expander to support avx512
11847 2021-06-23 Dimitar Dimitrov <dimitar@dinux.eu>
11849 * doc/lto.texi (Design Overview): Update that slim objects are
11852 2021-06-23 Aaron Sawdey <acsawdey@linux.ibm.com>
11854 * config/rs6000/rs6000-cpus.def: Take OPTION_MASK_PCREL_OPT out
11855 of OTHER_POWER10_MASKS so it will not be enabled by default.
11857 2021-06-23 Richard Biener <rguenther@suse.de>
11858 Martin Jambor <mjambor@suse.cz>
11860 * tree-inline.c (setup_one_parameter): Set TREE_READONLY of the
11861 param replacement unconditionally. Adjust comment.
11863 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11865 * Makefile.in (OBJS): Add gimple-range-fold.o
11866 * gimple-range-fold.cc: New.
11867 * gimple-range-fold.h: New.
11868 * gimple-range-gori.cc (gimple_range_calc_op1): Move to here.
11869 (gimple_range_calc_op2): Ditto.
11870 * gimple-range-gori.h: Move prototypes to here.
11871 * gimple-range.cc: Adjust include files.
11872 (fur_source:fur_source): Relocate to gimple-range-fold.cc.
11873 (fur_source::get_operand): Ditto.
11874 (fur_source::get_phi_operand): Ditto.
11875 (fur_source::query_relation): Ditto.
11876 (fur_source::register_relation): Ditto.
11877 (class fur_edge): Ditto.
11878 (fur_edge::fur_edge): Ditto.
11879 (fur_edge::get_operand): Ditto.
11880 (fur_edge::get_phi_operand): Ditto.
11881 (fur_stmt::fur_stmt): Ditto.
11882 (fur_stmt::get_operand): Ditto.
11883 (fur_stmt::get_phi_operand): Ditto.
11884 (fur_stmt::query_relation): Ditto.
11885 (class fur_depend): Relocate to gimple-range-fold.h.
11886 (fur_depend::fur_depend): Relocate to gimple-range-fold.cc.
11887 (fur_depend::register_relation): Ditto.
11888 (fur_depend::register_relation): Ditto.
11889 (class fur_list): Ditto.
11890 (fur_list::fur_list): Ditto.
11891 (fur_list::get_operand): Ditto.
11892 (fur_list::get_phi_operand): Ditto.
11893 (fold_range): Ditto.
11894 (adjust_pointer_diff_expr): Ditto.
11895 (gimple_range_adjustment): Ditto.
11896 (gimple_range_base_of_assignment): Ditto.
11897 (gimple_range_operand1): Ditto.
11898 (gimple_range_operand2): Ditto.
11899 (gimple_range_calc_op1): Relocate to gimple-range-gori.cc.
11900 (gimple_range_calc_op2): Ditto.
11901 (fold_using_range::fold_stmt): Relocate to gimple-range-fold.cc.
11902 (fold_using_range::range_of_range_op): Ditto.
11903 (fold_using_range::range_of_address): Ditto.
11904 (fold_using_range::range_of_phi): Ditto.
11905 (fold_using_range::range_of_call): Ditto.
11906 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
11907 (fold_using_range::range_of_builtin_call): Ditto.
11908 (fold_using_range::range_of_cond_expr): Ditto.
11909 (fold_using_range::range_of_ssa_name_with_loop_info): Ditto.
11910 (fold_using_range::relation_fold_and_or): Ditto.
11911 (fold_using_range::postfold_gcond_edges): Ditto.
11912 * gimple-range.h: Add gimple-range-fold.h to include files. Change
11913 GIMPLE_RANGE_STMT_H to GIMPLE_RANGE_H.
11914 (gimple_range_handler): Relocate to gimple-range-fold.h.
11915 (gimple_range_ssa_p): Ditto.
11916 (range_compatible_p): Ditto.
11917 (class fur_source): Ditto.
11918 (class fur_stmt): Ditto.
11919 (class fold_using_range): Ditto.
11920 (gimple_range_calc_op1): Relocate to gimple-range-gori.h
11921 (gimple_range_calc_op2): Ditto.
11923 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11925 PR tree-optimization/101148
11926 PR tree-optimization/101014
11927 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
11928 (ranger_cache::~ranger_cache): Adjust.
11929 (ranger_cache::block_range): Check if propagation disallowed.
11930 (ranger_cache::propagate_cache): Disallow propagation if new value
11931 can't be stored properly.
11932 * gimple-range-cache.h (ranger_cache::m_propfail): New member.
11934 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11936 * gimple-range-cache.cc (class ssa_block_ranges): Adjust prototype.
11937 (sbr_vector::set_bb_range): Return true.
11938 (class sbr_sparse_bitmap): Adjust.
11939 (sbr_sparse_bitmap::set_bb_range): Return value.
11940 (block_range_cache::set_bb_range): Return value.
11941 (ranger_cache::propagate_cache): Use return value to print msg.
11942 * gimple-range-cache.h (class block_range_cache): Adjust.
11944 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
11946 * gimple-range.cc (dump_bb): Use range_on_edge from the cache.
11948 2021-06-23 Jeff Law <jeffreyalaw@gmail.com>
11950 * config/h8300/logical.md (<code><mode>3<ccnz>): Use <cczn>
11951 so this pattern can be used for test/compare removal. Pass
11952 current insn to compute_logical_op_length and output_logical_op.
11953 * config/h8300/h8300.c (compute_logical_op_cc): Remove.
11954 (h8300_and_costs): Add argument to compute_logical_op_length.
11955 (output_logical_op): Add new argument. Use it to determine if the
11956 condition codes are used and adjust the output accordingly.
11957 (compute_logical_op_length): Add new argument and update length
11958 computations when condition codes are used.
11959 * config/h8300/h8300-protos.h (compute_logical_op_length): Update
11961 (output_logical_op): Likewise.
11963 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
11966 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
11967 Handle 64bit modes for TARGET_XOP. Use indirect gen_* functions.
11968 * config/i386/mmx.md (mmx_ppermv64): New insn pattern.
11969 * config/i386/i386.md (unspec): Move UNSPEC_XOP_PERMUTE from ...
11970 * config/i386/sse.md (unspec): ... here.
11972 2021-06-23 Martin Liska <mliska@suse.cz>
11975 * optc-save-gen.awk: Put back arm_fp16_format to
11978 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
11981 * config/i386/i386.md (bsr_rex64): Add zero-flag setting RTX.
11984 (clz<mode>2): Update RTX pattern for additions.
11986 2021-06-23 Jakub Jelinek <jakub@redhat.com>
11988 PR middle-end/101167
11989 * omp-low.c (lower_omp_regimplify_p): Regimplify also PARM_DECLs
11990 and RESULT_DECLs that have DECL_HAS_VALUE_EXPR_P set.
11992 2021-06-22 Sergei Trofimovich <siarheit@google.com>
11994 * doc/rtl.texi: drop unbalanced parenthesis.
11996 2021-06-22 Richard Biener <rguenther@suse.de>
11998 PR middle-end/101156
11999 * gimplify.c (gimplify_expr): Remove premature incorrect
12002 2021-06-22 Jakub Jelinek <jakub@redhat.com>
12004 PR tree-optimization/101159
12005 * tree-vect-patterns.c (vect_recog_popcount_pattern): Fix some
12008 2021-06-22 Jakub Jelinek <jakub@redhat.com>
12010 PR middle-end/101160
12011 * function.c (assign_parms): For decl_result with TYPE_EMPTY_P type
12012 clear crtl->return_rtx instead of keeping it referencing a pseudo.
12014 2021-06-22 Jakub Jelinek <jakub@redhat.com>
12015 Andrew Pinski <apinski@marvell.com>
12017 PR tree-optimization/101162
12018 * fold-const.c (range_check_type): Handle OFFSET_TYPE like pointer
12021 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12023 * range-op.cc (range_relational_tests): New.
12024 (range_op_tests): Call range_relational_tests.
12026 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12028 * range-op.cc (operator_cast::lhs_op1_relation): New.
12029 (operator_identity::lhs_op1_relation): Mew.
12031 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12033 * range-op.cc (operator_minus::op1_op2_relation_effect): New.
12035 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12037 * range-op.cc (operator_plus::lhs_op1_relation): New.
12038 (operator_plus::lhs_op2_relation): New.
12040 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12042 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a
12043 relation_oracle if dominators exist.
12044 (ranger_cache::~ranger_cache): Dispose of oracle.
12045 (ranger_cache::dump_bb): Dump oracle.
12046 * gimple-range.cc (fur_source::fur_source): New.
12047 (fur_source::get_operand): Use mmeber query.
12048 (fur_source::get_phi_operand): Use member_query.
12049 (fur_source::query_relation): New.
12050 (fur_source::register_dependency): Delete.
12051 (fur_source::register_relation): New.
12052 (fur_edge::fur_edge): Adjust.
12053 (fur_edge::get_phi_operand): Fix comment.
12054 (fur_edge::query): Delete.
12055 (fur_stmt::fur_stmt): Adjust.
12056 (fur_stmt::query): Delete.
12057 (fur_depend::fur_depend): Adjust.
12058 (fur_depend::register_relation): New.
12059 (fur_depend::register_relation): New.
12060 (fur_list::fur_list): Adjust.
12061 (fur_list::get_operand): Use member query.
12062 (fold_using_range::range_of_range_op): Process and query relations.
12063 (fold_using_range::range_of_address): Adjust dependency call.
12064 (fold_using_range::range_of_phi): Ditto.
12065 (gimple_ranger::gimple_ranger): New. Use ranger_ache oracle.
12066 (fold_using_range::relation_fold_and_or): New.
12067 (fold_using_range::postfold_gcond_edges): New.
12068 * gimple-range.h (class gimple_ranger): Adjust.
12069 (class fur_source): Adjust members.
12070 (class fur_stmt): Ditto.
12071 (class fold_using_range): Ditto.
12073 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12075 * range-op.cc (range_operator::wi_fold): Apply relation effect.
12076 (range_operator::fold_range): Adjust and apply relation effect.
12077 (*::fold_range): Add relation parameters.
12078 (*::op1_range): Ditto.
12079 (*::op2_range): Ditto.
12080 (range_operator::lhs_op1_relation): New.
12081 (range_operator::lhs_op2_relation): New.
12082 (range_operator::op1_op2_relation): New.
12083 (range_operator::op1_op2_relation_effect): New.
12084 (relop_early_resolve): New.
12085 (operator_equal::op1_op2_relation): New.
12086 (operator_equal::fold_range): Call relop_early_resolve.
12087 (operator_not_equal::op1_op2_relation): New.
12088 (operator_not_equal::fold_range): Call relop_early_resolve.
12089 (operator_lt::op1_op2_relation): New.
12090 (operator_lt::fold_range): Call relop_early_resolve.
12091 (operator_le::op1_op2_relation): New.
12092 (operator_le::fold_range): Call relop_early_resolve.
12093 (operator_gt::op1_op2_relation): New.
12094 (operator_gt::fold_range): Call relop_early_resolve.
12095 (operator_ge::op1_op2_relation): New.
12096 (operator_ge::fold_range): Call relop_early_resolve.
12097 * range-op.h (class range_operator): Adjust parameters and methods.
12099 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
12101 * Makefile.in (OBJS): Add value-relation.o.
12102 * gimple-range.h: Adjust include files.
12103 * tree-data-ref.c: Adjust include file order.
12104 * value-query.cc (range_query::get_value_range): Default to no oracle.
12105 (range_query::query_relation): New.
12106 (range_query::query_relation): New.
12107 * value-query.h (class range_query): Adjust.
12108 * value-relation.cc: New.
12109 * value-relation.h: New.
12111 2021-06-22 Richard Biener <rguenther@suse.de>
12113 PR tree-optimization/101151
12114 * tree-ssa-sink.c (statement_sink_location): Expand irreducible
12117 2021-06-22 Jojo R <rjiejie@linux.alibaba.com>
12119 * config/riscv/riscv.c (thead_c906_tune_info): New.
12120 (riscv_tune_info_table): Use new tune.
12122 2021-06-22 Richard Biener <rguenther@suse.de>
12124 PR tree-optimization/101158
12125 * tree-vect-slp.c (vect_build_slp_tree_1): Move same operand
12126 checking after checking for matching operation.
12128 2021-06-22 Richard Biener <rguenther@suse.de>
12130 PR tree-optimization/101159
12131 * tree-vect-patterns.c (vect_recog_popcount_pattern): Add
12132 missing NULL vectype check.
12134 2021-06-22 Richard Biener <rguenther@suse.de>
12136 PR tree-optimization/101154
12137 * tree-vect-slp.c (vect_build_slp_tree_2): Fix out-of-bound access.
12139 2021-06-22 Jakub Jelinek <jakub@redhat.com>
12142 * config/i386/i386-protos.h (ix86_last_zero_store_uid): Declare.
12143 * config/i386/i386-expand.c (ix86_last_zero_store_uid): New variable.
12144 * config/i386/i386.c (ix86_expand_prologue): Clear it.
12145 * config/i386/i386.md (peephole2s for 1/2/4 stores of const0_rtx):
12146 Remove "" from match_operand. Emit new insns using emit_move_insn and
12147 set ix86_last_zero_store_uid to INSN_UID of the last store.
12148 Add peephole2s for 1/2/4 stores of const0_rtx following previous
12151 2021-06-22 Martin Liska <mliska@suse.cz>
12153 * auto-profile.c (AUTO_PROFILE_VERSION): Bump as string format
12156 2021-06-22 Martin Liska <mliska@suse.cz>
12158 * gcov-io.h: Remove padding entries.
12160 2021-06-22 liuhongt <hongtao.liu@intel.com>
12162 PR tree-optimization/97770
12163 * tree-vect-patterns.c (vect_recog_popcount_pattern):
12165 (vect_recog_func vect_vect_recog_func_ptrs): Add new pattern.
12167 2021-06-22 liuhongt <hongtao.liu@intel.com>
12170 * config/i386/i386-builtin.def (BDESC): Adjust builtin name.
12171 * config/i386/sse.md (<avx512>_expand<mode>_mask): Rename to ..
12172 (expand<mode>_mask): this ..
12173 (*expand<mode>_mask): New pre_reload splitter to transform
12174 v{,p}expand* to vmov* when mask is zero, all ones, or has all
12175 ones in it's lower part, otherwise still generate
12178 2021-06-22 liuhongt <hongtao.liu@intel.com>
12181 * config/i386/i386-expand.c
12182 (ix86_expand_special_args_builtin): Keep constm1_operand only
12183 if it satisfies insn's operand predicate.
12185 2021-06-21 Jason Merrill <jason@redhat.com>
12188 * df-scan.c (df_ref_record): Check that regno < endregno.
12189 * function.c (assign_parms, expand_function_end): Do nothing with a
12190 TYPE_EMPTY_P result.
12192 2021-06-21 Richard Biener <rguenther@suse.de>
12194 PR tree-optimization/101120
12195 * tree-vect-data-refs.c (bump_vector_ptr): Fold the
12197 * tree-vect-slp.c (vect_transform_slp_perm_load): Add
12198 DR chain DCE capability.
12199 * tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
12200 * tree-vect-stmts.c (vectorizable_load): Remove unused
12201 loads in the DR chain for SLP.
12203 2021-06-21 Jakub Jelinek <jakub@redhat.com>
12205 PR inline-asm/100785
12206 * gimplify.c (gimplify_asm_expr): Don't diagnose errors if
12207 output or input operands were already error_mark_node.
12208 * cfgexpand.c (expand_asm_stmt): If errors are emitted,
12209 remove all inputs, outputs and clobbers from the asm and
12210 set template to "".
12212 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
12214 * config/arm/arm_neon.h (vceq_s8): Replace builtin with __a == __b.
12215 (vceq_s16): Likewise.
12216 (vceq_s32): Likewise.
12217 (vceq_u8): Likewise.
12218 (vceq_u16): Likewise.
12219 (vceq_u32): Likewise.
12220 (vceq_p8): Likewise.
12221 (vceqq_s8): Likewise.
12222 (vceqq_s16): Likewise.
12223 (vceqq_s32): Likewise.
12224 (vceqq_u8): Likewise.
12225 (vceqq_u16): Likewise.
12226 (vceqq_u32): Likewise.
12227 (vceqq_p8): Likewise.
12228 (vceq_f32): Gate __a == __b on __FAST_MATH__.
12229 (vceqq_f32): Likewise.
12230 (vceq_f16): Likewise.
12231 (vceqq_f16): Likewise.
12233 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
12236 * config/arm/iterators.md (NEON_VACMP): Remove.
12237 * config/arm/neon.md (neon_vca<cmp_op><mode>): Use GLTE instead of GTGE
12239 (neon_vca<cmp_op><mode>_insn): Likewise.
12240 (neon_vca<cmp_op_unsp><mode>_insn_unspec): Use NEON_VAGLTE instead of
12243 2021-06-21 Richard Biener <rguenther@suse.de>
12245 PR tree-optimization/101121
12246 * tree-vect-slp.c (vect_build_slp_tree_2): To not fail fatally
12247 when we just lack a stmt with the desired op when doing permutation.
12248 (vect_build_slp_tree): When caching a failed SLP build attempt
12249 assert that at least one lane is marked as not matching.
12251 2021-06-21 liuhongt <hongtao.liu@intel.com>
12254 * config/i386/i386.md: (*anddi_1): Disparage slightly the mask
12255 register alternative.
12256 (*and<mode>_1): Ditto.
12258 (*andn<mode>_1): Ditto.
12259 (*<code><mode>_1): Ditto.
12260 (*<code>qi_1): Ditto.
12261 (*one_cmpl<mode>2_1): Ditto.
12262 (*one_cmplsi2_1_zext): Ditto.
12263 (*one_cmplqi2_1): Ditto.
12264 * config/i386/i386.c (x86_order_regs_for_local_alloc): Change
12265 the order of mask registers to be before general registers.
12267 2021-06-21 Roger Sayle <roger@nextmovesoftware.com>
12270 * config/i386/i386.md: New define_peephole2s to shrink writing
12271 1, 2 or 4 consecutive zeros to memory when optimizing for size.
12273 2021-06-18 Jeff Law <jeffreyalaw@gmail.com>
12275 * config/h8300/h8300.c (h8300_select_cc_mode): Handle SYMBOL_REF.
12276 * config/h8300/logical.md (<code><mode>3 logcial expander): Generate
12277 more efficient code when the source can be trivially simplified.
12279 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
12281 * gimple-range-cache.cc (ranger_cache::range_of_def): Calculate
12282 a range if global is not available.
12283 (ranger_cache::entry_range): Fallback to range_of_def.
12284 * gimple-range-cache.h (range_of_def): Adjust prototype.
12286 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
12288 PR tree-optimization/101014
12289 * gimple-range-cache.cc (ranger_cache::ranger_cache): Remove poor
12291 (ranger_cache::~ranger_cache): Ditto.
12292 (ranger_cache::enable_new_values): Delete.
12293 (ranger_cache::push_poor_value): Delete.
12294 (ranger_cache::range_of_def): Remove poor value processing.
12295 (ranger_cache::entry_range): Ditto.
12296 (ranger_cache::fill_block_cache): Ditto.
12297 * gimple-range-cache.h (class ranger_cache): Remove poor value members.
12298 * gimple-range.cc (gimple_ranger::range_of_expr): Remove call.
12299 * gimple-range.h (class gimple_ranger): Adjust.
12301 2021-06-18 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
12304 * common/config/arm/arm-common.c (arm_canon_arch_option_1): New function
12305 derived from arm_canon_arch.
12306 (arm_canon_arch_option): Call it.
12307 (arm_canon_arch_multilib_option): New function.
12308 * config/arm/arm-cpus.in (IGNORE_FOR_MULTILIB): New fgroup.
12309 * config/arm/arm.h (arm_canon_arch_multilib_option): New prototype.
12310 (CANON_ARCH_MULTILIB_SPEC_FUNCTION): New macro.
12311 (MULTILIB_ARCH_CANONICAL_SPECS): New macro.
12312 (DRIVER_SELF_SPECS): Add MULTILIB_ARCH_CANONICAL_SPECS.
12313 * config/arm/arm.opt (mlibarch): New option.
12314 * config/arm/t-rmprofile (MULTILIB_MATCHES): For armv8*-m, replace use
12315 of march on RHS with mlibarch.
12317 2021-06-18 Marcel Vollweiler <marcel@codesourcery.com>
12319 * config.in: Regenerate.
12320 * config/gcn/gcn.c (print_operand_address): Fix for global_load assembler
12322 * configure: Regenerate.
12323 * configure.ac: Fix for global_load assembler functions.
12325 2021-06-18 Richard Biener <rguenther@suse.de>
12327 PR tree-optimization/101112
12328 * tree-vect-slp.c (vect_slp_linearize_chain): Fix condition
12329 to lookup a pattern stmt def.
12331 2021-06-18 Jakub Jelinek <jakub@redhat.com>
12333 PR middle-end/101062
12334 * stor-layout.c (finish_bitfield_layout): Don't add bitfield
12335 representatives in QUAL_UNION_TYPE.
12337 2021-06-18 Andrew Pinski <apinski@marvell.com>
12339 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
12340 Add counting of how many times it is done.
12341 (factor_out_conditional_conversion): Likewise.
12342 (match_simplify_replacement): Likewise.
12343 (value_replacement): Likewise.
12344 (spaceship_replacement): Likewise.
12345 (cond_store_replacement): Likewise.
12346 (cond_if_else_store_replacement_1): Likewise.
12347 (hoist_adjacent_loads): Likewise.
12349 2021-06-18 Andrew Pinski <apinski@marvell.com>
12351 * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
12352 types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and CNONJ_EXPR.
12353 (verify_gimple_assign_binary): Reject point and offset types on
12354 MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
12355 FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
12356 FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.
12358 2021-06-18 Michael Meissner <meissner@linux.ibm.com>
12360 * config/rs6000/rs6000.c (rs6000_emit_minmax): Add support for ISA
12361 3.1 IEEE 128-bit floating point xsmaxcqp/xsmincqp instructions.
12362 * config/rs6000/rs6000.md (s<minmax><mode>3, IEEE128 iterator):
12365 2021-06-17 Aaron Sawdey <acsawdey@linux.ibm.com>
12367 * config/rs6000/genfusion.pl (gen_logical_addsubf): Add
12368 earlyclobber to alts 0/1.
12369 (gen_addadd): Add earlyclobber to alts 0/1.
12370 * config/rs6000/fusion.md: Regenerate file.
12372 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12374 * cfgloopanal.c (get_loop_hot_path): Make path an auto_vec.
12376 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
12378 * gimple-range-cache.cc: Comment cleanups.
12379 * gimple-range-gori.cc: Comment cleanups.
12380 * gimple-range.cc: Comment/spacing cleanups
12381 * value-range.h: Comment cleanups.
12383 2021-06-17 H.J. Lu <hjl.tools@gmail.com>
12386 * calls.c (expand_call): Replace PUSH_ARGS with
12387 targetm.calls.push_argument (0).
12388 (emit_library_call_value_1): Likewise.
12389 * defaults.h (PUSH_ARGS): Removed.
12390 (PUSH_ARGS_REVERSED): Replace PUSH_ARGS with
12391 targetm.calls.push_argument (0).
12392 * expr.c (block_move_libcall_safe_for_call_parm): Likewise.
12393 (emit_push_insn): Pass the number bytes to push to
12394 targetm.calls.push_argument and pass 0 if ARGS_ADDR is 0.
12395 * hooks.c (hook_bool_uint_true): New.
12396 * hooks.h (hook_bool_uint_true): Likewise.
12397 * rtlanal.c (nonzero_bits1): Replace PUSH_ARGS with
12398 targetm.calls.push_argument (0).
12399 * target.def (push_argument): Add a targetm.calls hook.
12400 * targhooks.c (default_push_argument): New.
12401 * targhooks.h (default_push_argument): Likewise.
12402 * config/bpf/bpf.h (PUSH_ARGS): Removed.
12403 * config/cr16/cr16.c (TARGET_PUSH_ARGUMENT): New.
12404 * config/cr16/cr16.h (PUSH_ARGS): Removed.
12405 * config/i386/i386.c (ix86_push_argument): New.
12406 (TARGET_PUSH_ARGUMENT): Likewise.
12407 * config/i386/i386.h (PUSH_ARGS): Removed.
12408 * config/m32c/m32c.c (TARGET_PUSH_ARGUMENT): New.
12409 * config/m32c/m32c.h (PUSH_ARGS): Removed.
12410 * config/nios2/nios2.h (PUSH_ARGS): Likewise.
12411 * config/pru/pru.h (PUSH_ARGS): Likewise.
12412 * doc/tm.texi.in: Remove PUSH_ARGS documentation. Add
12413 TARGET_PUSH_ARGUMENT hook.
12414 * doc/tm.texi: Regenerated.
12416 2021-06-17 Uroš Bizjak <ubizjak@gmail.com>
12419 * config/i386/i386-expand.c (expand_vector_set_var):
12420 Handle V2FS mode remapping. Pass TARGET_MMX_WITH_SSE to
12421 ix86_expand_vector_init_duplicate.
12422 (ix86_expand_vector_init_duplicate): Emit insv_1 for
12423 QImode for !TARGET_PARTIAL_REG_STALL.
12424 * config/i386/predicates.md (vec_setm_mmx_operand): New predicate.
12425 * config/i386/mmx.md (vec_setv2sf): Use vec_setm_mmx_operand
12426 as operand 2 predicate. Call ix86_expand_vector_set_var
12427 for non-constant index operand.
12428 (vec_setv2si): Ditto.
12429 (vec_setv4hi): Ditto.
12430 (vec_setv8qi): ditto.
12432 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
12434 PR tree-optimization/100790
12435 * gimple-range.cc (range_of_builtin_call): Cleanup clz and ctz
12438 2021-06-17 Martin Liska <mliska@suse.cz>
12440 * doc/invoke.texi: Use consistently -O1 instead of -O.
12442 2021-06-17 Martin Liska <mliska@suse.cz>
12444 * gcov-io.h: Update documentation entry about string format.
12446 2021-06-17 Marius Hillenbrand <mhillen@linux.ibm.com>
12449 * config/s390/vecintrin.h (vec_doublee): Fix to use
12450 __builtin_s390_vflls.
12451 (vec_floate): Fix to use __builtin_s390_vflrd.
12453 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12455 * dominance.c (get_dominated_to_depth): Return auto_vec<basic_block>.
12456 * dominance.h (get_dominated_to_depth): Likewise.
12457 (get_all_dominated_blocks): Likewise.
12458 * cfgcleanup.c (delete_unreachable_blocks): Adjust.
12459 * gcse.c (hoist_code): Likewise.
12460 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
12461 * tree-parloops.c (oacc_entry_exit_ok): Likewise.
12462 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Likewise.
12463 * tree-ssa-phiprop.c (pass_phiprop::execute): Likewise.
12465 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12467 * dominance.c (get_dominated_by_region): Return auto_vec<basic_block>.
12468 * dominance.h (get_dominated_by_region): Likewise.
12469 * tree-cfg.c (gimple_duplicate_sese_region): Adjust.
12470 (gimple_duplicate_sese_tail): Likewise.
12471 (move_sese_region_to_fn): Likewise.
12473 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12475 * dominance.c (get_dominated_by): Return auto_vec<basic_block>.
12476 * dominance.h (get_dominated_by): Likewise.
12477 * auto-profile.c (afdo_find_equiv_class): Adjust.
12478 * cfgloopmanip.c (duplicate_loop_to_header_edge): Likewise.
12479 * loop-unroll.c (unroll_loop_runtime_iterations): Likewise.
12480 * tree-cfg.c (test_linear_chain): Likewise.
12481 (test_diamond): Likewise.
12483 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12485 * cfgloop.h (get_loop_hot_path): Return auto_vec<basic_block>.
12486 * cfgloopanal.c (get_loop_hot_path): Likewise.
12487 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
12489 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12491 * cgraph.c (cgraph_node::collect_callers): Return
12492 auto_vec<cgraph_edge *>.
12493 * cgraph.h (cgraph_node::collect_callers): Likewise.
12494 * ipa-cp.c (create_specialized_node): Adjust.
12495 (decide_about_value): Likewise.
12496 (decide_whether_version_node): Likewise.
12497 * ipa-sra.c (process_isra_node_results): Likewise.
12499 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
12501 * vec.h (vl_ptr>::using_auto_storage): Handle null m_vec.
12502 (auto_vec<T, 0>::auto_vec): Define move constructor, and delete copy
12504 (auto_vec<T, 0>::operator=): Define move assignment and delete copy
12507 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
12509 * gimple-range.cc (debug_seed_ranger): New.
12510 (dump_ranger): New.
12511 (debug_ranger): New.
12513 2021-06-17 Richard Biener <rguenther@suse.de>
12515 PR tree-optimization/54400
12516 * tree-vectorizer.h (enum slp_instance_kind): Add
12517 slp_inst_kind_bb_reduc.
12518 (reduction_fn_for_scalar_code): Declare.
12519 * tree-vect-data-refs.c (vect_slp_analyze_instance_dependence):
12520 Check SLP_INSTANCE_KIND instead of looking at the
12522 (vect_slp_analyze_instance_alignment): Likewise.
12523 * tree-vect-loop.c (reduction_fn_for_scalar_code): Export.
12524 * tree-vect-slp.c (vect_slp_linearize_chain): Split out
12525 chain linearization from vect_build_slp_tree_2 and generalize
12526 for the use of BB reduction vectorization.
12527 (vect_build_slp_tree_2): Adjust accordingly.
12528 (vect_optimize_slp): Elide permutes at the root of BB reduction
12530 (vectorizable_bb_reduc_epilogue): New function.
12531 (vect_slp_prune_covered_roots): Likewise.
12532 (vect_slp_analyze_operations): Use them.
12533 (vect_slp_check_for_constructors): Recognize associatable
12534 chains for BB reduction vectorization.
12535 (vectorize_slp_instance_root_stmt): Generate code for the
12536 BB reduction epilogue.
12538 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
12540 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Check with
12542 (gori_compute::may_recompute_p): New.
12543 (gori_compute::outgoing_edge_range_p): Perform recomputations.
12544 * gimple-range-gori.h (class gori_compute): Add prototype.
12546 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
12548 * gimple-range-cache.cc (ranger_cache::range_on_edge): Always return
12549 true when a range can be calculated.
12550 * gimple-range.cc (gimple_ranger::dump_bb): Check has_edge_range_p.
12552 2021-06-16 Martin Sebor <msebor@redhat.com>
12554 * doc/invoke.texi (-Wmismatched-dealloc, -Wmismatched-new-delete):
12555 Correct documented defaults.
12557 2021-06-16 Andrew MacLeod <amacleod@redhat.com>
12559 * gimple-range-cache.cc (ranger_cache::ranger_cache): Initialize
12560 m_new_value_p directly.
12562 2021-06-16 Uroš Bizjak <ubizjak@gmail.com>
12565 * config/i386/i386-expand.c (expand_vec_perm_2perm_pblendv):
12566 Handle 64bit modes for TARGET_SSE4_1.
12567 (expand_vec_perm_pshufb2): Handle 64bit modes for TARGET_SSSE3.
12568 (expand_vec_perm_even_odd_pack): Handle V4HI mode.
12569 (expand_vec_perm_even_odd_1) <case E_V4HImode>: Expand via
12570 expand_vec_perm_pshufb2 for TARGET_SSSE3 and via
12571 expand_vec_perm_even_odd_pack for TARGET_SSE4_1.
12572 * config/i386/mmx.md (mmx_packusdw): New insn pattern.
12574 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12576 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn<mode>):
12577 Change to an expander that emits the correct instruction
12578 depending on endianness.
12579 (aarch64_<sur><addsub>hn<mode>_insn_le): Define.
12580 (aarch64_<sur><addsub>hn<mode>_insn_be): Define.
12582 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12584 * config/aarch64/aarch64-simd-builtins.def: Split generator
12585 for aarch64_<su>qmovn builtins into scalar and vector
12587 * config/aarch64/aarch64-simd.md (aarch64_<su>qmovn<mode>_insn_le):
12589 (aarch64_<su>qmovn<mode>_insn_be): Define.
12590 (aarch64_<su>qmovn<mode>): Split into scalar and vector
12591 variants. Change vector variant to an expander that emits the
12592 correct instruction depending on endianness.
12594 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12596 * config/aarch64/aarch64-simd-builtins.def: Split generator
12597 for aarch64_sqmovun builtins into scalar and vector variants.
12598 * config/aarch64/aarch64-simd.md (aarch64_sqmovun<mode>):
12599 Split into scalar and vector variants. Change vector variant
12600 to an expander that emits the correct instruction depending
12602 (aarch64_sqmovun<mode>_insn_le): Define.
12603 (aarch64_sqmovun<mode>_insn_be): Define.
12605 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
12607 * config/aarch64/aarch64-simd.md (aarch64_xtn<mode>_insn_le):
12608 Define - modeling zero-high-half semantics.
12609 (aarch64_xtn<mode>): Change to an expander that emits the
12610 appropriate instruction depending on endianness.
12611 (aarch64_xtn<mode>_insn_be): Define - modeling zero-high-half
12613 (aarch64_xtn2<mode>_le): Rename to...
12614 (aarch64_xtn2<mode>_insn_le): This.
12615 (aarch64_xtn2<mode>_be): Rename to...
12616 (aarch64_xtn2<mode>_insn_be): This.
12617 (vec_pack_trunc_<mode>): Emit truncation instruction instead
12619 * config/aarch64/iterators.md (Vnarrowd): Add Vnarrowd mode
12620 attribute iterator.
12622 2021-06-16 Martin Jambor <mjambor@suse.cz>
12624 PR tree-optimization/100453
12625 * tree-sra.c (create_access): Disqualify any const candidates
12626 which are written to.
12627 (sra_modify_expr): Do not store sub-replacements back to a const base.
12628 (handle_unscalarized_data_in_subtree): Likewise.
12629 (sra_modify_assign): Likewise. Earlier, use TREE_READONLy test
12630 instead of constant_decl_p.
12632 2021-06-16 Jakub Jelinek <jakub@redhat.com>
12634 PR middle-end/101062
12635 * stor-layout.c (finish_bitfield_representative): For fields in unions
12636 assume nextf is always NULL.
12637 (finish_bitfield_layout): Compute bit field representatives also in
12638 unions, but handle it as if each bitfield was the only field in the
12641 2021-06-16 Richard Biener <rguenther@suse.de>
12643 PR tree-optimization/101088
12644 * tree-ssa-loop-im.c (sm_seq_valid_bb): Only look for
12645 supported refs on edges. Do not assert same ref but
12646 different kind stores are unsuported but mark them so.
12647 (hoist_memory_references): Only look for supported refs
12650 2021-06-16 Roger Sayle <roger@nextmovesoftware.com>
12652 PR rtl-optimization/46235
12653 * config/i386/i386.md: New define_split for bt followed by cmov.
12654 (*bt<mode>_setcqi): New define_insn_and_split for bt followed by setc.
12655 (*bt<mode>_setncqi): New define_insn_and_split for bt then setnc.
12656 (*bt<mode>_setnc<mode>): New define_insn_and_split for bt followed
12657 by setnc with zero extension.
12659 2021-06-16 Richard Biener <rguenther@suse.de>
12661 PR tree-optimization/101083
12662 * tree-vect-slp.c (vect_slp_build_two_operator_nodes): Get
12663 vectype as argument.
12664 (vect_build_slp_tree_2): Adjust.
12666 2021-06-15 Martin Sebor <msebor@redhat.com>
12668 PR middle-end/100876
12669 * builtins.c: (gimple_call_return_array): Account for size_t
12670 mangling as either unsigned int or unsigned long
12672 2021-06-15 Jeff Law <jeffreyalaw@gmail.com>
12674 * compare-elim.c (try_eliminate_compare): Run DCE to clean things
12675 up before eliminating comparisons.
12677 2021-06-15 Aldy Hernandez <aldyh@redhat.com>
12679 * range-op.cc (operator_bitwise_or::wi_fold): Make sure
12680 nonzero|X is nonzero.
12681 (range_op_bitwise_and_tests): Add tests for above.
12683 2021-06-15 Carl Love <cel@us.ibm.com>
12686 * config/rs6000/rs6000-builtin.def (VCMPEQUT): Fix the ICODE for the
12688 (VRLQ, VSLQ, VSRQ, VSRAQ): Remove unused BU_P10_OVERLOAD_2
12691 2021-06-15 Tobias Burnus <tobias@codesourcery.com>
12694 * gimplify.c (enum gimplify_defaultmap_kind): Add GDMK_SCALAR_TARGET.
12695 (struct gimplify_omp_ctx): Extend defaultmap array by one.
12696 (new_omp_context): Init defaultmap[GDMK_SCALAR_TARGET].
12697 (omp_notice_variable): Update type classification for Fortran.
12698 (gimplify_scan_omp_clauses): Update calls for new argument; handle
12699 GDMK_SCALAR_TARGET; for Fortran, GDMK_POINTER avoid GOVD_MAP_0LEN_ARRAY.
12700 * langhooks-def.h (lhd_omp_scalar_p): Add 'ptr_ok' argument.
12701 * langhooks.c (lhd_omp_scalar_p): Likewise.
12702 (LANG_HOOKS_OMP_ALLOCATABLE_P, LANG_HOOKS_OMP_SCALAR_TARGET_P): New.
12703 (LANG_HOOKS_DECLS): Add them.
12704 * langhooks.h (struct lang_hooks_for_decls): Add new hooks, update
12705 omp_scalar_p pointer type to include the new bool argument.
12707 2021-06-15 David Malcolm <dmalcolm@redhat.com>
12709 * doc/analyzer.texi
12710 (Special Functions for Debugging the Analyzer): Add
12711 __analyzer_dump_capacity.
12713 2021-06-15 Jakub Jelinek <jakub@redhat.com>
12716 * expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
12717 case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.
12719 2021-06-15 Richard Biener <rguenther@suse.de>
12721 * cfgloopanal.c (mark_irreducible_loops): Use a dominance
12722 check to identify loop latches.
12723 * cfgloop.c (verify_loop_structure): Likewise.
12724 * loop-init.c (apply_loop_flags): Allow marked irreducible
12725 regions even with multiple latches.
12726 * predict.c (rebuild_frequencies): Simplify.
12728 2021-06-15 Richard Biener <rguenther@suse.de>
12730 * tree-ssa-threadupdate.c
12731 (jump_thread_path_registry::mark_threaded_blocks): Assert we
12732 have marked irreducible regions.
12734 2021-06-14 Martin Sebor <msebor@redhat.com>
12737 * builtins.c (gimple_call_return_array): Check for attribute fn spec.
12738 Handle calls to placement new.
12739 (ndecl_dealloc_argno): Avoid placement delete.
12741 2021-06-14 Peter Bergner <bergner@linux.ibm.com>
12744 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
12745 create_tmp_reg_or_ssa_name().
12747 2021-06-14 Andrew MacLeod <amacleod@redhat.com>
12749 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
12750 (ranger_cache::enable_new_values): Set to specified value and
12751 return the old value.
12752 (ranger_cache::disable_new_values): Delete.
12753 (ranger_cache::fill_block_cache): Disable non 1st order derived
12755 * gimple-range-cache.h (ranger_cache): Adjust prototypes.
12756 * gimple-range.cc (gimple_ranger::range_of_expr): Adjust.
12758 2021-06-14 Uroš Bizjak <ubizjak@gmail.com>
12761 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
12762 Return true early when testing with V2HImode.
12763 * config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.
12765 2021-06-14 Christophe Lyon <christophe.lyon@linaro.org>
12767 * config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
12768 (mve_vec_unpack<US>_hi_<mode>): New pattern.
12769 (@mve_vec_pack_trunc_lo_<mode>): New pattern.
12770 (mve_vmovntq_<supf><mode>): Prefix with '@'.
12771 * config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
12773 (vec_unpack<US>_lo_<mode>): Likewise.
12774 (vec_pack_trunc_<mode>): Rename to
12775 neon_quad_vec_pack_trunc_<mode>.
12776 * config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
12778 (vec_unpack<US>_lo_<mode>): New.
12779 (vec_pack_trunc_<mode>): New.
12781 2021-06-14 Richard Biener <rguenther@suse.de>
12783 PR tree-optimization/100934
12784 * tree-ssa-dom.c (pass_dominator::execute): Properly
12785 mark irreducible regions.
12787 2021-06-14 Martin Liska <mliska@suse.cz>
12789 * doc/invoke.texi: Put r{...} on the same line as @item.
12791 2021-06-14 Martin Liska <mliska@suse.cz>
12793 * doc/invoke.texi: Add missing newline.
12795 2021-06-14 Martin Liska <mliska@suse.cz>
12797 * doc/invoke.texi: Remove '+' charasters.
12799 2021-06-14 Claudiu Zissulescu <claziss@synopsys.com>
12801 * config.gcc (arc): Add support for with_cpu option.
12802 * config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.
12804 2021-06-14 Richard Biener <rguenther@suse.de>
12806 PR tree-optimization/101031
12807 * tree-ssa-strlen.c (maybe_invalidate): Increment max_size
12808 instead of size when accounting for a possibly string
12811 2021-06-14 Martin Liska <mliska@suse.cz>
12813 * gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].
12815 2021-06-14 Aldy Hernandez <aldyh@redhat.com>
12817 * value-query.cc (gimple_range_global): Call get_range_global
12818 if called after inlining.
12820 2021-06-13 Uroš Bizjak <ubizjak@gmail.com>
12823 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
12824 Emit constant permutation insn directly from here.
12826 2021-06-13 Trevor Saunders <tbsaunde@tbsaunde.org>
12828 * attribs.c (find_attribute_namespace): Iterate over vec<> with
12830 * auto-profile.c (afdo_find_equiv_class): Likewise.
12831 * gcc.c (do_specs_vec): Likewise.
12832 (do_spec_1): Likewise.
12833 (driver::set_up_specs): Likewise.
12834 * gimple-loop-jam.c (any_access_function_variant_p): Likewise.
12835 * gimple-ssa-store-merging.c (compatible_load_p): Likewise.
12836 (imm_store_chain_info::try_coalesce_bswap): Likewise.
12837 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
12838 (get_location_for_stmts): Likewise.
12839 * graphite-poly.c (print_iteration_domains): Likewise.
12840 (free_poly_bb): Likewise.
12841 (remove_gbbs_in_scop): Likewise.
12842 (free_scop): Likewise.
12843 (dump_gbb_cases): Likewise.
12844 (dump_gbb_conditions): Likewise.
12845 (print_pdrs): Likewise.
12846 (print_scop): Likewise.
12847 * ifcvt.c (cond_move_process_if_block): Likewise.
12848 * lower-subreg.c (decompose_multiword_subregs): Likewise.
12849 * regcprop.c (pass_cprop_hardreg::execute): Likewise.
12850 * sanopt.c (sanitize_rewrite_addressable_params): Likewise.
12851 * sel-sched-dump.c (dump_insn_vector): Likewise.
12852 * store-motion.c (store_ops_ok): Likewise.
12853 (store_killed_in_insn): Likewise.
12854 * timevar.c (timer::named_items::print): Likewise.
12855 * tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
12856 (cleanup_tree_cfg_noloop): Likewise.
12857 * tree-data-ref.c (dump_data_references): Likewise.
12858 (print_dir_vectors): Likewise.
12859 (print_dist_vectors): Likewise.
12860 (dump_data_dependence_relations): Likewise.
12861 (dump_dist_dir_vectors): Likewise.
12862 (dump_ddrs): Likewise.
12863 (create_runtime_alias_checks): Likewise.
12864 (free_subscripts): Likewise.
12865 (save_dist_v): Likewise.
12866 (save_dir_v): Likewise.
12867 (invariant_access_functions): Likewise.
12868 (same_access_functions): Likewise.
12869 (access_functions_are_affine_or_constant_p): Likewise.
12870 (find_data_references_in_stmt): Likewise.
12871 (graphite_find_data_references_in_stmt): Likewise.
12872 (free_dependence_relations): Likewise.
12873 (free_data_refs): Likewise.
12874 * tree-inline.c (copy_debug_stmts): Likewise.
12875 * tree-into-ssa.c (dump_currdefs): Likewise.
12876 (rewrite_update_phi_arguments): Likewise.
12877 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
12878 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
12880 (vect_slp_analyze_node_dependences): Likewise.
12881 (vect_slp_analyze_instance_dependence): Likewise.
12882 (vect_record_base_alignments): Likewise.
12883 (vect_get_peeling_costs_all_drs): Likewise.
12884 (vect_peeling_supportable): Likewise.
12885 * tree-vectorizer.c (vec_info::~vec_info): Likewise.
12886 (vec_info::free_stmt_vec_infos): Likewise.
12888 2021-06-13 Jeff Law <jeffreyalaw@gmail.com>
12890 * config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
12891 (andqi3_1<cczn>): Removed.
12892 (<ors>qi3_1): Do not split for IOR/XOR a single bit.
12893 (H8/SX bit logicals): Split out from other patterns.
12894 * config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
12895 mulqihi3_const_clobber_flags.
12896 (mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly
12898 2021-06-13 H.J. Lu <hjl.tools@gmail.com>
12901 * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
12902 to true if red zone is used.
12903 (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
12904 ix86_red_zone_used.
12905 * config/i386/i386.h (machine_function): Add red_zone_used.
12906 (ix86_red_zone_size): Removed.
12907 (ix86_red_zone_used): New.
12908 * config/i386/i386.md (peephole2 patterns): Replace
12909 ix86_red_zone_size with ix86_red_zone_used.
12911 2021-06-12 Jason Merrill <jason@redhat.com>
12913 * doc/extend.texi (unused variable attribute): Applies to
12914 structure fields as well.
12916 2021-06-12 Eugene Rozenfeld <erozen@microsoft.com>
12918 * auto-profile.c (read_profile): fix a typo in an error string
12920 2021-06-11 Thomas Schwinge <thomas@codesourcery.com>
12922 * tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
12924 * tree-pretty-print.c (dump_omp_clauses): Update.
12925 (dump_generic_node) <OMP_CLAUSE>: Use it.
12927 2021-06-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
12930 * config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
12931 int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
12932 the polymorphic variants matching code.
12933 (__arm_vld1q_z): Likewise.
12934 (__arm_vld2q): Likewise.
12935 (__arm_vld4q): Likewise.
12936 (__arm_vldrbq_gather_offset): Likewise.
12937 (__arm_vldrbq_gather_offset_z): Likewise.
12939 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
12941 PR tree-optimization/96392
12942 * fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.
12944 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
12946 PR tree-optimization/96392
12947 * fold-const.c (fold_real_zero_addition_p): Take both arguments
12948 of the addition or subtraction, not just the zero. Use this
12949 other argument in tests for signaling NaNs and signed zeros.
12950 (tree_expr_maybe_real_minus_zero_p): New predicate.
12951 * fold-const.h (fold_real_zero_addition_p): Update prototype.
12952 (tree_expr_maybe_real_minus_zero_p): New function prototype.
12953 * match.pd: Update calls to fold_real_zero_addition_p.
12954 Replace HONOR_NANS with tree_expr_maybe_nan_p.
12955 Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
12956 Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
12957 * tree-ssa-reassoc.c (eliminate_using_constants): Update
12958 call to fold_real_zero_addition_p.
12960 2021-06-11 Richard Biener <rguenther@suse.de>
12962 PR tree-optimization/101025
12963 * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
12964 all refs that require dependence checking.
12966 2021-06-11 Richard Biener <rguenther@suse.de>
12968 PR tree-optimization/101028
12969 * tree-vect-slp.c (vect_build_slp_tree_2): When SLP
12970 reassoc discovery fails fatally, mark appropriate lanes
12973 2021-06-11 Richard Biener <rguenther@suse.de>
12975 PR tree-optimization/101026
12976 * tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
12977 have a representative for the associated chain nodes.
12979 2021-06-11 Jakub Jelinek <jakub@redhat.com>
12981 PR rtl-optimization/101008
12982 * simplify-rtx.c (relational_result): New function.
12983 (simplify_logical_relational_operation,
12984 simplify_relational_operation): Use it.
12986 2021-06-11 Jakub Jelinek <jakub@redhat.com>
12989 * config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.
12991 2021-06-11 Uroš Bizjak <ubizjak@gmail.com>
12994 * config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
12995 false if the permutation can be implemented with constant
12996 permutation instruction in wider mode.
12997 (canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
12998 Handle V8QImode and V4HImode.
13000 2021-06-11 Martin Liska <mliska@suse.cz>
13002 PR gcov-profile/100788
13003 * common.opt: Add new option.
13004 * coverage.c (coverage_begin_function): Emit warning instead on
13005 the internal compiler error.
13006 * doc/invoke.texi: Document the option.
13007 * toplev.c (process_options): Enable it by default.
13009 2021-06-11 Richard Biener <rguenther@suse.de>
13011 PR middle-end/101009
13012 * tree-data-ref.c (build_classic_dist_vector_1): Make sure
13013 to set *init_b to true when we encounter a constant equal
13015 (compute_affine_dependence): Also dump the actual DR_REF.
13017 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
13019 PR tree-optimization/100984
13020 * gimple-ssa-evrp.c (ssa_equiv_stack): Use auto_vec for
13021 replacements table.
13022 (ssa_equiv_stack::~ssa_equiv_stack): Remove.
13024 2021-06-11 Kewen Lin <linkw@linux.ibm.com>
13026 * config/rs6000/rs6000.md
13027 (floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
13028 define_insn_and_split.
13030 2021-06-11 Richard Biener <rguenther@suse.de>
13032 * tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
13033 to sort operands of the associative chain.
13035 2021-06-11 Richard Biener <rguenther@suse.de>
13037 * system.h (gcc_stablesort_r): Declare.
13038 * sort.cc (gcc_sort_r): Support stable sort.
13039 (gcc_stablesort_r): Define.
13040 * vec.h (vec<>::stablesort): Add.
13042 2021-06-10 Uroš Bizjak <ubizjak@gmail.com>
13045 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
13046 Handle V2SF mode. Emit SHUFPS to fixup unpack-high for V2SF mode.
13047 (expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
13048 (expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
13049 (expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
13050 (expand_vec_perm_interleave2): Handle 64bit modes.
13051 (expand_vec_perm_even_odd_pack): Handle V8QI mode.
13052 (expand_vec_perm_even_odd_1): Ditto.
13053 (ix86_vectorize_vec_perm_const): Ditto.
13054 * config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
13055 * config/i386/sse.md: ... here.
13056 * config/i386/mmx.md (*vec_interleave_lowv2sf):
13057 New insn_and_split pattern.
13058 (*vec_interleave_highv2sf): Ditto.
13059 (mmx_pshufbv8qi3): New insn pattern.
13060 (*mmx_pblendw): Ditto.
13062 2021-06-10 Peter Bergner <bergner@linux.ibm.com>
13064 * config/rs6000/rs6000-builtin.def (build_pair): New built-in.
13065 (build_acc): Likewise.
13066 * config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
13067 source operands in little-endian mode.
13068 (rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
13069 (mma_init_builtins): Likewise.
13070 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
13071 ordering for the MMA assemble and build source operands.
13072 * doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
13074 (__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
13077 2021-06-10 Jeff Law <jeffreyalaw@gmail.com>
13079 * config/h8300/h8300.c (select_cc_mode): Handle MEM. Use
13081 * config/h8300/extensions.md: Replace _clobber_flags patterns
13084 2021-06-10 Robin Dapp <rdapp@linux.ibm.com>
13086 * config/s390/vector.md (vcond_mask_<mode><mode>): Change to
13087 (vcond_mask_<mode><tointvec>): this.
13089 2021-06-10 Andrew Stubbs <ams@codesourcery.com>
13090 Thomas Schwinge <thomas@codesourcery.com>
13092 * omp-builtins.def (BUILT_IN_GOACC_ENTER_EXIT_DATA): Split into...
13093 (BUILT_IN_GOACC_ENTER_DATA, BUILT_IN_GOACC_EXIT_DATA): ... these.
13094 * gimple.h (enum gf_mask): Split
13095 'GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA' into
13096 'GF_OMP_TARGET_KIND_OACC_ENTER_DATA' and
13097 'GF_OMP_TARGET_KIND_OACC_EXIT_DATA'.
13098 (is_gimple_omp_oacc): Update.
13099 * gimple-pretty-print.c (dump_gimple_omp_target): Likewise.
13100 * gimplify.c (gimplify_omp_target_update): Likewise.
13101 * omp-expand.c (expand_omp_target, build_omp_regions_1)
13102 (omp_make_gimple_edges): Likewise.
13103 * omp-low.c (check_omp_nesting_restrictions, lower_omp_target):
13106 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
13108 * value-query.cc (value_query::value_on_edge): Rename name to
13110 (range_query::range_on_edge): Same.
13111 (range_query::value_of_expr): Same.
13112 (range_query::value_on_edge): Same.
13113 * value-query.h (class value_query): Same.
13114 (class range_query): Same.
13116 2021-06-10 Richard Biener <rguenther@suse.de>
13118 PR tree-optimization/101003
13119 * tree-vect-slp.c (vect_build_slp_tree_2): Appropriately
13120 use the pattern stmt defs when linearizing a chain.
13122 2021-06-10 Jakub Jelinek <jakub@redhat.com>
13125 * ifcvt.c (noce_get_alt_condition, noce_try_abs): Use
13126 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
13128 2021-06-10 Clement Chigot <clement.chigot@atos.net>
13130 * config/rs6000/aix71.h (ASM_CPU_SPEC): Add Power10 directive.
13131 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
13133 2021-06-09 Andrew Pinski <apinski@marvell.com>
13135 PR tree-optimization/100925
13136 * match.pd (a ? CST1 : CST2): Limit transformations
13137 that would produce a negative to integeral types only.
13138 Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.
13140 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
13143 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
13145 * doc/tm.texi: Correctly update.
13147 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
13149 * doc/tm.texi: Correctly update.
13151 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
13154 * doc/tm.texi.in (Trampolines): Add a missing blank line.
13156 2021-06-09 Paul Eggert <eggert@cs.ucla.edu>
13159 * doc/invoke.texi (Code Gen Options); Document that -fno-trampolines
13160 and -ftrampolines work only with Ada.
13161 * doc/tm.texi.in (Trampolines): Likewise.
13162 * doc/tm.texi: Regenerated.
13164 2021-06-09 Carl Love <cel@us.ibm.com>
13166 * config/rs6000/altivec.h (vec_signextll, vec_signexti, vec_signextq):
13167 Add define for new builtins.
13168 * config/rs6000/altivec.md(altivec_vreveti2): Add define_expand.
13169 * config/rs6000/rs6000-builtin.def (VSIGNEXTI, VSIGNEXTLL): Add
13170 overloaded builtin definitions.
13171 (VSIGNEXTSB2W, VSIGNEXTSH2W, VSIGNEXTSB2D, VSIGNEXTSH2D,VSIGNEXTSW2D,
13172 VSIGNEXTSD2Q): Add builtin expansions.
13173 (SIGNEXT): Add P10 overload definition.
13174 * config/rs6000/rs6000-call.c (P9V_BUILTIN_VEC_VSIGNEXTI, P9V_BUILTIN_VEC_VSIGNEXTLL,
13175 P10_BUILTIN_VEC_SIGNEXT): Add overloaded argument definitions.
13176 * config/rs6000/vsx.md (vsx_sign_extend_v2di_v1ti): Add define_insn.
13177 (vsignextend_v2di_v1ti, vsignextend_qi_<mode>, vsignextend_hi_<mode>,
13178 vsignextend_si_v2di)[VIlong]: Add define_expand.
13179 Make define_insn vsx_sign_extend_si_v2di visible.
13180 * doc/extend.texi: Add documentation for the vec_signexti,
13181 vec_signextll builtins and vec_signextq.
13183 2021-06-09 Carl Love <cel@us.ibm.com>
13185 * config/rs6000/rs6000.c (__fixkfti, __fixunskfti, __floattikf,
13186 __floatuntikf): Names changed to __fixkfti_sw, __fixunskfti_sw,
13187 __floattikf_sw, __floatuntikf_sw respectively.
13188 * config/rs6000/rs6000.md (floatti<mode>2, floatunsti<mode>2,
13189 fix_trunc<mode>ti2, fixuns_trunc<mode>ti2): Add
13190 define_insn for mode IEEE 128.
13192 2021-06-09 Carl Love <cel@us.ibm.com>
13194 * config/rs6000/altivec.md (altivec_vslq, altivec_vsrq):
13195 Rename to altivec_vslq_<mode>, altivec_vsrq_<mode>, mode VEC_TI.
13196 * config/rs6000/vector.md (VEC_TI): Was named VSX_TI in vsx.md.
13197 (vashlv1ti3): Change to vashl<mode>3, mode VEC_TI.
13198 (vlshrv1ti3): Change to vlshr<mode>3, mode VEC_TI.
13199 * config/rs6000/vsx.md (VSX_TI): Remove define_mode_iterator. Update
13200 uses of VSX_TI to VEC_TI.
13202 2021-06-09 Carl Love <cel@us.ibm.com>
13204 * config/rs6000/dfp.md (floattitd2, fixtdti2): New define_insns.
13206 2021-06-09 Carl Love <cel@us.ibm.com>
13208 * config/rs6000/altivec.h (vec_dive, vec_mod): Add define for new
13210 * config/rs6000/altivec.md (UNSPEC_VMULEUD, UNSPEC_VMULESD,
13211 UNSPEC_VMULOUD, UNSPEC_VMULOSD): New unspecs.
13212 (altivec_eqv1ti, altivec_gtv1ti, altivec_gtuv1ti, altivec_vmuleud,
13213 altivec_vmuloud, altivec_vmulesd, altivec_vmulosd, altivec_vrlq,
13214 altivec_vrlqmi, altivec_vrlqmi_inst, altivec_vrlqnm,
13215 altivec_vrlqnm_inst, altivec_vslq, altivec_vsrq, altivec_vsraq,
13216 altivec_vcmpequt_p, altivec_vcmpgtst_p, altivec_vcmpgtut_p): New
13218 (vec_widen_umult_even_v2di, vec_widen_smult_even_v2di,
13219 vec_widen_umult_odd_v2di, vec_widen_smult_odd_v2di, altivec_vrlqmi,
13220 altivec_vrlqnm): New define_expands.
13221 * config/rs6000/rs6000-builtin.def (VCMPEQUT_P, VCMPGTST_P,
13222 VCMPGTUT_P): Add macro expansions.
13223 (BU_P10V_AV_P): Add builtin predicate definition.
13224 (VCMPGTUT, VCMPGTST, VCMPEQUT, CMPNET, CMPGE_1TI,
13225 CMPGE_U1TI, CMPLE_1TI, CMPLE_U1TI, VNOR_V1TI_UNS, VNOR_V1TI, VCMPNET_P,
13226 VCMPAET_P, VMULEUD, VMULESD, VMULOUD, VMULOSD, VRLQ,
13227 VSLQ, VSRQ, VSRAQ, VRLQNM, DIV_V1TI, UDIV_V1TI, DIVES_V1TI, DIVEU_V1TI,
13228 MODS_V1TI, MODU_V1TI, VRLQMI): New macro expansions.
13229 (VRLQ, VSLQ, VSRQ, VSRAQ, DIVE, MOD): New overload expansions.
13230 * config/rs6000/rs6000-call.c (P10_BUILTIN_VCMPEQUT,
13231 P10V_BUILTIN_CMPGE_1TI, P10V_BUILTIN_CMPGE_U1TI,
13232 P10V_BUILTIN_VCMPGTUT, P10V_BUILTIN_VCMPGTST,
13233 P10V_BUILTIN_CMPLE_1TI, P10V_BUILTIN_VCMPLE_U1TI,
13234 P10V_BUILTIN_DIV_V1TI, P10V_BUILTIN_UDIV_V1TI,
13235 P10V_BUILTIN_VMULESD, P10V_BUILTIN_VMULEUD,
13236 P10V_BUILTIN_VMULOSD, P10V_BUILTIN_VMULOUD,
13237 P10V_BUILTIN_VNOR_V1TI, P10V_BUILTIN_VNOR_V1TI_UNS,
13238 P10V_BUILTIN_VRLQ, P10V_BUILTIN_VRLQMI,
13239 P10V_BUILTIN_VRLQNM, P10V_BUILTIN_VSLQ,
13240 P10V_BUILTIN_VSRQ, P10V_BUILTIN_VSRAQ,
13241 P10V_BUILTIN_VCMPGTUT_P, P10V_BUILTIN_VCMPGTST_P,
13242 P10V_BUILTIN_VCMPEQUT_P, P10V_BUILTIN_VCMPGTUT_P,
13243 P10V_BUILTIN_VCMPGTST_P, P10V_BUILTIN_CMPNET,
13244 P10V_BUILTIN_VCMPNET_P, P10V_BUILTIN_VCMPAET_P,
13245 P10V_BUILTIN_DIVES_V1TI, P10V_BUILTIN_MODS_V1TI,
13246 P10V_BUILTIN_MODU_V1TI):
13247 New overloaded definitions.
13248 (rs6000_gimple_fold_builtin) [P10V_BUILTIN_VCMPEQUT,
13249 P10V_BUILTIN_CMPNET, P10V_BUILTIN_CMPGE_1TI,
13250 P10V_BUILTIN_CMPGE_U1TI, P10V_BUILTIN_VCMPGTUT,
13251 P10V_BUILTIN_VCMPGTST, P10V_BUILTIN_CMPLE_1TI,
13252 P10V_BUILTIN_CMPLE_U1TI]: New case statements.
13253 (rs6000_init_builtins) [bool_V1TI_type_node, int_ftype_int_v1ti_v1ti]:
13255 (altivec_init_builtins): New E_V1TImode case statement.
13256 (builtin_function_type)[P10_BUILTIN_128BIT_VMULEUD,
13257 P10_BUILTIN_128BIT_VMULOUD, P10_BUILTIN_128BIT_DIVEU_V1TI,
13258 P10_BUILTIN_128BIT_MODU_V1TI, P10_BUILTIN_CMPGE_U1TI,
13259 P10_BUILTIN_VCMPGTUT, P10_BUILTIN_VCMPEQUT]: New case statements.
13260 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute) [E_TImode,
13261 E_V1TImode]: New case statements.
13262 * config/rs6000/rs6000.h (rs6000_builtin_type_index): New enum
13263 value RS6000_BTI_bool_V1TI.
13264 * config/rs6000/vector.md (vector_gtv1ti,vector_nltv1ti,
13265 vector_gtuv1ti, vector_nltuv1ti, vector_ngtv1ti, vector_ngtuv1ti,
13266 vector_eq_v1ti_p, vector_ne_v1ti_p, vector_ae_v1ti_p,
13267 vector_gt_v1ti_p, vector_gtu_v1ti_p, vrotlv1ti3, vashlv1ti3,
13268 vlshrv1ti3, vashrv1ti3): New define_expands.
13269 * config/rs6000/vsx.md (UNSPEC_VSX_DIVSQ, UNSPEC_VSX_DIVUQ,
13270 UNSPEC_VSX_DIVESQ, UNSPEC_VSX_DIVEUQ, UNSPEC_VSX_MODSQ,
13271 UNSPEC_VSX_MODUQ): New unspecs.
13272 (mulv2di3, vsx_div_v1ti, vsx_udiv_v1ti, vsx_dives_v1ti,
13273 vsx_diveu_v1ti, vsx_mods_v1ti, vsx_modu_v1ti, xxswapd_v1ti): New
13275 (vcmpnet): New define_expand.
13276 * doc/extend.texi: Add documentation for the new builtins vec_rl,
13277 vec_rlmi, vec_rlnm, vec_sl, vec_sr, vec_sra, vec_mule, vec_mulo,
13278 vec_div, vec_dive, vec_mod, vec_cmpeq, vec_cmpne, vec_cmpgt, vec_cmplt,
13279 vec_cmpge, vec_cmple, vec_all_eq, vec_all_ne, vec_all_gt, vec_all_lt,
13280 vec_all_ge, vec_all_le, vec_any_eq, vec_any_ne, vec_any_gt, vec_any_lt,
13281 vec_any_ge, vec_any_le.
13283 2021-06-09 Carl Love <cel@us.ibm.com>
13285 * config/rs6000/altivec.md (altivec_vrl<VI_char>mi): Fix
13286 bug in argument generation.
13288 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
13290 * config/arm/iterators.md (<supf>): Remove VCLZQ_U, VCLZQ_S.
13292 * config/arm/mve.md (mve_vclzq_<supf><mode>): Add '@' prefix,
13293 remove <supf> iterator.
13294 (mve_vclzq_u<mode>): New.
13295 * config/arm/neon.md (clz<mode>2): Rename to neon_vclz<mode>.
13296 (neon_vclz<mode): Move to ...
13297 * config/arm/unspecs.md (VCLZQ_U, VCLZQ_S): Remove.
13298 * config/arm/vec-common.md: ... here. Add support for MVE.
13300 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
13302 * config/arm/mve.md (mve_vhaddq_<supf><mode>): Prefix with '@'.
13303 (@mve_vrhaddq_<supf><mode): Likewise.
13304 * config/arm/neon.md (neon_v<r>hadd<sup><mode>): Likewise.
13305 * config/arm/vec-common.md (avg<mode>3_floor, uavg<mode>3_floor)
13306 (avg<mode>3_ceil", uavg<mode>3_ceil): New patterns.
13308 2021-06-09 imba-tjd <109224573@qq.com>
13310 * doc/invoke.texi: Fix typo.
13312 2021-06-09 Roger Sayle <roger@nextmovesoftware.com>
13314 PR middle-end/53267
13315 * fold-const-call.c (fold_const_call_sss) [CASE_CFN_FMOD]:
13316 Support evaluation of fmod/fmodf/fmodl at compile-time.
13318 2021-06-09 Richard Biener <rguenther@suse.de>
13320 PR tree-optimization/100981
13321 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
13322 gimple_get_lhs to also handle calls.
13323 * tree-vect-slp-patterns.c (complex_pattern::build): Transfer
13326 2021-06-09 Richard Biener <rguenther@suse.de>
13328 PR tree-optimization/97832
13329 * tree-vectorizer.h (_slp_tree::failed): New.
13330 * tree-vect-slp.c (_slp_tree::_slp_tree): Initialize
13332 (_slp_tree::~_slp_tree): Free failed.
13333 (vect_build_slp_tree): Retain failed nodes and record
13334 matches in them, copying that back out when running
13335 into a cached fail. Dump start and end of discovery.
13336 (dt_sort_cmp): New.
13337 (vect_build_slp_tree_2): Handle associatable chains
13338 together doing more aggressive operand swapping.
13340 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
13343 * config.gcc (gcc_cv_initfini_array): Set to yes for Linux and
13345 * doc/install.texi: Require glibc 2.1 and binutils 2.12 for
13346 Linux and GNU targets.
13348 2021-06-09 Richard Biener <rguenther@suse.de>
13350 * tree-vect-stmts.c (vect_is_simple_use): Always get dt
13353 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
13355 * config/arc/arc.md (loop_end): Change it to
13356 define_insn_and_split.
13358 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
13360 * config/arc/arc.md (maddhisi4): Use VMAC2H instruction.
13361 (machi): New pattern.
13362 (umaddhisi4): Use VMAC2HU instruction.
13363 (umachi): New pattern.
13365 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
13367 * config/arc/arc-protos.h (arc_split_move_p): New prototype.
13368 * config/arc/arc.c (arc_split_move_p): New function.
13369 (arc_split_move): Clean up.
13370 * config/arc/arc.md (movdi_insn): Clean up, use arc_split_move_p.
13371 (movdf_insn): Likewise.
13372 * config/arc/simdext.md (mov<VWH>_insn): Likewise.
13374 2021-06-09 Uroš Bizjak <ubizjak@gmail.com>
13377 * config/i386/i386.c (print_operand_address_as): Rename "no_rip"
13378 argument to "raw". Do not emit segment overrides when "raw" is true.
13380 2021-06-09 Martin Liska <mliska@suse.cz>
13382 * doc/gcov.texi: Create a proper JSON files.
13383 * doc/invoke.texi: Remove dots in order to make it a valid
13386 2021-06-09 Xionghu Luo <luoxhu@linux.ibm.com>
13388 * config/rs6000/rs6000-p8swap.c (pattern_is_rotate64): New.
13389 (insn_is_load_p): Use pattern_is_rotate64.
13390 (insn_is_swap_p): Likewise.
13391 (quad_aligned_load_p): Likewise.
13392 (const_load_sequence_p): Likewise.
13393 (replace_swapped_aligned_load): Likewise.
13394 (recombine_lvx_pattern): Likewise.
13395 (recombine_stvx_pattern): Likewise.
13397 2021-06-09 Andrew MacLeod <amacleod@redhat.com>
13399 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Use a
13400 fur_stmt source record.
13401 * gimple-range.cc (fur_source::get_operand): Generic range query.
13402 (fur_source::get_phi_operand): New.
13403 (fur_source::register_dependency): New.
13404 (fur_source::query): New.
13405 (class fur_edge): New. Edge source for operands.
13406 (fur_edge::fur_edge): New.
13407 (fur_edge::get_operand): New.
13408 (fur_edge::get_phi_operand): New.
13409 (fur_edge::query): New.
13410 (fur_stmt::fur_stmt): New.
13411 (fur_stmt::get_operand): New.
13412 (fur_stmt::get_phi_operand): New.
13413 (fur_stmt::query): New.
13414 (class fur_depend): New. Statement source and process dependencies.
13415 (fur_depend::fur_depend): New.
13416 (fur_depend::register_dependency): New.
13417 (class fur_list): New. List source for operands.
13418 (fur_list::fur_list): New.
13419 (fur_list::get_operand): New.
13420 (fur_list::get_phi_operand): New.
13421 (fold_range): New. Instantiate appropriate fur_source class and fold.
13422 (fold_using_range::range_of_range_op): Use new API.
13423 (fold_using_range::range_of_address): Ditto.
13424 (fold_using_range::range_of_phi): Ditto.
13425 (imple_ranger::fold_range_internal): Use fur_depend class.
13426 (fold_using_range::range_of_ssa_name_with_loop_info): Use new API.
13427 * gimple-range.h (class fur_source): Now a base class.
13428 (class fur_stmt): New.
13429 (fold_range): New prototypes.
13430 (fur_source::fur_source): Delete.
13432 2021-06-08 Andrew Pinski <apinski@marvell.com>
13434 PR tree-optimization/25290
13435 * tree-ssa-phiopt.c (xor_replacement): Delete.
13436 (tree_ssa_phiopt_worker): Delete use of xor_replacement.
13437 (match_simplify_replacement): Allow one cheap preparation
13438 statement that can be moved to before the if.
13440 2021-06-08 Pat Haugen <pthaugen@linux.ibm.com>
13442 * config/rs6000/power10.md (power10-fused-load, power10-fused-store,
13443 power10-fused_alu, power10-fused-vec, power10-fused-branch): New.
13445 2021-06-08 Jeff Law <jeffreyalaw@gmail.com>
13447 * config/h8300/logical.md (andqi3_1): Move BCLR case into define_insn_and_split.
13448 Create length attribute on define_insn_and_split. Only split for cases which we
13450 (andqi3_1<cczn>): Renamed from andqi3_1_clobber_flags. Only handle AND here and
13451 fix length computation.
13452 (b<code><mode>msx): Combine QImode and HImode H8/SX patterns using iterator.
13454 2021-06-08 Richard Biener <rguenther@suse.de>
13456 PR tree-optimization/100923
13457 * tree-ssa-sccvn.c (valueize_refs_1): Take a pointer to
13458 the operand vector to be valueized.
13459 (valueize_refs): Likewise.
13460 (valueize_shared_reference_ops_from_ref): Adjust.
13461 (valueize_shared_reference_ops_from_call): Likewise.
13462 (vn_reference_lookup_3): Likewise.
13463 (vn_reference_lookup_pieces): Likewise. Re-valueize
13464 with honoring availability when we are about to create
13465 the ao_ref and valueized before.
13466 (vn_reference_lookup): Likewise.
13467 (vn_reference_insert_pieces): Adjust.
13469 2021-06-08 Richard Biener <rguenther@suse.de>
13471 * tree-vectorizer.h (_slp_instance::root_stmt): Change to...
13472 (_slp_instance::root_stmts): ... a vector.
