1 2021-11-08 Bill Schmidt <wschmidt@linux.ibm.com>
3 * config/rs6000/rs6000-builtin-new.def (VEC_INIT_V16QI): Use
4 escape-newline support.
5 (VEC_INIT_V4SI): Likewise.
6 (VEC_INIT_V8HI): Likewise.
9 (VFIRSTMISMATCHOREOSINDEX_V16QI): Likewise.
10 (VFIRSTMISMATCHOREOSINDEX_V8HI): Likewise.
11 (VFIRSTMISMATCHOREOSINDEX_V4SI): Likewise.
15 (FMAF128_ODD): Likewise.
22 (TABORTDCI): Likewise.
23 (SE_LXVRBX): Likewise.
24 (SE_LXVRHX): Likewise.
25 (SE_LXVRWX): Likewise.
26 (SE_LXVRDX): Likewise.
27 (VREPLACE_UN_UV2DI): Likewise.
28 (VREPLACE_UN_UV4SI): Likewise.
29 (VREPLACE_UN_V2DI): Likewise.
30 (VREPLACE_ELT_UV2DI): Likewise.
31 (VREPLACE_ELT_V2DI): Likewise.
32 (ZE_LXVRBX): Likewise.
33 (ZE_LXVRHX): Likewise.
34 (ZE_LXVRWX): Likewise.
35 (ZE_LXVRDX): Likewise.
41 (PMXVBF16GER2): Likewise.
42 (PMXVBF16GER2_INTERNAL): Likewise.
43 (PMXVBF16GER2NN): Likewise.
44 (PMXVBF16GER2NN_INTERNAL): Likewise.
45 (PMXVBF16GER2NP): Likewise.
46 (PMXVBF16GER2NP_INTERNAL): Likewise.
47 (PMXVBF16GER2PN): Likewise.
48 (PMXVBF16GER2PN_INTERNAL): Likewise.
49 (PMXVBF16GER2PP): Likewise.
50 (PMXVBF16GER2PP_INTERNAL): Likewise.
51 (PMXVF16GER2): Likewise.
52 (PMXVF16GER2_INTERNAL): Likewise.
53 (PMXVF16GER2NN): Likewise.
54 (PMXVF16GER2NN_INTERNAL): Likewise.
55 (PMXVF16GER2NP): Likewise.
56 (PMXVF16GER2NP_INTERNAL): Likewise.
57 (PMXVF16GER2PN): Likewise.
58 (PMXVF16GER2PN_INTERNAL): Likewise.
59 (PMXVF16GER2PP): Likewise.
60 (PMXVF16GER2PP_INTERNAL): Likewise.
61 (PMXVF32GER_INTERNAL): Likewise.
62 (PMXVF32GERNN): Likewise.
63 (PMXVF32GERNN_INTERNAL): Likewise.
64 (PMXVF32GERNP): Likewise.
65 (PMXVF32GERNP_INTERNAL): Likewise.
66 (PMXVF32GERPN): Likewise.
67 (PMXVF32GERPN_INTERNAL): Likewise.
68 (PMXVF32GERPP): Likewise.
69 (PMXVF32GERPP_INTERNAL): Likewise.
70 (PMXVF64GER): Likewise.
71 (PMXVF64GER_INTERNAL): Likewise.
72 (PMXVF64GERNN): Likewise.
73 (PMXVF64GERNN_INTERNAL): Likewise.
74 (PMXVF64GERNP): Likewise.
75 (PMXVF64GERNP_INTERNAL): Likewise.
76 (PMXVF64GERPN): Likewise.
77 (PMXVF64GERPN_INTERNAL): Likewise.
78 (PMXVF64GERPP): Likewise.
79 (PMXVF64GERPP_INTERNAL): Likewise.
80 (PMXVI16GER2): Likewise.
81 (PMXVI16GER2_INTERNAL): Likewise.
82 (PMXVI16GER2PP): Likewise.
83 (PMXVI16GER2PP_INTERNAL): Likewise.
84 (PMXVI16GER2S): Likewise.
85 (PMXVI16GER2S_INTERNAL): Likewise.
86 (PMXVI16GER2SPP): Likewise.
87 (PMXVI16GER2SPP_INTERNAL): Likewise.
88 (PMXVI4GER8): Likewise.
89 (PMXVI4GER8_INTERNAL): Likewise.
90 (PMXVI4GER8PP): Likewise.
91 (PMXVI4GER8PP_INTERNAL): Likewise.
92 (PMXVI8GER4): Likewise.
93 (PMXVI8GER4_INTERNAL): Likewise.
94 (PMXVI8GER4PP): Likewise.
95 (PMXVI8GER4PP_INTERNAL): Likewise.
96 (PMXVI8GER4SPP): Likewise.
97 (PMXVI8GER4SPP_INTERNAL): Likewise.
98 * config/rs6000/rs6000-gen-builtins.c (MAXLINES): New macro.
99 (linebuf): Increase size.
100 (lines): New variable.
101 (lastline): Likewise.
102 (real_line_pos): New function.
103 (diag): Change signature.
104 (bif_diag): Change signature; support escape-newline handling.
105 (ovld_diag): Likewise.
106 (fatal): Move earlier.
107 (consume_whitespace): Adjust diag call.
108 (advance_line): Add escape-newline handling; call fatal.
109 (safe_inc_pos): Adjust diag call.
110 (match_identifier): Likewise.
111 (match_integer): Likewise.
112 (match_to_right_bracket): Call fatal instead of diag; adjust diag
114 (match_basetype): Adjust diag calls.
115 (match_bracketed_pair): Likewise.
116 (match_const_restriction): Likewise.
117 (match_type): Likewise.
118 (parse_args): Likewise.
119 (parse_bif_attrs): Likewise.
120 (complete_vector_type): Likewise.
121 (complete_base_type): Likewise.
122 (parse_prototype): Likewise.
123 (parse_bif_entry): Likewise.
124 (parse_bif_stanza): Likewise.
125 (parse_ovld_entry): Likewise.
126 (parse_ovld_stanza): Likewise.
127 (main): Allocate buffers for lines[].
129 2021-11-08 Eric Botcazou <ebotcazou@adacore.com>
131 * config/i386/i386.c (legitimize_pic_address): Adjust comment and
132 use the REG argument on the CM_LARGE_PIC code path as well.
133 * config/i386/predicates.md (gotoff_operand): Do not treat VxWorks
134 specially with the large code models.
136 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
138 * gimple.c (gimple_call_static_chain_flags): Revert the workaround
139 allowing interposition since issues with binds_to_local_def were
142 2021-11-08 Andrew MacLeod <amacleod@redhat.com>
144 PR tree-optimization/103122
145 * gimple-range.cc (gimple_ranger::range_of_expr): Request the cache
146 entry with "calulate new values" set to false.
148 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
150 * builtins.c (is_simple_builtin): Add builitin_dwarf_cfa
151 and builtin_return_address.
152 (builtin_fnspec): Annotate builtin_return,
153 bulitin_eh_pointer, builtin_eh_filter, builtin_unwind_resume,
154 builtin_cxa_end_cleanup, builtin_eh_copy_values,
155 builtin_frame_address, builtin_apply_args,
156 builtin_asan_before_dynamic_init, builtin_asan_after_dynamic_init,
157 builtin_prefetch, builtin_dwarf_cfa, builtin_return_addrss
159 * ipa-pure-const.c (special_builtin_state): Add builtin_dwarf_cfa
160 and builtin_return_address.
162 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
164 PR tree-optimization/103177
165 * passes.def: Move uncprop after pure/const and modref.
167 2021-11-08 Martin Jambor <mjambor@suse.cz>
171 * tree-inline.c (remap_gimple_stmt): Unshare the expression without
172 location before invoking remap_with_debug_expressions on it.
173 * ipa-param-manipulation.c
174 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
176 2021-11-08 David Edelsohn <dje.gcc@gmail.com>
178 * config/rs6000/vsx.md (vsx_splat_v4si_di): Revert "wa"
181 2021-11-08 Richard Biener <rguenther@suse.de>
183 * cfgloop.c (verify_loop_structure): Use a temporary BB flag
184 instead of an sbitmap to cache irreducible state.
186 2021-11-08 Aldy Hernandez <aldyh@redhat.com>
188 PR tree-optimization/103120
189 * gimple-range-path.cc (path_range_query::range_defined_in_block):
190 Bail if there's a cache entry.
192 2021-11-08 Bill Schmidt <wschmidt@linux.ibm.com>
194 * config/rs6000/rs6000.c (rs6000_builtin_reciprocal): Use
195 rs6000_builtin_decls_x when appropriate.
196 (add_condition_to_bb): Likewise.
197 (rs6000_atomic_assign_expand_fenv): Likewise.
199 2021-11-08 Bill Schmidt <wschmidt@linux.ibm.com>
201 * config/rs6000/rs6000-call.c (rs6000_new_builtin_decl): New function.
202 (rs6000_builtin_decl): Call it.
204 2021-11-08 Martin Liska <mliska@suse.cz>
206 * cgraph.c (cgraph_node::dump): Dump it from decl.
208 2021-11-08 Martin Liska <mliska@suse.cz>
210 PR gcov-profile/100520
211 * coverage.c (coverage_compute_profile_id): Strip .gk when
212 compare debug is used.
213 * system.h (endswith): New function.
215 2021-11-08 Martin Liska <mliska@suse.cz>
217 * cgraph.c (cgraph_node::dump): Dump static_chain_decl.
219 2021-11-08 Thomas Schwinge <thomas@codesourcery.com>
221 * config/rs6000/rbtree.c: Fix 'Copyright (C) 2020-21' into '2020-2021'
222 * config/rs6000/rbtree.h: Likewise.
223 * config/rs6000/rs6000-builtin-new.def: Likewise.
224 * config/rs6000/rs6000-gen-builtins.c: Likewise.
225 * config/rs6000/rs6000-overload.def: Likewise.
227 2021-11-08 Richard Sandiford <richard.sandiford@arm.com>
229 * config/aarch64/aarch64.c (aarch64_ldn_stn_vectors): New function.
230 (aarch64_address_cost): Use it instead of testing for CImode and
233 2021-11-08 Richard Sandiford <richard.sandiford@arm.com>
235 * genmodes.c (emit_insn_modes_h): Define NUM_MODE_* macros.
236 * expmed.h (NUM_MODE_INT): Delete in favor of genmodes definitions.
237 (NUM_MODE_PARTIAL_INT, NUM_MODE_VECTOR_INT): Likewise.
238 * real.h (real_format_for_mode): Use NUM_MODE_FLOAT and
239 NUM_MODE_DECIMAL_FLOAT.
240 (REAL_MODE_FORMAT): Likewise.
242 2021-11-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
244 * tree-vect-loop.c (vect_better_loop_vinfo_p): Change how epilogue loop
247 2021-11-08 Richard Biener <rguenther@suse.de>
249 * tree-vectorizer.h (vect_create_loop_vinfo): Add main_loop_info
251 * tree-vect-loop.c (vect_create_loop_vinfo): Likewise. Set
252 LOOP_VINFO_ORIG_LOOP_INFO and conditionalize set of
253 LOOP_VINFO_NITERS_ASSUMPTIONS.
254 (vect_analyze_loop_1): Adjust.
255 (vect_analyze_loop): Move loop constraint setting and
256 SCEV/niter reset here from vect_create_loop_vinfo to perform
258 (vect_analyze_loop_form): Move dumping of symbolic niters
259 here from vect_create_loop_vinfo.
261 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
263 * ipa-modref-tree.h (enum modref_special_parms): New enum.
264 (struct modref_access_node): update for special parms.
265 (struct modref_ref_node): Likewise.
266 (struct modref_parm_map): Likewise.
267 (struct modref_tree): Likewise.
268 * ipa-modref.c (dump_access): Likewise.
269 (get_access): Detect static chain.
270 (parm_map_for_arg): Take tree as arg instead of
272 (merge_call_side_effects): Compute map for static chain.
273 (process_fnspec): Update.
274 (struct escape_point): Remove retslot_arg and static_chain_arg.
275 (analyze_parms): Update.
276 (compute_parm_map): Update.
277 (propagate_unknown_call): Update.
278 (modref_propagate_in_scc): Update.
279 (modref_merge_call_site_flags): Update.
280 (ipa_merge_modref_summary_after_inlining): Update.
281 * tree-ssa-alias.c (modref_may_conflict): Handle static chain.
282 * ipa-modref-tree.c (test_merge): Update.
284 2021-11-08 Haochen Gui <guihaoc@gcc.gnu.org>
286 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin): Disable
287 gimple fold for VSX_BUILTIN_XVMINDP, ALTIVEC_BUILTIN_VMINFP,
288 VSX_BUILTIN_XVMAXDP, ALTIVEC_BUILTIN_VMAXFP when fast-math is not
291 2021-11-08 liuhongt <hongtao.liu@intel.com>
293 PR tree-optimization/103077
294 * doc/invoke.texi (Options That Control Optimization):
295 Update documentation for -ftree-loop-vectorize and
296 -ftree-slp-vectorize which are enabled by default at -02.
298 2021-11-08 liuhongt <hongtao.liu@intel.com>
301 * match.pd (Simplifcation (trunc)copysign((extend)a, (extend)b)
302 to .COPYSIGN (a, b)): Add !HONOR_SNANS.
304 2021-11-08 liuhongt <hongtao.liu@intel.com>
308 (trunc)fma ((extend)a, (extend)b, (extend)c) to IFN_FMA (a, b,
309 c) under flag_unsafe_math_optimizations.
311 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
314 * params.opt (modref-max-depth): Add range.
315 (modref-max-adjustments): Fix range.
317 2021-11-07 Aldy Hernandez <aldyh@redhat.com>
319 * tree-pass.h (make_pass_vrp_threader): Remove.
320 * tree-ssa-threadbackward.c
321 (back_threader_profitability::profitable_path_p): Remove
322 ASSERT_EXPR references.
323 * tree-ssa-threadedge.c (jt_state::register_equivs_stmt): Same.
324 * tree-vrp.c (vrp_folder::simplify_casted_conds): Same.
326 (class hybrid_threader): Remove.
327 (hybrid_threader::hybrid_threader): Remove.
328 (hybrid_threader::~hybrid_threader): Remove.
329 (hybrid_threader::before_dom_children): Remove.
330 (hybrid_threader::after_dom_children): Remove.
331 (execute_vrp_threader): Remove.
332 (class pass_vrp_threader): Remove.
333 (make_pass_vrp_threader): Remove.
335 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
337 * gimple.c (gimple_call_arg_flags): Use interposable_eaf_flags.
338 (gimple_call_retslot_flags): Likewise.
339 (gimple_call_static_chain_flags): Likewise.
340 * ipa-modref.c (remove_useless_eaf_flags): Do not remove everything for
342 (modref_summary::useful_p): Likewise.
343 (modref_summary_lto::useful_p): Likewise.
344 (analyze_parms): Do not give up on NOVOPS.
345 (analyze_function): When dumping report chnages in EAF flags
346 between IPA and local pass.
347 (modref_merge_call_site_flags): Compute implicit eaf flags
348 based on callee ecf_flags and fnspec; if the function does not
349 bind to current defs use interposable_eaf_flags.
350 (modref_propagate_flags_in_scc): Update.
351 * ipa-modref.h (interposable_eaf_flags): New function.
353 2021-11-07 Bill Schmidt <wschmidt@linux.ibm.com>
355 * config/rs6000/rs6000-call.c (rs6000_expand_new_builtin): New
357 (rs6000_invalid_new_builtin): New function.
358 (rs6000_expand_builtin): Call rs6000_expand_new_builtin.
359 (rs6000_expand_ldst_mask): New function.
360 (new_cpu_expand_builtin): Likewise.
361 (elemrev_icode): Likewise.
362 (ldv_expand_builtin): Likewise.
363 (lxvrse_expand_builtin): Likewise.
364 (lxvrze_expand_builtin): Likewise.
365 (stv_expand_builtin): Likewise.
366 (new_mma_expand_builtin): Likewise.
367 (new_htm_spr_num): Likewise.
368 (new_htm_expand_builtin): Likewise.
369 (rs6000_expand_new_builtin): Likewise.
370 (rs6000_init_builtins): Initialize altivec_builtin_mask_for_load.
372 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
374 * ipa-modref.c (modref_lattice): Add do_dataflow,
375 changed and propagate_to fields.
376 (modref_lattice::release): Free propagate_to
377 (modref_lattice::merge): Do not give up early on unknown
379 (modref_lattice::merge_deref): Likewise.
380 (modref_eaf_analysis): Update toplevel comment.
381 (modref_eaf_analysis::analyze_ssa_name): Record postponned ssa names;
382 do optimistic dataflow initialization.
383 (modref_eaf_analysis::merge_with_ssa_name): Build dataflow graph.
384 (modref_eaf_analysis::propagate): New member function.
385 (analyze_parms): Update to new API of modref_eaf_analysis.
387 2021-11-06 Jan Hubicka <hubicka@ucw.cz>
389 * cgraph.h (cgraph_node::can_be_discarded_p): Do not
390 return true on functions from other partition.
392 2021-11-06 David Edelsohn <dje.gcc@gmail.com>
394 * config/rs6000/vsx.md (vsx_splat_v4si): Change constraints to "wa".
395 (vsx_splat_v4si_di): Change constraint to "wa".
397 2021-11-06 Aldy Hernandez <aldyh@redhat.com>
399 PR tree-optimization/103061
400 * value-relation.cc (path_oracle::path_oracle): Initialize
402 (path_oracle::killing_def): Set m_killed_defs.
403 (path_oracle::query_relation): Do not look at the root oracle for
405 * value-relation.h (class path_oracle): Add m_killed_defs.
407 2021-11-06 Aldy Hernandez <aldyh@redhat.com>
409 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
410 Remove gotos and other cleanups.
412 2021-11-05 Jan Hubicka <hubicka@ucw.cz>
415 * ipa-modref-tree.h (modref_tree::insert): Do nothing for
416 paradoxical and zero sized accesses.
418 2021-11-05 Jan Hubicka <hubicka@ucw.cz>
421 * ipa-modref-tree.h (struct modref_access_node): Avoid left shift
424 2021-11-05 Iain Sandoe <iain@sandoe.co.uk>
426 * config/darwin.h (ASM_GENERATE_INTERNAL_LABEL): Add LTRAMP
427 to the list of symbol prefixes that must be made linker-
430 2021-11-05 Iain Sandoe <iain@sandoe.co.uk>
432 * config.host: Add support for aarch64-*-darwin.
433 * config/aarch64/host-aarch64-darwin.c: New file.
434 * config/aarch64/x-darwin: New file.
436 2021-11-05 Andrew MacLeod <amacleod@redhat.com>
438 PR tree-optimization/103093
439 * gimple-range-gori.cc (range_def_chain::get_imports): Remove assert.
441 2021-11-05 Andrew MacLeod <amacleod@redhat.com>
443 PR tree-optimization/102943
444 * gimple-range-cache.cc (class update_list): New.
445 (update_list::add): Replace add_to_update.
446 (update_list::pop): New.
447 (ranger_cache::ranger_cache): Adjust.
448 (ranger_cache::~ranger_cache): Adjust.
449 (ranger_cache::add_to_update): Delete.
450 (ranger_cache::propagate_cache): Adjust to new class.
451 (ranger_cache::propagate_updated_value): Ditto.
452 (ranger_cache::fill_block_cache): Ditto.
453 * gimple-range-cache.h (class ranger_cache): Adjust to update class.
455 2021-11-05 Richard Biener <rguenther@suse.de>
457 * tree-vect-loop.c (vect_analyze_loop): Remove obsolete
458 comment and expand on another one. Combine nested if.
460 2021-11-05 John David Anglin <danglin@gcc.gnu.org>
463 * config/pa/pa-modes.def: Add OImode integer type.
464 * config/pa/pa.c (pa_scalar_mode_supported_p): Allow TImode
466 * config/pa/pa.h (MIN_UNITS_PER_WORD) Define to MIN_UNITS_PER_WORD
467 to UNITS_PER_WORD if IN_LIBGCC2.
468 * config/pa/pa.md (addti3, addvti3, subti3, subvti3, negti2,
469 negvti2, ashlti3, shrpd_internal): New patterns.
470 Change some multi instruction types to multi.
472 2021-11-05 Jakub Jelinek <jakub@redhat.com>
476 (stringop_algs::stringop_strategy::stringop_strategy): Make the ctor
479 2021-11-05 Wilco Dijkstra <wdijkstr@arm.com>
482 * config/aarch64/aarch64.c (aarch64_mov_operand_p): Strip the salt
484 * config/aarch64/constraints.md: Support const in Usw.
486 2021-11-05 John David Anglin <danglin@gcc.gnu.org>
488 * config/pa/pa.h (PREFERRED_DEBUGGING_TYPE): Define to DWARF2_DEBUG.
489 * config/pa/pa64-hpux.h (PREFERRED_DEBUGGING_TYPE): Remove define.
491 2021-11-05 Richard Biener <rguenther@suse.de>
493 * tree-vectorizer.h (vec_info_shared::n_stmts): Add.
494 (LOOP_VINFO_N_STMTS): Likewise.
495 (vec_info_for_bb): Remove unused function.
496 * tree-vectorizer.c (vec_info_shared::vec_info_shared):
497 Initialize n_stmts member.
498 * tree-vect-loop.c: Remove INCLUDE_FUNCTIONAL.
499 (vect_create_loop_vinfo): Do not set loop->aux.
500 (vect_analyze_loop_2): Do not get n_stmts as argument,
501 instead use LOOP_VINFO_N_STMTS. Set LOOP_VINFO_VECTORIZABLE_P
503 (vect_analyze_loop_1): Remove callback, get the mode iterator
504 and autodetected_vector_mode as argument, advancing the
505 iterator and initializing autodetected_vector_mode here.
506 (vect_analyze_loop): Split analysis loop into two, first
507 processing main loops only and then epilogues.
509 2021-11-05 Martin Jambor <mjambor@suse.cz>
511 * ipa-prop.c (compute_complex_assign_jump_func): Remove
512 unnecessary check for RECORD_TYPE.
514 2021-11-05 Gerald Pfeifer <gerald@pfeifer.com>
516 * doc/install.texi2html: Do not generate old.html any longer.
518 2021-11-05 Martin Liska <mliska@suse.cz>
521 * opts.c (finish_options): Reset flag_gtoggle when it is used.
523 2021-11-05 Jakub Jelinek <jakub@redhat.com>
526 * dwarf2out.c (add_const_value_attribute): Add MODE argument, use it
527 in CONST_WIDE_INT handling. Adjust recursive calls.
528 (add_location_or_const_value_attribute): Pass DECL_MODE (decl) to
529 new add_const_value_attribute argument.
530 (tree_add_const_value_attribute): Pass TYPE_MODE (type) to new
531 add_const_value_attribute argument.
533 2021-11-05 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
535 * config/vx-common.h: Test value of TARGET_VXWORKS7 rather
538 2021-11-05 Richard Biener <rguenther@suse.de>
540 * tree-vectorizer.h (struct vect_loop_form_info): New.
541 (vect_analyze_loop_form): Adjust.
542 (vect_create_loop_vinfo): New.
543 * tree-parloops.c (gather_scalar_reductions): Adjust for
544 vect_analyze_loop_form API change.
545 * tree-vect-loop.c: Include <functional>.
546 (vect_analyze_loop_form_1): Rename to vect_analyze_loop_form,
547 take struct vect_loop_form_info as output parameter and adjust.
548 (vect_analyze_loop_form): Rename to vect_create_loop_vinfo and
549 split out call to the original vect_analyze_loop_form_1.
550 (vect_reanalyze_as_main_loop): Rename to...
551 (vect_analyze_loop_1): ... this, factor out the call to
552 vect_analyze_loop_form and generalize to be able to use it twice ...
553 (vect_analyze_loop): ... here. Perform vect_analyze_loop_form
556 2021-11-05 Xionghu Luo <luoxhu@linux.ibm.com>
559 * config/rs6000/fusion.md: Regenerate.
560 * config/rs6000/genfusion.pl: Fix incorrect clobber constraint.
562 2021-11-04 Andreas Krebbel <krebbel@linux.ibm.com>
564 * config/s390/s390.h (STACK_CHECK_MOVING_SP): New macro
567 2021-11-04 Tamar Christina <tamar.christina@arm.com>
569 * config/aarch64/aarch64-builtins.c
570 (aarch64_general_gimple_fold_builtin): Add ashl, sshl, ushl, ashr,
571 ashr_simd, lshr, lshr_simd.
572 * config/aarch64/aarch64-simd-builtins.def (lshr): Use USHIFTIMM.
573 * config/aarch64/arm_neon.h (vshr_n_u8, vshr_n_u16, vshr_n_u32,
574 vshrq_n_u8, vshrq_n_u16, vshrq_n_u32, vshrq_n_u64): Fix type hack.
576 2021-11-04 Tamar Christina <tamar.christina@arm.com>
578 * match.pd: New negate+shift pattern.
580 2021-11-04 Andrew MacLeod <amacleod@redhat.com>
582 PR tree-optimization/103079
583 * gimple-range-gori.cc (gimple_range_calc_op1): Treat undefined as
585 (gimple_range_calc_op2): Ditto.
587 2021-11-04 Martin Jambor <mjambor@suse.cz>
590 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
591 members remap_with_debug_expressions, m_dead_ssa_debug_equiv,
592 m_dead_stmt_debug_equiv and prepare_debug_expressions. Added
593 parameter to mark_dead_statements.
594 * ipa-param-manipulation.c: Include tree-phinodes.h and cfgexpand.h.
595 (ipa_param_body_adjustments::mark_dead_statements): New parameter
596 debugstack, push into it all SSA names used in debug statements,
597 produce m_dead_ssa_debug_equiv mapping for the removed param.
598 (replace_with_mapped_expr): New function.
599 (ipa_param_body_adjustments::remap_with_debug_expressions): Likewise.
600 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
601 (ipa_param_body_adjustments::common_initialization): Gather and
602 procecc SSA which will be removed but are in debug statements. Simplify.
603 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
605 * tree-inline.c (remap_gimple_stmt): Create a debug bind when possible
606 when avoiding a copy of an unnecessary statement. Remap removed SSA
607 names in existing debug statements.
608 (tree_function_versioning): Do not create DEBUG_EXPR_DECL for removed
609 parameters if we have already done so.
611 2021-11-04 Jan Hubicka <hubicka@ucw.cz>
614 * gimple.c (gimple_call_static_chain_flags): Handle case when
615 nested function does not bind locally.
617 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
619 * config/aarch64/aarch64.c (aarch64_function_value): Generate
620 a register rtx for Neon vector-tuple modes.
621 (aarch64_layout_arg): Likewise.
623 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
625 * lower-subreg.c (simple_move): Prevent decomposition if
626 modes are not tieable.
628 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
629 Richard Sandiford <richard.sandiford@arm.com>
631 * config/aarch64/aarch64-builtins.c (v2x8qi_UP): Define.
632 (v2x4hi_UP): Likewise.
633 (v2x4hf_UP): Likewise.
634 (v2x4bf_UP): Likewise.
635 (v2x2si_UP): Likewise.
636 (v2x2sf_UP): Likewise.
637 (v2x1di_UP): Likewise.
638 (v2x1df_UP): Likewise.
639 (v2x16qi_UP): Likewise.
640 (v2x8hi_UP): Likewise.
641 (v2x8hf_UP): Likewise.
642 (v2x8bf_UP): Likewise.
643 (v2x4si_UP): Likewise.
644 (v2x4sf_UP): Likewise.
645 (v2x2di_UP): Likewise.
646 (v2x2df_UP): Likewise.
647 (v3x8qi_UP): Likewise.
648 (v3x4hi_UP): Likewise.
649 (v3x4hf_UP): Likewise.
650 (v3x4bf_UP): Likewise.
651 (v3x2si_UP): Likewise.
652 (v3x2sf_UP): Likewise.
653 (v3x1di_UP): Likewise.
654 (v3x1df_UP): Likewise.
655 (v3x16qi_UP): Likewise.
656 (v3x8hi_UP): Likewise.
657 (v3x8hf_UP): Likewise.
658 (v3x8bf_UP): Likewise.
659 (v3x4si_UP): Likewise.
660 (v3x4sf_UP): Likewise.
661 (v3x2di_UP): Likewise.
662 (v3x2df_UP): Likewise.
663 (v4x8qi_UP): Likewise.
664 (v4x4hi_UP): Likewise.
665 (v4x4hf_UP): Likewise.
666 (v4x4bf_UP): Likewise.
667 (v4x2si_UP): Likewise.
668 (v4x2sf_UP): Likewise.
669 (v4x1di_UP): Likewise.
670 (v4x1df_UP): Likewise.
671 (v4x16qi_UP): Likewise.
672 (v4x8hi_UP): Likewise.
673 (v4x8hf_UP): Likewise.
674 (v4x8bf_UP): Likewise.
675 (v4x4si_UP): Likewise.
676 (v4x4sf_UP): Likewise.
677 (v4x2di_UP): Likewise.
678 (v4x2df_UP): Likewise.
679 (TYPES_GETREGP): Delete.
680 (TYPES_SETREGP): Likewise.
681 (TYPES_LOADSTRUCT_U): Define.
682 (TYPES_LOADSTRUCT_P): Likewise.
683 (TYPES_LOADSTRUCT_LANE_U): Likewise.
684 (TYPES_LOADSTRUCT_LANE_P): Likewise.
685 (TYPES_STORE1P): Move for consistency.
686 (TYPES_STORESTRUCT_U): Define.
687 (TYPES_STORESTRUCT_P): Likewise.
688 (TYPES_STORESTRUCT_LANE_U): Likewise.
689 (TYPES_STORESTRUCT_LANE_P): Likewise.
690 (aarch64_simd_tuple_types): Define.
691 (aarch64_lookup_simd_builtin_type): Handle tuple type lookup.
692 (aarch64_init_simd_builtin_functions): Update frontend lookup
693 for builtin functions after handling arm_neon.h pragma.
694 (register_tuple_type): Manually set modes of single-integer
695 tuple types. Record tuple types.
696 * config/aarch64/aarch64-modes.def
697 (ADV_SIMD_D_REG_STRUCT_MODES): Define D-register tuple modes.
698 (ADV_SIMD_Q_REG_STRUCT_MODES): Define Q-register tuple modes.
699 (SVE_MODES): Give single-vector modes priority over vector-
701 (VECTOR_MODES_WITH_PREFIX): Set partial-vector mode order to
702 be after all single-vector modes.
703 * config/aarch64/aarch64-simd-builtins.def: Update builtin
704 generator macros to reflect modifications to the backend
706 * config/aarch64/aarch64-simd.md (aarch64_simd_ld2<mode>):
707 Use vector-tuple mode iterator and rename to...
708 (aarch64_simd_ld2<vstruct_elt>): This.
709 (aarch64_simd_ld2r<mode>): Use vector-tuple mode iterator and
711 (aarch64_simd_ld2r<vstruct_elt>): This.
712 (aarch64_vec_load_lanesoi_lane<mode>): Use vector-tuple mode
713 iterator and rename to...
714 (aarch64_vec_load_lanes<mode>_lane<vstruct_elt>): This.
715 (vec_load_lanesoi<mode>): Use vector-tuple mode iterator and
717 (vec_load_lanes<mode><vstruct_elt>): This.
718 (aarch64_simd_st2<mode>): Use vector-tuple mode iterator and
720 (aarch64_simd_st2<vstruct_elt>): This.
721 (aarch64_vec_store_lanesoi_lane<mode>): Use vector-tuple mode
722 iterator and rename to...
723 (aarch64_vec_store_lanes<mode>_lane<vstruct_elt>): This.
724 (vec_store_lanesoi<mode>): Use vector-tuple mode iterator and
726 (vec_store_lanes<mode><vstruct_elt>): This.
727 (aarch64_simd_ld3<mode>): Use vector-tuple mode iterator and
729 (aarch64_simd_ld3<vstruct_elt>): This.
730 (aarch64_simd_ld3r<mode>): Use vector-tuple mode iterator and
732 (aarch64_simd_ld3r<vstruct_elt>): This.
733 (aarch64_vec_load_lanesci_lane<mode>): Use vector-tuple mode
734 iterator and rename to...
735 (vec_load_lanesci<mode>): This.
736 (aarch64_simd_st3<mode>): Use vector-tuple mode iterator and
738 (aarch64_simd_st3<vstruct_elt>): This.
739 (aarch64_vec_store_lanesci_lane<mode>): Use vector-tuple mode
740 iterator and rename to...
741 (vec_store_lanesci<mode>): This.
742 (aarch64_simd_ld4<mode>): Use vector-tuple mode iterator and
744 (aarch64_simd_ld4<vstruct_elt>): This.
745 (aarch64_simd_ld4r<mode>): Use vector-tuple mode iterator and
747 (aarch64_simd_ld4r<vstruct_elt>): This.
748 (aarch64_vec_load_lanesxi_lane<mode>): Use vector-tuple mode
749 iterator and rename to...
750 (vec_load_lanesxi<mode>): This.
751 (aarch64_simd_st4<mode>): Use vector-tuple mode iterator and
753 (aarch64_simd_st4<vstruct_elt>): This.
754 (aarch64_vec_store_lanesxi_lane<mode>): Use vector-tuple mode
755 iterator and rename to...
756 (vec_store_lanesxi<mode>): This.
757 (mov<mode>): Define for Neon vector-tuple modes.
758 (aarch64_ld1x3<VALLDIF:mode>): Use vector-tuple mode iterator
760 (aarch64_ld1x3<vstruct_elt>): This.
761 (aarch64_ld1_x3_<mode>): Use vector-tuple mode iterator and
763 (aarch64_ld1_x3_<vstruct_elt>): This.
764 (aarch64_ld1x4<VALLDIF:mode>): Use vector-tuple mode iterator
766 (aarch64_ld1x4<vstruct_elt>): This.
767 (aarch64_ld1_x4_<mode>): Use vector-tuple mode iterator and
769 (aarch64_ld1_x4_<vstruct_elt>): This.
770 (aarch64_st1x2<VALLDIF:mode>): Use vector-tuple mode iterator
772 (aarch64_st1x2<vstruct_elt>): This.
773 (aarch64_st1_x2_<mode>): Use vector-tuple mode iterator and
775 (aarch64_st1_x2_<vstruct_elt>): This.
776 (aarch64_st1x3<VALLDIF:mode>): Use vector-tuple mode iterator
778 (aarch64_st1x3<vstruct_elt>): This.
779 (aarch64_st1_x3_<mode>): Use vector-tuple mode iterator and
781 (aarch64_st1_x3_<vstruct_elt>): This.
782 (aarch64_st1x4<VALLDIF:mode>): Use vector-tuple mode iterator
784 (aarch64_st1x4<vstruct_elt>): This.
785 (aarch64_st1_x4_<mode>): Use vector-tuple mode iterator and
787 (aarch64_st1_x4_<vstruct_elt>): This.
788 (*aarch64_mov<mode>): Define for vector-tuple modes.
789 (*aarch64_be_mov<mode>): Likewise.
790 (aarch64_ld<VSTRUCT:nregs>r<VALLDIF:mode>): Use vector-tuple
791 mode iterator and rename to...
792 (aarch64_ld<nregs>r<vstruct_elt>): This.
793 (aarch64_ld2<mode>_dreg): Use vector-tuple mode iterator and
795 (aarch64_ld2<vstruct_elt>_dreg): This.
796 (aarch64_ld3<mode>_dreg): Use vector-tuple mode iterator and
798 (aarch64_ld3<vstruct_elt>_dreg): This.
799 (aarch64_ld4<mode>_dreg): Use vector-tuple mode iterator and
801 (aarch64_ld4<vstruct_elt>_dreg): This.
802 (aarch64_ld<VSTRUCT:nregs><VDC:mode>): Use vector-tuple mode
803 iterator and rename to...
804 (aarch64_ld<nregs><vstruct_elt>): Use vector-tuple mode
805 iterator and rename to...
806 (aarch64_ld<VSTRUCT:nregs><VQ:mode>): Use vector-tuple mode
807 (aarch64_ld1x2<VQ:mode>): Delete.
808 (aarch64_ld1x2<VDC:mode>): Use vector-tuple mode iterator and
810 (aarch64_ld1x2<vstruct_elt>): This.
811 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use vector-
812 tuple mode iterator and rename to...
813 (aarch64_ld<nregs>_lane<vstruct_elt>): This.
814 (aarch64_get_dreg<VSTRUCT:mode><VDC:mode>): Delete.
815 (aarch64_get_qreg<VSTRUCT:mode><VQ:mode>): Likewise.
816 (aarch64_st2<mode>_dreg): Use vector-tuple mode iterator and
818 (aarch64_st2<vstruct_elt>_dreg): This.
819 (aarch64_st3<mode>_dreg): Use vector-tuple mode iterator and
821 (aarch64_st3<vstruct_elt>_dreg): This.
822 (aarch64_st4<mode>_dreg): Use vector-tuple mode iterator and
824 (aarch64_st4<vstruct_elt>_dreg): This.
825 (aarch64_st<VSTRUCT:nregs><VDC:mode>): Use vector-tuple mode
826 iterator and rename to...
827 (aarch64_st<nregs><vstruct_elt>): This.
828 (aarch64_st<VSTRUCT:nregs><VQ:mode>): Use vector-tuple mode
829 iterator and rename to aarch64_st<nregs><vstruct_elt>.
830 (aarch64_st<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use vector-
831 tuple mode iterator and rename to...
832 (aarch64_st<nregs>_lane<vstruct_elt>): This.
833 (aarch64_set_qreg<VSTRUCT:mode><VQ:mode>): Delete.
834 (aarch64_simd_ld1<mode>_x2): Use vector-tuple mode iterator
836 (aarch64_simd_ld1<vstruct_elt>_x2): This.
837 * config/aarch64/aarch64.c (aarch64_advsimd_struct_mode_p):
838 Refactor to include new vector-tuple modes.
839 (aarch64_classify_vector_mode): Add cases for new vector-
841 (aarch64_advsimd_partial_struct_mode_p): Define.
842 (aarch64_advsimd_full_struct_mode_p): Likewise.
843 (aarch64_advsimd_vector_array_mode): Likewise.
844 (aarch64_sve_data_mode): Change location in file.
845 (aarch64_array_mode): Handle case of Neon vector-tuple modes.
846 (aarch64_hard_regno_nregs): Handle case of partial Neon
848 (aarch64_classify_address): Refactor to include handling of
849 Neon vector-tuple modes.
850 (aarch64_print_operand): Print "d" for "%R" for a partial
851 Neon vector structure.
852 (aarch64_expand_vec_perm_1): Use new vector-tuple mode.
853 (aarch64_modes_tieable_p): Prevent tieing Neon partial struct
854 modes with scalar machines modes larger than 8 bytes.
855 (aarch64_can_change_mode_class): Don't allow changes between
856 partial and full Neon vector-structure modes.
857 * config/aarch64/arm_neon.h (vst2_lane_f16): Use updated
858 builtin and remove boiler-plate code for opaque mode.
859 (vst2_lane_f32): Likewise.
860 (vst2_lane_f64): Likewise.
861 (vst2_lane_p8): Likewise.
862 (vst2_lane_p16): Likewise.
863 (vst2_lane_p64): Likewise.
864 (vst2_lane_s8): Likewise.
865 (vst2_lane_s16): Likewise.
866 (vst2_lane_s32): Likewise.
867 (vst2_lane_s64): Likewise.
868 (vst2_lane_u8): Likewise.
869 (vst2_lane_u16): Likewise.
870 (vst2_lane_u32): Likewise.
871 (vst2_lane_u64): Likewise.
872 (vst2q_lane_f16): Likewise.
873 (vst2q_lane_f32): Likewise.
874 (vst2q_lane_f64): Likewise.
875 (vst2q_lane_p8): Likewise.
876 (vst2q_lane_p16): Likewise.
877 (vst2q_lane_p64): Likewise.
878 (vst2q_lane_s8): Likewise.
879 (vst2q_lane_s16): Likewise.
880 (vst2q_lane_s32): Likewise.
881 (vst2q_lane_s64): Likewise.
882 (vst2q_lane_u8): Likewise.
883 (vst2q_lane_u16): Likewise.
884 (vst2q_lane_u32): Likewise.
885 (vst2q_lane_u64): Likewise.
886 (vst3_lane_f16): Likewise.
887 (vst3_lane_f32): Likewise.
888 (vst3_lane_f64): Likewise.
889 (vst3_lane_p8): Likewise.
890 (vst3_lane_p16): Likewise.
891 (vst3_lane_p64): Likewise.
892 (vst3_lane_s8): Likewise.
893 (vst3_lane_s16): Likewise.
894 (vst3_lane_s32): Likewise.
895 (vst3_lane_s64): Likewise.
896 (vst3_lane_u8): Likewise.
897 (vst3_lane_u16): Likewise.
898 (vst3_lane_u32): Likewise.
899 (vst3_lane_u64): Likewise.
900 (vst3q_lane_f16): Likewise.
901 (vst3q_lane_f32): Likewise.
902 (vst3q_lane_f64): Likewise.
903 (vst3q_lane_p8): Likewise.
904 (vst3q_lane_p16): Likewise.
905 (vst3q_lane_p64): Likewise.
906 (vst3q_lane_s8): Likewise.
907 (vst3q_lane_s16): Likewise.
908 (vst3q_lane_s32): Likewise.
909 (vst3q_lane_s64): Likewise.
910 (vst3q_lane_u8): Likewise.
911 (vst3q_lane_u16): Likewise.
912 (vst3q_lane_u32): Likewise.
913 (vst3q_lane_u64): Likewise.
914 (vst4_lane_f16): Likewise.
915 (vst4_lane_f32): Likewise.
916 (vst4_lane_f64): Likewise.
917 (vst4_lane_p8): Likewise.
918 (vst4_lane_p16): Likewise.
919 (vst4_lane_p64): Likewise.
920 (vst4_lane_s8): Likewise.
921 (vst4_lane_s16): Likewise.
922 (vst4_lane_s32): Likewise.
923 (vst4_lane_s64): Likewise.
924 (vst4_lane_u8): Likewise.
925 (vst4_lane_u16): Likewise.
926 (vst4_lane_u32): Likewise.
927 (vst4_lane_u64): Likewise.
928 (vst4q_lane_f16): Likewise.
929 (vst4q_lane_f32): Likewise.
930 (vst4q_lane_f64): Likewise.
931 (vst4q_lane_p8): Likewise.
932 (vst4q_lane_p16): Likewise.
933 (vst4q_lane_p64): Likewise.
934 (vst4q_lane_s8): Likewise.
935 (vst4q_lane_s16): Likewise.
936 (vst4q_lane_s32): Likewise.
937 (vst4q_lane_s64): Likewise.
938 (vst4q_lane_u8): Likewise.
939 (vst4q_lane_u16): Likewise.
940 (vst4q_lane_u32): Likewise.
941 (vst4q_lane_u64): Likewise.
942 (vtbl3_s8): Likewise.
943 (vtbl3_u8): Likewise.
944 (vtbl3_p8): Likewise.
945 (vtbl4_s8): Likewise.
946 (vtbl4_u8): Likewise.
947 (vtbl4_p8): Likewise.
948 (vld1_u8_x3): Likewise.
949 (vld1_s8_x3): Likewise.
950 (vld1_u16_x3): Likewise.
951 (vld1_s16_x3): Likewise.
952 (vld1_u32_x3): Likewise.
953 (vld1_s32_x3): Likewise.
954 (vld1_u64_x3): Likewise.
955 (vld1_s64_x3): Likewise.
956 (vld1_f16_x3): Likewise.
957 (vld1_f32_x3): Likewise.
958 (vld1_f64_x3): Likewise.
959 (vld1_p8_x3): Likewise.
960 (vld1_p16_x3): Likewise.
961 (vld1_p64_x3): Likewise.
962 (vld1q_u8_x3): Likewise.
963 (vld1q_s8_x3): Likewise.
964 (vld1q_u16_x3): Likewise.
965 (vld1q_s16_x3): Likewise.
966 (vld1q_u32_x3): Likewise.
967 (vld1q_s32_x3): Likewise.
968 (vld1q_u64_x3): Likewise.
969 (vld1q_s64_x3): Likewise.
970 (vld1q_f16_x3): Likewise.
971 (vld1q_f32_x3): Likewise.
972 (vld1q_f64_x3): Likewise.
973 (vld1q_p8_x3): Likewise.
974 (vld1q_p16_x3): Likewise.
975 (vld1q_p64_x3): Likewise.
976 (vld1_u8_x2): Likewise.
977 (vld1_s8_x2): Likewise.
978 (vld1_u16_x2): Likewise.
979 (vld1_s16_x2): Likewise.
980 (vld1_u32_x2): Likewise.
981 (vld1_s32_x2): Likewise.
982 (vld1_u64_x2): Likewise.
983 (vld1_s64_x2): Likewise.
984 (vld1_f16_x2): Likewise.
985 (vld1_f32_x2): Likewise.
986 (vld1_f64_x2): Likewise.
987 (vld1_p8_x2): Likewise.
988 (vld1_p16_x2): Likewise.
989 (vld1_p64_x2): Likewise.
990 (vld1q_u8_x2): Likewise.
991 (vld1q_s8_x2): Likewise.
992 (vld1q_u16_x2): Likewise.
993 (vld1q_s16_x2): Likewise.
994 (vld1q_u32_x2): Likewise.
995 (vld1q_s32_x2): Likewise.
996 (vld1q_u64_x2): Likewise.
997 (vld1q_s64_x2): Likewise.
998 (vld1q_f16_x2): Likewise.
999 (vld1q_f32_x2): Likewise.
1000 (vld1q_f64_x2): Likewise.
1001 (vld1q_p8_x2): Likewise.
1002 (vld1q_p16_x2): Likewise.
1003 (vld1q_p64_x2): Likewise.
1004 (vld1_s8_x4): Likewise.
1005 (vld1q_s8_x4): Likewise.
1006 (vld1_s16_x4): Likewise.
1007 (vld1q_s16_x4): Likewise.
1008 (vld1_s32_x4): Likewise.
1009 (vld1q_s32_x4): Likewise.
1010 (vld1_u8_x4): Likewise.
1011 (vld1q_u8_x4): Likewise.
1012 (vld1_u16_x4): Likewise.
1013 (vld1q_u16_x4): Likewise.
1014 (vld1_u32_x4): Likewise.
1015 (vld1q_u32_x4): Likewise.
1016 (vld1_f16_x4): Likewise.
1017 (vld1q_f16_x4): Likewise.
1018 (vld1_f32_x4): Likewise.
1019 (vld1q_f32_x4): Likewise.
1020 (vld1_p8_x4): Likewise.
1021 (vld1q_p8_x4): Likewise.
1022 (vld1_p16_x4): Likewise.
1023 (vld1q_p16_x4): Likewise.
1024 (vld1_s64_x4): Likewise.
1025 (vld1_u64_x4): Likewise.
1026 (vld1_p64_x4): Likewise.
1027 (vld1q_s64_x4): Likewise.
1028 (vld1q_u64_x4): Likewise.
1029 (vld1q_p64_x4): Likewise.
1030 (vld1_f64_x4): Likewise.
1031 (vld1q_f64_x4): Likewise.
1032 (vld2_s64): Likewise.
1033 (vld2_u64): Likewise.
1034 (vld2_f64): Likewise.
1035 (vld2_s8): Likewise.
1036 (vld2_p8): Likewise.
1037 (vld2_p64): Likewise.
1038 (vld2_s16): Likewise.
1039 (vld2_p16): Likewise.
1040 (vld2_s32): Likewise.
1041 (vld2_u8): Likewise.
1042 (vld2_u16): Likewise.
1043 (vld2_u32): Likewise.
1044 (vld2_f16): Likewise.
1045 (vld2_f32): Likewise.
1046 (vld2q_s8): Likewise.
1047 (vld2q_p8): Likewise.
1048 (vld2q_s16): Likewise.
1049 (vld2q_p16): Likewise.
1050 (vld2q_p64): Likewise.
1051 (vld2q_s32): Likewise.
1052 (vld2q_s64): Likewise.
1053 (vld2q_u8): Likewise.
1054 (vld2q_u16): Likewise.
1055 (vld2q_u32): Likewise.
1056 (vld2q_u64): Likewise.
1057 (vld2q_f16): Likewise.
1058 (vld2q_f32): Likewise.
1059 (vld2q_f64): Likewise.
1060 (vld3_s64): Likewise.
1061 (vld3_u64): Likewise.
1062 (vld3_f64): Likewise.
1063 (vld3_s8): Likewise.
1064 (vld3_p8): Likewise.
1065 (vld3_s16): Likewise.
1066 (vld3_p16): Likewise.
1067 (vld3_s32): Likewise.
1068 (vld3_u8): Likewise.
1069 (vld3_u16): Likewise.
1070 (vld3_u32): Likewise.
1071 (vld3_f16): Likewise.
1072 (vld3_f32): Likewise.
1073 (vld3_p64): Likewise.
1074 (vld3q_s8): Likewise.
1075 (vld3q_p8): Likewise.
1076 (vld3q_s16): Likewise.
1077 (vld3q_p16): Likewise.
1078 (vld3q_s32): Likewise.
1079 (vld3q_s64): Likewise.
1080 (vld3q_u8): Likewise.
1081 (vld3q_u16): Likewise.
1082 (vld3q_u32): Likewise.
1083 (vld3q_u64): Likewise.
1084 (vld3q_f16): Likewise.
1085 (vld3q_f32): Likewise.
1086 (vld3q_f64): Likewise.
1087 (vld3q_p64): Likewise.
1088 (vld4_s64): Likewise.
1089 (vld4_u64): Likewise.
1090 (vld4_f64): Likewise.
1091 (vld4_s8): Likewise.
1092 (vld4_p8): Likewise.
1093 (vld4_s16): Likewise.
1094 (vld4_p16): Likewise.
1095 (vld4_s32): Likewise.
1096 (vld4_u8): Likewise.
1097 (vld4_u16): Likewise.
1098 (vld4_u32): Likewise.
1099 (vld4_f16): Likewise.
1100 (vld4_f32): Likewise.
1101 (vld4_p64): Likewise.
1102 (vld4q_s8): Likewise.
1103 (vld4q_p8): Likewise.
1104 (vld4q_s16): Likewise.
1105 (vld4q_p16): Likewise.
1106 (vld4q_s32): Likewise.
1107 (vld4q_s64): Likewise.
1108 (vld4q_u8): Likewise.
1109 (vld4q_u16): Likewise.
1110 (vld4q_u32): Likewise.
1111 (vld4q_u64): Likewise.
1112 (vld4q_f16): Likewise.
1113 (vld4q_f32): Likewise.
1114 (vld4q_f64): Likewise.
1115 (vld4q_p64): Likewise.
1116 (vld2_dup_s8): Likewise.
1117 (vld2_dup_s16): Likewise.
1118 (vld2_dup_s32): Likewise.
1119 (vld2_dup_f16): Likewise.
1120 (vld2_dup_f32): Likewise.
1121 (vld2_dup_f64): Likewise.
1122 (vld2_dup_u8): Likewise.
1123 (vld2_dup_u16): Likewise.
1124 (vld2_dup_u32): Likewise.
1125 (vld2_dup_p8): Likewise.
1126 (vld2_dup_p16): Likewise.
1127 (vld2_dup_p64): Likewise.
1128 (vld2_dup_s64): Likewise.
1129 (vld2_dup_u64): Likewise.
1130 (vld2q_dup_s8): Likewise.
1131 (vld2q_dup_p8): Likewise.
1132 (vld2q_dup_s16): Likewise.
1133 (vld2q_dup_p16): Likewise.
1134 (vld2q_dup_s32): Likewise.
1135 (vld2q_dup_s64): Likewise.
1136 (vld2q_dup_u8): Likewise.
1137 (vld2q_dup_u16): Likewise.
1138 (vld2q_dup_u32): Likewise.
1139 (vld2q_dup_u64): Likewise.
1140 (vld2q_dup_f16): Likewise.
1141 (vld2q_dup_f32): Likewise.
1142 (vld2q_dup_f64): Likewise.
1143 (vld2q_dup_p64): Likewise.
1144 (vld3_dup_s64): Likewise.
1145 (vld3_dup_u64): Likewise.
1146 (vld3_dup_f64): Likewise.
1147 (vld3_dup_s8): Likewise.
1148 (vld3_dup_p8): Likewise.
1149 (vld3_dup_s16): Likewise.
1150 (vld3_dup_p16): Likewise.
1151 (vld3_dup_s32): Likewise.
1152 (vld3_dup_u8): Likewise.
1153 (vld3_dup_u16): Likewise.
1154 (vld3_dup_u32): Likewise.
1155 (vld3_dup_f16): Likewise.
1156 (vld3_dup_f32): Likewise.
1157 (vld3_dup_p64): Likewise.
1158 (vld3q_dup_s8): Likewise.
1159 (vld3q_dup_p8): Likewise.
1160 (vld3q_dup_s16): Likewise.
1161 (vld3q_dup_p16): Likewise.
1162 (vld3q_dup_s32): Likewise.
1163 (vld3q_dup_s64): Likewise.
1164 (vld3q_dup_u8): Likewise.
1165 (vld3q_dup_u16): Likewise.
1166 (vld3q_dup_u32): Likewise.
1167 (vld3q_dup_u64): Likewise.
1168 (vld3q_dup_f16): Likewise.
1169 (vld3q_dup_f32): Likewise.
1170 (vld3q_dup_f64): Likewise.
1171 (vld3q_dup_p64): Likewise.
1172 (vld4_dup_s64): Likewise.
1173 (vld4_dup_u64): Likewise.
1174 (vld4_dup_f64): Likewise.
1175 (vld4_dup_s8): Likewise.
1176 (vld4_dup_p8): Likewise.
1177 (vld4_dup_s16): Likewise.
1178 (vld4_dup_p16): Likewise.
1179 (vld4_dup_s32): Likewise.
1180 (vld4_dup_u8): Likewise.
1181 (vld4_dup_u16): Likewise.
1182 (vld4_dup_u32): Likewise.
1183 (vld4_dup_f16): Likewise.
1184 (vld4_dup_f32): Likewise.
1185 (vld4_dup_p64): Likewise.
1186 (vld4q_dup_s8): Likewise.
1187 (vld4q_dup_p8): Likewise.
1188 (vld4q_dup_s16): Likewise.
1189 (vld4q_dup_p16): Likewise.
1190 (vld4q_dup_s32): Likewise.
1191 (vld4q_dup_s64): Likewise.
1192 (vld4q_dup_u8): Likewise.
1193 (vld4q_dup_u16): Likewise.
1194 (vld4q_dup_u32): Likewise.
1195 (vld4q_dup_u64): Likewise.
1196 (vld4q_dup_f16): Likewise.
1197 (vld4q_dup_f32): Likewise.
1198 (vld4q_dup_f64): Likewise.
1199 (vld4q_dup_p64): Likewise.
1200 (vld2_lane_u8): Likewise.
1201 (vld2_lane_u16): Likewise.
1202 (vld2_lane_u32): Likewise.
1203 (vld2_lane_u64): Likewise.
1204 (vld2_lane_s8): Likewise.
1205 (vld2_lane_s16): Likewise.
1206 (vld2_lane_s32): Likewise.
1207 (vld2_lane_s64): Likewise.
1208 (vld2_lane_f16): Likewise.
1209 (vld2_lane_f32): Likewise.
1210 (vld2_lane_f64): Likewise.
1211 (vld2_lane_p8): Likewise.
1212 (vld2_lane_p16): Likewise.
1213 (vld2_lane_p64): Likewise.
1214 (vld2q_lane_u8): Likewise.
1215 (vld2q_lane_u16): Likewise.
1216 (vld2q_lane_u32): Likewise.
1217 (vld2q_lane_u64): Likewise.
1218 (vld2q_lane_s8): Likewise.
1219 (vld2q_lane_s16): Likewise.
1220 (vld2q_lane_s32): Likewise.
1221 (vld2q_lane_s64): Likewise.
1222 (vld2q_lane_f16): Likewise.
1223 (vld2q_lane_f32): Likewise.
1224 (vld2q_lane_f64): Likewise.
1225 (vld2q_lane_p8): Likewise.
1226 (vld2q_lane_p16): Likewise.
1227 (vld2q_lane_p64): Likewise.
1228 (vld3_lane_u8): Likewise.
1229 (vld3_lane_u16): Likewise.
1230 (vld3_lane_u32): Likewise.
1231 (vld3_lane_u64): Likewise.
1232 (vld3_lane_s8): Likewise.
1233 (vld3_lane_s16): Likewise.
1234 (vld3_lane_s32): Likewise.
1235 (vld3_lane_s64): Likewise.
1236 (vld3_lane_f16): Likewise.
1237 (vld3_lane_f32): Likewise.
1238 (vld3_lane_f64): Likewise.
1239 (vld3_lane_p8): Likewise.
1240 (vld3_lane_p16): Likewise.
1241 (vld3_lane_p64): Likewise.
1242 (vld3q_lane_u8): Likewise.
1243 (vld3q_lane_u16): Likewise.
1244 (vld3q_lane_u32): Likewise.
1245 (vld3q_lane_u64): Likewise.
1246 (vld3q_lane_s8): Likewise.
1247 (vld3q_lane_s16): Likewise.
1248 (vld3q_lane_s32): Likewise.
1249 (vld3q_lane_s64): Likewise.
1250 (vld3q_lane_f16): Likewise.
1251 (vld3q_lane_f32): Likewise.
1252 (vld3q_lane_f64): Likewise.
1253 (vld3q_lane_p8): Likewise.
1254 (vld3q_lane_p16): Likewise.
1255 (vld3q_lane_p64): Likewise.
1256 (vld4_lane_u8): Likewise.
1257 (vld4_lane_u16): Likewise.
1258 (vld4_lane_u32): Likewise.
1259 (vld4_lane_u64): Likewise.
1260 (vld4_lane_s8): Likewise.
1261 (vld4_lane_s16): Likewise.
1262 (vld4_lane_s32): Likewise.
1263 (vld4_lane_s64): Likewise.
1264 (vld4_lane_f16): Likewise.
1265 (vld4_lane_f32): Likewise.
1266 (vld4_lane_f64): Likewise.
1267 (vld4_lane_p8): Likewise.
1268 (vld4_lane_p16): Likewise.
1269 (vld4_lane_p64): Likewise.
1270 (vld4q_lane_u8): Likewise.
1271 (vld4q_lane_u16): Likewise.
1272 (vld4q_lane_u32): Likewise.
1273 (vld4q_lane_u64): Likewise.
1274 (vld4q_lane_s8): Likewise.
1275 (vld4q_lane_s16): Likewise.
1276 (vld4q_lane_s32): Likewise.
1277 (vld4q_lane_s64): Likewise.
1278 (vld4q_lane_f16): Likewise.
1279 (vld4q_lane_f32): Likewise.
1280 (vld4q_lane_f64): Likewise.
1281 (vld4q_lane_p8): Likewise.
1282 (vld4q_lane_p16): Likewise.
1283 (vld4q_lane_p64): Likewise.
1284 (vqtbl2_s8): Likewise.
1285 (vqtbl2_u8): Likewise.
1286 (vqtbl2_p8): Likewise.
1287 (vqtbl2q_s8): Likewise.
1288 (vqtbl2q_u8): Likewise.
1289 (vqtbl2q_p8): Likewise.
1290 (vqtbl3_s8): Likewise.
1291 (vqtbl3_u8): Likewise.
1292 (vqtbl3_p8): Likewise.
1293 (vqtbl3q_s8): Likewise.
1294 (vqtbl3q_u8): Likewise.
1295 (vqtbl3q_p8): Likewise.
1296 (vqtbl4_s8): Likewise.
1297 (vqtbl4_u8): Likewise.
1298 (vqtbl4_p8): Likewise.
1299 (vqtbl4q_s8): Likewise.
1300 (vqtbl4q_u8): Likewise.
1301 (vqtbl4q_p8): Likewise.
1302 (vqtbx2_s8): Likewise.
1303 (vqtbx2_u8): Likewise.
1304 (vqtbx2_p8): Likewise.
1305 (vqtbx2q_s8): Likewise.
1306 (vqtbx2q_u8): Likewise.
1307 (vqtbx2q_p8): Likewise.
1308 (vqtbx3_s8): Likewise.
1309 (vqtbx3_u8): Likewise.
1310 (vqtbx3_p8): Likewise.
1311 (vqtbx3q_s8): Likewise.
1312 (vqtbx3q_u8): Likewise.
1313 (vqtbx3q_p8): Likewise.
1314 (vqtbx4_s8): Likewise.
1315 (vqtbx4_u8): Likewise.
1316 (vqtbx4_p8): Likewise.
1317 (vqtbx4q_s8): Likewise.
1318 (vqtbx4q_u8): Likewise.
1319 (vqtbx4q_p8): Likewise.
1320 (vst1_s64_x2): Likewise.
1321 (vst1_u64_x2): Likewise.
1322 (vst1_f64_x2): Likewise.
1323 (vst1_s8_x2): Likewise.
1324 (vst1_p8_x2): Likewise.
1325 (vst1_s16_x2): Likewise.
1326 (vst1_p16_x2): Likewise.
1327 (vst1_s32_x2): Likewise.
1328 (vst1_u8_x2): Likewise.
1329 (vst1_u16_x2): Likewise.
1330 (vst1_u32_x2): Likewise.
1331 (vst1_f16_x2): Likewise.
1332 (vst1_f32_x2): Likewise.
1333 (vst1_p64_x2): Likewise.
1334 (vst1q_s8_x2): Likewise.
1335 (vst1q_p8_x2): Likewise.
1336 (vst1q_s16_x2): Likewise.
1337 (vst1q_p16_x2): Likewise.
1338 (vst1q_s32_x2): Likewise.
1339 (vst1q_s64_x2): Likewise.
1340 (vst1q_u8_x2): Likewise.
1341 (vst1q_u16_x2): Likewise.
1342 (vst1q_u32_x2): Likewise.
1343 (vst1q_u64_x2): Likewise.
1344 (vst1q_f16_x2): Likewise.
1345 (vst1q_f32_x2): Likewise.
1346 (vst1q_f64_x2): Likewise.
1347 (vst1q_p64_x2): Likewise.
1348 (vst1_s64_x3): Likewise.
1349 (vst1_u64_x3): Likewise.
1350 (vst1_f64_x3): Likewise.
1351 (vst1_s8_x3): Likewise.
1352 (vst1_p8_x3): Likewise.
1353 (vst1_s16_x3): Likewise.
1354 (vst1_p16_x3): Likewise.
1355 (vst1_s32_x3): Likewise.
1356 (vst1_u8_x3): Likewise.
1357 (vst1_u16_x3): Likewise.
1358 (vst1_u32_x3): Likewise.
1359 (vst1_f16_x3): Likewise.
1360 (vst1_f32_x3): Likewise.
1361 (vst1_p64_x3): Likewise.
1362 (vst1q_s8_x3): Likewise.
1363 (vst1q_p8_x3): Likewise.
1364 (vst1q_s16_x3): Likewise.
1365 (vst1q_p16_x3): Likewise.
1366 (vst1q_s32_x3): Likewise.
1367 (vst1q_s64_x3): Likewise.
1368 (vst1q_u8_x3): Likewise.
1369 (vst1q_u16_x3): Likewise.
1370 (vst1q_u32_x3): Likewise.
1371 (vst1q_u64_x3): Likewise.
1372 (vst1q_f16_x3): Likewise.
1373 (vst1q_f32_x3): Likewise.
1374 (vst1q_f64_x3): Likewise.
1375 (vst1q_p64_x3): Likewise.
1376 (vst1_s8_x4): Likewise.
1377 (vst1q_s8_x4): Likewise.
1378 (vst1_s16_x4): Likewise.
1379 (vst1q_s16_x4): Likewise.
1380 (vst1_s32_x4): Likewise.
1381 (vst1q_s32_x4): Likewise.
1382 (vst1_u8_x4): Likewise.
1383 (vst1q_u8_x4): Likewise.
1384 (vst1_u16_x4): Likewise.
1385 (vst1q_u16_x4): Likewise.
1386 (vst1_u32_x4): Likewise.
1387 (vst1q_u32_x4): Likewise.
1388 (vst1_f16_x4): Likewise.
1389 (vst1q_f16_x4): Likewise.
1390 (vst1_f32_x4): Likewise.
1391 (vst1q_f32_x4): Likewise.
1392 (vst1_p8_x4): Likewise.
1393 (vst1q_p8_x4): Likewise.
1394 (vst1_p16_x4): Likewise.
1395 (vst1q_p16_x4): Likewise.
1396 (vst1_s64_x4): Likewise.
1397 (vst1_u64_x4): Likewise.
1398 (vst1_p64_x4): Likewise.
1399 (vst1q_s64_x4): Likewise.
1400 (vst1q_u64_x4): Likewise.
1401 (vst1q_p64_x4): Likewise.
1402 (vst1_f64_x4): Likewise.
1403 (vst1q_f64_x4): Likewise.
1404 (vst2_s64): Likewise.
1405 (vst2_u64): Likewise.
1406 (vst2_f64): Likewise.
1407 (vst2_s8): Likewise.
1408 (vst2_p8): Likewise.
1409 (vst2_s16): Likewise.
1410 (vst2_p16): Likewise.
1411 (vst2_s32): Likewise.
1412 (vst2_u8): Likewise.
1413 (vst2_u16): Likewise.
1414 (vst2_u32): Likewise.
1415 (vst2_f16): Likewise.
1416 (vst2_f32): Likewise.
1417 (vst2_p64): Likewise.
1418 (vst2q_s8): Likewise.
1419 (vst2q_p8): Likewise.
1420 (vst2q_s16): Likewise.
1421 (vst2q_p16): Likewise.
1422 (vst2q_s32): Likewise.
1423 (vst2q_s64): Likewise.
1424 (vst2q_u8): Likewise.
1425 (vst2q_u16): Likewise.
1426 (vst2q_u32): Likewise.
1427 (vst2q_u64): Likewise.
1428 (vst2q_f16): Likewise.
1429 (vst2q_f32): Likewise.
1430 (vst2q_f64): Likewise.
1431 (vst2q_p64): Likewise.
1432 (vst3_s64): Likewise.
1433 (vst3_u64): Likewise.
1434 (vst3_f64): Likewise.
1435 (vst3_s8): Likewise.
1436 (vst3_p8): Likewise.
1437 (vst3_s16): Likewise.
1438 (vst3_p16): Likewise.
1439 (vst3_s32): Likewise.
1440 (vst3_u8): Likewise.
1441 (vst3_u16): Likewise.
1442 (vst3_u32): Likewise.
1443 (vst3_f16): Likewise.
1444 (vst3_f32): Likewise.
1445 (vst3_p64): Likewise.
1446 (vst3q_s8): Likewise.
1447 (vst3q_p8): Likewise.
1448 (vst3q_s16): Likewise.
1449 (vst3q_p16): Likewise.
1450 (vst3q_s32): Likewise.
1451 (vst3q_s64): Likewise.
1452 (vst3q_u8): Likewise.
1453 (vst3q_u16): Likewise.
1454 (vst3q_u32): Likewise.
1455 (vst3q_u64): Likewise.
1456 (vst3q_f16): Likewise.
1457 (vst3q_f32): Likewise.
1458 (vst3q_f64): Likewise.
1459 (vst3q_p64): Likewise.
1460 (vst4_s64): Likewise.
1461 (vst4_u64): Likewise.
1462 (vst4_f64): Likewise.
1463 (vst4_s8): Likewise.
1464 (vst4_p8): Likewise.
1465 (vst4_s16): Likewise.
1466 (vst4_p16): Likewise.
1467 (vst4_s32): Likewise.
1468 (vst4_u8): Likewise.
1469 (vst4_u16): Likewise.
1470 (vst4_u32): Likewise.
1471 (vst4_f16): Likewise.
1472 (vst4_f32): Likewise.
1473 (vst4_p64): Likewise.
1474 (vst4q_s8): Likewise.
1475 (vst4q_p8): Likewise.
1476 (vst4q_s16): Likewise.
1477 (vst4q_p16): Likewise.
1478 (vst4q_s32): Likewise.
1479 (vst4q_s64): Likewise.
1480 (vst4q_u8): Likewise.
1481 (vst4q_u16): Likewise.
1482 (vst4q_u32): Likewise.
1483 (vst4q_u64): Likewise.
1484 (vst4q_f16): Likewise.
1485 (vst4q_f32): Likewise.
1486 (vst4q_f64): Likewise.
1487 (vst4q_p64): Likewise.
1488 (vtbx4_s8): Likewise.
1489 (vtbx4_u8): Likewise.
1490 (vtbx4_p8): Likewise.
1491 (vld1_bf16_x2): Likewise.
1492 (vld1q_bf16_x2): Likewise.
1493 (vld1_bf16_x3): Likewise.
1494 (vld1q_bf16_x3): Likewise.
1495 (vld1_bf16_x4): Likewise.
1496 (vld1q_bf16_x4): Likewise.
1497 (vld2_bf16): Likewise.
1498 (vld2q_bf16): Likewise.
1499 (vld2_dup_bf16): Likewise.
1500 (vld2q_dup_bf16): Likewise.
1501 (vld3_bf16): Likewise.
1502 (vld3q_bf16): Likewise.
1503 (vld3_dup_bf16): Likewise.
1504 (vld3q_dup_bf16): Likewise.
1505 (vld4_bf16): Likewise.
1506 (vld4q_bf16): Likewise.
1507 (vld4_dup_bf16): Likewise.
1508 (vld4q_dup_bf16): Likewise.
1509 (vst1_bf16_x2): Likewise.
1510 (vst1q_bf16_x2): Likewise.
1511 (vst1_bf16_x3): Likewise.
1512 (vst1q_bf16_x3): Likewise.
1513 (vst1_bf16_x4): Likewise.
1514 (vst1q_bf16_x4): Likewise.
1515 (vst2_bf16): Likewise.
1516 (vst2q_bf16): Likewise.
1517 (vst3_bf16): Likewise.
1518 (vst3q_bf16): Likewise.
1519 (vst4_bf16): Likewise.
1520 (vst4q_bf16): Likewise.
1521 (vld2_lane_bf16): Likewise.
1522 (vld2q_lane_bf16): Likewise.
1523 (vld3_lane_bf16): Likewise.
1524 (vld3q_lane_bf16): Likewise.
1525 (vld4_lane_bf16): Likewise.
1526 (vld4q_lane_bf16): Likewise.
1527 (vst2_lane_bf16): Likewise.
1528 (vst2q_lane_bf16): Likewise.
1529 (vst3_lane_bf16): Likewise.
1530 (vst3q_lane_bf16): Likewise.
1531 (vst4_lane_bf16): Likewise.
1532 (vst4q_lane_bf16): Likewise.
1533 * config/aarch64/geniterators.sh: Modify iterator regex to
1534 match new vector-tuple modes.
1535 * config/aarch64/iterators.md (insn_count): Extend mode
1536 attribute with vector-tuple type information.
1538 (Vendreg): Likewise.
1541 (VSTRUCT_2D): New mode iterator.
1542 (VSTRUCT_2DNX): Likewise.
1543 (VSTRUCT_2DX): Likewise.
1544 (VSTRUCT_2Q): Likewise.
1545 (VSTRUCT_2QD): Likewise.
1546 (VSTRUCT_3D): Likewise.
1547 (VSTRUCT_3DNX): Likewise.
1548 (VSTRUCT_3DX): Likewise.
1549 (VSTRUCT_3Q): Likewise.
1550 (VSTRUCT_3QD): Likewise.
1551 (VSTRUCT_4D): Likewise.
1552 (VSTRUCT_4DNX): Likewise.
1553 (VSTRUCT_4DX): Likewise.
1554 (VSTRUCT_4Q): Likewise.
1555 (VSTRUCT_4QD): Likewise.
1556 (VSTRUCT_D): Likewise.
1557 (VSTRUCT_Q): Likewise.
1558 (VSTRUCT_QD): Likewise.
1559 (VSTRUCT_ELT): New mode attribute.
1560 (vstruct_elt): Likewise.
1561 * genmodes.c (VECTOR_MODE): Add default prefix and order
1563 (VECTOR_MODE_WITH_PREFIX): Define.
1564 (make_vector_mode): Add mode prefix and order parameters.
1566 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
1568 * expmed.c (extract_bit_field_1): Ensure modes are tieable.
1570 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
1572 * expr.c (emit_group_load_1): Remove historic workaround.
1574 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
1576 * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
1577 Factor out main loop to...
1578 (aarch64_init_simd_builtin_functions): This new function.
1579 (register_tuple_type): Define.
1580 (aarch64_scalar_builtin_type_p): Define.
1581 (handle_arm_neon_h): Define.
1582 * config/aarch64/aarch64-c.c (aarch64_pragma_aarch64): Handle
1583 pragma for arm_neon.h.
1584 * config/aarch64/aarch64-protos.h (aarch64_advsimd_struct_mode_p):
1586 (handle_arm_neon_h): Likewise.
1587 * config/aarch64/aarch64.c (aarch64_advsimd_struct_mode_p):
1588 Remove static modifier.
1589 * config/aarch64/arm_neon.h (target): Remove Neon vector
1590 structure type definitions.
1592 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
1594 PR tree-optimization/102943
1595 * gimple-range-path.cc (path_range_query::range_on_path_entry):
1596 Prefer range_of_expr unless there are no statements in the BB.
1598 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
1600 PR tree-optimization/102943
1601 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
1602 Avoid duplicate calculation of paths.
1604 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
1606 PR tree-optimization/102943
1607 * gimple-range-path.cc (path_range_query::compute_phi_relations):
1608 Only compute relations for SSA names in the import list.
1609 (path_range_query::compute_outgoing_relations): Same.
1610 * gimple-range-path.h (path_range_query::import_p): New.
1612 2021-11-04 Richard Biener <rguenther@suse.de>
1614 PR rtl-optimization/103075
1615 * simplify-rtx.c (exact_int_to_float_conversion_p): Return
1616 false for a VOIDmode operand.
1618 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
1620 * config/aarch64/aarch64.c (aarch64_vector_costs): Make member
1621 variables private and add "m_" to their names. Remove is_loop.
1622 (aarch64_record_potential_advsimd_unrolling): Replace with...
1623 (aarch64_vector_costs::record_potential_advsimd_unrolling): ...this.
1624 (aarch64_analyze_loop_vinfo): Replace with...
1625 (aarch64_vector_costs::analyze_loop_vinfo): ...this.
1626 Move initialization of (m_)vec_flags to add_stmt_cost.
1627 (aarch64_analyze_bb_vinfo): Delete.
1628 (aarch64_count_ops): Replace with...
1629 (aarch64_vector_costs::count_ops): ...this.
1630 (aarch64_vector_costs::add_stmt_cost): Set m_vec_flags,
1631 using m_costing_for_scalar to test whether we're costing
1632 scalar or vector code.
1633 (aarch64_adjust_body_cost_sve): Replace with...
1634 (aarch64_vector_costs::adjust_body_cost_sve): ...this.
1635 (aarch64_adjust_body_cost): Replace with...
1636 (aarch64_vector_costs::adjust_body_cost): ...this.
1637 (aarch64_vector_costs::finish_cost): Use m_vinfo instead of is_loop.
1639 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
1641 * target.def (targetm.vectorize.init_cost): Replace with...
1642 (targetm.vectorize.create_costs): ...this.
1643 (targetm.vectorize.add_stmt_cost): Delete.
1644 (targetm.vectorize.finish_cost): Likewise.
1645 (targetm.vectorize.destroy_cost_data): Likewise.
1646 * doc/tm.texi.in (TARGET_VECTORIZE_INIT_COST): Replace with...
1647 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
1648 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
1649 (TARGET_VECTORIZE_FINISH_COST): Likewise.
1650 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
1651 * doc/tm.texi: Regenerate.
1652 * tree-vectorizer.h (vec_info::vec_info): Remove target_cost_data
1654 (vec_info::target_cost_data): Change from a void * to a vector_costs *.
1655 (vector_costs): New class.
1656 (init_cost): Take a vec_info and return a vector_costs.
1657 (dump_stmt_cost): Remove data parameter.
1658 (add_stmt_cost): Replace vinfo and data parameters with a vector_costs.
1659 (add_stmt_costs): Likewise.
1660 (finish_cost): Replace data parameter with a vector_costs.
1661 (destroy_cost_data): Delete.
1662 * tree-vectorizer.c (dump_stmt_cost): Remove data argument and
1664 (vec_info::vec_info): Remove the target_cost_data parameter and
1665 initialize the member variable to null instead.
1666 (vec_info::~vec_info): Delete target_cost_data instead of calling
1668 (vector_costs::add_stmt_cost): New function.
1669 (vector_costs::finish_cost): Likewise.
1670 (vector_costs::record_stmt_cost): Likewise.
1671 (vector_costs::adjust_cost_for_freq): Likewise.
1672 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
1673 call to vec_info::vec_info.
1674 (vect_compute_single_scalar_iteration_cost): Update after above
1675 changes to costing interface.
1676 (vect_analyze_loop_operations): Likewise.
1677 (vect_estimate_min_profitable_iters): Likewise.
1678 (vect_analyze_loop_2): Initialize LOOP_VINFO_TARGET_COST_DATA
1679 at the start_over point, where it needs to be recreated after
1680 trying without slp. Update retry code accordingly.
1681 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Update call
1682 to vec_info::vec_info.
1683 (vect_slp_analyze_operation): Update after above changes to costing
1685 (vect_bb_vectorization_profitable_p): Likewise.
1686 * targhooks.h (default_init_cost): Replace with...
1687 (default_vectorize_create_costs): ...this.
1688 (default_add_stmt_cost): Delete.
1689 (default_finish_cost, default_destroy_cost_data): Likewise.
1690 * targhooks.c (default_init_cost): Replace with...
1691 (default_vectorize_create_costs): ...this.
1692 (default_add_stmt_cost): Delete, moving logic to vector_costs instead.
1693 (default_finish_cost, default_destroy_cost_data): Delete.
1694 * config/aarch64/aarch64.c (aarch64_vector_costs): Inherit from
1695 vector_costs. Add a constructor.
1696 (aarch64_init_cost): Replace with...
1697 (aarch64_vectorize_create_costs): ...this.
1698 (aarch64_add_stmt_cost): Replace with...
1699 (aarch64_vector_costs::add_stmt_cost): ...this. Use record_stmt_cost
1700 to adjust the cost for inner loops.
1701 (aarch64_finish_cost): Replace with...
1702 (aarch64_vector_costs::finish_cost): ...this.
1703 (aarch64_destroy_cost_data): Delete.
1704 (TARGET_VECTORIZE_INIT_COST): Replace with...
1705 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
1706 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
1707 (TARGET_VECTORIZE_FINISH_COST): Likewise.
1708 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
1709 * config/i386/i386.c (ix86_vector_costs): New structure.
1710 (ix86_init_cost): Replace with...
1711 (ix86_vectorize_create_costs): ...this.
1712 (ix86_add_stmt_cost): Replace with...
1713 (ix86_vector_costs::add_stmt_cost): ...this. Use adjust_cost_for_freq
1714 to adjust the cost for inner loops.
1715 (ix86_finish_cost, ix86_destroy_cost_data): Delete.
1716 (TARGET_VECTORIZE_INIT_COST): Replace with...
1717 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
1718 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
1719 (TARGET_VECTORIZE_FINISH_COST): Likewise.
1720 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
1721 * config/rs6000/rs6000.c (TARGET_VECTORIZE_INIT_COST): Replace with...
1722 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
1723 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
1724 (TARGET_VECTORIZE_FINISH_COST): Likewise.
1725 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
1726 (rs6000_cost_data): Inherit from vector_costs.
1727 Add a constructor. Drop loop_info, cost and costing_for_scalar
1728 in favor of the corresponding vector_costs member variables.
1729 Add "m_" to the names of the remaining member variables and
1731 (rs6000_density_test): Replace with...
1732 (rs6000_cost_data::density_test): ...this.
1733 (rs6000_init_cost): Replace with...
1734 (rs6000_vectorize_create_costs): ...this.
1735 (rs6000_update_target_cost_per_stmt): Replace with...
1736 (rs6000_cost_data::update_target_cost_per_stmt): ...this.
1737 (rs6000_add_stmt_cost): Replace with...
1738 (rs6000_cost_data::add_stmt_cost): ...this. Use adjust_cost_for_freq
1739 to adjust the cost for inner loops.
1740 (rs6000_adjust_vect_cost_per_loop): Replace with...
1741 (rs6000_cost_data::adjust_vect_cost_per_loop): ...this.
1742 (rs6000_finish_cost): Replace with...
1743 (rs6000_cost_data::finish_cost): ...this. Group loop code
1744 into a single if statement and pass the loop_vinfo down to
1746 (rs6000_destroy_cost_data): Delete.
1748 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
1750 PR tree-optimization/103062
1751 PR tree-optimization/103062
1752 * value-pointer-equiv.cc (ssa_equiv_stack::ssa_equiv_stack):
1753 Increase size of allocation by 1.
1754 (ssa_equiv_stack::push_replacement): Grow as needed.
1755 (ssa_equiv_stack::get_replacement): Same.
1756 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
1757 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Remove delete.
1758 (pointer_equiv_analyzer::set_global_equiv): Grow as needed.
1759 (pointer_equiv_analyzer::get_equiv): Same.
1760 (pointer_equiv_analyzer::get_equiv_expr): Remove const.
1761 * value-pointer-equiv.h (class pointer_equiv_analyzer): Remove
1762 const markers. Use auto_vec instead of tree *.
1764 2021-11-04 Richard Biener <rguenther@suse.de>
1766 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Remove always
1767 true parameter and inline valueization.
1768 (vn_nary_op_lookup_1): Inline valueization from ...
1769 (vn_nary_op_compute_hash): ... here and remove it here.
1770 * tree-ssa-pre.c (phi_translate_1): Do not valueize
1771 before vn_nary_lookup_pieces.
1772 (get_representative_for): Mark created SSA representatives
1775 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
1777 * simplify-rtx.c (simplify_context::simplify_gen_vec_select): Assert
1778 that the operand has a vector mode. Use subreg_lowpart_offset
1779 to test whether an index corresponds to the low part.
1781 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
1783 * read-rtl.c: Remove dead !GENERATOR_FILE block.
1784 * read-rtl-function.c (function_reader::consolidate_singletons):
1785 Generate canonical CONST_VECTORs.
1787 2021-11-04 liuhongt <hongtao.liu@intel.com>
1790 * config/i386/predicates.md (reg_or_notreg_operand): Rename to ..
1791 (regmem_or_bitnot_regmem_operand): .. and extend to handle
1793 * config/i386/sse.md (*<avx512>_vpternlog<mode>_1): Force_reg
1794 the operands which are required to be register_operand.
1795 (*<avx512>_vpternlog<mode>_2): Ditto.
1796 (*<avx512>_vpternlog<mode>_3): Ditto.
1797 (*<avx512>_vternlog<mode>_all): Disallow embeded broadcast for
1798 vector HFmodes since it's not a real AVX512FP16 instruction.
1800 2021-11-04 liuhongt <hongtao.liu@intel.com>
1803 * match.pd: simplify (trunc)copysign((extend)a, (extend)b) to
1804 .COPYSIGN (a,b) when a and b are same type as the truncation
1805 type and has less precision than extend type.
1807 2021-11-04 Richard Biener <rguenther@suse.de>
1809 * doc/generic.texi: Update TARGET_MEM_REF and MEM_REF
1812 2021-11-04 Hongyu Wang <hongyu.wang@intel.com>
1814 * config/i386/sse.md (VI2_AVX512VNNIBW): New mode iterator.
1815 (VI1_AVX512VNNI): Likewise.
1816 (SDOT_VPDP_SUF): New mode_attr.
1819 (sdot_prod<mode>): Use VI2_AVX512F iterator, expand to
1820 vpdpwssd when VNNI targets available.
1821 (usdot_prod<mode>): New expander for vector QImode.
1823 2021-11-04 Hongyu Wang <hongyu.wang@intel.com>
1825 * config/i386/amxtileintrin.h (_tile_loadd_internal): Add
1826 parentheses to base and stride.
1827 (_tile_stream_loadd_internal): Likewise.
1828 (_tile_stored_internal): Likewise.
1830 2021-11-03 Maciej W. Rozycki <macro@embecosm.com>
1832 * config/riscv/riscv.c (riscv_class_max_nregs): Swap the
1833 arguments to `reg_class_subset_p'.
1835 2021-11-03 Joseph Myers <joseph@codesourcery.com>
1838 * fold-const.c (fold_init): New function.
1839 * fold-const.h (fold_init): New prototype.
1841 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
1843 * params.opt (param_vrp2_mode): Make ranger the default for VRP2.
1845 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
1847 * gimple-range-fold.cc (fold_using_range::range_of_phi): Don't import
1848 a range from edge if arg == phidef.
1850 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
1852 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Test
1853 for constant before any other processing.
1855 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
1857 * flag-types.h (RANGER_DEBUG_ALL): Fix values.
1859 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
1861 * gimple-range.cc (gimple_ranger::gimple_ranger): Initialize current_bb.
1862 (gimple_ranger::range_of_expr): Pick up range_on_entry when there is
1863 no explcit context and current_bb is set.
1864 (gimple_ranger::fold_stmt): New.
1865 * gimple-range.h (current_bb, fold_stmt): New.
1866 * tree-vrp.c (rvrp_folder::fold_stmt): Call ranger's fold_stmt.
1868 2021-11-03 Richard Biener <rguenther@suse.de>
1870 PR tree-optimization/102970
1871 * tree-ssa-pre.c (phi_translate_1): Drop clique and base
1872 when translating a MEM_REF over a backedge.
1874 2021-11-03 Philipp Tomsich <philipp.tomsich@vrull.eu>
1876 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New Ampere-1 core.
1877 * config/aarch64/aarch64-tune.md: Regenerate.
1878 * config/aarch64/aarch64-cost-tables.h: Add extra costs for Ampere-1.
1879 * config/aarch64/aarch64.c: Add tuning structures for Ampere-1.
1880 * doc/invoke.texi: Add documentation for Ampere-1 core.
1882 2021-11-03 Wilco Dijkstra <wdijkstr@arm.com>
1884 * config/aarch64/aarch64.md (movsi): Add alternative for GOT accesses.
1886 (ldr_got_small_<mode>): Remove pattern.
1887 (ldr_got_small_sidi): Likewise.
1888 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately): Keep
1889 GOT accesses as moves.
1890 (aarch64_print_operand): Correctly print got_lo12 in L specifier.
1891 (aarch64_mov_operand_p): Make GOT accesses valid move operands.
1892 * config/aarch64/constraints.md: Add new constraint Usw for GOT access.
1894 2021-11-03 Martin Liska <mliska@suse.cz>
1896 * gcov.c (read_line): Remove dead variable.
1898 2021-11-03 Martin Liska <mliska@suse.cz>
1901 * ipa-fnsummary.c (edge_predicate_pool): Rename predicate class to ipa_predicate.
1902 (ipa_fn_summary::account_size_time): Likewise.
1903 (edge_set_predicate): Likewise.
1904 (set_hint_predicate): Likewise.
1905 (add_freqcounting_predicate): Likewise.
1906 (evaluate_conditions_for_known_args): Likewise.
1907 (evaluate_properties_for_edge): Likewise.
1908 (remap_freqcounting_preds_after_dup): Likewise.
1909 (ipa_fn_summary_t::duplicate): Likewise.
1910 (set_cond_stmt_execution_predicate): Likewise.
1911 (set_switch_stmt_execution_predicate): Likewise.
1912 (compute_bb_predicates): Likewise.
1913 (will_be_nonconstant_expr_predicate): Likewise.
1914 (will_be_nonconstant_predicate): Likewise.
1915 (phi_result_unknown_predicate): Likewise.
1916 (predicate_for_phi_result): Likewise.
1917 (analyze_function_body): Likewise.
1918 (compute_fn_summary): Likewise.
1919 (summarize_calls_size_and_time): Likewise.
1920 (estimate_calls_size_and_time): Likewise.
1921 (ipa_call_context::estimate_size_and_time): Likewise.
1922 (remap_edge_summaries): Likewise.
1923 (remap_freqcounting_predicate): Likewise.
1924 (ipa_merge_fn_summary_after_inlining): Likewise.
1925 (ipa_update_overall_fn_summary): Likewise.
1926 (read_ipa_call_summary): Likewise.
1927 (inline_read_section): Likewise.
1928 * ipa-fnsummary.h (struct ipa_freqcounting_predicate): Likewise.
1929 * ipa-predicate.c (predicate::add_clause): Likewise.
1930 (ipa_predicate::add_clause): Likewise.
1931 (predicate::or_with): Likewise.
1932 (ipa_predicate::or_with): Likewise.
1933 (predicate::evaluate): Likewise.
1934 (ipa_predicate::evaluate): Likewise.
1935 (predicate::probability): Likewise.
1936 (ipa_predicate::probability): Likewise.
1937 (dump_condition): Likewise.
1938 (dump_clause): Likewise.
1939 (predicate::dump): Likewise.
1940 (ipa_predicate::dump): Likewise.
1941 (predicate::debug): Likewise.
1942 (ipa_predicate::debug): Likewise.
1943 (predicate::remap_after_duplication): Likewise.
1944 (ipa_predicate::remap_after_duplication): Likewise.
1945 (predicate::remap_after_inlining): Likewise.
1946 (ipa_predicate::remap_after_inlining): Likewise.
1947 (predicate::stream_in): Likewise.
1948 (ipa_predicate::stream_in): Likewise.
1949 (predicate::stream_out): Likewise.
1950 (ipa_predicate::stream_out): Likewise.
1951 (add_condition): Likewise.
1952 * ipa-predicate.h (class predicate): Likewise.
1953 (class ipa_predicate): Likewise.
1954 (add_condition): Likewise.
1956 2021-11-03 Richard Biener <rguenther@suse.de>
1958 * bitmap.h (bitmap_bit_p): Change the return type to bool.
1959 * bitmap.c (bitmap_bit_p): Likewise.
1960 * sbitmap.h (bitmap_bit_p): Likewise.
1961 (bitmap_set_bit): Return whether the bit changed.
1962 (bitmap_clear_bit): Likewise.
1963 * tree-ssa.c (verify_vssa): Make use of the changed state
1964 from bitmap_set_bit.
1966 2021-11-03 Richard Biener <rguenther@suse.de>
1968 PR middle-end/103033
1969 * internal-fn.c (expand_DEFERRED_INIT): Elide the
1970 native_interpret_expr path in favor of folding the
1971 VIEW_CONVERT_EXPR generated when punning the RHS.
1973 2021-11-03 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
1975 * config/s390/s390.c (s390_loop_unroll_adjust): In case of early
1978 2021-11-03 Jan Hubicka <hubicka@ucw.cz>
1981 * ipa-modref.c (callee_to_caller_flags): New function.
1982 (modref_eaf_analysis::analyze_ssa_name): Use it.
1983 (ipa_merge_modref_summary_after_inlining): Fix whitespace.
1985 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
1987 * ipa-modref.c (modref_eaf_analysis::analyze_ssa_name): Revert
1990 2021-11-02 Roger Sayle <roger@nextmovesoftware.com>
1991 Uroš Bizjak <ubizjak@gmail.com>
1993 * config/i386/i386.md (<any_rotate>ti3): Provide expansion for
1994 rotations by non-constant amounts.
1996 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
1998 * ipa-modref.c: Fix anonymous namespace placement.
1999 (class modref_eaf_analysis): New class.
2000 (analyze_ssa_name_flags): Turn to ...
2001 (modref_eaf_analysis::analyze_ssa_name): ... this one.
2002 (merge_call_lhs_flags): Turn to ...
2003 (modref_eaf_analysis::merge_call_lhs_flags): .. this one
2004 (modref_eaf_analysis::merge_with_ssa_name): New member function.
2005 (record_escape_points): Turn to ...
2006 (modref_eaf_analysis::record_escape_points): ... this one.
2007 (analyze_parms): Updat
2008 (ipa_merge_modref_summary_after_inlining): Move to the end of file.
2010 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
2012 * gimple.c (gimple_call_static_chain_flags): New function.
2013 * gimple.h (gimple_call_static_chain_flags): Declare
2014 * ipa-modref.c (modref_summary::modref_summary): Initialize
2016 (modref_summary_lto::modref_summary_lto): Likewise.
2017 (modref_summary::useful_p): Test static_chain_flags.
2018 (modref_summary_lto::useful_p): Likewise.
2019 (struct modref_summary_lto): Add static_chain_flags.
2020 (modref_summary::dump): Dump static_chain_flags.
2021 (modref_summary_lto::dump): Likewise.
2022 (struct escape_point): Add static_cahin_arg.
2023 (analyze_ssa_name_flags): Use gimple_call_static_chain_flags.
2024 (analyze_parms): Handle static chains.
2025 (modref_summaries::duplicate): Duplicate static_chain_flags.
2026 (modref_summaries_lto::duplicate): Likewise.
2027 (modref_write): Stream static_chain_flags.
2028 (read_section): Likewise.
2029 (modref_merge_call_site_flags): Handle static_chain_flags.
2030 * ipa-modref.h (struct modref_summary): Add static_chain_flags.
2031 * tree-ssa-structalias.c (handle_rhs_call): Use
2032 gimple_static_chain_flags.
2034 2021-11-02 Richard Biener <rguenther@suse.de>
2036 PR tree-optimization/103029
2037 * tree-vect-loop-manip.c (vect_loop_versioning): Ensure
2038 the PHI nodes in the loop maintain their original operand
2041 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
2043 * tree-core.h (EAF_NOT_RETURNED_DIRECTLY): New flag.
2044 (EAF_NOREAD): Renumber.
2045 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOT_RETURNED_DIRECTLY.
2046 (remove_useless_eaf_flags): Handle EAF_NOT_RETURNED_DIRECTLY
2047 (deref_flags): Likewise.
2048 (modref_lattice::init): Likewise.
2049 (modref_lattice::merge): Likewise.
2050 (merge_call_lhs_flags): Likewise.
2051 (analyze_ssa_name_flags): Likewise.
2052 (modref_merge_call_site_flags): Likewise.
2053 * tree-ssa-structalias.c (handle_call_arg): Likewise.
2055 2021-11-02 Maciej W. Rozycki <macro@embecosm.com>
2057 * config/riscv/riscv.c (riscv_rtx_costs): Correct a CONST_INT_P
2058 check and remove an unused local variable with shNadd/shNadd.uw
2061 2021-11-02 Tamar Christina <tamar.christina@arm.com>
2063 PR tree-optimization/103007
2064 * tree-vect-slp-patterns.c (complex_fms_pattern::matches): Add elem
2067 2021-11-02 Richard Biener <rguenther@suse.de>
2069 PR middle-end/103038
2070 * fold-const.c (native_interpret_expr): Handle OFFSET_TYPE.
2071 (can_native_interpret_type_p): Likewise.
2072 * internal-fn.c (expand_DEFERRED_INIT): View-convert the
2073 RHS if the LHS is an SSA name.
2075 2021-11-02 Richard Sandiford <richard.sandiford@arm.com>
2077 * langhooks.h (lang_hooks_for_types::simulate_record_decl): New hook.
2078 * langhooks-def.h (lhd_simulate_record_decl): Declare.
2079 (LANG_HOOKS_SIMULATE_RECORD_DECL): Define.
2080 (LANG_HOOKS_FOR_TYPES_INITIALIZER): Include it.
2081 * langhooks.c (lhd_simulate_record_decl): New function.
2083 2021-11-02 Jakub Jelinek <jakub@redhat.com>
2086 * config/i386/i386.c (ix86_vector_mode_supported_p): Reject vector
2087 modes with TImode inner mode if 32-bit.
2089 2021-11-02 liuhongt <hongtao.liu@intel.com>
2091 * doc/sourcebuild.texi (vect_slp_v4qi_store_unalign,
2092 vect_slp_v2hi_store_unalign, vect_slp_v4hi_store_unalign,
2093 vect_slp_v4si_store_unalign): Document efficient target.
2094 (vect_slp_v4qi_store_unalign_1, vect_slp_v8qi_store_unalign_1,
2095 vect_slp_v16qi_store_unalign_1): Ditto.
2096 (vect_slp_v2hi_store_align,vect_slp_v2qi_store_align,
2097 vect_slp_v2si_store_align, vect_slp_v4qi_store_align): Ditto.
2098 (struct_4char_block_move, struct_8char_block_move,
2099 struct_16char_block_move): Ditto.
2101 2021-11-02 Roger Sayle <roger@nextmovesoftware.com>
2102 Jakub Jelinek <jakub@redhat.com>
2105 * config/i386/i386-expand.c (ix86_expand_v1ti_to_ti,
2106 ix86_expand_ti_to_v1ti): New helper functions.
2107 (ix86_expand_v1ti_shift): Check if the amount operand is an
2108 integer constant, and expand as a TImode shift if it isn't.
2109 (ix86_expand_v1ti_rotate): Check if the amount operand is an
2110 integer constant, and expand as a TImode rotate if it isn't.
2111 (ix86_expand_v1ti_ashiftrt): New function to expand arithmetic
2112 right shifts of V1TImode quantities.
2113 * config/i386/i386-protos.h (ix86_expand_v1ti_ashift): Prototype.
2114 * config/i386/sse.md (ashlv1ti3, lshrv1ti3): Change constraints
2115 to QImode general_operand, and let the helper functions lower
2116 shifts by non-constant operands, as TImode shifts. Make
2117 conditional on TARGET_64BIT.
2118 (ashrv1ti3): New expander calling ix86_expand_v1ti_ashiftrt.
2119 (rotlv1ti3, rotrv1ti3): Change shift operand to QImode.
2120 Make conditional on TARGET_64BIT.
2122 2021-11-02 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
2124 * config/s390/s390.md ("*cc_to_int", "tabort", "*tabort_1",
2125 "*tabort_1_plus"): Remove operands non-null check.
2127 2021-11-01 Martin Liska <mliska@suse.cz>
2129 * opt-functions.awk: Add new sanity checking.
2130 * optc-gen.awk: Add new argument to integer_range_info.
2131 * params.opt: Update 2 params which have negative IntegerRange.
2133 2021-11-01 qing zhao <qing.zhao@oracle.com>
2135 * gimplify.c (gimplify_decl_expr): Do not add call to
2136 __builtin_clear_padding when a variable is a gimple register
2137 or it might not have padding.
2138 (gimplify_init_constructor): Likewise.
2140 2021-11-01 Tamar Christina <tamar.christina@arm.com>
2142 * config/arm/aarch-common-protos.h (struct vector_cost_table): Add
2143 movi, dup and extract costing fields.
2144 * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs,
2145 thunderx_extra_costs, thunderx2t99_extra_costs,
2146 thunderx3t110_extra_costs, tsv110_extra_costs, a64fx_extra_costs): Use
2148 * config/arm/aarch-cost-tables.h (generic_extra_costs,
2149 cortexa53_extra_costs, cortexa57_extra_costs, cortexa76_extra_costs,
2150 exynosm1_extra_costs, xgene1_extra_costs): Likewise
2151 * config/aarch64/aarch64-simd.md (aarch64_simd_dup<mode>): Add r->w dup.
2152 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add extra costs.
2154 2021-11-01 Tamar Christina <tamar.christina@arm.com>
2156 * cse.c (add_to_set): New.
2157 (find_sets_in_insn): Register constants in sets.
2158 (canonicalize_insn): Use auto_vec instead.
2159 (cse_insn): Try materializing using vec_dup.
2160 * rtl.h (simplify_context::simplify_gen_vec_select,
2161 simplify_gen_vec_select): New.
2162 * simplify-rtx.c (simplify_context::simplify_gen_vec_select): New.
2164 2021-11-01 David Malcolm <dmalcolm@redhat.com>
2166 * common.opt (fdiagnostics-escape-format=): New.
2167 (diagnostics_escape_format): New enum.
2168 (DIAGNOSTICS_ESCAPE_FORMAT_UNICODE): New enum value.
2169 (DIAGNOSTICS_ESCAPE_FORMAT_BYTES): Likewise.
2170 * diagnostic-format-json.cc (json_end_diagnostic): Add
2171 "escape-source" attribute.
2172 * diagnostic-show-locus.c
2173 (exploc_with_display_col::exploc_with_display_col): Replace
2174 "tabstop" param with a cpp_char_column_policy and add an "aspect"
2175 param. Use these to compute m_display_col accordingly.
2176 (struct char_display_policy): New struct.
2177 (layout::m_policy): New field.
2178 (layout::m_escape_on_output): New field.
2179 (def_policy): New function.
2180 (make_range): Update for changes to exploc_with_display_col ctor.
2181 (default_print_decoded_ch): New.
2182 (width_per_escaped_byte): New.
2183 (escape_as_bytes_width): New.
2184 (escape_as_bytes_print): New.
2185 (escape_as_unicode_width): New.
2186 (escape_as_unicode_print): New.
2188 (layout::layout): Initialize new fields. Update m_exploc ctor
2189 call for above change to ctor.
2190 (layout::maybe_add_location_range): Update for changes to
2191 exploc_with_display_col ctor.
2192 (layout::calculate_x_offset_display): Update for change to
2194 (layout::print_source_line): Pass policy
2195 to cpp_display_width_computation. Capture cpp_decoded_char when
2196 calling process_next_codepoint. Move printing of source code to
2197 m_policy.m_print_cb.
2198 (line_label::line_label): Pass in policy rather than context.
2199 (layout::print_any_labels): Update for change to line_label ctor.
2200 (get_affected_range): Pass in policy rather than context, updating
2201 calls to location_compute_display_column accordingly.
2202 (get_printed_columns): Likewise, also for cpp_display_width.
2203 (correction::correction): Pass in policy rather than tabstop.
2204 (correction::compute_display_cols): Pass m_policy rather than
2205 m_tabstop to cpp_display_width.
2206 (correction::m_tabstop): Replace with...
2207 (correction::m_policy): ...this.
2208 (line_corrections::line_corrections): Pass in policy rather than
2210 (line_corrections::m_context): Replace with...
2211 (line_corrections::m_policy): ...this.
2212 (line_corrections::add_hint): Update to use m_policy rather than
2214 (line_corrections::add_hint): Likewise.
2215 (layout::print_trailing_fixits): Likewise.
2216 (selftest::test_display_widths): New.
2217 (selftest::test_layout_x_offset_display_utf8): Update to use
2218 policy rather than tabstop.
2219 (selftest::test_one_liner_labels_utf8): Add test of escaping
2221 (selftest::test_diagnostic_show_locus_one_liner_utf8): Update to
2222 use policy rather than tabstop.
2223 (selftest::test_overlapped_fixit_printing): Likewise.
2224 (selftest::test_overlapped_fixit_printing_utf8): Likewise.
2225 (selftest::test_overlapped_fixit_printing_2): Likewise.
2226 (selftest::test_tab_expansion): Likewise.
2227 (selftest::test_escaping_bytes_1): New.
2228 (selftest::test_escaping_bytes_2): New.
2229 (selftest::diagnostic_show_locus_c_tests): Call the new tests.
2230 * diagnostic.c (diagnostic_initialize): Initialize
2231 context->escape_format.
2232 (convert_column_unit): Update to use default character width policy.
2233 (selftest::test_diagnostic_get_location_text): Likewise.
2234 * diagnostic.h (enum diagnostics_escape_format): New enum.
2235 (diagnostic_context::escape_format): New field.
2236 * doc/invoke.texi (-fdiagnostics-escape-format=): New option.
2237 (-fdiagnostics-format=): Add "escape-source" attribute to examples
2238 of JSON output, and document it.
2239 * input.c (location_compute_display_column): Pass in "policy"
2240 rather than "tabstop", passing to
2241 cpp_byte_column_to_display_column.
2242 (selftest::test_cpp_utf8): Update to use cpp_char_column_policy.
2243 * input.h (class cpp_char_column_policy): New forward decl.
2244 (location_compute_display_column): Pass in "policy" rather than
2246 * opts.c (common_handle_option): Handle
2247 OPT_fdiagnostics_escape_format_.
2248 * selftest.c (temp_source_file::temp_source_file): New ctor
2249 overload taking a size_t.
2250 * selftest.h (temp_source_file::temp_source_file): Likewise.
2252 2021-11-01 Aldy Hernandez <aldyh@redhat.com>
2254 * dbgcnt.def: Add debug counter for back_thread[12] and
2255 back_threadfull[12].
2256 * passes.def: Pass "first" argument to each back threading pass.
2257 * tree-ssa-threadbackward.c (back_threader::back_threader): Add
2259 (back_threader::debug_counter): New.
2260 (back_threader::maybe_register_path): Call debug_counter.
2262 2021-11-01 Aldy Hernandez <aldyh@redhat.com>
2264 * tree-ssa-threadbackward.c (BT_NONE): New.
2267 (back_threader::back_threader): Add flags.
2268 Move loop initialization here.
2269 (back_threader::~back_threader): New.
2270 (back_threader::find_taken_edge_switch): Change solver and ranger
2272 (back_threader::find_taken_edge_cond): Same.
2273 (back_threader::find_paths_to_names): Same.
2274 (back_threader::find_paths): Same.
2275 (back_threader::dump): Same.
2276 (try_thread_blocks): Merge into thread_blocks.
2277 (back_threader::thread_blocks): New.
2278 (do_early_thread_jumps): Merge into thread_blocks.
2279 (do_thread_jumps): Merge into thread_blocks.
2280 (back_threader::thread_through_all_blocks): Remove.
2282 2021-11-01 Andrew MacLeod <amacleod@redhat.com>
2284 PR tree-optimization/103003
2285 * value-relation.cc (dom_oracle::register_relation): If the 2
2286 ssa names are the same, don't register any relation.
2288 2021-11-01 Dan Li <ashimida@linux.alibaba.com>
2290 * config/aarch64/aarch64.c (aarch64_expand_epilogue): Remove
2291 redundant check for calls_eh_return.
2292 * config/aarch64/aarch64.md (*do_return): Likewise.