13473 (SLP_INSTANCE_ROOT_STMT): Rename to ...
13474 (SLP_INSTANCE_ROOT_STMTS): ... this.
13475 (slp_root::root): Change to...
13476 (slp_root::roots): ... a vector.
13477 (slp_root::slp_root): Adjust.
13478 * tree-vect-slp.c (_slp_instance::location): Adjust.
13479 (vect_free_slp_instance): Release the root stmt vector.
13480 (vect_build_slp_instance): Adjust.
13481 (vect_analyze_slp): Likewise.
13482 (_bb_vec_info::~_bb_vec_info): Likewise.
13483 (vect_slp_analyze_operations): Likewise.
13484 (vect_bb_vectorization_profitable_p): Likewise. Adjust
13485 costs for the root stmt.
13486 (vect_slp_check_for_constructors): Gather all BIT_INSERT_EXPRs
13488 (vect_slp_analyze_bb_1): Simplify by marking all root stmts
13490 (vectorize_slp_instance_root_stmt): Adjust.
13491 (vect_schedule_slp): Likewise.
13493 2021-06-08 Aldy Hernandez <aldyh@redhat.com>
13495 * gimple-ssa-evrp.c (class ssa_equiv_stack): New.
13496 (ssa_equiv_stack::ssa_equiv_stack): New.
13497 (ssa_equiv_stack::~ssa_equiv_stack): New.
13498 (ssa_equiv_stack::enter): New.
13499 (ssa_equiv_stack::leave): New.
13500 (ssa_equiv_stack::push_replacement): New.
13501 (ssa_equiv_stack::get_replacement): New.
13502 (is_pointer_ssa): New.
13503 (class pointer_equiv_analyzer): New.
13504 (pointer_equiv_analyzer::pointer_equiv_analyzer): New.
13505 (pointer_equiv_analyzer::~pointer_equiv_analyzer): New.
13506 (pointer_equiv_analyzer::set_global_equiv): New.
13507 (pointer_equiv_analyzer::set_cond_equiv): New.
13508 (pointer_equiv_analyzer::get_equiv): New.
13509 (pointer_equiv_analyzer::enter): New.
13510 (pointer_equiv_analyzer::leave): New.
13511 (pointer_equiv_analyzer::get_equiv_expr): New.
13512 (pta_valueize): New.
13513 (pointer_equiv_analyzer::visit_stmt): New.
13514 (pointer_equiv_analyzer::visit_edge): New.
13515 (hybrid_folder::value_of_expr): Call PTA.
13516 (hybrid_folder::value_on_edge): Same.
13517 (hybrid_folder::pre_fold_bb): New.
13518 (hybrid_folder::post_fold_bb): New.
13519 (hybrid_folder::pre_fold_stmt): New.
13520 (rvrp_folder::pre_fold_bb): New.
13521 (rvrp_folder::post_fold_bb): New.
13522 (rvrp_folder::pre_fold_stmt): New.
13523 (rvrp_folder::value_of_expr): Call PTA.
13524 (rvrp_folder::value_on_edge): Same.
13526 2021-06-08 Jakub Jelinek <jakub@redhat.com>
13529 * tree-inline.c (copy_tree_body_r): For OMP_CLAUSE_DEPEND don't
13530 check TREE_CODE if OMP_CLAUSE_DECL is NULL.
13532 2021-06-08 Richard Biener <rguenther@suse.de>
13534 PR middle-end/100951
13535 * tree-vect-generic.c (expand_vector_piecewise): Build a
13536 VECTOR_CST if all elements are constant.
13537 (expand_vector_condition): Likewise.
13538 (lower_vec_perm): Likewise.
13539 (expand_vector_conversion): Likewise.
13541 2021-06-08 Martin Liska <mliska@suse.cz>
13543 * doc/invoke.texi: Document new param evrp-sparse-threshold.
13545 2021-06-08 Martin Liska <mliska@suse.cz>
13547 * genautomata.c (create_automata): Fix typo.
13549 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
13551 PR tree-optimization/100794
13552 * tree-predcom.c (tree_predictive_commoning_loop): Add parameter
13553 allow_unroll_p and only allow unrolling when it's true.
13554 (tree_predictive_commoning): Add parameter allow_unroll_p and
13556 (run_tree_predictive_commoning): Likewise.
13557 (pass_predcom::gate): Check flag_tree_loop_vectorize and
13558 global_options_set.x_flag_predictive_commoning.
13559 (pass_predcom::execute): Adjust for allow_unroll_p.
13561 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
13563 * tree-predcom.c (execute_pred_commoning): Remove update_ssa call.
13564 (tree_predictive_commoning_loop): Factor some cleanup stuffs into
13565 lambda function cleanup, remove scev_reset call, and adjust return
13567 (tree_predictive_commoning): Adjust for different changed values,
13568 only set flag TODO_update_ssa_only_virtuals if changed.
13569 (pass_data pass_data_predcom): Remove TODO_update_ssa_only_virtuals
13570 from todo_flags_finish.
13572 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
13574 * gimple-range-cache.cc (class sbr_sparse_bitmap): New.
13575 (sbr_sparse_bitmap::sbr_sparse_bitmap): New.
13576 (sbr_sparse_bitmap::bitmap_set_quad): New.
13577 (sbr_sparse_bitmap::bitmap_get_quad): New.
13578 (sbr_sparse_bitmap::set_bb_range): New.
13579 (sbr_sparse_bitmap::get_bb_range): New.
13580 (sbr_sparse_bitmap::bb_range_p): New.
13581 (block_range_cache::block_range_cache): initialize bitmap obstack.
13582 (block_range_cache::~block_range_cache): Destruct obstack.
13583 (block_range_cache::set_bb_range): Decide when to utilze the
13584 sparse on entry cache.
13585 * gimple-range-cache.h (block_range_cache): Add bitmap obstack.
13586 * params.opt (-param=evrp-sparse-threshold): New.
13588 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
13590 * bitmap.c (bitmap_set_aligned_chunk): New.
13591 (bitmap_get_aligned_chunk): New.
13592 (test_aligned_chunk): New.
13593 (bitmap_c_tests): Call test_aligned_chunk.
13594 * bitmap.h (bitmap_set_aligned_chunk, bitmap_get_aligned_chunk): New.
13596 2021-06-07 Uroš Bizjak <ubizjak@gmail.com>
13599 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
13601 (ix86_expand_vector_init_one_nonzero): Ditto.
13602 (ix86_expand_vector_init_one_var): Ditto.
13603 (ix86_expand_vector_init_general): Ditto.
13604 * config/i386/mmx.md (vec_initv4qiqi): New expander.
13606 2021-06-07 Jeff Law <jeffreyalaw@gmail.com>
13608 * config/h8300/movepush.md: Change most _clobber_flags
13609 patterns to instead use <cczn> subst.
13610 (movsi_cczn): New pattern with usable CC cases split out.
13611 (movsi_h8sx_cczn): Likewise.
13613 2021-06-07 Martin Liska <mliska@suse.cz>
13615 * common/common-target.def: Split long lines and replace them
13617 * target.def: Likewise.
13618 * doc/tm.texi: Re-generated.
13620 2021-06-07 Jakub Jelinek <jakub@redhat.com>
13623 * fold-const.c (fold_read_from_vector): Return NULL if trying to
13624 read from a CONSTRUCTOR with vector type elements.
13626 2021-06-07 Jakub Jelinek <jakub@redhat.com>
13628 PR middle-end/100898
13629 * tree-inline.c (copy_bb): Only use gimple_call_arg_ptr if memcpy
13630 should copy any arguments. Don't call gimple_call_num_args
13631 on id->call_stmt or call_stmt more than once.
13633 2021-06-07 liuhongt <hongtao.liu@intel.com>
13636 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): Refine
13638 (<insn>v4siv4di2): Delete constraints for define_expand.
13640 2021-06-07 liuhongt <hongtao.liu@intel.com>
13643 * config/i386/i386-expand.c (ix86_expand_builtin): Remove
13644 assignment of cfun->machine->has_explicit_vzeroupper.
13645 * config/i386/i386-features.c
13646 (ix86_add_reg_usage_to_vzerouppers): Delete.
13647 (ix86_add_reg_usage_to_vzeroupper): Ditto.
13648 (rest_of_handle_insert_vzeroupper): Remove
13649 ix86_add_reg_usage_to_vzerouppers, add df_analyze at the end
13651 (gate): Remove cfun->machine->has_explicit_vzeroupper.
13652 * config/i386/i386-protos.h (ix86_expand_avx_vzeroupper):
13654 * config/i386/i386.c (ix86_insn_callee_abi): New function.
13655 (ix86_initialize_callee_abi): Ditto.
13656 (ix86_expand_avx_vzeroupper): Ditto.
13657 (ix86_hard_regno_call_part_clobbered): Adjust for vzeroupper
13659 (TARGET_INSN_CALLEE_ABI): Define as ix86_insn_callee_abi.
13660 (ix86_emit_mode_set): Call ix86_expand_avx_vzeroupper
13662 * config/i386/i386.h (struct GTY(()) machine_function): Delete
13663 has_explicit_vzeroupper.
13664 * config/i386/i386.md (enum unspec): New member
13666 (ABI_DEFAULT,ABI_VZEROUPPER,ABI_UNKNOWN): New
13667 define_constants for insn callee abi index.
13668 * config/i386/predicates.md (vzeroupper_pattern): Adjust.
13669 * config/i386/sse.md (UNSPECV_VZEROUPPER): Deleted.
13670 (avx_vzeroupper): Call ix86_expand_avx_vzeroupper.
13671 (*avx_vzeroupper): Rename to ..
13672 (avx_vzeroupper_callee_abi): .. this, and adjust pattern as
13673 call_insn which has a special vzeroupper ABI.
13674 (*avx_vzeroupper_1): Deleted.
13676 2021-06-07 liuhongt <hongtao.liu@intel.com>
13679 * df-scan.c (df_get_call_refs): When call_insn is a fake call,
13680 it won't use stack pointer reg.
13681 * final.c (leaf_function_p): When call_insn is a fake call, it
13682 won't affect caller as a leaf function.
13683 * reg-stack.c (callee_clobbers_any_stack_reg): New.
13684 (subst_stack_regs): When call_insn doesn't clobber any stack
13685 reg, don't clear the arguments.
13686 * rtl.c (shallow_copy_rtx): Don't clear flag used when orig is
13688 * shrink-wrap.c (requires_stack_frame_p): No need for stack
13689 frame for a fake call.
13690 * rtl.h (FAKE_CALL_P): New macro.
13692 2021-06-06 Eric Botcazou <ebotcazou@adacore.com>
13694 * config/sparc/sparc-protos.h (order_regs_for_local_alloc): Rename
13696 (sparc_order_regs_for_local_alloc): ...this.
13697 (sparc_leaf_reg_remap): Declare.
13698 * config/sparc/sparc.h (ADJUST_REG_ALLOC_ORDER): Adjust.
13699 (LEAF_REG_REMAP): Reimplement as call to sparc_leaf_reg_remap.
13700 * config/sparc/sparc.c (leaf_reg_remap): Delete.
13701 (order_regs_for_local_alloc): Rename to...
13702 (sparc_order_regs_for_local_alloc): ...this.
13703 (sparc_leaf_reg_remap): New function.
13704 (sparc_conditional_register_usage): Do not modify leaf_reg_remap.
13706 2021-06-06 David Edelsohn <dje.gcc@gmail.com>
13708 * config/rs6000/rs6000.c (rs6000_xcoff_asm_output_aligned_decl_common):
13709 Use assemble_name to output BSS section name.
13711 2021-06-06 Uroš Bizjak <ubizjak@gmail.com>
13713 * config/i386/constraints.md (Bs):
13714 Remove boolean operators from match_test RTX.
13717 (M): Use "mode" variable instead of GET_MODE (op) in match_test RTX.
13720 2021-06-06 Martin Liska <mliska@suse.cz>
13722 * doc/extend.texi: Add missing @headitem.
13723 * doc/invoke.texi: Likewise.
13724 * doc/objc.texi: Likewise.
13726 2021-06-06 Martin Liska <mliska@suse.cz>
13728 * genhooks.c (emit_findices): Remove unused function.
13729 (emit_documentation): Do not call emit_findices
13730 and do not search for @Fcode directives.
13732 2021-06-06 Martin Liska <mliska@suse.cz>
13734 * doc/invoke.texi: Remove extra character.
13736 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13738 * config/sh/sh.md (doloop_end_split): Fix empty split condition.
13740 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13742 * config/sparc/sparc.md (*snedi<W:mode>_zero_vis3,
13743 *neg_snedi<W:mode>_zero_subxc, *plus_snedi<W:mode>_zero,
13744 *plus_plus_snedi<W:mode>_zero, *minus_snedi<W:mode>_zero,
13745 *minus_minus_snedi<W:mode>_zero): Fix empty split condition.
13747 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13749 * config/or1k/or1k.md (*movdi): Fix empty split condition.
13751 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13753 * config/mips/mips.md (<anonymous>, bswapsi2, bswapdi2): Fix empty
13756 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
13758 * config/m68k/m68k.md (*zero_extend_inc, *zero_extend_dec,
13759 *zero_extendsidi2): Fix empty split condition.
13761 2021-06-05 Jeff Law <jeffreyalaw@gmail.com>
13763 * config/h8300/addsub.md: Fix split condition in define_insn_and_split
13765 * config/h8300/bitfield.md: Likewise.
13766 * config/h8300/combiner.md: Likewise.
13767 * config/h8300/divmod.md: Likewise.
13768 * config/h8300/extensions.md: Likewise.
13769 * config/h8300/jumpcall.md: Likewise.
13770 * config/h8300/movepush.md: Likewise.
13771 * config/h8300/multiply.md: Likewise.
13772 * config/h8300/other.md: Likewise.
13773 * config/h8300/shiftrotate.md: Likewise.
13774 * config/h8300/logical.md: Likewise. Fix split pattern to use
13775 code iterator that somehow slipped through.
13777 2021-06-04 Tobias Burnus <tobias@codesourcery.com>
13779 PR middle-end/100905
13780 * tree-nested.c (convert_nonlocal_omp_clauses,
13781 convert_local_omp_clauses): Handle OMP_CLAUSE_BIND.
13783 2021-06-04 Martin Sebor <msebor@redhat.com>
13785 PR middle-end/100732
13786 * gimple-fold.c (gimple_fold_builtin_sprintf): Avoid folding calls
13787 with either source or destination argument of invalid type.
13788 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Avoid checking
13789 calls with arguments of invalid type.
13791 2021-06-04 Martin Sebor <msebor@redhat.com>
13793 * attribs.c (init_attr_rdwr_indices): Use VLA bounds in the expected
13795 (attr_access::vla_bounds): Also handle VLA bounds.
13797 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
13799 * config/i386/predicates.md (GOT_memory_operand):
13800 Implement using match_code RTXes.
13801 (GOT32_symbol_operand): Ditto.
13803 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
13806 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
13808 (ix86_expand_vector_init_general): Ditto.
13809 Use SImode instead of word_mode for logic operations
13810 when GET_MODE_SIZE (mode) < UNITS_PER_WORD.
13811 (expand_vec_perm_even_odd_1): Assert that V2HI mode should be
13812 implemented by expand_vec_perm_1.
13813 (expand_vec_perm_broadcast_1): Assert that V2HI and V4HI modes
13814 should be implemented using standard shuffle patterns.
13815 (ix86_vectorize_vec_perm_const): Handle V2HImode. Add V4HI and
13816 V2HI modes to modes, implementable with shuffle for one operand.
13817 * config/i386/mmx.md (*punpckwd): New insn_and_split pattern.
13818 (*pshufw_1): New insn pattern.
13819 (*vec_dupv2hi): Ditto.
13820 (vec_initv2hihi): New expander.
13822 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
13824 * config/arm/vfp.md (no_literal_pool_df_immediate,
13825 no_literal_pool_sf_immediate): Fix empty split condition.
13827 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
13829 * config/i386/i386.md (*load_tp_x32_zext, *add_tp_x32_zext,
13830 *tls_dynamic_gnu2_combine_32): Fix empty split condition.
13831 * config/i386/sse.md (*<sse2_avx2>_pmovmskb_lt,
13832 *<sse2_avx2>_pmovmskb_zext_lt, *sse2_pmovmskb_ext_lt,
13833 *<sse4_1_avx2>_pblendvb_lt): Likewise.
13835 2021-06-04 Jakub Jelinek <jakub@redhat.com>
13838 * config/i386/i386-expand.c (ix86_expand_vector_init): Handle
13839 concatenation from half-sized modes with TImode elements.
13841 2021-06-04 Claudiu Zissulescu <claziss@synopsys.com>
13843 * config/arc/arc.c (arc_override_options): Disable millicode
13844 thunks when RF16 is on.
13846 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
13848 * config/rs6000/rs6000.h (PROMOTE_MODE): Remove.
13850 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
13852 * config/rs6000/rs6000-call.c (rs6000_promote_function_mode):
13853 Replace PROMOTE_MODE marco with its content.
13855 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
13857 * config/cris/cris.md (*addi_reload): Fix empty split condition.
13859 2021-06-03 Jim Wilson <jimw@sifive.com>
13861 * config.gcc (riscv*-*-*): If --with-riscv-attribute not used,
13862 turn it on for all riscv targets.
13864 2021-06-03 Uroš Bizjak <ubizjak@gmail.com>
13867 * config/i386/i386-expand.c (ix86_expand_vector_set):
13868 Handle V2HI and V4QI modes.
13869 (ix86_expand_vector_extract): Ditto.
13870 * config/i386/mmx.md (*pinsrw): New insn pattern.
13873 (*pextrw_zext): Ditto.
13875 (*pextrb_zext): Ditto.
13876 (vec_setv2hi): New expander.
13877 (vec_extractv2hihi): Ditto.
13878 (vec_setv4qi): Ditto.
13879 (vec_extractv4qiqi): Ditto.
13880 (vec_setv8qi): Enable only for TARGET_SSE4_1.
13881 (vec_extractv8qiqi): Ditto.
13883 2021-06-03 Aaron Sawdey <acsawdey@linux.ibm.com>
13885 * config/rs6000/genfusion.pl (gen_logical_addsubf): Fix input
13886 order to subf instruction.
13887 * config/rs6000/fusion.md: Regenerate.
13889 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
13891 * calls.c (get_size_range): Use range_of_expr instead of
13892 determine_value_range.
13893 * tree-affine.c (expr_to_aff_combination): Same.
13894 * tree-data-ref.c (split_constant_offset): Same.
13895 * tree-vrp.c (determine_value_range_1): Remove.
13896 (determine_value_range): Remove.
13897 * tree-vrp.h (determine_value_range): Remove.
13899 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
13901 * function-tests.c (test_ranges): Call gimple_range_tests.
13902 * gimple-range-cache.cc (ranger_cache::range_of_expr): Pass stmt
13904 * gimple-range.cc (fur_source::get_operand): Do not call
13905 get_tree_range or gimple_range_global.
13907 (get_tree_range): Move to value-query.cc.
13908 Call get_arith_expr_range.
13909 (gimple_ranger::range_of_expr): Add argument to get_tree_range.
13910 Include gimple-range-tests.cc.
13911 * gimple-range.h (fold_range): Add argument.
13912 (get_tree_range): Remove.
13913 * selftest.h (gimple_range_tests): New.
13914 * value-query.cc (global_range_query::range_of_expr): Add
13916 (range_query::get_tree_range): Move from gimple-range.cc.
13917 * value-query.h (class range_query): Add get_tree_range and
13918 get_arith_expr_range. Make fur_source a friend.
13919 * vr-values.c (vr_values::range_of_expr): Pass stmt to
13921 * gimple-range-tests.cc: New file.
13923 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
13925 * gimple-range.cc (gimple_ranger::export_global_ranges): Call
13926 update_global_range.
13927 * value-query.cc (update_global_range): New.
13928 * value-query.h (update_global_range): New.
13930 2021-06-03 David Malcolm <dmalcolm@redhat.com>
13932 * diagnostic-show-locus.c (diagnostic_show_locus): Don't reject
13933 printing the same location twice if there are fix-it hints,
13934 multiple locations, or a label.
13936 2021-06-03 Andre Vieira <andre.simoesdiasvieira@arm.com>
13938 * tree-vect-loop.c (vect_transform_loop): Use main loop's various'
13939 thresholds to narrow the upper bound on epilogue iterations.
13941 2021-06-03 Christophe Lyon <christophe.lyon@linaro.org>
13943 * config/arm/mve.md (mve_vabsq_f<mode>): Use 'abs' instead of unspec.
13944 (mve_vabsq_s<mode>): Likewise.
13945 * config/arm/neon.md (abs<mode>2): Rename to neon_abs<mode>2.
13946 * config/arm/unspecs.md (VABSQ_F, VABSQ_S): Delete.
13947 * config/arm/vec-common.md (neg<mode>2): Rename to
13948 <absneg_str><mode>2.
13950 2021-06-03 Claudiu Zissulescu <claziss@synopsys.com>
13952 * common/config/arc/arc-common.c (arc_option_optimization_table):
13953 Remove malign-call.
13954 * config/arc/arc.c (arc_unalign_branch_p): Remove unused function.
13955 * config/arc/arc.h (TARGET_MIXED_CODE): Remove macro.
13956 (INDEX_REG_CLASS): Only refer to GENERAL_REGS.
13957 * config/arc/arc.md (abssi2_mixed): Remove pattern.
13958 * config/arc/arc.opt (munalign-prob-threshold): Mark it obsolete.
13959 (malign-call): Likewise.
13960 (mmixed-code): Likewise.
13961 * doc/invoke.texi (ARC): Update doc.
13963 2021-06-03 Martin Liska <mliska@suse.cz>
13965 * common.opt: Use proper Enum values.
13966 * opts.c (COVERAGE_SANITIZER_OPT): Remove.
13967 (parse_sanitizer_options): Handle only sanitizer_opts.
13968 (common_handle_option): Just assign value.
13970 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13973 * tree-inline.c (inline_forbidden_p): Remove test on return type.
13975 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13977 * dwarf2out.c (loc_list_from_tree_1) <FUNCTION_DECL>: Also generate
13978 DW_OP_GNU_variable_value referencing an existing DIE at file scope.
13979 (type_byte_size): Inline into...
13980 (add_byte_size_attribute): ...this and call add_scalar_info.
13982 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13984 * dwarf2out.c (mem_loc_descriptor) <UDIV>: Fix typo.
13985 (typed_binop_from_tree): New function.
13986 (loc_list_from_tree_1) <EXACT_DIV_EXPR>: For an unsigned type,
13987 turn a divide by a power of 2 into a shift.
13988 <CEIL_DIV_EXPR>: For an unsigned type, use a signed divide if the
13989 size of the mode is lower than DWARF2_ADDR_SIZE; otherwise, do a
13990 typed divide by calling typed_binop_from_tree.
13992 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
13994 * dwarf2out.c (scompare_loc_descriptor): Fix head comment.
13995 (is_handled_procedure_type): Likewise.
13996 (struct loc_descr_context): Add strict_signedness field.
13997 (resolve_args_picking_1): Deal with DW_OP_[GNU_]deref_type,
13998 DW_OP_[GNU_]convert and DW_OP_[GNU_]reinterpret.
13999 (resolve_args_picking): Minor tweak.
14000 (function_to_dwarf_procedure): Initialize strict_signedness field.
14001 (type_byte_size): Likewise.
14002 (field_byte_offset): Likewise.
14003 (gen_descr_array_type_die): Likewise.
14004 (gen_variant_part): Likewise.
14005 (loc_list_from_tree_1) <CALL_EXPR>: Tidy up and set strict_signedness
14006 to true when a context is present before evaluating the arguments.
14007 <COND_EXPR>: Do not generate a useless comparison with zero.
14008 When dereferencing an address, if strict_signedness is true and the
14009 type is small and signed, use DW_OP_deref_type to do the dereference
14010 and then DW_OP_convert to convert back to the generic type.
14012 2021-06-03 Jakub Jelinek <jakub@redhat.com>
14015 * tree-inline.c (copy_tree_body_r): Handle iterators on
14016 OMP_CLAUSE_AFFINITY or OMP_CLAUSE_DEPEND.
14018 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
14020 * config/arc/arc.md (*bbit_di): Remove.
14022 2021-06-02 Christoph Muellner <cmuellner@gcc.gnu.org>
14024 PR rtl-optimization/100264
14025 * ree.c (get_sub_rtx): Ignore SET expressions without register
14026 destinations and remove assertion, as it is not valid anymore
14027 with this new behaviour.
14028 (merge_def_and_ext): Eliminate destination check for register
14029 as such SET expressions can't occur anymore.
14030 (combine_reaching_defs): Likewise.
14032 2021-06-02 Jakub Jelinek <jakub@redhat.com>
14035 * config/xtensa/xtensa.h (LEAF_REG_REMAP): Cast REGNO to int to avoid
14036 -Wtype-limits warnings.
14037 (DWARF_FRAME_REGISTER): Rewrite into ternary operator with addition
14038 in operands to avoid -Wsign-compare warnings.
14040 2021-06-02 Pat Haugen <pthaugen@linux.ibm.com>
14042 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): Use
14045 2021-06-02 Vineet Gupta <vgupta@synopsys.com>
14047 * config/arc/arc.h (TARGET_CPU_DEFAULT): Change to hs38_linux.
14049 2021-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
14051 * config/s390/s390.md(*ashrdi3_31<setcc><cconly>): Use a single
14053 * config/s390/subst.md(cconly_subst): Use a single constraint
14054 in (match_scratch).
14056 2021-06-02 Martin Liska <mliska@suse.cz>
14058 * ipa-icf.h: Use auto_vec for memory_access_types.
14060 2021-06-02 Jeff Law <jeffreyalaw@gmail.com>
14062 * config/h8300/h8300-protos.h (compute_a_shift_length): Drop unused
14063 argument from prototype.
14064 (output_logical_op): Add rtx_code argument.
14065 (compute_logical_op_length): Likewise.
14066 * config/h8300/h8300.c (h8300_and_costs): Pass additional argument
14067 to compute_a_shift_length.
14068 (output_logical_op); New argument with the rtx code rather than
14069 extracting it from an operand. Handle QImode too.
14070 (compute_logical_op_length): Similary.
14071 (compute_a_shift_length): Drop unused argument.
14072 * config/h8300/h8300.md (logicals): New code iterator.
14073 * config/h8300/logical.md (<code><mode>3 expander): Combine
14074 the "and" expander with the "ior"/"xor" expander.
14075 (bclr<mode>msx): Combine the QI/HI mode patterns.
14076 (<logical><mode>3 insns): Use code iterator rather than match_operator.
14077 Handle QImode as well. Update call to output_logical_op and
14078 compute_logical_op_length to pass in rtx_code
14079 Fix split condition on all define_insn_and_split patterns.
14080 (one_cmpl<mode>2<cczn>): Use <cczn> to support both clobbering
14081 the flags and setting ZN via existing define_subst.
14082 * config/h8300/shiftrotate.md: Drop unused argument from
14083 calls to compute_a_shift_length.
14084 Signed-off-by: Jeff Law <jeffreyalaw@gmail.com>
14086 2021-06-01 Andrew Pinski <apinski@marvell.com>
14088 PR tree-optimization/25290
14089 * tree-ssa-phiopt.c (match_simplify_replacement):
14091 (tree_ssa_phiopt_worker): Use match_simplify_replacement.
14092 (two_value_replacement): Change the comment about
14093 conditional_replacement.
14094 (conditional_replacement): Delete.
14096 2021-06-01 Andrew Pinski <apinski@marvell.com>
14098 PR tree-optimization/95481
14099 * tree-tailcall.c (find_tail_calls): Handle empty typed
14102 2021-06-01 Andrew Pinski <apinski@marvell.com>
14104 * gimplify.c (zero_sized_field_decl): Delete
14105 (zero_sized_type): Delete
14106 (gimplify_init_ctor_eval): Use is_empty_type instead
14107 of zero_sized_field_decl.
14108 (gimplify_modify_expr): Use is_empty_type instead of
14111 2021-06-01 Jason Merrill <jason@redhat.com>
14114 * tree.h (CALL_FROM_NEW_OR_DELETE_P): Adjust comment.
14116 2021-06-01 Jason Merrill <jason@redhat.com>
14119 * diagnostic.h (warning_enabled_at): Declare.
14120 * diagnostic.c (diagnostic_enabled): Factor out from...
14121 (diagnostic_report_diagnostic): ...here.
14122 (warning_enabled_at): New.
14124 2021-06-01 Aldy Hernandez <aldyh@redhat.com>
14126 * gimple-ssa-evrp.c: Enable exporting of global ranges.
14128 2021-06-01 Martin Liska <mliska@suse.cz>
14131 * doc/invoke.texi: Mention that -fgcse-after-reload
14132 is enabled with -O3.
14134 2021-06-01 liuhongt <hongtao.liu@intel.com>
14136 PR tree-optimization/98365
14137 * tree-if-conv.c (strip_nop_cond_scalar_reduction): New function.
14138 (is_cond_scalar_reduction): Handle nop_expr in cond scalar reduction.
14139 (convert_scalar_cond_reduction): Ditto.
14140 (predicate_scalar_phi): Ditto.
14142 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
14144 PR tree-optimization/100781
14145 * gimple-range-cache.cc (ranger_cache::ranger_cache): Enable new
14146 value calculation by default.
14147 (ranger_cache::enable_new_values): New.
14148 (ranger_cache::disable_new_values): New.
14149 (ranger_cache::push_poor_value): Check if new values are allowed.
14150 * gimple-range-cache.h (class ranger_cache): New member/methods.
14151 * gimple-range.cc (gimple_ranger::range_of_expr): Check for debug
14152 statement, and disable/renable new value calculation.
14154 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
14156 * gimple-range-cache.cc (ranger_cache::ssa_range_in_bb): Delete.
14157 (ranger_cache::range_of_def): New.
14158 (ranger_cache::entry_range): New.
14159 (ranger_cache::exit_range): New.
14160 (ranger_cache::range_of_expr): Adjust.
14161 (ranger_cache::range_on_edge): Adjust.
14162 (ranger_cache::propagate_cache): Call exit_range directly.
14163 * gimple-range-cache.h (class ranger_cache): Adjust.
14165 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
14167 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust for
14168 gori_compute being a member rather than base class.
14169 dervied call to member call.
14170 (ranger_cache::dump): No longer dump gori_map.
14171 (ranger_cache::dump_bb): New.
14172 (ranger_cache::get_non_stale_global_range): Adjust for gori_compute
14173 being a member rather than base class.
14174 (ranger_cache::set_global_range): Ditto.
14175 (ranger_cache::ssa_range_in_bb): Ditto.
14176 (ranger_cache::range_of_expr): New.
14177 (ranger_cache::range_on_edge): New.
14178 (ranger_cache::block_range): Adjust for gori_computes. Debug changes.
14179 (ranger_cache::propagate_cache): Adjust debugging output.
14180 (ranger_cache::fill_block_cache): Adjust for gori_computes. Debug
14182 * gimple-range-cache.h (class ranger_cache): Make gori_compute a
14183 member, and inherit from range_query instead.
14184 (ranger_cache::dump_bb): New. split from dump.
14185 * gimple-range-gori.cc (gori_compute::ssa_range_in_bb): Delete.
14186 (gori_compute::expr_range_at_stmt): Delete.
14187 (gori_compute::compute_name_range_op): Delete.
14188 (gori_compute::compute_operand_range_switch): Add fur_source.
14189 (gori_compute::compute_operand_range): Add fur_source param, inline
14190 old compute_name_range_op and optimize_logical_operands.
14191 (struct tf_range): Delete.
14192 (gori_compute::logical_combine): Adjust
14193 (gori_compute::optimize_logical_operands): Delete.
14194 (gori_compute::compute_logical_operands_in_chain): Delete.
14195 (gori_compute::compute_logical_operands): Adjust.
14196 (gori_compute::compute_operand1_range): Adjust to fur_source.
14197 (gori_compute::compute_operand2_range): Ditto.
14198 (gori_compute::compute_operand1_and_operand2_range): Ditto.
14199 (gori_compute::outgoing_edge_range_p): Add range_query parameter,
14200 and adjust to fur_source.
14201 * gimple-range-gori.h (class gori_compute): Simplify and adjust to
14202 range_query and fur_source.
14203 * gimple-range.cc (gimple_ranger::range_on_edge): Query range_on_edge
14204 from the ranger_cache..
14205 (gimple_ranger::fold_range_internal): Adjust to base class change of
14207 (gimple_ranger::dump_bb): Adjust dump.
14208 * gimple-range.h (gimple_ranger):export gori computes object.
14210 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
14212 PR tree-optimization/100774
14213 * gimple-range-cache.cc (ranger_cache::get_non_stale_global_range):
14214 Constant values are also not stale.
14215 (ranger_cache::set_global_range): Range invariant values should also
14216 have the correct timestamp.
14218 2021-05-31 Martin Liska <mliska@suse.cz>
14220 * tree-streamer-in.c (unpack_ts_function_decl_value_fields):
14221 Unpack FUNCTION_DECL_DECL_TYPE.
14222 * tree-streamer-out.c (pack_ts_function_decl_value_fields):
14223 Stream FUNCTION_DECL_DECL_TYPE instead of
14224 DECL_IS_OPERATOR_NEW_P.
14225 * tree.h (set_function_decl_type): Use FUNCTION_DECL_DECL_TYPE
14227 (DECL_IS_OPERATOR_NEW_P): Likewise.
14228 (DECL_IS_OPERATOR_DELETE_P): Likewise.
14229 (DECL_LAMBDA_FUNCTION_P): Likewise.
14231 2021-05-31 Richard Biener <rguenther@suse.de>
14234 * internal-fn.c (expand_SHUFFLEVECTOR): Define.
14235 * internal-fn.def (SHUFFLEVECTOR): New.
14236 * internal-fn.h (expand_SHUFFLEVECTOR): Declare.
14237 * doc/extend.texi: Document __builtin_shufflevector.
14239 2021-05-31 Peter Bergner <bergner@linux.ibm.com>
14242 * config/rs6000/predicates.md(mma_assemble_input_operand): Allow
14243 indexed form addresses.
14245 2021-05-29 Jeff Law <jlaw@tachyum.com>
14247 * config/h8300/h8300.c (h8300_emit_stack_adjustment): Drop unused
14248 parameter. Call callers fixed.
14250 (output_plussi): Add FALLTHRU markers.
14251 (h8300_shift_needs_scratch_p): Add gcc_unreachable marker.
14253 2021-05-29 Jakub Jelinek <jakub@redhat.com>
14255 PR middle-end/99928
14256 * gimplify.c (gimplify_scan_omp_clauses): For taskloop simd
14257 combined with parallel, make sure to add shared clause to
14258 parallel for explicit linear clause.
14260 2021-05-29 Aldy Hernandez <aldyh@redhat.com>
14262 PR tree-optimization/100787
14263 * gimple-ssa-evrp.c: Disable exporting of global ranges.
14265 2021-05-28 Jason Merrill <jason@redhat.com>
14267 * tree-iterator.h (struct tree_stmt_iterator): Add operator++,
14268 operator--, operator*, operator==, and operator!=.
14269 (class tsi_range): New.
14271 2021-05-28 Richard Biener <rguenther@suse.de>
14273 PR tree-optimization/100778
14274 * tree-vect-slp.c (vect_build_slp_tree_1): Prevent possibly
14275 trapping ops in different BBs.
14277 2021-05-28 Richard Biener <rguenther@suse.de>
14280 * tree-inline.c (copy_bb): When processing __builtin_va_arg_pack
14281 copy fntype from original call.
14283 2021-05-28 Martin Liska <mliska@suse.cz>
14285 PR gcov-profile/100751
14286 * doc/gcov.texi: Revert partially a hunk that was wrong.
14288 2021-05-28 Cooper Qu <cooper.qu@linux.alibaba.com>
14290 * config/csky/csky-linux-elf.h (HAVE_sync_compare_and_swapqi):
14292 (HAVE_sync_compare_and_swaphi): Likewise.
14293 (HAVE_sync_compare_and_swapsi): Likewise.
14295 2021-05-28 Jakub Jelinek <jakub@redhat.com>
14297 PR middle-end/99928
14298 * tree.h (OMP_CLAUSE_MAP_IMPLICIT): Define.
14300 2021-05-28 Tobias Burnus <tobias@codesourcery.com>
14302 * gimplify.c (gimplify_omp_affinity): New.
14303 (gimplify_scan_omp_clauses): Call it; remove affinity clause afterwards.
14304 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_AFFINITY.
14305 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_AFFINITY.
14306 * tree.c (omp_clause_num_ops, omp_clause_code_name): Add clause.
14307 (walk_tree_1): Handle OMP_CLAUSE_AFFINITY.
14309 2021-05-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
14310 Richard Biener <rguenther@suse.de>
14312 * match.pd <popcount & / + pattern matching>:
14313 When generating popcount directly fails, try doing it in two halves.
14315 2021-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
14317 * Makefile.in (generated_files): Add gimple-match.c and
14320 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
14322 * gensupport.c (alter_predicate_for_insn): Handle MATCH_DUP.
14324 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
14326 * gensupport.c (alter_constraints): Add MATCH_SCRATCH case.
14328 2021-05-28 Kewen Lin <linkw@linux.ibm.com>
14330 PR tree-optimization/99398
14331 * tree-ssa-forwprop.c (simplify_permutation): Optimize some cases
14332 where the fed operands are CTOR/CST and propagated through
14333 VIEW_CONVERT_EXPR. Call vec_perm_indices::new_shrunk_vector.
14334 * vec-perm-indices.c (vec_perm_indices::new_shrunk_vector): New
14336 * vec-perm-indices.h (vec_perm_indices::new_shrunk_vector): New
14339 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
14341 * config/i386/mmx.md (addv2sf3): Do not call
14342 ix86_fixup_binary_operands_no_copy.
14345 (<smaxmin:code>v2sf3): Ditto.
14346 (<plusminus:insn><MMXMODEI:mode>3): Ditto.
14347 (<plusminus:insn><VI_32:mode>3): Remove expander.
14348 (<plusminus:insn><VI_32:mode>3): Rename from
14349 "*<plusminus:insn><VI_32:mode>3".
14350 (mulv4hi): Do not call ix86_fixup_binary_operands_no_copy.
14351 (mulv2hi3): Remove expander.
14352 (mulv2hi3): Rename from *mulv2hi3.
14353 (<s>mulv2hi3_highpart): Remove expander.
14354 (<s>mulv2hi3_highpart): Rename from *<s>mulv2hi3_highpart.
14355 (<smaxmin:code><MMXMODE14:mode>3): Rename from
14356 "*mmx_<smaxmin:code><MMXMODE14:mode>3".
14357 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Remove expander.
14358 (SMAXMIN_MMXMODEI): Remove mode iterator.
14359 (<smaxmin:code>v4hi3): New expander.
14360 (<smaxmin:code>v4qi3): Rename from *<smaxmin:code>v4qi3.
14361 (<smaxmin:code>v2hi3): Rename from *<smaxmin:code>v2hi3.
14362 (<smaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
14363 (SMAXMIN_VI_32): Remove mode iterator.
14364 (<umaxmin:code><MMXMODE24:mode>3): Rename from
14365 "*mmx_<umaxmin:code><MMXMODE24:mode>3".
14366 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Remove expander.
14367 (UMAXMIN_MMXMODEI): Remove mode iterator.
14368 (<umaxmin:code>v8qi3): New expander.
14369 (<umaxmin:code>v4qi3): Rename from *<umaxmin:code>v4qi3.
14370 (<umaxmin:code>v2hi3): Rename from *<umaxmin:code>v2hi3.
14371 (<umaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
14372 (UMAXMIN_VI_32): Remove mode iterator.
14373 (<any_shift:insn>v2hi3): Remove expander.
14374 (<any_shift:insn>v2hi3): Rename from *<any_shift:insn>v2hi3.
14375 (<any_logic:code><MMXMODEI:mode>3): Do not call
14376 ix86_fixup_binary_operands_no_copy.
14377 (<any_logic:code><VI_32:mode>3): Remove expander.
14378 (<any_logic:code><VI_32:mode>3): Rename from
14379 "*<any_logic:code><VI_32:mode>3".
14380 (uavg<mode>3_ceil): Do not call ix86_fixup_binary_operands_no_copy.
14381 * config/i386/sse.md (div<VF2:mode>3): Do not call
14382 ix86_fixup_binary_operands_no_copy.
14383 (div<VF1:mode>3): Ditto.
14384 (<maxmin:code><VI8_AVX2_AVX512F:mode>3): Ditto.
14385 (smulhrsv4hi3): Ditto.
14386 (smulhrsv2hi3): Ditto.
14388 2021-05-27 Martin Sebor <msebor@redhat.com>
14390 * ggc.h (gt_ggc_mx): Add overloads for all integers.
14392 * hash-map.h (class hash_map): Add pch_nx_helper overloads for all
14394 (hash_map::operator==): New function.
14396 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
14399 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
14400 For TARGET_XOP bypass SSE comparisons for all supported vector modes.
14401 * config/i386/mmx.md (*xop_maskcmp<MMXMODEI:mode>3): New insn pattern.
14402 (*xop_maskcmp<VI_32:mode>3): Ditto.
14403 (*xop_maskcmp_uns<MMXMODEI:mode>3): Ditto.
14404 (*xop_maskcmp_uns<VI_32:mode>3): Ditto.
14406 2021-05-27 Richard Earnshaw <rearnsha@arm.com>
14409 * config/arm/arm.c (arm_configure_build_target): Remove parameter
14410 opts_set, directly check opts parameters for being non-null.
14411 (arm_option_restore): Update call to arm_configure_build_target.
14412 (arm_option_override): Likewise.
14413 (arm_can_inline_p): Likewise.
14414 (arm_valid_target_attribute_tree): Likewise.
14415 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
14416 * config/arm/arm-protos.h (arm_configure_build_target): Adjust
14419 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
14421 * vr-values.c (simplify_conversion_using_ranges): Use
14422 get_range_query instead of get_global_range_query.
14424 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
14426 * gimple-range.cc (get_range_global): Move to value-query.cc.
14427 (gimple_range_global): Same.
14428 (get_global_range_query): Same.
14429 (global_range_query::range_of_expr): Same.
14430 * gimple-range.h (class global_range_query): Move to
14432 (gimple_range_global): Same.
14433 * tree-ssanames.c (get_range_info): Move to value-query.cc.
14434 (get_ptr_nonnull): Same.
14435 * tree-ssanames.h (get_range_info): Remove.
14436 (get_ptr_nonnull): Remove.
14437 * value-query.cc (get_ssa_name_range_info): Move from
14439 (get_ssa_name_ptr_info_nonnull): Same.
14440 (get_range_global): Move from gimple-range.cc.
14441 (gimple_range_global): Same.
14442 (get_global_range_query): Same.
14443 (global_range_query::range_of_expr): Same.
14444 * value-query.h (class global_range_query): Move from
14446 (gimple_range_global): Same.
14448 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
14451 * config/i386/mmx.md (uavgv4qi3_ceil): New insn pattern.
14452 (uavgv2hi3_ceil): Ditto.
14454 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
14457 * doc/extend.texi (scalar_storage_order): Rephrase slightly.
14459 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14461 * tree-ssanames.c (get_range_info): Merge both copies of
14462 get_range_info into one that works with irange.
14463 * tree-ssanames.h (get_range_info): Remove version that works on
14466 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14468 * builtins.c (check_nul_terminated_array): Convert to get_range_query.
14469 (expand_builtin_strnlen): Same.
14470 (determine_block_size): Same.
14471 * fold-const.c (expr_not_equal_to): Same.
14472 * gimple-fold.c (size_must_be_zero_p): Same.
14473 * gimple-match-head.c: Include gimple-range.h.
14474 * gimple-pretty-print.c (dump_ssaname_info): Convert to get_range_query.
14475 * gimple-ssa-warn-restrict.c
14476 (builtin_memref::extend_offset_range): Same.
14477 * graphite-sese-to-poly.c (add_param_constraints): Same.
14478 * internal-fn.c (get_min_precision): Same.
14479 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): Same.
14480 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Same.
14482 * tree-data-ref.c (split_constant_offset): Same.
14483 (dr_step_indicator): Same.
14484 * tree-dfa.c (get_ref_base_and_extent): Same.
14485 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
14486 * tree-ssa-loop-niter.c (refine_value_range_using_guard): Same.
14487 (determine_value_range): Same.
14488 (record_nonwrapping_iv): Same.
14489 (infer_loop_bounds_from_signedness): Same.
14490 (scev_var_range_cant_overflow): Same.
14491 * tree-ssa-phiopt.c (two_value_replacement): Same.
14492 * tree-ssa-pre.c (insert_into_preds_of_block): Same.
14493 * tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Same.
14494 * tree-ssa-strlen.c (handle_builtin_stxncpy_strncat): Same.
14496 (dump_strlen_info): Same.
14497 (set_strlen_range): Same.
14498 (maybe_diag_stxncpy_trunc): Same.
14499 (get_len_or_size): Same.
14500 (handle_integral_assign): Same.
14501 * tree-ssa-structalias.c (find_what_p_points_to): Same.
14502 * tree-ssa-uninit.c (find_var_cmp_const): Same.
14503 * tree-switch-conversion.c (bit_test_cluster::emit): Same.
14504 * tree-vect-patterns.c (vect_get_range_info): Same.
14505 (vect_recog_divmod_pattern): Same.
14506 * tree-vrp.c (intersect_range_with_nonzero_bits): Same.
14507 (register_edge_assert_for_2): Same.
14508 (determine_value_range_1): Same.
14509 * tree.c (get_range_pos_neg): Same.
14510 * vr-values.c (vr_values::get_lattice_entry): Same.
14511 (vr_values::update_value_range): Same.
14512 (simplify_conversion_using_ranges): Same.
14514 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14516 * gimple-ssa-warn-alloca.c (alloca_call_type): Use
14517 get_range_query instead of query argument.
14518 (pass_walloca::execute): Enable and disable global ranger.
14520 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14522 * gimple-ssa-evrp.c (rvrp_folder::rvrp_folder): Call
14524 (rvrp_folder::~rvrp_folder): Call disable_ranger.
14525 (hybrid_folder::hybrid_folder): Call enable_ranger.
14526 (hybrid_folder::~hybrid_folder): Call disable_ranger.
14528 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
14530 * function.c (allocate_struct_function): Set cfun->x_range_query.
14531 * function.h (struct function): Declare x_range_query.
14532 (get_range_query): New.
14533 (get_global_range_query): New.
14534 * gimple-range-cache.cc (ssa_global_cache::ssa_global_cache):
14535 Remove call to safe_grow_cleared.
14536 * gimple-range.cc (get_range_global): New.
14537 (gimple_range_global): Move from gimple-range.h.
14538 (get_global_range_query): New.
14539 (global_range_query::range_of_expr): New.
14540 (enable_ranger): New.
14541 (disable_ranger): New.
14542 * gimple-range.h (gimple_range_global): Move to gimple-range.cc.
14543 (class global_range_query): New.
14544 (enable_ranger): New.
14545 (disable_ranger): New.
14546 * gimple-ssa-evrp.c (evrp_folder::~evrp_folder): Rename
14547 dump_all_value_ranges to dump.
14548 * tree-vrp.c (vrp_prop::finalize): Same.
14549 * value-query.cc (range_query::dump): New.
14550 * value-query.h (range_query::dump): New.
14551 * vr-values.c (vr_values::dump_all_value_ranges): Rename to...
14552 (vr_values::dump): ...this.
14553 * vr-values.h (class vr_values): Rename dump_all_value_ranges to
14554 dump and make virtual.
14556 2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
14558 * config/i386/i386.c (ix86_autovectorize_vector_modes):
14559 Add V4QImode and V16QImode for TARGET_SSE2.
14560 * doc/sourcebuild.texi (Vector-specific attributes):
14561 Add vect64 and vect32 description.
14563 2021-05-26 Bernd Edlinger <bernd.edlinger@hotmail.de>
14565 * gimple-range-gori.cc (range_def_chain::register_dependency):
14566 Resize m_def_chain when needed.
14568 2021-05-26 Christophe Lyon <christophe.lyon@linaro.org>
14570 * config/arm/mve.md (mve_vaddvq_<supf><mode>): Prefix with '@'.
14571 * config/arm/neon.md (reduc_plus_scal_<mode>): Move to ..
14572 * config/arm/vec-common.md: .. here. Add support for MVE.
14574 2021-05-26 Jakub Jelinek <jakub@redhat.com>
14576 * config/epiphany/epiphany.c (epiphany_print_operand_address): Remove
14578 * config/microblaze/microblaze.c (microblaze_legitimize_address,
14580 microblaze_option_override, print_operand): Likewise.
14581 * config/microblaze/microblaze.md (call_internal_plt,
14582 call_value_intern_plt, call_value_intern): Likewise.
14583 * config/arm/aout.h (ASM_OUTPUT_ALIGN): Likewise.
14584 * config/iq2000/iq2000.md (call_internal1, call_value_internal1,
14585 call_value_multiple_internal1): Likewise.
14586 * config/bfin/bfin.c (symbolic_reference_mentioned_p): Likewise.
14588 2021-05-26 Jan-Benedict Glaw <jbglaw@lug-owl.de>
14590 * config/arc/arc.c (arc_address_cost, arc_print_operand_address,
14591 arc_ccfsm_advance, symbolic_reference_mentioned_p,
14592 arc_raw_symbolic_reference_mentioned_p): Remove register
14595 2021-05-26 Jakub Jelinek <jakub@redhat.com>
14598 * omp-low.c: Include omp-offload.h.
14599 (create_omp_child_function): If current_function_decl has
14600 "omp declare target" attribute and is_gimple_omp_offloaded,
14601 remove that attribute from the copy of attribute list and
14602 add "omp target entrypoint" attribute instead.
14603 (lower_omp_target): Mark .omp_data_sizes.* and .omp_data_kinds.*
14604 variables for offloading if in omp_maybe_offloaded_ctx.
14605 * omp-offload.c (pass_omp_target_link::execute): Nullify second
14606 argument to GOMP_target_data_ext in offloaded code.
14608 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14610 * config/csky/csky.c (csky_can_change_mode_class): Delete.
14611 For csky, HF/SF mode use the low bits of VREGS.
14613 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
14615 * gimplify.c (gimplify_decl_expr): Do not clear TREE_READONLY on a
14616 DECL which is a reference for OMP.
14618 2021-05-26 Martin Liska <mliska@suse.cz>
14620 PR gcov-profile/100751
14621 * doc/gcov.texi: Document that __gcov_dump can be called just
14622 once and that __gcov_reset resets run-time counters.
14624 2021-05-26 Martin Liska <mliska@suse.cz>
14626 * doc/install.texi: Port relevant part from install-old.texi
14627 and re-generate list of CPUs and systems.
14629 2021-05-26 Martin Liska <mliska@suse.cz>
14631 * Makefile.in: Remove it.
14632 * doc/include/fdl.texi: Update next/previous chapters.
14633 * doc/install.texi: Likewise.
14634 * doc/install-old.texi: Removed.
14636 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14638 * config/csky/csky.c (ck810_legitimate_index_p): Support
14639 "base + index" with DF mode.
14640 * config/csky/constraints.md ("Y"): New constraint for memory operands
14641 without index register.
14642 * config/csky/csky_insn_fpuv2.md (fpuv3_movdf): Use "Y" instead of "m"
14643 when mov between memory and general registers, and lower their priority.
14644 * config/csky/csky_insn_fpuv3.md (fpuv2_movdf): Likewise.
14646 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14648 * config/csky/csky.c (TARGET_PROMOTE_PROTOTYPES): Delete.
14650 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14652 * config/csky/csky.md (untyped_call): Emit clobber for return
14653 registers to mark them used.
14655 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
14657 * config/csky/csky.md (cskyv2_sextend_ldbs): New.
14659 2021-05-26 Andrew Pinski <apinski@marvell.com>
14661 * match.pd (x < 0 ? ~y : y): New patterns.
14663 2021-05-26 Andrew Pinski <apinski@marvell.com>
14665 * match.pd (A?CST1:CST2): Add simplifcations for A?0:+-1, A?+-1:0,
14666 A?POW2:0 and A?0:POW2.
14668 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14670 * gimple-range-gori.cc (class logical_stmt_cache): Delete
14671 (logical_stmt_cache::logical_stmt_cache ): Delete.
14672 (logical_stmt_cache::~logical_stmt_cache): Delete.
14673 (logical_stmt_cache::cache_entry::dump): Delete.
14674 (logical_stmt_cache::get_range): Delete.
14675 (logical_stmt_cache::cached_name ): Delete.
14676 (logical_stmt_cache::same_cached_name): Delete.
14677 (logical_stmt_cache::cacheable_p): Delete.
14678 (logical_stmt_cache::slot_diagnostics ): Delete.
14679 (logical_stmt_cache::dump): Delete.
14680 (gori_compute_cache::gori_compute_cache): Delete.
14681 (gori_compute_cache::~gori_compute_cache): Delete.
14682 (gori_compute_cache::compute_operand_range): Delete.
14683 (gori_compute_cache::cache_stmt): Delete.
14684 * gimple-range-gori.h (gori_compute::compute_operand_range): Remove
14686 (class gori_compute_cache): Delete.
14688 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14690 * gimple-range.cc (fold_using_range::range_of_range_op): Use m_gori
14692 (fold_using_range::range_of_address): Adjust.
14693 (fold_using_range::range_of_phi): Adjust.
14694 * gimple-range.h (class fur_source): Adjust.
14695 (fur_source::fur_source): Adjust.
14697 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14699 * gimple-range-gori.cc (gori_compute::expr_range_at_stmt): Rename
14700 from expr_range_in_bb and adjust.
14701 (gori_compute::compute_name_range_op): Adjust.
14702 (gori_compute::optimize_logical_operands): Adjust.
14703 (gori_compute::compute_logical_operands_in_chain): Adjust.
14704 (gori_compute::compute_operand1_range): Adjust.
14705 (gori_compute::compute_operand2_range): Adjust.
14706 (ori_compute_cache::cache_stmt): Adjust.
14707 * gimple-range-gori.h (gori_compute): Rename prototype.
14709 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14711 * gimple-range.cc (gimple_ranger::range_of_expr): Non-null should be
14712 checked only after range_of_stmt, not range_on_entry.
14713 (gimple_ranger::range_on_entry): Check for non-null in any
14714 predecessor block, if it is not already non-null.
14715 (gimple_ranger::range_on_exit): DOnt check for non-null after
14716 range on entry call.
14717 (gimple_ranger::dump_bb): New. Split from dump.
14718 (gimple_ranger::dump): Adjust.
14719 * gimple-range.h (class gimple_ranger): Adjust.
14721 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14723 * gimple-range-cache.cc (struct range_timestamp): Delete.
14724 (class temporal_cache): Adjust.
14725 (temporal_cache::get_timestamp): Delete.
14726 (temporal_cache::set_dependency): Delete.
14727 (temporal_cache::temporal_value): Adjust.
14728 (temporal_cache::current_p): Take dependencies as params.
14729 (temporal_cache::set_timestamp): Adjust.
14730 (temporal_cache::set_always_current): Adjust.
14731 (ranger_cache::get_non_stale_global_range): Adjust.
14732 (ranger_cache::register_dependency): Delete.
14733 * gimple-range-cache.h (class range_cache): Adjust.
14735 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14737 * gimple-range-gori.cc (range_def_chain::range_def_chain): init
14739 (range_def_chain::~range_def_chain): Dispose of obstack rather than
14740 each individual bitmap.
14741 (range_def_chain::set_import): New.
14742 (range_def_chain::get_imports): New.
14743 (range_def_chain::chain_import_p): New.
14744 (range_def_chain::register_dependency): Rename from build_def_chain
14746 (range_def_chain::def_chain_in_bitmap_p): New.
14747 (range_def_chain::add_def_chain_to_bitmap): New.
14748 (range_def_chain::has_def_chain): Just check first depenedence.
14749 (range_def_chain::get_def_chain): Process imports, use generic
14750 register_dependency routine.
14751 (range_def_chain::dump): New.
14752 (gori_map::gori_map): Allocate import list.
14753 (gori_map::~gori_map): Release imports.
14754 (gori_map::exports): Check for past allocated block size.
14755 (gori_map::imports): New.
14756 (gori_map::def_chain_in_export_p): Delete.
14757 (gori_map::is_import_p): New.
14758 (gori_map::maybe_add_gori): Handle imports.
14759 (gori_map::dump): Adjust output, add imports.
14760 (gori_compute::has_edge_range_p): Remove def_chain_in_export call.
14761 (gori_export_iterator::gori_export_iterator): New.
14762 (gori_export_iterator::next): New.
14763 (gori_export_iterator::get_name): New.
14764 * gimple-range-gori.h (range_def_chain): Add imports and direct
14765 dependecies via struct rdc.
14766 (range_def_chain::depend1): New.
14767 (range_def_chain::depend2): New.
14768 (class gori_map): Adjust.
14769 (FOR_EACH_GORI_IMPORT_NAME): New.
14770 (FOR_EACH_GORI_EXPORT_NAME): New.
14771 (class gori_export_iterator): New.
14773 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14775 * gimple-range-cache.cc (ranger_cache::ranger_cache): Move initial
14776 export cache filling to here.
14777 * gimple-range-gori.cc (gori_compute::gori_compute) : From Here.
14779 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
14781 * gimple-range-gori.cc (range_def_chain): Move to gimple-range-gori.h.
14782 (gori_map): Move to gimple-range-gori.h.
14783 (gori_compute::gori_compute): Adjust.
14784 (gori_compute::~gori_compute): Delete.
14785 (gori_compute::compute_operand_range_switch): Adjust.
14786 (gori_compute::compute_operand_range): Adjust.
14787 (gori_compute::compute_logical_operands): Adjust.
14788 (gori_compute::has_edge_range_p ): Adjust.
14789 (gori_compute::set_range_invariant): Delete.
14790 (gori_compute::dump): Adjust.
14791 (gori_compute::outgoing_edge_range_p): Adjust.
14792 * gimple-range-gori.h (class range_def_chain): Relocate here.
14793 (class gori_map): Relocate here.
14794 (class gori_compute): Inherit from gori_map, and adjust.
14796 2021-05-25 Aldy Hernandez <aldyh@redhat.com>
14798 * value-range.cc (range_tests_legacy): Use
14799 build_nonstandard_integer_type instead of int and short.
14801 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
14803 * gimplify.c (gimplify_decl_expr): Clear TREE_READONLY on the DECL
14804 when really creating an initialization statement for it.
14806 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
14808 * tree-inline.c (setup_one_parameter): Fix thinko in new condition.
14810 2021-05-25 Kito Cheng <kito.cheng@sifive.com>
14812 * config/riscv/riscv.h (ASM_SPEC): Pass -mno-relax.
14814 2021-05-25 Martin Liska <mliska@suse.cz>
14816 PR tree-optimization/92860
14818 * optc-save-gen.awk: Remove exceptions.
14820 2021-05-25 Martin Liska <mliska@suse.cz>
14822 * asan.h (sanitize_coverage_p): New function.
14823 * doc/extend.texi: Document it.
14824 * fold-const.c (fold_range_test): Use sanitize_flags_p
14825 instead of flag_sanitize_coverage.
14826 (fold_truth_andor): Likewise.
14827 * sancov.c: Likewise.
14828 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
14829 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Handle
14830 -fsanitize-coverage when inlining.
14832 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
14834 * config/csky/csky-modes.def : Fix copyright.
14836 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
14838 * config/csky/csky-modes.def : Amend copyright.
14839 * config/csky/csky_insn_fpuv2.md : Likewise.
14840 * config/csky/csky_insn_fpuv3.md : Likewise.
14842 2021-05-25 Richard Biener <rguenther@suse.de>
14844 PR middle-end/100727
14845 * calls.c (initialize_argument_information): Explicitely test
14846 for WITH_SIZE_EXPR.
14847 * gimple-expr.c (mark_addressable): Skip outer WITH_SIZE_EXPR.
14849 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14851 * config/csky/csky.h (FRAME_POINTER_REGNUM): Use
14852 HARD_FRAME_POINTER_REGNUM and FRAME_POINTER_REGNUM instead of
14853 the signle definition. The signle definition may not work well
14854 at simplify_subreg_regno().
14855 (HARD_FRAME_POINTER_REGNUM): New.
14856 (ELIMINABLE_REGS): Add for HARD_FRAME_POINTER_REGNUM.
14857 * config/csky/csky.c (get_csky_live_regs, csky_can_eliminate,
14858 csky_initial_elimination_offset, csky_expand_prologue,
14859 csky_expand_epilogue): Add for HARD_FRAME_POINTER_REGNUM.
14861 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14863 * config/csky/csky.c (csky_option_override):
14864 Init csky_arch_isa_features[] in advance, so TARGET_DSP
14865 and TARGET_DIV can be set well.
14867 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14869 * config/csky/constraints.md ("l", "h"): Delete.
14870 * config/csky/csky.h (reg_class, REG_CLASS_NAMES,
14871 REG_CLASS_CONTENTS): Delete LO_REGS and HI_REGS.
14872 * config/csky/csky.c (regno_reg_classm,
14873 csky_secondary_reload, csky_register_move_cost):
14874 Use HILO_REGS instead of LO_REGS and HI_REGS.
14876 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
14878 * config/csky/constraints.md ("W"): New constriant for mem operand
14879 with base reg, index register.
14880 ("Q"): Renamed and modified "csky_valid_fpuv2_mem_operand" to
14881 "csky_valid_mem_constraint_operand" to deal with both "Q" and "W"
14883 ("Dv"): New constraint for const double value that can be used at
14885 * config/csky/csky-modes.def (HFmode): New mode.
14886 * config/csky/csky-protos.h (csky_valid_fpuv2_mem_operand): Rename
14887 to "csky_valid_mem_constraint_operand" and support new constraint
14889 (csky_get_movedouble_length): New.
14890 (fpuv3_output_move): New.
14891 (fpuv3_const_double): New.
14892 * config/csky/csky.c (csky_option_override): New arch CK860 with fpv3.
14893 (decompose_csky_address): Refine.
14894 (csky_print_operand): New "CONST_DOUBLE" operand.
14895 (csky_output_move): Support fpv3 instructions.
14896 (csky_get_movedouble_length): New.
14897 (fpuv3_output_move): New.
14898 (fpuv3_const_double): New.
14899 (csky_emit_compare): Cover float comparsion.
14900 (csky_emit_compare_float): Refine.
14901 (csky_vaild_fpuv2_mem_operand): Rename to
14902 "csky_valid_mem_constraint_operand" and support new constraint "W".
14903 (ck860_rtx_costs): New.
14904 (csky_rtx_costs): Add the cost calculation of CK860.
14905 (regno_reg_class): New vregs for fpuv3.
14906 (csky_dbx_regno): Likewise.
14907 (csky_cpu_cpp_builtins): New builtin macro for fpuv3.
14908 (csky_conditional_register_usage): Suporrot fpuv3.
14909 (csky_dwarf_register_span): Suporrot fpuv3.
14910 (csky_init_builtins, csky_mangle_type): Support "__fp16" type.
14911 (ck810_legitimate_index_p): Support fp16.
14912 * config/csky/csky.h (TARGET_TLS): ADD CK860.
14913 (CSKY_VREG_P, CSKY_VREG_LO_P, CSKY_VREG_HI_P): Support fpuv3.
14914 (TARGET_SINGLE_FPU): Support fpuv3.
14915 (TARGET_SUPPORT_FPV3): New.
14916 (FIRST_PSEUDO_REGISTER): Change to 202 to hold the new fpuv3 registers.
14917 (FIXED_REGISTERS, CALL_REALLY_USED_REGISTERS, REGISTER_NAMES,
14918 REG_CLASS_CONTENTS): Support fpuv3.
14919 * config/csky/csky.md (movsf): Move to cksy_insn_fpu.md and refine.
14920 (csky_movsf_fpv2): Likewise.
14921 (ck801_movsf): Likewise.
14922 (csky_movsf): Likewise.
14924 (csky_movdf_fpv2): Likewise.
14925 (ck801_movdf): Likewise.
14926 (csky_movdf): Likewise.
14927 (movsicc): Refine. Use "comparison_operatior" instead of
14928 "ordered_comparison_operatior".
14929 (addsicc): Likewise.
14930 (CSKY_FIRST_VFP3_REGNUM, CSKY_LAST_VFP3_REGNUM): New constant.
14931 (call_value_internal_vh): New.
14932 * config/csky/csky_cores.def (CK860): New arch and cpu.
14937 * config/csky/csky_insn_fpu.md: Refactor. Separate all float patterns
14938 into emit-patterns and match-patterns, remain the emit-patterns here,
14939 and move the match-patterns to csky_insn_fpuv2.md or
14940 csky_insn_fpuv3.md.
14941 * config/csky/csky_insn_fpuv2.md: New file for fpuv2 instructions.
14942 * config/csky/csky_insn_fpuv3.md: New file and new patterns for fpuv3
14944 * config/csky/csky_isa.def (fcr): New.
14949 (CK860): New definition for ck860.
14950 * config/csky/csky_tables.opt (ck860): New processors ck860,
14951 ck860f. And new arch ck860.
14956 * config/csky/predicates.md (csky_float_comparsion_operator): Delete
14957 "geu", "gtu", "leu", "ltu", which will never appear at float comparison.
14958 * config/csky/t-csky-elf: Support 860.
14959 * config/csky/t-csky-linux: Likewise.
14960 * doc/md.texi: Add "Q" and "W" constraints for C-SKY.
14962 2021-05-24 Aaron Sawdey <acsawdey@linux.ibm.com>
14964 * config/rs6000/genfusion.pl (gen_logical_addsubf): Refactor to
14965 add generation of logical-add and add-logical fusion pairs.
14966 * config/rs6000/rs6000-cpus.def: Add new fusion to ISA 3.1 mask
14968 * config/rs6000/rs6000.c (rs6000_option_override_internal): Turn on
14969 logical-add and add-logical fusion by default.
14970 * config/rs6000/rs6000.opt: Add -mpower10-fusion-logical-add and
14971 -mpower10-fusion-add-logical options.
14972 * config/rs6000/fusion.md: Regenerate file.
14974 2021-05-24 Aldy Hernandez <aldyh@redhat.com>
14976 * value-range.cc (irange::legacy_equal_p): Check type when
14977 comparing VR_VARYING types.
14978 (range_tests_legacy): Test comparing VARYING ranges of different
14981 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
14983 * config/aarch64/aarch64.c (neoversen1_tunings):
14984 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
14986 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
14988 * config/aarch64/aarch64.c (aarch64_classify_symbol): Use GOT for
14989 extern weak symbols. Limit symbol offsets for non-GOT symbols with
14992 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
14994 * config/arm/neon.md (vec_load_lanesxi<mode>)
14995 (vec_store_lanexoi<mode>): Move ...
14996 * config/arm/vec-common.md: here.
14998 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
15000 * config/arm/neon.md (vec_load_lanesoi<mode>)
15001 (vec_store_lanesoi<mode>): Move ...
15002 * config/arm/vec-common.md: here.
15004 2021-05-24 liuhongt <hongtao.liu@intel.com>
15007 * config/i386/i386.c (ix86_gimple_fold_builtin): Replacing
15008 stmt with GIMPLE_NOP when lhs doesn't exist.
15010 2021-05-23 Uroš Bizjak <ubizjak@gmail.com>
15013 * config/i386/mmx.md (*push<VI_32:mode>2_rex64):
15014 New instruction pattern.
15015 (*push<VI_32:mode>2): Ditto.
15016 (push splitter for SSE registers): New splitter.
15018 2021-05-23 Andrew Pinski <apinski@marvell.com>
15020 * match.pd ((A & C) != 0 ? D : 0): Limit to non pointer types.
15022 2021-05-22 Aaron Sawdey <acsawdey@linux.ibm.com>
15024 * config/rs6000/genfusion.pl (gen_addadd): Fix incorrect attr types.
15025 * config/rs6000/fusion.md: Regenerate file.
15027 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
15029 * config/rs6000/genfusion.pl (gen_addadd): New function.
15030 * config/rs6000/fusion.md: Regenerate file.
15031 * config/rs6000/rs6000-cpus.def: Add
15032 OPTION_MASK_P10_FUSION_2ADD to masks.
15033 * config/rs6000/rs6000.c (rs6000_option_override_internal):
15034 Handle default value of OPTION_MASK_P10_FUSION_2ADD.
15035 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2add.
15037 2021-05-21 Jakub Jelinek <jakub@redhat.com>
15039 PR middle-end/99928
15040 * tree.h (OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET): Define.
15041 * gimplify.c (enum gimplify_omp_var_data): Fix up
15042 GOVD_MAP_HAS_ATTACHMENTS value, add GOVD_FIRSTPRIVATE_IMPLICIT.
15043 (omp_lastprivate_for_combined_outer_constructs): If combined target
15044 has GOVD_FIRSTPRIVATE_IMPLICIT set for the decl, change it to
15045 GOVD_MAP | GOVD_SEEN.
15046 (gimplify_scan_omp_clauses): Set GOVD_FIRSTPRIVATE_IMPLICIT for
15047 firstprivate clauses with OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT.
15048 (gimplify_adjust_omp_clauses): For firstprivate clauses with
15049 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT either clear that bit and
15050 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET too, or remove it and
15051 let it be replaced by implicit map clause.
15053 2021-05-21 Jakub Jelinek <jakub@redhat.com>
15055 PR middle-end/99928
15056 * gimplify.c (omp_lastprivate_for_combined_outer_constructs): New
15058 (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LASTPRIVATE>: Use it.
15059 (gimplify_omp_for): Likewise.
15061 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15063 PR middle-end/90115
15064 * omp-low.c (oacc_privatization_candidate_p): Reject 'static',
15065 'external' in blocks.
15067 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15069 PR middle-end/90115
15070 * flag-types.h (enum openacc_privatization): New.
15071 * params.opt (-param=openacc-privatization): New.
15072 * doc/invoke.texi (openacc-privatization): Document it.
15073 * omp-general.h (get_openacc_privatization_dump_flags): New
15075 * omp-low.c (oacc_privatization_candidate_p): Add diagnostics.
15076 * omp-offload.c (execute_oacc_device_lower)
15077 <IFN_UNIQUE_OACC_PRIVATE>: Re-work diagnostics.
15078 * target.def (goacc.adjust_private_decl): Add 'location_t'
15080 * doc/tm.texi: Regenerate.
15081 * config/gcn/gcn-protos.h (gcn_goacc_adjust_private_decl): Adjust.
15082 * config/gcn/gcn-tree.c (gcn_goacc_adjust_private_decl): Likewise.
15083 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl):
15084 Likewise. Preserve it for...
15085 (nvptx_goacc_expand_var_decl): ... use here.
15087 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15089 * doc/sourcebuild.texi (Other attributes): Document '__OPTIMIZE__'
15092 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15094 PR middle-end/90115
15095 * omp-low.c (oacc_privatization_candidate_p): New function.
15096 (oacc_privatization_scan_clause_chain)
15097 (oacc_privatization_scan_decl_chain): Use it. Also
15098 'gcc_checking_assert' that we're not seeing duplicates.
15100 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15102 PR middle-end/90115
15103 * omp-offload.c (execute_oacc_device_lower): Skip processing if no
15106 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15108 PR middle-end/90115
15109 * omp-offload.c (execute_oacc_device_lower): Explain.
15111 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15113 PR middle-end/90115
15114 * omp-offload.c (execute_oacc_device_lower)
15115 <IFN_UNIQUE_OACC_PRIVATE>: Diagnose and handle for 'level == -1'
15117 * internal-fn.c (expand_UNIQUE): Don't expect
15118 'IFN_UNIQUE_OACC_PRIVATE'.
15120 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15122 PR middle-end/90115
15123 * omp-low.c (lower_omp_for): Don't evaluate OpenMP 'for' clauses.
15125 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
15127 PR middle-end/90115
15128 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl)
15129 (nvptx_goacc_expand_var_decl): Tighten.
15131 2021-05-21 Julian Brown <julian@codesourcery.com>
15132 Chung-Lin Tang <cltang@codesourcery.com>
15133 Thomas Schwinge <thomas@codesourcery.com>
15135 PR middle-end/90115
15136 * doc/tm.texi.in (TARGET_GOACC_EXPAND_VAR_DECL)
15137 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Add documentation hooks.