2294 2021-11-01 Xionghu Luo <luoxhu@linux.ibm.com>
2296 * cfghooks.c (cfg_hook_duplicate_loop_to_header_edge): Rename
2297 duplicate_loop_to_header_edge to
2298 duplicate_loop_body_to_header_edge.
2299 (cfg_hook_duplicate_loop_body_to_header_edge): Likewise.
2300 * cfghooks.h (struct cfg_hooks): Likewise.
2301 (cfg_hook_duplicate_loop_body_to_header_edge): Likewise.
2302 * cfgloopmanip.c (duplicate_loop_body_to_header_edge): Likewise.
2303 (clone_loop_to_header_edge): Likewise.
2304 * cfgloopmanip.h (duplicate_loop_body_to_header_edge): Likewise.
2305 * cfgrtl.c (struct cfg_hooks): Likewise.
2306 * doc/loop.texi: Likewise.
2307 * loop-unroll.c (unroll_loop_constant_iterations): Likewise.
2308 (unroll_loop_runtime_iterations): Likewise.
2309 (unroll_loop_stupid): Likewise.
2310 (apply_opt_in_copies): Likewise.
2311 * tree-cfg.c (struct cfg_hooks): Likewise.
2312 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
2313 (try_peel_loop): Likewise.
2314 * tree-ssa-loop-manip.c (copy_phi_node_args): Likewise.
2315 (gimple_duplicate_loop_body_to_header_edge): Likewise.
2316 (tree_transform_and_unroll_loop): Likewise.
2317 * tree-ssa-loop-manip.h (gimple_duplicate_loop_body_to_header_edge):
2320 2021-11-01 Xionghu Luo <luoxhu@linux.ibm.com>
2322 * cfgloopmanip.c (loop_version): Refactor loopify to
2323 loop_version. Move condition generation after loopify.
2325 * cfgloopmanip.h (loopify): Delete.
2327 2021-10-31 Jan Hubicka <hubicka@ucw.cz>
2329 * ipa-fnsummary.c: Include tree-dfa.h.
2330 (points_to_local_or_readonly_memory_p): Return true on return
2332 * ipa-modref.c (analyze_ssa_name_flags): Fix handling of copy
2335 2021-10-30 Tobias Burnus <tobias@codesourcery.com>
2337 PR middle-end/102972
2338 * omp-low.c (omp_runtime_api_call): Use DECL_ASSEMBLER_NAME to get
2339 internal Fortran name; new permit_num_teams arg to permit
2340 omp_get_num_teams and omp_get_team_num.
2341 (scan_omp_1_stmt): Update call to it, add missing call for
2342 reverse offload, and check for strictly nested API calls in teams.
2344 2021-10-30 Jakub Jelinek <jakub@redhat.com>
2346 * gimplify.c (gimplify_omp_for): Diagnose threadprivate iterators.
2348 2021-10-29 David Malcolm <dmalcolm@redhat.com>
2350 * selftest.c (assert_streq): Add newlines when emitting non-equal
2353 2021-10-29 David Malcolm <dmalcolm@redhat.com>
2355 * Makefile.in: Fix syntax for reference to LIBDEPS in
2358 2021-10-29 Bernhard Reutner-Fischer <aldot@gcc.gnu.org>
2360 * doc/install.texi: Bump required minimum DejaGnu version.
2362 2021-10-29 Aldy Hernandez <aldyh@redhat.com>
2363 Andrew MacLeod <amacleod@redhat.com>
2365 * value-relation.cc (path_oracle::killing_def): Add a
2366 self-equivalence so we don't look to the root oracle.
2368 2021-10-29 Aldy Hernandez <aldyh@redhat.com>
2370 * passes.def: Replace the pass_thread_jumps before VRP* with
2371 pass_thread_jumps_full. Remove all pass_vrp_threader instances.
2372 * tree-ssa-threadbackward.c (pass_data_thread_jumps_full):
2373 Remove hyphen from "thread-full" name.
2375 2021-10-29 Andrew MacLeod <amacleod@redhat.com>
2377 PR tree-optimization/102983
2378 * gimple-range-cache.h (propagate_updated_value): Make public.
2379 * gimple-range.cc (gimple_ranger::range_of_stmt): Propagate exports
2380 when processing gcond stmts.
2382 2021-10-29 Jan Hubicka <hubicka@ucw.cz>
2384 * gimple.c (gimple_call_retslot_flags): New function.
2385 * gimple.h (gimple_call_retslot_flags): Declare.
2386 * ipa-modref.c: Include tree-cfg.h.
2387 (struct escape_entry): Turn parm_index to signed.
2388 (modref_summary_lto::modref_summary_lto): Add retslot_flags.
2389 (modref_summary::modref_summary): Initialize retslot_flags.
2390 (struct modref_summary_lto): Likewise.
2391 (modref_summary::useful_p): Check retslot_flags.
2392 (modref_summary_lto::useful_p): Likewise.
2393 (modref_summary::dump): Dump retslot_flags.
2394 (modref_summary_lto::dump): Likewise.
2395 (struct escape_point): Add hidden_args enum.
2396 (analyze_ssa_name_flags): Ignore return slot return;
2397 use gimple_call_retslot_flags.
2398 (record_escape_points): Break out from ...
2399 (analyze_parms): ... here; handle retslot_flags.
2400 (modref_summaries::duplicate): Duplicate retslot_flags.
2401 (modref_summaries_lto::duplicate): Likewise.
2402 (modref_write_escape_summary): Stream parm_index as signed.
2403 (modref_read_escape_summary): Likewise.
2404 (modref_write): Stream retslot_flags.
2405 (read_section): Likewise.
2406 (struct escape_map): Fix typo in comment.
2407 (update_escape_summary_1): Fix whitespace.
2408 (ipa_merge_modref_summary_after_inlining): Drop retslot_flags.
2409 (modref_merge_call_site_flags): Merge retslot_flags.
2410 * ipa-modref.h (struct modref_summary): Add retslot_flags.
2411 * tree-ssa-structalias.c (handle_rhs_call): Handle retslot_flags.
2413 2021-10-29 Tamar Christina <tamar.christina@arm.com>
2415 PR tree-optimization/102977
2416 * tree-vect-slp-patterns.c (vect_match_call_p): Remove.
2417 (vect_detect_pair_op): Add crosslane check.
2418 (vect_match_call_complex_mla): Remove.
2419 (class complex_mul_pattern): Update comment.
2420 (complex_mul_pattern::matches): Update detection.
2421 (class complex_fma_pattern): Remove.
2422 (complex_fma_pattern::matches): Remove.
2423 (complex_fma_pattern::recognize): Remove.
2424 (complex_fma_pattern::build): Remove.
2425 (class complex_fms_pattern): Update comment.
2426 (complex_fms_pattern::matches): Remove.
2427 (complex_operations_pattern::recognize): Remove complex_fma_pattern
2429 2021-10-29 Jakub Jelinek <jakub@redhat.com>
2431 * gimple-fold.c (gimple_fold_builtin_memset): Copy over location from
2434 2021-10-29 Xionghu Luo <luoxhu@linux.ibm.com>
2437 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Add
2438 patterns match and emit for VSX xxpermdi.
2440 2021-10-29 liuhongt <hongtao.liu@intel.com>
2443 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF): New
2445 (V16HF_FTYPE_V16HF): Ditto.
2446 (V32HF_FTYPE_V32HF): Ditto.
2447 (V8HF_FTYPE_V8HF_ROUND): Ditto.
2448 (V16HF_FTYPE_V16HF_ROUND): Ditto.
2449 (V32HF_FTYPE_V32HF_ROUND): Ditto.
2450 * config/i386/i386-builtin.def ( IX86_BUILTIN_FLOORPH,
2451 IX86_BUILTIN_CEILPH, IX86_BUILTIN_TRUNCPH,
2452 IX86_BUILTIN_FLOORPH256, IX86_BUILTIN_CEILPH256,
2453 IX86_BUILTIN_TRUNCPH256, IX86_BUILTIN_FLOORPH512,
2454 IX86_BUILTIN_CEILPH512, IX86_BUILTIN_TRUNCPH512): New builtin.
2455 * config/i386/i386-builtins.c
2456 (ix86_builtin_vectorized_function): Enable vectorization for
2457 HFmode FLOOR/CEIL/TRUNC operation.
2458 * config/i386/i386-expand.c (ix86_expand_args_builtin): Handle
2460 * config/i386/sse.md (rint<mode>2, nearbyint<mode>2): Extend
2463 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
2464 Andrew MacLeod <amacleod@redhat.com>
2466 * value-relation.cc (path_oracle::killing_def): Walk the
2467 equivalency list and remove SSA from any equivalencies.
2469 2021-10-28 Stafford Horne <shorne@gmail.com>
2471 * config/or1k/or1k.h (PROFILE_HOOK): Add return address argument
2474 2021-10-28 Jakub Jelinek <jakub@redhat.com>
2476 PR tree-optimization/102951
2477 * fold-const.h (address_compare): Declare.
2478 * fold-const.c (address_compare): New function.
2479 * match.pd (cmp (convert1?@2 addr@0) (convert2? addr@1)): Use
2480 address_compare helper.
2481 (minmax cmp (convert1?@2 addr@0) (convert2?@3 addr@1)): New
2484 2021-10-28 Andrew MacLeod <amacleod@redhat.com>
2486 * vr-values.c (simplify_using_ranges::fold_cond): Change fold message.
2488 2021-10-28 Andrew MacLeod <amacleod@redhat.com>
2490 PR tree-optimization/102940
2491 * tree-vrp.c (execute_ranger_vrp): Reset scev.
2493 2021-10-28 Richard Purdie <richard.purdie@linuxfoundation.org>
2495 * config/nios2/linux.h (MUSL_DYNAMIC_LINKER): Add musl linker
2497 2021-10-28 Richard Purdie <richard.purdie@linuxfoundation.org>
2499 * configure: Regenerate.
2500 * configure.ac: Use CPPFLAGS_FOR_BUILD for GMPINC
2502 2021-10-28 Eric Botcazou <ebotcazou@adacore.com>
2504 * doc/invoke.texi (%X): Remove obsolete reference to -Wl.
2506 2021-10-28 Richard Biener <rguenther@suse.de>
2509 * fold-const.c (fold_convert_const): Avoid int to float
2510 constant folding with -frounding-math and inexact result.
2511 * simplify-rtx.c (simplify_const_unary_operation): Likewise
2512 for both float and unsigned_float.
2514 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
2516 * tree-ssa-threadbackward.c
2517 (back_threader::find_taken_edge_switch): Use find_case_label_range
2518 instead of find_taken_edge.
2520 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
2522 * tree-ssa-threadbackward.c (class back_threader_registry):
2523 Inherit from back_jt_path_registry.
2524 (back_threader_registry::thread_through_all_blocks): Remove.
2525 (back_threader_registry::register_path): Remove
2526 m_lowlevel_registry prefix.
2528 2021-10-28 Richard Biener <rguenther@suse.de>
2531 * fold-const.c (fold_convert_const_real_from_real): Honor
2532 -frounding-math if the conversion is not exact.
2533 * simplify-rtx.c (simplify_const_unary_operation): Do not
2534 simplify FLOAT_TRUNCATE with sign dependent rounding.
2536 2021-10-28 Richard Biener <rguenther@suse.de>
2538 PR tree-optimization/102949
2539 * tree-vect-stmts.c (ensure_base_align): Look at the
2540 dr_info of a group leader and assert we are looking at
2541 one with analyzed alignment.
2543 2021-10-28 Kewen Lin <linkw@linux.ibm.com>
2546 * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Consider
2547 V1T1 mode for unaligned load and store.
2549 2021-10-28 Kito Cheng <kito.cheng@sifive.com>
2551 * config/riscv/riscv.md (zero_extendsidi2_internal): Allow ZBB
2554 2021-10-28 Kito Cheng <kito.cheng@sifive.com>
2556 * config/riscv/arch-canonicalize (CANONICAL_ORDER): Add `i` to
2559 2021-10-28 Alexandre Oliva <oliva@adacore.com>
2561 * common.opt (fharden-compares): New.
2562 (fharden-conditional-branches): New.
2563 * doc/invoke.texi: Document new options.
2564 * gimple-harden-conditionals.cc: New.
2565 * Makefile.in (OBJS): Build it.
2566 * passes.def: Add new passes.
2567 * tree-pass.h (make_pass_harden_compares): Declare.
2568 (make_pass_harden_conditional_branches): Declare.
2570 2021-10-28 Xionghu Luo <luoxhu@linux.ibm.com>
2573 * config/rs6000/altivec.md: Add vsx register constraints.
2574 * config/rs6000/vsx.md (vsx_xxsel<mode>): Delete.
2575 (vsx_xxsel<mode>2): Likewise.
2576 (vsx_xxsel<mode>3): Likewise.
2577 (vsx_xxsel<mode>4): Likewise.
2579 2021-10-28 Xionghu Luo <luoxhu@linux.ibm.com>
2582 * config/rs6000/altivec.md (*altivec_vsel<mode>): Change to ...
2583 (altivec_vsel<mode>): ... this and update define.
2584 (*altivec_vsel<mode>_uns): Delete.
2585 (altivec_vsel<mode>2): New define_insn.
2586 (altivec_vsel<mode>3): Likewise.
2587 (altivec_vsel<mode>4): Likewise.
2588 * config/rs6000/rs6000-call.c (altivec_expand_vec_sel_builtin): New.
2589 (altivec_expand_builtin): Call altivec_expand_vec_sel_builtin to expand
2591 * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Use bit-wise
2592 selection instead of per element.
2593 * config/rs6000/vector.md:
2594 * config/rs6000/vsx.md (*vsx_xxsel<mode>): Change to ...
2595 (vsx_xxsel<mode>): ... this and update define.
2596 (*vsx_xxsel<mode>_uns): Delete.
2597 (vsx_xxsel<mode>2): New define_insn.
2598 (vsx_xxsel<mode>3): Likewise.
2599 (vsx_xxsel<mode>4): Likewise.
2601 2021-10-28 Hongyu Wang <hongyu.wang@intel.com>
2603 * config/i386/i386.c (use_rsqrt_p): Add mode parameter, enable
2604 HFmode rsqrt without TARGET_SSE_MATH.
2605 (ix86_optab_supported_p): Refactor rint, adjust floor, ceil,
2606 btrunc condition to be restricted by -ftrapping-math, adjust
2607 use_rsqrt_p function call.
2608 * config/i386/i386.md (rcphf2): New define_insn.
2609 (rsqrthf2): Likewise.
2610 * config/i386/sse.md (div<mode>3): Change VF2H to VF2.
2611 (div<mode>3): New expander for HF mode.
2612 (rsqrt<mode>2): Likewise.
2613 (*avx512fp16_vmrcpv8hf2): New define_insn for rpad pass.
2614 (*avx512fp16_vmrsqrtv8hf2): Likewise.
2616 2021-10-27 Saagar Jha <saagar@saagarjha.com>
2618 * config.gcc: Adjust for Darwin21.
2619 * config/darwin-c.c (macosx_version_as_macro): Likewise.
2620 * config/darwin-driver.c (validate_macosx_version_min):
2622 (darwin_find_version_from_kernel): Likewise.
2624 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
2626 * gimple-range-path.cc
2627 (path_range_query::range_defined_in_block): Call killing_def.
2629 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
2631 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Dump
2632 operands as well as relation.
2633 * gimple-range-path.cc
2634 (path_range_query::compute_ranges_in_block): Compute PHI relations
2635 first. Compute outgoing relations at the end.
2636 (path_range_query::compute_ranges): Remove call to compute_relations.
2637 (path_range_query::compute_relations): Remove.
2638 (path_range_query::maybe_register_phi_relation): New.
2639 (path_range_query::compute_phi_relations): Abstract out
2640 registering one PHI relation to...
2641 (path_range_query::compute_outgoing_relations): ...here.
2642 * gimple-range-path.h (class path_range_query): Remove
2644 Add maybe_register_phi_relation.
2646 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
2647 Andrew MacLeod <amacleod@redhat.com>
2649 * value-relation.cc (path_oracle::killing_def): Kill second
2652 2021-10-27 Martin Jambor <mjambor@suse.cz>
2654 * ipa-cp.c (good_cloning_opportunity_p): Decide whether to use
2655 profile feedback depending on their local availability.
2657 2021-10-27 Martin Jambor <mjambor@suse.cz>
2659 * params.opt (param_ipa_cp_profile_count_base): New parameter.
2660 * doc/invoke.texi (Optimize Options): Add entry for
2661 ipa-cp-profile-count-base.
2662 * ipa-cp.c (max_count): Replace with base_count, replace all
2663 occurrences too, unless otherwise stated.
2664 (ipcp_cloning_candidate_p): identify mostly-directly called
2665 functions based on their counts, not max_count.
2666 (compare_edge_profile_counts): New function.
2667 (ipcp_propagate_stage): Instead of setting max_count, find the
2668 appropriate edge count in a sorted vector of counts of eligible
2669 edges and make it the base_count.
2671 2021-10-27 Martin Jambor <mjambor@suse.cz>
2673 * ipa-cp.c (struct caller_statistics): New fields rec_count_sum,
2674 n_nonrec_calls and itself, document all fields.
2675 (init_caller_stats): Initialize the above new fields.
2676 (gather_caller_stats): Gather self-recursive counts and calls number.
2677 (get_info_about_necessary_edges): Gather counts of self-recursive and
2678 other edges bringing in the requested value separately.
2679 (dump_profile_updates): Rework to dump info about a single node only.
2680 (lenient_count_portion_handling): New function.
2681 (struct gather_other_count_struct): New type.
2682 (gather_count_of_non_rec_edges): New function.
2683 (struct desc_incoming_count_struct): New type.
2684 (analyze_clone_icoming_counts): New function.
2685 (adjust_clone_incoming_counts): Likewise.
2686 (update_counts_for_self_gen_clones): Likewise.
2687 (update_profiling_info): Rewritten.
2688 (update_specialized_profile): Adjust call to dump_profile_updates.
2689 (create_specialized_node): Do not update profiling info.
2690 (decide_about_value): New parameter self_gen_clones, either push new
2691 clones into it or updat their profile counts. For self-recursively
2692 generated values, use a portion of the node count instead of count
2693 from self-recursive edges to estimate goodness.
2694 (decide_whether_version_node): Gather clones for self-generated values
2695 in a new vector, update their profiles at once at the end.
2697 2021-10-27 Richard Biener <rguenther@suse.de>
2699 * tree-vectorizer.c (vect_transform_loops): New function,
2701 (try_vectorize_loop_1): ... here. Simplify as epilogues
2702 are now fully handled in the split part.
2704 2021-10-27 Jakub Jelinek <jakub@redhat.com>
2706 * omp-expand.c (expand_omp_for_init_counts): Handle non-rectangular
2707 iterators with pointer types.
2708 (expand_omp_for_init_vars, extract_omp_for_update_vars): Likewise.
2710 2021-10-26 Martin Sebor <msebor@redhat.com>
2712 PR tree-optimization/102238
2713 PR tree-optimization/102919
2714 * gimple-ssa-sprintf.c (get_string_length): Add an argument.
2715 (array_elt_at_offset): Move to pointer-query.
2716 (set_aggregate_size_and_offset): New function.
2717 (field_at_offset): Move to pointer-query.
2718 (get_origin_and_offset): Rename...
2719 (get_origin_and_offset_r): this. Add an argument. Make aggregate
2720 handling more robust.
2721 (get_origin_and_offset): New.
2722 (alias_offset): Add an argument.
2723 (format_string): Use subobject size determined by get_origin_and_offset.
2724 * pointer-query.cc (field_at_offset): Move from gimple-ssa-sprintf.c.
2725 Improve/correct handling of aggregates.
2726 (array_elt_at_offset): Same.
2727 * pointer-query.h (field_at_offset): Declare.
2728 (array_elt_at_offset): Declare.
2730 2021-10-26 Martin Sebor <msebor@redhat.com>
2732 * builtins.c (check_strncat_sizes): Pass access_data ctor additional
2734 (expand_builtin_memcmp): Move code to gimple-ssa-warn-access.cc.
2735 (expand_builtin_fork_or_exec): Same.
2736 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Pass
2737 compute_objsize additional arguments.
2738 (inbounds_memaccess_p): Same.
2739 (array_bounds_checker::check_array_bounds): Add an assert. Stash
2740 statement in a member.
2741 (check_array_bounds_dom_walker::before_dom_children): Same.
2742 * gimple-array-bounds.h (array_bounds_checker::m_stmt): New member.
2743 * gimple-ssa-sprintf.c (get_destination_size): Add an argument.
2744 (handle_printf_call): Pass a new argument.
2745 * gimple-ssa-warn-access.cc (get_size_range): Add an argument.
2746 (check_access): Add an argument and pass it along to callees.
2747 (check_read_access): Make a member function.
2748 (pass_waccess::check_strcat): Pass access_data ctor additional
2750 (pass_waccess::check_strncat): Same.
2751 (pass_waccess::check_stxcpy): Same.
2752 (pass_waccess::check_stxncpy): Same.
2753 (pass_waccess::check_strncmp): Same.
2754 (pass_waccess::check_read_access): Same.
2755 (pass_waccess::check_builtin): Same.
2756 (pass_waccess::maybe_check_access_sizes): Same.
2757 (pass_waccess::maybe_check_dealloc_call): Same.
2758 * gimple-ssa-warn-access.h (check_read_access): Declare a new
2760 * pointer-query.cc (compute_objsize_r): Add an argument.
2761 (gimple_call_return_array): Same.
2762 (gimple_call_alloc_size): Same.
2763 (access_ref::access_ref): Same.
2764 (access_ref::get_ref): Same.
2765 (pointer_query::get_ref): Same.
2766 (handle_min_max_size): Pass an arguments to callees.
2767 (handle_array_ref): Add an argument.
2768 (handle_mem_ref): Same.
2769 (compute_objsize): Same.
2770 * pointer-query.h (struct access_ref): Adjust signatures.
2771 (struct access_data): Same.
2772 (gimple_call_alloc_size): Add an argument.
2773 (gimple_parm_array_size): Same.
2774 (compute_objsize): Same.
2775 * tree-ssa-strlen.c (strlen_pass::adjust_last_stmt): Pass an additional
2776 argument to compute_objsize.
2777 (strlen_pass::maybe_warn_overflow): Same.
2778 (maybe_diag_stxncpy_trunc): Same.
2780 2021-10-26 Martin Sebor <msebor@redhat.com>
2782 PR middle-end/102453
2783 * gimple-ssa-warn-access.cc (pass_waccess::check_atomic_builtin): New.
2784 (pass_waccess::check_atomic_builtin): Call it.
2786 2021-10-26 Vladimir N. Makarov <vmakarov@redhat.com>
2788 PR rtl-optimization/102842
2789 * lra-constraints.c (match_reload): Ignore out in checking values
2791 (curr_insn_transform): Collect outputs before doing reloads of operands.
2793 2021-10-26 Paul A. Clarke <pc@us.ibm.com>
2796 * config/rs6000/x86intrin.h: Move some included headers to new
2797 headers. Include new immintrin.h instead of those headers.
2798 * config/rs6000/immintrin.h: New.
2799 * config/rs6000/x86gprintrin.h: New.
2800 * config.gcc (powerpc*-*-*): Add new headers to extra_headers.
2802 2021-10-26 Richard Biener <rguenther@suse.de>
2804 * tree-vectorizer.h (vect_create_addr_base_for_vector_ref):
2805 Remove byte_offset parameter.
2806 (vect_create_data_ref_ptr): Likewise.
2807 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
2809 (vect_create_data_ref_ptr): Likewise.
2810 * tree-vect-stmts.c (vectorizable_store): Adjust.
2811 (vectorizable_load): Likewise.
2813 2021-10-26 Richard Biener <rguenther@suse.de>
2815 PR tree-optimization/96109
2816 * tree-vectorizer.h (dr_misalignment): Add optional offset
2818 * tree-vect-data-refs.c (dr_misalignment): Likewise. Remove
2819 offset applied for negative stride accesses.
2820 (vect_enhance_data_refs_alignment): Compute negative stride
2821 access offset and pass it to dr_misalignment.
2822 * tree-vect-stmts.c (get_negative_load_store_type): Pass
2823 negative offset to dr_misalignment.
2824 (get_group_load_store_type): Likewise.
2825 (get_load_store_type): Likewise.
2826 (vectorizable_store): Remove asserts about alignment.
2827 (vectorizable_load): Likewise.
2829 2021-10-26 Kewen Lin <linkw@linux.ibm.com>
2831 PR tree-optimization/102897
2832 * tree-ssa-forwprop.c (simplify_permutation): Remove a wrong assertion.
2834 2021-10-26 Richard Biener <rguenther@suse.de>
2836 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
2837 Take offset in bytes.
2838 (vect_create_data_ref_ptr): Likewise.
2839 * tree-vect-loop-manip.c (get_misalign_in_elems): Multiply
2840 offset by element size.
2841 (vect_create_cond_for_align_checks): Likewise.
2842 * tree-vect-stmts.c (get_negative_load_store_type): Likewise.
2843 (vectorizable_load): Remove duplicate leftover from merge
2846 2021-10-26 Roger Sayle <roger@nextmovesoftware.com>
2848 * config/i386/i386-expand.c (ix86_expand_v1ti_shift): New helper
2849 function to expand V1TI mode logical shifts by integer constants.
2850 (ix86_expand_v1ti_rotate): New helper function to expand V1TI
2851 mode rotations by integer constants.
2852 * config/i386/i386-protos.h (ix86_expand_v1ti_shift,
2853 ix86_expand_v1ti_rotate): Prototype new functions here.
2854 * config/i386/sse.md (ashlv1ti3, lshrv1ti3, rotlv1ti3, rotrv1ti3):
2855 New TARGET_SSE2 expanders to implement V1TI shifts and rotations.
2857 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
2859 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
2860 Avoid threading circular paths.
2862 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
2864 * tree-ssa-threadbackward.c (back_threader::resolve_phi):
2865 Attempt to resolve all incoming paths to a PHI.
2866 (back_threader::resolve_def): Always return true for PHIs.
2868 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
2870 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
2871 Always try to resolve path without looking back.
2872 * tree-ssa-threadupdate.c (dump_jump_thread): Indidicate whether
2873 edge is a back edge.
2875 2021-10-26 Kewen Lin <linkw@linux.ibm.com>
2877 PR tree-optimization/102789
2878 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
2879 update inits of simd_lane_access.
2881 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
2883 * tree-vrp.c (vrp_simplify_cond_using_ranges): Add return type and
2884 move to vr-values.c.
2885 (simplify_casted_conds): Move to vrp_folder class.
2886 (execute_vrp): Call via vrp_folder now.
2887 * vr-values.c (simplify_cond_using_ranges_1): Call simplify_casted_cond.
2888 (simplify_using_ranges::simplify_casted_cond): Relocate from tree-vrp.c.
2889 * vr-values.h (simplify_casted_cond): Add prototype.
2891 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
2893 * tree-vrp.c (rvrp_folder::fold_stmt): If simplification fails, try
2896 2021-10-25 Paul A. Clarke <pc@us.ibm.com>
2898 * config/rs6000/smmintrin.h (_mm_testz_si128): Add "extern" to
2900 (_mm_testc_si128): Likewise.
2901 (_mm_testnzc_si128): Likewise.
2902 (_mm_blend_ps): Likewise.
2903 (_mm_blendv_ps): Likewise.
2904 (_mm_blend_pd): Likewise.
2905 (_mm_blendv_pd): Likewise.
2906 (_mm_ceil_pd): Likewise.
2907 (_mm_ceil_sd): Likewise.
2908 (_mm_ceil_ps): Likewise.
2909 (_mm_ceil_ss): Likewise.
2910 (_mm_floor_pd): Likewise.
2911 (_mm_floor_sd): Likewise.
2912 (_mm_floor_ps): Likewise.
2913 (_mm_floor_ss): Likewise.
2914 (_mm_minpos_epu16): Likewise.
2915 (_mm_mul_epi32): Likewise.
2916 (_mm_cvtepi8_epi16): Likewise.
2917 (_mm_packus_epi32): Likewise.
2918 (_mm_cmpgt_epi64): Likewise.
2920 2021-10-25 Roger Sayle <roger@nextmovesoftware.com>
2922 * simplify-rtx.c (simplify_binary_operation_1) [SS_ASHIFT]: Simplify
2923 shifts of the mode's smin_value and smax_value when the bit count
2924 operand doesn't have side-effects.
2925 [US_ASHIFT]: Likewise, simplify shifts of the mode's umax_value
2926 when the bit count operand doesn't have side-effects.
2927 (simplify_const_binary_operation) [SS_ASHIFT, US_ASHIFT]: Perform
2928 compile-time evaluation of saturating left shifts with constant
2931 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
2933 * gimple-range.cc (gimple_ranger::export_global_ranges): Remove check
2936 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
2938 * flag-types.h (enum ranger_debug): Adjust values.
2939 * params.opt (ranger_debug): Ditto.
2941 2021-10-25 Martin Jambor <mjambor@suse.cz>
2943 PR tree-optimization/102886
2944 * tree-sra.c (totally_scalarize_subtree): Fix the out of
2947 2021-10-25 Andrew Pinski <apinski@marvell.com>
2949 * tree-ssa-dce.c (simple_dce_from_worklist):
2950 Check stmt_unremovable_because_of_non_call_eh_p also
2951 before removing the statement.
2953 2021-10-25 Richard Biener <rguenther@suse.de>
2955 PR tree-optimization/102905
2956 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
2957 Use vect_supportable_dr_alignment again to determine whether
2958 an access is supported when not aligned.
2960 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
2962 * config/riscv/riscv.c (riscv_rtx_costs): Handle cost model
2965 2021-10-25 Jim Wilson <jimw@sifive.com>
2966 Kito Cheng <kito.cheng@sifive.com>
2967 Jia-Wei Chen <jiawei@iscas.ac.cn>
2968 Shi-Hua Liao <shihua@iscas.ac.cn>
2970 * config/riscv/bitmanip.md (shiftm1): New.
2971 (*bset<mode>): Ditto.
2972 (*bset<mode>_mask): Ditto.
2973 (*bset<mode>_1): Ditto.
2974 (*bset<mode>_1_mask): Ditto.
2975 (*bseti<mode>): Ditto.
2976 (*bclr<mode>): Ditto.
2977 (*bclri<mode>): Ditto.
2978 (*binv<mode>): Ditto.
2979 (*binvi<mode>): Ditto.
2980 (*bext<mode>): Ditto.
2982 * config/riscv/predicates.md (splittable_const_int_operand):
2984 (single_bit_mask_operand): New.
2985 (not_single_bit_mask_operand): Ditto.
2986 (const31_operand): Ditto.
2987 (const63_operand): Ditto.
2988 * config/riscv/riscv.c (riscv_build_integer_1): Handle bseti.
2989 (riscv_output_move): Ditto.
2990 (riscv_print_operand): Handle new operand type: T and S.
2991 * config/riscv/riscv.h (SINGLE_BIT_MASK_OPERAND): New.
2993 2021-10-25 Jim Wilson <jimw@sifive.com>
2995 * config/riscv/riscv.c (riscv_build_integer_1): Build integer
2998 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
3000 * config/riscv/riscv.c (riscv_extend_cost): Handle cost model
3002 (riscv_rtx_costs): Ditto.
3004 2021-10-25 Jim Wilson <jimw@sifive.com>
3005 Kito Cheng <kito.cheng@sifive.com>
3006 Jia-Wei Chen <jiawei@iscas.ac.cn>
3008 * config/riscv/bitmanip.md (bitmanip_bitwise): New.
3009 (bitmanip_minmax): New.
3010 (clz_ctz_pcnt): New.
3011 (bitmanip_optab): New.
3012 (bitmanip_insn): New.
3013 (*<optab>_not<mode>): New.
3014 (*xor_not<mode>): New.
3015 (<bitmanip_optab>si2): New.
3016 (*<bitmanip_optab>disi2): New.
3017 (<bitmanip_optab>di2): New.
3018 (*zero_extendhi<GPR:mode>2_bitmanip): New.
3019 (*extend<SHORT:mode><SUPERQI:mode>2_zbb): New.
3020 (*zero_extendhi<GPR:mode>2_zbb): New.
3023 (rotrsi3_sext): New.
3026 (rotlsi3_sext): New.
3027 (bswap<mode>2): New.
3028 (<bitmanip_optab><mode>3): New.
3029 * config/riscv/riscv.md (type): Add rotate.
3030 (zero_extendhi<GPR:mode>2): Change to define_expand pattern.
3031 (*zero_extendhi<GPR:mode>2): New.
3032 (extend<SHORT:mode><SUPERQI:mode>2): Change to define_expand pattern.
3033 (*extend<SHORT:mode><SUPERQI:mode>2): New.
3035 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
3037 * config/riscv/riscv.c (riscv_extend_cost): Handle cost model
3039 (riscv_rtx_costs): Ditto.
3041 2021-10-25 Jim Wilson <jimw@sifive.com>
3042 Kito Cheng <kito.cheng@sifive.com>
3043 Jia-Wei Chen <jiawei@iscas.ac.cn>
3045 * config/riscv/bitmanip.md (*zero_extendsidi2_bitmanip): New.
3050 (riscv_rtx_costs): Ditto.
3051 * config/riscv/riscv.md: Include bitmanip.md
3052 (type): Add bitmanip bype.
3053 (zero_extendsidi2): Change to define_expand pattern.
3054 (*zero_extendsidi2_internal): New.
3055 (zero_extendsidi2_shifted): Disable for ZBA.
3057 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
3059 * common/config/riscv/riscv-common.c (riscv_ext_version_table):
3060 Add zba, zbb, zbc and zbs.
3061 (riscv_ext_flag_table): Ditto.
3062 * config/riscv/riscv-opts.h (MASK_ZBA): New.
3066 (TARGET_ZBA): Ditto.
3067 (TARGET_ZBB): Ditto.
3068 (TARGET_ZBC): Ditto.
3069 (TARGET_ZBS): Ditto.
3070 * config/riscv/riscv.opt (riscv_zb_subext): New.
3072 2021-10-25 liuhongt <hongtao.liu@intel.com>
3075 * match.pd: Simplify (_Float16) sqrtf((float) a) to .SQRT(a)
3076 when direct_internal_fn_supported_p, similar for sqrt/sqrtl.
3078 2021-10-25 Richard Biener <rguenther@suse.de>
3080 PR tree-optimization/102920
3081 * tree-ssa-sccvn.h (expressions_equal_p): Add argument
3082 controlling VN_TOP matching behavior.
3083 * tree-ssa-sccvn.c (expressions_equal_p): Likewise.
3084 (vn_phi_eq): Do not optimistically match VN_TOP.
3086 2021-10-25 konglin1 <lingling.kong@intel.com>
3088 * config/i386/sse.md (fma_<mode>_fadd_fmul): Add new
3089 define_insn_and_split.
3090 (fma_<mode>_fadd_fcmul):Likewise
3091 (fma_<complexopname>_<mode>_fma_zero):Likewise
3093 2021-10-24 John David Anglin <danglin@gcc.gnu.org>
3095 * config/pa/pa-d.c (pa_d_handle_target_float_abi): Don't check
3096 TARGET_DISABLE_FPREGS.
3097 * config/pa/pa.c (fix_range): Use MASK_SOFT_FLOAT instead of
3098 MASK_DISABLE_FPREGS.
3099 (hppa_rtx_costs): Don't check TARGET_DISABLE_FPREGS. Adjust
3100 cost of hardware integer multiplication.
3101 (pa_conditional_register_usage): Don't check TARGET_DISABLE_FPREGS.
3102 * config/pa/pa.h (INT14_OK_STRICT): Likewise.
3103 * config/pa/pa.md: Don't check TARGET_DISABLE_FPREGS. Check
3104 TARGET_SOFT_FLOAT in patterns that use xmpyu instruction.
3105 * config/pa/pa.opt (mdisable-fpregs): Change target mask to
3106 SOFT_FLOAT. Revise comment.
3107 (msoft-float): New option.
3109 2021-10-24 John David Anglin <danglin@gcc.gnu.org>
3111 * config/pa/pa.md: Don't use 'G' constraint in integer move patterns.
3113 2021-10-24 Gerald Pfeifer <gerald@pfeifer.com>
3115 * doc/install.texi (Specific): Remove obsolete details
3116 around GNU/Linux on Itanium.
3117 (Specific): Remove reference to Windows for Itanium.
3119 2021-10-23 Bernhard Reutner-Fischer <aldot@gcc.gnu.org>
3121 * config/i386/x86-tune-sched-bd.c (dispatch_group): Commentary
3124 2021-10-23 Jan Hubicka <hubicka@ucw.cz>
3126 * tree-ssa-structalias.c (compute_points_to_sets): Cleanup.
3128 2021-10-23 Roger Sayle <roger@nextmovesoftware.com>
3130 * config/i386/sse.md (<any_logic>v1ti3): New define_insn to
3131 implement V1TImode AND, IOR and XOR on TARGET_SSE2 (and above).
3132 (one_cmplv1ti2): New define expand.
3134 2021-10-22 Eric Gallager <egallager@gcc.gnu.org>
3137 * Makefile.in: Handle dvidir and install-dvi target.
3138 * configure: Regenerate.
3139 * configure.ac: Add install-dvi to target_list.
3141 2021-10-22 Gerald Pfeifer <gerald@pfeifer.com>
3143 * doc/install.texi (Binaries): Convert mingw-w64.org to https.
3146 2021-10-22 Richard Biener <rguenther@suse.de>
3148 PR tree-optimization/102893
3149 * tree-ssa-dce.c (find_obviously_necessary_stmts): Fix the
3150 test for an exit edge.
3152 2021-10-22 Aldy Hernandez <aldyh@redhat.com>
3153 Andrew MacLeod <amacleod@redhat.com>
3155 * gimple-range-path.cc (path_range_query::compute_phi_relations):
3156 Kill any global relations we may know before registering a new
3158 * value-relation.cc (path_oracle::killing_def): New.
3159 * value-relation.h (path_oracle::killing_def): New.
3161 2021-10-22 Richard Biener <rguenther@suse.de>
3164 * tree-ssa-sccvn.c (vn_phi_insert): For undefined SSA args
3166 (vn_phi_lookup): Likewise.
3168 2021-10-21 H.J. Lu <hjl.tools@gmail.com>
3171 * doc/invoke.texi: Document -fcf-protection requires i686 or
3174 2021-10-21 Eric Botcazou <ebotcazou@adacore.com>
3176 PR middle-end/102764
3177 * cfgexpand.c (expand_gimple_basic_block): Robustify latest change.
3179 2021-10-21 Jonathan Wright <jonathan.wright@arm.com>
3181 * config/aarch64/arm_neon.h (__STRUCTN): Delete function
3182 macro and all invocations.
3184 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
3186 * doc/invoke.texi (ranger-debug): Document.
3187 * flag-types.h (enum ranger_debug): New.
3188 (enum evrp_mode): Remove debug values.
3189 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Use new debug flag.
3190 * gimple-range-gori.cc (gori_compute::gori_compute): Ditto.
3191 * gimple-range.cc (gimple_ranger::gimple_ranger): Ditto.
3192 * gimple-ssa-evrp.c (hybrid_folder::choose_value): Ditto.
3193 (execute_early_vrp): Use evrp-mode directly.
3194 * params.opt (enum evrp_mode): Remove debug values.
3195 (ranger-debug): New.
3196 (ranger-logical-depth): Relocate to be in alphabetical order.
3198 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
3200 * doc/invoke.texi: (vrp1-mode, vrp2-mode): Document.
3201 * flag-types.h: (enum vrp_mode): New.
3202 * params.opt: (vrp1-mode, vrp2-mode): New.
3203 * tree-vrp.c (vrp_pass_num): New.
3204 (pass_vrp::pass_vrp): Set pass number.
3205 (pass_vrp::execute): Choose which VRP mode to execute.
3207 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
3209 * gimple-ssa-evrp.c (class rvrp_folder): Move to tree-vrp.c.
3210 (execute_early_vrp): For ranger only mode, invoke ranger_vrp.
3211 * tree-vrp.c (class rvrp_folder): Relocate here.
3212 (execute_ranger_vrp): New.
3213 * tree-vrp.h (execute_ranger_vrp): Export.
3215 2021-10-21 Martin Liska <mliska@suse.cz>
3219 * opts.c (finish_options): Process flag_var_tracking* options
3220 here as they can be adjusted by optimize attribute.
3221 Process also flag_syntax_only and flag_gtoggle.
3222 * toplev.c (process_options): Remove it here.
3223 * common.opt: Make debug_nonbind_markers_p as PerFunction
3224 attribute as it depends on optimization level.
3226 2021-10-21 Martin Jambor <mjambor@suse.cz>
3228 PR tree-optimization/102505
3229 * tree-sra.c (totally_scalarize_subtree): Check that the
3230 encountered field fits within the acces we would like to put it
3233 2021-10-21 Aldy Hernandez <aldyh@redhat.com>
3235 * tree-ssa-threadbackward.c
3236 (back_threader::maybe_register_path): Remove circular paths check.
3238 2021-10-21 Richard Biener <rguenther@suse.de>
3240 * toplev.c (process_options): Move the initial debug_hooks
3242 (toplev::main): ... before the call of the post_options
3245 2021-10-21 Richard Biener <rguenther@suse.de>
3247 PR tree-optimization/102847
3248 * tree-vect-stmts.c (vect_model_load_cost): Add the scalar
3249 load cost in the prologue for VMAT_INVARIANT.
3251 2021-10-21 Richard Biener <rguenther@suse.de>
3253 PR tree-optimization/102847
3254 * tree-vect-stmts.c (vect_model_load_cost): Explicitely
3255 handle VMAT_INVARIANT as a splat in the prologue.
3257 2021-10-21 Hongyu Wang <hongyu.wang@intel.com>
3260 * config/i386/i386.c (ix86_get_ssemov): Adjust HFmode vector
3261 move to use the same logic as HImode.
3263 2021-10-21 Richard Biener <rguenther@suse.de>
3265 * tree-vect-slp.c (vect_build_slp_tree_1): Remove
3266 superfluous gimple_call_nothrow_p check.
3268 2021-10-21 Andrew Pinski <apinski@marvell.com>
3270 * tree-cfg.c (maybe_remove_writeonly_store): Add dce_ssa_names argument.
3271 Mark the ssa-name of the rhs as one to be removed.
3272 (execute_fixup_cfg): Update call to maybe_remove_writeonly_store.
3273 Call simple_dce_from_worklist at the end to a simple dce.
3275 2021-10-21 Andrew Pinski <apinski@marvell.com>
3277 * tree-cfg.c (maybe_remove_writeonly_store): New function
3278 factored out from ...
3279 (execute_fixup_cfg): Here. Call maybe_remove_writeonly_store.
3281 2021-10-21 Andrew Pinski <apinski@marvell.com>
3283 * tree-cfg.c (execute_fixup_cfg): Remove comment
3284 about standalone pass.
3286 2021-10-21 Andrew Pinski <apinski@marvell.com>
3288 * tree-cfg.c (execute_fixup_cfg): Output when the statement
3289 is removed when it is a write only var.
3291 2021-10-21 Aldy Hernandez <aldyh@redhat.com>
3293 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
3294 Avoid threading circular paths.
3296 2021-10-20 Alex Coplan <alex.coplan@arm.com>
3298 * calls.c (initialize_argument_information): Remove some dead
3299 code, remove handling for function_arg returning const_int.
3300 * doc/tm.texi: Delete documentation for unused target hooks.
3301 * doc/tm.texi.in: Likewise.
3302 * target.def (load_bounds_for_arg): Delete.
3303 (store_bounds_for_arg): Delete.
3304 (load_returned_bounds): Delete.
3305 (store_returned_bounds): Delete.
3306 * targhooks.c (default_load_bounds_for_arg): Delete.
3307 (default_store_bounds_for_arg): Delete.
3308 (default_load_returned_bounds): Delete.
3309 (default_store_returned_bounds): Delete.
3310 * targhooks.h (default_load_bounds_for_arg): Delete.
3311 (default_store_bounds_for_arg): Delete.
3312 (default_load_returned_bounds): Delete.
3313 (default_store_returned_bounds): Delete.
3315 2021-10-20 Jonathan Wakely <jwakely@redhat.com>
3317 * doc/extend.texi (Basic Asm): Clarify that asm is not an
3319 * doc/invoke.texi (-fno-asm): Fix description for C++.
3321 2021-10-20 Jonathan Wakely <jwakely@redhat.com>
3323 * doc/install.texi: Remove link to old.html
3325 2021-10-20 Tamar Christina <tamar.christina@arm.com>
3327 * config/aarch64/aarch64-simd.md (*aarch64_cmtst_same_<mode>): New.
3329 2021-10-20 Tamar Christina <tamar.christina@arm.com>
3331 * config/aarch64/aarch64-simd.md (*aarch64_narrow_trunc<mode>): New.
3333 2021-10-20 Tamar Christina <tamar.christina@arm.com>
3335 * config/aarch64/aarch64-simd.md (aarch64_simd_ashr<mode>): Add case cmp
3337 * config/aarch64/constraints.md (D1): New.
3339 2021-10-20 Tamar Christina <tamar.christina@arm.com>
3341 * config/aarch64/aarch64-simd.md
3342 (*aarch64_<srn_op>topbits_shuffle<mode>_le): New.
3343 (*aarch64_topbits_shuffle<mode>_le): New.
3344 (*aarch64_<srn_op>topbits_shuffle<mode>_be): New.
3345 (*aarch64_topbits_shuffle<mode>_be): New.
3346 * config/aarch64/predicates.md
3347 (aarch64_simd_shift_imm_vec_exact_top): New.
3349 2021-10-20 Tamar Christina <tamar.christina@arm.com>
3351 * config/aarch64/aarch64-simd.md (*aarch64_<srn_op>shrn<mode>_vect,
3352 *aarch64_<srn_op>shrn<mode>2_vect_le,
3353 *aarch64_<srn_op>shrn<mode>2_vect_be): New.
3354 * config/aarch64/iterators.md (srn_op): New.
3356 2021-10-20 Chung-Lin Tang <cltang@codesourcery.com>
3358 * omp-low.c (omp_copy_decl_2): For !ctx, use record_vars to add new copy
3360 (scan_sharing_clauses): Place copy of OMP_CLAUSE_IN_REDUCTION decl in
3361 ctx->outer instead of ctx.
3363 2021-10-20 Martin Liska <mliska@suse.cz>
3366 2021-10-19 Martin Liska <mliska@suse.cz>
3369 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
3370 * system.h (strip_whilespaces): New function.
3372 2021-10-20 Martin Liska <mliska@suse.cz>
3375 2021-10-19 Martin Liska <mliska@suse.cz>
3378 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
3381 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
3383 * config/aarch64/aarch64.c (aarch64_case_values_threshold):
3384 Change to 8 with -Os, 11 otherwise.
3386 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
3388 * config/aarch64/aarch64.c (neoversev1_tunings):
3389 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
3390 (neoversen2_tunings): Likewise.
3392 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
3395 * config/aarch64/aarch64.md (frint_pattern): Update comment.
3396 * config/aarch64/aarch64-simd-builtins.def: Change frintn to roundeven.
3397 * config/aarch64/arm_fp16.h: Change frintn to roundeven.
3398 * config/aarch64/arm_neon.h: Likewise.
3399 * config/aarch64/iterators.md (frint_pattern): Use roundeven for FRINTN.
3401 2021-10-20 Martin Liska <mliska@suse.cz>
3403 * config/arm/arm.c (arm_unwind_emit_sequence): Do not declare
3404 already declared global variable.
3405 (arm_unwind_emit_set): Use out_file as function argument.
3406 (arm_unwind_emit): Likewise.
3407 * config/darwin.c (machopic_output_data_section_indirection): Likewise.
3408 (machopic_output_stub_indirection): Likewise.
3409 (machopic_output_indirection): Likewise.
3410 (machopic_finish): Likewise.
3411 * config/i386/i386.c (ix86_asm_output_function_label): Likewise.
3412 * config/i386/winnt.c (i386_pe_seh_unwind_emit): Likewise.
3413 * config/ia64/ia64.c (process_epilogue): Likewise.
3414 (process_cfa_adjust_cfa): Likewise.
3415 (process_cfa_register): Likewise.
3416 (process_cfa_offset): Likewise.
3417 (ia64_asm_unwind_emit): Likewise.
3418 * config/s390/s390.c (s390_asm_output_function_label): Likewise.
3420 2021-10-20 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
3422 * config/aarch64/aarch64-builtins.c
3423 (aarch64_general_gimple_fold_builtin):
3424 lower vld1 and vst1 variants of the neon builtins
3425 * config/aarch64/aarch64-protos.h:
3426 (aarch64_general_gimple_fold_builtin): Add gsi parameter.
3427 * config/aarch64/aarch64.c (aarch64_general_gimple_fold_builtin):
3430 2021-10-20 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
3432 * match.pd: Generate IFN_TRUNC.
3434 2021-10-20 Richard Biener <rguenther@suse.de>
3436 PR tree-optimization/102853
3437 * tree-data-ref.c (split_constant_offset_1): Bail out
3438 immediately if the expression traps on overflow.
3440 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
3442 * tree-ssa-threadbackward.c (back_threader::~back_threader): Remove.
3444 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
3446 * tree-ssa-threadupdate.c (back_jt_path_registry::adjust_paths_after_duplication):
3447 Remove superflous debugging message.
3448 (back_jt_path_registry::duplicate_thread_path): Same.
3450 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
3452 * tree-ssa-threadbackward.c (back_threader_registry::back_threader_registry):
3454 (back_threader_registry::register_path): Remove m_threaded_paths.
3456 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
3458 PR tree-optimization/102814
3459 * doc/invoke.texi: Document --param=max-fsm-thread-length.
3460 * params.opt: Add --param=max-fsm-thread-length.
3461 * tree-ssa-threadbackward.c
3462 (back_threader_profitability::profitable_path_p): Fail on paths
3463 longer than max-fsm-thread-length.
3465 2021-10-20 Eric Botcazou <ebotcazou@adacore.com>
3467 PR middle-end/102764
3468 * cfgexpand.c (expand_gimple_basic_block): Disregard a final debug
3469 statement to reset the current location for the outgoing edges.
3471 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
3472 Richard Biener <rguenther@suse.de>
3474 * tree-ssa-threadupdate.c (cancel_thread): Dump threading reason
3475 on the same line as the threading cancellation.
3476 (jt_path_registry::cancel_invalid_paths): Avoid rotating loops.
3477 Avoid threading through loop headers where the path remains in the
3480 2021-10-20 Siddhesh Poyarekar <siddhesh@gotplt.org>
3482 * tree-object-size.c (unknown): Make into a function. Adjust
3484 (unknown_object_size): Simplify implementation.
3486 2021-10-20 Hongtao Liu <hongtao.liu@intel.com>
3487 Kewen Lin <linkw@linux.ibm.com>
3489 * doc/sourcebuild.texi (Effective-Target Keywords): Document
3490 vect_slp_v2qi_store, vect_slp_v4qi_store, vect_slp_v8qi_store,
3491 vect_slp_v16qi_store, vect_slp_v2hi_store,
3492 vect_slp_v4hi_store, vect_slp_v2si_store, vect_slp_v4si_store.
3494 2021-10-19 Jonathan Wakely <jwakely@redhat.com>
3496 * doc/extend.texi (Basic PowerPC Built-in Functions): Fix typo.
3498 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
3502 * config/rs6000/emmintrin.h: Guard POWER8 intrinsics.
3503 * config/rs6000/pmmintrin.h: Same.
3504 * config/rs6000/smmintrin.h: Same.
3505 * config/rs6000/tmmintrin.h: Same.
3507 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
3509 * config.gcc (extra_headers): Add nmmintrin.h.
3511 2021-10-19 Richard Biener <rguenther@suse.de>
3513 * tree-vectorizer.h (vect_supportable_dr_alignment): Add
3514 misalignment parameter.
3515 * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
3516 Do not change DR_MISALIGNMENT in place, instead pass the
3517 adjusted misalignment to vect_supportable_dr_alignment.
3518 (vect_peeling_supportable): Likewise.
3519 (vect_peeling_hash_get_lowest_cost): Adjust.
3520 (vect_enhance_data_refs_alignment): Likewise.
3521 (vect_vfa_access_size): Likewise.
3522 (vect_supportable_dr_alignment): Add misalignment
3523 parameter and simplify.
3524 * tree-vect-stmts.c (get_negative_load_store_type): Adjust.
3525 (get_group_load_store_type): Likewise.
3526 (get_load_store_type): Likewise.
3528 2021-10-19 Clément Chigot <clement.chigot@atos.net>
3530 * config/rs6000/rs6000.c (rs6000_xcoff_file_end): Move
3531 __tls_get_addr reference to .text csect.
3533 2021-10-19 Martin Liska <mliska@suse.cz>
3536 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
3539 2021-10-19 Richard Biener <rguenther@suse.de>
3541 * tree-vectorizer.h (vect_get_store_cost): Adjust signature.
3542 (vect_get_load_cost): Likewise.
3543 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3544 alignment support scheme and misalignment as arguments
3546 (vect_get_peeling_costs_all_drs): Compute that info here
3547 and note that we shouldn't need to.
3548 * tree-vect-stmts.c (vect_model_store_cost): Get
3549 alignment support scheme and misalignment as arguments.
3550 (vect_get_store_cost): Likewise.
3551 (vect_model_load_cost): Likewise.
3552 (vect_get_load_cost): Likewise.
3553 (vectorizable_store): Pass down alignment support scheme
3554 and misalignment to costing.
3555 (vectorizable_load): Likewise.
3557 2021-10-19 Richard Biener <rguenther@suse.de>
3559 * tree-vect-stmts.c (get_negative_load_store_type): Add
3560 offset output parameter and initialize it.
3561 (get_group_load_store_type): Likewise.
3562 (get_load_store_type): Likewise.
3563 (vectorizable_store): Use offset as computed by
3564 get_load_store_type.
3565 (vectorizable_load): Likewise.
3567 2021-10-19 Richard Biener <rguenther@suse.de>
3569 PR tree-optimization/102827
3570 * tree-if-conv.c (predicate_statements): Add pe parameter
3571 and use that edge to insert invariant stmts on.
3572 (combine_blocks): Pass through pe.
3573 (tree_if_conversion): Compute the edge to insert invariant
3574 stmts on and pass it along.
3576 2021-10-19 Roger Sayle <roger@nextmovesoftware.com>
3579 * config/bfin/bfin.md (addsubv2hi3, subaddv2hi3, ssaddsubv2hi3,
3580 sssubaddv2hi3): Swap the order of operators in vec_concat.
3582 2021-10-19 Xionghu Luo <luoxhu@linux.ibm.com>
3584 * config/rs6000/altivec.md (*altivec_vmrghb_internal): Delete.
3585 (altivec_vmrghb_direct): New.
3586 (*altivec_vmrghh_internal): Delete.
3587 (altivec_vmrghh_direct): New.
3588 (*altivec_vmrghw_internal): Delete.
3589 (altivec_vmrghw_direct_<mode>): New.
3590 (altivec_vmrghw_direct): Delete.
3591 (*altivec_vmrglb_internal): Delete.
3592 (altivec_vmrglb_direct): New.
3593 (*altivec_vmrglh_internal): Delete.
3594 (altivec_vmrglh_direct): New.
3595 (*altivec_vmrglw_internal): Delete.
3596 (altivec_vmrglw_direct_<mode>): New.
3597 (altivec_vmrglw_direct): Delete.
3598 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
3599 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):
3601 * config/rs6000/vsx.md (vsx_xxmrghw_<mode>): Adjust.
3602 (vsx_xxmrglw_<mode>): Adjust.
3604 2021-10-19 Aldy Hernandez <aldyh@redhat.com>
3606 * passes.def: Change threading comment before pass_ccp pass.
3608 2021-10-19 Haochen Gui <guihaoc@gcc.gnu.org>
3610 * config/rs6000/rs6000-call.c (altivec_expand_lxvr_builtin):
3611 Modify the expansion for sign extension. All extensions are done
3612 within VSX registers.
3614 2021-10-19 Richard Biener <rguenther@suse.de>
3616 * tree-vect-stmts.c (get_group_load_store_type): Add
3617 misalignment output parameter and initialize it.
3618 (get_group_load_store_type): Likewise.
3619 (vectorizable_store): Remove now redundant queries.
3620 (vectorizable_load): Likewise.
3622 2021-10-19 Richard Biener <rguenther@suse.de>
3624 * tree-vectorizer.h (vect_supportable_dr_alignment): Remove
3625 check_aligned argument.
3626 * tree-vect-data-refs.c (vect_supportable_dr_alignment):
3628 (vect_peeling_hash_insert): Add supportable_if_not_aligned
3629 argument and do not call vect_supportable_dr_alignment here.
3630 (vect_peeling_supportable): Adjust.
3631 (vect_enhance_data_refs_alignment): Compute whether the
3632 access is supported with different alignment here and
3633 pass that down to vect_peeling_hash_insert.
3634 (vect_vfa_access_size): Adjust.
3635 * tree-vect-stmts.c (vect_get_store_cost): Likewise.
3636 (vect_get_load_cost): Likewise.
3637 (get_negative_load_store_type): Likewise.
3638 (get_group_load_store_type): Likewise.
3639 (get_load_store_type): Likewise.
3641 2021-10-19 Martin Liska <mliska@suse.cz>
3644 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
3645 * system.h (strip_whilespaces): New function.
3647 2021-10-19 dianhong xu <dianhong.xu@intel.com>
3649 * config/i386/avx512fp16intrin.h:
3650 (_mm512_set1_pch): New intrinsic.
3651 * config/i386/avx512fp16vlintrin.h:
3652 (_mm256_set1_pch): New intrinsic.
3653 (_mm_set1_pch): Ditto.
3655 2021-10-18 Andrew MacLeod <amacleod@redhat.com>
3657 PR tree-optimization/102796
3658 * gimple-range.cc (gimple_ranger::range_on_edge): Process EH edges
3659 normally. Return get_tree_range for non gimple_range_ssa_p names.
3660 (gimple_ranger::range_of_stmt): Use get_tree_range for non
3661 gimple_range_ssa_p names.
3663 2021-10-18 Uroš Bizjak <ubizjak@gmail.com>
3666 * config/i386/i386.c (ix86_print_operand_address):
3667 Error out for non-address_operand asm operands.
3669 2021-10-18 Richard Biener <rguenther@suse.de>
3671 * tree-vect-data-refs.c (vect_peeling_hash_insert): Do
3672 not auto-convert dr_alignment_support to bool.
3673 (vect_peeling_supportable): Likewise.
3674 (vect_enhance_data_refs_alignment): Likewise.
3675 (vect_supportable_dr_alignment): Commonize read/write case.
3676 * tree-vect-stmts.c (vect_get_store_cost): Use
3677 dr_alignment_support, not int, for the vect_supportable_dr_alignment
3679 (vect_get_load_cost): Likewise.
3681 2021-10-18 Siddhesh Poyarekar <siddhesh@gotplt.org>
3683 * tree-object-size.c (object_sizes_execute): Consolidate LHS
3684 null check and do it early.
3686 2021-10-18 Richard Biener <rguenther@suse.de>
3688 * tree-vect-stmts.c (vectorizable_store): Use the
3689 computed alignment scheme instead of querying
3692 2021-10-18 Richard Biener <rguenther@suse.de>
3694 * tree-vect-stmts.c (vectorizable_store): Do not recompute
3695 alignment scheme already determined by get_load_store_type.
3697 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
3699 * tree-ssa-threadbackward.c (class pass_thread_jumps_full):
3700 Clone corresponding pass.
3702 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
3704 * combine.c (recog_for_combine): For an unrecognized move/set of
3705 a constant, try force_const_mem to place it in the constant pool.
3706 * cse.c (constant_pool_entries_cost, constant_pool_entries_regcost):
3707 Delete global variables (that are no longer assigned a cost value).
3708 (cse_insn): Simplify logic for deciding whether to place a folded
3709 constant in the constant pool using force_const_mem.
3710 (cse_main): Remove zero initialization of constant_pool_entries_cost
3711 and constant_pool_entries_regcost.
3712 * config/i386/i386.c (ix86_rtx_costs): Make memory accesses
3713 fractionally more expensive, when optimizing for speed.
3715 2021-10-18 Martin Liska <mliska@suse.cz>
3717 PR gcov-profile/102746
3718 PR gcov-profile/102747
3719 * gcov.c (main): Return return_code.
3720 (output_gcov_file): Mark return_code when error happens.
3721 (generate_results): Likewise.
3722 (read_graph_file): Likewise.
3723 (read_count_file): Likewise.
3725 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
3727 * config/bfin/bfin.md (define_constants): Remove UNSPEC_ONES.
3728 (define_insn "ones"): Replace UNSPEC_ONES with a truncate of
3729 a popcount, allowing compile-time evaluation/simplification.
3730 (popcountsi2, popcounthi2): New expanders using a "ones" insn.
3732 2021-10-18 Richard Biener <rguenther@suse.de>
3734 PR tree-optimization/102788
3735 * tree-vect-patterns.c (vect_init_pattern_stmt): Allow
3737 (vect_pattern_recog_1): Likewise.
3738 (vect_recog_bool_pattern): Continue matching the pattern
3739 even if we do not have a vector type for a conversion
3742 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
3744 * simplify-rtx.c (simplify_const_unary_operation) [SS_NEG, SS_ABS]:
3745 Evalute SS_NEG and SS_ABS of a constant argument.
3747 2021-10-18 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
3750 * gimple-match-head.c (try_conditional_simplification): Add case for single operand.
3751 * internal-fn.def: Add entry for COND_NEG internal function.
3752 * internal-fn.c (FOR_EACH_CODE_MAPPING): Add entry for
3753 NEGATE_EXPR, COND_NEG mapping.
3754 * optabs.def: Add entry for cond_neg_optab.
3755 * match.pd (UNCOND_UNARY, COND_UNARY): New operator lists.
3756 (vec_cond COND (foo A) B) -> (IFN_COND_FOO COND A B): New pattern.
3757 (vec_cond COND B (foo A)) -> (IFN_COND_FOO ~COND A B): Likewise.
3759 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
3761 * tree-ssa-strlen.c (class strlen_pass): Rename from
3763 (handle_builtin_stxncpy_strncat): Move to strlen_pass.
3764 (handle_assign): Same.
3765 (adjust_last_stmt): Same.
3766 (maybe_warn_overflow): Same.
3767 (handle_builtin_strlen): Same.
3768 (handle_builtin_strchr): Same.
3769 (handle_builtin_strcpy): Same.
3770 (handle_builtin_strncat): Same.
3771 (handle_builtin_stxncpy_strncat): Same.
3772 (handle_builtin_memcpy): Same.
3773 (handle_builtin_strcat): Same.
3774 (handle_alloc_call): Same.
3775 (handle_builtin_memset): Same.
3776 (handle_builtin_memcmp): Same.
3777 (get_len_or_size): Same.
3778 (strxcmp_eqz_result): Same.
3779 (handle_builtin_string_cmp): Same.
3780 (handle_pointer_plus): Same.
3781 (count_nonzero_bytes_addr): Same.
3782 (count_nonzero_bytes): Same.
3783 (handle_store): Same.
3784 (strlen_check_and_optimize_call): Same.
3785 (handle_integral_assign): Same.
3786 (check_and_optimize_stmt): Same.
3787 (printf_strlen_execute): Rename strlen_dom_walker to strlen_pass.
3789 2021-10-18 Richard Biener <rguenther@suse.de>
3791 PR tree-optimization/102798
3792 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
3793 Only copy points-to info to newly generated SSA names.
3795 2021-10-18 Martin Liska <mliska@suse.cz>
3797 * dbgcnt.c (dbg_cnt_process_opt): Remove unused but set variable.
3798 * gcov.c (get_cycles_count): Likewise.
3799 * lto-compress.c (lto_compression_zlib): Likewise.
3800 (lto_uncompression_zlib): Likewise.
3801 * targhooks.c (default_pch_valid_p): Likewise.
3803 2021-10-17 Aldy Hernandez <aldyh@redhat.com>
3805 * tree-pass.h (make_pass_thread_jumps_full): New.
3806 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Inline.
3807 (try_thread_blocks): Add resolve and speed arguments.
3808 (pass_thread_jumps::execute): Inline.
3809 (do_early_thread_jumps): New.
3810 (do_thread_jumps): New.
3811 (make_pass_thread_jumps): Move.
3812 (pass_early_thread_jumps::gate): Inline.
3813 (pass_early_thread_jumps::execute): Inline.
3814 (class pass_thread_jumps_full): New.
3816 2021-10-16 Piotr Kubaj <pkubaj@FreeBSD.org>
3818 * configure.ac: Treat powerpc64*-*-freebsd* the same as
3819 powerpc64-*-freebsd*.
3820 * configure: Regenerate.
3822 2021-10-16 H.J. Lu <hjl.tools@gmail.com>
3824 * value-query.cc (get_ssa_name_ptr_info_nonnull): Change
3825 set_ptr_nonull to set_ptr_nonnull in comments.
3827 2021-10-16 Jan Hubicka <hubicka@ucw.cz>
3829 PR tree-optimization/102720
3830 * tree-ssa-structalias.c (compute_points_to_sets): Fix producing
3831 of call used and clobbered sets.
3833 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
3835 * config/darwin.h (LINK_COMMAND_SPEC_A): Update 'r' handling to
3836 skip gomp and itm when r or nodefaultlibs is given.
3837 (DSYMUTIL_SPEC): Do not call dsymutil for '-r' link lines.
3838 Update ordering of exclusions, remove duplicate 'v' addition
3839 (collect2 will add this from the main command line).
3841 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
3843 * config/darwin-driver.c (darwin_driver_init): Revise comments, handle
3844 filelist and framework options in specs instead of code.
3845 * config/darwin.h (SUBTARGET_DRIVER_SELF_SPECS): Update to handle link
3846 specs that are really driver ones.
3847 (DARWIN_CC1_SPEC): Likewise.
3848 (CPP_SPEC): Likewise.
3849 (SYSROOT_SPEC): Append space.
3850 (LINK_SYSROOT_SPEC): Remove most driver link specs.
3851 (STANDARD_STARTFILE_PREFIX_2): Update link-related specs.
3852 (STARTFILE_SPEC): Likewise.
3853 (ASM_MMACOSX_VERSION_MIN_SPEC): Fix line wrap.
3854 (ASM_SPEC): Update driver-related specs.
3855 (ASM_FINAL_SPEC): Likewise.
3856 * config/darwin.opt: Remove now unused option aliases.
3857 * config/i386/darwin.h (EXTRA_ASM_OPTS): Ensure space after opt.
3858 (ASM_SPEC): Update driver-related specs.
3860 2021-10-15 Roger Sayle <roger@nextmovesoftware.com>
3862 * config/i386/i386.c (ix86_hardreg_mov_ok): For vector modes,
3863 allow standard_sse_constant_p immediate constants.
3865 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
3867 * config.gcc: Add tm-dwarf2.h to tm_d-file.
3869 2021-10-15 Andrew MacLeod <amacleod@redhat.com>
3871 * gimple-range-fold.h (gimple_range_ssa_p): Don't process names
3872 that occur in abnormal phis.
3873 * gimple-range.cc (gimple_ranger::range_on_edge): Return false for
3874 abnormal and EH edges.
3875 * gimple-ssa-evrp.c (rvrp_folder::value_of_expr): Ditto.
3876 (rvrp_folder::value_on_edge): Ditto.
3877 (rvrp_folder::value_of_stmt): Ditto.
3878 (hybrid_folder::value_of_expr): Ditto for ranger queries.
3879 (hybrid_folder::value_on_edge): Ditto.
3880 (hybrid_folder::value_of_stmt): Ditto.
3881 * value-query.cc (gimple_range_global): Always return a range if
3882 the type is supported.
3884 2021-10-15 John David Anglin <danglin@gcc.gnu.org>
3886 * config/pa/pa.md: Consistently use "rG" constraint for copy
3887 instruction in move patterns.
3889 2021-10-15 Andrew Stubbs <ams@codesourcery.com>
3891 * config/gcn/gcn-hsa.h (S_FIJI): Set unconditionally.
3894 * config/gcn/gcn.c: Hard code SRAM ECC settings for old architectures.
3895 * config/gcn/mkoffload.c (ELFABIVERSION_AMDGPU_HSA): Rename to ...
3896 (ELFABIVERSION_AMDGPU_HSA_V3): ... this.
3897 (ELFABIVERSION_AMDGPU_HSA_V4): New.
3898 (SET_SRAM_ECC_UNSUPPORTED): New.
3899 (copy_early_debug_info): Create elf flags to match the other objects.
3900 (main): Just let the attribute flags pass through.
3902 2021-10-15 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
3904 * tree-loop-distribution.c (reduction_var_overflows_first):
3905 Pass the type of reduction_var as first argument as it is also
3906 done for the load type.
3907 (loop_distribution::transform_reduction_loop): Add missing
3908 TREE_TYPE while determining precission of reduction_var.
3910 2021-10-15 Richard Biener <rguenther@suse.de>
3912 * defaults.h (PREFERRED_DEBUGGING_TYPE): Choose DWARF2_DEBUG
3914 * toplev.c (process_options): Warn when STABS debugging is
3915 enabled but not the preferred format.
3916 * config/pa/som.h (PREFERRED_DEBUGGING_TYPE): Define to
3918 * config/pdp11/pdp11.h (PREFERRED_DEBUGGING_TYPE): Likewise.
3920 2021-10-15 Richard Biener <rguenther@suse.de>
3923 * tree-inline.c (copy_bb): Avoid underflowing nargs.
3925 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
3927 * config/i386/i386-expand.c (ix86_expand_vec_perm): Convert
3928 HFmode input operand to HImode.
3929 (ix86_vectorize_vec_perm_const): Likewise.
3930 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1_hf):
3932 (*avx512f_permvar_truncv8siv8hi_1_hf):
3935 2021-10-15 Richard Biener <rguenther@suse.de>
3937 PR middle-end/102682
3938 * expmed.c (store_bit_field_1): Ensure a LHS subreg would
3939 not create a paradoxical subreg.
3941 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
3943 * config/i386/i386-expand.c (ix86_expand_vector_init):
3944 For half_vector concat for HFmode, handle them like HImode.
3946 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
3948 * tree-ssa-threadbackward.c (class back_threader): Add m_resolve.
3949 (back_threader::back_threader): Same.
3950 (back_threader::resolve_phi): Try to solve without looking back if
3952 (back_threader::find_paths_to_names): Same.
3953 (try_thread_blocks): Pass resolve argument to back threader.
3954 (pass_early_thread_jumps::execute): Same.
3956 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
3958 * doc/invoke.texi: Remove max-fsm-thread-length,
3959 max-fsm-thread-paths, and fsm-maximum-phi-arguments.
3961 * tree-ssa-threadbackward.c (back_threader::back_threader): Remove
3963 (back_threader_registry::back_threader_registry): Same.
3964 (back_threader_profitability::profitable_path_p): Remove
3965 param_max_fsm_thread-length.
3966 (back_threader_registry::register_path): Remove
3967 m_max_allowable_paths.
3969 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
3971 * tree-ssa-threadbackward.c (class back_threader): Make m_imports
3973 (back_threader::~back_threader): Do not release m_path.
3975 2021-10-14 Andrew MacLeod <amacleod@redhat.com>
3977 PR tree-optimization/102738
3978 * vr-values.c (simplify_using_ranges::simplify): Handle RSHIFT_EXPR.
3980 2021-10-14 Kwok Cheung Yeung <kcy@codesourcery.com>
3982 * omp-general.c (omp_check_context_selector): Move from c-omp.c.
3983 (omp_mark_declare_variant): Move from c-omp.c.
3984 (omp_context_name_list_prop): Update for Fortran strings.
3985 * omp-general.h (omp_check_context_selector): New prototype.
3986 (omp_mark_declare_variant): New prototype.
3988 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
3990 * config/rs6000/rs6000.c (rs6000_density_test): Move early
3991 exit test further up the function.
3993 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
3995 * config/arm/arm.c (arm_add_stmt_cost): Delete.
3996 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
3998 2021-10-14 Martin Jambor <mjambor@suse.cz>
4000 * doc/invoke.texi (Optimize Options): Add entry for
4001 ipa-cp-recursive-freq-factor.
4003 2021-10-14 Tamar Christina <tamar.christina@arm.com>
4005 * match.pd: New rule.
4007 2021-10-14 Jan Hubicka <hubicka@ucw.cz>
4010 * ipa-modref-tree.h (modref_access_node::update2):
4011 Also check that parm_offset is unchanged.
4012 (modref_ref_node::insert_access): Fix updating of
4015 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
4017 * tree-ssa-threadbackward.c (back_threader::resolve_phi): Add
4020 2021-10-14 Richard Biener <rguenther@suse.de>
4022 PR tree-optimization/102659
4023 * tree-if-conv.c (if_convertible_gimple_assign_stmt_p): Also
4024 rewrite pointer typed undefined overflow operations.
4025 (predicate_statements): Likewise. Make sure to emit invariant
4026 conversions in the preheader.
4027 * tree-vectorizer.c (vect_loop_vectorized_call): Look through
4028 non-empty preheaders.
4029 * tree-data-ref.c (dr_analyze_indices): Strip useless
4030 conversions to the MEM_REF base type.
4032 2021-10-14 Martin Liska <mliska@suse.cz>
4034 * common.opt: Stop using AUTODETECT_VALUE
4035 and use EnabledBy where possible.
4036 * opts.c: Enable OPT_fvar_tracking with optimize >= 1.
4037 * toplev.c (AUTODETECT_VALUE): Remove macro.
4038 (process_options): Simplify by using EnabledBy and
4039 OPT_fvar_tracking. Use OPTION_SET_P macro instead of
4042 2021-10-14 Jonathan Wright <jonathan.wright@arm.com>
4044 * config/aarch64/arm_neon.h (vld1_s8_x3): Use signed type for
4046 (vld1_s32_x3): Likewise.
4048 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
4050 PR tree-optimization/102736
4051 PR tree-optimization/102736
4052 * gimple-range-path.cc (path_range_query::range_on_path_entry):
4053 Assert that the requested range is defined outside the path.
4054 (path_range_query::ssa_range_in_phi): Do not call
4055 range_on_path_entry for SSA names that are defined within the
4058 2021-10-14 Iain Sandoe <iain@sandoe.co.uk>
4060 * config/darwin-driver.c (darwin_find_version_from_kernel):
4061 Quote internal identifiers and avoid contractions in
4063 (darwin_default_min_version): Likewise.
4064 (darwin_driver_init): Likewise.
4066 2021-10-14 Martin Jambor <mjambor@suse.cz>
4068 * params.opt (ipa-cp-recursive-freq-factor): New.
4069 * ipa-cp.c (ipcp_value): Switch to inline initialization. New members
4070 scc_no, self_recursion_generated_level, same_scc and
4071 self_recursion_generated_p.
4072 (ipcp_lattice::add_value): Replaced parameter unlimited with
4073 same_lat_gen_level, usit it determine limit of values and store it to
4075 (ipcp_lattice<valtype>::print): Dump the new fileds.
4076 (allocate_and_init_ipcp_value): Take same_lat_gen_level as a new
4077 parameter and store it to the new value.
4078 (self_recursively_generated_p): Removed.
4079 (propagate_vals_across_arith_jfunc): Use self_recursion_generated_p
4080 instead of self_recursively_generated_p, store self generation level
4082 (value_topo_info<valtype>::add_val): Set scc_no.
4083 (value_topo_info<valtype>::propagate_effects): Multiply frequencies of
4084 recursively feeding values and self generated values by appropriate
4087 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
4089 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Remove
4090 redundant test for flag_vect_cost_model.
4092 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
4094 * bitmap.c (debug): New overloaded function for auto_bitmaps.
4095 * bitmap.h (debug): Same.
4097 2021-10-14 Hongyu Wang <hongyu.wang@intel.com>
4099 * config/i386/avx512fp16intrin.h (_mm512_mask_fcmadd_pch):
4100 Adjust builtin call.
4101 (_mm512_mask3_fcmadd_pch): Likewise.
4102 (_mm512_mask_fmadd_pch): Likewise
4103 (_mm512_mask3_fmadd_pch): Likewise
4104 (_mm512_mask_fcmadd_round_pch): Likewise
4105 (_mm512_mask3_fcmadd_round_pch): Likewise
4106 (_mm512_mask_fmadd_round_pch): Likewise
4107 (_mm512_mask3_fmadd_round_pch): Likewise
4108 (_mm_mask_fcmadd_sch): Likewise
4109 (_mm_mask3_fcmadd_sch): Likewise
4110 (_mm_mask_fmadd_sch): Likewise
4111 (_mm_mask3_fmadd_sch): Likewise
4112 (_mm_mask_fcmadd_round_sch): Likewise
4113 (_mm_mask3_fcmadd_round_sch): Likewise
4114 (_mm_mask_fmadd_round_sch): Likewise
4115 (_mm_mask3_fmadd_round_sch): Likewise
4116 (_mm_fcmadd_round_sch): Likewise
4117 * config/i386/avx512fp16vlintrin.h (_mm_mask_fmadd_pch):
4118 Adjust builtin call.
4119 (_mm_mask3_fmadd_pch): Likewise
4120 (_mm256_mask_fmadd_pch): Likewise
4121 (_mm256_mask3_fmadd_pch): Likewise
4122 (_mm_mask_fcmadd_pch): Likewise
4123 (_mm_mask3_fcmadd_pch): Likewise
4124 (_mm256_mask_fcmadd_pch): Likewise
4125 (_mm256_mask3_fcmadd_pch): Likewise
4126 * config/i386/i386-builtin.def: Add mask3 builtin for complex
4127 fma, and adjust mask_builtin to corresponding expander.
4128 * config/i386/i386-expand.c (ix86_expand_round_builtin):
4129 Skip eraseing embedded rounding for expanders that emits
4131 * config/i386/sse.md (complexmove): New mode_attr.
4132 (<avx512>_fmaddc_<mode>_mask1<round_expand_name>): New expander.
4133 (<avx512>_fcmaddc_<mode>_mask1<round_expand_name>): Likewise.
4134 (avx512fp16_fmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
4135 (avx512fp16_fcmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
4136 (avx512fp16_fcmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
4137 (avx512fp16_fmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
4138 * config/i386/subst.md (round_embedded_complex): New subst.
4140 2021-10-14 John David Anglin <danglin@gcc.gnu.org>
4142 * config/pa/pa.md (cbranchsf4): Disable if TARGET_SOFT_FLOAT.
4143 (cbranchdf4): Likewise.
4144 Add missing move patterns for TARGET_SOFT_FLOAT.
4146 2021-10-13 Roger Sayle <roger@nextmovesoftware.com>
4148 * config/i386/i386-expand.c (ix86_expand_vector_move): Use a
4149 pseudo intermediate when moving a SUBREG into a hard register,
4150 by checking ix86_hardreg_mov_ok.
4151 (ix86_expand_vector_extract): Store zero-extended SImode
4152 intermediate in a pseudo, then set target using a SUBREG_PROMOTED
4154 * config/i386/sse.md (mov<VMOVE>_internal): Prevent CSE creating
4155 complex (SUBREG) sets of (vector) hard registers before reload, by
4156 checking ix86_hardreg_mov_ok.
4158 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
4160 * ctfc.h (enum ctf_dtu_d_union_enum): Remove redundant comma.
4162 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
4164 * dwarf2ctf.c (gen_ctf_array_type): Fix typo in comment.
4166 2021-10-13 Martin Sebor <msebor@redhat.com>
4168 PR middle-end/102630
4169 * pointer-query.cc (compute_objsize_r): Handle named address spaces.
4171 2021-10-13 Iain Sandoe <iain@sandoe.co.uk>
4173 * collect2.c (is_lto_object_file): Release simple-object
4174 resources, close files.
4176 2021-10-13 Andre Vieira <andre.simoesdiasvieira@arm.com>
4178 * config/arm/arm.c (thumb2_legitimate_address_p): Use VALID_MVE_MODE
4179 when checking mve addressing modes.
4180 (mve_vector_mem_operand): Fix the way we handle pre, post and offset
4182 (arm_print_operand): Fix printing of POST_ and PRE_MODIFY.
4183 * config/arm/mve.md: Use mve_memory_operand predicate everywhere where
4184 there is a single Ux constraint.
4186 2021-10-13 John David Anglin <danglin@gcc.gnu.org>
4188 * config/pa/pa.md (muldi3): Add support for inlining 64-bit
4189 multiplication on 32-bit PA 1.1 and 2.0 targets.
4191 2021-10-13 Martin Liska <mliska@suse.cz>
4193 PR gcov-profile/90364
4194 * coverage.c (build_info): Emit checksum to the global variable.
4195 (build_info_type): Add new field for checksum.
4196 (coverage_obj_finish): Pass object_checksum.
4197 (coverage_init): Use 0 as checksum for .gcno files.
4198 * gcov-dump.c (dump_gcov_file): Dump also new checksum field.
4199 * gcov.c (read_graph_file): Read also checksum.
4200 * doc/invoke.texi: Document the behaviour change.
4202 2021-10-13 Richard Biener <rguenther@suse.de>
4204 * gimple-iterator.h (gsi_iterator_update): Add GSI_LAST_NEW_STMT,
4205 start at integer value 2.
4206 * gimple-iterator.c (gsi_insert_seq_nodes_before): Update
4207 the iterator for GSI_LAST_NEW_STMT.
4208 (gsi_insert_seq_nodes_after): Likewise.
4209 * tree-if-conv.c (predicate_statements): Use GSI_LAST_NEW_STMT.
4210 * tree-ssa.c (execute_update_addresses_taken): Correct bogus
4211 arguments to gsi_replace.
4213 2021-10-13 Martin Liska <mliska@suse.cz>
4216 * common.opt: Use EnabledBy instead of detection in
4217 finish_options and process_options.
4218 * opts.c (finish_options): Remove handling of
4219 x_flag_unroll_all_loops.
4220 * toplev.c (process_options): Likewise for flag_web and
4221 flag_rename_registers.
4223 2021-10-13 Richard Biener <rguenther@suse.de>
4225 PR tree-optimization/102659
4226 * tree-if-conv.c (need_to_rewrite_undefined): New flag.
4227 (if_convertible_gimple_assign_stmt_p): Mark the loop for
4228 rewrite when stmts with undefined behavior on integer
4230 (combine_blocks): Predicate also when we need to rewrite stmts.
4231 (predicate_statements): Rewrite affected stmts to something
4232 with well-defined behavior on overflow.
4233 (tree_if_conversion): Initialize need_to_rewrite_undefined.
4235 2021-10-13 Richard Biener <rguenther@suse.de>
4238 * ipa-sra.c (ptr_parm_has_nonarg_uses): Fix volatileness
4241 2021-10-13 Jose E. Marchesi <jose.marchesi@oracle.com>
4243 * dwarf2ctf.c: Fix typo in comment.
4245 2021-10-12 Bill Schmidt <wschmidt@linux.ibm.com>
4248 * config/rs6000/altivec.h (vec_cpsgn): Swap operand order.
4249 * config/rs6000/rs6000-overload.def (VEC_COPYSIGN): Use SKIP to
4250 avoid generating an automatic #define of vec_cpsgn. Use the
4251 correct built-in for V4SFmode that doesn't depend on VSX.
4253 2021-10-12 Uroš Bizjak <ubizjak@gmail.com>
4257 * config/i386/i386.md (*add<mode>_1_slp): Rewrite as
4258 define_insn_and_split pattern. Add alternative 1 and split it
4259 post reload to insert operand 1 into the low part of operand 0.
4260 (*sub<mode>_1_slp): Ditto.
4261 (*and<mode>_1_slp): Ditto.
4262 (*<any_or:code><mode>_1_slp): Ditto.
4263 (*ashl<mode>3_1_slp): Ditto.
4264 (*<any_shiftrt:insn><mode>3_1_slp): Ditto.
4265 (*<any_rotate:insn><mode>3_1_slp): Ditto.
4266 (*neg<mode>_1_slp): New insn_and_split pattern.
4267 (*one_cmpl<mode>_1_slp): Ditto.
4269 2021-10-12 David Edelsohn <dje.gcc@gmail.com>
4271 * doc/install.texi: Update MinGW and mingw-64 Binaries
4274 2021-10-12 Daniel Le Duc Khoi Nguyen <greenrecyclebin@gmail.com>
4276 * doc/extend.texi (Common Variable Attributes): Fix typos in
4277 alloc_size documentation.
4279 2021-10-12 Richard Biener <rguenther@suse.de>
4281 PR tree-optimization/102696
4282 * tree-vect-slp.c (vect_build_slp_tree_2): Properly mark
4283 the tree fatally failed when we reject a BIT_FIELD_REF.
4285 2021-10-12 Richard Biener <rguenther@suse.de>
4287 PR tree-optimization/102572
4288 * tree-vect-stmts.c (vect_build_gather_load_calls): When
4289 gathering the vectorized defs for the mask pass in the
4290 desired mask vector type so invariants will be handled
4293 2021-10-12 Tamar Christina <tamar.christina@arm.com>
4295 * config/aarch64/aarch64-sve.md (*fcm<cmp_op><mode>_bic_combine,
4296 *fcm<cmp_op><mode>_nor_combine, *fcmuo<mode>_bic_combine,
4297 *fcmuo<mode>_nor_combine): New.
4299 2021-10-12 Eric Botcazou <ebotcazou@adacore.com>
4302 * config/sparc/sparc-modes.def (OI): New integer mode.
4304 2021-10-12 Jakub Jelinek <jakub@redhat.com>
4306 * gimple-fold.h (clear_padding_type_may_have_padding_p): Declare.
4307 * gimple-fold.c (clear_padding_type_may_have_padding_p): No longer
4310 2021-10-12 Jakub Jelinek <jakub@redhat.com>
4312 * tree-vectorizer.h (loop_cost_model): New function.
4313 (unlimited_cost_model): Use it.
4314 * tree-vect-loop.c (vect_analyze_loop_costing): Use loop_cost_model
4315 call instead of flag_vect_cost_model.
4316 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
4317 (vect_prune_runtime_alias_test_list): Likewise. Also use it instead
4318 of flag_simd_cost_model.
4320 2021-10-12 liuhongt <hongtao.liu@intel.com>
4323 * config/i386/i386-expand.c (emit_reduc_half): Handle
4325 * config/i386/mmx.md (reduc_<code>_scal_v4qi): New expander.
4326 (reduc_plus_scal_v4qi): Ditto.
4328 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
4330 * config/rs6000/smmintrin.h (_mm_cmpeq_epi64, _mm_cmpgt_epi64,
4331 _mm_mullo_epi32, _mm_mul_epi32, _mm_packus_epi32): New.
4332 * config/rs6000/nmmintrin.h: Copy from i386, tweak to suit.
4334 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
4336 * config/rs6000/smmintrin.h (_mm_cvtepi8_epi16, _mm_cvtepi8_epi32,
4337 _mm_cvtepi8_epi64, _mm_cvtepi16_epi32, _mm_cvtepi16_epi64,
4338 _mm_cvtepi32_epi64, _mm_cvtepu8_epi16, _mm_cvtepu8_epi32,
4339 _mm_cvtepu8_epi64, _mm_cvtepu16_epi32, _mm_cvtepu16_epi64,
4340 _mm_cvtepu32_epi64): New.
4342 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
4344 * config/rs6000/smmintrin.h (_mm_test_all_zeros,
4345 _mm_test_all_ones, _mm_test_mix_ones_zeros): Rewrite as macro.
4347 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
4349 * config/rs6000/smmintrin.h (_mm_min_epi8, _mm_min_epu16,
4350 _mm_min_epi32, _mm_min_epu32, _mm_max_epi8, _mm_max_epu16,
4351 _mm_max_epi32, _mm_max_epu32): New.
4353 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
4355 * ipa-modref-tree.h (struct modref_access_node): Revert
4357 (struct modref_ref_node): Likewise.
4359 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
4361 * ipa-modref-tree.h (modref_tree::global_access_p): New member
4364 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
4365 ignore_stores_eaf_flags): Move to ipa-modref.h
4366 (remove_useless_eaf_flags): Remove early exit on NOCLOBBER.
4367 (modref_summary::global_memory_read_p): New member function.
4368 (modref_summary::global_memory_written_p): New member function.
4369 * ipa-modref.h (modref_summary::global_memory_read_p,
4370 modref_summary::global_memory_written_p): Declare.
4371 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
4372 ignore_stores_eaf_flags): move here.
4373 * tree-ssa-structalias.c: Include ipa-modref-tree.h, ipa-modref.h
4375 (handle_rhs_call): Rewrite.
4376 (handle_call_arg): New function.
4377 (determine_global_memory_access): New function.
4378 (handle_const_call): Remove
4379 (handle_pure_call): Remove
4380 (find_func_aliases_for_call): Update use of handle_rhs_call.
4381 (compute_points_to_sets): Handle global memory acccesses
4384 2021-10-11 Diane Meirowitz <diane.meirowitz@oracle.com>
4386 * doc/invoke.texi: Add link to UndefinedBehaviorSanitizer
4387 documentation, mention UBSAN_OPTIONS, similar to what is done
4388 for AddressSanitizer.
4390 2021-10-11 Richard Biener <rguenther@suse.de>
4392 PR middle-end/102683
4393 * internal-fn.c (expand_DEFERRED_INIT): Check for mode
4394 availability before building an integer type for storage
4397 2021-10-11 Richard Biener <rguenther@suse.de>
4399 PR middle-end/101480
4400 * gimple.c (gimple_call_fnspec): Do not mark operator new/delete
4403 2021-10-11 Martin Liska <mliska@suse.cz>
4405 * common.opt: Remove Init(2) for some options.
4406 * toplev.c (process_options): Do not use AUTODETECT_VALUE, but
4407 use rather OPTION_SET_P.
4409 2021-10-11 Martin Liska <mliska@suse.cz>
4411 * common.opt: Remove usage of IRA_REGION_AUTODETECT.
4412 * flag-types.h (enum ira_region): Likewise.
4413 * toplev.c (process_options): Use OPTION_SET_P instead of
4414 IRA_REGION_AUTODETECT.
4416 2021-10-11 Jakub Jelinek <jakub@redhat.com>
4418 * omp-low.c (omp_runtime_api_call): Handle omp_get_max_teams,
4419 omp_[sg]et_teams_thread_limit and omp_set_num_teams.
4421 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
4423 * config/s390/s390-protos.h (s390_rawmemchr): Add prototype.
4424 * config/s390/s390.c (s390_rawmemchr): New function.
4425 * config/s390/s390.md (rawmemchr<SINT:mode>): New expander.
4426 * config/s390/vector.md (@vec_vfees<mode>): Basically a copy of
4427 the pattern vfees<mode> from vx-builtins.md.
4428 * config/s390/vx-builtins.md (*vfees<mode>): Remove.
4430 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
4432 * builtins.c (get_memory_rtx): Change to external linkage.
4433 * builtins.h (get_memory_rtx): Add function prototype.
4434 * doc/md.texi (rawmemchr<mode>): Document.
4435 * internal-fn.c (expand_RAWMEMCHR): Define.
4436 * internal-fn.def (RAWMEMCHR): Add.
4437 * optabs.def (rawmemchr_optab): Add.
4438 * tree-loop-distribution.c (find_single_drs): Change return code
4439 behaviour by also returning true if no single store was found
4441 (loop_distribution::classify_partition): Respect the new return
4442 code behaviour of function find_single_drs.
4443 (loop_distribution::execute): Call new function
4444 transform_reduction_loop in order to replace rawmemchr or strlen
4445 like loops by calls into builtins.
4446 (generate_reduction_builtin_1): New function.
4447 (generate_rawmemchr_builtin): New function.
4448 (generate_strlen_builtin_1): New function.
4449 (generate_strlen_builtin): New function.
4450 (generate_strlen_builtin_using_rawmemchr): New function.
4451 (reduction_var_overflows_first): New function.
4452 (determine_reduction_stmt_1): New function.
4453 (determine_reduction_stmt): New function.
4454 (loop_distribution::transform_reduction_loop): New function.
4456 2021-10-11 Martin Liska <mliska@suse.cz>
4458 * tree.c (cl_option_hasher::hash): Use cl_optimization_hash
4459 and remove legacy hashing code.
4461 2021-10-11 Kito Cheng <kito.cheng@sifive.com>
4464 * builtins.c (maybe_emit_call_builtin___clear_cache): Allow
4465 CONST_INT for BEGIN and END, and use gcc_assert rather than
4468 2021-10-10 Jakub Jelinek <jakub@redhat.com>
4471 * var-tracking.c (add_stores): For cselib_sp_derived_value_p values
4472 use MO_VAL_SET if loc is not sp.
4474 2021-10-10 Andrew Pinski <apinski@marvell.com>
4476 PR tree-optimization/102622
4477 * match.pd: Swap the order of a?pow2cst:0 and a?-1:0 transformations.
4478 Swap the order of a?0:pow2cst and a?0:-1 transformations.
4480 2021-10-09 liuhongt <hongtao.liu@intel.com>
4483 * config/i386/i386-expand.c (ix86_valid_mask_cmp_mode): Handle
4485 (ix86_use_mask_cmp_p): Ditto.
4486 (ix86_expand_sse_movcc): Ditto.
4487 * config/i386/i386.md (setcc_hf_mask): New define_insn.
4488 (movhf_mask): Ditto.
4489 (UNSPEC_MOVCC_MASK): New unspec.
4490 * config/i386/sse.md (UNSPEC_PCMP): Move to i386.md.
4492 2021-10-08 Vladimir N. Makarov <vmakarov@redhat.com>
4494 PR rtl-optimization/102627
4495 * lra-constraints.c (split_reg): Use at least natural mode of hard reg.
4497 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
4499 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Grow
4502 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
4504 * value-range.cc (irange::debug): New.
4505 * value-range.h (irange::debug): New.
4507 2021-10-08 Richard Sandiford <richard.sandiford@arm.com>
4509 PR tree-optimization/102385
4510 * predict.h (change_edge_frequency): Declare.
4511 * predict.c (change_edge_frequency): New function.
4512 * tree-ssa-loop-manip.h (tree_transform_and_unroll_loop): Remove
4514 (tree_unroll_loop): Likewise.
4515 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Update accordingly.
4516 * tree-predcom.c (pcom_worker::tree_predictive_commoning_loop):
4518 * tree-ssa-loop-prefetch.c (loop_prefetch_arrays): Likewise.
4519 * tree-ssa-loop-manip.c (tree_unroll_loop): Likewise.
4520 (tree_transform_and_unroll_loop): Likewise. Use single_dom_exit
4521 to retrieve the exit edges. Make all the old profile update code
4522 conditional on !single_loop_p -- the case it was written for --
4523 and use a different approach for the single-loop case.
4525 2021-10-08 Martin Liska <mliska@suse.cz>
4527 * config/alpha/alpha.c (alpha_option_override): Use new macro
4529 * config/arc/arc.c (arc_override_options): Likewise.
4530 * config/arm/arm.c (arm_option_override): Likewise.
4531 * config/bfin/bfin.c (bfin_load_pic_reg): Likewise.
4532 * config/c6x/c6x.c (c6x_option_override): Likewise.
4533 * config/csky/csky.c: Likewise.
4534 * config/darwin.c (darwin_override_options): Likewise.
4535 * config/frv/frv.c (frv_option_override): Likewise.
4536 * config/i386/djgpp.h: Likewise.
4537 * config/i386/i386.c (ix86_stack_protect_guard): Likewise.
4538 (ix86_max_noce_ifcvt_seq_cost): Likewise.
4539 * config/ia64/ia64.c (ia64_option_override): Likewise.
4540 (ia64_override_options_after_change): Likewise.
4541 * config/m32c/m32c.c (m32c_option_override): Likewise.
4542 * config/m32r/m32r.c (m32r_init): Likewise.
4543 * config/m68k/m68k.c (m68k_option_override): Likewise.
4544 * config/microblaze/microblaze.c (microblaze_option_override): Likewise.
4545 * config/mips/mips.c (mips_option_override): Likewise.
4546 * config/nios2/nios2.c (nios2_option_override): Likewise.
4547 * config/nvptx/nvptx.c (nvptx_option_override): Likewise.
4548 * config/pa/pa.c (pa_option_override): Likewise.
4549 * config/riscv/riscv.c (riscv_option_override): Likewise.
4550 * config/rs6000/aix71.h: Likewise.
4551 * config/rs6000/aix72.h: Likewise.
4552 * config/rs6000/aix73.h: Likewise.
4553 * config/rs6000/rs6000.c (darwin_rs6000_override_options): Likewise.
4554 (rs6000_override_options_after_change): Likewise.
4555 (rs6000_linux64_override_options): Likewise.
4556 (glibc_supports_ieee_128bit): Likewise.
4557 (rs6000_option_override_internal): Likewise.
4558 (rs6000_file_start): Likewise.
4559 (rs6000_darwin_file_start): Likewise.
4560 * config/rs6000/rtems.h: Likewise.
4561 * config/rs6000/sysv4.h: Likewise.
4562 * config/rs6000/vxworks.h (SUB3TARGET_OVERRIDE_OPTIONS): Likewise.
4563 * config/s390/s390.c (s390_option_override): Likewise.
4564 * config/sh/linux.h: Likewise.
4565 * config/sh/netbsd-elf.h (while): Likewise.
4566 * config/sh/sh.c (sh_option_override): Likewise.
4567 * config/sol2.c (solaris_override_options): Likewise.
4568 * config/sparc/sparc.c (sparc_option_override): Likewise.
4569 * config/tilegx/tilegx.c (tilegx_option_override): Likewise.
4570 * config/visium/visium.c (visium_option_override): Likewise.
4571 * config/vxworks.c (vxworks_override_options): Likewise.
4572 * lto-opts.c (lto_write_options): Likewise.
4573 * omp-expand.c (expand_omp_simd): Likewise.
4574 * omp-general.c (omp_max_vf): Likewise.
4575 * omp-offload.c (oacc_xform_loop): Likewise.
4576 * opts.h (OPTION_SET_P): Likewise.
4577 * targhooks.c (default_max_noce_ifcvt_seq_cost): Likewise.
4578 * toplev.c (process_options): Likewise.
4579 * tree-predcom.c: Likewise.
4580 * tree-sra.c (analyze_all_variable_accesses): Likewise.
4582 2021-10-08 liuhongt <hongtao.liu@intel.com>
4585 * config/i386/i386.c (ix86_optab_supported_p):
4586 Return true for HFmode.
4587 * match.pd: Simplify (_Float16) ceil ((double) x) to
4588 __builtin_ceilf16 (a) when a is _Float16 type and
4589 direct_internal_fn_supported_p.
4591 2021-10-08 liuhongt <hongtao.liu@intel.com>
4594 * config/i386/i386-expand.c (emit_reduc_half): Hanlde V4HImode.
4595 * config/i386/mmx.md (reduc_plus_scal_v4hi): New.
4596 (reduc_<code>_scal_v4hi): New.
4598 2021-10-08 liuhongt <hongtao.liu@intel.com>
4600 * common.opt (ftree-vectorize): Add Var(flag_tree_vectorize).
4601 * doc/invoke.texi (Options That Control Optimization): Update
4603 * opts.c (default_options_table): Enable auto-vectorization at
4604 O2 with very-cheap cost model.
4605 (finish_options): Use cheap cost model for
4606 explicit -ftree{,-loop}-vectorize.
4608 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
4610 * ctfc.c (ctfc_delete_container): Free hash table contents.
4612 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
4614 * toplev.c (process_options): Do not warn for GNU GIMPLE.
4616 2021-10-07 Siddhesh Poyarekar <siddhesh@gotplt.org>
4618 * tree-object-size.c (addr_object_size,
4619 compute_builtin_object_size): Drop PDECL and POFF arguments.
4620 (addr_object_size): Adjust calls.
4621 * tree-object-size.h (compute_builtin_object_size): Drop PDECL
4624 2021-10-07 Roger Sayle <roger@nextmovesoftware.com>
4626 * rtl.def (SMUL_HIGHPART, UMUL_HIGHPART): New RTX codes for
4627 representing signed and unsigned high-part multiplication resp.
4628 * simplify-rtx.c (simplify_binary_operation_1) [SMUL_HIGHPART,
4629 UMUL_HIGHPART]: Simplify high-part multiplications by zero.
4630 [SS_PLUS, US_PLUS, SS_MINUS, US_MINUS, SS_MULT, US_MULT,
4631 SS_DIV, US_DIV]: Similar simplifications for saturating
4633 (simplify_const_binary_operation) [SS_PLUS, US_PLUS, SS_MINUS,
4634 US_MINUS, SS_MULT, US_MULT, SMUL_HIGHPART, UMUL_HIGHPART]:
4635 Implement compile-time evaluation for constant operands.
4636 * dwarf2out.c (mem_loc_descriptor): Skip SMUL_HIGHPART and
4638 * doc/rtl.texi (smul_highpart, umul_highpart): Document RTX codes.
4639 * doc/md.texi (smul@var{m}3_highpart, umul@var{m3}_highpart):
4640 Mention the new smul_highpart and umul_highpart RTX codes.
4641 * doc/invoke.texi: Silence @xref "compilation" warnings.
4643 2021-10-07 Martin Jambor <mjambor@suse.cz>
4646 * ipa-prop.c (ipa_edge_args_sum_t::duplicate): Also handle the
4647 case when the source reference description corresponds to a
4648 referance taken in a function src->caller is inlined to.
4650 2021-10-07 Jan Hubicka <hubicka@ucw.cz>
4653 * ipa-modref-tree.h (modref_access_node::contains_p): Handle offsets
4655 (modref_access_node::try_merge_with): Add sanity check that there
4656 are no redundant entries in the list.
4658 2021-10-07 Richard Biener <rguenther@suse.de>
4660 PR tree-optimization/102608
4661 * tree-ssa-sccvn.c (visit_stmt): Drop .DEFERRED_INIT to
4664 2021-10-07 Martin Liska <mliska@suse.cz>
4666 * toplev.c (toplev::main): Make
4667 save_opt_decoded_options a pointer type
4668 * toplev.h: Likewise.
4670 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
4672 * config/gcn/gcn-valu.md (gather<mode>_insn_2offsets<exec>): Apply
4673 HAVE_GCN_ASM_GLOBAL_LOAD_FIXED.
4674 (scatter<mode>_insn_2offsets<exec_scatter>): Likewise.
4676 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
4678 * config/gcn/gcn-hsa.h (SRAMOPT): Include the whole option string.
4679 Adjust for new -msram-ecc=any behaviour.
4680 (ASM_SPEC): Adjust -mxnack and -msram-ecc usage.
4681 * config/gcn/gcn.c (output_file_start): Implement -msram-ecc=any.
4682 * config/gcn/mkoffload.c (EF_AMDGPU_XNACK): Rename to ...
4683 (EF_AMDGPU_XNACK_V3): ... this.
4684 (EF_AMDGPU_SRAM_ECC): Rename to ...
4685 (EF_AMDGPU_SRAM_ECC_V3): ... this.
4686 (EF_AMDGPU_FEATURE_XNACK_V4): New.
4687 (EF_AMDGPU_FEATURE_XNACK_UNSUPPORTED_V4): New.
4688 (EF_AMDGPU_FEATURE_XNACK_ANY_V4): New.
4689 (EF_AMDGPU_FEATURE_XNACK_OFF_V4): New.
4690 (EF_AMDGPU_FEATURE_XNACK_ON_V4): New.
4691 (EF_AMDGPU_FEATURE_SRAMECC_V4): New.
4692 (EF_AMDGPU_FEATURE_SRAMECC_UNSUPPORTED_V4): New.
4693 (EF_AMDGPU_FEATURE_SRAMECC_ANY_V4): New.
4694 (EF_AMDGPU_FEATURE_SRAMECC_OFF_V4): New.
4695 (EF_AMDGPU_FEATURE_SRAMECC_ON_V4): New.
4696 (SET_XNACK_ON): New.
4697 (SET_XNACK_OFF): New.
4699 (SET_SRAM_ECC_ON): New.
4700 (SET_SRAM_ECC_ANY): New.
4701 (SET_SRAM_ECC_OFF): New.
4702 (TEST_SRAM_ECC_ANY): New.
4703 (TEST_SRAM_ECC_ON): New.
4704 (main): Implement HSACOv4 and -msram-ecc=any.
4706 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
4708 * config.in: Regenerate.
4709 * config/gcn/gcn-hsa.h (X_FIJI): New macro.
4713 (A_FIJI): Rename to ...
4715 (A_900): Rename to ...
4717 (A_906): Rename to ...
4719 (A_908): Rename to ...
4721 (SRAMOPT): New macro.
4722 (ASM_SPEC): Adjust xnack option usage.
4723 * config/gcn/gcn.c (output_file_start): Adjust amdgcn_target usage.
4724 * configure: Regenerate.
4725 * configure.ac: Detect LLVM assembler dialect.
4727 2021-10-07 Richard Biener <rguenther@suse.de>
4729 * tree-pretty-print.c (dump_generic_node): Do not elide
4730 printing '&' when dumping with -gimple.
4732 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
4734 * gimple-range-cache.cc (non_null_ref::adjust_range): Call new
4736 * gimple-range-fold.cc (adjust_pointer_diff_expr): Ditto.
4737 (adjust_imagpart_expr): Ditto.
4738 * value-range.cc (irange::irange_intersect): Call new routine if
4739 RHS is a single pair.
4740 (irange::intersect): New wide_int version.
4741 * value-range.h (class irange): New prototype.
4743 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
4745 * gimple-range-edge.cc (gimple_outgoing_range::gimple_outgoing_range):
4746 Add parameter to limit size when recognizing switches.
4747 (gimple_outgoing_range::edge_range_p): Check size limit.
4748 * gimple-range-edge.h (gimple_outgoing_range): Add size field.
4749 * gimple-range-gori.cc (gori_map::calculate_gori): Ignore switches
4750 that exceed the size limit.
4751 (gori_compute::gori_compute): Add initializer.
4752 * params.opt (evrp-switch-limit): New.
4753 * doc/invoke.texi: Update docs.