15138 * doc/tm.texi: Regenerate.
15139 * expr.c (expand_expr_real_1): Expand decls using the
15140 expand_var_decl OpenACC hook if defined.
15141 * internal-fn.c (expand_UNIQUE): Handle IFN_UNIQUE_OACC_PRIVATE.
15142 * internal-fn.h (IFN_UNIQUE_CODES): Add OACC_PRIVATE.
15143 * omp-low.c (omp_context): Add oacc_privatization_candidates
15145 (lower_oacc_reductions): Add PRIVATE_MARKER parameter. Insert
15147 (lower_oacc_head_tail): Add PRIVATE_MARKER parameter. Modify
15148 private marker's gimple call arguments, and pass it to
15149 lower_oacc_reductions.
15150 (oacc_privatization_scan_clause_chain)
15151 (oacc_privatization_scan_decl_chain, lower_oacc_private_marker):
15153 (lower_omp_for, lower_omp_target, lower_omp_1): Use these.
15154 * omp-offload.c (convert.h): Include.
15155 (oacc_loop_xform_head_tail): Treat private-variable markers like
15156 fork/join when transforming head/tail sequences.
15157 (struct var_decl_rewrite_info): Add struct.
15158 (oacc_rewrite_var_decl, is_sync_builtin_call): New functions.
15159 (execute_oacc_device_lower): Support rewriting gang-private
15160 variables using target hook, and fix up addr_expr and var_decl
15162 * target.def (adjust_private_decl, expand_var_decl): New hooks.
15163 * config/gcn/gcn-protos.h (gcn_goacc_adjust_gangprivate_decl):
15165 (gcn_goacc_adjust_private_decl): ...this.
15166 * config/gcn/gcn-tree.c (gcn_goacc_adjust_gangprivate_decl):
15168 (gcn_goacc_adjust_private_decl): ...this. Add LEVEL parameter.
15169 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_GANGPRIVATE_DECL): Rename
15170 definition using gcn_goacc_adjust_gangprivate_decl...
15171 (TARGET_GOACC_ADJUST_PRIVATE_DECL): ...to this, using
15172 gcn_goacc_adjust_private_decl.
15173 * config/nvptx/nvptx.c (tree-pretty-print.h): Include.
15174 (gang_private_shared_size): New global variable.
15175 (gang_private_shared_align): Likewise.
15176 (gang_private_shared_sym): Likewise.
15177 (gang_private_shared_hmap): Likewise.
15178 (nvptx_option_override): Initialize these.
15179 (nvptx_file_end): Output gang_private_shared_sym.
15180 (nvptx_goacc_adjust_private_decl, nvptx_goacc_expand_var_decl):
15182 (nvptx_set_current_function): Clear gang_private_shared_hmap.
15183 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Define hook.
15184 (TARGET_GOACC_EXPAND_VAR_DECL): Likewise.
15186 2021-05-21 H.J. Lu <hjl.tools@gmail.com>
15188 * config/i386/i386-modes.def (MAX_BITSIZE_MODE_ANY_INT): Removed.
15190 2021-05-21 Richard Biener <rguenther@suse.de>
15191 H.J. Lu <hjl.tools@gmail.com>
15193 PR middle-end/90773
15194 * expr.c (expand_constructor): Elide expand_constructor if
15195 move by pieces is preferred.
15197 2021-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15199 * config/aarch64/aarch64-builtins.c (aarch64_call_properties):
15200 Take a flag and mode value as arguments.
15201 (aarch64_modifies_global_state_p): Likewise.
15202 (aarch64_reads_global_state_p): Likewise.
15203 (aarch64_could_trap_p): Likewise.
15204 (aarch64_get_attributes): Likewise.
15205 (aarch64_init_simd_builtins): Adjust callsite of above.
15206 (aarch64_init_fcmla_laneq_builtins): Use aarch64_get_attributes to get
15207 function attributes to apply to builtins.
15208 (aarch64_init_crc32_builtins): Likewise.
15209 (aarch64_init_builtin_rsqrt): Likewise.
15211 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
15213 * config/rs6000/rs6000.md (define_attr "type"): Add types for fusion.
15214 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10): Use new fusion types.
15215 (gen_2logical): Use new fusion types.
15216 * config/rs6000/fusion.md: Regenerate.
15218 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
15221 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
15222 Handle V4QI and V2HI modes.
15223 (ix86_expand_sse_movcc): Ditto.
15224 * config/i386/mmx.md (*<sat_plusminus:insn><VI_32:mode>3):
15225 New instruction pattern.
15226 (*eq<VI_32:mode>3): Ditto.
15227 (*gt<VI_32:mode>3): Ditto.
15228 (*xop_pcmov_<VI_32:mode>): Ditto.
15229 (mmx_pblendvb32): Ditto.
15230 (mmx_pblendvb64): Rename from mmx_pblendvb.
15231 (vec_cmp<VI_32:mode><VI_32:mode>): New expander.
15232 (vec_cmpu<VI_32:mode><VI_32:mode>): Ditto.
15233 (vcond<VI_32:mode><VI_32:mode>): Ditto.
15234 (vcondu<VI_32:mode><VI_32:mode>): Ditto.
15235 (vcond_mask_<VI_32:mode><VI_32:mode>): Ditto.
15237 2021-05-21 Jakub Jelinek <jakub@redhat.com>
15239 PR tree-optimization/94589
15240 * tree-ssa-phiopt.c (spaceship_replacement): For integral rhs1 and
15241 rhs2, treat x <= 4 equivalently to x < 5 etc. In cmp1 and cmp2 (if
15242 not the same as cmp3) treat <= the same as < and >= the same as >.
15243 Don't require that cond2_phi_edge is true edge, instead take
15244 false/true edges into account based on cmp1/cmp2 comparison kinds.
15246 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
15249 * config/i386/mmx.md (SMAXMIN_MMXMODEI): New mode iterator.
15250 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Macroize expander
15251 from <smaxmin:code>v4hi3> and <smaxmin:code><MMXMODE14:mode>3
15252 using SMAXMIN_MMXMODEI mode iterator.
15253 (*<smaxmin:code>v4qi3): New insn pattern.
15254 (*<smaxmin:code>v2hi3): Ditto.
15255 (SMAXMIN_VI_32): New mode iterator.
15256 (<smaxmin:code><SMAXMIN_VI_32>mode3): New expander.
15257 (UMAXMIN_MMXMODEI): New mode iterator.
15258 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Macroize expander
15259 from <umaxmin:code>v8qi3> and <umaxmin:code><MMXMODE24:mode>3
15260 using UMAXMIN_MMXMODEI mode iterator.
15261 (*<umaxmin:code>v4qi3): New insn pattern.
15262 (*<umaxmin:code>v2hi3): Ditto.
15263 (UMAXMIN_VI_32): New mode iterator.
15264 (<umaxmin:code><UMAXMIN_VI_32>mode3): New expander.
15265 (abs<VI_32:mode>2): New insn pattern.
15266 (ssse3_abs<MMXMODEI:mode>2, abs<MMXMODEI:mode>2): Move from ...
15267 * config/i386/sse.md: ... here.
15269 2021-05-20 Clement Chigot <clement.chigot@atos.net>
15270 David Edelsohn <dje.gcc@gmail.com>
15272 * collect2.c (scan_prog_file): Issue non-fatal warning for
15275 2021-05-20 Jonathan Wakely <jwakely@redhat.com>
15277 * doc/invoke.texi (-Wno-c++11-extensions)
15278 (-Wno-c++14-extensions, -Wno-c++17-extensions)
15279 (-Wno-c++20-extensions, -Wno-c++23-extensions): Document
15282 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
15284 * config/c6x/c6x.c (c6x_output_file_unwind): Use dwarf_debuginfo_p.
15285 * config/darwin.c (darwin_override_options): Likewise.
15286 * config/i386/cygming.h (DBX_REGISTER_NUMBER): Likewise.
15287 * config/i386/darwin.h (DBX_REGISTER_NUMBER): Likewise.
15288 (DWARF2_FRAME_REG_OUT): Likewise.
15289 * config/mips/mips.c (mips_output_filename): Likewise.
15290 * config/rs6000/rs6000.c (rs6000_xcoff_declare_function_name):
15292 (rs6000_dbx_register_number): Likewise.
15293 * dbxout.c: Include flags.h.
15294 * dwarf2cfi.c (cfi_label_required_p): Likewise.
15295 (dwarf2out_do_frame): Likewise.
15296 * except.c: Include flags.h.
15297 * final.c (dwarf2_debug_info_emitted_p): Likewise.
15298 (final_scan_insn_1): Likewise.
15299 * flags.h (dwarf_debuginfo_p): New function declaration.
15300 * opts.c (dwarf_debuginfo_p): New function definition.
15301 * targhooks.c (default_debug_unwind_info): Use dwarf_debuginfo_p.
15302 * toplev.c (process_options): Likewise.
15304 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
15306 * common.opt: Change type to support bitmasks.
15307 * flag-types.h (enum debug_info_type): Rename enumerator constants.
15308 (NO_DEBUG): New bitmask.
15309 (DBX_DEBUG): Likewise.
15310 (DWARF2_DEBUG): Likewise.
15311 (XCOFF_DEBUG): Likewise.
15312 (VMS_DEBUG): Likewise.
15313 (VMS_AND_DWARF2_DEBUG): Likewise.
15314 * flags.h (debug_set_to_format): New function declaration.
15315 (debug_set_count): Likewise.
15316 (debug_set_names): Likewise.
15317 * opts.c (debug_type_masks): Array of bitmasks for debug formats.
15318 (debug_set_to_format): New function definition.
15319 (debug_set_count): Likewise.
15320 (debug_set_names): Likewise.
15321 (set_debug_level): Update access to debug_type_names.
15322 * toplev.c: Likewise.
15324 2021-05-20 Martin Sebor <msebor@redhat.com>
15326 PR middle-end/100684
15327 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Handle C++ lambda.
15329 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
15332 * config/i386/i386.md (isa): Remove x64_bmi.
15333 (enabled): Remove x64_bmi.
15334 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
15335 Remove general register alternative.
15336 (*andnot<VI_32:mode>3): Ditto.
15337 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
15338 (*<any_logic:code><VI_32:mode>3): Ditto.
15340 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
15342 * config/arm/arm.c: Include head files tree-vectorizer.h and
15345 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
15348 * config/i386/mmx.md (Yv_Yw): Revert adding V4QI and V2HI modes.
15349 (*<plusminus:insn><VI_32:mode>3): Use Yw instad of <Yv_Yw> constrint.
15350 (<s>mulv4hi3_highpart): New expander.
15351 (*<s>mulv2hi3_highpart): New insn pattern.
15352 (<s>mulv2hi3_higpart): New expander.
15353 (*<any_shift:insn>v2hi3): New insn pattern.
15354 (<any_shift:insn>v2hi3): New expander.
15355 * config/i386/sse.md (smulhrsv2hi3): New expander.
15356 (*smulhrsv2hi3): New insn pattern.
15358 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
15360 * doc/invoke.texi (vect-inner-loop-cost-factor): Document new
15362 * params.opt (vect-inner-loop-cost-factor): New.
15363 * targhooks.c (default_add_stmt_cost): Replace hardcoded factor
15364 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR, include head file
15365 tree-vectorizer.h and its required ones.
15366 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Replace
15367 hardcoded factor 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR.
15368 * config/arm/arm.c (arm_add_stmt_cost): Likewise.
15369 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
15370 * config/rs6000/rs6000.c (rs6000_add_stmt_cost): Likewise.
15371 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
15373 (_loop_vec_info::_loop_vec_info): Init inner_loop_cost_factor.
15374 * tree-vectorizer.h (_loop_vec_info): Add inner_loop_cost_factor.
15375 (LOOP_VINFO_INNER_LOOP_COST_FACTOR): New macro.
15377 2021-05-20 Christophe Lyon <christophe.lyon@linaro.org>
15378 Torbjörn Svensson <torbjorn.svensson@st.com>
15381 * doc/cpp.texi (Common Predefined Macros): Document __FILE_NAME__.
15383 2021-05-20 Jakub Jelinek <jakub@redhat.com>
15385 PR middle-end/99928
15386 * gimplify.c (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LINEAR>: For
15387 explicit linear clause when combined with target, make it map(tofrom:)
15388 instead of no clause or firstprivate.
15390 2021-05-20 Jakub Jelinek <jakub@redhat.com>
15392 PR tree-optimization/94589
15393 * match.pd ((X & Y) == X -> (X & ~Y) == 0): Simplify even in presence
15394 of integral conversions.
15396 2021-05-19 Andrew MacLeod <amacleod@redhat.com>
15398 * gimple-range.cc (fur_source::get_operand): New.
15399 (gimple_range_fold): Delete.
15400 (fold_using_range::fold_stmt): Move from gimple_ranger::calc_stmt.
15401 (fold_using_range::range_of_range_op): Move from gimple_ranger.
15402 (fold_using_range::range_of_address): Ditto.
15403 (fold_using_range::range_of_phi): Ditto.
15404 (fold_using_range::range_of_call): Ditto.
15405 (fold_using_range::range_of_builtin_ubsan_call): Move from
15406 range_of_builtin_ubsan_call.
15407 (fold_using_range::range_of_builtin_call): Move from
15408 range_of_builtin_call.
15409 (gimple_ranger::range_of_builtin_call): Delete.
15410 (fold_using_range::range_of_cond_expr): Move from gimple_ranger.
15411 (gimple_ranger::fold_range_internal): New.
15412 (gimple_ranger::range_of_stmt): Use new fold_using_range API.
15413 (fold_using_range::range_of_ssa_name_with_loop_info): Move from
15414 gimple_ranger. Improve ranges of SSA_NAMES when possible.
15415 * gimple-range.h (gimple_ranger): Remove various range_of routines.
15416 (class fur_source): New.
15417 (class fold_using_range): New.
15418 (fur_source::fur_source): New.
15420 * vr-values.c (vr_values::extract_range_basic): Use fold_using_range
15421 instead of range_of_builtin_call.
15423 2021-05-19 Jonathan Wakely <jwakely@redhat.com>
15425 * doc/cpp.texi (Common Predefined Macros): Update documentation
15426 for the __GXX_EXPERIMENTAL_CXX0X__ macro.
15428 2021-05-19 Alex Coplan <alex.coplan@arm.com>
15431 * config/arm/arm.md (nonsecure_call_internal): Always ensure
15432 callee's address is in a register.
15434 2021-05-19 Geng Qi <gengqi@linux.alibaba.com>
15436 * common/config/riscv/riscv-common.c
15437 (riscv_subset_list::parsing_subset_version): Properly parse the letter
15439 (riscv_subset_list::parse_std_ext,
15440 riscv_subset_list::parse_multiletter_ext): To handle errors generated
15441 in riscv_subset_list::parsing_subset_version.
15443 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15445 * config/aarch64/aarch64-simd.md: Use "neon_move_narrow_q"
15446 type attribute in patterns generating XTN(2).
15448 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15450 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_pack_trunc_<mode>):
15451 Remove as duplicate of...
15452 (aarch64_xtn<mode>): This.
15453 (aarch64_xtn2<mode>_le): Move position in file.
15454 (aarch64_xtn2<mode>_be): Move position in file.
15455 (aarch64_xtn2<mode>): Move position in file.
15456 (vec_pack_trunc_<mode>): Define as an expander.
15458 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15460 * config/aarch64/aarch64-simd-builtins.def: Split builtin
15461 generation for aarch64_<sur>q<r>shr<u>n_n<mode> pattern into
15462 separate scalar and vector generators.
15463 * config/aarch64/aarch64-simd.md
15464 (aarch64_<sur>q<r>shr<u>n_n<mode>): Define as an expander and
15466 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_le): This and...
15467 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_be): This.
15468 * config/aarch64/iterators.md: Define SD_HSDI iterator.
15470 2021-05-19 Jonathn Wright <jonathan.wright@arm.com>
15472 * config/aarch64/aarch64-simd.md: Use UNSPEC_SQXTUN instead
15474 * config/aarch64/iterators.md: Remove UNSPEC_SQXTUN2.
15476 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15478 * config/aarch64/aarch64-simd.md (aarch64_<sur>q<r>shr<u>n2_n<mode>):
15479 Implement as an expand emitting a big/little endian
15480 instruction pattern.
15481 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_le): Define.
15482 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_be): Define.
15484 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
15486 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn2<mode>):
15487 Implement as an expand emitting a big/little endian
15488 instruction pattern.
15489 (aarch64_<sur><addsub>hn2<mode>_insn_le): Define.
15490 (aarch64_<sur><addsub>hn2<mode>_insn_be): Define.
15491 * config/aarch64/iterators.md: Remove UNSPEC_[R]ADDHN2 and
15492 UNSPEC_[R]SUBHN2 unspecs and ADDSUBHN2 iterator.
15494 2021-05-19 Richard Biener <rguenther@suse.de>
15496 PR middle-end/100672
15497 * fold-const.c (fold_negate_expr_1): Use element_precision.
15498 (negate_expr_p): Likewise.
15500 2021-05-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
15502 * config/aarch64/iterators.md (SVE_PRED_LOAD): New iterator.
15503 (pred_load): New int attribute.
15504 * config/aarch64/aarch64-sve.md
15505 (aarch64_load_<ANY_EXTEND:optab><SVE_HSDI:mode><SVE_PARTIAL_I:mode>): Use
15506 SVE_PRED_LOAD enum iterator and corresponding pred_load attribute.
15507 * config/aarch64/aarch64-sve-builtins-base.cc (expand): Update call to
15508 code_for_aarch64_load.
15510 2021-05-19 Richard Biener <rguenther@suse.de>
15512 * cfgexpand.c (discover_nonconstant_array_refs_r): Make
15513 sure TARGET_MEM_REF bases are expanded as memory.
15514 * tree-ssa-operands.c (operands_scanner::get_tmr_operands):
15515 Do not mark TARGET_MEM_REF bases addressable.
15516 * tree-ssa.c (non_rewritable_mem_ref_base): Handle
15517 TARGET_MEM_REF bases as never rewritable.
15518 * gimple-walk.c (walk_stmt_load_store_addr_ops): Do not
15519 walk TARGET_MEM_REF bases as address-takens.
15520 * tree-ssa-dce.c (ref_may_be_aliased): Handle TARGET_MEM_REF.
15522 2021-05-19 Richard Biener <rguenther@suse.de>
15524 * builtins.c (get_object_alignment_1): Strip outer
15526 * tree-dfa.c (get_ref_base_and_extent): Handle outer
15527 WITH_SIZE_EXPR for size processing and process the
15529 * tree-ssa-alias.c (ao_ref_base_alias_set): Strip
15530 outer WITH_SIZE_EXPR.
15531 (ao_ref_base_alias_ptr_type): Likewise.
15532 (refs_may_alias_p_2): Allow WITH_SIZE_EXPR in ref->ref
15533 and handle that accordingly, stripping it for the
15534 core alias workers.
15535 * tree.c (get_base_address): Handle WITH_SIZE_EXPR by
15536 looking through it instead of returning NULL.
15538 2021-05-19 Jakub Jelinek <jakub@redhat.com>
15540 PR middle-end/100576
15541 * builtins.c (check_read_access): Convert bound to size_type_node if
15544 2021-05-19 Richard Biener <rguenther@suse.de>
15546 * tree-cfg.c (verify_types_in_gimple_min_lval): Inline...
15547 (verify_types_in_gimple_reference): ... here. Sanitize.
15548 (verify_gimple_call): Verify references in LHS and arguments.
15549 (verify_gimple_assign_single): Reject WITH_SIZE_EXPR.
15551 2021-05-19 Uroš Bizjak <ubizjak@gmail.com>
15553 * config/i386/i386.h (VALID_INT_MODE_P):
15554 Add V8QI, V4HI and V2SI modes for TARGET_64BIT.
15555 * config/i386/i386.md (isa): Add x64_bmi.
15556 (enabled): Handle x64_bmi.
15557 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
15558 Add alternative using 64bit general registers.
15559 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
15561 2021-05-19 Jakub Jelinek <jakub@redhat.com>
15563 PR middle-end/99928
15564 * tree.h (OMP_MASTER_COMBINED): Define.
15565 * gimplify.c (gimplify_scan_omp_clauses): Rewrite lastprivate
15566 handling for outer combined/composite constructs to a loop.
15567 Handle lastprivate on combined target.
15568 (gimplify_expr): Formatting fix.
15570 2021-05-19 Xionghu Luo <luoxhu@linux.ibm.com>
15572 * passes.def: Add sink_code pass before store_merging.
15573 * tree-ssa-sink.c (pass_sink_code:clone): New.
15575 2021-05-18 Bill Schmidt <wschmidt@linux.ibm.com>
15577 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Remove call to
15578 rs6000_special_adjust_field_align_p.
15579 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
15580 * config/rs6000/rs6000-call.c (rs6000_function_arg_boundary):
15581 Remove ABI warning.
15582 (rs6000_function_arg): Likewise.
15583 * config/rs6000/rs6000-protos.h
15584 (rs6000_special_adjust_field_align_p): Remove prototype.
15585 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
15587 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Remove call to
15588 rs6000_special_adjust_field_align_p.
15590 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
15593 * config/i386/i386.h (VALID_SSE2_REG_MODE):
15594 Add V4QI and V2HI modes.
15595 (VALID_INT_MODE_P): Ditto.
15596 * config/i386/mmx.md (VI_32): New mode iterator.
15597 (mmxvecsize): Handle V4QI and V2HI.
15599 (mov<VI_32:mode>): New expander.
15600 (*mov<mode>_internal): New insn pattern.
15601 (movmisalign<VI_32:mode>): New expander.
15602 (neg<VI_32:mode>): New expander.
15603 (<plusminus:insn><VI_32:mode>3): New expander.
15604 (*<plusminus:insn><VI_32:mode>3): New insn pattern.
15605 (mulv2hi3): New expander.
15606 (*mulv2hi3): New insn pattern.
15607 (one_cmpl<VI_32:mode>2): New expander.
15608 (*andnot<VI_32:mode>3): New insn pattern.
15609 (<any_logic:code><VI_32:mode>3): New expander.
15610 (*<any_logic:code><VI_32:mode>3): New insn pattern.
15612 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
15614 * config/i386/sse.md (<any_extend:insn>v4qiv4di2):
15615 Fix a mode mismatch with operand 1.
15617 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
15620 * config/i386/i386-expand.c (split_double_mode): Return
15621 temporary register when simplify_gen_subreg fails with
15622 the high half od the paradoxical subreg.
15624 2021-05-18 Richard Biener <rguenther@suse.de>
15626 * cfgexpand.c (expand_one_var): Pass in forced_stack_var
15627 and honor it when expanding.
15628 (expand_used_vars_for_block): Pass through forced_stack_var.
15629 (expand_used_vars): Likewise.
15630 (discover_nonconstant_array_refs_r): Set bits in
15631 forced_stack_vars instead of marking vars TREE_ADDRESSABLE.
15632 (avoid_type_punning_on_regs): Likewise.
15633 (discover_nonconstant_array_refs): Likewise.
15634 (pass_expand::execute): Create and pass down forced_stack_var
15635 bitmap. For parameters and returns temporarily set
15636 TREE_ADDRESSABLE when expand_function_start.
15638 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
15640 * doc/sourcebuild.texi: Document 'dg-note'.
15642 2021-05-18 Tobias Burnus <tobias@codesourcery.com>
15645 * configure: Regenerate.
15646 * configure.ac (BUILD_CFLAG, BUILD_CXXFLAGS): Add $(CFLAGS-$@).
15648 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
15650 * gimple.h (is_gimple_omp_oacc): Tighten.
15651 * omp-low.c (check_omp_nesting_restrictions): Adjust.
15653 2021-05-18 Richard Biener <rguenther@suse.de>
15655 * tree-ssa-operands.c (mark_address_taken): Simplify.
15657 2021-05-18 Martin Liska <mliska@suse.cz>
15659 * config/gcn/mkoffload.c (STR): Redefine.
15660 * config/i386/intelmic-mkoffload.c (STR): Likewise.
15661 * config/nvptx/mkoffload.c (STR): Likewise.
15663 2021-05-18 Martin Liska <mliska@suse.cz>
15665 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
15666 Use startswith function instead of strncmp.
15667 * common/config/bfin/bfin-common.c (bfin_handle_option): Likewise.
15668 * common/config/riscv/riscv-common.c (riscv_subset_list::parse): Likewise.
15669 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): Likewise.
15670 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Likewise.
15671 * config/alpha/alpha.c (alpha_elf_section_type_flags): Likewise.
15672 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
15673 * config/arm/arm.c (arm_file_start): Likewise.
15674 (arm_valid_target_attribute_rec): Likewise.
15675 (thumb1_md_asm_adjust): Likewise.
15676 * config/arm/driver-arm.c (host_detect_local_cpu): Likewise.
15677 * config/avr/avr.c (STR_PREFIX_P): Likewise.
15678 (avr_set_current_function): Likewise.
15679 (avr_handle_addr_attribute): Likewise.
15680 (avr_asm_output_aligned_decl_common): Likewise.
15681 (avr_asm_named_section): Likewise.
15682 (avr_section_type_flags): Likewise.
15683 (avr_asm_select_section): Likewise.
15684 * config/c6x/c6x.c (c6x_in_small_data_p): Likewise.
15685 (c6x_section_type_flags): Likewise.
15686 * config/darwin-c.c (darwin_cfstring_ref_p): Likewise.
15687 (darwin_objc_declare_unresolved_class_reference): Likewise.
15688 (darwin_objc_declare_class_definition): Likewise.
15689 * config/darwin.c (indirect_data): Likewise.
15690 (darwin_encode_section_info): Likewise.
15691 (darwin_objc2_section): Likewise.
15692 (darwin_objc1_section): Likewise.
15693 (machopic_select_section): Likewise.
15694 (darwin_globalize_label): Likewise.
15695 (darwin_label_is_anonymous_local_objc_name): Likewise.
15696 (darwin_asm_named_section): Likewise.
15697 (darwin_asm_output_dwarf_offset): Likewise.
15698 * config/frv/frv.c (frv_string_begins_with): Likewise.
15699 (frv_in_small_data_p): Likewise.
15700 * config/gcn/mkoffload.c (STR): Likewise.
15702 * config/i386/i386-builtins.c (get_builtin_code_for_version): Likewise.
15703 * config/i386/i386-options.c (ix86_option_override_internal): Likewise.
15704 * config/i386/i386.c (x86_64_elf_section_type_flags): Likewise.
15705 (ix86_md_asm_adjust): Likewise.
15706 * config/i386/intelmic-mkoffload.c (STR): Likewise.
15707 * config/i386/winnt.c (i386_pe_asm_named_section): Likewise.
15708 (i386_pe_file_end): Likewise.
15709 * config/ia64/ia64.c (ia64_in_small_data_p): Likewise.
15710 (ia64_section_type_flags): Likewise.
15711 * config/mips/driver-native.c (host_detect_local_cpu): Likewise.
15712 * config/mips/mips.c (mips_handle_interrupt_attr): Likewise.
15713 (mips16_stub_function_p): Likewise.
15714 (mips_function_rodata_section): Likewise.
15715 * config/msp430/msp430.c (msp430_mcu_name): Likewise.
15716 (msp430_function_section): Likewise.
15717 (msp430_section_type_flags): Likewise.
15718 (msp430_expand_helper): Likewise.
15719 * config/nios2/nios2.c (nios2_small_section_name_p): Likewise.
15720 (nios2_valid_target_attribute_rec): Likewise.
15721 * config/nvptx/mkoffload.c (process): Likewise.
15723 * config/pa/som.h: Likewise.
15724 * config/pdp11/pdp11.c (pdp11_output_ident): Likewise.
15725 * config/riscv/riscv.c (riscv_elf_select_rtx_section): Likewise.
15726 * config/rs6000/rs6000.c (VTABLE_NAME_P): Likewise.
15727 (rs6000_inner_target_options): Likewise.
15728 * config/s390/driver-native.c (s390_host_detect_local_cpu): Likewise.
15729 * config/sparc/driver-sparc.c (host_detect_local_cpu): Likewise.
15730 * config/vax/vax.c (vax_output_int_move): Likewise.
15731 * config/vms/vms-ld.c (startswith): Likewise.
15732 (process_args): Likewise.
15734 * config/vms/vms.c: Likewise.
15736 2021-05-18 Jakub Jelinek <jakub@redhat.com>
15738 PR rtl-optimization/100590
15739 * regcprop.c (copyprop_hardreg_forward_1): Only DCE dead sets if
15740 they are NONJUMP_INSN_P.
15742 2021-05-18 Jakub Jelinek <jakub@redhat.com>
15745 * function.c (push_dummy_function): Set DECL_ARTIFICIAL and
15746 DECL_ASSEMBLER_NAME on the fn_decl.
15748 2021-05-18 Jakub Jelinek <jakub@redhat.com>
15750 PR tree-optimization/94589
15751 * tree-ssa-phiopt.c (spaceship_replacement): Pattern match
15752 phi result used in (res & ~1) == 0 comparison as res >= 0 as
15753 res == 2 would be UB with -ffinite-math-only.
15755 2021-05-18 Martin Liska <mliska@suse.cz>
15757 * Makefile.in: genversion.o should depend on DATESTAMP.
15759 2021-05-18 Claudiu Zissulescu <claziss@synopsys.com>
15761 * config/arc/simdext.md (negv2si2): Remove round bracket.
15763 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
15765 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Define
15766 _Bool as macro expanding to _Bool.
15768 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
15771 * tree.c (build_reference_type_for_mode)
15772 (build_pointer_type_for_mode): Pick pointer mode if MODE argument
15774 (build_reference_type, build_pointer_type): Invoke
15775 build_*_type_for_mode with VOIDmode.
15777 2021-05-17 Andrew MacLeod <amacleod@redhat.com>
15779 PR tree-optimization/100512
15780 * gimple-range-cache.cc (ranger_cache::set_global_range): Mark const
15781 and non-zero pointer ranges as invariant.
15782 * gimple-range.cc (gimple_ranger::range_of_stmt): Remove pointer
15783 processing from here.
15785 2021-05-17 Tom de Vries <tdevries@suse.de>
15788 * config/nvptx/nvptx-protos.h (nvptx_output_atomic_insn): Declare
15789 * config/nvptx/nvptx.c (nvptx_output_barrier)
15790 (nvptx_output_atomic_insn): New function.
15791 (nvptx_print_operand): Add support for 'B'.
15792 * config/nvptx/nvptx.md: Use nvptx_output_atomic_insn for atomic
15795 2021-05-17 Aldy Hernandez <aldyh@redhat.com>
15797 PR tree-optimization/100349
15798 * vr-values.c (bounds_of_var_in_loop): Bail if scev returns
15801 2021-05-17 Tamar Christina <tamar.christina@arm.com>
15803 * config/aarch64/driver-aarch64.c (DEFAULT_ARCH): New.
15804 (host_detect_local_cpu): Use it.
15806 2021-05-17 Martin Liska <mliska@suse.cz>
15808 * doc/invoke.texi: Add 2 missing dots.
15810 2021-05-17 Marius Hillenbrand <mhillen@linux.ibm.com>
15812 PR bootstrap/100552
15813 * configure.ac: Replace pattern substitution with call to sed.
15814 * configure: Regenerate.
15816 2021-05-17 Richard Biener <rguenther@suse.de>
15818 PR middle-end/100582
15819 * tree.c (array_at_struct_end_p): Get to the base of the
15820 reference before looking for the underlying decl.
15822 2021-05-17 Joern Rennecke <joern.rennecke@embecosm.com>
15824 * genoutput.c (validate_insn_alternatives) Make "wrong number of
15825 alternatives" message more specific, and remove assumption on where
15828 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
15830 * config/arm/iterators.md (V16): New iterator.
15831 (VH_cvtto): New iterator.
15832 (v_cmp_result): Added V4HF and V8HF support.
15833 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result>): Use VDQWH.
15834 (vcond<mode><mode>): Likewise.
15835 (vcond_mask_<mode><v_cmp_result>): Likewise.
15836 (vcond<VH_cvtto><mode>): New expander.
15838 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
15840 * config/arm/arm-protos.h (arm_expand_vector_compare): Update
15842 * config/arm/arm.c (arm_expand_vector_compare): Add support for
15844 (arm_expand_vcond): Likewise.
15845 * config/arm/iterators.md (supf): Remove VCMPNEQ_S, VCMPEQQ_S,
15846 VCMPEQQ_N_S, VCMPNEQ_N_S.
15847 (VCMPNEQ, VCMPEQQ, VCMPEQQ_N, VCMPNEQ_N): Remove.
15848 * config/arm/mve.md (@mve_vcmp<mve_cmp_op>q_<mode>): Add '@' prefix.
15849 (@mve_vcmp<mve_cmp_op>q_f<mode>): Likewise.
15850 (@mve_vcmp<mve_cmp_op>q_n_f<mode>): Likewise.
15851 (@mve_vpselq_<supf><mode>): Likewise.
15852 (@mve_vpselq_f<mode>"): Likewise.
15853 * config/arm/neon.md (vec_cmp<mode><v_cmp_result): Enable for MVE
15854 and move to vec-common.md.
15855 (vec_cmpu<mode><mode>): Likewise.
15856 (vcond<mode><mode>): Likewise.
15857 (vcond<V_cvtto><mode>): Likewise.
15858 (vcondu<mode><v_cmp_result>): Likewise.
15859 (vcond_mask_<mode><v_cmp_result>): Likewise.
15860 * config/arm/unspecs.md (VCMPNEQ_U, VCMPNEQ_S, VCMPEQQ_S)
15861 (VCMPEQQ_N_S, VCMPNEQ_N_S, VCMPEQQ_U, CMPEQQ_N_U, VCMPNEQ_N_U)
15862 (VCMPGEQ_N_S, VCMPGEQ_S, VCMPGTQ_N_S, VCMPGTQ_S, VCMPLEQ_N_S)
15863 (VCMPLEQ_S, VCMPLTQ_N_S, VCMPLTQ_S, VCMPCSQ_N_U, VCMPCSQ_U)
15864 (VCMPHIQ_N_U, VCMPHIQ_U): Remove.
15865 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result): Moved
15867 (vec_cmpu<mode><mode>): Likewise.
15868 (vcond<mode><mode>): Likewise.
15869 (vcond<V_cvtto><mode>): Likewise.
15870 (vcondu<mode><v_cmp_result>): Likewise.
15871 (vcond_mask_<mode><v_cmp_result>): Likewise. Added unsafe math
15874 2021-05-17 liuhongt <hongtao.liu@intel.com>
15877 * config/i386/i386.c (ix86_gimple_fold_builtin): Use
15878 gsi_insert_seq_before instead.
15880 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
15882 * doc/sourcebuild.texi (arm_qbit_ok): Rename into...
15883 (arm_sat_ok): ...this.
15885 2021-05-17 Martin Liska <mliska@suse.cz>
15887 * lto-wrapper.c (merge_flto_options): Factor out a new function.
15888 (merge_and_complain): Use it.
15889 (run_gcc): Merge also linker command line -flto=foo argument
15892 2021-05-16 Christophe Lyon <christophe.lyon@linaro.org>
15894 * config/arm/arm.h (CPP_SPEC): Remove error message about
15895 -mlittle-endian/-mbig-endian conflict.
15897 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
15899 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
15900 __ROP_PROTECT__ if -mrop-protect is selected.
15902 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
15904 * config/rs6000/rs6000-internal.h (rs6000_stack): Add
15905 rop_hash_save_offset and rop_hash_size.
15906 * config/rs6000/rs6000-logue.c (rs6000_stack_info): Compute
15907 rop_hash_size and rop_hash_save_offset.
15908 (debug_stack_info): Dump rop_hash_save_offset and rop_hash_size.
15909 (rs6000_emit_prologue): Emit hashst[p] in prologue.
15910 (rs6000_emit_epilogue): Emit hashchk[p] in epilogue.
15911 * config/rs6000/rs6000.md (unspec): Add UNSPEC_HASHST and
15913 (hashst): New define_insn.
15914 (hashchk): Likewise.
15916 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
15918 * config/rs6000/rs6000.c (rs6000_option_override_internal):
15919 Disable shrink wrap when inserting ROP-protect instructions.
15920 * config/rs6000/rs6000.opt (mrop-protect): New option.
15921 (mprivileged): Likewise.
15922 * doc/invoke.texi: Document mrop-protect and mprivileged.
15924 2021-05-15 Hans-Peter Nilsson <hp@axis.com>
15926 * reorg.c (fill_slots_from_thread): Reinstate code typoed out in
15929 2021-05-15 Martin Jambor <mjambor@suse.cz>
15932 2021-05-13 Martin Jambor <mjambor@suse.cz>
15934 PR tree-optimization/100453
15935 * tree-sra.c (sra_modify_assign): All const base accesses do not
15936 need refreshing, not just those from decl_pool.
15937 (sra_modify_assign): Do not refresh into a const base decl.
15939 2021-05-15 Jakub Jelinek <jakub@redhat.com>
15941 PR rtl-optimization/100342
15942 * regcprop.c (copy_value): When copying a source reg in a wider
15943 mode than it has recorded for the value, adjust recorded destination
15944 mode too or punt if !REG_CAN_CHANGE_MODE_P.
15946 2021-05-14 Jason Merrill <jason@redhat.com>
15948 * intl.h: Add comments.
15950 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15952 * config/aarch64/aarch64-simd.md
15953 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Split into...
15954 (aarch64_sqdmlsl2_lane<mode>_internal): ... This...
15955 (aarch64_sqdmlal2_lane<mode>_internal): ... And this.
15956 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Split into ...
15957 (aarch64_sqdmlsl2_laneq<mode>_internal): ... This...
15958 (aarch64_sqdmlal2_laneq<mode>_internal): ... And this.
15959 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>_internal): Split into...
15960 (aarch64_sqdmlsl2_n<mode>_internal): ... This...
15961 (aarch64_sqdmlal2_n<mode>_internal): ... And this.
15963 2021-05-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15966 * config/arm/arm_neon.h (vtst_s8): Replace call to vtst builtin with it's
15967 boolean logic equivalent.
15968 (vtst_s16): Likewise.
15969 (vtst_s32): Likewise.
15970 (vtst_u8): Likewise.
15971 (vtst_u16): Likewise.
15972 (vtst_u32): Likewise.
15973 (vtst_p8): Likewise.
15974 (vtst_p16): Likewise.
15975 (vtstq_s8): Likewise.
15976 (vtstq_s16): Likewise.
15977 (vtstq_s32): Likewise.
15978 (vtstq_u8): Likewise.
15979 (vtstq_u16): Likewise.
15980 (vtstq_u32): Likewise.
15981 (vtstq_p8): Likewise.
15982 (vtstq_p16): Likewise.
15983 * config/arm/arm_neon_builtins.def: Remove entry for vtst.
15984 * config/arm/neon.md (neon_vtst<mode>): Remove pattern.
15986 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15988 * config/aarch64/aarch64-simd.md (aarch64_sqdmlal2<mode>): Merge into...
15989 (aarch64_sqdml<SBINQOPS:as>l2<mode>): ... This.
15990 (aarch64_sqdmlsl2<mode>): Delete.
15991 (aarch64_sqdmlal2_lane<mode>): Merge this...
15992 (aarch64_sqdmlsl2_lane<mode>): ... And this...
15993 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>): ... Into this.
15994 (aarch64_sqdmlal2_laneq<mode>): Merge this...
15995 (aarch64_sqdmlsl2_laneq<mode>): ... And this...
15996 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>): ... Into this.
15997 (aarch64_sqdmlal2_n<mode>): Merge this...
15998 (aarch64_sqdmlsl2_n<mode>): ... And this...
15999 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>): ... Into this.
16001 2021-05-13 Martin Sebor <msebor@redhat.com>
16003 PR middle-end/100574
16004 * builtins.c (access_ref::get_ref): Improve detection of PHIs with
16005 all null arguments.
16007 2021-05-13 Martin Sebor <msebor@redhat.com>
16009 PR tree-optimization/93100
16010 PR middle-end/98583
16011 * tree-ssa-uninit.c (check_defs): Exclude intrinsic functions that
16012 don't modify referenced objects.
16014 2021-05-13 Martin Jambor <mjambor@suse.cz>
16016 PR tree-optimization/100453
16017 * tree-sra.c (sra_modify_assign): All const base accesses do not
16018 need refreshing, not just those from decl_pool.
16019 (sra_modify_assign): Do not refresh into a const base decl.
16021 2021-05-13 Martin Liska <mliska@suse.cz>
16023 * tree-ssa-dom.c: Remove m_simplifier.
16025 2021-05-13 Richard Earnshaw <rearnsha@arm.com>
16028 * config/arm/arm.c (arm_canonicalize_comparison): Correctly
16029 canonicalize DImode inequality comparisons against the
16030 maximum integral value.
16032 2021-05-13 Jakub Jelinek <jakub@redhat.com>
16034 PR tree-optimization/98856
16035 * config/i386/i386.c (ix86_shift_rotate_cost): Add CODE argument.
16036 Expect V2DI and V4DI arithmetic right shifts to be emulated.
16037 (ix86_rtx_costs, ix86_add_stmt_cost): Adjust ix86_shift_rotate_cost
16039 * config/i386/i386-expand.c (expand_vec_perm_2perm_interleave,
16040 expand_vec_perm_2perm_pblendv): New functions.
16041 (ix86_expand_vec_perm_const_1): Use them.
16042 * config/i386/sse.md (ashr<mode>3<mask_name>): Rename to ...
16043 (<mask_codefor>ashr<mode>3<mask_name>): ... this.
16044 (ashr<mode>3): New define_expand with VI248_AVX512BW iterator.
16045 (ashrv4di3): New define_expand.
16046 (ashrv2di3): Change condition to TARGET_SSE2, handle !TARGET_XOP
16047 and !TARGET_AVX512VL expansion.
16049 2021-05-13 Uroš Bizjak <ubizjak@gmail.com>
16052 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Force mode
16053 sizes < 16 to a register when constructing vpcmov pattern.
16054 * config/i386/mmx.md (*xop_pcmov_<mode>): Use MMXMODE124 mode.
16056 2021-05-13 Martin Liska <mliska@suse.cz>
16058 * gcov-io.c (gcov_write_block): Remove.
16059 (gcov_write_words): Likewise.
16060 (gcov_read_words): Re-implement using gcov_read_bytes.
16061 (gcov_allocate): Remove.
16062 (GCOV_BLOCK_SIZE): Likewise.
16063 (struct gcov_var): Remove most of the fields.
16064 (gcov_position): Implement with ftell.
16065 (gcov_rewrite): Remove setting of start and offset fields.
16066 (from_file): Re-format.
16067 (gcov_open): Remove setbuf call. It should not be needed.
16068 (gcov_close): Remove internal buffer handling.
16069 (gcov_magic): Use __builtin_bswap32.
16070 (gcov_write_counter): Use directly gcov_write_unsigned.
16071 (gcov_write_string): Use direct fwrite and do not round
16073 (gcov_seek): Use directly fseek.
16074 (gcov_write_tag): Use gcov_write_unsigned directly.
16075 (gcov_write_length): Likewise.
16076 (gcov_write_tag_length): Likewise.
16077 (gcov_read_bytes): Use directly fread.
16078 (gcov_read_unsigned): Use gcov_read_words.
16079 (gcov_read_counter): Likewise.
16080 (gcov_read_string): Use gcov_read_bytes.
16081 * gcov-io.h (GCOV_WORD_SIZE): Adjust to reflect
16082 that size is not in bytes, but words (4B).
16083 (GCOV_TAG_FUNCTION_LENGTH): Likewise.
16084 (GCOV_TAG_ARCS_LENGTH): Likewise.
16085 (GCOV_TAG_ARCS_NUM): Likewise.
16086 (GCOV_TAG_COUNTER_LENGTH): Likewise.
16087 (GCOV_TAG_COUNTER_NUM): Likewise.
16088 (GCOV_TAG_SUMMARY_LENGTH): Likewise.
16090 2021-05-13 liuhongt <hongtao.liu@intel.com>
16093 * config/i386/sse.md (ssedoublevecmode): Add attribute for
16094 V64QI/V32HI/V16SI/V4DI.
16095 (ssehalfvecmode): Add attribute for V2DI/V2DF.
16096 (*vec_concatv4si_0): Extend to VI124_128.
16097 (*vec_concat<mode>_0): New pre-reload splitter.
16098 * config/i386/predicates.md (movq_parallel): New predicate.
16100 2021-05-13 Alexandre Oliva <oliva@adacore.com>
16102 * targhooks.c (default_zero_call_used_regs): Retry using
16103 successfully-zeroed registers as sources.
16105 2021-05-12 Tobias Burnus <tobias@codesourcery.com>
16107 * omp-low.c (finish_taskreg_scan): Use the proper detach decl.
16109 2021-05-12 Aldy Hernandez <aldyh@redhat.com>
16112 * gimple-range.cc (range_of_builtin_call): Skip out on
16113 processing __builtin_clz when varying.
16115 2021-05-12 Tom de Vries <tdevries@suse.de>
16118 * config/nvptx/nvptx-opts.h (enum ptx_version): New enum.
16119 * config/nvptx/nvptx.c (nvptx_file_start): Print .version according
16120 to ptx_version_option.
16121 * config/nvptx/nvptx.h (TARGET_PTX_6_3): Define.
16122 * config/nvptx/nvptx.md (define_insn "nvptx_shuffle<mode>")
16123 (define_insn "nvptx_vote_ballot"): Use sync variant for
16125 * config/nvptx/nvptx.opt (ptx_version): Add enum.
16126 (mptx): Add option.
16127 * doc/invoke.texi (Nvidia PTX Options): Add mptx item.
16129 2021-05-12 Richard Biener <rguenther@suse.de>
16131 PR tree-optimization/100566
16132 * tree-ssa-sccvn.c (dominated_by_p_w_unex): Properly handle
16133 allow_back for all edge queries.
16135 2021-05-12 liuhongt <hongtao.liu@intel.com>
16138 * config/i386/sse.md (<sse4_1_avx2>_pblendvb): Add
16139 splitters for pblendvb of NOT mask register.
16141 2021-05-12 Richard Biener <rguenther@suse.de>
16143 PR tree-optimization/100519
16144 * tree-ssa-reassoc.c (can_associate_p): Split into...
16145 (can_associate_op_p): ... this
16146 (can_associate_type_p): ... and this.
16147 (is_reassociable_op): Call can_associate_op_p.
16148 (break_up_subtract_bb): Call the appropriate predicates.
16149 (reassociate_bb): Likewise.
16151 2021-05-12 Martin Liska <mliska@suse.cz>
16153 * lto-wrapper.c (merge_and_complain): Merge -flto=arg options.
16154 (run_gcc): Use -flto argument detection for merged
16157 2021-05-12 Martin Liska <mliska@suse.cz>
16159 * lto-wrapper.c (print_lto_docs_link): New function.
16160 (run_gcc): Print warning about missing job server detection
16161 after we know NR of partitions. Do the same for -flto{,=1}.
16162 * opts.c (get_option_html_page): Support -flto option.
16164 2021-05-12 Martin Liska <mliska@suse.cz>
16166 * lto-wrapper.c (get_options_from_collect_gcc_options): Change
16168 (append_option): Remove.
16169 (find_option): Rework to use the vector type.
16170 (remove_option): Remove.
16171 (merge_and_complain): Use vectors for cl_decoded_option data
16173 (append_compiler_options): Likewise.
16174 (append_diag_options): Likewise.
16175 (append_linker_options): Likewise.
16176 (append_offload_options): Likewise.
16177 (compile_offload_image): Likewise.
16178 (compile_images_for_offload_targets): Likewise.
16179 (find_and_merge_options): Likewise.
16180 (run_gcc): Likewise.
16182 2021-05-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
16185 * dwarf2out.c (dwarf2out_finish): Set
16186 have_multiple_function_sections with multi-range text_section.
16188 2021-05-12 Martin Liska <mliska@suse.cz>
16190 PR bootstrap/100560
16191 * Makefile.in: Remove version.h from linker command line.
16193 2021-05-12 Richard Biener <rguenther@suse.de>
16195 PR middle-end/100547
16196 * rtl.h (rtvec_alloc): Make argument size_t.
16197 * rtl.c (rtvec_alloc): Verify the count is less than INT_MAX.
16199 2021-05-12 Jakub Jelinek <jakub@redhat.com>
16201 PR middle-end/100508
16202 * cfgexpand.c (expand_debug_expr): For DEBUG_EXPR_DECL with vector
16203 type, don't reuse DECL_RTL if it has different mode, instead force
16204 creation of a new DEBUG_EXPR.
16206 2021-05-12 Jakub Jelinek <jakub@redhat.com>
16207 Marc Glisse <marc.glisse@inria.fr>
16209 PR tree-optimization/94589
16210 * match.pd ((X & Y) == X -> (X & ~Y) == 0,
16211 (X | Y) == Y -> (X & ~Y) == 0): New GIMPLE simplifications.
16213 2021-05-12 Uroš Bizjak <ubizjak@gmail.com>
16216 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Handle V2SF mode.
16217 * config/i386/mmx.md (MMXMODE124): New mode iterator.
16219 (mmxintvecmode): New mode attribute.
16220 (mmxintvecmodelower): Ditto.
16221 (*mmx_maskcmpv2sf3_comm): New insn pattern.
16222 (*mmx_maskcmpv2sf3): Ditto.
16223 (vec_cmpv2sfv2si): New expander.
16224 (vcond<V2FI:mode>v2si): Ditto.
16225 (mmx_vlendvps): New insn pattern.
16226 (vcond<MMXMODE124:mode><MMXMODEI:mode>): Also handle V2SFmode.
16227 (vcondu<MMXMODE124:mode><MMXMODEI:mode>): Ditto.
16228 (vcond_mask_<mode><mmxintvecmodelower>): Ditto.
16230 2021-05-11 Martin Sebor <msebor@redhat.com>
16232 PR middle-end/21433
16233 * expr.c (expand_expr_real_1): Replace unreachable code with an assert.
16235 2021-05-11 Richard Biener <rguenther@suse.de>
16237 * gimple-fold.c (gimple_fold_call): Do not call
16238 maybe_fold_reference on call arguments or the static chain.
16239 (fold_stmt_1): Do not call maybe_fold_reference on GIMPLE_ASM
16242 2021-05-11 Martin Liska <mliska@suse.cz>
16244 * builtins.def (DEF_HSAIL_BUILTIN): Remove.
16245 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
16246 (DEF_HSAIL_SAT_BUILTIN): Likewise.
16247 (DEF_HSAIL_INTR_BUILTIN): Likewise.
16248 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
16249 * doc/frontends.texi: Remove BRIG.
16250 * doc/install.texi: Likewise.
16251 * doc/invoke.texi: Likewise.
16252 * doc/standards.texi: Likewise.
16253 * brig-builtins.def: Removed.
16254 * brig/ChangeLog: Removed.
16255 * brig/Make-lang.in: Removed.
16256 * brig/brig-builtins.h: Removed.
16257 * brig/brig-c.h: Removed.
16258 * brig/brig-lang.c: Removed.
16259 * brig/brigfrontend/brig-arg-block-handler.cc: Removed.
16260 * brig/brigfrontend/brig-atomic-inst-handler.cc: Removed.
16261 * brig/brigfrontend/brig-basic-inst-handler.cc: Removed.
16262 * brig/brigfrontend/brig-branch-inst-handler.cc: Removed.
16263 * brig/brigfrontend/brig-cmp-inst-handler.cc: Removed.
16264 * brig/brigfrontend/brig-code-entry-handler.cc: Removed.
16265 * brig/brigfrontend/brig-code-entry-handler.h: Removed.
16266 * brig/brigfrontend/brig-comment-handler.cc: Removed.
16267 * brig/brigfrontend/brig-control-handler.cc: Removed.
16268 * brig/brigfrontend/brig-copy-move-inst-handler.cc: Removed.
16269 * brig/brigfrontend/brig-cvt-inst-handler.cc: Removed.
16270 * brig/brigfrontend/brig-fbarrier-handler.cc: Removed.
16271 * brig/brigfrontend/brig-function-handler.cc: Removed.
16272 * brig/brigfrontend/brig-function.cc: Removed.
16273 * brig/brigfrontend/brig-function.h: Removed.
16274 * brig/brigfrontend/brig-inst-mod-handler.cc: Removed.
16275 * brig/brigfrontend/brig-label-handler.cc: Removed.
16276 * brig/brigfrontend/brig-lane-inst-handler.cc: Removed.
16277 * brig/brigfrontend/brig-machine.c: Removed.
16278 * brig/brigfrontend/brig-machine.h: Removed.
16279 * brig/brigfrontend/brig-mem-inst-handler.cc: Removed.
16280 * brig/brigfrontend/brig-module-handler.cc: Removed.
16281 * brig/brigfrontend/brig-queue-inst-handler.cc: Removed.
16282 * brig/brigfrontend/brig-seg-inst-handler.cc: Removed.
16283 * brig/brigfrontend/brig-signal-inst-handler.cc: Removed.
16284 * brig/brigfrontend/brig-to-generic.cc: Removed.
16285 * brig/brigfrontend/brig-to-generic.h: Removed.
16286 * brig/brigfrontend/brig-util.cc: Removed.
16287 * brig/brigfrontend/brig-util.h: Removed.
16288 * brig/brigfrontend/brig-variable-handler.cc: Removed.
16289 * brig/brigfrontend/hsa-brig-format.h: Removed.
16290 * brig/brigfrontend/phsa.h: Removed.
16291 * brig/brigspec.c: Removed.
16292 * brig/config-lang.in: Removed.
16293 * brig/gccbrig.texi: Removed.
16294 * brig/lang-specs.h: Removed.
16295 * brig/lang.opt: Removed.
16297 2021-05-11 Richard Biener <rguenther@suse.de>
16300 * ipa-param-manipulation.c
16301 (ipa_param_body_adjustments::modify_call_stmt): Avoid
16302 altering SSA_NAME_DEF_STMT by adjusting the calls LHS
16303 via gimple_call_lhs_ptr.
16305 2021-05-11 Alex Coplan <alex.coplan@arm.com>
16308 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear):
16309 Avoid emitting CFA adjusts on the sp if we have the fp.
16311 2021-05-11 Richard Sandiford <richard.sandiford@arm.com>
16313 * config/aarch64/iterators.md (VMUL_CHANGE_NLANES): Delete.
16314 (VMULD): New iterator.
16315 (VCOND): Handle V4HF and V8HF.
16316 (VCONQ): Fix entry for V2SF.
16317 * config/aarch64/aarch64-simd.md (mul_lane<mode>3): Use VMULD
16318 instead of VMUL. Use a 64-bit vector mode for the indexed operand.
16319 (*aarch64_mul3_elt_<vswap_width_name><mode>): Merge with...
16320 (mul_laneq<mode>3): ...this define_insn. Use VMUL instead of VDQSF.
16321 Use a 128-bit vector mode for the indexed operand. Use stype for
16322 the scheduling type.
16324 2021-05-11 Richard Biener <rguenther@suse.de>
16326 * gimple-fold.c (maybe_fold_reference): Only return
16327 is_gimple_min_invariant values.
16329 2021-05-11 Richard Biener <rguenther@suse.de>
16331 PR middle-end/100509
16332 * gimple-fold.c (fold_gimple_assign): Only call
16333 get_symbol_constant_value on register type symbols.
16335 2021-05-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
16336 Joe Ramsay <joe.ramsay@arm.com>
16339 * config/arm/arm_mve.h (__arm_vstrwq_scatter_offset): Fix wrong arguments.
16340 (__arm_vcmpneq): Remove duplicate definition.
16341 (__arm_vstrwq_scatter_offset_p): Likewise.
16342 (__arm_vmaxq_x): Likewise.
16343 (__arm_vmlsdavaq): Likewise.
16344 (__arm_vmlsdavaxq): Likewise.
16345 (__arm_vmlsdavq_p): Likewise.
16346 (__arm_vmlsdavxq_p): Likewise.
16347 (__arm_vrmlaldavhaq): Likewise.
16348 (__arm_vstrbq_p): Likewise.
16349 (__arm_vstrbq_scatter_offset): Likewise.
16350 (__arm_vstrbq_scatter_offset_p): Likewise.
16351 (__arm_vstrdq_scatter_offset): Likewise.
16352 (__arm_vstrdq_scatter_offset_p): Likewise.
16353 (__arm_vstrdq_scatter_shifted_offset): Likewise.
16354 (__arm_vstrdq_scatter_shifted_offset_p): Likewise.
16356 2021-05-11 Jakub Jelinek <jakub@redhat.com>
16358 PR middle-end/100471
16359 * omp-low.c (lower_omp_task_reductions): For OMP_TASKLOOP, if data
16360 is 0, bypass the reduction loop including
16361 GOMP_taskgroup_reduction_unregister call.
16363 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
16365 * config/rs6000/rs6000.c (struct rs6000_cost_data): New member
16366 costing_for_scalar.
16367 (rs6000_density_test): Early return if costing_for_scalar is true.
16368 (rs6000_init_cost): Init costing_for_scalar of rs6000_cost_data.
16370 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
16372 * doc/tm.texi: Regenerated.
16373 * target.def (init_cost): Add new parameter costing_for_scalar.
16374 * targhooks.c (default_init_cost): Adjust for new parameter.
16375 * targhooks.h (default_init_cost): Likewise.
16376 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
16377 (vect_compute_single_scalar_iteration_cost): Likewise.
16378 (vect_analyze_loop_2): Likewise.
16379 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
16380 (vect_bb_vectorization_profitable_p): Likewise.
16381 * tree-vectorizer.h (init_cost): Likewise.
16382 * config/aarch64/aarch64.c (aarch64_init_cost): Likewise.
16383 * config/i386/i386.c (ix86_init_cost): Likewise.
16384 * config/rs6000/rs6000.c (rs6000_init_cost): Likewise.
16386 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
16388 * config/rs6000/rs6000.c (rs6000_vect_nonmem): Renamed to
16389 vect_nonmem and moved into...
16390 (struct rs6000_cost_data): ...here.
16391 (rs6000_init_cost): Use vect_nonmem of cost_data instead.
16392 (rs6000_add_stmt_cost): Likewise.
16393 (rs6000_finish_cost): Likewise.
16395 2021-05-10 Eric Botcazou <ebotcazou@adacore.com>
16397 * range-op.cc (get_bool_state): Adjust head comment.
16398 (operator_not_equal::op1_range): Fix comment.
16399 (operator_bitwise_xor::op1_range): Remove call to gcc_unreachable.
16401 2021-05-10 Martin Sebor <msebor@redhat.com>
16403 PR middle-end/100425
16404 PR middle-end/100510
16405 * gimple-ssa-warn-alloca.c (pass_walloca::firast_time_p): Rename...
16406 (pass_walloca::xlimit_certain_p): ...to this.
16407 (pass_walloca::gate): Execute for any kind of handled warning.
16408 (pass_walloca::execute): Avoid issuing "maybe" and "unbounded"
16409 warnings when xlimit_certain_p is set.
16411 2021-05-10 Pat Haugen <pthaugen@linux.ibm.com>
16413 * config/rs6000/rs6000.c (rs6000_ira_change_pseudo_allocno_class):
16414 Return ALTIVEC_REGS if that is best_class.
16415 (rs6000_compute_pressure_classes): Add ALTIVEC_REGS.
16417 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16419 * config/arm/arm.h (CPP_SPEC): Remove error message about
16422 2021-05-10 Martin Jambor <mjambor@suse.cz>
16424 * ipa-prop.h (IPA_NODE_REF): Removed.
16425 (IPA_NODE_REF_GET_CREATE): Likewise.
16426 (IPA_EDGE_REF): Likewise.
16427 (IPA_EDGE_REF_GET_CREATE): Likewise.
16428 (IS_VALID_JUMP_FUNC_INDEX): Likewise.
16429 * ipa-cp.c (print_all_lattices): Replaced IPA_NODE_REF with a direct
16430 use of ipa_node_params_sum.
16431 (ipcp_versionable_function_p): Likewise.
16432 (push_node_to_stack): Likewise.
16433 (pop_node_from_stack): Likewise.
16434 (set_single_call_flag): Replaced two IPA_NODE_REF with one single
16435 direct use of ipa_node_params_sum.
16436 (initialize_node_lattices): Replaced IPA_NODE_REF with a direct use of
16437 ipa_node_params_sum.
16438 (ipa_context_from_jfunc): Replaced IPA_EDGE_REF with a direct use of
16440 (ipcp_verify_propagated_values): Replaced IPA_NODE_REF with a direct
16441 use of ipa_node_params_sum.
16442 (self_recursively_generated_p): Likewise.
16443 (propagate_scalar_across_jump_function): Likewise.
16444 (propagate_context_across_jump_function): Replaced IPA_EDGE_REF with a
16445 direct use of ipa_edge_args_sum, moved the lookup after the early
16446 exit. Replaced IPA_NODE_REF with a direct use of ipa_node_params_sum.
16447 (propagate_bits_across_jump_function): Replaced IPA_NODE_REF with
16448 direct uses of ipa_node_params_sum.
16449 (propagate_vr_across_jump_function): Likewise.
16450 (propagate_aggregate_lattice): Likewise.
16451 (propagate_aggs_across_jump_function): Likewise.
16452 (propagate_constants_across_call): Likewise, also replaced
16453 IPA_EDGE_REF with a direct use of ipa_edge_args_sum.
16454 (good_cloning_opportunity_p): Replaced IPA_NODE_REF with a direct use
16455 of ipa_node_params_sum.
16456 (estimate_local_effects): Likewise.
16457 (add_all_node_vals_to_toposort): Likewise.
16458 (propagate_constants_topo): Likewise.
16459 (ipcp_propagate_stage): Likewise.
16460 (ipcp_discover_new_direct_edges): Likewise.
16461 (calls_same_node_or_its_all_contexts_clone_p): Likewise.
16462 (cgraph_edge_brings_value_p): Likewise (in both overloaded functions).
16463 (get_info_about_necessary_edges): Likewise.
16464 (want_remove_some_param_p): Likewise.
16465 (create_specialized_node): Likewise.
16466 (self_recursive_pass_through_p): Likewise.
16467 (self_recursive_agg_pass_through_p): Likewise.
16468 (find_more_scalar_values_for_callers_subset): Likewise and also
16469 replaced IPA_EDGE_REF with direct uses of ipa_edge_args_sum, in one
16470 case replacing two of those with a single query.
16471 (find_more_contexts_for_caller_subset): Likewise for the
16472 ipa_polymorphic_call_context overload.
16473 (intersect_aggregates_with_edge): Replaced IPA_EDGE_REF with a direct
16474 use of ipa_edge_args_sum. Replaced IPA_NODE_REF with direct uses of
16475 ipa_node_params_sum.
16476 (find_aggregate_values_for_callers_subset): Likewise, also reusing
16477 results of ipa_edge_args_sum->get.
16478 (cgraph_edge_brings_all_scalars_for_node): Replaced IPA_NODE_REF with
16479 direct uses of ipa_node_params_sum, replaced IPA_EDGE_REF with a
16480 direct use of ipa_edge_args_sum.
16481 (cgraph_edge_brings_all_agg_vals_for_node): Likewise, moved node
16482 summary query after the early exit and reused the result later.
16483 (decide_about_value): Replaced IPA_NODE_REF with a direct use of
16484 ipa_node_params_sum.
16485 (decide_whether_version_node): Likewise. Removed re-querying for
16486 summaries after cloning.
16487 (spread_undeadness): Replaced IPA_NODE_REF with a direct use of
16488 ipa_node_params_sum.
16489 (has_undead_caller_from_outside_scc_p): Likewise, reusing results of
16491 (identify_dead_nodes): Likewise.
16492 (ipcp_store_bits_results): Replaced IPA_NODE_REF with direct uses of
16493 ipa_node_params_sum.
16494 (ipcp_store_vr_results): Likewise.
16495 * ipa-fnsummary.c (evaluate_properties_for_edge): Likewise.
16496 (ipa_fn_summary_t::duplicate): Likewise.
16497 (analyze_function_body): Likewise.
16498 (estimate_calls_size_and_time): Likewise.
16499 (ipa_cached_call_context::duplicate_from): Likewise.
16500 (ipa_call_context::equal_to): Likewise.
16501 (remap_edge_params): Likewise.
16502 (ipa_merge_fn_summary_after_inlining): Likewise.
16503 (inline_read_section): Likewise.
16504 * ipa-icf.c (sem_function::param_used_p): Likewise.
16505 * ipa-modref.c (compute_parm_map): Likewise.
16506 (compute_parm_map): Replaced IPA_EDGE_REF with a direct use of
16508 (get_access_for_fnspec): Replaced IPA_NODE_REF with a direct use of
16509 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
16511 * ipa-profile.c (check_argument_count): Likewise.
16512 * ipa-prop.c (ipa_alloc_node_params): Replaced IPA_NODE_REF_GET_CREATE
16513 with a direct use of ipa_node_params_sum.
16514 (ipa_initialize_node_params): Likewise.
16515 (ipa_print_node_jump_functions_for_edge): Replaced IPA_EDGE_REF with a
16516 direct use of ipa_edge_args_sum and reused the query result.
16517 (ipa_compute_jump_functions_for_edge): Replaced IPA_NODE_REF with a
16518 direct use of ipa_node_params_sum and replaced IPA_EDGE_REF with a
16519 direct use of ipa_edge_args_sum.
16520 (ipa_note_param_call): Replaced IPA_NODE_REF with a direct use of
16521 ipa_node_params_sum and reused the result of the query.
16522 (ipa_analyze_node): Likewise.
16523 (ipa_analyze_controlled_uses): Replaced IPA_NODE_REF with a direct use
16524 of ipa_node_params_sum.
16525 (update_jump_functions_after_inlining): Replaced IPA_EDGE_REF with
16526 direct uses of ipa_edge_args_sum.
16527 (update_indirect_edges_after_inlining): Replaced IPA_NODE_REF with
16528 direct uses of ipa_node_params_sum and replaced IPA_EDGE_REF with a
16529 direct use of ipa_edge_args_sum. Removed superficial re-querying the
16531 (propagate_controlled_uses): Replaced IPA_NODE_REF with direct uses of
16532 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
16534 (ipa_propagate_indirect_call_infos): Replaced IPA_EDGE_REF with a
16535 direct use of ipa_edge_args_sum.
16536 (ipa_edge_args_sum_t::duplicate): Replaced IPA_NODE_REF with a direct
16537 use of ipa_node_params_sum.
16538 (ipa_print_node_params): Likewise.
16539 (ipa_write_node_info): Likewise and also replaced IPA_EDGE_REF with
16540 direct uses of ipa_edge_args_sum.
16541 (ipa_read_edge_info): Replaced IPA_EDGE_REF with a direct use of
16543 (ipa_read_node_info): Replaced IPA_NODE_REF with a direct use of
16544 ipa_node_params_sum.
16545 (ipa_prop_write_jump_functions): Likewise. Move variable node to the
16546 scopes where it is used.
16548 2021-05-10 Uroš Bizjak <ubizjak@gmail.com>
16550 * config/i386/i386-expand.c (ix86_expand_sse_movcc)
16551 <case E_V2SImode>: Force op_true to register.
16553 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16555 * config/arm/iterators.md (MVE_FP_COMPARISONS): New.
16556 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_f<mode>)
16557 (mve_vcmp<mve_cmp_op>q_n_f<mode>): New, merge all vcmp_*f*
16559 (mve_vcmpeqq_f<mode>, mve_vcmpeqq_n_f<mode>, mve_vcmpgeq_f<mode>)
16560 (mve_vcmpgeq_n_f<mode>, mve_vcmpgtq_f<mode>)
16561 (mve_vcmpgtq_n_f<mode>, mve_vcmpleq_f<mode>)
16562 (mve_vcmpleq_n_f<mode>, mve_vcmpltq_f<mode>)
16563 (mve_vcmpltq_n_f<mode>, mve_vcmpneq_f<mode>)
16564 (mve_vcmpneq_n_f<mode>): Remove.
16565 * config/arm/unspecs.md (VCMPEQQ_F, VCMPEQQ_N_F, VCMPGEQ_F)
16566 (VCMPGEQ_N_F, VCMPGTQ_F, VCMPGTQ_N_F, VCMPLEQ_F, VCMPLEQ_N_F)
16567 (VCMPLTQ_F, VCMPLTQ_N_F, VCMPNEQ_F, VCMPNEQ_N_F): Remove.
16569 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16571 * config/arm/iterators.md (MVE_COMPARISONS): New.
16573 (mve_cmp_type): New.
16574 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_<mode>): New, merge all
16576 (mve_vcmpneq_<mode>, mve_vcmpcsq_n_<mode>, mve_vcmpcsq_<mode>)
16577 (mve_vcmpeqq_n_<mode>, mve_vcmpeqq_<mode>, mve_vcmpgeq_n_<mode>)
16578 (mve_vcmpgeq_<mode>, mve_vcmpgtq_n_<mode>, mve_vcmpgtq_<mode>)
16579 (mve_vcmphiq_n_<mode>, mve_vcmphiq_<mode>, mve_vcmpleq_n_<mode>)
16580 (mve_vcmpleq_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
16581 (mve_vcmpneq_n_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
16582 (mve_vcmpneq_n_<mode>): Remove.
16584 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16586 * config/arm/arm_mve.h (__arm_vcmp*): Remove 's' suffix.
16587 * config/arm/arm_mve_builtins.def (vcmp*): Remove 's' suffix.
16588 * config/arm/mve.md (mve_vcmp*): Remove 's' suffix in pattern
16591 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16593 * config/arm/arm_mve_builtins.def (vcmpneq_u): Remove.
16594 (vcmpneq_n_u): Likewise.
16595 (vcmpeqq_u,): Likewise.
16596 (vcmpeqq_n_u): Likewise.
16597 * config/arm/iterators.md (supf): Remove VCMPNEQ_U, VCMPEQQ_U,
16598 VCMPEQQ_N_U and VCMPNEQ_N_U.
16599 * config/arm/mve.md (mve_vcmpneq): Remove <supf> iteration.
16600 (mve_vcmpeqq_n): Likewise.
16601 (mve_vcmpeqq): Likewise.
16602 (mve_vcmpneq_n): Likewise.
16604 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
16606 * config/arm/arm_mve.h (__arm_vcmpeq*u*, __arm_vcmpne*u*): Call
16607 the 's' version of the builtin.
16609 2021-05-10 Richard Biener <rguenther@suse.de>
16611 PR tree-optimization/100492
16612 * tree-loop-distribution.c (find_seed_stmts_for_distribution):
16613 Find nothing when the loop contains an irreducible region.
16615 2021-05-10 Richard Biener <rguenther@suse.de>
16617 PR middle-end/100464
16619 * gimple-fold.c (canonicalize_constructor_val): Do not set
16622 2021-05-10 Richard Biener <rguenther@suse.de>
16624 PR tree-optimization/100434
16625 * tree-ssa-dse.c (initialize_ao_ref_for_dse): Handle
16627 (dse_optimize_stmt): Handle call LHS by dropping the
16628 LHS or the whole call if it doesn't have other
16630 (pass_dse::execute): Adjust.
16632 2021-05-10 Martin Liska <mliska@suse.cz>
16634 * Makefile.in: Add missing genversion rule.
16636 2021-05-10 Alex Coplan <alex.coplan@arm.com>
16639 * config/arm/mve.md (*mve_mov<mode>): Simplify output code. Use
16640 vldrw.u32 and vstrw.32 for V2D[IF]mode loads and stores.
16642 2021-05-10 Martin Liska <mliska@suse.cz>
16644 * builtins.c (is_builtin_name): Use startswith
16645 function instead of strncmp.
16646 * collect2.c (main): Likewise.
16647 (has_lto_section): Likewise.
16648 (scan_libraries): Likewise.
16649 * coverage.c (coverage_checksum_string): Likewise.
16650 (coverage_init): Likewise.
16651 * dwarf2out.c (is_cxx): Likewise.
16652 (gen_compile_unit_die): Likewise.
16653 * gcc-ar.c (main): Likewise.
16654 * gcc.c (init_spec): Likewise.
16655 (read_specs): Likewise.
16656 (execute): Likewise.
16657 (check_live_switch): Likewise.
16658 * genattrtab.c (write_attr_case): Likewise.
16659 (IS_ATTR_GROUP): Likewise.
16660 * gencfn-macros.c (main): Likewise.
16661 * gengtype.c (type_for_name): Likewise.
16662 (gen_rtx_next): Likewise.
16663 (get_file_langdir): Likewise.
16664 (write_local): Likewise.
16665 * genmatch.c (get_operator): Likewise.
16666 (get_operand_type): Likewise.
16667 (expr::gen_transform): Likewise.
16668 * genoutput.c (validate_optab_operands): Likewise.
16669 * incpath.c (add_sysroot_to_chain): Likewise.
16670 * langhooks.c (lang_GNU_C): Likewise.
16671 (lang_GNU_CXX): Likewise.
16672 (lang_GNU_Fortran): Likewise.
16673 (lang_GNU_OBJC): Likewise.
16674 * lto-wrapper.c (run_gcc): Likewise.
16675 * omp-general.c (omp_max_simt_vf): Likewise.
16676 * omp-low.c (omp_runtime_api_call): Likewise.
16677 * opts-common.c (parse_options_from_collect_gcc_options): Likewise.
16678 * read-rtl-function.c (function_reader::read_rtx_operand_r): Likewise.
16679 * real.c (real_from_string): Likewise.
16680 * selftest.c (assert_str_startswith): Likewise.
16681 * timevar.c (timer::validate_phases): Likewise.
16682 * tree.c (get_file_function_name): Likewise.
16683 * ubsan.c (ubsan_use_new_style_p): Likewise.
16684 * varasm.c (default_function_rodata_section): Likewise.
16685 (incorporeal_function_p): Likewise.
16686 (default_section_type_flags): Likewise.
16687 * system.h (startswith): Define startswith.
16689 2021-05-10 Martin Liska <mliska@suse.cz>
16691 * bitmap.h (class auto_bitmap): Remove
16692 __cplusplus >= 201103.
16693 * config/aarch64/aarch64.c: Likewise.
16694 * gimple-ssa-store-merging.c (store_immediate_info::store_immediate_info):
16696 * sbitmap.h: Likewise.
16698 2021-05-10 Martin Liska <mliska@suse.cz>
16700 * Makefile.in: Rename gcov-iov to genversion and depend
16701 on version.h (instead of gcov-iov.h).
16702 * gcov-io.h: Include version.h instread of gcov-iov.h.
16703 * gengtype-state.c (read_state_version): Likewise.
16704 * gcov-iov.c: Moved to...
16705 * genversion.c: ...here.
16706 * lto-streamer.h (LTO_major_version): Define it with
16708 * version.c: Removed.
16709 * version.h: Removed.
16711 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16713 * config/arc/arc.md (UNSPEC_ARC_DMPYWH): Define.
16714 * config/arc/simdext.md (VCT): Add predicates for iterator
16717 (voptab): Likewise.
16718 (vec_widen_<V_US>mult_hi_v4hi): Change pattern predicate.
16719 (<voptab>v2si3): New patterns.
16721 (reduc_plus_scal_v4hi): Likewise.
16722 (reduc_plus_scal_v2si): Likewise.
16723 (vec_duplicatev2si): Likewise.
16724 (vec_duplicatev4hi): Likewise.
16726 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16728 * config/arc/simdext.md: Format and cleanup file.
16730 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16732 * config/arc/simdext.md (movmisalignv2hi): Allow misaligned access
16733 only when munaligned-access option is on.
16734 (movmisalign<mode>): Likewise.
16736 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16738 * common/config/arc/arc-common.c (arc_handle_option): Remove dot
16740 * config/arc/arc.c (arc_reorg): Remove underscore from string.
16742 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16744 * config/arc/arc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
16745 (CTZ_DEFINED_VALUE_AT_ZERO): Likewise.
16746 * config/arc/arc.md (clrsbsi2): Cleanup pattern.
16747 (norm_f): Likewise.
16750 (clzsi2): Use fls instruction when available.
16751 (arc_clzsi2): Likewise.
16753 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16755 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add r26 and r27.
16757 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
16759 * doc/extend.texi (__builtin_arc_sr): Swap arguments.
16761 2021-05-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
16763 PR middle-end/100467
16764 * toplev.c (compile_file): Call insn_locations_init before
16765 targetm.asm_out.code_end.
16767 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
16770 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
16772 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
16774 2021-05-07 Jakub Jelinek <jakub@redhat.com>
16775 Andrew Stubbs <amd@codesourcery.com>
16778 * builtins.c (try_store_by_multiple_pieces): Use force_operand for
16779 emit_move_insn operands.
16781 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
16783 * cfgexpand.c (expand_gimple_basic_block): Do not inherit a current
16784 location for the outgoing edges of an empty block.
16785 * dwarf2out.c (add_subscript_info): Retrieve the bounds and index
16786 type by means of the get_array_descr_info langhook, if it is set and
16787 returns true. Remove obsolete code dealing with unnamed subtypes.
16789 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16791 * gimple-range-cache.cc (ssa_block_ranges): Virtualize.
16792 (sbr_vector): Renamed from ssa_block_cache.
16793 (sbr_vector::sbr_vector): Allocate from obstack abd initialize.
16794 (ssa_block_ranges::~ssa_block_ranges): Remove.
16795 (sbr_vector::set_bb_range): Use varying and undefined cached values.
16796 (ssa_block_ranges::set_bb_varying): Remove.
16797 (sbr_vector::get_bb_range): Adjust assert.
16798 (sbr_vector::bb_range_p): Adjust assert.
16799 (~block_range_cache): No freeing loop required.
16800 (block_range_cache::get_block_ranges): Remove.
16801 (block_range_cache::set_bb_range): Inline get_block_ranges.
16802 (block_range_cache::set_bb_varying): Remove.
16803 * gimple-range-cache.h (set_bb_varying): Remove prototype.
16804 * value-range.h (irange_allocator::get_memory): New.
16806 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16808 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Search
16809 dominator tree is available and requested.
16810 (ranger_cache::ssa_range_in_bb): Don't search dom tree here.
16811 (ranger_cache::fill_block_cache): Don't search dom tree here either.
16812 * gimple-range-cache.h (non_null_deref_p): Add dom_search param.
16814 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16816 * gimple-range.cc (gimple_ranger::range_on_exit): Handle block with
16817 only PHI nodes better.
16819 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16821 * gimple-range-edge.h (gimple_outgoing_range): Rename from
16823 (gcond_edge_range): Export prototype.
16824 * gimple-range-edge.cc (gcond_edge_range): New.
16825 (gimple_outgoing_range::edge_range_p): Use gcond_edge_range.
16826 * gimple-range-gori.h (gori_compute): Use gimple_outgoing_range.
16828 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16830 * gimple-range-edge.cc (outgoing_range::calc_switch_ranges): Compute
16831 default range into a temp and allocate only what is needed.
16833 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16835 * range-op.cc (operator_trunc_mod::wi_fold): x % 0 is UNDEFINED.
16837 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
16839 * gimple-range.h (gimple_range_global): Pick up parameter initial
16840 values, and use-before defined locals are UNDEFINED.
16842 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
16844 * doc/extend.texi (scalar_storage_order): Mention effect on pointer
16846 * tree.h (reverse_storage_order_for_component_p): Return false if
16847 the type is a pointer.
16849 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
16851 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
16853 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
16856 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
16857 Handle V8QI, V4HI and V2SI modes.
16858 * config/i386/mmx.md (mmx_pblendvb): New insn pattern.
16859 * config/i386/sse.md (unspec): Move UNSPEC_BLENDV ...
16860 * config/i386/i386.md (unspec): ... here.
16862 2021-05-07 Tobias Burnus <tobias@codesourcery.com>
16863 Tom de Vries <tdevries@suse.de>
16865 * omp-low.c (lower_rec_simd_input_clauses): Set max_vf = 1 if
16866 a truth_value_p reduction variable is nonintegral.
16868 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
16871 * config/i386/i386-expand.c (ix86_use_mask_cmp_p):
16872 Return false for mode sizes < 16.
16874 2021-05-07 Jakub Jelinek <jakub@redhat.com>
16877 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn.
16879 2021-05-06 Martin Jambor <mjambor@suse.cz>
16881 * ipa-sra.c (ipa_sra_dump_all_summaries): Dump edge summaries even
16882 when there is no function summary.
16883 (ipa_sra_summarize_function): produce edge summaries even when
16886 2021-05-06 Tom Tromey <tom@tromey.com>
16888 * godump.c (string_hash_eq): Remove.
16889 (go_finish): Use htab_eq_string.
16891 2021-05-06 Tom Tromey <tom@tromey.com>
16893 * gengtype-state.c (read_state): Use htab_eq_string.
16894 (string_eq): Remove.
16896 2021-05-06 Tom Tromey <tom@tromey.com>
16898 * gensupport.c (htab_eq_string): Remove.
16900 2021-05-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
16903 * debug.h (gcc_debug_hooks): Add set_ignored_loc function pointer.
16904 * dwarf2out.h (dw_fde_node::ignored_debug): New data item.
16905 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Add dummy
16906 set_ignored_loc callbacks.
16907 * debug.c (do_nothing_debug_hooks): Likewise.
16908 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
16909 * dwarf2out.c (text_section_used, cold_text_section_used): Remove.
16910 (in_text_section_p, last_text_label, last_cold_label,
16911 switch_text_ranges, switch_cold_ranges): New data items.
16912 (dwarf2out_note_section_used): Remove.
16913 (dwarf2out_begin_prologue): Set fde->ignored_debug and
16915 (mark_ignored_debug_section): New helper function.
16916 (dwarf2out_end_epilogue, dwarf2out_switch_text_section): Call
16917 mark_ignored_debug_section.
16918 (dwarf2_debug_hooks): Use dwarf2out_set_ignored_loc.
16919 (dwarf2_lineno_debug_hooks): Use dummy for set_ignored_loc.
16920 (size_of_aranges): Adjust formula for multi-part text ranges size.
16921 (output_aranges): Output multi-part text ranges.
16922 (dwarf2out_set_ignored_loc): New callback function.
16923 (dwarf2out_finish): Output multi-part text ranges.
16924 (dwarf2out_c_finalize): Clear new data items.
16925 * final.c (final_start_function_1): Call set_ignored_loc callback.
16926 (final_scan_insn_1): Likewise.
16927 * ggc-page.c (gt_ggc_mx): New helper function.
16928 * stringpool.c (gt_pch_nx): Likewise.
16930 2021-05-06 Richard Biener <rguenther@suse.de>
16932 * timevar.def (TV_TREE_INSERT_PHI_NODES): Remove.
16933 (TV_TREE_SSA_REWRITE_BLOCKS): Likewise.
16934 (TV_TREE_INTO_SSA): New.
16935 * tree-into-ssa.c (insert_phi_nodes): Do not account separately.
16936 (rewrite_blocks): Likewise.
16937 (pass_data_build_ssa): Account to TV_TREE_INTO_SSA.
16939 2021-05-06 Jakub Jelinek <jakub@redhat.com>
16941 * tree-ssa-phiopt.c (value_replacement, minmax_replacement,
16942 abs_replacement, xor_replacement,
16943 cond_removal_in_popcount_clz_ctz_pattern,
16944 replace_phi_edge_with_variable): Change type of phi argument from
16945 gimple * to gphi *.
16947 2021-05-06 Richard Biener <rguenther@suse.de>
16949 * tree-ssa-loop-split.c (split_loop): Delay updating SSA form.
16950 Output an opt-info message.
16951 (do_split_loop_on_cond): Likewise.
16952 (tree_ssa_split_loops): Update SSA form here.
16954 2021-05-06 Richard Biener <rguenther@suse.de>
16956 * tree-inline.c (tree_function_versioning): Fix DECL_BY_REFERENCE
16957 return variable removal.
16959 2021-05-06 Marius Hillenbrand <mhillen@linux.ibm.com>
16961 * config/s390/s390-builtins.def (O_M5, O1_M5, ...): Remove unused macros.
16962 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
16963 (s390_vec_permi_dbl, s390_vpdi): Use the O3_U2 type for the immediate
16965 * config/s390/s390.c (s390_const_operand_ok): Remove unused
16968 2021-05-06 Jakub Jelinek <jakub@redhat.com>
16970 PR tree-optimization/94589
16971 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Call
16972 spaceship_replacement.
16973 (cond_only_block_p, spaceship_replacement): New functions.
16975 2021-05-06 Richard Biener <rguenther@suse.de>
16978 * tree-emutls.c (gen_emutls_addr): Pass in whether we're
16979 dealing with a debug use and only query existing addresses
16981 (lower_emutls_1): Avoid splitting out addresses for debug
16982 stmts, reset the debug stmt when we fail to find existing
16984 (lower_emutls_phi_arg): Set wi.stmt.
16986 2021-05-06 Christoph Muellner <cmuellner@gcc.gnu.org>
16989 * config/riscv/riscv.c (riscv_block_move_loop): Use cbranch helper.
16990 * config/riscv/riscv.md (cbranch<mode>4): Generate helpers.
16991 (stack_protect_test): Use cbranch helper.
16993 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
16996 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
16997 always return the establisher frame for __builtin_frame_address (0).
16999 2021-05-05 Ivan Sorokin <vanyacpp@gmail.com>
17002 * config/i386/i386-builtins.c (ix86_cpu_model_type_node): New.
17003 (ix86_cpu_model_var): Likewise.
17004 (ix86_cpu_features2_type_node): Likewise.
17005 (ix86_cpu_features2_var): Likewise.
17006 (fold_builtin_cpu): Cache __cpu_model and __cpu_features2 with
17009 2021-05-05 Martin Sebor <msebor@redhat.com>
17011 * passes.def (pass_warn_printf): Run after SSA.
17013 2021-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17015 * config/arm/neon.md (neon_vtst_combine<mode>): New pattern.
17016 * config/arm/predicates.md (minus_one_operand): New predicate.
17018 2021-05-05 Jeff Law <jlaw@tachyum.com>
17020 * config/avr/avr.md: Remove references to CC_STATUS_INIT.
17022 2021-05-05 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
17024 PR rtl-optimization/100263
17025 * postreload.c (move2add_valid_value_p): Ensure register can
17028 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
17030 PR rtl-optimization/100411
17031 * cfgcleanup.c (try_crossjump_to_edge): Also skip end of prologue
17032 and beginning of function markers.
17034 2021-05-05 Jeff Law <jlaw@tachyum.com>
17036 * config/cr16/cr16.h (NOTICE_UPDATE_CC): Remove.
17037 * config/cr16/cr16.c (notice_update_cc): Remove.
17038 * config/cr16/cr16-protos.h (notice_update_cc): Remove.
17040 2021-05-05 Uroš Bizjak <ubizjak@gmail.com>
17043 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
17044 Handle V8QI, V4HI and V2SI modes.
17045 * config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
17046 (ix86_build_signbit_mask): Ditto.
17047 * config/i386/mmx.md (MMXMODE14): New mode iterator.
17048 (<smaxmin:code><MMXMODE14:mode>3): New expander.
17049 (*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
17050 (<umaxmin:code><MMXMODE24:mode>3): New expander.
17051 (*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
17052 (vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
17053 (vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
17054 (vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
17055 (vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
17056 (vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
17058 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
17060 * dwarf2out.c (loc_list_from_tree_1) <DECL>: During early DWARF, do
17061 not expand the VALUE_EXPR of variables put in the non-local frame.
17062 * gimplify.c (gimplify_type_sizes) <RECORD_TYPE>: If the type is not
17063 to be ignored for debug info, ensure its variable offsets are not.
17065 2021-05-05 Richard Biener <rguenther@suse.de>
17067 PR tree-optimization/79333
17068 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
17069 Fold stmt following SSA edges.
17071 2021-05-05 Richard Biener <rguenther@suse.de>
17073 PR middle-end/100394
17074 * calls.c (expand_call): Preserve possibly throwing calls.
17075 * cfgexpand.c (expand_call_stmt): When a call can throw signal
17076 RTL expansion there are side-effects.
17077 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Simplify,
17078 mark all possibly throwing stmts necessary unless we can elide
17080 * tree-ssa-dse.c (pass_dse::execute): Preserve exceptions unless
17081 -fdelete-dead-exceptions.
17082 * tree.h (DECL_PURE_P): Add note about exceptions.
17084 2021-05-05 Alexandre Oliva <oliva@adacore.com>
17086 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Make it
17089 2021-05-04 David Edelsohn <dje.gcc@gmail.com>
17091 * config/rs6000/rs6000-call.c (rs6000_output_mi_thunk): Use
17092 get_fnname_from_decl for name of thunk.
17093 * config/rs6000/rs6000.c (rs6000_declare_alias): Use assemble_name
17094 and ASM_OUTPUT_LABEL.
17095 (rs6000_xcoff_declare_function_name): Use assemble_name and
17097 (rs6000_xcoff_declare_object_name): Use ASM_OUTPUT_LABEL.
17098 (rs6000_xcoff_encode_section_info): Don't add mapping class
17099 for aliases. Always add [DS] mapping class to primary
17101 (rs6000_asm_weaken_decl): Don't explicitly add [DS].
17103 2021-05-04 Martin Sebor <msebor@redhat.com>
17105 PR middle-end/100307
17106 * builtins.c (compute_objsize_r): Clear base0 for pointers.
17108 2021-05-04 Jeff Law <jlaw@tachyum.com>
17110 * config/bfin/bfin.h (NOTICE_UPDATE_CC): Remove.
17112 2021-05-04 Segher Boessenkool <segher@kernel.crashing.org>
17114 * caller-save.c: Remove CC0.
17115 * cfgcleanup.c: Remove CC0.
17116 * cfgrtl.c: Remove CC0.
17117 * combine.c: Remove CC0.
17118 * compare-elim.c: Remove CC0.
17119 * conditions.h: Remove CC0.
17120 * config/h8300/h8300.h: Remove CC0.
17121 * config/h8300/h8300-protos.h: Remove CC0.
17122 * config/h8300/peepholes.md: Remove CC0.
17123 * config/i386/x86-tune-sched.c: Remove CC0.
17124 * config/m68k/m68k.c: Remove CC0.
17125 * config/rl78/rl78.c: Remove CC0.
17126 * config/sparc/sparc.c: Remove CC0.
17127 * config/xtensa/xtensa.c: Remove CC0.
17128 (gen_conditional_move): Use pc_rtx instead of cc0_rtx in a piece of
17129 RTL where that is used as a placeholder only.
17130 * cprop.c: Remove CC0.
17131 * cse.c: Remove CC0.
17132 * cselib.c: Remove CC0.
17133 * df-problems.c: Remove CC0.
17134 * df-scan.c: Remove CC0.
17135 * doc/md.texi: Remove CC0. Adjust an example.
17136 * doc/rtl.texi: Remove CC0. Adjust an example.
17137 * doc/tm.texi: Regenerate.
17138 * doc/tm.texi.in: Remove CC0.
17139 * emit-rtl.c: Remove CC0.
17140 * final.c: Remove CC0.
17141 * fwprop.c: Remove CC0.
17142 * gcse-common.c: Remove CC0.
17143 * gcse.c: Remove CC0.
17144 * genattrtab.c: Remove CC0.
17145 * genconfig.c: Remove CC0.
17146 * genemit.c: Remove CC0.
17147 * genextract.c: Remove CC0.
17148 * gengenrtl.c: Remove CC0.
17149 * genrecog.c: Remove CC0.
17150 * haifa-sched.c: Remove CC0.
17151 * ifcvt.c: Remove CC0.
17152 * ira-costs.c: Remove CC0.
17153 * ira.c: Remove CC0.
17154 * jump.c: Remove CC0.
17155 * loop-invariant.c: Remove CC0.
17156 * lra-constraints.c: Remove CC0.
17157 * lra-eliminations.c: Remove CC0.
17158 * optabs.c: Remove CC0.
17159 * postreload-gcse.c: Remove CC0.
17160 * postreload.c: Remove CC0.
17161 * print-rtl.c: Remove CC0.
17162 * read-rtl-function.c: Remove CC0.
17163 * reg-notes.def: Remove CC0.
17164 * reg-stack.c: Remove CC0.
17165 * reginfo.c: Remove CC0.
17166 * regrename.c: Remove CC0.
17167 * reload.c: Remove CC0.
17168 * reload1.c: Remove CC0.
17169 * reorg.c: Remove CC0.
17170 * resource.c: Remove CC0.
17171 * rtl.c: Remove CC0.
17172 * rtl.def: Remove CC0.
17173 * rtl.h: Remove CC0.
17174 * rtlanal.c: Remove CC0.
17175 * sched-deps.c: Remove CC0.
17176 * sched-rgn.c: Remove CC0.
17177 * shrink-wrap.c: Remove CC0.
17178 * simplify-rtx.c: Remove CC0.
17179 * system.h: Remove CC0. Poison NOTICE_UPDATE_CC, CC_STATUS_MDEP_INIT,
17180 CC_STATUS_MDEP, and CC_STATUS.
17181 * target.def: Remove CC0.
17182 * valtrack.c: Remove CC0.
17183 * var-tracking.c: Remove CC0.
17185 2021-05-04 Richard Biener <rguenther@suse.de>
17187 PR tree-optimization/100414
17188 * tree-ssa-phiopt.c (get_non_trapping): Do not compute dominance
17190 (tree_ssa_phiopt_worker): But unconditionally here.
17192 2021-05-04 Tobias Burnus <tobias@codesourcery.com>
17194 * omp-low.c (lower_rec_input_clauses, lower_reduction_clauses): Handle
17195 && and || with floating-point and complex arguments.
17197 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
17199 * tree-inline.c (insert_debug_decl_map): Delete.
17200 (copy_debug_stmt): Minor tweak.
17201 (setup_one_parameter): Do not use a variable if the value is either
17202 a read-only DECL or a non-addressable local variable in the caller.
17203 In this case, insert the debug-only variable in the map manually.
17204 (expand_call_inline): Do not generate a CLOBBER for these values.
17205 * tree-inline.h (debug_map): Minor tweak.
17207 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
17209 * builtins.c (builtin_with_linkage_p): Return true for stp[n]cpy.
17210 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Tidy up.
17212 2021-05-04 Richard Biener <rguenther@suse.de>
17214 PR tree-optimization/100329
17215 * tree-ssa-reassoc.c (can_reassociate_p): Do not reassociate
17217 (insert_stmt_after): Assert we're not running into asm goto.
17219 2021-05-04 Richard Biener <rguenther@suse.de>
17221 PR tree-optimization/100398
17222 * tree-ssa-dse.c (pass_dse::execute): Preserve control
17225 2021-05-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17227 * builtins.c (try_store_by_multiple_pieces): Fix constfun's prototype.
17229 2021-05-04 Alexandre Oliva <oliva@adacore.com>
17231 * builtins.c (try_store_by_multiple_pieces): New.
17232 (expand_builtin_memset_args): Use it. If target_char_cast
17233 fails, proceed as for non-constant val. Pass len's ctz to...
17234 * expr.c (clear_storage_hints): ... this. Try store by
17235 multiple pieces after setmem.
17236 (clear_storage): Adjust.
17237 * expr.h (clear_storage_hints): Likewise.
17238 (try_store_by_multiple_pieces): Declare.
17239 * passes.def: Replace the last copy_prop with ccp.
17241 2021-05-03 Tom de Vries <tdevries@suse.de>
17244 * omp-low.c (lower_rec_input_clauses): Disable SIMT for user-defined
17247 2021-05-03 Richard Biener <rguenther@suse.de>
17249 * tree-ssa-dse.c (dse_classify_store): Track two PHI defs.
17251 2021-05-03 Richard Biener <rguenther@suse.de>
17253 * tree-ssa-dse.c: Do not include domwalk.h but cfganal.h.
17254 (dse_dom_walker): Remove.
17255 (dse_dom_walker::dse_optimize_stmt): Rename...
17256 (dse_optimize_stmt): ... to this, pass in live_bytes sbitmap.
17257 (dse_dom_walker::before_dom_children): Inline ...
17258 (pass_dse::execute): ... here. Perform a reverse program
17261 2021-05-03 H.J. Lu <hjl.tools@gmail.com>
17264 * configure: Regenerated.
17266 2021-05-03 Ilya Leoshkevich <iii@linux.ibm.com>
17269 * config/s390/s390.c (s390_hard_fp_reg_p): New function.
17270 (s390_md_asm_adjust): Handle hard registers.
17272 2021-05-03 Jakub Jelinek <jakub@redhat.com>
17274 PR tree-optimization/100382
17275 * tree-ssa-dse.c: Include tree-eh.h.
17276 (dse_dom_walker::before_dom_children): Don't remove stmts if
17277 stmt_unremovable_because_of_non_call_eh_p is true.
17279 2021-05-02 David Edelsohn <dje.gcc@gmail.com>
17281 * varasm.c (compute_reloc_for_var): Split out from...
17282 (get_variable_section): Use it.
17283 * output.h (compute_reloc_for_var): Declare.
17284 * config/rs6000/rs6000-protos.h
17285 (rs6000_xcoff_asm_output_aligned_decl_common): Change alignment to
17287 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address_aix):
17288 Don't append storage mapping class to symbol.
17289 (rs6000_xcoff_asm_named_section): Add BS and UL mapping classes.
17290 Don't convert TLS BSS to common.
17291 (rs6000_xcoff_unique_section): Don't fall back to select_secton.
17292 (rs6000_xcoff_section_type_flags): Add SECTION_BSS if DECL is
17294 (rs6000_xcoff_asm_globalize_decl_name): Don't strip storage
17296 (rs6000_xcoff_asm_output_aligned_decl_common): Align is unsigned int.
17297 If align is 0 from TLS class, use the same rules as varasm.c
17298 If not common, switch to BSS section manually.
17299 If common, emit appropriate comm or lcomm directive.
17300 (rs6000_xcoff_encode_section_info): Add logic to append all
17301 storage mapping classes.
17302 (rs6000_asm_weaken_decl): Adjust for qualname symbols.
17303 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
17304 rs6000_xcoff_asm_output_aligned_decl_common.
17305 (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
17306 rs6000_xcoff_asm_output_aligned_decl_common.
17307 (ASM_OUTPUT_TLS_COMMON): Use
17308 rs6000_xcoff_asm_output_aligned_decl_common.
17310 2021-05-02 Jakub Jelinek <jakub@redhat.com>
17313 * config/nvptx/nvptx.c (nvptx_sese_pseudo): Use nullptr instead of 0
17314 as first argument of pseudo_node_t constructors.
17316 2021-05-02 Jakub Jelinek <jakub@redhat.com>
17319 * config/i386/t-i386 (TM_H): Add $(srcdir)/config/i386/i386-isa.def.
17321 2021-05-01 Aldy Hernandez <aldyh@redhat.com>
17323 * value-range.cc (DEFINE_INT_RANGE_GC_STUBS): Remove.
17324 (gt_pch_nx (int_range<1> *&)): New.
17325 (gt_ggc_mx (int_range<1> *&)): New.
17326 * value-range.h (class irange): Add GTY support for
17329 2021-05-01 Geng Qi <gengqi@linux.alibaba.com>
17331 * doc/options.texi (Negative): Change either or to both and.
17333 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17335 * config/aarch64/aarch64-simd-builtins.def: Add
17336 float_ml[as][q]_laneq builtin generator macros.
17337 * config/aarch64/aarch64-simd.md (mul_laneq<mode>3): Define.
17338 (aarch64_float_mla_laneq<mode>): Define.
17339 (aarch64_float_mls_laneq<mode>): Define.
17340 * config/aarch64/arm_neon.h (vmla_laneq_f32): Use RTL builtin
17341 instead of GCC vector extensions.
17342 (vmlaq_laneq_f32): Likewise.
17343 (vmls_laneq_f32): Likewise.
17344 (vmlsq_laneq_f32): Likewise.
17346 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17348 * config/aarch64/aarch64-simd-builtins.def: Add
17349 float_ml[as]_lane builtin generator macros.
17350 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt<mode>):
17352 (mul_lane<mode>3): This, and re-order arguments.
17353 (aarch64_float_mla_lane<mode>): Define.
17354 (aarch64_float_mls_lane<mode>): Define.
17355 * config/aarch64/arm_neon.h (vmla_lane_f32): Use RTL builtin
17356 instead of GCC vector extensions.
17357 (vmlaq_lane_f32): Likewise.
17358 (vmls_lane_f32): Likewise.
17359 (vmlsq_lane_f32): Likewise.
17361 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17363 * config/aarch64/aarch64-simd-builtins.def: Add float_ml[as]
17364 builtin generator macros.
17365 * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>):
17367 (aarch64_float_mls<mode>): Define.
17368 * config/aarch64/arm_neon.h (vmla_f32): Use RTL builtin
17369 instead of relying on GCC vector extensions.
17370 (vmla_f64): Likewise.
17371 (vmlaq_f32): Likewise.
17372 (vmlaq_f64): Likewise.
17373 (vmls_f32): Likewise.
17374 (vmls_f64): Likewise.
17375 (vmlsq_f32): Likewise.
17376 (vmlsq_f64): Likewise.
17377 * config/aarch64/iterators.md: Define VDQF_DF mode iterator.
17379 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
17381 * config/aarch64/aarch64-simd-builtins.def: Add
17382 float_ml[as]_n_builtin generator macros.
17383 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt_from_dup<mode>):
17385 (mul_n<mode>3): This, and re-order arguments.
17386 (aarch64_float_mla_n<mode>): Define.
17387 (aarch64_float_mls_n<mode>): Define.
17388 * config/aarch64/arm_neon.h (vmla_n_f32): Use RTL builtin
17389 instead of inline asm.
17390 (vmlaq_n_f32): Likewise.
17391 (vmls_n_f32): Likewise.
17392 (vmlsq_n_f32): Likewise.
17394 2021-04-30 Jonathan Wright <joanthan.wright@arm.com>
17396 * config/aarch64/aarch64-simd-builtins.def: Add pmull[2]
17397 builtin generator macros.
17398 * config/aarch64/aarch64-simd.md (aarch64_pmullv8qi): Define.
17399 (aarch64_pmull_hiv16qi_insn): Define.
17400 (aarch64_pmull_hiv16qi): Define.
17401 * config/aarch64/arm_neon.h (vmull_high_p8): Use RTL builtin
17402 instead of inline asm.
17403 (vmull_p8): Likewise.
17405 2021-04-30 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
17407 * config/avr/avr.md: Adjust peepholes to match and
17408 generate parallels with clobber of REG_CC.
17409 (mov<mode>_insn): Rename to mov<mode>_insn_split.
17410 (*mov<mode>_insn): Rename to mov<mode>_insn.
17412 2021-04-30 David Edelsohn <dje.gcc@gmail.com>
17414 * varasm.c (use_blocks_for_decl_p): Don't use section anchors
17415 for VAR_DECLs if -fdata-sections enabled.
17417 2021-04-30 Michael Meissner <meissner@linux.ibm.com>
17419 PR bootstrap/100327
17420 * config/rs6000/rs6000.c
17421 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Define.
17422 (rs6000_libgcc_floating_mode_supported_p): New target hook.
17424 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
17426 * tree-ssa-threadbackward.c (class thread_jumps): Split out code
17428 (class back_threader_registry): ...to here...
17429 (class back_threader_profitability): ...and here...
17430 (thread_jumps::thread_through_all_blocks): Remove argument.
17431 (back_threader_registry::back_threader_registry): New.
17432 (back_threader_registry::~back_threader_registry): New.
17433 (back_threader_registry::thread_through_all_blocks): New.
17434 (thread_jumps::profitable_jump_thread_path): Move from here...
17435 (back_threader_profitability::profitable_path_p): ...to here.
17436 (thread_jumps::find_taken_edge): New.
17437 (thread_jumps::convert_and_register_current_path): Move...
17438 (back_threader_registry::register_path): ...to here.
17439 (thread_jumps::register_jump_thread_path_if_profitable): Move...
17440 (thread_jumps::maybe_register_path): ...to here.
17441 (thread_jumps::handle_phi): Call find_taken_edge and
17442 maybe_register_path.
17443 (thread_jumps::handle_assignment): Same.
17444 (thread_jumps::fsm_find_control_statement_thread_paths): Remove
17445 tree argument to handle_phi and handle_assignment.
17446 (thread_jumps::find_jump_threads_backwards): Set m_name. Remove
17447 set of m_speed_p and m_max_threaded_paths.
17448 (pass_thread_jumps::execute): Remove second argument from
17449 find_jump_threads_backwards.
17450 (pass_early_thread_jumps::execute): Same.
17452 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
17454 * tree-ssa-dom.c (class dom_jump_threader_simplifier): New.
17455 (class dom_opt_dom_walker): Initialize some class variables.
17456 (pass_dominator::execute): Pass evrp_range_analyzer and
17457 dom_jump_threader_simplifier to dom_opt_dom_walker.
17458 Adjust for some functions moving into classes.
17459 (simplify_stmt_for_jump_threading): Adjust and move to...
17460 (jump_threader_simplifier::simplify): ...here.
17461 (dom_opt_dom_walker::before_dom_children): Adjust for
17462 m_evrp_range_analyzer.
17463 (dom_opt_dom_walker::after_dom_children): Remove x_vr_values hack.
17464 (test_for_singularity): Place in dom_opt_dom_walker class.
17465 (dom_opt_dom_walker::optimize_stmt): The argument
17466 evrp_range_analyzer is now a class field.
17467 * tree-ssa-threadbackward.c (class thread_jumps): Add m_registry.
17468 (thread_jumps::thread_through_all_blocks): New.
17469 (thread_jumps::convert_and_register_current_path): Use m_registry.
17470 (pass_thread_jumps::execute): Adjust for thread_through_all_blocks
17471 being in the threader class.
17472 (pass_early_thread_jumps::execute): Same.
17473 * tree-ssa-threadedge.c (threadedge_initialize_values): Move...
17474 (jump_threader::jump_threader): ...here.
17475 (threadedge_finalize_values): Move...
17476 (jump_threader::~jump_threader): ...here.
17477 (jump_threader::remove_jump_threads_including): New.
17478 (jump_threader::thread_through_all_blocks): New.
17479 (record_temporary_equivalences_from_phis): Move...
17480 (jump_threader::record_temporary_equivalences_from_phis): ...here.
17481 (record_temporary_equivalences_from_stmts_at_dest): Move...
17482 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
17484 (simplify_control_stmt_condition_1): Move to jump_threader class.
17485 (simplify_control_stmt_condition): Move...
17486 (jump_threader::simplify_control_stmt_condition): ...here.
17487 (thread_around_empty_blocks): Move...
17488 (jump_threader::thread_around_empty_blocks): ...here.
17489 (thread_through_normal_block): Move...
17490 (jump_threader::thread_through_normal_block): ...here.
17491 (thread_across_edge): Move...
17492 (jump_threader::thread_across_edge): ...here.
17493 (thread_outgoing_edges): Move...
17494 (jump_threader::thread_outgoing_edges): ...here.
17495 * tree-ssa-threadedge.h: Move externally facing functings...
17496 (class jump_threader): ...here...
17497 (class jump_threader_simplifier): ...and here.
17498 * tree-ssa-threadupdate.c (struct redirection_data): Remove comment.
17499 (jump_thread_path_allocator::jump_thread_path_allocator): New.
17500 (jump_thread_path_allocator::~jump_thread_path_allocator): New.
17501 (jump_thread_path_allocator::allocate_thread_edge): New.
17502 (jump_thread_path_allocator::allocate_thread_path): New.
17503 (jump_thread_path_registry::jump_thread_path_registry): New.
17504 (jump_thread_path_registry::~jump_thread_path_registry): New.
17505 (jump_thread_path_registry::allocate_thread_edge): New.
17506 (jump_thread_path_registry::allocate_thread_path): New.
17507 (dump_jump_thread_path): Make extern.
17508 (debug (const vec<jump_thread_edge *> &path)): New.
17509 (struct removed_edges): Move to tree-ssa-threadupdate.h.
17510 (struct thread_stats_d): Remove.
17511 (remove_ctrl_stmt_and_useless_edges): Make static.
17512 (lookup_redirection_data): Move...
17513 (jump_thread_path_registry::lookup_redirection_data): ...here.
17514 (ssa_redirect_edges): Make static.
17515 (thread_block_1): Move...
17516 (jump_thread_path_registry::thread_block_1): ...here.
17517 (thread_block): Move...
17518 (jump_thread_path_registry::thread_block): ...here.
17519 (thread_through_loop_header): Move...
17520 (jump_thread_path_registry::thread_through_loop_header): ...here.
17521 (mark_threaded_blocks): Move...
17522 (jump_thread_path_registry::mark_threaded_blocks): ...here.
17523 (debug_path): Move...
17524 (jump_thread_path_registry::debug_path): ...here.
17525 (debug_all_paths): Move...
17526 (jump_thread_path_registry::dump): ..here.
17527 (rewire_first_differing_edge): Move...
17528 (jump_thread_path_registry::rewire_first_differing_edge): ...here.
17529 (adjust_paths_after_duplication): Move...
17530 (jump_thread_path_registry::adjust_paths_after_duplication): ...here.
17531 (duplicate_thread_path): Move...
17532 (jump_thread_path_registry::duplicate_thread_path): ..here.
17533 (remove_jump_threads_including): Move...
17534 (jump_thread_path_registry::remove_jump_threads_including): ...here.
17535 (thread_through_all_blocks): Move to...
17536 (jump_thread_path_registry::thread_through_all_blocks): ...here.
17537 (delete_jump_thread_path): Remove.
17538 (register_jump_thread): Move...
17539 (jump_thread_path_registry::register_jump_thread): ...here.
17540 * tree-ssa-threadupdate.h: Move externally facing functions...
17541 (class jump_thread_path_allocator): ...here...
17542 (class jump_thread_path_registry): ...and here.
17543 (thread_through_all_blocks): Remove.
17544 (struct removed_edges): New.
17545 (register_jump_thread): Remove.
17546 (remove_jump_threads_including): Remove.
17547 (delete_jump_thread_path): Remove.
17548 (remove_ctrl_stmt_and_useless_edges): Remove.
17549 (free_dom_edge_info): New prototype.
17550 * tree-vrp.c: Remove x_vr_values hack.
17551 (class vrp_jump_threader_simplifier): New.
17552 (vrp_jump_threader_simplifier::simplify): New.
17553 (vrp_jump_threader::vrp_jump_threader): Adjust method signature.
17554 Remove m_dummy_cond.
17555 Instantiate m_simplifier and m_threader.
17556 (vrp_jump_threader::thread_through_all_blocks): New.
17557 (vrp_jump_threader::simplify_stmt): Remove.
17558 (vrp_jump_threader::after_dom_children): Do not set m_dummy_cond.
17559 Remove x_vr_values hack.
17560 (execute_vrp): Adjust for thread_through_all_blocks being in a
17563 2021-04-30 Christophe Lyon <christophe.lyon@linaro.org>
17565 * genflags.c (gen_insn): Print failed expansion string.
17567 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
17569 * expr.c (alignment_for_piecewise_move): Call mode_for_size
17570 without limit to MAX_FIXED_MODE_SIZE.
17572 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
17574 PR middle-end/90773
17575 * builtins.c (builtin_memset_gen_str): Don't use return from
17576 simplify_gen_subreg.
17578 2021-04-30 Uroš Bizjak <ubizjak@gmail.com>
17581 * config/i386/i386.md (*add<mode>3_carry_0r): New insn pattern.
17582 (*addsi3_carry_zext_0r): Ditto.
17583 (*sub<mode>3_carry_0): Ditto.
17584 (*subsi3_carry_zext_0r): Ditto.
17585 * config/i386/predicates.md (ix86_carry_flag_unset_operator):
17587 * config/i386/i386.c (ix86_rtx_costs) <case PLUS, case MINUS>:
17588 Also consider ix86_carry_flag_unset_operator to calculate
17589 the cost of adc/sbb insn.
17591 2021-04-30 Roman Zhuykov <zhroma@ispras.ru>
17593 PR rtl-optimization/100225
17594 PR rtl-optimization/84878
17595 * modulo-sched.c (sms_schedule): Use note_stores to skip loops
17596 where we have an instruction which touches (writes) any hard
17597 register from df->regular_block_artificial_uses set.
17598 Allow not-single-set instruction only right before basic block
17601 2021-04-30 Geng Qi <gengqi@linux.alibaba.com>
17603 * config/riscv/riscv.opt (march=,mabi=): Negative itself.
17605 2021-04-30 LevyHsu <admin@levyhsu.com>
17607 * config/riscv/riscv.c (riscv_min_arithmetic_precision): New.
17608 * config/riscv/riscv.h (TARGET_MIN_ARITHMETIC_PRECISION): New.
17609 * config/riscv/riscv.md (addv<mode>4, uaddv<mode>4): New.
17610 (subv<mode>4, usubv<mode>4, mulv<mode>4, umulv<mode>4): New.
17612 2021-04-29 Alexandre Oliva <oliva@adacore.com>
17614 * config.gcc: Merged x86 and x86_64 cpu_type-setting cases.
17616 2021-04-29 Alexandre Oliva <oliva@adacore.com>
17618 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Rename to...
17619 (ASM_OUTPUT_MAX_SKIP_ALIGN): ... this. Enclose in do/while(0).
17620 * config/i386/i386.c: Adjust.
17621 * config/i386/i386.md: Adjust.
17622 * config/i386/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Drop.
17623 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17624 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17625 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17626 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17627 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17628 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17629 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17630 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17631 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
17632 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
17634 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
17636 * config/i386/i386-expand.c (ix86_expand_int_compare):
17637 Swap operands of GTU and LEU comparison to emit carry flag comparison.
17638 * config/i386/i386.md (*add<mode>3_carry_0): Change insn
17639 predicate to allow more combine opportunities with memory operands.
17640 (*sub<mode>3_carry_0): Ditto.
17642 2021-04-29 Richard Sandiford <richard.sandiford@arm.com>
17644 PR rtl-optimization/100303
17645 * rtl-ssa/accesses.cc (function_info::make_use_available): Take a
17646 boolean that indicates whether the use will only be used in
17647 debug instructions. Treat it in the same way that existing
17648 cross-EBB debug references would be handled if so.
17649 (function_info::make_uses_available): Likewise.
17650 * rtl-ssa/functions.h (function_info::make_uses_available): Update
17651 prototype accordingly.
17652 (function_info::make_uses_available): Likewise.
17653 * fwprop.c (try_fwprop_subst): Update call accordingly.
17655 2021-04-29 Jeff Law <jlaw@tachyum.com>
17657 * config/nios2/nios2-protos.h (nios2_fpu_insn_enabled): Move outside
17660 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
17661 Richard Biener <rguenther@suse.de>
17664 * config/i386/i386-builtin.def (IX86_BUILTIN_MASKLOADPD)
17665 (IX86_BUILTIN_MASKLOADPS, IX86_BUILTIN_MASKLOADPD256)
17666 (IX86_BUILTIN_MASKLOADPS256, IX86_BUILTIN_MASKLOADD)
17667 (IX86_BUILTIN_MASKLOADQ, IX86_BUILTIN_MASKLOADD256)
17668 (IX86_BUILTIN_MASKLOADQ256): Move from SPECIAL_ARGS
17669 to PURE_ARGS category.
17670 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
17671 Handle PURE_ARGS category.
17672 * config/i386/i386-expand.c (ix86_expand_builtin): Ditto.
17674 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
17676 * configure.ac: Check for the presence of sys/locking.h header and
17677 for whether _LK_LOCK is supported by _locking.
17678 * configure: Regenerate.
17679 * config.in: Likewise.
17680 * gcov-io.h: Define GCOV_LOCKED_WITH_LOCKING if HOST_HAS_LK_LOCK.
17681 * gcov-io.c (gcov_open): Add support for GCOV_LOCKED_WITH_LOCKING.
17682 * system.h: Include <sys/locking.h> if HAVE_SYS_LOCKING_H.
17684 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
17686 * config/i386/predicates.md (fcmov_comparison_operator):
17687 Do not check for trivial FP comparison operator.
17688 <case GEU, case LTU>: Allow CCGZmode.
17689 <case GTU, case LEU>: Do not allow CCCmode.
17690 (ix86_comparison_operator) <case GTU, case LEU>: Allow only CCmode.
17691 (ix86_carry_flag_operator): Match only LTU and UNLT code.
17692 Do not check for trivial FP comparison operator. Allow CCGZmode.
17694 2021-04-29 Tom de Vries <tdevries@suse.de>
17696 * omp-expand.c (expand_omp_simd): Add step_orig, and replace uses of
17697 fd->loop.step by either step or orig_step.
17699 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
17701 * config/sparc/sparc.c (gen_load_pcrel_sym): Delete.
17702 (load_got_register): Do the PIC dance here.
17703 (sparc_legitimize_tls_address): Simplify.
17704 (sparc_emit_probe_stack_range): Likewise.
17705 (sparc32_initialize_trampoline): Likewise.
17706 (sparc64_initialize_trampoline): Likewise.
17707 * config/sparc/sparc.md (load_pcrel_sym<P:mode>): Add @ marker.
17708 (probe_stack_range<P:mode>): Likewise.
17709 (flush<P:mode>): Likewise.
17710 (tgd_hi22<P:mode>): Likewise.
17711 (tgd_lo10<P:mode>): Likewise.
17712 (tgd_add<P:mode>): Likewise.
17713 (tgd_call<P:mode>): Likewise.
17714 (tldm_hi22<P:mode>): Likewise.
17715 (tldm_lo10<P:mode>): Likewise.
17716 (tldm_add<P:mode>): Likewise.
17717 (tldm_call<P:mode>): Likewise.
17718 (tldo_hix22<P:mode>): Likewise.
17719 (tldo_lox10<P:mode>): Likewise.
17720 (tldo_add<P:mode>): Likewise.
17721 (tie_hi22<P:mode>): Likewise.
17722 (tie_lo10<P:mode>): Likewise.
17723 (tie_add<P:mode>): Likewise.
17724 (tle_hix22<P:mode>): Likewise.
17725 (tle_lox10<P:mode>): Likewise.
17726 (stack_protect_setsi): Rename to...
17727 (stack_protect_set32): ...this.
17728 (stack_protect_setdi): Rename to...
17729 (stack_protect_set64): ...this.
17730 (stack_protect_set): Adjust calls to above.
17731 (stack_protect_testsi): Rename to...
17732 (stack_protect_test32): ...this.
17733 (stack_protect_testdi): Rename to...
17734 (stack_protect_test64): ...this.
17735 (stack_protect_test): Adjust calls to above.
17737 2021-04-29 H.J. Lu <hjl.tools@gmail.com>
17739 PR middle-end/90773
17740 * builtins.c (builtin_memcpy_read_str): Add a dummy argument.
17741 (builtin_strncpy_read_str): Likewise.
17742 (builtin_memset_read_str): Add an argument for the previous RTL
17743 information and generate the new RTL from the previous RTL info.
17744 (builtin_memset_gen_str): Likewise.
17745 * builtins.h (builtin_strncpy_read_str): Update the prototype.
17746 (builtin_memset_read_str): Likewise.
17747 * expr.c (by_pieces_ninsns): If targetm.overlap_op_by_pieces_p()
17748 returns true, round up size and alignment to the widest integer
17749 mode for maximum size.
17750 (pieces_addr::adjust): Add a pointer to by_pieces_prev argument
17751 and pass it to m_constfn.
17752 (op_by_pieces_d): Add m_push and m_overlap_op_by_pieces.
17753 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
17754 initialize m_push. Initialize m_overlap_op_by_pieces with
17755 targetm.overlap_op_by_pieces_p ().
17756 (op_by_pieces_d::run): Pass the previous RTL information to
17757 pieces_addr::adjust and generate overlapping operations if
17758 m_overlap_op_by_pieces is true.
17760 (move_by_pieces_d::move_by_pieces_d): Updated for op_by_pieces_d
17762 (store_by_pieces_d::store_by_pieces_d): Updated for op_by_pieces_d
17764 (can_store_by_pieces): Use by_pieces_constfn on constfun.
17765 (store_by_pieces): Use by_pieces_constfn on constfun. Updated
17766 for op_by_pieces_d change.
17767 (clear_by_pieces_1): Add a dummy argument.
17768 (clear_by_pieces): Updated for op_by_pieces_d change.
17769 (compare_by_pieces_d::compare_by_pieces_d): Likewise.
17770 (string_cst_read_str): Add a dummy argument.
17771 * expr.h (by_pieces_constfn): Add a dummy argument.
17772 (by_pieces_prev): New.
17773 * target.def (overlap_op_by_pieces_p): New target hook.
17774 * config/i386/i386.c (TARGET_OVERLAP_OP_BY_PIECES_P): New.
17775 * doc/tm.texi.in: Add TARGET_OVERLAP_OP_BY_PIECES_P.
17776 * doc/tm.texi: Regenerated.
17778 2021-04-29 Richard Biener <rguenther@suse.de>
17780 PR tree-optimization/100253
17781 * tree-vect-stmts.c (vectorizable_load): Do not assume
17782 element alignment when DR_MISALIGNMENT is -1.
17783 (vectorizable_store): Likewise.
17785 2021-04-29 Jakub Jelinek <jakub@redhat.com>
17788 * config/aarch64/aarch64.c (aarch64_add_offset_1_temporaries): Use
17789 absu_hwi instead of abs_hwi.
17791 2021-04-29 Richard Biener <rguenther@suse.de>
17793 PR middle-end/38474
17794 * tree-ssa-structalias.c (add_graph_edge): Avoid direct
17795 forwarding when indirect forwarding through ESCAPED
17798 2021-04-29 Tom de Vries <tdevries@suse.de>
17801 * internal-fn.c (expand_GOMP_SIMT_ENTER_ALLOC)
17802 (expand_GOMP_SIMT_LAST_LANE, expand_GOMP_SIMT_ORDERED_PRED)
17803 (expand_GOMP_SIMT_VOTE_ANY, expand_GOMP_SIMT_XCHG_BFLY)
17804 (expand_GOMP_SIMT_XCHG_IDX): Ensure target is assigned to.
17806 2021-04-29 Richard Biener <rguenther@suse.de>
17808 PR tree-optimization/99912
17809 * tree-ssa-dse.c (dse_dom_walker::m_need_cfg_cleanup): New.
17810 (dse_dom_walker::todo): Likewise.
17811 (dse_dom_walker::dse_optimize_stmt): Move VDEF check to the
17813 (dse_dom_walker::before_dom_children): Remove trivially
17814 dead SSA defs and schedule CFG cleanup if we removed all
17816 (pass_dse::execute): Get TODO as computed by the DOM walker
17817 and return it. Wipe dominator info earlier.
17819 2021-04-29 Richard Biener <rguenther@suse.de>
17822 * ipa-prop.c (ipcp_modif_dom_walker::before_dom_children):
17823 Track blocks to cleanup EH in new m_need_eh_cleanup.
17824 (ipcp_modif_dom_walker::cleanup_eh): New.
17825 (ipcp_transform_function): Release dominator info before
17828 2021-04-29 Martin Sebor <msebor@redhat.com>
17830 PR middle-end/100250
17831 * attribs.c (attr_access::array_as_string): Avoid dereferencing
17832 a pointer when it's null.
17834 2021-04-29 Martin Sebor <msebor@redhat.com>
17836 * Makefile.in (OBJS): Add ipa-free-lang-data.o.
17837 * ipa-free-lang-data.cc: New file.
17838 * tree.c: Move pass free_lang_data to file above.
17839 (build_array_type_1): Declare extern.
17840 * tree.h (build_array_type_1): Declare.
17842 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17844 * config/aarch64/aarch64-simd-builtins.def: Modify comment to
17845 make consistent with updated RTL pattern.
17846 * config/aarch64/aarch64-simd.md (aarch64_<sur>qmovn<mode>):
17847 Implement using ss_truncate and us_truncate rather than
17849 * config/aarch64/iterators.md: Remove redundant unspecs and
17850 iterator: UNSPEC_[SU]QXTN and SUQMOVN respectively.
17852 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17854 * config/aarch64/arm_acle.h (__attribute__): Make intrinsic
17855 attributes consistent with those defined in arm_neon.h.
17857 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17859 * config/aarch64/arm_fp16.h (__attribute__): Make intrinsic
17860 attributes consistent with those defined in arm_neon.h.
17862 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17864 * config/aarch64/aarch64-simd-builtins.def: Add
17865 float_trunc_rodd builtin generator macros.
17866 * config/aarch64/aarch64-simd.md (aarch64_float_trunc_rodd_df):
17868 (aarch64_float_trunc_rodd_lo_v2sf): Define.
17869 (aarch64_float_trunc_rodd_hi_v4sf_le): Define.
17870 (aarch64_float_trunc_rodd_hi_v4sf_be): Define.
17871 (aarch64_float_trunc_rodd_hi_v4sf): Define.
17872 * config/aarch64/arm_neon.h (vcvtx_f32_f64): Use RTL builtin
17873 instead of inline asm.
17874 (vcvtx_high_f32_f64): Likewise.
17875 (vcvtxd_f32_f64): Likewise.
17876 * config/aarch64/iterators.md: Add FCVTXN unspec.
17878 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17880 * config/aarch64/aarch64-simd-builtins.def: Add tbx1 builtin
17882 * config/aarch64/aarch64-simd.md (aarch64_tbx1<mode>):
17884 * config/aarch64/arm_neon.h (vqtbx1_s8): USE RTL builtin
17885 instead of inline asm.
17886 (vqtbx1_u8): Likewise.
17887 (vqtbx1_p8): Likewise.
17888 (vqtbx1q_s8): Likewise.
17889 (vqtbx1q_u8): Likewise.
17890 (vqtbx1q_p8): Likewise.
17891 (vtbx2_s8): Likewise.
17892 (vtbx2_u8): Likewise.
17893 (vtbx2_p8): Likewise.
17895 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17897 * config/aarch64/aarch64-simd-builtins.def: Add tbl1 builtin
17899 * config/aarch64/arm_neon.h (vqtbl1_p8): Use RTL builtin
17900 instead of inline asm.
17901 (vqtbl1_s8): Likewise.
17902 (vqtbl1_u8): Likewise.
17903 (vqtbl1q_p8): Likewise.
17904 (vqtbl1q_s8): Likewise.
17905 (vqtbl1q_u8): Likewise.
17906 (vtbl1_s8): Likewise.
17907 (vtbl1_u8): Likewise.
17908 (vtbl1_p8): Likewise.
17909 (vtbl2_s8): Likewise.
17910 (vtbl2_u8): Likewise.
17911 (vtbl2_p8): Likewise.
17913 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17915 * config/aarch64/aarch64-simd-builtins.def: Add polynomial
17916 ssri_n buitin generator macro.
17917 * config/aarch64/arm_neon.h (vsri_n_p8): Use RTL builtin
17918 instead of inline asm.
17919 (vsri_n_p16): Likewise.
17920 (vsri_n_p64): Likewise.
17921 (vsriq_n_p8): Likewise.
17922 (vsriq_n_p16): Likewise.
17923 (vsriq_n_p64): Likewise.
17925 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17927 * config/aarch64/aarch64-simd-builtins.def: Use VALLP mode
17928 iterator for polynomial ssli_n builtin generator macro.
17929 * config/aarch64/arm_neon.h (vsli_n_p8): Use RTL builtin
17930 instead of inline asm.
17931 (vsli_n_p16): Likewise.
17932 (vsliq_n_p8): Likewise.
17933 (vsliq_n_p16): Likewise.
17934 * config/aarch64/iterators.md: Define VALLP mode iterator.
17936 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17938 * config/aarch64/aarch64-simd-builtins.def: Use VDQV_L
17939 iterator to generate [su]adalp RTL builtins.
17940 * config/aarch64/aarch64-simd.md: Use VDQV_L iterator in
17941 [su]adalp RTL pattern.
17942 * config/aarch64/arm_neon.h (vpadal_s32): Use RTL builtin
17943 instead of inline asm.
17944 (vpadal_u32): Likewise.
17946 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17948 * config/aarch64/aarch64-simd-builtins.def: Add [su]addlp
17949 builtin generator macros.
17950 * config/aarch64/aarch64-simd.md (aarch64_<su>addlp<mode>):
17952 * config/aarch64/arm_neon.h (vpaddl_s8): Use RTL builtin
17953 instead of inline asm.
17954 (vpaddl_s16): Likewise.
17955 (vpaddl_s32): Likewise.
17956 (vpaddl_u8): Likewise.
17957 (vpaddl_u16): Likewise.
17958 (vpaddl_u32): Likewise.
17959 (vpaddlq_s8): Likewise.
17960 (vpaddlq_s16): Likewise.
17961 (vpaddlq_s32): Likewise.
17962 (vpaddlq_u8): Likewise.
17963 (vpaddlq_u16): Likewise.
17964 (vpaddlq_u32): Liwewise.
17965 * config/aarch64/iterators.md: Define [SU]ADDLP unspecs with
17966 appropriate attributes.
17968 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17970 * config/aarch64/aarch64-simd-builtins.def: Use VDQ_I iterator
17971 for aarch64_addp<mode> builtin macro generator.
17972 * config/aarch64/aarch64-simd.md: Use VDQ_I iterator in
17973 aarch64_addp<mode> RTL pattern.
17974 * config/aarch64/arm_neon.h (vpaddq_s8): Use RTL builtin
17975 instead of inline asm.
17976 (vpaddq_s16): Likewise.
17977 (vpaddq_s32): Likewise.
17978 (vpaddq_s64): Likewise.
17979 (vpaddq_u8): Likewise.
17980 (vpaddq_u16): Likewise.
17981 (vpaddq_u32): Likewise.
17982 (vpaddq_u64): Likewise.
17984 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
17986 * config/aarch64/aarch64-simd-builtins.def: Add sq[r]dmulh_n
17987 builtin generator macros.
17988 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_n<mode>):
17990 * config/aarch64/arm_neon.h (vqdmulh_n_s16): Use RTL builtin
17991 instead of inline asm.
17992 (vqdmulh_n_s32): Likewise.
17993 (vqdmulhq_n_s16): Likewise.
17994 (vqdmulhq_n_s32): Likewise.
17995 (vqrdmulh_n_s16): Likewise.
17996 (vqrdmulh_n_s32): Likewise.
17997 (vqrdmulhq_n_s16): Likewise.
17998 (vqrdmulhq_n_s32): Likewise.
18000 2021-04-28 Tobias Burnus <tobias@codesourcery.com>
18002 * doc/install.texi (--enable-offload-defaulted): Document.
18004 2021-04-28 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
18006 * config/avr/avr-dimode.md: Turn existing patterns into
18007 define_insn_and_split style patterns where the splitter
18008 adds a clobber of the condition code register. Drop "cc"
18009 attribute. Add new patterns to match output of
18011 * config/avr/avr-fixed.md: Likewise.
18012 * config/avr/avr.c (cc_reg_rtx): New.
18013 (avr_parallel_insn_from_insns): Adjust insn count
18014 for removal of set of cc0.
18015 (avr_is_casesi_sequence): Likewise.
18016 (avr_casei_sequence_check_operands): Likewise.
18017 (avr_optimize_casesi): Likewise. Also insert
18018 new insns after jump_insn.
18019 (avr_pass_casesi::avr_rest_of_handle_casesi): Adjust
18020 for removal of set of cc0.
18021 (avr_init_expanders): Initialize cc_reg_rtx.
18022 (avr_regno_reg_class): Handle REG_CC.
18023 (cond_string): Remove usage of CC_OVERFLOW_UNUSABLE.
18024 (avr_notice_update_cc): Remove function.
18025 (ret_cond_branch): Remove usage of CC_OVERFLOW_UNUSABLE.
18026 (compare_condition): Adjust for PARALLEL with
18028 (out_shift_with_cnt): Likewise.
18029 (ashlhi3_out): Likewise.
18030 (ashrhi3_out): Likewise.
18031 (lshrhi3_out): Likewise.
18032 (avr_class_max_nregs): Return single reg for REG_CC.
18033 (avr_compare_pattern): Check for REG_CC instead
18035 (avr_reorg_remove_redundant_compare): Likewise.
18036 (avr_reorg):Adjust for PARALLEL with REG_CC clobber.
18037 (avr_hard_regno_nregs): Return single reg for REG_CC.
18038 (avr_hard_regno_mode_ok): Allow only CCmode for REG_CC.
18039 (avr_md_asm_adjust): Clobber REG_CC.
18040 (TARGET_HARD_REGNO_NREGS): Define.
18041 (TARGET_CLASS_MAX_NREGS): Define.
18042 (TARGET_MD_ASM_ADJUST): Define.
18043 * config/avr/avr.h (FIRST_PSEUDO_REGISTER): Adjust
18045 (enum reg_class): Add CC_REG class.
18046 (NOTICE_UPDATE_CC): Remove.
18047 (CC_OVERFLOW_UNUSABLE): Remove.
18048 (CC_NO_CARRY): Remove.
18049 * config/avr/avr.md: Turn existing patterns into
18050 define_insn_and_split style patterns where the splitter
18051 adds a clobber of the condition code register. Drop "cc"
18052 attribute. Add new patterns to match output of
18054 (sez): Remove unused pattern.
18056 2021-04-28 Richard Earnshaw <rearnsha@arm.com>
18059 * config/arm/arm.c (arm_hard_regno_mode_ok): Only allow VPR to be
18062 2021-04-28 Richard Sandiford <richard.sandiford@arm.com>
18065 * config/aarch64/constraints.md (Utq): Require the address to
18066 be valid for both the element mode and for V2DImode.
18068 2021-04-28 Jakub Jelinek <jakub@redhat.com>
18069 Tobias Burnus <tobias@codesourcery.com>
18071 * configure.ac (OFFLOAD_DEFAULTED): AC_DEFINE if offload-defaulted.
18072 * gcc.c (process_command): New variable.
18073 (driver::maybe_putenv_OFFLOAD_TARGETS): If OFFLOAD_DEFAULTED,
18074 set it if -foffload is defaulted.
18075 * lto-wrapper.c (OFFLOAD_TARGET_DEFAULT_ENV): Define.
18076 (compile_offload_image): If OFFLOAD_DEFAULTED and
18077 OFFLOAD_TARGET_DEFAULT is in the environment, don't fail
18078 if corresponding mkoffload can't be found.
18079 (compile_images_for_offload_targets): Likewise. Free and clear
18080 offload_names if no valid offload is found.
18081 * config.in: Regenerate.
18082 * configure: Regenerate.
18084 2021-04-28 Richard Biener <rguenther@suse.de>
18086 PR tree-optimization/100292
18087 * tree-vect-generic.c (expand_vector_condition): Do not fold
18090 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
18092 * config/rs6000/aix.h (SUBTARGET_DRIVER_SELF_SPECS): New.
18093 * config/rs6000/aix64.opt (m64): New.
18096 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
18098 * config/vax/vax.c (print_operand_address, vax_address_cost_1)
18099 (index_term_p): Handle ASHIFT too.
18101 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
18103 * config/vax/builtins.md (jbb<ccss>i<mode>): Remove operand #3.
18104 (sync_lock_test_and_set<mode>): Adjust accordingly.
18105 (sync_lock_release<mode>): Likewise.
18107 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
18109 * config/vax/vax-protos.h (adjacent_operands_p): Remove
18111 * config/vax/vax.c (adjacent_operands_p): Remove.
18113 2021-04-27 Maciej W. Rozycki <macro@linux-mips.org>
18115 * ifcvt.c (dead_or_predicable) [!IFCVT_MODIFY_TESTS]: Fall
18116 through to the non-conditional execution case if getting the
18117 condition for conditional execution has failed.
18119 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
18121 PR middle-end/100284
18122 * gimple.c (gimple_could_trap_p_1): Remove VEC_COND_EXPR test.
18123 * tree-eh.c (operation_could_trap_p): Handle VEC_COND_EXPR rather
18124 than asserting on it.
18126 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
18128 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): Protect
18129 with TARGET_AIX_OS.
18131 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
18134 * calls.c (precompute_register_parameters): Additionally test
18135 targetm.precompute_tls_p to pre-compute argument.
18136 * config/rs6000/aix.h (TARGET_PRECOMPUTE_TLS_P): Define.
18137 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): New.
18138 * target.def (precompute_tls_p): New.
18139 * doc/tm.texi.in (TARGET_PRECOMPUTE_TLS_P): Add hook documentation.
18140 * doc/tm.texi: Regenerated.
18142 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18145 * config/aarch64/aarch64.c (aarch64_print_operand): Cast -UINTVAL
18146 back to HOST_WIDE_INT.
18148 2021-04-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
18151 * simplify-rtx.c (simplify_context::simplify_subreg): Check the
18152 memory alignment for the outer mode.
18154 2021-04-27 H.J. Lu <hjl.tools@gmail.com>
18156 PR middle-end/90773
18157 * expr.c (op_by_pieces_d::get_usable_mode): New member function.
18158 (op_by_pieces_d::run): Cange a while loop to a do-while loop.
18160 2021-04-27 Alex Coplan <alex.coplan@arm.com>
18163 * config/arm/arm.c (arm_split_compare_and_swap): Fix up codegen
18164 with negative immediates: ensure we expand cbranchsi4_scratch
18165 correctly and ensure we satisfy its constraints.
18166 * config/arm/sync.md
18167 (@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): Don't
18168 attempt to tie two output operands together with constraints;
18169 collapse two alternatives.
18170 (@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
18171 * config/arm/thumb1.md (cbranchsi4_neg_late): New.
18173 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18176 * config/aarch64/predicates.md (aarch64_sub_immediate,
18177 aarch64_plus_immediate): Use -UINTVAL instead of -INTVAL.
18178 * config/aarch64/aarch64.md (casesi, rotl<mode>3): Likewise.
18179 * config/aarch64/aarch64.c (aarch64_print_operand,
18180 aarch64_split_atomic_op, aarch64_expand_subvti): Likewise.
18182 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18184 PR tree-optimization/100239
18185 * tree-vect-generic.c (lower_vec_perm): Don't accept constant
18186 permutations with all indices from the first zero element as vec_shl.
18188 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18190 PR rtl-optimization/100254
18191 * cfgcleanup.c (outgoing_edges_match): Check REG_EH_REGION on
18192 last1 and last2 insns rather than BB_END (bb1) and BB_END (bb2) insns.
18194 2021-04-27 Richard Biener <rguenther@suse.de>
18196 PR tree-optimization/99912
18197 * passes.def: Add comment about new TODO_remove_unused_locals.
18198 * tree-stdarg.c (pass_data_stdarg): Run TODO_remove_unused_locals
18201 2021-04-27 Richard Biener <rguenther@suse.de>
18203 PR tree-optimization/99912
18204 * passes.def (pass_all_optimizations): Add pass_dse before
18205 the first pass_dce, move the first pass_dse before the
18206 pass_dce following pass_pre.
18208 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18210 PR tree-optimization/95527
18211 * generic-match-head.c: Include tm.h.
18212 * gimple-match-head.c: Include tm.h.
18213 * match.pd (CLZ == INTEGER_CST): Don't use
18214 #ifdef CLZ_DEFINED_VALUE_AT_ZERO, only test CLZ_DEFINED_VALUE_AT_ZERO
18215 if clz == CFN_CLZ. Add missing val declaration.
18216 (CTZ cmp CST): New simplifications.
18218 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18220 PR tree-optimization/96696
18221 * expr.c (expand_expr_divmod): New function.
18222 (expand_expr_real_2) <case TRUNC_DIV_EXPR>: Use it for truncations and
18223 divisions. Formatting fixes.
18224 <case MULT_EXPR>: Optimize x / y * y as x - x % y if the latter is
18227 2021-04-27 Martin Jambor <mjambor@suse.cz>
18230 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
18231 If removing a call statement LHS SSA name, release it.
18233 2021-04-27 Richard Earnshaw <rearnsha@arm.com>
18236 * config/arm/arm.c (THUMB2_WORK_REGS): Check PIC_OFFSET_TABLE_REGNUM
18237 is valid before including it in the mask.
18239 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
18242 * config/aarch64/aarch64.c (aarch64_comp_type_attributes): Handle
18245 2021-04-27 Richard Biener <rguenther@suse.de>
18247 PR tree-optimization/100051
18248 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Add
18249 disambiguator based on access size vs. decl size.
18251 2021-04-27 Richard Biener <rguenther@suse.de>
18253 PR tree-optimization/100278
18254 * tree-ssa-pre.c (compute_avail): Give up when we cannot
18255 adjust TBAA beacuse of mismatching bases.
18257 2021-04-27 Jakub Jelinek <jakub@redhat.com>
18260 * config/i386/i386.md (*<insn><mode>3_mask, *<insn><mode>3_mask_1):
18261 For any_rotate define_insn_split and following splitters, use
18262 SWI iterator instead of SWI48.
18264 2021-04-27 Richard Biener <rguenther@suse.de>
18266 PR tree-optimization/99776
18267 * match.pd (bit_field_ref (ctor)): Relax element extract
18268 type compatibility checks.
18270 2021-04-27 Cui,Lili <lili.cui@intel.com>
18272 * common/config/i386/i386-common.c (processor_names):
18273 Sync processor_names with processor_type.
18274 * config/i386/i386-options.c (processor_cost_table):
18275 Sync processor_cost_table with processor_type.
18277 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18279 * value-range.cc (irange::irange_set_1bit_anti_range): Add assert.
18280 (irange::set): Call irange_set_1bit_anti_range for handling all
18281 1-bit ranges. Fall through on ~[MIN,MAX].
18283 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18285 * value-range.cc (irange::legacy_num_pairs): Remove.
18286 (irange::invert): Change gcc_assert to gcc_checking_assert.
18287 * value-range.h (irange::num_pairs): Adjust for a cached
18288 num_pairs(). Also, rename all gcc_assert's to
18289 gcc_checking_assert's.
18291 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18293 * value-range.cc (irange::operator=): Set m_kind.
18294 (irange::copy_to_legacy): Handle varying and undefined sources
18295 as a legacy copy since they can be easily copied.
18296 (irange::irange_set): Set m_kind.
18297 (irange::irange_set_anti_range): Same.
18298 (irange::set): Rename normalize_min_max to normalize_kind.
18299 (irange::verify_range): Adjust for multi-ranges having the
18301 (irange::irange_union): Set m_kind.
18302 (irange::irange_intersect): Same.
18303 (irange::invert): Same.
18304 * value-range.h (irange::kind): Always return m_kind.
18305 (irange::varying_p): Rename to...
18306 (irange::varying_comptaible_p): ...this.
18307 (irange::undefined_p): Only look at m_kind.
18308 (irange::irange): Always set VR_UNDEFINED if applicable.
18309 (irange::set_undefined): Always set VR_UNDEFINED.
18310 (irange::set_varying): Always set m_kind to VR_VARYING.
18311 (irange::normalize_min_max): Rename to...
18312 (irange::normalize_kind): ...this.
18314 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18316 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
18317 Adjust for constant_p including varying_p.
18318 * tree-vrp.c (vrp_prop::finalize): Same.
18319 (determine_value_range): Same.
18320 * vr-values.c (vr_values::range_of_expr): Same.
18321 * value-range.cc (irange::symbolic_p): Do not check varying_p.
18322 (irange::constant_p): Same.
18324 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
18326 * value-range.cc (irange::legacy_lower_bound): Replace
18327 !undefined_p check with num_ranges > 0.
18328 (irange::legacy_upper_bound): Same.
18329 * value-range.h (irange::type): Same.
18330 (irange::lower_bound): Same.
18331 (irange::upper_bound): Same.
18333 2021-04-26 Richard Biener <rguenther@suse.de>
18335 PR tree-optimization/99956
18336 * gimple-loop-interchange.cc (compute_access_stride):
18337 Try instantiating the access in a shallower loop nest
18338 if instantiating failed.
18339 (compute_access_strides): Pass adjustable loop_nest
18340 to compute_access_stride.
18342 2021-04-26 Christophe Lyon <christophe.lyon@linaro.org>
18344 * doc/sourcebuild.texi (arm_cmse_hw): Document.
18346 2021-04-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18348 * config/aarch64/iterators.md (vwcore): Handle V4BF, V8BF.
18350 2021-04-26 Thomas Schwinge <thomas@codesourcery.com>
18351 Nathan Sidwell <nathan@codesourcery.com>
18352 Tom de Vries <vries@codesourcery.com>
18353 Julian Brown <julian@codesourcery.com>
18354 Kwok Cheung Yeung <kcy@codesourcery.com>
18356 * omp-offload.c (oacc_validate_dims): Implement
18357 '-Wopenacc-parallelism'.