4755 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
4757 * value-range.h (irange::set_varying): Use TYPE_MIN_VALUE and
4758 TYPE_MAX_VALUE instead of creating new trees when possible.
4760 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
4762 * gimple-range-cache.cc (non_null_ref::adjust_range): Check for
4763 zero and non-zero more efficently.
4765 2021-10-06 Richard Biener <rguenther@suse.de>
4768 * dumpfile.h (TDF_GIMPLE_VAL): New.
4769 (dump_flag): Re-order and adjust TDF_* flags. Make
4770 the enum uint32_t. Use std::underlying_type in the
4772 (optgroup_flag): Likewise for the operator overloads.
4773 * tree-pretty-print.c (dump_generic_node): Wrap ADDR_EXPR
4774 in _Literal if TDF_GIMPLE_VAL.
4775 * gimple-pretty-print.c (dump_gimple_assign): Add
4776 TDF_GIMPLE_VAL to flags when dumping operands where only
4777 is_gimple_val are allowed.
4778 (dump_gimple_cond): Likewise.
4780 2021-10-06 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
4782 * gimple-isel.cc (gimple_expand_vec_cond_expr): Remove redundant if
4785 2021-10-05 qing zhao <qing.zhao@oracle.com>
4787 PR middle-end/102359
4788 * gimplify.c (gimplify_decl_expr): Not add initialization for an
4789 auto variable when it has been initialized by frontend.
4791 2021-10-05 Aldy Hernandez <aldyh@redhat.com>
4793 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
4796 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
4798 * common/config/avr/avr-common.c (avr_handle_option): Mark
4799 argument as ATTRIBUTE_UNUSED.
4801 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
4803 * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC):
4804 Undefine before redefinition.
4806 2021-10-05 Richard Biener <rguenther@suse.de>
4808 * toplev.c (no_backend): Remove global var.
4809 (process_options): Pass in no_backend, move post_options
4810 langhook call to toplev::main.
4811 (do_compile): Pass in no_backend, move process_options call
4813 (toplev::run_self_tests): Check no_backend at the caller.
4814 (toplev::main): Call post_options and process_options
4815 split out from do_compile, do self-tests only if
4816 no_backend is initialized.
4818 2021-10-05 Richard Biener <rguenther@suse.de>
4820 * tree-cfg.c (dump_function_to_file): Dump the UID of the
4821 function as part of the name when requested.
4822 * tree-pretty-print.c (dump_function_name): Dump the UID when
4823 requested and the langhook produced the actual name.
4825 2021-10-05 Richard Biener <rguenther@suse.de>
4827 PR middle-end/102587
4828 PR middle-end/102285
4829 * internal-fn.c (expand_DEFERRED_INIT): Fall back to
4830 zero-initialization as last resort, use the constant
4831 size as given by the DEFERRED_INIT argument to build
4834 2021-10-04 Marek Polacek <polacek@redhat.com>
4837 * doc/invoke.texi: Document -Warray-compare.
4839 2021-10-04 Richard Biener <rguenther@suse.de>
4841 * gimplify.c (is_var_need_auto_init): DECL_HARD_REGISTER
4842 variables are not to be initialized.
4844 2021-10-04 Richard Biener <rguenther@suse.de>
4846 * expr.h (non_mem_decl_p): Declare.
4847 (mem_ref_refers_to_non_mem_p): Likewise.
4848 * expr.c (non_mem_decl_p): Export.
4849 (mem_ref_refers_to_non_mem_p): Likewise.
4850 * internal-fn.c (expand_DEFERRED_INIT): Do not expand the LHS
4851 but check the base with mem_ref_refers_to_non_mem_p
4854 2021-10-04 Richard Biener <rguenther@suse.de>
4856 PR tree-optimization/102570
4857 * tree-ssa-sccvn.h (vn_reference_op_struct): Document
4858 we are using clique for the internal function code.
4859 * tree-ssa-sccvn.c (vn_reference_op_eq): Compare the
4860 internal function code.
4861 (print_vn_reference_ops): Print the internal function code.
4862 (vn_reference_op_compute_hash): Hash it.
4863 (copy_reference_ops_from_call): Record it.
4864 (visit_stmt): Remove the restriction around internal function
4866 (fully_constant_vn_reference_p): Use fold_const_call and handle
4868 (vn_reference_eq): Compare call return types.
4869 * tree-ssa-pre.c (create_expression_by_pieces): Handle
4870 generating calls to internal functions.
4871 (compute_avail): Remove the restriction around internal function
4874 2021-10-04 Aldy Hernandez <aldyh@redhat.com>
4876 PR tree-optimization/102560
4877 * gimple-ssa-warn-alloca.c (alloca_call_type): Remove static
4878 marker for invalid_range.
4880 2021-10-04 Richard Biener <rguenther@suse.de>
4882 PR middle-end/102587
4883 * internal-fn.c (expand_DEFERRED_INIT): Guard register
4884 initialization path an avoid initializing VLA registers
4887 2021-10-04 Eric Botcazou <ebotcazou@adacore.com>
4889 * config/rs6000/vxworks.h (TARGET_INIT_LIBFUNCS): Delete.
4891 2021-10-03 Martin Liska <mliska@suse.cz>
4893 * toplev.c (toplev::main): Check opt_index if it is a part
4896 2021-10-02 Aldy Hernandez <aldyh@redhat.com>
4898 PR tree-optimization/102563
4899 * range-op.cc (operator_lshift::op1_range): Do not clobber
4902 2021-10-02 Martin Liska <mliska@suse.cz>
4904 * toplev.c (toplev::main): save_decoded_options[0] is program
4905 name and so it should be skipped.
4907 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
4909 PR tree-optimization/102546
4910 * range-op.cc (operator_lshift::op1_range): Teach range-ops that
4911 X << Y is non-zero implies X is also non-zero.
4913 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
4915 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
4917 * config/aarch64/aarch64-tune.md: Regenerate.
4918 * doc/invoke.texi: Update docs.
4920 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
4922 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
4924 * config/aarch64/aarch64-tune.md: Regenerate.
4925 * doc/invoke.texi: Update docs.
4927 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
4929 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
4931 * config/aarch64/aarch64-tune.md: Regenerate.
4932 * doc/invoke.texi: Update docs.
4934 2021-10-01 Martin Sebor <msebor@redhat.com>
4937 * doc/invoke.texi (-Waddress): Update.
4938 * gengtype.c (write_types): Avoid -Waddress.
4939 * poly-int.h (POLY_SET_COEFF): Avoid using null.
4941 2021-10-01 John David Anglin <danglin@gcc.gnu.org>
4944 * config/pa/pa.c (pa_option_override): Default to dwarf version 4
4947 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
4949 * config/aarch64/aarch64.h (AARCH64_FL_V9): Update value.
4951 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
4953 * gimple-range-path.cc (path_range_query::compute_ranges): Use
4955 * gimple-range-path.h (class path_range_query): Remove shadowed
4957 (path_range_query::get_path_oracle): New.
4959 2021-10-01 Jakub Jelinek <jakub@redhat.com>
4960 Richard Biener <rguenther@suse.de>
4963 * doc/invoke.texi (-fsanitize=integer-divide-by-zero): Remove
4964 INT_MIN / -1 division detection from here ...
4965 (-fsanitize=signed-integer-overflow): ... and add it here.
4967 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
4969 * config/aarch64/aarch64-arches.def (AARCH64_ARCH): Added
4971 * config/aarch64/aarch64.h (AARCH64_FL_V9): New.
4972 (AARCH64_FL_FOR_ARCH9): New flags for Armv9-A.
4973 (AARCH64_ISA_V9): New ISA flag.
4974 * doc/invoke.texi: Update docs.
4976 2021-10-01 Martin Liska <mliska@suse.cz>
4978 * toplev.c (toplev::main): Save decoded optimization options.
4979 * toplev.h (save_opt_decoded_options): New.
4980 * doc/extend.texi: Be more clear about optimize and target
4983 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
4985 * explow.c: Include langhooks.h.
4986 (set_stack_check_libfunc): Build a proper function type.
4988 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
4991 * config/i386/i386.c (legitimate_pic_address_disp_p): For PE-COFF do
4992 not return true for external weak function symbols in medium model.
4994 2021-10-01 Jakub Jelinek <jakub@redhat.com>
4996 * tree.h (OMP_CLAUSE_ORDER_REPRODUCIBLE): Define.
4997 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_ORDER>: Print
4998 reproducible: for OMP_CLAUSE_ORDER_REPRODUCIBLE.
4999 * omp-general.c (omp_extract_for_data): If OMP_CLAUSE_ORDER is seen
5000 without OMP_CLAUSE_ORDER_UNCONSTRAINED, overwrite sched_kind to
5001 OMP_CLAUSE_SCHEDULE_STATIC.
5003 2021-10-01 Richard Biener <rguenther@suse.de>
5005 PR middle-end/102518
5006 * tree-inline.c (setup_one_parameter): Avoid substituting
5007 an invariant into contexts where a GIMPLE register is not valid.
5009 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
5011 * config/arm/arm-cpus.in: Add Cortex-R52+ CPU.
5012 * config/arm/arm-tables.opt: Regenerate.
5013 * config/arm/arm-tune.md: Regenerate.
5014 * doc/invoke.texi: Update docs.
5016 2021-09-30 Uroš Bizjak <ubizjak@gmail.com>
5019 * config/i386/i386.md
5020 (sign_extend:WIDE (any_logic:NARROW (memory, immediate)) splitters):
5023 2021-09-30 Tobias Burnus <tobias@codesourcery.com>
5025 * omp-low.c (omp_runtime_api_call): Add omp_aligned_{,c}alloc and
5026 omp_{c,re}alloc, fix omp_alloc/omp_free.
5028 2021-09-30 Martin Liska <mliska@suse.cz>
5030 * defaults.h (ASM_OUTPUT_ASCII): Do not hide global variable
5031 asm_out_file and stream directly to MYFILE.
5033 2021-09-30 Richard Biener <rguenther@suse.de>
5035 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
5036 Restore and fix condition under which we apply npeel to
5037 the DRs misalignment value.
5039 2021-09-30 Richard Biener <rguenther@suse.de>
5041 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
5042 Fix npeel check for variable amount of peeling.
5044 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
5046 * lto-wrapper.c (run_gcc): Plug snprintf overflow.
5048 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
5050 * gimple-range.cc (gimple_ranger::debug): New.
5051 * gimple-range.h (class gimple_ranger): Add debug.
5053 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
5055 PR middle-end/102519
5056 * tree-vrp.c (hybrid_threader::~hybrid_threader): Free m_query.
5058 2021-09-29 Indu Bhagat <indu.bhagat@oracle.com>
5061 * btfout.c (GTY): Add GTY (()) albeit for cosmetic only purpose.
5062 (btf_finalize): Empty the hash_map btf_var_ids.
5064 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
5066 * tree-vrp.c (thread_through_all_blocks): Return bool.
5067 (execute_vrp_threader): Return TODO_* flags.
5068 (pass_data_vrp_threader): Set todo_flags_finish to 0.
5070 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
5072 * timevar.def (TV_TREE_VRP_THREADER): New.
5073 * tree-vrp.c: Use TV_TREE_VRP_THREADER for VRP threader pass.
5075 2021-09-29 David Faust <david.faust@oracle.com>
5077 * config.gcc (bpf-*-*): Do not overwrite extra_headers.
5079 2021-09-29 Jonathan Wright <jonathan.wright@arm.com>
5081 * config/aarch64/aarch64-builtins.c (TYPES_BINOP_PPU): Define
5082 new type qualifier enum.
5083 (TYPES_TERNOP_SSSU): Likewise.
5084 (TYPES_TERNOP_PPPU): Likewise.
5085 * config/aarch64/aarch64-simd-builtins.def: Define PPU, SSU,
5086 PPPU and SSSU builtin generator macros for qtbl1 and qtbx1
5088 * config/aarch64/arm_neon.h (vqtbl1_p8): Use type-qualified
5089 builtin and remove casts.
5090 (vqtbl1_s8): Likewise.
5091 (vqtbl1q_p8): Likewise.
5092 (vqtbl1q_s8): Likewise.
5093 (vqtbx1_s8): Likewise.
5094 (vqtbx1_p8): Likewise.
5095 (vqtbx1q_s8): Likewise.
5096 (vqtbx1q_p8): Likewise.
5097 (vtbl1_p8): Likewise.
5098 (vtbl2_p8): Likewise.
5099 (vtbx2_p8): Likewise.
5101 2021-09-29 Richard Biener <rguenther@suse.de>
5103 * tree-vect-data-refs.c (vect_dr_misalign_for_aligned_access):
5105 (vect_update_misalignment_for_peel): Use it to update
5106 misaligned to the value necessary for an aligned access.
5107 (vect_get_peeling_costs_all_drs): Likewise.
5108 (vect_enhance_data_refs_alignment): Likewise.
5110 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5112 * config/aarch64/aarch64.c (aarch64_expand_cpymem): Count number of
5113 emitted operations and adjust heuristic for code size.
5115 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5117 * config/aarch64/aarch64.c (aarch64_expand_setmem): Count number of
5118 emitted operations and adjust heuristic for code size.
5120 2021-09-29 Jakub Jelinek <jakub@redhat.com>
5122 PR middle-end/102504
5123 * gimplify.c (gimplify_scan_omp_clauses): Use omp_check_private even
5124 in OMP_SCOPE clauses, not just on worksharing construct clauses.
5126 2021-09-28 Geng Qi <gengqi@linux.alibaba.com>
5128 * config/riscv/riscv.md (mulv<mode>4): Call gen_smul<mode>3_highpart.
5129 (<u>mulditi3): Call <su>muldi3_highpart.
5130 (<u>muldi3_highpart): Rename to <su>muldi3_highpart.
5131 (<u>mulsidi3): Call <su>mulsi3_highpart.
5132 (<u>mulsi3_highpart): Rename to <su>mulsi3_highpart.
5134 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
5136 * config/darwin.h (DSYMUTIL_SPEC): Recognize D sources.
5138 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
5140 * config/rs6000/darwin.h (FIXED_R13): Add for PPC64.
5141 (FIRST_SAVED_GP_REGNO): Save from R13 even when it is one
5144 2021-09-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5146 * config/aarch64/aarch64.h (AARCH64_FL_LS64): Define
5147 (AARCH64_FL_V8_7): Likewise.
5148 (AARCH64_FL_FOR_ARCH8_7): Likewise.
5149 * config/aarch64/aarch64-arches.def (armv8.7-a): Define.
5150 * config/aarch64/aarch64-option-extensions.def (ls64): Define.
5151 * doc/invoke.texi: Document the above.
5153 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
5155 * dbgcnt.c (dbg_cnt_counter): New.
5156 * dbgcnt.h (dbg_cnt_counter): New.
5157 * dumpfile.c (dump_options): Add entry for TDF_THREADING.
5158 * dumpfile.h (enum dump_flag): Add TDF_THREADING.
5159 * gimple-range-path.cc (DEBUG_SOLVER): Use TDF_THREADING.
5160 * tree-ssa-threadupdate.c (dump_jump_thread_path): Dump out
5163 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
5165 * cfgcleanup.c (pass_jump::execute): Check
5166 flag_expensive_optimizations.
5167 (pass_jump_after_combine::gate): Same.
5168 * doc/invoke.texi (-fthread-jumps): Enable for -O1.
5169 * opts.c (default_options_table): Enable -fthread-jumps at -O1.
5170 * tree-ssa-threadupdate.c
5171 (fwd_jt_path_registry::remove_jump_threads_including): Bail unless
5174 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
5176 * tree-ssa-reassoc.c (biased_names): New global.
5177 (propagate_bias_p): New function.
5178 (loop_carried_phi): Remove.
5179 (propagate_rank): Propagate bias along single uses.
5180 (get_rank): Update biased_names when needed.
5182 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
5184 * passes.def (pass_reassoc): Rename parameter to early_p.
5185 * tree-ssa-reassoc.c (reassoc_bias_loop_carried_phi_ranks_p):
5187 (phi_rank): Don't bias loop-carried phi ranks
5188 before vectorization pass.
5189 (execute_reassoc): Add bias_loop_carried_phi_ranks_p parameter.
5190 (pass_reassoc::pass_reassoc): Add bias_loop_carried_phi_ranks_p
5192 (pass_reassoc::set_param): Set bias_loop_carried_phi_ranks_p
5194 (pass_reassoc::execute): Pass bias_loop_carried_phi_ranks_p to
5196 (pass_reassoc::bias_loop_carried_phi_ranks_p): New member.
5198 2021-09-28 Jakub Jelinek <jakub@redhat.com>
5201 * config/i386/i386.c (standard_80387_constant_p): Don't recognize
5202 special 80387 instruction XFmode constants if flag_rounding_math.
5204 2021-09-28 Richard Biener <rguenther@suse.de>
5206 PR tree-optimization/100112
5207 * tree-ssa-sccvn.c (visit_reference_op_load): Record the
5208 referece into the hashtable twice in case last_vuse is
5209 different from the original vuse on the stmt.
5211 2021-09-28 Jakub Jelinek <jakub@redhat.com>
5213 PR middle-end/102492
5214 * gimplify.c (gimplify_adjust_omp_clauses_1): Don't call the
5215 omp_finish_clause langhook on implicitly added OMP_CLAUSE_PRIVATE
5216 clauses on SIMD constructs.
5218 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
5220 PR tree-optimization/102511
5221 * gimple-range-path.cc (path_range_query::range_on_path_entry):
5222 Return VARYING when nothing found.
5224 2021-09-28 Hongyu Wang <hongyu.wang@intel.com>
5227 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): Add
5229 (VALID_SSE2_REG_VHF_MODE): Add V4HFmode and V2HFmode.
5230 (VALID_MMX_REG_MODE): Add V4HFmode.
5231 (SSE_REG_MODE_P): Replace VALID_AVX512FP16_REG_MODE with
5232 vector mode condition.
5233 * config/i386/i386.c (classify_argument): Parse V4HF/V2HF
5235 (function_arg_32): Add V4HFmode.
5236 (function_arg_advance_32): Likewise.
5237 * config/i386/i386.md (mode): Add V4HF/V2HF.
5238 (MODE_SIZE): Likewise.
5239 * config/i386/mmx.md (MMXMODE): Add V4HF mode.
5240 (V_32): Add V2HF mode.
5241 (VHF_32_64): New mode iterator.
5242 (*mov<mode>_internal): Adjust sse alternatives to support
5244 (*mov<mode>_internal): Adjust sse alternatives to support
5246 (<insn><mode>3): New define_insn for add/sub/mul/div.
5248 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
5250 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Check
5252 (pass_early_thread_jumps::gate): Same.
5253 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
5254 Return if !flag_thread_jumps.
5255 * tree-ssa-threadupdate.c
5256 (jt_path_registry::register_jump_thread): Assert that
5257 flag_thread_jumps is true.
5259 2021-09-28 liuhongt <hongtao.liu@intel.com>
5262 (simplify_context::simplify_binary_operation_1): Relax
5263 condition of simplifying (vec_concat:M (vec_select op0
5264 index0)(vec_select op1 index1)) to allow different modes
5265 between op0 and M, but have same inner mode.
5267 2021-09-28 liuhongt <hongtao.liu@intel.com>
5269 * config/i386/i386-expand.c (emit_reduc_half): Handle
5270 V8HF/V16HF/V32HFmode.
5271 * config/i386/sse.md (REDUC_SSE_PLUS_MODE): Add V8HF.
5272 (REDUC_SSE_SMINMAX_MODE): Ditto.
5273 (REDUC_PLUS_MODE): Add V16HF and V32HF.
5274 (REDUC_SMINMAX_MODE): Ditto.
5276 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
5278 * gimple-range-path.cc
5279 (path_range_query::precompute_ranges_in_block): Rename to...
5280 (path_range_query::compute_ranges_in_block): ...this.
5281 (path_range_query::precompute_ranges): Rename to...
5282 (path_range_query::compute_ranges): ...this.
5283 (path_range_query::precompute_relations): Rename to...
5284 (path_range_query::compute_relations): ...this.
5285 (path_range_query::precompute_phi_relations): Rename to...
5286 (path_range_query::compute_phi_relations): ...this.
5287 * gimple-range-path.h: Rename precompute* to compute*.
5288 * tree-ssa-threadbackward.c
5289 (back_threader::find_taken_edge_switch): Same.
5290 (back_threader::find_taken_edge_cond): Same.
5291 * tree-ssa-threadedge.c
5292 (hybrid_jt_simplifier::compute_ranges_from_state): Same.
5293 (hybrid_jt_state::register_equivs_stmt): Inline...
5294 * tree-ssa-threadedge.h: ...here.
5296 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
5298 * tree-vrp.c (lhs_of_dominating_assert): Remove.
5299 (class vrp_jt_state): Remove.
5300 (class vrp_jt_simplifier): Remove.
5301 (vrp_jt_simplifier::simplify): Remove.
5302 (class vrp_jump_threader): Remove.
5303 (vrp_jump_threader::vrp_jump_threader): Remove.
5304 (vrp_jump_threader::~vrp_jump_threader): Remove.
5305 (vrp_jump_threader::before_dom_children): Remove.
5306 (vrp_jump_threader::after_dom_children): Remove.
5308 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
5310 * passes.def (pass_vrp_threader): New.
5311 * tree-pass.h (make_pass_vrp_threader): Add make_pass_vrp_threader.
5312 * tree-ssa-threadedge.c (hybrid_jt_state::register_equivs_stmt): New.
5313 (hybrid_jt_simplifier::hybrid_jt_simplifier): New.
5314 (hybrid_jt_simplifier::simplify): New.
5315 (hybrid_jt_simplifier::compute_ranges_from_state): New.
5316 * tree-ssa-threadedge.h (class hybrid_jt_state): New.
5317 (class hybrid_jt_simplifier): New.
5318 * tree-vrp.c (execute_vrp): Remove ASSERT_EXPR based jump
5320 (class hybrid_threader): New.
5321 (hybrid_threader::hybrid_threader): New.
5322 (hybrid_threader::~hybrid_threader): New.
5323 (hybrid_threader::before_dom_children): New.
5324 (hybrid_threader::after_dom_children): New.
5325 (execute_vrp_threader): New.
5326 (class pass_vrp_threader): New.
5327 (make_pass_vrp_threader): New.
5329 2021-09-27 Martin Liska <mliska@suse.cz>
5331 * output.h (enum section_flag): New.
5332 (SECTION_FORGET): Remove.
5333 (SECTION_ENTSIZE): Make it (1UL << 8) - 1.
5334 (SECTION_STYLE_MASK): Define it based on other enum
5336 * varasm.c (switch_to_section): Remove unused handling of
5339 2021-09-27 Martin Liska <mliska@suse.cz>
5341 * common.opt: Add new variable flag_default_complex_method.
5342 * opts.c (finish_options): Handle flags related to
5343 x_flag_complex_method.
5344 * toplev.c (process_options): Remove option handling related
5345 to flag_complex_method.
5347 2021-09-27 Richard Biener <rguenther@suse.de>
5349 PR middle-end/102450
5350 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid using
5351 type_for_size, instead use int_mode_for_size.
5353 2021-09-27 Andrew Pinski <apinski@marvell.com>
5356 * gimplify.c (gimplify_save_expr): Return early
5357 if the type of val is error_mark_node.
5359 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
5361 * tree-ssanames.c (ssa_name_has_boolean_range): Use
5364 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
5366 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
5367 vrp_visit_cond_stmt.
5368 * tree-ssa-dom.c (cprop_operand): Convert to range_query API.
5369 (cprop_into_stmt): Same.
5370 (dom_opt_dom_walker::optimize_stmt): Same.
5372 2021-09-27 Richard Biener <rguenther@suse.de>
5374 PR tree-optimization/97351
5375 PR tree-optimization/97352
5376 PR tree-optimization/82426
5377 * tree-vectorizer.h (dr_misalignment): Add vector type
5379 (aligned_access_p): Likewise.
5380 (known_alignment_for_access_p): Likewise.
5381 (vect_supportable_dr_alignment): Likewise.
5382 (vect_known_alignment_in_bytes): Likewise. Refactor.
5383 (DR_MISALIGNMENT): Remove.
5384 (vect_update_shared_vectype): Likewise.
5385 * tree-vect-data-refs.c (dr_misalignment): Refactor, handle
5386 a vector type with larger alignment requirement and apply
5387 the negative step adjustment here.
5388 (vect_calculate_target_alignment): Remove.
5389 (vect_compute_data_ref_alignment): Get explicit vector type
5390 argument, do not apply a negative step alignment adjustment
5392 (vect_slp_analyze_node_alignment): Re-analyze alignment
5393 when we re-visit the DR with a bigger desired alignment but
5394 keep more precise results from smaller alignments.
5395 * tree-vect-slp.c (vect_update_shared_vectype): Remove.
5396 (vect_slp_analyze_node_operations_1): Do not update the
5397 shared vector type on stmts.
5398 * tree-vect-stmts.c (vect_analyze_stmt): Push/pop the
5399 vector type of an SLP node to the representative stmt-info.
5400 (vect_transform_stmt): Likewise.
5402 2021-09-27 liuhongt <hongtao.liu@intel.com>
5405 2021-09-09 liuhongt <hongtao.liu@intel.com>
5408 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
5409 (reduc_plus_scal_v4sf): .. this, New define_expand.
5410 (reduc_plus_scal_v2df): .. and this, New define_expand.
5412 2021-09-26 liuhongt <hongtao.liu@intel.com>
5414 * doc/extend.texi (Half-Precision): Remove storage only
5415 description for _Float16 w/o avx512fp16.
5417 2021-09-25 Dimitar Dimitrov <dimitar@dinux.eu>
5419 * config/pru/constraints.md (Rrio): New constraint.
5420 * config/pru/predicates.md (regio_operand): New predicate.
5421 * config/pru/pru-pragma.c (pru_register_pragmas): Register
5422 the __regio_symbol address space.
5423 * config/pru/pru-protos.h (pru_symref2ioregno): Declaration.
5424 * config/pru/pru.c (pru_symref2ioregno): New helper function.
5425 (pru_legitimate_address_p): Remove.
5426 (pru_addr_space_legitimate_address_p): Use the address space
5428 (pru_nongeneric_pointer_addrspace): New helper function.
5429 (pru_insert_attributes): New function to validate __regio_symbol
5431 (TARGET_INSERT_ATTRIBUTES): New macro.
5432 (TARGET_LEGITIMATE_ADDRESS_P): Remove.
5433 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): New macro.
5434 * config/pru/pru.h (enum reg_class): Add REGIO_REGS class.
5435 * config/pru/pru.md (*regio_readsi): New pattern to read I/O
5437 (*regio_nozext_writesi): New pattern to write to I/O registers.
5438 (*regio_zext_write_r30<EQS0:mode>): Ditto.
5439 * doc/extend.texi: Document the new PRU Named Address Space.
5441 2021-09-24 Patrick Palka <ppalka@redhat.com>
5445 * real.c (encode_ieee_double): Avoid unwanted sign extension.
5446 (encode_ieee_quad): Likewise.
5448 2021-09-24 Vladimir Makarov <vmakarov@redhat.com>
5450 PR rtl-optimization/102147
5451 * ira-build.c (ira_conflict_vector_profitable_p): Make
5452 profitability calculation independent of host compiler pointer and
5455 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
5457 * gimple-range-path.cc (path_range_query::path_range_query):
5458 Move debugging header...
5459 (path_range_query::precompute_ranges): ...here.
5460 (path_range_query::internal_range_of_expr): Do not call
5461 range_on_path_entry if NAME is defined in the current block.
5463 2021-09-24 Richard Biener <rguenther@suse.de>
5465 * cfghooks.c (verify_flow_info): Verify unallocated BB and
5466 edge flags are not set.
5468 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
5470 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
5472 (jt_path_registry::register_jump_thread): Call
5473 cancel_invalid_paths.
5474 * tree-ssa-threadupdate.h (class jt_path_registry): Add
5475 cancel_invalid_paths.
5477 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
5479 PR tree-optimization/102400
5480 * tree-ssa-sccvn.c (vn_reference_insert_pieces): Initialize
5481 result_vdef to zero value.
5483 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
5485 PR tree-optimization/102451
5486 * tree-ssa-dse.c (delete_dead_or_redundant_call): Record bb of stmt
5489 2021-09-24 Hongyu Wang <hongyu.wang@intel.com>
5491 * config/i386/sse.md (cond_<insn><mode>): Extend to support
5493 (cond_mul<mode>): Likewise.
5494 (cond_div<mode>): Likewise.
5495 (cond_<code><mode>): Likewise.
5496 (cond_fma<mode>): Likewise.
5497 (cond_fms<mode>): Likewise.
5498 (cond_fnma<mode>): Likewise.
5499 (cond_fnms<mode>): Likewise.
5501 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
5503 PR tree-optimization/102463
5504 * gimple-range-fold.cc (fold_using_range::relation_fold_and_or): If
5505 there is no range-ops handler, don't look for a relation.
5507 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
5509 * gimple-range-cache.cc (ranger_cache::ranger_cache): Take
5510 non-executable_edge flag as parameter.
5511 * gimple-range-cache.h (ranger_cache): Adjust prototype.
5512 * gimple-range-gori.cc (gori_compute::gori_compute): Take
5513 non-executable_edge flag as parameter.
5514 (gori_compute::outgoing_edge_range_p): Check new flag.
5515 * gimple-range-gori.h (gori_compute): Adjust prototype.
5516 * gimple-range.cc (gimple_ranger::gimple_ranger): Create new flag.
5517 (gimple_ranger::range_on_edge): Check new flag.
5518 * gimple-range.h (gimple_ranger::non_executable_edge_flag): New.
5519 * gimple-ssa-evrp.c (rvrp_folder): Pass ranger flag to simplifer.
5520 (hybrid_folder::hybrid_folder): Set ranger non-executable flag value.
5521 (hybrid_folder::fold_stmt): Set flag value in the simplifer.
5522 * vr-values.c (simplify_using_ranges::set_and_propagate_unexecutable):
5523 Use not_executable flag if provided inmstead of EDGE_EXECUTABLE.
5524 (simplify_using_ranges::simplify_switch_using_ranges): Clear
5525 EDGE_EXECUTABLE like it originally did.
5526 (simplify_using_ranges::cleanup_edges_and_switches): Clear any
5527 NON_EXECUTABLE flags.
5528 (simplify_using_ranges::simplify_using_ranges): Adjust.
5529 * vr-values.h (class simplify_using_ranges): Adjust.
5530 (simplify_using_ranges::set_range_query): Add non-executable flag param.
5532 2021-09-23 Bill Schmidt <wschmidt@linux.ibm.com>
5535 * config/rs6000/rs6000-call.c (rs6000_aggregate_candidate): Detect
5536 zero-width bit fields and return indicator.
5537 (rs6000_discover_homogeneous_aggregate): Diagnose when the
5538 presence of a zero-width bit field changes parameter passing in
5541 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
5543 * gimple-range-fold.cc (fold_using_range::range_of_phi):
5544 Remove dominator check.
5546 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
5548 * gimple-range-path.cc (path_range_query::precompute_relations):
5549 Hoist edge calculations before using EDGE_SUCC.
5551 2021-09-23 Jonathan Wakely <jwakely@redhat.com>
5553 * configure.ac: Fix --with-multilib-list description.
5554 * configure: Regenerate.
5556 2021-09-23 Richard Biener <rguenther@suse.de>
5558 PR tree-optimization/102448
5559 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
5560 Clear alignment info copied from DR_PTR_INFO.
5562 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
5564 * config/i386/i386-expand.c (ix86_use_mask_cmp_p): Enable
5566 * config/i386/sse.md (sseintvecmodelower): Add HF vector modes.
5567 (<avx512>_store<mode>_mask): Extend to support HF vector modes.
5568 (vec_cmp<mode><avx512fmaskmodelower>): Likewise.
5569 (vcond_mask_<mode><avx512fmaskmodelower>): Likewise.
5570 (vcond<mode><mode>): New expander.
5571 (vcond<mode><sseintvecmodelower>): Likewise.
5572 (vcond<sseintvecmodelower><mode>): Likewise.
5573 (vcondu<mode><sseintvecmodelower>): Likewise.
5575 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
5577 * config/i386/sse.md (extend<ssePHmodelower><mode>2):
5579 (extendv4hf<mode>2): Likewise.
5580 (extendv2hfv2df2): Likewise.
5581 (trunc<mode><ssePHmodelower>2): Likewise.
5582 (avx512fp16_vcvt<castmode>2ph_<mode>): Rename to ...
5583 (trunc<mode>v4hf2): ... this, and drop constraints.
5584 (avx512fp16_vcvtpd2ph_v2df): Rename to ...
5585 (truncv2dfv2hf2): ... this, and likewise.
5587 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
5589 * config/i386/sse.md (float<floatunssuffix><mode><ssePHmodelower>2):
5591 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>):
5593 (float<floatunssuffix><mode>v4hf2): ... this, and drop constraints.
5594 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Rename to ...
5595 (float<floatunssuffix>v2div2hf2): ... this, and likewise.
5597 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
5599 * config/i386/i386.md (fix<fixunssuffix>_trunchf<mode>2): New expander.
5600 (fixuns_trunchfhi2): Likewise.
5601 (*fixuns_trunchfsi2zext): New define_insn.
5602 * config/i386/sse.md (ssePHmodelower): New mode_attr.
5603 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
5604 New expander for same element vector fix_truncate.
5605 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
5606 Likewise for V4HF to V4SI/V4DI fix_truncate.
5607 (fix<fixunssuffix>_truncv2hfv2di2):
5608 Likeise for V2HF to V2DI fix_truncate.
5610 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
5612 * config/i386/i386.md (<code>hf3): New expander.
5614 2021-09-23 liuhongt <hongtao.liu@intel.com>
5616 * config/i386/sse.md (FMAMODEM): extend to handle FP16.
5617 (VFH_SF_AVX512VL): Extend to handle HFmode.
5618 (VF_SF_AVX512VL): Deleted.
5620 2021-09-23 liuhongt <hongtao.liu@intel.com>
5622 * config/i386/i386.md (rinthf2): New expander.
5623 (nearbyinthf2): New expander.
5625 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
5627 * tree-ssa-dom.c (class dom_jump_threader_simplifier): Rename...
5628 (class dom_jt_state): ...this and provide virtual overrides.
5629 (dom_jt_state::register_equiv): New.
5630 (class dom_jt_simplifier): Rename from
5631 dom_jump_threader_simplifier.
5632 (dom_jump_threader_simplifier::simplify): Rename...
5633 (dom_jt_simplifier::simplify): ...to this.
5634 (pass_dominator::execute): Use dom_jt_simplifier and
5636 * tree-ssa-threadedge.c (jump_threader::jump_threader):
5638 (jt_state::register_equivs_stmt): Abstract out...
5639 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
5641 (jump_threader::thread_around_empty_blocks): Update state.
5642 (jump_threader::thread_through_normal_block): Same.
5643 (jt_state::jt_state): Remove.
5644 (jt_state::push): Remove pass specific bits. Keep block vector
5646 (jt_state::append_path): New.
5647 (jt_state::pop): Remove pass specific bits.
5648 (jt_state::register_equiv): Same.
5649 (jt_state::record_ranges_from_stmt): Same.
5650 (jt_state::register_equivs_on_edge): Same. Rename...
5651 (jt_state::register_equivs_edge): ...to this.
5652 (jt_state::dump): New.
5653 (jt_state::debug): New.
5654 (jump_threader_simplifier::simplify): Remove.
5655 (jt_state::get_path): New.
5656 * tree-ssa-threadedge.h (class jt_simplifier): Make into a base
5657 class. Expose common functionality as virtual methods.
5658 (class jump_threader_simplifier): Same. Rename...
5659 (class jt_simplifier): ...to this.
5660 * tree-vrp.c (class vrp_jump_threader_simplifier): Rename...
5661 (class vrp_jt_simplifier): ...to this. Provide pass specific
5663 (class vrp_jt_state): New.
5664 (vrp_jump_threader_simplifier::simplify): Rename...
5665 (vrp_jt_simplifier::simplify): ...to this. Inline code from
5666 what used to be the base class.
5667 (vrp_jump_threader::vrp_jump_threader): Use vrp_jt_state and
5670 2021-09-22 Tobias Burnus <tobias@codesourcery.com>
5673 * doc/invoke.texi (-Wno-missing-include-dirs.): Document Fortran
5676 2021-09-22 Roger Sayle <roger@nextmovesoftware.com>
5677 Richard Biener <rguenther@suse.de>
5679 * match.pd (negation simplifications): Implement some negation
5680 folding transformations from fold-const.c's fold_negate_expr.
5681 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Add a SIMPLIFY
5682 argument, to control whether the op should be simplified prior
5683 to looking up/assigning a value number.
5684 (vn_nary_build_or_lookup): Update call to vn_nary_build_or_lookup_1.
5685 (vn_nary_simplify): Likewise.
5686 (visit_nary_op): Likewise, but when constructing a NEGATE_EXPR
5687 now call vn_nary_build_or_lookup_1 disabling simplification.
5689 2021-09-22 Jiufu Guo <guojiufu@linux.ibm.com>
5691 PR tree-optimization/102087
5692 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
5693 Update bound/cmp/control for niter.
5695 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
5697 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
5698 Move check for non-empty BB here.
5699 (fur_source::register_outgoing_edges): ...from here.
5701 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
5703 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
5704 Remove call to improve_range_with_equivs.
5705 (path_range_query::improve_range_with_equivs): Remove
5706 * gimple-range-path.h: Remove improve_range_with_equivs.
5708 2021-09-22 dianhong xu <dianhong.xu@intel.com>
5710 * config/i386/avx512fp16intrin.h:
5711 (_mm512_mask_blend_ph): New intrinsic.
5712 (_mm512_permutex2var_ph): Ditto.
5713 (_mm512_permutexvar_ph): Ditto.
5714 * config/i386/avx512fp16vlintrin.h:
5715 (_mm256_mask_blend_ph): New intrinsic.
5716 (_mm256_permutex2var_ph): Ditto.
5717 (_mm256_permutexvar_ph): Ditto.
5718 (_mm_mask_blend_ph): Ditto.
5719 (_mm_permutex2var_ph): Ditto.
5720 (_mm_permutexvar_ph): Ditto.
5722 2021-09-22 dianhong xu <dianhong.xu@intel.com>
5724 * config/i386/avx512fp16intrin.h: Add new intrinsics.
5725 (_mm512_conj_pch): New intrinsic.
5726 (_mm512_mask_conj_pch): Ditto.
5727 (_mm512_maskz_conj_pch): Ditto.
5728 * config/i386/avx512fp16vlintrin.h: Add new intrinsics.
5729 (_mm256_conj_pch): New intrinsic.
5730 (_mm256_mask_conj_pch): Ditto.
5731 (_mm256_maskz_conj_pch): Ditto.
5732 (_mm_conj_pch): Ditto.
5733 (_mm_mask_conj_pch): Ditto.
5734 (_mm_maskz_conj_pch): Ditto.
5736 2021-09-22 dianhong xu <dianhong.xu@intel.com>
5738 * config/i386/avx512fp16intrin.h (_MM512_REDUCE_OP): New macro
5739 (_mm512_reduce_add_ph): New intrinsic.
5740 (_mm512_reduce_mul_ph): Ditto.
5741 (_mm512_reduce_min_ph): Ditto.
5742 (_mm512_reduce_max_ph): Ditto.
5743 * config/i386/avx512fp16vlintrin.h
5744 (_MM256_REDUCE_OP/_MM_REDUCE_OP): New macro.
5745 (_mm256_reduce_add_ph): New intrinsic.
5746 (_mm256_reduce_mul_ph): Ditto.
5747 (_mm256_reduce_min_ph): Ditto.
5748 (_mm256_reduce_max_ph): Ditto.
5749 (_mm_reduce_add_ph): Ditto.
5750 (_mm_reduce_mul_ph): Ditto.
5751 (_mm_reduce_min_ph): Ditto.
5752 (_mm_reduce_max_ph): Ditto.
5754 2021-09-22 dianhong xu <dianhong.xu@intel.com>
5756 * config/i386/avx512fp16intrin.h (__m512h_u, __m256h_u,
5757 __m128h_u): New typedef.
5758 (_mm512_load_ph): New intrinsic.
5759 (_mm256_load_ph): Ditto.
5760 (_mm_load_ph): Ditto.
5761 (_mm512_loadu_ph): Ditto.
5762 (_mm256_loadu_ph): Ditto.
5763 (_mm_loadu_ph): Ditto.
5764 (_mm512_store_ph): Ditto.
5765 (_mm256_store_ph): Ditto.
5766 (_mm_store_ph): Ditto.
5767 (_mm512_storeu_ph): Ditto.
5768 (_mm256_storeu_ph): Ditto.
5769 (_mm_storeu_ph): Ditto.
5770 (_mm512_abs_ph): Ditto.
5771 * config/i386/avx512fp16vlintrin.h
5772 (_mm_abs_ph): Ditto.
5773 (_mm256_abs_ph): Ditto.
5775 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
5777 * config/s390/tpf.md (prologue_tpf, epilogue_tpf): Add cc clobber.
5779 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
5782 * config/s390/s390.c (s390_expand_insv): Emit a normal move if it
5783 is actually a full copy of the source operand into the target.
5784 Don't emit a strict low part move if source and target mode match.
5786 2021-09-22 Jakub Jelinek <jakub@redhat.com>
5788 PR middle-end/102415
5789 * omp-expand.c (expand_omp_single): If region->exit is NULL,
5790 assert region->entry is GIMPLE_OMP_SCOPE region and return.
5792 2021-09-22 Jakub Jelinek <jakub@redhat.com>
5794 * tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
5795 * tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
5796 arguments from 2 to 3.
5797 * tree-pretty-print.c (dump_omp_clause): Print allocator() around
5798 allocate clause allocator and print align if present.
5799 * omp-low.c (scan_sharing_clauses): Force allocate_map entry even
5800 for omp_default_mem_alloc if align modifier is present. If align
5801 modifier is present, use TREE_LIST to encode both allocator and
5803 (lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
5804 Handle align modifier on allocator clause if present.
5806 2021-09-22 liuhongt <hongtao.liu@intel.com>
5808 * config/i386/i386.md (define_attr "isa"): Add
5810 (define_attr "enabled"): Correspond fma_or_avx512vl to
5811 TARGET_FMA || TARGET_AVX512VL.
5812 * config/i386/mmx.md (fmav2sf4): Extend to AVX512 fma.
5817 2021-09-22 liuhongt <hongtao.liu@intel.com>
5819 * config/i386/i386.md (cstorehf3): New define_expand.
5821 2021-09-22 liuhongt <hongtao.liu@intel.com>
5823 * config/i386/i386.md (<rounding_insn>hf2): New expander.
5824 (sse4_1_round<mode>2): Extend from MODEF to MODEFH.
5825 * config/i386/sse.md (*sse4_1_round<ssescalarmodesuffix>):
5826 Extend from VF_128 to VFH_128.
5828 2021-09-22 liuhongt <hongtao.liu@intel.com>
5830 * config/i386/i386-features.c (i386-features.c): Handle
5832 * config/i386/i386.md (sqrthf2): New expander.
5833 (*sqrthf2): New define_insn.
5834 * config/i386/sse.md
5835 (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
5838 2021-09-22 liuhongt <hongtao.liu@intel.com>
5840 * config/i386/avx512fp16intrin.h (_mm_mask_fcmadd_sch):
5842 (_mm_mask3_fcmadd_sch): Likewise.
5843 (_mm_maskz_fcmadd_sch): Likewise.
5844 (_mm_fcmadd_sch): Likewise.
5845 (_mm_mask_fmadd_sch): Likewise.
5846 (_mm_mask3_fmadd_sch): Likewise.
5847 (_mm_maskz_fmadd_sch): Likewise.
5848 (_mm_fmadd_sch): Likewise.
5849 (_mm_mask_fcmadd_round_sch): Likewise.
5850 (_mm_mask3_fcmadd_round_sch): Likewise.
5851 (_mm_maskz_fcmadd_round_sch): Likewise.
5852 (_mm_fcmadd_round_sch): Likewise.
5853 (_mm_mask_fmadd_round_sch): Likewise.
5854 (_mm_mask3_fmadd_round_sch): Likewise.
5855 (_mm_maskz_fmadd_round_sch): Likewise.
5856 (_mm_fmadd_round_sch): Likewise.
5857 (_mm_fcmul_sch): Likewise.
5858 (_mm_mask_fcmul_sch): Likewise.
5859 (_mm_maskz_fcmul_sch): Likewise.
5860 (_mm_fmul_sch): Likewise.
5861 (_mm_mask_fmul_sch): Likewise.
5862 (_mm_maskz_fmul_sch): Likewise.
5863 (_mm_fcmul_round_sch): Likewise.
5864 (_mm_mask_fcmul_round_sch): Likewise.
5865 (_mm_maskz_fcmul_round_sch): Likewise.
5866 (_mm_fmul_round_sch): Likewise.
5867 (_mm_mask_fmul_round_sch): Likewise.
5868 (_mm_maskz_fmul_round_sch): Likewise.
5869 * config/i386/i386-builtin.def: Add corresponding new builtins.
5870 * config/i386/sse.md
5871 (avx512fp16_fmaddcsh_v8hf_maskz<round_expand_name>): New expander.
5872 (avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>): Ditto.
5873 (avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>):
5875 (avx512fp16_<complexopname>sh_v8hf_mask<round_name>): Ditto.
5876 (avx512fp16_<complexopname>sh_v8hf<mask_scalarc_name><round_scalarcz_name>):
5878 * config/i386/subst.md (mask_scalarcz_name): New.
5879 (mask_scalarc_name): Ditto.
5880 (mask_scalarc_operand3): Ditto.
5881 (mask_scalarcz_operand4): Ditto.
5882 (round_scalarcz_name): Ditto.
5883 (round_scalarc_mask_operand3): Ditto.
5884 (round_scalarcz_mask_operand4): Ditto.
5885 (round_scalarc_mask_op3): Ditto.
5886 (round_scalarcz_mask_op4): Ditto.
5887 (round_scalarcz_constraint): Ditto.
5888 (round_scalarcz_nimm_predicate): Ditto.
5889 (mask_scalarcz): Ditto.
5890 (mask_scalarc): Ditto.
5891 (round_scalarcz): Ditto.
5893 2021-09-22 liuhongt <hongtao.liu@intel.com>
5895 * config/i386/avx512fp16intrin.h (_mm512_fcmadd_pch):
5897 (_mm512_mask_fcmadd_pch): Likewise.
5898 (_mm512_mask3_fcmadd_pch): Likewise.
5899 (_mm512_maskz_fcmadd_pch): Likewise.
5900 (_mm512_fmadd_pch): Likewise.
5901 (_mm512_mask_fmadd_pch): Likewise.
5902 (_mm512_mask3_fmadd_pch): Likewise.
5903 (_mm512_maskz_fmadd_pch): Likewise.
5904 (_mm512_fcmadd_round_pch): Likewise.
5905 (_mm512_mask_fcmadd_round_pch): Likewise.
5906 (_mm512_mask3_fcmadd_round_pch): Likewise.
5907 (_mm512_maskz_fcmadd_round_pch): Likewise.
5908 (_mm512_fmadd_round_pch): Likewise.
5909 (_mm512_mask_fmadd_round_pch): Likewise.
5910 (_mm512_mask3_fmadd_round_pch): Likewise.
5911 (_mm512_maskz_fmadd_round_pch): Likewise.
5912 (_mm512_fcmul_pch): Likewise.
5913 (_mm512_mask_fcmul_pch): Likewise.
5914 (_mm512_maskz_fcmul_pch): Likewise.
5915 (_mm512_fmul_pch): Likewise.
5916 (_mm512_mask_fmul_pch): Likewise.
5917 (_mm512_maskz_fmul_pch): Likewise.
5918 (_mm512_fcmul_round_pch): Likewise.
5919 (_mm512_mask_fcmul_round_pch): Likewise.
5920 (_mm512_maskz_fcmul_round_pch): Likewise.
5921 (_mm512_fmul_round_pch): Likewise.
5922 (_mm512_mask_fmul_round_pch): Likewise.
5923 (_mm512_maskz_fmul_round_pch): Likewise.
5924 * config/i386/avx512fp16vlintrin.h (_mm_fmadd_pch):
5926 (_mm_mask_fmadd_pch): Likewise.
5927 (_mm_mask3_fmadd_pch): Likewise.
5928 (_mm_maskz_fmadd_pch): Likewise.
5929 (_mm256_fmadd_pch): Likewise.
5930 (_mm256_mask_fmadd_pch): Likewise.
5931 (_mm256_mask3_fmadd_pch): Likewise.
5932 (_mm256_maskz_fmadd_pch): Likewise.
5933 (_mm_fcmadd_pch): Likewise.
5934 (_mm_mask_fcmadd_pch): Likewise.
5935 (_mm_mask3_fcmadd_pch): Likewise.
5936 (_mm_maskz_fcmadd_pch): Likewise.
5937 (_mm256_fcmadd_pch): Likewise.
5938 (_mm256_mask_fcmadd_pch): Likewise.
5939 (_mm256_mask3_fcmadd_pch): Likewise.
5940 (_mm256_maskz_fcmadd_pch): Likewise.
5941 (_mm_fmul_pch): Likewise.
5942 (_mm_mask_fmul_pch): Likewise.
5943 (_mm_maskz_fmul_pch): Likewise.
5944 (_mm256_fmul_pch): Likewise.
5945 (_mm256_mask_fmul_pch): Likewise.
5946 (_mm256_maskz_fmul_pch): Likewise.
5947 (_mm_fcmul_pch): Likewise.
5948 (_mm_mask_fcmul_pch): Likewise.
5949 (_mm_maskz_fcmul_pch): Likewise.
5950 (_mm256_fcmul_pch): Likewise.
5951 (_mm256_mask_fcmul_pch): Likewise.
5952 (_mm256_maskz_fcmul_pch): Likewise.
5953 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF_V8HF_V8HF,
5954 V8HF_FTYPE_V16HF_V16HF_V16HF, V16HF_FTYPE_V16HF_V16HF_V16HF_UQI,
5955 V32HF_FTYPE_V32HF_V32HF_V32HF_INT,
5956 V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT): Add new builtin types.
5957 * config/i386/i386-builtin.def: Add new builtins.
5958 * config/i386/i386-expand.c: Handle new builtin types.
5959 * config/i386/subst.md (SUBST_CV): New.
5960 (maskc_name): Ditto.
5961 (maskc_operand3): Ditto.
5963 (sdc_maskz_name): Ditto.
5964 (sdc_mask_op4): Ditto.
5965 (sdc_mask_op5): Ditto.
5966 (sdc_mask_mode512bit_condition): Ditto.
5968 (round_maskc_operand3): Ditto.
5969 (round_sdc_mask_operand4): Ditto.
5970 (round_maskc_op3): Ditto.
5971 (round_sdc_mask_op4): Ditto.
5972 (round_saeonly_sdc_mask_operand5): Ditto.
5973 * config/i386/sse.md (unspec): Add complex fma unspecs.
5974 (avx512fmaskcmode): New.
5975 (UNSPEC_COMPLEX_F_C_MA): Ditto.
5976 (UNSPEC_COMPLEX_F_C_MUL): Ditto.
5977 (complexopname): Ditto.
5978 (<avx512>_fmaddc_<mode>_maskz<round_expand_name>): New expander.
5979 (<avx512>_fcmaddc_<mode>_maskz<round_expand_name>): Ditto.
5980 (fma_<complexopname>_<mode><sdc_maskz_name><round_name>): New
5982 (<avx512>_<complexopname>_<mode>_mask<round_name>): Ditto.
5983 (<avx512>_<complexopname>_<mode><maskc_name><round_name>): Ditto.
5985 2021-09-22 Kewen Lin <linkw@linux.ibm.com>
5987 * config/rs6000/rs6000.opt (rs6000-density-pct-threshold,
5988 rs6000-density-size-threshold, rs6000-density-penalty,
5989 rs6000-density-load-pct-threshold,
5990 rs6000-density-load-num-threshold): New parameter.
5991 * config/rs6000/rs6000.c (rs6000_density_test): Adjust with
5992 corresponding parameters.
5994 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
5996 * gimple-range-path.cc (path_range_query::defined_outside_path):
5998 (path_range_query::range_on_path_entry): New.
5999 (path_range_query::internal_range_of_expr): Resolve unknowns
6001 (path_range_query::improve_range_with_equivs): New.
6002 (path_range_query::ssa_range_in_phi): Resolve unknowns with
6004 * gimple-range-path.h (class path_range_query): Add
6005 defined_outside_path, range_on_path_entry, and
6006 improve_range_with_equivs.
6008 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
6010 * gimple-range-path.cc (path_range_query::add_to_imports): New.
6011 (path_range_query::add_copies_to_imports): New.
6012 (path_range_query::precompute_ranges): Call
6013 add_copies_to_imports.
6014 * gimple-range-path.h (class path_range_query): Add prototypes
6015 for add_copies_to_imports and add_to_imports.
6017 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
6019 * gimple-range-path.cc (path_range_query::range_defined_in_block):
6020 Remove useless code.
6022 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
6024 * gimple-range-fold.h (class fur_source): Make oracle protected.
6025 * gimple-range-path.cc (path_range_query::path_range_query): Add
6026 resolve argument. Initialize oracle.
6027 (path_range_query::~path_range_query): Delete oracle.
6028 (path_range_query::range_of_stmt): Adapt to use relations.
6029 (path_range_query::precompute_ranges): Pre-compute relations.
6030 (class jt_fur_source): New
6031 (jt_fur_source::jt_fur_source): New.
6032 (jt_fur_source::register_relation): New.
6033 (jt_fur_source::query_relation): New.
6034 (path_range_query::precompute_relations): New.
6035 (path_range_query::precompute_phi_relations): New.
6036 * gimple-range-path.h (path_range_query): Add resolve argument.
6037 Add oracle, precompute_relations, precompute_phi_relations.
6038 * tree-ssa-threadbackward.c (back_threader::back_threader): Pass
6039 resolve argument to solver.
6041 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
6043 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
6044 Rename postfold_gcond_edges to register_outgoing_edges and
6046 (fold_using_range::postfold_gcond_edges): Rename...
6047 (fur_source::register_outgoing_edges): ...to this.
6048 * gimple-range-fold.h (postfold_gcond_edges): Rename to
6049 register_outgoing_edges and move to fur_source.
6051 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
6053 * gimple-range-fold.cc (fold_using_range::range_of_phi): Check
6054 dom_info_available_p.
6056 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
6058 * gimple-range-cache.cc (non_null_ref::non_null_ref): Use create
6059 and quick_grow_cleared instead of safe_grow_cleared.
6061 2021-09-21 Thomas Schwinge <thomas@codesourcery.com>
6064 * omp-oacc-neuter-broadcast.cc (oacc_do_neutering): Evaluate
6067 2021-09-21 Richard Earnshaw <rearnsha@arm.com>
6069 * configure.ac: Detect when the assembler supports new-style
6070 architecture extensions.
6071 * common/config/arm/arm-common.c (arm_rewrite_mcpu): Return
6072 the full CPU string if the assembler can grok it.
6073 (arm_rewrite_march): Likewise but for the architecture.
6074 * config.in: Regenerate.
6075 * configure: Regenerate.
6077 2021-09-21 Richard Biener <rguenther@suse.de>
6079 PR tree-optimization/102421
6080 * tree-vect-loop.c (vect_dissolve_slp_only_groups): Copy and
6081 adjust alignment info.
6083 2021-09-21 Kewen Lin <linkw@linux.ibm.com>
6085 * ipa-fnsummary.c (ipa_fn_summary_write): Remove inconsistent
6086 bitfield stream out.
6088 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
6090 * gimple-range-fold.cc (fold_using_range::range_of_phi): Ignore
6091 undefined edges, apply an equivalence if appropriate.
6092 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Return
6093 UNDEFINED if EDGE_EXECUTABLE is not set.
6094 * gimple-range.cc (gimple_ranger::gimple_ranger): Set all edges
6095 as EXECUTABLE upon startup.
6096 (gimple_ranger::range_on_edge): Return UNDEFINED for edges without
6097 EDGE_EXECUTABLE set.
6098 * vr-values.c (set_and_propagate_unexecutable): New.
6099 (simplify_using_ranges::fold_cond): Call set_and_propagate.
6100 (simplify_using_ranges::simplify_switch_using_ranges): Ditto.
6101 * vr-values.h: Add prototype.
6103 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
6105 * value-relation.cc (equiv_oracle::register_initial_def): New.
6106 (equiv_oracle::register_relation): Call register_initial_def.
6107 (equiv_oracle::add_equiv_to_block): New. Split register_relation.
6108 (relation_oracle::register_stmt): Check def block of PHI arguments.
6109 * value-relation.h (equiv_oracle): Add new prototypes.
6111 2021-09-20 Matthias Kretz <m.kretz@gsi.de>
6113 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
6114 Define __RECIPROCAL_MATH__, __NO_SIGNED_ZEROS__,
6115 __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__, and
6116 __ROUNDING_MATH__ according to their corresponding flags.
6117 * doc/cpp.texi: Document __RECIPROCAL_MATH__,
6118 __NO_SIGNED_ZEROS__, __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__,
6119 and __ROUNDING_MATH__.
6121 2021-09-20 Richard Biener <rguenther@suse.de>
6123 * tree-vect-stmts.c (vectorizable_load): Use the vectype
6126 2021-09-20 Richard Biener <rguenther@suse.de>
6128 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
6129 Do not compute alignment of the vectorized access here.
6131 2021-09-20 Richard Biener <rguenther@suse.de>
6133 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
6134 Store -1 for runtime alias peeling iterations.
6136 2021-09-20 Richard Biener <rguenther@suse.de>
6138 * config.gcc: Obsolete hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11*.
6140 2021-09-20 Thomas Schwinge <thomas@codesourcery.com>
6142 * input.c (string_concat_db::record_string_concatenation)
6143 (string_concat_db::get_string_concatenation): Skip for
6144 'RESERVED_LOCATION_P'.
6146 2021-09-20 Richard Biener <rguenther@suse.de>
6148 PR tree-optimization/65206
6149 * tree-data-ref.h (struct data_reference): Add alt_indices,
6151 * tree-data-ref.c (free_data_ref): Release alt_indices.
6152 (dr_analyze_indices): Work on struct indices and get DR_REF as tree.
6153 (create_data_ref): Adjust.
6154 (initialize_data_dependence_relation): Split into head
6155 and tail. When the base objects fail to match up try
6156 again with pointer-based analysis of indices.
6157 * tree-vectorizer.c (vec_info_shared::check_datarefs): Do
6158 not compare the lazily computed alternate set of indices.
6160 2021-09-20 Iain Sandoe <iain@sandoe.co.uk>
6162 * gcc.c: Test for execute OK when we find the
6163 programs for assembler linker and dsymutil and those
6164 were specified at configure-time.
6166 2021-09-19 Martin Sebor <msebor@redhat.com>
6168 PR middle-end/102403
6169 * gimple-predicate-analysis.cc (predicate::init_from_control_deps):
6170 Correct a function pre/postcondition.
6172 2021-09-19 Martin Sebor <msebor@redhat.com>
6174 PR middle-end/102243
6175 * tree-ssa-strlen.c (get_range): Handle null cfun.
6177 2021-09-19 Iain Sandoe <iain@sandoe.co.uk>
6179 * config/darwin.h (LINK_COMMAND_SPEC_A): Use Darwin10
6180 unwinder shim as a convenience library.
6182 2021-09-19 Andrew Pinski <apinski@marvell.com>
6184 * doc/install.texi: Add note about
6185 binutils 2.35 is required for LTO usage.
6187 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
6189 * tree-ssa-threadbackward.c
6190 (back_threader_registry::register_path): Use push_edge.
6191 * tree-ssa-threadedge.c
6192 (jump_threader::thread_around_empty_blocks): Same.
6193 (jump_threader::thread_through_normal_block): Same.
6194 (jump_threader::thread_across_edge): Same. Also, use auto_bitmap.
6196 * tree-ssa-threadupdate.c
6197 (jt_path_registry::allocate_thread_edge): Remove.
6198 (jt_path_registry::push_edge): New.
6199 (dump_jump_thread_path): Make static.
6200 * tree-ssa-threadupdate.h (allocate_thread_edge): Remove.
6203 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
6205 * gimple-range-path.cc (path_range_query::path_range_query): Add
6207 (path_range_query::dump): Remove extern declaration of dump_ranger.
6208 * gimple-range-trace.cc (dump_ranger): Add DEBUG_FUNCTION marker.
6209 * gimple-range-trace.h (dump_ranger): Add prototype.
6211 2021-09-19 John Ericson <git@JohnEricson.me>
6213 * gcc.c (find_a_program): New function, factored out of...
6214 (find_a_file): Here.
6215 (execute): Use find_a_program when looking for programs rather
6218 2021-09-19 Matwey V. Kornilov <matwey.kornilov@gmail.com>
6220 * config/avr/avr-mcus.def: Add atmega324pb.
6221 * doc/avr-mmcu.texi: Corresponding changes.
6223 2021-09-19 Roger Sayle <roger@nextmovesoftware.com>
6226 * match.pd (cmp @0 REAL_CST@1): When @0 is also REAL_CST, apply
6227 the same transformations as to @1. For comparisons against NaN,
6228 don't check HONOR_SNANS but confirm that neither operand is a
6231 2021-09-19 Benjamin Peterson <benjamin@locrian.net>
6233 * attribs.c (make_unique_name): Delete.
6234 * attribs.h (make_unique_name): Delete.
6236 2021-09-19 Andrew Pinski <apinski@marvell.com>
6238 * lra-constraints.c (check_and_process_move): Assert
6239 that dclass and sclass are greater than or equal to NO_REGS.
6241 2021-09-18 Jakub Jelinek <jakub@redhat.com>
6243 * tree.h (OMP_CLAUSE_ORDER_UNCONSTRAINED): Define.
6244 * tree-pretty-print.c (dump_omp_clause): Print unconstrained:
6245 for OMP_CLAUSE_ORDER_UNCONSTRAINED.
6247 2021-09-18 liuhongt <hongtao.liu@intel.com>
6249 * config/i386/i386-features.c (remove_partial_avx_dependency):
6250 Restrict TARGET_USE_VECTOR_FP_CONVERTS and
6251 TARGET_USE_VECTOR_CONVERTS to conversion instructions only.
6253 2021-09-18 Jakub Jelinek <jakub@redhat.com>
6255 * gimplify.c (omp_default_clause): For C/C++ default({,first}private),
6256 if file/namespace scope variable doesn't have predetermined sharing,
6257 treat it as if there was default(none).
6259 2021-09-18 liuhongt <hongtao.liu@intel.com>
6261 * config/i386/avx512fp16intrin.h (_mm_fmadd_sh):
6263 (_mm_mask_fmadd_sh): Likewise.
6264 (_mm_mask3_fmadd_sh): Likewise.
6265 (_mm_maskz_fmadd_sh): Likewise.
6266 (_mm_fmadd_round_sh): Likewise.
6267 (_mm_mask_fmadd_round_sh): Likewise.
6268 (_mm_mask3_fmadd_round_sh): Likewise.
6269 (_mm_maskz_fmadd_round_sh): Likewise.
6270 (_mm_fnmadd_sh): Likewise.
6271 (_mm_mask_fnmadd_sh): Likewise.
6272 (_mm_mask3_fnmadd_sh): Likewise.
6273 (_mm_maskz_fnmadd_sh): Likewise.
6274 (_mm_fnmadd_round_sh): Likewise.
6275 (_mm_mask_fnmadd_round_sh): Likewise.
6276 (_mm_mask3_fnmadd_round_sh): Likewise.
6277 (_mm_maskz_fnmadd_round_sh): Likewise.
6278 (_mm_fmsub_sh): Likewise.
6279 (_mm_mask_fmsub_sh): Likewise.
6280 (_mm_mask3_fmsub_sh): Likewise.
6281 (_mm_maskz_fmsub_sh): Likewise.
6282 (_mm_fmsub_round_sh): Likewise.
6283 (_mm_mask_fmsub_round_sh): Likewise.
6284 (_mm_mask3_fmsub_round_sh): Likewise.
6285 (_mm_maskz_fmsub_round_sh): Likewise.
6286 (_mm_fnmsub_sh): Likewise.
6287 (_mm_mask_fnmsub_sh): Likewise.
6288 (_mm_mask3_fnmsub_sh): Likewise.
6289 (_mm_maskz_fnmsub_sh): Likewise.
6290 (_mm_fnmsub_round_sh): Likewise.
6291 (_mm_mask_fnmsub_round_sh): Likewise.
6292 (_mm_mask3_fnmsub_round_sh): Likewise.
6293 (_mm_maskz_fnmsub_round_sh): Likewise.
6294 * config/i386/i386-builtin-types.def
6295 (V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT): New builtin type.
6296 * config/i386/i386-builtin.def: Add new builtins.
6297 * config/i386/i386-expand.c: Handle new builtin type.
6298 * config/i386/sse.md (fmai_vmfmadd_<mode><round_name>):
6299 Ajdust to support FP16.
6300 (fmai_vmfmsub_<mode><round_name>): Ditto.
6301 (fmai_vmfnmadd_<mode><round_name>): Ditto.
6302 (fmai_vmfnmsub_<mode><round_name>): Ditto.
6303 (*fmai_fmadd_<mode>): Ditto.
6304 (*fmai_fmsub_<mode>): Ditto.
6305 (*fmai_fnmadd_<mode><round_name>): Ditto.
6306 (*fmai_fnmsub_<mode><round_name>): Ditto.
6307 (avx512f_vmfmadd_<mode>_mask<round_name>): Ditto.
6308 (avx512f_vmfmadd_<mode>_mask3<round_name>): Ditto.
6309 (avx512f_vmfmadd_<mode>_maskz<round_expand_name>): Ditto.
6310 (avx512f_vmfmadd_<mode>_maskz_1<round_name>): Ditto.
6311 (*avx512f_vmfmsub_<mode>_mask<round_name>): Ditto.
6312 (avx512f_vmfmsub_<mode>_mask3<round_name>): Ditto.
6313 (*avx512f_vmfmsub_<mode>_maskz_1<round_name>): Ditto.
6314 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
6315 (*avx512f_vmfnmsub_<mode>_mask3<round_name>): Ditto.
6316 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
6317 (*avx512f_vmfnmadd_<mode>_mask<round_name>): Renamed to ...
6318 (avx512f_vmfnmadd_<mode>_mask<round_name>) ... this, and
6319 adjust to support FP16.
6320 (avx512f_vmfnmadd_<mode>_mask3<round_name>): Ditto.
6321 (avx512f_vmfnmadd_<mode>_maskz_1<round_name>): Ditto.
6322 (avx512f_vmfnmadd_<mode>_maskz<round_expand_name>): New
6325 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
6327 * config/i386/sse.md (avx512fmaskmodelower): Extend to support
6329 (maskload<mode><avx512fmaskmodelower>): Ditto.
6330 (maskstore<mode><avx512fmaskmodelower>): Ditto.
6332 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
6334 * config/i386/i386-expand.c (ix86_expand_fp_absneg_operator):
6336 (ix86_expand_copysign): Ditto.
6337 (ix86_expand_xorsign): Ditto.
6338 * config/i386/i386.c (ix86_build_const_vector): Handle HF vector
6340 (ix86_build_signbit_mask): Ditto.
6341 (ix86_can_change_mode_class): Ditto.
6342 * config/i386/i386.md
6343 (SSEMODEF): Add HFmode.
6344 (ssevecmodef): Ditto.
6345 (<code>hf2): New define_expand.
6346 (*<code>hf2_1): New define_insn_and_split.
6347 (copysign<mode>): Extend to support HFmode under AVX512FP16.
6348 (xorsign<mode>): Ditto.
6349 * config/i386/sse.md (VFB): New mode iterator.
6350 (VFB_128_256): Ditto.
6352 (sseintvecmode2): Support HF vector mode.
6353 (<code><mode>2): Use new mode iterator.
6354 (*<code><mode>2): Ditto.
6355 (copysign<mode>3): Ditto.
6356 (xorsign<mode>3): Ditto.
6357 (<code><mode>3<mask_name>): Ditto.
6358 (<code><mode>3<mask_name>): Ditto.
6359 (<sse>_andnot<mode>3<mask_name>): Adjust for HF vector mode.
6360 (<sse>_andnot<mode>3<mask_name>): Ditto.
6361 (*<code><mode>3<mask_name>): Ditto.
6362 (*<code><mode>3<mask_name>): Ditto.
6364 2021-09-18 liuhongt <hongtao.liu@intel.com>
6366 * config/i386/avx512fp16intrin.h (_mm512_mask_fmadd_ph):
6368 (_mm512_mask3_fmadd_ph): Likewise.
6369 (_mm512_maskz_fmadd_ph): Likewise.
6370 (_mm512_fmadd_round_ph): Likewise.
6371 (_mm512_mask_fmadd_round_ph): Likewise.
6372 (_mm512_mask3_fmadd_round_ph): Likewise.
6373 (_mm512_maskz_fmadd_round_ph): Likewise.
6374 (_mm512_fnmadd_ph): Likewise.
6375 (_mm512_mask_fnmadd_ph): Likewise.
6376 (_mm512_mask3_fnmadd_ph): Likewise.
6377 (_mm512_maskz_fnmadd_ph): Likewise.
6378 (_mm512_fnmadd_round_ph): Likewise.
6379 (_mm512_mask_fnmadd_round_ph): Likewise.
6380 (_mm512_mask3_fnmadd_round_ph): Likewise.
6381 (_mm512_maskz_fnmadd_round_ph): Likewise.
6382 (_mm512_fmsub_ph): Likewise.
6383 (_mm512_mask_fmsub_ph): Likewise.
6384 (_mm512_mask3_fmsub_ph): Likewise.
6385 (_mm512_maskz_fmsub_ph): Likewise.
6386 (_mm512_fmsub_round_ph): Likewise.
6387 (_mm512_mask_fmsub_round_ph): Likewise.
6388 (_mm512_mask3_fmsub_round_ph): Likewise.
6389 (_mm512_maskz_fmsub_round_ph): Likewise.
6390 (_mm512_fnmsub_ph): Likewise.
6391 (_mm512_mask_fnmsub_ph): Likewise.
6392 (_mm512_mask3_fnmsub_ph): Likewise.
6393 (_mm512_maskz_fnmsub_ph): Likewise.
6394 (_mm512_fnmsub_round_ph): Likewise.
6395 (_mm512_mask_fnmsub_round_ph): Likewise.
6396 (_mm512_mask3_fnmsub_round_ph): Likewise.
6397 (_mm512_maskz_fnmsub_round_ph): Likewise.
6398 * config/i386/avx512fp16vlintrin.h (_mm256_fmadd_ph):
6400 (_mm256_mask_fmadd_ph): Likewise.
6401 (_mm256_mask3_fmadd_ph): Likewise.
6402 (_mm256_maskz_fmadd_ph): Likewise.
6403 (_mm_fmadd_ph): Likewise.
6404 (_mm_mask_fmadd_ph): Likewise.
6405 (_mm_mask3_fmadd_ph): Likewise.
6406 (_mm_maskz_fmadd_ph): Likewise.
6407 (_mm256_fnmadd_ph): Likewise.
6408 (_mm256_mask_fnmadd_ph): Likewise.
6409 (_mm256_mask3_fnmadd_ph): Likewise.
6410 (_mm256_maskz_fnmadd_ph): Likewise.
6411 (_mm_fnmadd_ph): Likewise.
6412 (_mm_mask_fnmadd_ph): Likewise.
6413 (_mm_mask3_fnmadd_ph): Likewise.
6414 (_mm_maskz_fnmadd_ph): Likewise.
6415 (_mm256_fmsub_ph): Likewise.
6416 (_mm256_mask_fmsub_ph): Likewise.
6417 (_mm256_mask3_fmsub_ph): Likewise.
6418 (_mm256_maskz_fmsub_ph): Likewise.
6419 (_mm_fmsub_ph): Likewise.
6420 (_mm_mask_fmsub_ph): Likewise.
6421 (_mm_mask3_fmsub_ph): Likewise.
6422 (_mm_maskz_fmsub_ph): Likewise.
6423 (_mm256_fnmsub_ph): Likewise.
6424 (_mm256_mask_fnmsub_ph): Likewise.
6425 (_mm256_mask3_fnmsub_ph): Likewise.
6426 (_mm256_maskz_fnmsub_ph): Likewise.
6427 (_mm_fnmsub_ph): Likewise.
6428 (_mm_mask_fnmsub_ph): Likewise.
6429 (_mm_mask3_fnmsub_ph): Likewise.
6430 (_mm_maskz_fnmsub_ph): Likewise.
6431 * config/i386/i386-builtin.def: Add corresponding new builtins.
6432 * config/i386/sse.md
6433 (<avx512>_fmadd_<mode>_maskz<round_expand_name>): Adjust to
6434 support HF vector modes.
6435 (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
6437 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
6438 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
6439 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
6440 (<avx512>_fmadd_<mode>_mask<round_name>): Ditto.
6441 (<avx512>_fmadd_<mode>_mask3<round_name>): Ditto.
6442 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Ditto.
6443 (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>):
6445 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
6446 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
6447 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
6448 (<avx512>_fmsub_<mode>_mask<round_name>): Ditto.
6449 (<avx512>_fmsub_<mode>_mask3<round_name>): Ditto.
6450 (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>):
6452 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
6453 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
6454 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
6455 (<avx512>_fnmadd_<mode>_mask<round_name>): Ditto.
6456 (<avx512>_fnmadd_<mode>_mask3<round_name>): Ditto.
6457 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Ditto.
6458 (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>):
6460 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
6461 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
6462 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
6463 (<avx512>_fnmsub_<mode>_mask<round_name>): Ditto.
6464 (<avx512>_fnmsub_<mode>_mask3<round_name>): Ditto.
6466 2021-09-18 liuhongt <hongtao.liu@intel.com>
6468 * config/i386/avx512fp16intrin.h (_mm512_fmaddsub_ph):
6470 (_mm512_mask_fmaddsub_ph): Likewise.
6471 (_mm512_mask3_fmaddsub_ph): Likewise.
6472 (_mm512_maskz_fmaddsub_ph): Likewise.
6473 (_mm512_fmaddsub_round_ph): Likewise.
6474 (_mm512_mask_fmaddsub_round_ph): Likewise.
6475 (_mm512_mask3_fmaddsub_round_ph): Likewise.
6476 (_mm512_maskz_fmaddsub_round_ph): Likewise.
6477 (_mm512_mask_fmsubadd_ph): Likewise.
6478 (_mm512_mask3_fmsubadd_ph): Likewise.
6479 (_mm512_maskz_fmsubadd_ph): Likewise.
6480 (_mm512_fmsubadd_round_ph): Likewise.
6481 (_mm512_mask_fmsubadd_round_ph): Likewise.
6482 (_mm512_mask3_fmsubadd_round_ph): Likewise.
6483 (_mm512_maskz_fmsubadd_round_ph): Likewise.
6484 * config/i386/avx512fp16vlintrin.h (_mm256_fmaddsub_ph):
6486 (_mm256_mask_fmaddsub_ph): Likewise.
6487 (_mm256_mask3_fmaddsub_ph): Likewise.
6488 (_mm256_maskz_fmaddsub_ph): Likewise.
6489 (_mm_fmaddsub_ph): Likewise.
6490 (_mm_mask_fmaddsub_ph): Likewise.
6491 (_mm_mask3_fmaddsub_ph): Likewise.
6492 (_mm_maskz_fmaddsub_ph): Likewise.
6493 (_mm256_fmsubadd_ph): Likewise.
6494 (_mm256_mask_fmsubadd_ph): Likewise.
6495 (_mm256_mask3_fmsubadd_ph): Likewise.
6496 (_mm256_maskz_fmsubadd_ph): Likewise.
6497 (_mm_fmsubadd_ph): Likewise.
6498 (_mm_mask_fmsubadd_ph): Likewise.
6499 (_mm_mask3_fmsubadd_ph): Likewise.
6500 (_mm_maskz_fmsubadd_ph): Likewise.
6501 * config/i386/i386-builtin.def: Add corresponding new builtins.
6502 * config/i386/sse.md (VFH_SF_AVX512VL): New mode iterator.
6503 * (<avx512>_fmsubadd_<mode>_maskz<round_expand_name>): New expander.
6504 * (<avx512>_fmaddsub_<mode>_maskz<round_expand_name>): Use
6506 * (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
6508 * (<avx512>_fmaddsub_<mode>_mask<round_name>): Ditto.
6509 * (<avx512>_fmaddsub_<mode>_mask3<round_name>): Ditto.
6510 * (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
6512 * (<avx512>_fmsubadd_<mode>_mask<round_name>): Ditto.
6513 * (<avx512>_fmsubadd_<mode>_mask3<round_name>): Ditto.
6515 2021-09-18 liuhongt <hongtao.liu@intel.com>
6518 * config/i386/i386.c (ix86_print_operand): Handle
6519 V8HF/V16HF/V32HFmode.
6520 * config/i386/i386.h (VALID_BCST_MODE_P): Add HFmode.
6521 * config/i386/sse.md (avx512bcst): Remove.
6523 2021-09-17 Martin Sebor <msebor@redhat.com>
6525 * Makefile.in (OBJS): Add gimple-predicate-analysis.o.
6526 * tree-ssa-uninit.c (max_phi_args): Move to gimple-predicate-analysis.
6527 (MASK_SET_BIT, MASK_TEST_BIT, MASK_EMPTY): Same.
6528 (check_defs): Add comment.
6529 (can_skip_redundant_opnd): Update comment.
6530 (compute_uninit_opnds_pos): Adjust to namespace change.
6531 (find_pdom): Move to gimple-predicate-analysis.cc.
6533 (struct uninit_undef_val_t): New.
6534 (is_non_loop_exit_postdominating): Move to gimple-predicate-analysis.cc.
6535 (find_control_equiv_block): Same.
6536 (MAX_NUM_CHAINS, MAX_CHAIN_LEN, MAX_POSTDOM_CHECK): Same.
6537 (MAX_SWITCH_CASES): Same.
6538 (compute_control_dep_chain): Same.
6539 (find_uninit_use): Use predicate analyzer.
6540 (struct pred_info): Move to gimple-predicate-analysis.
6541 (convert_control_dep_chain_into_preds): Same.
6542 (find_predicates): Same.
6543 (collect_phi_def_edges): Same.
6544 (warn_uninitialized_phi): Use predicate analyzer.
6545 (find_def_preds): Move to gimple-predicate-analysis.
6546 (dump_pred_info): Same.
6547 (dump_pred_chain): Same.
6548 (dump_predicates): Same.
6549 (destroy_predicate_vecs): Remove.
6550 (execute_late_warn_uninitialized): New.
6551 (get_cmp_code): Move to gimple-predicate-analysis.
6552 (is_value_included_in): Same.
6553 (value_sat_pred_p): Same.
6554 (find_matching_predicate_in_rest_chains): Same.
6555 (is_use_properly_guarded): Same.
6556 (prune_uninit_phi_opnds): Same.
6557 (find_var_cmp_const): Same.
6558 (use_pred_not_overlap_with_undef_path_pred): Same.
6559 (pred_equal_p): Same.
6560 (is_neq_relop_p): Same.
6561 (is_neq_zero_form_p): Same.
6562 (pred_expr_equal_p): Same.
6563 (is_pred_expr_subset_of): Same.
6564 (is_pred_chain_subset_of): Same.
6565 (is_included_in): Same.
6566 (is_superset_of): Same.
6568 (simplify_pred): Same.
6569 (simplify_preds_2): Same.
6570 (simplify_preds_3): Same.
6571 (simplify_preds_4): Same.
6572 (simplify_preds): Same.
6574 (push_to_worklist): Same.
6575 (get_pred_info_from_cmp): Same.
6576 (is_degenerated_phi): Same.
6577 (normalize_one_pred_1): Same.
6578 (normalize_one_pred): Same.
6579 (normalize_one_pred_chain): Same.
6580 (normalize_preds): Same.
6581 (can_one_predicate_be_invalidated_p): Same.
6582 (can_chain_union_be_invalidated_p): Same.
6583 (uninit_uses_cannot_happen): Same.
6584 (pass_late_warn_uninitialized::execute): Define.
6585 * gimple-predicate-analysis.cc: New file.
6586 * gimple-predicate-analysis.h: New file.
6588 2021-09-17 Julian Brown <julian@codesourcery.com>
6590 * config/gcn/gcn.c (gimple.h): Include.
6591 (gcn_fork_join): Emit barrier for worker-level joins.
6592 * omp-oacc-neuter-broadcast.cc (find_local_vars_to_propagate): Add
6593 writes_gang_private bitmap parameter. Set bit for blocks
6594 containing gang-private variable writes.
6595 (worker_single_simple): Don't emit barrier after predicated block.
6596 (worker_single_copy): Don't emit barrier if we're not broadcasting
6597 anything and the block contains no gang-private writes.
6598 (neuter_worker_single): Don't predicate blocks that only contain
6599 NOPs or internal marker functions. Pass has_gang_private_write
6600 argument to worker_single_copy.
6601 (oacc_do_neutering): Add writes_gang_private bitmap handling.
6603 2021-09-17 Julian Brown <julian@codesourcery.com>
6605 * config/gcn/gcn-protos.h
6606 (gcn_goacc_create_worker_broadcast_record): Update prototype.
6607 * config/gcn/gcn-tree.c (gcn_goacc_get_worker_red_decl): Use
6608 preallocated block of LDS memory. Do not cache/share decls for
6609 reduction temporaries between invocations.
6610 (gcn_goacc_reduction_teardown): Unshare VAR on second use.
6611 (gcn_goacc_create_worker_broadcast_record): Add OFFSET parameter
6612 and return temporary LDS space at that offset. Return pointer in
6614 * config/gcn/gcn.c (acc_lds_size, gang_private_hwm, lds_allocs):
6616 (ACC_LDS_SIZE): Define as acc_lds_size.
6617 (gcn_init_machine_status): Don't initialise lds_allocated,
6618 lds_allocs, reduc_decls fields of machine function struct.
6619 (gcn_option_override): Handle default size for gang-private
6620 variables and -mgang-private-size option.
6621 (gcn_expand_prologue): Use LDS_SIZE instead of LDS_SIZE-1 when
6622 initialising M0_REG.
6623 (gcn_shared_mem_layout): New function.
6624 (gcn_print_lds_decl): Update comment. Use global lds_allocs map and
6625 gang_private_hwm variable.
6626 (TARGET_GOACC_SHARED_MEM_LAYOUT): Define target hook.
6627 * config/gcn/gcn.h (machine_function): Remove lds_allocated,
6628 lds_allocs, reduc_decls. Add reduction_base, reduction_limit.
6629 * config/gcn/gcn.opt (gang_private_size_opt): New global.
6630 (mgang-private-size=): New option.
6631 * doc/tm.texi.in (TARGET_GOACC_SHARED_MEM_LAYOUT): Place
6633 * doc/tm.texi: Regenerate.
6634 * omp-oacc-neuter-broadcast.cc (targhooks.h, diagnostic-core.h):
6636 (build_sender_ref): Handle sender_decl being pointer.
6637 (worker_single_copy): Add PLACEMENT and ISOLATE_BROADCASTS
6638 parameters. Pass placement argument to
6639 create_worker_broadcast_record hook invocations. Handle
6640 sender_decl being pointer and isolate_broadcasts inserting extra
6642 (blk_offset_map_t): Add typedef.
6643 (neuter_worker_single): Add BLK_OFFSET_MAP parameter. Pass
6644 preallocated range to worker_single_copy call.
6645 (dfs_broadcast_reachable_1): New function.
6646 (idx_decl_pair_t, used_range_vec_t): New typedefs.
6647 (sort_size_descending): New function.
6648 (addr_range): New class.
6649 (splay_tree_compare_addr_range, splay_tree_free_key)
6650 (first_fit_range, merge_ranges_1, merge_ranges): New functions.
6651 (execute_omp_oacc_neuter_broadcast): Rename to...
6652 (oacc_do_neutering): ... this. Add BOUNDS_LO, BOUNDS_HI
6653 parameters. Arrange layout of shared memory for broadcast
6655 (execute_omp_oacc_neuter_broadcast): New function.
6656 (pass_omp_oacc_neuter_broadcast::gate): Remove num_workers==1
6657 handling from here. Enable pass for all OpenACC routines in order
6658 to call shared memory-layout hook.
6659 * target.def (create_worker_broadcast_record): Add OFFSET
6661 (shared_mem_layout): New hook.
6663 2021-09-17 Julian Brown <julian@codesourcery.com>
6664 Thomas Schwinge <thomas@codesourcery.com>
6666 * omp-oacc-neuter-broadcast.cc
6667 (pass_omp_oacc_neuter_broadcast::gate): Disable if num_workers is
6669 (execute_omp_oacc_neuter_broadcast): Adjust.
6671 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
6673 * value-relation.cc (class equiv_chain): Move to header file.
6674 (path_oracle::path_oracle): New.
6675 (path_oracle::~path_oracle): New.
6676 (path_oracle::register_relation): New.
6677 (path_oracle::query_relation): New.
6678 (path_oracle::reset_path): New.
6679 (path_oracle::dump): New.
6680 * value-relation.h (class equiv_chain): Move to here.
6681 (class path_oracle): New.
6683 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
6685 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a DOM
6687 * gimple-range-fold.cc (fur_depend::register_relation): Use
6688 register_stmt/edge routines.
6689 * value-relation.cc (equiv_chain::find): Relocate from equiv_oracle.
6690 (equiv_oracle::equiv_oracle): Create self equivalence cache.
6691 (equiv_oracle::~equiv_oracle): Release same.
6692 (equiv_oracle::equiv_set): Return entry from self equiv cache if there
6693 are no equivalences.
6694 (equiv_oracle::find_equiv_block): Move list find to equiv_chain.
6695 (equiv_oracle::register_relation): Rename from register_equiv.
6696 (relation_chain_head::find_relation): Relocate from dom_oracle.
6697 (relation_oracle::register_stmt): New.
6698 (relation_oracle::register_edge): New.
6699 (dom_oracle::*): Rename from relation_oracle.
6700 (dom_oracle::register_relation): Adjust to call equiv_oracle.
6701 (dom_oracle::set_one_relation): Split from register_relation.
6702 (dom_oracle::register_transitives): Consolidate 2 methods.
6703 (dom_oracle::find_relation_block): Move core to relation_chain.
6704 (dom_oracle::query_relation): Rename from find_relation_dom and adjust.
6705 * value-relation.h (class relation_oracle): New pure virtual base.
6706 (class equiv_oracle): Inherit from relation_oracle and adjust.
6707 (class dom_oracle): Rename from old relation_oracle and adjust.
6709 2021-09-17 Martin Sebor <msebor@redhat.com>
6711 PR middle-end/102200
6712 * pointer-query.cc (access_ref::inform_access): Handle MIN/MAX_EXPR.
6713 (handle_min_max_size): Change argument. Store original SSA_NAME for
6714 operands to potentially distinct (sub)objects.
6715 (compute_objsize_r): Adjust call to the above.
6717 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
6719 * config/rs6000/rs6000.c (rs6000-builtins.h): New include.
6720 (rs6000_new_builtin_vectorized_function): New function.
6721 (rs6000_new_builtin_md_vectorized_function): Likewise.
6722 (rs6000_builtin_vectorized_function): Call
6723 rs6000_new_builtin_vectorized_function.
6724 (rs6000_builtin_md_vectorized_function): Call
6725 rs6000_new_builtin_md_vectorized_function.
6727 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
6729 * config/rs6000/rs6000-builtin-new.def (ASSEMBLE_ACC): Add mmaint flag.
6730 (ASSEMBLE_PAIR): Likewise.
6731 (BUILD_ACC): Likewise.
6732 (DISASSEMBLE_ACC): Likewise.
6733 (DISASSEMBLE_PAIR): Likewise.
6734 (PMXVBF16GER2): Likewise.
6735 (PMXVBF16GER2NN): Likewise.
6736 (PMXVBF16GER2NP): Likewise.
6737 (PMXVBF16GER2PN): Likewise.
6738 (PMXVBF16GER2PP): Likewise.
6739 (PMXVF16GER2): Likewise.
6740 (PMXVF16GER2NN): Likewise.
6741 (PMXVF16GER2NP): Likewise.
6742 (PMXVF16GER2PN): Likewise.
6743 (PMXVF16GER2PP): Likewise.
6744 (PMXVF32GER): Likewise.
6745 (PMXVF32GERNN): Likewise.
6746 (PMXVF32GERNP): Likewise.
6747 (PMXVF32GERPN): Likewise.
6748 (PMXVF32GERPP): Likewise.
6749 (PMXVF64GER): Likewise.
6750 (PMXVF64GERNN): Likewise.
6751 (PMXVF64GERNP): Likewise.
6752 (PMXVF64GERPN): Likewise.
6753 (PMXVF64GERPP): Likewise.
6754 (PMXVI16GER2): Likewise.
6755 (PMXVI16GER2PP): Likewise.
6756 (PMXVI16GER2S): Likewise.
6757 (PMXVI16GER2SPP): Likewise.
6758 (PMXVI4GER8): Likewise.
6759 (PMXVI4GER8PP): Likewise.
6760 (PMXVI8GER4): Likewise.
6761 (PMXVI8GER4PP): Likewise.
6762 (PMXVI8GER4SPP): Likewise.
6763 (XVBF16GER2): Likewise.
6764 (XVBF16GER2NN): Likewise.
6765 (XVBF16GER2NP): Likewise.
6766 (XVBF16GER2PN): Likewise.
6767 (XVBF16GER2PP): Likewise.
6768 (XVF16GER2): Likewise.
6769 (XVF16GER2NN): Likewise.
6770 (XVF16GER2NP): Likewise.
6771 (XVF16GER2PN): Likewise.
6772 (XVF16GER2PP): Likewise.
6773 (XVF32GER): Likewise.
6774 (XVF32GERNN): Likewise.
6775 (XVF32GERNP): Likewise.
6776 (XVF32GERPN): Likewise.
6777 (XVF32GERPP): Likewise.
6778 (XVF64GER): Likewise.
6779 (XVF64GERNN): Likewise.
6780 (XVF64GERNP): Likewise.
6781 (XVF64GERPN): Likewise.
6782 (XVF64GERPP): Likewise.
6783 (XVI16GER2): Likewise.
6784 (XVI16GER2PP): Likewise.
6785 (XVI16GER2S): Likewise.
6786 (XVI16GER2SPP): Likewise.
6787 (XVI4GER8): Likewise.
6788 (XVI4GER8PP): Likewise.
6789 (XVI8GER4): Likewise.
6790 (XVI8GER4PP): Likewise.
6791 (XVI8GER4SPP): Likewise.
6792 (XXMFACC): Likewise.
6793 (XXMTACC): Likewise.
6794 (XXSETACCZ): Likewise.
6795 (ASSEMBLE_PAIR_V): Likewise.
6796 (BUILD_PAIR): Likewise.
6797 (DISASSEMBLE_PAIR_V): Likewise.
6800 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_mma_builtin):
6801 Handle RS6000_BIF_LXVP and RS6000_BIF_STXVP.
6802 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add ismmaint.
6803 (parse_bif_attrs): Handle ismmaint.
6804 (write_decls): Add bif_mmaint_bit and bif_is_mmaint.
6805 (write_bif_static_init): Handle ismmaint.
6807 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
6809 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin): New
6811 (rs6000_gimple_fold_builtin): Call rs6000_gimple_fold_new_builtin.
6812 (rs6000_new_builtin_valid_without_lhs): New function.
6813 (rs6000_gimple_fold_new_mma_builtin): Likewise.
6814 (rs6000_gimple_fold_new_builtin): Likewise.
6816 2021-09-17 Thomas Schwinge <thomas@codesourcery.com>
6818 * hash-table.h (hash_table<Descriptor, Lazy, Allocator>::expand):
6819 Destruct stale Value objects.
6820 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
6823 2021-09-17 Roger Sayle <roger@nextmovesoftware.com>
6826 * match.pd (shift optimizations): Disable recent sign-changing
6827 optimization for shifts by zero, these will be folded later.
6829 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
6831 * config/rs6000/rs6000-builtin-new.def (__builtin_mffsl): Move from
6832 [power9] to [always].
6834 2021-09-17 Richard Biener <rguenther@suse.de>
6836 * tree-vect-stmts.c (vectorizable_load): Do not frob
6839 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
6841 * config/i386/i386-features.c (remove_partial_avx_dependency):
6842 Also check TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY and
6843 and TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY before generating
6845 * config/i386/i386.h (TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY):
6847 (TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
6848 * config/i386/i386.md (SSE FP to FP splitters): Replace
6849 TARGET_SSE_PARTIAL_REG_DEPENDENCY with
6850 TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY.
6851 (SSE INT to FP splitter): Replace TARGET_SSE_PARTIAL_REG_DEPENDENCY
6852 with TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY.
6853 * config/i386/x86-tune.def
6854 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): New.
6855 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
6857 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
6860 * config/i386/i386-features.c (remove_partial_avx_dependency):
6861 Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
6862 before generating vxorps.
6864 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
6866 * config/i386/i386-options.c (processor_cost_table): Use
6867 tremont_cost for Tremont.
6868 * config/i386/x86-tune-costs.h (tremont_memcpy): New.
6869 (tremont_memset): Likewise.
6870 (tremont_cost): Likewise.
6871 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
6874 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
6876 * common/config/i386/i386-common.c: Use Haswell scheduling model
6878 * config/i386/i386.c (ix86_sched_init_global): Prepare for Tremont
6880 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Tremont
6882 (ix86_adjust_cost): Handle Tremont.
6883 * config/i386/x86-tune.def (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY):
6885 (X86_TUNE_USE_LEAVE): Likewise.
6886 (X86_TUNE_PUSH_MEMORY): Likewise.
6887 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
6888 (X86_TUNE_USE_CLTD): Likewise.
6889 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
6890 (X86_TUNE_AVOID_MFENCE): Likewise.
6891 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
6892 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
6893 (X86_TUNE_ACCUMULATE_OUTGOING_ARGS): Disable for Tremont.
6894 (X86_TUNE_FOUR_JUMP_LIMIT): Likewise.
6895 (X86_TUNE_OPT_AGU): Likewise.
6896 (X86_TUNE_AVOID_LEA_FOR_ADDR): Likewise.
6897 (X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE): Likewise.
6898 (X86_TUNE_EXPAND_ABS): Likewise.
6899 (X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS): Likewise.
6900 (X86_TUNE_SLOW_PSHUFB): Likewise.
6902 2021-09-17 Eric Botcazou <ebotcazou@adacore.com>
6904 PR rtl-optimization/102306
6905 * combine.c (try_combine): Abort the combination if we are about to
6906 duplicate volatile references.
6908 2021-09-17 liuhongt <hongtao.liu@intel.com>
6910 * config/i386/avx512fp16intrin.h (_mm_undefined_ph):
6912 (_mm256_undefined_ph): Likewise.
6913 (_mm512_undefined_ph): Likewise.
6914 (_mm_cvtsh_h): Likewise.
6915 (_mm256_cvtsh_h): Likewise.
6916 (_mm512_cvtsh_h): Likewise.
6917 (_mm512_castph_ps): Likewise.
6918 (_mm512_castph_pd): Likewise.
6919 (_mm512_castph_si512): Likewise.
6920 (_mm512_castph512_ph128): Likewise.
6921 (_mm512_castph512_ph256): Likewise.
6922 (_mm512_castph128_ph512): Likewise.
6923 (_mm512_castph256_ph512): Likewise.
6924 (_mm512_zextph128_ph512): Likewise.
6925 (_mm512_zextph256_ph512): Likewise.
6926 (_mm512_castps_ph): Likewise.
6927 (_mm512_castpd_ph): Likewise.
6928 (_mm512_castsi512_ph): Likewise.
6929 * config/i386/avx512fp16vlintrin.h (_mm_castph_ps):
6931 (_mm256_castph_ps): Likewise.
6932 (_mm_castph_pd): Likewise.
6933 (_mm256_castph_pd): Likewise.
6934 (_mm_castph_si128): Likewise.
6935 (_mm256_castph_si256): Likewise.
6936 (_mm_castps_ph): Likewise.
6937 (_mm256_castps_ph): Likewise.
6938 (_mm_castpd_ph): Likewise.
6939 (_mm256_castpd_ph): Likewise.
6940 (_mm_castsi128_ph): Likewise.
6941 (_mm256_castsi256_ph): Likewise.
6942 (_mm256_castph256_ph128): Likewise.
6943 (_mm256_castph128_ph256): Likewise.
6944 (_mm256_zextph128_ph256): Likewise.
6946 2021-09-17 liuhongt <hongtao.liu@intel.com>
6948 * config/i386/avx512fp16intrin.h (_mm_cvtsh_ss):
6950 (_mm_mask_cvtsh_ss): Likewise.
6951 (_mm_maskz_cvtsh_ss): Likewise.
6952 (_mm_cvtsh_sd): Likewise.
6953 (_mm_mask_cvtsh_sd): Likewise.
6954 (_mm_maskz_cvtsh_sd): Likewise.
6955 (_mm_cvt_roundsh_ss): Likewise.
6956 (_mm_mask_cvt_roundsh_ss): Likewise.
6957 (_mm_maskz_cvt_roundsh_ss): Likewise.
6958 (_mm_cvt_roundsh_sd): Likewise.
6959 (_mm_mask_cvt_roundsh_sd): Likewise.
6960 (_mm_maskz_cvt_roundsh_sd): Likewise.
6961 (_mm_cvtss_sh): Likewise.
6962 (_mm_mask_cvtss_sh): Likewise.
6963 (_mm_maskz_cvtss_sh): Likewise.
6964 (_mm_cvtsd_sh): Likewise.
6965 (_mm_mask_cvtsd_sh): Likewise.
6966 (_mm_maskz_cvtsd_sh): Likewise.
6967 (_mm_cvt_roundss_sh): Likewise.
6968 (_mm_mask_cvt_roundss_sh): Likewise.
6969 (_mm_maskz_cvt_roundss_sh): Likewise.
6970 (_mm_cvt_roundsd_sh): Likewise.
6971 (_mm_mask_cvt_roundsd_sh): Likewise.
6972 (_mm_maskz_cvt_roundsd_sh): Likewise.
6973 * config/i386/i386-builtin-types.def
6974 (V8HF_FTYPE_V2DF_V8HF_V8HF_UQI_INT,
6975 V8HF_FTYPE_V4SF_V8HF_V8HF_UQI_INT,
6976 V2DF_FTYPE_V8HF_V2DF_V2DF_UQI_INT,
6977 V4SF_FTYPE_V8HF_V4SF_V4SF_UQI_INT): Add new builtin types.
6978 * config/i386/i386-builtin.def: Add corrresponding new builtins.
6979 * config/i386/i386-expand.c: Handle new builtin types.
6980 * config/i386/sse.md (VF48_128): New mode iterator.
6981 (avx512fp16_vcvtsh2<ssescalarmodesuffix><mask_scalar_name><round_saeonly_scalar_name>):
6983 (avx512fp16_vcvt<ssescalarmodesuffix>2sh<mask_scalar_name><round_scalar_name>):
6986 2021-09-17 liuhongt <hongtao.liu@intel.com>
6988 * config/i386/avx512fp16intrin.h (_mm512_cvtph_pd):
6990 (_mm512_mask_cvtph_pd): Likewise.
6991 (_mm512_maskz_cvtph_pd): Likewise.
6992 (_mm512_cvt_roundph_pd): Likewise.
6993 (_mm512_mask_cvt_roundph_pd): Likewise.
6994 (_mm512_maskz_cvt_roundph_pd): Likewise.
6995 (_mm512_cvtxph_ps): Likewise.
6996 (_mm512_mask_cvtxph_ps): Likewise.
6997 (_mm512_maskz_cvtxph_ps): Likewise.
6998 (_mm512_cvtx_roundph_ps): Likewise.
6999 (_mm512_mask_cvtx_roundph_ps): Likewise.
7000 (_mm512_maskz_cvtx_roundph_ps): Likewise.
7001 (_mm512_cvtxps_ph): Likewise.
7002 (_mm512_mask_cvtxps_ph): Likewise.
7003 (_mm512_maskz_cvtxps_ph): Likewise.
7004 (_mm512_cvtx_roundps_ph): Likewise.
7005 (_mm512_mask_cvtx_roundps_ph): Likewise.
7006 (_mm512_maskz_cvtx_roundps_ph): Likewise.
7007 (_mm512_cvtpd_ph): Likewise.
7008 (_mm512_mask_cvtpd_ph): Likewise.
7009 (_mm512_maskz_cvtpd_ph): Likewise.
7010 (_mm512_cvt_roundpd_ph): Likewise.
7011 (_mm512_mask_cvt_roundpd_ph): Likewise.
7012 (_mm512_maskz_cvt_roundpd_ph): Likewise.
7013 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_pd):
7015 (_mm_mask_cvtph_pd): Likewise.
7016 (_mm_maskz_cvtph_pd): Likewise.
7017 (_mm256_cvtph_pd): Likewise.
7018 (_mm256_mask_cvtph_pd): Likewise.
7019 (_mm256_maskz_cvtph_pd): Likewise.
7020 (_mm_cvtxph_ps): Likewise.
7021 (_mm_mask_cvtxph_ps): Likewise.
7022 (_mm_maskz_cvtxph_ps): Likewise.
7023 (_mm256_cvtxph_ps): Likewise.
7024 (_mm256_mask_cvtxph_ps): Likewise.
7025 (_mm256_maskz_cvtxph_ps): Likewise.
7026 (_mm_cvtxps_ph): Likewise.
7027 (_mm_mask_cvtxps_ph): Likewise.
7028 (_mm_maskz_cvtxps_ph): Likewise.
7029 (_mm256_cvtxps_ph): Likewise.
7030 (_mm256_mask_cvtxps_ph): Likewise.
7031 (_mm256_maskz_cvtxps_ph): Likewise.
7032 (_mm_cvtpd_ph): Likewise.
7033 (_mm_mask_cvtpd_ph): Likewise.
7034 (_mm_maskz_cvtpd_ph): Likewise.
7035 (_mm256_cvtpd_ph): Likewise.
7036 (_mm256_mask_cvtpd_ph): Likewise.
7037 (_mm256_maskz_cvtpd_ph): Likewise.
7038 * config/i386/i386-builtin.def: Add corresponding new builtins.
7039 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7040 * config/i386/i386-expand.c: Handle new builtin types.
7041 * config/i386/sse.md
7042 (VF4_128_8_256): New.
7043 (VF48H_AVX512VL): Ditto.
7044 (ssePHmode): Add HF vector modes.
7045 (castmode): Add new convertable modes.
7048 (avx512fp16_vcvt<castmode>2ph_<mode><mask_name><round_name>): Ditto.
7049 (avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
7050 (*avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
7051 (avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
7052 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
7053 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask_1): Ditto.
7054 (avx512fp16_float_extend_ph<mode>2<mask_name><round_saeonly_name>):
7056 (avx512fp16_float_extend_ph<mode>2<mask_name>): Ditto.
7057 (*avx512fp16_float_extend_ph<mode>2_load<mask_name>): Ditto.
7058 (avx512fp16_float_extend_phv2df2<mask_name>): Ditto.
7059 (*avx512fp16_float_extend_phv2df2_load<mask_name>): Ditto.
7061 2021-09-17 liuhongt <hongtao.liu@intel.com>
7063 * config/i386/avx512fp16intrin.h (_mm_cvttsh_i32):
7065 (_mm_cvttsh_u32): Likewise.
7066 (_mm_cvtt_roundsh_i32): Likewise.
7067 (_mm_cvtt_roundsh_u32): Likewise.
7068 (_mm_cvttsh_i64): Likewise.
7069 (_mm_cvttsh_u64): Likewise.
7070 (_mm_cvtt_roundsh_i64): Likewise.
7071 (_mm_cvtt_roundsh_u64): Likewise.
7072 * config/i386/i386-builtin.def: Add corresponding new builtins.
7073 * config/i386/sse.md
7074 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<round_saeonly_name>):
7077 2021-09-17 liuhongt <hongtao.liu@intel.com>
7079 * config/i386/avx512fp16intrin.h (_mm512_cvttph_epi32):
7081 (_mm512_mask_cvttph_epi32): Likewise.
7082 (_mm512_maskz_cvttph_epi32): Likewise.
7083 (_mm512_cvtt_roundph_epi32): Likewise.
7084 (_mm512_mask_cvtt_roundph_epi32): Likewise.
7085 (_mm512_maskz_cvtt_roundph_epi32): Likewise.
7086 (_mm512_cvttph_epu32): Likewise.
7087 (_mm512_mask_cvttph_epu32): Likewise.
7088 (_mm512_maskz_cvttph_epu32): Likewise.
7089 (_mm512_cvtt_roundph_epu32): Likewise.
7090 (_mm512_mask_cvtt_roundph_epu32): Likewise.
7091 (_mm512_maskz_cvtt_roundph_epu32): Likewise.
7092 (_mm512_cvttph_epi64): Likewise.
7093 (_mm512_mask_cvttph_epi64): Likewise.
7094 (_mm512_maskz_cvttph_epi64): Likewise.
7095 (_mm512_cvtt_roundph_epi64): Likewise.
7096 (_mm512_mask_cvtt_roundph_epi64): Likewise.
7097 (_mm512_maskz_cvtt_roundph_epi64): Likewise.
7098 (_mm512_cvttph_epu64): Likewise.
7099 (_mm512_mask_cvttph_epu64): Likewise.
7100 (_mm512_maskz_cvttph_epu64): Likewise.
7101 (_mm512_cvtt_roundph_epu64): Likewise.
7102 (_mm512_mask_cvtt_roundph_epu64): Likewise.
7103 (_mm512_maskz_cvtt_roundph_epu64): Likewise.
7104 (_mm512_cvttph_epi16): Likewise.
7105 (_mm512_mask_cvttph_epi16): Likewise.