18358 * doc/invoke.texi (-Wopenacc-parallelism): Document.
18360 2021-04-26 Richard Biener <rguenther@suse.de>
18362 * tree-cfg.h (gimplify_build1): Remove.
18363 (gimplify_build2): Likewise.
18364 (gimplify_build3): Likewise.
18365 * tree-cfg.c (gimplify_build1): Move to tree-vect-generic.c.
18366 (gimplify_build2): Likewise.
18367 (gimplify_build3): Likewise.
18368 * tree-vect-generic.c (gimplify_build1): Move from tree-cfg.c.
18370 (gimplify_build2): Likewise.
18371 (gimplify_build3): Likewise.
18372 (tree_vec_extract): Use resimplify with following SSA edges.
18373 (expand_vector_parallel): Avoid passing NULL size/bitpos
18374 to tree_vec_extract.
18375 * expr.c (store_constructor): Deal with zero-element CTORs.
18376 * match.pd (bit_field_ref <vector CTOR>): Make sure to
18377 produce vector constants when possible.
18379 2021-04-26 Richard Biener <rguenther@suse.de>
18381 * tree-complex.c: Include gimple-fold.h.
18382 (expand_complex_addition): Use gimple_build.
18383 (expand_complex_multiplication_components): Likewise.
18384 (expand_complex_multiplication): Likewise.
18385 (expand_complex_div_straight): Likewise.
18386 (expand_complex_div_wide): Likewise.
18387 (expand_complex_division): Likewise.
18388 (expand_complex_conjugate): Likewise.
18389 (expand_complex_comparison): Likewise.
18391 2021-04-26 Richard Biener <rguenther@suse.de>
18393 * tree-ssa-phiopt.c (two_value_replacement): Remove use
18394 of legacy gimplify_buildN API.
18396 2021-04-26 Richard Biener <rguenther@suse.de>
18398 PR tree-optimization/99473
18399 * tree-ssa-phiopt.c (cond_store_replacement): Handle all
18402 2021-04-26 Richard Biener <rguenther@suse.de>
18404 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin):
18405 Use replace_call_with_value.
18407 2021-04-26 Richard Biener <rguenther@suse.de>
18409 * tree-ssa-propagate.h (valid_gimple_rhs_p): Remove.
18410 (update_gimple_call): Likewise.
18411 (update_call_from_tree): Likewise.
18412 * tree-ssa-propagate.c (valid_gimple_rhs_p): Remove.
18413 (valid_gimple_call_p): Likewise.
18414 (move_ssa_defining_stmt_for_defs): Likewise.
18415 (finish_update_gimple_call): Likewise.
18416 (update_gimple_call): Likewise.
18417 (update_call_from_tree): Likewise.
18418 (propagate_tree_value_into_stmt): Use replace_call_with_value.
18419 * gimple-fold.h (update_gimple_call): Declare.
18420 * gimple-fold.c (valid_gimple_rhs_p): Move here from
18421 tree-ssa-propagate.c.
18422 (update_gimple_call): Likewise.
18423 (valid_gimple_call_p): Likewise.
18424 (finish_update_gimple_call): Likewise, and simplify.
18425 (gimplify_and_update_call_from_tree): Implement
18426 update_call_from_tree functionality, avoid excessive
18427 push/pop_gimplify_context.
18428 (gimple_fold_builtin): Use only gimplify_and_update_call_from_tree.
18429 (gimple_fold_call): Likewise.
18430 * gimple-ssa-sprintf.c (try_substitute_return_value): Likewise.
18431 * tree-ssa-ccp.c (ccp_folder::fold_stmt): Likewise.
18432 (pass_fold_builtins::execute): Likewise.
18433 (optimize_stack_restore): Use replace_call_with_value.
18434 * tree-cfg.c (fold_loop_internal_call): Likewise.
18435 * tree-ssa-dce.c (maybe_optimize_arith_overflow): Use
18436 only gimplify_and_update_call_from_tree.
18437 * tree-ssa-strlen.c (handle_builtin_strlen): Likewise.
18438 (handle_builtin_strchr): Likewise.
18439 * tsan.c: Include gimple-fold.h instead of tree-ssa-propagate.h.
18441 2021-04-26 Jakub Jelinek <jakub@redhat.com>
18444 * vmsdbgout.c (ASM_OUTPUT_DEBUG_STRING, vmsdbgout_begin_block,
18445 vmsdbgout_end_block, lookup_filename, vmsdbgout_source_line): Remove
18448 2021-04-25 liuhongt <hongtao.liu@intel.com>
18451 * config/i386/i386-builtin.def (BDESC): Change the icode of
18452 the following builtins to CODE_FOR_nothing.
18453 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
18454 IX86_BUILTIN_PCMPEQB128, IX86_BUILTIN_PCMPEQW128,
18455 IX86_BUILTIN_PCMPEQD128, IX86_BUILTIN_PCMPEQQ,
18456 IX86_BUILTIN_PCMPEQB256, IX86_BUILTIN_PCMPEQW256,
18457 IX86_BUILTIN_PCMPEQD256, IX86_BUILTIN_PCMPEQQ256,
18458 IX86_BUILTIN_PCMPGTB128, IX86_BUILTIN_PCMPGTW128,
18459 IX86_BUILTIN_PCMPGTD128, IX86_BUILTIN_PCMPGTQ,
18460 IX86_BUILTIN_PCMPGTB256, IX86_BUILTIN_PCMPGTW256,
18461 IX86_BUILTIN_PCMPGTD256, IX86_BUILTIN_PCMPGTQ256.
18462 * config/i386/sse.md (avx2_eq<mode>3): Deleted.
18463 (sse2_eq<mode>3): Ditto.
18464 (sse4_1_eqv2di3): Ditto.
18465 (sse2_gt<mode>3): Rename to ..
18466 (*sse2_gt<mode>3): .. this.
18468 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
18471 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
18474 * config/darwin.c (darwin_binds_local_p): Assume that any
18475 public symbol might be interposed for PIC code. Update function
18476 header comment to reflect current Darwin capability.
18478 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
18481 * config/darwin.c (darwin_binds_local_p): Assume that any
18482 public symbol might be interposed for PIC code. Update function
18483 header comment to reflect current Darwin capability.
18485 2021-04-24 Richard Sandiford <richard.sandiford@arm.com>
18487 * doc/sourcebuild.texi: Document no-opts and any-opts target
18490 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
18492 * config/bpf/bpf.h (ASM_OUTPUT_ALIGNED_BSS): Use .type and .lcomm.
18494 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
18496 * config/bpf/bpf.h (FUNCTION_BOUNDARY): Set to 64.
18498 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
18501 * config/i386/i386-options.c (ix86_option_override_internal):
18502 Error out when -m96bit-long-double is used with 64bit targets.
18503 * config/i386/i386.md (*pushxf_rounded): Remove pattern.
18505 2021-04-23 Martin Liska <mliska@suse.cz>
18507 * lto-wrapper.c: Remove FIXME about usage of
18508 hardware_concurrency. The function is not on par with
18511 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
18514 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
18515 Copy operand 3 to operand 4. Use sse_reg_operand
18516 as operand 3 predicate.
18517 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): Ditto.
18518 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
18519 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
18520 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2):
18521 Copy operand 1 to operand 0.
18522 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): Ditto.
18523 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
18524 (LDX_ATOMIC/LDX_ATOMIC FP store peephole2 with mem blockage): Ditto.
18526 2021-04-23 Alex Coplan <alex.coplan@arm.com>
18528 PR rtl-optimization/100230
18529 * early-remat.c (early_remat::sort_candidates): Use delete[]
18530 instead of delete for array allocated with new[].
18532 2021-04-23 Richard Biener <rguenther@suse.de>
18534 * genmatch.c (lower_cond): Remove VEC_COND_EXPR special-casing.
18535 (capture_info::capture_info): Likewise.
18536 (capture_info::walk_match): Likewise.
18537 (expr::gen_transform): Likewise.
18538 (dt_simplify::gen_1): Likewise.
18539 * gimple-match-head.c (maybe_resimplify_conditional_op):
18540 Remove VEC_COND_EXPR special-casing.
18541 (gimple_simplify): Likewise.
18542 * gimple.c (gimple_could_trap_p_1): Adjust.
18543 * tree-ssa-pre.c (compute_avail): Allow VEC_COND_EXPR
18544 to participate in PRE.
18546 2021-04-23 Richard Biener <rguenther@suse.de>
18548 * cfganal.c (connect_infinite_loops_to_exit): First call
18549 add_noreturn_fake_exit_edges.
18550 * ipa-sra.c (process_scan_results): Do not call the now redundant
18551 add_noreturn_fake_exit_edges.
18552 * predict.c (tree_estimate_probability): Likewise.
18553 (rebuild_frequencies): Likewise.
18554 * store-motion.c (one_store_motion_pass): Likewise.
18556 2021-04-23 Richard Biener <rguenther@suse.de>
18558 PR tree-optimization/100222
18559 * predict.c (pass_profile::execute): Remove redundant call to
18560 mark_irreducible_loops.
18561 (report_predictor_hitrates): Likewise.
18563 2021-04-23 Richard Biener <rguenther@suse.de>
18565 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Avoid
18566 valid_gimple_rhs_p by instead gimplifying to one.
18568 2021-04-23 Richard Biener <rguenther@suse.de>
18570 PR tree-optimization/99971
18571 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
18572 Always use TBAA for loads.
18574 2021-04-23 liuhongt <hongtao.liu@intel.com>
18577 * config/i386/i386-options.c (ix86_option_override_internal):
18578 Clear MASK_AVX256_SPLIT_UNALIGNED_LOAD/STORE in x_target_flags
18579 when X86_TUNE_AVX256_UNALIGNED_LOAD/STORE_OPTIMAL is enabled
18580 by target attribute.
18582 2021-04-23 David Edelsohn <dje.gcc@gmail.com>
18584 * config/rs6000/aix71.h (PREFERRED_DEBUGGING_TYPE): Change to
18586 * config/rs6000/aix72.h (PREFERRED_DEBUGGING_TYPE): Same.
18588 2021-04-22 David Edelsohn <dje.gcc@gmail.com>
18590 * config.gcc (powerpc-ibm-aix6.*): Remove.
18591 * config/rs6000/aix61.h: Delete.
18593 2021-04-22 Martin Liska <mliska@suse.cz>
18595 PR testsuite/100159
18596 PR testsuite/100192
18597 * builtins.c (expand_builtin): Fix typos and missing comments.
18598 * dwarf2out.c (gen_subprogram_die): Likewise.
18599 (gen_struct_or_union_type_die): Likewise.
18601 2021-04-22 Uroš Bizjak <ubizjak@gmail.com>
18604 * config/i386/i386-expand.c (ix86_expand_convert_uns_sidf_sse):
18605 Remove the sign with FE_DOWNWARD, where x - x = -0.0.
18607 2021-04-21 Iain Sandoe <iain@sandoe.co.uk>
18609 * config/i386/darwin.h (TARGET_64BIT): Remove definition
18610 based on TARGET_ISA_64BIT.
18611 (TARGET_64BIT_P): Remove definition based on
18612 TARGET_ISA_64BIT_P().
18614 2021-04-21 Martin Liska <mliska@suse.cz>
18617 2021-04-21 Martin Liska <mliska@suse.cz>
18619 * lto-wrapper.c (cpuset_popcount): Remove.
18620 (init_num_threads): Remove and use hardware_concurrency.
18622 2021-04-21 Martin Liska <mliska@suse.cz>
18625 * main.c (main): Call toplev::finalize in CHECKING_P mode.
18626 * ipa-modref.c (ipa_modref_c_finalize): summaries are NULL
18627 when incremental LTO linking happens.
18629 2021-04-21 Martin Liska <mliska@suse.cz>
18631 * lto-wrapper.c (run_gcc): When -flto=jobserver is used, but the
18632 makeserver cannot be detected, then use -flto=N fallback.
18634 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
18636 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): When cross-compiling,
18637 default to yes for aarch64-linux-gnu.
18638 * configure: Regenerate.
18640 2021-04-21 Martin Liska <mliska@suse.cz>
18642 * lto-wrapper.c (cpuset_popcount): Remove.
18643 (init_num_threads): Remove and use hardware_concurrency.
18645 2021-04-21 Martin Liska <mliska@suse.cz>
18647 * config/i386/i386.c: Remove superfluous || TARGET_MACHO
18648 which remains to be '(... || 0)' and clang complains about it.
18649 * dwarf2out.c (AT_vms_delta): Declare conditionally.
18650 (add_AT_vms_delta): Likewise.
18651 * tree.c (fld_simplified_type): Use rather more common pattern
18652 for disabling of something (#if 0).
18653 (get_tree_code_name): Likewise.
18654 (verify_type_variant): Likewise.
18656 2021-04-21 Martin Liska <mliska@suse.cz>
18658 * config/i386/i386-expand.c (decide_alignment): Use newly named
18659 macro TARGET_CPU_P.
18660 * config/i386/i386.c (ix86_decompose_address): Likewise.
18661 (ix86_address_cost): Likewise.
18662 (ix86_lea_outperforms): Likewise.
18663 (ix86_avoid_lea_for_addr): Likewise.
18664 (ix86_add_stmt_cost): Likewise.
18665 * config/i386/i386.h (TARGET_*): Remove.
18666 (TARGET_CPU_P): New macro.
18667 * config/i386/i386.md: Use newly named macro TARGET_CPU_P.
18668 * config/i386/x86-tune-sched-atom.c (do_reorder_for_imul): Likewise.
18669 (swap_top_of_ready_list): Likewise.
18670 (ix86_atom_sched_reorder): Likewise.
18671 * config/i386/x86-tune-sched-bd.c (ix86_bd_has_dispatch): Likewise.
18672 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Likewise.
18674 2021-04-21 Martin Liska <mliska@suse.cz>
18676 * config/i386/i386-options.c (TARGET_EXPLICIT_NO_SAHF_P):
18678 (SET_TARGET_NO_SAHF): Likewise.
18679 (TARGET_EXPLICIT_PREFETCH_SSE_P): Likewise.
18680 (SET_TARGET_PREFETCH_SSE): Likewise.
18681 (TARGET_EXPLICIT_NO_TUNE_P): Likewise.
18682 (SET_TARGET_NO_TUNE): Likewise.
18683 (TARGET_EXPLICIT_NO_80387_P): Likewise.
18684 (SET_TARGET_NO_80387): Likewise.
18686 * config/i386/i386.h (TARGET_*): Remove.
18687 * opth-gen.awk: Generate new used macros.
18689 2021-04-21 Martin Liska <mliska@suse.cz>
18691 * config/i386/i386.h (PTA_*): Remove.
18692 (enum pta_flag): New.
18693 (DEF_PTA): Generate PTA_* values from i386-isa.def.
18694 * config/i386/i386-isa.def: New file.
18696 2021-04-21 Alex Coplan <alex.coplan@arm.com>
18699 * config/aarch64/aarch64-bti-insert.c (aarch64_bti_j_insn_p): New.
18700 (rest_of_insert_bti): Avoid inserting duplicate bti j insns for
18701 jump table targets.
18703 2021-04-21 H.J. Lu <hjl.tools@gmail.com>
18705 * config.gcc: Install mwaitintrin.h for i[34567]86-*-* and
18706 x86_64-*-* targets.
18707 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_MWAIT_SET):
18709 (OPTION_MASK_ISA2_MWAIT_UNSET): Likewise.
18710 (ix86_handle_option): Handle -mmwait.
18711 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
18712 Replace OPTION_MASK_ISA_SSE3 with OPTION_MASK_ISA2_MWAIT on
18713 __builtin_ia32_monitor and __builtin_ia32_mwait.
18714 * config/i386/i386-options.c (isa2_opts): Add -mmwait.
18715 (ix86_valid_target_attribute_inner_p): Likewise.
18716 (ix86_option_override_internal): Enable mwait/monitor
18717 instructions for -msse3.
18718 * config/i386/i386.h (TARGET_MWAIT): New.
18719 (TARGET_MWAIT_P): Likewise.
18720 * config/i386/i386.opt: Add -mmwait.
18721 * config/i386/mwaitintrin.h: New file.
18722 * config/i386/pmmintrin.h: Include <mwaitintrin.h>.
18723 * config/i386/sse.md (sse3_mwait): Replace TARGET_SSE3 with
18725 (@sse3_monitor_<mode>): Likewise.
18726 * config/i386/x86gprintrin.h: Include <mwaitintrin.h>.
18727 * doc/extend.texi: Document mwait target attribute.
18728 * doc/invoke.texi: Document -mmwait.
18730 2021-04-21 Martin Liska <mliska@suse.cz>
18732 * config/i386/i386-options.c (DEF_ENUM): Remove it.
18733 * config/i386/i386-opts.h (DEF_ENUM): Likewise.
18734 * config/i386/stringop.def (DEF_ENUM): Likewise.
18736 2021-04-21 Martin Liska <mliska@suse.cz>
18738 * tree-cfg.c (gimple_verify_flow_info): Use qD instead
18739 of print_generic_expr.
18741 2021-04-21 Jakub Jelinek <jakub@redhat.com>
18743 PR rtl-optimization/100148
18744 * cprop.c (constprop_register): Use next_nondebug_insn instead of
18747 2021-04-21 Martin Liska <mliska@suse.cz>
18750 * cgraphunit.c (cgraph_node::analyze): Remove duplicate
18751 free_dominance_info calls.
18753 2021-04-21 Richard Biener <rguenther@suse.de>
18755 * gimple-fold.c (maybe_fold_reference): Remove is_lhs
18756 parameter (and assume it to be false).
18757 (fold_gimple_assign): Adjust, remove all callers of
18758 maybe_fold_reference calling it with is_lhs true.
18759 (gimple_fold_call): Likewise.
18760 (fold_stmt_1): Likewise.
18762 2021-04-21 Richard Biener <rguenther@suse.de>
18764 * fold-const.c (pedantic_non_lvalue_loc): Remove.
18765 (fold_binary_loc): Adjust.
18766 (fold_ternary_loc): Likewise.
18768 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
18770 PR middle-end/100130
18771 * varasm.c (get_block_for_decl): Make sure that any use of the
18772 retain attribute matches the section's retain flag.
18773 (switch_to_section): Check for retain mismatches even when
18774 changing sections, but do not warn if the given decl is the
18775 section's named.decl.
18776 (output_object_block): Pass the first decl in the block (if any)
18777 to switch_to_section.
18779 2021-04-20 H.J. Lu <hjl.tools@gmail.com>
18781 * config/i386/i386-c.c (ix86_target_macros_internal): Define
18782 __CRC32__ for -mcrc32.
18783 * config/i386/i386-options.c (ix86_option_override_internal):
18784 Enable crc32 instruction for -msse4.2.
18785 * config/i386/i386.md (sse4_2_crc32<mode>): Remove TARGET_SSE4_2
18787 (sse4_2_crc32di): Likewise.
18788 * config/i386/ia32intrin.h: Use crc32 target option for CRC32
18791 2021-04-20 Segher Boessenkool <segher@kernel.crashing.org>
18794 * config/rs6000/rs6000.c (rs6000_machine_from_flags): Do not consider
18797 2021-04-20 Martin Liska <mliska@suse.cz>
18799 * doc/invoke.texi: Fix typo.
18800 * params.opt: Likewise.
18802 2021-04-20 Martin Liska <mliska@suse.cz>
18804 * doc/invoke.texi: Document new param.
18806 2021-04-19 Andrew MacLeod <amacleod@redhat.com>
18808 PR tree-optimization/100081
18809 * gimple-range-cache.h (ranger_cache): Inherit from gori_compute
18810 rather than gori_compute_cache.
18811 * gimple-range-gori.cc (is_gimple_logical_p): Move to top of file.
18812 (range_def_chain::m_logical_depth): New member.
18813 (range_def_chain::range_def_chain): Initialize m_logical_depth.
18814 (range_def_chain::get_def_chain): Don't build defchains through more
18815 than LOGICAL_LIMIT logical expressions.
18816 * params.opt (param_ranger_logical_depth): New.
18818 2021-04-19 Richard Earnshaw <rearnsha@arm.com>
18821 * config/arm/arm.c (arm_configure_build_target): Do not strip
18822 extended FPU/SIMD feature bits from the target ISA when -mfpu
18823 is specified (partial revert of r11-8168).
18825 2021-04-19 Thomas Schwinge <thomas@codesourcery.com>
18827 * params.opt (-param=openacc-kernels=): Add.
18828 * omp-oacc-kernels-decompose.cc
18829 (pass_omp_oacc_kernels_decompose::gate): Use it.
18830 * doc/invoke.texi (-fopenacc-kernels=@var{mode}): Move...
18831 (--param): ... here, 'openacc-kernels'.
18833 2021-04-19 Martin Liska <mliska@suse.cz>
18836 * gengtype.c (finish_root_table): Align function arguments
18837 in between declaration and definition.
18839 2021-04-19 Eric Botcazou <ebotcazou@adacore.com>
18841 * config/i386/winnt.c (i386_pe_seh_cold_init): Properly deal with
18842 frames larger than the SEH maximum frame size.
18844 2021-04-18 Segher Boessenkool <segher@kernel.crashing.org>
18846 PR rtl-optimization/99927
18847 * combine.c (distribute_notes) [REG_UNUSED]: If the register already
18848 is dead, just drop it.
18850 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
18853 * config/i386/winnt-d.c (TARGET_D_TEMPLATES_ALWAYS_COMDAT): Define.
18854 * doc/tm.texi: Regenerate.
18855 * doc/tm.texi.in (D language and ABI): Add @hook for
18856 TARGET_D_TEMPLATES_ALWAYS_COMDAT.
18858 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
18860 * config/darwin-d.c (darwin_d_handle_target_object_format): New
18862 (darwin_d_register_target_info): New function.
18863 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18864 * config/dragonfly-d.c (dragonfly_d_handle_target_object_format): New
18866 (dragonfly_d_register_target_info): New function.
18867 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18868 * config/freebsd-d.c (freebsd_d_handle_target_object_format): New
18870 (freebsd_d_register_target_info): New function.
18871 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18872 * config/glibc-d.c (glibc_d_handle_target_object_format): New
18874 (glibc_d_register_target_info): New function.
18875 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18876 * config/i386/i386-d.c (ix86_d_handle_target_object_format): New
18878 (ix86_d_register_target_info): Add ix86_d_handle_target_object_format
18879 as handler for objectFormat key.
18880 * config/i386/winnt-d.c (winnt_d_handle_target_object_format): New
18882 (winnt_d_register_target_info): New function.
18883 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18884 * config/netbsd-d.c (netbsd_d_handle_target_object_format): New
18886 (netbsd_d_register_target_info): New function.
18887 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18888 * config/openbsd-d.c (openbsd_d_handle_target_object_format): New
18890 (openbsd_d_register_target_info): New function.
18891 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18892 * config/pa/pa-d.c (pa_d_handle_target_object_format): New function.
18893 (pa_d_register_target_info): Add pa_d_handle_target_object_format as
18894 handler for objectFormat key.
18895 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_object_format): New
18897 (rs6000_d_register_target_info): Add
18898 rs6000_d_handle_target_object_format as handler for objectFormat key.
18899 * config/sol2-d.c (solaris_d_handle_target_object_format): New
18901 (solaris_d_register_target_info): New function.
18902 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
18904 2021-04-16 Jakub Jelinek <jakub@redhat.com>
18907 * config/aarch64/aarch64.c (aarch64_function_arg_alignment): Change
18908 abi_break argument from bool * to unsigned *, store there the pre-GCC 9
18910 (aarch64_layout_arg, aarch64_gimplify_va_arg_expr): Adjust callers.
18911 (aarch64_function_arg_regno_p): Likewise. Only emit -Wpsabi note if
18912 the old and new alignment after applying MIN/MAX to it is different.
18914 2021-04-16 Tamar Christina <tamar.christina@arm.com>
18917 * config/aarch64/aarch64-sve.md (@aarch64_sve_trn1_conv<mode>): New.
18918 * config/aarch64/aarch64.c (aarch64_expand_sve_const_pred_trn): Use new
18920 * config/aarch64/iterators.md (UNSPEC_TRN1_CONV): New.
18922 2021-04-16 Bill Schmidt <wschmidt@linux.ibm.com>
18924 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Revise
18925 this section and its subsections.
18927 2021-04-16 Jakub Jelinek <jakub@redhat.com>
18930 * config/aarch64/aarch64.md (*neg_asr_si2_extr, *extrsi5_insn_di): New
18931 define_insn patterns.
18933 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
18935 PR rtl-optimization/98689
18936 * reg-notes.def (UNTYPED_CALL): New note.
18937 * combine.c (distribute_notes): Handle it.
18938 * emit-rtl.c (try_split): Likewise.
18939 * rtlanal.c (rtx_properties::try_to_add_insn): Likewise. Assume
18940 that calls with the note implicitly set all return value registers.
18941 * builtins.c (expand_builtin_apply): Add a REG_UNTYPED_CALL
18944 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
18946 PR rtl-optimization/99596
18947 * rtlanal.c (rtx_properties::try_to_add_insn): Don't add global
18948 register accesses for const calls. Assume that pure functions
18949 can only read from global registers. Ignore cases in which
18950 the stack pointer has been marked global.
18952 2021-04-16 Jakub Jelinek <jakub@redhat.com>
18955 * tree-vect-loop.c (vect_transform_loop): Don't remove just
18956 dead scalar .MASK_LOAD calls, but also dead .COND_* calls - replace
18957 them by their last argument.
18959 2021-04-15 Martin Liska <mliska@suse.cz>
18961 * doc/invoke.texi: Other params don't use it, remove it.
18963 2021-04-15 Richard Biener <rguenther@suse.de>
18965 * gimple-builder.h: Add deprecation note.
18967 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
18970 * attribs.h (restrict_type_identity_attributes_to): Declare.
18971 * attribs.c (restrict_type_identity_attributes_to): New function.
18973 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
18976 * attribs.h (affects_type_identity_attributes): Declare.
18977 * attribs.c (remove_attributes_matching): New function.
18978 (affects_type_identity_attributes): Likewise.
18980 2021-04-15 Jakub Jelinek <jakub@redhat.com>
18983 * config/aarch64/aarch64.md (*<LOGICAL:optab>_<SHIFT:optab><mode>3):
18984 Add combine splitters for *<LOGICAL:optab>_ashl<mode>3 with
18985 ZERO_EXTEND, SIGN_EXTEND or AND.
18987 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
18989 PR rtl-optimization/99929
18990 * rtl.h (same_vector_encodings_p): New function.
18991 * cse.c (exp_equiv_p): Check that CONST_VECTORs have the same encoding.
18992 * cselib.c (rtx_equal_for_cselib_1): Likewise.
18993 * jump.c (rtx_renumbered_equal_p): Likewise.
18994 * lra-constraints.c (operands_match_p): Likewise.
18995 * reload.c (operands_match_p): Likewise.
18996 * rtl.c (rtx_equal_p_cb, rtx_equal_p): Likewise.
18998 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
19000 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Print
19001 more information about variable-length CONST_VECTORs.
19003 2021-04-14 Vladimir N. Makarov <vmakarov@redhat.com>
19005 PR rtl-optimization/100066
19006 * lra-constraints.c (split_reg): Check paradoxical_subreg_p for
19007 ordered modes when choosing splitting mode for hard reg.
19009 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
19012 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector_sel):
19014 (aarch64_expand_sve_const_vector): Use it for nelts_per_pattern==2.
19016 2021-04-14 Andreas Krebbel <krebbel@linux.ibm.com>
19018 * config/s390/s390-builtins.def (O_M5, O_M12, ...): Add new macros
19019 for mask operand types.
19020 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
19021 (s390_vec_permi_dbl, s390_vpdi): Use the M5 type for the immediate
19023 (s390_vec_msum_u128, s390_vmslg): Use the M12 type for the
19025 * config/s390/s390.c (s390_const_operand_ok): Check the new
19026 operand types and generate a list of valid values.
19028 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
19030 * doc/tm.texi: Regenerate.
19031 * doc/tm.texi.in (D language and ABI): Add @hook for
19032 TARGET_D_REGISTER_OS_TARGET_INFO.
19034 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
19036 * config/aarch64/aarch64-d.c (aarch64_d_handle_target_float_abi): New
19038 (aarch64_d_register_target_info): New function.
19039 * config/aarch64/aarch64-protos.h (aarch64_d_register_target_info):
19041 * config/aarch64/aarch64.h (TARGET_D_REGISTER_CPU_TARGET_INFO):
19043 * config/arm/arm-d.c (arm_d_handle_target_float_abi): New function.
19044 (arm_d_register_target_info): New function.
19045 * config/arm/arm-protos.h (arm_d_register_target_info): Declare.
19046 * config/arm/arm.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19047 * config/i386/i386-d.c (ix86_d_handle_target_float_abi): New function.
19048 (ix86_d_register_target_info): New function.
19049 * config/i386/i386-protos.h (ix86_d_register_target_info): Declare.
19050 * config/i386/i386.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19051 * config/mips/mips-d.c (mips_d_handle_target_float_abi): New function.
19052 (mips_d_register_target_info): New function.
19053 * config/mips/mips-protos.h (mips_d_register_target_info): Declare.
19054 * config/mips/mips.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19055 * config/pa/pa-d.c (pa_d_handle_target_float_abi): New function.
19056 (pa_d_register_target_info): New function.
19057 * config/pa/pa-protos.h (pa_d_register_target_info): Declare.
19058 * config/pa/pa.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19059 * config/riscv/riscv-d.c (riscv_d_handle_target_float_abi): New
19061 (riscv_d_register_target_info): New function.
19062 * config/riscv/riscv-protos.h (riscv_d_register_target_info): Declare.
19063 * config/riscv/riscv.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19064 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_float_abi): New
19066 (rs6000_d_register_target_info): New function.
19067 * config/rs6000/rs6000-protos.h (rs6000_d_register_target_info):
19069 * config/rs6000/rs6000.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19070 * config/s390/s390-d.c (s390_d_handle_target_float_abi): New function.
19071 (s390_d_register_target_info): New function.
19072 * config/s390/s390-protos.h (s390_d_register_target_info): Declare.
19073 * config/s390/s390.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19074 * config/sparc/sparc-d.c (sparc_d_handle_target_float_abi): New
19076 (sparc_d_register_target_info): New function.
19077 * config/sparc/sparc-protos.h (sparc_d_register_target_info): Declare.
19078 * config/sparc/sparc.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
19079 * doc/tm.texi: Regenerate.
19080 * doc/tm.texi.in (D language and ABI): Add @hook for
19081 TARGET_D_REGISTER_CPU_TARGET_INFO.
19083 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
19085 * config/i386/i386-d.c (ix86_d_has_stdcall_convention): New function.
19086 * config/i386/i386-protos.h (ix86_d_has_stdcall_convention): Declare.
19087 * config/i386/i386.h (TARGET_D_HAS_STDCALL_CONVENTION): Define.
19088 * doc/tm.texi: Regenerate.
19089 * doc/tm.texi.in (D language and ABI): Add @hook for
19090 TARGET_D_HAS_STDCALL_CONVENTION.
19092 2021-04-14 Richard Biener <rguenther@suse.de>
19094 * tree-cfg.c (verify_gimple_assign_ternary): Verify that
19095 VEC_COND_EXPRs have a gimple_val condition.
19096 * tree-ssa-propagate.c (valid_gimple_rhs_p): VEC_COND_EXPR
19097 can no longer have a GENERIC condition.
19099 2021-04-14 Richard Earnshaw <rearnsha@arm.com>
19102 * config/arm/arm.c (arm_configure_build_target): Strip isa_all_fpbits
19103 from the isa_delta when -mfpu has been used.
19104 (arm_options_perform_arch_sanity_checks): It's the architecture that
19105 lacks an FPU not the processor.
19107 2021-04-13 Richard Biener <rguenther@suse.de>
19109 PR tree-optimization/100053
19110 * tree-ssa-sccvn.c (vn_nary_op_get_predicated_value): Do
19111 not use optimistic dominance queries for backedges to validate
19113 (dominated_by_p_w_unex): Add parameter to ignore executable
19114 state on backedges.
19115 (rpo_elim::eliminate_avail): Adjust.
19117 2021-04-13 Jakub Jelinek <jakub@redhat.com>
19120 * config/aarch64/aarch64.md (*aarch64_bfxil<mode>_extr,
19121 *aarch64_bfxilsi_extrdi): New define_insn patterns.
19123 2021-04-13 Jakub Jelinek <jakub@redhat.com>
19126 * simplify-rtx.c (simplify_immed_subreg): For MODE_COMPOSITE_P
19127 outermode, return NULL if the result doesn't encode back to the
19128 original byte sequence.
19129 (simplify_gen_subreg): Don't create SUBREGs from constants to
19130 MODE_COMPOSITE_P outermode.
19132 2021-04-12 Jakub Jelinek <jakub@redhat.com>
19134 PR rtl-optimization/99905
19135 * combine.c (expand_compound_operation): If pos + len > modewidth,
19136 perform the right shift by pos in inner_mode and then convert to mode,
19137 instead of trying to simplify a shift of rtx with inner_mode by pos
19138 as if it was a shift in mode.
19140 2021-04-12 Jakub Jelinek <jakub@redhat.com>
19143 * combine.c (simplify_and_const_int_1): Don't optimize varop
19144 away if it has side-effects.
19146 2021-04-12 Martin Liska <mliska@suse.cz>
19148 * doc/extend.texi: Escape @smallexample content.
19150 2021-04-12 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
19152 * config/s390/s390.md ("*movdi_31", "*movdi_64"): Add
19153 alternative in order to load a DFP zero.
19155 2021-04-12 Martin Liska <mliska@suse.cz>
19157 * doc/extend.texi: Be more precise in documentation
19158 of symver attribute.
19160 2021-04-12 Martin Liska <mliska@suse.cz>
19163 * gimplify.c (gimplify_expr): Right now, we unpoison all
19164 variables before a goto <dest>. We should not do it if we are
19167 2021-04-12 Cui,Lili <lili.cui@intel.com>
19169 * common/config/i386/cpuinfo.h (get_intel_cpu): Handle
19171 * common/config/i386/i386-common.c (processor_names): Add
19173 (processor_alias_table): Add rocketlake.
19174 * common/config/i386/i386-cpuinfo.h (processor_subtypes): Add
19175 INTEL_COREI7_ROCKETLAKE.
19176 * config.gcc: Add -march=rocketlake.
19177 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19179 * config/i386/i386-options.c (m_ROCKETLAKE) : Define.
19180 (processor_cost_table): Add rocketlake cost.
19181 * config/i386/i386.h (ix86_size_cost) : Define
19183 (processor_type) : Add PROCESSOR_ROCKETLAKE.
19184 (PTA_ROCKETLAKE): Ditto.
19185 * doc/extend.texi: Add rocketlake.
19186 * doc/invoke.texi: Add rocketlake.
19188 2021-04-12 Cui,Lili <lili.cui@intel.com>
19190 * config/i386/i386.h (PTA_ALDERLAKE): Change alderlake ISA list.
19191 * config/i386/i386-options.c (m_CORE_AVX2): Add m_ALDERLAKE.
19192 * common/config/i386/cpuinfo.h (get_intel_cpu): Add AlderLake model.
19193 * doc/invoke.texi: Change alderlake ISA list.
19195 2021-04-11 Hafiz Abid Qadeer <abidh@codesourcery.com>
19197 PR middle-end/98088
19198 * omp-expand.c (expand_oacc_collapse_init): Update condition in
19201 2021-04-10 H.J. Lu <hjl.tools@gmail.com>
19204 * config/i386/serializeintrin.h (_serialize): Defined as macro.
19206 2021-04-10 Jakub Jelinek <jakub@redhat.com>
19209 * expr.c (expand_expr_addr_expr_1): Test is_global_var rather than
19210 just TREE_STATIC on COMPOUND_LITERAL_EXPR_DECLs.
19212 2021-04-10 Jakub Jelinek <jakub@redhat.com>
19214 PR middle-end/99989
19215 * gimple-ssa-warn-alloca.c
19216 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit to
19217 0 with integer precision unconditionally.
19219 2021-04-10 Jakub Jelinek <jakub@redhat.com>
19221 PR rtl-optimization/98601
19222 * rtlanal.c (rtx_addr_can_trap_p_1): Allow in assert unknown size
19223 not just for BLKmode, but also for VOIDmode. For STRICT_ALIGNMENT
19224 unaligned_mems handle VOIDmode like BLKmode.
19226 2021-04-10 Jan Hubicka <hubicka@ucw.cz>
19229 * tree.c (free_lang_data_in_decl): Do not release body of
19230 declare_variant_alt.
19232 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
19234 * config/aarch64/aarch64.c (aarch64_option_restore): If the
19235 architecture was specified explicitly and the tuning wasn't,
19236 tune for the architecture rather than the configured default CPU.
19238 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
19240 * config/aarch64/aarch64.md (tlsdesc_small_sve_<mode>): Use X30
19241 as the temporary register.
19243 2021-04-09 Martin Liska <mliska@suse.cz>
19245 * doc/extend.texi: Move non-target attributes on the top level.
19247 2021-04-09 Martin Liska <mliska@suse.cz>
19249 * doc/invoke.texi: Document minimum and maximum value of the
19250 argument for both supported compression algorithms.
19252 2021-04-08 David Edelsohn <dje.gcc@gmail.com>
19254 * config/rs6000/rs6000.c (rs6000_xcoff_select_section): Select
19255 TLS BSS before TLS data.
19256 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .comm.
19258 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
19260 * doc/sourcebuild.texi (stdint_types_mbig_endian): Document.
19262 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
19264 * match.pd: Extend vec_cond folds to handle shifts.
19266 2021-04-08 Maciej W. Rozycki <macro@orcam.me.uk>
19268 * config/vax/vax.md: Fix comment for `*bit<mode>' pattern's
19271 2021-04-08 Alex Coplan <alex.coplan@arm.com>
19274 * config/arm/iterators.md (MVE_vecs): New.
19275 (V_elem): Also handle V2DF.
19276 * config/arm/mve.md (*mve_mov<mode>): Rename to ...
19277 (*mve_vdup<mode>): ... this. Remove second alternative since
19278 vec_duplicate of const_int is not canonical RTL, and we don't
19279 want to match symbol_refs.
19280 (*mve_vec_duplicate<mode>): Delete (pattern is redundant).
19282 2021-04-08 Xionghu Luo <luoxhu@linux.ibm.com>
19284 * fold-const.c (fold_single_bit_test): Fix typo.
19285 * print-rtl.c (print_rtx_insn_vec): Call print_rtl_single
19288 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
19290 PR tree-optimization/97513
19291 * tree-vect-slp.c (vect_add_slp_permutation): New function,
19293 (vectorizable_slp_permutation): ...here. Detect cases in which
19294 all VEC_PERM_EXPRs are guaranteed to have the same stepped
19295 permute vector and only generate one permute vector for that case.
19296 Extend that case to handle variable-length vectors.
19298 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
19300 PR tree-optimization/99873
19301 * tree-vect-slp.c (vect_slp_prefer_store_lanes_p): New function.
19302 (vect_build_slp_instance): Don't split store groups that could
19303 use IFN_STORE_LANES.
19305 2021-04-07 Jakub Jelinek <jakub@redhat.com>
19308 * varasm.c (output_constant_pool_contents): Don't strip name encoding
19309 from XSTR (desc->sym, 0) or from label before passing those to
19312 2021-04-07 Richard Biener <rguenther@suse.de>
19314 PR tree-optimization/99954
19315 * tree-loop-distribution.c: Include tree-affine.h.
19316 (generate_memcpy_builtin): Try using tree-affine to prove
19318 (loop_distribution::classify_builtin_ldst): Always classify
19321 2021-04-07 Richard Biener <rguenther@suse.de>
19323 PR tree-optimization/99947
19324 * tree-vect-loop.c (vectorizable_induction): Pre-allocate
19325 steps vector to avoid pushing elements from the reallocated
19328 2021-04-07 Richard Biener <rguenther@suse.de>
19330 * tree-ssa-sccvn.h (print_vn_reference_ops): Declare.
19331 * tree-ssa-pre.c (print_pre_expr): Factor out VN reference operand
19333 * tree-ssa-sccvn.c (print_vn_reference_ops): ... into this new
19335 (debug_vn_reference_ops): New.
19337 2021-04-07 Bin Cheng <bin.cheng@linux.alibaba.com>
19339 PR tree-optimization/98736
19340 * tree-loop-distribution.c
19341 * (loop_distribution::bb_top_order_init):
19342 Compute RPO with programing order preserved by calling function
19343 rev_post_order_and_mark_dfs_back_seme.
19345 2021-04-06 Vladimir N. Makarov <vmakarov@redhat.com>
19348 * lra-constraints.c (split_reg): Don't check paradoxical_subreg_p.
19349 * lra-lives.c (clear_sparseset_regnos, regnos_in_sparseset_p): New
19351 (process_bb_lives): Don't update biggest mode of hard reg for
19352 implicit in multi-register group. Use the new functions for
19353 updating dead_set and unused_set by register notes.
19355 2021-04-06 Xianmiao Qu <xianmiao_qu@c-sky.com>
19357 * config/csky/csky_pipeline_ck802.md : Use insn reservation name
19360 2021-04-06 H.J. Lu <hjl.tools@gmail.com>
19362 * config/i386/x86-tune-costs.h (skylake_memcpy): Updated.
19363 (skylake_memset): Likewise.
19364 (skylake_cost): Change CLEAR_RATIO to 17.
19365 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
19366 Replace m_CANNONLAKE, m_ICELAKE_CLIENT, m_ICELAKE_SERVER,
19367 m_TIGERLAKE and m_SAPPHIRERAPIDS with m_SKYLAKE and m_CORE_AVX512.
19369 2021-04-06 Richard Biener <rguenther@suse.de>
19371 PR tree-optimization/99880
19372 * tree-vect-loop.c (maybe_set_vectorized_backedge_value): Only
19373 set vectorized defs of relevant PHIs.
19375 2021-04-06 Richard Biener <rguenther@suse.de>
19377 PR tree-optimization/99924
19378 * tree-vect-slp.c (vect_bb_partition_graph_r): Do not mark
19379 nodes w/o scalar stmts as visited.
19381 2021-04-06 Alex Coplan <alex.coplan@arm.com>
19384 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Also use base
19385 PCS for [su]fix_optab.
19387 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
19389 * config/darwin.c (machopic_legitimize_pic_address): Check
19390 that the current pic register is one of the hard reg set
19391 before setting liveness.
19393 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
19395 * config/darwin.c (machopic_legitimize_pic_address): Fix
19396 whitespace, remove unused code.
19398 2021-04-03 Jakub Jelinek <jakub@redhat.com>
19400 PR tree-optimization/99882
19401 * gimple-ssa-store-merging.c (bswap_view_convert): Handle val with
19404 2021-04-03 Jakub Jelinek <jakub@redhat.com>
19406 PR rtl-optimization/99863
19407 * dse.c (replace_read): Drop regs_live argument. Instead of
19408 regs_live, use store_insn->fixed_regs_live if non-NULL,
19409 otherwise punt if insns sequence clobbers or sets any hard
19412 2021-04-03 Jakub Jelinek <jakub@redhat.com>
19415 * targhooks.h (default_print_patchable_function_entry_1): Declare.
19416 * targhooks.c (default_print_patchable_function_entry_1): New function,
19417 copied from default_print_patchable_function_entry with an added flags
19419 (default_print_patchable_function_entry): Rewritten into a small
19420 wrapper around default_print_patchable_function_entry_1.
19421 * config/rs6000/rs6000.c (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
19423 (rs6000_print_patchable_function_entry): New function.
19425 2021-04-02 Eric Botcazou <ebotcazou@adacore.com>
19427 * doc/invoke.texi (fdelete-dead-exceptions): Minor tweak.
19429 2021-04-01 Jason Merrill <jason@redhat.com>
19432 * common.opt: Document v15 and v16.
19434 2021-04-01 Richard Biener <rguenther@suse.de>
19436 PR tree-optimization/99863
19437 * gimplify.c (gimplify_init_constructor): Recompute vector
19440 2021-04-01 Jakub Jelinek <jakub@redhat.com>
19442 * doc/extend.texi (symver attribute): Fix up syntax errors
19445 2021-04-01 Jakub Jelinek <jakub@redhat.com>
19447 PR tree-optimization/96573
19448 * gimple-ssa-store-merging.c (init_symbolic_number): Handle
19449 also pointer types.
19451 2021-04-01 Richard Biener <rguenther@suse.de>
19453 PR tree-optimization/99856
19454 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Promote
19455 precision to vector element precision.
19457 2021-04-01 Martin Jambor <mjambor@suse.cz>
19459 PR tree-optimization/97009
19460 * tree-sra.c (access_or_its_child_written): New function.
19461 (propagate_subaccesses_from_rhs): Use it instead of a simple grp_write
19464 2021-03-31 Jan Hubicka <hubicka@ucw.cz>
19467 * cif-code.def (USES_COMDAT_LOCAL): Make CIF_FINAL_NORMAL.
19469 2021-03-31 Pat Haugen <pthaugen@linux.ibm.com>
19472 * config/rs6000/altivec.md (xxspltiw_v4si, xxspltiw_v4sf_inst,
19473 xxspltidp_v2df_inst, xxsplti32dx_v4si_inst, xxsplti32dx_v4sf_inst,
19474 xxblend_<mode>, xxpermx_inst, xxeval): Mark prefixed.
19475 * config/rs6000/mma.md (mma_<vvi4i4i8>, mma_<avvi4i4i8>,
19476 mma_<vvi4i4i2>, mma_<avvi4i4i2>, mma_<vvi4i4>, mma_<avvi4i4>,
19477 mma_<pvi4i2>, mma_<apvi4i2>, mma_<vvi4i4i4>, mma_<avvi4i4i4>):
19479 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Adjust test.
19480 * config/rs6000/rs6000.md (define_attr "maybe_prefixed"): New.
19481 (define_attr "prefixed"): Update initializer.
19483 2021-03-31 Jakub Jelinek <jakub@redhat.com>
19486 * dwarf2out.c (debug_ranges_dwo_section): New variable.
19487 (DW_RANGES_IDX_SKELETON): Define.
19488 (struct dw_ranges): Add begin_entry and end_entry members.
19489 (DEBUG_DWO_RNGLISTS_SECTION): Define.
19490 (add_ranges_num): Adjust r initializer for addition of *_entry
19492 (add_ranges_by_labels): For -gsplit-dwarf and force_direct,
19493 set idx to DW_RANGES_IDX_SKELETON.
19494 (use_distinct_base_address_for_range): New function.
19495 (index_rnglists): Don't set r->idx if it is equal to
19496 DW_RANGES_IDX_SKELETON. Initialize r->begin_entry and
19497 r->end_entry for -gsplit-dwarf if those will be needed by
19499 (output_rnglists): Add DWO argument. If true, switch to
19500 debug_ranges_dwo_section rather than debug_ranges_section.
19501 Adjust l1/l2 label indexes. Only output the offset table when
19502 dwo is true and don't include in there the skeleton range
19503 entry if present. For -gsplit-dwarf, skip ranges that belong
19504 to the other rnglists section. Change return type from void
19505 to bool and return true if there are any range entries for
19506 the other section. For dwarf_split_debug_info use
19507 DW_RLE_startx_endx, DW_RLE_startx_length and DW_RLE_base_addressx
19508 entries instead of DW_RLE_start_end, DW_RLE_start_length and
19509 DW_RLE_base_address. Use use_distinct_base_address_for_range.
19510 (init_sections_and_labels): Initialize debug_ranges_dwo_section
19511 if -gsplit-dwarf and DWARF >= 5. Adjust ranges_section_label
19512 and range_base_label indexes.
19513 (dwarf2out_finish): Call index_rnglists earlier before finalizing
19514 .debug_addr. Never emit DW_AT_rnglists_base attribute. For
19515 -gsplit-dwarf and DWARF >= 5 call output_rnglists up to twice
19516 with different dwo arguments.
19517 (dwarf2out_c_finalize): Clear debug_ranges_dwo_section.
19519 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19521 PR tree-optimization/98268
19522 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Call
19523 recompute_tree_invariant_for_addr_expr after successfully
19524 folding a TARGET_MEM_REF that occurs inside an ADDR_EXPR.
19526 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19528 PR tree-optimization/99726
19529 * tree-data-ref.c (create_intersect_range_checks_index): Bail
19530 out if there is more than one access function SCEV for the loop
19533 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19535 PR rtl-optimization/97141
19536 PR rtl-optimization/98726
19537 * emit-rtl.c (valid_for_const_vector_p): Return true for
19539 * rtx-vector-builder.h (rtx_vector_builder::step): Return a
19540 poly_wide_int instead of a wide_int.
19541 (rtx_vector_builder::apply_set): Take a poly_wide_int instead
19543 * rtx-vector-builder.c (rtx_vector_builder::apply_set): Likewise.
19544 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Return
19545 false for CONST_VECTORs that cannot be forced to memory.
19546 * config/aarch64/aarch64-simd.md (mov<mode>): If a CONST_VECTOR
19547 is too complex to force to memory, build it up from individual
19550 2021-03-31 Jan Hubicka <jh@suse.cz>
19553 * cgraph.c (cgraph_node::release_body): Fix overactive check.
19555 2021-03-31 Christophe Lyon <christophe.lyon@linaro.org>
19558 * config/arm/vec-common.md (mul<mode>3): Disable on iwMMXT, expect
19561 2021-03-31 H.J. Lu <hjl.tools@gmail.com>
19563 * config/i386/i386-expand.c (expand_set_or_cpymem_via_rep):
19564 For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, don't convert QImode
19566 (decide_alg): For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, use
19567 "rep movsb/stosb" only for known sizes.
19568 * config/i386/i386-options.c (processor_cost_table): Use Ice
19569 Lake cost for Cannon Lake, Ice Lake, Tiger Lake, Sapphire
19570 Rapids and Alder Lake.
19571 * config/i386/i386.h (TARGET_PREFER_KNOWN_REP_MOVSB_STOSB): New.
19572 * config/i386/x86-tune-costs.h (icelake_memcpy): New.
19573 (icelake_memset): Likewise.
19574 (icelake_cost): Likewise.
19575 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
19578 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
19581 * config/aarch64/aarch64.c
19582 (aarch64_vectorize_preferred_vector_alignment): Query the size
19583 of the provided SVE vector; do not assume that all SVE vectors
19584 have the same size.
19586 2021-03-31 Jan Hubicka <jh@suse.cz>
19589 * cgraph.c (cgraph_node::release_body): Remove all callers and
19591 * cgraphclones.c (cgraph_node::materialize_clone): Do not do it here.
19592 * cgraphunit.c (cgraph_node::expand): And here.
19594 2021-03-31 Martin Liska <mliska@suse.cz>
19596 * ipa-modref.c (analyze_ssa_name_flags): Fix coding style
19597 and one negated condition.
19599 2021-03-31 Jakub Jelinek <jakub@redhat.com>
19600 Richard Sandiford <richard.sandiford@arm.com>
19603 * config/aarch64/aarch64.md (*add<mode>3_poly_1): Swap Uai and Uav
19604 constraints on operands[2] and similarly 0 and rk constraints
19605 on operands[1] corresponding to that.
19607 2021-03-31 Jakub Jelinek <jakub@redhat.com>
19610 * configure.ac (HAVE_LD_BROKEN_PE_DWARF5): New AC_DEFINE if PECOFF
19611 linker doesn't support DWARF sections new in DWARF5.
19612 * config/i386/i386-options.c (ix86_option_override_internal): Default
19613 to dwarf_version 4 if HAVE_LD_BROKEN_PE_DWARF5 for TARGET_PECOFF
19615 * config.in: Regenerated.
19616 * configure: Regenerated.
19618 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19621 * config/aarch64/aarch64.c (aarch64_analyze_loop_vinfo): Check for
19622 available issue_info before using it.
19624 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19627 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): Do not allow zero
19630 2021-03-30 Xionghu Luo <luoxhu@linux.ibm.com>
19633 * config/rs6000/altivec.md (altivec_lvsl_reg): Change to ...
19634 (altivec_lvsl_reg_<mode>): ... this.
19635 (altivec_lvsr_reg): Change to ...
19636 (altivec_lvsr_reg_<mode>): ... this.
19637 * config/rs6000/predicates.md (vec_set_index_operand): New.
19638 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
19639 Enable 32bit variable vec_insert for all TARGET_VSX.
19640 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
19641 Enable 32bit variable vec_insert for p9 and above.
19642 (rs6000_expand_vector_set_var_p8): Rename to ...
19643 (rs6000_expand_vector_set_var_p7): ... this.
19644 (rs6000_expand_vector_set): Use TARGET_VSX and adjust assert
19646 * config/rs6000/vector.md (vec_set<mode>): Use vec_set_index_operand.
19647 * config/rs6000/vsx.md (xl_len_r): Use gen_altivec_lvsl_reg_di and
19648 gen_altivec_lvsr_reg_di.
19650 2021-03-30 H.J. Lu <hjl.tools@gmail.com>
19653 * config/i386/ia32intrin.h (__rdtsc): Defined as macro.
19654 (__rdtscp): Likewise.
19656 2021-03-30 Tamar Christina <tamar.christina@arm.com>
19658 PR tree-optimization/99825
19659 * tree-vect-slp-patterns.c (vect_check_evenodd_blend):
19660 Reject non-mult 2 lanes.
19662 2021-03-30 Richard Earnshaw <rearnsha@arm.com>
19665 * config/arm/arm.c (arm_file_start): Fix emission of
19666 Tag_ABI_VFP_args attribute.
19668 2021-03-30 Richard Biener <rguenther@suse.de>
19670 PR tree-optimization/99824
19671 * stor-layout.c (set_min_and_max_values_for_integral_type):
19672 Assert the precision is within the bounds of
19673 WIDE_INT_MAX_PRECISION.
19674 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
19675 the outermost component ref only to lower the access size
19676 and initialize that from the access type.
19678 2021-03-30 Richard Sandiford <richard.sandiford@arm.com>
19681 * config/aarch64/aarch64.md (mov<mode>): Pass multi-instruction
19682 CONST_INTs to aarch64_expand_mov_immediate when called after RA.
19684 2021-03-30 Mihailo Stojanovic <mihailo.stojanovic@typhoon-hil.com>
19686 * config/aarch64/aarch64.md
19687 (<optab>_trunc<fcvt_target><GPI:mode>2): Set the "arch"
19688 attribute to disambiguate between SIMD and FP variants of the
19691 2021-03-29 Jan Hubicka <hubicka@ucw.cz>
19693 * ipa-modref.c (merge_call_lhs_flags): Correct handling of deref.
19694 (analyze_ssa_name_flags): Fix typo in comment.
19696 2021-03-29 Alex Coplan <alex.coplan@arm.com>
19699 * config/aarch64/aarch64-sve-builtins.cc
19700 (function_builder::add_function): Add placeholder_p argument, use
19701 placeholder decls if this is set.
19702 (function_builder::add_unique_function): Instead of conditionally adding
19703 direct overloads, unconditionally add either a direct overload or a
19705 (function_builder::add_overloaded_function): Set placeholder_p if we're
19706 using C++ overloads. Use the obstack for string storage instead
19707 of relying on the tree nodes.
19708 (function_builder::add_overloaded_functions): Don't return early for
19709 m_direct_overloads: we need to add placeholders.
19710 * config/aarch64/aarch64-sve-builtins.h
19711 (function_builder::add_function): Add placeholder_p argument.
19713 2021-03-29 Richard Biener <rguenther@suse.de>
19715 PR tree-optimization/99807
19716 * tree-vect-slp.c (vect_slp_analyze_node_operations_1): Move
19717 assert below VEC_PERM handling.
19719 2021-03-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19722 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>): Use
19723 aarch64_simd_or_scalar_imm_zero to match zeroes. Remove pattern
19724 matching const_int 0.
19725 (move_lo_quad_internal_be_<mode>): Likewise.
19726 (move_lo_quad_<mode>): Update for the above.
19727 * config/aarch64/iterators.md (VQ_2E): Delete.
19729 2021-03-29 Jakub Jelinek <jakub@redhat.com>
19731 PR tree-optimization/99777
19732 * fold-const.c (extract_muldiv_1): For conversions, punt on casts from
19733 types other than scalar integral types.
19735 2021-03-28 David Edelsohn <dje.gcc@gmail.com>
19737 * config/rs6000/rs6000.c (rs6000_output_dwarf_dtprel): Do not add
19738 XCOFF TLS reloc decorations.
19740 2021-03-28 Gerald Pfeifer <gerald@pfeifer.com>
19742 * doc/analyzer.texi (Analyzer Internals): Update link to
19743 "A Memory Model for Static Analysis of C Programs".
19745 2021-03-26 David Edelsohn <dje.gcc@gmail.com>
19747 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Call function.
19748 * config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align):
19750 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align): New.
19751 (rs6000_special_round_type_align): Recursively check innermost first
19754 2021-03-26 Jakub Jelinek <jakub@redhat.com>
19757 * dwarf2out.h (struct dw_fde_node): Add rule18 member.
19758 * dwarf2cfi.c (dwarf2out_frame_debug_expr): When handling (set hfp sp)
19759 assignment with drap_reg active, queue reg save for hfp with offset 0
19760 and flush queued reg saves. When handling a push with rule18,
19761 defer queueing reg save for hfp and just assert the offset is 0.
19762 (scan_trace): Assert that fde->rule18 is false.
19764 2021-03-26 Vladimir Makarov <vmakarov@redhat.com>
19767 * ira-costs.c (record_reg_classes): Put case with
19768 CT_RELAXED_MEMORY adjacent to one with CT_MEMORY.
19769 * ira.c (ira_setup_alts): Ditto.
19770 * lra-constraints.c (process_alt_operands): Ditto.
19771 * recog.c (asm_operand_ok): Ditto.
19772 * reload.c (find_reloads): Ditto.
19774 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19776 * config/aarch64/aarch64-protos.h
19777 (cpu_addrcost_table::post_modify_ld3_st3): New member variable.
19778 (cpu_addrcost_table::post_modify_ld4_st4): Likewise.
19779 * config/aarch64/aarch64.c (generic_addrcost_table): Update
19780 accordingly, using the same costs as for post_modify.
19781 (exynosm1_addrcost_table, xgene1_addrcost_table): Likewise.
19782 (thunderx2t99_addrcost_table, thunderx3t110_addrcost_table):
19783 (tsv110_addrcost_table, qdf24xx_addrcost_table): Likewise.
19784 (a64fx_addrcost_table): Likewise.
19785 (neoversev1_addrcost_table): New.
19786 (neoversev1_tunings): Use neoversev1_addrcost_table.
19787 (aarch64_address_cost): Use the new post_modify costs for CImode
19790 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19792 * config/aarch64/aarch64.opt
19793 (-param=aarch64-loop-vect-issue-rate-niters=): New parameter.
19794 * doc/invoke.texi: Document it.
19795 * config/aarch64/aarch64-protos.h (aarch64_base_vec_issue_info)
19796 (aarch64_scalar_vec_issue_info, aarch64_simd_vec_issue_info)
19797 (aarch64_advsimd_vec_issue_info, aarch64_sve_vec_issue_info)
19798 (aarch64_vec_issue_info): New structures.
19799 (cpu_vector_cost): Write comments above the variables rather
19801 (cpu_vector_cost::issue_info): New member variable.
19802 * config/aarch64/aarch64.c: Include gimple-pretty-print.h
19803 and tree-ssa-loop-niter.h.
19804 (generic_vector_cost, a64fx_vector_cost, qdf24xx_vector_cost)
19805 (thunderx_vector_cost, tsv110_vector_cost, cortexa57_vector_cost)
19806 (exynosm1_vector_cost, xgene1_vector_cost, thunderx2t99_vector_cost)
19807 (thunderx3t110_vector_cost): Initialize issue_info to null.
19808 (neoversev1_scalar_issue_info, neoversev1_advsimd_issue_info)
19809 (neoversev1_sve_issue_info, neoversev1_vec_issue_info): New structures.
19810 (neoversev1_vector_cost): Use them.
19811 (aarch64_vec_op_count, aarch64_sve_op_count): New structures.
19812 (aarch64_vector_costs::saw_sve_only_op): New member variable.
19813 (aarch64_vector_costs::num_vector_iterations): Likewise.
19814 (aarch64_vector_costs::scalar_ops): Likewise.
19815 (aarch64_vector_costs::advsimd_ops): Likewise.
19816 (aarch64_vector_costs::sve_ops): Likewise.
19817 (aarch64_vector_costs::seen_loads): Likewise.
19818 (aarch64_simd_vec_costs_for_flags): New function.
19819 (aarch64_analyze_loop_vinfo): Initialize num_vector_iterations.
19820 Count the number of predicate operations required by SVE WHILE
19822 (aarch64_comparison_type, aarch64_multiply_add_p): New functions.
19823 (aarch64_sve_only_stmt_p, aarch64_in_loop_reduction_latency): Likewise.
19824 (aarch64_count_ops): Likewise.
19825 (aarch64_add_stmt_cost): Record whether see an SVE operation
19826 that cannot currently be implementing using Advanced SIMD.
19827 Record issue information about the scalar, Advanced SIMD
19828 and (where relevant) SVE versions of a loop.
19829 (aarch64_vec_op_count::dump): New function.
19830 (aarch64_sve_op_count::dump): Likewise.
19831 (aarch64_estimate_min_cycles_per_iter): Likewise.
19832 (aarch64_adjust_body_cost): If issue information is available,
19833 try to compare the issue rates of the various loop implementations
19834 and increase or decrease the vector body cost accordingly.
19836 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19838 * config/aarch64/aarch64.c (aarch64_detect_vector_stmt_subtype):
19839 Assume a zero cost for induction phis.
19841 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19843 * config/aarch64/aarch64.c (aarch64_embedded_comparison_type): New
19845 (aarch64_adjust_stmt_cost): Add the costs of embedded scalar and
19846 vector comparisons.
19848 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19850 * config/aarch64/aarch64.c (aarch64_detect_scalar_stmt_subtype):
19852 (aarch64_add_stmt_cost): Call it.
19854 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19856 * config/aarch64/aarch64-tuning-flags.def (matched_vector_throughput):
19857 New tuning parameter.
19858 * config/aarch64/aarch64.c (neoversev1_tunings): Use it.
19859 (aarch64_estimated_sve_vq): New function.
19860 (aarch64_vector_costs::analyzed_vinfo): New member variable.
19861 (aarch64_vector_costs::is_loop): Likewise.
19862 (aarch64_vector_costs::unrolled_advsimd_niters): Likewise.
19863 (aarch64_vector_costs::unrolled_advsimd_stmts): Likewise.
19864 (aarch64_record_potential_advsimd_unrolling): New function.
19865 (aarch64_analyze_loop_vinfo, aarch64_analyze_bb_vinfo): Likewise.
19866 (aarch64_add_stmt_cost): Call aarch64_analyze_loop_vinfo or
19867 aarch64_analyze_bb_vinfo on the first use of a costs structure.
19868 Detect whether we're vectorizing a loop for SVE that might be
19869 completely unrolled if it used Advanced SIMD instead.
19870 (aarch64_adjust_body_cost_for_latency): New function.
19871 (aarch64_finish_cost): Call it.
19873 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19875 * config/aarch64/aarch64.c (aarch64_vector_costs): New structure.
19876 (aarch64_init_cost): New function.
19877 (aarch64_add_stmt_cost): Use aarch64_vector_costs instead of
19878 the default unsigned[3].
19879 (aarch64_finish_cost, aarch64_destroy_cost_data): New functions.
19880 (TARGET_VECTORIZE_INIT_COST): Override.
19881 (TARGET_VECTORIZE_FINISH_COST): Likewise.
19882 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
19884 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19886 * config/aarch64/aarch64.c (neoversev1_advsimd_vector_cost)
19887 (neoversev1_sve_vector_cost): New cost structures.
19888 (neoversev1_vector_cost): Likewise.
19889 (neoversev1_tunings): Use them. Enable use_new_vector_costs.
19891 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19893 * config/aarch64/aarch64-protos.h
19894 (sve_vec_cost::scatter_store_elt_cost): New member variable.
19895 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
19896 accordingly, taking the cost from the cost of a scalar_store.
19897 (a64fx_sve_vector_cost): Likewise.
19898 (aarch64_detect_vector_stmt_subtype): Detect scatter stores.
19900 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19902 * config/aarch64/aarch64-protos.h
19903 (simd_vec_cost::store_elt_extra_cost): New member variable.
19904 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
19905 accordingly, using the vec_to_scalar cost for the new field.
19906 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
19907 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
19908 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
19909 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
19910 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
19911 (thunderx3t110_advsimd_vector_cost): Likewise.
19912 (aarch64_detect_vector_stmt_subtype): Detect single-element stores.
19914 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19916 * config/aarch64/aarch64-protos.h (simd_vec_cost::ld2_st2_permute_cost)
19917 (simd_vec_cost::ld3_st3_permute_cost): New member variables.
19918 (simd_vec_cost::ld4_st4_permute_cost): Likewise.
19919 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
19920 accordingly, using zero for the new costs.
19921 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
19922 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
19923 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
19924 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
19925 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
19926 (thunderx3t110_advsimd_vector_cost): Likewise.
19927 (aarch64_ld234_st234_vectors): New function.
19928 (aarch64_adjust_stmt_cost): Likewise.
19929 (aarch64_add_stmt_cost): Call aarch64_adjust_stmt_cost if using
19930 the new vector costs.
19932 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19934 * config/aarch64/aarch64-protos.h (sve_vec_cost): Turn into a
19935 derived class of simd_vec_cost. Add information about CLAST[AB]
19936 and FADDA instructions.
19937 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
19938 accordingly, using the vec_to_scalar costs for the new fields.
19939 (a64fx_sve_vector_cost): Likewise.
19940 (aarch64_reduc_type): New function.
19941 (aarch64_sve_in_loop_reduction_latency): Likewise.
19942 (aarch64_detect_vector_stmt_subtype): Take a vinfo parameter.
19943 Use aarch64_sve_in_loop_reduction_latency to handle SVE reductions
19944 that occur in the loop body.
19945 (aarch64_add_stmt_cost): Update call accordingly.
19947 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
19949 * config/aarch64/aarch64-tuning-flags.def (use_new_vector_costs):
19951 * config/aarch64/aarch64-protos.h (simd_vec_cost): Put comments
19952 above the fields rather than to the right.
19953 (simd_vec_cost::reduc_i8_cost): New member variable.
19954 (simd_vec_cost::reduc_i16_cost): Likewise.
19955 (simd_vec_cost::reduc_i32_cost): Likewise.
19956 (simd_vec_cost::reduc_i64_cost): Likewise.
19957 (simd_vec_cost::reduc_f16_cost): Likewise.
19958 (simd_vec_cost::reduc_f32_cost): Likewise.
19959 (simd_vec_cost::reduc_f64_cost): Likewise.
19960 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
19961 accordingly, using the vec_to_scalar_cost for the new fields.
19962 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
19963 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
19964 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
19965 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
19966 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
19967 (thunderx3t110_advsimd_vector_cost): Likewise.
19968 (aarch64_use_new_vector_costs_p): New function.
19969 (aarch64_simd_vec_costs): New function, split out from...
19970 (aarch64_builtin_vectorization_cost): ...here.
19971 (aarch64_is_reduction): New function.
19972 (aarch64_detect_vector_stmt_subtype): Likewise.
19973 (aarch64_add_stmt_cost): Call aarch64_detect_vector_stmt_subtype if
19974 using the new vector costs.
19976 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19979 * tree-emutls.c (get_emutls_init_templ_addr): Mark initializer of weak
19980 TLS declarations as public.
19982 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19984 * config/aarch64/aarch64-d.c (IN_TARGET_CODE): Define.
19985 * config/arm/arm-d.c (IN_TARGET_CODE): Likewise.
19986 * config/i386/i386-d.c (IN_TARGET_CODE): Likewise.
19987 * config/mips/mips-d.c (IN_TARGET_CODE): Likewise.
19988 * config/pa/pa-d.c (IN_TARGET_CODE): Likewise.
19989 * config/riscv/riscv-d.c (IN_TARGET_CODE): Likewise.
19990 * config/rs6000/rs6000-d.c (IN_TARGET_CODE): Likewise.
19991 * config/s390/s390-d.c (IN_TARGET_CODE): Likewise.
19992 * config/sparc/sparc-d.c (IN_TARGET_CODE): Likewise.
19994 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
19997 * config.gcc (*-*-cygwin*): Add winnt-d.o
19998 (*-*-mingw*): Likewise.
19999 * config/i386/cygwin.h (EXTRA_TARGET_D_OS_VERSIONS): New macro.
20000 * config/i386/mingw32.h (EXTRA_TARGET_D_OS_VERSIONS): Likewise.
20001 * config/i386/t-cygming: Add winnt-d.o.
20002 * config/i386/winnt-d.c: New file.
20004 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
20006 * config/freebsd-d.c: Include memmodel.h.
20008 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
20011 * config.gcc (*-*-openbsd*): Add openbsd-d.o.
20012 * config/t-openbsd: Add openbsd-d.o.
20013 * config/openbsd-d.c: New file.
20015 2021-03-25 Stam Markianos-Wright <stam.markianos-wright@arm.com>
20017 PR tree-optimization/96974
20018 * tree-vect-stmts.c (vect_get_vector_types_for_stmt): Replace assert
20019 with graceful exit.
20021 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
20024 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
20028 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
20029 always_inline in system headers.
20031 2021-03-25 Kewen Lin <linkw@linux.ibm.com>
20033 * tree-vect-loop.c (vect_model_reduction_cost): Init inside_cost.
20035 2021-03-25 Jakub Jelinek <jakub@redhat.com>
20038 * tree-core.h (enum operand_equal_flag): Add OEP_ADDRESS_OF_SAME_FIELD.
20039 * fold-const.c (operand_compare::operand_equal_p): Don't compare
20040 field offsets if OEP_ADDRESS_OF_SAME_FIELD.
20042 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
20046 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
20047 always_inline in system headers.
20049 2021-03-25 Richard Biener <rguenther@suse.de>
20051 PR tree-optimization/99746
20052 * tree-vect-slp-patterns.c (complex_pattern::build): Do not mark
20053 the scalar stmt as patterned. Instead set up required things
20056 2021-03-25 Xionghu Luo <luoxhu@linux.ibm.com>
20058 * config/rs6000/rs6000.c (power8_costs): Change l2 cache
20061 2021-03-24 Martin Liska <mliska@suse.cz>
20064 * common/config/i386/i386-common.c (ARRAY_SIZE): Fix off-by-one
20066 * config/i386/i386-options.c (ix86_option_override_internal):
20067 Add run-time assert.
20069 2021-03-24 Martin Jambor <mjambor@suse.cz>
20072 * ipa-cp.c (initialize_node_lattices): Mark as bottom all
20073 parameters with unknown type.
20074 (ipacp_value_safe_for_type): New function.
20075 (propagate_vals_across_arith_jfunc): Verify that the constant type
20076 can be used for a type of the formal parameter.
20077 (propagate_vals_across_ancestor): Likewise.
20078 (propagate_scalar_across_jump_function): Likewise. Pass the type
20079 also to propagate_vals_across_ancestor.
20081 2021-03-24 Christophe Lyon <christophe.lyon@linaro.org>
20084 * config/arm/mve.md (movmisalign<mode>_mve_store): Use Ux
20086 (movmisalign<mode>_mve_load): Likewise.
20088 2021-03-24 Jakub Jelinek <jakub@redhat.com>
20091 * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
20092 movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.
20094 2021-03-24 Alexandre Oliva <oliva@adacore.com>
20096 * doc/sourcebuild.texi (sysconf): New effective target.
20098 2021-03-24 Alexandre Oliva <oliva@adacore.com>
20100 * config/i386/predicates.md (reg_or_const_vec_operand): New.
20101 * config/i386/sse.md (ssse3_pshufbv8qi3): Add an expander for
20102 the now *-prefixed insn_and_split, turn the splitter const vec
20103 into an input for the insn, making it an ignored immediate for
20104 non-split cases, and loaded into the scratch register
20107 2021-03-23 Vladimir N. Makarov <vmakarov@redhat.com>
20110 * config/aarch64/constraints.md (Utq, UOb, UOh, UOw, UOd, UOty):
20111 Use define_relaxed_memory_constraint for them.
20113 2021-03-23 Iain Sandoe <iain@sandoe.co.uk>
20116 * config/host-darwin.c (darwin_gt_pch_use_address): Add a
20117 colon to the diagnostic message.