7106 (_mm512_maskz_cvttph_epi16): Likewise.
7107 (_mm512_cvtt_roundph_epi16): Likewise.
7108 (_mm512_mask_cvtt_roundph_epi16): Likewise.
7109 (_mm512_maskz_cvtt_roundph_epi16): Likewise.
7110 (_mm512_cvttph_epu16): Likewise.
7111 (_mm512_mask_cvttph_epu16): Likewise.
7112 (_mm512_maskz_cvttph_epu16): Likewise.
7113 (_mm512_cvtt_roundph_epu16): Likewise.
7114 (_mm512_mask_cvtt_roundph_epu16): Likewise.
7115 (_mm512_maskz_cvtt_roundph_epu16): Likewise.
7116 * config/i386/avx512fp16vlintrin.h (_mm_cvttph_epi32):
7118 (_mm_mask_cvttph_epi32): Likewise.
7119 (_mm_maskz_cvttph_epi32): Likewise.
7120 (_mm256_cvttph_epi32): Likewise.
7121 (_mm256_mask_cvttph_epi32): Likewise.
7122 (_mm256_maskz_cvttph_epi32): Likewise.
7123 (_mm_cvttph_epu32): Likewise.
7124 (_mm_mask_cvttph_epu32): Likewise.
7125 (_mm_maskz_cvttph_epu32): Likewise.
7126 (_mm256_cvttph_epu32): Likewise.
7127 (_mm256_mask_cvttph_epu32): Likewise.
7128 (_mm256_maskz_cvttph_epu32): Likewise.
7129 (_mm_cvttph_epi64): Likewise.
7130 (_mm_mask_cvttph_epi64): Likewise.
7131 (_mm_maskz_cvttph_epi64): Likewise.
7132 (_mm256_cvttph_epi64): Likewise.
7133 (_mm256_mask_cvttph_epi64): Likewise.
7134 (_mm256_maskz_cvttph_epi64): Likewise.
7135 (_mm_cvttph_epu64): Likewise.
7136 (_mm_mask_cvttph_epu64): Likewise.
7137 (_mm_maskz_cvttph_epu64): Likewise.
7138 (_mm256_cvttph_epu64): Likewise.
7139 (_mm256_mask_cvttph_epu64): Likewise.
7140 (_mm256_maskz_cvttph_epu64): Likewise.
7141 (_mm_cvttph_epi16): Likewise.
7142 (_mm_mask_cvttph_epi16): Likewise.
7143 (_mm_maskz_cvttph_epi16): Likewise.
7144 (_mm256_cvttph_epi16): Likewise.
7145 (_mm256_mask_cvttph_epi16): Likewise.
7146 (_mm256_maskz_cvttph_epi16): Likewise.
7147 (_mm_cvttph_epu16): Likewise.
7148 (_mm_mask_cvttph_epu16): Likewise.
7149 (_mm_maskz_cvttph_epu16): Likewise.
7150 (_mm256_cvttph_epu16): Likewise.
7151 (_mm256_mask_cvttph_epu16): Likewise.
7152 (_mm256_maskz_cvttph_epu16): Likewise.
7153 * config/i386/i386-builtin.def: Add new builtins.
7154 * config/i386/sse.md
7155 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name><round_saeonly_name>):
7157 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name>): Ditto.
7158 (*avx512fp16_fix<fixunssuffix>_trunc<mode>2_load<mask_name>): Ditto.
7159 (avx512fp16_fix<fixunssuffix>_truncv2di2<mask_name>): Ditto.
7160 (avx512fp16_fix<fixunssuffix>_truncv2di2_load<mask_name>): Ditto.
7162 2021-09-17 liuhongt <hongtao.liu@intel.com>
7164 * config/i386/avx512fp16intrin.h (_mm_cvtsh_i32): New intrinsic.
7165 (_mm_cvtsh_u32): Likewise.
7166 (_mm_cvt_roundsh_i32): Likewise.
7167 (_mm_cvt_roundsh_u32): Likewise.
7168 (_mm_cvtsh_i64): Likewise.
7169 (_mm_cvtsh_u64): Likewise.
7170 (_mm_cvt_roundsh_i64): Likewise.
7171 (_mm_cvt_roundsh_u64): Likewise.
7172 (_mm_cvti32_sh): Likewise.
7173 (_mm_cvtu32_sh): Likewise.
7174 (_mm_cvt_roundi32_sh): Likewise.
7175 (_mm_cvt_roundu32_sh): Likewise.
7176 (_mm_cvti64_sh): Likewise.
7177 (_mm_cvtu64_sh): Likewise.
7178 (_mm_cvt_roundi64_sh): Likewise.
7179 (_mm_cvt_roundu64_sh): Likewise.
7180 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7181 * config/i386/i386-builtin.def: Add corresponding new builtins.
7182 * config/i386/i386-expand.c (ix86_expand_round_builtin):
7183 Handle new builtin types.
7184 * config/i386/sse.md
7185 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix><round_name>):
7187 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix>_2): Likewise.
7188 (avx512fp16_vcvt<floatsuffix>si2sh<rex64namesuffix><round_name>): Likewise.
7190 2021-09-16 Bill Schmidt <wschmidt@linux.ibm.com>
7192 * config/rs6000/rs6000-c.c (rs6000-builtins.h): New include.
7193 (altivec_resolve_new_overloaded_builtin): New forward decl.
7194 (rs6000_new_builtin_type_compatible): New function.
7195 (altivec_resolve_overloaded_builtin): Call
7196 altivec_resolve_new_overloaded_builtin.
7197 (altivec_build_new_resolved_builtin): New function.
7198 (altivec_resolve_new_overloaded_builtin): Likewise.
7199 * config/rs6000/rs6000-call.c (rs6000_new_builtin_is_supported):
7201 * config/rs6000/rs6000-gen-builtins.c (write_decls): Remove _p from
7202 name of rs6000_new_builtin_is_supported.
7204 2021-09-16 Uroš Bizjak <ubizjak@gmail.com>
7206 * config/i386/i386-protos.h (ix86_decompose_address):
7207 Change return type to bool.
7208 * config/i386/i386.c (ix86_decompose_address): Ditto.
7210 2021-09-16 Tobias Burnus <tobias@codesourcery.com>
7213 * config/rs6000/t-rs6000 (build/rs6000-gen-builtins.o, build/rbtree.o):
7214 Added 'build/' to target, use build/%.o rule.
7215 (build/rs6000-gen-builtins$(build_exeext)): Add 'build/' and
7216 '$(build_exeext)' to target and 'build/' for the *.o files.
7217 (rs6000-builtins.c): Update for those changes; run rs6000-gen-builtins
7220 2021-09-16 Martin Jambor <mjambor@suse.cz>
7222 * cgraph.c (cgraph_node::dump): Do not check caller count sums if
7223 the body has been removed. Remove trailing whitespace.
7225 2021-09-16 Richard Biener <rguenther@suse.de>
7227 PR middle-end/102360
7228 * internal-fn.c (expand_DEFERRED_INIT): Make pattern-init
7229 of non-memory more robust.
7231 2021-09-16 Daniel Cederman <cederman@gaisler.com>
7233 * config/sparc/sparc-opts.h (enum sparc_processor_type): Add LEON5
7234 * config/sparc/sparc.c (struct processor_costs): Add LEON5 costs
7235 (leon5_adjust_cost): Increase cost of store with data dependency
7236 on ALU instruction and FPU anti-dependencies.
7237 (sparc_option_override): Add LEON5 costs
7238 (sparc_adjust_cost): Add LEON5 cost adjustments
7239 * config/sparc/sparc.h: Add LEON5
7240 * config/sparc/sparc.md: Include LEON5 scheduling information
7241 * config/sparc/sparc.opt: Add LEON5
7242 * doc/invoke.texi: Add LEON5
7243 * config/sparc/leon5.md: New file.
7245 2021-09-16 Daniel Cederman <cederman@gaisler.com>
7247 * config/sparc/sparc.md (stack_protect_set32): Add NOP to prevent
7248 sensitive sequence for B2BST errata workaround.
7250 2021-09-16 Daniel Cederman <cederman@gaisler.com>
7252 * config/sparc/sparc.c (sparc_do_work_around_errata): Do not begin
7253 functions with atomic instruction in the UT700 errata workaround.
7255 2021-09-16 Daniel Cederman <cederman@gaisler.com>
7257 * config/sparc/sparc.c (next_active_non_empty_insn): New function
7258 that returns next active non empty assembly instruction.
7259 (sparc_do_work_around_errata): Use new function.
7261 2021-09-16 Daniel Cederman <cederman@gaisler.com>
7263 * config/sparc/sparc.c (store_insn_p): Add predicate for store
7265 (load_insn_p): Add predicate for load attributes.
7266 (sparc_do_work_around_errata): Use new predicates.
7268 2021-09-16 Andreas Larsson <andreas@gaisler.com>
7270 * config/sparc/sparc.c (dump_target_flag_bits): Print bit names for
7273 2021-09-16 Martin Liska <mliska@suse.cz>
7275 * config/mips/netbsd.h: Fix typo in name of a macro.
7277 2021-09-16 liuhongt <hongtao.liu@intel.com>
7279 PR middle-end/102080
7280 * match.pd: Check mask type when doing cond_op related gimple
7282 * tree.c (is_truth_type_for): New function.
7283 * tree.h (is_truth_type_for): New declaration.
7285 2021-09-16 liuhongt <hongtao.liu@intel.com>
7287 * config/i386/avx512fp16intrin.h (_mm512_cvtepi32_ph): New
7289 (_mm512_mask_cvtepi32_ph): Likewise.
7290 (_mm512_maskz_cvtepi32_ph): Likewise.
7291 (_mm512_cvt_roundepi32_ph): Likewise.
7292 (_mm512_mask_cvt_roundepi32_ph): Likewise.
7293 (_mm512_maskz_cvt_roundepi32_ph): Likewise.
7294 (_mm512_cvtepu32_ph): Likewise.
7295 (_mm512_mask_cvtepu32_ph): Likewise.
7296 (_mm512_maskz_cvtepu32_ph): Likewise.
7297 (_mm512_cvt_roundepu32_ph): Likewise.
7298 (_mm512_mask_cvt_roundepu32_ph): Likewise.
7299 (_mm512_maskz_cvt_roundepu32_ph): Likewise.
7300 (_mm512_cvtepi64_ph): Likewise.
7301 (_mm512_mask_cvtepi64_ph): Likewise.
7302 (_mm512_maskz_cvtepi64_ph): Likewise.
7303 (_mm512_cvt_roundepi64_ph): Likewise.
7304 (_mm512_mask_cvt_roundepi64_ph): Likewise.
7305 (_mm512_maskz_cvt_roundepi64_ph): Likewise.
7306 (_mm512_cvtepu64_ph): Likewise.
7307 (_mm512_mask_cvtepu64_ph): Likewise.
7308 (_mm512_maskz_cvtepu64_ph): Likewise.
7309 (_mm512_cvt_roundepu64_ph): Likewise.
7310 (_mm512_mask_cvt_roundepu64_ph): Likewise.
7311 (_mm512_maskz_cvt_roundepu64_ph): Likewise.
7312 (_mm512_cvtepi16_ph): Likewise.
7313 (_mm512_mask_cvtepi16_ph): Likewise.
7314 (_mm512_maskz_cvtepi16_ph): Likewise.
7315 (_mm512_cvt_roundepi16_ph): Likewise.
7316 (_mm512_mask_cvt_roundepi16_ph): Likewise.
7317 (_mm512_maskz_cvt_roundepi16_ph): Likewise.
7318 (_mm512_cvtepu16_ph): Likewise.
7319 (_mm512_mask_cvtepu16_ph): Likewise.
7320 (_mm512_maskz_cvtepu16_ph): Likewise.
7321 (_mm512_cvt_roundepu16_ph): Likewise.
7322 (_mm512_mask_cvt_roundepu16_ph): Likewise.
7323 (_mm512_maskz_cvt_roundepu16_ph): Likewise.
7324 * config/i386/avx512fp16vlintrin.h (_mm_cvtepi32_ph): New
7326 (_mm_mask_cvtepi32_ph): Likewise.
7327 (_mm_maskz_cvtepi32_ph): Likewise.
7328 (_mm256_cvtepi32_ph): Likewise.
7329 (_mm256_mask_cvtepi32_ph): Likewise.
7330 (_mm256_maskz_cvtepi32_ph): Likewise.
7331 (_mm_cvtepu32_ph): Likewise.
7332 (_mm_mask_cvtepu32_ph): Likewise.
7333 (_mm_maskz_cvtepu32_ph): Likewise.
7334 (_mm256_cvtepu32_ph): Likewise.
7335 (_mm256_mask_cvtepu32_ph): Likewise.
7336 (_mm256_maskz_cvtepu32_ph): Likewise.
7337 (_mm_cvtepi64_ph): Likewise.
7338 (_mm_mask_cvtepi64_ph): Likewise.
7339 (_mm_maskz_cvtepi64_ph): Likewise.
7340 (_mm256_cvtepi64_ph): Likewise.
7341 (_mm256_mask_cvtepi64_ph): Likewise.
7342 (_mm256_maskz_cvtepi64_ph): Likewise.
7343 (_mm_cvtepu64_ph): Likewise.
7344 (_mm_mask_cvtepu64_ph): Likewise.
7345 (_mm_maskz_cvtepu64_ph): Likewise.
7346 (_mm256_cvtepu64_ph): Likewise.
7347 (_mm256_mask_cvtepu64_ph): Likewise.
7348 (_mm256_maskz_cvtepu64_ph): Likewise.
7349 (_mm_cvtepi16_ph): Likewise.
7350 (_mm_mask_cvtepi16_ph): Likewise.
7351 (_mm_maskz_cvtepi16_ph): Likewise.
7352 (_mm256_cvtepi16_ph): Likewise.
7353 (_mm256_mask_cvtepi16_ph): Likewise.
7354 (_mm256_maskz_cvtepi16_ph): Likewise.
7355 (_mm_cvtepu16_ph): Likewise.
7356 (_mm_mask_cvtepu16_ph): Likewise.
7357 (_mm_maskz_cvtepu16_ph): Likewise.
7358 (_mm256_cvtepu16_ph): Likewise.
7359 (_mm256_mask_cvtepu16_ph): Likewise.
7360 (_mm256_maskz_cvtepu16_ph): Likewise.
7361 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7362 * config/i386/i386-builtin.def: Add corresponding new builtins.
7363 * config/i386/i386-expand.c
7364 (ix86_expand_args_builtin): Handle new builtin types.
7365 (ix86_expand_round_builtin): Ditto.
7366 * config/i386/i386-modes.def: Declare V2HF and V6HF.
7367 * config/i386/sse.md (VI2H_AVX512VL): New.
7369 (sseintvecmode): Add HF vector modes.
7370 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode><mask_name><round_name>):
7372 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
7373 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
7374 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
7375 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
7376 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask_1): Ditto.
7377 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
7378 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
7379 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
7380 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
7381 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask_1): Ditto.
7382 * config/i386/subst.md (round_qq2phsuff): New subst_attr.
7384 2021-09-16 liuhongt <hongtao.liu@intel.com>
7386 * config/i386/avx512fp16intrin.h (_mm512_cvtph_epi32):
7388 (_mm512_mask_cvtph_epi32): Likewise.
7389 (_mm512_maskz_cvtph_epi32): Likewise.
7390 (_mm512_cvt_roundph_epi32): Likewise.
7391 (_mm512_mask_cvt_roundph_epi32): Likewise.
7392 (_mm512_maskz_cvt_roundph_epi32): Likewise.
7393 (_mm512_cvtph_epu32): Likewise.
7394 (_mm512_mask_cvtph_epu32): Likewise.
7395 (_mm512_maskz_cvtph_epu32): Likewise.
7396 (_mm512_cvt_roundph_epu32): Likewise.
7397 (_mm512_mask_cvt_roundph_epu32): Likewise.
7398 (_mm512_maskz_cvt_roundph_epu32): Likewise.
7399 (_mm512_cvtph_epi64): Likewise.
7400 (_mm512_mask_cvtph_epi64): Likewise.
7401 (_mm512_maskz_cvtph_epi64): Likewise.
7402 (_mm512_cvt_roundph_epi64): Likewise.
7403 (_mm512_mask_cvt_roundph_epi64): Likewise.
7404 (_mm512_maskz_cvt_roundph_epi64): Likewise.
7405 (_mm512_cvtph_epu64): Likewise.
7406 (_mm512_mask_cvtph_epu64): Likewise.
7407 (_mm512_maskz_cvtph_epu64): Likewise.
7408 (_mm512_cvt_roundph_epu64): Likewise.
7409 (_mm512_mask_cvt_roundph_epu64): Likewise.
7410 (_mm512_maskz_cvt_roundph_epu64): Likewise.
7411 (_mm512_cvtph_epi16): Likewise.
7412 (_mm512_mask_cvtph_epi16): Likewise.
7413 (_mm512_maskz_cvtph_epi16): Likewise.
7414 (_mm512_cvt_roundph_epi16): Likewise.
7415 (_mm512_mask_cvt_roundph_epi16): Likewise.
7416 (_mm512_maskz_cvt_roundph_epi16): Likewise.
7417 (_mm512_cvtph_epu16): Likewise.
7418 (_mm512_mask_cvtph_epu16): Likewise.
7419 (_mm512_maskz_cvtph_epu16): Likewise.
7420 (_mm512_cvt_roundph_epu16): Likewise.
7421 (_mm512_mask_cvt_roundph_epu16): Likewise.
7422 (_mm512_maskz_cvt_roundph_epu16): Likewise.
7423 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_epi32):
7425 (_mm_mask_cvtph_epi32): Likewise.
7426 (_mm_maskz_cvtph_epi32): Likewise.
7427 (_mm256_cvtph_epi32): Likewise.
7428 (_mm256_mask_cvtph_epi32): Likewise.
7429 (_mm256_maskz_cvtph_epi32): Likewise.
7430 (_mm_cvtph_epu32): Likewise.
7431 (_mm_mask_cvtph_epu32): Likewise.
7432 (_mm_maskz_cvtph_epu32): Likewise.
7433 (_mm256_cvtph_epu32): Likewise.
7434 (_mm256_mask_cvtph_epu32): Likewise.
7435 (_mm256_maskz_cvtph_epu32): Likewise.
7436 (_mm_cvtph_epi64): Likewise.
7437 (_mm_mask_cvtph_epi64): Likewise.
7438 (_mm_maskz_cvtph_epi64): Likewise.
7439 (_mm256_cvtph_epi64): Likewise.
7440 (_mm256_mask_cvtph_epi64): Likewise.
7441 (_mm256_maskz_cvtph_epi64): Likewise.
7442 (_mm_cvtph_epu64): Likewise.
7443 (_mm_mask_cvtph_epu64): Likewise.
7444 (_mm_maskz_cvtph_epu64): Likewise.
7445 (_mm256_cvtph_epu64): Likewise.
7446 (_mm256_mask_cvtph_epu64): Likewise.
7447 (_mm256_maskz_cvtph_epu64): Likewise.
7448 (_mm_cvtph_epi16): Likewise.
7449 (_mm_mask_cvtph_epi16): Likewise.
7450 (_mm_maskz_cvtph_epi16): Likewise.
7451 (_mm256_cvtph_epi16): Likewise.
7452 (_mm256_mask_cvtph_epi16): Likewise.
7453 (_mm256_maskz_cvtph_epi16): Likewise.
7454 (_mm_cvtph_epu16): Likewise.
7455 (_mm_mask_cvtph_epu16): Likewise.
7456 (_mm_maskz_cvtph_epu16): Likewise.
7457 (_mm256_cvtph_epu16): Likewise.
7458 (_mm256_mask_cvtph_epu16): Likewise.
7459 (_mm256_maskz_cvtph_epu16): Likewise.
7460 * config/i386/i386-builtin-types.def: Add new builtin types.
7461 * config/i386/i386-builtin.def: Add new builtins.
7462 * config/i386/i386-expand.c
7463 (ix86_expand_args_builtin): Handle new builtin types.
7464 (ix86_expand_round_builtin): Ditto.
7465 * config/i386/sse.md (sseintconvert): New.
7467 (UNSPEC_US_FIX_NOTRUNC): Ditto.
7468 (sseintconvertsignprefix): Ditto.
7469 (avx512fp16_vcvtph2<sseintconvertsignprefix><sseintconvert>_<mode><mask_name><round_name>):
7472 2021-09-16 liuhongt <hongtao.liu@intel.com>
7474 * config/i386/avx512fp16intrin.h: (_mm_cvtsi16_si128):
7476 (_mm_cvtsi128_si16): Likewise.
7477 (_mm_mask_load_sh): Likewise.
7478 (_mm_maskz_load_sh): Likewise.
7479 (_mm_mask_store_sh): Likewise.
7480 (_mm_move_sh): Likewise.
7481 (_mm_mask_move_sh): Likewise.
7482 (_mm_maskz_move_sh): Likewise.
7483 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7484 * config/i386/i386-builtin.def: Add corresponding new builtins.
7485 * config/i386/i386-expand.c
7486 (ix86_expand_special_args_builtin): Handle new builtin types.
7487 (ix86_expand_vector_init_one_nonzero): Adjust for FP16 target.
7488 * config/i386/sse.md (VI2F): New mode iterator.
7489 (vec_set<mode>_0): Use new mode iterator.
7490 (avx512f_mov<ssescalarmodelower>_mask): Adjust for HF vector mode.
7491 (avx512f_store<mode>_mask): Ditto.
7493 2021-09-16 Kewen Lin <linkw@linux.ibm.com>
7495 * config/rs6000/rs6000.opt (-mtoc-fusion): Remove.
7497 2021-09-15 David Edelsohn <dje.gcc@gmail.com>
7499 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
7500 Proceed if no symbol summary or the symbol alias flag is false.
7502 2021-09-15 Jakub Jelinek <jakub@redhat.com>
7506 * varasm.c (output_constructor_regular_field): Instead of assertion
7507 that array_size_for_constructor result is equal to size of
7508 TREE_TYPE (local->val) in bytes, assert that the type size is greater
7509 or equal to array_size_for_constructor result and use type size as
7512 2021-09-15 Martin Liska <mliska@suse.cz>
7515 * config/i386/vxworks.h: Use new macro TARGET_CPU_P.
7517 2021-09-15 Martin Liska <mliska@suse.cz>
7520 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
7521 Check that we have a symbol summary for a symbol.
7523 2021-09-15 Richard Biener <rguenther@suse.de>
7526 * config/rs6000/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
7527 to inherit from elfos.h
7529 2021-09-15 liuhongt <hongtao.liu@intel.com>
7532 * config/i386/i386-expand.c
7533 (ix86_expand_vector_init_interleave): Use puncklwd to pack 2
7535 (ix86_expand_vector_set): Use blendw instead of pinsrw.
7536 * config/i386/i386.c (ix86_can_change_mode_class): Adjust for
7537 AVX512FP16 which supports 16bit vector load.
7538 * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>):
7540 (avx512bw_interleave_high<mode><mask_name>): .. this, and
7541 extend to V32HFmode.
7542 (avx2_interleave_highv16hi<mask_name>): Rename to ..
7543 (avx2_interleave_high<mode><mask_name>): .. this, and extend
7545 (vec_interleave_highv8hi<mask_name>): Rename to ..
7546 (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode.
7547 (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>):
7549 (<mask_codefor>avx512bw_interleave_low<mode><mask_name>):
7550 this, and extend to V32HFmode.
7551 (avx2_interleave_lowv16hi<mask_name>): Rename to ..
7552 (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode.
7553 (vec_interleave_lowv8hi<mask_name>): Rename to ..
7554 (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode.
7555 (sse4_1_pblendw): Rename to ..
7556 (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode.
7557 (avx2_pblendph): New define_expand.
7558 (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use
7559 sseintmodesuffix instead of ssemodesuffix.
7560 (blendsuf): New mode attr.
7562 2021-09-15 Richard Biener <rguenther@suse.de>
7564 * tree-vectorizer.h (dr_misalignment): Move out of line.
7565 (dr_target_alignment): New.
7566 (DR_TARGET_ALIGNMENT): Wrap dr_target_alignment.
7567 (set_dr_target_alignment): New.
7568 (SET_DR_TARGET_ALIGNMENT): Wrap set_dr_target_alignment.
7569 * tree-vect-data-refs.c (dr_misalignment): Compute and
7570 return the group members misalignment.
7571 (vect_compute_data_ref_alignment): Use SET_DR_TARGET_ALIGNMENT.
7572 (vect_analyze_data_refs_alignment): Compute alignment only
7573 for the first element of a DR group.
7574 (vect_slp_analyze_node_alignment): Likewise.
7576 2021-09-15 Hongyu Wang <hongyu.wang@intel.com>
7578 * config/i386/avx512fp16intrin.h: Adjust all builtin calls.
7579 * config/i386/avx512fp16vlintrin.h: Likewise.
7580 * config/i386/i386-builtin.def: Adjust builtin name and
7581 enumeration to match AVX512F style.
7583 2021-09-15 Richard Biener <rguenther@suse.de>
7585 PR tree-optimization/102318
7586 * tree-vect-loop.c (vect_transform_cycle_phi): Revert
7587 previous change and do the mode conversion separately from
7588 the sign conversion.
7590 2021-09-15 Hongtao Liu <hongtao.liu@intel.com>
7591 Peter Cordes <peter@cordes.ca>
7594 * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
7595 (*vec_extract<mode><ssescalarmodelower>_valign): Output
7596 vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
7599 2021-09-15 Richard Biener <rguenther@suse.de>
7601 * config.gcc: Remove vax-*-openbsd* configuration.
7603 2021-09-15 Richard Biener <rguenther@suse.de>
7605 * config.gcc: Remove m68k-openbsd.
7607 2021-09-15 Max Filippov <jcmvbkbc@gmail.com>
7610 * config/xtensa/t-xtensa (TM_H): Add include/xtensa-config.h.
7612 2021-09-14 Peter Bergner <bergner@linux.ibm.com>
7614 * config/rs6000/mma.md (unspec): Delete UNSPEC_MMA_XXSETACCZ.
7615 (unspecv): Add UNSPECV_MMA_XXSETACCZ.
7616 (*mma_xxsetaccz): Delete.
7617 (mma_xxsetaccz): Change to define_insn. Remove operand 1.
7618 Use UNSPECV_MMA_XXSETACCZ. Update comment.
7619 * config/rs6000/rs6000.c (rs6000_rtx_costs): Use UNSPECV_MMA_XXSETACCZ.
7621 2021-09-14 Iain Sandoe <iain@sandoe.co.uk>
7623 * Makefile.in: Remove variables related to applying no-PIE
7624 to the exes on $build.
7625 * configure: Regenerate.
7626 * configure.ac: Remove configuration related to applying
7627 no-PIE to the exes on $build.
7629 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
7631 * config/arc/arc.md (doloop_end): Add missing mode.
7632 (loop_end): Likewise.
7634 2021-09-14 Jakub Jelinek <jakub@redhat.com>
7636 * gimplify.c (goa_stabilize_expr): Add depth argument, propagate
7637 it to recursive calls, for depth above 7 just gimplify or return.
7638 Perform a test even for MODIFY_EXPR, ADDR_EXPR, COMPOUND_EXPR with
7639 __builtin_clear_padding and TARGET_EXPR.
7640 (gimplify_omp_atomic): Adjust goa_stabilize_expr callers.
7642 2021-09-14 liuhongt <hongtao.liu@intel.com>
7644 * config/i386/avx512fp16intrin.h (_mm_fpclass_sh_mask):
7646 (_mm_mask_fpclass_sh_mask): Likewise.
7647 (_mm512_mask_fpclass_ph_mask): Likewise.
7648 (_mm512_fpclass_ph_mask): Likewise.
7649 (_mm_getexp_sh): Likewise.
7650 (_mm_mask_getexp_sh): Likewise.
7651 (_mm_maskz_getexp_sh): Likewise.
7652 (_mm512_getexp_ph): Likewise.
7653 (_mm512_mask_getexp_ph): Likewise.
7654 (_mm512_maskz_getexp_ph): Likewise.
7655 (_mm_getexp_round_sh): Likewise.
7656 (_mm_mask_getexp_round_sh): Likewise.
7657 (_mm_maskz_getexp_round_sh): Likewise.
7658 (_mm512_getexp_round_ph): Likewise.
7659 (_mm512_mask_getexp_round_ph): Likewise.
7660 (_mm512_maskz_getexp_round_ph): Likewise.
7661 (_mm_getmant_sh): Likewise.
7662 (_mm_mask_getmant_sh): Likewise.
7663 (_mm_maskz_getmant_sh): Likewise.
7664 (_mm512_getmant_ph): Likewise.
7665 (_mm512_mask_getmant_ph): Likewise.
7666 (_mm512_maskz_getmant_ph): Likewise.
7667 (_mm_getmant_round_sh): Likewise.
7668 (_mm_mask_getmant_round_sh): Likewise.
7669 (_mm_maskz_getmant_round_sh): Likewise.
7670 (_mm512_getmant_round_ph): Likewise.
7671 (_mm512_mask_getmant_round_ph): Likewise.
7672 (_mm512_maskz_getmant_round_ph): Likewise.
7673 * config/i386/avx512fp16vlintrin.h (_mm_mask_fpclass_ph_mask):
7675 (_mm_fpclass_ph_mask): Likewise.
7676 (_mm256_mask_fpclass_ph_mask): Likewise.
7677 (_mm256_fpclass_ph_mask): Likewise.
7678 (_mm256_getexp_ph): Likewise.
7679 (_mm256_mask_getexp_ph): Likewise.
7680 (_mm256_maskz_getexp_ph): Likewise.
7681 (_mm_getexp_ph): Likewise.
7682 (_mm_mask_getexp_ph): Likewise.
7683 (_mm_maskz_getexp_ph): Likewise.
7684 (_mm256_getmant_ph): Likewise.
7685 (_mm256_mask_getmant_ph): Likewise.
7686 (_mm256_maskz_getmant_ph): Likewise.
7687 (_mm_getmant_ph): Likewise.
7688 (_mm_mask_getmant_ph): Likewise.
7689 (_mm_maskz_getmant_ph): Likewise.
7690 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7691 * config/i386/i386-builtin.def: Add corresponding new builtins.
7692 * config/i386/i386-expand.c
7693 (ix86_expand_args_builtin): Handle new builtin types.
7694 (ix86_expand_round_builtin): Ditto.
7695 * config/i386/sse.md (vecmemsuffix): Add HF vector modes.
7696 (<avx512>_getexp<mode><mask_name><round_saeonly_name>): Adjust
7697 to support HF vector modes.
7698 (avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name):
7700 (avx512dq_fpclass<mode><mask_scalar_merge_name>): Ditto.
7701 (avx512dq_vmfpclass<mode><mask_scalar_merge_name>): Ditto.
7702 (<avx512>_getmant<mode><mask_name><round_saeonly_name>): Ditto.
7703 (avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
7706 2021-09-14 liuhongt <hongtao.liu@intel.com>
7708 * config/i386/avx512fp16intrin.h (_mm512_reduce_ph):
7710 (_mm512_mask_reduce_ph): Likewise.
7711 (_mm512_maskz_reduce_ph): Likewise.
7712 (_mm512_reduce_round_ph): Likewise.
7713 (_mm512_mask_reduce_round_ph): Likewise.
7714 (_mm512_maskz_reduce_round_ph): Likewise.
7715 (_mm_reduce_sh): Likewise.
7716 (_mm_mask_reduce_sh): Likewise.
7717 (_mm_maskz_reduce_sh): Likewise.
7718 (_mm_reduce_round_sh): Likewise.
7719 (_mm_mask_reduce_round_sh): Likewise.
7720 (_mm_maskz_reduce_round_sh): Likewise.
7721 (_mm512_roundscale_ph): Likewise.
7722 (_mm512_mask_roundscale_ph): Likewise.
7723 (_mm512_maskz_roundscale_ph): Likewise.
7724 (_mm512_roundscale_round_ph): Likewise.
7725 (_mm512_mask_roundscale_round_ph): Likewise.
7726 (_mm512_maskz_roundscale_round_ph): Likewise.
7727 (_mm_roundscale_sh): Likewise.
7728 (_mm_mask_roundscale_sh): Likewise.
7729 (_mm_maskz_roundscale_sh): Likewise.
7730 (_mm_roundscale_round_sh): Likewise.
7731 (_mm_mask_roundscale_round_sh): Likewise.
7732 (_mm_maskz_roundscale_round_sh): Likewise.
7733 * config/i386/avx512fp16vlintrin.h: (_mm_reduce_ph):
7735 (_mm_mask_reduce_ph): Likewise.
7736 (_mm_maskz_reduce_ph): Likewise.
7737 (_mm256_reduce_ph): Likewise.
7738 (_mm256_mask_reduce_ph): Likewise.
7739 (_mm256_maskz_reduce_ph): Likewise.
7740 (_mm_roundscale_ph): Likewise.
7741 (_mm_mask_roundscale_ph): Likewise.
7742 (_mm_maskz_roundscale_ph): Likewise.
7743 (_mm256_roundscale_ph): Likewise.
7744 (_mm256_mask_roundscale_ph): Likewise.
7745 (_mm256_maskz_roundscale_ph): Likewise.
7746 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7747 * config/i386/i386-builtin.def: Add corresponding new builtins.
7748 * config/i386/i386-expand.c
7749 (ix86_expand_args_builtin): Handle new builtin types.
7750 (ix86_expand_round_builtin): Ditto.
7751 * config/i386/sse.md (<mask_codefor>reducep<mode><mask_name>):
7753 (<mask_codefor>reducep<mode><mask_name><round_saeonly_name>):
7754 ... this, and adjust for round operands.
7755 (reduces<mode><mask_scalar_name>): Likewise, with ...
7756 (reduces<mode><mask_scalar_name><round_saeonly_scalar_name):
7758 (<avx512>_rndscale<mode><mask_name><round_saeonly_name>):
7759 Adjust for HF vector modes.
7760 (avx512f_rndscale<mode><mask_scalar_name><round_saeonly_scalar_name>):
7762 (*avx512f_rndscale<mode><round_saeonly_name>): Ditto.
7764 2021-09-14 liuhongt <hongtao.liu@intel.com>
7766 * config/i386/avx512fp16intrin.h: (_mm512_rcp_ph):
7768 (_mm512_mask_rcp_ph): Likewise.
7769 (_mm512_maskz_rcp_ph): Likewise.
7770 (_mm_rcp_sh): Likewise.
7771 (_mm_mask_rcp_sh): Likewise.
7772 (_mm_maskz_rcp_sh): Likewise.
7773 (_mm512_scalef_ph): Likewise.
7774 (_mm512_mask_scalef_ph): Likewise.
7775 (_mm512_maskz_scalef_ph): Likewise.
7776 (_mm512_scalef_round_ph): Likewise.
7777 (_mm512_mask_scalef_round_ph): Likewise.
7778 (_mm512_maskz_scalef_round_ph): Likewise.
7779 (_mm_scalef_sh): Likewise.
7780 (_mm_mask_scalef_sh): Likewise.
7781 (_mm_maskz_scalef_sh): Likewise.
7782 (_mm_scalef_round_sh): Likewise.
7783 (_mm_mask_scalef_round_sh): Likewise.
7784 (_mm_maskz_scalef_round_sh): Likewise.
7785 * config/i386/avx512fp16vlintrin.h (_mm_rcp_ph):
7787 (_mm256_rcp_ph): Likewise.
7788 (_mm_mask_rcp_ph): Likewise.
7789 (_mm256_mask_rcp_ph): Likewise.
7790 (_mm_maskz_rcp_ph): Likewise.
7791 (_mm256_maskz_rcp_ph): Likewise.
7792 (_mm_scalef_ph): Likewise.
7793 (_mm256_scalef_ph): Likewise.
7794 (_mm_mask_scalef_ph): Likewise.
7795 (_mm256_mask_scalef_ph): Likewise.
7796 (_mm_maskz_scalef_ph): Likewise.
7797 (_mm256_maskz_scalef_ph): Likewise.
7798 * config/i386/i386-builtin.def: Add new builtins.
7799 * config/i386/sse.md (VFH_AVX512VL): New.
7800 (avx512fp16_rcp<mode>2<mask_name>): Ditto.
7801 (avx512fp16_vmrcpv8hf2<mask_scalar_name>): Ditto.
7802 (avx512f_vmscalef<mode><mask_scalar_name><round_scalar_name>):
7803 Adjust to support HF vector modes.
7804 (<avx512>_scalef<mode><mask_name><round_name>): Ditto.
7806 2021-09-14 liuhongt <hongtao.liu@intel.com>
7808 * config/i386/avx512fp16intrin.h: (_mm512_sqrt_ph):
7810 (_mm512_mask_sqrt_ph): Likewise.
7811 (_mm512_maskz_sqrt_ph): Likewise.
7812 (_mm512_sqrt_round_ph): Likewise.
7813 (_mm512_mask_sqrt_round_ph): Likewise.
7814 (_mm512_maskz_sqrt_round_ph): Likewise.
7815 (_mm512_rsqrt_ph): Likewise.
7816 (_mm512_mask_rsqrt_ph): Likewise.
7817 (_mm512_maskz_rsqrt_ph): Likewise.
7818 (_mm_rsqrt_sh): Likewise.
7819 (_mm_mask_rsqrt_sh): Likewise.
7820 (_mm_maskz_rsqrt_sh): Likewise.
7821 (_mm_sqrt_sh): Likewise.
7822 (_mm_mask_sqrt_sh): Likewise.
7823 (_mm_maskz_sqrt_sh): Likewise.
7824 (_mm_sqrt_round_sh): Likewise.
7825 (_mm_mask_sqrt_round_sh): Likewise.
7826 (_mm_maskz_sqrt_round_sh): Likewise.
7827 * config/i386/avx512fp16vlintrin.h (_mm_sqrt_ph): New intrinsic.
7828 (_mm256_sqrt_ph): Likewise.
7829 (_mm_mask_sqrt_ph): Likewise.
7830 (_mm256_mask_sqrt_ph): Likewise.
7831 (_mm_maskz_sqrt_ph): Likewise.
7832 (_mm256_maskz_sqrt_ph): Likewise.
7833 (_mm_rsqrt_ph): Likewise.
7834 (_mm256_rsqrt_ph): Likewise.
7835 (_mm_mask_rsqrt_ph): Likewise.
7836 (_mm256_mask_rsqrt_ph): Likewise.
7837 (_mm_maskz_rsqrt_ph): Likewise.
7838 (_mm256_maskz_rsqrt_ph): Likewise.
7839 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
7840 * config/i386/i386-builtin.def: Add corresponding new builtins.
7841 * config/i386/i386-expand.c
7842 (ix86_expand_args_builtin): Handle new builtins.
7843 (ix86_expand_round_builtin): Ditto.
7844 * config/i386/sse.md (VF_AVX512FP16VL): New.
7845 (sqrt<mode>2): Adjust for HF vector modes.
7846 (<sse>_sqrt<mode>2<mask_name><round_name>): Likewise.
7847 (<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
7849 (<sse>_rsqrt<mode>2<mask_name>): New.
7850 (avx512fp16_vmrsqrtv8hf2<mask_scalar_name>): Likewise.
7852 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
7855 * diagnostic-spec.c (warning_suppressed_at, copy_warning): Handle
7856 'RESERVED_LOCATION_P' locations.
7857 * warning-control.cc (get_nowarn_spec, suppress_warning)
7858 (copy_warning): Likewise.
7860 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
7862 * diagnostic-spec.h (typedef xint_hash_t): Use 'location_t' instead of...
7863 (typedef key_type_t): ... this. Remove.
7864 (nowarn_map): Document.
7865 * diagnostic-spec.c (nowarn_map): Likewise.
7866 * warning-control.cc (convert_to_key): Evolve functions into...
7867 (get_location): ... these. Adjust all users.
7869 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
7871 * warning-control.cc (copy_warning): Remove 'nowarn_map' setup.
7873 2021-09-13 Jason Merrill <jason@redhat.com>
7875 * params.opt: Add destructive-interference-size and
7876 constructive-interference-size.
7877 * doc/invoke.texi: Document them.
7878 * config/aarch64/aarch64.c (aarch64_override_options_internal):
7880 * config/arm/arm.c (arm_option_override): Set them.
7881 * config/i386/i386-options.c (ix86_option_override_internal):
7884 2021-09-13 Martin Liska <mliska@suse.cz>
7885 H.J. Lu <hjl.tools@gmail.com>
7888 * common/config/i386/cpuinfo.h (cpu_indicator_init): Add support
7889 for x86-64 micro levels for __builtin_cpu_supports.
7890 * common/config/i386/i386-cpuinfo.h (enum feature_priority):
7891 Add priorities for the micro-arch levels.
7892 (enum processor_features): Add new features.
7893 * common/config/i386/i386-isas.h: Add micro-arch features.
7894 * config/i386/i386-builtins.c (get_builtin_code_for_version):
7895 Support the micro-arch levels by callsing
7896 __builtin_cpu_supports.
7897 * doc/extend.texi: Document that the levels are support by
7898 __builtin_cpu_supports.
7900 2021-09-13 Andrew Pinski <apinski@marvell.com>
7903 * config/aarch64/aarch64-builtins.c (aarch64_fold_builtin_lane_check):
7905 (aarch64_general_fold_builtin): Handle AARCH64_SIMD_BUILTIN_LANE_CHECK.
7906 (aarch64_general_gimple_fold_builtin): Likewise.
7908 2021-09-13 Andrew Pinski <apinski@marvell.com>
7910 * config.gcc: Add m32r-*-linux* and m32rle-*-linux*
7911 to the Unsupported targets list.
7912 Remove support for m32r-*-linux* and m32rle-*-linux*.
7913 * config/m32r/linux.h: Removed.
7914 * config/m32r/t-linux: Removed.
7916 2021-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7919 * config/aarch64/aarch64.c (aarch64_classify_address): Don't allow
7920 register index for SVE predicate modes.
7922 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
7924 * tree-ssa-threadbackward.c
7925 (back_threader_profitability::profitable_path_p): Remove FSM
7927 (back_threader_registry::register_path): Same.
7928 * tree-ssa-threadedge.c
7929 (jump_threader::simplify_control_stmt_condition): Same.
7930 * tree-ssa-threadupdate.c (jt_path_registry::jt_path_registry):
7931 Add backedge_threads argument.
7932 (fwd_jt_path_registry::fwd_jt_path_registry): Pass
7933 backedge_threads argument.
7934 (back_jt_path_registry::back_jt_path_registry): Same.
7935 (dump_jump_thread_path): Adjust for FSM removal.
7936 (back_jt_path_registry::rewire_first_differing_edge): Same.
7937 (back_jt_path_registry::adjust_paths_after_duplication): Same.
7938 (back_jt_path_registry::update_cfg): Same.
7939 (jt_path_registry::register_jump_thread): Same.
7940 * tree-ssa-threadupdate.h (enum jump_thread_edge_type): Remove
7942 (class back_jt_path_registry): Add backedge_threads to
7945 2021-09-13 Martin Liska <mliska@suse.cz>
7948 * asan.h (sanitize_coverage_p): Handle when fn == NULL.
7950 2021-09-13 H.J. Lu <hjl.tools@gmail.com>
7953 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): New.
7954 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
7955 (MOVE_MAX): Check TARGET_AVX256_MOVE_BY_PIECES and
7956 TARGET_AVX256_STORE_BY_PIECES instead of
7957 TARGET_AVX256_SPLIT_UNALIGNED_LOAD and
7958 TARGET_AVX256_SPLIT_UNALIGNED_STORE.
7959 (STORE_MAX_PIECES): Check TARGET_AVX256_STORE_BY_PIECES instead
7960 of TARGET_AVX256_SPLIT_UNALIGNED_STORE.
7961 * config/i386/x86-tune.def (X86_TUNE_AVX256_MOVE_BY_PIECES): New.
7962 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
7964 2021-09-13 liuhongt <hongtao.liu@intel.com>
7967 * expmed.c (extract_bit_field_using_extv): Use
7968 gen_lowpart_if_possible instead of gen_lowpart to avoid ICE.
7970 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
7972 * Makefile.in (OBJS): Add value-pointer-equiv.o.
7973 * gimple-ssa-evrp.c (class ssa_equiv_stack): Move to
7974 value-pointer-equiv.*.
7975 (ssa_equiv_stack::ssa_equiv_stack): Same.
7976 (ssa_equiv_stack::enter): Same.
7977 (ssa_equiv_stack::leave): Same.
7978 (ssa_equiv_stack::push_replacement): Same.
7979 (ssa_equiv_stack::get_replacement): Same.
7980 (is_pointer_ssa): Same.
7981 (class pointer_equiv_analyzer): Same.
7982 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
7983 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Same.
7984 (pointer_equiv_analyzer::set_global_equiv): Same.
7985 (pointer_equiv_analyzer::set_cond_equiv): Same.
7986 (pointer_equiv_analyzer::get_equiv): Same.
7987 (pointer_equiv_analyzer::enter): Same.
7988 (pointer_equiv_analyzer::leave): Same.
7989 (pointer_equiv_analyzer::get_equiv_expr): Same.
7990 (pta_valueize): Same.
7991 (pointer_equiv_analyzer::visit_stmt): Same.
7992 (pointer_equiv_analyzer::visit_edge): Same.
7993 (hybrid_folder::value_of_expr): Same.
7994 (hybrid_folder::value_on_edge): Same.
7995 * value-pointer-equiv.cc: New file.
7996 * value-pointer-equiv.h: New file.
7998 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
8001 * gimple-fold.c (gimple_fold_builtin_memory_op): Allow folding
8002 memcpy if the size is not more than MOVE_MAX * MOVE_RATIO.
8004 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
8007 * config/arm/arm.md (movmisaligndi): New define_expand.
8008 * config/arm/vec-common.md (movmisalign<mode>): Iterate over VDQ mode.
8010 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
8013 * emit-rtl.c (gen_highpart): Use adjust_address to handle
8014 MEM rather than calling simplify_gen_subreg.
8016 2021-09-13 Jan-Benedict Glaw <jbglaw@ług-owl.de>
8018 * config/alpha/vms.h (INIT_CUMULATIVE_ARGS): Wrap multi-statment
8019 define into a block.
8021 2021-09-13 Richard Biener <rguenther@suse.de>
8023 * config/darwin.h (DARWIN_PREFER_DWARF): Do not define.
8024 * config/i386/darwin.h (PREFERRED_DEBUGGING_TYPE): Do not
8025 change based on DARWIN_PREFER_DWARF not being defined.
8027 2021-09-13 Richard Biener <rguenther@suse.de>
8029 * config/i386/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
8030 to inherit from elfos.h
8032 2021-09-13 Richard Biener <rguenther@suse.de>
8034 * config.gcc: Add cr16-*-* to the list of obsoleted targets.
8036 2021-09-13 Richard Biener <rguenther@suse.de>
8038 * config/avr/elf.h (PREFERRED_DEBUGGING_TYPE): Remove
8039 override, pick up DWARF2_DEBUG define from elfos.h
8041 2021-09-13 Richard Biener <rguenther@suse.de>
8043 * config/rx/rx.h (PREFERRED_DEBUGGING_TYPE): Always define to
8046 2021-09-13 Richard Biener <rguenther@suse.de>
8048 * config/alpha/vms.h (PREFERRED_DEBUGGING_TYPE): Define to
8051 2021-09-13 Richard Biener <rguenther@suse.de>
8053 * config/i386/cygming.h: Always default to DWARF2 debugging.
8054 Do not define DBX_DEBUGGING_INFO, that's done via dbxcoff.h
8056 * doc/install.texi: Document binutils 2.16 as minimum
8057 requirement for mingw.
8059 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
8061 * config/rs6000/rs6000.c (struct rs6000_cost_data): New members
8062 nstmts, nloads and extra_ctor_cost.
8063 (rs6000_density_test): Add load density related heuristics. Do
8064 extra costing on vector construction statements if need.
8065 (rs6000_init_cost): Init new members.
8066 (rs6000_update_target_cost_per_stmt): New function.
8067 (rs6000_add_stmt_cost): Factor vect_nonmem hunk out to function
8068 rs6000_update_target_cost_per_stmt and call it.
8070 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
8072 * config/rs6000/rs6000.c (struct rs6000_cost_data): Remove typedef.
8073 (rs6000_init_cost): Adjust.
8075 2021-09-13 liuhongt <hongtao.liu@intel.com>
8077 * config/i386/i386.md: (UNSPEC_COPYSIGN): Remove.
8078 (UNSPEC_XORSIGN): Ditto.
8080 2021-09-12 Roger Sayle <roger@nextmovesoftware.com>
8082 * expr.c (convert_move): Preserve SUBREG_PROMOTED_VAR_P when
8083 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
8086 2021-09-11 Aldy Hernandez <aldyh@redhat.com>
8088 * tree-ssa-threadbackward.c (class back_threader_registry): Use
8089 back_jt_path_registry.
8090 * tree-ssa-threadedge.c (jump_threader::jump_threader): Use
8091 fwd_jt_path_registry.
8092 * tree-ssa-threadedge.h (class jump_threader): Same..
8093 * tree-ssa-threadupdate.c
8094 (jump_thread_path_registry::jump_thread_path_registry): Rename...
8095 (jt_path_registry::jt_path_registry): ...to this.
8096 (jump_thread_path_registry::~jump_thread_path_registry): Rename...
8097 (jt_path_registry::~jt_path_registry): ...this.
8098 (fwd_jt_path_registry::fwd_jt_path_registry): New.
8099 (fwd_jt_path_registry::~fwd_jt_path_registry): New.
8100 (jump_thread_path_registry::allocate_thread_edge): Rename...
8101 (jt_path_registry::allocate_thread_edge): ...to this.
8102 (jump_thread_path_registry::allocate_thread_path): Rename...
8103 (jt_path_registry::allocate_thread_path): ...to this.
8104 (jump_thread_path_registry::lookup_redirection_data): Rename...
8105 (fwd_jt_path_registry::lookup_redirection_data): ...to this.
8106 (jump_thread_path_registry::thread_block_1): Rename...
8107 (fwd_jt_path_registry::thread_block_1): ...to this.
8108 (jump_thread_path_registry::thread_block): Rename...
8109 (fwd_jt_path_registry::thread_block): ...to this.
8110 (jt_path_registry::thread_through_loop_header): Rename...
8111 (fwd_jt_path_registry::thread_through_loop_header): ...to this.
8112 (jump_thread_path_registry::mark_threaded_blocks): Rename...
8113 (fwd_jt_path_registry::mark_threaded_blocks): ...to this.
8114 (jump_thread_path_registry::debug_path): Rename...
8115 (jt_path_registry::debug_path): ...to this.
8116 (jump_thread_path_registry::dump): Rename...
8117 (jt_path_registry::debug): ...to this.
8118 (jump_thread_path_registry::rewire_first_differing_edge): Rename...
8119 (back_jt_path_registry::rewire_first_differing_edge): ...to this.
8120 (jump_thread_path_registry::adjust_paths_after_duplication): Rename...
8121 (back_jt_path_registry::adjust_paths_after_duplication): ...to this.
8122 (jump_thread_path_registry::duplicate_thread_path): Rename...
8123 (back_jt_path_registry::duplicate_thread_path): ...to this. Also,
8124 drop ill-formed candidates.
8125 (jump_thread_path_registry::remove_jump_threads_including): Rename...
8126 (fwd_jt_path_registry::remove_jump_threads_including): ...to this.
8127 (jt_path_registry::thread_through_all_blocks): New.
8128 (back_jt_path_registry::update_cfg): New.
8129 (fwd_jt_path_registry::update_cfg): New.
8130 (jump_thread_path_registry::register_jump_thread): Rename...
8131 (jt_path_registry::register_jump_thread): ...to this.
8132 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
8134 (class jt_path_registry): ...here.
8135 (class fwd_jt_path_registry): New.
8136 (class back_jt_path_registry): New.
8138 2021-09-10 liuhongt <hongtao.liu@intel.com>
8141 2021-09-01 liuhongt <hongtao.liu@intel.com>
8143 * emit-rtl.c (validate_subreg): Get rid of all float-int
8146 2021-09-10 Jakub Jelinek <jakub@redhat.com>
8148 * tree-core.h (enum omp_memory_order): Add OMP_MEMORY_ORDER_MASK,
8149 OMP_FAIL_MEMORY_ORDER_UNSPECIFIED, OMP_FAIL_MEMORY_ORDER_RELAXED,
8150 OMP_FAIL_MEMORY_ORDER_ACQUIRE, OMP_FAIL_MEMORY_ORDER_RELEASE,
8151 OMP_FAIL_MEMORY_ORDER_ACQ_REL, OMP_FAIL_MEMORY_ORDER_SEQ_CST and
8152 OMP_FAIL_MEMORY_ORDER_MASK enumerators.
8153 (OMP_FAIL_MEMORY_ORDER_SHIFT): Define.
8154 * gimple-pretty-print.c (dump_gimple_omp_atomic_load,
8155 dump_gimple_omp_atomic_store): Print [weak] for weak atomic
8157 * gimple.h (enum gf_mask): Change GF_OMP_ATOMIC_MEMORY_ORDER
8158 to 6-bit mask, adjust GF_OMP_ATOMIC_NEED_VALUE value and add
8160 (gimple_omp_atomic_weak_p, gimple_omp_atomic_set_weak): New inline
8162 * tree.h (OMP_ATOMIC_WEAK): Define.
8163 * tree-pretty-print.c (dump_omp_atomic_memory_order): Adjust for
8164 fail memory order being encoded in the same enum and also print
8165 fail clause if present.
8166 (dump_generic_node): Print weak clause if OMP_ATOMIC_WEAK.
8167 * gimplify.c (goa_stabilize_expr): Add target_expr and rhs arguments,
8168 handle pre_p == NULL case as a test mode that only returns value
8169 but doesn't change gimplify nor change anything otherwise, adjust
8170 recursive calls, add MODIFY_EXPR, ADDR_EXPR, COND_EXPR, TARGET_EXPR
8171 and CALL_EXPR handling, adjust COMPOUND_EXPR handling for
8172 __builtin_clear_padding calls, for !rhs gimplify as lvalue rather
8174 (gimplify_omp_atomic): Adjust goa_stabilize_expr caller. Handle
8175 COND_EXPR rhs. Set weak flag on gimple load/store for
8177 * omp-expand.c (omp_memory_order_to_fail_memmodel): New function.
8178 (omp_memory_order_to_memmodel): Adjust for fail clause encoded
8180 (expand_omp_atomic_cas): New function.
8181 (expand_omp_atomic_pipeline): Use omp_memory_order_to_fail_memmodel
8183 (expand_omp_atomic): Attempt to optimize atomic compare and exchange
8184 using expand_omp_atomic_cas.
8186 2021-09-10 Aldy Hernandez <aldyh@redhat.com>
8187 Michael Matz <matz@suse.de>
8189 * tree-pass.h (PROP_loop_opts_done): New.
8190 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
8191 Intersect with global range.
8192 * tree-ssa-loop.c (tree_ssa_loop_done): Set PROP_loop_opts_done.
8193 * tree-ssa-threadbackward.c
8194 (back_threader_profitability::profitable_path_p): Disable
8195 threading through latches until after loop optimizations have run.
8197 2021-09-10 David Faust <david.faust@oracle.com>
8199 * doc/invoke.texi: Document BPF -mcpu, -mjmpext, -mjmp32 and -malu32
8202 2021-09-10 David Faust <david.faust@oracle.com>
8204 * config/bpf/bpf-opts.h (bpf_isa_version): New enum.
8205 * config/bpf/bpf-protos.h (bpf_expand_cbranch): New.
8206 * config/bpf/bpf.c (bpf_option_override): Handle -mcpu option.
8207 (bpf_expand_cbranch): New function.
8208 * config/bpf/bpf.md (AM mode iterator): Conditionalize support for SI
8210 (zero_extendsidi2): Only use mov32 instruction if it is available.
8211 (SIM mode iterator): Conditionalize support for SI mode.
8212 (JM mode iterator): New.
8213 (cbranchdi4): Update name, use new JM iterator. Use bpf_expand_cbranch.
8214 (*branch_on_di): Update name, use new JM iterator.
8215 * config/bpf/bpf.opt: (mjmpext): New option.
8219 (bpf_isa): New enum.
8221 2021-09-10 David Faust <david.faust@oracle.com>
8223 * config/bpf/bpf.md (zero_extendhidi2): Add new output template
8224 for register-to-register extensions.
8225 (zero_extendqidi2): Likewise.
8227 2021-09-10 Richard Biener <rguenther@suse.de>
8229 PR middle-end/102273
8230 * internal-fn.c (expand_DEFERRED_INIT): Always expand non-SSA vars.
8232 2021-09-10 Richard Biener <rguenther@suse.de>
8234 PR middle-end/102269
8235 * gimplify.c (is_var_need_auto_init): Empty types do not need
8238 2021-09-10 Richard Biener <rguenther@suse.de>
8240 * configure.ac (--with-stabs): Remove.
8241 * configure: Regenerate.
8242 * doc/install.texi: Remove --with-stabs documentation.
8244 2021-09-10 liuhongt <hongtao.liu@intel.com>
8246 * config/i386/avx512fp16intrin.h: (_mm512_cmp_ph_mask):
8248 (_mm512_mask_cmp_ph_mask): Likewise.
8249 (_mm512_cmp_round_ph_mask): Likewise.
8250 (_mm512_mask_cmp_round_ph_mask): Likewise.
8251 (_mm_cmp_sh_mask): Likewise.
8252 (_mm_mask_cmp_sh_mask): Likewise.
8253 (_mm_cmp_round_sh_mask): Likewise.
8254 (_mm_mask_cmp_round_sh_mask): Likewise.
8255 (_mm_comieq_sh): Likewise.
8256 (_mm_comilt_sh): Likewise.
8257 (_mm_comile_sh): Likewise.
8258 (_mm_comigt_sh): Likewise.
8259 (_mm_comige_sh): Likewise.
8260 (_mm_comineq_sh): Likewise.
8261 (_mm_ucomieq_sh): Likewise.
8262 (_mm_ucomilt_sh): Likewise.
8263 (_mm_ucomile_sh): Likewise.
8264 (_mm_ucomigt_sh): Likewise.
8265 (_mm_ucomige_sh): Likewise.
8266 (_mm_ucomineq_sh): Likewise.
8267 (_mm_comi_round_sh): Likewise.
8268 (_mm_comi_sh): Likewise.
8269 * config/i386/avx512fp16vlintrin.h (_mm_cmp_ph_mask): New intrinsic.
8270 (_mm_mask_cmp_ph_mask): Likewise.
8271 (_mm256_cmp_ph_mask): Likewise.
8272 (_mm256_mask_cmp_ph_mask): Likewise.
8273 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
8274 * config/i386/i386-builtin.def: Add corresponding new builtins.
8275 * config/i386/i386-expand.c
8276 (ix86_expand_args_builtin): Handle new builtin types.
8277 (ix86_expand_round_builtin): Ditto.
8278 * config/i386/i386.md (ssevecmode): Add HF mode.
8279 (MODEFH): New mode iterator.
8280 * config/i386/sse.md
8281 (V48H_AVX512VL): New mode iterator to support HF vector modes.
8282 Ajdust corresponding description.
8283 (ssecmpintprefix): New.
8284 (VI12_AVX512VL): Adjust to support HF vector modes.
8285 (cmp_imm_predicate): Likewise.
8286 (<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
8288 (avx512f_vmcmp<mode>3<round_saeonly_name>): Likewise.
8289 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Likewise.
8290 (<sse>_<unord>comi<round_saeonly_name>): Likewise.
8292 2021-09-10 liuhongt <hongtao.liu@intel.com>
8294 * config/i386/avx512fp16intrin.h: (_mm512_max_ph): New intrinsic.
8295 (_mm512_mask_max_ph): Likewise.
8296 (_mm512_maskz_max_ph): Likewise.
8297 (_mm512_min_ph): Likewise.
8298 (_mm512_mask_min_ph): Likewise.
8299 (_mm512_maskz_min_ph): Likewise.
8300 (_mm512_max_round_ph): Likewise.
8301 (_mm512_mask_max_round_ph): Likewise.
8302 (_mm512_maskz_max_round_ph): Likewise.
8303 (_mm512_min_round_ph): Likewise.
8304 (_mm512_mask_min_round_ph): Likewise.
8305 (_mm512_maskz_min_round_ph): Likewise.
8306 (_mm_max_sh): Likewise.
8307 (_mm_mask_max_sh): Likewise.
8308 (_mm_maskz_max_sh): Likewise.
8309 (_mm_min_sh): Likewise.
8310 (_mm_mask_min_sh): Likewise.
8311 (_mm_maskz_min_sh): Likewise.
8312 (_mm_max_round_sh): Likewise.
8313 (_mm_mask_max_round_sh): Likewise.
8314 (_mm_maskz_max_round_sh): Likewise.
8315 (_mm_min_round_sh): Likewise.
8316 (_mm_mask_min_round_sh): Likewise.
8317 (_mm_maskz_min_round_sh): Likewise.
8318 * config/i386/avx512fp16vlintrin.h (_mm_max_ph): New intrinsic.
8319 (_mm256_max_ph): Likewise.
8320 (_mm_mask_max_ph): Likewise.
8321 (_mm256_mask_max_ph): Likewise.
8322 (_mm_maskz_max_ph): Likewise.
8323 (_mm256_maskz_max_ph): Likewise.
8324 (_mm_min_ph): Likewise.
8325 (_mm256_min_ph): Likewise.
8326 (_mm_mask_min_ph): Likewise.
8327 (_mm256_mask_min_ph): Likewise.
8328 (_mm_maskz_min_ph): Likewise.
8329 (_mm256_maskz_min_ph): Likewise.
8330 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
8331 * config/i386/i386-builtin.def: Add corresponding new builtins.
8332 * config/i386/i386-expand.c
8333 (ix86_expand_args_builtin): Handle new builtin types.
8334 * config/i386/sse.md
8335 (<code><mode>3<mask_name><round_saeonly_name>): Adjust to
8336 support HF vector modes.
8337 (*<code><mode>3<mask_name><round_saeonly_name>): Likewise.
8338 (ieee_<ieee_maxmin><mode>3<mask_name><round_saeonly_name>):
8340 (<sse>_vm<code><mode>3<mask_scalar_name><round_saeonly_scalar_name>):
8342 * config/i386/subst.md (round_saeonly_mode512bit_condition):
8343 Adjust for HF vector modes.
8345 2021-09-10 Liu, Hongtao <hongtao.liu@intel.com>
8347 * config/i386/avx512fp16intrin.h (_mm_add_sh): New intrinsic.
8348 (_mm_mask_add_sh): Likewise.
8349 (_mm_maskz_add_sh): Likewise.
8350 (_mm_sub_sh): Likewise.
8351 (_mm_mask_sub_sh): Likewise.
8352 (_mm_maskz_sub_sh): Likewise.
8353 (_mm_mul_sh): Likewise.
8354 (_mm_mask_mul_sh): Likewise.
8355 (_mm_maskz_mul_sh): Likewise.
8356 (_mm_div_sh): Likewise.
8357 (_mm_mask_div_sh): Likewise.
8358 (_mm_maskz_div_sh): Likewise.
8359 (_mm_add_round_sh): Likewise.
8360 (_mm_mask_add_round_sh): Likewise.
8361 (_mm_maskz_add_round_sh): Likewise.
8362 (_mm_sub_round_sh): Likewise.
8363 (_mm_mask_sub_round_sh): Likewise.
8364 (_mm_maskz_sub_round_sh): Likewise.
8365 (_mm_mul_round_sh): Likewise.
8366 (_mm_mask_mul_round_sh): Likewise.
8367 (_mm_maskz_mul_round_sh): Likewise.
8368 (_mm_div_round_sh): Likewise.
8369 (_mm_mask_div_round_sh): Likewise.
8370 (_mm_maskz_div_round_sh): Likewise.
8371 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
8372 * config/i386/i386-builtin.def: Add corresponding new builtins.
8373 * config/i386/i386-expand.c
8374 (ix86_expand_round_builtin): Handle new builtins.
8375 * config/i386/sse.md (VF_128): Change description.
8376 (<sse>_vm<plusminus_insn><mode>3<mask_scalar_name><round_scalar_name>):
8377 Adjust to support HF vector modes.
8378 (<sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name><round_scalar_name>):
8381 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
8383 * config/i386/i386-expand.c
8384 (ix86_avx256_split_vector_move_misalign): Handle V16HF mode.
8385 * config/i386/i386.c
8386 (ix86_preferred_simd_mode): Handle HF mode.
8387 * config/i386/sse.md (V_256H): New mode iterator.
8388 (avx_vextractf128<mode>): Use it.
8389 (VEC_INIT_MODE): Align vector HFmode condition to vector
8390 HImodes since there're no real HF instruction used.
8391 (VEC_INIT_HALF_MODE): Ditto.
8393 (VIHF_AVX512BW): Ditto.
8394 (*vec_extracthf): Ditto.
8395 (VEC_EXTRACT_MODE): Ditto.
8397 2021-09-10 Richard Biener <rguenther@suse.de>
8400 * config/dbx.h: Remove.
8401 * config/dbxcoff.h: Do not define PREFERRED_DEBUGGING_TYPE.
8402 * config/lynx.h: Likewise.
8404 2021-09-10 liuhongt <hongtao.liu@intel.com>
8406 * config/i386/i386-expand.c (ix86_expand_copysign): Expand
8407 right into ANDNOT + AND + IOR, using paradoxical subregs.
8408 (ix86_split_copysign_const): Remove.
8409 (ix86_split_copysign_var): Ditto.
8410 * config/i386/i386-protos.h (ix86_split_copysign_const): Dotto.
8411 (ix86_split_copysign_var): Ditto.
8412 * config/i386/i386.md (@copysign<mode>3_const): Ditto.
8413 (@copysign<mode>3_var): Ditto.
8415 2021-09-09 qing zhao <qing.zhao@oracle.com>
8417 * builtins.c (expand_builtin_memset): Make external visible.
8418 * builtins.h (expand_builtin_memset): Declare extern.
8419 * common.opt (ftrivial-auto-var-init=): New option.
8420 * doc/extend.texi: Document the uninitialized attribute.
8421 * doc/invoke.texi: Document -ftrivial-auto-var-init.
8422 * flag-types.h (enum auto_init_type): New enumerated type
8424 * gimple-fold.c (clear_padding_type): Add one new parameter.
8425 (clear_padding_union): Likewise.
8426 (clear_padding_emit_loop): Likewise.
8427 (clear_type_padding_in_mask): Likewise.
8428 (gimple_fold_builtin_clear_padding): Handle this new parameter.
8429 * gimplify.c (gimple_add_init_for_auto_var): New function.
8430 (gimple_add_padding_init_for_auto_var): New function.
8431 (is_var_need_auto_init): New function.
8432 (gimplify_decl_expr): Add initialization to automatic variables per
8434 (gimplify_call_expr): Add one new parameter for call to
8435 __builtin_clear_padding.
8436 (gimplify_init_constructor): Add padding initialization in the end.
8437 * internal-fn.c (INIT_PATTERN_VALUE): New macro.
8438 (expand_DEFERRED_INIT): New function.
8439 * internal-fn.def (DEFERRED_INIT): New internal function.
8440 * tree-cfg.c (verify_gimple_call): Verify calls to .DEFERRED_INIT.
8441 * tree-sra.c (generate_subtree_deferred_init): New function.
8442 (scan_function): Avoid setting cannot_scalarize_away_bitmap for
8443 calls to .DEFERRED_INIT.
8444 (sra_modify_deferred_init): New function.
8445 (sra_modify_function_body): Handle calls to DEFERRED_INIT specially.
8446 * tree-ssa-structalias.c (find_func_aliases_for_call): Likewise.
8447 * tree-ssa-uninit.c (warn_uninit): Handle calls to DEFERRED_INIT
8449 (check_defs): Likewise.
8450 (warn_uninitialized_vars): Likewise.
8451 * tree-ssa.c (ssa_undefined_value_p): Likewise.
8452 * tree.c (build_common_builtin_nodes): Build tree node for
8453 BUILT_IN_CLEAR_PADDING when needed.
8455 2021-09-09 Richard Biener <rguenther@suse.de>
8457 * tree-ssa-loop-im.c (fill_always_executed_in_1): Walk
8460 2021-09-09 Richard Biener <rguenther@suse.de>
8462 * tree-ssa-loop-im.c (fill_always_executed_in_1): Integrate
8463 DOM walk from get_loop_body_in_dom_order using a worklist
8466 2021-09-09 liuhongt <hongtao.liu@intel.com>
8468 * config.gcc: Add avx512fp16vlintrin.h.
8469 * config/i386/avx512fp16intrin.h: (_mm512_add_ph): New intrinsic.
8470 (_mm512_mask_add_ph): Likewise.
8471 (_mm512_maskz_add_ph): Likewise.
8472 (_mm512_sub_ph): Likewise.
8473 (_mm512_mask_sub_ph): Likewise.
8474 (_mm512_maskz_sub_ph): Likewise.
8475 (_mm512_mul_ph): Likewise.
8476 (_mm512_mask_mul_ph): Likewise.
8477 (_mm512_maskz_mul_ph): Likewise.
8478 (_mm512_div_ph): Likewise.
8479 (_mm512_mask_div_ph): Likewise.
8480 (_mm512_maskz_div_ph): Likewise.
8481 (_mm512_add_round_ph): Likewise.
8482 (_mm512_mask_add_round_ph): Likewise.
8483 (_mm512_maskz_add_round_ph): Likewise.
8484 (_mm512_sub_round_ph): Likewise.
8485 (_mm512_mask_sub_round_ph): Likewise.
8486 (_mm512_maskz_sub_round_ph): Likewise.
8487 (_mm512_mul_round_ph): Likewise.
8488 (_mm512_mask_mul_round_ph): Likewise.
8489 (_mm512_maskz_mul_round_ph): Likewise.
8490 (_mm512_div_round_ph): Likewise.
8491 (_mm512_mask_div_round_ph): Likewise.
8492 (_mm512_maskz_div_round_ph): Likewise.
8493 * config/i386/avx512fp16vlintrin.h: New header.
8494 * config/i386/i386-builtin-types.def (V16HF, V8HF, V32HF):
8495 Add new builtin types.
8496 * config/i386/i386-builtin.def: Add corresponding builtins.
8497 * config/i386/i386-expand.c
8498 (ix86_expand_args_builtin): Handle new builtin types.
8499 (ix86_expand_round_builtin): Likewise.
8500 * config/i386/immintrin.h: Include avx512fp16vlintrin.h
8501 * config/i386/sse.md (VFH): New mode_iterator.
8503 (avx512fmaskmode): Add HF vector modes.
8504 (avx512fmaskhalfmode): Likewise.
8505 (<plusminus_insn><mode>3<mask_name><round_name>): Adjust to for
8507 (*<plusminus_insn><mode>3<mask_name><round_name>): Likewise.
8508 (mul<mode>3<mask_name><round_name>): Likewise.
8509 (*mul<mode>3<mask_name><round_name>): Likewise.
8510 (div<mode>3): Likewise.
8511 (<sse>_div<mode>3<mask_name><round_name>): Likewise.
8512 * config/i386/subst.md (SUBST_V): Add HF vector modes.
8513 (SUBST_A): Likewise.
8514 (round_mode512bit_condition): Adjust for V32HFmode.
8516 2021-09-09 liuhongt <hongtao.liu@intel.com>
8519 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
8520 (reduc_plus_scal_v4sf): .. this, New define_expand.
8521 (reduc_plus_scal_v2df): .. and this, New define_expand.
8523 2021-09-09 liuhongt <hongtao.liu@intel.com>
8526 * config/i386/sse.md (*vec_extract<mode><ssescalarmodelower>_valign):
8529 2021-09-08 Jonathan Wakely <jwakely@redhat.com>
8532 * doc/trouble.texi (Copy Assignment): Fix description of
8533 behaviour and fix code in example.
8535 2021-09-08 Segher Boessenkool <segher@kernel.crashing.org>
8538 * config/rs6000/rs6000-logue.c (rs6000_emit_epilogue): For ELFv2 use
8539 r11 instead of r12 for restoring CR.
8541 2021-09-08 Jakub Jelinek <jakub@redhat.com>
8542 liuhongt <hongtao.liu@intel.com>
8545 * config/i386/i386.md (@xorsign<mode>3_1): Remove.
8546 * config/i386/i386-expand.c (ix86_expand_xorsign): Expand right away
8547 into AND with mask and XOR, using paradoxical subregs.
8548 (ix86_split_xorsign): Remove.
8549 * config/i386/i386-protos.h (ix86_split_xorsign): Remove.
8551 2021-09-08 Di Zhao <dizhao@os.amperecomputing.com>
8553 * tree-ssa-sccvn.c (vn_nary_op_insert_into): fix result compare
8555 2021-09-08 Jakub Jelinek <jakub@redhat.com>
8558 * config/i386/i386.md (xorsign<mode>3): If operands[1] is equal to
8559 operands[2], emit abs<mode>2 instead.
8560 (@xorsign<mode>3_1): Add early-clobbers for output operand, enable
8561 first alternative even for avx, add another alternative with
8562 =&Yv <- 0, Yv, Yvm constraints.
8563 * config/i386/i386-expand.c (ix86_split_xorsign): If op0 is equal
8564 to op1, emit vpandn instead.
8566 2021-09-08 liuhongt <hongtao.liu@intel.com>
8568 * config/i386/avx512fp16intrin.h (_mm_set_ph): New intrinsic.
8569 (_mm256_set_ph): Likewise.
8570 (_mm512_set_ph): Likewise.
8571 (_mm_setr_ph): Likewise.
8572 (_mm256_setr_ph): Likewise.
8573 (_mm512_setr_ph): Likewise.
8574 (_mm_set1_ph): Likewise.
8575 (_mm256_set1_ph): Likewise.
8576 (_mm512_set1_ph): Likewise.
8577 (_mm_setzero_ph): Likewise.
8578 (_mm256_setzero_ph): Likewise.
8579 (_mm512_setzero_ph): Likewise.
8580 (_mm_set_sh): Likewise.
8581 (_mm_load_sh): Likewise.
8582 (_mm_store_sh): Likewise.
8583 * config/i386/i386-builtin-types.def (V8HF): New type.
8584 (DEF_FUNCTION_TYPE (V8HF, V8HI)): New builtin function type
8585 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
8586 Support vector HFmodes.
8587 (ix86_expand_vector_init_one_nonzero): Likewise.
8588 (ix86_expand_vector_init_one_var): Likewise.
8589 (ix86_expand_vector_init_interleave): Likewise.
8590 (ix86_expand_vector_init_general): Likewise.
8591 (ix86_expand_vector_set): Likewise.
8592 (ix86_expand_vector_extract): Likewise.
8593 (ix86_expand_vector_init_concat): Likewise.
8594 (ix86_expand_sse_movcc): Handle vector HFmodes.
8595 (ix86_expand_vector_set_var): Ditto.
8596 * config/i386/i386-modes.def: Add HF vector modes in comment.
8597 * config/i386/i386.c (classify_argument): Add HF vector modes.
8598 (ix86_hard_regno_mode_ok): Allow HF vector modes for AVX512FP16.
8599 (ix86_vector_mode_supported_p): Likewise.
8600 (ix86_set_reg_reg_cost): Handle vector HFmode.
8601 (ix86_get_ssemov): Handle vector HFmode.
8602 (function_arg_advance_64): Pass unamed V16HFmode and V32HFmode
8604 (function_arg_advance_32): Pass V8HF/V16HF/V32HF by sse reg for 32bit
8606 (function_arg_advance_32): Ditto.
8607 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): New.
8608 (VALID_AVX256_REG_OR_OI_MODE): Rename to ..
8609 (VALID_AVX256_REG_OR_OI_VHF_MODE): .. this, and add V16HF.
8610 (VALID_SSE2_REG_VHF_MODE): New.
8611 (VALID_AVX512VL_128_REG_MODE): Add V8HF and TImode.
8612 (SSE_REG_MODE_P): Add vector HFmode.
8613 * config/i386/i386.md (mode): Add HF vector modes.
8614 (MODE_SIZE): Likewise.
8615 (ssemodesuffix): Add ph suffix for HF vector modes.
8616 * config/i386/sse.md (VFH_128): New mode iterator.
8617 (VMOVE): Adjust for HF vector modes.
8619 (V_256_512): Likewise.
8621 (avx512fmaskmode): Likewise.
8622 (shuffletype): Likewise.
8623 (sseinsnmode): Likewise.
8624 (ssedoublevecmode): Likewise.
8625 (ssehalfvecmode): Likewise.
8626 (ssehalfvecmodelower): Likewise.
8627 (ssePScmode): Likewise.
8628 (ssescalarmode): Likewise.
8629 (ssescalarmodelower): Likewise.
8630 (sseintprefix): Likewise.
8632 (bcstscalarsuff): Likewise.
8633 (xtg_mode): Likewise.
8634 (VI12HF_AVX512VL): New mode_iterator.
8635 (VF_AVX512FP16): Likewise.
8637 (VIHF_256): Likewise.
8638 (VIHF_AVX512BW): Likewise.
8639 (V16_256): Likewise.
8640 (V32_512): Likewise.
8641 (sseintmodesuffix): New mode_attr.
8642 (sse): Add scalar and vector HFmodes.
8643 (ssescalarmode): Add vector HFmode mapping.
8644 (ssescalarmodesuffix): Add sh suffix for HFmode.
8645 (*<sse>_vm<insn><mode>3): Use VFH_128.
8646 (*<sse>_vm<multdiv_mnemonic><mode>3): Likewise.
8647 (*ieee_<ieee_maxmin><mode>3): Likewise.
8648 (<avx512>_blendm<mode>): New define_insn.
8649 (vec_setv8hf): New define_expand.
8650 (vec_set<mode>_0): New define_insn for HF vector set.
8651 (*avx512fp16_movsh): Likewise.
8652 (avx512fp16_movsh): Likewise.
8653 (vec_extract_lo_v32hi): Rename to ...
8654 (vec_extract_lo_<mode>): ... this, and adjust to allow HF
8656 (vec_extract_hi_v32hi): Likewise.
8657 (vec_extract_hi_<mode>): Likewise.
8658 (vec_extract_lo_v16hi): Likewise.
8659 (vec_extract_lo_<mode>): Likewise.
8660 (vec_extract_hi_v16hi): Likewise.
8661 (vec_extract_hi_<mode>): Likewise.
8662 (vec_set_hi_v16hi): Likewise.
8663 (vec_set_hi_<mode>): Likewise.
8664 (vec_set_lo_v16hi): Likewise.
8665 (vec_set_lo_<mode>): Likewise.
8666 (*vec_extract<mode>_0): New define_insn_and_split for HF
8668 (*vec_extracthf): New define_insn.
8669 (VEC_EXTRACT_MODE): Add HF vector modes.
8670 (PINSR_MODE): Add V8HF.
8671 (sse2p4_1): Likewise.
8672 (pinsr_evex_isa): Likewise.
8673 (<sse2p4_1>_pinsr<ssemodesuffix>): Adjust to support
8674 insert for V8HFmode.
8675 (pbroadcast_evex_isa): Add HF vector modes.
8676 (AVX2_VEC_DUP_MODE): Likewise.
8677 (VEC_INIT_MODE): Likewise.
8678 (VEC_INIT_HALF_MODE): Likewise.
8679 (avx2_pbroadcast<mode>): Adjust to support HF vector mode
8681 (avx2_pbroadcast<mode>_1): Likewise.
8682 (<avx512>_vec_dup<mode>_1): Likewise.
8683 (<avx512>_vec_dup<mode><mask_name>): Likewise.
8684 (<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>):
8687 2021-09-08 Guo, Xuepeng <xuepeng.guo@intel.com>
8688 H.J. Lu <hongjiu.lu@intel.com>
8689 Liu Hongtao <hongtao.liu@intel.com>
8690 Wang Hongyu <hongyu.wang@intel.com>
8691 Xu Dianhong <dianhong.xu@intel.com>
8693 * common/config/i386/cpuinfo.h (get_available_features):
8694 Detect FEATURE_AVX512FP16.
8695 * common/config/i386/i386-common.c
8696 (OPTION_MASK_ISA_AVX512FP16_SET,
8697 OPTION_MASK_ISA_AVX512FP16_UNSET,
8698 OPTION_MASK_ISA2_AVX512FP16_SET,
8699 OPTION_MASK_ISA2_AVX512FP16_UNSET): New.
8700 (OPTION_MASK_ISA2_AVX512BW_UNSET,
8701 OPTION_MASK_ISA2_AVX512BF16_UNSET): Add AVX512FP16.
8702 (ix86_handle_option): Handle -mavx512fp16.
8703 * common/config/i386/i386-cpuinfo.h (enum processor_features):
8704 Add FEATURE_AVX512FP16.
8705 * common/config/i386/i386-isas.h: Add entry for AVX512FP16.
8706 * config.gcc: Add avx512fp16intrin.h.
8707 * config/i386/avx512fp16intrin.h: New intrinsic header.
8708 * config/i386/cpuid.h: Add bit_AVX512FP16.
8709 * config/i386/i386-builtin-types.def: (FLOAT16): New primitive type.
8710 * config/i386/i386-builtins.c: Support _Float16 type for i386
8712 (ix86_register_float16_builtin_type): New function.
8713 (ix86_float16_type_node): New.
8714 * config/i386/i386-c.c (ix86_target_macros_internal): Define
8716 * config/i386/i386-expand.c (ix86_expand_branch): Support
8718 (ix86_prepare_fp_compare_args): Adjust TARGET_SSE_MATH &&
8719 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
8720 (ix86_expand_fp_movcc): Ditto.
8721 * config/i386/i386-isa.def: Add PTA define for AVX512FP16.
8722 * config/i386/i386-options.c (isa2_opts): Add -mavx512fp16.
8723 (ix86_valid_target_attribute_inner_p): Add avx512fp16 attribute.
8724 * config/i386/i386.c (ix86_get_ssemov): Use
8725 vmovdqu16/vmovw/vmovsh for HFmode/HImode scalar or vector.
8726 (ix86_get_excess_precision): Use
8727 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when TARGET_AVX512FP16
8729 (sse_store_index): Use SFmode cost for HFmode cost.
8730 (inline_memory_move_cost): Add HFmode, and perfer SSE cost over
8731 GPR cost for HFmode.
8732 (ix86_hard_regno_mode_ok): Allow HImode in sse register.
8733 (ix86_mangle_type): Add manlging for _Float16 type.
8734 (inline_secondary_memory_needed): No memory is needed for
8735 16bit movement between gpr and sse reg under
8737 (ix86_multiplication_cost): Adjust TARGET_SSE_MATH &&
8738 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
8739 (ix86_division_cost): Ditto.
8740 (ix86_rtx_costs): Ditto.
8741 (ix86_add_stmt_cost): Ditto.
8742 (ix86_optab_supported_p): Ditto.
8743 * config/i386/i386.h (VALID_AVX512F_SCALAR_MODE): Add HFmode.
8744 (SSE_FLOAT_MODE_SSEMATH_OR_HF_P): Add HFmode.
8745 (PTA_SAPPHIRERAPIDS): Add PTA_AVX512FP16.
8746 * config/i386/i386.md (mode): Add HFmode.
8747 (MODE_SIZE): Add HFmode.
8748 (isa): Add avx512fp16.
8749 (enabled): Handle avx512fp16.
8750 (ssemodesuffix): Add sh suffix for HFmode.
8751 (comm): Add mult, div.
8752 (plusminusmultdiv): New code iterator.
8753 (insn): Add mult, div.
8754 (*movhf_internal): Adjust for avx512fp16 instruction.
8755 (*movhi_internal): Ditto.
8756 (*cmpi<unord>hf): New define_insn for HFmode.
8757 (*ieee_s<ieee_maxmin>hf3): Likewise.
8758 (extendhf<mode>2): Likewise.
8759 (trunc<mode>hf2): Likewise.
8760 (float<floatunssuffix><mode>hf2): Likewise.
8761 (*<insn>hf): Likewise.
8762 (cbranchhf4): New expander.
8763 (movhfcc): Likewise.
8764 (<insn>hf3): Likewise.
8767 * config/i386/i386.opt: Add mavx512fp16.
8768 * config/i386/immintrin.h: Include avx512fp16intrin.h.
8769 * doc/invoke.texi: Add mavx512fp16.
8770 * doc/extend.texi: Add avx512fp16 Usage Notes.
8772 2021-09-08 liuhongt <hongtao.liu@intel.com>
8774 * common.opt: Support -fexcess-precision=16.
8775 * config/aarch64/aarch64.c (aarch64_excess_precision): Return
8776 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when
8777 EXCESS_PRECISION_TYPE_FLOAT16.
8778 * config/arm/arm.c (arm_excess_precision): Ditto.
8779 * config/i386/i386.c (ix86_get_excess_precision): Ditto.
8780 * config/m68k/m68k.c (m68k_excess_precision): Issue an error
8781 when EXCESS_PRECISION_TYPE_FLOAT16.
8782 * config/s390/s390.c (s390_excess_precision): Ditto.
8783 * coretypes.h (enum excess_precision_type): Add
8784 EXCESS_PRECISION_TYPE_FLOAT16.
8785 * doc/tm.texi (TARGET_C_EXCESS_PRECISION): Update documents.
8786 * doc/tm.texi.in (TARGET_C_EXCESS_PRECISION): Ditto.
8787 * doc/extend.texi (Half-Precision): Document
8788 -fexcess-precision=16.
8789 * flag-types.h (enum excess_precision): Add
8790 EXCESS_PRECISION_FLOAT16.
8791 * target.def (excess_precision): Update document.
8792 * tree.c (excess_precision_type): Set excess_precision_type to
8793 EXCESS_PRECISION_FLOAT16 when -fexcess-precision=16.
8795 2021-09-08 liuhongt <hongtao.liu@intel.com>
8797 * doc/extend.texi: (@node Floating Types): Adjust the wording.
8798 (@node Half-Precision): Ditto.
8800 2021-09-07 Takayuki 'January June' Suwa <jjsuwa_sys3175@yahoo.co.jp>
8803 * config/xtensa/xtensa.c (xtensa_emit_move_sequence): Add
8804 'CONST_INT_P (src)' to the condition of the block that tries to
8805 eliminate literal when loading integer contant.
8807 2021-09-07 David Faust <david.faust@oracle.com>
8809 * doc/extend.texi (BPF Type Attributes) New node.
8810 Document new preserve_access_index attribute.
8811 Document new preserve_access_index builtin.
8812 * doc/invoke.texi: Document -mco-re and -mno-co-re options.
8814 2021-09-07 David Faust <david.faust@oracle.com>
8816 * config/bpf/bpf.c: Adjust includes.
8817 (bpf_handle_preserve_access_index_attribute): New function.
8818 (bpf_attribute_table): Use it here.
8819 (bpf_builtins): Add BPF_BUILTIN_PRESERVE_ACCESS_INDEX.
8820 (bpf_option_override): Handle "-mco-re" option.
8821 (bpf_asm_init_sections): New.
8822 (TARGET_ASM_INIT_SECTIONS): Redefine.
8823 (bpf_file_end): New.
8824 (TARGET_ASM_FILE_END): Redefine.
8825 (bpf_init_builtins): Add "__builtin_preserve_access_index".
8826 (bpf_core_compute, bpf_core_get_index): New.
8827 (is_attr_preserve_access): New.
8828 (bpf_expand_builtin): Handle new builtins.
8829 (bpf_core_newdecl, bpf_core_is_maybe_aggregate_access): New.
8830 (bpf_core_walk): New.
8831 (bpf_resolve_overloaded_builtin): New.
8832 (TARGET_RESOLVE_OVERLOADED_BUILTIN): Redefine.
8834 (pass_bpf_core_attr): New RTL pass.
8835 * config/bpf/bpf-passes.def: New file.
8836 * config/bpf/bpf-protos.h (make_pass_bpf_core_attr): New.
8837 * config/bpf/coreout.c: New file.
8838 * config/bpf/coreout.h: Likewise.
8839 * config/bpf/t-bpf (TM_H): Add $(srcdir)/config/bpf/coreout.h.
8840 (coreout.o): New rule.
8841 (PASSES_EXTRA): Add $(srcdir)/config/bpf/bpf-passes.def.
8842 * config.gcc (bpf): Add coreout.h to extra_headers.
8843 Add coreout.o to extra_objs.
8844 Add $(srcdir)/config/bpf/coreout.c to target_gtfiles.
8846 2021-09-07 David Faust <david.faust@oracle.com>
8848 * btfout.c (get_btf_id): Function is no longer static.
8849 * ctfc.h: Expose it here.
8851 2021-09-07 David Faust <david.faust@oracle.com>
8853 * ctfc.c (ctf_lookup_tree_type): New function.
8856 2021-09-07 David Faust <david.faust@oracle.com>
8858 * ctfc.c (ctf_dtd_lookup): Function is no longer static.
8859 * ctfc.h: Analogous change.
8861 2021-09-07 David Faust <david.faust@oracle.com>
8863 * dwarf2out.c (lookup_type_die): Function is no longer static.
8864 * dwarf2out.h: Expose it here.
8866 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
8868 * dwarf2ctf.c (ctf_debug_finalize): Make it static.
8869 (ctf_debug_early_finish): New definition.
8870 (ctf_debug_finish): Likewise.
8871 * dwarf2ctf.h (ctf_debug_finalize): Remove declaration.
8872 (ctf_debug_early_finish): New declaration.
8873 (ctf_debug_finish): Likewise.
8874 * dwarf2out.c (dwarf2out_finish): Invoke ctf_debug_finish.
8875 (dwarf2out_early_finish): Invoke ctf_debug_early_finish.
8877 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
8879 * config/bpf/bpf.c (bpf_option_override): For BPF backend, disable LTO
8880 support when compiling for CO-RE.
8881 * config/bpf/bpf.opt: Add new command line option -mco-re.
8883 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
8885 * flag-types.h (enum debug_info_type): Add new enum
8886 DINFO_TYPE_BTF_WITH_CORE.
8887 (BTF_WITH_CORE_DEBUG): New bitmask.
8888 * flags.h (btf_with_core_debuginfo_p): New declaration.
8889 * opts.c (btf_with_core_debuginfo_p): New definition.
8891 2021-09-07 Jason Merrill <jason@redhat.com>
8893 * tree.h (error_operand_p): Change to inline function.
8895 2021-09-07 Aldy Hernandez <aldyh@redhat.com>
8897 * tree-ssa-threadedge.c (forwarder_block_p): Rename to...
8898 (empty_block_with_phis_p): ...this.
8899 (potentially_threadable_block): Same.
8900 (jump_threader::thread_through_normal_block): Same.
8902 2021-09-07 Eric Botcazou <ebotcazou@adacore.com>
8905 * dwarf2out.c (mark_base_types): New overloaded function.
8906 (dwarf2out_early_finish): Invoke it on the COMDAT type list as well
8907 as the compilation unit, and call move_marked_base_types afterward.
8909 2021-09-07 H.J. Lu <hjl.tools@gmail.com>
8912 * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
8914 (ix86_expand_vector_convert_uns_vsivsf): Likewise.
8916 2021-09-07 Richard Biener <rguenther@suse.de>
8918 PR tree-optimization/102226
8919 * tree-vect-loop.c (vect_transform_cycle_phi): Record
8920 the converted value for the epilogue PHI use.
8922 2021-09-07 Martin Liska <mliska@suse.cz>
8924 PR gcov-profile/80223
8925 * ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
8926 options, do not inline when no_profile_instrument_function
8927 attributes are different in early inliner. It's fine to inline
8928 it after PGO instrumentation.
8930 2021-09-07 Richard Biener <rguenther@suse.de>
8932 PR tree-optimization/101555
8933 * tree-ssa-pre.c (translate_vuse_through_block): Do not
8934 perform an alias walk to determine the validity of the
8935 mem at the start of the block which is already guaranteed
8936 by means of prune_clobbered_mems.
8937 (phi_translate_1): Pass edge to translate_vuse_through_block.
8939 2021-09-07 Xionghu Luo <luoxhu@linux.ibm.com>
8942 * config/rs6000/rs6000.md (fmod<mode>3): New define_expand.
8943 (remainder<mode>3): Likewise.
8945 2021-09-07 YunQiang Su <yunqiang.su@cipunited.com>
8947 * config/mips/mips.c (mips_file_start): add .module for
8950 2021-09-06 Roger Sayle <roger@nextmovesoftware.com>
8952 * wide-int.cc (wi::clz): Reorder tests to ensure the result
8953 is zero for all negative values.
8955 2021-09-06 Tobias Burnus <tobias@codesourcery.com>
8957 * doc/invoke.texi (-foffload-options): Fix @opindex.
8959 2021-09-06 H.J. Lu <hjl.tools@gmail.com>
8962 * config/i386/i386-expand.c (ix86_split_xorsign): Use operands[2].
8963 * config/i386/i386.md (@xorsign<mode>3_1): Add non-destructive
8964 source alternative for AVX.
8966 2021-09-06 liuhongt <hongtao.liu@intel.com>
8968 PR middle-end/102182
8969 * optabs.c (expand_fix): Add from1 to avoid from being
8972 2021-09-06 Eric Botcazou <ebotcazou@adacore.com>
8974 * dwarf2out.c (modified_type_die): Deal with all array types earlier
8975 and use local variable consistently throughout the function.
8977 2021-09-06 Jakub Jelinek <jakub@redhat.com>
8979 PR tree-optimization/102207
8980 * match.pd: Don't demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
8981 were promoted from signed to wider unsigned type.
8983 2021-09-06 Andrew Pinski <apinski@marvell.com>
8985 PR tree-optimization/63184
8986 * match.pd: Add simplification of pointer_diff of two pointer_plus
8987 with addr_expr in the first operand of each pointer_plus.
8988 Add simplificatoin of ne/eq of two pointer_plus with addr_expr
8989 in the first operand of each pointer_plus.
8991 2021-09-06 Richard Biener <rguenther@suse.de>
8993 PR tree-optimization/102176
8994 * tree-vect-slp.c (vect_slp_gather_vectorized_scalar_stmts):
8996 (vect_bb_slp_scalar_cost): Use the computed set of
8997 vectorized scalar stmts instead of relying on the out-of-date
8998 and not accurate PURE_SLP_STMT.
8999 (vect_bb_vectorization_profitable_p): Compute the set
9000 of vectorized scalar stmts.
9002 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
9004 * gimple-range-path.cc (path_range_query::range_of_stmt): Remove
9005 GIMPLE_COND special casing.
9006 (path_range_query::range_defined_in_block): Use range_of_stmt
9007 instead of calling fold_range directly.
9009 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
9011 * gimple-range-path.cc (path_range_query::range_of_expr): Set
9012 m_undefined_path when appropriate.
9013 (path_range_query::internal_range_of_expr): Copy from range_of_expr.
9014 (path_range_query::unreachable_path_p): New.
9015 (path_range_query::precompute_ranges): Set m_undefined_path.
9016 * gimple-range-path.h (path_range_query::unreachable_path_p): New.
9017 (path_range_query::internal_range_of_expr): New.
9018 * tree-ssa-threadbackward.c (back_threader::find_taken_edge_cond):
9019 Use unreachable_path_p.
9021 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
9023 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
9024 Remove argument and call find_taken_edge.
9025 (back_threader::resolve_phi): Do not calculate taken edge before
9026 calling maybe_register_path.
9027 (back_threader::find_paths_to_names): Same.
9029 2021-09-05 Jeff Law <jlaw@localhost.localdomain>
9031 * config/h8300/h8300.md (QHSI2 mode iterator): New mode iterator.
9032 * config/h8300/testcompare.md (store_c): Update name, use new
9034 (store_neg_c, store_shifted_c): New patterns.
9036 2021-09-03 Segher Boessenkool <segher@kernel.crashing.org>
9039 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): On ELFv2 use r11
9040 instead of r12 for CR save, in all cases.
9042 2021-09-03 Andrew Pinski <apinski@marvell.com>
9044 * config/aarch64/aarch64-sve-builtins.cc (register_vector_type):
9045 Handle error_mark_node as the type of the type_decl.
9047 2021-09-03 Andrew Pinski <apinski@marvell.com>
9049 * config/aarch64/aarch64-builtins.c (struct aarch64_simd_type_info):
9051 (aarch64_simd_types): Likewise.
9052 (aarch64_simd_intOI_type_node): Likewise.
9053 (aarch64_simd_intCI_type_node): Likewise.
9054 (aarch64_simd_intXI_type_node): Likewise.
9055 * config/aarch64/aarch64.h (aarch64_fp16_type_node): Likewise.
9056 (aarch64_fp16_ptr_type_node): Likewise.
9057 (aarch64_bf16_type_node): Likewise.
9058 (aarch64_bf16_ptr_type_node): Likewise.
9060 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
9062 * range-op.cc (operator_minus::op1_op2_relation_effect): Abstract
9064 (minus_op1_op2_relation_effect): ...here.
9065 (class operator_pointer_diff): New.
9066 (operator_pointer_diff::op1_op2_relation_effect): Call
9067 minus_op1_op2_relation_effect.
9068 (integral_table::integral_table): Add entry for POINTER_DIFF_EXPR.
9070 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
9072 * tree-ssa-threadbackward.c (back_threader::thread_through_all_blocks):
9073 Add may_peel_loop_headers.
9074 (back_threader_registry::thread_through_all_blocks): Same.
9075 (try_thread_blocks): Pass may_peel_loop_headers argument.
9076 (pass_early_thread_jumps::execute): Same.
9078 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
9080 * tree-ssa-threadedge.c (has_phis_p): New.
9081 (forwarder_block_p): New.
9082 (potentially_threadable_block): Call forwarder_block_p.
9083 (jump_threader::thread_around_empty_blocks): Call has_phis_p.
9084 (jump_threader::thread_through_normal_block): Call
9087 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
9089 * tree-ssa-threadbackward.c (back_threader::dump): New.
9090 (back_threader::debug): New.
9091 (back_threader_profitability::profitable_path_p): Dump blocks
9092 even if we are bailing early.
9094 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
9096 * tree-ssa-threadupdate.c (cancel_thread): New.
9097 (jump_thread_path_registry::thread_block_1): Use cancel_thread.
9098 (jump_thread_path_registry::mark_threaded_blocks): Same.
9099 (jump_thread_path_registry::register_jump_thread): Same.
9101 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
9103 * tree-ssa-threadedge.c (jt_state::push): Only call methods for
9104 which objects are available.
9105 (jt_state::pop): Same.
9106 (jt_state::register_equiv): Same.
9107 (jt_state::register_equivs_on_edge): Same.
9109 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
9111 * tree-ssa-threadedge.c (jump_threader::thread_across_edge):
9112 Move pop until after a thread is registered.
9114 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
9116 * tree-ssa-threadupdate.c (debug): New.
9118 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
9120 * gimple-range-trace.cc (push_dump_file::push_dump_file): New.
9121 (push_dump_file::~push_dump_file): New.
9122 (dump_ranger): Change dump_file temporarily while dumping
9124 * gimple-range-trace.h (class push_dump_file): New.
9126 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
9128 * gimple-range-trace.cc (debug_seed_ranger): Remove static.
9129 (dump_ranger): Dump function name.
9131 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
9133 * gimple-range-path.cc (path_range_query::range_defined_in_block):
9134 Adjust for non-null.
9135 (path_range_query::adjust_for_non_null_uses): New.
9136 (path_range_query::precompute_ranges): Call
9137 adjust_for_non_null_uses.
9138 * gimple-range-path.h: Add m_non_null and
9139 adjust_for_non_null_uses.
9141 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
9143 * gimple-range-path.cc (path_range_query::dump): Dump path
9145 (path_range_query::precompute_ranges): Dump entire path.
9147 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
9149 * value-relation.cc (relation_oracle::debug): New.
9150 * value-relation.h (relation_oracle::debug): New.
9152 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
9154 * tree-ssa-loop-ch.c: Remove unnecessary include file.
9156 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
9158 * gimple-range-fold.cc (fold_using_range::postfold_gcond_edges):
9159 Skip statements with no defining BB.
9160 * gimple-range-path.cc (path_range_query::range_defined_in_block):
9161 Do not get confused by statements with no defining BB.
9163 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
9165 * gimple-range-fold.cc (adjust_imagpart_expr): Move from
9166 gimple_range_adjustment. Add support for constants.
9167 (adjust_realpart_expr): New.
9168 (gimple_range_adjustment): Move IMAGPART_EXPR code to
9169 adjust_imagpart_expr.
9170 * range-op.cc (integral_table::integral_table): Add entry for
9173 2021-09-03 Jakub Jelinek <jakub@redhat.com>
9175 * omp-expand.c (expand_omp_atomic_pipeline): Use
9176 IFN_ATOMIC_COMPARE_EXCHANGE instead of
9177 BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_? so that memory order
9180 2021-09-03 Jakub Jelinek <jakub@redhat.com>
9183 * tree.h (DECL_FIELD_ABI_IGNORED): Changed into rvalue only macro
9184 that is false if DECL_BIT_FIELD.
9185 (SET_DECL_FIELD_ABI_IGNORED, DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD,
9186 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD): Define.
9187 * tree-streamer-out.c (pack_ts_decl_common_value_fields): For
9188 DECL_BIT_FIELD stream DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead
9189 of DECL_FIELD_ABI_IGNORED.
9190 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
9191 SET_DECL_FIELD_ABI_IGNORED instead of writing to
9192 DECL_FIELD_ABI_IGNORED and for DECL_BIT_FIELD use
9193 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead.
9194 * lto-streamer-out.c (hash_tree): For DECL_BIT_FIELD hash
9195 DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead of DECL_FIELD_ABI_IGNORED.
9197 2021-09-03 liuhongt <hongtao.liu@intel.com>
9200 * config/i386/amxbf16intrin.h : Remove macro check for __AMX_BF16__.
9201 * config/i386/amxint8intrin.h : Remove macro check for __AMX_INT8__.
9202 * config/i386/amxtileintrin.h : Remove macro check for __AMX_TILE__.
9204 2021-09-02 Martin Sebor <msebor@redhat.com>
9206 PR tree-optimization/17506
9208 * tree-ssa-uninit.c (warn_uninit): Remove conditional guarding note.
9210 2021-09-02 Richard Biener <rguenther@suse.de>
9212 * tree-ssa-loop-im.c (fill_always_executed_in_1): Refine
9213 fix for PR78185 and continue processing when leaving
9216 2021-09-02 Jakub Jelinek <jakub@redhat.com>
9218 PR tree-optimization/99591
9219 * match.pd: Demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
9222 2021-09-02 Richard Biener <rguenther@suse.de>
9225 2021-09-02 Richard Biener <rguenther@suse.de>
9227 PR tree-optimization/102155
9228 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
9229 over a part of the RPO array and do not recurse here.
9230 Dump blocks marked as always executed.
9231 (fill_always_executed_in): Walk over the RPO array and
9232 process loops whose header we run into.
9233 (loop_invariant_motion_in_fun): Compute the first RPO
9234 using rev_post_order_and_mark_dfs_back_seme in iteration
9235 order and pass that to fill_always_executed_in.
9237 2021-09-02 liuhongt <hongtao.liu@intel.com>
9239 * config/i386/i386-modes.def (FLOAT_MODE): Define ieee HFmode.
9240 * config/i386/i386.c (enum x86_64_reg_class): Add
9242 (merge_classes): Handle X86_64_SSEHF_CLASS.
9243 (examine_argument): Ditto.
9244 (construct_container): Ditto.
9245 (classify_argument): Ditto, and set HFmode/HCmode to
9247 (function_value_32): Return _FLoat16/Complex Float16 by
9249 (function_value_64): Return _Float16/Complex Float16 by SSE
9251 (ix86_print_operand): Handle CONST_DOUBLE HFmode.
9252 (ix86_secondary_reload): Require gpr as intermediate register
9253 to store _Float16 from sse register when sse4 is not
9255 (ix86_libgcc_floating_mode_supported_p): Enable _FLoat16 under
9257 (ix86_scalar_mode_supported_p): Ditto.
9258 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Defined.
9259 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add HFmode.
9260 (VALID_INT_MODE_P): Add HFmode and HCmode.
9261 * config/i386/i386.md (*pushhf_rex64): New define_insn.
9263 (*movhf_internal): Ditto.
9264 * doc/extend.texi (Half-Precision Floating Point): Documemt
9267 2021-09-02 Richard Biener <rguenther@suse.de>
9269 PR tree-optimization/102155
9270 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
9271 over a part of the RPO array and do not recurse here.
9272 Dump blocks marked as always executed.
9273 (fill_always_executed_in): Walk over the RPO array and
9274 process loops whose header we run into.
9275 (loop_invariant_motion_in_fun): Compute the first RPO
9276 using rev_post_order_and_mark_dfs_back_seme in iteration
9277 order and pass that to fill_always_executed_in.
9279 2021-09-02 YunQiang Su <syq@debian.org>
9282 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
9284 * config/mips/mips.c (mips_module_isa_name): New.
9285 mips_file_start: add .module mipsREV to all asm output
9287 2021-09-01 Jeff Law <jlaw@localhost.localdomain>
9289 PR tree-optimization/102152
9290 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Reduce a vector
9291 comparison to a scalar comparison before calling
9292 update_stmt_if_modified.
9294 2021-09-01 Andrew Pinski <apinski@marvell.com>
9297 * config/aarch64/aarch64.c (aarch64_expand_setmem):
9298 Check STRICT_ALIGNMENT before creating an overlapping
9301 2021-09-01 Martin Sebor <msebor@redhat.com>
9303 * gimple-ssa-warn-access.cc (get_size_range): Add argument.
9304 (check_access): Pass additional argument.
9305 (check_memop_access): Remove template and make a member function.
9306 (maybe_check_dealloc_call): Make a pass_waccess member function.
9307 (class pass_waccess): Add, rename, and remove members.
9308 (pass_waccess::pass_waccess): Adjust to name change.
9309 (pass_waccess::~pass_waccess): Same.
9310 (check_alloca): Make a member function.
9311 (check_alloc_size_call): Same.