20119 2021-03-23 Ilya Leoshkevich <iii@linux.ibm.com>
20121 * fwprop.c (fwprop_propagation::fwprop_propagation): Look at
20123 (try_fwprop_subst_note): Use set_info instead of insn_info.
20124 (try_fwprop_subst_pattern): Likewise.
20125 (try_fwprop_subst_notes): Likewise.
20126 (try_fwprop_subst): Likewise.
20127 (forward_propagate_subreg): Likewise.
20128 (forward_propagate_and_simplify): Likewise.
20129 (forward_propagate_into): Likewise.
20130 * rtl-ssa/accesses.h (set_info::single_nondebug_use) New
20132 (set_info::single_nondebug_insn_use): Likewise.
20133 (set_info::single_phi_use): Likewise.
20134 * rtl-ssa/member-fns.inl (set_info::single_nondebug_use) New
20136 (set_info::single_nondebug_insn_use): Likewise.
20137 (set_info::single_phi_use): Likewise.
20139 2021-03-23 Christophe Lyon <christophe.lyon@linaro.org>
20141 * doc/sourcebuild.texi (arm_dsp_ok, arm_dsp): Document.
20143 2021-03-23 Jakub Jelinek <jakub@redhat.com>
20146 * config/aarch64/aarch64.c (aarch64_add_offset): Tell
20147 expand_mult to perform an unsigned rather than a signed
20150 2021-03-23 H.J. Lu <hjl.tools@gmail.com>
20153 * config/i386/cpuid.h (__cpuid): Add __volatile__.
20154 (__cpuid_count): Likewise.
20156 2021-03-23 Richard Biener <rguenther@suse.de>
20158 PR tree-optimization/99721
20159 * tree-vect-slp.c (vect_slp_analyze_node_operations):
20160 Make sure we can schedule the node.
20162 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20164 * config/riscv/riscv.c (riscv_subword): Take endianness into
20165 account when calculating the byte offset.
20167 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20169 * config/riscv/predicates.md (subreg_lowpart_operator): New predicate
20170 * config/riscv/riscv.md (*addsi3_extended2, *subsi3_extended2)
20171 (*negsi2_extended2, *mulsi3_extended2, *<optab>si3_mask)
20172 (*<optab>si3_mask_1, *<optab>di3_mask, *<optab>di3_mask_1)
20173 (*<optab>si3_extend_mask, *<optab>si3_extend_mask_1): Use
20174 new predicate "subreg_lowpart_operator"
20176 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20178 * config/riscv/riscv.c (riscv_swap_instruction): New function
20179 to byteswap an SImode rtx containing an instruction.
20180 (riscv_trampoline_init): Byteswap the generated instructions
20183 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20185 * common/config/riscv/riscv-common.c
20186 (TARGET_DEFAULT_TARGET_FLAGS): Set default endianness.
20187 * config.gcc (riscv32be-*, riscv64be-*): Set
20188 TARGET_BIG_ENDIAN_DEFAULT to 1.
20189 * config/riscv/elf.h (LINK_SPEC): Change -melf* value
20190 depending on default endianness.
20191 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
20192 * config/riscv/linux.h (LINK_SPEC): Likewise.
20193 * config/riscv/riscv.c (TARGET_DEFAULT_TARGET_FLAGS): Set
20194 default endianness.
20195 * config/riscv/riscv.h (DEFAULT_ENDIAN_SPEC): New macro.
20197 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
20199 * config/riscv/elf.h (LINK_SPEC): Pass linker endianness flag.
20200 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
20201 * config/riscv/linux.h (LINK_SPEC): Likewise.
20202 * config/riscv/riscv.h (ASM_SPEC): Pass -mbig-endian and
20204 (BYTES_BIG_ENDIAN): Handle big endian.
20205 (WORDS_BIG_ENDIAN): Define to BYTES_BIG_ENDIAN.
20206 * config/riscv/riscv.opt (-mbig-endian, -mlittle-endian): New
20208 * doc/invoke.texi (-mbig-endian, -mlittle-endian): Document.
20210 2021-03-23 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
20212 * regcprop.c (find_oldest_value_reg): Ask target whether
20213 different mode is fine for replacement register.
20215 2021-03-23 Aldy Hernandez <aldyh@redhat.com>
20217 PR tree-optimization/99296
20218 * value-range.cc (irange::irange_set_1bit_anti_range): New.
20219 (irange::irange_set_anti_range): Call irange_set_1bit_anti_range
20220 * value-range.h (irange::irange_set_1bit_anti_range): New.
20222 2021-03-22 Vladimir N. Makarov <vmakarov@redhat.com>
20225 * config/aarch64/constraints.md (UtQ): Use
20226 define_relaxed_memory_constraint for it.
20227 * doc/md.texi (define_relaxed_memory_constraint): Describe it.
20228 * genoutput.c (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
20229 * genpreds.c (constraint_data): Add bitfield is_relaxed_memory.
20230 (have_relaxed_memory_constraints): New static var.
20231 (relaxed_memory_start, relaxed_memory_end): Ditto.
20232 (add_constraint): Add arg is_relaxed_memory. Check name for
20233 relaxed memory. Set up is_relaxed_memory in constraint_data and
20234 have_relaxed_memory_constraints. Adjust calls.
20235 (choose_enum_order): Process relaxed memory.
20236 (write_tm_preds_h): Ditto.
20237 (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
20238 * gensupport.c (process_rtx): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
20239 * ira-costs.c (record_reg_classes): Process CT_RELAXED_MEMORY.
20240 * ira-lives.c (single_reg_class): Use
20241 insn_extra_relaxed_memory_constraint.
20242 * ira.c (ira_setup_alts): CT_RELAXED_MEMORY.
20243 * lra-constraints.c (valid_address_p): Use
20244 insn_extra_relaxed_memory_constraint instead of other memory
20246 (process_alt_operands): Process CT_RELAXED_MEMORY.
20247 (curr_insn_transform): Use insn_extra_relaxed_memory_constraint.
20248 * recog.c (asm_operand_ok, preprocess_constraints): Process
20250 * reload.c (find_reloads): Ditto.
20251 * rtl.def (DEFINE_RELAXED_MEMORY_CONSTRAINT): New.
20252 * stmt.c (parse_input_constraint): Use
20253 insn_extra_relaxed_memory_constraint.
20255 2021-03-22 Segher Boessenkool <segher@kernel.crashing.org>
20258 * ubsan.c (ubsan_instrument_float_cast): Don't test for unordered if
20261 2021-03-22 Alex Coplan <alex.coplan@arm.com>
20264 * config/arm/arm-protos.h (neon_make_constant): Add generate
20265 argument to guard emitting insns, default to true.
20266 * config/arm/arm.c (arm_legitimate_constant_p_1): Reject
20267 CONST_VECTORs which neon_make_constant can't handle.
20268 (neon_vdup_constant): Add generate argument, avoid emitting
20269 insns if it's not set.
20270 (neon_make_constant): Plumb new generate argument through.
20271 * config/arm/constraints.md (Ui): New. Use it...
20272 * config/arm/mve.md (*mve_mov<mode>): ... here.
20273 * config/arm/vec-common.md (movv8hf): Use neon_make_constant to
20274 synthesize constants.
20276 2021-03-22 Richard Biener <rguenther@suse.de>
20278 * debug.h: Add deprecation warning.
20280 2021-03-22 Richard Biener <rguenther@suse.de>
20282 PR tree-optimization/99694
20283 * tree-ssa-sccvn.c (visit_phi): Ignore edges with the
20286 2021-03-22 Kito Cheng <kito.cheng@sifive.com>
20289 * config/riscv/riscv.c (riscv_expand_block_move): Get RTL value
20290 after type checking.
20292 2021-03-22 Jakub Jelinek <jakub@redhat.com>
20296 * dwarf2out.c (get_full_len): Use get_precision rather than
20298 (add_const_value_attribute): Make sure add_AT_wide argument has
20299 precision prec rather than some very wide one.
20301 2021-03-22 Kewen Lin <linkw@linux.ibm.com>
20303 * config/rs6000/rs6000.md (*rotldi3_insert_sf,
20304 *mov<SFDF:mode><SFDF2:mode>cc_p9, floatsi<mode>2_lfiwax,
20305 floatsi<mode>2_lfiwax_mem, floatunssi<mode>2_lfiwzx,
20306 floatunssi<mode>2_lfiwzx_mem, *floatsidf2_internal,
20307 *floatunssidf2_internal, fix_trunc<mode>si2_stfiwx,
20308 fix_trunc<mode>si2_internal, fixuns_trunc<mode>si2_stfiwx,
20309 *round32<mode>2_fprs, *roundu32<mode>2_fprs,
20310 *fix_trunc<mode>si2_internal): Fix empty split condition.
20311 * config/rs6000/vsx.md (*vsx_le_undo_permute_<mode>,
20312 vsx_reduc_<VEC_reduc_name>_v2df, vsx_reduc_<VEC_reduc_name>_v4sf,
20313 *vsx_reduc_<VEC_reduc_name>_v2df_scalar,
20314 *vsx_reduc_<VEC_reduc_name>_v4sf_scalar): Likewise.
20316 2021-03-22 Xionghu Luo <luoxhu@linux.ibm.com>
20319 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
20320 Convert idx to DImode.
20321 (rs6000_expand_vector_set_var_p8): Likewise.
20323 2021-03-21 Jakub Jelinek <jakub@redhat.com>
20326 * dwarf2out.c (insert_float): Change return type from void to
20327 unsigned, handle GET_MODE_SIZE (mode) == 2 and return element size.
20328 (mem_loc_descriptor, loc_descriptor, add_const_value_attribute):
20331 2021-03-20 H.J. Lu <hjl.tools@gmail.com>
20334 * config/i386/i386.c (construct_container): Check cfun != NULL
20335 before accessing silent_p.
20337 2021-03-20 Ahamed Husni <ahamedhusni73@gmail.com>
20339 * asan.c: Fix typos in comments.
20341 2021-03-20 Vladimir N. Makarov <vmakarov@redhat.com>
20343 PR rtl-optimization/99680
20344 * lra-constraints.c (skip_contraint_modifiers): Rename to skip_constraint_modifiers.
20345 (process_address_1): Check empty constraint before using
20348 2021-03-19 Pat Haugen <pthaugen@linux.ibm.com>
20350 * config/rs6000/rs6000.c (power10_cost): New.
20351 (rs6000_option_override_internal): Set Power10 costs.
20352 (rs6000_issue_rate): Set Power10 issue rate.
20353 * config/rs6000/power10.md: Rewrite for Power10.
20355 2021-03-19 Vladimir N. Makarov <vmakarov@redhat.com>
20358 * lra-constraints.c (process_address_1): Don't use unknown
20359 constraint for address constraint.
20361 2021-03-19 Iain Sandoe <iain@sandoe.co.uk>
20364 * config.gcc (powerpc-*-darwin8): Delete the reference to
20365 the now removed darwin8.h.
20367 2021-03-19 Olivier Hainque <hainque@adacore.com>
20370 * config/vxworksae.h (VX_CPU_PREFIX): Define.
20372 2021-03-19 John David Anglin <danglin@gcc.gnu.org>
20374 * config/pa/pa.c (import_milli): Use memcpy instead of strncpy.
20376 2021-03-19 Tamar Christina <tamar.christina@arm.com>
20378 PR tree-optimization/99656
20379 * tree-vect-slp-patterns.c (linear_loads_p,
20380 complex_add_pattern::matches, is_eq_or_top,
20381 vect_validate_multiplication, complex_mul_pattern::matches,
20382 complex_fms_pattern::matches): Remove complex_perm_kinds_t.
20383 * tree-vectorizer.h: (complex_load_perm_t): Removed.
20384 (slp_tree_to_load_perm_map_t): Use complex_perm_kinds_t instead of
20385 complex_load_perm_t.
20387 2021-03-19 H.J. Lu <hjl.tools@gmail.com>
20390 * config/i386/i386-options.c (ix86_init_machine_status): Set
20392 * config/i386/i386.c (init_cumulative_args): Set silent_p to
20394 (construct_container): Return early for return and argument
20395 errors if silent_p is true.
20396 * config/i386/i386.h (machine_function): Add silent_p.
20398 2021-03-19 Jakub Jelinek <jakub@redhat.com>
20401 * config/arm/constraints.md (Ds): New constraint.
20402 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): Use w,Ds
20403 constraint instead of w,Dm.
20405 2021-03-19 Andrew Stubbs <ams@codesourcery.com>
20407 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Fix quotes
20410 2021-03-19 Eric Botcazou <ebotcazou@adacore.com>
20412 PR middle-end/99641
20413 * fold-const.c (native_encode_initializer) <CONSTRUCTOR>: For an
20414 array type, do the computation of the current position in sizetype.
20416 2021-03-18 Vladimir N. Makarov <vmakarov@redhat.com>
20419 * lra-constraints.c (process_address_1): Use lookup_constraint
20420 only for a single constraint.
20422 2021-03-18 Martin Sebor <msebor@redhat.com>
20424 PR middle-end/99502
20425 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): Rename...
20426 (inbounds_memaccess_p): ...to this. Check the ending offset of
20427 the accessed member.
20429 2021-03-18 Andrew Stubbs <ams@codesourcery.com>
20431 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Add %< and
20432 %> quote markers to error messages.
20433 (gcn_goacc_validate_dims): Likewise.
20434 (gcn_conditional_register_usage): Remove exclaimation mark from error
20436 (gcn_vectorize_vec_perm_const): Ensure perm is fully uninitialized.
20438 2021-03-18 Jan Hubicka <hubicka@ucw.cz>
20440 * config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
20443 2021-03-18 Sinan Lin <sinan@isrc.iscas.ac.cn>
20444 Kito Cheng <kito.cheng@sifive.com>
20446 * config/riscv/riscv.c (riscv_block_move_straight): Change type
20447 to unsigned HOST_WIDE_INT for parameter and local variable with
20448 HOST_WIDE_INT type.
20449 (riscv_adjust_block_mem): Ditto.
20450 (riscv_block_move_loop): Ditto.
20451 (riscv_expand_block_move): Ditto.
20453 2021-03-18 Nick Clifton <nickc@redhat.com>
20455 * config/v850/v850.c (construct_restore_jr): Increase static
20457 (construct_save_jarl): Likewise.
20458 * config/v850/v850.h (DWARF2_DEBUGGING_INFO): Define.
20460 2021-03-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20462 * config/aarch64/aarch64.c (aarch64_adjust_generic_arch_tuning): Define.
20463 (aarch64_override_options_internal): Use it.
20464 (generic_tunings): Add AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS to
20467 2021-03-17 Sandra Loosemore <sandra@codesourcery.com>
20469 * config/nios2/nios2.c (nios2_custom_check_insns): Clean up
20470 error message format issues.
20471 (nios2_option_override): Likewise.
20472 (nios2_expand_fpu_builtin): Likewise.
20473 (nios2_init_custom_builtins): Adjust to avoid bogus strncpy
20474 truncation warning.
20475 (nios2_expand_custom_builtin): More error message format fixes.
20476 (nios2_expand_rdwrctl_builtin): Likewise.
20477 (nios2_expand_rdprs_builtin): Likewise.
20478 (nios2_expand_eni_builtin): Likewise.
20479 (nios2_expand_builtin): Likewise.
20480 (nios2_register_custom_code): Likewise.
20481 (nios2_valid_target_attribute_rec): Likewise.
20482 (nios2_add_insn_asm): Fix uninitialized variable warning.
20484 2021-03-17 Jan Hubicka <jh@suse.cz>
20486 * config/i386/x86-tune-costs.h (struct processor_costs): Update costs
20487 of gather to match reality.
20488 * config/i386/x86-tune.def (X86_TUNE_USE_GATHER): Enable for znver3.
20490 2021-03-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20492 * config/aarch64/aarch64-builtins.c (aarch64_expand_rng_builtin): Use EQ
20493 to compare against CC_REG rather than NE.
20495 2021-03-17 H.J. Lu <hjl.tools@gmail.com>
20498 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support
20499 inline assembly statements.
20500 (ix86_print_operand): Update 'P' handling for -fno-plt.
20502 2021-03-17 Tamar Christina <tamar.christina@arm.com>
20505 * config/aarch64/aarch64.c
20506 (aarch64_simd_clone_compute_vecsize_and_simdlen): Remove unused var.
20508 2021-03-16 Segher Boessenkool <segher@kernel.crashing.org>
20511 * config/rs6000/predicates.md (branch_comparison_operator): Allow
20512 ordered and unordered for CCFPmode, if flag_finite_math_only.
20514 2021-03-16 Jakub Jelinek <jakub@redhat.com>
20517 * config/i386/i386-expand.c (ix86_split_lea_for_addr): Emit a MULT
20518 rather than ASHIFT.
20519 * config/i386/i386.md (mult by 1248 into ashift): New splitter.
20521 2021-03-16 Martin Liska <mliska@suse.cz>
20524 * optc-save-gen.awk: Add flag_ipa_ra to exceptions for
20525 cl_optimization_compare function.
20527 2021-03-16 Ilya Leoshkevich <iii@linux.ibm.com>
20529 * config/s390/s390.c (f_constraint_p): Treat "fv" constraints
20532 2021-03-16 Jakub Jelinek <jakub@redhat.com>
20535 * config/i386/i386.h (struct machine_function): Add
20536 has_explicit_vzeroupper bitfield.
20537 * config/i386/i386-expand.c (ix86_expand_builtin): Set
20538 cfun->machine->has_explicit_vzeroupper when expanding
20539 IX86_BUILTIN_VZEROUPPER.
20540 * config/i386/i386-features.c (rest_of_handle_insert_vzeroupper):
20541 Do the mode switching only when TARGET_VZEROUPPER, expensive
20542 optimizations turned on and not optimizing for size.
20543 (pass_insert_vzeroupper::gate): Enable even when
20544 cfun->machine->has_explicit_vzeroupper is set.
20546 2021-03-16 Jakub Jelinek <jakub@redhat.com>
20549 * config/aarch64/aarch64.c
20550 (aarch64_simd_clone_compute_vecsize_and_simdlen): If not a function
20551 definition, walk TYPE_ARG_TYPES list if non-NULL for argument types
20552 instead of DECL_ARGUMENTS. Ignore types for uniform arguments.
20554 2021-03-15 Richard Biener <rguenther@suse.de>
20556 PR tree-optimization/98834
20557 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle missing
20558 subsetting by truncating the access size.
20560 2021-03-15 Jan Hubicka <hubicka@ucw.cz>
20562 * config/i386/i386-options.c (processor_cost_table): Add znver3_cost.
20563 * config/i386/x86-tune-costs.h (znver3_cost): New gobal variable; copy
20566 2021-03-15 Martin Liska <mliska@suse.cz>
20568 * spellcheck.c: Add missing comma in initialization.
20570 2021-03-14 Uroš Bizjak <ubizjak@gmail.com>
20572 * config/i386/sse.md (*vec_extract<mode>): Merge alternative 0 with
20573 alternative 2 and alternative 1 with alternative 3 using
20574 YW register constraint.
20575 (*vec_extract<PEXTR_MODE12:mode>_zext): Merge alternatives
20576 using YW register constraint.
20577 (*vec_extractv16qi_zext): Ditto.
20578 (*vec_extractv4si): Merge alternatives 4 and 5
20579 using Yw register constraint.
20580 (*ssse3_palignr<mode>_perm): Use Yw instead of v for alternative 3.
20582 2021-03-13 Martin Sebor <msebor@redhat.com>
20584 PR tree-optimization/99489
20585 * builtins.c (gimple_call_alloc_size): Fail gracefully when argument
20586 is not a call statement.
20588 2021-03-13 Jakub Jelinek <jakub@redhat.com>
20590 PR tree-optimization/99544
20591 * match.pd (X + (X << C) -> X * (1 + (1 << C))): Don't simplify
20592 if for vector types multiplication can't be done in type's mode.
20594 2021-03-12 Eric Botcazou <ebotcazou@adacore.com>
20597 * config/sparc/constraints.md (w): Rename to...
20598 (W): ... this and ditch previous implementation.
20599 * config/sparc/sparc.md (*movdi_insn_sp64): Replace W with m.
20600 (*movdf_insn_sp64): Likewise.
20601 (*mov<VM64:mode>_insn_sp64): Likewise.
20602 * config/sparc/sync.md (*atomic_compare_and_swap<mode>_1): Replace
20604 (atomic_compare_and_swap_leon3_1): Likewise.
20605 (*atomic_compare_and_swapdi_v8plus): Likewise.
20606 * config/sparc/sparc.c (memory_ok_for_ldd): Remove useless test on
20607 architecture and add missing address validity check during LRA.
20609 2021-03-12 Tobias Burnus <tobias@codesourcery.com>
20612 * gimplify.c (omp_add_variable): Handle NULL_TREE as size
20613 occuring for assumed-size arrays in use_device_{ptr,addr}.
20615 2021-03-12 Jakub Jelinek <jakub@redhat.com>
20618 * config/i386/constraints.md (YW): New internal constraint.
20619 * config/i386/sse.md (v_Yw): Add V4TI, V2TI, V1TI and TI cases.
20620 (*<sse2_avx2>_<insn><mode>3<mask_name>,
20621 *<sse2_avx2>_uavg<mode>3<mask_name>, *abs<mode>2,
20622 *<s>mul<mode>3_highpart<mask_name>): Use <v_Yw> instead of v in
20624 (<sse2_avx2>_psadbw): Use YW instead of v in constraints.
20625 (*avx2_pmaddwd, *sse2_pmaddwd, *<code>v8hi3, *<code>v16qi3,
20626 avx2_pmaddubsw256, ssse3_pmaddubsw128): Merge last two alternatives
20627 into one, use Yw instead of former x,v.
20628 (ashr<mode>3, <insn><mode>3): Use <v_Yw> instead of x in constraints of
20629 the last alternative.
20630 (<sse2_avx2>_packsswb<mask_name>, <sse2_avx2>_packssdw<mask_name>,
20631 <sse2_avx2>_packuswb<mask_name>, <sse4_1_avx2>_packusdw<mask_name>,
20632 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, <ssse3_avx2>_palignr<mode>,
20633 <ssse3_avx2>_pshufb<mode>3<mask_name>): Merge last two alternatives
20634 into one, use <v_Yw> instead of former x,v.
20635 (avx2_interleave_highv32qi<mask_name>,
20636 vec_interleave_highv16qi<mask_name>): Use Yw instead of v in
20637 constraints. Add && <mask_avx512bw_condition> to condition.
20638 (avx2_interleave_lowv32qi<mask_name>,
20639 vec_interleave_lowv16qi<mask_name>,
20640 avx2_interleave_highv16hi<mask_name>,
20641 vec_interleave_highv8hi<mask_name>,
20642 avx2_interleave_lowv16hi<mask_name>, vec_interleave_lowv8hi<mask_name>,
20643 avx2_pshuflw_1<mask_name>, sse2_pshuflw_1<mask_name>,
20644 avx2_pshufhw_1<mask_name>, sse2_pshufhw_1<mask_name>,
20645 avx2_<code>v16qiv16hi2<mask_name>, sse4_1_<code>v8qiv8hi2<mask_name>,
20646 *sse4_1_<code>v8qiv8hi2<mask_name>_1, <sse2_avx2>_<insn><mode>3): Use
20647 Yw instead of v in constraints.
20648 * config/i386/mmx.md (Yv_Yw): New define_mode_attr.
20649 (*mmx_<insn><mode>3, mmx_ashr<mode>3, mmx_<insn><mode>3): Use <Yv_Yw>
20650 instead of Yv in constraints.
20651 (*mmx_<insn><mode>3, *mmx_mulv4hi3, *mmx_smulv4hi3_highpart,
20652 *mmx_umulv4hi3_highpart, *mmx_pmaddwd, *mmx_<code>v4hi3,
20653 *mmx_<code>v8qi3, mmx_pack<s_trunsuffix>swb, mmx_packssdw,
20654 mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
20655 *mmx_uavgv8qi3, *mmx_uavgv4hi3, mmx_psadbw): Use Yw instead of Yv in
20657 (*mmx_pinsrw, *mmx_pinsrb, *mmx_pextrw, *mmx_pextrw_zext, *mmx_pextrb,
20658 *mmx_pextrb_zext): Use YW instead of Yv in constraints.
20659 (*mmx_eq<mode>3, mmx_gt<mode>3): Use x instead of Yv in constraints.
20660 (mmx_andnot<mode>3, *mmx_<code><mode>3): Split last alternative into
20661 two, one with just x, another isa avx512vl with v.
20663 2021-03-12 Martin Liska <mliska@suse.cz>
20665 * doc/invoke.texi: Add missing param documentation.
20667 2021-03-11 David Malcolm <dmalcolm@redhat.com>
20670 * Makefile.in (ANALYZER_OBJS): Add analyzer/feasible-graph.o and
20671 analyzer/trimmed-graph.o.
20672 * doc/analyzer.texi (Analyzer Paths): Rewrite description of
20673 feasibility checking to reflect new implementation.
20674 * doc/invoke.texi (-fdump-analyzer-feasibility): Document new
20676 * shortest-paths.h (shortest_paths::get_shortest_distance): New.
20678 2021-03-11 David Malcolm <dmalcolm@redhat.com>
20680 * digraph.cc (selftest::test_shortest_paths): Update
20681 shortest_paths init for new param. Add test of
20682 SPS_TO_GIVEN_TARGET.
20683 * shortest-paths.h (enum shortest_path_sense): New.
20684 (shortest_paths::shortest_paths): Add "sense" param.
20685 Update for renamings. Generalize to use "sense" param.
20686 (shortest_paths::get_shortest_path): Rename param.
20687 (shortest_paths::m_sense): New field.
20688 (shortest_paths::m_prev): Rename...
20689 (shortest_paths::m_best_edge): ...to this.
20690 (shortest_paths::get_shortest_path): Update for renamings.
20691 Conditionalize flipping of path on sense of traversal.
20693 2021-03-11 David Malcolm <dmalcolm@redhat.com>
20695 * digraph.cc (selftest::test_shortest_paths): Add test coverage
20696 for paths from B and C.
20697 * shortest-paths.h (shortest_paths::shortest_paths): Handle
20698 unreachable nodes, rather than asserting.
20700 2021-03-11 David Edelsohn <dje.gcc@gmail.com>
20703 * config/rs6000/rs6000.c (rs6000_xcoff_file_start): Don't create
20704 xcoff_tbss_section_name.
20705 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .lcomm.
20706 * xcoffout.c (xcoff_tbss_section_name): Delete.
20707 * xcoffout.h (xcoff_tbss_section_name): Delete.
20709 2021-03-11 Richard Biener <rguenther@suse.de>
20711 PR tree-optimization/99523
20712 * tree-cfg.c (dump_function_to_file): Dump SSA names
20713 w/o identifier to the decls section as well, not only those
20714 without a VAR_DECL.
20716 2021-03-11 Jakub Jelinek <jakub@redhat.com>
20719 * ipa-icf-gimple.c (func_checker::compare_gimple_call): For internal
20720 function calls with lhs fail if the lhs don't have compatible types.
20722 2021-03-11 Hans-Peter Nilsson <hp@axis.com>
20724 * config/cris/cris.h (HARD_FRAME_POINTER_REGNUM): Define.
20725 Change FRAME_POINTER_REGNUM to correspond to a new faked
20726 register faked_fp, part of GENNONACR_REGS like faked_ap.
20727 (CRIS_FAKED_REGS_CONTENTS): New helper macro.
20728 (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS, CALL_USED_REGISTERS):
20729 (REG_ALLOC_ORDER, REG_CLASS_CONTENTS, REGNO_OK_FOR_BASE_P)
20730 (ELIMINABLE_REGS, REGISTER_NAMES): Adjust accordingly.
20731 * config/cris/cris.md (CRIS_FP_REGNUM): Renumber to new faked
20733 (CRIS_REAL_FP_REGNUM): New constant.
20734 * config/cris/cris.c (cris_reg_saved_in_regsave_area): Check
20735 for HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM.
20736 (cris_initial_elimination_offset): Handle elimination changes
20737 to HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM
20738 and add one from FRAME_POINTER_REGNUM to
20739 HARD_FRAME_POINTER_REGNUM.
20740 (cris_expand_prologue, cris_expand_epilogue): Emit code for
20741 hard_frame_pointer_rtx instead of frame_pointer_rtx.
20743 2021-03-10 David Edelsohn <dje.gcc@gmail.com>
20746 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Add check for DCmode.
20747 * config/rs6000/rs6000.c (rs6000_special_round_type_align): Same.
20749 2021-03-10 Vladimir N. Makarov <vmakarov@redhat.com>
20752 * lra-constraints.c (process_address_1): Don't check unknown
20753 constraint, use X for empty constraint.
20755 2021-03-10 Alex Coplan <alex.coplan@arm.com>
20757 * config/aarch64/aarch64.c (aarch64_vfp_is_call_or_return_candidate):
20758 Fix typo in comment describing "is_ha" argument.
20760 2021-03-10 John David Anglin <danglin@gcc.gnu.org>
20762 * doc/sourcebuild.texi: Document LRA target selector.
20764 2021-03-10 David Malcolm <dmalcolm@redhat.com>
20766 * doc/ux.texi: Add subsection contrasting interactive versus
20767 batch usage of GCC.
20769 2021-03-10 Joel Hutton <joel.hutton@arm.com>
20772 * tree-vect-stmts.c (vectorizable_store): Fix scatter store mask
20774 (vectorizable_load): Fix gather load mask check condition.
20776 2021-03-10 Richard Biener <rguenther@suse.de>
20778 PR tree-optimization/99510
20779 * tree.c (check_aligned_type): Check that the candidate
20780 has TYPE_USER_ALIGN set instead of matching with the
20783 2021-03-10 Eric Botcazou <ebotcazou@adacore.com>
20785 * config/sparc/sparc.c (sparc_regmode_natural_size): Return 4 for
20786 float and vector integer modes only if the mode is not larger.
20788 2021-03-10 Hans-Peter Nilsson <hp@axis.com>
20790 * config/cris/cris.h (DWARF_FRAME_REGISTERS): Define.
20792 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
20794 * ira.c (ira_setup_alts, ira_get_dup_out_num): Process digital
20796 * ira-lives.c (single_reg_class): Ditto.
20798 2021-03-09 Sebastian Huber <sebastian.huber@embedded-brains.de>
20800 * config.gcc (aarch64-*-rtems*): Include general rtems.h after
20801 the architecture-specific rtems.h.
20802 (aarch64-*-rtems*): Likewise.
20803 (arm*-*-rtems*): Likewise.
20804 (epiphany-*-rtems*): Likewise.
20805 (riscv*-*-rtems*): Likewise.
20807 2021-03-09 Jakub Jelinek <jakub@redhat.com>
20809 PR tree-optimization/99305
20810 * tree-ssa-phiopt.c (conditional_replacement): Test integer_pow2p
20811 before integer_all_onesp instead of vice versa.
20813 2021-03-09 Richard Earnshaw <rearnsha@arm.com>
20815 * common/config/arm/arm-common.c (arm_config_default): Change type
20816 of 'i' to unsigned.
20818 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
20821 * lra-constraints.c (process_address_1): Process constraint 'g'
20822 separately and digital constraints containing more one digit.
20824 2021-03-09 Nick Clifton <nickc@redhat.com>
20826 * config/rx/rx.h (DBX_DEBUGGING_INFO): Define.
20827 (DWARF"_DEBUGGING_INFO): Define.
20829 2021-03-09 Eric Botcazou <ebotcazou@adacore.com>
20832 * calls.c (initialize_argument_information): When the argument
20833 is passed by reference, do not make a copy in a thunk only if
20834 the argument is already in memory. Remove redundant test for
20835 the case of callee copy.
20837 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
20840 * lra-constraints.c (process_address_1): Process 0..9 constraints
20841 in process_address_1.
20843 2021-03-09 Andreas Krebbel <krebbel@linux.ibm.com>
20845 * config/s390/s390.c (struct s390_processor processor_table):
20846 Binutils name string must not be empty.
20848 2021-03-09 Claudiu Zissulescu <claziss@synopsys.com>
20850 * config/arc/arc.c (arc_attr_type): Remove function.
20852 2021-03-09 Martin Liska <mliska@suse.cz>
20855 * config/i386/i386-options.c (ix86_option_override_internal):
20856 Set isa_flags for OPTS argument and not for the global
20859 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
20861 * config/rs6000/predicates.md (ds_form_mem_operand): Check
20864 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
20867 * config/rs6000/predicates.md (ds_form_mem_operand) New
20869 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10) Use
20870 ds_form_mem_operand in ld/lwa patterns.
20871 * config/rs6000/fusion.md: Regenerate file.
20873 2021-03-08 Martin Sebor <msebor@redhat.com>
20875 PR middle-end/98266
20876 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): New function.
20877 (array_bounds_checker::check_array_bounds): Call it.
20879 2021-03-08 Martin Sebor <msebor@redhat.com>
20881 PR middle-end/97631
20882 * tree-ssa-strlen.c (maybe_warn_overflow): Test rawmem.
20883 (handle_builtin_stxncpy_strncat): Rename locals. Determine
20884 destination size from allocation calls. Issue a more appropriate
20886 (handle_builtin_memcpy): Pass true as rawmem to maybe_warn_overflow.
20887 (handle_builtin_memset): Same.
20889 2021-03-08 Peter Bergner <bergner@linux.ibm.com>
20892 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_permute): Add an assert
20893 to ensure we do not have an Altivec style address.
20894 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): Disable if passed
20895 an Altivec style address.
20896 (*vsx_le_perm_store_<mode>): Likewise.
20897 (splitters after *vsx_le_perm_store_<mode>): Likewise.
20898 (vsx_load_<mode>): Disable special expander if passed an Altivec
20900 (vsx_store_<mode>): Likewise.
20902 2021-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20905 * config/aarch64/predicates.md (aarch64_simd_shift_imm_vec_qi): Define.
20906 (aarch64_simd_shift_imm_vec_hi): Likewise.
20907 (aarch64_simd_shift_imm_vec_si): Likewise.
20908 (aarch64_simd_shift_imm_vec_di): Likewise.
20909 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le): Use
20910 predicate from above.
20911 (aarch64_shrn<mode>_insn_be): Likewise.
20912 (aarch64_rshrn<mode>_insn_le): Likewise.
20913 (aarch64_rshrn<mode>_insn_be): Likewise.
20914 (aarch64_shrn2<mode>_insn_le): Likewise.
20915 (aarch64_shrn2<mode>_insn_be): Likewise.
20916 (aarch64_rshrn2<mode>_insn_le): Likewise.
20917 (aarch64_rshrn2<mode>_insn_be): Likewise.
20919 2021-03-08 Vladimir N. Makarov <vmakarov@redhat.com>
20922 * lra-constraints.c (skip_contraint_modifiers): New function.
20923 (process_address_1): Use it before lookup_constraint call.
20925 2021-03-08 Martin Liska <mliska@suse.cz>
20928 * config/i386/i386-options.c (ix86_option_override_internal):
20929 Enable UINTR and HRESET for -march that supports it.
20931 2021-03-08 Ilya Leoshkevich <iii@linux.ibm.com>
20933 * config/s390/s390.c (f_constraint_p): New function.
20934 (s390_md_asm_adjust): Implement TARGET_MD_ASM_ADJUST.
20935 (TARGET_MD_ASM_ADJUST): Likewise.
20937 2021-03-08 Tobias Burnus <tobias@codesourcery.com>
20940 * tree-nested.c (convert_local_reference_stmt): Avoid calling
20941 lookup_field_for_decl for Fortran module (= namespace context).
20943 2021-03-08 Andreas Krebbel <krebbel@linux.ibm.com>
20945 * config/s390/s390.c (s390_expand_vec_compare): Implement <0
20946 comparison with arithmetic right shift.
20947 (s390_expand_vcond): No need for a force_reg anymore.
20948 s390_vec_compare will do it.
20949 * config/s390/vector.md ("vec_cmp<mode><tointvec>"): Accept also
20950 immediate operands.
20952 2021-03-07 Jakub Jelinek <jakub@redhat.com>
20955 * config/i386/constraints.md (Yw): Use SSE_REGS if TARGET_SSE
20956 but TARGET_AVX512BW or TARGET_AVX512VL is not set. Adjust description
20958 * config/i386/sse.md (v_Yw): New define_mode_attr.
20959 (*<insn><mode>3, *mul<mode>3<mask_name>, *avx2_<code><mode>3,
20960 *sse4_1_<code><mode>3<mask_name>): Use <v_Yw> instead of v
20962 * config/i386/mmx.md (mmx_pshufw_1, *vec_dupv4hi): Use Yw instead of
20963 xYw in constraints.
20965 2021-03-06 Julian Brown <julian@codesourcery.com>
20967 * tree-pretty-print.c (dump_generic_node): Emit non-generic
20968 address space info for aggregates.
20970 2021-03-06 Hans-Peter Nilsson <hp@axis.com>
20972 * config/cris/cris.h (MAX_FIXED_MODE_SIZE): Don't define.
20974 2021-03-05 Jakub Jelinek <jakub@redhat.com>
20976 PR middle-end/99322
20977 * tree-cfg.c (bb_to_omp_idx): New variable.
20978 (execute_build_cfg): Release the bb_to_omp_idx vector after
20979 cleanup_tree_cfg returns.
20980 (handle_abnormal_edges): Remove bb_to_omp_idx argument, adjust
20981 for bb_to_omp_idx being a vec<int> instead of pointer to array
20983 (make_edges): Remove bb_to_omp_idx local variable, don't pass
20984 it to handle_abnormal_edges, adjust for bb_to_omp_idx being a
20985 vec<int> instead of pointer to array of ints and don't free/release
20987 (remove_bb): When removing a bb and placing forced label somewhere
20988 else, ensure it is put into the same OpenMP region during cfg
20989 pass if possible or to entry successor as fallback. Unregister
20990 bb from bb_to_omp_idx.
20992 2021-03-05 Vladimir N. Makarov <vmakarov@redhat.com>
20995 * lra-constraints.c (process_address_1): Skip decomposing address
20996 for asm insn operand with unknown constraint.
20998 2021-03-05 Martin Jambor <mjambor@suse.cz>
21001 * cgraph.c (cgraph_edge::set_call_stmt): Do not update all
21002 corresponding speculative edges if we are about to resolve
21003 sepculation. Make edge direct (and so resolve speculations) before
21004 removing it from call_site_hash.
21005 (cgraph_edge::make_direct): Relax the initial assert to allow calling
21006 the function on speculative direct edges.
21008 2021-03-05 Eric Botcazou <ebotcazou@adacore.com>
21010 PR rtl-optimization/99376
21011 * rtlanal.c (nonzero_bits1) <arithmetic operators>: If the number
21012 of low-order zero bits is too large, set the result to 0 directly.
21014 2021-03-04 Jakub Jelinek <jakub@redhat.com>
21016 PR middle-end/93235
21017 * expmed.c (store_bit_field_using_insv): Return false of xop0 is a
21018 SUBREG and a SUBREG to op_mode can't be created.
21020 2021-03-04 Alex Coplan <alex.coplan@arm.com>
21023 * config/aarch64/aarch64-sve-builtins.cc
21024 (function_resolver::require_vector_type): Handle error_mark_node.
21026 2021-03-04 Ilya Leoshkevich <iii@linux.ibm.com>
21028 * cfgexpand.c (expand_asm_loc): Pass new parameter.
21029 (expand_asm_stmt): Likewise.
21030 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Add new
21032 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
21033 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
21034 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
21035 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
21036 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
21037 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
21038 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
21039 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
21040 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
21041 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
21042 * doc/tm.texi (md_asm_adjust): Likewise.
21043 * target.def (md_asm_adjust): Likewise.
21045 2021-03-04 Richard Biener <rguenther@suse.de>
21047 PR middle-end/97855
21048 * tree-pretty-print.c: Poison pp_printf.
21049 (dump_decl_name): Avoid use of pp_printf.
21050 (dump_block_node): Likewise.
21051 (dump_generic_node): Likewise.
21053 2021-03-04 Martin Sebor <msebor@redhat.com>
21055 PR middle-end/96963
21056 PR middle-end/94655
21057 * builtins.c (handle_array_ref): New helper.
21058 (handle_mem_ref): New helper.
21059 (compute_objsize_r): Factor out ARRAY_REF and MEM_REF handling
21060 into new helper functions. Correct a workaround for vectorized
21063 2021-03-03 Pat Haugen <pthaugen@linux.ibm.com>
21065 * config/rs6000/dfp.md (extendddtd2, trunctddd2, *cmp<mode>_internal1,
21066 floatditd2, ftrunc<mode>2, fix<mode>di2, dfp_ddedpd_<mode>,
21067 dfp_denbcd_<mode>, dfp_dxex_<mode>, dfp_diex_<mode>,
21068 *dfp_sgnfcnc_<mode>, dfp_dscli_<mode>, dfp_dscri_<mode>): Update size
21069 attribute for Power10.
21070 * config/rs6000/mma.md (*movoo): Likewise.
21071 * config/rs6000/rs6000.md (define_attr "size"): Add 256.
21072 (define_mode_attr bits): Add DD/TD modes.
21073 * config/rs6000/sync.md (load_quadpti, store_quadpti, load_lockedpti,
21074 store_conditionalpti): Update size attribute for Power10.
21076 2021-03-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21079 * config/sparc/t-sparc (tree-ssanames.o-warn): Don't error for
21080 -Wuninitialized, -Wmaybe-uninitialized.
21081 (wide-int.o-warn): Likewise.
21083 2021-03-03 Richard Earnshaw <rearnsha@arm.com>
21085 * common/config/arm/arm-common.c: Include configargs.h.
21086 (arm_config_default): New function.
21087 (arm_target_mode): Renamed from arm_target_thumb_only. Handle
21088 processors that do not support Thumb. Take into account the
21089 --with-mode configuration setting for selecting the default.
21090 * config/arm/arm.h (OPTION_DEFAULT_SPECS): Remove entry for 'mode'.
21091 (TARGET_MODE_SPEC_FUNCTIONS): Update for function name change.
21093 2021-03-03 Martin Liska <mliska@suse.cz>
21095 PR gcov-profile/97461
21096 * gcov-io.h (GCOV_PREALLOCATED_KVP): Remove.
21098 2021-03-03 Eric Botcazou <ebotcazou@adacore.com>
21101 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
21102 point back the hard frame pointer to its default location when the
21103 frame is larger than SEH_MAX_FRAME_SIZE.
21105 2021-03-03 Jakub Jelinek <jakub@redhat.com>
21108 * config/i386/predicates.md (logic_operator): New define_predicate.
21109 * config/i386/i386.md (mov + mem using comm arith peephole2):
21110 Punt if operands[1] is EXT_REX_SSE_REGNO_P, AVX512BW is not enabled
21111 and the inner mode is [QH]Imode.
21113 2021-03-03 Jakub Jelinek <jakub@redhat.com>
21116 * dwarf2out.c (dw_loc_list_struct): Add end_entry member.
21117 (new_loc_list): Clear end_entry.
21118 (output_loc_list): Only use DW_LLE_startx_length for -gsplit-dwarf
21119 if HAVE_AS_LEB128, otherwise use DW_LLE_startx_endx. Fix comment
21121 (index_location_lists): For dwarf_version >= 5 without HAVE_AS_LEB128,
21122 initialize also end_entry.
21124 2021-03-03 Jakub Jelinek <jakub@redhat.com>
21127 * cfgrtl.c (fixup_partitions): When changing some bbs from hot to cold
21128 partitions, if in non-layout mode after reorder_blocks also move
21129 affected blocks to ensure a single partition transition.
21131 2021-03-03 Jason Merrill <jason@redhat.com>
21134 * cgraphunit.c (process_function_and_variable_attributes): Don't
21135 warn about flatten on an alias if the target also has it.
21136 * cgraph.h (symtab_node::get_alias_target_tree): New.
21138 2021-03-02 David Edelsohn <dje.gcc@gmail.com>
21140 * config/rs6000/rs6000.md (tls_get_tpointer_internal): Prepend
21141 period to symbol name.
21142 (tls_get_addr_internal<mode>): Same.
21144 2021-03-02 David Malcolm <dmalcolm@redhat.com>
21147 * diagnostic-show-locus.c
21148 (selftest::test_one_liner_many_fixits_2): Fix accidental usage of
21151 2021-03-02 Martin Sebor <msebor@redhat.com>
21153 PR middle-end/99276
21154 * builtins.c (warn_for_access): Remove stray warning text.
21156 2021-03-02 Martin Sebor <msebor@redhat.com>
21158 PR middle-end/99295
21159 * doc/extend.texi (attribute malloc): Reword and clarify nonaliasing
21162 2021-03-02 Jakub Jelinek <jakub@redhat.com>
21165 * dwarf2out.c (output_macinfo_op): Use DW_MACRO_*_str* even with
21166 -gdwarf-5 -gstrict-dwarf. For -gsplit-dwarf -gdwarf-5 use
21167 DW_MACRO_*_strx instead of DW_MACRO_*_strp. Handle
21168 DW_MACRO_define_strx and DW_MACRO_undef_strx.
21169 (save_macinfo_strings): Use DW_MACRO_*_str* even with
21170 -gdwarf-5 -gstrict-dwarf. Handle DW_MACRO_define_strx and
21171 DW_MACRO_undef_strx.
21173 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
21175 * config/s390/s390-builtin-types.def (BT_FN_V4SF_V8HI_UINT): New
21177 (BT_FN_V8HI_V8HI_UINT): Likewise.
21178 (BT_FN_V8HI_V4SF_V4SF_UINT): Likewise.
21179 * config/s390/s390-builtins.def (B_NNPA): New macro definition.
21180 (s390_vclfnhs, s390_vclfnls, s390_vcrnfs, s390_vcfn, s390_vcnf):
21181 New builtin definitions.
21182 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Bump
21183 vector extension version.
21184 * config/s390/s390.c (s390_expand_builtin): Check if builtins are
21185 available with current -march level.
21186 * config/s390/s390.md (UNSPEC_NNPA_VCLFNHS_V8HI)
21187 (UNSPEC_NNPA_VCLFNLS_V8HI, UNSPEC_NNPA_VCRNFS_V8HI)
21188 (UNSPEC_NNPA_VCFN_V8HI, UNSPEC_NNPA_VCNF_V8HI): New constants.
21189 * config/s390/vecintrin.h (vec_extend_to_fp32_hi): New macro.
21190 (vec_extend_to_fp32_lo): Likewise.
21191 (vec_round_from_fp32): Likewise.
21192 (vec_convert_to_fp16): Likewise.
21193 (vec_convert_from_fp16): Likewise.
21194 * config/s390/vx-builtins.md (vclfnhs_v8hi): New insn pattern.
21195 (vclfnls_v8hi): Likewise.
21196 (vcrnfs_v8hi): Likewise.
21197 (vcfn_v8hi): Likewise.
21198 (vcnf_v8hi): Likewise.
21200 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
21202 * common/config/s390/s390-common.c (processor_flags_table): New entry.
21203 * config.gcc: Enable arch14 for --with-arch and --with-tune.
21204 * config/s390/driver-native.c (s390_host_detect_local_cpu): Pick
21205 arch14 for unknown CPU models.
21206 * config/s390/s390-opts.h (enum processor_type): Add PROCESSOR_ARCH14.
21207 * config/s390/s390.c (s390_issue_rate): Add case for PROCESSOR_ARCH14.
21208 (s390_get_sched_attrmask): Likewise.
21209 (s390_get_unit_mask): Likewise.
21210 * config/s390/s390.h (enum processor_flags): Add PF_NNPA and PF_ARCH14.
21211 (TARGET_CPU_ARCH14, TARGET_CPU_ARCH14_P, TARGET_CPU_NNPA)
21212 (TARGET_CPU_NNPA_P, TARGET_ARCH14, TARGET_ARCH14_P, TARGET_NNPA)
21213 (TARGET_NNPA_P): New macro definitions.
21214 * config/s390/s390.md ("cpu_facility", "enabled"): Add arch14 and nnpa.
21215 * config/s390/s390.opt: Add PROCESSOR_ARCH14.
21217 2021-03-02 Jakub Jelinek <jakub@redhat.com>
21219 PR middle-end/95757
21220 * tree-vrp.c (register_edge_assert_for): Remove superfluous ()s around
21221 condition. Call register_edge_assert_for_1 for == 0, != 0, == 1 and
21222 != 1 comparisons if name is lhs of a comparison.
21224 2021-03-01 Iain Sandoe <iain@sandoe.co.uk>
21228 * config/darwin-protos.h (darwin_should_restore_cfa_state): New.
21229 * config/darwin.c (darwin_should_restore_cfa_state): New.
21230 * config/darwin.h (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New.
21231 * doc/tm.texi: Regenerated.
21232 * doc/tm.texi.in: Document TARGET_ASM_SHOULD_RESTORE_CFA_STATE.
21233 * dwarf2cfi.c (connect_traces): If the target requests, restore
21234 the CFA expression after a DW_CFA_restore.
21235 * target.def (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New hook.
21237 2021-03-01 Martin Liska <mliska@suse.cz>
21240 * optc-save-gen.awk: Add 4 more exceptions.
21242 2021-03-01 Nathan Sidwell <nathan@acm.org>
21245 * tree.h (TYPE_ALIGN_RAW): New accessor.
21246 (TYPE_ALIGN): Use it.
21248 2021-03-01 Jan Hubicka <jh@suse.cz>
21251 * ipa-fnsummary.c (compute_fn_summary): Fix sanity check.
21253 2021-03-01 Eric Botcazou <ebotcazou@adacore.com>
21256 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
21257 point the hard frame pointer to the SSE register save area instead
21258 of the general register save area. Perform only minimal adjustment
21259 for small frames if it is initially not correctly aligned.
21260 (ix86_expand_prologue): Remove early saves for a SEH target.
21261 * config/i386/winnt.c (struct seh_frame_state): Document constraint.
21263 2021-02-28 Jakub Jelinek <jakub@redhat.com>
21266 * ipa.c (symbol_table::remove_unreachable_nodes): Fix a comment
21267 typo - referneced -> referenced.
21268 * tree.c (component_ref_size): Fix comment typo -
21269 refernce -> reference.
21270 * tree-ssa-alias.c (access_path_may_continue_p): Fix comment typo -
21271 traling -> trailing.
21272 (aliasing_component_refs_p): Fix comment typos -
21273 refernce -> reference and refernece -> reference and
21274 traling -> trailing.
21275 (nonoverlapping_refs_since_match_p): Fix comment typo -
21276 referneces -> references.
21277 * doc/invoke.texi (--param modref-max-bases): Fix a typo -
21278 referneces -> references.
21280 2021-02-27 Iain Sandoe <iain@sandoe.co.uk>
21282 * config/host-darwin.c (darwin_gt_pch_use_address): Modify
21283 diagnostic message to avoid use of a contraction and format
21286 2021-02-27 Jakub Jelinek <jakub@redhat.com>
21289 * gcse.c (gcse_or_cprop_is_too_expensive): Use %wu instead of
21290 HOST_WIDE_INT_PRINT_UNSIGNED in warning format string.
21291 * ipa-devirt.c (ipa_odr_read_section): Use %wd instead of
21292 HOST_WIDE_INT_PRINT_DEC in inform format string. Fix comment
21295 2021-02-26 Richard Biener <rguenther@suse.de>
21297 PR middle-end/99281
21298 * expr.c (store_field): For calls with return-slot optimization
21299 and addressable return type expand the store directly.
21301 2021-02-26 Richard Biener <rguenther@suse.de>
21304 * builtins.c (warn_string_no_nul): Fix diagnostic formatting.
21306 2021-02-26 Peter Bergner <bergner@linux.ibm.com>
21309 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Replace assert
21312 2021-02-26 Aaron Sawdey <acsawdey@linux.ibm.com>
21314 * config.gcc: Add rs6000-pcrel-opt.o.
21315 * config/rs6000/rs6000-pcrel-opt.c: New file.
21316 * config/rs6000/pcrel-opt.md: New file.
21317 * config/rs6000/predicates.md: Add d_form_memory predicate.
21318 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_PCREL_OPT.
21319 * config/rs6000/rs6000-passes.def: Add pass_pcrel_opt.
21320 * config/rs6000/rs6000-protos.h: Add reg_to_non_prefixed(),
21321 pcrel_opt_valid_mem_p(), output_pcrel_opt_reloc(),
21322 and make_pass_pcrel_opt().
21323 * config/rs6000/rs6000.c (reg_to_non_prefixed): Make global.
21324 (rs6000_option_override_internal): Add pcrel-opt.
21325 (rs6000_delegitimize_address): Support pcrel-opt.
21326 (rs6000_opt_masks): Add pcrel-opt.
21327 (pcrel_opt_valid_mem_p): New function.
21328 (reg_to_non_prefixed): Make global.
21329 (rs6000_asm_output_opcode): Reset prepend_p_to_next_insn.
21330 (output_pcrel_opt_reloc): New function.
21331 * config/rs6000/rs6000.md (loads_extern_addr): New attr.
21332 (pcrel_extern_addr): Set loads_extern_addr.
21333 Add include for pcrel-opt.md.
21334 * config/rs6000/rs6000.opt: Add -mpcrel-opt.
21335 * config/rs6000/t-rs6000: Add rules for pcrel-opt.c and
21338 2021-02-26 YunQiang Su <yunqiang.su@cipunited.com>
21341 * config/mips/mips.c (mips_expand_ext_as_unaligned_load):
21342 If TARGET_64BIT and dest is SUBREG, we check the width, if it
21343 equal to SImode, we use SImode operation, just like what we are
21346 2021-02-26 Marek Polacek <polacek@redhat.com>
21348 * builtins.c (warn_for_access): Fix typos.
21350 2021-02-25 Iain Sandoe <iain@sandoe.co.uk>
21352 * config/aarch64/aarch64.md (<optab>_rol<mode>3): Add a '#'
21353 mark in front of the immediate quantity.
21354 (<optab>_rolsi3_uxtw): Likewise.
21356 2021-02-25 Richard Earnshaw <rearnsha@arm.com>
21359 * config/arm/thumb2.md (nonsecure_call_reg_thumb2_fpcxt): New pattern.
21360 (nonsecure_call_value_reg_thumb2_fpcxt): Likewise.
21361 (nonsecure_call_reg_thumb2): Restrict to using r4 for the callee
21362 address and disable when the FPCXT is not available.
21363 (nonsecure_call_value_reg_thumb2): Likewise.
21365 2021-02-25 Nathan Sidwell <nathan@acm.org>
21368 * doc/invoke.texi (flang-info-module-cmi): Renamed option.
21370 2021-02-25 Tamar Christina <tamar.christina@arm.com>
21372 * tree-vect-slp.c (optimize_load_redistribution_1): Abort on NULL nodes.
21374 2021-02-25 Richard Biener <rguenther@suse.de>
21376 PR tree-optimization/99253
21377 * tree-vect-loop.c (check_reduction_path): First compute
21378 code, then verify out-of-loop uses.
21380 2021-02-25 Jakub Jelinek <jakub@redhat.com>
21383 * match.pd ((T)(A) + CST -> (T)(A + CST)): Add :s to convert.
21385 2021-02-25 Jakub Jelinek <jakub@redhat.com>
21387 PR tree-optimization/80635
21388 * tree-vrp.c (vrp_simplify_cond_using_ranges): Also handle
21389 VIEW_CONVERT_EXPR if modes are the same, innerop is integral and
21390 has mode precision.
21392 2021-02-25 Richard Biener <rguenther@suse.de>
21394 * tree-vect-slp.c (optimize_load_redistribution_1): Delay
21395 load_map population.
21396 (vect_match_slp_patterns_2): Revert part of last change.
21397 (vect_analyze_slp): Do not interleave optimize_load_redistribution
21398 with pattern detection but do it afterwards. Dump the
21399 whole SLP graph after pattern recognition and load
21400 redistribution optimization finished.
21402 2021-02-24 Jakub Jelinek <jakub@redhat.com>
21405 * omp-low.c (struct omp_context): Add teams_nested_p and
21406 nonteams_nested_p members.
21407 (scan_omp_target): Diagnose teams nested inside of target with other
21408 directives strictly nested inside of the same target.
21409 (check_omp_nesting_restrictions): Set ctx->teams_nested_p or
21410 ctx->nonteams_nested_p as needed.
21412 2021-02-24 Vladimir N. Makarov <vmakarov@redhat.com>
21414 PR inline-asm/99123
21415 * lra-constraints.c (uses_hard_regs_p): Don't use decompose_mem_address.
21417 2021-02-24 Hans-Peter Nilsson <hp@axis.com>
21419 * config/cris/cris.c (cris_expand_prologue): Set
21420 current_function_static_stack_size, if flag_stack_usage_info.
21422 2021-02-24 Pat Haugen <pthaugen@linux.ibm.com>
21424 * config/rs6000/rs6000.c (next_insn_prefixed_p): Rename.
21425 (rs6000_final_prescan_insn): Adjust.
21426 (rs6000_asm_output_opcode): Likewise.
21428 2021-02-24 Martin Sebor <msebor@redhat.com>
21430 PR middle-end/97172
21431 * attribs.c (attr_access::free_lang_data): Clear attribute arg spec
21432 from function arguments.
21434 2021-02-24 Tamar Christina <tamar.christina@arm.com>
21436 PR tree-optimization/99220
21437 * tree-vect-slp.c (optimize_load_redistribution_1): Remove
21438 node from cache when it's about to be deleted.
21440 2021-02-24 Jakub Jelinek <jakub@redhat.com>
21442 PR tree-optimization/99225
21443 * fold-const.c (fold_binary_loc) <case NE_EXPR>: In (x & (1 << y)) != 0
21444 to ((x >> y) & 1) != 0 simplifications use build_one_cst instead of
21445 build_int_cst (..., 1). Formatting fixes.
21447 2021-02-24 Tamar Christina <tamar.christina@arm.com>
21449 PR tree-optimization/99149
21450 * tree-vect-slp-patterns.c (vect_detect_pair_op): Don't recreate the
21452 (vect_slp_reset_pattern): Remove.
21453 (complex_fma_pattern::matches): Remove call to vect_slp_reset_pattern.
21454 (complex_mul_pattern::build, complex_fma_pattern::build,
21455 complex_fms_pattern::build): Fix ref counts.
21456 * tree-vect-slp.c (vect_free_slp_tree): Undo SLP only pattern relevancy
21457 when node is being deleted.
21458 (vect_match_slp_patterns_2): Correct result of cache hit on patterns.
21459 (vect_schedule_slp): Invalidate SLP_TREE_REPRESENTATIVE of removed
21461 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize value.
21463 2021-02-24 Matthias Klose <doko@ubuntu.com>
21466 2020-12-07 Matthias Klose <doko@ubuntu.com>
21468 * genextract.c (print_header): Undefine ENABLE_RTL_CHECKING
21469 and ENABLE_RTL_FLAG_CHECKING.
21471 2021-02-24 Richard Biener <rguenther@suse.de>
21474 * builtins.c (fold_builtin_next_arg): Avoid NULL arg.
21476 2021-02-23 Peter Bergner <bergner@linux.ibm.com>
21478 * config/rs6000/mma.md (mma_assemble_pair): Rename from this...
21479 (vsx_assemble_pair): ...to this.
21480 (*mma_assemble_pair): Rename from this...
21481 (*vsx_assemble_pair): ...to this.
21482 (mma_disassemble_pair): Rename from this...
21483 (vsx_disassemble_pair): ...to this.
21484 (*mma_disassemble_pair): Rename from this...
21485 (*vsx_disassemble_pair): ...to this.
21486 * config/rs6000/rs6000-builtin.def (BU_MMA_V2, BU_MMA_V3,
21487 BU_COMPAT): New macros.
21488 (mma_assemble_pair): Rename from this...
21489 (vsx_assemble_pair): ...to this.
21490 (mma_disassemble_pair): Rename from this...
21491 (vsx_disassemble_pair): ...to this.
21492 (mma_assemble_pair): New compatibility built-in.
21493 (mma_disassemble_pair): Likewise.
21494 * config/rs6000/rs6000-call.c (struct builtin_compatibility): New.
21495 (RS6000_BUILTIN_COMPAT): Define.
21496 (bdesc_compat): New.
21497 (mma_expand_builtin): Use VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
21498 (rs6000_gimple_fold_mma_builtin): Use MMA_BUILTIN_DISASSEMBLE_PAIR
21499 and VSX_BUILTIN_ASSEMBLE_PAIR.
21500 (rs6000_init_builtins): Register compatibility built-ins.
21501 (mma_init_builtins): Use VSX_BUILTIN_ASSEMBLE_PAIR,
21502 VSX_BUILTIN_ASSEMBLE_PAIR_INTERNAL, VSX_BUILTIN_DISASSEMBLE_PAIR and
21503 VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
21504 * doc/extend.texi (__builtin_mma_assemble_pair): Rename from this...
21505 (__builtin_vsx_assemble_pair): ...to this.
21506 (__builtin_mma_disassemble_pair): Rename from this...
21507 (__builtin_vsx_disassemble_pair): ...to this.
21509 2021-02-23 Martin Liska <mliska@suse.cz>
21512 * ipa-icf.c (sem_variable::merge): Do not merge 2 variables
21513 with different alignment. That leads to an invalid red zone
21514 size allocated in runtime.
21516 2021-02-23 Jakub Jelinek <jakub@redhat.com>
21518 PR tree-optimization/99204
21519 * fold-const.c (fold_read_from_constant_string): Check that
21520 tree_fits_uhwi_p (index) rather than just that index is INTEGER_CST.
21522 2021-02-23 Segher Boessenkool <segher@kernel.crashing.org>
21523 Kewen Lin <linkw@gcc.gnu.org>
21525 * config/rs6000/rs6000.md (*rotl<mode>3_insert_3): Renamed to...
21526 (rotl<mode>3_insert_3): ...this.
21527 (plus_ior_xor): New code_iterator.
21528 (define_split for GPR rl*imi): New splitter.
21529 * config/rs6000/vsx.md (vsx_init_v4si): Use gen_rotldi3_insert_3
21530 for integer merging.
21532 2021-02-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21534 * config/aarch64/aarch64-tuning-flags.def (cse_sve_vl_constants):
21536 * config/aarch64/aarch64.md (add<mode>3): Force CONST_POLY_INT immediates
21537 into a register when the above is enabled.
21538 * config/aarch64/aarch64.c (neoversev1_tunings):
21539 AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
21540 (aarch64_rtx_costs): Use AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
21542 2021-02-22 Hans-Peter Nilsson <hp@axis.com>
21544 * config/cris/cris.c (cris_print_operand) <'T'>: Change
21545 valid operand from is now an addi mult-value to shift-value.
21546 * config/cris/cris.md (*addi): Change expression of scaled
21547 operand from mult to ashift.
21548 * config/cris/cris.md (*addi_reload): New insn_and_split.
21550 2021-02-22 John David Anglin <danglin@gcc.gnu.org>
21553 * config/pa/pa.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Define as
21554 hook_bool_const_tree_hwi_hwi_const_tree_true.
21555 (pa_asm_output_mi_thunk): Add support for nonzero vcall_offset.
21557 2021-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
21559 PR rtl-optimization/98791
21560 * ira-conflicts.c (process_regs_for_copy): Don't create allocno copies
21561 for unordered modes.
21563 2021-02-22 Martin Liska <mliska@suse.cz>
21565 * tree-inline.c (inline_forbidden_p): Set
21566 inline_forbidden_reason.
21568 2021-02-22 Richard Biener <rguenther@suse.de>
21570 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Dump
21573 2021-02-22 Richard Biener <rguenther@suse.de>
21575 PR tree-optimization/99165
21576 * gimple-ssa-store-merging.c (pass_store_merging::process_store):
21577 Accumulate changed to ret.
21579 2021-02-21 Uros Bizjak <ubizjak@gmail.com>
21582 2020-12-09 Uroš Bizjak <ubizjak@gmail.com>
21584 * config/i386/i386.h (REG_ALLOC_ORDER): Remove
21586 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
21589 * config/s390/vector.md (trunctf<DFP_ALL:mode>2_vr): New
21591 (trunctf<DFP_ALL:mode>2): Likewise.
21592 (trunctdtf2_vr): Likewise.
21593 (trunctdtf2): Likewise.
21594 (extend<DFP_ALL:mode>tf2_vr): Likewise.
21595 (extend<DFP_ALL:mode>tf2): Likewise.
21596 (extendtftd2_vr): Likewise.
21597 (extendtftd2): Likewise.
21599 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
21601 * config/s390/vector.md (*fprx2_to_tf): Rename to fprx2_to_tf,
21602 add memory alternative.
21603 (tf_to_fprx2): New pattern.
21605 2021-02-19 Martin Sebor <msebor@redhat.com>
21608 * attribs.c (init_attr_rdwr_indices): Guard vblist use.
21609 (attr_access::free_lang_data): Remove a spurious test.
21611 2021-02-19 Nathan Sidwell <nathan@acm.org>
21613 * doc/invoke.texi (flang-info-module-read): Document.
21615 2021-02-19 Martin Liska <mliska@suse.cz>
21617 PR translation/99167
21618 * params.opt: Fix typo.
21620 2021-02-19 Richard Biener <rguenther@suse.de>
21622 PR middle-end/99122
21623 * tree-inline.c (inline_forbidden_p): Do not inline functions
21624 with VLA arguments or return value.
21626 2021-02-19 Jakub Jelinek <jakub@redhat.com>
21629 * config/arm/arm.md (*stack_protect_combined_set_insn,
21630 *stack_protect_combined_test_insn): If force_const_mem result
21631 is not valid general operand, force its address into the destination
21634 2021-02-19 Jakub Jelinek <jakub@redhat.com>
21637 * tree-cfg.c (gimple_merge_blocks): If bb a starts with eh landing
21638 pad or non-local label, put FORCED_LABELs from bb b after that label
21639 rather than before it.
21641 2021-02-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
21644 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3): Use
21645 expand_vector_broadcast' to emit the vec_duplicate operand.
21647 2021-02-18 Vladimir N. Makarov <vmakarov@redhat.com>
21649 PR rtl-optimization/96264
21650 * lra-remat.c (reg_overlap_for_remat_p): Check also output insn
21653 2021-02-18 H.J. Lu <hjl.tools@gmail.com>
21656 * varasm.c (get_section): Replace SUPPORTS_SHF_GNU_RETAIN with
21657 looking up the retain attribute.
21658 (resolve_unique_section): Likewise.
21659 (get_variable_section): Likewise.
21660 (switch_to_section): Likewise. Warn when a symbol without the
21661 retain attribute and a symbol with the retain attribute are
21662 placed in the section with the same name, instead of the used
21664 * doc/extend.texi: Document the "retain" attribute.
21666 2021-02-18 Nathan Sidwell <nathan@acm.org>
21669 * doc/invoke.texi (flang-info-include-translate): Document header
21672 2021-02-18 Richard Biener <rguenther@suse.de>
21674 PR middle-end/99122
21675 * ipa-fnsummary.c (analyze_function_body): Set
21676 CIF_FUNCTION_NOT_INLINABLE for VLA parameter calls.
21677 * tree-inline.c (insert_init_debug_bind): Pass NULL for
21678 error_mark_node values.
21679 (force_value_to_type): Do not build V_C_Es for WITH_SIZE_EXPR
21681 (setup_one_parameter): Delay force_value_to_type until when
21684 2021-02-18 Hans-Peter Nilsson <hp@axis.com>
21686 PR tree-optimization/99142
21687 * match.pd (clz cmp 0): Gate replacement on single_use of clz result.
21689 2021-02-18 Jakub Jelinek <jakub@redhat.com>
21691 * wide-int-bitmask.h (wide_int_bitmask::wide_int_bitmask (),
21692 wide_int_bitmask::wide_int_bitmask (uint64_t),
21693 wide_int_bitmask::wide_int_bitmask (uint64_t, uint64_t),
21694 wide_int_bitmask::operator ~ () const,
21695 wide_int_bitmask::operator | (wide_int_bitmask) const,
21696 wide_int_bitmask::operator & (wide_int_bitmask) const): Use constexpr
21698 * config/i386/i386.h (PTA_3DNOW, PTA_3DNOW_A, PTA_64BIT, PTA_ABM,
21699 PTA_AES, PTA_AVX, PTA_BMI, PTA_CX16, PTA_F16C, PTA_FMA, PTA_FMA4,
21700 PTA_FSGSBASE, PTA_LWP, PTA_LZCNT, PTA_MMX, PTA_MOVBE, PTA_NO_SAHF,
21701 PTA_PCLMUL, PTA_POPCNT, PTA_PREFETCH_SSE, PTA_RDRND, PTA_SSE, PTA_SSE2,
21702 PTA_SSE3, PTA_SSE4_1, PTA_SSE4_2, PTA_SSE4A, PTA_SSSE3, PTA_TBM,
21703 PTA_XOP, PTA_AVX2, PTA_BMI2, PTA_RTM, PTA_HLE, PTA_PRFCHW, PTA_RDSEED,
21704 PTA_ADX, PTA_FXSR, PTA_XSAVE, PTA_XSAVEOPT, PTA_AVX512F, PTA_AVX512ER,
21705 PTA_AVX512PF, PTA_AVX512CD, PTA_NO_TUNE, PTA_SHA, PTA_PREFETCHWT1,
21706 PTA_CLFLUSHOPT, PTA_XSAVEC, PTA_XSAVES, PTA_AVX512DQ, PTA_AVX512BW,
21707 PTA_AVX512VL, PTA_AVX512IFMA, PTA_AVX512VBMI, PTA_CLWB, PTA_MWAITX,
21708 PTA_CLZERO, PTA_NO_80387, PTA_PKU, PTA_AVX5124VNNIW, PTA_AVX5124FMAPS,
21709 PTA_AVX512VPOPCNTDQ, PTA_SGX, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES,
21710 PTA_AVX512VBMI2, PTA_VPCLMULQDQ, PTA_AVX512BITALG, PTA_RDPID,
21711 PTA_PCONFIG, PTA_WBNOINVD, PTA_AVX512VP2INTERSECT, PTA_PTWRITE,
21712 PTA_AVX512BF16, PTA_WAITPKG, PTA_MOVDIRI, PTA_MOVDIR64B, PTA_ENQCMD,
21713 PTA_CLDEMOTE, PTA_SERIALIZE, PTA_TSXLDTRK, PTA_AMX_TILE, PTA_AMX_INT8,
21714 PTA_AMX_BF16, PTA_UINTR, PTA_HRESET, PTA_KL, PTA_WIDEKL, PTA_AVXVNNI,
21715 PTA_X86_64_BASELINE, PTA_X86_64_V2, PTA_X86_64_V3, PTA_X86_64_V4,
21716 PTA_CORE2, PTA_NEHALEM, PTA_WESTMERE, PTA_SANDYBRIDGE, PTA_IVYBRIDGE,
21717 PTA_HASWELL, PTA_BROADWELL, PTA_SKYLAKE, PTA_SKYLAKE_AVX512,
21718 PTA_CASCADELAKE, PTA_COOPERLAKE, PTA_CANNONLAKE, PTA_ICELAKE_CLIENT,
21719 PTA_ICELAKE_SERVER, PTA_TIGERLAKE, PTA_SAPPHIRERAPIDS, PTA_ALDERLAKE,
21720 PTA_KNL, PTA_BONNELL, PTA_SILVERMONT, PTA_GOLDMONT, PTA_GOLDMONT_PLUS,
21721 PTA_TREMONT, PTA_KNM): Use constexpr instead of const.
21723 2021-02-18 Jakub Jelinek <jakub@redhat.com>
21725 PR middle-end/99109
21726 * gimple-array-bounds.cc (build_zero_elt_array_type): Rename to ...
21727 (build_printable_array_type): ... this. Add nelts argument. For
21728 overaligned eltype, use TYPE_MAIN_VARIANT (eltype) instead. If
21729 nelts, call build_array_type_nelts.
21730 (array_bounds_checker::check_mem_ref): Use build_printable_array_type
21731 instead of build_zero_elt_array_type and build_array_type_nelts.