9312 (check_strcat): Same.
9313 (check_strncat): Same.
9314 (check_stxcpy): Same.
9315 (check_stxncpy): Same.
9316 (check_strncmp): Same.
9317 (maybe_warn_rdwr_sizes): Rename...
9318 (pass_waccess::maybe_check_access_sizes): ...to this.
9319 (pass_waccess::check_call): Adjust to name changes.
9320 (pass_waccess::maybe_check_dealloc_call): Make a pass_waccess member
9322 (pass_waccess::execute): Adjust to name changes.
9323 * gimple-ssa-warn-access.h (check_memop_access): Remove.
9324 * pointer-query.cc (access_ref::phi): Handle null pointer.
9325 (access_ref::inform_access): Same.
9326 (pointer_query::put_ref): Modify a cached value, not a copy of it.
9327 (pointer_query::dump): New function.
9328 (compute_objsize_r): Avoid overwriting access_ref::bndrng. Cache
9330 * pointer-query.h (pointer_query::dump): Declare.
9331 * tree-ssa-strlen.c (get_range): Simplify. Use function query.
9332 (dump_strlen_info): Use function query.
9333 (printf_strlen_execute): Factor code out into pointer_query::put_ref.
9335 2021-09-01 Thomas Schwinge <thomas@codesourcery.com>
9337 * tree.c (walk_tree_1) <OMP_CLAUSE>: Simplify.
9339 2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
9341 * doc/extend.texi: Document unavailable attribute.
9342 * print-tree.c (print_node): Handle unavailable attribute.
9343 * tree-core.h (struct tree_base): Add a bit to carry unavailability.
9344 * tree.c (error_unavailable_use): New.
9345 * tree.h (TREE_UNAVAILABLE): New.
9346 (error_unavailable_use): New.
9348 2021-09-01 Jakub Jelinek <jakub@redhat.com>
9350 PR tree-optimization/102124
9351 * tree-vect-patterns.c (vect_recog_widen_op_pattern): For ORIG_CODE
9352 MINUS_EXPR, if itype is unsigned with smaller precision than type,
9353 add an extra cast to signed variant of itype to ensure sign-extension.
9355 2021-09-01 Martin Liska <mliska@suse.cz>
9357 * graph.c (draw_cfg_node_succ_edges): Do not color fallthru
9358 edges and rather use colors for TRUE and FALSE edges.
9360 2021-09-01 Richard Biener <rguenther@suse.de>
9362 PR tree-optimization/93491
9363 * tree-ssa-pre.c (compute_avail): Set BB_MAY_NOTRETURN
9364 after processing the stmt itself. Do not consider
9365 pure functions possibly not returning. Properly avoid
9366 adding possibly trapping calls to EXP_GEN when there's
9367 a preceeding possibly not returning call.
9368 * tree-ssa-sccvn.c (vn_reference_may_trap): Conservatively
9371 2021-09-01 Richard Biener <rguenther@suse.de>
9373 PR tree-optimization/102139
9374 * tree-vectorizer.h (vec_base_alignments): Adjust hash-map
9375 type to record a std::pair of the stmt-info and the innermost
9377 (dr_vec_info::group): New member.
9378 * tree-vect-data-refs.c (vect_record_base_alignment): Adjust.
9379 (vect_compute_data_ref_alignment): Verify the recorded
9380 base alignment can be used.
9381 (data_ref_pair): Remove.
9382 (dr_group_sort_cmp): Adjust.
9383 (vect_analyze_data_ref_accesses): Store the group-ID in the
9384 dr_vec_info and operate on a vector of dr_vec_infos.
9386 2021-09-01 YunQiang Su <yunqiang.su@cipunited.com>
9388 * read-md.c (md_reader::handle_enum): support value assignation.
9389 * doc/md.texi: record define_c_enum value assignation support.
9391 2021-09-01 Jakub Jelinek <jakub@redhat.com>
9393 PR tree-optimization/102141
9394 * gimple-ssa-store-merging.c (bswap_view_convert): Add BEFORE
9395 argument. If false, emit stmts after gsi instead of before, and
9397 (bswap_replace): Adjust callers. When converting output of bswap,
9398 emit VIEW_CONVERT prepratation stmts after a copy of gsi instead
9401 2021-09-01 liuhongt <hongtao.liu@intel.com>
9403 * emit-rtl.c (validate_subreg): Get rid of all float-int
9406 2021-09-01 liuhongt <hongtao.liu@intel.com>
9409 2021-08-30 liuhongt <hongtao.liu@intel.com>
9411 * expmed.c (extract_bit_field_1): Make sure we're playing with
9412 integral modes before call extract_integral_bit_field.
9413 (extract_integral_bit_field): Add a parameter of type
9414 scalar_int_mode which corresponds to of tmode.
9415 And call extract_and_convert_fixed_bit_field instead of
9416 extract_fixed_bit_field and convert_extracted_bit_field.
9417 (extract_and_convert_fixed_bit_field): New function, it's a
9418 combination of extract_fixed_bit_field and
9419 convert_extracted_bit_field.
9421 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
9423 * tree.c (walk_tree_1) <OMP_CLAUSE_TILE>: Handle three operands.
9425 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
9427 * omp-general.h (omp_is_reference): Rename to...
9428 (omp_privatize_by_reference): ... this. Adjust all users...
9429 * omp-general.c: ... here, ...
9430 * gimplify.c: ... here, ...
9431 * omp-expand.c: ... here, ...
9432 * omp-low.c: ... here.
9434 2021-08-31 Martin Sebor <msebor@redhat.com>
9436 * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Test
9437 pointer element for equality to zero, not that of the cotaining
9440 2021-08-31 Martin Sebor <msebor@redhat.com>
9442 * gcc-rich-location.h (gcc_rich_location): Make ctor explicit.
9444 2021-08-31 Martin Sebor <msebor@redhat.com>
9446 * function.h (function): Add comments.
9447 (get_range_query): Same. Add attribute returns nonnull.
9449 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
9451 * expr.c (convert_modes): Don't use subreg_promoted_mode on a
9452 SUBREG if it can't be guaranteed to a SUBREG_PROMOTED_VAR_P set.
9453 Instead use the standard (safer) is_a <scalar_int_mode> idiom.
9455 2021-08-31 Jeff Law <jlaw@localhost.localdomain>
9457 * config.gcc (cris-*-elf, cris-*-none): Remove dbxelf.h from
9459 (m32r-*-elf, m32rle-*-elf, m32r-*-linux): Likewise.
9460 (mn10300-*-*, am33_2.0-*-linux*): Likewise.
9461 (xtensa*-*-elf, xtensa*-*-linux, xtensa*-*-uclinux): Likewise.
9462 (m32c-*-elf*, m32c-*-rtems*): Likewise.
9463 * config/cris/cris.h (DBX_NO_XREFS): Remove.
9464 (DBX_CONTIN_LENGTH, DBX_CONTIN_CHAR): Likewise.
9465 * config/m32r/m32r.h (DBXOUT_SOURCE_LINE): Likewise.
9466 (DBX_DEBUGGING_INFO, DBX_CONTIN_LENGTH): Likewise.
9467 * config/mn10300/mn10300.h (DEFAULT_GDB_EXTENSIONS): Likewise.
9468 * config/mn10300/linux.h (DBX_REGISTER_NAMES): Likewise.
9470 2021-08-31 Marcel Vollweiler <marcel@codesourcery.com>
9472 * gimplify.c (gimplify_scan_omp_clauses): Error handling. 'ancestor' only
9473 allowed on target constructs and only with particular other clauses.
9474 * omp-expand.c (expand_omp_target): Output of 'sorry, not supported' if
9476 * omp-low.c (check_omp_nesting_restrictions): Error handling. No nested OpenMP
9477 structs when 'ancestor' is used.
9478 (scan_omp_1_stmt): No usage of OpenMP runtime routines in a target region when
9480 * tree-pretty-print.c (dump_omp_clause): Append 'ancestor'.
9481 * tree.h (OMP_CLAUSE_DEVICE_ANCESTOR): Define macro.
9483 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
9485 * expr.c (convert_modes): Preserve SUBREG_PROMOTED_VAR_P when
9486 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
9488 * simplify-rtx.c (simplify_unary_operation_1) [SIGN_EXTEND]:
9489 Likewise, preserve SUBREG_PROMOTED_VAR_P when creating a (wider)
9490 partial subreg from a SUBREG_PROMOTED_VAR_P subreg. Generate
9491 SIGN_EXTEND of the SUBREG_REG when a subreg would be paradoxical.
9492 [ZERO_EXTEND]: Likewise, preserve SUBREG_PROMOTED_VAR_P when
9493 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
9494 subreg. Generate ZERO_EXTEND of the SUBREG_REG when a subreg
9495 would be paradoxical.
9497 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
9499 * combine.c (combine_simplify_rtx): Avoid converting an explicit
9500 TRUNCATE into a lowpart SUBREG on !TRULY_NOOP_TRUNCATION targets.
9501 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
9503 2021-08-31 Richard Biener <rguenther@suse.de>
9505 PR tree-optimization/102142
9506 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Fix
9507 condition under which to unset the visited flag.
9509 2021-08-31 Richard Biener <rguenther@suse.de>
9511 PR middle-end/102129
9512 * tree-ssa-ter.c (find_replaceable_in_bb): Do not move
9513 possibly trapping expressions across calls.
9515 2021-08-31 Jakub Jelinek <jakub@redhat.com>
9517 PR tree-optimization/102134
9518 * tree-ssa-ccp.c (bit_value_binop) <case RSHIFT_EXPR>: If sgn is
9519 UNSIGNED and r1val | r1mask has MSB set, ensure lzcount doesn't
9522 2021-08-31 Andrew Pinski <apinski@marvell.com>
9525 * collect-utils.c (setup_signals): New declaration.
9526 * collect-utils.h (setup_signals): New function.
9527 * collect2.c (handler): Delete.
9528 (main): Instead of manually setting up the signals,
9529 just call setup_signals.
9530 * lto-wrapper.c (main): Likewise.
9532 2021-08-31 Andrew Pinski <apinski@marvell.com>
9535 * config/i386/i386-protos.h (x86_output_aligned_bss):
9536 Change align argument to unsigned type.
9537 (x86_elf_aligned_decl_common): Likewise.
9538 * config/i386/i386.c (x86_elf_aligned_decl_common): Likewise.
9539 (x86_output_aligned_bss): Likewise.
9541 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
9543 * config/mips/mips.c (mips_module_isa_name): New.
9544 mips_file_start: add .module mipsREV to all asm output
9546 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
9548 * config/mips/mips.h (struct mips_cpu_info): define enum mips_isa;
9549 use enum instead of int for 'isa' member.
9550 * config.gcc, config/mips/mips.c, config/mips/mips-cpus.def,
9551 config/mips/netbsd.h: replace hardcoded numbers with enum.
9553 2021-08-31 liuhongt <hongtao.liu@intel.com>
9555 * config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
9556 define_split to define_insn_and_split.
9557 (*avx2_eq<mode>3): Removed.
9558 (<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
9559 (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
9560 (*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
9562 (*avx2_gt<mode>3): Removed.
9563 (<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
9564 define_insn to define_expand, and adjust pattern.
9565 (UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.
9567 2021-08-30 David Malcolm <dmalcolm@redhat.com>
9570 * Makefile.in (ANALYZER_OBJS): Add analyzer/call-info.o.
9572 2021-08-30 Jason Merrill <jason@redhat.com>
9574 * doc/invoke.texi: Document -Wmissing-requires.
9576 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
9578 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Remove
9579 TARGET_EXTRA_BUILTINS guard.
9581 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
9583 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Change
9584 initialization of V2DI_type_node and unsigned_V2DI_type_node.
9586 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
9588 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Use the new
9589 decl when new_builtins_are_live.
9590 * config/rs6000/rs6000-builtin-new.def (__builtin_cfstring): New
9593 2021-08-30 Pat Haugen <pthaugen@linux.ibm.com>
9595 * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add
9596 OPTION_MASK_P10_FUSION_2STORE.
9597 (POWERPC_MASKS): Likewise.
9598 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
9599 store fusion for Power10.
9600 (is_fusable_store): New.
9601 (power10_sched_reorder): Likewise.
9602 (rs6000_sched_reorder): Do Power10 specific reordering.
9603 (rs6000_sched_reorder2): Likewise.
9604 * config/rs6000/rs6000.opt: Add new option.
9606 2021-08-30 Richard Biener <rguenther@suse.de>
9608 PR tree-optimization/102128
9609 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
9610 Move scanning for if-converted scalar code to the caller
9611 and instead delay clearing the visited flag for profitable
9613 (vect_slp_region): Cost all subgraphs before scheduling.
9614 For if-converted BB vectorization scan for scalar COND_EXPRs
9615 and do not vectorize if any found and the cost model is
9618 2021-08-30 Richard Biener <rguenther@suse.de>
9620 * common.opt (fexceptions): Mark
9621 EnabledBy(fnon-call-exceptions).
9622 * doc/invoke.texi (fnon-call-exceptions): Document this
9623 enables -fexceptions.
9625 2021-08-30 Sebastian Huber <sebastian.huber@embedded-brains.de>
9627 * tsystem.h (abort): Define abort() if inhibit_libc is defined and it
9628 is not already defined.
9630 2021-08-30 liuhongt <hongtao.liu@intel.com>
9632 * expmed.c (extract_bit_field_1): Make sure we're playing with
9633 integral modes before call extract_integral_bit_field.
9634 (extract_integral_bit_field): Add a parameter of type
9635 scalar_int_mode which corresponds to of tmode.
9636 And call extract_and_convert_fixed_bit_field instead of
9637 extract_fixed_bit_field and convert_extracted_bit_field.
9638 (extract_and_convert_fixed_bit_field): New function, it's a
9639 combination of extract_fixed_bit_field and
9640 convert_extracted_bit_field.
9642 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
9644 * config/darwin.c (darwin_libc_has_function): Do not run
9645 the checks for x86 or modern Darwin. Make sure that there
9646 is a value set for darwin_macosx_version_min before testing.
9648 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
9650 * config/i386/darwin.h (CLEAR_INSN_CACHE): New.
9652 2021-08-28 Jan Hubicka <hubicka@ucw.cz>
9654 * ipa-modref-tree.h (modref_access_node::merge): Break out
9655 logic combining offsets and logic merging ranges to ...
9656 (modref_access_node::combined_offsets): ... here
9657 (modref_access_node::update2): ... here
9658 (modref_access_node::closer_pair_p): New member function.
9659 (modref_access_node::forced_merge): New member function.
9660 (modre_ref_node::insert): Do merging when table is full.
9662 2021-08-28 YunQiang Su <yunqiang.su@cipunited.com>
9665 * config.gcc: MIPS: use N64 ABI by default if the triple end
9666 with -gnuabi64, which is used by Debian since 2013.
9668 2021-08-28 Alexandre Oliva <oliva@adacore.com>
9670 * ipa-modref.c (analyze_function): Skip debug stmts.
9671 * tree-inline.c (estimate_num_insn): Consider builtins even
9672 without a cgraph_node.
9674 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
9676 * config/h8300/bitfield.md (cstore<mode>4): Remove expander.
9677 * config/h8300/h8300.c (h8300_expand_branch): Remove function.
9678 * config/h8300/h8300-protos.h (h8300_expadn_branch): Remove prototype.
9679 * config/h8300/h8300.md (eqne): New code iterator.
9680 (geultu, geultu_to_c): Similarly.
9681 * config/h8300/testcompare.md (cstore<mode>4): Dummy expander.
9682 (store_c_<mode>, store_c_i_<mode>): New define_insn_and_splits
9683 (cmp<mode>_c): New pattern
9685 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
9687 * tree-ssa-dom.c (reduce_vector_comparison_to_scalar_comparison): New
9689 (dom_opt_dom_walker::optimize_stmt): Use it.
9691 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
9693 * config/darwin.c (finalize_ctors): Add a section-start linker-
9695 (finalize_dtors): Likewise.
9696 * config/darwin.h (MIN_LD64_INIT_TERM_START_LABELS): New.
9698 2021-08-27 Bill Schmidt <wschmidt@linux.ibm.com>
9700 * config/rs6000/rs6000-call.c (rs6000-builtins.h): New #include.
9701 (rs6000_init_builtins): Call rs6000_init_generated_builtins. Skip the
9702 old initialization logic when new builtins are enabled.
9703 * config/rs6000/rs6000-gen-builtins.c (write_decls): Rename
9704 rs6000_autoinit_builtins to rs6000_init_generated_builtins.
9705 (write_init_file): Likewise.
9707 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
9709 * configure.ac (darwin2[[0-9]]* | darwin19*): Alter use of
9710 gcc_GAS_CHECK_FEATURE to remove an extraneous parameter.
9711 (amdgcn-* | gcn-*) Likewise.
9713 2021-08-27 Anthony Sharp <anthonysharp15@gmail.com>
9715 * symbol-summary.h: Added missing template keyword.
9717 2021-08-27 Richard Biener <rguenther@suse.de>
9719 PR tree-optimization/45178
9720 * tree-ssa-dce.c (find_obviously_necessary_stmts): For
9721 infinite loops without exit do not mark control dependent
9722 edges of the latch necessary.
9724 2021-08-27 konglin1 <lingling.kong@intel.com>
9727 * config/i386/sse.md: (<avx512>scattersi<mode>): Add mask operand to
9729 (<avx512>scattersi<mode>): Likewise.
9730 (*avx512f_scattersi<VI48F:mode>): Merge mask operand to set_dest.
9731 (*avx512f_scatterdi<VI48F:mode>): Likewise
9733 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
9735 * config/rs6000/rs6000.c (rs6000_builtin_md_vectorized_function): Add
9736 support for built-in functions MISC_BUILTIN_DIVWE, MISC_BUILTIN_DIVWEU,
9737 MISC_BUILTIN_DIVDE, MISC_BUILTIN_DIVDEU, P10_BUILTIN_CFUGED,
9738 P10_BUILTIN_CNTLZDM, P10_BUILTIN_CNTTZDM, P10_BUILTIN_PDEPD and
9739 P10_BUILTIN_PEXTD on Power10.
9741 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
9743 * config/rs6000/rs6000-call.c (builtin_function_type): Add unsigned
9744 signedness for some Power10 bifs.
9746 2021-08-27 David Edelsohn <dje.gcc@gmail.com>
9749 * config/rs6000/rs6000.c (rs6000_adjust_field_align): Use
9750 computed alignment if the entire struct has attribute packed.
9752 2021-08-27 liuhongt <hongtao.liu@intel.com>
9756 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
9757 IX86_BUILTIN_SHUFPD512, IX86_BUILTIN_SHUFPS512,
9758 IX86_BUILTIN_SHUFPD256, IX86_BUILTIN_SHUFPS,
9759 IX86_BUILTIN_SHUFPS256.
9760 (ix86_masked_all_ones): New function.
9762 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
9764 * config/i386/i386.md (*btr<mode>_1): Call force_reg unconditionally.
9765 (conditional moves with memory inputs splitters): Ditto.
9766 * config/i386/sse.md (one_cmpl<mode>2): Simplify.
9768 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
9770 * ipa-modref-tree.h (modref_access_node::try_merge_with): Restart
9771 search after merging.
9773 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
9775 * config/rs6000/rs6000-overload.def: Add remaining overloads.
9777 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
9779 * config/rs6000/rs6000-builtin-new.def: Add cell stanza.
9781 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
9783 * config/rs6000/rs6000-builtin-new.def: Add ieee128-hw, dfp,
9784 crypto, and htm stanzas.
9786 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
9788 * config/rs6000/rs6000-builtin-new.def: Add mma stanza.
9790 2021-08-26 Martin Sebor <msebor@redhat.com>
9792 * tree-ssa-uninit.c (warn_uninit): Refactor and simplify.
9793 (warn_uninit_phi_uses): Remove argument from calls to warn_uninit.
9794 (warn_uninitialized_vars): Same. Reduce visibility of locals.
9795 (warn_uninitialized_phi): Same.
9797 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
9799 * tree-ssa-ccp.c (get_individual_bits): Helper function to
9800 extract the individual bits from a widest_int constant (mask).
9801 (gray_code_bit_flips): New read-only table for effiently
9802 enumerating permutations/combinations of bits.
9803 (bit_value_binop) [LROTATE_EXPR, RROTATE_EXPR]: Handle rotates
9804 by unknown counts that are guaranteed less than the target
9805 precision and four or fewer unknown bits by enumeration.
9806 [LSHIFT_EXPR, RSHIFT_EXPR]: Likewise, also handle shifts by
9807 enumeration under the same conditions. Handle remaining
9808 shifts as a mask based upon the minimum possible shift value.
9810 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
9811 Richard Biener <rguenther@suse.de>
9813 * match.pd (shift transformations): Remove a redundant
9814 !POINTER_TYPE_P check.
9816 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
9819 * config/i386/i386.md (cmove reg-reg move elimination peephole2s):
9820 Set all_regs to true in the call to replace_rtx.
9822 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
9824 * ipa-modref-tree.c (test_insert_search_collapse): Update test.
9825 * ipa-modref-tree.h (modref_base_node::insert): Be smarter when
9826 hiting --param modref-max-refs limit.
9827 (modref_tree:insert_base): Be smarter when hitting
9828 --param modref-max-bases limit. Add new parameter REF.
9829 (modref_tree:insert): Update.
9830 (modref_tree:merge): Update.
9831 * ipa-modref.c (read_modref_records): Update.
9833 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
9835 * params.opt: (modref-max-adjustments): Add full stop.
9837 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
9839 * ipa-modref-tree.h (modref_ref_node::verify): New member
9841 (modref_ref_node::insert): Use it.
9842 (modref_ref_node::try_mere_with): Fix off by one error.
9844 2021-08-26 Martin Liska <mliska@suse.cz>
9845 Stefan Kneifel <stefan.kneifel@bluewin.ch>
9847 * cgraph.h (create_version_clone_with_body): Add new parameter.
9848 * cgraphclones.c: Likewise.
9849 * multiple_target.c (create_dispatcher_calls): Do not use
9851 (create_target_clone): Likewise here.
9853 2021-08-26 Jonathan Yong <10walls@gmail.com>
9855 * doc/extend.texi: Add note about reserved priorities
9856 to the constructor attribute.
9858 2021-08-25 Martin Sebor <msebor@redhat.com>
9860 * gimple-range-cache.cc (ssa_global_cache::dump): Avoid printing
9861 range table header alone.
9862 * gimple-range.cc (gimple_ranger::export_global_ranges): Same.
9864 2021-08-25 Jan Hubicka <hubicka@ucw.cz>
9866 * doc/invoke.texi: Document --param modref-max-adjustments.
9867 * ipa-modref-tree.c (test_insert_search_collapse): Update.
9868 (test_merge): Update.
9869 * ipa-modref-tree.h (struct modref_access_node): Add adjustments;
9870 (modref_access_node::operator==): Fix handling of access ranges.
9871 (modref_access_node::contains): Constify parameter; handle also
9872 mismatched parm offsets.
9873 (modref_access_node::update): New function.
9874 (modref_access_node::merge): New function.
9875 (unspecified_modref_access_node): Update constructor.
9876 (modref_ref_node::insert_access): Add record_adjustments parameter;
9878 (modref_ref_node::try_merge_with): New private function.
9879 (modref_tree::insert): New record_adjustments parameter.
9880 (modref_tree::merge): New record_adjustments parameter.
9881 (modref_tree::copy_from): Update.
9882 * ipa-modref.c (dump_access): Dump adjustments field.
9883 (get_access): Update constructor.
9884 (record_access): Update call of insert.
9885 (record_access_lto): Update call of insert.
9886 (merge_call_side_effects): Add record_adjustments parameter.
9887 (get_access_for_fnspec): Update.
9888 (process_fnspec): Update.
9889 (analyze_call): Update.
9890 (analyze_function): Update.
9891 (read_modref_records): Update.
9892 (ipa_merge_modref_summary_after_inlining): Update.
9893 (propagate_unknown_call): Update.
9894 (modref_propagate_in_scc): Update.
9895 * params.opt (param-max-modref-adjustments=): New.
9897 2021-08-25 Michael Meissner <meissner@linux.ibm.com>
9899 * config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
9901 (xxspltiw_v4si): Use vecperm type attribute.
9902 (xxspltiw_v4si_inst): Use vecperm type attribute.
9903 (xxspltiw_v4sf_inst): Likewise.
9904 (xxspltidp_v2df): Use vecperm type attribute. Use
9905 UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
9906 (xxspltidp_v2df_inst): Likewise.
9907 (xxsplti32dx_v4si): Use vecperm type attribute.
9908 (xxsplti32dx_v4si_inst): Likewise.
9909 (xxsplti32dx_v4sf_inst): Likewise.
9910 (xxblend_<mode>): Likewise.
9911 (xxpermx): Likewise.
9912 (xxpermx_inst): Likewise.
9915 2021-08-25 Lewis Hyatt <lhyatt@gmail.com>
9918 * coretypes.h (typedef diagnostic_input_charset_callback): Declare.
9919 * diagnostic.c (diagnostic_initialize_input_context): New function.
9920 * diagnostic.h (diagnostic_initialize_input_context): Declare.
9921 * input.c (default_charset_callback): New function.
9922 (file_cache::initialize_input_context): New function.
9923 (file_cache_slot::create): Added ability to convert the input
9924 according to the input context.
9925 (file_cache::file_cache): Initialize the new input context.
9926 (class file_cache_slot): Added new m_alloc_offset member.
9927 (file_cache_slot::file_cache_slot): Initialize the new member.
9928 (file_cache_slot::~file_cache_slot): Handle potentially offset buffer.
9929 (file_cache_slot::maybe_grow): Likewise.
9930 (file_cache_slot::needs_read_p): Handle NULL fp, which is now possible.
9931 (file_cache_slot::get_next_line): Likewise.
9932 * input.h (class file_cache): Added input context member.
9934 2021-08-25 Richard Biener <rguenther@suse.de>
9936 PR tree-optimization/102046
9937 * tree-vect-slp.c (vect_build_slp_tree_2): Conservatively
9938 update ->any_pattern when swapping operands.
9940 2021-08-25 Hongyu Wang <hongyu.wang@intel.com>
9943 * config/i386/i386.c (ix86_live_on_entry): Adjust comment.
9944 (ix86_decompose_address): Remove retval check for ASHIFT,
9945 allow non-canonical zero extend if AND mask covers ASHIFT
9947 (ix86_legitimate_address_p): Adjust condition for decompose.
9948 (ix86_rtx_costs): Adjust cost for lea with non-canonical
9950 Co-Authored by: Uros Bizjak <ubizjak@gmail.com>
9952 2021-08-25 Jiufu Guo <guojiufu@linux.ibm.com>
9954 PR tree-optimization/101145
9955 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
9957 (number_of_iterations_lt): Invoke above function.
9958 (adjust_cond_for_loop_until_wrap):
9959 Merge to number_of_iterations_until_wrap.
9960 (number_of_iterations_cond): Update invokes for
9961 adjust_cond_for_loop_until_wrap and number_of_iterations_lt.
9963 2021-08-25 konglin1 <lingling.kong@intel.com>
9966 * config/i386/avx512dqintrin.h (_mm512_fpclass_ps_mask): Fix
9968 (_mm512_mask_fpclass_ps_mask): Ditto.
9970 2021-08-25 Kewen Lin <linkw@linux.ibm.com>
9972 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Remove.
9973 (vec_unpacku_hi_v8hi): Likewise.
9974 (vec_unpacku_lo_v16qi): Likewise.
9975 (vec_unpacku_lo_v8hi): Likewise.
9976 (vec_unpacku_hi_<VP_small_lc>): New define_expand.
9977 (vec_unpacku_lo_<VP_small_lc>): Likewise.
9979 2021-08-24 David Edelsohn <dje.gcc@gmail.com>
9981 * config/rs6000/aix.h (SYSTEM_IMPLICIT_EXTERN_C): Delete.
9982 * config/rs6000/aix71.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
9983 * config/rs6000/aix72.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
9984 * config/rs6000/aix73.h (TARGET_AIX_VERSION): Increase to 73.
9986 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
9988 PR middle-end/102031
9989 * simplify-rtx.c (simplify_truncation): When comparing precisions
9990 use "subreg_prec" variable, not "subreg_mode".
9992 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
9994 * config/rs6000/rs6000-builtin-new.def: Add power10 and power10-64
9997 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
9999 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Initialize
10000 various pointer type nodes.
10001 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add enum
10002 values for various pointer types.
10003 (ptr_V16QI_type_node): New macro.
10004 (ptr_V1TI_type_node): New macro.
10005 (ptr_V2DI_type_node): New macro.
10006 (ptr_V2DF_type_node): New macro.
10007 (ptr_V4SI_type_node): New macro.
10008 (ptr_V4SF_type_node): New macro.
10009 (ptr_V8HI_type_node): New macro.
10010 (ptr_unsigned_V16QI_type_node): New macro.
10011 (ptr_unsigned_V1TI_type_node): New macro.
10012 (ptr_unsigned_V8HI_type_node): New macro.
10013 (ptr_unsigned_V4SI_type_node): New macro.
10014 (ptr_unsigned_V2DI_type_node): New macro.
10015 (ptr_bool_V16QI_type_node): New macro.
10016 (ptr_bool_V8HI_type_node): New macro.
10017 (ptr_bool_V4SI_type_node): New macro.
10018 (ptr_bool_V2DI_type_node): New macro.
10019 (ptr_bool_V1TI_type_node): New macro.
10020 (ptr_pixel_type_node): New macro.
10021 (ptr_intQI_type_node): New macro.
10022 (ptr_uintQI_type_node): New macro.
10023 (ptr_intHI_type_node): New macro.
10024 (ptr_uintHI_type_node): New macro.
10025 (ptr_intSI_type_node): New macro.
10026 (ptr_uintSI_type_node): New macro.
10027 (ptr_intDI_type_node): New macro.
10028 (ptr_uintDI_type_node): New macro.
10029 (ptr_intTI_type_node): New macro.
10030 (ptr_uintTI_type_node): New macro.
10031 (ptr_long_integer_type_node): New macro.
10032 (ptr_long_unsigned_type_node): New macro.
10033 (ptr_float_type_node): New macro.
10034 (ptr_double_type_node): New macro.
10035 (ptr_long_double_type_node): New macro.
10036 (ptr_dfloat64_type_node): New macro.
10037 (ptr_dfloat128_type_node): New macro.
10038 (ptr_ieee128_type_node): New macro.
10039 (ptr_ibm128_type_node): New macro.
10040 (ptr_vector_pair_type_node): New macro.
10041 (ptr_vector_quad_type_node): New macro.
10042 (ptr_long_long_integer_type_node): New macro.
10043 (ptr_long_long_unsigned_type_node): New macro.
10045 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
10047 * config/rs6000/rs6000-builtin-new.def: Add power9-vector, power9,
10048 and power9-64 stanzas.
10050 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
10051 Tom de Vries <tdevries@suse.de>
10053 * config.gcc (nvptx-*-*): Define {c,c++}_target_objs.
10054 * config/nvptx/nvptx-protos.h (nvptx_cpu_cpp_builtins): Prototype.
10055 * config/nvptx/nvptx.h (TARGET_CPU_CPP_BUILTINS): Implement with
10056 a call to the new nvptx_cpu_cpp_builtins function in nvptx-c.c.
10057 * config/nvptx/t-nvptx (nvptx-c.o): New rule.
10058 * config/nvptx/nvptx-c.c: New source file.
10059 (nvptx_cpu_cpp_builtins): Move implementation here.
10061 2021-08-24 Martin Sebor <msebor@redhat.com>
10063 PR middle-end/101600
10064 PR middle-end/101977
10065 * gimple-ssa-warn-access.cc (maybe_warn_for_bound): Tighten up
10066 the phrasing of a warning.
10067 (check_access): Use the remaining size after subtracting any offset
10068 rather than the whole object size.
10069 * pointer-query.cc (access_ref::get_ref): Clear BASE0 flag if it's
10070 clear for any nonnull PHI argument.
10071 (compute_objsize): Clear argument.
10073 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
10075 * config/rs6000/rs6000-builtin-new.def: Add power8-vector stanza.
10077 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
10079 * config/rs6000/rs6000-builtin-new.def: Add power7 and power7-64
10082 2021-08-24 Andrew MacLeod <amacleod@redhat.com>
10084 * value-relation.cc (rr_transitive_table): New.
10085 (relation_transitive): New.
10086 (value_relation::swap): Remove.
10087 (value_relation::apply_transitive): New.
10088 (relation_oracle::relation_oracle): Allocate a new tmp bitmap.
10089 (relation_oracle::register_relation): Call register_transitives.
10090 (relation_oracle::register_transitives): New.
10091 * value-relation.h (relation_oracle): Add new temporary bitmap and
10094 2021-08-24 H.J. Lu <hjl.tools@gmail.com>
10097 * config/i386/i386-expand.c (ix86_expand_vector_move): Broadcast
10098 from integer to a pseudo vector register.
10100 2021-08-24 Richard Biener <rguenther@suse.de>
10102 PR tree-optimization/100089
10103 * tree-vectorizer.h (vect_slp_bb): Rename to ...
10104 (vect_slp_if_converted_bb): ... this and get the original
10105 loop as new argument.
10106 * tree-vectorizer.c (try_vectorize_loop_1): Revert previous fix,
10107 pass original loop to vect_slp_if_converted_bb.
10108 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
10109 If orig_loop was passed scan the not vectorized stmts
10110 for COND_EXPRs and force not profitable if found.
10111 (vect_slp_region): Pass down all SLP instances to costing
10112 if orig_loop was specified.
10113 (vect_slp_bbs): Pass through orig_loop.
10114 (vect_slp_bb): Rename to ...
10115 (vect_slp_if_converted_bb): ... this and get the original
10116 loop as new argument.
10117 (vect_slp_function): Adjust.
10119 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
10122 * config/arm/arm.md (attribute arch): Add fix_vlldm.
10123 (arch_enabled): Use it.
10124 * config/arm/vfp.md (lazy_store_multiple_insn): Add alternative to
10125 use when erratum mitigation is needed.
10127 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
10130 * config/arm/arm.opt (mfix-cmse-cve-2021-35465): New option.
10131 * doc/invoke.texi (Arm Options): Document it.
10132 * config/arm/arm-cpus.in (quirk_vlldm): New feature bit.
10133 (ALL_QUIRKS): Add quirk_vlldm.
10134 (cortex-m33): Add quirk_vlldm.
10135 (cortex-m35p, cortex-m55): Likewise.
10136 * config/arm/arm.c (arm_option_override): Enable fix_vlldm if
10137 targetting an affected CPU and not explicitly controlled on
10140 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
10142 * config/arm/vfp.md (lazy_store_multiple_insn): Rewrite as valid RTL.
10143 (lazy_load_multiple_insn): Likewise.
10145 2021-08-24 liuhongt <hongtao.liu@intel.com>
10148 * config/i386/sse.md (<avx512>_vternlog<mode><sd_maskz_name>):
10149 Enable avx512 embedded broadcast.
10150 (*<avx512>_vternlog<mode>_all): Ditto.
10151 (<avx512>_vternlog<mode>_mask): Ditto.
10153 2021-08-24 liuhongt <hongtao.liu@intel.com>
10156 * config/i386/i386.c (ix86_rtx_costs): Define cost for
10158 * config/i386/i386.h (STRIP_UNARY): New macro.
10159 * config/i386/predicates.md (reg_or_notreg_operand): New
10161 * config/i386/sse.md (*<avx512>_vternlog<mode>_all): New define_insn.
10162 (*<avx512>_vternlog<mode>_1): New pre_reload
10163 define_insn_and_split.
10164 (*<avx512>_vternlog<mode>_2): Ditto.
10165 (*<avx512>_vternlog<mode>_3): Ditto.
10166 (any_logic1,any_logic2): New code iterator.
10167 (logic_op): New code attribute.
10168 (ternlogsuffix): Extend to VNxDF and VNxSF.
10170 2021-08-24 Richard Biener <rguenther@suse.de>
10172 * doc/invoke.texi (vect-inner-loop-cost-factor): Adjust.
10173 * params.opt (--param vect-inner-loop-cost-factor): Adjust
10175 * tree-vect-loop.c (vect_analyze_loop_form): Initialize
10176 inner_loop_cost_factor to the minimum of the estimated number
10177 of iterations of the inner loop and vect-inner-loop-cost-factor.
10179 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
10180 Richard Biener <rguenther@suse.de>
10182 * config/i386/i386-features.c (compute_convert_gain): Provide
10183 more accurate values for CONST_INT, when optimizing for size.
10184 * config/i386/i386.c (COSTS_N_BYTES): Move definition from here...
10185 * config/i386/i386.h (COSTS_N_BYTES): to here.
10187 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
10188 Jakub Jelinek <jakub@redhat.com>
10190 PR middle-end/102029
10191 * match.pd (shift transformations): Add an additional check for
10192 !POINTER_TYPE_P in the recently added left shift transformation.
10194 2021-08-24 liuhongt <hongtao.liu@intel.com>
10196 PR tree-optimization/100089
10197 * tree-vectorizer.c (try_vectorize_loop_1): Disable slp in
10198 loop vectorizer when cost model is very-cheap.
10200 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
10202 * config/rs6000/rs6000-gen-builtins.c (parse_bif_entry): Don't call
10203 asprintf, which is not available on AIX.
10205 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
10207 * config.gcc (target_gtfiles): Add ./rs6000-builtins.h.
10208 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Set.
10210 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
10212 * config.gcc (powerpc*-*-*): Add rs6000-builtins.o to extra_objs.
10213 * config/rs6000/rs6000-gen-builtins.c (main): Close init_file
10215 * config/rs6000/t-rs6000 (rs6000-gen-builtins.o): New target.
10216 (rbtree.o): Likewise.
10217 (rs6000-gen-builtins): Likewise.
10218 (rs6000-builtins.c): Likewise.
10219 (rs6000-builtins.h): Likewise.
10220 (rs6000.o): Add dependency.
10221 (EXTRA_HEADERS): Add rs6000-vecdefines.h.
10222 (rs6000-vecdefines.h): New target.
10223 (rs6000-builtins.o): Likewise.
10224 (rs6000-call.o): Add rs6000-builtins.h as a dependency.
10225 (rs6000-c.o): Likewise.
10227 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
10230 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace):
10231 Diagnose buffer overrun.
10232 (safe_inc_pos): Fix overrun detection.
10233 (match_identifier): Diagnose buffer overrun.
10234 (match_integer): Likewise.
10235 (match_to_right_bracket): Likewise.
10237 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
10239 * ipa-modref-tree.h (modref_access_node::range_info_useful_p):
10240 Improve range compare.
10241 (modref_access_node::contains): New member function.
10242 (modref_access_node::search): Remove.
10243 (modref_access_node::insert): Be smarter about subaccesses.
10245 2021-08-23 Thomas Schwinge <thomas@codesourcery.com>
10247 * config/i386/i386-options.c (ix86_omp_device_kind_arch_isa)
10248 <omp_device_arch> [ACCEL_COMPILER]: Match "intel_mic".
10249 * config/i386/t-omp-device (omp-device-properties-i386) <arch>:
10252 2021-08-23 Jeff Law <jlaw@localhost.localdomain>
10254 * config/h8300/h8300-protos.h (h8300_expand_epilogue): Add new
10256 * config/h8300/jumpcall.md (call, call_value): Restrict to
10257 !SIBLING_CALL_P cases.
10258 (subcall, sibcall_value): New patterns & expanders.
10259 * config/h8300/proepi.md (epilogue): Pass new argument to
10260 h8300_expand_epilogue.
10261 (sibcall_epilogue): New expander.
10262 * config/h8300/h8300.c (h8300_expand_epilogue): Handle sibcall
10264 (h8300_ok_for_sibcall_p): New function.
10265 (TARGET_FUNCTION_OK_FOR_SIBCALL): define.
10267 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
10269 * simplify-rtx.c (simplify_unary_operation_1): [TRUNCATE]:
10270 Handle case where the operand is already the desired mode.
10272 2021-08-23 Richard Biener <rguenther@suse.de>
10275 * tree-ssa-structalias.c (ipa_pta_execute): Check in_other_partition
10276 in addition to has_gimple_body.
10278 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
10280 PR middle-end/101949
10281 * ipa-modref.c (analyze_ssa_name_flags): Fix merging of
10284 2021-08-23 Martin Liska <mliska@suse.cz>
10286 * doc/invoke.texi: Put the option out of -mxl-mode-app-model
10289 2021-08-23 Richard Biener <rguenther@suse.de>
10291 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
10292 Properly scale the inner loop cost only once.
10294 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
10296 * tree-ssa-ccp.c (bit_value_binop) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]:
10297 Provide bounds for unsigned (and signed with non-negative operands)
10298 division and modulus.
10300 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
10302 * simplify-rtx.c (simplify_truncation): Generalize simplification
10303 of (truncate:A (subreg:B X)).
10304 (simplify_unary_operation_1) [FLOAT_TRUNCATE, FLOAT_EXTEND,
10305 SIGN_EXTEND, ZERO_EXTEND]: Handle cases where the operand
10306 already has the desired machine mode.
10307 (test_scalar_int_ops): Add tests that useless extensions and
10308 truncations are optimized away.
10309 (test_scalar_int_ext_ops): New self-test function to confirm
10310 that truncations of extensions are correctly simplified.
10311 (test_scalar_int_ext_ops2): New self-test function to check
10312 truncations of truncations, extensions of extensions, and
10313 truncations of extensions.
10314 (test_scalar_ops): Call the above two functions with a
10315 representative sampling of integer machine modes.
10317 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
10319 * match.pd (shift transformations): Change the sign of an
10320 LSHIFT_EXPR if it reduces the number of explicit conversions.
10322 2021-08-23 Jakub Jelinek <jakub@redhat.com>
10324 PR tree-optimization/86723
10325 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
10326 cast64_to_32 argument, set *cast64_to_32 to false, unless n is
10327 non-memory permutation of 64-bit src which only has bytes of
10328 0 or [5..8] and n->range is 4.
10329 (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
10330 find_bswap_or_nop_finalize caller, support bswap with some bytes
10331 zeroed, as long as at least two bytes are not zeroed.
10332 (bswap_replace): Add mask argument and handle masking of bswap
10334 (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
10335 caller, punt if cast64_to_32 or mask is not all ones.
10336 (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
10337 caller, for now punt if cast64_to_32.
10339 2021-08-23 Richard Biener <rguenther@suse.de>
10341 PR tree-optimization/79334
10342 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Record
10343 a type also for COMPONENT_REFs.
10344 (vn_reference_may_trap): Check ARRAY_REF with constant index
10345 against the array domain.
10347 2021-08-23 liuhongt <hongtao.liu@intel.com>
10350 * config/i386/sse.md (*avx512f_pshufb_truncv8hiv8qi_1): Add
10351 TARGET_AVX512BW to condition.
10353 2021-08-23 Jakub Jelinek <jakub@redhat.com>
10356 * dwarf2out.c (gen_variable_die): Add DW_AT_location for global
10357 register variables already during early_dwarf if possible.
10359 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
10361 * config/arm/arm_mve.h: Fix __arm_vctp16q return type.
10363 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
10366 * config/arm/arm.opt: Fix typo.
10367 * config/arm/t-rmprofile: Fix typo.
10369 2021-08-23 Jakub Jelinek <jakub@redhat.com>
10371 * tree.h (OMP_CLAUSE_GRAINSIZE_STRICT): Define.
10372 (OMP_CLAUSE_NUM_TASKS_STRICT): Define.
10373 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_GRAINSIZE,
10374 case OMP_CLAUSE_NUM_TASKS>: Print strict: modifier.
10375 * omp-expand.c (expand_task_call): Use GOMP_TASK_FLAG_STRICT in iflags
10376 if either grainsize or num_tasks clause has the strict modifier.
10378 2021-08-23 Martin Liska <mliska@suse.cz>
10380 * dbgcnt.def (DEBUG_COUNTER): New counter.
10381 * gimple.c (gimple_call_arg_flags): Use it in IPA PTA.
10383 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
10385 * ipa-modref.c (analyze_ssa_name_flags): Improve handling of return slot.
10387 2021-08-23 Xi Ruoyao <xry111@mengyan1223.wang>
10390 * config/mips/mips-protos.h (mips_msa_output_shift_immediate):
10392 * config/mips/mips.c (mips_msa_output_shift_immediate): New
10394 * config/mips/mips-msa.md (vashl<mode>3, vashr<mode>3,
10395 vlshr<mode>3): Call it.
10397 2021-08-22 Jan Hubicka <hubicka@ucw.cz>
10398 Martin Liska <mliska@suse.cz>
10400 PR middle-end/101949
10401 * ipa-modref.c (analyze_ssa_name_flags): Indirect call implies
10404 2021-08-21 Dragan Mladjenovic <OT_Dragan.Mladjenovic@mediatek.com>
10406 * config/mips/mips.c (mips_function_rodata_section,
10407 TARGET_ASM_FUNCTION_RODATA_SECTION): Removed.
10409 2021-08-21 John David Anglin <danglin@gcc.gnu.org>
10411 * config/pa/pa.c (pa_asm_output_aligned_common): Remove warning.
10413 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
10415 * configure.ac (thread-local storage support): Remove tls_first_major
10416 and tls_first_minor. Use "$conftest_s" to check support.
10417 * configure: Regenerate.
10419 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
10421 * configure.ac: Fixup formatting.
10423 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
10425 * acinclude.m4 (gcc_GAS_CHECK_FEATURE): Remove third argument and ...
10426 * configure.ac: ... update all callers.
10428 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
10431 * acinclude.m4 (_gcc_COMPUTE_GAS_VERSION, _gcc_GAS_VERSION_GTE_IFELSE)
10432 (gcc_GAS_VERSION_GTE_IFELSE): Remove.
10433 (gcc_GAS_CHECK_FEATURE): Do not handle in-tree case specially.
10434 * configure.ac: Remove gcc_cv_gas_major_version, gcc_cv_gas_minor_version.
10435 Remove remaining checks for in-tree assembler.
10436 * configure: Regenerate.
10438 2021-08-20 Jeff Law <jlaw@localhost.localdomain>
10440 * config/h8300/h8300.c (shift_alg_hi): Improve arithmetic shift right
10441 by 15 bits for H8/300H and H8/S. Improve logical shifts by 12
10443 (shift_alg_si): Improve arithmetic right shift by 28-30 bits for
10444 H8/300H. Improve arithmetic shift right by 15 bits for H8/S.
10445 Improve logical shifts by 27 bits for H8/S.
10446 (get_shift_alg): Corresponding changes.
10447 (h8300_option_override): Revert to loops for -Os when profitable.
10449 2021-08-20 Richard Biener <rguenther@suse.de>
10451 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
10453 (vect_analyze_data_ref_accesses): Likewise. Assign the BB
10454 index as group_id when dataref_groups were not computed.
10455 * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
10456 we advace to the next BB.
10458 2021-08-20 Jakub Jelinek <jakub@redhat.com>
10460 * omp-builtins.def (BUILT_IN_GOMP_WARNING, BUILT_IN_GOMP_ERROR): New
10463 2021-08-20 Martin Liska <mliska@suse.cz>
10465 PR gcov-profile/89961
10466 * gcov.c (make_gcov_file_name): Rewrite using std::string.
10467 (mangle_name): Simplify, do not used the second argument.
10468 (strip_extention): New function.
10469 (get_md5sum): Likewise.
10470 (get_gcov_intermediate_filename): Handle properly -p and -x
10472 (output_gcov_file): Use string type.
10473 (generate_results): Likewise.
10474 (md5sum_to_hex): Remove.
10476 2021-08-20 Michael Meissner <meissner@linux.ibm.com>
10478 * config/rs6000/altivec.md (UNSPEC_XXEVAL): Move to vsx.md.
10479 (UNSPEC_XXSPLTIW): Move to vsx.md.
10480 (UNSPEC_XXSPLTID): Move to vsx.md.
10481 (UNSPEC_XXSPLTI32DX): Move to vsx.md.
10482 (UNSPEC_XXBLEND): Move to vsx.md.
10483 (UNSPEC_XXPERMX): Move to vsx.md.
10484 (VM3): Move to vsx.md.
10485 (VM3_char): Move to vsx.md.
10486 (xxspltiw_v4si): Move to vsx.md.
10487 (xxspltiw_v4sf): Move to vsx.md.
10488 (xxspltiw_v4sf_inst): Move to vsx.md.
10489 (xxspltidp_v2df): Move to vsx.md.
10490 (xxspltidp_v2df_inst): Move to vsx.md.
10491 (xxsplti32dx_v4si_inst): Move to vsx.md.
10492 (xxsplti32dx_v4sf): Move to vsx.md.
10493 (xxsplti32dx_v4sf_inst): Move to vsx.md.
10494 (xxblend_<mode>): Move to vsx.md.
10495 (xxpermx): Move to vsx.md.
10496 (xxpermx_inst): Move to vsx.md.
10497 * config/rs6000/vsx.md (UNSPEC_XXEVAL): Move from altivec.md.
10498 (UNSPEC_XXSPLTIW): Move from altivec.md.
10499 (UNSPEC_XXSPLTID): Move from altivec.md.
10500 (UNSPEC_XXSPLTI32DX): Move from altivec.md.
10501 (UNSPEC_XXBLEND): Move from altivec.md.
10502 (UNSPEC_XXPERMX): Move from altivec.md.
10503 (VM3): Move from altivec.md.
10504 (VM3_char): Move from altivec.md.
10505 (xxspltiw_v4si): Move from altivec.md.
10506 (xxspltiw_v4sf): Move from altivec.md.
10507 (xxspltiw_v4sf_inst): Move from altivec.md.
10508 (xxspltidp_v2df): Move from altivec.md.
10509 (xxspltidp_v2df_inst): Move from altivec.md.
10510 (xxsplti32dx_v4si_inst): Move from altivec.md.
10511 (xxsplti32dx_v4sf): Move from altivec.md.
10512 (xxsplti32dx_v4sf_inst): Move from altivec.md.
10513 (xxblend_<mode>): Move from altivec.md.
10514 (xxpermx): Move from altivec.md.
10515 (xxpermx_inst): Move from altivec.md.
10517 2021-08-19 Roger Sayle <roger@nextmovesoftware.com>
10519 * tree-vect-generic.c (expand_vector_operations_1): Use either
10520 gimplify_build1 or gimplify_build2 instead of gimple_build_assign
10521 when constructing scalar splat expressions.
10523 2021-08-19 Peter Bergner <bergner@linux.ibm.com>
10526 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Cast
10527 pointer to __vector_pair *.
10529 2021-08-19 Martin Sebor <msebor@redhat.com>
10531 * gimple-range.cc: Add comments.
10532 * gimple-range.h: Same.
10534 2021-08-19 Martin Sebor <msebor@redhat.com>
10536 PR middle-end/101984
10537 * gimple-ssa-warn-access.cc (pass_waccess::execute): Also call
10540 2021-08-19 Jeff Law <jlaw@localhost.localdomain>
10542 * config.gcc (h8300-*-elf*): Do not include dbxelf.h.
10543 (h8300-*-linux*, v850-*-rtems*, v850*-elf*): Likewise.
10544 * config/v850/v850.h (DEFAULT_GDB_EXTENSIONS): Remove.
10546 2021-08-19 Jakub Jelinek <jakub@redhat.com>
10548 PR middle-end/101950
10549 * optabs.c (expand_clrsb_using_clz): New function.
10550 (expand_unop): Use it as another clrsb expansion fallback.
10552 2021-08-19 liuhongt <hongtao.liu@intel.com>
10555 2021-07-28 liuhongt <hongtao.liu@intel.com>
10558 * config/i386/i386.h (processor_costs): Add new member
10560 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
10561 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
10562 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
10563 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
10564 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
10565 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
10566 generic_cost, core_cost): Initialize integer_to_sse same value
10568 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
10569 * config/i386/i386.c (ix86_builtin_vectorization_cost):
10570 Use integer_to_sse instead of sse_op to calculate the cost of
10573 2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
10575 * config.gcc: Include rpath.opt for Darwin.
10576 * config/darwin.h (DRIVER_SELF_SPECS): Handle -rpath.
10578 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
10580 PR bootstrap/101959
10581 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
10584 2021-08-18 Jonathan Wright <jonathan.wright@arm.com>
10586 * config/aarch64/arm_neon.h (vld3_lane_f64): Use float RTL
10587 pattern and type cast.
10588 (vld4_lane_f32): Use float RTL pattern.
10589 (vld4q_lane_f64): Use float type cast.
10591 2021-08-18 Jan Hubicka <hubicka@ucw.cz>
10593 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Check also
10596 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
10598 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor): Extend.
10599 (test_map_of_type_with_ctor_and_dtor_expand): Add function.
10600 (hash_map_tests_c_tests): Call it.
10602 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
10604 * ggc.h (enum ggc_collect): New.
10605 (ggc_collect): Use it.
10606 * ggc-page.c: Adjust.
10607 * ggc-common.c: Likewise.
10608 * ggc-tests.c: Likewise.
10609 * read-rtl-function.c: Likewise.
10610 * selftest-run-tests.c: Likewise.
10611 * doc/gty.texi (Invoking the garbage collector): Likewise.
10613 2021-08-18 liuhongt <hongtao.liu@intel.com>
10616 * config/i386/i386.h (TARGET_V2DF_REDUCTION_PREFER_HADDPD):
10618 * config/i386/sse.md (*sse3_haddv2df3_low): Add
10619 TARGET_V2DF_REDUCTION_PREFER_HADDPD.
10620 (*sse3_hsubv2df3_low): Ditto.
10621 * config/i386/x86-tune.def
10622 (X86_TUNE_V2DF_REDUCTION_PREFER_HADDPD): New tune.
10624 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
10626 * gimple-range-gori.cc (gori_compute::gori_compute): Enable tracing.
10627 (gori_compute::compute_operand_range): Add tracing.
10628 (gori_compute::logical_combine): Ditto.
10629 (gori_compute::compute_logical_operands): Ditto.
10630 (gori_compute::compute_operand1_range): Ditto.
10631 (gori_compute::compute_operand2_range): Ditto.
10632 (gori_compute::outgoing_edge_range_p): Ditto.
10633 * gimple-range-gori.h (class gori_compute): Add range_tracer.
10635 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
10637 * flag-types.h (enum evrp_mode): Adjust evrp-mode values.
10638 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Relocate from.
10639 * gimple-range-trace.h (DEBUG_RANGE_CACHE): Here.
10640 * params.opt (--param=evrp-mode): Adjust options.
10642 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
10644 * Makefile.in (OBJS): Add gimple-range-trace.o.
10645 * gimple-range-cache.h (enable_new_values): Remove unused prototype.
10646 * gimple-range-fold.cc: Adjust headers.
10647 * gimple-range-trace.cc: New.
10648 * gimple-range-trace.h: New.
10649 * gimple-range.cc (gimple_ranger::gimple_ranger): Enable tracer.
10650 (gimple_ranger::range_of_expr): Add tracing.
10651 (gimple_ranger::range_on_entry): Ditto.
10652 (gimple_ranger::range_on_exit): Ditto.
10653 (gimple_ranger::range_on_edge): Ditto.
10654 (gimple_ranger::fold_range_internal): Ditto.
10655 (gimple_ranger::dump_bb): Do not calculate edge range twice.
10656 (trace_ranger::*): Remove.
10657 (enable_ranger): Never create a trace_ranger.
10658 (debug_seed_ranger): Move to gimple-range-trace.cc.
10659 (dump_ranger): Ditto.
10660 (debug_ranger): Ditto.
10661 * gimple-range.h: Include gimple-range-trace.h.
10662 (range_on_entry, range_on_exit): No longer virtual.
10663 (class trace_ranger): Remove.
10664 (DEBUG_RANGE_CACHE): Move to gimple-range-trace.h.
10666 2021-08-17 Martin Sebor <msebor@redhat.com>
10668 PR middle-end/101854
10669 * builtins.c (expand_builtin_alloca): Move warning code to check_alloca
10670 in gimple-ssa-warn-access.cc.
10671 * calls.c (alloc_max_size): Move code to check_alloca.
10672 (get_size_range): Move to pointer-query.cc.
10673 (maybe_warn_alloc_args_overflow): Move to gimple-ssa-warn-access.cc.
10674 (get_attr_nonstring_decl): Move to tree.c.
10675 (fntype_argno_type): Move to gimple-ssa-warn-access.cc.
10676 (append_attrname): Same.
10677 (maybe_warn_rdwr_sizes): Same.
10678 (initialize_argument_information): Move code to
10679 gimple-ssa-warn-access.cc.
10680 * calls.h (maybe_warn_alloc_args_overflow): Move to
10681 gimple-ssa-warn-access.h.
10682 (get_attr_nonstring_decl): Move to tree.h.
10683 (maybe_warn_nonstring_arg): Move to gimple-ssa-warn-access.h.
10684 (enum size_range_flags): Move to pointer-query.h.
10685 (get_size_range): Same.
10686 * gimple-ssa-warn-access.cc (has_location): Remove unused overload
10687 to avoid Clang -Wunused-function.
10688 (get_size_range): Declare static.
10689 (maybe_emit_free_warning): Rename...
10690 (maybe_check_dealloc_call): ...to this for consistency.
10691 (class pass_waccess): Add members.
10692 (pass_waccess::~pass_waccess): Defined.
10693 (alloc_max_size): Move here from calls.c.
10694 (maybe_warn_alloc_args_overflow): Same.
10695 (check_alloca): New function.
10696 (check_alloc_size_call): New function.
10697 (check_strncat): Handle another warning flag.
10698 (pass_waccess::check_builtin): Handle alloca.
10699 (fntype_argno_type): Move here from calls.c.
10700 (append_attrname): Same.
10701 (maybe_warn_rdwr_sizes): Same.
10702 (pass_waccess::check_call): Define.
10703 (check_nonstring_args): New function.
10704 (pass_waccess::check): Call new member functions.
10705 (pass_waccess::execute): Enable ranger.
10706 * gimple-ssa-warn-access.h (get_size_range): Move here from calls.h.
10707 (maybe_warn_nonstring_arg): Same.
10708 * gimple-ssa-warn-restrict.c: Remove #include.
10709 * pointer-query.cc (get_size_range): Move here from calls.c.
10710 * pointer-query.h (enum size_range_flags): Same.
10711 (get_size_range): Same.
10712 * tree.c (get_attr_nonstring_decl): Move here from calls.c.
10713 * tree.h (get_attr_nonstring_decl): Move here from calls.h.
10715 2021-08-17 Thomas Schwinge <thomas@codesourcery.com>
10717 * ggc.h (ggc_collect): Add 'force_collect' parameter.
10718 * ggc-page.c (ggc_collect): Use that one instead of global
10719 'ggc_force_collect'. Adjust all users.
10720 * doc/gty.texi (Invoking the garbage collector): Update.
10721 * ggc-internal.h (ggc_force_collect): Remove.
10722 * ggc-common.c (ggc_force_collect): Likewise.
10723 * selftest.h (forcibly_ggc_collect): Remove.
10724 * ggc-tests.c (selftest::forcibly_ggc_collect): Likewise.
10725 * read-rtl-function.c (test_loading_labels): Adjust.
10726 * selftest-run-tests.c (run_tests): Likewise.
10728 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
10730 * config/darwin.c (darwin_file_end): Reset and reclaim the
10731 section names table at the end of compile.
10733 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
10736 * config.in: Regenerate.
10737 * config/i386/darwin.h (EXTRA_ASM_OPTS): New
10738 (ASM_SPEC): Pass options to disable branch shortening where
10740 * configure: Regenerate.
10741 * configure.ac: Detect versions of 'as' that support the
10742 optimisation which has the bug.
10744 2021-08-17 Richard Biener <rguenther@suse.de>
10746 * optabs-query.c (supports_vec_gather_load_p): Also check
10748 (supports_vec_scatter_store_p): Likewise.
10749 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Fall
10750 back to masked variants if non-masked are not supported.
10751 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
10752 When we need to use masked gather/scatter but do not have
10753 a mask set up a constant true one.
10754 * tree-vect-stmts.c (vect_check_scalar_mask): Also allow
10755 non-SSA_NAME masks.
10757 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
10759 * tree-ssa-ccp.c (bit_value_binop) [MINUS_EXPR]: Use same
10760 algorithm as PLUS_EXPR to improve subtraction bit bounds.
10761 [POINTER_DIFF_EXPR]: Treat as synonymous with MINUS_EXPR.
10763 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
10765 * tree-ssa-ccp.c (bit_value_mult_const): New helper function to
10766 calculate the mask-value pair result of a multiplication by an
10768 (bit_value_binop) [MULT_EXPR]: Call it from here for
10769 multiplications by (sparse) non-negative constants.
10771 2021-08-17 Christophe Lyon <christophe.lyon@foss.st.com>
10774 * config.gcc (gcc_cv_initfini_array): Leave undefined for
10775 uclinuxfdpiceabi targets.
10777 2021-08-17 Alexandre Oliva <oliva@adacore.com>
10779 * tree-inline.c (maybe_move_debug_stmts_to_successors): Don't
10780 reverse debug stmts.
10782 2021-08-17 Alexandre Oliva <oliva@adacore.com>
10784 * tree-cfg.c (dump_function_to_file): Use fun, not cfun.
10786 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
10788 * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete.
10789 (__LD4Q_LANE_FUNC): Likewise.
10790 (vld4_lane_u8): Define without macro.
10791 (vld4_lane_u16): Likewise.
10792 (vld4_lane_u32): Likewise.
10793 (vld4_lane_u64): Likewise.
10794 (vld4_lane_s8): Likewise.
10795 (vld4_lane_s16): Likewise.
10796 (vld4_lane_s32): Likewise.
10797 (vld4_lane_s64): Likewise.
10798 (vld4_lane_f16): Likewise.
10799 (vld4_lane_f32): Likewise.
10800 (vld4_lane_f64): Likewise.
10801 (vld4_lane_p8): Likewise.
10802 (vld4_lane_p16): Likewise.
10803 (vld4_lane_p64): Likewise.
10804 (vld4q_lane_u8): Likewise.
10805 (vld4q_lane_u16): Likewise.
10806 (vld4q_lane_u32): Likewise.
10807 (vld4q_lane_u64): Likewise.
10808 (vld4q_lane_s8): Likewise.
10809 (vld4q_lane_s16): Likewise.
10810 (vld4q_lane_s32): Likewise.
10811 (vld4q_lane_s64): Likewise.
10812 (vld4q_lane_f16): Likewise.
10813 (vld4q_lane_f32): Likewise.
10814 (vld4q_lane_f64): Likewise.
10815 (vld4q_lane_p8): Likewise.
10816 (vld4q_lane_p16): Likewise.
10817 (vld4q_lane_p64): Likewise.
10818 (vld4_lane_bf16): Likewise.
10819 (vld4q_lane_bf16): Likewise.
10821 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
10823 * config/aarch64/arm_neon.h (__LD3_LANE_FUNC): Delete.
10824 (__LD3Q_LANE_FUNC): Delete.
10825 (vld3_lane_u8): Define without macro.
10826 (vld3_lane_u16): Likewise.
10827 (vld3_lane_u32): Likewise.
10828 (vld3_lane_u64): Likewise.
10829 (vld3_lane_s8): Likewise.
10830 (vld3_lane_s16): Likewise.
10831 (vld3_lane_s32): Likewise.
10832 (vld3_lane_s64): Likewise.
10833 (vld3_lane_f16): Likewise.
10834 (vld3_lane_f32): Likewise.
10835 (vld3_lane_f64): Likewise.
10836 (vld3_lane_p8): Likewise.
10837 (vld3_lane_p16): Likewise.
10838 (vld3_lane_p64): Likewise.
10839 (vld3q_lane_u8): Likewise.
10840 (vld3q_lane_u16): Likewise.
10841 (vld3q_lane_u32): Likewise.
10842 (vld3q_lane_u64): Likewise.
10843 (vld3q_lane_s8): Likewise.
10844 (vld3q_lane_s16): Likewise.
10845 (vld3q_lane_s32): Likewise.
10846 (vld3q_lane_s64): Likewise.
10847 (vld3q_lane_f16): Likewise.
10848 (vld3q_lane_f32): Likewise.
10849 (vld3q_lane_f64): Likewise.
10850 (vld3q_lane_p8): Likewise.
10851 (vld3q_lane_p16): Likewise.
10852 (vld3q_lane_p64): Likewise.
10853 (vld3_lane_bf16): Likewise.
10854 (vld3q_lane_bf16): Likewise.
10856 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
10858 * config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Delete.
10859 (__LD2Q_LANE_FUNC): Likewise.
10860 (vld2_lane_u8): Define without macro.
10861 (vld2_lane_u16): Likewise.
10862 (vld2_lane_u32): Likewise.
10863 (vld2_lane_u64): Likewise.
10864 (vld2_lane_s8): Likewise.
10865 (vld2_lane_s16): Likewise.
10866 (vld2_lane_s32): Likewise.
10867 (vld2_lane_s64): Likewise.
10868 (vld2_lane_f16): Likewise.
10869 (vld2_lane_f32): Likewise.
10870 (vld2_lane_f64): Likewise.
10871 (vld2_lane_p8): Likewise.
10872 (vld2_lane_p16): Likewise.
10873 (vld2_lane_p64): Likewise.
10874 (vld2q_lane_u8): Likewise.
10875 (vld2q_lane_u16): Likewise.
10876 (vld2q_lane_u32): Likewise.
10877 (vld2q_lane_u64): Likewise.
10878 (vld2q_lane_s8): Likewise.
10879 (vld2q_lane_s16): Likewise.
10880 (vld2q_lane_s32): Likewise.
10881 (vld2q_lane_s64): Likewise.
10882 (vld2q_lane_f16): Likewise.
10883 (vld2q_lane_f32): Likewise.
10884 (vld2q_lane_f64): Likewise.
10885 (vld2q_lane_p8): Likewise.
10886 (vld2q_lane_p16): Likewise.
10887 (vld2q_lane_p64): Likewise.
10888 (vld2_lane_bf16): Likewise.
10889 (vld2q_lane_bf16): Likewise.
10891 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
10893 * haifa-sched.c (advance_one_cycle): Output more context-synchronization
10896 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
10898 * haifa-sched.c (enum rfs_decision, rfs_str): Add RFS_AUTOPREF.
10899 (rank_for_schedule): Use it.
10901 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
10903 PR rtl-optimization/91598
10904 * haifa-sched.c (autopref_rank_for_schedule): Prioritize "irrelevant"
10905 insns after memory reads and before memory writes.
10907 2021-08-17 Alistair_Lee <alistair.lee@arm.com>
10909 * rtl.h (CONST_VECTOR_P): New macro.
10910 * config/aarch64/aarch64.c (aarch64_get_sve_pred_bits): Use RTL
10911 code testing macros.
10912 (aarch64_ptrue_all_mode): Likewise.
10913 (aarch64_expand_mov_immediate): Likewise.
10914 (aarch64_const_vec_all_in_range_p): Likewise.
10915 (aarch64_rtx_costs): Likewise.
10916 (aarch64_legitimate_constant_p): Likewise.
10917 (aarch64_simd_valid_immediate): Likewise.
10918 (aarch64_simd_make_constant): Likewise.
10919 (aarch64_convert_mult_to_shift): Likewise.
10920 (aarch64_expand_sve_vec_perm): Likewise.
10921 (aarch64_vec_fpconst_pow_of_2): Likewise.
10923 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
10925 PR tree-optimization/101938
10926 * range-op.cc (operator_abs::op1_range): Special case
10927 -TYPE_MIN_VALUE for flag_wrapv.
10929 2021-08-17 Kewen Lin <linkw@linux.ibm.com>
10931 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Add the cost for
10934 2021-08-17 Jakub Jelinek <jakub@redhat.com>
10936 * tree.def (OMP_SCOPE): New tree code.
10937 * tree.h (OMP_SCOPE_BODY, OMP_SCOPE_CLAUSES): Define.
10938 * tree-nested.c (convert_nonlocal_reference_stmt,
10939 convert_local_reference_stmt, convert_gimple_call): Handle
10941 * tree-pretty-print.c (dump_generic_node): Handle OMP_SCOPE.
10942 * gimple.def (GIMPLE_OMP_SCOPE): New gimple code.
10943 * gimple.c (gimple_build_omp_scope): New function.
10944 (gimple_copy): Handle GIMPLE_OMP_SCOPE.
10945 * gimple.h (gimple_build_omp_scope): Declare.
10946 (gimple_has_substatements): Handle GIMPLE_OMP_SCOPE.
10947 (gimple_omp_scope_clauses, gimple_omp_scope_clauses_ptr,
10948 gimple_omp_scope_set_clauses): New inline functions.
10949 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_SCOPE.
10950 * gimple-pretty-print.c (dump_gimple_omp_scope): New function.
10951 (pp_gimple_stmt_1): Handle GIMPLE_OMP_SCOPE.
10952 * gimple-walk.c (walk_gimple_stmt): Likewise.
10953 * gimple-low.c (lower_stmt): Likewise.
10954 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
10955 (gimplify_scan_omp_clauses): For task reductions, handle OMP_SCOPE
10956 like ORT_WORKSHARE constructs. Adjust diagnostics for %<scope%>
10957 allowing task reductions. Reject inscan reductions on scope.
10958 (omp_find_stores_stmt): Handle GIMPLE_OMP_SCOPE.
10959 (gimplify_omp_workshare, gimplify_expr): Handle OMP_SCOPE.
10960 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_SCOPE.
10961 (estimate_num_insns): Likewise.
10962 * omp-low.c (build_outer_var_ref): Look through GIMPLE_OMP_SCOPE
10963 contexts if var isn't privatized there.
10964 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_SCOPE.
10965 (scan_omp_1_stmt): Likewise.
10966 (maybe_add_implicit_barrier_cancel): Look through outer
10968 (lower_omp_scope): New function.
10969 (lower_omp_task_reductions): Handle OMP_SCOPE.
10970 (lower_omp_1): Handle GIMPLE_OMP_SCOPE.
10971 (diagnose_sb_1, diagnose_sb_2): Likewise.
10972 * omp-expand.c (expand_omp_single): Support also GIMPLE_OMP_SCOPE.
10973 (expand_omp): Handle GIMPLE_OMP_SCOPE.
10974 (omp_make_gimple_edges): Likewise.
10975 * omp-builtins.def (BUILT_IN_GOMP_SCOPE_START): New built-in.
10977 2021-08-17 Richard Biener <rguenther@suse.de>
10979 PR tree-optimization/101925
10980 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Set
10981 reverse on COMPONENT_REF and ARRAY_REF according to
10982 what reverse_storage_order_for_component_p does.
10983 (vn_reference_eq): Compare reversed on reference ops.
10984 (reverse_storage_order_for_component_p): New overload.
10985 (vn_reference_lookup_3): Check reverse_storage_order_for_component_p
10986 on the reference looked up.
10988 2021-08-17 Jeff Law <jlaw@localhost.localdomain>
10990 * config/h8300/h8300.c (shift_alg_si): Avoid loops for most SImode
10991 shifts on the H8/S.
10992 (h8300_option_override): Use loops on H8/S more often when optimizing
10994 (get_shift_alg): Handle new "special" cases on H8/S. Simplify
10995 accordingly. Handle various arithmetic right shifts with special
10996 sequences that we couldn't handle before.
10998 2021-08-16 Jeff Law <jlaw@localhost.localdomain>
11000 * config.gcc (rl78-*-elf*): Do not include dbxelf.h.
11002 2021-08-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
11004 * config/sparc/rtemself.h (SPARC_GCOV_TYPE_SIZE): Define.
11005 * config/sparc/sparc.c (sparc_gcov_type_size): New.
11006 (TARGET_GCOV_TYPE_SIZE): Redefine if SPARC_GCOV_TYPE_SIZE is defined.
11007 * coverage.c (get_gcov_type): Use targetm.gcov_type_size().
11008 * doc/tm.texi (TARGET_GCOV_TYPE_SIZE): Add hook under "Misc".
11009 * doc/tm.texi.in: Regenerate.
11010 * target.def (gcov_type_size): New target hook.
11011 * targhooks.c (default_gcov_type_size): New.
11012 * targhooks.h (default_gcov_type_size): Declare.
11013 * tree-profile.c (gimple_gen_edge_profiler): Use precision of
11015 (gimple_gen_time_profiler): Likewise.
11017 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
11019 * dwarf2out.c (add_scalar_info): Deal with DW_AT_data_bit_offset.
11021 2021-08-16 Tobias Burnus <tobias@codesourcery.com>
11023 PR middle-end/101931
11024 * omp-low.c (omp_runtime_api_call): Update for routines
11025 added in the meanwhile.
11027 2021-08-16 Martin Liska <mliska@suse.cz>
11029 PR tree-optimization/100393
11030 * tree-switch-conversion.c (group_cluster::dump): Use
11031 get_comparison_count.
11032 (jump_table_cluster::find_jump_tables): Pre-compute number of
11033 comparisons and then decrement it. Cache also max_ratio.
11034 (jump_table_cluster::can_be_handled): Change signature.
11035 * tree-switch-conversion.h (get_comparison_count): New.
11037 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
11039 * dwarf2out.c (add_data_member_location_attribute): Use GNAT
11040 encodings only when -fgnat-encodings=all is specified.
11041 (add_bound_info): Likewise.
11042 (add_byte_size_attribute): Likewise.
11043 (gen_member_die): Likewise.
11045 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
11047 * omp-oacc-neuter-broadcast.cc
11048 (execute_omp_oacc_neuter_broadcast): Plug 'par' memory leak.
11050 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
11052 * omp-oacc-neuter-broadcast.cc
11053 (execute_omp_oacc_neuter_broadcast): Clarify memory management for
11056 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
11058 * omp-oacc-neuter-broadcast.cc (field_map): Move variable into...
11059 (execute_omp_oacc_neuter_broadcast): ... here.
11060 (install_var_field, build_receiver_ref, build_sender_ref): Take
11061 'field_map_t *' parameter. Adjust all users.
11062 (worker_single_copy, neuter_worker_single): Take a
11063 'record_field_map_t *' parameter. Adjust all users.
11065 2021-08-16 liuhongt <hongtao.liu@intel.com>
11068 * config/i386/i386.md (ldexp<mode>3): Force operands[1] to
11071 2021-08-16 Martin Liska <mliska@suse.cz>
11074 * multiple_target.c (create_dispatcher_calls): Make default
11075 function local only if it is a definition.
11077 2021-08-16 Martin Liska <mliska@suse.cz>
11080 * ipa-icf-gimple.c (func_checker::compare_ssa_name): Do not
11081 consider equal SSA_NAMEs when one is a param.
11083 2021-08-16 liuhongt <hongtao.liu@intel.com>
11086 * config/i386/i386-expand.c (ix86_expand_vec_perm_vpermt2):
11087 Support vpermi2b for V32QI/V16QImode.
11088 (ix86_extract_perm_from_pool_constant): New function.
11089 (ix86_expand_vec_one_operand_perm_avx512): Support
11090 vpermw/vpermb under TARGET_AVX512BW/TARGET_AVX512VBMI.
11091 (expand_vec_perm_1): Adjust comments for upper.
11092 * config/i386/i386-protos.h (ix86_extract_perm_from_pool_constant):
11094 * config/i386/predicates.md (permvar_truncate_operand): New predicate.
11095 (pshufb_truncv4siv4hi_operand): Ditto.
11096 (pshufb_truncv8hiv8qi_operand): Ditto.
11097 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1):
11098 New pre_reload define_insn_and_split.
11099 (*avx512f_permvar_truncv8siv8hi_1): Ditto.
11100 (*avx512f_vpermvar_truncv8div8si_1): Ditto.
11101 (*avx512f_permvar_truncv32hiv32qi_1): Ditto.
11102 (*avx512f_permvar_truncv16hiv16qi_1): Ditto.
11103 (*avx512f_permvar_truncv4div4si_1): Ditto.
11104 (*avx512f_pshufb_truncv8hiv8qi_1): Ditto.
11105 (*avx512f_pshufb_truncv4siv4hi_1): Ditto.
11106 (*avx512f_pshufd_truncv2div2si_1): Ditto.
11108 2021-08-16 Kito Cheng <kito.cheng@sifive.com>
11110 * config/riscv/multilib-generator: Support code model option for
11112 * doc/install.texi: Add document of new option for
11113 --with-multilib-generator.
11115 2021-08-15 Clément Chigot <clement.chigot@atos.net>
11117 * config/rs6000/rs6000.c (xcoff_tls_exec_model_detected): New.
11118 (rs6000_legitimize_tls_address_aix): Use it.
11119 (rs6000_xcoff_file_end): Add ".ref __tls_get_addr" when
11120 xcoff_tls_exec_model_detected is true.
11122 2021-08-15 Jeff Law <jlaw@localhost.localdomain>
11124 * config/h8300/h8300.c (shift_alg_si): Retune H8/300H shifts
11125 to allow a bit more code growth, saving many dozens of cycles.
11126 (h8300_option_override): Adjus shift_alg_si if optimizing for
11128 (get_shift_alg): Use special + inline shifts for residuals
11131 2021-08-14 Stafford Horne <shorne@gmail.com>
11134 * config/or1k/or1k-opts.h: New file.
11135 * config/or1k/or1k.c (or1k_legitimize_address_1, print_reloc):
11136 Support generating gotha relocations if -mcmodel=large is
11138 * config/or1k/or1k.h (TARGET_CMODEL_SMALL, TARGET_CMODEL_LARGE):
11140 * config/or1k/or1k.opt (mcmodel=): New option.
11141 * doc/invoke.texi (OpenRISC Options): Document mcmodel.
11143 2021-08-14 Martin Sebor <msebor@redhat.com>
11145 PR middle-end/101791
11146 * gimple-ssa-warn-access.cc (new_delete_mismatch_p): Use new argument
11147 to valid_new_delete_pair_p.
11148 * tree.c (valid_new_delete_pair_p): Add argument.
11149 * tree.h (valid_new_delete_pair_p): Same.
11151 2021-08-14 Jakub Jelinek <jakub@redhat.com>
11154 * config/i386/i386-expand.c (expand_vec_perm_broadcast_1)
11155 <case E_V64QImode>: For this mode assert
11156 !TARGET_AVX512BW || d->perm[0] rather than !TARGET_AVX2 || d->perm[0].
11158 2021-08-13 Michael Meissner <meissner@linux.ibm.com>
11161 * config/rs6000/altivec.md (xxeval): Use register_predicate
11162 instead of altivec_register_predicate.
11164 2021-08-13 Martin Sebor <msebor@redhat.com>
11166 PR middle-end/101734
11167 * tree-ssa-uninit.c (maybe_warn_read_write_only): New function.
11168 (maybe_warn_operand): Call it.
11170 2021-08-13 Martin Liska <mliska@suse.cz>
11173 * attribs.c (decl_attributes): Make naked functions "noipa"
11176 2021-08-13 Martin Liska <mliska@suse.cz>
11179 * symtab.c (symtab_node::noninterposable_alias): Do not create
11180 local aliases for target_clone functions as the clonning pass
11183 2021-08-13 Martin Liska <mliska@suse.cz>
11185 * opts.c (LIVE_PATCHING_OPTION): Define.
11186 (control_options_for_live_patching): Use it in error messages.
11188 2021-08-13 Jan Hubicka <hubicka@ucw.cz>
11190 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOREAD.
11191 (implicit_const_eaf_flags, implicit_pure_eaf_flags,
11192 ignore_stores_eaf_flags): New constants.
11193 (remove_useless_eaf_flags): New function.
11194 (eaf_flags_useful_p): Use it.
11195 (deref_flags): Add EAF_NOT_RETURNED if flag is unused;
11197 (modref_lattice::init): Add EAF_NOREAD.
11198 (modref_lattice::add_escape_point): Do not reacord escape point if
11200 (modref_lattice::merge): EAF_NOESCAPE implies EAF_NODIRECTESCAPE;
11201 use remove_useless_eaf_flags.
11202 (modref_lattice::merge_deref): Use ignore_stores_eaf_flags.
11203 (modref_lattice::merge_direct_load): Add EAF_NOREAD
11204 (analyze_ssa_name_flags): Fix handling EAF_NOT_RETURNED
11205 (analyze_parms): Use remove_useless_eaf_flags.
11206 (ipa_merge_modref_summary_after_inlining): Use ignore_stores_eaf_flags.
11207 (modref_merge_call_site_flags): Add caller and ecf_flags parameter;
11208 use remove_useless_eaf_flags.
11209 (modref_propagate_flags_in_scc): Update.
11210 * ipa-modref.h: Turn eaf_flags_t back to char.
11211 * tree-core.h (EAF_NOT_RETURNED): Fix.
11212 (EAF_NOREAD): New constant
11213 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1): Check for
11215 * tree-ssa-structalias.c (handle_rhs_call): Handle new flags.
11216 (handle_pure_call): Likewise.
11218 2021-08-12 Jakub Jelinek <jakub@redhat.com>
11220 * tree.def (OMP_MASKED): New tree code.
11221 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_FILTER.
11222 * tree.h (OMP_MASKED_BODY, OMP_MASKED_CLAUSES, OMP_MASKED_COMBINED,
11223 OMP_CLAUSE_FILTER_EXPR): Define.
11224 * tree.c (omp_clause_num_ops): Add OMP_CLAUSE_FILTER entry.
11225 (omp_clause_code_name): Likewise.
11226 (walk_tree_1): Handle OMP_CLAUSE_FILTER.
11227 * tree-nested.c (convert_nonlocal_omp_clauses,
11228 convert_local_omp_clauses): Handle OMP_CLAUSE_FILTER.
11229 (convert_nonlocal_reference_stmt, convert_local_reference_stmt,
11230 convert_gimple_call): Handle GIMPLE_OMP_MASTER.
11231 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_FILTER.
11232 (dump_generic_node): Handle OMP_MASTER.
11233 * gimple.def (GIMPLE_OMP_MASKED): New gimple code.
11234 * gimple.c (gimple_build_omp_masked): New function.
11235 (gimple_copy): Handle GIMPLE_OMP_MASKED.
11236 * gimple.h (gimple_build_omp_masked): Declare.
11237 (gimple_has_substatements): Handle GIMPLE_OMP_MASKED.
11238 (gimple_omp_masked_clauses, gimple_omp_masked_clauses_ptr,
11239 gimple_omp_masked_set_clauses): New inline functions.
11240 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_MASKED.
11241 * gimple-pretty-print.c (dump_gimple_omp_masked): New function.
11242 (pp_gimple_stmt_1): Handle GIMPLE_OMP_MASKED.
11243 * gimple-walk.c (walk_gimple_stmt): Likewise.
11244 * gimple-low.c (lower_stmt): Likewise.
11245 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
11246 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_FILTER. For clauses
11247 that take one expression rather than decl or constant, force
11248 gimplification of that into a SSA_NAME or temporary unless min
11250 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_FILTER.
11251 (gimplify_expr): Handle OMP_MASKED.
11252 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_MASKED.
11253 (estimate_num_insns): Likewise.
11254 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_FILTER.
11255 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_MASKED. Adjust
11256 diagnostics for existence of masked construct.
11257 (scan_omp_1_stmt, lower_omp_master, lower_omp_1, diagnose_sb_1,
11258 diagnose_sb_2): Handle GIMPLE_OMP_MASKED.
11259 * omp-expand.c (expand_omp_synch, expand_omp, omp_make_gimple_edges):
11262 2021-08-12 Uroš Bizjak <ubizjak@gmail.com>
11265 * config/i386/i386.md (avx512f_scalef<mode>2): New insn pattern.
11266 (ldexp<mode>3): Use avx512f_scalef<mode>2.
11267 (UNSPEC_SCALEF): Move from ...
11268 * config/i386/sse.md (UNSPEC_SCALEF): ... here.
11270 2021-08-12 Jan Hubicka <hubicka@ucw.cz>
11272 * ipa-split.c (consider_split): Fix condition testing void functions.
11274 2021-08-12 Aldy Hernandez <aldyh@redhat.com>
11276 * doc/invoke.texi: Remove docs for threader-mode param.
11277 * flag-types.h (enum threader_mode): Remove.
11278 * params.opt: Remove threader-mode param.
11279 * tree-ssa-threadbackward.c (class back_threader): Remove
11280 path_is_unreachable_p.
11281 Make find_paths private.
11282 Add maybe_thread and thread_through_all_blocks.
11283 Remove reference marker for m_registry.
11284 Remove reference marker for m_profit.
11285 (back_threader::back_threader): Adjust for registry and profit not
11287 (dump_path): Move down.
11288 (debug): Move down.
11289 (class thread_jumps): Remove.
11290 (class back_threader_registry): Remove m_all_paths.
11292 (thread_jumps::thread_through_all_blocks): Move to back_threader
11294 (fsm_find_thread_path): Remove
11295 (back_threader::maybe_thread): New.
11296 (back_threader::thread_through_all_blocks): Move from
11298 (back_threader_registry::back_threader_registry): Remove
11300 (back_threader_registry::~back_threader_registry): Remove.
11301 (thread_jumps::find_taken_edge): Remove.
11302 (thread_jumps::check_subpath_and_update_thread_path): Remove.
11303 (thread_jumps::maybe_register_path): Remove.
11304 (thread_jumps::handle_phi): Remove.
11305 (handle_assignment_p): Remove.
11306 (thread_jumps::handle_assignment): Remove.
11307 (thread_jumps::fsm_find_control_statement_thread_paths): Remove.
11308 (thread_jumps::find_jump_threads_backwards): Remove.
11309 (thread_jumps::find_jump_threads_backwards_with_ranger): Remove.
11310 (try_thread_blocks): Rename find_jump_threads_backwards to
11312 (pass_early_thread_jumps::execute): Same.
11314 2021-08-12 Tobias Burnus <tobias@codesourcery.com>
11316 * tree-core.h (omp_clause_proc_bind_kind): Add
11317 OMP_CLAUSE_PROC_BIND_PRIMARY.
11318 * tree-pretty-print.c (dump_omp_clause): Add TODO comment to
11319 change 'master' to 'primary' in proc_bind for OpenMP 5.1.
11321 2021-08-12 Claudiu Zissulescu <claziss@synopsys.com>
11323 * common/config/arc/arc-common.c (arc_option_init_struct): Remove
11324 fno-common reference.
11325 * config/arc/arc.c (arc_override_options): Remove overriding of
11328 2021-08-12 Jakub Jelinek <jakub@redhat.com>
11331 * config/i386/i386-expand.c (ix86_expand_vec_one_operand_perm_avx512):
11332 If d->testing_p, return true after performing checks instead of
11333 actually expanding the insn.
11334 (expand_vec_perm_broadcast_1): Handle V32HImode - assert
11335 !TARGET_AVX512BW and return false.
11337 2021-08-12 Eric Botcazou <ebotcazou@gcc.gnu.org>
11339 * configure.ac (PE linker --disable-dynamicbase support): New check.
11340 * configure: Regenerate.
11341 * config.in: Likewise.
11342 * config/i386/mingw32.h (LINK_SPEC_DISABLE_DYNAMICBASE): New define.
11343 (LINK_SPEC): Use it.
11344 * config/i386/mingw-w64.h (LINK_SPEC_DISABLE_DYNAMICBASE): Likewise.
11345 (LINK_SPEC): Likewise.
11347 2021-08-12 liuhongt <hongtao.liu@intel.com>
11350 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_2): New
11351 post_reload define_insn_and_split.
11352 (*avx512bw_zero_extendv32qiv32hi2_2): Ditto.
11353 (*sse4_1_zero_extendv8qiv8hi2_4): Ditto.
11354 (*avx512f_zero_extendv16hiv16si2_2): Ditto.
11355 (*avx2_zero_extendv8hiv8si2_2): Ditto.
11356 (*sse4_1_zero_extendv4hiv4si2_4): Ditto.
11357 (*avx512f_zero_extendv8siv8di2_2): Ditto.
11358 (*avx2_zero_extendv4siv4di2_2): Ditto.
11359 (*sse4_1_zero_extendv2siv2di2_4): Ditto.
11360 (VI248_256, VI248_512, VI148_512, VI148_256, VI148_128): New
11363 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
11365 * config/rs6000/rs6000-builtin-new.def: Add always, power5, and
11368 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
11370 * config/rs6000/rs6000-builtin-new.def: Add vsx stanza.
11372 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
11374 * config/rs6000/rs6000-builtin-new.def: Finish altivec stanza.
11375 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
11376 initialization of pcvoid_type_node here...
11377 (altivec_init_builtins): ...from here.
11378 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add
11379 RS6000_BTI_const_ptr_void.
11380 (pcvoid_type_node): New macro.
11382 2021-08-11 Richard Biener <rguenther@suse.de>
11385 * tree-ssa-forwprop.c (pass_forwprop::execute): Do not decompose
11386 hard-register accesses.
11388 2021-08-11 Richard Biener <rguenther@suse.de>
11390 * tree-ssa-operands.c (operands_scanner::get_expr_operands):
11391 Do not look at COMPONENT_REF FIELD_DECLs TREE_THIS_VOLATILE
11392 to determine has_volatile_ops.
11394 2021-08-11 Eric Botcazou <ebotcazou@gcc.gnu.org>
11396 * cfgexpand.c (expand_used_vars): Reuse attribs local variable.
11398 2021-08-11 Jan Hubicka <hubicka@ucw.cz>
11399 Alexandre Oliva <oliva@adacore.com>
11401 * ipa-modref.c (modref_lattice::dump): Fix escape_point's min_flags
11403 (modref_lattice::merge_deref): Fix handling of indirect scape points.
11404 (update_escape_summary_1): Likewise.
11405 (update_escape_summary): Likewise.
11406 (ipa_merge_modref_summary_after_inlining): Likewise.
11408 2021-08-11 Richard Biener <rguenther@suse.de>
11410 PR middle-end/101858
11411 * fold-const.c (fold_binary_loc): Guard simplification
11412 of X < (cast) (1 << Y) to integer types.
11414 2021-08-11 Richard Biener <rguenther@suse.de>
11416 PR tree-optimization/101861
11417 * tree-vect-stmts.c (vectorizable_load): Fix error in
11418 previous change with regard to gather vectorization.
11420 2021-08-11 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
11423 * config/arm/arm_neon.h (vdup_n_s8): Replace call to builtin
11425 (vdup_n_s16): Likewise.
11426 (vdup_n_s32): Likewise.
11427 (vdup_n_s64): Likewise.
11428 (vdup_n_u8): Likewise.
11429 (vdup_n_u16): Likewise.
11430 (vdup_n_u32): Likewise.
11431 (vdup_n_u64): Likewise.
11432 (vdup_n_p8): Likewise.
11433 (vdup_n_p16): Likewise.
11434 (vdup_n_p64): Likewise.
11435 (vdup_n_f16): Likewise.
11436 (vdup_n_f32): Likewise.
11437 (vdupq_n_s8): Likewise.
11438 (vdupq_n_s16): Likewise.
11439 (vdupq_n_s32): Likewise.
11440 (vdupq_n_s64): Likewise.
11441 (vdupq_n_u8): Likewise.
11442 (vdupq_n_u16): Likewise.
11443 (vdupq_n_u32): Likewise.
11444 (vdupq_n_u64): Likewise.
11445 (vdupq_n_p8): Likewise.
11446 (vdupq_n_p16): Likewise.
11447 (vdupq_n_p64): Likewise.
11448 (vdupq_n_f16): Likewise.
11449 (vdupq_n_f32): Likewise.
11450 (vmov_n_s8): Replace call to builtin with call to corresponding
11452 (vmov_n_s16): Likewise.
11453 (vmov_n_s32): Likewise.
11454 (vmov_n_s64): Likewise.
11455 (vmov_n_u8): Likewise.
11456 (vmov_n_u16): Likewise.
11457 (vmov_n_u32): Likewise.
11458 (vmov_n_u64): Likewise.
11459 (vmov_n_p8): Likewise.
11460 (vmov_n_p16): Likewise.
11461 (vmov_n_f16): Likewise.
11462 (vmov_n_f32): Likewise.
11463 (vmovq_n_s8): Likewise.
11464 (vmovq_n_s16): Likewise.
11465 (vmovq_n_s32): Likewise.
11466 (vmovq_n_s64): Likewise.
11467 (vmovq_n_u8): Likewise.
11468 (vmovq_n_u16): Likewise.
11469 (vmovq_n_u32): Likewise.
11470 (vmovq_n_u64): Likewise.
11471 (vmovq_n_p8): Likewise.
11472 (vmovq_n_p16): Likewise.
11473 (vmovq_n_f16): Likewise.
11474 (vmovq_n_f32): Likewise.
11475 * config/arm/arm_neon_builtins.def: Remove entries for vdup_n.
11477 2021-08-11 liuhongt <hongtao.liu@intel.com>
11480 * config/i386/i386.md (ldexp<mode>3): Extend to vscalefs[sd]
11481 when TARGET_AVX512F and TARGET_SSE_MATH.
11483 2021-08-10 Jakub Jelinek <jakub@redhat.com>
11486 * config/i386/i386-expand.c (expand_vec_perm_even_odd): Return false
11487 for V32HImode if !TARGET_AVX512BW.
11488 (ix86_vectorize_vec_perm_const) <case E_V32HImode, case E_V64QImode>:
11489 If !TARGET_AVX512BW and TARGET_AVX512F and d.testing_p, don't fail
11490 early, but actually check the permutation.
11492 2021-08-10 Richard Biener <rguenther@suse.de>
11494 PR tree-optimization/101809
11495 * tree-vect-stmts.c (get_load_store_type): Allow emulated
11496 gathers with offset vector nunits being a constant multiple
11497 of the data vector nunits.
11498 (vect_get_gather_scatter_ops): Use the appropriate nunits
11499 for the offset vector defs.
11500 (vectorizable_store): Adjust call to
11501 vect_get_gather_scatter_ops.
11502 (vectorizable_load): Likewise. Handle the case of less
11503 offset vectors than data vectors.
11505 2021-08-10 Jakub Jelinek <jakub@redhat.com>
11508 * config/i386/sse.md (*avx512f_shuf_<shuffletype>64x2_1<mask_name>_1,
11509 *avx512f_shuf_<shuffletype>32x4_1<mask_name>_1): New define_insn
11512 2021-08-10 Richard Biener <rguenther@suse.de>
11514 PR tree-optimization/101801
11515 PR tree-optimization/101819
11516 * tree-vectorizer.h (vect_emulated_vector_p): Declare.
11517 * tree-vect-loop.c (vect_emulated_vector_p): New function.
11518 (vectorizable_reduction): Re-instantiate a check for emulated
11520 * tree-vect-stmts.c (vectorizable_shift): Likewise.
11521 (vectorizable_operation): Likewise. Cost emulated vector
11522 operations according to the scalar sequence synthesized by
11525 2021-08-10 Richard Biener <rguenther@suse.de>
11527 PR middle-end/101824
11528 * tree-nested.c (get_frame_field): Mark the COMPONENT_REF as
11529 volatile in case the variable was.
11531 2021-08-10 H.J. Lu <hjl.tools@gmail.com>
11534 * config/i386/constraints.md (BC): Document for integer SSE
11535 constant all bits set operand.
11536 (BF): New constraint for const floating-point all bits set
11538 * config/i386/i386.c (standard_sse_constant_p): Likewise.
11539 (standard_sse_constant_opcode): Likewise.
11540 * config/i386/sse.md (sseconstm1): New mode attribute.
11541 (mov<mode>_internal): Replace BC with <sseconstm1>.
11543 2021-08-10 liuhongt <hongtao.liu@intel.com>
11545 * config/i386/sse.md (cond_<insn><mode>): New expander.
11546 (VI248_AVX512VLBW): New mode iterator.
11547 * config/i386/predicates.md
11548 (nonimmediate_or_const_vec_dup_operand): New predicate.
11550 2021-08-09 Andrew MacLeod <amacleod@redhat.com>
11552 PR tree-optimization/101741
11553 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Check
11554 type of parameter for toupper/tolower.
11556 2021-08-09 Martin Jambor <mjambor@suse.cz>
11558 PR testsuite/101654
11559 * ipa-prop.c (propagate_controlled_uses): Removed a spurious space.
11561 2021-08-09 Pat Haugen <pthaugen@linux.ibm.com>
11563 * config/rs6000/rs6000.c (is_load_insn1): Verify destination is a
11565 (is_store_insn1): Verify source is a register.
11567 2021-08-09 Uroš Bizjak <ubizjak@gmail.com>
11570 * config/i386/mmx.md (<any_logic:code>v2sf3):
11571 Rename from *mmx_<any_logic:code>v2sf3
11573 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11575 * config/nvptx/nvptx.c: Cross-reference parts adapted in
11576 'gcc/omp-oacc-neuter-broadcast.cc'.
11577 * omp-low.c: Likewise.
11578 * omp-oacc-neuter-broadcast.cc: Cross-reference parts adapted from
11581 2021-08-09 Julian Brown <julian@codesourcery.com>
11582 Kwok Cheung Yeung <kcy@codesourcery.com>
11583 Thomas Schwinge <thomas@codesourcery.com>
11585 * config/gcn/gcn.c (gcn_init_builtins): Override decls for
11586 BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START,
11587 BUILT_IN_GOACC_SINGLE_COPY_END and BUILT_IN_GOACC_BARRIER.
11588 (gcn_goacc_validate_dims): Turn on worker partitioning unconditionally.
11589 (gcn_fork_join): Update comment.
11590 * config/gcn/gcn.opt (flag_worker_partitioning): Remove.
11591 (macc_experimental_workers): Remove unused option.
11593 2021-08-09 Julian Brown <julian@codesourcery.com>
11594 Nathan Sidwell <nathan@codesourcery.com> (via 'gcc/config/nvptx/nvptx.c' master)
11595 Kwok Cheung Yeung <kcy@codesourcery.com>
11596 Thomas Schwinge <thomas@codesourcery.com>
11598 * Makefile.in (OBJS): Add omp-oacc-neuter-broadcast.o.
11599 * doc/tm.texi.in (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD):
11600 Add documentation hook.
11601 * doc/tm.texi: Regenerate.
11602 * omp-oacc-neuter-broadcast.cc: New file.
11603 * omp-builtins.def (BUILT_IN_GOACC_BARRIER)
11604 (BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START)
11605 (BUILT_IN_GOACC_SINGLE_COPY_END): New builtins.
11606 * passes.def (pass_omp_oacc_neuter_broadcast): Add pass.
11607 * target.def (goacc.create_worker_broadcast_record): Add target
11609 * tree-pass.h (make_pass_omp_oacc_neuter_broadcast): Add
11611 * config/gcn/gcn-protos.h (gcn_goacc_adjust_propagation_record):
11612 Rename prototype to...
11613 (gcn_goacc_create_worker_broadcast_record): ... this.
11614 * config/gcn/gcn-tree.c (gcn_goacc_adjust_propagation_record): Rename
11616 (gcn_goacc_create_worker_broadcast_record): ... this.
11617 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_PROPAGATION_RECORD):
11619 (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD): ... this.
11621 2021-08-09 Tejas Belagod <tejas.belagod@arm.com>
11624 * config/aarch64/aarch64-simd.md (vlshr<mode>3, vashr<mode>3): Use
11625 the right iterator.
11627 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11629 * Makefile.in (GTFILES): Remove '$(srcdir)/omp-offload.c'.
11631 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11633 * builtins.def (DEF_GOACC_BUILTIN, DEF_GOMP_BUILTIN): Don't
11634 consider '-foffload-abi'.
11635 * common.opt (-foffload-abi): Remove 'Var', 'Init'.
11636 * opts.c (common_handle_option) <-foffload-abi> [ACCEL_COMPILER]:
11639 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11641 * optc-gen.awk: Sanity check that 'Init' doesn't appear without
11644 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11646 * omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE): Remove.
11648 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11650 * doc/gty.texi (Files): Update.
11652 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
11654 * doc/gty.texi (Files): Fix GTY header file example.
11656 2021-08-09 Roger Sayle <roger@nextmovesoftware.com>
11658 * tree-ssa-ccp.c (value_mask_to_min_max): Helper function to
11659 determine the upper and lower bounds from a mask-value pair.
11660 (bit_value_unop) [ABS_EXPR, ABSU_EXPR]: Add support for
11661 absolute value and unsigned absolute value expressions.
11662 (bit_value_binop): Initialize *VAL's precision.
11663 [LT_EXPR, LE_EXPR]: Use value_mask_to_min_max to determine
11664 upper and lower bounds of operands. Add LE_EXPR/GE_EXPR
11665 support when the operands are unknown but potentially equal.
11666 [MIN_EXPR, MAX_EXPR]: Support minimum/maximum expressions.
11668 2021-08-09 Bin Cheng <bin.cheng@linux.alibaba.com>
11670 * config/aarch64/aarch64.md
11671 (*extend<SHORT:mode><GPI:mode>2_aarch64): Use %<GPI:w>0.
11673 2021-08-08 Sergei Trofimovich <siarheit@google.com>
11675 * lra-constraints.c: Fix s/otput/output/ typo.
11677 2021-08-06 Martin Sebor <msebor@redhat.com>
11679 * builtins.c (expand_builtin_memchr): Move to gimple-ssa-warn-access.cc.
11680 (expand_builtin_strcat): Same.
11681 (expand_builtin_stpncpy): Same.
11682 (expand_builtin_strncat): Same.
11683 (check_read_access): Same.
11684 (check_memop_access): Same.
11685 (expand_builtin_strlen): Move checks to gimple-ssa-warn-access.cc.
11686 (expand_builtin_strnlen): Same.
11687 (expand_builtin_memcpy): Same.
11688 (expand_builtin_memmove): Same.
11689 (expand_builtin_mempcpy): Same.
11690 (expand_builtin_strcpy): Same.
11691 (expand_builtin_strcpy_args): Same.
11692 (expand_builtin_stpcpy_1): Same.
11693 (expand_builtin_strncpy): Same.
11694 (expand_builtin_memset): Same.
11695 (expand_builtin_bzero): Same.
11696 (expand_builtin_strcmp): Same.
11697 (expand_builtin_strncmp): Same.
11698 (expand_builtin): Remove handlers.
11699 (fold_builtin_strlen): Add a comment.
11700 * builtins.h (check_access): Move to gimple-ssa-warn-access.cc.
11701 * calls.c (maybe_warn_nonstring_arg): Same.
11702 * diagnostic-spec.c (nowarn_spec_t::nowarn_spec_t): Add warning option.
11703 * gimple-fold.c (gimple_fold_builtin_strcpy): Pass argument to callee.
11704 (gimple_fold_builtin_stpcpy): Same.
11705 * gimple-ssa-warn-access.cc (has_location): New function.
11706 (get_location): Same.
11707 (get_callee_fndecl): Same.
11708 (call_nargs): Same.
11710 (warn_string_no_nul): Define.
11711 (unterminated_array): Same.
11712 (check_nul_terminated_array): Same.
11713 (maybe_warn_nonstring_arg): Same.
11714 (maybe_warn_for_bound): Same.
11715 (warn_for_access): Same.
11716 (check_access): Same.
11717 (check_memop_access): Same.
11718 (check_read_access): Same.
11719 (warn_dealloc_offset): Use helper functions.
11720 (maybe_emit_free_warning): Same.
11721 (class pass_waccess): Add members.
11722 (check_strcat): New function.
11723 (check_strncat): New function.
11724 (check_stxcpy): New function.
11725 (check_stxncpy): New function.
11726 (check_strncmp): New function.
11727 (pass_waccess::check_builtin): New function.
11728 (pass_waccess::check): Call it.
11729 * gimple-ssa-warn-access.h (warn_string_no_nul): Move here from
11731 (maybe_warn_for_bound): Same.
11732 (check_access): Same.
11733 (check_memop_access): Same.
11734 (check_read_access): Same.
11735 * pointer-query.h (struct access_data): Define a ctor overload.
11737 2021-08-06 Richard Biener <rguenther@suse.de>
11739 PR tree-optimization/101801
11740 * tree-vectorizer.h (vect_worthwhile_without_simd_p): Rename...
11741 (vect_can_vectorize_without_simd_p): ... to this.
11742 * tree-vect-loop.c (vect_worthwhile_without_simd_p): Rename...
11743 (vect_can_vectorize_without_simd_p): ... to this and fold
11744 in vect_min_worthwhile_factor.
11745 (vect_min_worthwhile_factor): Remove.
11746 (vectorizable_reduction): Adjust and remove the cost part.
11747 * tree-vect-stmts.c (vectorizable_shift): Likewise.
11748 (vectorizable_operation): Likewise.
11750 2021-08-06 Uroš Bizjak <ubizjak@gmail.com>
11753 * config/i386/i386.md (cmove reg-to-reg move elimination peephole2s):
11754 Add general_gr_operand predicate to operand 3.
11756 2021-08-06 Roger Sayle <roger@nextmovesoftware.com>
11758 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern): Use
11759 CFN_BUILT_IN_CLRSB* instead of BUILT_IN_CLRSB* for consistency.
11761 2021-08-06 Tamar Christina <tamar.christina@arm.com>
11763 * config/aarch64/aarch64-sve-builtins.cc (register_svpattern,
11764 register_svprfop): Pass vec<> by pointer.
11765 * langhooks-def.h (lhd_simulate_enum_decl): Likewise.
11766 * langhooks.c (lhd_simulate_enum_decl): Likewise.
11767 * langhooks.h (struct lang_hooks_for_types): Likewise.
11769 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
11771 * config/aarch64/arm_neon.h (vst1_bf16_x2): Use
11772 __builtin_memcpy instead of constructing an additional
11773 __builtin_aarch64_simd_oi one vector at a time.
11774 (vst1q_bf16_x2): Likewise.
11775 (vst1_bf16_x3): Use __builtin_memcpy instead of constructing
11776 an additional __builtin_aarch64_simd_ci one vector at a time.
11777 (vst1q_bf16_x3): Likewise.
11778 (vst1_bf16_x4): Use __builtin_memcpy instead of a union.
11779 (vst1q_bf16_x4): Likewise.
11780 (vst2_bf16): Use __builtin_memcpy instead of constructing an
11781 additional __builtin_aarch64_simd_oi one vector at a time.
11782 (vst2q_bf16): Likewise.
11783 (vst3_bf16): Use __builtin_memcpy instead of constructing an
11784 additional __builtin_aarch64_simd_ci mode one vector at a
11786 (vst3q_bf16): Likewise.
11787 (vst4_bf16): Use __builtin_memcpy instead of constructing an
11788 additional __builtin_aarch64_simd_xi one vector at a time.
11789 (vst4q_bf16): Likewise.
11791 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
11793 * config/aarch64/arm_neon.h (__ST2_LANE_FUNC): Delete.
11794 (__ST2Q_LANE_FUNC): Delete.
11795 (vst2_lane_f16): Use __builtin_memcpy to copy vector
11796 structure instead of constructing __builtin_aarch64_simd_oi
11797 one vector at a time.
11798 (vst2_lane_f32): Likewise.
11799 (vst2_lane_f64): Likewise.
11800 (vst2_lane_p8): Likewise.
11801 (vst2_lane_p16): Likewise.
11802 (vst2_lane_p64): Likewise.
11803 (vst2_lane_s8): Likewise.
11804 (vst2_lane_s16): Likewise.
11805 (vst2_lane_s32): Likewise.
11806 (vst2_lane_s64): Likewise.
11807 (vst2_lane_u8): Likewise.
11808 (vst2_lane_u16): Likewise.
11809 (vst2_lane_u32): Likewise.
11810 (vst2_lane_u64): Likewise.
11811 (vst2_lane_bf16): Likewise.
11812 (vst2q_lane_f16): Use __builtin_memcpy to copy vector
11813 structure instead of using a union.
11814 (vst2q_lane_f32): Likewise.
11815 (vst2q_lane_f64): Likewise.
11816 (vst2q_lane_p8): Likewise.
11817 (vst2q_lane_p16): Likewise.
11818 (vst2q_lane_p64): Likewise.
11819 (vst2q_lane_s8): Likewise.
11820 (vst2q_lane_s16): Likewise.
11821 (vst2q_lane_s32): Likewise.
11822 (vst2q_lane_s64): Likewise.
11823 (vst2q_lane_u8): Likewise.
11824 (vst2q_lane_u16): Likewise.
11825 (vst2q_lane_u32): Likewise.
11826 (vst2q_lane_u64): Likewise.
11827 (vst2q_lane_bf16): Likewise.
11829 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
11831 * config/aarch64/arm_neon.h (__ST3_LANE_FUNC): Delete.
11832 (__ST3Q_LANE_FUNC): Delete.
11833 (vst3_lane_f16): Use __builtin_memcpy to copy vector
11834 structure instead of constructing __builtin_aarch64_simd_ci
11835 one vector at a time.
11836 (vst3_lane_f32): Likewise.
11837 (vst3_lane_f64): Likewise.
11838 (vst3_lane_p8): Likewise.
11839 (vst3_lane_p16): Likewise.
11840 (vst3_lane_p64): Likewise.
11841 (vst3_lane_s8): Likewise.
11842 (vst3_lane_s16): Likewise.
11843 (vst3_lane_s32): Likewise.
11844 (vst3_lane_s64): Likewise.
11845 (vst3_lane_u8): Likewise.
11846 (vst3_lane_u16): Likewise.
11847 (vst3_lane_u32): Likewise.
11848 (vst3_lane_u64): Likewise.
11849 (vst3_lane_bf16): Likewise.
11850 (vst3q_lane_f16): Use __builtin_memcpy to copy vector
11851 structure instead of using a union.
11852 (vst3q_lane_f32): Likewise.
11853 (vst3q_lane_f64): Likewise.
11854 (vst3q_lane_p8): Likewise.
11855 (vst3q_lane_p16): Likewise.
11856 (vst3q_lane_p64): Likewise.
11857 (vst3q_lane_s8): Likewise.
11858 (vst3q_lane_s16): Likewise.
11859 (vst3q_lane_s32): Likewise.
11860 (vst3q_lane_s64): Likewise.
11861 (vst3q_lane_u8): Likewise.
11862 (vst3q_lane_u16): Likewise.
11863 (vst3q_lane_u32): Likewise.
11864 (vst3q_lane_u64): Likewise.
11865 (vst3q_lane_bf16): Likewise.
11867 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
11869 * config/aarch64/arm_neon.h (__ST4_LANE_FUNC): Delete.
11870 (__ST4Q_LANE_FUNC): Delete.
11871 (vst4_lane_f16): Use __builtin_memcpy to copy vector
11872 structure instead of constructing __builtin_aarch64_simd_xi
11873 one vector at a time.
11874 (vst4_lane_f32): Likewise.
11875 (vst4_lane_f64): Likewise.
11876 (vst4_lane_p8): Likewise.
11877 (vst4_lane_p16): Likewise.
11878 (vst4_lane_p64): Likewise.
11879 (vst4_lane_s8): Likewise.
11880 (vst4_lane_s16): Likewise.
11881 (vst4_lane_s32): Likewise.
11882 (vst4_lane_s64): Likewise.
11883 (vst4_lane_u8): Likewise.
11884 (vst4_lane_u16): Likewise.
11885 (vst4_lane_u32): Likewise.
11886 (vst4_lane_u64): Likewise.
11887 (vst4_lane_bf16): Likewise.
11888 (vst4q_lane_f16): Use __builtin_memcpy to copy vector
11889 structure instead of using a union.
11890 (vst4q_lane_f32): Likewise.
11891 (vst4q_lane_f64): Likewise.
11892 (vst4q_lane_p8): Likewise.
11893 (vst4q_lane_p16): Likewise.
11894 (vst4q_lane_p64): Likewise.
11895 (vst4q_lane_s8): Likewise.
11896 (vst4q_lane_s16): Likewise.
11897 (vst4q_lane_s32): Likewise.
11898 (vst4q_lane_s64): Likewise.
11899 (vst4q_lane_u8): Likewise.
11900 (vst4q_lane_u16): Likewise.
11901 (vst4q_lane_u32): Likewise.
11902 (vst4q_lane_u64): Likewise.
11903 (vst4q_lane_bf16): Likewise.
11905 2021-08-06 Martin Liska <mliska@suse.cz>
11907 * config/rs6000/rs6000.c (rs6000_option_override_internal): When
11908 a target option is restored, it can have
11909 rs6000_long_double_type_size set to FLOAT_PRECISION_TFmode
11910 and error should not be emitted.
11912 2021-08-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
11914 * gcov-io.h (gcov_write): Declare.
11915 * gcov-io.c (gcov_write): New.
11916 (gcov_write_counter): Remove.
11917 (gcov_write_tag_length): Likewise.
11918 (gcov_write_summary): Replace gcov_write_tag_length() with calls to
11919 gcov_write_unsigned().
11920 * doc/invoke.texi (fprofile-info-section): Mention
11921 __gcov_info_to_gdca().
11923 2021-08-06 Martin Sebor <msebor@redhat.com>
11925 * dominance.c (prune_bbs_to_update_dominators): Adjust by-value vec
11926 arguments to by-reference.
11927 (iterate_fix_dominators): Same.
11928 * dominance.h (iterate_fix_dominators): Same.
11929 * ipa-prop.h: Call auto_vec::to_vec_legacy.
11930 * tree-data-ref.c (dump_data_dependence_relation): Adjust by-value vec
11931 arguments to by-reference.
11932 (debug_data_dependence_relation): Same.
11933 (dump_data_dependence_relations): Same.
11934 * tree-data-ref.h (debug_data_dependence_relation): Same.
11935 (dump_data_dependence_relations): Same.
11936 * tree-predcom.c (dump_chains): Same.
11937 (initialize_root_vars_lm): Same.
11938 (determine_unroll_factor): Same.
11939 (replace_phis_by_defined_names): Same.
11940 (insert_init_seqs): Same.
11941 (pcom_worker::tree_predictive_commoning_loop): Call
11942 auto_vec::to_vec_legacy.
11943 * tree-ssa-pre.c (insert_into_preds_of_block): Adjust by-value vec
11944 arguments to by-reference.
11945 * tree-ssa-threadbackward.c (populate_worklist): Same.
11946 (back_threader::resolve_def): Same.
11947 * tree-vect-data-refs.c (vect_check_nonzero_value): Same.
11948 (vect_enhance_data_refs_alignment): Same.
11949 (vect_check_lower_bound): Same.
11950 (vect_prune_runtime_alias_test_list): Same.
11951 (vect_permute_store_chain): Same.
11952 * tree-vect-slp-patterns.c (vect_normalize_conj_loc): Same.
11953 * tree-vect-stmts.c (vect_create_vectorized_demotion_stmts): Same.
11954 * tree-vectorizer.h (vect_permute_store_chain): Same.
11955 * vec.c (test_init): New function.
11956 (vec_c_tests): Call new function.
11957 * vec.h (vec): Declare ctors, dtor, and assignment.
11958 (auto_vec::vec_to_legacy): New function.
11959 (vec::copy): Adjust initialization.
11961 2021-08-05 H.J. Lu <hjl.tools@gmail.com>
11964 * config/i386/i386.c (ix86_can_inline_p): Ignore MASK_80387 if
11965 callee only uses GPRs.
11966 * config/i386/ia32intrin.h: Revert commit 5463cee2770.
11967 * config/i386/serializeintrin.h: Revert commit 71958f740f1.
11968 * config/i386/x86gprintrin.h: Add
11969 #pragma GCC target("general-regs-only") and #pragma GCC pop_options
11970 to disable non-GPR ISAs.
11972 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
11974 PR middle-end/101787
11975 * doc/md.texi (cond_ashl, cond_ashr, cond_lshr): Document.
11977 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
11979 * tree-vectorizer.h (vect_is_store_elt_extraction, vect_is_reduction)
11980 (vect_reduc_type, vect_embedded_comparison_type, vect_comparison_type)
11981 (vect_is_extending_load, vect_is_integer_truncation): New functions,
11982 moved from aarch64.c but given different names.
11983 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction)
11984 (aarch64_is_reduction, aarch64_reduc_type)
11985 (aarch64_embedded_comparison_type, aarch64_comparison_type)
11986 (aarch64_extending_load_p, aarch64_integer_truncation_p): Delete
11987 in favor of the above. Update callers accordingly.
11989 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
11992 * config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
11993 writing .cpu directive in asm output.
11994 * config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
11995 (arm_last_printed_arch_string): Delete.
11996 (arm_last-printed_fpu_string): Delete.
11997 (arm_configure_build_target): If use of floating-point/SIMD is
11998 disabled, remove all fp/simd related features from the target ISA.
11999 (last_arm_targ_options): New variable.
12000 (arm_print_asm_arch_directives): Add new parameters. Change order
12001 of emitted directives and handle all cases here.
12002 (arm_file_start): Always call arm_print_asm_arch_directives, move
12003 all generation of .arch/.arch_extension here.
12004 (arm_file_end): Call arm_print_asm_arch.
12005 (arm_declare_function_name): Call arm_print_asm_arch_directives
12006 instead of printing .arch/.fpu directives directly.
12008 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
12010 * config/arm/arm.c (arm_configure_build_target): Don't call
12011 arm_option_reconfigure_globals.
12012 (arm_option_restore): Call arm_option_reconfigure_globals after
12013 reconfiguring the target.
12014 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
12016 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
12018 * config/arm/arm.c (arm_configure_build_target): Ensure the target's
12019 arch_name is always set.
12021 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
12023 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
12024 of vec_select high-half from being added into Neon subtract
12027 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
12029 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
12030 of vec_select high-half from being added into Neon add cost.
12032 2021-08-05 Kewen Lin <linkw@linux.ibm.com>
12034 * cfgloop.h (loops_list::loops_list): Add one optional argument
12035 root and adjust accordingly, update loop tree walking and factor
12037 * cfgloop.c (loops_list::walk_loop_tree): ... this. New function.
12039 2021-08-05 Eric Botcazou <ebotcazou@gcc.gnu.org>
12041 PR tree-optimization/101626
12042 * tree-sra.c (propagate_subaccesses_from_rhs): Do not set the
12043 reverse scalar storage order on a pointer or vector component.
12045 2021-08-05 liuhongt <hongtao.liu@intel.com>
12047 * config/i386/sse.md (cond_<code><mode>): New expander.
12049 2021-08-05 liuhongt <hongtao.liu@intel.com>
12051 * config/i386/sse.md (cond_<code><mode>): New expander.
12053 2021-08-05 liuhongt <hongtao.liu@intel.com>
12055 * config/i386/sse.md (cond_<code><mode>): New expander.
12057 2021-08-04 David Malcolm <dmalcolm@redhat.com>
12060 * Makefile.in (ANALYZER_OBJS): Add analyzer/region-model-asm.o.
12062 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
12065 * config/i386/i386.h (STORE_MAX_PIECES): Allow 16/32/64 bytes
12066 only if TARGET_INTER_UNIT_MOVES_TO_VEC is true.
12068 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
12071 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
12072 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
12073 data with SSE register from one memory location to another.
12075 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
12077 * config/s390/s390.c (expand_perm_with_vpdi): New function.
12078 (vectorize_vec_perm_const_1): Call expand_perm_with_vpdi.
12079 * config/s390/vector.md (*vpdi1<mode>, @vpdi1<mode>): Enable a
12080 parameterized expander.
12081 (*vpdi4<mode>, @vpdi4<mode>): Likewise.
12083 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
12085 * config/s390/s390.c (MAX_VECT_LEN): Define macro.
12086 (struct expand_vec_perm_d): Define struct.
12087 (expand_perm_with_merge): New function.
12088 (vectorize_vec_perm_const_1): New function.
12089 (s390_vectorize_vec_perm_const): New function.
12090 (TARGET_VECTORIZE_VEC_PERM_CONST): Define target macro.
12092 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
12094 * config/s390/vector.md (V_HW_64): Remove mode iterator.
12095 (*vec_load_pair<mode>): Use V_HW_2 instead of V_HW_64.
12096 * config/s390/vx-builtins.md
12097 (vec_scatter_element<V_HW_2:mode>_SI): Use V_HW_2 instead of
12100 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
12102 * config/s390/s390.md (UNSPEC_VEC_PERMI): Remove constant
12104 * config/s390/vector.md (*vpdi1<mode>, *vpdi4<mode>): New pattern
12106 * config/s390/vx-builtins.md (*vec_permi<mode>): Emit generic rtx
12107 instead of an unspec.
12109 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
12111 * config/s390/s390-modes.def: Add more vector modes to support
12112 concatenation of two vectors.
12113 * config/s390/s390-protos.h (s390_expand_merge_perm_const): Add
12115 (s390_expand_merge): Likewise.
12116 * config/s390/s390.c (s390_expand_merge_perm_const): New function.
12117 (s390_expand_merge): New function.
12118 * config/s390/s390.md (UNSPEC_VEC_MERGEH, UNSPEC_VEC_MERGEL):
12119 Remove constant definitions.
12120 * config/s390/vector.md (V_HW_2): Add mode iterators.
12121 (VI_HW_4, V_HW_4): Rename VI_HW_4 to V_HW_4.
12122 (vec_2x_nelts, vec_2x_wide): New mode attributes.
12123 (*vmrhb, *vmrlb, *vmrhh, *vmrlh, *vmrhf, *vmrlf, *vmrhg, *vmrlg):
12124 New pattern definitions.
12125 (vec_widen_umult_lo_<mode>, vec_widen_umult_hi_<mode>)
12126 (vec_widen_smult_lo_<mode>, vec_widen_smult_hi_<mode>)
12127 (vec_unpacks_lo_v4sf, vec_unpacks_hi_v4sf, vec_unpacks_lo_v2df)
12128 (vec_unpacks_hi_v2df): Adjust expanders to emit non-unspec RTX for
12130 * config/s390/vx-builtins.md (V_HW_4): Remove mode iterator. Now
12132 (vec_mergeh<mode>, vec_mergel<mode>): Use s390_expand_merge to
12133 emit vec merge pattern.
12135 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
12137 * config/aarch64/aarch64.c (aarch64_strip_extend_vec_half):
12139 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent cost of
12140 vec_select high-half from being added into Neon multiply
12142 * rtlanal.c (vec_series_highpart_p): Define.
12143 * rtlanal.h (vec_series_highpart_p): Declare.
12145 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
12147 * config/aarch64/aarch64.c (aarch64_strip_duplicate_vec_elt):
12149 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent
12150 vec_select cost from being added into Neon multiply cost.
12152 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
12154 * tree-vect-loop.c (vect_better_loop_vinfo_p): Detect cases in
12155 which old_loop_vinfo is an epilogue loop that handles a constant
12156 number of iterations.
12158 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
12160 * tree-vect-loop.c (vect_analyze_loop): Print a dump message
12161 when a reanalyzed loop fails to be cheaper than the current
12164 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
12166 * config/aarch64/aarch64.c: Fix a typo.
12168 2021-08-04 Vincent Lefèvre <vincent-gcc@vinc17.net>
12170 PR gcov-profile/101773
12171 * gcov-io.c (gcov_close): Check return code of a fclose.
12173 2021-08-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
12176 * dwarf2out.c (dwarf2out_assembly_start): Emit a dummy
12177 .file statement when needed.
12179 2021-08-04 Richard Biener <rguenther@suse.de>
12181 * tree-vect-data-refs.c (vect_check_gather_scatter):
12182 Include widening conversions only when the result is
12183 still handed by native gather or the current offset
12184 size not already matches the data size.
12185 Also succeed analysis in case there's no native support,
12186 noted by a IFN_LAST ifn and a NULL decl.
12187 (vect_analyze_data_refs): Always consider gathers.
12188 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
12189 Test for no IFN gather rather than decl gather.
12190 * tree-vect-stmts.c (vect_model_load_cost): Pass in the
12191 gather-scatter info and cost emulated gathers accordingly.
12192 (vect_truncate_gather_scatter_offset): Properly test for
12194 (vect_use_strided_gather_scatters_p): Likewise.
12195 (get_load_store_type): Handle emulated gathers and its
12197 (vectorizable_load): Likewise. Emulate them by extracting
12198 scalar offsets, doing scalar loads and a vector construct.
12200 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
12203 * expr.c (op_by_pieces_d::op_by_pieces_d): Add a max_pieces
12204 argument to set m_max_size.
12205 (move_by_pieces_d): Pass MOVE_MAX_PIECES to op_by_pieces_d.
12206 (store_by_pieces_d): Pass STORE_MAX_PIECES to op_by_pieces_d.
12207 (compare_by_pieces_d): Pass COMPARE_MAX_PIECES to op_by_pieces_d.
12209 2021-08-04 Roger Sayle <roger@nextmovesoftware.com>
12210 Marc Glisse <marc.glisse@inria.fr>
12212 * match.pd (bit_ior, bit_xor): Canonicalize (X*C1)|(X*C2) and
12213 (X*C1)^(X*C2) as X*(C1+C2), and related variants, using
12214 tree_nonzero_bits to ensure that operands are bit-wise disjoint.
12216 2021-08-04 Richard Biener <rguenther@suse.de>
12218 * tree-ssa-forwprop.c (pass_forwprop::execute): Split
12219 out code to decompose vector loads ...
12220 (optimize_vector_load): ... here. Generalize it to
12221 handle intermediate widening and TARGET_MEM_REF loads
12222 and apply it to loads with a supported vector mode as well.
12224 2021-08-04 Richard Biener <rguenther@suse.de>
12226 PR tree-optimization/101756
12227 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Make sure
12228 the result of the reduction epilogue is compatible to the original
12231 2021-08-04 liuhongt <hongtao.liu@intel.com>
12234 * config/i386/i386.md (peephole2): Refine predicate from
12235 register_operand to general_reg_operand.
12237 2021-08-04 Aldy Hernandez <aldyh@redhat.com>
12239 * gimple-range-path.h (path_range_query::dump): Mark override.
12241 2021-08-04 Richard Biener <rguenther@suse.de>
12243 PR tree-optimization/101769
12244 * tree-tailcall.c (eliminate_tail_call): Add the created loop
12245 for the first recursion and return it via the new output parameter.
12246 (optimize_tail_call): Pass through new output param.
12247 (tree_optimize_tail_calls_1): After creating all latches,
12248 add the created loop to the loop tree. Do not mark loops for fixup.
12250 2021-08-04 Martin Liska <mliska@suse.cz>
12252 * doc/invoke.texi: Document threader-mode param.
12254 2021-08-04 liuhongt <hongtao.liu@intel.com>
12256 * config/i386/sse.md (cond_fma<mode>): New expander.
12257 (cond_fms<mode>): Ditto.
12258 (cond_fnma<mode>): Ditto.
12259 (cond_fnms<mode>): Ditto.
12261 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
12263 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode>): Use && instead of &.
12265 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
12267 * config/rs6000/constraints.md: Remove "e" from the list of available
12268 constraint characters.
12270 2021-08-03 Eugene Rozenfeld <erozen@microsoft.com>
12272 PR gcov-profile/71672
12273 * auto-profile.c (afdo_indirect_call): Fix setup of the historgram value for indirect calls.
12275 2021-08-03 Paul A. Clarke <pc@us.ibm.com>
12277 * config/rs6000/smmintrin.h (_mm_minpos_epu16): New.
12279 2021-08-03 H.J. Lu <hjl.tools@gmail.com>
12281 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): In 64-bit mode,
12282 try XMM31 to avoid vzeroupper.
12284 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
12286 * doc/invoke.texi: Document -mtune=neoverse-512tvb and
12287 -mcpu=neoverse-512tvb.
12288 * config/aarch64/aarch64-cores.def (neoverse-512tvb): New entry.
12289 * config/aarch64/aarch64-tune.md: Regenerate.
12290 * config/aarch64/aarch64.c (neoverse512tvb_sve_vector_cost)
12291 (neoverse512tvb_sve_issue_info, neoverse512tvb_vec_issue_info)
12292 (neoverse512tvb_vector_cost, neoverse512tvb_tunings): New structures.
12293 (aarch64_adjust_body_cost_sve): Handle -mtune=neoverse-512tvb.
12294 (aarch64_adjust_body_cost): Likewise.
12296 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
12298 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Only
12299 record issue information for operations that occur in the
12302 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
12304 * config/aarch64/aarch64.c (aarch64_multiply_add_p): Add a vec_flags
12305 parameter. Detect cases in which an Advanced SIMD MLA would almost
12306 certainly require a MOV.
12307 (aarch64_count_ops): Update accordingly.
12309 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
12311 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction): New
12312 function, split out from...
12313 (aarch64_detect_vector_stmt_subtype): ...here.
12314 (aarch64_add_stmt_cost): Treat extracting element 0 as free.
12316 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
12318 * config/aarch64/aarch64-protos.h (sve_vec_cost):
12319 Add gather_load_x32_cost and gather_load_x64_cost.
12320 * config/aarch64/aarch64.c (generic_sve_vector_cost)
12321 (a64fx_sve_vector_cost, neoversev1_sve_vector_cost): Update
12322 accordingly, using the values given by the scalar_load * number
12323 of elements calculation that we used previously.
12324 (aarch64_detect_vector_stmt_subtype): Use the new fields.
12326 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
12328 * config/aarch64/aarch64.c (aarch64_adjust_body_cost_sve): New
12329 function, split out from...
12330 (aarch64_adjust_body_cost): ...here.
12332 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
12334 * config/aarch64/fractional-cost.h: New file.
12335 * config/aarch64/aarch64.c: Include <algorithm> (indirectly)
12336 and cost_fraction.h.
12337 (vec_cost_fraction): New typedef.
12338 (aarch64_detect_scalar_stmt_subtype): Use it for statement costs.
12339 (aarch64_detect_vector_stmt_subtype): Likewise.
12340 (aarch64_sve_adjust_stmt_cost, aarch64_adjust_stmt_cost): Likewise.
12341 (aarch64_estimate_min_cycles_per_iter): Use vec_cost_fraction
12343 (aarch64_adjust_body_cost): Likewise.
12344 (aarch64_test_cost_fraction): New function.
12345 (aarch64_run_selftests): Call it.
12347 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
12349 * config/aarch64/aarch64-protos.h (tune_params::sve_width): Turn
12351 * config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Update
12353 (aarch64_estimated_poly_value): Likewise. Use the least significant
12354 set bit for the minimum and likely values. Use the most significant
12355 set bit for the maximum value.
12357 2021-08-03 liuhongt <hongtao.liu@intel.com>
12359 * config/i386/sse.md (cond_<insn><mode>): New expander.
12360 (cond_mul<mode>): Ditto.
12362 2021-08-03 Kewen Lin <linkw@linux.ibm.com>
12364 * tree-cfg.c (move_sese_region_to_fn): Fix typos on dloop.
12366 2021-08-03 liuhongt <hongtao.liu@intel.com>
12368 * config/i386/sse.md (cond_<insn><mode>):New expander.
12369 (cond_mul<mode>): Ditto.
12370 (cond_div<mode>): Ditto.
12372 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
12374 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Also
12375 check stack_realign_needed for stack realignment.
12376 (ix86_legitimate_constant_p): Always allow CONST_WIDE_INT smaller
12377 than the largest integer supported by vector register.
12378 * config/i386/i386.h (MAX_MOVE_MAX): New. Set to 64.
12379 (MOVE_MAX): Set to bytes of the largest integer supported by
12381 (STORE_MAX_PIECES): New.
12383 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
12385 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
12386 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
12387 data from one memory location to another.
12389 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
12391 PR middle-end/90773
12392 * config/i386/i386.c (TARGET_GEN_MEMSET_SCRATCH_RTX): New.
12394 2021-08-02 Aldy Hernandez <aldyh@redhat.com>
12396 PR tree-optimization/101724
12397 * params.opt: Remove --param=threader-iterative.
12398 * tree-ssa-threadbackward.c (pass_thread_jumps::execute): Remove
12401 2021-08-02 Tom de Vries <tdevries@suse.de>
12403 PR middle-end/101665
12404 * doc/extend.texi (nonnull attribute): Improve documentation.
12406 2021-08-02 Andrew Pinski <apinski@marvell.com>
12408 PR rtl-optimization/101683
12409 * rtlanal.c (may_trap_p_1): Handle UNSIGNED_FIX.
12411 2021-08-02 Roger Sayle <roger@nextmovesoftware.com>
12413 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern):
12414 Renamed from cond_removal_in_popcount_clz_ctz_pattern.
12415 Add support for BSWAP, FFS, PARITY and CLRSB builtins.
12416 (tree_ssa_phiop_worker): Update call to function above.
12418 2021-08-01 H.J. Lu <hjl.tools@gmail.com>
12421 * config/i386/i386.md (bsr_rex64_1_zext): New.
12422 (combine splitter for constant - clzll): Replace gen_bsr_rex64_1
12423 with gen_bsr_rex64_1_zext.
12425 2021-07-31 Jakub Jelinek <jakub@redhat.com>
12428 * config/i386/i386.md (bsr_rex64_1, bsr_1, bsr_zext_1): New
12429 define_insn patterns.
12430 (*bsr_rex64_2, *bsr_2): New define_insn_and_split patterns.
12431 Add combine splitters for constant - clz.
12432 (clz<mode>2): Use a temporary pseudo for bsr result.
12434 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
12436 * config/rs6000/smmintrin.h (_mm_floor_pd, _mm_floor_ps,
12437 _mm_floor_sd, _mm_floor_ss): New.
12439 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
12441 * config/rs6000/smmintrin.h (_mm_ceil_pd, _mm_ceil_ps,
12442 _mm_ceil_sd, _mm_ceil_ss): New.
12444 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
12446 * config/rs6000/smmintrin.h (_mm_blend_pd, _mm_blendv_pd,
12447 _mm_blend_ps, _mm_blendv_ps): New.
12449 2021-07-30 Roger Sayle <roger@nextmovesoftware.com>
12450 Uroš Bizjak <ubizjak@gmail.com>
12452 * config/i386/i386.md (*dec_cmov<mode>): New define_insn_and_split
12453 to generate a conditional move using the carry flag after sub $1.
12454 (peephole2): Eliminate a register-to-register move by inverting
12455 the condition of a conditional move.
12457 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
12459 * config/mmix/mmix.md ("call", "call_value", "*call_real")
12460 ("*call_value_real"): Don't generate rtx mentioning the generic
12461 operands 1 and 2 to "call", and similarly for "call_value".
12462 * config/mmix/mmix.c (mmix_print_operand_punct_valid_p)
12463 (mmix_print_operand): Use '!' instead of 'p'.
12465 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
12467 * doc/md.texi (call): Correct information about operand 2.
12468 * config/mmix/mmix.md ("call", "call_value"): Remove fixed FIXMEs.
12470 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
12472 * range-op.cc (operator_trunc_mod::wi_fold): Fold constants.
12474 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
12476 * range-op.cc (operator_div::wi_fold): Return UNDEFINED for [0, 0] divisor.
12478 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
12480 * gimple-range-cache.cc (*::set_bb_range): Change const basic_block to
12481 const_basic_block..
12482 (*::get_bb_range): Ditto.
12483 (*::bb_range_p): Ditto.
12484 * gimple-range-cache.h: Change prototypes.
12486 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
12488 PR middle-end/90773
12489 * builtins.c (builtin_memcpy_read_str): Change the mode argument
12490 from scalar_int_mode to fixed_size_mode.
12491 (builtin_strncpy_read_str): Likewise.
12492 (gen_memset_value_from_prev): New function.
12493 (builtin_memset_read_str): Change the mode argument from
12494 scalar_int_mode to fixed_size_mode. Use gen_memset_value_from_prev
12495 and support CONST_VECTOR.
12496 (builtin_memset_gen_str): Likewise.
12497 (try_store_by_multiple_pieces): Use by_pieces_constfn to declare
12499 * builtins.h (builtin_strncpy_read_str): Replace scalar_int_mode
12500 with fixed_size_mode.
12501 (builtin_memset_read_str): Likewise.
12502 * expr.c (widest_int_mode_for_size): Renamed to ...
12503 (widest_fixed_size_mode_for_size): Add a bool argument to
12504 indicate if QI vector mode can be used.
12505 (by_pieces_ninsns): Call widest_fixed_size_mode_for_size
12506 instead of widest_int_mode_for_size.
12507 (pieces_addr::adjust): Change the mode argument from
12508 scalar_int_mode to fixed_size_mode.
12509 (op_by_pieces_d): Make m_len read-only. Add a bool member,
12510 m_qi_vector_mode, to indicate that QI vector mode can be used.
12511 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
12512 initialize m_qi_vector_mode. Call widest_fixed_size_mode_for_size
12513 instead of widest_int_mode_for_size.
12514 (op_by_pieces_d::get_usable_mode): Change the mode argument from
12515 scalar_int_mode to fixed_size_mode. Call
12516 widest_fixed_size_mode_for_size instead of
12517 widest_int_mode_for_size.
12518 (op_by_pieces_d::smallest_fixed_size_mode_for_size): New member
12519 function to return the smallest integer or QI vector mode.
12520 (op_by_pieces_d::run): Call widest_fixed_size_mode_for_size
12521 instead of widest_int_mode_for_size. Call
12522 smallest_fixed_size_mode_for_size instead of
12523 smallest_int_mode_for_size.
12524 (store_by_pieces_d::store_by_pieces_d): Add a bool argument to
12525 indicate that QI vector mode can be used and pass it to
12526 op_by_pieces_d::op_by_pieces_d.
12527 (can_store_by_pieces): Call widest_fixed_size_mode_for_size
12528 instead of widest_int_mode_for_size. Pass memsetp to
12529 widest_fixed_size_mode_for_size to support QI vector mode.
12530 Allow all CONST_VECTORs for memset if vec_duplicate is supported.
12531 (store_by_pieces): Pass memsetp to
12532 store_by_pieces_d::store_by_pieces_d.
12533 (clear_by_pieces_1): Removed.
12534 (clear_by_pieces): Replace clear_by_pieces_1 with
12535 builtin_memset_read_str and pass true to store_by_pieces_d to
12536 support vector mode broadcast.
12537 (string_cst_read_str): Change the mode argument from
12538 scalar_int_mode to fixed_size_mode.
12539 * expr.h (by_pieces_constfn): Change scalar_int_mode to
12541 (by_pieces_prev): Likewise.
12542 * rtl.h (lowpart_subreg_regno): New.
12543 * rtlanal.c (lowpart_subreg_regno): New. A wrapper around
12544 simplify_subreg_regno.
12545 * target.def (gen_memset_scratch_rtx): New hook.
12546 * doc/tm.texi.in: Add TARGET_GEN_MEMSET_SCRATCH_RTX.
12547 * doc/tm.texi: Regenerated.
12549 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
12552 * config/mips/mips.c (mips_atomic_assign_expand_fenv): Use
12553 TARGET_EXPR instead of MODIFY_EXPR.
12555 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
12558 * config/mips/mips-protos.h (mips_expand_vec_cmp_expr): Declare.
12559 * config/mips/mips.c (mips_expand_vec_cmp_expr): New function.
12560 * config/mips/mips-msa.md (vec_cmp<MSA:mode><mode_i>): New
12562 (vec_cmpu<IMSA:mode><mode_i>): New expander.
12564 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
12567 * config/i386/i386-options.c (ix86_option_override_internal):
12568 Don't enable LZCNT/POPCNT if they have been disabled explicitly.
12570 2021-07-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
12573 * config/arm/arm_neon.h (vld1_p64): Replace call to builtin by
12574 explicitly dereferencing __a.
12575 (vld1_s64): Likewise.
12576 (vld1_u64): Likewise.
12577 * config/arm/arm_neon_builtins.def (vld1): Remove entry for di
12578 and change to VAR13.
12580 2021-07-30 Aldy Hernandez <aldyh@redhat.com>
12582 * gimple-loop-versioning.cc (lv_dom_walker::lv_dom_walker): Remove
12583 use of m_range_analyzer.
12584 (loop_versioning::lv_dom_walker::before_dom_children): Same.
12585 (loop_versioning::lv_dom_walker::after_dom_children): Remove.
12586 (loop_versioning::prune_loop_conditions): Replace vr_values use
12587 with range_query interface.
12588 (pass_loop_versioning::execute): Use ranger.
12590 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
12593 * ipa-devirt.c (ipa_odr_read_section): Compare the precision of
12594 enum values, and emit a warning if they mismatch.
12596 2021-07-30 Kewen Lin <linkw@linux.ibm.com>
12598 * cfgloop.h (as_const): New function.
12599 (class loop_iterator): Rename to ...
12600 (class loops_list): ... this.
12601 (loop_iterator::next): Rename to ...
12602 (loops_list::Iter::fill_curr_loop): ... this and adjust.
12603 (loop_iterator::loop_iterator): Rename to ...
12604 (loops_list::loops_list): ... this and adjust.
12605 (loops_list::Iter): New class.
12606 (loops_list::iterator): New type.
12607 (loops_list::const_iterator): New type.
12608 (loops_list::begin): New function.
12609 (loops_list::end): Likewise.
12610 (loops_list::begin const): Likewise.
12611 (loops_list::end const): Likewise.
12612 (FOR_EACH_LOOP): Remove.
12613 (FOR_EACH_LOOP_FN): Remove.
12614 * cfgloop.c (flow_loops_dump): Adjust FOR_EACH_LOOP* with range-based
12615 for loop with loops_list instance.
12616 (sort_sibling_loops): Likewise.
12617 (disambiguate_loops_with_multiple_latches): Likewise.
12618 (verify_loop_structure): Likewise.
12619 * cfgloopmanip.c (create_preheaders): Likewise.
12620 (force_single_succ_latches): Likewise.
12621 * config/aarch64/falkor-tag-collision-avoidance.c
12622 (execute_tag_collision_avoidance): Likewise.
12623 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Likewise.
12624 * config/s390/s390.c (s390_adjust_loops): Likewise.
12625 * doc/loop.texi: Likewise.
12626 * gimple-loop-interchange.cc (pass_linterchange::execute): Likewise.
12627 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
12628 * gimple-loop-versioning.cc (loop_versioning::analyze_blocks): Likewise.
12629 (loop_versioning::make_versioning_decisions): Likewise.
12630 * gimple-ssa-split-paths.c (split_paths): Likewise.
12631 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
12632 * graphite.c (canonicalize_loop_form): Likewise.
12633 (graphite_transform_loops): Likewise.
12634 * ipa-fnsummary.c (analyze_function_body): Likewise.
12635 * ipa-pure-const.c (analyze_function): Likewise.
12636 * loop-doloop.c (doloop_optimize_loops): Likewise.
12637 * loop-init.c (loop_optimizer_finalize): Likewise.
12638 (fix_loop_structure): Likewise.
12639 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
12640 (move_loop_invariants): Likewise.
12641 * loop-unroll.c (decide_unrolling): Likewise.
12642 (unroll_loops): Likewise.
12643 * modulo-sched.c (sms_schedule): Likewise.
12644 * predict.c (predict_loops): Likewise.
12645 (pass_profile::execute): Likewise.
12646 * profile.c (branch_prob): Likewise.
12647 * sel-sched-ir.c (sel_finish_pipelining): Likewise.
12648 (sel_find_rgns): Likewise.
12649 * tree-cfg.c (replace_loop_annotate): Likewise.
12650 (replace_uses_by): Likewise.
12651 (move_sese_region_to_fn): Likewise.
12652 * tree-if-conv.c (pass_if_conversion::execute): Likewise.
12653 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
12654 * tree-parloops.c (parallelize_loops): Likewise.
12655 * tree-predcom.c (tree_predictive_commoning): Likewise.
12656 * tree-scalar-evolution.c (scev_initialize): Likewise.
12657 (scev_reset): Likewise.
12658 * tree-ssa-dce.c (find_obviously_necessary_stmts): Likewise.
12659 * tree-ssa-live.c (remove_unused_locals): Likewise.
12660 * tree-ssa-loop-ch.c (ch_base::copy_headers): Likewise.
12661 * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
12662 (tree_ssa_lim_initialize): Likewise.
12663 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Likewise.
12664 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize): Likewise.
12665 * tree-ssa-loop-manip.c (get_loops_exits): Likewise.
12666 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations): Likewise.
12667 (free_numbers_of_iterations_estimates): Likewise.
12668 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Likewise.
12669 * tree-ssa-loop-split.c (tree_ssa_split_loops): Likewise.
12670 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Likewise.
12671 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
12672 (pass_scev_cprop::execute): Likewise.
12673 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
12674 * tree-ssa-sccvn.c (do_rpo_vn): Likewise.
12675 * tree-ssa-threadupdate.c
12676 (jump_thread_path_registry::thread_through_all_blocks): Likewise.
12677 * tree-vectorizer.c (vectorize_loops): Likewise.
12678 * tree-vrp.c (vrp_asserts::find_assert_locations): Likewise.
12680 2021-07-29 Hans-Peter Nilsson <hp@bitrange.com>
12682 * config/mmix/mmix.c (mmix_function_arg_1): Avoid
12683 generating a VOIDmode register for e.g the
12684 function_arg_info::end_marker.
12686 2021-07-29 Jeff Law <jeffreyalaw@gmail.com>
12688 * config/h8300/h8300-modes.def: Add CCZ, CCV and CCC, drop CCZNV.
12689 * config/h8300/h8300.md (H8cc mode iterator): Add CCZ.
12690 (cc mode_attr): Similarly.
12691 (ccz subst_attr): Similarly.
12692 * config/h8300/jumpcall.md: Add new patterns for branch-on-bit.
12693 * config/h8300/testcompare.md: Remove various cc0 based patterns
12694 that had been commented out. Add pattern to set CCZ from a bit
12697 2021-07-29 Thomas Schwinge <thomas@codesourcery.com>
12698 Julian Brown <julian@codesourcery.com>
12699 Kwok Cheung Yeung <kcy@codesourcery.com>
12701 * omp-offload.c (oacc_loop_xform_head_tail, oacc_loop_process):
12702 'update_stmt' after modification.
12703 (pass_oacc_loop_designation): New function, extracted out of...
12704 (pass_oacc_device_lower): ... this.
12705 (pass_data_oacc_loop_designation, pass_oacc_loop_designation)
12706 (make_pass_oacc_loop_designation): New
12707 * passes.def: Add it.
12708 * tree-parloops.c (create_parallel_loop): Adjust.
12709 * tree-pass.h (make_pass_oacc_loop_designation): New.
12711 2021-07-29 Aldy Hernandez <aldyh@redhat.com>
12713 * flag-types.h (enum threader_mode): New.
12714 * params.opt: Add entry for --param=threader-mode.
12715 * tree-ssa-threadbackward.c (THREADER_ITERATIVE_MODE): New.
12716 (class back_threader): New.
12717 (back_threader::back_threader): New.
12718 (back_threader::~back_threader): New.
12719 (back_threader::maybe_register_path): New.
12720 (back_threader::find_taken_edge): New.
12721 (back_threader::find_taken_edge_switch): New.
12722 (back_threader::find_taken_edge_cond): New.
12723 (back_threader::resolve_def): New.
12724 (back_threader::resolve_phi): New.
12725 (back_threader::find_paths_to_names): New.
12726 (back_threader::find_paths): New.
12729 (thread_jumps::find_jump_threads_backwards): Call ranger threader.
12730 (thread_jumps::find_jump_threads_backwards_with_ranger): New.
12731 (pass_thread_jumps::execute): Abstract out code...
12732 (try_thread_blocks): ...here.
12733 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
12734 Abstract out threading candidate code to...
12735 (single_succ_to_potentially_threadable_block): ...here.
12736 * tree-ssa-threadedge.h (single_succ_to_potentially_threadable_block):
12738 * tree-ssa-threadupdate.c (register_jump_thread): Return boolean.
12739 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
12740 Return bool from register_jump_thread.
12742 2021-07-29 Andreas Krebbel <krebbel@linux.ibm.com>
12744 * target.def: in0 and in1 do not need to be registers.
12745 * doc/tm.texi: Regenerate.
12747 2021-07-29 liuhongt <hongtao.liu@intel.com>
12750 * config/i386/i386.c (ix86_widen_mult_cost): New function.
12751 (ix86_add_stmt_cost): Use ix86_widen_mult_cost for
12754 2021-07-29 Jiufu Guo <guojiufu@linux.ibm.com>
12757 * config/rs6000/rs6000.c (TARGET_PREFERRED_DOLOOP_MODE): New hook.
12758 (rs6000_preferred_doloop_mode): New hook.
12759 * doc/tm.texi: Regenerate.
12760 * doc/tm.texi.in: Add hook preferred_doloop_mode.
12761 * target.def (preferred_doloop_mode): New hook.
12762 * targhooks.c (default_preferred_doloop_mode): New hook.
12763 * targhooks.h (default_preferred_doloop_mode): New hook.
12764 * tree-ssa-loop-ivopts.c (compute_doloop_base_on_mode): New function.
12765 (add_iv_candidate_for_doloop): Call targetm.preferred_doloop_mode
12766 and compute_doloop_base_on_mode.
12768 2021-07-28 Martin Sebor <msebor@redhat.com>
12770 PR middle-end/101494
12771 * tree-ssa-uninit.c (maybe_warn_operand): Correct object offset
12772 and size computation.
12774 2021-07-28 Martin Sebor <msebor@redhat.com>
12776 PR middle-end/101601
12777 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Remove
12779 Handle pointers to functions.
12781 2021-07-28 Martin Sebor <msebor@redhat.com>
12783 * Makefile.in (OBJS): Add gimple-ssa-warn-access.o and pointer-query.o.
12784 * attribs.h (fndecl_dealloc_argno): Move fndecl_dealloc_argno to tree.h.
12785 * builtins.c (compute_objsize_r): Move to pointer-query.cc.
12786 (access_ref::access_ref): Same.
12787 (access_ref::phi): Same.
12788 (access_ref::get_ref): Same.
12789 (access_ref::size_remaining): Same.
12790 (access_ref::offset_in_range): Same.
12791 (access_ref::add_offset): Same.
12792 (access_ref::inform_access): Same.
12793 (ssa_name_limit_t::visit_phi): Same.
12794 (ssa_name_limit_t::leave_phi): Same.
12795 (ssa_name_limit_t::next): Same.
12796 (ssa_name_limit_t::next_phi): Same.
12797 (ssa_name_limit_t::~ssa_name_limit_t): Same.
12798 (pointer_query::pointer_query): Same.
12799 (pointer_query::get_ref): Same.
12800 (pointer_query::put_ref): Same.
12801 (pointer_query::flush_cache): Same.
12802 (warn_string_no_nul): Move to gimple-ssa-warn-access.cc.
12803 (check_nul_terminated_array): Same.
12804 (unterminated_array): Same.
12805 (maybe_warn_for_bound): Same.
12806 (check_read_access): Same.
12807 (warn_for_access): Same.
12808 (get_size_range): Same.
12809 (check_access): Same.
12810 (gimple_call_alloc_size): Move to tree.c.
12811 (gimple_parm_array_size): Move to pointer-query.cc.
12812 (get_offset_range): Same.
12813 (gimple_call_return_array): Same.
12814 (handle_min_max_size): Same.
12815 (handle_array_ref): Same.
12816 (handle_mem_ref): Same.
12817 (compute_objsize): Same.
12818 (gimple_call_alloc_p): Move to gimple-ssa-warn-access.cc.
12819 (call_dealloc_argno): Same.
12820 (fndecl_dealloc_argno): Same.
12821 (new_delete_mismatch_p): Same.
12822 (matching_alloc_calls_p): Same.
12823 (warn_dealloc_offset): Same.
12824 (maybe_emit_free_warning): Same.
12825 * builtins.h (check_nul_terminated_array): Move to
12826 gimple-ssa-warn-access.h.
12827 (check_nul_terminated_array): Same.
12828 (warn_string_no_nul): Same.
12829 (unterminated_array): Same.
12830 (class ssa_name_limit_t): Same.
12831 (class pointer_query): Same.
12832 (struct access_ref): Same.
12833 (class range_query): Same.
12834 (struct access_data): Same.
12835 (gimple_call_alloc_size): Same.
12836 (gimple_parm_array_size): Same.
12837 (compute_objsize): Same.
12838 (class access_data): Same.
12839 (maybe_emit_free_warning): Same.
12840 * calls.c (initialize_argument_information): Remove call to
12841 maybe_emit_free_warning.
12842 * gimple-array-bounds.cc: Include new header..
12843 * gimple-fold.c: Same.
12844 * gimple-ssa-sprintf.c: Same.
12845 * gimple-ssa-warn-restrict.c: Same.
12846 * passes.def: Add pass_warn_access.
12847 * tree-pass.h (make_pass_warn_access): Declare.
12848 * tree-ssa-strlen.c: Include new headers.
12849 * tree.c (fndecl_dealloc_argno): Move here from builtins.c.
12850 * tree.h (fndecl_dealloc_argno): Move here from attribs.h.
12851 * gimple-ssa-warn-access.cc: New file.
12852 * gimple-ssa-warn-access.h: New file.
12853 * pointer-query.cc: New file.
12854 * pointer-query.h: New file.
12856 2021-07-28 Jakub Jelinek <jakub@redhat.com>
12858 PR middle-end/101624
12859 * ubsan.c (maybe_instrument_pointer_overflow,
12860 instrument_object_size): Only test DECL_REGISTER on VAR_DECLs,
12861 PARM_DECLs or RESULT_DECLs.
12862 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
12864 2021-07-28 Jakub Jelinek <jakub@redhat.com>
12866 PR middle-end/101642
12867 * match.pd (bswap16 (x) == bswap16 (y)): Cast both operands
12868 to type of bswap16 for comparison.
12869 (bswap16 (x) == cst): Cast bswap16 operand to type of cst.
12871 2021-07-28 Richard Biener <rguenther@suse.de>
12873 PR tree-optimization/101615
12874 * tree-vect-slp.c (vect_optimize_slp): Materialize permutes
12875 at CTOR SLP graph entries.
12877 2021-07-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12879 * config/aarch64/aarch64.md (*extend<SHORT:mode><GPI:mode>2_aarch64):
12880 Add "r,w" alternative.
12882 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
12885 * config/i386/i386.c (ix86_avx_u128_mode_needed): Don't set
12886 AVX_U128_DIRTY when all bits are zero.
12888 2021-07-28 Richard Biener <rguenther@suse.de>
12890 PR tree-optimization/101615
12891 * tree-vect-slp.c (vect_optimize_slp): Pre-existing vector
12892 external nodes cannot be permuted so make them perm_out 0.
12894 2021-07-28 Andrew Stubbs <ams@codesourcery.com>
12897 * config.in: Regenerate.
12898 * config/gcn/gcn-hsa.h (A_FIJI): New define.
12899 (A_900): New define.
12900 (A_906): New define.
12901 (A_908): New define.
12902 (ASM_SPEC): Use A_FIJI, A_900, A_906 and A_908.
12903 * config/gcn/gcn.c (output_file_start): Adjust attributes according
12904 to the assembler capabilities.
12905 * config/gcn/mkoffload.c (main): Likewise.
12906 * configure: Regenerate.
12907 * configure.ac: Add tests for LLVM assembler attribute features.
12909 2021-07-28 Andrew MacLeod <amacleod@redhat.com>
12911 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Check for
12912 cond_false and cond_true on branches.
12914 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
12916 * config/aarch64/aarch64.c (aarch64_gen_adjusted_ldpstp): use
12919 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
12921 * alias.c (init_alias_analysis): Don't skip prologue/epilogue.
12923 2021-07-28 Jakub Jelinek <jakub@redhat.com>
12926 * config/i386/sse.md (vashr<mode>3): Split into vashrv8di3 expander
12927 and vashrv4di3 expander, where the latter requires just TARGET_AVX2
12928 and has special !TARGET_AVX512VL expansion.
12929 (vashrv2di3<mask_name>): Rename to ...
12930 (vashrv2di3): ... this. Change condition to TARGET_XOP || TARGET_AVX2
12931 and add special !TARGET_XOP && !TARGET_AVX512VL expansion.
12933 2021-07-28 Martin Uecker <muecker@gwdg.de>
12935 * calls.c (maybe_warn_rdwr_sizes): Correct argument
12936 numbers in warning that were switched.
12938 2021-07-28 Kewen Lin <linkw@linux.ibm.com>
12940 PR tree-optimization/101596
12941 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Fix wrong check
12942 by using new_type's precision instead.
12944 2021-07-28 liuhongt <hongtao.liu@intel.com>
12947 * config/i386/i386.h (processor_costs): Add new member
12949 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
12950 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
12951 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
12952 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
12953 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
12954 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
12955 generic_cost, core_cost): Initialize integer_to_sse same value
12957 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
12958 * config/i386/i386.c (ix86_builtin_vectorization_cost):
12959 Use integer_to_sse instead of sse_op to calculate the cost of
12962 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
12964 * config/rs6000/rs6000-gen-builtins.c (write_ovld_static_init): New
12966 (write_init_file): Call write_ovld_static_init.
12968 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
12970 * config/rs6000/rs6000-gen-builtins.c (write_bif_static_init): New
12972 (write_init_file): Call write_bif_static_init.
12974 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
12976 * config/rs6000/rs6000-gen-builtins.c (typemap): New struct.
12977 (TYPE_MAP_SIZE): New macro.
12978 (type_map): New initialized variable.
12979 (typemap_cmp): New function.
12980 (write_type_node): Likewise.
12981 (write_fntype_init): Implement.
12983 2021-07-27 Martin Sebor <msebor@redhat.com>
12985 PR tree-optimization/101584
12986 * tree-ssa-uninit.c (builtin_call_nomodifying_p): New function.
12987 (check_defs): Call it.
12989 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
12991 * tree-ssa-dom.c (dom_jump_threader_simplifier):
12992 Put avail_exprs_stack in the class, instead of passing it to
12993 jump_threader_simplifier.
12994 (dom_jump_threader_simplifier::simplify): Add state argument.
12995 (dom_opt_dom_walker): Add state.
12996 (pass_dominator::execute): Pass state to threader.
12997 (dom_opt_dom_walker::before_dom_children): Use state.
12998 * tree-ssa-threadedge.c (jump_threader::jump_threader): Replace
12999 arguments by state.
13000 (jump_threader::record_temporary_equivalences_from_phis):
13001 Register equivalences through the state variable.
13002 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
13003 Record ranges in a statement through the state variable.
13004 (jump_threader::simplify_control_stmt_condition): Pass state to
13006 (jump_threader::simplify_control_stmt_condition_1): Same.
13007 (jump_threader::thread_around_empty_blocks): Remove obsolete
13009 (jump_threader::thread_through_normal_block): Record equivalences
13010 on edge through the state variable.
13011 (jump_threader::thread_across_edge): Abstract state pushing.
13012 (jt_state::jt_state): New.
13013 (jt_state::push): New.
13014 (jt_state::pop): New.
13015 (jt_state::register_equiv): New.
13016 (jt_state::record_ranges_from_stmt): New.
13017 (jt_state::register_equivs_on_edge): New.
13018 (jump_threader_simplifier::jump_threader_simplifier): Move from
13020 (jump_threader_simplifier::simplify): Add state argument.
13021 * tree-ssa-threadedge.h (class jt_state): New.
13022 (class jump_threader): Add state to constructor.
13023 (class jump_threader_simplifier): Add state to simplify. Remove
13024 avail_exprs_stack from class.
13025 * tree-vrp.c (vrp_jump_threader_simplifier::simplify): Add state
13027 (vrp_jump_threader::vrp_jump_threader): Add state.
13028 (vrp_jump_threader::~vrp_jump_threader): Cleanup state.
13030 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
13032 * Makefile.in (OBJS): Add gimple-range-path.o.
13033 * gimple-range-path.cc: New file.
13034 * gimple-range-path.h: New file.
13036 2021-07-27 Jonathan Wright <jonathan.wright@arm.com>
13038 * config/aarch64/aarch64-simd.md: Push sign/zero-extension
13039 inside vec_duplicate for all patterns.
13040 * simplify-rtx.c (simplify_context::simplify_unary_operation_1):
13041 Push sign/zero-extension inside vec_duplicate.
13043 2021-07-27 Richard Biener <rguenther@suse.de>
13045 PR tree-optimization/101573
13046 * tree-ssa-uninit.c (warn_uninit_phi_uses): New function
13047 looking at uninitialized PHI arg defs in some constrained cases.
13048 (warn_uninitialized_vars): Call it.
13049 (execute_early_warn_uninitialized): Calculate dominators.
13051 2021-07-27 Richard Biener <rguenther@suse.de>
13053 PR tree-optimization/39821
13054 * tree-vect-stmts.c (vect_model_promotion_demotion_cost): Use
13055 vector_stmt for widening arithmetic.
13056 (vectorizable_conversion): Adjust.
13058 2021-07-27 Martin Jambor <mjambor@suse.cz>
13060 * cgraph.h (ipa_replace_map): New field force_load_ref.
13061 * ipa-prop.h (ipa_param_descriptor): Reduce precision of move_cost,
13062 aded new flag load_dereferenced, adjusted comments.
13063 (ipa_get_param_dereferenced): New function.
13064 (ipa_set_param_dereferenced): Likewise.
13065 * cgraphclones.c (cgraph_node::create_virtual_clone): Follow it.
13066 * ipa-cp.c: Include gimple.h.
13067 (ipcp_discover_new_direct_edges): Take into account dereferenced flag.
13068 (get_replacement_map): New parameter force_load_ref, set the
13069 appropriate flag in ipa_replace_map if set.
13070 (struct symbol_and_index_together): New type.
13071 (adjust_refs_in_act_callers): New function.
13072 (adjust_references_in_caller): Likewise.
13073 (create_specialized_node): When appropriate, call
13074 adjust_references_in_caller and force only load references.
13075 * ipa-prop.c (load_from_dereferenced_name): New function.
13076 (ipa_analyze_controlled_uses): Also detect loads from a
13077 dereference, harden testing of call statements.
13078 (ipa_write_node_info): Stream the dereferenced flag.
13079 (ipa_read_node_info): Likewise.
13080 (ipa_set_jf_constant): Also create refdesc when jump function
13081 references a variable.
13082 (cgraph_node_for_jfunc): Rename to symtab_node_for_jfunc, work
13083 also on references of variables and return a symtab_node. Adjust
13085 (propagate_controlled_uses): Also remove references to VAR_DECLs.
13087 2021-07-27 Jakub Jelinek <jakub@redhat.com>
13089 PR middle-end/101586
13090 * gimple-fold.c (clear_padding_type): Ignore FIELD_DECLs with byte
13091 positions above or equal to sz except for diagnostics of flexible
13094 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
13096 PR tree-optimization/78888
13097 * gimple-range-fold.cc (get_letter_range): New.
13098 (fold_using_range::range_of_builtin_call): Call get_letter_range.
13100 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
13102 PR tree-optimization/78888
13103 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Add cases
13104 for CFN_BUILT_IN_TOUPPER and CFN_BUILT_IN_TOLOWER.
13106 2021-07-26 Roger Sayle <roger@nextmovesoftware.com>
13107 Marc Glisse <marc.glisse@inria.fr>
13109 * match.pd (rotate): Simplify equality/inequality of rotations.
13110 (bswap): Simplify equality/inequality tests of byte swapping.
13112 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
13114 * range-op.cc (operator_bitwise_xor::op1_op2_relation_effect):
13117 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
13119 * range-op.cc (operator_lshift::fold_range): Pass rel to
13120 base class fold_range.
13121 (operator_rshift::fold_range): Same.
13123 2021-07-26 Ashimida <ashimida@linux.alibaba.com>
13126 * toplev.h (min_align_loops_log): Remove declaration.
13127 (min_align_jumps_log, min_align_labels_log): Likewise.
13128 (min_align_functions_log): Likewise.
13130 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
13132 * tree-vrp.c (vrp_simplify_cond_using_ranges): Rename vr_values
13134 (execute_vrp): Abstract out simplification of conditionals...
13135 (simplify_casted_conds): ...here.
13137 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
13139 * gimple-array-bounds.cc (array_bounds_checker::get_value_range):
13140 Add gimple argument.
13141 (array_bounds_checker::check_array_ref): Same.
13142 (array_bounds_checker::check_addr_expr): Same.
13143 (array_bounds_checker::check_array_bounds): Pass statement to
13144 check_array_bounds and check_addr_expr.
13145 * gimple-array-bounds.h (check_array_bounds): Add gimple argument.
13146 (check_addr_expr): Same.
13147 (get_value_range): Same.
13149 2021-07-26 Tamar Christina <tamar.christina@arm.com>
13151 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): Rename to..
13152 (sdot_prod, udot_prod): ... This.
13153 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): Merged
13155 (<sur>dot_prod<vsi2qi>): ... this.
13156 (aarch64_<sur>dot_lane<vsi2qi>, aarch64_<sur>dot_laneq<vsi2qi>):
13157 Change operands order.
13158 (<sur>sadv16qi): Use new operands order.
13159 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32,
13160 vdotq_s32): Use new RTL ordering.
13162 2021-07-26 Tamar Christina <tamar.christina@arm.com>
13164 * config/aarch64/aarch64-builtins.c (TYPES_TERNOP_SUSS,
13165 aarch64_types_ternop_suss_qualifiers): New.
13166 * config/aarch64/aarch64-simd-builtins.def (usdot_prod): Use it.
13167 * config/aarch64/aarch64-simd.md (usdot_prod<vsi2qi>): Re-organize RTL.
13168 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Use it.
13170 2021-07-23 Jakub Jelinek <jakub@redhat.com>
13172 PR rtl-optimization/101562
13173 * expmed.c (store_integral_bit_field): Only use movstrict_optab
13174 if the operand isn't paradoxical.
13176 2021-07-23 Aldy Hernandez <aldyh@redhat.com>
13178 * gimple-array-bounds.h (class array_bounds_checker): Change
13179 ranges type to range_query.
13181 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
13183 * config/aarch64/arm_neon.h (vst1_s64_x2): Use
13184 __builtin_memcpy instead of constructing
13185 __builtin_aarch64_simd_oi one vector at a time.
13186 (vst1_u64_x2): Likewise.
13187 (vst1_f64_x2): Likewise.
13188 (vst1_s8_x2): Likewise.
13189 (vst1_p8_x2): Likewise.
13190 (vst1_s16_x2): Likewise.
13191 (vst1_p16_x2): Likewise.
13192 (vst1_s32_x2): Likewise.
13193 (vst1_u8_x2): Likewise.
13194 (vst1_u16_x2): Likewise.
13195 (vst1_u32_x2): Likewise.
13196 (vst1_f16_x2): Likewise.
13197 (vst1_f32_x2): Likewise.
13198 (vst1_p64_x2): Likewise.
13199 (vst1q_s8_x2): Likewise.
13200 (vst1q_p8_x2): Likewise.
13201 (vst1q_s16_x2): Likewise.
13202 (vst1q_p16_x2): Likewise.
13203 (vst1q_s32_x2): Likewise.
13204 (vst1q_s64_x2): Likewise.
13205 (vst1q_u8_x2): Likewise.
13206 (vst1q_u16_x2): Likewise.
13207 (vst1q_u32_x2): Likewise.
13208 (vst1q_u64_x2): Likewise.
13209 (vst1q_f16_x2): Likewise.
13210 (vst1q_f32_x2): Likewise.
13211 (vst1q_f64_x2): Likewise.
13212 (vst1q_p64_x2): Likewise.
13214 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
13216 * config/aarch64/arm_neon.h (vst1_s64_x3): Use
13217 __builtin_memcpy instead of constructing
13218 __builtin_aarch64_simd_ci one vector at a time.
13219 (vst1_u64_x3): Likewise.
13220 (vst1_f64_x3): Likewise.
13221 (vst1_s8_x3): Likewise.
13222 (vst1_p8_x3): Likewise.
13223 (vst1_s16_x3): Likewise.
13224 (vst1_p16_x3): Likewise.
13225 (vst1_s32_x3): Likewise.
13226 (vst1_u8_x3): Likewise.
13227 (vst1_u16_x3): Likewise.
13228 (vst1_u32_x3): Likewise.
13229 (vst1_f16_x3): Likewise.
13230 (vst1_f32_x3): Likewise.
13231 (vst1_p64_x3): Likewise.
13232 (vst1q_s8_x3): Likewise.
13233 (vst1q_p8_x3): Likewise.
13234 (vst1q_s16_x3): Likewise.
13235 (vst1q_p16_x3): Likewise.
13236 (vst1q_s32_x3): Likewise.
13237 (vst1q_s64_x3): Likewise.
13238 (vst1q_u8_x3): Likewise.
13239 (vst1q_u16_x3): Likewise.
13240 (vst1q_u32_x3): Likewise.
13241 (vst1q_u64_x3): Likewise.
13242 (vst1q_f16_x3): Likewise.
13243 (vst1q_f32_x3): Likewise.
13244 (vst1q_f64_x3): Likewise.
13245 (vst1q_p64_x3): Likewise.
13247 2021-07-23 H.J. Lu <hjl.tools@gmail.com>
13250 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): Don't return
13251 hard register when LRA is in progress.
13253 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
13255 * config/aarch64/arm_neon.h (vst1_s8_x4): Use
13256 __builtin_memcpy instead of using a union.
13257 (vst1q_s8_x4): Likewise.
13258 (vst1_s16_x4): Likewise.
13259 (vst1q_s16_x4): Likewise.
13260 (vst1_s32_x4): Likewise.
13261 (vst1q_s32_x4): Likewise.
13262 (vst1_u8_x4): Likewise.
13263 (vst1q_u8_x4): Likewise.
13264 (vst1_u16_x4): Likewise.
13265 (vst1q_u16_x4): Likewise.
13266 (vst1_u32_x4): Likewise.
13267 (vst1q_u32_x4): Likewise.
13268 (vst1_f16_x4): Likewise.
13269 (vst1q_f16_x4): Likewise.
13270 (vst1_f32_x4): Likewise.
13271 (vst1q_f32_x4): Likewise.
13272 (vst1_p8_x4): Likewise.
13273 (vst1q_p8_x4): Likewise.
13274 (vst1_p16_x4): Likewise.
13275 (vst1q_p16_x4): Likewise.
13276 (vst1_s64_x4): Likewise.
13277 (vst1_u64_x4): Likewise.
13278 (vst1_p64_x4): Likewise.
13279 (vst1q_s64_x4): Likewise.
13280 (vst1q_u64_x4): Likewise.
13281 (vst1q_p64_x4): Likewise.
13282 (vst1_f64_x4): Likewise.
13283 (vst1q_f64_x4): Likewise.
13285 2021-07-23 Jonathan Wrightt <jonathan.wright@arm.com>
13287 * config/aarch64/arm_neon.h (vst2_s64): Use __builtin_memcpy
13288 instead of constructing __builtin_aarch64_simd_oi one vector
13290 (vst2_u64): Likewise.
13291 (vst2_f64): Likewise.
13292 (vst2_s8): Likewise.
13293 (vst2_p8): Likewise.
13294 (vst2_s16): Likewise.
13295 (vst2_p16): Likewise.
13296 (vst2_s32): Likewise.
13297 (vst2_u8): Likewise.
13298 (vst2_u16): Likewise.
13299 (vst2_u32): Likewise.
13300 (vst2_f16): Likewise.
13301 (vst2_f32): Likewise.
13302 (vst2_p64): Likewise.
13303 (vst2q_s8): Likewise.
13304 (vst2q_p8): Likewise.
13305 (vst2q_s16): Likewise.
13306 (vst2q_p16): Likewise.
13307 (vst2q_s32): Likewise.
13308 (vst2q_s64): Likewise.
13309 (vst2q_u8): Likewise.
13310 (vst2q_u16): Likewise.
13311 (vst2q_u32): Likewise.
13312 (vst2q_u64): Likewise.
13313 (vst2q_f16): Likewise.
13314 (vst2q_f32): Likewise.
13315 (vst2q_f64): Likewise.
13316 (vst2q_p64): Likewise.
13318 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
13320 * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy
13321 instead of constructing __builtin_aarch64_simd_ci one vector
13323 (vst3_u64): Likewise.
13324 (vst3_f64): Likewise.
13325 (vst3_s8): Likewise.
13326 (vst3_p8): Likewise.
13327 (vst3_s16): Likewise.
13328 (vst3_p16): Likewise.
13329 (vst3_s32): Likewise.
13330 (vst3_u8): Likewise.
13331 (vst3_u16): Likewise.
13332 (vst3_u32): Likewise.
13333 (vst3_f16): Likewise.
13334 (vst3_f32): Likewise.
13335 (vst3_p64): Likewise.
13336 (vst3q_s8): Likewise.
13337 (vst3q_p8): Likewise.
13338 (vst3q_s16): Likewise.
13339 (vst3q_p16): Likewise.
13340 (vst3q_s32): Likewise.
13341 (vst3q_s64): Likewise.
13342 (vst3q_u8): Likewise.
13343 (vst3q_u16): Likewise.
13344 (vst3q_u32): Likewise.
13345 (vst3q_u64): Likewise.
13346 (vst3q_f16): Likewise.
13347 (vst3q_f32): Likewise.
13348 (vst3q_f64): Likewise.
13349 (vst3q_p64): Likewise.
13351 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
13353 * config/aarch64/arm_neon.h (vst4_s64): Use __builtin_memcpy
13354 instead of constructing __builtin_aarch64_simd_xi one vector
13356 (vst4_u64): Likewise.
13357 (vst4_f64): Likewise.
13358 (vst4_s8): Likewise.
13359 (vst4_p8): Likewise.
13360 (vst4_s16): Likewise.
13361 (vst4_p16): Likewise.
13362 (vst4_s32): Likewise.
13363 (vst4_u8): Likewise.
13364 (vst4_u16): Likewise.
13365 (vst4_u32): Likewise.
13366 (vst4_f16): Likewise.
13367 (vst4_f32): Likewise.
13368 (vst4_p64): Likewise.
13369 (vst4q_s8): Likewise.
13370 (vst4q_p8): Likewise.
13371 (vst4q_s16): Likewise.
13372 (vst4q_p16): Likewise.
13373 (vst4q_s32): Likewise.
13374 (vst4q_s64): Likewise.
13375 (vst4q_u8): Likewise.
13376 (vst4q_u16): Likewise.
13377 (vst4q_u32): Likewise.
13378 (vst4q_u64): Likewise.
13379 (vst4q_f16): Likewise.
13380 (vst4q_f32): Likewise.
13381 (vst4q_f64): Likewise.
13382 (vst4q_p64): Likewise.
13384 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
13386 * config/aarch64/arm_neon.h (vtbx4_s8): Use __builtin_memcpy
13387 instead of constructing __builtin_aarch64_simd_oi one vector
13389 (vtbx4_u8): Likewise.
13390 (vtbx4_p8): Likewise.
13392 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
13394 * config/aarch64/arm_neon.h (vtbl3_s8): Use __builtin_memcpy
13395 instead of constructing __builtin_aarch64_simd_oi one vector
13397 (vtbl3_u8): Likewise.
13398 (vtbl3_p8): Likewise.
13399 (vtbl4_s8): Likewise.
13400 (vtbl4_u8): Likewise.
13401 (vtbl4_p8): Likewise.
13403 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
13405 * config/aarch64/arm_neon.h (vqtbx2_s8): Use __builtin_memcpy
13406 instead of constructing __builtin_aarch64_simd_oi one vector
13408 (vqtbx2_u8): Likewise.
13409 (vqtbx2_p8): Likewise.
13410 (vqtbx2q_s8): Likewise.
13411 (vqtbx2q_u8): Likewise.
13412 (vqtbx2q_p8): Likewise.
13413 (vqtbx3_s8): Use __builtin_memcpy instead of constructing
13414 __builtin_aarch64_simd_ci one vector at a time.
13415 (vqtbx3_u8): Likewise.
13416 (vqtbx3_p8): Likewise.
13417 (vqtbx3q_s8): Likewise.
13418 (vqtbx3q_u8): Likewise.
13419 (vqtbx3q_p8): Likewise.
13420 (vqtbx4_s8): Use __builtin_memcpy instead of constructing
13421 __builtin_aarch64_simd_xi one vector at a time.
13422 (vqtbx4_u8): Likewise.
13423 (vqtbx4_p8): Likewise.
13424 (vqtbx4q_s8): Likewise.
13425 (vqtbx4q_u8): Likewise.
13426 (vqtbx4q_p8): Likewise.
13428 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
13430 * config/aarch64/arm_neon.h (vqtbl2_s8): Use __builtin_memcpy
13431 instead of constructing __builtin_aarch64_simd_oi one vector
13433 (vqtbl2_u8): Likewise.
13434 (vqtbl2_p8): Likewise.
13435 (vqtbl2q_s8): Likewise.
13436 (vqtbl2q_u8): Likewise.
13437 (vqtbl2q_p8): Likewise.
13438 (vqtbl3_s8): Use __builtin_memcpy instead of constructing
13439 __builtin_aarch64_simd_ci one vector at a time.
13440 (vqtbl3_u8): Likewise.
13441 (vqtbl3_p8): Likewise.
13442 (vqtbl3q_s8): Likewise.
13443 (vqtbl3q_u8): Likewise.
13444 (vqtbl3q_p8): Likewise.
13445 (vqtbl4_s8): Use __builtin_memcpy instead of constructing
13446 __builtin_aarch64_simd_xi one vector at a time.
13447 (vqtbl4_u8): Likewise.
13448 (vqtbl4_p8): Likewise.
13449 (vqtbl4q_s8): Likewise.
13450 (vqtbl4q_u8): Likewise.
13451 (vqtbl4q_p8): Likewise.
13453 2021-07-23 Haochen Gui <guihaoc@gcc.gnu.org>
13456 * config/rs6000/rs6000.md (cstore<mode>4): Fix wrong fall through.
13458 2021-07-22 Andrew Pinski <apinski@marvell.com>
13460 PR tree-optimization/10153
13461 * tree-tailcall.c (create_tailcall_accumulator):
13462 Don't call fold_convert as the type should be correct already.
13463 (tree_optimize_tail_calls_1): Use build_{one,zero}_cst instead
13464 of integer_{one,zero}_node for the call of create_tailcall_accumulator.
13466 2021-07-22 Aldy Hernandez <aldyh@redhat.com>
13468 * gimple-range-cache.cc (non_null_ref::adjust_range): Replace
13469 varying_p check for null/non-null check.
13471 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
13473 PR tree-optimization/101511
13474 * value-relation.cc (relation_oracle::query_relation): Check if ssa1
13475 is in ssa2's equiv set, and don't trap if so.
13477 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
13479 PR tree-optimization/101497
13480 * gimple-range-fold.cc (fold_using_range::range_of_cond_expr): Check
13483 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
13485 PR tree-optimization/101496
13486 * vr-values.c (simplify_using_ranges::fold_cond): Call range_of_stmt
13487 first, then vrp_visit_cond_Stmt.
13489 2021-07-22 liuhongt <hongtao.liu@intel.com>
13491 * config/i386/i386-expand.c
13492 (ix86_broadcast_from_integer_constant): Rename to ..
13493 (ix86_broadcast_from_constant): .. this, and extend it to
13495 (ix86_expand_vector_move): Extend to float mode.
13496 * config/i386/i386-features.c
13497 (replace_constant_pool_with_broadcast): Remove.
13498 (remove_partial_avx_dependency_gate): Ditto.
13499 (constant_pool_broadcast): Ditto.
13500 (class pass_constant_pool_broadcast): Ditto.
13501 (make_pass_constant_pool_broadcast): Ditto.
13502 (remove_partial_avx_dependency): Adjust gate.
13503 * config/i386/i386-passes.def: Remove pass_constant_pool_broadcast.
13504 * config/i386/i386-protos.h
13505 (make_pass_constant_pool_broadcast): Remove.
13507 2021-07-22 liuhongt <hongtao.liu@intel.com>
13509 * config/i386/constraints.md (Wb): New constraint.
13511 * config/i386/i386.md (*ashlhi3_1): Extend to avx512 mask
13513 (*ashlqi3_1): Ditto.
13514 (*<insn><mode>3_1): Split to ..
13515 (*ashr<mode>3_1): this, ...
13516 (*lshr<mode>3_1): and this, also extend this pattern to avx512
13518 (*<insn><mode>3_1): Split to ..
13519 (*ashr<mode>3_1): this, ...
13520 (*lshrqi3_1): and this, also extend this pattern to avx512
13522 (*lshrhi3_1): And this, also extend this pattern to avx512
13524 * config/i386/sse.md (k<code><mode>): New define_split after
13525 it to convert generic shift pattern to mask shift ones.
13527 2021-07-21 Thomas Schwinge <thomas@codesourcery.com>
13528 Joseph Myers <joseph@codesourcery.com>
13529 Cesar Philippidis <cesar@codesourcery.com>
13531 * tree-core.h (omp_clause_code): Add 'OMP_CLAUSE_NOHOST'.
13532 * tree.c (omp_clause_num_ops, omp_clause_code_name, walk_tree_1):
13534 * tree-pretty-print.c (dump_omp_clause): Likewise.
13535 * omp-general.c (oacc_verify_routine_clauses): Likewise.
13536 * gimplify.c (gimplify_scan_omp_clauses)
13537 (gimplify_adjust_omp_clauses): Likewise.
13538 * tree-nested.c (convert_nonlocal_omp_clauses)
13539 (convert_local_omp_clauses): Likewise.
13540 * omp-low.c (scan_sharing_clauses): Likewise.
13541 * omp-offload.c (execute_oacc_device_lower): Update.
13543 2021-07-21 Martin Sebor <msebor@redhat.com>
13545 * tree-ssa-alias.c (walk_aliased_vdefs_1): Fix typos in a comment.
13547 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13549 * config/rs6000/rs6000-gen-builtins.c (write_init_bif_table):
13552 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13554 * config/rs6000/rs6000-gen-builtins.c (write_fntype): New
13556 (write_fntype_init): New stub function.
13557 (write_init_bif_table): Likewise.
13558 (write_init_ovld_table): New function.
13559 (write_init_file): Implement.
13561 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13563 * config/rs6000/rs6000-gen-builtins.c
13564 (write_autogenerated_header): New function.
13565 (write_decls): Likewise.
13566 (write_extern_fntype): New callback function.
13567 (write_header_file): Implement.
13569 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13571 * config/rs6000/rs6000-gen-builtins.c (write_defines_file):
13574 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13576 * config/rs6000/rs6000-gen-builtins.c (complete_vector_type): New
13578 (complete_base_type): Likewise.
13579 (construct_fntype_id): Likewise.
13580 (parse_bif_entry): Call contruct_fntype_id.
13581 (parse_ovld_entry): Likewise.
13583 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13585 * config/rs6000/rs6000-gen-builtins.c (ovld_stanza): New struct.
13586 (MAXOVLDSTANZAS): New macro.
13587 (ovld_stanzas): New variable.
13588 (curr_ovld_stanza): Likewise.
13589 (MAXOVLDS): New macro.
13590 (ovlddata): New struct.
13591 (ovlds): New variable.
13592 (curr_ovld): Likewise.
13593 (max_ovld_args): Likewise.
13594 (parse_ovld_entry): New function.
13595 (parse_ovld_stanza): Likewise.
13596 (parse_ovld): Implement.
13598 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13600 * config/rs6000/rs6000-gen-builtins.c (parse_bif_attrs):
13603 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13605 * config/rs6000/rs6000-gen-builtins.c (parse_args): New function.
13606 (parse_prototype): Implement.
13608 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13610 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): New enum.
13611 (curr_bif_stanza): New variable.
13612 (stanza_entry): New struct.
13613 (stanza_map): New initialized variable.
13614 (enable_string): Likewise.
13615 (fnkinds): New enum.
13616 (typelist): New struct.
13617 (attrinfo): Likewise.
13618 (MAXRESTROPNDS): New macro.
13619 (prototype): New struct.
13620 (MAXBIFS): New macro.
13621 (bifdata): New struct.
13622 (bifs): New variable.
13623 (curr_bif): Likewise.
13624 (bif_order): Likewise.
13625 (bif_index): Likewise.
13626 (fatal): New function.
13627 (stanza_name_to_stanza): Likewise.
13628 (parse_bif_attrs): New stub function.
13629 (parse_prototype): Likewise.
13630 (parse_bif_entry): New function.
13631 (parse_bif_stanza): Likewise.
13632 (parse_bif): Implement.
13633 (set_bif_order): New function.
13634 (create_bif_order): Implement.
13636 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
13638 * config/rs6000/rs6000-gen-builtins.c (rbtree.h): New #include.
13639 (num_bifs): New variable.
13640 (num_ovld_stanzas): Likewise.
13641 (num_ovlds): Likewise.
13642 (parse_codes): New enum.
13643 (bif_rbt): New variable.
13644 (ovld_rbt): Likewise.
13645 (fntype_rbt): Likewise.
13646 (bifo_rbt): Likewise.
13647 (parse_bif): New stub function.
13648 (create_bif_order): Likewise.
13649 (parse_ovld): Likewise.
13650 (write_header_file): Likewise.
13651 (write_init_file): Likewise.
13652 (write_defines_file): Likewise.
13653 (delete_output_files): New function.
13656 2021-07-21 H.J. Lu <hjl.tools@gmail.com>
13659 * config/i386/i386-builtin.def: Remove OPTION_MASK_ISA_SSE4_2
13660 from CRC32 _builtin functions.
13662 2021-07-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
13664 * coverage.c (build_gcov_info_var_registration): Mark the object placed
13665 in the linker set as referenced so that it does not get optimized away.
13667 2021-07-21 Kito Cheng <kito.cheng@sifive.com>
13670 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
13672 * config.gcc (riscv*-*-*): Detect which python is available.
13674 2021-07-21 Jakub Jelinek <jakub@redhat.com>
13676 PR middle-end/101535
13677 * gimplify.c (omp_check_private): Properly skip ORT_TARGET_DATA
13678 contexts in which decl isn't privatized and for ORT_TARGET return
13679 false if decl is mapped.
13681 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
13683 * gimple-loop-jam.c: Include tree-ssa-sccvn.h.
13684 (tree_loop_unroll_and_jam): Run value-numbering on a loop that
13685 has been successfully unrolled.
13687 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
13689 * tree-ssa-loop-manip.c (determine_exit_conditions): Return a null
13690 exit condition if no tail loop is needed, and if the original exit
13691 condition should therefore be kept as-is.
13692 (tree_transform_and_unroll_loop): Handle that case here too.
13694 2021-07-21 Kewen Lin <linkw@linux.ibm.com>
13696 * tree-data-ref.c (free_dependence_relations): Adjust to pass vec
13698 (free_data_refs): Likewise.
13699 * tree-data-ref.h (free_dependence_relations): Likewise.
13700 (free_data_refs): Likewise.
13701 * tree-predcom.c (struct chain): Use auto_vec instead of vec for
13703 (struct component): Likewise.
13704 (pcom_worker::pcom_worker): Adjust for auto_vec and renaming changes.
13705 (pcom_worker::~pcom_worker): Likewise.
13706 (pcom_worker::release_chain): Adjust as auto_vec changes.
13707 (pcom_worker::loop): Rename to ...
13708 (pcom_worker::m_loop): ... this.
13709 (pcom_worker::datarefs): Rename to ...
13710 (pcom_worker::m_datarefs): ... this. Use auto_vec instead of vec.
13711 (pcom_worker::dependences): Rename to ...
13712 (pcom_worker::m_dependences): ... this. Use auto_vec instead of vec.
13713 (pcom_worker::chains): Rename to ...
13714 (pcom_worker::m_chains): ... this. Use auto_vec instead of vec.
13715 (pcom_worker::looparound_phis): Rename to ...
13716 (pcom_worker::m_looparound_phis): ... this. Use auto_vec instead of
13718 (pcom_worker::cache): Rename to ...
13719 (pcom_worker::m_cache): ... this. Use auto_vec instead of vec.
13720 (pcom_worker::release_chain): Adjust for auto_vec changes.
13721 (pcom_worker::release_chains): Adjust for auto_vec and renaming
13723 (release_component): Remove.
13724 (release_components): Adjust for release_component removal.
13725 (component_of): Adjust to use vec.
13726 (merge_comps): Likewise.
13727 (pcom_worker::aff_combination_dr_offset): Adjust for renaming changes.
13728 (pcom_worker::determine_offset): Likewise.
13729 (class comp_ptrs): Remove.
13730 (pcom_worker::split_data_refs_to_components): Adjust for renaming
13731 changes, for comp_ptrs removal with auto_vec.
13732 (pcom_worker::suitable_component_p): Adjust for renaming changes.
13733 (pcom_worker::filter_suitable_components): Adjust for release_component
13735 (pcom_worker::valid_initializer_p): Adjust for renaming changes.
13736 (pcom_worker::find_looparound_phi): Likewise.
13737 (pcom_worker::add_looparound_copies): Likewise.
13738 (pcom_worker::determine_roots_comp): Likewise.
13739 (pcom_worker::single_nonlooparound_use): Likewise.
13740 (pcom_worker::execute_pred_commoning_chain): Likewise.
13741 (pcom_worker::execute_pred_commoning): Likewise.
13742 (pcom_worker::try_combine_chains): Likewise.
13743 (pcom_worker::prepare_initializers_chain): Likewise.
13744 (pcom_worker::prepare_initializers): Likewise.
13745 (pcom_worker::prepare_finalizers_chain): Likewise.
13746 (pcom_worker::prepare_finalizers): Likewise.
13747 (pcom_worker::tree_predictive_commoning_loop): Likewise.
13749 2021-07-20 Martin Sebor <msebor@redhat.com>
13751 PR middle-end/101397
13752 * builtins.c (gimple_call_return_array): Add argument. Correct
13753 offsets for memchr, mempcpy, stpcpy, and stpncpy.
13754 (compute_objsize_r): Adjust offset computation for argument returning
13757 2021-07-20 Martin Sebor <msebor@redhat.com>
13759 PR middle-end/101300
13760 * tree-ssa-uninit.c (check_defs): Handle UBSAN built-ins.
13762 2021-07-20 Jeff Law <jlaw@localhost.localdomain>
13764 * function.c (assign_parm_setup_block): Use adjust_address instead
13765 of change_address to preserve MEM_EXPR and friends.
13767 2021-07-20 Martin Sebor <msebor@redhat.com>
13769 * cfgloop.h (single_likely_exit): Adjust by-value argument to
13770 by-const-reference.
13771 * cfgloopanal.c (single_likely_exit): Same.
13772 * cgraph.h (struct cgraph_node): Same.
13773 * cgraphclones.c (cgraph_node::create_virtual_clone): Same.
13774 * genautomata.c (merge_states): Same.
13775 * genextract.c (VEC_char_to_string): Same.
13776 * genmatch.c (dt_node::gen_kids_1): Same.
13777 (walk_captures): Adjust by-value argument to by-reference.
13778 * gimple-ssa-store-merging.c (check_no_overlap): Adjust by-value argument
13779 to by-const-reference.
13780 * gimple.c (gimple_build_call_vec): Same.
13781 (gimple_build_call_internal_vec): Same.
13782 (gimple_build_switch): Same.
13783 (sort_case_labels): Same.
13784 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
13786 * gimple.h (gimple_build_call_vec): Adjust by-value argument to
13787 by-const-reference.
13788 (gimple_build_call_internal_vec): Same.
13789 (gimple_build_switch): Same.
13790 (sort_case_labels): Same.
13791 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
13793 * haifa-sched.c (calc_priorities): Adjust by-value argument to
13794 by-const-reference.
13795 (sched_init_luids): Same.
13796 (haifa_init_h_i_d): Same.
13797 * ipa-cp.c (ipa_get_indirect_edge_target_1): Same.
13798 (adjust_callers_for_value_intersection): Adjust by-value argument to
13800 (find_more_scalar_values_for_callers_subset): Adjust by-value argument to
13801 by-const-reference.
13802 (find_more_contexts_for_caller_subset): Same.
13803 (find_aggregate_values_for_callers_subset): Same.
13804 (copy_useful_known_contexts): Same.
13805 * ipa-fnsummary.c (remap_edge_summaries): Same.
13806 (remap_freqcounting_predicate): Same.
13807 * ipa-inline.c (add_new_edges_to_heap): Adjust by-value argument to
13809 * ipa-predicate.c (predicate::remap_after_inlining): Adjust by-value argument
13810 to by-const-reference.
13811 * ipa-predicate.h (predicate::remap_after_inlining): Same.
13812 * ipa-prop.c (ipa_find_agg_cst_for_param): Same.
13813 * ipa-prop.h (ipa_find_agg_cst_for_param): Same.
13814 * ira-build.c (ira_loop_tree_body_rev_postorder): Same.
13815 * read-rtl.c (add_overload_instance): Same.
13816 * rtl.h (native_decode_rtx): Same.
13817 (native_decode_vector_rtx): Same.
13818 * sched-int.h (sched_init_luids): Same.
13819 (haifa_init_h_i_d): Same.
13820 * simplify-rtx.c (native_decode_vector_rtx): Same.
13821 (native_decode_rtx): Same.
13822 * tree-call-cdce.c (gen_shrink_wrap_conditions): Same.
13823 (shrink_wrap_one_built_in_call_with_conds): Same.
13824 (shrink_wrap_conditional_dead_built_in_calls): Same.
13825 * tree-data-ref.c (create_runtime_alias_checks): Same.
13826 (compute_all_dependences): Same.
13827 * tree-data-ref.h (compute_all_dependences): Same.
13828 (create_runtime_alias_checks): Same.
13829 (index_in_loop_nest): Same.
13830 * tree-if-conv.c (mask_exists): Same.
13831 * tree-loop-distribution.c (class loop_distribution): Same.
13832 (loop_distribution::create_rdg_vertices): Same.
13833 (dump_rdg_partitions): Same.
13834 (debug_rdg_partitions): Same.
13835 (partition_contains_all_rw): Same.
13836 (loop_distribution::distribute_loop): Same.
13837 * tree-parloops.c (oacc_entry_exit_ok_1): Same.
13838 (oacc_entry_exit_single_gang): Same.
13839 * tree-ssa-loop-im.c (hoist_memory_references): Same.
13840 (loop_suitable_for_sm): Same.
13841 * tree-ssa-loop-niter.c (bound_index): Same.
13842 * tree-ssa-reassoc.c (update_ops): Same.
13843 (swap_ops_for_binary_stmt): Same.
13844 (rewrite_expr_tree): Same.
13845 (rewrite_expr_tree_parallel): Same.
13846 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Same.
13847 * tree-ssa-sccvn.h (ao_ref_init_from_vn_reference): Same.
13848 * tree-ssa-structalias.c (process_all_all_constraints): Same.
13849 (make_constraints_to): Same.
13850 (handle_lhs_call): Same.
13851 (find_func_aliases_for_builtin_call): Same.
13852 (sort_fieldstack): Same.
13853 (check_for_overlaps): Same.
13854 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Same.
13855 (vect_create_cond_for_unequal_addrs): Same.
13856 (vect_create_cond_for_lower_bounds): Same.
13857 (vect_create_cond_for_alias_checks): Same.
13858 * tree-vect-slp-patterns.c (vect_validate_multiplication): Same.
13859 * tree-vect-slp.c (vect_analyze_slp_instance): Same.
13860 (vect_make_slp_decision): Same.
13861 (vect_slp_bbs): Same.
13862 (duplicate_and_interleave): Same.
13863 (vect_transform_slp_perm_load): Same.
13864 (vect_schedule_slp): Same.
13865 * tree-vectorizer.h (vect_transform_slp_perm_load): Same.
13866 (vect_schedule_slp): Same.
13867 (duplicate_and_interleave): Same.
13868 * tree.c (build_vector_from_ctor): Same.
13869 (build_vector): Same.
13870 (check_vector_cst): Same.
13871 (check_vector_cst_duplicate): Same.
13872 (check_vector_cst_fill): Same.
13873 (check_vector_cst_stepped): Same.
13874 * tree.h (build_vector_from_ctor): Same.
13876 2021-07-20 Jakub Jelinek <jakub@redhat.com>
13879 * config/rs6000/rs6000-protos.h (easy_altivec_constant): Change return
13880 type from bool to int.
13881 * config/rs6000/rs6000.c (vspltis_constant): Fix up handling the
13882 EASY_VECTOR_MSB case if either step or copies is not 1.
13883 (vspltis_shifted): Fix comment typo.
13884 (easy_altivec_constant): Change return type from bool to int, instead
13885 of returning true return byte size of the element mode that should be
13886 used to synthetize the constant.
13887 * config/rs6000/predicates.md (easy_vector_constant_msb): Require
13888 that vspltis_shifted is 0, handle the case where easy_altivec_constant
13889 assumes using different vector mode from CONST_VECTOR's mode.
13890 * config/rs6000/altivec.md (easy_vector_constant_msb splitter): Use
13891 easy_altivec_constant to determine mode in which -1 >> -1 should be
13892 performed, use rs6000_expand_vector_init instead of gen_vec_initv4sisi.
13894 2021-07-20 Richard Biener <rguenther@suse.de>
13897 * dwarf2out.h (dwarf_file_data): Add key member.
13898 * dwarf2out.c (dwarf_file_hasher::equal): Compare key.
13899 (dwarf_file_hasher::hash): Hash key.
13900 (lookup_filename): Remap the filename and store it in the
13901 filename member of dwarf_file_data when creating a new
13903 (file_name_acquire): Do not remap the filename again.
13904 (maybe_emit_file): Likewise.
13906 2021-07-20 Jonathan Wright <jonathan.wright@arm.com>
13908 * config/aarch64/aarch64-simd-builtins.def: Use two variant
13909 generators for all TBL/TBX intrinsics and rename to
13910 consistent forms: qtbl[1234] or qtbx[1234].
13911 * config/aarch64/aarch64-simd.md (aarch64_tbl1<mode>):
13913 (aarch64_qtbl1<mode>): This.
13914 (aarch64_tbx1<mode>): Rename to...
13915 (aarch64_qtbx1<mode>): This.
13916 (aarch64_tbl2v16qi): Delete.
13917 (aarch64_tbl3<mode>): Rename to...
13918 (aarch64_qtbl2<mode>): This.
13919 (aarch64_tbx4<mode>): Rename to...
13920 (aarch64_qtbx2<mode>): This.
13921 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_1): Use
13922 renamed qtbl1 and qtbl2 RTL patterns.
13923 * config/aarch64/arm_neon.h (vqtbl1_p8): Use renamed qtbl1
13925 (vqtbl1_s8): Likewise.
13926 (vqtbl1_u8): Likewise.
13927 (vqtbl1q_p8): Likewise.
13928 (vqtbl1q_s8): Likewise.
13929 (vqtbl1q_u8): Likewise.
13930 (vqtbx1_s8): Use renamed qtbx1 RTL pattern.
13931 (vqtbx1_u8): Likewise.
13932 (vqtbx1_p8): Likewise.
13933 (vqtbx1q_s8): Likewise.
13934 (vqtbx1q_u8): Likewise.
13935 (vqtbx1q_p8): Likewise.
13936 (vtbl1_s8): Use renamed qtbl1 RTL pattern.
13937 (vtbl1_u8): Likewise.
13938 (vtbl1_p8): Likewise.
13939 (vtbl2_s8): Likewise
13940 (vtbl2_u8): Likewise.
13941 (vtbl2_p8): Likewise.
13942 (vtbl3_s8): Use renamed qtbl2 RTL pattern.
13943 (vtbl3_u8): Likewise.
13944 (vtbl3_p8): Likewise.
13945 (vtbl4_s8): Likewise.
13946 (vtbl4_u8): Likewise.
13947 (vtbl4_p8): Likewise.
13948 (vtbx2_s8): Use renamed qtbx2 RTL pattern.
13949 (vtbx2_u8): Likewise.
13950 (vtbx2_p8): Likewise.
13951 (vqtbl2_s8): Use renamed qtbl2 RTL pattern.
13952 (vqtbl2_u8): Likewise.
13953 (vqtbl2_p8): Likewise.
13954 (vqtbl2q_s8): Likewise.
13955 (vqtbl2q_u8): Likewise.
13956 (vqtbl2q_p8): Likewise.
13957 (vqtbx2_s8): Use renamed qtbx2 RTL pattern.
13958 (vqtbx2_u8): Likewise.
13959 (vqtbx2_p8): Likewise.
13960 (vqtbx2q_s8): Likewise.
13961 (vqtbx2q_u8): Likewise.
13962 (vqtbx2q_p8): Likewise.
13963 (vtbx4_s8): Likewise.
13964 (vtbx4_u8): Likewise.
13965 (vtbx4_p8): Likewise.
13967 2021-07-20 Uroš Bizjak <ubizjak@gmail.com>
13970 * config/i386/sync.md (define_peephole2 atomic_storedi_fpu):
13972 (define_peephole2 atomic_loaddi_fpu): Ditto.
13974 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
13976 * config.gcc (riscv*-*-*): Detect which python is available.
13978 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
13980 * config/rs6000/vsx.md (mulhs_<mode>): Rename to...
13981 (smul<mode>3_highpart): ... this.
13982 (mulhu_<mode>): Rename to...
13983 (umul<mode>3_highpart): ... this.
13984 * config/rs6000/rs6000-builtin.def (MULHS_V2DI, MULHS_V4SI,
13985 MULHU_V2DI, MULHU_V4SI): Adjust.
13987 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
13989 PR tree-optimization/100696
13990 * internal-fn.c (first_commutative_argument): Add info for IFN_MULH.
13991 * internal-fn.def (IFN_MULH): New internal function.
13992 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Add support to
13993 recog normal multiply highpart as IFN_MULH.
13994 * config/i386/i386.c (ix86_add_stmt_cost): Adjust for combined
13997 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
13999 * config/elfos.h (CTF_DEBUGGING_INFO): New definition.
14000 (BTF_DEBUGGING_INFO): Likewise.
14001 * doc/tm.texi.in: Document the new macros.
14002 * doc/tm.texi: Regenerated.
14003 * toplev.c: Guard initialization of debug hooks.
14005 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
14007 * flags.h (ctf_debuginfo_p): New function declaration.
14008 * opts.c (ctf_debuginfo_p): New function definition.
14010 2021-07-19 Andrew Stubbs <ams@codesourcery.com>
14013 * config/gcn/gcn-hsa.h (DRIVER_SELF_SPECS): New.
14014 (ASM_SPEC): Set -mattr for xnack and sram-ecc.
14015 * config/gcn/gcn-opts.h (enum sram_ecc_type): New.
14016 * config/gcn/gcn-valu.md: Add a warning comment.
14017 * config/gcn/gcn.c (gcn_option_override): Add "sorry" for -mxnack.
14018 (output_file_start): Add xnack and sram-ecc state to ".amdgcn_target".
14019 * config/gcn/gcn.md: Add a warning comment.
14020 * config/gcn/gcn.opt: Add -mxnack and -msram-ecc.
14021 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): Remove
14023 (EF_AMDGPU_XNACK): New.
14024 (EF_AMDGPU_SRAM_ECC): New.
14026 (copy_early_debug_info): Use elf_flags.
14027 (main): Handle -mxnack and -msram-ecc options.
14028 * doc/invoke.texi: Document -mxnack and -msram-ecc.
14030 2021-07-19 Andrew Pinski <apinski@marvell.com>
14033 * config/aarch64/aarch64.md (csneg3_uxtw_insn): Rename to ...
14034 (*cs<neg_not_cs>3_uxtw_insn4): and extend to NEG_NOT.
14036 2021-07-19 Richard Biener <rguenther@suse.de>
14038 PR tree-optimization/101505
14039 * tree-vect-patterns.c (vect_determine_precisions): Walk
14040 PHIs also for loop vectorization.
14042 2021-07-19 Richard Biener <rguenther@suse.de>
14044 * gimple.h (gimple_expr_type): Remove.
14045 * doc/gimple.texi: Remove gimple_expr_type documentation.
14047 2021-07-19 Richard Biener <rguenther@suse.de>
14049 * tree-ssa-sccvn.c (vn_reference_eq): Handle NULL vr->type.
14050 (ao_ref_init_from_vn_reference): Likewise.
14051 (fully_constant_reference): Likewise.
14052 (vn_reference_lookup_call): Do not set vr->type to random
14054 * tree-ssa-pre.c (compute_avail): Do not try to PRE calls
14056 * tree-vect-generic.c (expand_vector_piecewise): Pass in
14057 whether we expanded parallel.
14058 (expand_vector_parallel): Adjust.
14059 (expand_vector_addition): Likewise.
14060 (expand_vector_comparison): Likewise.
14061 (expand_vector_operation): Likewise.
14062 (expand_vector_scalar_condition): Likewise.
14063 (expand_vector_conversion): Likewise.
14065 2021-07-19 Richard Biener <rguenther@suse.de>
14067 * tree-vrp.c (register_edge_assert_for_2): Use the
14069 (vrp_folder::fold_predicate_in): Likewise.
14070 * vr-values.c (gimple_assign_nonzero_p): Likewise.
14071 (vr_values::extract_range_from_comparison): Likewise.
14072 (vr_values::extract_range_from_ubsan_builtin): Use the
14073 type of the first operand.
14074 (vr_values::extract_range_basic): Push down type
14075 computation, use the appropriate LHS.
14076 (vr_values::extract_range_from_assignment): Use the
14079 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
14082 * common/config/i386/i386-common.c (ix86_handle_option): For
14083 -mgeneral-regs-only, enable the GPR only instructions which are
14084 enabled implicitly by SSE ISAs unless they have been disabled
14087 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
14090 * config/i386/i386.c (ix86_check_avx_upper_stores): Moved before
14091 ix86_avx_u128_mode_needed.
14092 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY if callee
14093 returns AVX register.
14095 2021-07-17 Jan Hubicka <hubicka@ucw.cz>
14097 * tree-ssa-structalias.c (handle_rhs_call): Support EAF_NOT_RETURNED.
14098 (handle_const_call): Liekise
14099 (handle_pure_call): Liekise
14101 2021-07-17 Andrew MacLeod <amacleod@redhat.com>
14103 PR tree-optimization/96542
14104 * range-op.cc (range_operator::wi_fold_in_parts): New.
14105 (range_operator::fold_range): Call wi_fold_in_parts.
14106 (operator_lshift::wi_fold): Fix broken lshift by [0,0].
14107 * range-op.h (wi_fold_in_parts): Add prototype.
14109 2021-07-16 David Malcolm <dmalcolm@redhat.com>
14111 * doc/analyzer.texi: Add __analyzer_dump_state.
14113 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
14115 * config/rs6000/rbtree.c: New file.
14116 * config/rs6000/rbtree.h: New file.
14118 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
14120 * config/rs6000/rs6000-gen-builtins.c (restriction): New enum.
14121 (typeinfo): Add restr field.
14122 (match_bracketed_pair): New function.
14123 (match_const_restriction): Implement.
14125 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
14127 * config/rs6000/rs6000-gen-builtins.c (match_basetype): Implement.
14129 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
14131 * config/rs6000/rs6000-gen-builtins.c (void_status): New enum.
14132 (basetype): Likewise.
14133 (typeinfo): Likewise.
14134 (handle_pointer): New function.
14135 (match_basetype): New stub function.
14136 (match_const_restriction): Likewise.
14137 (match_type): New function.
14139 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
14141 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace): New
14143 (advance_line): Likewise.
14144 (safe_inc_pos): Likewise.
14145 (match_identifier): Likewise.
14146 (match_integer): Likewise.
14147 (match_to_right_bracket): Likewise.
14149 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
14151 * config/rs6000/rs6000-gen-builtins.c (bif_file): New variable.
14152 (ovld_file): Likewise.
14153 (header_file): Likewise.
14154 (init_file): Likewise.
14155 (defines_file): Likewise.
14156 (pgm_path): Likewise.
14157 (bif_path): Likewise.
14158 (ovld_path): Likewise.
14159 (header_path): Likewise.
14160 (init_path): Likewise.
14161 (defines_path): Likewise.
14162 (LINELEN): New macro.
14163 (linebuf): New variable.
14167 (bif_diag): New function.
14168 (ovld_diag): Likewise.
14170 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
14172 * config/rs6000/rs6000-builtin-new.def: New.
14173 * config/rs6000/rs6000-overload.def: New.
14175 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
14177 * config/rs6000/rs6000-gen-builtins.c: New.
14179 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
14181 * Makefile.in (EXTRA_GTYPE_DEPS): New variable.
14182 (s-gtype): Depend on EXTRA_GTYPE_DEPS.
14183 * gengtype-state.c (state_writer::write_state_file_list): Add a
14184 parameter to the fileslist expression for the number of build
14186 (read_state_files_list): Detect build headers and strip the
14187 initial "./" or ".\" from their names.
14188 * gengtype.c (build_headers): New global variable.
14189 (num_build_headers): Likewise.
14190 (open_base_files): Emit #include for each build header.
14191 (main): Detect and count build headers.
14192 * gengtype.h (build_headers): New extern variable.
14193 (num_build_headers): Likewise.
14195 2021-07-16 Richard Biener <rguenther@suse.de>
14197 * gimple-ssa-store-merging.c (verify_symbolic_number_p): Use
14198 the type of the LHS.
14199 (find_bswap_or_nop_1): Likewise.
14200 (find_bswap_or_nop): Likewise.
14201 * tree-vectorizer.h (vect_get_smallest_scalar_type): Adjust
14203 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
14204 Remove unused parameters, pass in the scalar type. Fix
14205 internal store function handling.
14206 * tree-vect-stmts.c (vect_analyze_stmt): Remove assert.
14207 (vect_get_vector_types_for_stmt): Move down check for
14208 existing vector stmt after we've determined a scalar type.
14209 Pass down the used scalar type to vect_get_smallest_scalar_type.
14210 * tree-vect-generic.c (expand_vector_condition): Use
14211 the type of the LHS.
14212 (expand_vector_scalar_condition): Likewise.
14213 (expand_vector_operations_1): Likewise.
14214 * tree-vect-patterns.c (vect_widened_op_tree): Likewise.
14215 (vect_recog_dot_prod_pattern): Likewise.
14216 (vect_recog_sad_pattern): Likewise.
14217 (vect_recog_widen_op_pattern): Likewise.
14218 (vect_recog_widen_sum_pattern): Likewise.
14219 (vect_recog_mixed_size_cond_pattern): Likewise.
14221 2021-07-16 Jan Hubicka <hubicka@ucw.cz>
14223 * ipa-modref.c (struct escape_entry): Use eaf_fleags_t.
14224 (dump_eaf_flags): Dump EAF_NOT_RETURNED
14225 (eaf_flags_useful_p): Use eaf_fleags_t; handle const functions
14226 and EAF_NOT_RETURNED.
14227 (modref_summary::useful_p): Likewise.
14228 (modref_summary_lto::useful_p): Likewise.
14229 (struct) modref_summary_lto: Use eaf_fleags_t.
14230 (deref_flags): Handle EAF_NOT_RETURNED.
14231 (struct escape_point): Use min_flags.
14232 (modref_lattice::init): Add EAF_NOT_RETURNED.
14233 (merge_call_lhs_flags): Ignore EAF_NOT_RETURNED functions
14234 (analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
14236 (analyze_parms): Also analyze const functions; update conition on
14238 (modref_write): Update streaming.
14239 (read_section): Update streaming.
14240 (remap_arg_flags): Use eaf_flags_t.
14241 (modref_merge_call_site_flags): Hanlde EAF_NOT_RETURNED.
14242 * ipa-modref.h: (eaf_flags_t): New typedef.
14243 (struct modref_summary): Use eaf_flags_t.
14244 * tree-core.h (EAF_NOT_RETURNED): New constant.
14246 2021-07-16 Richard Biener <rguenther@suse.de>
14248 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Use
14249 the type of the LHS.
14250 (gimple_assign_nonnegative_warnv_p): Likewise.
14251 (gimple_call_nonnegative_warnv_p): Likewise. Return false
14252 if the call has no LHS.
14253 * gimple.c (gimple_could_trap_p_1): Use the type of the LHS.
14254 * tree-eh.c (stmt_could_throw_1_p): Likewise.
14255 * tree-inline.c (insert_init_stmt): Likewise.
14256 * tree-ssa-loop-niter.c (get_val_for): Likewise.
14257 * tree-outof-ssa.c (ssa_is_replaceable_p): Use the type of
14259 * tree-ssa-sccvn.c (init_vn_nary_op_from_stmt): Take a
14260 gassign *. Use the type of the lhs.
14261 (vn_nary_op_lookup_stmt): Adjust.