21733 2021-02-18 Jakub Jelinek <jakub@redhat.com>
21736 * config/i386/i386.c (distance_non_agu_define): Don't call
21737 extract_insn_cached here.
21738 (ix86_lea_outperforms): Save and restore recog_data around call
21739 to distance_non_agu_define and distance_agu_use.
21740 (ix86_ok_to_clobber_flags): Remove.
21741 (ix86_avoid_lea_for_add): Don't call ix86_ok_to_clobber_flags.
21742 (ix86_avoid_lea_for_addr): Likewise. Adjust function comment.
21743 * config/i386/i386.md (*lea<mode>): Change from define_insn_and_split
21744 into define_insn. Move the splitting to define_peephole2 and
21745 check there using peep2_regno_dead_p if FLAGS_REG is dead.
21747 2021-02-17 Julian Brown <julian@codesourcery.com>
21749 * gimplify.c (gimplify_scan_omp_clauses): Handle ATTACH_DETACH
21752 2021-02-17 Xi Ruoyao <xry111@mengyan1223.wang>
21755 * config/mips/mips.c (mips_symbol_insns): Do not use
21756 MSA_SUPPORTED_MODE_P if mode is MAX_MACHINE_MODE.
21758 2021-02-16 Vladimir N. Makarov <vmakarov@redhat.com>
21760 PR inline-asm/98096
21761 * stmt.c (resolve_operand_name_1): Take inout operands into account
21762 for access to labels by names.
21763 * doc/extend.texi: Describe counting operands for accessing labels.
21765 2021-02-16 Richard Biener <rguenther@suse.de>
21767 PR tree-optimization/38474
21768 * tree-ssa-structalias.c (variable_info::address_taken): New.
21769 (new_var_info): Initialize address_taken.
21770 (process_constraint): Set address_taken.
21771 (solve_constraints): Use the new address_taken flag rather
21772 than is_reg_var for sorting variables.
21773 (dump_constraint): Dump the variable number if the name
21776 2021-02-16 Jakub Jelinek <jakub@redhat.com>
21779 * tree-vect-stmts.c (vectorizable_simd_clone_call): For num_calls != 1
21780 multiply by 4096 and for inbranch by 8192.
21781 * config/i386/i386.c (ix86_simd_clone_usable): For TARGET_AVX512F,
21782 return 3, 2 or 1 for mangle letters 'b', 'c' or 'd'.
21784 2021-02-15 Maya Rashish <coypu@sdf.org>
21786 * config/aarch64/aarch64.c (aarch64_init_builtins):
21787 Call SUBTARGET_INIT_BUILTINS.
21789 2021-02-15 Peter Bergner <bergner@linux.ibm.com>
21791 PR rtl-optimization/98872
21792 * init-regs.c (initialize_uninitialized_regs): Skip initialization
21793 if CONST0_RTX is NULL.
21795 2021-02-15 Richard Sandiford <richard.sandiford@arm.com>
21797 PR rtl-optimization/98863
21798 * rtl-ssa/functions.h (function_info::bb_live_out_info): Delete.
21799 (function_info::build_info): Turn into a declaration, moving the
21800 definition to internals.h.
21801 (function_info::bb_walker): Declare.
21802 (function_info::create_reg_use): Likewise.
21803 (function_info::calculate_potential_phi_regs): Take a build_info
21805 (function_info::place_phis, function_info::create_ebbs): Declare.
21806 (function_info::calculate_ebb_live_in_for_debug): Likewise.
21807 (function_info::populate_backedge_phis): Delete.
21808 (function_info::start_block, function_info::end_block): Declare.
21809 (function_info::populate_phi_inputs): Delete.
21810 (function_info::m_potential_phi_regs): Move information to build_info.
21811 * rtl-ssa/internals.h: New file.
21812 (function_info::bb_phi_info): New class.
21813 (function_info::build_info): Moved from functions.h.
21814 Add a constructor and destructor.
21815 (function_info::build_info::ebb_use): Delete.
21816 (function_info::build_info::ebb_def): Likewise.
21817 (function_info::build_info::bb_live_out): Likewise.
21818 (function_info::build_info::tmp_ebb_live_in_for_debug): New variable.
21819 (function_info::build_info::potential_phi_regs): Likewise.
21820 (function_info::build_info::potential_phi_regs_for_debug): Likewise.
21821 (function_info::build_info::ebb_def_regs): Likewise.
21822 (function_info::build_info::bb_phis): Likewise.
21823 (function_info::build_info::bb_mem_live_out): Likewise.
21824 (function_info::build_info::bb_to_rpo): Likewise.
21825 (function_info::build_info::def_stack): Likewise.
21826 (function_info::build_info::old_def_stack_limit): Likewise.
21827 * rtl-ssa/internals.inl (function_info::build_info::record_reg_def):
21828 Remove the regno argument. Push the previous definition onto the
21829 definition stack where necessary.
21830 * rtl-ssa/accesses.cc: Include internals.h.
21831 * rtl-ssa/changes.cc: Likewise.
21832 * rtl-ssa/blocks.cc: Likewise.
21833 (function_info::build_info::build_info): Define.
21834 (function_info::build_info::~build_info): Likewise.
21835 (function_info::bb_walker): New class.
21836 (function_info::bb_walker::bb_walker): Define.
21837 (function_info::add_live_out_use): Convert a logarithmic-complexity
21838 test into a linear one. Allow the same definition to be passed
21840 (function_info::calculate_potential_phi_regs): Moved from
21841 functions.cc. Take a build_info parameter and store the
21842 information there instead.
21843 (function_info::place_phis): New function.
21844 (function_info::add_entry_block_defs): Update call to record_reg_def.
21845 (function_info::calculate_ebb_live_in_for_debug): New function.
21846 (function_info::add_phi_nodes): Use bb_phis to decide which
21847 registers need phi nodes and initialize ebb_def_regs accordingly.
21848 Do not add degenerate phis here.
21849 (function_info::add_artificial_accesses): Use create_reg_use.
21850 Assert that all definitions are listed in the DF LR sets.
21851 Update call to record_reg_def.
21852 (function_info::record_block_live_out): Record live-out register
21853 values in the phis of successor blocks. Use the live-out set
21854 when processing the last block in an EBB, instead of always
21855 using the live-in sets of successor blocks. AND the live sets
21856 with the set of registers that have been defined in the EBB,
21857 rather than with all potential phi registers. Cope correctly
21858 with branches back to the start of the current EBB.
21859 (function_info::start_block): New function.
21860 (function_info::end_block): Likewise.
21861 (function_info::populate_phi_inputs): Likewise.
21862 (function_info::create_ebbs): Likewise.
21863 (function_info::process_all_blocks): Rewrite into a multi-phase
21865 * rtl-ssa/functions.cc: Include internals.h.
21866 (function_info::calculate_potential_phi_regs): Move to blocks.cc.
21867 (function_info::init_function_data): Remove caller.
21868 * rtl-ssa/insns.cc: Include internals.h
21869 (function_info::create_reg_use): New function. Lazily any
21870 degenerate phis needed by the linear RPO view.
21871 (function_info::record_use): Use create_reg_use. When processing
21872 debug uses, use potential_phi_regs and test it before checking
21873 whether the register is live on entry to the current EBB. Lazily
21874 calculate ebb_live_in_for_debug.
21875 (function_info::record_call_clobbers): Update call to record_reg_def.
21876 (function_info::record_def): Likewise.
21878 2021-02-15 Martin Liska <mliska@suse.cz>
21880 * toplev.c (init_asm_output): Free output of
21881 gen_command_line_string function.
21882 (process_options): Likewise.
21884 2021-02-15 Martin Liska <mliska@suse.cz>
21886 * params.opt: Add 2 missing Param keywords.
21888 2021-02-15 Eric Botcazou <ebotcazou@adacore.com>
21890 * df-core.c (df_worklist_dataflow_doublequeue): Use proper cast.
21892 2021-02-15 Jakub Jelinek <jakub@redhat.com>
21894 PR tree-optimization/99079
21895 * match.pd (A % (pow2pcst << N) -> A & ((pow2pcst << N) - 1)): Remove
21896 useless tree_nop_conversion_p (type, TREE_TYPE (@3)) check. Instead
21897 require both type and TREE_TYPE (@1) to be integral types and either
21898 type having smaller or equal precision, or TREE_TYPE (@1) being
21899 unsigned type, or type being signed type. If TREE_TYPE (@1)
21900 doesn't have wrapping overflow, perform the subtraction of one in
21903 2021-02-14 Jan Hubicka <hubicka@ucw.cz>
21904 Richard Biener <rguether@suse.de>
21907 * ipa-reference.c (ipa_init): Only conditinally initialize
21908 reference_vars_to_consider.
21909 (propagate): Conditionally deninitialize reference_vars_to_consider.
21910 (ipa_reference_write_optimization_summary): Sanity check that
21911 reference_vars_to_consider is not allocated.
21913 2021-02-13 Levy Hsu <admin@levyhsu.com>
21916 * config/riscv/riscv-shorten-memrefs.c (pass_shorten_memrefs): Add
21917 extend parameter to get_si_mem_base_reg declaration.
21918 (get_si_mem_base_reg): Add extend parameter. Set it.
21919 (analyze): Pass extend arg to get_si_mem_base_reg.
21920 (transform): Likewise. Use it when rewriting mems.
21921 * config/riscv/riscv.c (riscv_legitimize_move): Check for subword
21922 loads and emit sign/zero extending load followed by subreg move.
21924 2021-02-13 Jim Wilson <jimw@sifive.com>
21927 * config/riscv/riscv.c (riscv_compressed_lw_address_p): Drop early
21928 exit when !reload_completed. Only perform check for compressed reg
21929 if reload_completed.
21930 (riscv_rtx_costs): In MEM case, when optimizing for size and
21931 shorten memrefs, if not compressible, then increase cost.
21933 2021-02-13 Jakub Jelinek <jakub@redhat.com>
21935 PR rtl-optimization/98439
21936 * recog.c (pass_split_before_regstack::gate): Enable even when
21937 pass_split_before_sched2 is enabled if -fselective-scheduling2 is
21940 2021-02-13 Jakub Jelinek <jakub@redhat.com>
21943 * config/i386/mmx.md (*mmx_pshufd_1): Add a combine splitter for
21944 swap of V2SImode elements in memory into DImode memory rotate by 32.
21946 2021-02-12 Martin Sebor <msebor@redhat.com>
21948 * tree-pretty-print.c (print_generic_expr_to_str): Update comment.
21950 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
21952 * rtl-ssa/accesses.cc (function_info::make_use_available): Use
21953 m_temp_obstack rather than m_obstack to allocate the temporary use.
21955 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
21957 * df-problems.c (df_lr_bb_local_compute): Treat partial definitions
21958 as read-modify operations.
21960 2021-02-12 Richard Biener <rguenther@suse.de>
21962 PR middle-end/38474
21963 * ipa-fnsummary.c (unmodified_parm_1): Only walk when
21964 fbi->aa_walk_budget is bigger than zero. Update
21965 fbi->aa_walk_budget.
21966 (param_change_prob): Likewise.
21967 * ipa-prop.c (detect_type_change_from_memory_writes):
21968 Properly account walk_aliased_vdefs.
21969 (parm_preserved_before_stmt_p): Canonicalize updates.
21970 (parm_ref_data_preserved_p): Likewise.
21971 (parm_ref_data_pass_through_p): Likewise.
21972 (determine_known_aggregate_parts): Account own alias queries.
21974 2021-02-12 Martin Liska <mliska@suse.cz>
21976 * opts-common.c (decode_cmdline_option): Release werror_arg.
21977 * opts.c (gen_producer_string): Release output of
21978 gen_command_line_string.
21980 2021-02-12 Richard Biener <rguenther@suse.de>
21982 PR tree-optimization/38474
21983 * params.opt (-param=max-store-chains-to-track=): New param.
21984 (-param=max-stores-to-track=): Likewise.
21985 * doc/invoke.texi (max-store-chains-to-track): Document.
21986 (max-stores-to-track): Likewise.
21987 * gimple-ssa-store-merging.c (pass_store_merging::m_n_chains):
21989 (pass_store_merging::m_n_stores): Likewise.
21990 (pass_store_merging::terminate_and_process_chain): Update
21991 m_n_stores and m_n_chains.
21992 (pass_store_merging::process_store): Likewise. Terminate
21993 oldest chains if the number of stores or chains get too large.
21994 (imm_store_chain_info::terminate_and_process_chain): Dump
21997 2021-02-11 Eric Botcazou <ebotcazou@adacore.com>
21999 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
22000 the cold section, emit a nop before the directive if the previous
22001 active instruction can throw.
22003 2021-02-11 Peter Bergner <bergner@linux.ibm.com>
22006 * config/rs6000/predicates.md (mma_assemble_input_operand): Restrict
22007 memory addresses that are legal for quad word accesses.
22009 2021-02-11 Andrea Corallo <andrea.corallo@arm.com>
22012 * config/arm/thumb2.md (*doloop_end_internal): Generate
22013 alternative sequence to handle long range branches.
22015 2021-02-11 Joel Hutton <joel.hutton@arm.com>
22017 PR tree-optimization/98772
22018 * optabs-tree.c (supportable_half_widening_operation): New function
22019 to check for supportable V8QI->V8HI widening patterns.
22020 * optabs-tree.h (supportable_half_widening_operation): New function.
22021 * tree-vect-stmts.c (vect_create_half_widening_stmts): New function
22022 to create promotion stmts for V8QI->V8HI widening patterns.
22023 (vectorizable_conversion): Add case for V8QI->V8HI.
22025 2021-02-11 Richard Biener <rguenther@suse.de>
22027 * sparseset.h (SPARSESET_ELT_BITS): Remove.
22028 (SPARSESET_ELT_TYPE): Use unsigned int.
22029 * fwprop.c: Do not include sparseset.h.
22031 2021-02-10 Jakub Jelinek <jakub@redhat.com>
22034 * varasm.c (declare_weak): For -fsyntax-only, allow even
22035 TREE_ASM_WRITTEN function decls.
22037 2021-02-10 Jakub Jelinek <jakub@redhat.com>
22040 * config/i386/sse.md (fix<fixunssuffix>_truncv2sfv2di2,
22041 <insn>v8qiv8hi2, <insn>v8qiv8si2, <insn>v4qiv4si2, <insn>v4hiv4si2,
22042 <insn>v8qiv8di2, <insn>v4qiv4di2, <insn>v2qiv2di2, <insn>v4hiv4di2,
22043 <insn>v2hiv2di2, <insn>v2siv2di2): Force operands[1] into REG before
22044 calling simplify_gen_subreg on it.
22046 2021-02-10 Martin Liska <mliska@suse.cz>
22048 * config/nvptx/nvptx.c (nvptx_option_override): Use
22049 flag_patchable_function_entry instead of the removed
22050 function_entry_patch_area_size.
22052 2021-02-10 Martin Liska <mliska@suse.cz>
22054 PR tree-optimization/99002
22055 PR tree-optimization/99026
22056 * gimple-if-to-switch.cc (if_chain::is_beneficial): Fix memory
22057 leak when adjacent cases are merged.
22058 * tree-switch-conversion.c (switch_decision_tree::analyze_switch_statement): Use
22060 (make_pass_lower_switch): Remove trailing whitespace.
22061 * tree-switch-conversion.h (release_clusters): New.
22063 2021-02-10 Richard Biener <rguenther@suse.de>
22065 PR rtl-optimization/99054
22066 * cfgrtl.c (rtl-optimization/99054): Return an auto_vec.
22067 (fixup_partitions): Adjust.
22068 (rtl_verify_edges): Likewise.
22070 2021-02-10 Jakub Jelinek <jakub@redhat.com>
22072 PR middle-end/99007
22073 * gimplify.c (gimplify_scan_omp_clauses): For MEM_REF on reductions,
22074 temporarily disable gimplify_ctxp->into_ssa around gimplify_expr
22077 2021-02-10 Richard Biener <rguenther@suse.de>
22080 * ipa-pure-const.c (propagate_malloc): Use an auto_vec<>
22083 2021-02-10 Richard Biener <rguenther@suse.de>
22085 PR tree-optimization/99024
22086 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Only
22087 clear loop->aux if it is associated with the destroyed loop_vinfo.
22089 2021-02-10 Martin Liska <mliska@suse.cz>
22091 PR tree-optimization/99002
22092 * gimple-if-to-switch.cc (find_conditions): Fix memory leak
22095 2021-02-10 Martin Liska <mliska@suse.cz>
22098 * ipa-icf.c (sem_item::add_reference): Fix memory leak when
22099 a reference exists.
22101 2021-02-10 Jakub Jelinek <jakub@redhat.com>
22104 * dwarf2out.c (prune_unused_types_walk): Mark DW_TAG_variable DIEs
22105 at class scope for DWARF5+.
22107 2021-02-09 Eric Botcazou <ebotcazou@adacore.com>
22109 PR rtl-optimization/96015
22110 * reorg.c (skip_consecutive_labels): Minor comment tweaks.
22111 (relax_delay_slots): When deleting a jump to the next active
22112 instruction over a barrier, first delete the barrier if the
22113 jump is the only way to reach the target label.
22115 2021-02-09 Andre Vieira <andre.simoesdiasvieira@arm.com>
22117 * config/aarch64/aarch64-cost-tables.h: Add entries for vect.mul.
22118 * config/aarch64/aarch64.c (aarch64_rtx_mult_cost): Use vect.mul for
22119 vector multiplies and vect.alu for SSRA.
22120 * config/arm/aarch-common-protos.h (struct vector_cost_table): Define
22121 vect.mul cost field.
22122 * config/arm/aarch-cost-tables.h: Add entries for vect.mul.
22123 * config/arm/arm.c: Likewise.
22125 2021-02-09 Richard Biener <rguenther@suse.de>
22127 PR tree-optimization/98863
22128 * tree-ssa-sccvn.h (vn_avail::next_undo): Add.
22129 * tree-ssa-sccvn.c (last_pushed_avail): New global.
22130 (rpo_elim::eliminate_push_avail): Chain pushed avails.
22131 (unwind_state::avail_top): Add.
22132 (do_unwind): Rewrite unwinding of avail entries.
22133 (do_rpo_vn): Initialize last_pushed_avail and
22134 avail_top of the undo state.
22136 2021-02-09 Jakub Jelinek <jakub@redhat.com>
22138 PR middle-end/99004
22139 * calls.c (maybe_warn_rdwr_sizes): Change s0 and s1 type from
22140 const char * to char * and free those pointers after use.
22142 2021-02-09 Richard Biener <rguenther@suse.de>
22144 PR tree-optimization/99017
22145 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Allow
22146 zero vector cost entries.
22148 2021-02-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
22150 PR middle-end/98974
22151 * tree-vect-stmts.c (vectorizable_condition): Remove shadow vec_num
22152 parameter in vectorizable_condition.
22154 2021-02-08 Richard Biener <rguenther@suse.de>
22157 * tree.c (walk_tree_1): Walk VECTOR_CST elements.
22159 2021-02-08 Martin Liska <mliska@suse.cz>
22162 * cfgexpand.c (pass_expand::execute): Parse per-function option
22163 flag_patchable_function_entry and use it.
22164 * common.opt: Remove function_entry_patch_area_size and
22165 function_entry_patch_area_start global variables.
22166 * opts.c (parse_and_check_patch_area): New function.
22167 (common_handle_option): Use it.
22168 * opts.h (parse_and_check_patch_area): New function.
22169 * toplev.c (process_options): Parse and use
22170 function_entry_patch_area_size.
22172 2021-02-08 Martin Sebor <msebor@redhat.com>
22174 * doc/extend.texi (attribute malloc): Correct typos.
22176 2021-02-05 Nathan Sidwell <nathan@acm.org>
22179 * gcc.c (driver::maybe_run_linker): Check for input file
22180 accessibility if not linking.
22182 2021-02-05 Richard Biener <rguenther@suse.de>
22184 PR tree-optimization/98855
22185 * tree-vectorizer.h (add_stmt_cost): New overload.
22186 * tree-vect-slp.c (li_cost_vec_cmp): New.
22187 (vect_bb_slp_scalar_cost): Cost individual loop regions
22188 separately. Account for the scalar instance root stmt.
22190 2021-02-05 Tom de Vries <tdevries@suse.de>
22193 * tree-switch-conversion.c (jump_table_cluster::emit): Add loc
22195 (bit_test_cluster::emit): Reuse location_t for newly created
22197 (switch_decision_tree::try_switch_expansion): Preserve
22199 * tree-switch-conversion.h: Change function signatures.
22201 2021-02-05 Jakub Jelinek <jakub@redhat.com>
22204 * config/i386/i386-options.c (m_NONE, m_ALL): Define.
22205 * config/i386/x86-tune.def (X86_TUNE_BRANCH_PREDICTION_HINTS,
22206 X86_TUNE_PROMOTE_QI_REGS): Use m_NONE instead of 0U.
22207 (X86_TUNE_QIMODE_MATH): Use m_ALL instead of ~0U.
22209 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22211 * config/aarch64/aarch64-simd-builtins.def (get_high): Define builtin.
22212 * config/aarch64/aarch64-simd.md (aarch64_get_high<mode>): Define.
22213 * config/aarch64/arm_neon.h (__GET_HIGH): Delete.
22214 (vget_high_f16): Reimplement using new builtin.
22215 (vget_high_f32): Likewise.
22216 (vget_high_f64): Likewise.
22217 (vget_high_p8): Likewise.
22218 (vget_high_p16): Likewise.
22219 (vget_high_p64): Likewise.
22220 (vget_high_s8): Likewise.
22221 (vget_high_s16): Likewise.
22222 (vget_high_s32): Likewise.
22223 (vget_high_s64): Likewise.
22224 (vget_high_u8): Likewise.
22225 (vget_high_u16): Likewise.
22226 (vget_high_u32): Likewise.
22227 (vget_high_u64): Likewise.
22229 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22231 * config/aarch64/aarch64-simd-builtins.def (get_low): Define builtin.
22232 * config/aarch64/aarch64-simd.md (aarch64_get_low<mode>): Define.
22233 * config/aarch64/arm_neon.h (__GET_LOW): Delete.
22234 (vget_low_f16): Reimplement using new builtin.
22235 (vget_low_f32): Likewise.
22236 (vget_low_f64): Likewise.
22237 (vget_low_p8): Likewise.
22238 (vget_low_p16): Likewise.
22239 (vget_low_p64): Likewise.
22240 (vget_low_s8): Likewise.
22241 (vget_low_s16): Likewise.
22242 (vget_low_s32): Likewise.
22243 (vget_low_s64): Likewise.
22244 (vget_low_u8): Likewise.
22245 (vget_low_u16): Likewise.
22246 (vget_low_u32): Likewise.
22247 (vget_low_u64): Likewise.
22249 2021-02-05 Kito Cheng <kito.cheng@sifive.com>
22251 * gcc.c (print_multilib_info): Check all required argument is provided
22254 2021-02-05 liuhongt <hongtao.liu@intel.com>
22257 * config/i386/i386-expand.c (ix86_expand_sse_cmp): Don't
22258 generate integer mask comparison for 128/256-bits vector when
22259 op_true/op_false is NULL_RTX or CONSTM1_RTX/CONST0_RTX. Also
22260 delete redundant !maskcmp condition.
22261 (ix86_expand_int_vec_cmp): Ditto but no redundant deletion
22263 (ix86_expand_sse_movcc): Delete definition of maskcmp, add the
22264 condition directly to if (maskcmp), add extra check for
22265 cmpmode, it should be MODE_INT.
22266 (ix86_expand_fp_vec_cmp): Pass NULL to ix86_expand_sse_cmp's
22267 parameters op_true/op_false.
22268 (ix86_use_mask_cmp_p): New.
22270 2021-02-05 liuhongt <hongtao.liu@intel.com>
22273 * config/i386/x86-tune.def (X86_TUNE_AVX256_UNALIGNED_LOAD_OPTIMAL):
22274 Remove m_GENERIC from ~list.
22275 (X86_TUNE_AVX256_UNALIGNED_STORE_OPTIMAL): Ditto.
22277 2021-02-04 David Malcolm <dmalcolm@redhat.com>
22280 * diagnostic-show-locus.c (compatible_locations_p): Require
22281 locations in the same macro map to be either both from the
22282 macro definition, or both from the macro arguments.
22284 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
22286 * config/aarch64/aarch64-simd-builtins.def: Add
22287 [su]mull_hi_lane[q] builtin generator macros.
22288 * config/aarch64/aarch64-simd.md
22289 (aarch64_<su>mull_hi_lane<mode>_insn): Define.
22290 (aarch64_<su>mull_hi_lane<mode>): Define.
22291 (aarch64_<su>mull_hi_laneq<mode>_insn): Define.
22292 (aarch64_<su>mull_hi_laneq<mode>): Define.
22293 * config/aarch64/arm_neon.h (vmull_high_lane_s16): Use RTL
22294 builtin instead of inline asm.
22295 (vmull_high_lane_s32): Likewise.
22296 (vmull_high_lane_u16): Likewise.
22297 (vmull_high_lane_u32): Likewise.
22298 (vmull_high_laneq_s16): Likewise.
22299 (vmull_high_laneq_s32): Likewise.
22300 (vmull_high_laneq_u16): Likewise.
22301 (vmull_high_laneq_u32): Liekwise.
22303 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
22305 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_hi_n
22306 builtin generator macros.
22307 * config/aarch64/aarch64-simd.md
22308 (aarch64_<su>mull_hi_n<mode>_insn): Define.
22309 (aarch64_<su>mull_hi_n<mode>): Define.
22310 * config/aarch64/arm_neon.h (vmull_high_n_s16): Use RTL builtin
22311 instead of inline asm.
22312 (vmull_high_n_s32): Likewise.
22313 (vmull_high_n_u16): Likewise.
22314 (vmull_high_n_u32): Likewise.
22316 2021-02-04 Richard Biener <rguenther@suse.de>
22318 PR tree-optimization/98855
22319 * tree-vect-loop.c (vectorizable_phi): Do not cost
22320 single-argument PHIs.
22321 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Likewise.
22322 * tree-vect-stmts.c (vectorizable_bswap): Also perform
22323 costing for SLP operation.
22325 2021-02-04 Martin Liska <mliska@suse.cz>
22327 * doc/extend.texi: Mention -mprefer-vector-width in target
22330 2021-02-03 Martin Sebor <msebor@redhat.com>
22332 PR tree-optimization/98937
22333 * tree-ssa-strlen.c (strlen_dom_walker::~strlen_dom_walker): Define.
22334 Flush pointer_query cache.
22336 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
22338 * config/rs6000/genfusion.pl (gen_2logical): Add missing
22339 fixes based on patch review.
22340 * config/rs6000/fusion.md: Regenerate file.
22342 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
22344 * config/rs6000/t-rs6000: Comment out auto generation of
22347 2021-02-03 Andrew Stubbs <ams@codesourcery.com>
22349 * config/gcn/gcn-opts.h (enum processor_type): Add PROCESSOR_GFX908.
22350 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Add gfx908.
22351 (output_file_start): Add gfx908.
22352 * config/gcn/gcn.opt (gpu_type): Add gfx908.
22353 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Add march=gfx908.
22354 (MULTILIB_DIRNAMES): Add gfx908.
22355 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): New define.
22356 (main): Recognize gfx908.
22357 * config/gcn/t-omp-device: Add gfx908.
22359 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22361 * config/aarch64/aarch64-simd-builtins.def: Add
22362 [su]mlsl_hi_lane[q] builtin macro generators.
22363 * config/aarch64/aarch64-simd.md
22364 (aarch64_<su>mlsl_hi_lane<mode>_insn): Define.
22365 (aarch64_<su>mlsl_hi_lane<mode>): Define.
22366 (aarch64_<su>mlsl_hi_laneq<mode>_insn): Define.
22367 (aarch64_<su>mlsl_hi_laneq<mode>): Define.
22368 * config/aarch64/arm_neon.h (vmlsl_high_lane_s16): Use RTL
22369 builtin instead of inline asm.
22370 (vmlsl_high_lane_s32): Likewise.
22371 (vmlsl_high_lane_u16): Likewise.
22372 (vmlsl_high_lane_u32): Likewise.
22373 (vmlsl_high_laneq_s16): Likewise.
22374 (vmlsl_high_laneq_s32): Likewise.
22375 (vmlsl_high_laneq_u16): Likewise.
22376 (vmlsl_high_laneq_u32): Likewise.
22377 (vmlal_high_laneq_u32): Likewise.
22379 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22381 * config/aarch64/aarch64-simd-builtins.def: Add
22382 [su]mlal_hi_lane[q] builtin generator macros.
22383 * config/aarch64/aarch64-simd.md
22384 (aarch64_<su>mlal_hi_lane<mode>_insn): Define.
22385 (aarch64_<su>mlal_hi_lane<mode>): Define.
22386 (aarch64_<su>mlal_hi_laneq<mode>_insn): Define.
22387 (aarch64_<su>mlal_hi_laneq<mode>): Define.
22388 * config/aarch64/arm_neon.h (vmlal_high_lane_s16): Use RTL
22389 builtin instead of inline asm.
22390 (vmlal_high_lane_s32): Likewise.
22391 (vmlal_high_lane_u16): Likewise.
22392 (vmlal_high_lane_u32): Likewise.
22393 (vmlal_high_laneq_s16): Likewise.
22394 (vmlal_high_laneq_s32): Likewise.
22395 (vmlal_high_laneq_u16): Likewise.
22396 (vmlal_high_laneq_u32): Likewise.
22398 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22400 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_hi_n
22401 builtin generator macros.
22402 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_hi_n<mode>_insn):
22404 (aarch64_<su>mlsl_hi_n<mode>): Define.
22405 * config/aarch64/arm_neon.h (vmlsl_high_n_s16): Use RTL builtin
22406 instead of inline asm.
22407 (vmlsl_high_n_s32): Likewise.
22408 (vmlsl_high_n_u16): Likewise.
22409 (vmlsl_high_n_u32): Likewise.
22411 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22413 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_hi_n
22414 builtin generator macros.
22415 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_hi_n<mode>_insn):
22417 (aarch64_<su>mlal_hi_n<mode>): Define.
22418 * config/aarch64/arm_neon.h (vmlal_high_n_s16): Use RTL builtin
22419 instead of inline asm.
22420 (vmlal_high_n_s32): Likewise.
22421 (vmlal_high_n_u16): Likewise.
22422 (vmlal_high_n_u32): Likewise.
22424 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
22426 * config/aarch64/aarch64-simd-builtins.def: Add RTL builtin
22428 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal_hi<mode>):
22430 (aarch64_<su>mlal_hi<mode>_insn): This.
22431 (aarch64_<su>mlal_hi<mode>): Define.
22432 * config/aarch64/arm_neon.h (vmlal_high_s8): Use RTL builtin
22433 instead of inline asm.
22434 (vmlal_high_s16): Likewise.
22435 (vmlal_high_s32): Likewise.
22436 (vmlal_high_u8): Likewise.
22437 (vmlal_high_u16): Likewise.
22438 (vmlal_high_u32): Likewise.
22440 2021-02-03 Ilya Leoshkevich <iii@linux.ibm.com>
22442 * lra-spills.c (remove_pseudos): Call lra_update_insn_recog_data()
22443 after calling alter_subreg() on a (mem).
22445 2021-02-03 Martin Liska <mliska@suse.cz>
22448 * lto-streamer-out.c (produce_lto_section): Fill up missing
22450 * lto-streamer.h (struct lto_section): Add _padding field.
22452 2021-02-03 Richard Biener <rguenther@suse.de>
22454 * lto-streamer.c (lto_get_section_name): Free temporary
22456 * tree-loop-distribution.c
22457 (loop_distribution::merge_dep_scc_partitions): Free edge data.
22459 2021-02-03 Jakub Jelinek <jakub@redhat.com>
22461 PR middle-end/97487
22462 * ifcvt.c (noce_can_force_operand): New function.
22463 (noce_emit_move_insn): Use it.
22464 (noce_try_sign_mask): Likewise. Formatting fix.
22466 2021-02-03 Jakub Jelinek <jakub@redhat.com>
22468 PR middle-end/97971
22469 * lra-constraints.c (process_alt_operands): For inline asm, don't call
22470 fatal_insn, but instead return false.
22472 2021-02-03 Jakub Jelinek <jakub@redhat.com>
22474 PR tree-optimization/98287
22475 * config/i386/mmx.md (<insn><mode>3): For shifts don't enable expander
22478 2021-02-03 Tamar Christina <tamar.christina@arm.com>
22480 PR tree-optimization/98928
22481 * tree-vect-loop.c (vect_analyze_loop_2): Change
22482 STMT_VINFO_SLP_VECT_ONLY to STMT_VINFO_SLP_VECT_ONLY_PATTERN.
22483 * tree-vect-slp-patterns.c (complex_pattern::build): Likewise.
22484 * tree-vectorizer.h (STMT_VINFO_SLP_VECT_ONLY_PATTERN): New.
22485 (class _stmt_vec_info): Add slp_vect_pattern_only_p.
22487 2021-02-02 Richard Biener <rguenther@suse.de>
22489 * gimple-loop-interchange.cc (prepare_data_references):
22491 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
22492 * tree-ssa-loop-im.c (hoist_memory_references): Likewise.
22493 * tree-vect-stmts.c (vectorizable_condition): Do not
22495 (vectorizable_comparison): Likewise.
22497 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22499 * config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
22500 * config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
22501 * config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
22502 (vrsqrteq_u32): Likewise.
22504 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22506 * config/aarch64/aarch64-simd-builtins.def (sqxtun2): Define builtin.
22507 * config/aarch64/aarch64-simd.md (aarch64_sqxtun2<mode>_le): Define.
22508 (aarch64_sqxtun2<mode>_be): Likewise.
22509 (aarch64_sqxtun2<mode>): Likewise.
22510 * config/aarch64/arm_neon.h (vqmovun_high_s16): Reimplement using builtin.
22511 (vqmovun_high_s32): Likewise.
22512 (vqmovun_high_s64): Likewise.
22513 * config/aarch64/iterators.md (UNSPEC_SQXTUN2): Define.
22515 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22517 * config/aarch64/aarch64-simd-builtins.def (bfdot_lane, bfdot_laneq): Use
22519 (bfmlalb_lane, bfmlalt_lane, bfmlalb_lane_q, bfmlalt_lane_q): Use FP flags.
22521 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22523 * config/aarch64/aarch64-simd-builtins.def (fcmla_lane0, fcmla_lane90,
22524 fcmla_lane180, fcmla_lane270, fcmlaq_lane0, fcmlaq_lane90, fcmlaq_lane180,
22525 fcmlaq_lane270, scvtf, ucvtf, fcvtzs, fcvtzu, scvtfsi, scvtfdi, ucvtfsi,
22526 ucvtfdi, fcvtzshf, fcvtzuhf, fmlal_lane_low, fmlsl_lane_low,
22527 fmlal_laneq_low, fmlsl_laneq_low, fmlalq_lane_low, fmlslq_lane_low,
22528 fmlalq_laneq_low, fmlslq_laneq_low, fmlal_lane_high, fmlsl_lane_high,
22529 fmlal_laneq_high, fmlsl_laneq_high, fmlalq_lane_high, fmlslq_lane_high,
22530 fmlalq_laneq_high, fmlslq_laneq_high): Use FP flags.
22532 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22534 * config/aarch64/aarch64-builtins.c (FLAG_LOAD): Define.
22535 * config/aarch64/aarch64-simd-builtins.def (ld1x2, ld2, ld3, ld4, ld2r,
22536 ld3r, ld4r, ld1, ld1x3, ld1x4): Use LOAD flags.
22538 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22540 * config/aarch64/aarch64-simd-builtins.def (combine, zip1, zip2,
22541 uzp1, uzp2, trn1, trn2, simd_bsl): Use AUTO_FP flags.
22543 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22545 * config/aarch64/aarch64-simd-builtins.def (clrsb, clz, ctz, popcount,
22546 vec_smult_lane_, vec_smlal_lane_, vec_smult_laneq_, vec_smlal_laneq_,
22547 vec_umult_lane_, vec_umlal_lane_, vec_umult_laneq_, vec_umlal_laneq_,
22548 ashl, sshl, ushl, srshl, urshl, sdot_lane, udot_lane, sdot_laneq,
22549 udot_laneq, usdot_lane, usdot_laneq, sudot_lane, sudot_laneq, ashr,
22550 ashr_simd, lshr, lshr_simd, srshr_n, urshr_n, ssra_n, usra_n, srsra_n,
22551 ursra_n, sshll_n, ushll_n, sshll2_n, ushll2_n, ssri_n, usri_n, ssli_n,
22552 ssli_n, usli_n, bswap, rbit, simd_bsl, eor3q, rax1q, xarq, bcaxq): Use
22553 NONE builtin flags.
22555 2021-02-02 Jakub Jelinek <jakub@redhat.com>
22557 PR tree-optimization/98848
22558 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
22559 STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
22561 2021-02-02 Kito Cheng <kito.cheng@sifive.com>
22564 * expr.c: Check mode before calling store_expr.
22566 2021-02-02 Christophe Lyon <christophe.lyon@linaro.org>
22568 * config/arm/iterators.md (supf): Remove VORNQ_S and VORNQ_U.
22570 * config/arm/mve.md (mve_vornq_s<mode>): New entry for vorn
22571 instruction using expression ior.
22572 (mve_vornq_u<mode>): New expander.
22573 (mve_vornq_f<mode>): Use ior code instead of unspec.
22574 * config/arm/unspecs.md (VORNQ_S, VORNQ_U, VORNQ_F): Remove.
22576 2021-02-02 Alexandre Oliva <oliva@adacore.com>
22578 * tree-nested.c (convert_nonlocal_reference_op): Move
22579 current_function_decl restore after re-gimplification.
22580 (convert_local_reference_op): Likewise.
22582 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22584 * config/aarch64/aarch64-simd-builtins.def (rshrn, rshrn2):
22586 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>_insn_le):
22588 (aarch64_rshrn<mode>_insn_be): Likewise.
22589 (aarch64_rshrn<mode>): Likewise.
22590 (aarch64_rshrn2<mode>_insn_le): Likewise.
22591 (aarch64_rshrn2<mode>_insn_be): Likewise.
22592 (aarch64_rshrn2<mode>): Likewise.
22593 * config/aarch64/aarch64.md (unspec): Add UNSPEC_RSHRN.
22594 * config/aarch64/arm_neon.h (vrshrn_high_n_s16): Reimplement
22596 (vrshrn_high_n_s32): Likewise.
22597 (vrshrn_high_n_s64): Likewise.
22598 (vrshrn_high_n_u16): Likewise.
22599 (vrshrn_high_n_u32): Likewise.
22600 (vrshrn_high_n_u64): Likewise.
22601 (vrshrn_n_s16): Likewise.
22602 (vrshrn_n_s32): Likewise.
22603 (vrshrn_n_s64): Likewise.
22604 (vrshrn_n_u16): Likewise.
22605 (vrshrn_n_u32): Likewise.
22606 (vrshrn_n_u64): Likewise.
22608 2021-02-01 Sergei Trofimovich <siarheit@google.com>
22610 PR tree-optimization/98499
22611 * ipa-modref.c (analyze_ssa_name_flags): treat RVO
22612 conservatively and assume all possible side-effects.
22614 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22616 * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi,
22617 vec_unpacku_hi_): Define builtins.
22618 * config/aarch64/arm_neon.h (vmovl_high_s8): Reimplement using
22620 (vmovl_high_s16): Likewise.
22621 (vmovl_high_s32): Likewise.
22622 (vmovl_high_u8): Likewise.
22623 (vmovl_high_u16): Likewise.
22624 (vmovl_high_u32): Likewise.
22626 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22628 * config/aarch64/aarch64-simd-builtins.def (sabdl, uabdl):
22630 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl<mode>): New
22632 * config/aarch64/aarch64.md (unspec): Define UNSPEC_SABDL,
22634 * config/aarch64/arm_neon.h (vabdl_s8): Reimplemet using
22636 (vabdl_s16): Likewise.
22637 (vabdl_s32): Likewise.
22638 (vabdl_u8): Likewise.
22639 (vabdl_u16): Likewise.
22640 (vabdl_u32): Likewise.
22641 * config/aarch64/iterators.md (ABDL): New int iterator.
22642 (sur): Handle UNSPEC_SABDL, UNSPEC_UABDL.
22644 2021-02-01 Martin Sebor <msebor@redhat.com>
22646 * tree.h (BLOCK_VARS): Add comment.
22647 (BLOCK_SUBBLOCKS): Same.
22648 (BLOCK_SUPERCONTEXT): Same.
22649 (BLOCK_ABSTRACT_ORIGIN): Same.
22650 (inlined_function_outer_scope_p): Same.
22652 2021-02-01 Martin Sebor <msebor@redhat.com>
22654 PR middle-end/97172
22655 * attribs.c (attr_access::free_lang_data): Define new function.
22656 * attribs.h (attr_access::free_lang_data): Declare new function.
22658 2021-02-01 Richard Biener <rguenther@suse.de>
22660 * vec.h (auto_vec::auto_vec): Add memory stat parameters
22662 * bitmap.h (auto_bitmap::auto_bitmap): Likewise.
22664 2021-02-01 Tamar Christina <tamar.christina@arm.com>
22666 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>,
22667 aarch64_<su>mlsl<mode>, aarch64_<su>mlsl_n<mode>): Flip mult operands.
22669 2021-02-01 Richard Biener <rguenther@suse.de>
22671 PR rtl-optimization/98863
22672 * config/i386/i386-features.c (convert_scalars_to_vector):
22673 Set DF_RD_PRUNE_DEAD_DEFS.
22675 2021-01-31 Eric Botcazou <ebotcazou@adacore.com>
22677 * system.h (SIZE_MAX): Define if not already defined.
22679 2021-01-30 Aaron Sawdey <acsawdey@linux.ibm.com>
22681 * config/rs6000/genfusion.pl (gen_2logical): New function to
22682 generate patterns for logical-logical fusion.
22683 * config/rs6000/fusion.md: Regenerated patterns.
22684 * config/rs6000/rs6000-cpus.def: Add
22685 OPTION_MASK_P10_FUSION_2LOGICAL.
22686 * config/rs6000/rs6000.c (rs6000_option_override_internal):
22687 Enable logical-logical fusion for p10.
22688 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2logical.
22690 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
22692 * config/rs6000/rs6000.opt: Add periods to new AIX options.
22694 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
22696 * config/rs6000/rs6000.opt (mabi=vec-extabi): New.
22697 (mabi=vec-default): New.
22698 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
22699 __EXTABI__ for AIX Vector extended ABI.
22700 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print AIX Vector
22702 (conditional_register_usage): If AIX vec_extabi enabled, vs20-vs31
22704 * doc/invoke.texi (PowerPC mabi): Add AIX vec-extabi and vec-default.
22706 2021-01-30 Jakub Jelinek <jakub@redhat.com>
22708 * config/i386/i386-features.c (remove_partial_avx_dependency): Clear
22709 DF_DEFER_INSN_RESCAN after calling df_process_deferred_rescans.
22711 2021-01-29 Vladimir N. Makarov <vmakarov@redhat.com>
22714 * lra-constraints.c (in_class_p): Don't narrow class only for REG
22717 2021-01-29 Will Schmidt <will_schmidt@vnet.ibm.com>
22719 * config/rs6000/rs6000-call.c (rs6000_expand_binup_builtin): Add
22720 clauses for CODE_FOR_vsx_xvcvuxddp_scale and
22721 CODE_FOR_vsx_xvcvsxddp_scale to the parameter checking code.
22723 2021-01-29 Andrew MacLeod <amacleod@redhat.com>
22725 PR tree-optimization/98866
22726 * gimple-range-gori.h (gori_compute:set_range_invariant): New.
22727 * gimple-range-gori.cc (gori_map::set_range_invariant): New.
22728 (gori_map::m_maybe_invariant): Rename from all_outgoing.
22729 (gori_map::gori_map): Rename all_outgoing to m_maybe_invariant.
22730 (gori_map::is_export_p): Ditto.
22731 (gori_map::calculate_gori): Ditto.
22732 (gori_compute::set_range_invariant): New.
22733 * gimple-range.cc (gimple_ranger::range_of_stmt): Set range
22734 invariant for pointers evaluating to [1, +INF].
22736 2021-01-29 Richard Biener <rguenther@suse.de>
22738 PR rtl-optimization/98863
22739 * config/i386/i386-features.c (remove_partial_avx_dependency):
22740 Do not perform DF analysis.
22741 (pass_data_remove_partial_avx_dependency): Remove
22744 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
22746 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_n
22747 builtin generator macros.
22748 * config/aarch64/aarch64-simd.md (aarch64_<su>mull_n<mode>):
22750 * config/aarch64/arm_neon.h (vmull_n_s16): Use RTL builtin
22751 instead of inline asm.
22752 (vmull_n_s32): Likewise.
22753 (vmull_n_u16): Likewise.
22754 (vmull_n_u32): Likewise.
22756 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22758 * config/aarch64/aarch64-simd-builtins.def (sabdl2, uabdl2):
22760 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
22762 (aarch64_<sur>abdl2<mode>): ... This.
22763 (<sur>sadv16qi): Adjust use of above.
22764 * config/aarch64/arm_neon.h (vabdl_high_s8): Reimplement using
22766 (vabdl_high_s16): Likewise.
22767 (vabdl_high_s32): Likewise.
22768 (vabdl_high_u8): Likewise.
22769 (vabdl_high_u16): Likewise.
22770 (vabdl_high_u32): Likewise.
22772 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22774 * config/aarch64/aarch64-simd-builtins.def (sabal2): Define
22776 (uabal2): Likewise.
22777 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal2<mode>): New
22779 * config/aarch64/aarch64.md (unspec): Add UNSPEC_SABAL2 and
22781 * config/aarch64/arm_neon.h (vabal_high_s8): Reimplement using
22783 (vabal_high_s16): Likewise.
22784 (vabal_high_s32): Likewise.
22785 (vabal_high_u8): Likewise.
22786 (vabal_high_u16): Likewise.
22787 (vabal_high_u32): Likewise.
22788 * config/aarch64/iterators.md (ABAL2): New mode iterator.
22789 (sur): Handle UNSPEC_SABAL2, UNSPEC_UABAL2.
22791 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22793 * config/aarch64/aarch64-simd-builtins.def (sabal): Define
22796 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal<mode>_4):
22798 (aarch64_<sur>abal<mode>): ... This
22799 (<sur>sadv16qi): Adust use of the above.
22800 * config/aarch64/arm_neon.h (vabal_s8): Reimplement using
22802 (vabal_s16): Likewise.
22803 (vabal_s32): Likewise.
22804 (vabal_u8): Likewise.
22805 (vabal_u16): Likewise.
22806 (vabal_u32): Likewise.
22808 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22810 * config/aarch64/aarch64-simd-builtins.def (saddlv, uaddlv):
22812 * config/aarch64/aarch64-simd.md (aarch64_<su>addlv<mode>):
22814 * config/aarch64/arm_neon.h (vaddlv_s8): Reimplement using
22816 (vaddlv_s16): Likewise.
22817 (vaddlv_u8): Likewise.
22818 (vaddlv_u16): Likewise.
22819 (vaddlvq_s8): Likewise.
22820 (vaddlvq_s16): Likewise.
22821 (vaddlvq_s32): Likewise.
22822 (vaddlvq_u8): Likewise.
22823 (vaddlvq_u16): Likewise.
22824 (vaddlvq_u32): Likewise.
22825 (vaddlv_s32): Likewise.
22826 (vaddlv_u32): Likewise.
22827 * config/aarch64/iterators.md (VDQV_L): New mode iterator.
22828 (unspec): Add UNSPEC_SADDLV, UNSPEC_UADDLV.
22829 (Vwstype): New mode attribute.
22831 (VWIDE_S): Likewise.
22832 (USADDLV): New int iterator.
22833 (su): Handle UNSPEC_SADDLV, UNSPEC_UADDLV.
22835 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
22837 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_lane[q]
22838 builtin generator macros.
22839 * config/aarch64/aarch64-simd.md (aarch64_vec_<su>mlsl_lane<Qlane>):
22841 * config/aarch64/arm_neon.h (vmlsl_lane_s16): Use RTL builtin
22842 instead of inline asm.
22843 (vmlsl_lane_s32): Likewise.
22844 (vmlsl_lane_u16): Likewise.
22845 (vmlsl_lane_u32): Likewise.
22846 (vmlsl_laneq_s16): Likewise.
22847 (vmlsl_laneq_s32): Likewise.
22848 (vmlsl_laneq_u16): Likewise.
22849 (vmlsl_laneq_u32): Likewise.
22851 2021-01-29 Richard Biener <rguenther@suse.de>
22853 * doc/invoke.texi (--param max-gcse-memory): Document unit
22855 * gcse.c (gcse_or_cprop_is_too_expensive): Adjust.
22856 * params.opt (--param max-gcse-memory): Adjust default and
22857 document unit of size.
22859 2021-01-29 Richard Biener <rguenther@suse.de>
22861 PR rtl-optimization/98863
22862 * gcse.c (gcse_or_cprop_is_too_expensive): Use unsigned
22863 HOST_WIDE_INT for the memory estimate.
22865 2021-01-29 Bin Cheng <bin.cheng@linux.alibaba.com>
22866 Richard Biener <rguenther@suse.de>
22868 PR tree-optimization/97627
22869 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
22870 Do not analyze fake edges.
22872 2021-01-29 Richard Biener <rguenther@suse.de>
22874 PR rtl-optimization/98144
22875 * df.h (df_mir_bb_info): Add con_visited member.
22876 * df-problems.c (df_mir_alloc): Initialize con_visited,
22877 do not fully populate IN and OUT.
22878 (df_mir_reset): Likewise.
22879 (df_mir_confluence_0): Set con_visited.
22880 (df_mir_confluence_n): Properly handle implicitely
22881 fully populated IN and OUT as designated by con_visited
22882 and update con_visited accordingly.
22884 2021-01-29 Jakub Jelinek <jakub@redhat.com>
22887 * config/arm/vec-common.md (mve_vshlq_<supf><mode>,
22888 vashl<mode>3, vashr<mode>3, vlshr<mode>3): Add
22889 && !TARGET_REALLY_IWMMXT to conditions.
22891 2021-01-29 Jakub Jelinek <jakub@redhat.com>
22894 * cfgbuild.c (find_bb_boundaries): Reset debug_insn when seeing
22897 2021-01-28 Marek Polacek <polacek@redhat.com>
22900 * stor-layout.c (finalize_type_size): If we reset TYPE_USER_ALIGN in
22901 the main variant, maybe reset it in its variants too.
22902 * tree.c (check_base_type): Return true only if TYPE_USER_ALIGN match.
22903 (check_aligned_type): Check if TYPE_USER_ALIGN match.
22905 2021-01-28 Christophe Lyon <christophe.lyon@linaro.org>
22908 * config/arm/arm.c (arm_rtx_costs_internal): Adjust cost of vector
22909 of constant zero for comparisons.
22911 2021-01-28 Michael Meissner <meissner@linux.ibm.com>
22913 * config/rs6000/rs6000.c (rs6000_mangle_decl_assembler_name): Add
22914 support for mapping built-in function names for long double
22915 built-in functions if long double is IEEE 128-bit.
22917 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
22919 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_n
22920 builtin generator macros.
22921 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_n<mode>):
22923 * config/aarch64/arm_neon.h (vmlsl_n_s16): Use RTL builtin
22924 instead of inline asm.
22925 (vmlsl_n_s32): Likewise.
22926 (vmlsl_n_u16): Likewise.
22927 (vmlsl_n_u32): Likewise.
22929 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
22931 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_n
22932 builtin generator macros.
22933 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>):
22935 * config/aarch64/arm_neon.h (vmlal_n_s16): Use RTL builtin
22936 instead of inline asm.
22937 (vmlal_n_s32): Likewise.
22938 (vmlal_n_u16): Likewise.
22939 (vmlal_n_u32): Likewise.
22941 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22943 * config/aarch64/aarch64-simd-builtins.def (shrn2): Define
22945 * config/aarch64/aarch64-simd.md (aarch64_shrn2<mode>_insn_le):
22947 (aarch64_shrn2<mode>_insn_be): Likewise.
22948 (aarch64_shrn2<mode>): Likewise.
22949 * config/aarch64/arm_neon.h (vshrn_high_n_s16): Reimlplement
22951 (vshrn_high_n_s32): Likewise.
22952 (vshrn_high_n_s64): Likewise.
22953 (vshrn_high_n_u16): Likewise.
22954 (vshrn_high_n_u32): Likewise.
22955 (vshrn_high_n_u64): Likewise.
22957 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22959 * config/aarch64/aarch64-simd-builtins.def (shrn): Define
22961 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le):
22963 (aarch64_shrn<mode>_insn_be): Likewise.
22964 (aarch64_shrn<mode>): Likewise.
22965 * config/aarch64/arm_neon.h (vshrn_n_s16): Reimplement using
22967 (vshrn_n_s32): Likewise.
22968 (vshrn_n_s64): Likewise.
22969 (vshrn_n_u16): Likewise.
22970 (vshrn_n_u32): Likewise.
22971 (vshrn_n_u64): Likewise.
22972 * config/aarch64/iterators.md (vn_mode): New mode attribute.
22974 2021-01-28 Richard Biener <rguenther@suse.de>
22976 PR rtl-optimization/80960
22977 * dse.c (check_mem_read_rtx): Call get_addr on the
22980 2021-01-28 Xionghu Luo <luoxhu@linux.ibm.com>
22981 David Edelsohn <dje.gcc@gmail.com>
22984 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22985 Don't generate VIEW_CONVERT_EXPR for fcode ALTIVEC_BUILTIN_VEC_INSERT
22987 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
22989 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Remove the
22990 wrapper call rs6000_expand_vector_set_var for cleanup. Call
22991 rs6000_expand_vector_set_var_p9 and rs6000_expand_vector_set_var_p8
22993 (rs6000_expand_vector_set_var): Delete.
22994 (rs6000_expand_vector_set_var_p9): Make static.
22995 (rs6000_expand_vector_set_var_p8): Make static.
22997 2021-01-28 Xing GUO <higuoxing@gmail.com>
22999 * common/config/riscv/riscv-common.c
23000 (riscv_subset_list::parsing_subset_version): Fix -march option parsing
23001 when `p` extension exists.
23003 2021-01-27 Vladimir N. Makarov <vmakarov@redhat.com>
23005 PR rtl-optimization/97684
23006 * ira.c (ira): Call ira_set_pseudo_classes before
23007 update_equiv_regs when it is necessary.
23009 2021-01-27 Jakub Jelinek <jakub@redhat.com>
23012 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Use
23013 %w0, %w1 and %2 instead of %0, %1 and %2.
23015 2021-01-27 Aaron Sawdey <acsawdey@linux.ibm.com>
23017 * config/rs6000/genfusion.pl: New script to generate
23018 define_insn_and_split patterns so combine can arrange fused
23019 instructions next to each other.
23020 * config/rs6000/fusion.md: New file, generated fused instruction
23021 patterns for combine.
23022 * config/rs6000/predicates.md (const_m1_to_1_operand): New predicate.
23023 (non_update_memory_operand): New predicate.
23024 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_P10_FUSION and
23025 OPTION_MASK_P10_FUSION_LD_CMPI to ISA_3_1_MASKS_SERVER and
23027 * config/rs6000/rs6000-protos.h (address_is_non_pfx_d_or_x): Add
23029 * config/rs6000/rs6000.c (rs6000_option_override_internal):
23030 Automatically set OPTION_MASK_P10_FUSION and
23031 OPTION_MASK_P10_FUSION_LD_CMPI if target is power10.
23032 (rs600_opt_masks): Allow -mpower10-fusion
23033 in function attributes.
23034 (address_is_non_pfx_d_or_x): New function.
23035 * config/rs6000/rs6000.h: Add MASK_P10_FUSION.
23036 * config/rs6000/rs6000.md: Include fusion.md.
23037 * config/rs6000/rs6000.opt: Add -mpower10-fusion
23038 and -mpower10-fusion-ld-cmpi.
23039 * config/rs6000/t-rs6000: Add dependencies involving fusion.md.
23041 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
23043 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal
23044 builtin generator macros.
23045 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal<mode>):
23047 (aarch64_<su>mlal<mode>): This.
23048 * config/aarch64/arm_neon.h (vmlal_s8): Use RTL builtin
23049 instead of inline asm.
23050 (vmlal_s16): Likewise.
23051 (vmlal_s32): Likewise.
23052 (vmlal_u8): Likewise.
23053 (vmlal_u16): Likewise.
23054 (vmlal_u32): Likewise.
23056 2021-01-27 Richard Biener <rguenther@suse.de>
23058 PR tree-optimization/98854
23059 * tree-vect-slp.c (vect_build_slp_tree_2): Also build
23060 PHIs from scalars when the number of CTORs matches the
23061 number of children.
23063 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
23065 * config/aarch64/aarch64-simd-builtins.def: Add mls_n builtin
23067 * config/aarch64/aarch64-simd.md (*aarch64_mls_elt_merge<mode>):
23069 (aarch64_mls_n<mode>): This.
23070 * config/aarch64/arm_neon.h (vmls_n_s16): Use RTL builtin
23072 (vmls_n_s32): Likewise.
23073 (vmls_n_u16): Likewise.
23074 (vmls_n_u32): Likewise.
23075 (vmlsq_n_s16): Likewise.
23076 (vmlsq_n_s32): Likewise.
23077 (vmlsq_n_u16): Likewise.
23078 (vmlsq_n_u32): Likewise.
23080 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
23082 * config/aarch64/aarch64-simd-builtins.def: Add mls builtin
23084 * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather
23086 (vmls_s16): Likewise.
23087 (vmls_s32): Likewise.
23088 (vmls_u8): Likewise.
23089 (vmls_u16): Likewise.
23090 (vmls_u32): Likewise.
23091 (vmlsq_s8): Likewise.
23092 (vmlsq_s16): Likewise.
23093 (vmlsq_s32): Likewise.
23094 (vmlsq_u8): Likewise.
23095 (vmlsq_u16): Likewise.
23096 (vmlsq_u32): Likewise.
23098 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
23100 * config/aarch64/aarch64-simd-builtins.def: Add mla_n builtin
23102 * config/aarch64/aarch64-simd.md (*aarch64_mla_elt_merge<mode>):
23104 (aarch64_mla_n<mode>): This.
23105 * config/aarch64/arm_neon.h (vmla_n_s16): Use RTL builtin
23107 (vmla_n_s32): Likewise.
23108 (vmla_n_u16): Likewise.
23109 (vmla_n_u32): Likewise.
23110 (vmlaq_n_s16): Likewise.
23111 (vmlaq_n_s32): Likewise.
23112 (vmlaq_n_u16): Likewise.
23113 (vmlaq_n_u32): Likewise.
23115 2021-01-27 liuhongt <hongtao.liu@intel.com>
23118 * config/i386/sse.md (sse2_gt<mode>3): Drop !TARGET_XOP in condition.
23119 (*sse2_eq<mode>3): Ditto.
23121 2021-01-27 Jakub Jelinek <jakub@redhat.com>
23123 * tree-pass.h (PROP_trees): Rename to ...
23124 (PROP_gimple): ... this.
23125 * cfgexpand.c (pass_data_expand): Replace PROP_trees with PROP_gimple.
23126 * passes.c (execute_function_dump, execute_function_todo,
23127 execute_one_ipa_transform_pass, execute_one_pass): Likewise.
23128 * varpool.c (ctor_for_folding): Likewise.
23130 2021-01-27 Jakub Jelinek <jakub@redhat.com>
23132 PR tree-optimization/97260
23133 * varpool.c: Include tree-pass.h.
23134 (ctor_for_folding): In GENERIC return DECL_INITIAL for TREE_READONLY
23135 non-TREE_SIDE_EFFECTS automatic variables.
23137 2021-01-26 Paul Fee <paul.f.fee@gmail.com>
23139 * doc/cpp.texi (__cplusplus): Document value for -std=c++23
23141 * doc/invoke.texi: Document -std=c++23 and -std=gnu++23.
23142 * dwarf2out.c (highest_c_language): Recognise C++20 and C++23.
23143 (gen_compile_unit_die): Recognise C++23.
23145 2021-01-26 Jakub Jelinek <jakub@redhat.com>
23148 * dwarf2asm.c (dw2_assemble_integer): Cast DWARF2_ADDR_SIZE to int
23151 2021-01-26 Jakub Jelinek <jakub@redhat.com>
23154 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
23155 Use UINTVAL (shft_amnt) and UINTVAL (mask) instead of INTVAL (shft_amnt)
23156 and INTVAL (mask). Add && INTVAL (mask) > 0 condition.
23158 2021-01-26 Richard Biener <rguenther@suse.de>
23160 * gimple-pretty-print.c (dump_binary_rhs): Handle
23161 VEC_WIDEN_{PLUS,MINUS}_{LO,HI}_EXPR.
23163 2021-01-26 Richard Biener <rguenther@suse.de>
23165 PR middle-end/98726
23166 * tree.h (vector_cst_int_elt): Remove.
23167 * tree.c (vector_cst_int_elt): Use poly_wide_int for computations,
23170 2021-01-26 Andrew Stubbs <ams@codesourcery.com>
23172 * config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
23173 for V64DFmode min/max reductions.
23175 2021-01-26 Jakub Jelinek <jakub@redhat.com>
23177 * dwarf2asm.c (dw2_assemble_integer): Handle size twice as large
23178 as DWARF2_ADDR_SIZE if x is not a scalar int by emitting it as
23179 two halves, one with x and the other with const0_rtx, ordered
23180 depending on endianity.
23182 2021-01-26 Alexandre Oliva <oliva@adacore.com>
23184 * gimplify.c (gimplify_decl_expr): Skip asan marking calls for
23185 temporaries not seen in binding block, and not about to be
23186 added as gimple variables.
23188 2021-01-25 Martin Sebor <msebor@redhat.com>
23191 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust warning text.
23193 2021-01-25 Martin Liska <mliska@suse.cz>
23195 * value-prof.c (get_nth_most_common_value): Use %s instead
23198 2021-01-25 Jakub Jelinek <jakub@redhat.com>
23201 * configure.ac (HAVE_AS_GDWARF_5_DEBUG_FLAG): Only define if
23202 readelf -wi is able to read the emitted .debug_info back.
23203 * configure: Regenerated.
23205 2021-01-25 Martin Liska <mliska@suse.cz>
23207 PR gcov-profile/98739
23208 * common.opt: Add missing sign symbol.
23209 * value-prof.c (get_nth_most_common_value): Restore handling
23210 of PROFILE_REPRODUCIBILITY_PARALLEL_RUNS and
23211 PROFILE_REPRODUCIBILITY_MULTITHREADED.
23213 2021-01-25 Richard Biener <rguenther@suse.de>
23215 PR middle-end/98807
23216 * tree.c (vector_element_bits): Always use precision of
23217 the element type for boolean vectors.
23219 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
23221 * config/rtems.h (STARTFILE_SPEC): Remove qnolinkcmds.
23222 (ENDFILE_SPEC): Evaluate qnolinkcmds.
23224 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
23226 * config/rtems.h (STARTFILE_SPEC): Remove nostdlib and
23227 nostartfiles handling since this is already done by
23228 LINK_COMMAND_SPEC. Evaluate qnolinkcmds.
23229 (ENDFILE_SPEC): Remove nostdlib and nostartfiles handling since this
23230 is already done by LINK_COMMAND_SPEC.
23231 (LIB_SPECS): Remove nostdlib and nodefaultlibs handling since
23232 this is already done by LINK_COMMAND_SPEC. Remove qnolinkcmds
23235 2021-01-25 Jakub Jelinek <jakub@redhat.com>
23238 * fold-const-call.c (host_size_t_cst_p): Renamed to ...
23239 (size_t_cst_p): ... this. Check and store unsigned HOST_WIDE_INT
23240 value rather than host size_t.
23241 (fold_const_call): Change type of s2 from size_t to
23242 unsigned HOST_WIDE_INT. Use size_t_cst_p instead of
23243 host_size_t_cst_p. For strncmp calls, pass MIN (s2, SIZE_MAX)
23244 instead of s2 as last argument.
23246 2021-01-25 Tamar Christina <tamar.christina@arm.com>
23248 * config/arm/iterators.md (rotsplit1, rotsplit2, conj_op, fcmac1,
23249 VCMLA_OP, VCMUL_OP): New.
23250 * config/arm/mve.md (mve_vcmlaq<mve_rot><mode>): Support vec_dup 0.
23251 * config/arm/neon.md (cmul<conj_op><mode>3): New.
23252 * config/arm/unspecs.md (UNSPEC_VCMLA_CONJ, UNSPEC_VCMLA180_CONJ,
23253 UNSPEC_VCMUL_CONJ): New.
23254 * config/arm/vec-common.md (cmul<conj_op><mode>3, arm_vcmla<rot><mode>,
23255 cml<fcmac1><conj_op><mode>4): New.
23257 2021-01-23 Jakub Jelinek <jakub@redhat.com>
23260 * config/rs6000/mmintrin.h (__m64): Add __may_alias__ attribute.
23262 2021-01-22 Jonathan Wright <jonathan.wright@arm.com>
23264 * config/aarch64/aarch64-simd-builtins.def: Add mla builtin
23266 * config/aarch64/arm_neon.h (vmla_s8): Use RTL builtin rather
23268 (vmla_s16): Likewise.
23269 (vmla_s32): Likewise.
23270 (vmla_u8): Likewise.
23271 (vmla_u16): Likewise.
23272 (vmla_u32): Likewise.
23273 (vmlaq_s8): Likewise.
23274 (vmlaq_s16): Likewise.
23275 (vmlaq_s32): Likewise.
23276 (vmlaq_u8): Likewise.
23277 (vmlaq_u16): Likewise.
23278 (vmlaq_u32): Likewise.
23280 2021-01-22 David Malcolm <dmalcolm@redhat.com>
23282 * doc/invoke.texi (GCC_EXTRA_DIAGNOSTIC_OUTPUT): Add @findex
23285 2021-01-22 Jakub Jelinek <jakub@redhat.com>
23288 * dwarf2out.c (output_file_names): For -gdwarf-5, if there are no
23289 filenames to emit, still emit the required 0 index directory and
23290 filename entries that match DW_AT_comp_dir and DW_AT_name of the
23293 2021-01-22 Marek Polacek <polacek@redhat.com>
23296 * doc/invoke.texi: Update C++ ABI Version 15 description.
23298 2021-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23300 PR tree-optimization/98766
23301 * tree-ssa-math-opts.c (convert_mult_to_fma): Use maybe_le when
23302 comparing against type size with param_avoid_fma_max_bits.
23304 2021-01-22 Richard Biener <rguenther@suse.de>
23306 PR middle-end/98793
23307 * tree.c (vector_element_bits): Key single-bit bool vector on
23308 integer mode rather than not vector mode.
23310 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
23313 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
23314 Generate ARRAY_REF(VIEW_CONVERT_EXPR) for P8 and later
23316 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): Update
23317 to call different path for P8 and P9.
23318 (rs6000_expand_vector_set_var_p9): New function.
23319 (rs6000_expand_vector_set_var_p8): New function.
23321 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
23325 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
23326 Ajdust variable index vec_insert from address dereference to
23327 ARRAY_REF(VIEW_CONVERT_EXPR) tree expression.
23328 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
23330 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): New function.
23332 2021-01-22 Martin Liska <mliska@suse.cz>
23334 PR gcov-profile/98739
23335 * profile.c (compute_value_histograms): Drop time profile for
23336 -fprofile-reproducible=multithreaded.
23338 2021-01-22 Nathan Sidwell <nathan@acm.org>
23340 * gcc.c (process_command): Don't check OPT_SPECIAL_input_file
23343 2021-01-22 Richard Biener <rguenther@suse.de>
23345 PR middle-end/98773
23346 * tree-data-ref.c (initalize_matrix_A): Revert previous
23347 change, retaining failing on HOST_WIDE_INT_MIN CHREC_RIGHT.
23349 2021-01-22 Jakub Jelinek <jakub@redhat.com>
23351 PR tree-optimization/90248
23352 * match.pd (X cmp 0.0 ? 1.0 : -1.0 -> copysign(1, +-X),
23353 X cmp 0.0 ? -1.0 : +1.0 -> copysign(1, -+X)): Remove
23355 (X * (X cmp 0.0 ? 1.0 : -1.0) -> +-abs(X),
23356 X * (X cmp 0.0 ? -1.0 : 1.0) -> +-abs(X)): New simplifications.
23358 2021-01-22 Jakub Jelinek <jakub@redhat.com>
23360 PR tree-optimization/98255
23361 * tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
23362 extend index - low_bound from sizetype's precision rather than index
23364 (get_addr_base_and_unit_offset_1): Likewise.
23365 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
23366 * gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
23368 2021-01-22 Richard Biener <rguenther@suse.de>
23370 PR tree-optimization/98786
23371 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Avoid
23372 adding new uses of abnormals. Verify we deal with a conditional
23375 2021-01-22 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23378 * optc-save-gen.awk: Add arm_fp16_format to checked_options.
23380 2021-01-22 liuhongt <hongtao.liu@intel.com>
23384 * config/i386/sse.md (VI_128_256): New mode iterator.
23385 (*avx_cmp<mode>3_1, *avx_cmp<mode>3_2, *avx_cmp<mode>3_3,
23386 *avx_cmp<mode>3_4, *avx2_eq<mode>3, *avx2_pcmp<mode>3_1,
23387 *avx2_pcmp<mode>3_2, *avx2_gt<mode>3): New
23388 define_insn_and_split to lower avx512 vector comparison to avx
23389 version when dest is vector.
23390 (*<avx512>_cmp<mode>3,*<avx512>_cmp<mode>3,*<avx512>_ucmp<mode>3):
23391 define_insn_and_split for negating the comparison result.
23392 * config/i386/predicates.md (float_vector_all_ones_operand):
23394 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Use
23395 general NOT operator without UNSPEC_MASKOP.
23397 2021-01-21 Vladimir N. Makarov <vmakarov@redhat.com>
23399 PR rtl-optimization/98777
23400 * lra-int.h (lra_pmode_pseudo): New extern.
23401 * lra.c (lra_pmode_pseudo): New global.
23403 * lra-eliminations.c (eliminate_regs_in_insn): Use it.
23405 2021-01-21 Ilya Leoshkevich <iii@linux.ibm.com>
23407 * fwprop.c (fwprop_propagation::classify_result): Allow
23408 (subreg (mem)) simplifications.
23410 2021-01-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23412 * config/aarch64/aarch64-simd.md (aarch64_sqdml<SBINQOPS:as>l<mode>):
23414 (aarch64_sqdmlal<mode>): ... This...
23415 (aarch64_sqdmlsl<mode>): ... And this.
23416 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Split into...
23417 (aarch64_sqdmlal_lane<mode>): ... This...
23418 (aarch64_sqdmlsl_lane<mode>): ... And this.
23419 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Split into...
23420 (aarch64_sqdmlsl_laneq<mode>): ... This...
23421 (aarch64_sqdmlal_laneq<mode>): ... And this.
23422 (aarch64_sqdml<SBINQOPS:as>l_n<mode>): Split into...
23423 (aarch64_sqdmlsl_n<mode>): ... This...
23424 (aarch64_sqdmlal_n<mode>): ... And this.
23425 (aarch64_sqdml<SBINQOPS:as>l2<mode>_internal): Split into...
23426 (aarch64_sqdmlal2<mode>_internal): ... This...
23427 (aarch64_sqdmlsl2<mode>_internal): ... And this.
23429 2021-01-21 Christophe Lyon <christophe.lyon@linaro.org>
23431 * config/arm/arm_mve.h (__arm_vcmpneq_s8): Fix return type.
23433 2021-01-21 Andrea Corallo <andrea.corallo@arm.com>
23436 * doc/sourcebuild.texi (arm_thumb2_no_arm_v8_1_lob): Document.
23438 2021-01-21 liuhongt <hongtao.liu@intel.com>
23440 PR rtl-optimization/98694
23441 * regcprop.c (copy_value): If SRC had been assigned a mode
23442 narrower than the copy, we can't link DEST into the chain even
23443 they have same hard_regno_nregs(i.e. HImode/SImode in i386
23446 2021-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23448 * config/aarch64/aarch64-simd.md (aarch64_get_lane<mode>):
23449 Convert to define_insn_and_split. Split into simple move when moving
23452 2021-01-20 Segher Boessenkool <segher@kernel.crashing.org>
23454 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Change assert.