14262 (vn_nary_op_insert_stmt): Likewise.
14264 2021-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
14266 * config/s390/predicates.md (bras_sym_operand): Accept all
14267 functions in 64-bit mode, use UNSPEC_PLT31.
14268 (larl_operand): Use UNSPEC_PLT31.
14269 * config/s390/s390.c (s390_loadrelative_operand_p): Likewise.
14270 (legitimize_pic_address): Likewise.
14271 (s390_emit_tls_call_insn): Mark __tls_get_offset as function,
14273 (s390_delegitimize_address): Use UNSPEC_PLT31.
14274 (s390_output_addr_const_extra): Likewise.
14275 (print_operand): Add @PLT to TLS calls, handle %K.
14276 (s390_function_profiler): Mark __fentry__/_mcount as function,
14277 use %K, use UNSPEC_PLT31.
14278 (s390_output_mi_thunk): Use only UNSPEC_GOT, use %K.
14279 (s390_emit_call): Use UNSPEC_PLT31.
14280 (s390_emit_tpf_eh_return): Mark __tpf_eh_return as function.
14281 * config/s390/s390.md (UNSPEC_PLT31): Rename from UNSPEC_PLT.
14282 (*movdi_64): Use %K.
14283 (reload_base_64): Likewise.
14284 (*sibcall_brc): Likewise.
14285 (*sibcall_brcl): Likewise.
14286 (*sibcall_value_brc): Likewise.
14287 (*sibcall_value_brcl): Likewise.
14289 (*brasl): Likewise.
14290 (*bras_r): Likewise.
14291 (*brasl_r): Likewise.
14292 (*bras_tls): Likewise.
14293 (*brasl_tls): Likewise.
14294 (main_base_64): Likewise.
14295 (reload_base_64): Likewise.
14296 (@split_stack_call<mode>): Likewise.
14298 2021-07-16 Richard Biener <rguenther@suse.de>
14300 PR tree-optimization/101467
14301 * tree-vect-stmts.c (vect_gen_while): Properly guard
14302 make_temp_ssa_name usage.
14304 2021-07-16 Cooper Qu <cooper.qu@linux.alibaba.com>
14306 * config.gcc: Don't use forked print-sysroot-suffix.sh and
14307 t-sysroot-suffix for C-SKY.
14308 * config/csky/print-sysroot-suffix.sh: Delete.
14309 * config/csky/t-csky-linux: Delete.
14310 * config/csky/t-sysroot-suffix: Define MULTILIB_DIRNAMES
14311 instead of CSKY_MULTILIB_DIRNAMES.
14313 2021-07-16 Richard Biener <rguenther@suse.de>
14315 * tree-vect-loop.c (vect_transform_cycle_phi): Correct sign
14316 conversion issues with the partial reduction of the reused
14317 vector accumulator.
14319 2021-07-16 Richard Biener <rguenther@suse.de>
14321 * config/i386/i386-options.c (ix86_option_override_internal): Set
14322 param_vect_partial_vector_usage to zero if not set.
14324 2021-07-15 Uroš Bizjak <ubizjak@gmail.com>
14327 * config/i386/i386.h (VALID_SSE_REG_MODE): Add TDmode.
14328 (VALID_INT_MODE_P): Add SDmode and DDmode.
14329 Add TDmode for TARGET_64BIT.
14330 (VALID_DFP_MODE_P): Remove.
14331 * config/i386/i386.c (ix86_hard_regno_mode_ok):
14332 Do not use VALID_DFP_MODE_P.
14334 2021-07-15 Andrew MacLeod <amacleod@redhat.com>
14336 * gimple-range-fold.cc (adjust_pointer_diff_expr): Use
14338 (fold_using_range::fold_stmt): Ditto.
14339 (fold_using_range::range_of_range_op): Ditto.
14340 (fold_using_range::range_of_phi): Ditto.
14341 (fold_using_range::range_of_call): Ditto.
14342 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
14343 (fold_using_range::range_of_builtin_call): Ditto.
14344 (fold_using_range::range_of_cond_expr): Ditto.
14345 * gimple-range-fold.h (gimple_range_type): New.
14347 2021-07-15 Martin Sebor <msebor@redhat.com>
14349 PR middle-end/97027
14350 * tree-ssa-strlen.c (handle_assign): New function.
14351 (maybe_warn_overflow): Add argument.
14352 (nonzero_bytes_for_type): New function.
14353 (count_nonzero_bytes): Handle more tree types. Call
14354 nonzero_bytes_for_tye.
14355 (count_nonzero_bytes): Handle types.
14356 (handle_store): Handle stores from function calls.
14357 (strlen_check_and_optimize_call): Move code to handle_assign. Call
14358 it for assignments from function calls.
14360 2021-07-15 David Malcolm <dmalcolm@redhat.com>
14365 * doc/invoke.texi: Add -Wanalyzer-use-of-uninitialized-value.
14367 2021-07-15 David Malcolm <dmalcolm@redhat.com>
14369 * doc/invoke.texi (-fdump-analyzer-exploded-paths): New.
14371 2021-07-15 Martin Sebor <msebor@redhat.com>
14375 * fold-const.c (operand_compare::operand_equal_p): Handle OEP_DECL_NAME.
14376 (operand_compare::verify_hash_value): Same.
14377 * tree-core.h (OEP_DECL_NAME): New.
14379 2021-07-15 Martin Jambor <mjambor@suse.cz>
14381 * profile-count.h (profile_count::value): Change the return type to
14383 * gimple-pretty-print.c (dump_gimple_bb_header): Adjust print
14385 * tree-cfg.c (dump_function_to_file): Likewise.
14387 2021-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
14390 * config/rs6000/rs6000-p8swap.c (has_part_mult): New.
14391 (rs6000_analyze_swaps): Insns containing a subreg of a mult are
14394 2021-07-15 Richard Biener <rguenther@suse.de>
14396 * tree-vectorizer.h (vect_gen_while): Match up with
14397 vect_gen_while_not.
14398 * tree-vect-stmts.c (vect_gen_while): Adjust API to that
14399 of vect_gen_while_not.
14400 (vect_gen_while_not): Adjust.
14401 * tree-vect-loop-manip.c (vect_set_loop_controls_directly): Likewise.
14403 2021-07-15 Aldy Hernandez <aldyh@redhat.com>
14405 * gimple-range-cache.cc (non_null_ref::adjust_range): New.
14406 (ranger_cache::range_of_def): Call adjust_range.
14407 (ranger_cache::entry_range): Same.
14408 * gimple-range-cache.h (non_null_ref::adjust_range): New.
14409 * gimple-range.cc (gimple_ranger::range_of_expr): Call
14411 (gimple_ranger::range_on_entry): Same.
14413 2021-07-15 Tamar Christina <tamar.christina@arm.com>
14416 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14418 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
14420 2021-07-15 Tamar Christina <tamar.christina@arm.com>
14423 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14425 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
14426 (sdot_prod, udot_prod): ...These.
14427 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
14428 (aarch64_<sur>dot<vsi2qi>): Rename to...
14429 (<sur>dot_prod<vsi2qi>): ...This.
14430 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
14433 2021-07-15 Jakub Jelinek <jakub@redhat.com>
14435 PR middle-end/101437
14436 * gimplify.c (gimplify_expr): Throw away volatile reads from empty
14437 types even if they have non-BLKmode TYPE_MODE.
14439 2021-07-15 Richard Biener <rguenther@suse.de>
14442 * gcc.c (process_command): Process -gtoggle like process_options
14443 would after parsing options.
14445 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
14447 * cfgexpand.c (expand_asm_loc): Adjust.
14448 (expand_asm_stmt): Likewise.
14449 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Likewise.
14450 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
14451 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
14452 * config/avr/avr.c (avr_md_asm_adjust): Likewise.
14453 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
14454 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
14455 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
14456 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
14457 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
14458 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
14459 * config/s390/s390.c (s390_md_asm_adjust): Likewise.
14460 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
14461 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
14462 * doc/tm.texi: Regenerate.
14463 * target.def: Add location argument to md_asm_adjust.
14465 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
14467 * tree-diagnostic.c (diagnostic_report_current_function): Use the
14468 diagnostic's location, not input_location.
14470 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
14472 * cfgexpand.c (tree_conflicts_with_clobbers_p): Pass location to
14474 (expand_asm_stmt): Likewise.
14476 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
14478 * config/rs6000/rs6000.c (adjacent_mem_locations): Return the lower
14479 addressed memory rtx, if any.
14480 (rs6000_split_multireg_move): Fix code formatting.
14481 Handle MMA build built-ins with operands in adjacent memory locations.
14483 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
14485 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Move to later
14488 2021-07-14 Jason Merrill <jason@redhat.com>
14490 * sel-sched-ir.h (get_all_loop_exits): Use auto_vec.
14492 2021-07-14 Jason Merrill <jason@redhat.com>
14494 * doc/invoke.texi: -fdelete-dead-exceptions is on by default for
14497 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14499 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
14500 Remove erroneous line.
14502 2021-07-14 Andrew MacLeod <amacleod@redhat.com>
14504 * params.opt (param_evrp_mode): Change default.
14506 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14508 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
14509 (sdot_prod, udot_prod): ...These.
14510 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
14511 (aarch64_<sur>dot<vsi2qi>): Rename to...
14512 (<sur>dot_prod<vsi2qi>): ...This.
14513 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
14516 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14518 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
14520 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14522 * doc/sourcebuild.texi (arm_v8_2a_i8mm_neon_hw): Document.
14524 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14526 * config/arm/neon.md (usdot_prod<vsi2qi>): New.
14528 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14530 * config/aarch64/aarch64-simd.md (aarch64_usdot<vsi2qi>): Rename to...
14531 (usdot_prod<vsi2qi>): ... This.
14532 * config/aarch64/aarch64-simd-builtins.def (usdot): Rename to...
14533 (usdot_prod): ...This.
14534 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Likewise.
14535 * config/aarch64/aarch64-sve.md (@aarch64_<sur>dot_prod<vsi2qi>):
14537 (@<sur>dot_prod<vsi2qi>): ...This.
14538 * config/aarch64/aarch64-sve-builtins-base.cc
14539 (svusdot_impl::expand): Use it.
14541 2021-07-14 Tamar Christina <tamar.christina@arm.com>
14543 * optabs.def (usdot_prod_optab): New.
14544 * doc/md.texi: Document it and clarify other dot prod optabs.
14545 * optabs-tree.h (enum optab_subtype): Add optab_vector_mixed_sign.
14546 * optabs-tree.c (optab_for_tree_code): Support usdot_prod_optab.
14547 * optabs.c (expand_widen_pattern_expr): Likewise.
14548 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
14549 * tree-vect-loop.c (vectorizable_reduction): Query dot-product kind.
14550 * tree-vect-patterns.c (vect_supportable_direct_optab_p): Take optional
14552 (vect_widened_op_tree): Optionally ignore
14554 (vect_recog_dot_prod_pattern): Support usdot_prod_optab.
14556 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
14559 * config/i386/driver-i386.c (host_detect_local_cpu): Check
14560 "arch [32|64]" and "tune [32|64]" for 32-bit and 64-bit codegen.
14561 Enable UINTR only for 64-bit codegen.
14562 * config/i386/i386-options.c
14563 (ix86_option_override_internal::DEF_PTA): Skip PTA_UINTR if not
14565 * config/i386/i386.h (ARCH_ARG): New.
14566 (CC1_CPU_SPEC): Pass "[arch|tune] 32" for 32-bit codegen and
14567 "[arch|tune] 64" for 64-bit codegen.
14569 2021-07-14 Richard Biener <rguenther@suse.de>
14571 PR tree-optimization/101445
14572 * tree-vect-stmts.c (vectorizable_load): Do the gap adjustment
14573 of the IV in the correct direction for negative stride
14576 2021-07-14 Jakub Jelinek <jakub@redhat.com>
14579 * godump.c (godump_str_hash): New type.
14580 (godump_container::pot_dummy_types): Use string_hash instead of
14581 ptr_hash in the hash_set.
14583 2021-07-14 Richard Biener <rguenther@suse.de>
14585 * tree-vect-loop.c (vect_find_reusable_accumulator): Handle
14586 vector types where the old vector type has a multiple of
14587 the new vector type elements.
14588 (vect_create_partial_epilog): New function, split out from...
14589 (vect_create_epilog_for_reduction): ... here.
14590 (vect_transform_cycle_phi): Reduce the re-used accumulator
14591 to the new vector type.
14593 2021-07-14 Alexandre Oliva <oliva@adacore.com>
14595 * tree-ssa-alias.c (attr_fnspec::verify): Fix index in
14596 non-'t'-sized arg check.
14598 2021-07-14 Alexandre Oliva <oliva@adacore.com>
14600 * tree-cfg.c (cleanup_dead_labels_eh): Update
14601 post_landing_pad label upon change of landing pad block's
14603 (cleanup_dead_labels): Check that a removed label is not that
14606 2021-07-13 Jonathan Wright <jonathan.wright@arm.com>
14608 * combine.c (combine_simplify_rtx): Add vec_select -> subreg
14610 * config/aarch64/aarch64.md (*zero_extend<SHORT:mode><GPI:mode>2_aarch64):
14611 Add Neon to general purpose register case for zero-extend
14613 * config/arm/vfp.md (*arm_movsi_vfp): Remove "*" from *t -> r
14614 case to prevent some cases opting to go through memory.
14615 * cse.c (fold_rtx): Add vec_select -> subreg simplification.
14616 * rtl.c (rtvec_series_p): Define predicate to determine
14617 whether a vector contains a linear series of integers.
14618 * rtl.h (rtvec_series_p): Define.
14619 * rtlanal.c (vec_series_lowpart_p): Define predicate to
14620 determine if a vector selection is equivalent to the low part
14622 * rtlanal.h (vec_series_lowpart_p): Define.
14623 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
14624 Add vec_select -> subreg simplification.
14626 2021-07-13 Paul A. Clarke <pc@us.ibm.com>
14628 * config/rs6000/smmintrin.h (_mm_testz_si128, _mm_testc_si128,
14629 _mm_testnzc_si128, _mm_test_all_ones, _mm_test_all_zeros,
14630 _mm_test_mix_ones_zeros): New.
14632 2021-07-13 Roger Sayle <roger@nextmovesoftware.com>
14633 Richard Biener <rguenther@suse.de>
14635 * gimple.c (gimple_could_trap_p_1): Make S argument a
14636 "const gimple*". Preserve constness in call to
14637 gimple_asm_volatile_p.
14638 (gimple_could_trap_p): Make S argument a "const gimple*".
14639 * gimple.h (gimple_could_trap_p_1, gimple_could_trap_p):
14640 Update function prototypes.
14642 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14644 * tree-vectorizer.h (vect_reusable_accumulator): New structure.
14645 (_loop_vec_info::main_loop_edge): New field.
14646 (_loop_vec_info::skip_main_loop_edge): Likewise.
14647 (_loop_vec_info::skip_this_loop_edge): Likewise.
14648 (_loop_vec_info::reusable_accumulators): Likewise.
14649 (_stmt_vec_info::reduc_scalar_results): Likewise.
14650 (_stmt_vec_info::reused_accumulator): Likewise.
14651 (vect_get_main_loop_result): Declare.
14652 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize
14653 reduc_scalar_inputs.
14654 (vec_info::free_stmt_vec_info): Free reduc_scalar_inputs.
14655 * tree-vect-loop-manip.c (vect_get_main_loop_result): New function.
14656 (vect_do_peeling): Fill an epilogue loop's main_loop_edge,
14657 skip_main_loop_edge and skip_this_loop_edge fields.
14658 * tree-vect-loop.c (INCLUDE_ALGORITHM): Define.
14659 (vect_emit_reduction_init_stmts): New function.
14660 (get_initial_def_for_reduction): Use it.
14661 (get_initial_defs_for_reduction): Likewise. Change the vinfo
14662 parameter to a loop_vec_info.
14663 (vect_create_epilog_for_reduction): Store the scalar results
14664 in the reduc_info. If an epilogue loop is reusing an accumulator
14665 from the main loop, and if the epilogue loop can also be skipped,
14666 try to place the reduction code in the join block. Record
14667 accumulators that could potentially be reused by epilogue loops.
14668 (vect_transform_cycle_phi): When vectorizing epilogue loops,
14669 try to reuse accumulators from the main loop. Record the initial
14670 value in reduc_info for non-SLP reductions too.
14672 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14674 * tree-vect-loop.c (get_initial_def_for_reduction): Remove
14675 adjustment handling. Take the neutral value as an argument,
14676 in place of the code argument.
14677 (vect_transform_cycle_phi): Update accordingly. Handle the
14678 initial values of cond reductions separately from code reductions.
14679 Choose the adjustment here rather than in
14680 get_initial_def_for_reduction. Sink the splat of vec_initial_def.
14682 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14684 * tree-vect-loop.c (neutral_op_for_slp_reduction): Replace with...
14685 (neutral_op_for_reduction): ...this, providing a more general
14687 (vect_create_epilog_for_reduction): Update accordingly.
14688 (vectorizable_reduction): Likewise.
14689 (vect_transform_cycle_phi): Likewise.
14691 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14693 * tree-vect-loop.c (get_initial_def_for_reduction): Take the
14694 reduc_info instead of the original stmt_vec_info.
14695 (vect_transform_cycle_phi): Update accordingly.
14697 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14699 * tree-vect-loop.c (get_initial_defs_for_reduction): Take the
14700 reduc_info as an additional parameter.
14701 (vect_transform_cycle_phi): Update accordingly.
14703 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14705 * tree-vectorizer.h: Include tree-ssa-operands.h.
14706 (vect_phi_initial_value): New function.
14707 * tree-vect-loop.c (neutral_op_for_slp_reduction): Use it.
14708 (get_initial_defs_for_reduction, info_for_reduction): Likewise.
14709 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
14710 (vect_transform_cycle_phi, vectorizable_induction): Likewise.
14712 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14714 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
14715 the phi results to vectype after creating them. Remove later
14716 conversion code that thus becomes redundant.
14718 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14720 * tree-vect-loop.c (vect_create_epilog_for_reduction): Replace
14721 the new_phis vector with a reduc_inputs vector. Combine handling
14722 of reduction chains and ncopies > 1.
14724 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14726 * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
14727 scalar_results to group_size elements after reducing down from
14728 N*group_size elements. Construct an array_slice of the live-out
14729 stmts and assert that there is one stmt per scalar result.
14731 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14733 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
14734 nested_in_vect_loop and use double_reduc everywhere. Remove dead
14735 assignment to "loop".
14737 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
14739 * internal-fn.c (vectorized_internal_fn_supported_p): Handle
14740 vector types first. For scalar types, consider both the preferred
14741 vector mode and the alternative vector modes.
14742 * optabs-query.c (can_vec_mask_load_store_p): Use the same
14743 structure as above, in particular using related_vector_mode
14744 for modes provided by autovectorize_vector_modes.
14746 2021-07-13 Jakub Jelinek <jakub@redhat.com>
14747 Richard Biener <rguenther@suse.de>
14749 PR tree-optimization/101419
14750 * tree-pass.h (PROP_objsz): Define.
14751 (make_pass_early_object_sizes): Declare.
14752 * passes.def (pass_all_early_optimizations): Rename pass_object_sizes
14753 there to pass_early_object_sizes, drop parameter.
14754 (pass_all_optimizations): Move pass_object_sizes right after pass_ccp,
14755 drop parameter, move pass_post_ipa_warn right after that.
14756 * tree-object-size.c (pass_object_sizes::execute): Rename to...
14757 (object_sizes_execute): ... this. Add insert_min_max_p argument.
14758 (pass_data_object_sizes): Move after object_sizes_execute.
14759 (pass_object_sizes): Likewise. In execute method call
14760 object_sizes_execute, drop set_pass_param method and insert_min_max_p
14761 non-static data member and its initializer in the ctor.
14762 (pass_data_early_object_sizes, pass_early_object_sizes,
14763 make_pass_early_object_sizes): New.
14764 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Use
14765 (cfun->curr_properties & PROP_objsz) instead of cfun->after_inlining.
14767 2021-07-13 Kito Cheng <kito.cheng@sifive.com>
14770 * config/riscv/constraints.md ("S"): Update description and remove
14772 * doc/md.texi (Machine Constraints): Document the 'S' constraints
14775 2021-07-13 Richard Biener <rguenther@suse.de>
14778 2021-07-12 Richard Biener <rguenther@suse.de>
14780 * tree-vect-slp.c (vect_slp_region): Show the number of
14781 SLP graph entries in the optimization message.
14783 2021-07-13 Michael Meissner <meissner@linux.ibm.com>
14785 * config/rs6000/altivec.md (xxspltiw_v4sf): Change local variable
14787 * config/rs6000/rs6000-protos.h (rs6000_const_f32_to_i32): Change
14788 return type to long.
14789 * config/rs6000/rs6000.c (rs6000_const_f32_to_i32): Change return
14792 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
14794 * gimple-range-fold.cc (fold_using_range::range_of_builtin_ubsan_call):
14795 Query relation between the 2 operands and use it.
14797 2021-07-12 Sergei Trofimovich <siarheit@google.com>
14799 * doc/cfg.texi: Fix s/ei_safe_safe/ei_safe_edge/ typo.
14801 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
14804 * config/i386/predicates.md (vec_setm_sse41_operand):
14805 Rename from vec_setm_operand.
14806 (vec_setm_avx2_operand): New predicate.
14807 * config/i386/sse.md (vec_set<V_128:mode>): Use V_128 mode iterator.
14808 Use vec_setm_sse41_operand as operand 2 predicate.
14809 (vec_set<V_256_512:mode): New expander.
14810 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_sse41_operand
14811 as operand 2 predicate.
14813 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
14815 PR tree-optimization/101335
14816 * range-op.cc (operator_cast::lhs_op1_relation): Delete.
14818 2021-07-12 Andrew Pinski <apinski@marvell.com>
14820 * tree-ssa-phiopt.c (match_simplify_replacement): Move
14821 insert of the sequence before the movement of the
14822 statement. Check if to see if the statement is used
14823 outside of the original phi to see if we should move it.
14825 2021-07-12 Richard Biener <rguenther@suse.de>
14827 * dump-context.h (debug_dump_context::debug_dump_context):
14828 Add FILE * parameter defaulted to stderr.
14829 * dumpfile.c (debug_dump_context::debug_dump_context): Adjust.
14830 * tree-vect-slp.c (dot_slp_tree): New functions.
14832 2021-07-12 Richard Biener <rguenther@suse.de>
14834 PR tree-optimization/101373
14835 * tree-ssa-pre.c (prune_clobbered_mems): Also prune trapping
14836 references when the BB may not return.
14837 (compute_avail): Pass in the function we're working on and
14838 replace cfun references with it. Externally throwing
14839 const calls also possibly terminate the function.
14840 (pass_pre::execute): Pass down the function we're working on.
14841 * gcse.c (compute_hash_table_work): Externally throwing
14842 const/pure calls also need record_last_mem_set_info.
14843 * postreload-gcse.c (record_opr_changes): Looping or externally
14844 throwing const/pure calls also need record_last_mem_set_info.
14846 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
14848 * recog.c (memory_address_addr_space_p): Change the type to bool.
14849 Return true/false instead of 1/0.
14850 (offsettable_memref_p): Ditto.
14851 (offsettable_nonstrict_memref_p): Ditto.
14852 (offsettable_address_addr_space_p): Ditto.
14853 Change the type of addressp indirect function to bool.
14854 * recog.h (memory_address_addr_space_p): Change the type to bool.
14855 (strict_memory_address_addr_space_p): Ditto.
14856 (offsettable_memref_p): Ditto.
14857 (offsettable_nonstrict_memref_p): Ditto.
14858 (offsettable_address_addr_space_p): Ditto.
14859 * reload.c (maybe_memory_address_addr_space_p): Ditto.
14860 (strict_memory_address_addr_space_p): Change the type to bool.
14861 Return true/false instead of 1/0.
14862 (maybe_memory_address_addr_space_p): Change the type to bool.
14864 2021-07-12 Richard Biener <rguenther@suse.de>
14866 * tree-vect-slp.c (vect_slp_region): Show the number of
14867 SLP graph entries in the optimization message.
14869 2021-07-12 Richard Biener <rguenther@suse.de>
14871 PR tree-optimization/101394
14872 * tree-ssa-pre.c (do_pre_regular_insertion): Avoid inserting
14873 copies from abnormals for a full redundancy.
14875 2021-07-12 Richard Biener <rguenther@suse.de>
14877 PR middle-end/101423
14878 * gimple.c (gimple_could_trap_p_1): Internal function calls
14880 * tree-eh.c (tree_could_trap_p): Likewise.
14882 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
14885 * config/arm/arm_neon.h (vmul_n_u32): Replace call to builtin with
14887 (vmulq_n_u32): Likewise.
14888 (vmul_n_f32): Gate __a * __b on __FAST_MATH__.
14889 (vmulq_n_f32): Likewise.
14890 (vmul_n_f16): Likewise.
14891 (vmulq_n_f16): Likewise.
14893 2021-07-12 Martin Liska <mliska@suse.cz>
14895 PR sanitizer/101425
14896 * gcc.c (check_offload_target_name): Call
14897 candidates_list_and_hint only if we have a candidate.
14899 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
14902 * config/arm/neon.md (vec_init): Move to ...
14903 * config/arm/vec-common.md (vec_init): ... here.
14904 Change the pattern's mode to VDQX and gate it on VALID_MVE_MODE.
14906 2021-07-12 Roger Sayle <roger@nextmovesoftware.com>
14908 PR tree-optimization/101403
14909 * match.pd ((T)bswap(X)>>C): Correctly handle cases where
14910 signedness of the shift is not the same as the signedness of
14911 the type extension.
14913 2021-07-09 Roger Sayle <roger@nextmovesoftware.com>
14914 Uroš Bizjak <ubizjak@gmail.com>
14916 * config/i386/i386.md (*divmodsi4_const): Optimize SImode
14917 divmod of a constant numerator with new define_insn_and_split.
14919 2021-07-09 Iain Sandoe <iain@sandoe.co.uk>
14922 * config/i386/i386-expand.c (ix86_expand_call): If a call is
14923 to a non-local-binding, or local but to a public symbol, then
14924 assume that it might be indirected via the lazy symbol binder.
14925 Mark R10 and R10 as clobbered in that case.
14927 2021-07-09 Eric Botcazou <ebotcazou@adacore.com>
14930 * gcc.c (ASM_DEBUG_DWARF_OPTION): Set again to --gdwarf2 in
14931 the case where HAVE_AS_WORKING_DWARF_N_FLAG is not defined
14932 and HAVE_LD_BROKEN_PE_DWARF5 is defined.
14934 2021-07-09 Uroš Bizjak <ubizjak@gmail.com>
14936 * config/i386/i386.md (*udivmodsi4_pow2_zext_1): Limit the
14937 log2 range of operands[3] to [1,31].
14938 (*udivmodsi4_pow2_zext_2): Ditto. Correct insn RTX pattern.
14940 2021-07-09 Sergei Trofimovich <siarheit@google.com>
14942 * doc/md.texi: Don't split @smallexample in multiple @groups.
14944 2021-07-09 Sergei Trofimovich <siarheit@google.com>
14946 * doc/md.texi: Add missing 'see' word.
14948 2021-07-09 Andrew Pinski <apinski@marvell.com>
14950 * tree-ssa-phiopt.c (phiopt_early_allow): Change arguments
14951 to take sequence and gimple_match_op. Accept the case where
14952 op is a SSA_NAME and one statement in the sequence.
14953 Also allow constants.
14954 (gimple_simplify_phiopt): Always pass a sequence to resimplify.
14955 Update call to phiopt_early_allow. Discard the sequence if not
14958 2021-07-09 Xi Ruoyao <xry111@mengyan1223.wang>
14963 * config/mips/mips.c (mips_const_insns): Use MSA_SUPPORTED_MODE_P
14964 instead of ISA_HAS_MSA.
14965 (mips_expand_vec_unpack): Likewise.
14966 (mips_expand_vector_init): Likewise.
14968 2021-07-09 Kewen Lin <linkw@linux.ibm.com>
14970 * config/rs6000/vsx.md (mods_<mode>): Rename to...
14971 (mod<mode>3): ... this.
14972 (modu_<mode>): Rename to...
14973 (umod<mode>3): ... this.
14974 * config/rs6000/rs6000-builtin.def (MODS_V2DI, MODS_V4SI, MODU_V2DI,
14975 MODU_V4SI): Adjust.
14977 2021-07-08 Jeff Law <jeffreyalaw@gmail.com>
14979 * config/h8300/shiftrotate.md (variable shifts): Expose condition
14980 code handling for the test before the loop.
14982 2021-07-08 Martin Jambor <mjambor@suse.cz>
14985 * ipa-sra.c (class isra_call_summary): New member
14986 m_before_any_store, initialize it in the constructor.
14987 (isra_call_summary::dump): Dump the new field.
14988 (ipa_sra_call_summaries::duplicate): Copy it.
14989 (process_scan_results): Set it.
14990 (isra_write_edge_summary): Stream it.
14991 (isra_read_edge_summary): Likewise.
14992 (param_splitting_across_edge): Only override
14993 safe_to_import_accesses if m_before_any_store is set.
14995 2021-07-08 Martin Sebor <msebor@redhat.com>
14997 PR bootstrap/101374
14998 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref):
14999 Use Object Size Type 0 instead of 1.
15001 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
15003 * tree-vect-loop.c (vectorizable_reduction): Remove always-true
15006 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
15008 * match.pd: Simplify an extend-operate-truncate sequence involving
15011 2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
15012 Richard Biener <rguenther@suse.de>
15014 PR tree-optimization/40210
15015 * match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
15016 (x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
15017 when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
15019 2021-07-08 Uroš Bizjak <ubizjak@gmail.com>
15022 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
15024 * config/i386/mmx.md (V_32): New mode iterator.
15025 (mov<V_32:mode>): Use V_32 mode iterator.
15026 (*mov<V_32:mode>_internal): Ditto.
15027 (*push<V_32:mode>2_rex64): Ditto.
15028 (*push<V_32:mode>2): Ditto.
15029 (movmisalign<V_32:mode>): Ditto.
15030 (mmx_<any_shiftrt:insn>v1si3): New insn pattern.
15031 (sse4_1_<any_extend:code>v2qiv2hi2): Ditto.
15032 (vec_unpacks_lo_v4qi): New expander.
15033 (vec_unpacks_hi_v4qi): Ditto.
15034 (vec_unpacku_lo_v4qi): Ditto.
15035 (vec_unpacku_hi_v4qi): Ditto.
15036 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add V1SImode.
15037 (VALID_INT_MODE_P): Ditto.
15039 2021-07-08 Michael Meissner <meissner@linux.ibm.com>
15042 * config/rs6000/rs6000.md (udivti3): New insn.
15043 (divti3): New insn.
15044 (umodti3): New insn.
15045 (modti3): New insn.
15047 2021-07-07 Martin Sebor <msebor@redhat.com>
15049 PR tree-optimization/100137
15050 PR tree-optimization/99121
15051 PR tree-optimization/97027
15052 * builtins.c (access_ref::access_ref): Also set offmax.
15053 (access_ref::offset_in_range): Define new function.
15054 (access_ref::add_offset): Set offmax.
15055 (access_ref::inform_access): Handle access_none.
15056 (handle_mem_ref): Clear ostype.
15057 (compute_objsize_r): Handle ASSERT_EXPR.
15058 * builtins.h (struct access_ref): Add offmax member.
15059 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Use
15060 compute_objsize() and simplify.
15062 2021-07-07 Peter Bergner <bergner@linux.ibm.com>
15064 * config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
15065 and VSX_BUILTIN_STXVP.
15067 2021-07-07 Martin Sebor <msebor@redhat.com>
15070 * config/aarch64/aarch64.c (aarch64_simd_lane_bounds): Remove
15071 a stray %K from error_at() missed in r12-2088.
15073 2021-07-07 Richard Biener <rguenther@suse.de>
15075 PR tree-optimization/99728
15076 * tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
15078 (mem_refs_may_alias_p): Add assert we handled aggregate
15080 (sm_seq_valid_bb): Give up when running into aggregate copies.
15081 (ref_indep_loop_p): Handle aggregate copies as never
15082 being invariant themselves but allow other refs to be
15083 disambiguated against them.
15084 (can_sm_ref_p): Do not try to apply store-motion to aggregate
15087 2021-07-06 Indu Bhagat <indu.bhagat@oracle.com>
15090 * dwarf2ctf.c (ctf_get_AT_data_member_location): Multiply by 8 to get
15093 2021-07-06 Martin Sebor <msebor@redhat.com>
15095 * gimple-pretty-print.c (percent_G_format): Remove.
15096 * tree-diagnostic.c (default_tree_printer): Remove calls.
15097 * tree-pretty-print.c (percent_K_format): Remove.
15098 * tree-pretty-print.h (percent_K_format): Remove.
15100 2021-07-06 Martin Sebor <msebor@redhat.com>
15102 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_builtin):
15103 Remove %K and use error_at.
15104 (aarch64_expand_fcmla_builtin): Same.
15105 (aarch64_expand_builtin_tme): Same.
15106 (aarch64_expand_builtin_memtag): Same.
15107 * config/arm/arm-builtins.c (arm_expand_acle_builtin): Same.
15108 (arm_expand_builtin): Same.
15109 * config/arm/arm.c (bounds_check): Same.
15111 2021-07-06 Martin Sebor <msebor@redhat.com>
15113 * builtins.c (warn_string_no_nul): Remove %G.
15114 (maybe_warn_for_bound): Same.
15115 (warn_for_access): Same.
15116 (check_access): Same.
15117 (check_strncat_sizes): Same.
15118 (expand_builtin_strncat): Same.
15119 (expand_builtin_strncmp): Same.
15120 (expand_builtin): Same.
15121 (expand_builtin_object_size): Same.
15122 (warn_dealloc_offset): Same.
15123 (maybe_emit_free_warning): Same.
15124 * calls.c (maybe_warn_alloc_args_overflow): Same.
15125 (maybe_warn_nonstring_arg): Same.
15126 (maybe_warn_rdwr_sizes): Same.
15127 * expr.c (expand_expr_real_1): Remove %K.
15128 * gimple-fold.c (gimple_fold_builtin_strncpy): Remove %G.
15129 (gimple_fold_builtin_strncat): Same.
15130 * gimple-ssa-sprintf.c (format_directive): Same.
15131 (handle_printf_call): Same.
15132 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Same.
15133 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
15134 (maybe_diag_access_bounds): Same. Call gimple_location.
15135 (check_bounds_or_overlap): Same.
15136 * trans-mem.c (ipa_tm_scan_irr_block): Remove %K. Simplify.
15137 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Remove %G.
15138 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
15139 (maybe_diag_stxncpy_trunc): Same.
15140 (handle_builtin_stxncpy_strncat): Same.
15141 (maybe_warn_pointless_strcmp): Same.
15142 * tree-ssa-uninit.c (maybe_warn_operand): Same.
15144 2021-07-06 Uroš Bizjak <ubizjak@gmail.com>
15147 * config/i386/predicates.md (vec_setm_operand): Enable
15148 register_operand for TARGET_SSE4_1.
15149 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_operand
15150 as operand 2 predicate. Call ix86_expand_vector_set_var
15151 for non-constant index operand.
15152 (vec_setv4qi): Use vec_setm_mmx_operand as operand 2 predicate.
15153 Call ix86_expand_vector_set_var for non-constant index operand.
15155 2021-07-06 Jeff Law <jeffreyalaw@gmail.com>
15157 * config/h8300/jumpcall.md (*branch): When possible, generate
15158 the comparison in CCZN mode.
15159 * config/h8300/predicates.md (simple_memory_operand): Reject all
15160 auto-increment addressing modes.
15162 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
15164 PR bootstrap/100246
15165 * config/i386/i386.h (struct stringop_algs): Define a CTOR for
15168 2021-07-06 Richard Biener <rguenther@suse.de>
15170 * doc/md.texi (vec_fmaddsub<mode>4): Document.
15171 (vec_fmsubadd<mode>4): Likewise.
15172 * optabs.def (vec_fmaddsub$a4): Add.
15173 (vec_fmsubadd$a4): Likewise.
15174 * internal-fn.def (IFN_VEC_FMADDSUB): Add.
15175 (IFN_VEC_FMSUBADD): Likewise.
15176 * tree-vect-slp-patterns.c (addsub_pattern::recognize):
15177 Refactor to handle IFN_VEC_FMADDSUB and IFN_VEC_FMSUBADD.
15178 (addsub_pattern::build): Likewise.
15179 * tree-vect-slp.c (vect_optimize_slp): CFN_VEC_FMADDSUB
15180 and CFN_VEC_FMSUBADD are not transparent for permutes.
15181 * config/i386/sse.md (vec_fmaddsub<mode>4): New expander.
15182 (vec_fmsubadd<mode>4): Likewise.
15184 2021-07-06 Richard Biener <rguenther@suse.de>
15186 * doc/invoke.texi (fmove-loop-stores): Document.
15187 * common.opt (fmove-loop-stores): New option.
15188 * opts.c (default_options_table): Enable -fmove-loop-stores
15189 at -O1 but not -Og.
15190 * tree-ssa-loop-im.c (pass_lim::execute): Pass
15191 flag_move_loop_stores instead of true to
15192 loop_invariant_motion_in_fun.
15194 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
15196 * doc/install.texi: Document --with-dsymutil.
15198 2021-07-06 Andrew Pinski <apinski@marvell.com>
15200 PR tree-optimization/101256
15201 * dbgcnt.def (phiopt_edge_range): New counter.
15202 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
15203 Check to make sure the new name is defined in the same
15204 bb as the conditional before duplicating range info.
15205 Also add debug counter.
15207 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
15209 PR rtl-optimization/100328
15210 * config/i386/i386-options.c (ix86_option_override_internal):
15211 Set param_ira_consider_dup_in_all_alts to 0.
15213 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
15215 PR rtl-optimization/100328
15216 * doc/invoke.texi (ira-consider-dup-in-all-alts): Document new
15218 * ira.c (ira_get_dup_out_num): Adjust as parameter
15219 param_ira_consider_dup_in_all_alts.
15220 * params.opt (ira-consider-dup-in-all-alts): New.
15221 * ira-conflicts.c (process_regs_for_copy): Add one parameter
15222 single_input_op_has_cstr_p.
15223 (get_freq_for_shuffle_copy): New function.
15224 (add_insn_allocno_copies): Adjust as single_input_op_has_cstr_p.
15225 * ira-int.h (ira_get_dup_out_num): Add one bool parameter.
15227 2021-07-05 Jeff Law <jeffreyalaw@gmail.com>
15229 * config/h8300/shiftrotate.md (shift-by-variable patterns): Update to
15230 generate condition code aware RTL directly.
15232 2021-07-05 Andrew Pinski <apinski@marvell.com>
15234 PR tree-optimization/101039
15235 * match.pd (A CMP 0 ? A : -A): New patterns.
15236 * tree-ssa-phiopt.c (abs_replacement): Delete function.
15237 (tree_ssa_phiopt_worker): Don't call abs_replacement.
15238 Update comment about abs_replacement.
15240 2021-07-05 Andrew Pinski <apinski@marvell.com>
15242 * tree-ssa-phiopt.c (gimple_simplify_phiopt):
15243 If "A ? B : C" fails to simplify, try "(!A) ? C : B".
15245 2021-07-05 Andrew Pinski <apinski@marvell.com>
15247 * tree-ssa-phiopt.c (match_simplify_replacement):
15248 Add early_p argument. Call gimple_simplify_phiopt
15249 instead of gimple_simplify.
15250 (tree_ssa_phiopt_worker): Update call to
15251 match_simplify_replacement and allow unconditionally.
15252 (phiopt_early_allow): New function.
15253 (gimple_simplify_phiopt): New function.
15255 2021-07-05 Andrew Pinski <apinski@marvell.com>
15257 PR middle-end/101237
15258 * fold-const.c (negate_expr_p): Remove call to element_mode
15259 and TREE_MODE/TREE_TYPE when calling HONOR_SIGNED_ZEROS,
15260 HONOR_SIGN_DEPENDENT_ROUNDING, and HONOR_SNANS.
15261 (fold_negate_expr_1): Likewise.
15262 (const_unop): Likewise.
15263 (fold_cond_expr_with_comparison): Likewise.
15264 (fold_binary_loc): Likewise.
15265 (fold_ternary_loc): Likewise.
15266 (tree_call_nonnegative_warnv_p): Likewise.
15267 * match.pd (-(A + B) -> (-B) - A): Likewise.
15269 2021-07-05 Iain Sandoe <iain@sandoe.co.uk>
15271 * configure.ac: Handle --with-dsymutil in the same way as we
15272 do for the assembler and linker. (DEFAULT_DSYMUTIL): New.
15273 Extract the type and version for the dsymutil configured or
15274 found by the default searches.
15275 * config.in: Regenerated.
15276 * configure: Regenerated.
15277 * collect2.c (do_dsymutil): Handle locating dsymutil in the
15278 same way as for the assembler and linker.
15279 * config/darwin.h (DSYMUTIL): Delete.
15280 * gcc.c: Report a configured dsymutil correctly.
15281 * exec-tool.in: Allow for dsymutil.
15283 2021-07-05 Uroš Bizjak <ubizjak@gmail.com>
15285 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
15286 Handle V4QI and V2HI modes.
15287 (expand_vec_perm_blend): Allow 4-byte vector modes with TARGET_SSE4_1.
15288 Handle V4QI mode. Emit mmx_pblendvb32 for 4-byte modes.
15289 (expand_vec_perm_pshufb): Rewrite to use switch statemets.
15290 Handle 4-byte dual operands with TARGET_XOP and single operands
15291 with TARGET_SSSE3. Emit mmx_ppermv32 for TARGET_XOP and
15292 mmx_pshufbv4qi3 for TARGET_SSSE3.
15293 (expand_vec_perm_pblendv): Allow 4-byte vector modes with TARGET_SSE4_1.
15294 (expand_vec_perm_interleave2): Allow 4-byte vector modes.
15295 (expand_vec_perm_pshufb2): Allow 4-byte vector modes with TARGET_SSSE3.
15296 (expand_vec_perm_even_odd_1): Handle V4QI mode.
15297 (expand_vec_perm_broadcast_1): Handle V4QI mode.
15298 (ix86_vectorize_vec_perm_const): Handle V4QI mode.
15299 * config/i386/mmx.md (mmx_ppermv32): New insn pattern.
15300 (mmx_pshufbv4qi3): Ditto.
15301 (*mmx_pblendw32): Ditto.
15302 (*mmx_pblendw64): Rename from *mmx_pblendw.
15303 (mmx_punpckhbw_low): New insn_and_split pattern.
15304 (mmx_punpcklbw_low): Ditto.
15306 2021-07-05 Richard Biener <rguenther@suse.de>
15308 * tree-vect-loop-manip.c (vect_loop_versioning): Do not
15309 set LOOP_C_INFINITE on the vectorized loop.
15311 2021-07-05 Richard Biener <rguenther@suse.de>
15313 PR middle-end/101291
15314 * cfgloopmanip.c (loop_version): Set the loop copy of the
15315 versioned loop to the new loop.
15317 2021-07-04 Iain Sandoe <iain@sandoe.co.uk>
15320 * config.gcc: Ensure that Darwin biarch definitions are
15321 added before i386.h.
15322 * config/i386/darwin.h (TARGET_64BIT): Remove.
15323 (PR80556_WORKAROUND): New.
15324 (REAL_LIBGCC_SPEC): Amend to use PR80556_WORKAROUND.
15325 (DARWIN_SUBARCH_SPEC): New.
15326 * config/i386/darwin32-biarch.h (TARGET_64BIT_DEFAULT,
15327 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
15328 (REAL_LIBGCC_SPEC): Remove.
15329 * config/i386/darwin64-biarch.h (TARGET_64BIT_DEFAULT,
15330 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
15331 (REAL_LIBGCC_SPEC): Remove.
15333 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
15335 PR middle-end/101294
15336 * expr.c (store_constructor): Don't use vec_duplicate on vector.
15338 2021-07-02 Martin Sebor <msebor@redhat.com>
15340 PR middle-end/98871
15341 PR middle-end/98512
15342 * diagnostic.c (get_any_inlining_info): New.
15343 (update_effective_level_from_pragmas): Handle inlining context.
15344 (diagnostic_enabled): Same.
15345 (diagnostic_report_diagnostic): Same.
15346 * diagnostic.h (struct diagnostic_info): Add ctor.
15347 (struct diagnostic_context): Add new member.
15348 * tree-diagnostic.c (set_inlining_locations): New.
15349 (tree_diagnostics_defaults): Set new callback pointer.
15351 2021-07-02 Peter Bergner <bergner@linux.ibm.com>
15353 * config/rs6000/rs6000-builtin.def (BU_MMA_PAIR_LD, BU_MMA_PAIR_ST):
15355 (__builtin_vsx_lxvp, __builtin_vsx_stxvp): New built-ins.
15356 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Expand
15357 lxvp and stxvp built-ins.
15358 (mma_init_builtins): Handle lxvp and stxvp built-ins.
15359 (builtin_function_type): Likewise.
15360 * doc/extend.texi (__builtin_vsx_lxvp, __builtin_mma_stxvp): Document.
15362 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
15364 * config/h8300/h8300-protos.h (compute_a_shift_cc): Accept
15365 additional argument for the code.
15366 * config/h8300/h8300.c (compute_a_shift_cc): Accept additional
15367 argument for the code. Just return if the ZN bits are useful or
15368 not rather than the old style CC_* enums.
15369 * config/h8300/shiftrotate.md (shiftqi_noscratch): Move before
15370 more generic shiftqi patterns.
15371 (shifthi_noscratch, shiftsi_noscratch): Similarly.
15372 (shiftqi_noscratch_set_flags): New pattern.
15373 (shifthi_noscratch_set_flags, shiftsi_noscratch_set_flags): Likewise.
15375 2021-07-02 Andrew MacLeod <amacleod@redhat.com>
15377 PR tree-optimization/101223
15378 * range-op.cc (build_lt): Add -1 for signed values.
15379 (built_gt): Subtract -1 for signed values.
15381 2021-07-02 David Faust <david.faust@oracle.com>
15383 * btfout.c (get_btf_kind): Support BTF_KIND_FLOAT.
15384 (btf_asm_type): Likewise.
15386 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
15388 * config/h8300/h8300-protos.h (output_a_shift): Make first argument
15389 an array of rtx rather than a pointer to rtx. Add code argument.
15390 (compute_a_shift_length): Similarly.
15391 * config/h8300/h8300.c (h8300_shift_costs): Adjust now that the
15392 shift itself isn't an operand. Create dummy operand[0] to carry
15393 a mode and pass a suitable rtx code to compute_a_shift_length.
15394 (get_shift_alg): Adjust operand number of clobber in output templates.
15395 (output_a_shift): Make first argument an array of rtx rather than
15396 a pointer to rtx. Add code argument for the type of shift.
15397 Adjust now that the shift itself is no longer an operand.
15398 (compute_a_shift_length): Similarly.
15399 * config/h8300/shiftrotate.md (shiftqi, shifthi, shiftsi): Use an
15400 iterator rather than nshift_operator.
15401 (shiftqi_noscratch, shifthi_noscratch, shiftsi_noscratch): Likewise.
15402 (shiftqi_clobber_flags): Adjust to API changes in output_a_shift
15403 and compute_a_shift_length.
15404 (shiftqi_noscratch_clobber_flags): Likewise.
15405 (shifthi_noscratch_clobber_flags): Likewise.
15406 (shiftsi_noscratch_clobber_flags): Likewise.
15408 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
15411 * config/darwin.h (DSYMUTIL_SPEC): Do not try to run
15412 dsymutil for BTF/CTF.
15414 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
15417 * config/darwin.h (CTF_INFO_SECTION_NAME): Update the
15418 segment to include BTF.
15419 (BTF_INFO_SECTION_NAME): New.
15421 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
15423 * config/m32r/m32r-protos.h (call_operand): Adjust return type.
15424 (small_data_operand, memreg_operand, small_insn_p): Likewise.
15425 * config/m32r/m32r.c (call_operand): Adjust return type.
15426 (small_data_operand, memreg_operand): Likewise.
15428 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
15430 * config/frv/frv-protos.h (integer_register_operand): Adjust return
15432 (frv_load_operand, gpr_or_fpr_operand, gpr_no_subreg_operand): Likewise.
15433 (fpr_or_int6_operand, gpr_or_int_operand); Likewise.
15434 (gpr_or_int12_operand, gpr_or_int10_operand); Likewise.
15435 (move_source_operand, move_destination_operand): Likewise.
15436 (condexec_source_operand, condexec_dest_operand): Likewise.
15437 (lr_operand, gpr_or_memory_operand, fpr_or_memory_operand): Likewise.
15438 (reg_or_0_operand, fcc_operand, icc_operand, cc_operand): Likewise.
15439 (fcr_operand, icr_operand, cr_operand, call_operand): Likewise.
15440 (fpr_operand, even_reg_operand, odd_reg_operand): Likewise.
15441 (even_gpr_operand, odd_gpr_operand, quad_fpr_operand): Likewise.
15442 (even_fpr_operand, odd_fpr_operand): Likewise.
15443 (dbl_memory_one_insn_operand, dbl_memory_two_insn_operand): Likewise.
15444 (int12_operand, int6_operand, int5_operand, uint5_operand): Likewise.
15445 (uint4_operand, uint1_operand, int_2word_operand): Likewise
15446 (upper_int16_operand, uint16_operand, symbolic_operand): Likewise.
15447 (relational_operator, float_relational_operator): Likewise.
15448 (ccr_eqne_operator, minmax_operator): Likewise.
15449 (condexec_si_binary_operator, condexec_si_media_operator): Likewise.
15450 (condexec_si_divide_operator, condexec_si_unary_operator): Likewise.
15451 (condexec_sf_conv_operator, condexec_sf_add_operator): Likewise.
15452 (intop_compare_operator, acc_operand, even_acc_operand): Likewise.
15453 (quad_acc_operand, accg_operand): Likewise.
15455 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
15457 * config/stormy16/stormy16-protos.h (xstormy16_below_100_symbol): Change
15458 return type to a bool.
15459 (nonimmediate_nonstack_operand): Likewise.
15460 (xstormy16_splittable_below100_operand): Likewise.
15461 * config/stormy16/stormy16.c (xstormy16_below_100_symbol): Fix
15463 (xstormy16_splittable_below100_operand): Likewise.
15465 2021-07-02 Richard Biener <rguenther@suse.de>
15467 PR tree-optimization/101293
15468 * tree-ssa-loop-im.c (mem_ref_hasher::equal): Compare MEM_REF bases
15469 with combined offsets.
15470 (gather_mem_refs_stmt): Hash MEM_REFs as if their offset were
15471 combined with the rest of the offset.
15473 2021-07-02 Eric Botcazou <ebotcazou@adacore.com>
15475 * config/i386/i386.c (asm_preferred_eh_data_format): Always use the
15476 PIC encodings for PE-COFF targets.
15478 2021-07-02 Jakub Jelinek <jakub@redhat.com>
15481 * config/i386/i386-expand.c (ix86_broadcast_from_integer_constant):
15482 Return nullptr for TImode inner mode.
15484 2021-07-02 Richard Biener <rguenther@suse.de>
15486 PR tree-optimization/101280
15487 PR tree-optimization/101173
15488 * gimple-loop-interchange.cc
15489 (tree_loop_interchange::valid_data_dependences): Properly
15490 guard all dependence checks with DDR_REVERSED_P or its
15493 2021-07-02 Hongyu Wang <hongyu.wang@intel.com>
15495 * config/i386/i386-expand.c (ix86_expand_builtin):
15496 Add branch to clear odata when ZF is set for asedecenc_expand
15497 and wideaesdecenc_expand.
15499 2021-07-02 Eugene Rozenfeld <erozen@microsoft.com>
15501 * config/i386/gcc-auto-profile: regenerate
15503 2021-07-02 liuhongt <hongtao.liu@intel.com>
15505 * config/i386/sse.md (trunc<mode><pmov_dst_4>2): Refined to ..
15506 (trunc<mode><pmov_dst_4_lower>2): this.
15508 2021-07-01 David Malcolm <dmalcolm@redhat.com>
15510 * diagnostic.h (diagnostic_context::m_file_cache): New field.
15511 * input.c (class fcache): Rename to...
15512 (class file_cache_slot): ...this, making most members private and
15513 prefixing fields with "m_".
15514 (file_cache_slot::get_file_path): New accessor.
15515 (file_cache_slot::get_use_count): New accessor.
15516 (file_cache_slot::missing_trailing_newline_p): New accessor.
15517 (file_cache_slot::inc_use_count): New.
15518 (fcache_buffer_size): Move to...
15519 (file_cache_slot::buffer_size): ...here.
15520 (fcache_line_record_size): Move to...
15521 (file_cache_slot::line_record_size): ...here.
15522 (fcache_tab): Delete, in favor of global_dc->m_file_cache.
15523 (fcache_tab_size): Move to file_cache::num_file_slots.
15524 (diagnostic_file_cache_init): Update for move of fcache_tab
15525 to global_dc->m_file_cache.
15526 (diagnostic_file_cache_fini): Likewise.
15527 (lookup_file_in_cache_tab): Convert to...
15528 (file_cache::lookup_file): ...this.
15529 (diagnostics_file_cache_forcibly_evict_file): Update for move of
15530 fcache_tab to global_dc->m_file_cache, moving most of
15531 implementation to...
15532 (file_cache::forcibly_evict_file): ...this new function and...
15533 (file_cache_slot::evict): ...this new function.
15534 (evicted_cache_tab_entry): Convert to...
15535 (file_cache::evicted_cache_tab_entry): ...this.
15536 (add_file_to_cache_tab): Convert to...
15537 (file_cache::add_file): ...this, moving bulk of implementation
15539 (file_cache_slot::create): ..this new function.
15540 (file_cache::file_cache): New.
15541 (file_cache::~file_cache): New.
15542 (lookup_or_add_file_to_cache_tab): Convert to...
15543 (file_cache::lookup_or_add_file): ..this new function.
15544 (fcache::fcache): Rename to...
15545 (file_cache_slot::file_cache_slot): ...this, adding "m_" prefixes
15547 (fcache::~fcache): Rename to...
15548 (file_cache_slot::~file_cache_slot): ...this, adding "m_" prefixes
15550 (needs_read): Convert to...
15551 (file_cache_slot::needs_read_p): ...this.
15552 (needs_grow): Convert to...
15553 (file_cache_slot::needs_grow_p): ...this.
15554 (maybe_grow): Convert to...
15555 (file_cache_slot::maybe_grow): ...this.
15556 (read_data): Convert to...
15557 (file_cache_slot::read_data): ...this.
15558 (maybe_read_data): Convert to...
15559 (file_cache_slot::maybe_read_data): ...this.
15560 (get_next_line): Convert to...
15561 (file_cache_slot::get_next_line): ...this.
15562 (goto_next_line): Convert to...
15563 (file_cache_slot::goto_next_line): ...this.
15564 (read_line_num): Convert to...
15565 (file_cache_slot::read_line_num): ...this.
15566 (location_get_source_line): Update for moving of globals to
15567 global_dc->m_file_cache.
15568 (location_missing_trailing_newline): Likewise.
15569 * input.h (class file_cache_slot): New forward decl.
15570 (class file_cache): New.
15572 2021-07-01 Michael Meissner <meissner@linux.ibm.com>
15574 * config/rs6000/rs6000.c (rs6000_maybe_emit_fp_cmove): Add IEEE
15575 128-bit floating point conditional move support.
15576 (have_compare_and_set_mask): Add IEEE 128-bit floating point
15578 * config/rs6000/rs6000.md (mov<mode>cc, IEEE128 iterator): New insn.
15579 (mov<mode>cc_p10, IEEE128 iterator): New insn.
15580 (mov<mode>cc_invert_p10, IEEE128 iterator): New insn.
15581 (fpmask<mode>, IEEE128 iterator): New insn.
15582 (xxsel<mode>, IEEE128 iterator): New insn.
15584 2021-07-01 Iain Sandoe <iain@sandoe.co.uk>
15587 * config/darwin.h (CTF_INFO_SECTION_NAME): New.
15589 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
15591 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
15593 * config/i386/i386-protos.h (ix86_expand_vector_init_duplicate):
15595 * config/i386/sse.md (INT_BROADCAST_MODE): New mode iterator.
15596 (vec_duplicate<mode>): New expander.
15598 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
15601 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
15603 (ix86_byte_broadcast): New function.
15604 (ix86_convert_const_wide_int_to_broadcast): Likewise.
15605 (ix86_expand_move): Convert CONST_WIDE_INT to broadcast if mode
15606 size is 16 bytes or bigger.
15607 (ix86_broadcast_from_integer_constant): New function.
15608 (ix86_expand_vector_move): Convert CONST_WIDE_INT and CONST_VECTOR
15609 to broadcast if mode size is 16 bytes or bigger.
15610 * config/i386/i386-protos.h (ix86_gen_scratch_sse_rtx): New
15612 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): New function.
15614 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
15616 * config/i386/predicates.md (ix86_endbr_immediate_operand):
15617 Return true/false instead of 1/0.
15618 (movq_parallel): Ditto.
15620 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
15622 * recog.c (general_operand): Return true/false instead of 1/0.
15623 (register_operand): Ditto.
15624 (immediate_operand): Ditto.
15625 (const_int_operand): Ditto.
15626 (const_scalar_int_operand): Ditto.
15627 (const_double_operand): Ditto.
15628 (push_operand): Ditto.
15629 (pop_operand): Ditto.
15630 (memory_operand): Ditto.
15631 (indirect_operand): Ditto.
15633 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
15635 * genpreds.c (write_predicate_subfunction):
15636 Change the type of written subfunction to bool.
15637 (write_one_predicate_function):
15638 Change the type of written function to bool.
15639 (write_tm_preds_h): Ditto.
15640 * recog.h (*insn_operand_predicate_fn): Change the type to bool.
15641 * recog.c (general_operand): Change the type to bool.
15642 (address_operand): Ditto.
15643 (register_operand): Ditto.
15644 (pmode_register_operand): Ditto.
15645 (scratch_operand): Ditto.
15646 (immediate_operand): Ditto.
15647 (const_int_operand): Ditto.
15648 (const_scalar_int_operand): Ditto.
15649 (const_double_operand): Ditto.
15650 (nonimmediate_operand): Ditto.
15651 (nonmemory_operand): Ditto.
15652 (push_operand): Ditto.
15653 (pop_operand): Ditto.
15654 (memory_operand): Ditto.
15655 (indirect_operand): Ditto.
15656 (ordered_comparison_operator): Ditto.
15657 (comparison_operator): Ditto.
15658 * config/i386/i386-expand.c (ix86_expand_sse_cmp):
15659 Change the type of indirect predicate function to bool.
15660 * config/rs6000/rs6000.c (easy_vector_constant):
15661 Change the type to bool.
15662 * config/mips/mips-protos.h (m16_based_address_p):
15663 Change the type of operand 3 to bool.
15665 2021-07-01 Richard Biener <rguenther@suse.de>
15667 PR tree-optimization/101280
15668 PR tree-optimization/101173
15669 * gimple-loop-interchange.cc
15670 (tree_loop_interchange::valid_data_dependences): Revert
15671 previous change and instead correctly handle DDR_REVERSED_P
15674 2021-07-01 Richard Biener <rguenther@suse.de>
15676 PR tree-optimization/101278
15677 * tree-ssa-dse.c (dse_classify_store): First check for
15678 uses, then ignore stmt for chaining purposes.
15680 2021-07-01 Richard Biener <rguenther@suse.de>
15682 PR tree-optimization/100778
15683 * tree-vect-slp.c (vect_schedule_slp_node): Do not place trapping
15684 vectorized ops ahead of their scalar BB.
15686 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
15689 * config/i386/i386.md (*nabs<dwi>2_doubleword):
15690 New insn_and_split pattern.
15691 (*nabs<dwi>2_1): Ditto.
15692 * config/i386/i386-features.c
15693 (general_scalar_chain::compute_convert_gain):
15694 Handle (NEG (ABS (...))) RTX. Rewrite src code
15695 scanner as switch statement.
15696 (general_scalar_chain::convert_insn):
15697 Handle (NEG (ABS (...))) RTX.
15698 (general_scalar_to_vector_candidate_p):
15699 Detect (NEG (ABS (...))) RTX. Reorder case statements
15700 for (AND (NOT (...) ...)) fallthrough.
15702 2021-07-01 Richard Biener <rguenther@suse.de>
15704 PR tree-optimization/101178
15705 * tree-vect-slp.c (slpg_vertex::materialize): Remove.
15706 (slpg::perm_in): Add.
15707 (slpg::get_perm_in): Remove.
15708 (slpg::get_perm_materialized): Add.
15709 (vect_optimize_slp): Handle VEC_PERM nodes more optimally
15710 during permute propagation and materialization.
15712 2021-07-01 Jakub Jelinek <jakub@redhat.com>
15715 * dwarf2out.c (loc_list_from_tree_1): Handle COMPOUND_LITERAL_EXPR.
15717 2021-07-01 Jakub Jelinek <jakub@redhat.com>
15719 PR middle-end/94366
15720 * omp-low.c (lower_rec_input_clauses): Rename is_fp_and_or to
15721 is_truth_op, set it for TRUTH_*IF_EXPR regardless of new_var's type,
15722 use boolean_type_node instead of integer_type_node as NE_EXPR type.
15723 (lower_reduction_clauses): Likewise.
15725 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
15727 * config/gcn/gcn.c: Include dwarf2.h.
15728 (gcn_addr_space_debug): New function.
15729 (TARGET_ADDR_SPACE_DEBUG): New hook.
15731 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
15733 * common/config/gcn/gcn-common.c
15734 (gcn_option_optimization_table): Change OPT_fomit_frame_pointer to -O3.
15735 * config/gcn/gcn.c (gcn_expand_prologue): Prefer the frame pointer
15737 (gcn_expand_prologue): Prefer the frame pointer when emitting CFI.
15738 (gcn_frame_pointer_rqd): New function.
15739 (TARGET_FRAME_POINTER_REQUIRED): New hook.
15741 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
15743 * config/gcn/gcn.c (move_callee_saved_registers): Emit CFI notes for
15744 prologue register saves.
15745 (gcn_debug_unwind_info): Use UI_DWARF2.
15746 (gcn_dwarf_register_number): Map DWARF_LINK_REGISTER to DWARF PC.
15747 (gcn_dwarf_register_span): DWARF_LINK_REGISTER doesn't span.
15748 * config/gcn/gcn.h: (DWARF_FRAME_RETURN_COLUMN): New define.
15749 (DWARF_LINK_REGISTER): New define.
15750 (FIRST_PSEUDO_REGISTER): Increment.
15751 (FIXED_REGISTERS): Add entry for DWARF_LINK_REGISTER.
15752 (CALL_USED_REGISTERS): Likewise.
15753 (REGISTER_NAMES): Likewise.
15755 2021-06-30 Richard Biener <rguenther@suse.de>
15757 PR tree-optimization/101267
15758 * tree-vect-stmts.c (vect_check_scalar_mask): Adjust
15759 API and use SLP compatible interface of vect_is_simple_use.
15760 Reject not vectorized SLP defs for callers that do not support
15762 (vect_check_store_rhs): Handle masked stores and pass down
15763 the appropriate operator index.
15764 (vectorizable_call): Adjust.
15765 (vectorizable_store): Likewise.
15766 (vectorizable_load): Likewise. Handle SLP pecularity of
15768 (vect_is_simple_use): Remove special-casing of masked stores.
15770 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
15772 * common.opt (foffload): Remove help as Driver only.
15773 * gcc.c (display_help): Add -foffload.
15775 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
15777 * gcc.c (close_at_file, execute): Replace alloca by XALLOCAVEC.
15778 (check_offload_target_name): Fix splitting OFFLOAD_TARGETS into
15779 a candidate list; better inform no offload target is configured
15780 and fix hint extraction when passed target is not '\0' at [len].
15781 * common.opt (foffload): Add tailing '.'.
15782 (foffload-options): Likewise; fix flag name in the help string.
15784 2021-06-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
15787 * config/arm/arm_neon.h: Move vabs intrinsics before vcage_f32.
15788 (vcage_f32): Gate comparison on __FAST_MATH__.
15789 (vcageq_f32): Likewise.
15790 (vcale_f32): Likewise.
15791 (vcaleq_f32): Likewise.
15792 (vcagt_f32): Likewise.
15793 (vcagtq_f32): Likewise.
15794 (vcalt_f32): Likewise.
15795 (vcaltq_f32): Likewise.
15796 (vcage_f16): Likewise.
15797 (vcageq_f16): Likewise.
15798 (vcale_f16): Likewise.
15799 (vcaleq_f16): Likewise.
15800 (vcagt_f16): Likewise.
15801 (vcagtq_f16): Likewise.
15802 (vcalt_f16): Likewise.
15803 (vcaltq_f16): Likewise.
15805 2021-06-30 Richard Biener <rguenther@suse.de>
15807 PR tree-optimization/101264
15808 * tree-vect-slp.c (vect_optimize_slp): Propagate the
15809 computed perm_in to all "any" permute successors
15810 we cannot de-duplicate immediately.
15812 2021-06-30 liuhongt <hongtao.liu@intel.com>
15815 * config/i386/sse.md
15816 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>):
15818 (avx512f_sfixupimm<mode><maskz_scalar_name><round_saeonly_name>):
15820 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>"): Refined.
15821 * config/i386/subst.md (maskz_scalar): New define_subst.
15822 (maskz_scalar_name): New subst_attr.
15823 (maskz_scalar_op5): Ditto.
15824 (round_saeonly_maskz_scalar_op5): Ditto.
15825 (round_saeonly_maskz_scalar_operand5): Ditto.
15827 2021-06-30 David Edelsohn <dje.gcc@gmail.com>
15829 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
15830 Increase code CSECT alignment to at least 32 bytes.
15831 * config/rs6000/xcoff.h (TEXT_SECTION_ASM_OP): Add 32 byte
15832 alignment designation.
15834 2021-06-29 Sergei Trofimovich <siarheit@google.com>
15836 * doc/generic.texi: Fix s/net yet/not yet/ typo.
15838 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
15840 PR tree-optimization/101254
15841 * range-op.cc (operator_minus::op1_op2_relation_effect): Check for
15842 wrapping/non-wrapping when setting the result range.
15844 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
15846 * value-query.cc (gimple_range_global): Allow phis.
15848 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
15850 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): Use stmt.
15851 (simplify_using_ranges::op_with_boolean_value_range_p): Add a
15852 statement for location context.
15853 (check_for_binary_op_overflow): Ditto.
15854 (simplify_using_ranges::get_vr_for_comparison): Ditto.
15855 (simplify_using_ranges::compare_name_with_value): Ditto.
15856 (simplify_using_ranges::compare_names): Ditto.
15857 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Ditto.
15858 (simplify_using_ranges::simplify_truth_ops_using_ranges): Ditto.
15859 (simplify_using_ranges::simplify_min_or_max_using_ranges): Ditto.
15860 (simplify_using_ranges::simplify_internal_call_using_ranges): Ditto.
15861 (simplify_using_ranges::two_valued_val_range_p): Ditto.
15862 (simplify_using_ranges::simplify): Ditto.
15863 * vr-values.h: Adjust prototypes.
15865 2021-06-29 Uroš Bizjak <ubizjak@gmail.com>
15868 * config/i386/mmx.md (vec_addsubv2sf3): New insn pattern.
15870 2021-06-29 Julian Brown <julian@codesourcery.com>
15872 * config/gcn/gcn.c (gcn_init_libfuncs): New function.
15873 (TARGET_INIT_LIBFUNCS): Define target hook using above function.
15874 * config/gcn/gcn.h (UNITS_PER_WORD): Define to 8 for IN_LIBGCC2, 4
15876 (LIBGCC2_UNITS_PER_WORD, BITS_PER_WORD): Remove definitions.
15877 (MAX_FIXED_MODE_SIZE): Change to 128.
15879 2021-06-29 Julian Brown <julian@codesourcery.com>
15881 * config/gcn/gcn.md (UNSPEC_FLBIT_INT): New unspec constant.
15882 (s_mnemonic): Add clrsb.
15883 (gcn_flbit<mode>_int): Add insn pattern for SImode/DImode.
15884 (clrsb<mode>2): Add expander for SImode/DImode.
15886 2021-06-29 Julian Brown <julian@codesourcery.com>
15888 * config/gcn/gcn.md (<su>mulsidi3, <su>mulsidi3_reg, <su>mulsidi3_imm,
15889 muldi3): Add patterns.
15891 2021-06-29 Julian Brown <julian@codesourcery.com>
15893 * config/gcn/gcn.md (<su>mulsi3_highpart): Change to expander.
15894 (<su>mulsi3_highpart_reg, <su>mulsi3_highpart_imm): New patterns.
15896 2021-06-29 Julian Brown <julian@codesourcery.com>
15898 * config/gcn/gcn.md (mulsi3): Make s_mulk_i32 variant clobber SCC.
15900 2021-06-29 Joseph Myers <joseph@codesourcery.com>
15902 * btfout.c, ctfout.c: Include "memmodel.h".
15904 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
15906 * gcc.c (check_offload_target_name): Cast len argument to
15907 %q.*s to 'int'; avoid -Wstringop-truncation warning.
15909 2021-06-29 Richard Biener <rguenther@suse.de>
15911 * tree-vect-slp.c (vect_optimize_slp): Forward propagate
15912 to "any" permute nodes and relax "any" permute proapgation
15913 during iterative backward propagation.
15915 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
15918 * common.opt (-foffload=): Update description.
15919 (-foffload-options=): New.
15920 * doc/invoke.texi (C Language Options): Document
15921 -foffload and -foffload-options.
15922 * gcc.c (check_offload_target_name): New, split off from
15923 handle_foffload_option.
15924 (check_foffload_target_names): New.
15925 (handle_foffload_option): Handle -foffload=default.
15926 (driver_handle_option): Update for -foffload-options.
15927 * lto-opts.c (lto_write_options): Use -foffload-options
15928 instead of -foffload.
15929 * lto-wrapper.c (merge_and_complain, append_offload_options):
15931 * opts.c (common_handle_option): Likewise.
15933 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
15935 * doc/invoke.texi (C Language Options): Sort options
15936 alphabetically in optlist and also the description itself.
15937 Remove leftover -fallow-single-precision from and add missing
15938 -fgnu-tm to the optlist.
15940 2021-06-29 Richard Biener <rguenther@suse.de>
15942 * tree-vect-slp.c (slpg_vertex::visited): Remove.
15943 (vect_slp_perms_eq): Handle -1 permutes.
15944 (vect_optimize_slp): Rewrite permute propagation.
15946 2021-06-29 Jakub Jelinek <jakub@redhat.com>
15949 * match.pd ((intptr_t)x eq/ne CST to x eq/ne (typeof x) CST): Don't
15950 perform the optimization in GENERIC when sanitizing and x has a
15953 2021-06-29 Richard Biener <rguenther@suse.de>
15955 PR tree-optimization/101242
15956 * tree-vect-slp.c (vect_slp_build_vertices): Force-add
15957 PHIs with not represented initial values as leafs.
15959 2021-06-29 Jan-Benedict Glaw <jbglaw@getslash.de>
15961 * config/pdp11/pdp11.h (ASM_OUTPUT_SKIP): Fix signedness warning.
15962 * config/pdp11/pdp11.c (pdp11_asm_print_operand_punct_valid_p): Remove
15963 "register" keyword.
15964 (pdp11_initial_elimination_offset) Remove unused variable.
15965 (pdp11_cmp_length) Ditto.
15966 (pdp11_insn_cost): Ditto, and fix signedness warning.
15968 2021-06-29 David Edelsohn <dje.gcc@gmail.com>
15970 * btfout.c: Include tm_p.h.
15973 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
15975 * config/bpf/bpf.c (bpf_expand_prologue): Do not mark insns as
15977 (bpf_expand_epilogue): Likewise.
15978 * config/bpf/bpf.h (DWARF2_FRAME_INFO): Define to 0.
15979 Do not define DBX_DEBUGGING_INFO.
15981 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
15983 * doc/invoke.texi: Document the CTF and BTF debug info options.
15985 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
15986 David Faust <david.faust@oracle.com>
15987 Jose E. Marchesi <jose.marchesi@oracle.com>
15988 Weimin Pan <weimin.pan@oracle.com>
15990 * Makefile.in: Add ctfc.*, ctfout.c and btfout.c files to
15991 GTFILES. Add new object files.
15992 * common.opt: Add CTF and BTF debug info options.
15993 * btfout.c: New file.
15994 * ctfc.c: Likewise.
15995 * ctfc.h: Likewise.
15996 * ctfout.c: Likewise.
15997 * dwarf2ctf.c: Likewise.
15998 * dwarf2ctf.h: Likewise.
15999 * dwarf2cfi.c (dwarf2out_do_frame): Acknowledge CTF_DEBUG and
16001 * dwarf2out.c (dwarf2out_source_line): Likewise.
16002 (dwarf2out_finish): Skip emitting DWARF if CTF or BTF are to
16004 (debug_format_do_cu): New function.
16005 (dwarf2out_early_finish): Traverse DIEs and emit CTF/BTF for
16007 Include dwarf2ctf.c.
16008 * final.c (dwarf2_debug_info_emitted_p): Acknowledge DWARF-based debug
16010 * flag-types.h (enum debug_info_type): Add CTF_DEBUG and BTF_DEBUG.
16011 (CTF_DEBUG): New bitmask.
16012 (BTF_DEBUG): Likewise.
16013 (enum ctf_debug_info_levels): New enum.
16014 * gengtype.c (open_base_files): Handle ctfc.h.
16015 (main): Handle uint32_t type.
16016 * flags.h (btf_debuginfo_p): New definition.
16017 (dwarf_based_debuginfo_p): Likewise.
16018 * opts.c (debug_type_names): Add entries for CTF and BTF.
16019 (btf_debuginfo_p): New function.
16020 (dwarf_based_debuginfo_p): Likewise.
16021 (common_handle_option): Handle -gctfN and -gbtf options.
16022 (set_debug_level): Set CTF_DEBUG, BTF_DEBUG whenever appropriate.
16023 * toplev.c (process_options): Inform the user and ignore -gctfLEVEL if
16026 2021-06-28 Jose E. Marchesi <jose.marchesi@oracle.com>
16028 * dwarf2out.c (AT_class): Function is no longer static.
16029 (AT_int): Likewise.
16030 (AT_unsigned): Likewise.
16031 (AT_loc): Likewise.
16032 (get_AT): Likewise.
16033 (get_AT_string): Likewise.
16034 (get_AT_flag): Likewise.
16035 (get_AT_unsigned): Likewise.
16036 (get_AT_ref): Likewise.
16037 (new_die_raw): Likewise.
16038 (lookup_decl_die): Likewise.
16039 (base_type_die): Likewise.
16040 (add_name_attribute): Likewise.
16041 (add_AT_int): Likewise.
16042 (add_AT_unsigned): Likewise.
16043 (add_AT_loc): Likewise.
16044 (dw_get_die_tag): New function.
16045 (dw_get_die_child): Likewise.
16046 (dw_get_die_sib): Likewise.
16047 (struct dwarf_file_data): Move from here to dwarf2out.h
16048 (struct dw_attr_struct): Likewise.
16049 * dwarf2out.h: Analogous changes.
16051 2021-06-28 Martin Jambor <mjambor@suse.cz>
16054 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
16055 members m_dead_stmts and m_dead_ssas.
16056 * ipa-param-manipulation.c
16057 (ipa_param_body_adjustments::mark_dead_statements): New function.
16058 (ipa_param_body_adjustments::common_initialization): Call it on
16059 all removed but not split parameters.
16060 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
16062 (ipa_param_body_adjustments::modify_call_stmt): Remove arguments that
16064 * tree-inline.c (remap_gimple_stmt): Do not copy dead statements, reset
16065 dead debug statements.
16066 (copy_phis_for_bb): Do not copy dead PHI nodes.
16068 2021-06-28 Martin Jambor <mjambor@suse.cz>
16071 * symtab-clones.h (clone_info): Removed member param_adjustments.
16072 * ipa-param-manipulation.h: Adjust initial comment to reflect how we
16073 deal with pass-through splits now.
16074 (ipa_param_performed_split): Removed.
16075 (ipa_param_adjustments::modify_call): Adjusted parameters.
16076 (class ipa_param_body_adjustments): Adjusted parameters of
16077 register_replacement, modify_gimple_stmt and modify_call_stmt.
16078 (ipa_verify_edge_has_no_modifications): Declare.
16079 (ipa_edge_modifications_finalize): Declare.
16080 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Remove
16081 performed_splits processing, pas only edge to padjs->modify_call,
16082 check that call arguments were not modified if they should not have
16084 * cgraphclones.c (cgraph_node::create_clone): Do not copy performed
16086 * ipa-param-manipulation.c (struct pass_through_split_map): New type.
16087 (ipa_edge_modification_info): Likewise.
16088 (ipa_edge_modification_sum): Likewise.
16089 (ipa_edge_modifications): New edge summary.
16090 (ipa_verify_edge_has_no_modifications): New function.
16091 (transitive_split_p): Removed.
16092 (transitive_split_map): Likewise.
16093 (init_transitive_splits): Likewise.
16094 (ipa_param_adjustments::modify_call): Adjusted to use the new edge
16095 summary instead of performed_splits.
16096 (ipa_param_body_adjustments::register_replacement): Drop dummy
16097 parameter, set base_index of the created ipa_param_body_replacement.
16098 (phi_arg_will_live_p): New function.
16099 (ipa_param_body_adjustments::common_initialization): Do not create
16100 IPA_SRA dummy decls.
16101 (simple_tree_swap_info): Removed.
16102 (remap_split_decl_to_dummy): Likewise.
16103 (record_argument_state_1): New function.
16104 (record_argument_state): Likewise.
16105 (ipa_param_body_adjustments::modify_call_stmt): New parameter
16106 orig_stmt. Do not work with dummy decls, save necessary info about
16107 changes to ipa_edge_modifications.
16108 (ipa_param_body_adjustments::modify_gimple_stmt): New parameter
16109 orig_stmt, pass it to modify_call_stmt.
16110 (ipa_param_body_adjustments::modify_cfun_body): Adjust call to
16111 modify_gimple_stmt.
16112 (ipa_edge_modifications_finalize): New function.
16113 * tree-inline.c (remap_gimple_stmt): Pass original statement to
16114 modify_gimple_stmt.
16115 (copy_phis_for_bb): Do not copy dead PHI nodes.
16116 (expand_call_inline): Do not remap performed_splits.
16117 (update_clone_info): Likewise.
16118 * toplev.c: Include ipa-param-manipulation.h.
16119 (toplev::finalize): Call ipa_edge_modifications_finalize.
16121 2021-06-28 Andrew Pinski <apinski@marvell.com>
16123 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Duplicate range
16124 info if we're the only things setting the target PHI.
16125 (value_replacement): Don't duplicate range here.
16126 (minmax_replacement): Likewise.
16128 2021-06-28 Richard Biener <rguenther@suse.de>
16130 PR tree-optimization/101229
16131 * gimple-walk.c (gimple_walk_op): Handle PHIs.
16133 2021-06-28 Martin Liska <mliska@suse.cz>
16135 * config/v850/v850.c (construct_dispose_instruction): Allocate
16137 (construct_prepare_instruction): Likewise.
16139 2021-06-28 Martin Liska <mliska@suse.cz>
16141 * config/v850/v850.c (v850_option_override): Build default
16143 (v850_can_inline_p): New. Allow MASK_PROLOG_FUNCTION to be
16144 ignored for inlining.
16145 (TARGET_CAN_INLINE_P): New.
16147 2021-06-28 Richard Biener <rguenther@suse.de>
16149 PR tree-optimization/101207
16150 * tree-vect-slp.c (vect_optimize_slp): Do BB reduction
16151 permute eliding for load permutations properly.
16153 2021-06-28 Richard Biener <rguenther@suse.de>
16155 PR tree-optimization/101173
16156 * gimple-loop-interchange.cc
16157 (tree_loop_interchange::valid_data_dependences): Disallow outer
16158 loop dependence distance of zero.
16160 2021-06-28 liuhongt <hongtao.liu@intel.com>
16163 * config/i386/sse.md (*avx_cmp<mode>3_lt): New
16164 define_insn_and_split.
16165 (*avx_cmp<mode>3_ltint): Ditto.
16166 (*avx2_pcmp<mode>3_3): Ditto.
16167 (*avx2_pcmp<mode>3_4): Ditto.
16168 (*avx2_pcmp<mode>3_5): Ditto.
16170 2021-06-28 liuhongt <hongtao.liu@intel.com>
16172 * config/i386/i386-builtin.def (IX86_BUILTIN_BLENDVPD256,
16173 IX86_BUILTIN_BLENDVPS256, IX86_BUILTIN_PBLENDVB256,
16174 IX86_BUILTIN_BLENDVPD, IX86_BUILTIN_BLENDVPS,
16175 IX86_BUILTIN_PBLENDVB128): Replace icode with
16177 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold blendv
16179 * config/i386/sse.md (*<sse4_1_avx2>_pblendvb_lt_subreg_not):
16180 New pre_reload splitter.
16182 2021-06-27 Andrew Pinski <apinski@marvell.com>
16184 PR middle-end/101230
16185 * fold-const.c (fold_ternary_loc): Check
16186 the return value of invert_tree_comparison.
16188 2021-06-27 David Edelsohn <dje.gcc@gmail.com>
16190 * config.gcc: Add SPDX License Identifier.
16191 (powerpc-ibm-aix789): Default to aix73.h.
16192 (powerpc-ibm-aix7.2.*.*): New stanza.
16193 * config/rs6000/aix72.h: Add SPDX License Identifier.
16194 * config/rs6000/aix73.h: New file.
16196 2021-06-26 Jason Merrill <jason@redhat.com>
16198 * except.c: #include "dwarf2.h" instead of "dwarf2out.h".
16200 2021-06-26 Andrew Pinski <apinski@marvell.com>
16202 * genmatch.c (lower_cond): Copy for_subst_vec
16203 for the simplify also.
16204 (lower): Swap the order for lower_for and lower_cond.
16206 2021-06-26 Andrew Pinski <apinski@marvell.com>
16208 * tree-ssa-phiopt.c (match_simplify_replacement): Reset
16209 flow senatitive info on the moved ssa set.
16211 2021-06-26 Andrew Pinski <apinski@marvell.com>
16213 * fold-const.c (fold_cond_expr_with_comparison):
16214 Exand arg0 into comp_code, arg00, and arg01.
16215 (fold_ternary_loc): Use invert_tree_comparison
16216 instead of fold_invert_truthvalue for the case
16217 where we have A CMP B ? C : A.
16219 2021-06-25 Martin Sebor <msebor@redhat.com>
16221 PR middle-end/101216
16222 * calls.c (maybe_warn_rdwr_sizes): Use the no_warning constant.
16224 2021-06-25 Jeff Law <jeffreyalaw@gmail.com>
16226 * config/h8300/h8300.c (select_cc_mode): Handle ASHIFTRT and LSHIFTRT.
16228 2021-06-25 Richard Biener <rguenther@suse.de>
16230 PR tree-optimization/101202
16231 * tree-vect-slp.c (vect_optimize_slp): Explicitely handle
16234 2021-06-25 Richard Biener <rguenther@suse.de>
16236 * tree-vect-slp-patterns.c (addsub_pattern::build): Copy
16237 STMT_VINFO_REDUC_DEF from the original representative.
16239 2021-06-25 Martin Sebor <msebor@redhat.com>
16241 * builtins.c (warn_string_no_nul): Replace uses of TREE_NO_WARNING,
16242 gimple_no_warning_p and gimple_set_no_warning with
16243 warning_suppressed_p, and suppress_warning.
16245 (maybe_warn_for_bound): Same.
16246 (warn_for_access): Same.
16247 (check_access): Same.
16248 (expand_builtin_strncmp): Same.
16249 (fold_builtin_varargs): Same.
16250 * calls.c (maybe_warn_nonstring_arg): Same.
16251 (maybe_warn_rdwr_sizes): Same.
16252 * cfgexpand.c (expand_call_stmt): Same.
16253 * cgraphunit.c (check_global_declaration): Same.
16254 * fold-const.c (fold_undefer_overflow_warnings): Same.
16255 (fold_truth_not_expr): Same.
16256 (fold_unary_loc): Same.
16257 (fold_checksum_tree): Same.
16258 * gimple-array-bounds.cc (array_bounds_checker::check_array_ref): Same.
16259 (array_bounds_checker::check_mem_ref): Same.
16260 (array_bounds_checker::check_addr_expr): Same.
16261 (array_bounds_checker::check_array_bounds): Same.
16262 * gimple-expr.c (copy_var_decl): Same.
16263 * gimple-fold.c (gimple_fold_builtin_strcpy): Same.
16264 (gimple_fold_builtin_strncat): Same.
16265 (gimple_fold_builtin_stxcpy_chk): Same.
16266 (gimple_fold_builtin_stpcpy): Same.
16267 (gimple_fold_builtin_sprintf): Same.
16268 (fold_stmt_1): Same.
16269 * gimple-ssa-isolate-paths.c (diag_returned_locals): Same.
16270 * gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Same.
16271 * gimple-ssa-sprintf.c (handle_printf_call): Same.
16272 * gimple-ssa-store-merging.c (imm_store_chain_info::output_merged_store): Same.
16273 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
16274 * gimple-ssa-warn-restrict.h: Adjust declarations.
16275 (maybe_diag_access_bounds): Replace uses of TREE_NO_WARNING,
16276 gimple_no_warning_p and gimple_set_no_warning with
16277 warning_suppressed_p, and suppress_warning.
16278 (check_call): Same.
16279 (check_bounds_or_overlap): Same.
16280 * gimple.c (gimple_build_call_from_tree): Same.
16281 * gimplify.c (gimplify_return_expr): Same.
16282 (gimplify_cond_expr): Same.
16283 (gimplify_modify_expr_complex_part): Same.
16284 (gimplify_modify_expr): Same.
16285 (gimple_push_cleanup): Same.
16286 (gimplify_expr): Same.
16287 * omp-expand.c (expand_omp_for_generic): Same.
16288 (expand_omp_taskloop_for_outer): Same.
16289 * omp-low.c (lower_rec_input_clauses): Same.
16290 (lower_lastprivate_clauses): Same.
16291 (lower_send_clauses): Same.
16292 (lower_omp_target): Same.
16293 * tree-cfg.c (pass_warn_function_return::execute): Same.
16294 * tree-complex.c (create_one_component_var): Same.
16295 * tree-inline.c (remap_gimple_op_r): Same.
16296 (copy_tree_body_r): Same.
16297 (declare_return_variable): Same.
16298 (expand_call_inline): Same.
16299 * tree-nested.c (lookup_field_for_decl): Same.
16300 * tree-sra.c (create_access_replacement): Same.
16301 (generate_subtree_copies): Same.
16302 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Same.
16303 * tree-ssa-forwprop.c (combine_cond_expr_cond): Same.
16304 * tree-ssa-loop-ch.c (ch_base::copy_headers): Same.
16305 * tree-ssa-loop-im.c (execute_sm): Same.
16306 * tree-ssa-phiopt.c (cond_store_replacement): Same.
16307 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
16308 (handle_builtin_strcpy): Same.
16309 (maybe_diag_stxncpy_trunc): Same.
16310 (handle_builtin_stxncpy_strncat): Same.
16311 (handle_builtin_strcat): Same.
16312 * tree-ssa-uninit.c (get_no_uninit_warning): Same.
16313 (set_no_uninit_warning): Same.
16314 (uninit_undefined_value_p): Same.
16315 (warn_uninit): Same.
16316 (maybe_warn_operand): Same.
16317 * tree-vrp.c (compare_values_warnv): Same.
16318 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr): Same.
16319 (test_for_singularity): Same.
16320 * gimple.h (warning_suppressed_p): New function.
16321 (suppress_warning): Same.
16322 (copy_no_warning): Same.
16323 (gimple_set_block): Call gimple_set_location.
16324 (gimple_set_location): Call copy_warning.
16326 2021-06-25 Martin Sebor <msebor@redhat.com>
16328 * tree.h (warning_suppressed_at, copy_warning,
16329 warning_suppressed_p, suppress_warning): New functions.
16331 2021-06-25 Martin Sebor <msebor@redhat.com>
16333 * Makefile.in (OBJS-libcommon): Add diagnostic-spec.o.
16334 * gengtype.c (open_base_files): Add diagnostic-spec.h.
16335 * diagnostic-spec.c: New file.
16336 * diagnostic-spec.h: New file.
16337 * tree.h (no_warning, all_warnings, suppress_warning_at): New
16339 * warning-control.cc: New file.
16341 2021-06-25 liuhongt <hongtao.liu@intel.com>
16344 * config/i386/i386.c (x86_order_regs_for_local_alloc):
16347 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
16349 PR tree-optimization/101189
16350 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Pass
16351 LHS range of condition to postfold routine.
16352 (fold_using_range::postfold_gcond_edges): Only process the TRUE or
16353 FALSE edge if the LHS range supports it being taken.
16354 * gimple-range-fold.h (postfold_gcond_edges): Add range parameter.
16356 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
16358 * value-relation.cc (equiv_oracle::dump): Do not dump NULL blocks.
16359 (relation_oracle::find_relation_block): Check correct bitmap.
16360 (relation_oracle::dump): Do not dump NULL blocks.
16362 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
16364 * gimple-range-cache.cc (ranger_cache::propagate_cache): Call
16365 range_on_edge instead of manually calculating.
16367 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
16369 * range-op.cc: Fix comment.
16371 2021-06-24 Uroš Bizjak <ubizjak@gmail.com>
16374 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
16375 Handle V8QI and V4HI modes.
16376 * config/i386/mmx.md (sse4_1_<any_extend:code>v4qiv4hi2):
16378 (sse4_1_<any_extend:code>v4qiv4hi2): Ditto.
16379 (mmxpackmode): New mode attribute.
16380 (vec_pack_trunc_<mmxpackmode:mode>): New expander.
16381 (mmxunpackmode): New mode attribute.
16382 (vec_unpacks_lo_<mmxunpackmode:mode>): New expander.
16383 (vec_unpacks_hi_<mmxunpackmode:mode>): Ditto.
16384 (vec_unpacku_lo_<mmxunpackmode:mode>): Ditto.
16385 (vec_unpacku_hi_<mmxunpackmode:mode>): Ditto.
16386 * config/i386/i386.md (extsuffix): Move from ...
16387 * config/i386/sse.md: ... here.
16389 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
16391 * dwarf2out.c (dwarf2out_assembly_start): Emit .file 0 marker here..
16392 (dwarf2out_finish): ...instead of here.
16394 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
16396 * configure.ac (--gdwarf-5 option): Use objdump instead of readelf.
16397 (working --gdwarf-4/--gdwarf-5 for all sources): Likewise.
16398 (--gdwarf-4 not refusing generated .debug_line): Adjust for Windows.
16399 * configure: Regenerate.
16401 2021-06-24 Richard Biener <rguenther@suse.de>
16403 * config/i386/sse.md (vec_addsubv4df3, vec_addsubv2df3,
16404 vec_addsubv8sf3, vec_addsubv4sf3): Merge into ...
16405 (vec_addsub<mode>3): ... using a new addsub_cst mode attribute.
16407 2021-06-24 Richard Biener <rguenther@suse.de>
16409 * config/i386/sse.md (avx_addsubv4df3): Rename to
16411 (avx_addsubv8sf3): Rename to vec_addsubv8sf3.
16412 (sse3_addsubv2df3): Rename to vec_addsubv2df3.
16413 (sse3_addsubv4sf3): Rename to vec_addsubv4sf3.
16414 * config/i386/i386-builtin.def: Adjust.
16415 * internal-fn.def (VEC_ADDSUB): New internal optab fn.
16416 * optabs.def (vec_addsub_optab): New optab.
16417 * tree-vect-slp-patterns.c (class addsub_pattern): New.
16418 (slp_patterns): Add addsub_pattern.
16419 * tree-vect-slp.c (vect_optimize_slp): Disable propagation
16420 across CFN_VEC_ADDSUB.
16421 * tree-vectorizer.h (vect_pattern::vect_pattern): Make
16423 * doc/md.texi (vec_addsub<mode>3): Document.
16425 2021-06-24 Jakub Jelinek <jakub@redhat.com>
16427 PR middle-end/101170
16428 * df-scan.c (df_ref_record): For paradoxical big-endian SUBREGs
16429 where regno + subreg_regno_offset wraps around use 0 as starting
16432 2021-06-24 Jakub Jelinek <jakub@redhat.com>
16434 PR middle-end/101172
16435 * stor-layout.c (finish_bitfield_representative): If nextf has
16436 error_mark_node type, set repr type to error_mark_node too.
16438 2021-06-24 Ilya Leoshkevich <iii@linux.ibm.com>
16440 * config/s390/s390.c (s390_function_profiler): Ignore labelno
16442 * config/s390/s390.h (NO_PROFILE_COUNTERS): Define.
16444 2021-06-24 Richard Biener <rguenther@suse.de>
16446 * tree-vect-slp.c (vect_optimize_slp): Do not propagate
16447 across operations that have different semantics on different
16450 2021-06-24 Jakub Jelinek <jakub@redhat.com>
16452 * tree.h (OMP_CLAUSE_MAP_IN_REDUCTION): Document meaning for OpenMP.
16453 * gimplify.c (gimplify_scan_omp_clauses): For OpenMP map clauses
16454 with OMP_CLAUSE_MAP_IN_REDUCTION flag partially defer gimplification
16455 of non-decl OMP_CLAUSE_DECL. For OMP_CLAUSE_IN_REDUCTION on
16456 OMP_TARGET user outer_ctx instead of ctx for placeholders and
16457 initializer/combiner gimplification.
16458 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_MAP_IN_REDUCTION
16459 on target constructs.
16460 (lower_rec_input_clauses): Likewise.
16461 (lower_omp_target): Likewise.
16462 * omp-expand.c (expand_omp_target): Temporarily ignore nowait clause
16463 on target if in_reduction is present.
16465 2021-06-24 Kewen Lin <linkw@linux.ibm.com>
16467 * tree-predcom.c (class pcom_worker): New class.
16468 (release_chain): Renamed to...
16469 (pcom_worker::release_chain): ...this.
16470 (release_chains): Renamed to...
16471 (pcom_worker::release_chains): ...this.
16472 (aff_combination_dr_offset): Renamed to...
16473 (pcom_worker::aff_combination_dr_offset): ...this.
16474 (determine_offset): Renamed to...
16475 (pcom_worker::determine_offset): ...this.
16476 (class comp_ptrs): New class.
16477 (split_data_refs_to_components): Renamed to...
16478 (pcom_worker::split_data_refs_to_components): ...this,
16479 and update with class comp_ptrs.
16480 (suitable_component_p): Renamed to...
16481 (pcom_worker::suitable_component_p): ...this.
16482 (filter_suitable_components): Renamed to...
16483 (pcom_worker::filter_suitable_components): ...this.
16484 (valid_initializer_p): Renamed to...
16485 (pcom_worker::valid_initializer_p): ...this.
16486 (find_looparound_phi): Renamed to...
16487 (pcom_worker::find_looparound_phi): ...this.
16488 (add_looparound_copies): Renamed to...
16489 (pcom_worker::add_looparound_copies): ...this.
16490 (determine_roots_comp): Renamed to...
16491 (pcom_worker::determine_roots_comp): ...this.
16492 (determine_roots): Renamed to...
16493 (pcom_worker::determine_roots): ...this.
16494 (single_nonlooparound_use): Renamed to...
16495 (pcom_worker::single_nonlooparound_use): ...this.
16496 (remove_stmt): Renamed to...
16497 (pcom_worker::remove_stmt): ...this.
16498 (execute_pred_commoning_chain): Renamed to...
16499 (pcom_worker::execute_pred_commoning_chain): ...this.
16500 (execute_pred_commoning): Renamed to...
16501 (pcom_worker::execute_pred_commoning): ...this.
16502 (struct epcc_data): New member worker.
16503 (execute_pred_commoning_cbck): Call execute_pred_commoning
16504 with pcom_worker pointer.
16505 (find_use_stmt): Renamed to...
16506 (pcom_worker::find_use_stmt): ...this.
16507 (find_associative_operation_root): Renamed to...
16508 (pcom_worker::find_associative_operation_root): ...this.
16509 (find_common_use_stmt): Renamed to...
16510 (pcom_worker::find_common_use_stmt): ...this.
16511 (combinable_refs_p): Renamed to...
16512 (pcom_worker::combinable_refs_p): ...this.
16513 (reassociate_to_the_same_stmt): Renamed to...
16514 (pcom_worker::reassociate_to_the_same_stmt): ...this.
16515 (stmt_combining_refs): Renamed to...
16516 (pcom_worker::stmt_combining_refs): ...this.
16517 (combine_chains): Renamed to...
16518 (pcom_worker::combine_chains): ...this.
16519 (try_combine_chains): Renamed to...
16520 (pcom_worker::try_combine_chains): ...this.
16521 (prepare_initializers_chain): Renamed to...
16522 (pcom_worker::prepare_initializers_chain): ...this.
16523 (prepare_initializers): Renamed to...
16524 (pcom_worker::prepare_initializers): ...this.
16525 (prepare_finalizers_chain): Renamed to...
16526 (pcom_worker::prepare_finalizers_chain): ...this.
16527 (prepare_finalizers): Renamed to...
16528 (pcom_worker::prepare_finalizers): ...this.
16529 (tree_predictive_commoning_loop): Renamed to...
16530 (pcom_worker::tree_predictive_commoning_loop): ...this, adjust
16531 some calls and remove some cleanup code.
16532 (tree_predictive_commoning): Adjusted to use pcom_worker instance.
16533 (static variable looparound_phis): Remove.
16534 (static variable name_expansions): Remove.
16536 2021-06-24 Richard Biener <rguenther@suse.de>
16538 * tree-vect-slp.c (slpg_vertex): New struct.
16539 (vect_slp_build_vertices): Adjust.
16540 (vect_optimize_slp): Likewise. Maintain an outgoing permute
16541 and a materialized one.
16543 2021-06-24 Richard Biener <rguenther@suse.de>
16545 PR tree-optimization/101105
16546 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
16547 Only ignore steps when they are equal or scalar order is preserved.
16549 2021-06-24 liuhongt <hongtao.liu@intel.com>
16552 * config/i386/i386-expand.c (ix86_expand_vec_interleave):
16553 Adjust comments for ix86_expand_vecop_qihi2.
16554 (ix86_expand_vecmul_qihi): Renamed to ..
16555 (ix86_expand_vecop_qihi2): Adjust function prototype to
16556 support shift operation, add static to definition.
16557 (ix86_expand_vec_shift_qihi_constant): Add static to definition.
16558 (ix86_expand_vecop_qihi): Call ix86_expand_vecop_qihi2 and
16559 ix86_expand_vec_shift_qihi_constant.
16560 * config/i386/i386-protos.h (ix86_expand_vecmul_qihi): Deleted.
16561 (ix86_expand_vec_shift_qihi_constant): Deleted.
16562 * config/i386/sse.md (VI12_256_512_AVX512VL): New mode
16564 (mulv8qi3): Call ix86_expand_vecop_qihi directly, add
16565 condition TARGET_64BIT.
16566 (mul<mode>3): Ditto.
16567 (<insn><mode>3): Ditto.
16568 (vlshr<mode>3): Extend to support avx512 vlshr.
16569 (v<insn><mode>3): New expander for
16571 (v<insn>v8qi3): Ditto.
16572 (vashrv8hi3<mask_name>): Renamed to ..
16573 (vashr<mode>3): And extend to support V16QImode for avx512.
16574 (vashrv16qi3): Deleted.
16575 (vashrv2di3<mask_name>): Extend expander to support avx512
16578 2021-06-23 Dimitar Dimitrov <dimitar@dinux.eu>
16580 * doc/lto.texi (Design Overview): Update that slim objects are
16583 2021-06-23 Aaron Sawdey <acsawdey@linux.ibm.com>
16585 * config/rs6000/rs6000-cpus.def: Take OPTION_MASK_PCREL_OPT out
16586 of OTHER_POWER10_MASKS so it will not be enabled by default.
16588 2021-06-23 Richard Biener <rguenther@suse.de>
16589 Martin Jambor <mjambor@suse.cz>
16591 * tree-inline.c (setup_one_parameter): Set TREE_READONLY of the
16592 param replacement unconditionally. Adjust comment.
16594 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
16596 * Makefile.in (OBJS): Add gimple-range-fold.o
16597 * gimple-range-fold.cc: New.
16598 * gimple-range-fold.h: New.
16599 * gimple-range-gori.cc (gimple_range_calc_op1): Move to here.
16600 (gimple_range_calc_op2): Ditto.
16601 * gimple-range-gori.h: Move prototypes to here.
16602 * gimple-range.cc: Adjust include files.
16603 (fur_source:fur_source): Relocate to gimple-range-fold.cc.
16604 (fur_source::get_operand): Ditto.
16605 (fur_source::get_phi_operand): Ditto.
16606 (fur_source::query_relation): Ditto.
16607 (fur_source::register_relation): Ditto.
16608 (class fur_edge): Ditto.
16609 (fur_edge::fur_edge): Ditto.
16610 (fur_edge::get_operand): Ditto.
16611 (fur_edge::get_phi_operand): Ditto.
16612 (fur_stmt::fur_stmt): Ditto.
16613 (fur_stmt::get_operand): Ditto.
16614 (fur_stmt::get_phi_operand): Ditto.
16615 (fur_stmt::query_relation): Ditto.
16616 (class fur_depend): Relocate to gimple-range-fold.h.
16617 (fur_depend::fur_depend): Relocate to gimple-range-fold.cc.
16618 (fur_depend::register_relation): Ditto.
16619 (fur_depend::register_relation): Ditto.
16620 (class fur_list): Ditto.
16621 (fur_list::fur_list): Ditto.
16622 (fur_list::get_operand): Ditto.
16623 (fur_list::get_phi_operand): Ditto.
16624 (fold_range): Ditto.
16625 (adjust_pointer_diff_expr): Ditto.
16626 (gimple_range_adjustment): Ditto.
16627 (gimple_range_base_of_assignment): Ditto.
16628 (gimple_range_operand1): Ditto.
16629 (gimple_range_operand2): Ditto.
16630 (gimple_range_calc_op1): Relocate to gimple-range-gori.cc.
16631 (gimple_range_calc_op2): Ditto.
16632 (fold_using_range::fold_stmt): Relocate to gimple-range-fold.cc.
16633 (fold_using_range::range_of_range_op): Ditto.
16634 (fold_using_range::range_of_address): Ditto.
16635 (fold_using_range::range_of_phi): Ditto.
16636 (fold_using_range::range_of_call): Ditto.
16637 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
16638 (fold_using_range::range_of_builtin_call): Ditto.
16639 (fold_using_range::range_of_cond_expr): Ditto.
16640 (fold_using_range::range_of_ssa_name_with_loop_info): Ditto.
16641 (fold_using_range::relation_fold_and_or): Ditto.
16642 (fold_using_range::postfold_gcond_edges): Ditto.
16643 * gimple-range.h: Add gimple-range-fold.h to include files. Change
16644 GIMPLE_RANGE_STMT_H to GIMPLE_RANGE_H.
16645 (gimple_range_handler): Relocate to gimple-range-fold.h.
16646 (gimple_range_ssa_p): Ditto.
16647 (range_compatible_p): Ditto.
16648 (class fur_source): Ditto.
16649 (class fur_stmt): Ditto.
16650 (class fold_using_range): Ditto.
16651 (gimple_range_calc_op1): Relocate to gimple-range-gori.h
16652 (gimple_range_calc_op2): Ditto.
16654 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
16656 PR tree-optimization/101148
16657 PR tree-optimization/101014
16658 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
16659 (ranger_cache::~ranger_cache): Adjust.
16660 (ranger_cache::block_range): Check if propagation disallowed.
16661 (ranger_cache::propagate_cache): Disallow propagation if new value
16662 can't be stored properly.
16663 * gimple-range-cache.h (ranger_cache::m_propfail): New member.
16665 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
16667 * gimple-range-cache.cc (class ssa_block_ranges): Adjust prototype.
16668 (sbr_vector::set_bb_range): Return true.
16669 (class sbr_sparse_bitmap): Adjust.
16670 (sbr_sparse_bitmap::set_bb_range): Return value.
16671 (block_range_cache::set_bb_range): Return value.
16672 (ranger_cache::propagate_cache): Use return value to print msg.
16673 * gimple-range-cache.h (class block_range_cache): Adjust.
16675 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
16677 * gimple-range.cc (dump_bb): Use range_on_edge from the cache.
16679 2021-06-23 Jeff Law <jeffreyalaw@gmail.com>
16681 * config/h8300/logical.md (<code><mode>3<ccnz>): Use <cczn>
16682 so this pattern can be used for test/compare removal. Pass
16683 current insn to compute_logical_op_length and output_logical_op.
16684 * config/h8300/h8300.c (compute_logical_op_cc): Remove.
16685 (h8300_and_costs): Add argument to compute_logical_op_length.
16686 (output_logical_op): Add new argument. Use it to determine if the
16687 condition codes are used and adjust the output accordingly.
16688 (compute_logical_op_length): Add new argument and update length
16689 computations when condition codes are used.
16690 * config/h8300/h8300-protos.h (compute_logical_op_length): Update
16692 (output_logical_op): Likewise.
16694 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
16697 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
16698 Handle 64bit modes for TARGET_XOP. Use indirect gen_* functions.
16699 * config/i386/mmx.md (mmx_ppermv64): New insn pattern.
16700 * config/i386/i386.md (unspec): Move UNSPEC_XOP_PERMUTE from ...
16701 * config/i386/sse.md (unspec): ... here.