23455 Adjust comment. Simplify code.
23457 2021-01-20 Jakub Jelinek <jakub@redhat.com>
23460 * dwarf2out.c (reset_indirect_string): Also reset indirect strings
23461 with DW_FORM_line_strp form.
23462 (prune_unused_types_update_strings): Don't add into debug_str_hash
23463 indirect strings with DW_FORM_line_strp form.
23464 (adjust_name_comp_dir): New function.
23465 (dwarf2out_finish): Call it on CU DIEs after resetting
23466 debug_line_str_hash.
23468 2021-01-20 Vladimir N. Makarov <vmakarov@redhat.com>
23470 PR rtl-optimization/98722
23471 * lra-eliminations.c (eliminate_regs_in_insn): Check that target
23472 has no 3-op add insn to transform insns containing two pluses.
23474 2021-01-20 Richard Biener <rguenther@suse.de>
23476 * hwint.h (add_hwi): New function.
23477 (mul_hwi): Likewise.
23478 * tree-data-ref.c (initialize_matrix_A): Properly translate
23479 tree constants and avoid HOST_WIDE_INT_MIN.
23480 (lambda_matrix_row_add): Avoid undefined integer overflow
23481 and return true on such overflow.
23482 (lambda_matrix_right_hermite): Handle overflow from
23483 lambda_matrix_row_add gracefully. Simplify previous fix.
23484 (analyze_subscript_affine_affine): Likewise.
23486 2021-01-20 Eugene Rozenfeld <erozen@microsoft.com>
23488 PR tree-optimization/96674
23489 * match.pd: New patterns: x < y || y == XXX_MIN --> x <= y - 1
23490 x >= y && y != XXX_MIN --> x > y - 1
23492 2021-01-20 Richard Sandiford <richard.sandiford@arm.com>
23494 PR tree-optimization/98535
23495 * tree-vect-slp.c (duplicate_and_interleave): Use quick_grow_cleared.
23496 If the high and low permutes are the same, remove the high permutes
23497 from the working set and only continue with the low ones.
23499 2021-01-20 Jakub Jelinek <jakub@redhat.com>
23501 PR tree-optimization/98721
23502 * builtins.c (access_ref::inform_access): Don't assume
23503 SSA_NAME_IDENTIFIER must be non-NULL. Print messages about
23504 object whenever allocfn is NULL, rather than only when DECL_P
23505 is true. Use %qE instead of %qD for that. Formatting fixes.
23507 2021-01-20 Richard Biener <rguenther@suse.de>
23509 PR tree-optimization/98758
23510 * tree-data-ref.c (int_divides_p): Use lambda_int arguments.
23511 (lambda_matrix_right_hermite): Avoid undefinedness with
23512 signed integer abs and multiplication.
23513 (analyze_subscript_affine_affine): Use lambda_int.
23515 2021-01-20 David Malcolm <dmalcolm@redhat.com>
23518 * dwarf2out.c (output_line_info): Rename static variable
23519 "generation", moving it out of the function to...
23520 (output_line_info_generation): New.
23521 (init_sections_and_labels): Likewise, renaming the variable to...
23522 (init_sections_and_labels_generation): New.
23523 (dwarf2out_c_finalize): Reset the new variables.
23525 2021-01-19 Martin Sebor <msebor@redhat.com>
23527 PR middle-end/98664
23528 * tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
23529 all functions, even if they're not declared artificial or inline.
23530 * tree.c (tree_inlined_location): Use macro expansion location
23531 only if scope traversal fails to expose one.
23533 2021-01-19 Richard Sandiford <richard.sandiford@arm.com>
23535 PR rtl-optimization/92294
23536 * alias.c (compare_base_symbol_refs): Take an extra parameter
23537 and add the distance between two symbols to it. Enshrine in
23538 comments that -1 means "either 0 or 1, but we can't tell
23539 which at compile time".
23540 (memrefs_conflict_p): Update call accordingly.
23541 (rtx_equal_for_memref_p): Likewise. Take the distance between symbols
23544 2021-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23546 * config/aarch64/aarch64-simd-builtins.def (sqshl, uqshl,
23547 sqrshl, uqrshl, sqadd, uqadd, sqsub, uqsub, suqadd, usqadd, sqmovn,
23548 uqmovn, sqxtn2, uqxtn2, sqabs, sqneg, sqdmlal, sqdmlsl, sqdmlal_lane,
23549 sqdmlsl_lane, sqdmlal_laneq, sqdmlsl_laneq, sqdmlal_n, sqdmlsl_n,
23550 sqdmlal2, sqdmlsl2, sqdmlal2_lane, sqdmlsl2_lane, sqdmlal2_laneq,
23551 sqdmlsl2_laneq, sqdmlal2_n, sqdmlsl2_n, sqdmull, sqdmull_lane,
23552 sqdmull_laneq, sqdmull_n, sqdmull2, sqdmull2_lane, sqdmull2_laneq,
23553 sqdmull2_n, sqdmulh, sqrdmulh, sqdmulh_lane, sqdmulh_laneq,
23554 sqrdmulh_lane, sqrdmulh_laneq, sqshrun_n, sqrshrun_n, sqshrn_n,
23555 uqshrn_n, sqrshrn_n, uqrshrn_n, sqshlu_n, sqshl_n, uqshl_n, sqrdmlah,
23556 sqrdmlsh, sqrdmlah_lane, sqrdmlsh_lane, sqrdmlah_laneq, sqrdmlsh_laneq,
23557 sqmovun): Use NONE flags.
23559 2021-01-19 Richard Biener <rguenther@suse.de>
23562 * ipa-modref.c (analyze_stmt): Only record a summary for a
23565 2021-01-19 Richard Biener <rguenther@suse.de>
23567 PR middle-end/98638
23568 * tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
23570 2021-01-19 Daniel Hellstrom <daniel@gaisler.com>
23572 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
23573 built-in define __FIX_LEON3FT_TN0018.
23575 2021-01-19 Richard Biener <rguenther@suse.de>
23578 * tree-inline.c (tree_function_versioning): Set input_location
23579 to UNKNOWN_LOCATION throughout the function.
23581 2021-01-19 Tobias Burnus <tobias@codesourcery.com>
23584 * omp-low.c (lower_omp_target): Handle nonpointer is_device_ptr.
23586 2021-01-19 Martin Jambor <mjambor@suse.cz>
23589 * ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
23590 whether non-call exceptions allow removal of a statement.
23591 (isra_analyze_call): Pass the appropriate function to
23592 ssa_name_only_returned_p.
23594 2021-01-19 Geng Qi <gengqi@linux.alibaba.com>
23596 * config/riscv/arch-canonicalize (longext_sort): New function for
23597 sorting 'multi-letter'.
23598 * config/riscv/multilib-generator: Adjusting the loop of 'alt' in
23599 'alts'. The 'arch' may not be the first of 'alts'.
23600 (_expand_combination): Add underline for the 'ext' without '*'.
23601 This is because, a single-letter extension can always be treated well
23602 with a '_' prefix, but it cannot be separated out if it is appended
23605 2021-01-18 Vladimir N. Makarov <vmakarov@redhat.com>
23608 * ira.c (ira): Skip abnormal critical edge splitting.
23610 2021-01-18 Jakub Jelinek <jakub@redhat.com>
23612 PR tree-optimization/98727
23613 * tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
23614 second .MUL_OVERFLOW operand for signed multiplication with overflow
23615 checking if the second operand of multiplication is not constant.
23617 2021-01-18 David Edelsohn <dje.gcc@gmail.com>
23619 * doc/invoke.texi (-gdwarf): TPF defaults to version 2 and AIX
23620 defaults to version 4.
23622 2021-01-18 David Malcolm <dmalcolm@redhat.com>
23624 * attribs.h (fndecl_dealloc_argno): New decl.
23625 * builtins.c (call_dealloc_argno): Split out second half of
23627 (fndecl_dealloc_argno): New.
23628 * doc/extend.texi (Common Function Attributes): Document the
23629 interaction between the analyzer and the malloc attribute.
23630 * doc/invoke.texi (Static Analyzer Options): Likewise.
23632 2021-01-17 David Edelsohn <dje.gcc@gmail.com>
23634 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Override
23635 dwarf_version to 4.
23636 * config/rs6000/aix72.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
23638 2021-01-17 Martin Jambor <mjambor@suse.cz>
23641 * cgraph.c (clone_of_p): Check also former_clone_of as we climb
23644 2021-01-17 Mark Wielaard <mark@klomp.org>
23646 * common.opt (gdwarf-): Init(5).
23647 * doc/invoke.texi (-gdwarf): Document default to 5.
23649 2021-01-16 Kwok Cheung Yeung <kcy@codesourcery.com>
23651 * builtin-types.def
23652 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
23654 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
23655 ...this. Add extra argument.
23656 * gimplify.c (omp_default_clause): Ensure that event handle is
23657 firstprivate in a task region.
23658 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_DETACH.
23659 (gimplify_adjust_omp_clauses): Likewise.
23660 * omp-builtins.def (BUILT_IN_GOMP_TASK): Change function type to
23661 BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR.
23662 * omp-expand.c (expand_task_call): Add GOMP_TASK_FLAG_DETACH to flags
23663 if detach clause specified. Add detach argument when generating
23665 * omp-low.c (scan_sharing_clauses): Setup data environment for detach
23667 (finish_taskreg_scan): Move field for variable containing the event
23668 handle to the front of the struct.
23669 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_DETACH. Fix
23671 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
23672 OMP_CLAUSE_DETACH clause.
23673 (convert_local_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
23674 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_DETACH.
23675 * tree.c (omp_clause_num_ops): Add entry for OMP_CLAUSE_DETACH.
23677 (omp_clause_code_name): Add entry for OMP_CLAUSE_DETACH. Fix
23679 (walk_tree_1): Handle OMP_CLAUSE_DETACH.
23681 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
23683 * config/nios2/t-rtems: Reset all MULTILIB_* variables. Shorten
23684 multilib directory names. Use MULTILIB_REQUIRED instead of
23685 MULTILIB_EXCEPTIONS. Add -mhw-mul -mhw-mulx -mhw-div
23686 -mcustom-fpu-cfg=fph2 multilib.
23688 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
23690 * config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
23691 (nios2_init_fpu_configs): Provide register values for new
23692 -mcustom-fpu-cfg=fph2 option variant.
23693 * doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
23696 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
23698 * config/nios2/nios2.c (nios2_custom_check_insns): Remove
23699 custom instruction warnings.
23701 2021-01-16 Jakub Jelinek <jakub@redhat.com>
23703 PR tree-optimization/96669
23704 * match.pd ((CST << x) & 1 -> x == 0): New simplification.
23706 2021-01-16 Jakub Jelinek <jakub@redhat.com>
23708 PR tree-optimization/96271
23709 * passes.def: Pass false argument to first two pass_cd_dce
23710 instances and true to last instance. Add comment that
23711 last instance rewrites no longer addressed locals.
23712 * tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
23714 (pass_cd_dce::set_pass_param): New method.
23715 (pass_cd_dce::execute): Return TODO_update_address_taken from
23716 last cd_dce instance.
23718 2021-01-15 Carl Love <cel@us.ibm.com>
23720 * config/rs6000/altivec.h (vec_mulh, vec_div, vec_dive, vec_mod):
23722 * config/rs6000/altivec.md (VIlong): Move define to file vsx.md.
23723 * config/rs6000/rs6000-builtin.def (DIVES_V4SI, DIVES_V2DI,
23724 DIVEU_V4SI, DIVEU_V2DI, DIVS_V4SI, DIVS_V2DI, DIVU_V4SI,
23725 DIVU_V2DI, MODS_V2DI, MODS_V4SI, MODU_V2DI, MODU_V4SI,
23726 MULHS_V2DI, MULHS_V4SI, MULHU_V2DI, MULHU_V4SI, MULLD_V2DI):
23727 Add builtin define.
23728 (MULH, DIVE, MOD): Add new BU_P10_OVERLOAD_2 definitions.
23729 * config/rs6000/rs6000-call.c (VSX_BUILTIN_VEC_DIV,
23730 VSX_BUILTIN_VEC_DIVE, P10_BUILTIN_VEC_MOD, P10_BUILTIN_VEC_MULH):
23731 New overloaded definitions.
23732 (builtin_function_type) [P10V_BUILTIN_DIVEU_V4SI,
23733 P10V_BUILTIN_DIVEU_V2DI, P10V_BUILTIN_DIVU_V4SI,
23734 P10V_BUILTIN_DIVU_V2DI, P10V_BUILTIN_MODU_V2DI,
23735 P10V_BUILTIN_MODU_V4SI, P10V_BUILTIN_MULHU_V2DI,
23736 P10V_BUILTIN_MULHU_V4SI]: Add case
23737 statement for builtins.
23738 * config/rs6000/rs6000.md (bits): Add new attribute sizes V4SI, V2DI.
23739 * config/rs6000/vsx.md (VIlong): Moved from config/rs6000/altivec.md.
23740 (UNSPEC_VDIVES, UNSPEC_VDIVEU): New unspec definitions.
23741 (vsx_mul_v2di): Add if TARGET_POWER10 statement.
23742 (vsx_udiv_v2di): Add if TARGET_POWER10 statement.
23743 (dives_<mode>, diveu_<mode>, div<mode>3, uvdiv<mode>3,
23744 mods_<mode>, modu_<mode>, mulhs_<mode>, mulhu_<mode>, mulv2di3):
23745 Add define_insn, mode is VIlong.
23746 * doc/extend.texi (vec_mulh, vec_mul, vec_div, vec_dive, vec_mod):
23747 Add builtin descriptions.
23749 2021-01-15 Eric Botcazou <ebotcazou@adacore.com>
23751 * final.c (final_start_function_1): Reset force_source_line.
23753 2021-01-15 Jakub Jelinek <jakub@redhat.com>
23755 PR tree-optimization/96669
23756 * match.pd (((1 << A) & 1) != 0 -> A == 0,
23757 ((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
23758 possibly different power of two constants and to right shift too.
23760 2021-01-15 Jakub Jelinek <jakub@redhat.com>
23762 PR tree-optimization/96681
23763 * match.pd ((x < 0) ^ (y < 0) to (x ^ y) < 0): New simplification.
23764 ((x >= 0) ^ (y >= 0) to (x ^ y) < 0): Likewise.
23765 ((x < 0) ^ (y >= 0) to (x ^ y) >= 0): Likewise.
23766 ((x >= 0) ^ (y < 0) to (x ^ y) >= 0): Likewise.
23768 2021-01-15 Alexandre Oliva <oliva@adacore.com>
23770 * opts.c (gen_command_line_string): Exclude -dumpbase-ext.
23772 2021-01-15 Tamar Christina <tamar.christina@arm.com>
23774 * config/aarch64/aarch64-simd.md (cml<fcmac1><conj_op><mode>4,
23775 cmul<conj_op><mode>3): New.
23776 * config/aarch64/iterators.md (UNSPEC_FCMUL,
23777 UNSPEC_FCMUL180, UNSPEC_FCMLA_CONJ, UNSPEC_FCMLA180_CONJ,
23778 UNSPEC_CMLA_CONJ, UNSPEC_CMLA180_CONJ, UNSPEC_CMUL, UNSPEC_CMUL180,
23779 FCMLA_OP, FCMUL_OP, conj_op, rotsplit1, rotsplit2, fcmac1, sve_rot1,
23780 sve_rot2, SVE2_INT_CMLA_OP, SVE2_INT_CMUL_OP, SVE2_INT_CADD_OP): New.
23781 (rot): Add UNSPEC_FCMUL, UNSPEC_FCMUL180.
23782 (rot_op): Renamed to conj_op.
23783 * config/aarch64/aarch64-sve.md (cml<fcmac1><conj_op><mode>4,
23784 cmul<conj_op><mode>3): New.
23785 * config/aarch64/aarch64-sve2.md (cml<fcmac1><conj_op><mode>4,
23786 cmul<conj_op><mode>3): New.
23788 2021-01-15 David Malcolm <dmalcolm@redhat.com>
23792 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
23793 Escape the tempfile name when constructing the expected output.
23795 2021-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23797 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlsl_hi<mode>):
23799 (aarch64_<su>mlsl_hi<mode>): ... This.
23800 (aarch64_<su>mlsl_hi<mode>): Define.
23801 (*aarch64_<su>mlsl<mode): Rename to...
23802 (aarch64_<su>mlsl<mode): ... This.
23803 * config/aarch64/aarch64-simd-builtins.def (smlsl, umlsl,
23804 smlsl_hi, umlsl_hi): Define builtins.
23805 * config/aarch64/arm_neon.h (vmlsl_high_s8, vmlsl_high_s16,
23806 vmlsl_high_s32, vmlsl_high_u8, vmlsl_high_u16, vmlsl_high_u32,
23807 vmlsl_s8, vmlsl_s16, vmlsl_s32, vmlsl_u8,
23808 vmlsl_u16, vmlsl_u32): Reimplement with builtins.
23810 2021-01-15 Uroš Bizjak <ubizjak@gmail.com>
23812 * config/i386/i386-c.c (ix86_target_macros):
23813 Use cpp_define_formatted for __SIZEOF_FLOAT80__ definition.
23815 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
23818 * config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
23819 * Makefile.in (RTL_SSA_H): New variable.
23820 * config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
23821 * config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
23822 * config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
23824 * config/aarch64/aarch64-cc-fusion.cc: New file.
23826 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
23828 * recog.h (insn_change_watermark::~insn_change_watermark): Avoid
23829 calling cancel_changes for changes that no longer exist.
23831 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
23833 * rtl-ssa/functions.h (function_info::ref_defs): Rename to...
23834 (function_info::reg_defs): ...this.
23835 * rtl-ssa/member-fns.inl (function_info::ref_defs): Rename to...
23836 (function_info::reg_defs): ...this.
23838 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23841 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
23843 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23846 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23849 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
23851 2021-01-15 Richard Biener <rguenther@suse.de>
23853 PR tree-optimization/96376
23854 * tree-vect-stmts.c (get_load_store_type): Disregard alignment
23855 for VMAT_INVARIANT.
23857 2021-01-15 Martin Liska <mliska@suse.cz>
23859 * doc/install.texi: Document that some tests need pytest module.
23860 * doc/sourcebuild.texi: Likewise.
23862 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23865 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
23867 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23869 * config/arm/mve.md (mve_vshrq_n_s<mode>_imm): New entry.
23870 (mve_vshrq_n_u<mode>_imm): Likewise.
23871 * config/arm/neon.md (vashr<mode>3, vlshr<mode>3): Move to ...
23872 * config/arm/vec-common.md: ... here.
23874 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
23876 * config/arm/mve.md (mve_vshlq_<supf><mode>): Move to
23878 * config/arm/neon.md (vashl<mode>3): Delete.
23879 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): New.
23880 (vasl<mode>3): New expander.
23882 2021-01-15 Richard Biener <rguenther@suse.de>
23884 PR tree-optimization/98685
23885 * tree-vect-slp.c (vect_schedule_slp_node): Refactor handling
23886 of vector extern defs.
23888 2021-01-14 David Malcolm <dmalcolm@redhat.com>
23891 * diagnostic.c (diagnostic_kind_text): Break out this array
23893 (diagnostic_build_prefix): ...here.
23894 (fancy_abort): Detect when diagnostic_initialize has not yet been
23895 called and fall back to a minimal implementation of printing the
23896 ICE, rather than segfaulting in internal_error.
23898 2021-01-14 David Malcolm <dmalcolm@redhat.com>
23900 * diagnostic.c (diagnostic_initialize): Eliminate
23901 parseable_fixits_p in favor of initializing extra_output_kind from
23902 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
23903 (convert_column_unit): New function, split out from...
23904 (diagnostic_converted_column): ...this.
23905 (print_parseable_fixits): Add "column_unit" and "tabstop" params.
23906 Use them to call convert_column_unit on the column values.
23907 (diagnostic_report_diagnostic): Eliminate conditional on
23908 parseable_fixits_p in favor of a switch statement on
23909 extra_output_kind, passing the appropriate values to the new
23910 params of print_parseable_fixits.
23911 (selftest::test_print_parseable_fixits_none): Update for new
23912 params of print_parseable_fixits.
23913 (selftest::test_print_parseable_fixits_insert): Likewise.
23914 (selftest::test_print_parseable_fixits_remove): Likewise.
23915 (selftest::test_print_parseable_fixits_replace): Likewise.
23916 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
23918 (selftest::diagnostic_c_tests): Call it.
23919 * diagnostic.h (enum diagnostics_extra_output_kind): New.
23920 (diagnostic_context::parseable_fixits_p): Delete field in favor
23922 (diagnostic_context::extra_output_kind): ...this new field.
23923 * doc/invoke.texi (Environment Variables): Add
23924 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
23925 * opts.c (common_handle_option): Update handling of
23926 OPT_fdiagnostics_parseable_fixits for change to diagnostic_context
23929 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23931 * tree-vect-slp-patterns.c (class complex_operations_pattern,
23932 complex_operations_pattern::matches,
23933 complex_operations_pattern::recognize,
23934 complex_operations_pattern::build): New.
23935 (slp_patterns): Use it.
23937 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23939 * internal-fn.def (COMPLEX_FMS, COMPLEX_FMS_CONJ): New.
23940 * optabs.def (cmls_optab, cmls_conj_optab): New.
23941 * doc/md.texi: Document them.
23942 * tree-vect-slp-patterns.c (class complex_fms_pattern,
23943 complex_fms_pattern::matches, complex_fms_pattern::recognize,
23944 complex_fms_pattern::build): New.
23946 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23948 * internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
23949 * optabs.def (cmla_optab, cmla_conj_optab): New.
23950 * doc/md.texi: Document them.
23951 * tree-vect-slp-patterns.c (vect_match_call_p,
23952 class complex_fma_pattern, vect_slp_reset_pattern,
23953 complex_fma_pattern::matches, complex_fma_pattern::recognize,
23954 complex_fma_pattern::build): New.
23956 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23958 * internal-fn.def (COMPLEX_MUL, COMPLEX_MUL_CONJ): New.
23959 * optabs.def (cmul_optab, cmul_conj_optab): New.
23960 * doc/md.texi: Document them.
23961 * tree-vect-slp-patterns.c (vect_match_call_complex_mla,
23962 vect_normalize_conj_loc, is_eq_or_top, vect_validate_multiplication,
23963 vect_build_combine_node, class complex_mul_pattern,
23964 complex_mul_pattern::matches, complex_mul_pattern::recognize,
23965 complex_mul_pattern::build): New.
23967 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23969 * tree-vect-slp.c (optimize_load_redistribution_1): New.
23970 (optimize_load_redistribution, vect_is_slp_load_node): New.
23971 (vect_match_slp_patterns): Use it.
23973 2021-01-14 Tamar Christina <tamar.christina@arm.com>
23975 * tree-vect-slp-patterns.c (complex_add_pattern::build):
23978 2021-01-14 Thomas Schwinge <thomas@codesourcery.com>
23980 * config/gcn/mkoffload.c (main): Create an offload image only in
23981 64-bit configurations.
23983 2021-01-14 H.J. Lu <hjl.tools@gmail.com>
23986 * config/i386/i386-options.c (ix86_option_override_internal):
23987 Issue an error for -fcf-protection with CF_BRANCH when compiling
23988 for 32-bit non-TARGET_CMOV targets.
23990 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
23993 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p):
23994 Remove declaration and initialization of shadow variable "ret".
23995 (ix86_option_override_internal): Remove delcaration of
23996 shadow variable "i". Redeclare shadowed variable to unsigned.
23997 * common/config/i386/i386-common.c (pta_size): Redeclare to unsigned.
23998 * config/i386/i386-builtins.c (get_builtin_code_for_version):
23999 Update for redeclaration.
24000 * config/i386/i386.h (pta_size): Ditto.
24002 2021-01-14 Richard Biener <rguenther@suse.de>
24004 PR tree-optimization/98674
24005 * tree-data-ref.c (base_supports_access_fn_components_p): New.
24006 (initialize_data_dependence_relation): For two bases without
24007 possible access fns resort to type size equality when determining
24008 shape compatibility.
24010 2021-01-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24013 * config/arm/arm_neon.h: Replace calls to __builtin_vcge* by
24014 <=, >= operators in vcle and vcge intrinsics respectively.
24015 * config/arm/arm_neon_builtins.def: Remove entry for
24018 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
24021 * config/i386/i386-options.c (ix86_function_specific_save):
24022 Remove redundant assignment to opts->x_ix86_branch_cost.
24023 * config/i386/i386.c (ix86_prefetch_sse):
24024 Rename from x86_prefetch_sse. Update all uses.
24025 * config/i386/i386.h: Update for rename.
24026 * config/i386/i386-options.h: Ditto.
24028 2021-01-14 Jakub Jelinek <jakub@redhat.com>
24031 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3,
24032 *sse4_1_zero_extendv4hiv4si2_3, *sse4_1_zero_extendv2siv2di2_3):
24033 Use Bm instead of m for non-avx. Add isa attribute.
24035 2021-01-14 Jakub Jelinek <jakub@redhat.com>
24037 PR tree-optimization/96688
24038 * match.pd (~(X >> Y) -> ~X >> Y): New simplification if
24039 ~X can be simplified.
24041 2021-01-14 Richard Sandiford <richard.sandiford@arm.com>
24043 * tree-vect-stmts.c (vect_model_load_cost): Account for unused
24044 IFN_LOAD_LANES results.
24046 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24048 * config/aarch64/aarch64-simd.md (aarch64_<su>xtl<mode>):
24050 (aarch64_xtn<mode>): Likewise.
24051 * config/aarch64/aarch64-simd-builtins.def (sxtl, uxtl, xtn):
24054 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
24056 (vmovl_s16): Likewise.
24057 (vmovl_s32): Likewise.
24058 (vmovl_u8): Likewise.
24059 (vmovl_u16): Likewise.
24060 (vmovl_u32): Likewise.
24061 (vmovn_s16): Likewise.
24062 (vmovn_s32): Likewise.
24063 (vmovn_s64): Likewise.
24064 (vmovn_u16): Likewise.
24065 (vmovn_u32): Likewise.
24066 (vmovn_u64): Likewise.
24068 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24070 * config/aarch64/aarch64-simd.md (aarch64_<su>qxtn2<mode>_le):
24072 (aarch64_<su>qxtn2<mode>_be): Likewise.
24073 (aarch64_<su>qxtn2<mode>): Likewise.
24074 * config/aarch64/aarch64-simd-builtins.def (sqxtn2, uqxtn2):
24076 * config/aarch64/iterators.md (SAT_TRUNC): Define code_iterator.
24077 (su): Handle ss_truncate and us_truncate.
24078 * config/aarch64/arm_neon.h (vqmovn_high_s16): Reimplement using
24080 (vqmovn_high_s32): Likewise.
24081 (vqmovn_high_s64): Likewise.
24082 (vqmovn_high_u16): Likewise.
24083 (vqmovn_high_u32): Likewise.
24084 (vqmovn_high_u64): Likewise.
24086 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24088 * config/aarch64/aarch64-simd.md (aarch64_xtn2<mode>_le):
24090 (aarch64_xtn2<mode>_be): Likewise.
24091 (aarch64_xtn2<mode>): Likewise.
24092 * config/aarch64/aarch64-simd-builtins.def (xtn2): Define
24094 * config/aarch64/arm_neon.h (vmovn_high_s16): Reimplement using
24096 (vmovn_high_s32): Likewise.
24097 (vmovn_high_s64): Likewise.
24098 (vmovn_high_u16): Likewise.
24099 (vmovn_high_u32): Likewise.
24100 (vmovn_high_u64): Likewise.
24102 2021-01-13 Stafford Horne <shorne@gmail.com>
24104 * config/or1k/or1k.h (ASM_PREFERRED_EH_DATA_FORMAT): New macro.
24106 2021-01-13 Stafford Horne <shorne@gmail.com>
24108 * config/or1k/linux.h (TARGET_ASM_FILE_END): Define macro.
24110 2021-01-13 Stafford Horne <shorne@gmail.com>
24112 * config/or1k/or1k.h (TARGET_CPU_CPP_BUILTINS): Add builtin
24113 define for __or1k_hard_float__.
24115 2021-01-13 Stafford Horne <shorne@gmail.com>
24117 * config/or1k/or1k.h (NO_PROFILE_COUNTERS): Define as 1.
24118 (PROFILE_HOOK): Define to call _mcount.
24119 (FUNCTION_PROFILER): Change from abort to no-op.
24121 2021-01-13 Jakub Jelinek <jakub@redhat.com>
24123 PR tree-optimization/96691
24124 * match.pd ((~X | C) ^ D -> (X | C) ^ (~D ^ C),
24125 (~X & C) ^ D -> (X & C) ^ (D ^ C)): New simplifications if
24126 (~D ^ C) or (D ^ C) can be simplified.
24128 2021-01-13 Richard Biener <rguenther@suse.de>
24130 PR tree-optimization/92645
24131 * match.pd (BIT_FIELD_REF to conversion): Delay canonicalization
24132 until after vector lowering.
24134 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
24136 * config/aarch64/aarch64-sve.md (fnma<mode>4): Extend from SVE_FULL_I
24138 (@aarch64_pred_fnma<mode>, cond_fnma<mode>, *cond_fnma<mode>_2)
24139 (*cond_fnma<mode>_4, *cond_fnma<mode>_any): Likewise.
24141 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
24143 * config/aarch64/aarch64-sve.md (fma<mode>4): Extend from SVE_FULL_I
24145 (@aarch64_pred_fma<mode>, cond_fma<mode>, *cond_fma<mode>_2)
24146 (*cond_fma<mode>_4, *cond_fma<mode>_any): Likewise.
24148 2021-01-13 Richard Biener <rguenther@suse.de>
24150 PR tree-optimization/92645
24151 * tree-vect-slp.c (vect_build_slp_tree_1): Relax supported
24152 BIT_FIELD_REF argument.
24153 (vect_build_slp_tree_2): Record the desired vector type
24154 on the external vector def.
24155 (vectorizable_slp_permutation): Handle required punning
24156 of existing vector defs.
24158 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
24160 * rtl-ssa/accesses.h (def_lookup): Fix order of comparison results.
24162 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
24164 * config/sh/sh.md (movsf_ie): Remove operands[2] test.
24166 2021-01-13 Samuel Thibault <samuel.thibault@ens-lyon.org>
24168 * config.gcc [$target == *-*-gnu*]: Enable
24169 'default_gnu_indirect_function'.
24171 2021-01-13 Jakub Jelinek <jakub@redhat.com>
24174 * optabs.c (expand_vec_perm_const): Don't force v0 and v1 into
24175 registers before calling targetm.vectorize.vec_perm_const, only after
24177 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const): Handle
24178 two argument permutation when one operand is zero vector and only
24179 after that force operands into registers.
24180 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_1): New
24181 define_insn_and_split pattern.
24182 (*avx512bw_zero_extendv32qiv32hi2_1): Likewise.
24183 (*avx512f_zero_extendv16hiv16si2_1): Likewise.
24184 (*avx2_zero_extendv8hiv8si2_1): Likewise.
24185 (*avx512f_zero_extendv8siv8di2_1): Likewise.
24186 (*avx2_zero_extendv4siv4di2_1): Likewise.
24187 * config/mips/mips.c (mips_vectorize_vec_perm_const): Force operands
24189 * config/arm/arm.c (arm_vectorize_vec_perm_const): Likewise.
24190 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Likewise.
24191 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const): Likewise.
24192 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const): Likewise.
24193 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const): Likewise.
24194 * config/gcn/gcn.c (gcn_vectorize_vec_perm_const): Likewise. Use std::swap.
24196 2021-01-13 Martin Liska <mliska@suse.cz>
24198 PR tree-optimization/98455
24199 * gimple-if-to-switch.cc (condition_info::record_phi_mapping):
24200 Record also virtual PHIs.
24201 (pass_if_to_switch::execute): Return TODO_cleanup_cfg only
24204 2021-01-13 Jonathan Wakely <jwakely@redhat.com>
24206 * doc/invoke.texi (C++ Modules): Fix typos.
24208 2021-01-13 Richard Biener <rguenther@suse.de>
24210 PR tree-optimization/98640
24211 * tree-ssa-sccvn.c (visit_nary_op): Do not try to
24212 handle plus or minus from a truncated operand to be
24215 2021-01-13 Jakub Jelinek <jakub@redhat.com>
24218 * config/i386/i386.md (*btr<mode>_1, *btr<mode>_2): New
24219 define_insn_and_split patterns.
24220 (splitter after *btr<mode>_2): New splitter.
24222 2021-01-13 Martin Liska <mliska@suse.cz>
24225 * cgraphunit.c (analyze_functions): Remove dead code.
24227 2021-01-13 Qian Jianhua <qianjh@cn.fujitsu.com>
24229 * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
24230 * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
24231 (a64fx_regmove_cost, a64fx_vector_cost): New.
24232 (a64fx_tunings): Use the new added cost tables.
24234 2021-01-13 Jakub Jelinek <jakub@redhat.com>
24237 * config/i386/predicates.md (pmovzx_parallel): New predicate.
24238 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): New
24239 define_insn_and_split pattern.
24240 (*sse4_1_zero_extendv4hiv4si2_3): Likewise.
24241 (*sse4_1_zero_extendv2siv2di2_3): Likewise.
24243 2021-01-13 Julian Brown <julian@codesourcery.com>
24245 * config/gcn/gcn.c (gcn_conditional_register_usage): Remove dead code
24246 to fix v0 register.
24248 2021-01-13 Julian Brown <julian@codesourcery.com>
24250 * config/gcn/gcn.c (gcn_md_reorg): Fix case where EXEC reg is live
24253 2021-01-13 Julian Brown <julian@codesourcery.com>
24255 * config/gcn/gcn-valu.md (recip<mode>2<exec>, recip<mode>2): Use unspec
24256 for reciprocal-approximation instructions.
24257 (div<mode>3): Use fused multiply-accumulate operations for reciprocal
24258 refinement and division result.
24259 * config/gcn/gcn.md (UNSPEC_RCP): New unspec constant.
24261 2021-01-13 Julian Brown <julian@codesourcery.com>
24263 * config/gcn/gcn-valu.md (subdf): Rename to...
24266 2021-01-12 Martin Liska <mliska@suse.cz>
24268 * gcov.c (source_info::debug): Fix printf format for 32-bit hosts.
24270 2021-01-12 Andrea Corallo <andrea.corallo@arm.com>
24272 * function-abi.h: Fix typo.
24274 2021-01-12 Christophe Lyon <christophe.lyon@linaro.org>
24278 * config/arm/arm.h (ARM_HAVE_NEON_V8QI_LDST): New macro.
24279 (ARM_HAVE_NEON_V16QI_LDST, ARM_HAVE_NEON_V4HI_LDST): Likewise.
24280 (ARM_HAVE_NEON_V8HI_LDST, ARM_HAVE_NEON_V2SI_LDST): Likewise.
24281 (ARM_HAVE_NEON_V4SI_LDST, ARM_HAVE_NEON_V4HF_LDST): Likewise.
24282 (ARM_HAVE_NEON_V8HF_LDST, ARM_HAVE_NEON_V4BF_LDST): Likewise.
24283 (ARM_HAVE_NEON_V8BF_LDST, ARM_HAVE_NEON_V2SF_LDST): Likewise.
24284 (ARM_HAVE_NEON_V4SF_LDST, ARM_HAVE_NEON_DI_LDST): Likewise.
24285 (ARM_HAVE_NEON_V2DI_LDST): Likewise.
24286 (ARM_HAVE_V8QI_LDST, ARM_HAVE_V16QI_LDST): Likewise.
24287 (ARM_HAVE_V4HI_LDST, ARM_HAVE_V8HI_LDST): Likewise.
24288 (ARM_HAVE_V2SI_LDST, ARM_HAVE_V4SI_LDST, ARM_HAVE_V4HF_LDST): Likewise.
24289 (ARM_HAVE_V8HF_LDST, ARM_HAVE_V4BF_LDST, ARM_HAVE_V8BF_LDST): Likewise.
24290 (ARM_HAVE_V2SF_LDST, ARM_HAVE_V4SF_LDST, ARM_HAVE_DI_LDST): Likewise.
24291 (ARM_HAVE_V2DI_LDST): Likewise.
24292 * config/arm/mve.md (*movmisalign<mode>_mve_store): New pattern.
24293 (*movmisalign<mode>_mve_load): New pattern.
24294 * config/arm/neon.md (movmisalign<mode>): Move to ...
24295 * config/arm/vec-common.md: ... here.
24297 2021-01-12 Vladimir N. Makarov <vmakarov@redhat.com>
24300 * lra-eliminations.c (eliminate_regs_in_insn): Add transformation
24301 of pattern 'plus (plus (hard reg, const), pseudo)'.
24303 2021-01-12 Richard Biener <rguenther@suse.de>
24305 PR tree-optimization/98550
24306 * tree-vect-slp.c (vect_record_max_nunits): Check whether
24307 the group size is a multiple of the vector element count.
24308 (vect_build_slp_tree_1): When we need to fail because
24309 the vector type choosen causes unrolling do so lazily
24310 without affecting matches only at the end to guide group splitting.
24312 2021-01-12 Martin Liska <mliska@suse.cz>
24315 * optc-save-gen.awk: Compare also n_target_save vars with
24318 2021-01-12 Martin Liska <mliska@suse.cz>
24320 * gcov.c (source_info::debug): New.
24321 (print_usage): Add --debug (-D) option.
24322 (process_args): Likewise.
24323 (generate_results): Call src->debug after
24324 accumulate_line_counts.
24325 (read_graph_file): Properly assign id for EXIT_BLOCK.
24326 * profile.c (branch_prob): Dump function body before it is
24329 2021-01-12 Jakub Jelinek <jakub@redhat.com>
24331 PR tree-optimization/98629
24332 * tree-ssa-math-opts.c (arith_overflow_check_p): Don't update use_stmt
24333 unless returning non-zero.
24335 2021-01-12 Jakub Jelinek <jakub@redhat.com>
24337 PR tree-optimization/95731
24338 * tree-ssa-reassoc.c (optimize_range_tests_cmp_bitwise): Also optimize
24339 x < 0 && y < 0 && z < 0 into (x | y | z) < 0 for signed x, y, z.
24340 (optimize_range_tests): Call optimize_range_tests_cmp_bitwise
24341 only after optimize_range_tests_var_bound.
24343 2021-01-12 Jakub Jelinek <jakub@redhat.com>
24345 * configure.ac: Ensure c/Make-lang.in comes first in @all_lang_makefrags@.
24346 * configure: Regenerated.
24348 2021-01-12 liuhongt <hongtao.liu@intel.com>
24351 * config/i386/i386-builtins.h (BUILTIN_DESC_SWAP_OPERANDS):
24353 * config/i386/i386-expand.c (ix86_expand_sse_comi): Delete
24356 2021-01-12 Alexandre Oliva <oliva@adacore.com>
24358 * ssa-iterators.h (end_imm_use_stmt_traverse): Forward
24360 (auto_end_imm_use_stmt_traverse): New struct.
24361 (FOR_EACH_IMM_USE_STMT): Use it.
24362 (BREAK_FROM_IMM_USE_STMT, RETURN_FROM_IMM_USE_STMT): Remove,
24364 * gimple-ssa-strength-reduction.c: ... here, ...
24365 * graphite-scop-detection.c: ... here, ...
24366 * ipa-modref.c, ipa-pure-const.c, ipa-sra.c: ... here, ...
24367 * tree-predcom.c, tree-ssa-ccp.c: ... here, ...
24368 * tree-ssa-dce.c, tree-ssa-dse.c: ... here, ...
24369 * tree-ssa-loop-ivopts.c, tree-ssa-math-opts.c: ... here, ...
24370 * tree-ssa-phiprop.c, tree-ssa.c: ... here, ...
24371 * tree-vect-slp.c: ... and here, ...
24372 * doc/tree-ssa.texi: ... and the example here.
24374 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24376 * config/aarch64/aarch64-sve.md (sdiv_pow2<mode>3): Extend from
24377 SVE_FULL_I to SVE_I. Generate an UNSPEC_PRED_X.
24378 (*sdiv_pow2<mode>3): New pattern.
24379 (@cond_<sve_int_op><mode>): Extend from SVE_FULL_I to SVE_I.
24380 Wrap the ASRD in an UNSPEC_PRED_X.
24381 (*cond_<sve_int_op><mode>_2): Likewise. Replace the UNSPEC_PRED_X
24382 predicate with a constant PTRUE, if it isn't already.
24383 (*cond_<sve_int_op><mode>_z): Replace with...
24384 (*cond_<sve_int_op><mode>_any): ...this new pattern.
24386 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24388 * config/aarch64/aarch64-sve.md (*cond_bic<mode>_2): Extend from
24389 SVE_FULL_I to SVE_I.
24390 (*cond_bic<mode>_any): Likewise.
24392 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24394 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart)
24395 (@aarch64_pred_<MUL_HIGHPART:optab><mode>): Extend from SVE_FULL_I
24398 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24400 * config/aarch64/aarch64-sve.md (<su>abd<mode>_3): Extend from
24401 SVE_FULL_I to SVE_I.
24402 (*aarch64_cond_<su>abd<mode>_2): Likewise.
24403 (*aarch64_cond_<su>abd<mode>_any): Likewise.
24404 (@aarch64_pred_<su>abd<mode>): Likewise. Use UNSPEC_PRED_X
24405 for the max and min but not for the minus.
24406 (*aarch64_cond_<su>abd<mode>_3): New pattern.
24408 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24410 * config/aarch64/iterators.md (SVE_24I): New iterator.
24411 * config/aarch64/aarch64-sve.md (*aarch64_adr<mode>_shift): Extend from
24412 SVE_FULL_SDI to SVE_24I. Use containers rather than elements.
24414 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24416 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_BINARY:optab><mode>)
24417 (*cond_<SVE_INT_BINARY:optab><mode>_2): Extend from SVE_FULL_I
24419 (*cond_<SVE_INT_BINARY:optab><mode>_3): Likewise.
24420 (*cond_<SVE_INT_BINARY:optab><mode>_any): Likewise.
24421 (*cond_<SVE_INT_BINARY:optab><mode>_2_const): Likewise.
24422 (*cond_<SVE_INT_BINARY:optab><mode>_any_const): Likewise.
24424 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24426 * config/aarch64/aarch64-sve.md (<SVE_INT_BINARY_IMM:optab><mode>3)
24427 (@aarch64_pred_<SVE_INT_BINARY_IMM:optab><mode>)
24428 (*post_ra_<SVE_INT_BINARY_IMM:optab><mode>3): Extend from SVE_FULL_I
24431 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
24433 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3)
24434 (v<ASHIFT:optab><mode>3, @aarch64_pred_<optab><mode>)
24435 (*post_ra_v<ASHIFT:optab><mode>3): Extend from SVE_FULL_I to SVE_I.
24437 2021-01-11 Martin Liska <mliska@suse.cz>
24440 * symtab-clones.h (clone_info::release): Release
24441 symtab::m_clones with ggc_delete as it's a GGC memory.
24443 2021-01-11 Matthias Klose <doko@ubuntu.com>
24445 * Makefile.in (LINK_PROGRESS): Show the link target.
24447 2021-01-11 Richard Biener <rguenther@suse.de>
24449 PR tree-optimization/91403
24450 * tree-vect-data-refs.c (vect_analyze_group_access_1): Cap
24451 single-element interleaving group size at 4096 elements.
24453 2021-01-11 Richard Biener <rguenther@suse.de>
24455 PR tree-optimization/98526
24456 * tree-vect-loop.c (vect_model_reduction_cost): Remove costing
24457 of the actual reduction op for the regular case.
24458 (vectorizable_reduction): Cost the stmts
24459 vect_transform_reduction produces here.
24461 2021-01-11 Andreas Krebbel <krebbel@linux.ibm.com>
24463 * tree-ssa-forwprop.c (simplify_vector_constructor): For
24464 big-endian, use UNPACK[_FLOAT]_HI.
24466 2021-01-11 Tamar Christina <tamar.christina@arm.com>
24468 * tree-vect-slp-patterns.c (class complex_pattern,
24469 class complex_add_pattern): Add parameters to matches.
24470 (complex_add_pattern::build): Free memory.
24471 (complex_add_pattern::matches): Move validation end of match.
24472 (complex_add_pattern::recognize): Likewise.
24474 2021-01-11 Tamar Christina <tamar.christina@arm.com>
24476 * tree-vect-slp-patterns.c (linear_loads_p): Fix externals.
24478 2021-01-11 Tamar Christina <tamar.christina@arm.com>
24480 * tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
24482 2021-01-11 Jakub Jelinek <jakub@redhat.com>
24484 PR tree-optimization/95867
24485 * tree-ssa-math-opts.h: New header.
24486 * tree-ssa-math-opts.c: Include tree-ssa-math-opts.h.
24487 (powi_as_mults): No longer static. Use build_one_cst instead of
24488 build_real. Formatting fix.
24489 * tree-ssa-reassoc.c: Include tree-ssa-math-opts.h.
24490 (attempt_builtin_powi): Handle multiplication reassociation without
24491 powi_fndecl using powi_as_mults.
24492 (reassociate_bb): For integral types don't require
24493 -funsafe-math-optimizations to call attempt_builtin_powi.
24495 2021-01-11 Jakub Jelinek <jakub@redhat.com>
24497 PR tree-optimization/95852
24498 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): Change
24499 mul_stmts parameter type to vec<gimple *> &. Before cond_stmt
24500 allow in the bb any of the stmts in that vector, div_stmt and
24501 up to 3 cast stmts.
24502 (arith_cast_equal_p): New function.
24503 (arith_overflow_check_p): Add cast_stmt argument, handle signed
24504 multiply overflow checks.
24505 (match_arith_overflow): Adjust caller. Handle signed multiply
24508 2021-01-11 Jakub Jelinek <jakub@redhat.com>
24510 PR tree-optimization/95852
24511 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): New function.
24512 (uaddsub_overflow_check_p): Renamed to ...
24513 (arith_overflow_check_p): ... this. Handle also multiplication
24514 with overflow check.
24515 (match_uaddsub_overflow): Renamed to ...
24516 (match_arith_overflow): ... this. Add cfg_changed argument. Handle
24517 also multiplication with overflow check. Adjust function comment.
24518 (math_opts_dom_walker::after_dom_children): Adjust callers. Call
24519 match_arith_overflow also for MULT_EXPR.
24521 2021-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24523 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
24524 __builtin_convertvector.
24525 (vmovl_s16): Likewise.
24526 (vmovl_s32): Likewise.
24527 (vmovl_u8): Likewise.
24528 (vmovl_u16): Likewise.
24529 (vmovl_u32): Likewise.
24530 (vmovn_s16): Likewise.
24531 (vmovn_s32): Likewise.
24532 (vmovn_s64): Likewise.
24533 (vmovn_u16): Likewise.
24534 (vmovn_u32): Likewise.
24535 (vmovn_u64): Likewise.
24537 2021-01-11 Martin Liska <mliska@suse.cz>
24539 * gimple-if-to-switch.cc (struct condition_info): Use auto_var.
24540 (if_chain::is_beneficial): Delete clusters
24541 (find_conditions): Make second argument of conditions_in_bbs a
24542 pointer so that we control over it's lifetime.
24543 (pass_if_to_switch::execute): Delete them.
24545 2021-01-11 Kewen Lin <linkw@linux.ibm.com>
24547 * ira.c (move_unallocated_pseudos): Check other_reg and skip if
24550 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24552 * config/vax/vax.md (cc): Remove mode attribute.
24553 (subst_<cc>, subst_f<cc>): Rename to...
24554 (subst_<mode>, subst_f<VAXccnz:mode>): ... these respectively.
24555 (*cbranch<VAXint:mode>4_<VAXcc:mode>): Update for `cc' removal.
24556 (*cbranch<VAXfp:mode>4_<VAXccnz:mode>): Likewise.
24557 (*branch_<mode>, *branch_<mode>_reversed): Likewise.
24559 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24561 * config/vax/vax.md (subst_f<cc>): Add mode to operands and
24562 `const_double_zero'.
24564 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24566 * config/pdp11/pdp11.md (PDPfp): New mode iterator.
24567 (fcc_cc, fcc_ccnz): Use it. Add mode to `const_double_zero' and
24570 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
24572 * genemit.c (gen_exp) <CONST_DOUBLE>: Handle `const_double_zero'
24574 * read-rtl.c (rtx_reader::read_rtx_code): Handle machine mode
24575 with `const_double_zero'.
24576 * doc/rtl.texi (Constant Expression Types): Document it.
24578 2021-01-09 Jakub Jelinek <jakub@redhat.com>
24581 * tree-cfg.c (verify_gimple_assign_binary): Allow lhs of
24582 POINTER_DIFF_EXPR to be any integral type.
24584 2021-01-09 Jakub Jelinek <jakub@redhat.com>
24586 PR rtl-optimization/98603
24587 * function.c (instantiate_virtual_regs_in_insn): For asm goto
24588 with impossible constraints, drop all SETs, CLOBBERs, drop PARALLEL
24589 if any, set ASM_OPERANDS mode to VOIDmode and change
24590 ASM_OPERANDS_OUTPUT_CONSTRAINT and ASM_OPERANDS_OUTPUT_IDX.
24592 2021-01-09 Alexandre Oliva <oliva@gnu.org>
24595 * final.c (notice_source_line): Narrow down the condition to
24596 skip a line-0 marker.
24598 2021-01-08 Sergei Trofimovich <siarheit@google.com>
24600 * ipa-modref.c (merge_call_side_effects): Fix
24601 linebreak split by reordering two print calls.
24603 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
24605 * config/s390/vector.md (*tf_to_fprx2_0): Rename from
24606 "*mov_tf_to_fprx2_0" for consistency, fix constraint.
24607 (*tf_to_fprx2_1): Rename from "*mov_tf_to_fprx2_1" for
24608 consistency, fix constraint.
24610 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
24612 * config/s390/s390-c.c (s390_def_or_undef_macro): Accept
24613 callables instead of mask values.
24614 (struct target_flag_set_p): New predicate.
24615 (s390_cpu_cpp_builtins_internal): Define or undefine
24616 __LONG_DOUBLE_VX__ macro.
24618 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
24621 * config/i386/i386.c (x86_function_profiler): Use R10 and R11
24622 to call mcount in large model with PIC for NO_PROFILE_COUNTERS
24625 2021-01-08 Richard Biener <rguenther@suse.de>
24627 * tree-ssa-sccvn.c (pass_fre::execute): Reset the SCEV hash table.
24629 2021-01-08 Richard Biener <rguenther@suse.de>
24631 * tree-vect-slp.c (scalar_stmts_to_slp_tree_map_t): Fix.
24632 (vect_build_slp_tree): On cache hit release the matched
24633 scalar stmts vector.
24634 * tree-vect-stmts.c (vectorizable_store): Properly free
24635 vec_oprnds before possibly gathering them again.
24637 2021-01-08 Richard Biener <rguenther@suse.de>
24639 PR tree-optimization/98544
24640 * tree-vect-slp.c (vect_optimize_slp): Always materialize
24641 permutes at a permute node.
24643 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
24646 * config/i386/i386.c (x86_function_profiler): Use R10 to call
24647 mcount in large model. Sorry for large model with PIC.
24649 2021-01-08 Jakub Jelinek <jakub@redhat.com>
24652 * config/i386/i386.opt (ix86_cmodel, ix86_incoming_stack_boundary_arg,
24653 ix86_pmode, ix86_preferred_stack_boundary_arg, ix86_regparm,
24654 ix86_veclibabi_type): Remove x_ prefix, use TargetVariable instead of
24655 TargetSave and initialize for variables with enum types.
24656 (mfentry, mstack-protector-guard-reg=, mstack-protector-guard-offset=,
24657 mstack-protector-guard-symbol=): Add Save.
24658 * config/i386/i386-options.c (ix86_function_specific_save,
24659 ix86_function_specific_restore): Don't save or restore x_ix86_cmodel,
24660 x_ix86_incoming_stack_boundary_arg, x_ix86_pmode,
24661 x_ix86_preferred_stack_boundary_arg, x_ix86_regparm,
24662 x_ix86_veclibabi_type.
24664 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
24666 * config/aarch64/aarch64-sve.md (*cnot<mode>): Extend from
24667 SVE_FULL_I to SVE_I.
24668 (*cond_cnot<mode>_2, *cond_cnot<mode>_any): Likewise.
24670 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
24672 * config/aarch64/aarch64-sve.md (*cond_uxt<mode>_2): Extend from
24673 SVE_FULL_I to SVE_I.
24674 (*cond_uxt<mode>_any): Likewise.
24676 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24678 * config/aarch64/iterators.md (Vwhalf): New iterator.
24679 * config/aarch64/aarch64-simd.md (aarch64_<sur>adalp<mode>_3):
24681 (aarch64_<sur>adalp<mode>): ... This. Make more
24683 (<sur>sadv16qi): Adjust callsite of the above.
24684 * config/aarch64/aarch64-simd-builtins.def (sadalp, uadalp): New
24686 * config/aarch64/arm_neon.h (vpadal_s8): Reimplement using
24688 (vpadal_s16): Likewise.
24689 (vpadal_u8): Likewise.
24690 (vpadal_u16): Likewise.
24691 (vpadalq_s8): Likewise.
24692 (vpadalq_s16): Likewise.
24693 (vpadalq_s32): Likewise.
24694 (vpadalq_u8): Likewise.
24695 (vpadalq_u16): Likewise.
24696 (vpadalq_u32): Likewise.
24698 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24700 * config/aarch64/aarch64-simd.md (aarch64_<su>abd<mode>_3):
24702 (aarch64_<su>abd<mode>): ... This.
24703 (<sur>sadv16qi): Adjust callsite of the above.
24704 * config/aarch64/aarch64-simd-builtins.def (sabd, uabd): Define
24706 * config/aarch64/arm_neon.h (vabd_s8): Reimplement using
24708 (vabd_s16): Likewise.
24709 (vabd_s32): Likewise.
24710 (vabd_u8): Likewise.
24711 (vabd_u16): Likewise.
24712 (vabd_u32): Likewise.
24713 (vabdq_s8): Likewise.
24714 (vabdq_s16): Likewise.
24715 (vabdq_s32): Likewise.
24716 (vabdq_u8): Likewise.
24717 (vabdq_u16): Likewise.
24718 (vabdq_u32): Likewise.
24720 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24722 * config/aarch64/aarch64-simd-builtins.def (saba, uaba): Define
24724 * config/aarch64/arm_neon.h (vaba_s8): Implement using builtin.
24725 (vaba_s16): Likewise.
24726 (vaba_s32): Likewise.
24727 (vaba_u8): Likewise.
24728 (vaba_u16): Likewise.
24729 (vaba_u32): Likewise.
24730 (vabaq_s8): Likewise.
24731 (vabaq_s16): Likewise.
24732 (vabaq_s32): Likewise.
24733 (vabaq_u8): Likewise.
24734 (vabaq_u16): Likewise.
24735 (vabaq_u32): Likewise.
24737 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24739 * config/aarch64/aarch64-simd.md (aba<mode>_3): Rename to...
24740 (aarch64_<su>aba<mode>): ... This. Handle uaba as well.
24741 Change RTL pattern to match.
24743 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
24745 * common/config/riscv/riscv-common.c (riscv_current_subset_list): New.
24746 * config/riscv/riscv-c.c (riscv-subset.h): New.
24747 (INCLUDE_STRING): Define.
24748 (riscv_cpu_cpp_builtins): Add new style architecture extension
24750 * config/riscv/riscv-subset.h (riscv_subset_list::begin): New.
24751 (riscv_subset_list::end): New.
24752 (riscv_current_subset_list): New.
24754 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
24756 * common/config/riscv/riscv-common.c (RISCV_DONT_CARE_VERSION):
24757 Move to riscv-subset.h.
24758 (struct riscv_subset_t): Ditto.
24759 (class riscv_subset_list): Ditto.
24760 * config/riscv/riscv-subset.h (RISCV_DONT_CARE_VERSION): Move
24761 from riscv-common.c.
24762 (struct riscv_subset_t): Ditto.
24763 (class riscv_subset_list): Ditto.
24764 * config/riscv/t-riscv ($(common_out_file)): Add file
24767 2021-01-07 Jakub Jelinek <jakub@redhat.com>
24770 * config/i386/i386.md (*bmi_blsi_<mode>_cmp, *bmi_blsi_<mode>_ccno):
24771 New define_insn patterns.
24773 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
24775 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_UNARY:optab><mode>)
24776 (*cond_<SVE_INT_UNARY:optab><mode>_2): Extend from SVE_FULL_I to SVE_I.
24777 (*cond_<SVE_INT_UNARY:optab><mode>_any): Likewise.
24779 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
24781 PR tree-optimization/98560
24782 * internal-fn.def (IFN_VCONDU, IFN_VCONDEQ): Use type vec_cond.
24783 * internal-fn.c (vec_cond_mask_direct): Get the data mode from
24785 (vec_cond_direct): Likewise argument 2.
24786 (vec_condu_direct, vec_condeq_direct): Delete.
24787 (expand_vect_cond_optab_fn): Rename to...
24788 (expand_vec_cond_optab_fn): ...this, replacing old macro.
24789 (expand_vec_condu_optab_fn, expand_vec_condeq_optab_fn): Delete.
24790 (expand_vect_cond_mask_optab_fn): Rename to...
24791 (expand_vec_cond_mask_optab_fn): ...this, replacing old macro.
24792 (direct_vec_cond_mask_optab_supported_p): Treat the optab as a
24794 (direct_vec_cond_optab_supported_p): Likewise.
24795 (direct_vec_condu_optab_supported_p): Delete.
24796 (direct_vec_condeq_optab_supported_p): Delete.
24797 * gimple-isel.cc: Include internal-fn.h.
24798 (gimple_expand_vec_cond_expr): Check that IFN_VCONDEQ is supported
24801 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
24803 PR tree-optimization/98560
24804 * gimple-isel.cc (gimple_expand_vec_cond_expr): If we fail to use
24805 IFN_VCOND{,U,EQ}, fall back on IFN_VCOND_MASK.
24807 2021-01-07 Uroš Bizjak <ubizjak@gmail.com>
24809 * config/i386/i386.md (insn): Merge from plusminus_insn, shift_insn,
24810 rotate_insn and optab code attributes.
24811 Update all uses to merged code attribute.
24812 * config/i386/sse.md: Update all uses to merged code attribute.
24813 * config/i386/mmx.md: Update all uses to merged code attribute.
24815 2021-01-07 Jakub Jelinek <jakub@redhat.com>
24817 PR tree-optimization/98568
24818 * gimple-ssa-store-merging.c (bswap_view_convert): New function.
24819 (bswap_replace): Use it.
24821 2021-01-06 Vladimir N. Makarov <vmakarov@redhat.com>
24823 PR rtl-optimization/97978
24824 * lra-int.h (lra_hard_reg_split_p): New external.
24825 * lra.c (lra_hard_reg_split_p): New global.
24826 (lra): Set up lra_hard_reg_split_p after splitting a hard reg.
24827 * lra-assigns.c (lra_assign): Don't check allocation correctness
24828 after hard reg splitting.
24830 2021-01-06 Martin Sebor <msebor@redhat.com>
24833 * builtins.c (new_delete_mismatch_p): New overload.
24834 (new_delete_mismatch_p (tree, tree)): Call it.
24836 2021-01-06 Alexandre Oliva <oliva@adacore.com>
24838 * Makefile.in (T_GLIMITS_H): New.
24839 (stmp-int-hdrs): Depend on it, use it.
24840 * config/t-vxworks (T_GLIMITS_H): Override it.
24841 (vxw-glimits.h): New.
24843 2021-01-06 Richard Biener <rguenther@suse.de>
24845 PR tree-optimization/98513
24846 * value-range.cc (intersect_ranges): Compare the upper bounds
24847 for the expected relation.
24849 2021-01-06 Gerald Pfeifer <gerald@pfeifer.com>
24852 2020-12-28 Gerald Pfeifer <gerald@pfeifer.com>
24854 * doc/standards.texi (HSAIL): Remove section.
24856 2021-01-05 Samuel Thibault <samuel.thibault@ens-lyon.org>
24858 * configure: Re-generate.
24860 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24862 * doc/invoke.texi (-std=c++20): Adjust for the publication of
24863 ISO 14882:2020 standard.
24864 * doc/standards.texi: Likewise.
24866 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24868 PR tree-optimization/94802
24869 * expr.h (maybe_optimize_sub_cmp_0): Declare.
24870 * expr.c: Include tree-pretty-print.h and flags.h.
24871 (maybe_optimize_sub_cmp_0): New function.
24872 (do_store_flag): Use it.
24873 * cfgexpand.c (expand_gimple_cond): Likewise.
24875 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24877 * mux-utils.h (pointer_mux::m_ptr): Tweak description of contents.
24878 * rtlanal.c (simple_regno_set): Tweak description to clarify the
24881 2021-01-05 Richard Biener <rguenther@suse.de>
24883 PR tree-optimization/98516
24884 * tree-vect-slp.c (vect_optimize_slp): Permute the incoming
24885 lanes when materializing on a VEC_PERM node.
24886 (vectorizable_slp_permutation): Dump the permute properly.
24888 2021-01-05 Richard Biener <rguenther@suse.de>
24890 * tree-vect-slp.c (vect_slp_region): Move debug counter
24891 to cover individual subgraphs.
24893 2021-01-05 Richard Biener <rguenther@suse.de>
24895 PR tree-optimization/98428
24896 * tree-vect-slp.c (vect_build_slp_tree_1): Properly reject
24897 vector lane extracts for loop vectorization.
24899 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24901 PR tree-optimization/98514
24902 * tree-ssa-reassoc.c (bb_rank): Change type from long * to
24904 (operand_rank): Change type from hash_map<tree, long> to
24905 hash_map<tree, int64_t>.
24906 (phi_rank): Change return type from long to int64_t.
24907 (loop_carried_phi): Change block_rank variable type from long to
24909 (propagate_rank): Change return type, rank parameter type and
24910 op_rank variable type from long to int64_t.
24911 (find_operand_rank): Change return type from long to int64_t
24912 and change slot variable type from long * to int64_t *.
24913 (insert_operand_rank): Change rank parameter type from long to
24915 (get_rank): Change return type and rank variable type from long to
24916 int64_t. Use PRId64 instead of ld to print the rank.
24917 (init_reassoc): Change rank variable type from long to int64_t
24918 and adjust correspondingly bb_rank and operand_rank initialization.
24920 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24922 PR tree-optimization/96928
24923 * tree-ssa-phiopt.c (xor_replacement): New function.
24924 (tree_ssa_phiopt_worker): Call it.
24926 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24928 PR tree-optimization/96930
24929 * match.pd ((A / (1 << B)) -> (A >> B)): If A is extended
24930 from narrower value which has the same type as 1 << B, perform
24931 the right shift on the narrower value followed by extension.
24933 2021-01-05 Jakub Jelinek <jakub@redhat.com>
24935 PR tree-optimization/96239
24936 * gimple-ssa-store-merging.c (maybe_optimize_vector_constructor): New
24938 (get_status_for_store_merging): Don't return BB_INVALID for blocks
24939 with potential bswap optimizable CONSTRUCTORs.
24940 (pass_store_merging::execute): Optimize vector CONSTRUCTORs with bswap
24943 2021-01-05 Richard Biener <rguenther@suse.de>
24945 PR tree-optimization/98381
24946 * tree.c (vector_element_bits): Properly compute bool vector
24948 * tree-vect-loop.c (vectorizable_live_operation): Properly
24949 compute the last lane bit offset.
24951 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
24954 * config/i386/sse.md (sse_cvtps2pi): Redefine as define_insn_and_split.
24955 Clear the top 64 bytes of the input XMM register.
24956 (sse_cvttps2pi): Ditto.
24958 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
24961 * config/i386/xopintrin.h (_mm256_cmov_si256): New.
24963 2021-01-05 H.J. Lu <hjl.tools@gmail.com>
24966 * config/i386/xmmintrin.h (_mm_extract_pi16): Cast to unsigned
24969 2021-01-05 Claudiu Zissulescu <claziss@synopsys.com>
24971 * config/arc/arc.md (maddsidi4_split): Use ACC_REG_FIRST.
24972 (umaddsidi4_split): Likewise.
24974 2021-01-05 liuhongt <hongtao.liu@intel.com>
24977 * config/i386/sse.md (*sse2_pmovskb_zexthisi): New
24978 define_insn_and_split for zero_extend of subreg HI of pmovskb
24980 (*sse2_pmovskb_zexthisi): Add new combine splitters for
24981 zero_extend of not of subreg HI of pmovskb result.
24983 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24986 * explow.c (convert_memory_address_addr_space_1): Handle UNSPECs
24988 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate): Use
24989 convert_memory_address to convert symbolic immediates to ptr_mode
24990 before forcing them to memory.
24992 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
24994 PR rtl-optimization/97144
24995 * recog.c (constrain_operands): Initialize matching_operand
24996 for each alternative, rather than only doing it once.
24998 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
25000 PR rtl-optimization/98403
25001 * rtl-ssa/changes.cc (function_info::finalize_new_accesses): Explain
25002 why we don't remove call clobbers.
25003 (function_info::apply_changes_to_insn): Don't attempt to add
25004 call clobbers here.
25006 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
25008 PR tree-optimization/98371
25009 * tree-vect-loop.c (vect_reanalyze_as_main_loop): New function.
25010 (vect_analyze_loop): If an epilogue loop appears to be cheaper
25011 than the main loop, re-analyze it as a main loop before adopting
25014 2021-01-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
25017 * configure.ac (NETLIBS): Determine using AX_LIB_SOCKET_NSL.
25018 * aclocal.m4, configure: Regenerate.
25019 * Makefile.in (NETLIBS): Define.
25020 (BACKEND): Remove $(CODYLIB).
25022 2021-01-05 Jakub Jelinek <jakub@redhat.com>
25024 PR rtl-optimization/98334
25025 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
25026 Optimize (X - 1) * Y + Y to X * Y or (X + 1) * Y - Y to X * Y.
25028 2021-01-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
25030 * tree-inline.c (expand_call_inline): Restore input_location.
25031 Return result from recursive call.
25033 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
25035 PR tree-optimization/95401
25036 * config/aarch64/aarch64-sve-builtins.cc
25037 (gimple_folder::load_store_cookie): Use bits rather than bytes
25038 for the alignment argument to IFN_MASK_LOAD and IFN_MASK_STORE.
25039 * gimple-fold.c (gimple_fold_mask_load_store_mem_ref): Likewise.
25040 * tree-vect-stmts.c (vectorizable_store): Likewise.
25041 (vectorizable_load): Likewise.
25043 2021-01-04 Richard Biener <rguenther@suse.de>
25045 PR tree-optimization/98308
25046 * tree-vect-stmts.c (vectorizable_load): Set invariant mask
25049 2021-01-04 Jakub Jelinek <jakub@redhat.com>
25051 PR tree-optimization/95771
25052 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Handle types
25053 with precision smaller than int's precision and types with precision
25054 twice as large as long long. Formatting fixes.
25056 2021-01-04 Richard Biener <rguenther@suse.de>
25058 PR tree-optimization/98464
25059 * tree-ssa-sccvn.c (vn_valueize_for_srt): Rename from ...
25060 (vn_valueize_wrapper): ... this. Temporarily adjust vn_context_bb.
25061 (process_bb): Adjust.
25063 2021-01-04 Matthew Malcomson <matthew.malcomson@arm.com>
25066 * doc/invoke.texi (-fsanitize=address): Fix wording describing
25067 clash with -fsanitize=hwaddress.
25069 2021-01-04 Richard Biener <rguenther@suse.de>
25071 PR tree-optimization/98282
25072 * tree-ssa-sccvn.c (vn_get_stmt_kind): Classify tcc_reference on
25073 invariants as VN_NARY.
25075 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
25078 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Accept
25079 aarch64_simd_reg_or_zero for operand 2. Use the combinez patterns
25080 to handle zero operands.
25082 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
25084 * config/aarch64/aarch64.c (offset_6bit_signed_scaled_p): New function.
25085 (offset_6bit_unsigned_scaled_p): Fix typo in comment.
25086 (aarch64_sve_prefetch_operand_p): Accept MUL VLs in the range
25089 2021-01-04 Richard Biener <rguenther@suse.de>
25091 PR tree-optimization/98393
25092 * tree-vect-slp.c (vect_build_slp_tree): Properly zero matches
25093 when hitting the limit.
25095 2021-01-04 Richard Biener <rguenther@suse.de>
25097 PR tree-optimization/98291
25098 * tree-vect-loop.c (vectorizable_reduction): Bypass
25099 associativity check for SLP reductions with VF 1.
25101 2021-01-04 Jakub Jelinek <jakub@redhat.com>
25103 PR tree-optimization/96782
25104 * match.pd (x == ~x -> false, x != ~x -> true): New simplifications.
25106 2021-01-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
25108 * collect-utils.c (collect_execute): Check dumppfx.
25109 * collect2.c (maybe_run_lto_and_relink, do_link): Pass atsuffix
25110 to collect_execute.
25111 (do_link): Add new parameter atsuffix.
25112 (main): Handle -dumpdir option. Skip one argument for
25113 -o, -isystem and -B options.
25114 * gcc.c (make_at_file): New helper function.
25115 (close_at_file): Use it.
25117 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25119 * config/darwin.h (MIN_LD64_NO_COAL_SECTS): Adjust.
25120 Amend handling for LD64_VERSION fallback defaults.
25122 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25124 * config.gcc: Compute default version information
25125 from the configured target. Likewise defaults for
25127 * config/darwin10.h: Removed.
25128 * config/darwin12.h: Removed.
25129 * config/darwin9.h: Removed.
25130 * config/rs6000/darwin8.h: Removed.
25132 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25134 * config/darwin9.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
25136 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25138 * config/darwin9.h (STACK_CHECK_STATIC_BUILTIN): Move from here..
25139 * config/darwin.h (STACK_CHECK_STATIC_BUILTIN): .. to here.
25141 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25143 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move from
25145 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): ... to here.
25147 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25149 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move the spec
25150 for the Darwin10 unwinder stub from here ...
25151 * config/darwin.h (LINK_COMMAND_SPEC_A): ... to here.
25153 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
25155 * config/darwin.h (DSYMUTIL_SPEC): Default to DWARF
25156 (ASM_DEBUG_SPEC):Only define if the assembler supports
25158 (PREFERRED_DEBUGGING_TYPE): Default to DWARF.
25159 (DARWIN_PREFER_DWARF): Define.
25160 * config/darwin9.h (PREFERRED_DEBUGGING_TYPE): Remove.
25161 (DARWIN_PREFER_DWARF): Likewise
25162 (DSYMUTIL_SPEC): Likewise.
25163 (COLLECT_RUN_DSYMUTIL): Likewise.
25164 (ASM_DEBUG_SPEC): Likewise.
25165 (ASM_DEBUG_OPTION_SPEC): Likewise.
25167 2021-01-02 Jan Hubicka <jh@suse.cz>
25169 * cfg.c (free_block): ggc_free bb.
25171 2021-01-01 Jakub Jelinek <jakub@redhat.com>
25173 * gcc.c (process_command): Update copyright notice dates.
25174 * gcov-dump.c (print_version): Ditto.
25175 * gcov.c (print_version): Ditto.
25176 * gcov-tool.c (print_version): Ditto.
25177 * gengtype.c (create_file): Ditto.
25178 * doc/cpp.texi: Bump @copying's copyright year.
25179 * doc/cppinternals.texi: Ditto.
25180 * doc/gcc.texi: Ditto.
25181 * doc/gccint.texi: Ditto.
25182 * doc/gcov.texi: Ditto.
25183 * doc/install.texi: Ditto.
25184 * doc/invoke.texi: Ditto.
25186 2021-01-01 Jakub Jelinek <jakub@redhat.com>
25188 * ChangeLog-2020: Rotate ChangeLog. New file.
25191 Copyright (C) 2021 Free Software Foundation, Inc.
25193 Copying and distribution of this file, with or without modification,
25194 are permitted in any medium without royalty provided the copyright
25195 notice and this notice are preserved.