16703 2021-06-23 Martin Liska <mliska@suse.cz>
16706 * optc-save-gen.awk: Put back arm_fp16_format to
16709 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
16712 * config/i386/i386.md (bsr_rex64): Add zero-flag setting RTX.
16715 (clz<mode>2): Update RTX pattern for additions.
16717 2021-06-23 Jakub Jelinek <jakub@redhat.com>
16719 PR middle-end/101167
16720 * omp-low.c (lower_omp_regimplify_p): Regimplify also PARM_DECLs
16721 and RESULT_DECLs that have DECL_HAS_VALUE_EXPR_P set.
16723 2021-06-22 Sergei Trofimovich <siarheit@google.com>
16725 * doc/rtl.texi: drop unbalanced parenthesis.
16727 2021-06-22 Richard Biener <rguenther@suse.de>
16729 PR middle-end/101156
16730 * gimplify.c (gimplify_expr): Remove premature incorrect
16733 2021-06-22 Jakub Jelinek <jakub@redhat.com>
16735 PR tree-optimization/101159
16736 * tree-vect-patterns.c (vect_recog_popcount_pattern): Fix some
16739 2021-06-22 Jakub Jelinek <jakub@redhat.com>
16741 PR middle-end/101160
16742 * function.c (assign_parms): For decl_result with TYPE_EMPTY_P type
16743 clear crtl->return_rtx instead of keeping it referencing a pseudo.
16745 2021-06-22 Jakub Jelinek <jakub@redhat.com>
16746 Andrew Pinski <apinski@marvell.com>
16748 PR tree-optimization/101162
16749 * fold-const.c (range_check_type): Handle OFFSET_TYPE like pointer
16752 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16754 * range-op.cc (range_relational_tests): New.
16755 (range_op_tests): Call range_relational_tests.
16757 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16759 * range-op.cc (operator_cast::lhs_op1_relation): New.
16760 (operator_identity::lhs_op1_relation): Mew.
16762 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16764 * range-op.cc (operator_minus::op1_op2_relation_effect): New.
16766 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16768 * range-op.cc (operator_plus::lhs_op1_relation): New.
16769 (operator_plus::lhs_op2_relation): New.
16771 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16773 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a
16774 relation_oracle if dominators exist.
16775 (ranger_cache::~ranger_cache): Dispose of oracle.
16776 (ranger_cache::dump_bb): Dump oracle.
16777 * gimple-range.cc (fur_source::fur_source): New.
16778 (fur_source::get_operand): Use mmeber query.
16779 (fur_source::get_phi_operand): Use member_query.
16780 (fur_source::query_relation): New.
16781 (fur_source::register_dependency): Delete.
16782 (fur_source::register_relation): New.
16783 (fur_edge::fur_edge): Adjust.
16784 (fur_edge::get_phi_operand): Fix comment.
16785 (fur_edge::query): Delete.
16786 (fur_stmt::fur_stmt): Adjust.
16787 (fur_stmt::query): Delete.
16788 (fur_depend::fur_depend): Adjust.
16789 (fur_depend::register_relation): New.
16790 (fur_depend::register_relation): New.
16791 (fur_list::fur_list): Adjust.
16792 (fur_list::get_operand): Use member query.
16793 (fold_using_range::range_of_range_op): Process and query relations.
16794 (fold_using_range::range_of_address): Adjust dependency call.
16795 (fold_using_range::range_of_phi): Ditto.
16796 (gimple_ranger::gimple_ranger): New. Use ranger_ache oracle.
16797 (fold_using_range::relation_fold_and_or): New.
16798 (fold_using_range::postfold_gcond_edges): New.
16799 * gimple-range.h (class gimple_ranger): Adjust.
16800 (class fur_source): Adjust members.
16801 (class fur_stmt): Ditto.
16802 (class fold_using_range): Ditto.
16804 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16806 * range-op.cc (range_operator::wi_fold): Apply relation effect.
16807 (range_operator::fold_range): Adjust and apply relation effect.
16808 (*::fold_range): Add relation parameters.
16809 (*::op1_range): Ditto.
16810 (*::op2_range): Ditto.
16811 (range_operator::lhs_op1_relation): New.
16812 (range_operator::lhs_op2_relation): New.
16813 (range_operator::op1_op2_relation): New.
16814 (range_operator::op1_op2_relation_effect): New.
16815 (relop_early_resolve): New.
16816 (operator_equal::op1_op2_relation): New.
16817 (operator_equal::fold_range): Call relop_early_resolve.
16818 (operator_not_equal::op1_op2_relation): New.
16819 (operator_not_equal::fold_range): Call relop_early_resolve.
16820 (operator_lt::op1_op2_relation): New.
16821 (operator_lt::fold_range): Call relop_early_resolve.
16822 (operator_le::op1_op2_relation): New.
16823 (operator_le::fold_range): Call relop_early_resolve.
16824 (operator_gt::op1_op2_relation): New.
16825 (operator_gt::fold_range): Call relop_early_resolve.
16826 (operator_ge::op1_op2_relation): New.
16827 (operator_ge::fold_range): Call relop_early_resolve.
16828 * range-op.h (class range_operator): Adjust parameters and methods.
16830 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
16832 * Makefile.in (OBJS): Add value-relation.o.
16833 * gimple-range.h: Adjust include files.
16834 * tree-data-ref.c: Adjust include file order.
16835 * value-query.cc (range_query::get_value_range): Default to no oracle.
16836 (range_query::query_relation): New.
16837 (range_query::query_relation): New.
16838 * value-query.h (class range_query): Adjust.
16839 * value-relation.cc: New.
16840 * value-relation.h: New.
16842 2021-06-22 Richard Biener <rguenther@suse.de>
16844 PR tree-optimization/101151
16845 * tree-ssa-sink.c (statement_sink_location): Expand irreducible
16848 2021-06-22 Jojo R <rjiejie@linux.alibaba.com>
16850 * config/riscv/riscv.c (thead_c906_tune_info): New.
16851 (riscv_tune_info_table): Use new tune.
16853 2021-06-22 Richard Biener <rguenther@suse.de>
16855 PR tree-optimization/101158
16856 * tree-vect-slp.c (vect_build_slp_tree_1): Move same operand
16857 checking after checking for matching operation.
16859 2021-06-22 Richard Biener <rguenther@suse.de>
16861 PR tree-optimization/101159
16862 * tree-vect-patterns.c (vect_recog_popcount_pattern): Add
16863 missing NULL vectype check.
16865 2021-06-22 Richard Biener <rguenther@suse.de>
16867 PR tree-optimization/101154
16868 * tree-vect-slp.c (vect_build_slp_tree_2): Fix out-of-bound access.
16870 2021-06-22 Jakub Jelinek <jakub@redhat.com>
16873 * config/i386/i386-protos.h (ix86_last_zero_store_uid): Declare.
16874 * config/i386/i386-expand.c (ix86_last_zero_store_uid): New variable.
16875 * config/i386/i386.c (ix86_expand_prologue): Clear it.
16876 * config/i386/i386.md (peephole2s for 1/2/4 stores of const0_rtx):
16877 Remove "" from match_operand. Emit new insns using emit_move_insn and
16878 set ix86_last_zero_store_uid to INSN_UID of the last store.
16879 Add peephole2s for 1/2/4 stores of const0_rtx following previous
16882 2021-06-22 Martin Liska <mliska@suse.cz>
16884 * auto-profile.c (AUTO_PROFILE_VERSION): Bump as string format
16887 2021-06-22 Martin Liska <mliska@suse.cz>
16889 * gcov-io.h: Remove padding entries.
16891 2021-06-22 liuhongt <hongtao.liu@intel.com>
16893 PR tree-optimization/97770
16894 * tree-vect-patterns.c (vect_recog_popcount_pattern):
16896 (vect_recog_func vect_vect_recog_func_ptrs): Add new pattern.
16898 2021-06-22 liuhongt <hongtao.liu@intel.com>
16901 * config/i386/i386-builtin.def (BDESC): Adjust builtin name.
16902 * config/i386/sse.md (<avx512>_expand<mode>_mask): Rename to ..
16903 (expand<mode>_mask): this ..
16904 (*expand<mode>_mask): New pre_reload splitter to transform
16905 v{,p}expand* to vmov* when mask is zero, all ones, or has all
16906 ones in it's lower part, otherwise still generate
16909 2021-06-22 liuhongt <hongtao.liu@intel.com>
16912 * config/i386/i386-expand.c
16913 (ix86_expand_special_args_builtin): Keep constm1_operand only
16914 if it satisfies insn's operand predicate.
16916 2021-06-21 Jason Merrill <jason@redhat.com>
16919 * df-scan.c (df_ref_record): Check that regno < endregno.
16920 * function.c (assign_parms, expand_function_end): Do nothing with a
16921 TYPE_EMPTY_P result.
16923 2021-06-21 Richard Biener <rguenther@suse.de>
16925 PR tree-optimization/101120
16926 * tree-vect-data-refs.c (bump_vector_ptr): Fold the
16928 * tree-vect-slp.c (vect_transform_slp_perm_load): Add
16929 DR chain DCE capability.
16930 * tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
16931 * tree-vect-stmts.c (vectorizable_load): Remove unused
16932 loads in the DR chain for SLP.
16934 2021-06-21 Jakub Jelinek <jakub@redhat.com>
16936 PR inline-asm/100785
16937 * gimplify.c (gimplify_asm_expr): Don't diagnose errors if
16938 output or input operands were already error_mark_node.
16939 * cfgexpand.c (expand_asm_stmt): If errors are emitted,
16940 remove all inputs, outputs and clobbers from the asm and
16941 set template to "".
16943 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
16945 * config/arm/arm_neon.h (vceq_s8): Replace builtin with __a == __b.
16946 (vceq_s16): Likewise.
16947 (vceq_s32): Likewise.
16948 (vceq_u8): Likewise.
16949 (vceq_u16): Likewise.
16950 (vceq_u32): Likewise.
16951 (vceq_p8): Likewise.
16952 (vceqq_s8): Likewise.
16953 (vceqq_s16): Likewise.
16954 (vceqq_s32): Likewise.
16955 (vceqq_u8): Likewise.
16956 (vceqq_u16): Likewise.
16957 (vceqq_u32): Likewise.
16958 (vceqq_p8): Likewise.
16959 (vceq_f32): Gate __a == __b on __FAST_MATH__.
16960 (vceqq_f32): Likewise.
16961 (vceq_f16): Likewise.
16962 (vceqq_f16): Likewise.
16964 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
16967 * config/arm/iterators.md (NEON_VACMP): Remove.
16968 * config/arm/neon.md (neon_vca<cmp_op><mode>): Use GLTE instead of GTGE
16970 (neon_vca<cmp_op><mode>_insn): Likewise.
16971 (neon_vca<cmp_op_unsp><mode>_insn_unspec): Use NEON_VAGLTE instead of
16974 2021-06-21 Richard Biener <rguenther@suse.de>
16976 PR tree-optimization/101121
16977 * tree-vect-slp.c (vect_build_slp_tree_2): To not fail fatally
16978 when we just lack a stmt with the desired op when doing permutation.
16979 (vect_build_slp_tree): When caching a failed SLP build attempt
16980 assert that at least one lane is marked as not matching.
16982 2021-06-21 liuhongt <hongtao.liu@intel.com>
16985 * config/i386/i386.md: (*anddi_1): Disparage slightly the mask
16986 register alternative.
16987 (*and<mode>_1): Ditto.
16989 (*andn<mode>_1): Ditto.
16990 (*<code><mode>_1): Ditto.
16991 (*<code>qi_1): Ditto.
16992 (*one_cmpl<mode>2_1): Ditto.
16993 (*one_cmplsi2_1_zext): Ditto.
16994 (*one_cmplqi2_1): Ditto.
16995 * config/i386/i386.c (x86_order_regs_for_local_alloc): Change
16996 the order of mask registers to be before general registers.
16998 2021-06-21 Roger Sayle <roger@nextmovesoftware.com>
17001 * config/i386/i386.md: New define_peephole2s to shrink writing
17002 1, 2 or 4 consecutive zeros to memory when optimizing for size.
17004 2021-06-18 Jeff Law <jeffreyalaw@gmail.com>
17006 * config/h8300/h8300.c (h8300_select_cc_mode): Handle SYMBOL_REF.
17007 * config/h8300/logical.md (<code><mode>3 logcial expander): Generate
17008 more efficient code when the source can be trivially simplified.
17010 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
17012 * gimple-range-cache.cc (ranger_cache::range_of_def): Calculate
17013 a range if global is not available.
17014 (ranger_cache::entry_range): Fallback to range_of_def.
17015 * gimple-range-cache.h (range_of_def): Adjust prototype.
17017 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
17019 PR tree-optimization/101014
17020 * gimple-range-cache.cc (ranger_cache::ranger_cache): Remove poor
17022 (ranger_cache::~ranger_cache): Ditto.
17023 (ranger_cache::enable_new_values): Delete.
17024 (ranger_cache::push_poor_value): Delete.
17025 (ranger_cache::range_of_def): Remove poor value processing.
17026 (ranger_cache::entry_range): Ditto.
17027 (ranger_cache::fill_block_cache): Ditto.
17028 * gimple-range-cache.h (class ranger_cache): Remove poor value members.
17029 * gimple-range.cc (gimple_ranger::range_of_expr): Remove call.
17030 * gimple-range.h (class gimple_ranger): Adjust.
17032 2021-06-18 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
17035 * common/config/arm/arm-common.c (arm_canon_arch_option_1): New function
17036 derived from arm_canon_arch.
17037 (arm_canon_arch_option): Call it.
17038 (arm_canon_arch_multilib_option): New function.
17039 * config/arm/arm-cpus.in (IGNORE_FOR_MULTILIB): New fgroup.
17040 * config/arm/arm.h (arm_canon_arch_multilib_option): New prototype.
17041 (CANON_ARCH_MULTILIB_SPEC_FUNCTION): New macro.
17042 (MULTILIB_ARCH_CANONICAL_SPECS): New macro.
17043 (DRIVER_SELF_SPECS): Add MULTILIB_ARCH_CANONICAL_SPECS.
17044 * config/arm/arm.opt (mlibarch): New option.
17045 * config/arm/t-rmprofile (MULTILIB_MATCHES): For armv8*-m, replace use
17046 of march on RHS with mlibarch.
17048 2021-06-18 Marcel Vollweiler <marcel@codesourcery.com>
17050 * config.in: Regenerate.
17051 * config/gcn/gcn.c (print_operand_address): Fix for global_load assembler
17053 * configure: Regenerate.
17054 * configure.ac: Fix for global_load assembler functions.
17056 2021-06-18 Richard Biener <rguenther@suse.de>
17058 PR tree-optimization/101112
17059 * tree-vect-slp.c (vect_slp_linearize_chain): Fix condition
17060 to lookup a pattern stmt def.
17062 2021-06-18 Jakub Jelinek <jakub@redhat.com>
17064 PR middle-end/101062
17065 * stor-layout.c (finish_bitfield_layout): Don't add bitfield
17066 representatives in QUAL_UNION_TYPE.
17068 2021-06-18 Andrew Pinski <apinski@marvell.com>
17070 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
17071 Add counting of how many times it is done.
17072 (factor_out_conditional_conversion): Likewise.
17073 (match_simplify_replacement): Likewise.
17074 (value_replacement): Likewise.
17075 (spaceship_replacement): Likewise.
17076 (cond_store_replacement): Likewise.
17077 (cond_if_else_store_replacement_1): Likewise.
17078 (hoist_adjacent_loads): Likewise.
17080 2021-06-18 Andrew Pinski <apinski@marvell.com>
17082 * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
17083 types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and CNONJ_EXPR.
17084 (verify_gimple_assign_binary): Reject point and offset types on
17085 MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
17086 FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
17087 FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.
17089 2021-06-18 Michael Meissner <meissner@linux.ibm.com>
17091 * config/rs6000/rs6000.c (rs6000_emit_minmax): Add support for ISA
17092 3.1 IEEE 128-bit floating point xsmaxcqp/xsmincqp instructions.
17093 * config/rs6000/rs6000.md (s<minmax><mode>3, IEEE128 iterator):
17096 2021-06-17 Aaron Sawdey <acsawdey@linux.ibm.com>
17098 * config/rs6000/genfusion.pl (gen_logical_addsubf): Add
17099 earlyclobber to alts 0/1.
17100 (gen_addadd): Add earlyclobber to alts 0/1.
17101 * config/rs6000/fusion.md: Regenerate file.
17103 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
17105 * cfgloopanal.c (get_loop_hot_path): Make path an auto_vec.
17107 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
17109 * gimple-range-cache.cc: Comment cleanups.
17110 * gimple-range-gori.cc: Comment cleanups.
17111 * gimple-range.cc: Comment/spacing cleanups
17112 * value-range.h: Comment cleanups.
17114 2021-06-17 H.J. Lu <hjl.tools@gmail.com>
17117 * calls.c (expand_call): Replace PUSH_ARGS with
17118 targetm.calls.push_argument (0).
17119 (emit_library_call_value_1): Likewise.
17120 * defaults.h (PUSH_ARGS): Removed.
17121 (PUSH_ARGS_REVERSED): Replace PUSH_ARGS with
17122 targetm.calls.push_argument (0).
17123 * expr.c (block_move_libcall_safe_for_call_parm): Likewise.
17124 (emit_push_insn): Pass the number bytes to push to
17125 targetm.calls.push_argument and pass 0 if ARGS_ADDR is 0.
17126 * hooks.c (hook_bool_uint_true): New.
17127 * hooks.h (hook_bool_uint_true): Likewise.
17128 * rtlanal.c (nonzero_bits1): Replace PUSH_ARGS with
17129 targetm.calls.push_argument (0).
17130 * target.def (push_argument): Add a targetm.calls hook.
17131 * targhooks.c (default_push_argument): New.
17132 * targhooks.h (default_push_argument): Likewise.
17133 * config/bpf/bpf.h (PUSH_ARGS): Removed.
17134 * config/cr16/cr16.c (TARGET_PUSH_ARGUMENT): New.
17135 * config/cr16/cr16.h (PUSH_ARGS): Removed.
17136 * config/i386/i386.c (ix86_push_argument): New.
17137 (TARGET_PUSH_ARGUMENT): Likewise.
17138 * config/i386/i386.h (PUSH_ARGS): Removed.
17139 * config/m32c/m32c.c (TARGET_PUSH_ARGUMENT): New.
17140 * config/m32c/m32c.h (PUSH_ARGS): Removed.
17141 * config/nios2/nios2.h (PUSH_ARGS): Likewise.
17142 * config/pru/pru.h (PUSH_ARGS): Likewise.
17143 * doc/tm.texi.in: Remove PUSH_ARGS documentation. Add
17144 TARGET_PUSH_ARGUMENT hook.
17145 * doc/tm.texi: Regenerated.
17147 2021-06-17 Uroš Bizjak <ubizjak@gmail.com>
17150 * config/i386/i386-expand.c (expand_vector_set_var):
17151 Handle V2FS mode remapping. Pass TARGET_MMX_WITH_SSE to
17152 ix86_expand_vector_init_duplicate.
17153 (ix86_expand_vector_init_duplicate): Emit insv_1 for
17154 QImode for !TARGET_PARTIAL_REG_STALL.
17155 * config/i386/predicates.md (vec_setm_mmx_operand): New predicate.
17156 * config/i386/mmx.md (vec_setv2sf): Use vec_setm_mmx_operand
17157 as operand 2 predicate. Call ix86_expand_vector_set_var
17158 for non-constant index operand.
17159 (vec_setv2si): Ditto.
17160 (vec_setv4hi): Ditto.
17161 (vec_setv8qi): ditto.
17163 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
17165 PR tree-optimization/100790
17166 * gimple-range.cc (range_of_builtin_call): Cleanup clz and ctz
17169 2021-06-17 Martin Liska <mliska@suse.cz>
17171 * doc/invoke.texi: Use consistently -O1 instead of -O.
17173 2021-06-17 Martin Liska <mliska@suse.cz>
17175 * gcov-io.h: Update documentation entry about string format.
17177 2021-06-17 Marius Hillenbrand <mhillen@linux.ibm.com>
17180 * config/s390/vecintrin.h (vec_doublee): Fix to use
17181 __builtin_s390_vflls.
17182 (vec_floate): Fix to use __builtin_s390_vflrd.
17184 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
17186 * dominance.c (get_dominated_to_depth): Return auto_vec<basic_block>.
17187 * dominance.h (get_dominated_to_depth): Likewise.
17188 (get_all_dominated_blocks): Likewise.
17189 * cfgcleanup.c (delete_unreachable_blocks): Adjust.
17190 * gcse.c (hoist_code): Likewise.
17191 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
17192 * tree-parloops.c (oacc_entry_exit_ok): Likewise.
17193 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Likewise.
17194 * tree-ssa-phiprop.c (pass_phiprop::execute): Likewise.
17196 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
17198 * dominance.c (get_dominated_by_region): Return auto_vec<basic_block>.
17199 * dominance.h (get_dominated_by_region): Likewise.
17200 * tree-cfg.c (gimple_duplicate_sese_region): Adjust.
17201 (gimple_duplicate_sese_tail): Likewise.
17202 (move_sese_region_to_fn): Likewise.
17204 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
17206 * dominance.c (get_dominated_by): Return auto_vec<basic_block>.
17207 * dominance.h (get_dominated_by): Likewise.
17208 * auto-profile.c (afdo_find_equiv_class): Adjust.
17209 * cfgloopmanip.c (duplicate_loop_to_header_edge): Likewise.
17210 * loop-unroll.c (unroll_loop_runtime_iterations): Likewise.
17211 * tree-cfg.c (test_linear_chain): Likewise.
17212 (test_diamond): Likewise.
17214 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
17216 * cfgloop.h (get_loop_hot_path): Return auto_vec<basic_block>.
17217 * cfgloopanal.c (get_loop_hot_path): Likewise.
17218 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
17220 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
17222 * cgraph.c (cgraph_node::collect_callers): Return
17223 auto_vec<cgraph_edge *>.
17224 * cgraph.h (cgraph_node::collect_callers): Likewise.
17225 * ipa-cp.c (create_specialized_node): Adjust.
17226 (decide_about_value): Likewise.
17227 (decide_whether_version_node): Likewise.
17228 * ipa-sra.c (process_isra_node_results): Likewise.
17230 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
17232 * vec.h (vl_ptr>::using_auto_storage): Handle null m_vec.
17233 (auto_vec<T, 0>::auto_vec): Define move constructor, and delete copy
17235 (auto_vec<T, 0>::operator=): Define move assignment and delete copy
17238 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
17240 * gimple-range.cc (debug_seed_ranger): New.
17241 (dump_ranger): New.
17242 (debug_ranger): New.
17244 2021-06-17 Richard Biener <rguenther@suse.de>
17246 PR tree-optimization/54400
17247 * tree-vectorizer.h (enum slp_instance_kind): Add
17248 slp_inst_kind_bb_reduc.
17249 (reduction_fn_for_scalar_code): Declare.
17250 * tree-vect-data-refs.c (vect_slp_analyze_instance_dependence):
17251 Check SLP_INSTANCE_KIND instead of looking at the
17253 (vect_slp_analyze_instance_alignment): Likewise.
17254 * tree-vect-loop.c (reduction_fn_for_scalar_code): Export.
17255 * tree-vect-slp.c (vect_slp_linearize_chain): Split out
17256 chain linearization from vect_build_slp_tree_2 and generalize
17257 for the use of BB reduction vectorization.
17258 (vect_build_slp_tree_2): Adjust accordingly.
17259 (vect_optimize_slp): Elide permutes at the root of BB reduction
17261 (vectorizable_bb_reduc_epilogue): New function.
17262 (vect_slp_prune_covered_roots): Likewise.
17263 (vect_slp_analyze_operations): Use them.
17264 (vect_slp_check_for_constructors): Recognize associatable
17265 chains for BB reduction vectorization.
17266 (vectorize_slp_instance_root_stmt): Generate code for the
17267 BB reduction epilogue.
17269 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
17271 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Check with
17273 (gori_compute::may_recompute_p): New.
17274 (gori_compute::outgoing_edge_range_p): Perform recomputations.
17275 * gimple-range-gori.h (class gori_compute): Add prototype.
17277 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
17279 * gimple-range-cache.cc (ranger_cache::range_on_edge): Always return
17280 true when a range can be calculated.
17281 * gimple-range.cc (gimple_ranger::dump_bb): Check has_edge_range_p.
17283 2021-06-16 Martin Sebor <msebor@redhat.com>
17285 * doc/invoke.texi (-Wmismatched-dealloc, -Wmismatched-new-delete):
17286 Correct documented defaults.
17288 2021-06-16 Andrew MacLeod <amacleod@redhat.com>
17290 * gimple-range-cache.cc (ranger_cache::ranger_cache): Initialize
17291 m_new_value_p directly.
17293 2021-06-16 Uroš Bizjak <ubizjak@gmail.com>
17296 * config/i386/i386-expand.c (expand_vec_perm_2perm_pblendv):
17297 Handle 64bit modes for TARGET_SSE4_1.
17298 (expand_vec_perm_pshufb2): Handle 64bit modes for TARGET_SSSE3.
17299 (expand_vec_perm_even_odd_pack): Handle V4HI mode.
17300 (expand_vec_perm_even_odd_1) <case E_V4HImode>: Expand via
17301 expand_vec_perm_pshufb2 for TARGET_SSSE3 and via
17302 expand_vec_perm_even_odd_pack for TARGET_SSE4_1.
17303 * config/i386/mmx.md (mmx_packusdw): New insn pattern.
17305 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
17307 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn<mode>):
17308 Change to an expander that emits the correct instruction
17309 depending on endianness.
17310 (aarch64_<sur><addsub>hn<mode>_insn_le): Define.
17311 (aarch64_<sur><addsub>hn<mode>_insn_be): Define.
17313 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
17315 * config/aarch64/aarch64-simd-builtins.def: Split generator
17316 for aarch64_<su>qmovn builtins into scalar and vector
17318 * config/aarch64/aarch64-simd.md (aarch64_<su>qmovn<mode>_insn_le):
17320 (aarch64_<su>qmovn<mode>_insn_be): Define.
17321 (aarch64_<su>qmovn<mode>): Split into scalar and vector
17322 variants. Change vector variant to an expander that emits the
17323 correct instruction depending on endianness.
17325 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
17327 * config/aarch64/aarch64-simd-builtins.def: Split generator
17328 for aarch64_sqmovun builtins into scalar and vector variants.
17329 * config/aarch64/aarch64-simd.md (aarch64_sqmovun<mode>):
17330 Split into scalar and vector variants. Change vector variant
17331 to an expander that emits the correct instruction depending
17333 (aarch64_sqmovun<mode>_insn_le): Define.
17334 (aarch64_sqmovun<mode>_insn_be): Define.
17336 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
17338 * config/aarch64/aarch64-simd.md (aarch64_xtn<mode>_insn_le):
17339 Define - modeling zero-high-half semantics.
17340 (aarch64_xtn<mode>): Change to an expander that emits the
17341 appropriate instruction depending on endianness.
17342 (aarch64_xtn<mode>_insn_be): Define - modeling zero-high-half
17344 (aarch64_xtn2<mode>_le): Rename to...
17345 (aarch64_xtn2<mode>_insn_le): This.
17346 (aarch64_xtn2<mode>_be): Rename to...
17347 (aarch64_xtn2<mode>_insn_be): This.
17348 (vec_pack_trunc_<mode>): Emit truncation instruction instead
17350 * config/aarch64/iterators.md (Vnarrowd): Add Vnarrowd mode
17351 attribute iterator.
17353 2021-06-16 Martin Jambor <mjambor@suse.cz>
17355 PR tree-optimization/100453
17356 * tree-sra.c (create_access): Disqualify any const candidates
17357 which are written to.
17358 (sra_modify_expr): Do not store sub-replacements back to a const base.
17359 (handle_unscalarized_data_in_subtree): Likewise.
17360 (sra_modify_assign): Likewise. Earlier, use TREE_READONLy test
17361 instead of constant_decl_p.
17363 2021-06-16 Jakub Jelinek <jakub@redhat.com>
17365 PR middle-end/101062
17366 * stor-layout.c (finish_bitfield_representative): For fields in unions
17367 assume nextf is always NULL.
17368 (finish_bitfield_layout): Compute bit field representatives also in
17369 unions, but handle it as if each bitfield was the only field in the
17372 2021-06-16 Richard Biener <rguenther@suse.de>
17374 PR tree-optimization/101088
17375 * tree-ssa-loop-im.c (sm_seq_valid_bb): Only look for
17376 supported refs on edges. Do not assert same ref but
17377 different kind stores are unsuported but mark them so.
17378 (hoist_memory_references): Only look for supported refs
17381 2021-06-16 Roger Sayle <roger@nextmovesoftware.com>
17383 PR rtl-optimization/46235
17384 * config/i386/i386.md: New define_split for bt followed by cmov.
17385 (*bt<mode>_setcqi): New define_insn_and_split for bt followed by setc.
17386 (*bt<mode>_setncqi): New define_insn_and_split for bt then setnc.
17387 (*bt<mode>_setnc<mode>): New define_insn_and_split for bt followed
17388 by setnc with zero extension.
17390 2021-06-16 Richard Biener <rguenther@suse.de>
17392 PR tree-optimization/101083
17393 * tree-vect-slp.c (vect_slp_build_two_operator_nodes): Get
17394 vectype as argument.
17395 (vect_build_slp_tree_2): Adjust.
17397 2021-06-15 Martin Sebor <msebor@redhat.com>
17399 PR middle-end/100876
17400 * builtins.c: (gimple_call_return_array): Account for size_t
17401 mangling as either unsigned int or unsigned long
17403 2021-06-15 Jeff Law <jeffreyalaw@gmail.com>
17405 * compare-elim.c (try_eliminate_compare): Run DCE to clean things
17406 up before eliminating comparisons.
17408 2021-06-15 Aldy Hernandez <aldyh@redhat.com>
17410 * range-op.cc (operator_bitwise_or::wi_fold): Make sure
17411 nonzero|X is nonzero.
17412 (range_op_bitwise_and_tests): Add tests for above.
17414 2021-06-15 Carl Love <cel@us.ibm.com>
17417 * config/rs6000/rs6000-builtin.def (VCMPEQUT): Fix the ICODE for the
17419 (VRLQ, VSLQ, VSRQ, VSRAQ): Remove unused BU_P10_OVERLOAD_2
17422 2021-06-15 Tobias Burnus <tobias@codesourcery.com>
17425 * gimplify.c (enum gimplify_defaultmap_kind): Add GDMK_SCALAR_TARGET.
17426 (struct gimplify_omp_ctx): Extend defaultmap array by one.
17427 (new_omp_context): Init defaultmap[GDMK_SCALAR_TARGET].
17428 (omp_notice_variable): Update type classification for Fortran.
17429 (gimplify_scan_omp_clauses): Update calls for new argument; handle
17430 GDMK_SCALAR_TARGET; for Fortran, GDMK_POINTER avoid GOVD_MAP_0LEN_ARRAY.
17431 * langhooks-def.h (lhd_omp_scalar_p): Add 'ptr_ok' argument.
17432 * langhooks.c (lhd_omp_scalar_p): Likewise.
17433 (LANG_HOOKS_OMP_ALLOCATABLE_P, LANG_HOOKS_OMP_SCALAR_TARGET_P): New.
17434 (LANG_HOOKS_DECLS): Add them.
17435 * langhooks.h (struct lang_hooks_for_decls): Add new hooks, update
17436 omp_scalar_p pointer type to include the new bool argument.
17438 2021-06-15 David Malcolm <dmalcolm@redhat.com>
17440 * doc/analyzer.texi
17441 (Special Functions for Debugging the Analyzer): Add
17442 __analyzer_dump_capacity.
17444 2021-06-15 Jakub Jelinek <jakub@redhat.com>
17447 * expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
17448 case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.
17450 2021-06-15 Richard Biener <rguenther@suse.de>
17452 * cfgloopanal.c (mark_irreducible_loops): Use a dominance
17453 check to identify loop latches.
17454 * cfgloop.c (verify_loop_structure): Likewise.
17455 * loop-init.c (apply_loop_flags): Allow marked irreducible
17456 regions even with multiple latches.
17457 * predict.c (rebuild_frequencies): Simplify.
17459 2021-06-15 Richard Biener <rguenther@suse.de>
17461 * tree-ssa-threadupdate.c
17462 (jump_thread_path_registry::mark_threaded_blocks): Assert we
17463 have marked irreducible regions.
17465 2021-06-14 Martin Sebor <msebor@redhat.com>
17468 * builtins.c (gimple_call_return_array): Check for attribute fn spec.
17469 Handle calls to placement new.
17470 (ndecl_dealloc_argno): Avoid placement delete.
17472 2021-06-14 Peter Bergner <bergner@linux.ibm.com>
17475 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
17476 create_tmp_reg_or_ssa_name().
17478 2021-06-14 Andrew MacLeod <amacleod@redhat.com>
17480 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
17481 (ranger_cache::enable_new_values): Set to specified value and
17482 return the old value.
17483 (ranger_cache::disable_new_values): Delete.
17484 (ranger_cache::fill_block_cache): Disable non 1st order derived
17486 * gimple-range-cache.h (ranger_cache): Adjust prototypes.
17487 * gimple-range.cc (gimple_ranger::range_of_expr): Adjust.
17489 2021-06-14 Uroš Bizjak <ubizjak@gmail.com>
17492 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
17493 Return true early when testing with V2HImode.
17494 * config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.
17496 2021-06-14 Christophe Lyon <christophe.lyon@linaro.org>
17498 * config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
17499 (mve_vec_unpack<US>_hi_<mode>): New pattern.
17500 (@mve_vec_pack_trunc_lo_<mode>): New pattern.
17501 (mve_vmovntq_<supf><mode>): Prefix with '@'.
17502 * config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
17504 (vec_unpack<US>_lo_<mode>): Likewise.
17505 (vec_pack_trunc_<mode>): Rename to
17506 neon_quad_vec_pack_trunc_<mode>.
17507 * config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
17509 (vec_unpack<US>_lo_<mode>): New.
17510 (vec_pack_trunc_<mode>): New.
17512 2021-06-14 Richard Biener <rguenther@suse.de>
17514 PR tree-optimization/100934
17515 * tree-ssa-dom.c (pass_dominator::execute): Properly
17516 mark irreducible regions.
17518 2021-06-14 Martin Liska <mliska@suse.cz>
17520 * doc/invoke.texi: Put r{...} on the same line as @item.
17522 2021-06-14 Martin Liska <mliska@suse.cz>
17524 * doc/invoke.texi: Add missing newline.
17526 2021-06-14 Martin Liska <mliska@suse.cz>
17528 * doc/invoke.texi: Remove '+' charasters.
17530 2021-06-14 Claudiu Zissulescu <claziss@synopsys.com>
17532 * config.gcc (arc): Add support for with_cpu option.
17533 * config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.
17535 2021-06-14 Richard Biener <rguenther@suse.de>
17537 PR tree-optimization/101031
17538 * tree-ssa-strlen.c (maybe_invalidate): Increment max_size
17539 instead of size when accounting for a possibly string
17542 2021-06-14 Martin Liska <mliska@suse.cz>
17544 * gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].
17546 2021-06-14 Aldy Hernandez <aldyh@redhat.com>
17548 * value-query.cc (gimple_range_global): Call get_range_global
17549 if called after inlining.
17551 2021-06-13 Uroš Bizjak <ubizjak@gmail.com>
17554 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
17555 Emit constant permutation insn directly from here.
17557 2021-06-13 Trevor Saunders <tbsaunde@tbsaunde.org>
17559 * attribs.c (find_attribute_namespace): Iterate over vec<> with
17561 * auto-profile.c (afdo_find_equiv_class): Likewise.
17562 * gcc.c (do_specs_vec): Likewise.
17563 (do_spec_1): Likewise.
17564 (driver::set_up_specs): Likewise.
17565 * gimple-loop-jam.c (any_access_function_variant_p): Likewise.
17566 * gimple-ssa-store-merging.c (compatible_load_p): Likewise.
17567 (imm_store_chain_info::try_coalesce_bswap): Likewise.
17568 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
17569 (get_location_for_stmts): Likewise.
17570 * graphite-poly.c (print_iteration_domains): Likewise.
17571 (free_poly_bb): Likewise.
17572 (remove_gbbs_in_scop): Likewise.
17573 (free_scop): Likewise.
17574 (dump_gbb_cases): Likewise.
17575 (dump_gbb_conditions): Likewise.
17576 (print_pdrs): Likewise.
17577 (print_scop): Likewise.
17578 * ifcvt.c (cond_move_process_if_block): Likewise.
17579 * lower-subreg.c (decompose_multiword_subregs): Likewise.
17580 * regcprop.c (pass_cprop_hardreg::execute): Likewise.
17581 * sanopt.c (sanitize_rewrite_addressable_params): Likewise.
17582 * sel-sched-dump.c (dump_insn_vector): Likewise.
17583 * store-motion.c (store_ops_ok): Likewise.
17584 (store_killed_in_insn): Likewise.
17585 * timevar.c (timer::named_items::print): Likewise.
17586 * tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
17587 (cleanup_tree_cfg_noloop): Likewise.
17588 * tree-data-ref.c (dump_data_references): Likewise.
17589 (print_dir_vectors): Likewise.
17590 (print_dist_vectors): Likewise.
17591 (dump_data_dependence_relations): Likewise.
17592 (dump_dist_dir_vectors): Likewise.
17593 (dump_ddrs): Likewise.
17594 (create_runtime_alias_checks): Likewise.
17595 (free_subscripts): Likewise.
17596 (save_dist_v): Likewise.
17597 (save_dir_v): Likewise.
17598 (invariant_access_functions): Likewise.
17599 (same_access_functions): Likewise.
17600 (access_functions_are_affine_or_constant_p): Likewise.
17601 (find_data_references_in_stmt): Likewise.
17602 (graphite_find_data_references_in_stmt): Likewise.
17603 (free_dependence_relations): Likewise.
17604 (free_data_refs): Likewise.
17605 * tree-inline.c (copy_debug_stmts): Likewise.
17606 * tree-into-ssa.c (dump_currdefs): Likewise.
17607 (rewrite_update_phi_arguments): Likewise.
17608 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
17609 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
17611 (vect_slp_analyze_node_dependences): Likewise.
17612 (vect_slp_analyze_instance_dependence): Likewise.
17613 (vect_record_base_alignments): Likewise.
17614 (vect_get_peeling_costs_all_drs): Likewise.
17615 (vect_peeling_supportable): Likewise.
17616 * tree-vectorizer.c (vec_info::~vec_info): Likewise.
17617 (vec_info::free_stmt_vec_infos): Likewise.
17619 2021-06-13 Jeff Law <jeffreyalaw@gmail.com>
17621 * config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
17622 (andqi3_1<cczn>): Removed.
17623 (<ors>qi3_1): Do not split for IOR/XOR a single bit.
17624 (H8/SX bit logicals): Split out from other patterns.
17625 * config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
17626 mulqihi3_const_clobber_flags.
17627 (mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly
17629 2021-06-13 H.J. Lu <hjl.tools@gmail.com>
17632 * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
17633 to true if red zone is used.
17634 (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
17635 ix86_red_zone_used.
17636 * config/i386/i386.h (machine_function): Add red_zone_used.
17637 (ix86_red_zone_size): Removed.
17638 (ix86_red_zone_used): New.
17639 * config/i386/i386.md (peephole2 patterns): Replace
17640 ix86_red_zone_size with ix86_red_zone_used.
17642 2021-06-12 Jason Merrill <jason@redhat.com>
17644 * doc/extend.texi (unused variable attribute): Applies to
17645 structure fields as well.
17647 2021-06-12 Eugene Rozenfeld <erozen@microsoft.com>
17649 * auto-profile.c (read_profile): fix a typo in an error string
17651 2021-06-11 Thomas Schwinge <thomas@codesourcery.com>
17653 * tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
17655 * tree-pretty-print.c (dump_omp_clauses): Update.
17656 (dump_generic_node) <OMP_CLAUSE>: Use it.
17658 2021-06-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
17661 * config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
17662 int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
17663 the polymorphic variants matching code.
17664 (__arm_vld1q_z): Likewise.
17665 (__arm_vld2q): Likewise.
17666 (__arm_vld4q): Likewise.
17667 (__arm_vldrbq_gather_offset): Likewise.
17668 (__arm_vldrbq_gather_offset_z): Likewise.
17670 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
17672 PR tree-optimization/96392
17673 * fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.
17675 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
17677 PR tree-optimization/96392
17678 * fold-const.c (fold_real_zero_addition_p): Take both arguments
17679 of the addition or subtraction, not just the zero. Use this
17680 other argument in tests for signaling NaNs and signed zeros.
17681 (tree_expr_maybe_real_minus_zero_p): New predicate.
17682 * fold-const.h (fold_real_zero_addition_p): Update prototype.
17683 (tree_expr_maybe_real_minus_zero_p): New function prototype.
17684 * match.pd: Update calls to fold_real_zero_addition_p.
17685 Replace HONOR_NANS with tree_expr_maybe_nan_p.
17686 Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
17687 Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
17688 * tree-ssa-reassoc.c (eliminate_using_constants): Update
17689 call to fold_real_zero_addition_p.
17691 2021-06-11 Richard Biener <rguenther@suse.de>
17693 PR tree-optimization/101025
17694 * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
17695 all refs that require dependence checking.
17697 2021-06-11 Richard Biener <rguenther@suse.de>
17699 PR tree-optimization/101028
17700 * tree-vect-slp.c (vect_build_slp_tree_2): When SLP
17701 reassoc discovery fails fatally, mark appropriate lanes
17704 2021-06-11 Richard Biener <rguenther@suse.de>
17706 PR tree-optimization/101026
17707 * tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
17708 have a representative for the associated chain nodes.
17710 2021-06-11 Jakub Jelinek <jakub@redhat.com>
17712 PR rtl-optimization/101008
17713 * simplify-rtx.c (relational_result): New function.
17714 (simplify_logical_relational_operation,
17715 simplify_relational_operation): Use it.
17717 2021-06-11 Jakub Jelinek <jakub@redhat.com>
17720 * config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.
17722 2021-06-11 Uroš Bizjak <ubizjak@gmail.com>
17725 * config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
17726 false if the permutation can be implemented with constant
17727 permutation instruction in wider mode.
17728 (canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
17729 Handle V8QImode and V4HImode.
17731 2021-06-11 Martin Liska <mliska@suse.cz>
17733 PR gcov-profile/100788
17734 * common.opt: Add new option.
17735 * coverage.c (coverage_begin_function): Emit warning instead on
17736 the internal compiler error.
17737 * doc/invoke.texi: Document the option.
17738 * toplev.c (process_options): Enable it by default.
17740 2021-06-11 Richard Biener <rguenther@suse.de>
17742 PR middle-end/101009
17743 * tree-data-ref.c (build_classic_dist_vector_1): Make sure
17744 to set *init_b to true when we encounter a constant equal
17746 (compute_affine_dependence): Also dump the actual DR_REF.
17748 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
17750 PR tree-optimization/100984
17751 * gimple-ssa-evrp.c (ssa_equiv_stack): Use auto_vec for
17752 replacements table.
17753 (ssa_equiv_stack::~ssa_equiv_stack): Remove.
17755 2021-06-11 Kewen Lin <linkw@linux.ibm.com>
17757 * config/rs6000/rs6000.md
17758 (floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
17759 define_insn_and_split.
17761 2021-06-11 Richard Biener <rguenther@suse.de>
17763 * tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
17764 to sort operands of the associative chain.
17766 2021-06-11 Richard Biener <rguenther@suse.de>
17768 * system.h (gcc_stablesort_r): Declare.
17769 * sort.cc (gcc_sort_r): Support stable sort.
17770 (gcc_stablesort_r): Define.
17771 * vec.h (vec<>::stablesort): Add.
17773 2021-06-10 Uroš Bizjak <ubizjak@gmail.com>
17776 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
17777 Handle V2SF mode. Emit SHUFPS to fixup unpack-high for V2SF mode.
17778 (expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
17779 (expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
17780 (expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
17781 (expand_vec_perm_interleave2): Handle 64bit modes.
17782 (expand_vec_perm_even_odd_pack): Handle V8QI mode.
17783 (expand_vec_perm_even_odd_1): Ditto.
17784 (ix86_vectorize_vec_perm_const): Ditto.
17785 * config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
17786 * config/i386/sse.md: ... here.
17787 * config/i386/mmx.md (*vec_interleave_lowv2sf):
17788 New insn_and_split pattern.
17789 (*vec_interleave_highv2sf): Ditto.
17790 (mmx_pshufbv8qi3): New insn pattern.
17791 (*mmx_pblendw): Ditto.
17793 2021-06-10 Peter Bergner <bergner@linux.ibm.com>
17795 * config/rs6000/rs6000-builtin.def (build_pair): New built-in.
17796 (build_acc): Likewise.
17797 * config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
17798 source operands in little-endian mode.
17799 (rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
17800 (mma_init_builtins): Likewise.
17801 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
17802 ordering for the MMA assemble and build source operands.
17803 * doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
17805 (__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
17808 2021-06-10 Jeff Law <jeffreyalaw@gmail.com>
17810 * config/h8300/h8300.c (select_cc_mode): Handle MEM. Use
17812 * config/h8300/extensions.md: Replace _clobber_flags patterns
17815 2021-06-10 Robin Dapp <rdapp@linux.ibm.com>
17817 * config/s390/vector.md (vcond_mask_<mode><mode>): Change to
17818 (vcond_mask_<mode><tointvec>): this.
17820 2021-06-10 Andrew Stubbs <ams@codesourcery.com>
17821 Thomas Schwinge <thomas@codesourcery.com>
17823 * omp-builtins.def (BUILT_IN_GOACC_ENTER_EXIT_DATA): Split into...
17824 (BUILT_IN_GOACC_ENTER_DATA, BUILT_IN_GOACC_EXIT_DATA): ... these.
17825 * gimple.h (enum gf_mask): Split
17826 'GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA' into
17827 'GF_OMP_TARGET_KIND_OACC_ENTER_DATA' and
17828 'GF_OMP_TARGET_KIND_OACC_EXIT_DATA'.
17829 (is_gimple_omp_oacc): Update.
17830 * gimple-pretty-print.c (dump_gimple_omp_target): Likewise.
17831 * gimplify.c (gimplify_omp_target_update): Likewise.
17832 * omp-expand.c (expand_omp_target, build_omp_regions_1)
17833 (omp_make_gimple_edges): Likewise.
17834 * omp-low.c (check_omp_nesting_restrictions, lower_omp_target):
17837 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
17839 * value-query.cc (value_query::value_on_edge): Rename name to
17841 (range_query::range_on_edge): Same.
17842 (range_query::value_of_expr): Same.
17843 (range_query::value_on_edge): Same.
17844 * value-query.h (class value_query): Same.
17845 (class range_query): Same.
17847 2021-06-10 Richard Biener <rguenther@suse.de>
17849 PR tree-optimization/101003
17850 * tree-vect-slp.c (vect_build_slp_tree_2): Appropriately
17851 use the pattern stmt defs when linearizing a chain.
17853 2021-06-10 Jakub Jelinek <jakub@redhat.com>
17856 * ifcvt.c (noce_get_alt_condition, noce_try_abs): Use
17857 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
17859 2021-06-10 Clement Chigot <clement.chigot@atos.net>
17861 * config/rs6000/aix71.h (ASM_CPU_SPEC): Add Power10 directive.
17862 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
17864 2021-06-09 Andrew Pinski <apinski@marvell.com>
17866 PR tree-optimization/100925
17867 * match.pd (a ? CST1 : CST2): Limit transformations
17868 that would produce a negative to integeral types only.
17869 Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.
17871 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
17874 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
17876 * doc/tm.texi: Correctly update.
17878 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
17880 * doc/tm.texi: Correctly update.
17882 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
17885 * doc/tm.texi.in (Trampolines): Add a missing blank line.
17887 2021-06-09 Paul Eggert <eggert@cs.ucla.edu>
17890 * doc/invoke.texi (Code Gen Options); Document that -fno-trampolines
17891 and -ftrampolines work only with Ada.
17892 * doc/tm.texi.in (Trampolines): Likewise.
17893 * doc/tm.texi: Regenerated.
17895 2021-06-09 Carl Love <cel@us.ibm.com>
17897 * config/rs6000/altivec.h (vec_signextll, vec_signexti, vec_signextq):
17898 Add define for new builtins.
17899 * config/rs6000/altivec.md(altivec_vreveti2): Add define_expand.
17900 * config/rs6000/rs6000-builtin.def (VSIGNEXTI, VSIGNEXTLL): Add
17901 overloaded builtin definitions.
17902 (VSIGNEXTSB2W, VSIGNEXTSH2W, VSIGNEXTSB2D, VSIGNEXTSH2D,VSIGNEXTSW2D,
17903 VSIGNEXTSD2Q): Add builtin expansions.
17904 (SIGNEXT): Add P10 overload definition.
17905 * config/rs6000/rs6000-call.c (P9V_BUILTIN_VEC_VSIGNEXTI, P9V_BUILTIN_VEC_VSIGNEXTLL,
17906 P10_BUILTIN_VEC_SIGNEXT): Add overloaded argument definitions.
17907 * config/rs6000/vsx.md (vsx_sign_extend_v2di_v1ti): Add define_insn.
17908 (vsignextend_v2di_v1ti, vsignextend_qi_<mode>, vsignextend_hi_<mode>,
17909 vsignextend_si_v2di)[VIlong]: Add define_expand.
17910 Make define_insn vsx_sign_extend_si_v2di visible.
17911 * doc/extend.texi: Add documentation for the vec_signexti,
17912 vec_signextll builtins and vec_signextq.
17914 2021-06-09 Carl Love <cel@us.ibm.com>
17916 * config/rs6000/rs6000.c (__fixkfti, __fixunskfti, __floattikf,
17917 __floatuntikf): Names changed to __fixkfti_sw, __fixunskfti_sw,
17918 __floattikf_sw, __floatuntikf_sw respectively.
17919 * config/rs6000/rs6000.md (floatti<mode>2, floatunsti<mode>2,
17920 fix_trunc<mode>ti2, fixuns_trunc<mode>ti2): Add
17921 define_insn for mode IEEE 128.
17923 2021-06-09 Carl Love <cel@us.ibm.com>
17925 * config/rs6000/altivec.md (altivec_vslq, altivec_vsrq):
17926 Rename to altivec_vslq_<mode>, altivec_vsrq_<mode>, mode VEC_TI.
17927 * config/rs6000/vector.md (VEC_TI): Was named VSX_TI in vsx.md.
17928 (vashlv1ti3): Change to vashl<mode>3, mode VEC_TI.
17929 (vlshrv1ti3): Change to vlshr<mode>3, mode VEC_TI.
17930 * config/rs6000/vsx.md (VSX_TI): Remove define_mode_iterator. Update
17931 uses of VSX_TI to VEC_TI.
17933 2021-06-09 Carl Love <cel@us.ibm.com>
17935 * config/rs6000/dfp.md (floattitd2, fixtdti2): New define_insns.
17937 2021-06-09 Carl Love <cel@us.ibm.com>
17939 * config/rs6000/altivec.h (vec_dive, vec_mod): Add define for new
17941 * config/rs6000/altivec.md (UNSPEC_VMULEUD, UNSPEC_VMULESD,
17942 UNSPEC_VMULOUD, UNSPEC_VMULOSD): New unspecs.
17943 (altivec_eqv1ti, altivec_gtv1ti, altivec_gtuv1ti, altivec_vmuleud,
17944 altivec_vmuloud, altivec_vmulesd, altivec_vmulosd, altivec_vrlq,
17945 altivec_vrlqmi, altivec_vrlqmi_inst, altivec_vrlqnm,
17946 altivec_vrlqnm_inst, altivec_vslq, altivec_vsrq, altivec_vsraq,
17947 altivec_vcmpequt_p, altivec_vcmpgtst_p, altivec_vcmpgtut_p): New
17949 (vec_widen_umult_even_v2di, vec_widen_smult_even_v2di,
17950 vec_widen_umult_odd_v2di, vec_widen_smult_odd_v2di, altivec_vrlqmi,
17951 altivec_vrlqnm): New define_expands.
17952 * config/rs6000/rs6000-builtin.def (VCMPEQUT_P, VCMPGTST_P,
17953 VCMPGTUT_P): Add macro expansions.
17954 (BU_P10V_AV_P): Add builtin predicate definition.
17955 (VCMPGTUT, VCMPGTST, VCMPEQUT, CMPNET, CMPGE_1TI,
17956 CMPGE_U1TI, CMPLE_1TI, CMPLE_U1TI, VNOR_V1TI_UNS, VNOR_V1TI, VCMPNET_P,
17957 VCMPAET_P, VMULEUD, VMULESD, VMULOUD, VMULOSD, VRLQ,
17958 VSLQ, VSRQ, VSRAQ, VRLQNM, DIV_V1TI, UDIV_V1TI, DIVES_V1TI, DIVEU_V1TI,
17959 MODS_V1TI, MODU_V1TI, VRLQMI): New macro expansions.
17960 (VRLQ, VSLQ, VSRQ, VSRAQ, DIVE, MOD): New overload expansions.
17961 * config/rs6000/rs6000-call.c (P10_BUILTIN_VCMPEQUT,
17962 P10V_BUILTIN_CMPGE_1TI, P10V_BUILTIN_CMPGE_U1TI,
17963 P10V_BUILTIN_VCMPGTUT, P10V_BUILTIN_VCMPGTST,
17964 P10V_BUILTIN_CMPLE_1TI, P10V_BUILTIN_VCMPLE_U1TI,
17965 P10V_BUILTIN_DIV_V1TI, P10V_BUILTIN_UDIV_V1TI,
17966 P10V_BUILTIN_VMULESD, P10V_BUILTIN_VMULEUD,
17967 P10V_BUILTIN_VMULOSD, P10V_BUILTIN_VMULOUD,
17968 P10V_BUILTIN_VNOR_V1TI, P10V_BUILTIN_VNOR_V1TI_UNS,
17969 P10V_BUILTIN_VRLQ, P10V_BUILTIN_VRLQMI,
17970 P10V_BUILTIN_VRLQNM, P10V_BUILTIN_VSLQ,
17971 P10V_BUILTIN_VSRQ, P10V_BUILTIN_VSRAQ,
17972 P10V_BUILTIN_VCMPGTUT_P, P10V_BUILTIN_VCMPGTST_P,
17973 P10V_BUILTIN_VCMPEQUT_P, P10V_BUILTIN_VCMPGTUT_P,
17974 P10V_BUILTIN_VCMPGTST_P, P10V_BUILTIN_CMPNET,
17975 P10V_BUILTIN_VCMPNET_P, P10V_BUILTIN_VCMPAET_P,
17976 P10V_BUILTIN_DIVES_V1TI, P10V_BUILTIN_MODS_V1TI,
17977 P10V_BUILTIN_MODU_V1TI):
17978 New overloaded definitions.
17979 (rs6000_gimple_fold_builtin) [P10V_BUILTIN_VCMPEQUT,
17980 P10V_BUILTIN_CMPNET, P10V_BUILTIN_CMPGE_1TI,
17981 P10V_BUILTIN_CMPGE_U1TI, P10V_BUILTIN_VCMPGTUT,
17982 P10V_BUILTIN_VCMPGTST, P10V_BUILTIN_CMPLE_1TI,
17983 P10V_BUILTIN_CMPLE_U1TI]: New case statements.
17984 (rs6000_init_builtins) [bool_V1TI_type_node, int_ftype_int_v1ti_v1ti]:
17986 (altivec_init_builtins): New E_V1TImode case statement.
17987 (builtin_function_type)[P10_BUILTIN_128BIT_VMULEUD,
17988 P10_BUILTIN_128BIT_VMULOUD, P10_BUILTIN_128BIT_DIVEU_V1TI,
17989 P10_BUILTIN_128BIT_MODU_V1TI, P10_BUILTIN_CMPGE_U1TI,
17990 P10_BUILTIN_VCMPGTUT, P10_BUILTIN_VCMPEQUT]: New case statements.
17991 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute) [E_TImode,
17992 E_V1TImode]: New case statements.
17993 * config/rs6000/rs6000.h (rs6000_builtin_type_index): New enum
17994 value RS6000_BTI_bool_V1TI.
17995 * config/rs6000/vector.md (vector_gtv1ti,vector_nltv1ti,
17996 vector_gtuv1ti, vector_nltuv1ti, vector_ngtv1ti, vector_ngtuv1ti,
17997 vector_eq_v1ti_p, vector_ne_v1ti_p, vector_ae_v1ti_p,
17998 vector_gt_v1ti_p, vector_gtu_v1ti_p, vrotlv1ti3, vashlv1ti3,
17999 vlshrv1ti3, vashrv1ti3): New define_expands.
18000 * config/rs6000/vsx.md (UNSPEC_VSX_DIVSQ, UNSPEC_VSX_DIVUQ,
18001 UNSPEC_VSX_DIVESQ, UNSPEC_VSX_DIVEUQ, UNSPEC_VSX_MODSQ,
18002 UNSPEC_VSX_MODUQ): New unspecs.
18003 (mulv2di3, vsx_div_v1ti, vsx_udiv_v1ti, vsx_dives_v1ti,
18004 vsx_diveu_v1ti, vsx_mods_v1ti, vsx_modu_v1ti, xxswapd_v1ti): New
18006 (vcmpnet): New define_expand.
18007 * doc/extend.texi: Add documentation for the new builtins vec_rl,
18008 vec_rlmi, vec_rlnm, vec_sl, vec_sr, vec_sra, vec_mule, vec_mulo,
18009 vec_div, vec_dive, vec_mod, vec_cmpeq, vec_cmpne, vec_cmpgt, vec_cmplt,
18010 vec_cmpge, vec_cmple, vec_all_eq, vec_all_ne, vec_all_gt, vec_all_lt,
18011 vec_all_ge, vec_all_le, vec_any_eq, vec_any_ne, vec_any_gt, vec_any_lt,
18012 vec_any_ge, vec_any_le.
18014 2021-06-09 Carl Love <cel@us.ibm.com>
18016 * config/rs6000/altivec.md (altivec_vrl<VI_char>mi): Fix
18017 bug in argument generation.
18019 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
18021 * config/arm/iterators.md (<supf>): Remove VCLZQ_U, VCLZQ_S.
18023 * config/arm/mve.md (mve_vclzq_<supf><mode>): Add '@' prefix,
18024 remove <supf> iterator.
18025 (mve_vclzq_u<mode>): New.
18026 * config/arm/neon.md (clz<mode>2): Rename to neon_vclz<mode>.
18027 (neon_vclz<mode): Move to ...
18028 * config/arm/unspecs.md (VCLZQ_U, VCLZQ_S): Remove.
18029 * config/arm/vec-common.md: ... here. Add support for MVE.
18031 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
18033 * config/arm/mve.md (mve_vhaddq_<supf><mode>): Prefix with '@'.
18034 (@mve_vrhaddq_<supf><mode): Likewise.
18035 * config/arm/neon.md (neon_v<r>hadd<sup><mode>): Likewise.
18036 * config/arm/vec-common.md (avg<mode>3_floor, uavg<mode>3_floor)
18037 (avg<mode>3_ceil", uavg<mode>3_ceil): New patterns.
18039 2021-06-09 imba-tjd <109224573@qq.com>
18041 * doc/invoke.texi: Fix typo.
18043 2021-06-09 Roger Sayle <roger@nextmovesoftware.com>
18045 PR middle-end/53267
18046 * fold-const-call.c (fold_const_call_sss) [CASE_CFN_FMOD]:
18047 Support evaluation of fmod/fmodf/fmodl at compile-time.
18049 2021-06-09 Richard Biener <rguenther@suse.de>
18051 PR tree-optimization/100981
18052 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
18053 gimple_get_lhs to also handle calls.
18054 * tree-vect-slp-patterns.c (complex_pattern::build): Transfer
18057 2021-06-09 Richard Biener <rguenther@suse.de>
18059 PR tree-optimization/97832
18060 * tree-vectorizer.h (_slp_tree::failed): New.
18061 * tree-vect-slp.c (_slp_tree::_slp_tree): Initialize
18063 (_slp_tree::~_slp_tree): Free failed.
18064 (vect_build_slp_tree): Retain failed nodes and record
18065 matches in them, copying that back out when running
18066 into a cached fail. Dump start and end of discovery.
18067 (dt_sort_cmp): New.
18068 (vect_build_slp_tree_2): Handle associatable chains
18069 together doing more aggressive operand swapping.
18071 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
18074 * config.gcc (gcc_cv_initfini_array): Set to yes for Linux and
18076 * doc/install.texi: Require glibc 2.1 and binutils 2.12 for
18077 Linux and GNU targets.
18079 2021-06-09 Richard Biener <rguenther@suse.de>
18081 * tree-vect-stmts.c (vect_is_simple_use): Always get dt
18084 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
18086 * config/arc/arc.md (loop_end): Change it to
18087 define_insn_and_split.
18089 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
18091 * config/arc/arc.md (maddhisi4): Use VMAC2H instruction.
18092 (machi): New pattern.
18093 (umaddhisi4): Use VMAC2HU instruction.
18094 (umachi): New pattern.
18096 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
18098 * config/arc/arc-protos.h (arc_split_move_p): New prototype.
18099 * config/arc/arc.c (arc_split_move_p): New function.
18100 (arc_split_move): Clean up.
18101 * config/arc/arc.md (movdi_insn): Clean up, use arc_split_move_p.
18102 (movdf_insn): Likewise.
18103 * config/arc/simdext.md (mov<VWH>_insn): Likewise.
18105 2021-06-09 Uroš Bizjak <ubizjak@gmail.com>
18108 * config/i386/i386.c (print_operand_address_as): Rename "no_rip"
18109 argument to "raw". Do not emit segment overrides when "raw" is true.
18111 2021-06-09 Martin Liska <mliska@suse.cz>
18113 * doc/gcov.texi: Create a proper JSON files.
18114 * doc/invoke.texi: Remove dots in order to make it a valid
18117 2021-06-09 Xionghu Luo <luoxhu@linux.ibm.com>
18119 * config/rs6000/rs6000-p8swap.c (pattern_is_rotate64): New.
18120 (insn_is_load_p): Use pattern_is_rotate64.
18121 (insn_is_swap_p): Likewise.
18122 (quad_aligned_load_p): Likewise.
18123 (const_load_sequence_p): Likewise.
18124 (replace_swapped_aligned_load): Likewise.
18125 (recombine_lvx_pattern): Likewise.
18126 (recombine_stvx_pattern): Likewise.
18128 2021-06-09 Andrew MacLeod <amacleod@redhat.com>
18130 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Use a
18131 fur_stmt source record.
18132 * gimple-range.cc (fur_source::get_operand): Generic range query.
18133 (fur_source::get_phi_operand): New.
18134 (fur_source::register_dependency): New.
18135 (fur_source::query): New.
18136 (class fur_edge): New. Edge source for operands.
18137 (fur_edge::fur_edge): New.
18138 (fur_edge::get_operand): New.
18139 (fur_edge::get_phi_operand): New.
18140 (fur_edge::query): New.
18141 (fur_stmt::fur_stmt): New.
18142 (fur_stmt::get_operand): New.
18143 (fur_stmt::get_phi_operand): New.
18144 (fur_stmt::query): New.
18145 (class fur_depend): New. Statement source and process dependencies.
18146 (fur_depend::fur_depend): New.
18147 (fur_depend::register_dependency): New.
18148 (class fur_list): New. List source for operands.
18149 (fur_list::fur_list): New.
18150 (fur_list::get_operand): New.
18151 (fur_list::get_phi_operand): New.
18152 (fold_range): New. Instantiate appropriate fur_source class and fold.
18153 (fold_using_range::range_of_range_op): Use new API.
18154 (fold_using_range::range_of_address): Ditto.
18155 (fold_using_range::range_of_phi): Ditto.
18156 (imple_ranger::fold_range_internal): Use fur_depend class.
18157 (fold_using_range::range_of_ssa_name_with_loop_info): Use new API.
18158 * gimple-range.h (class fur_source): Now a base class.
18159 (class fur_stmt): New.
18160 (fold_range): New prototypes.
18161 (fur_source::fur_source): Delete.
18163 2021-06-08 Andrew Pinski <apinski@marvell.com>
18165 PR tree-optimization/25290
18166 * tree-ssa-phiopt.c (xor_replacement): Delete.
18167 (tree_ssa_phiopt_worker): Delete use of xor_replacement.
18168 (match_simplify_replacement): Allow one cheap preparation
18169 statement that can be moved to before the if.
18171 2021-06-08 Pat Haugen <pthaugen@linux.ibm.com>
18173 * config/rs6000/power10.md (power10-fused-load, power10-fused-store,
18174 power10-fused_alu, power10-fused-vec, power10-fused-branch): New.
18176 2021-06-08 Jeff Law <jeffreyalaw@gmail.com>
18178 * config/h8300/logical.md (andqi3_1): Move BCLR case into define_insn_and_split.
18179 Create length attribute on define_insn_and_split. Only split for cases which we
18181 (andqi3_1<cczn>): Renamed from andqi3_1_clobber_flags. Only handle AND here and
18182 fix length computation.
18183 (b<code><mode>msx): Combine QImode and HImode H8/SX patterns using iterator.
18185 2021-06-08 Richard Biener <rguenther@suse.de>
18187 PR tree-optimization/100923
18188 * tree-ssa-sccvn.c (valueize_refs_1): Take a pointer to
18189 the operand vector to be valueized.
18190 (valueize_refs): Likewise.
18191 (valueize_shared_reference_ops_from_ref): Adjust.
18192 (valueize_shared_reference_ops_from_call): Likewise.
18193 (vn_reference_lookup_3): Likewise.
18194 (vn_reference_lookup_pieces): Likewise. Re-valueize
18195 with honoring availability when we are about to create
18196 the ao_ref and valueized before.
18197 (vn_reference_lookup): Likewise.
18198 (vn_reference_insert_pieces): Adjust.
18200 2021-06-08 Richard Biener <rguenther@suse.de>
18202 * tree-vectorizer.h (_slp_instance::root_stmt): Change to...
18203 (_slp_instance::root_stmts): ... a vector.
18204 (SLP_INSTANCE_ROOT_STMT): Rename to ...
18205 (SLP_INSTANCE_ROOT_STMTS): ... this.
18206 (slp_root::root): Change to...
18207 (slp_root::roots): ... a vector.
18208 (slp_root::slp_root): Adjust.
18209 * tree-vect-slp.c (_slp_instance::location): Adjust.
18210 (vect_free_slp_instance): Release the root stmt vector.
18211 (vect_build_slp_instance): Adjust.
18212 (vect_analyze_slp): Likewise.
18213 (_bb_vec_info::~_bb_vec_info): Likewise.
18214 (vect_slp_analyze_operations): Likewise.
18215 (vect_bb_vectorization_profitable_p): Likewise. Adjust
18216 costs for the root stmt.
18217 (vect_slp_check_for_constructors): Gather all BIT_INSERT_EXPRs
18219 (vect_slp_analyze_bb_1): Simplify by marking all root stmts
18221 (vectorize_slp_instance_root_stmt): Adjust.
18222 (vect_schedule_slp): Likewise.
18224 2021-06-08 Aldy Hernandez <aldyh@redhat.com>
18226 * gimple-ssa-evrp.c (class ssa_equiv_stack): New.
18227 (ssa_equiv_stack::ssa_equiv_stack): New.
18228 (ssa_equiv_stack::~ssa_equiv_stack): New.
18229 (ssa_equiv_stack::enter): New.
18230 (ssa_equiv_stack::leave): New.
18231 (ssa_equiv_stack::push_replacement): New.
18232 (ssa_equiv_stack::get_replacement): New.
18233 (is_pointer_ssa): New.
18234 (class pointer_equiv_analyzer): New.
18235 (pointer_equiv_analyzer::pointer_equiv_analyzer): New.
18236 (pointer_equiv_analyzer::~pointer_equiv_analyzer): New.
18237 (pointer_equiv_analyzer::set_global_equiv): New.
18238 (pointer_equiv_analyzer::set_cond_equiv): New.
18239 (pointer_equiv_analyzer::get_equiv): New.
18240 (pointer_equiv_analyzer::enter): New.
18241 (pointer_equiv_analyzer::leave): New.
18242 (pointer_equiv_analyzer::get_equiv_expr): New.
18243 (pta_valueize): New.
18244 (pointer_equiv_analyzer::visit_stmt): New.
18245 (pointer_equiv_analyzer::visit_edge): New.
18246 (hybrid_folder::value_of_expr): Call PTA.
18247 (hybrid_folder::value_on_edge): Same.
18248 (hybrid_folder::pre_fold_bb): New.
18249 (hybrid_folder::post_fold_bb): New.
18250 (hybrid_folder::pre_fold_stmt): New.
18251 (rvrp_folder::pre_fold_bb): New.
18252 (rvrp_folder::post_fold_bb): New.
18253 (rvrp_folder::pre_fold_stmt): New.
18254 (rvrp_folder::value_of_expr): Call PTA.
18255 (rvrp_folder::value_on_edge): Same.
18257 2021-06-08 Jakub Jelinek <jakub@redhat.com>
18260 * tree-inline.c (copy_tree_body_r): For OMP_CLAUSE_DEPEND don't
18261 check TREE_CODE if OMP_CLAUSE_DECL is NULL.
18263 2021-06-08 Richard Biener <rguenther@suse.de>
18265 PR middle-end/100951
18266 * tree-vect-generic.c (expand_vector_piecewise): Build a
18267 VECTOR_CST if all elements are constant.
18268 (expand_vector_condition): Likewise.
18269 (lower_vec_perm): Likewise.
18270 (expand_vector_conversion): Likewise.
18272 2021-06-08 Martin Liska <mliska@suse.cz>
18274 * doc/invoke.texi: Document new param evrp-sparse-threshold.
18276 2021-06-08 Martin Liska <mliska@suse.cz>
18278 * genautomata.c (create_automata): Fix typo.
18280 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
18282 PR tree-optimization/100794
18283 * tree-predcom.c (tree_predictive_commoning_loop): Add parameter
18284 allow_unroll_p and only allow unrolling when it's true.
18285 (tree_predictive_commoning): Add parameter allow_unroll_p and
18287 (run_tree_predictive_commoning): Likewise.
18288 (pass_predcom::gate): Check flag_tree_loop_vectorize and
18289 global_options_set.x_flag_predictive_commoning.
18290 (pass_predcom::execute): Adjust for allow_unroll_p.
18292 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
18294 * tree-predcom.c (execute_pred_commoning): Remove update_ssa call.
18295 (tree_predictive_commoning_loop): Factor some cleanup stuffs into
18296 lambda function cleanup, remove scev_reset call, and adjust return
18298 (tree_predictive_commoning): Adjust for different changed values,
18299 only set flag TODO_update_ssa_only_virtuals if changed.
18300 (pass_data pass_data_predcom): Remove TODO_update_ssa_only_virtuals
18301 from todo_flags_finish.
18303 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
18305 * gimple-range-cache.cc (class sbr_sparse_bitmap): New.
18306 (sbr_sparse_bitmap::sbr_sparse_bitmap): New.
18307 (sbr_sparse_bitmap::bitmap_set_quad): New.
18308 (sbr_sparse_bitmap::bitmap_get_quad): New.
18309 (sbr_sparse_bitmap::set_bb_range): New.
18310 (sbr_sparse_bitmap::get_bb_range): New.
18311 (sbr_sparse_bitmap::bb_range_p): New.
18312 (block_range_cache::block_range_cache): initialize bitmap obstack.
18313 (block_range_cache::~block_range_cache): Destruct obstack.
18314 (block_range_cache::set_bb_range): Decide when to utilze the
18315 sparse on entry cache.
18316 * gimple-range-cache.h (block_range_cache): Add bitmap obstack.
18317 * params.opt (-param=evrp-sparse-threshold): New.
18319 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
18321 * bitmap.c (bitmap_set_aligned_chunk): New.
18322 (bitmap_get_aligned_chunk): New.
18323 (test_aligned_chunk): New.
18324 (bitmap_c_tests): Call test_aligned_chunk.
18325 * bitmap.h (bitmap_set_aligned_chunk, bitmap_get_aligned_chunk): New.
18327 2021-06-07 Uroš Bizjak <ubizjak@gmail.com>
18330 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
18332 (ix86_expand_vector_init_one_nonzero): Ditto.
18333 (ix86_expand_vector_init_one_var): Ditto.
18334 (ix86_expand_vector_init_general): Ditto.
18335 * config/i386/mmx.md (vec_initv4qiqi): New expander.
18337 2021-06-07 Jeff Law <jeffreyalaw@gmail.com>
18339 * config/h8300/movepush.md: Change most _clobber_flags
18340 patterns to instead use <cczn> subst.
18341 (movsi_cczn): New pattern with usable CC cases split out.
18342 (movsi_h8sx_cczn): Likewise.
18344 2021-06-07 Martin Liska <mliska@suse.cz>
18346 * common/common-target.def: Split long lines and replace them
18348 * target.def: Likewise.
18349 * doc/tm.texi: Re-generated.
18351 2021-06-07 Jakub Jelinek <jakub@redhat.com>
18354 * fold-const.c (fold_read_from_vector): Return NULL if trying to
18355 read from a CONSTRUCTOR with vector type elements.
18357 2021-06-07 Jakub Jelinek <jakub@redhat.com>
18359 PR middle-end/100898
18360 * tree-inline.c (copy_bb): Only use gimple_call_arg_ptr if memcpy
18361 should copy any arguments. Don't call gimple_call_num_args
18362 on id->call_stmt or call_stmt more than once.
18364 2021-06-07 liuhongt <hongtao.liu@intel.com>
18367 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): Refine
18369 (<insn>v4siv4di2): Delete constraints for define_expand.
18371 2021-06-07 liuhongt <hongtao.liu@intel.com>
18374 * config/i386/i386-expand.c (ix86_expand_builtin): Remove
18375 assignment of cfun->machine->has_explicit_vzeroupper.
18376 * config/i386/i386-features.c
18377 (ix86_add_reg_usage_to_vzerouppers): Delete.
18378 (ix86_add_reg_usage_to_vzeroupper): Ditto.
18379 (rest_of_handle_insert_vzeroupper): Remove
18380 ix86_add_reg_usage_to_vzerouppers, add df_analyze at the end
18382 (gate): Remove cfun->machine->has_explicit_vzeroupper.
18383 * config/i386/i386-protos.h (ix86_expand_avx_vzeroupper):
18385 * config/i386/i386.c (ix86_insn_callee_abi): New function.
18386 (ix86_initialize_callee_abi): Ditto.
18387 (ix86_expand_avx_vzeroupper): Ditto.
18388 (ix86_hard_regno_call_part_clobbered): Adjust for vzeroupper
18390 (TARGET_INSN_CALLEE_ABI): Define as ix86_insn_callee_abi.
18391 (ix86_emit_mode_set): Call ix86_expand_avx_vzeroupper
18393 * config/i386/i386.h (struct GTY(()) machine_function): Delete
18394 has_explicit_vzeroupper.
18395 * config/i386/i386.md (enum unspec): New member
18397 (ABI_DEFAULT,ABI_VZEROUPPER,ABI_UNKNOWN): New
18398 define_constants for insn callee abi index.
18399 * config/i386/predicates.md (vzeroupper_pattern): Adjust.
18400 * config/i386/sse.md (UNSPECV_VZEROUPPER): Deleted.
18401 (avx_vzeroupper): Call ix86_expand_avx_vzeroupper.
18402 (*avx_vzeroupper): Rename to ..
18403 (avx_vzeroupper_callee_abi): .. this, and adjust pattern as
18404 call_insn which has a special vzeroupper ABI.
18405 (*avx_vzeroupper_1): Deleted.
18407 2021-06-07 liuhongt <hongtao.liu@intel.com>
18410 * df-scan.c (df_get_call_refs): When call_insn is a fake call,
18411 it won't use stack pointer reg.
18412 * final.c (leaf_function_p): When call_insn is a fake call, it
18413 won't affect caller as a leaf function.
18414 * reg-stack.c (callee_clobbers_any_stack_reg): New.
18415 (subst_stack_regs): When call_insn doesn't clobber any stack
18416 reg, don't clear the arguments.
18417 * rtl.c (shallow_copy_rtx): Don't clear flag used when orig is
18419 * shrink-wrap.c (requires_stack_frame_p): No need for stack
18420 frame for a fake call.
18421 * rtl.h (FAKE_CALL_P): New macro.
18423 2021-06-06 Eric Botcazou <ebotcazou@adacore.com>
18425 * config/sparc/sparc-protos.h (order_regs_for_local_alloc): Rename
18427 (sparc_order_regs_for_local_alloc): ...this.
18428 (sparc_leaf_reg_remap): Declare.
18429 * config/sparc/sparc.h (ADJUST_REG_ALLOC_ORDER): Adjust.
18430 (LEAF_REG_REMAP): Reimplement as call to sparc_leaf_reg_remap.
18431 * config/sparc/sparc.c (leaf_reg_remap): Delete.
18432 (order_regs_for_local_alloc): Rename to...
18433 (sparc_order_regs_for_local_alloc): ...this.
18434 (sparc_leaf_reg_remap): New function.
18435 (sparc_conditional_register_usage): Do not modify leaf_reg_remap.
18437 2021-06-06 David Edelsohn <dje.gcc@gmail.com>
18439 * config/rs6000/rs6000.c (rs6000_xcoff_asm_output_aligned_decl_common):
18440 Use assemble_name to output BSS section name.
18442 2021-06-06 Uroš Bizjak <ubizjak@gmail.com>
18444 * config/i386/constraints.md (Bs):
18445 Remove boolean operators from match_test RTX.
18448 (M): Use "mode" variable instead of GET_MODE (op) in match_test RTX.
18451 2021-06-06 Martin Liska <mliska@suse.cz>
18453 * doc/extend.texi: Add missing @headitem.
18454 * doc/invoke.texi: Likewise.
18455 * doc/objc.texi: Likewise.
18457 2021-06-06 Martin Liska <mliska@suse.cz>
18459 * genhooks.c (emit_findices): Remove unused function.
18460 (emit_documentation): Do not call emit_findices
18461 and do not search for @Fcode directives.
18463 2021-06-06 Martin Liska <mliska@suse.cz>
18465 * doc/invoke.texi: Remove extra character.
18467 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
18469 * config/sh/sh.md (doloop_end_split): Fix empty split condition.
18471 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
18473 * config/sparc/sparc.md (*snedi<W:mode>_zero_vis3,
18474 *neg_snedi<W:mode>_zero_subxc, *plus_snedi<W:mode>_zero,
18475 *plus_plus_snedi<W:mode>_zero, *minus_snedi<W:mode>_zero,
18476 *minus_minus_snedi<W:mode>_zero): Fix empty split condition.
18478 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
18480 * config/or1k/or1k.md (*movdi): Fix empty split condition.
18482 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
18484 * config/mips/mips.md (<anonymous>, bswapsi2, bswapdi2): Fix empty
18487 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
18489 * config/m68k/m68k.md (*zero_extend_inc, *zero_extend_dec,
18490 *zero_extendsidi2): Fix empty split condition.
18492 2021-06-05 Jeff Law <jeffreyalaw@gmail.com>
18494 * config/h8300/addsub.md: Fix split condition in define_insn_and_split
18496 * config/h8300/bitfield.md: Likewise.
18497 * config/h8300/combiner.md: Likewise.
18498 * config/h8300/divmod.md: Likewise.
18499 * config/h8300/extensions.md: Likewise.
18500 * config/h8300/jumpcall.md: Likewise.
18501 * config/h8300/movepush.md: Likewise.
18502 * config/h8300/multiply.md: Likewise.
18503 * config/h8300/other.md: Likewise.
18504 * config/h8300/shiftrotate.md: Likewise.
18505 * config/h8300/logical.md: Likewise. Fix split pattern to use
18506 code iterator that somehow slipped through.
18508 2021-06-04 Tobias Burnus <tobias@codesourcery.com>
18510 PR middle-end/100905
18511 * tree-nested.c (convert_nonlocal_omp_clauses,
18512 convert_local_omp_clauses): Handle OMP_CLAUSE_BIND.
18514 2021-06-04 Martin Sebor <msebor@redhat.com>
18516 PR middle-end/100732
18517 * gimple-fold.c (gimple_fold_builtin_sprintf): Avoid folding calls
18518 with either source or destination argument of invalid type.
18519 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Avoid checking
18520 calls with arguments of invalid type.
18522 2021-06-04 Martin Sebor <msebor@redhat.com>
18524 * attribs.c (init_attr_rdwr_indices): Use VLA bounds in the expected
18526 (attr_access::vla_bounds): Also handle VLA bounds.
18528 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
18530 * config/i386/predicates.md (GOT_memory_operand):
18531 Implement using match_code RTXes.
18532 (GOT32_symbol_operand): Ditto.
18534 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
18537 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
18539 (ix86_expand_vector_init_general): Ditto.
18540 Use SImode instead of word_mode for logic operations
18541 when GET_MODE_SIZE (mode) < UNITS_PER_WORD.
18542 (expand_vec_perm_even_odd_1): Assert that V2HI mode should be
18543 implemented by expand_vec_perm_1.
18544 (expand_vec_perm_broadcast_1): Assert that V2HI and V4HI modes
18545 should be implemented using standard shuffle patterns.
18546 (ix86_vectorize_vec_perm_const): Handle V2HImode. Add V4HI and
18547 V2HI modes to modes, implementable with shuffle for one operand.
18548 * config/i386/mmx.md (*punpckwd): New insn_and_split pattern.
18549 (*pshufw_1): New insn pattern.
18550 (*vec_dupv2hi): Ditto.
18551 (vec_initv2hihi): New expander.
18553 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
18555 * config/arm/vfp.md (no_literal_pool_df_immediate,
18556 no_literal_pool_sf_immediate): Fix empty split condition.
18558 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
18560 * config/i386/i386.md (*load_tp_x32_zext, *add_tp_x32_zext,
18561 *tls_dynamic_gnu2_combine_32): Fix empty split condition.
18562 * config/i386/sse.md (*<sse2_avx2>_pmovmskb_lt,
18563 *<sse2_avx2>_pmovmskb_zext_lt, *sse2_pmovmskb_ext_lt,
18564 *<sse4_1_avx2>_pblendvb_lt): Likewise.
18566 2021-06-04 Jakub Jelinek <jakub@redhat.com>
18569 * config/i386/i386-expand.c (ix86_expand_vector_init): Handle
18570 concatenation from half-sized modes with TImode elements.
18572 2021-06-04 Claudiu Zissulescu <claziss@synopsys.com>
18574 * config/arc/arc.c (arc_override_options): Disable millicode
18575 thunks when RF16 is on.
18577 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
18579 * config/rs6000/rs6000.h (PROMOTE_MODE): Remove.
18581 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
18583 * config/rs6000/rs6000-call.c (rs6000_promote_function_mode):
18584 Replace PROMOTE_MODE marco with its content.
18586 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
18588 * config/cris/cris.md (*addi_reload): Fix empty split condition.
18590 2021-06-03 Jim Wilson <jimw@sifive.com>
18592 * config.gcc (riscv*-*-*): If --with-riscv-attribute not used,
18593 turn it on for all riscv targets.
18595 2021-06-03 Uroš Bizjak <ubizjak@gmail.com>
18598 * config/i386/i386-expand.c (ix86_expand_vector_set):
18599 Handle V2HI and V4QI modes.
18600 (ix86_expand_vector_extract): Ditto.
18601 * config/i386/mmx.md (*pinsrw): New insn pattern.
18604 (*pextrw_zext): Ditto.
18606 (*pextrb_zext): Ditto.
18607 (vec_setv2hi): New expander.
18608 (vec_extractv2hihi): Ditto.
18609 (vec_setv4qi): Ditto.
18610 (vec_extractv4qiqi): Ditto.
18611 (vec_setv8qi): Enable only for TARGET_SSE4_1.
18612 (vec_extractv8qiqi): Ditto.
18614 2021-06-03 Aaron Sawdey <acsawdey@linux.ibm.com>
18616 * config/rs6000/genfusion.pl (gen_logical_addsubf): Fix input
18617 order to subf instruction.
18618 * config/rs6000/fusion.md: Regenerate.
18620 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
18622 * calls.c (get_size_range): Use range_of_expr instead of
18623 determine_value_range.
18624 * tree-affine.c (expr_to_aff_combination): Same.
18625 * tree-data-ref.c (split_constant_offset): Same.
18626 * tree-vrp.c (determine_value_range_1): Remove.
18627 (determine_value_range): Remove.
18628 * tree-vrp.h (determine_value_range): Remove.
18630 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
18632 * function-tests.c (test_ranges): Call gimple_range_tests.
18633 * gimple-range-cache.cc (ranger_cache::range_of_expr): Pass stmt
18635 * gimple-range.cc (fur_source::get_operand): Do not call
18636 get_tree_range or gimple_range_global.
18638 (get_tree_range): Move to value-query.cc.
18639 Call get_arith_expr_range.
18640 (gimple_ranger::range_of_expr): Add argument to get_tree_range.
18641 Include gimple-range-tests.cc.
18642 * gimple-range.h (fold_range): Add argument.
18643 (get_tree_range): Remove.
18644 * selftest.h (gimple_range_tests): New.
18645 * value-query.cc (global_range_query::range_of_expr): Add
18647 (range_query::get_tree_range): Move from gimple-range.cc.
18648 * value-query.h (class range_query): Add get_tree_range and
18649 get_arith_expr_range. Make fur_source a friend.
18650 * vr-values.c (vr_values::range_of_expr): Pass stmt to
18652 * gimple-range-tests.cc: New file.
18654 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
18656 * gimple-range.cc (gimple_ranger::export_global_ranges): Call
18657 update_global_range.
18658 * value-query.cc (update_global_range): New.
18659 * value-query.h (update_global_range): New.
18661 2021-06-03 David Malcolm <dmalcolm@redhat.com>
18663 * diagnostic-show-locus.c (diagnostic_show_locus): Don't reject
18664 printing the same location twice if there are fix-it hints,
18665 multiple locations, or a label.
18667 2021-06-03 Andre Vieira <andre.simoesdiasvieira@arm.com>
18669 * tree-vect-loop.c (vect_transform_loop): Use main loop's various'
18670 thresholds to narrow the upper bound on epilogue iterations.
18672 2021-06-03 Christophe Lyon <christophe.lyon@linaro.org>
18674 * config/arm/mve.md (mve_vabsq_f<mode>): Use 'abs' instead of unspec.
18675 (mve_vabsq_s<mode>): Likewise.
18676 * config/arm/neon.md (abs<mode>2): Rename to neon_abs<mode>2.
18677 * config/arm/unspecs.md (VABSQ_F, VABSQ_S): Delete.
18678 * config/arm/vec-common.md (neg<mode>2): Rename to
18679 <absneg_str><mode>2.
18681 2021-06-03 Claudiu Zissulescu <claziss@synopsys.com>
18683 * common/config/arc/arc-common.c (arc_option_optimization_table):
18684 Remove malign-call.
18685 * config/arc/arc.c (arc_unalign_branch_p): Remove unused function.
18686 * config/arc/arc.h (TARGET_MIXED_CODE): Remove macro.
18687 (INDEX_REG_CLASS): Only refer to GENERAL_REGS.
18688 * config/arc/arc.md (abssi2_mixed): Remove pattern.
18689 * config/arc/arc.opt (munalign-prob-threshold): Mark it obsolete.
18690 (malign-call): Likewise.
18691 (mmixed-code): Likewise.
18692 * doc/invoke.texi (ARC): Update doc.
18694 2021-06-03 Martin Liska <mliska@suse.cz>
18696 * common.opt: Use proper Enum values.
18697 * opts.c (COVERAGE_SANITIZER_OPT): Remove.
18698 (parse_sanitizer_options): Handle only sanitizer_opts.
18699 (common_handle_option): Just assign value.
18701 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
18704 * tree-inline.c (inline_forbidden_p): Remove test on return type.
18706 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
18708 * dwarf2out.c (loc_list_from_tree_1) <FUNCTION_DECL>: Also generate
18709 DW_OP_GNU_variable_value referencing an existing DIE at file scope.
18710 (type_byte_size): Inline into...
18711 (add_byte_size_attribute): ...this and call add_scalar_info.
18713 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
18715 * dwarf2out.c (mem_loc_descriptor) <UDIV>: Fix typo.
18716 (typed_binop_from_tree): New function.
18717 (loc_list_from_tree_1) <EXACT_DIV_EXPR>: For an unsigned type,
18718 turn a divide by a power of 2 into a shift.
18719 <CEIL_DIV_EXPR>: For an unsigned type, use a signed divide if the
18720 size of the mode is lower than DWARF2_ADDR_SIZE; otherwise, do a
18721 typed divide by calling typed_binop_from_tree.
18723 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
18725 * dwarf2out.c (scompare_loc_descriptor): Fix head comment.
18726 (is_handled_procedure_type): Likewise.
18727 (struct loc_descr_context): Add strict_signedness field.
18728 (resolve_args_picking_1): Deal with DW_OP_[GNU_]deref_type,
18729 DW_OP_[GNU_]convert and DW_OP_[GNU_]reinterpret.
18730 (resolve_args_picking): Minor tweak.
18731 (function_to_dwarf_procedure): Initialize strict_signedness field.
18732 (type_byte_size): Likewise.
18733 (field_byte_offset): Likewise.
18734 (gen_descr_array_type_die): Likewise.
18735 (gen_variant_part): Likewise.
18736 (loc_list_from_tree_1) <CALL_EXPR>: Tidy up and set strict_signedness
18737 to true when a context is present before evaluating the arguments.
18738 <COND_EXPR>: Do not generate a useless comparison with zero.
18739 When dereferencing an address, if strict_signedness is true and the
18740 type is small and signed, use DW_OP_deref_type to do the dereference
18741 and then DW_OP_convert to convert back to the generic type.
18743 2021-06-03 Jakub Jelinek <jakub@redhat.com>
18746 * tree-inline.c (copy_tree_body_r): Handle iterators on
18747 OMP_CLAUSE_AFFINITY or OMP_CLAUSE_DEPEND.
18749 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
18751 * config/arc/arc.md (*bbit_di): Remove.
18753 2021-06-02 Christoph Muellner <cmuellner@gcc.gnu.org>
18755 PR rtl-optimization/100264
18756 * ree.c (get_sub_rtx): Ignore SET expressions without register
18757 destinations and remove assertion, as it is not valid anymore
18758 with this new behaviour.
18759 (merge_def_and_ext): Eliminate destination check for register
18760 as such SET expressions can't occur anymore.
18761 (combine_reaching_defs): Likewise.
18763 2021-06-02 Jakub Jelinek <jakub@redhat.com>
18766 * config/xtensa/xtensa.h (LEAF_REG_REMAP): Cast REGNO to int to avoid
18767 -Wtype-limits warnings.
18768 (DWARF_FRAME_REGISTER): Rewrite into ternary operator with addition
18769 in operands to avoid -Wsign-compare warnings.
18771 2021-06-02 Pat Haugen <pthaugen@linux.ibm.com>
18773 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): Use
18776 2021-06-02 Vineet Gupta <vgupta@synopsys.com>
18778 * config/arc/arc.h (TARGET_CPU_DEFAULT): Change to hs38_linux.
18780 2021-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
18782 * config/s390/s390.md(*ashrdi3_31<setcc><cconly>): Use a single
18784 * config/s390/subst.md(cconly_subst): Use a single constraint
18785 in (match_scratch).
18787 2021-06-02 Martin Liska <mliska@suse.cz>
18789 * ipa-icf.h: Use auto_vec for memory_access_types.
18791 2021-06-02 Jeff Law <jeffreyalaw@gmail.com>
18793 * config/h8300/h8300-protos.h (compute_a_shift_length): Drop unused
18794 argument from prototype.
18795 (output_logical_op): Add rtx_code argument.
18796 (compute_logical_op_length): Likewise.
18797 * config/h8300/h8300.c (h8300_and_costs): Pass additional argument
18798 to compute_a_shift_length.
18799 (output_logical_op); New argument with the rtx code rather than
18800 extracting it from an operand. Handle QImode too.
18801 (compute_logical_op_length): Similary.
18802 (compute_a_shift_length): Drop unused argument.
18803 * config/h8300/h8300.md (logicals): New code iterator.
18804 * config/h8300/logical.md (<code><mode>3 expander): Combine
18805 the "and" expander with the "ior"/"xor" expander.
18806 (bclr<mode>msx): Combine the QI/HI mode patterns.
18807 (<logical><mode>3 insns): Use code iterator rather than match_operator.
18808 Handle QImode as well. Update call to output_logical_op and
18809 compute_logical_op_length to pass in rtx_code
18810 Fix split condition on all define_insn_and_split patterns.
18811 (one_cmpl<mode>2<cczn>): Use <cczn> to support both clobbering
18812 the flags and setting ZN via existing define_subst.
18813 * config/h8300/shiftrotate.md: Drop unused argument from
18814 calls to compute_a_shift_length.
18815 Signed-off-by: Jeff Law <jeffreyalaw@gmail.com>
18817 2021-06-01 Andrew Pinski <apinski@marvell.com>
18819 PR tree-optimization/25290
18820 * tree-ssa-phiopt.c (match_simplify_replacement):
18822 (tree_ssa_phiopt_worker): Use match_simplify_replacement.
18823 (two_value_replacement): Change the comment about
18824 conditional_replacement.
18825 (conditional_replacement): Delete.
18827 2021-06-01 Andrew Pinski <apinski@marvell.com>
18829 PR tree-optimization/95481
18830 * tree-tailcall.c (find_tail_calls): Handle empty typed
18833 2021-06-01 Andrew Pinski <apinski@marvell.com>
18835 * gimplify.c (zero_sized_field_decl): Delete
18836 (zero_sized_type): Delete
18837 (gimplify_init_ctor_eval): Use is_empty_type instead
18838 of zero_sized_field_decl.
18839 (gimplify_modify_expr): Use is_empty_type instead of
18842 2021-06-01 Jason Merrill <jason@redhat.com>
18845 * tree.h (CALL_FROM_NEW_OR_DELETE_P): Adjust comment.
18847 2021-06-01 Jason Merrill <jason@redhat.com>
18850 * diagnostic.h (warning_enabled_at): Declare.
18851 * diagnostic.c (diagnostic_enabled): Factor out from...
18852 (diagnostic_report_diagnostic): ...here.
18853 (warning_enabled_at): New.
18855 2021-06-01 Aldy Hernandez <aldyh@redhat.com>
18857 * gimple-ssa-evrp.c: Enable exporting of global ranges.
18859 2021-06-01 Martin Liska <mliska@suse.cz>
18862 * doc/invoke.texi: Mention that -fgcse-after-reload
18863 is enabled with -O3.
18865 2021-06-01 liuhongt <hongtao.liu@intel.com>
18867 PR tree-optimization/98365
18868 * tree-if-conv.c (strip_nop_cond_scalar_reduction): New function.
18869 (is_cond_scalar_reduction): Handle nop_expr in cond scalar reduction.
18870 (convert_scalar_cond_reduction): Ditto.
18871 (predicate_scalar_phi): Ditto.
18873 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
18875 PR tree-optimization/100781
18876 * gimple-range-cache.cc (ranger_cache::ranger_cache): Enable new
18877 value calculation by default.
18878 (ranger_cache::enable_new_values): New.
18879 (ranger_cache::disable_new_values): New.
18880 (ranger_cache::push_poor_value): Check if new values are allowed.
18881 * gimple-range-cache.h (class ranger_cache): New member/methods.
18882 * gimple-range.cc (gimple_ranger::range_of_expr): Check for debug
18883 statement, and disable/renable new value calculation.
18885 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
18887 * gimple-range-cache.cc (ranger_cache::ssa_range_in_bb): Delete.
18888 (ranger_cache::range_of_def): New.
18889 (ranger_cache::entry_range): New.
18890 (ranger_cache::exit_range): New.
18891 (ranger_cache::range_of_expr): Adjust.
18892 (ranger_cache::range_on_edge): Adjust.
18893 (ranger_cache::propagate_cache): Call exit_range directly.
18894 * gimple-range-cache.h (class ranger_cache): Adjust.
18896 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
18898 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust for
18899 gori_compute being a member rather than base class.
18900 dervied call to member call.
18901 (ranger_cache::dump): No longer dump gori_map.
18902 (ranger_cache::dump_bb): New.
18903 (ranger_cache::get_non_stale_global_range): Adjust for gori_compute
18904 being a member rather than base class.
18905 (ranger_cache::set_global_range): Ditto.
18906 (ranger_cache::ssa_range_in_bb): Ditto.
18907 (ranger_cache::range_of_expr): New.
18908 (ranger_cache::range_on_edge): New.
18909 (ranger_cache::block_range): Adjust for gori_computes. Debug changes.
18910 (ranger_cache::propagate_cache): Adjust debugging output.
18911 (ranger_cache::fill_block_cache): Adjust for gori_computes. Debug
18913 * gimple-range-cache.h (class ranger_cache): Make gori_compute a
18914 member, and inherit from range_query instead.
18915 (ranger_cache::dump_bb): New. split from dump.
18916 * gimple-range-gori.cc (gori_compute::ssa_range_in_bb): Delete.
18917 (gori_compute::expr_range_at_stmt): Delete.
18918 (gori_compute::compute_name_range_op): Delete.
18919 (gori_compute::compute_operand_range_switch): Add fur_source.
18920 (gori_compute::compute_operand_range): Add fur_source param, inline
18921 old compute_name_range_op and optimize_logical_operands.
18922 (struct tf_range): Delete.
18923 (gori_compute::logical_combine): Adjust
18924 (gori_compute::optimize_logical_operands): Delete.
18925 (gori_compute::compute_logical_operands_in_chain): Delete.
18926 (gori_compute::compute_logical_operands): Adjust.
18927 (gori_compute::compute_operand1_range): Adjust to fur_source.
18928 (gori_compute::compute_operand2_range): Ditto.
18929 (gori_compute::compute_operand1_and_operand2_range): Ditto.
18930 (gori_compute::outgoing_edge_range_p): Add range_query parameter,
18931 and adjust to fur_source.
18932 * gimple-range-gori.h (class gori_compute): Simplify and adjust to
18933 range_query and fur_source.
18934 * gimple-range.cc (gimple_ranger::range_on_edge): Query range_on_edge
18935 from the ranger_cache..
18936 (gimple_ranger::fold_range_internal): Adjust to base class change of
18938 (gimple_ranger::dump_bb): Adjust dump.
18939 * gimple-range.h (gimple_ranger):export gori computes object.
18941 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
18943 PR tree-optimization/100774
18944 * gimple-range-cache.cc (ranger_cache::get_non_stale_global_range):
18945 Constant values are also not stale.
18946 (ranger_cache::set_global_range): Range invariant values should also
18947 have the correct timestamp.
18949 2021-05-31 Martin Liska <mliska@suse.cz>
18951 * tree-streamer-in.c (unpack_ts_function_decl_value_fields):
18952 Unpack FUNCTION_DECL_DECL_TYPE.
18953 * tree-streamer-out.c (pack_ts_function_decl_value_fields):
18954 Stream FUNCTION_DECL_DECL_TYPE instead of
18955 DECL_IS_OPERATOR_NEW_P.
18956 * tree.h (set_function_decl_type): Use FUNCTION_DECL_DECL_TYPE
18958 (DECL_IS_OPERATOR_NEW_P): Likewise.
18959 (DECL_IS_OPERATOR_DELETE_P): Likewise.
18960 (DECL_LAMBDA_FUNCTION_P): Likewise.
18962 2021-05-31 Richard Biener <rguenther@suse.de>
18965 * internal-fn.c (expand_SHUFFLEVECTOR): Define.
18966 * internal-fn.def (SHUFFLEVECTOR): New.
18967 * internal-fn.h (expand_SHUFFLEVECTOR): Declare.
18968 * doc/extend.texi: Document __builtin_shufflevector.
18970 2021-05-31 Peter Bergner <bergner@linux.ibm.com>
18973 * config/rs6000/predicates.md(mma_assemble_input_operand): Allow
18974 indexed form addresses.
18976 2021-05-29 Jeff Law <jlaw@tachyum.com>
18978 * config/h8300/h8300.c (h8300_emit_stack_adjustment): Drop unused
18979 parameter. Call callers fixed.
18981 (output_plussi): Add FALLTHRU markers.
18982 (h8300_shift_needs_scratch_p): Add gcc_unreachable marker.
18984 2021-05-29 Jakub Jelinek <jakub@redhat.com>
18986 PR middle-end/99928
18987 * gimplify.c (gimplify_scan_omp_clauses): For taskloop simd
18988 combined with parallel, make sure to add shared clause to
18989 parallel for explicit linear clause.
18991 2021-05-29 Aldy Hernandez <aldyh@redhat.com>
18993 PR tree-optimization/100787
18994 * gimple-ssa-evrp.c: Disable exporting of global ranges.
18996 2021-05-28 Jason Merrill <jason@redhat.com>
18998 * tree-iterator.h (struct tree_stmt_iterator): Add operator++,
18999 operator--, operator*, operator==, and operator!=.
19000 (class tsi_range): New.
19002 2021-05-28 Richard Biener <rguenther@suse.de>
19004 PR tree-optimization/100778
19005 * tree-vect-slp.c (vect_build_slp_tree_1): Prevent possibly
19006 trapping ops in different BBs.
19008 2021-05-28 Richard Biener <rguenther@suse.de>
19011 * tree-inline.c (copy_bb): When processing __builtin_va_arg_pack
19012 copy fntype from original call.
19014 2021-05-28 Martin Liska <mliska@suse.cz>
19016 PR gcov-profile/100751
19017 * doc/gcov.texi: Revert partially a hunk that was wrong.
19019 2021-05-28 Cooper Qu <cooper.qu@linux.alibaba.com>
19021 * config/csky/csky-linux-elf.h (HAVE_sync_compare_and_swapqi):
19023 (HAVE_sync_compare_and_swaphi): Likewise.
19024 (HAVE_sync_compare_and_swapsi): Likewise.
19026 2021-05-28 Jakub Jelinek <jakub@redhat.com>
19028 PR middle-end/99928
19029 * tree.h (OMP_CLAUSE_MAP_IMPLICIT): Define.
19031 2021-05-28 Tobias Burnus <tobias@codesourcery.com>
19033 * gimplify.c (gimplify_omp_affinity): New.
19034 (gimplify_scan_omp_clauses): Call it; remove affinity clause afterwards.
19035 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_AFFINITY.
19036 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_AFFINITY.
19037 * tree.c (omp_clause_num_ops, omp_clause_code_name): Add clause.
19038 (walk_tree_1): Handle OMP_CLAUSE_AFFINITY.
19040 2021-05-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
19041 Richard Biener <rguenther@suse.de>
19043 * match.pd <popcount & / + pattern matching>:
19044 When generating popcount directly fails, try doing it in two halves.
19046 2021-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
19048 * Makefile.in (generated_files): Add gimple-match.c and
19051 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
19053 * gensupport.c (alter_predicate_for_insn): Handle MATCH_DUP.
19055 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
19057 * gensupport.c (alter_constraints): Add MATCH_SCRATCH case.
19059 2021-05-28 Kewen Lin <linkw@linux.ibm.com>
19061 PR tree-optimization/99398
19062 * tree-ssa-forwprop.c (simplify_permutation): Optimize some cases
19063 where the fed operands are CTOR/CST and propagated through
19064 VIEW_CONVERT_EXPR. Call vec_perm_indices::new_shrunk_vector.
19065 * vec-perm-indices.c (vec_perm_indices::new_shrunk_vector): New
19067 * vec-perm-indices.h (vec_perm_indices::new_shrunk_vector): New
19070 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
19072 * config/i386/mmx.md (addv2sf3): Do not call
19073 ix86_fixup_binary_operands_no_copy.
19076 (<smaxmin:code>v2sf3): Ditto.
19077 (<plusminus:insn><MMXMODEI:mode>3): Ditto.
19078 (<plusminus:insn><VI_32:mode>3): Remove expander.
19079 (<plusminus:insn><VI_32:mode>3): Rename from
19080 "*<plusminus:insn><VI_32:mode>3".
19081 (mulv4hi): Do not call ix86_fixup_binary_operands_no_copy.
19082 (mulv2hi3): Remove expander.
19083 (mulv2hi3): Rename from *mulv2hi3.
19084 (<s>mulv2hi3_highpart): Remove expander.
19085 (<s>mulv2hi3_highpart): Rename from *<s>mulv2hi3_highpart.
19086 (<smaxmin:code><MMXMODE14:mode>3): Rename from
19087 "*mmx_<smaxmin:code><MMXMODE14:mode>3".
19088 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Remove expander.
19089 (SMAXMIN_MMXMODEI): Remove mode iterator.
19090 (<smaxmin:code>v4hi3): New expander.
19091 (<smaxmin:code>v4qi3): Rename from *<smaxmin:code>v4qi3.
19092 (<smaxmin:code>v2hi3): Rename from *<smaxmin:code>v2hi3.
19093 (<smaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
19094 (SMAXMIN_VI_32): Remove mode iterator.
19095 (<umaxmin:code><MMXMODE24:mode>3): Rename from
19096 "*mmx_<umaxmin:code><MMXMODE24:mode>3".
19097 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Remove expander.
19098 (UMAXMIN_MMXMODEI): Remove mode iterator.
19099 (<umaxmin:code>v8qi3): New expander.
19100 (<umaxmin:code>v4qi3): Rename from *<umaxmin:code>v4qi3.
19101 (<umaxmin:code>v2hi3): Rename from *<umaxmin:code>v2hi3.
19102 (<umaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
19103 (UMAXMIN_VI_32): Remove mode iterator.
19104 (<any_shift:insn>v2hi3): Remove expander.
19105 (<any_shift:insn>v2hi3): Rename from *<any_shift:insn>v2hi3.
19106 (<any_logic:code><MMXMODEI:mode>3): Do not call
19107 ix86_fixup_binary_operands_no_copy.
19108 (<any_logic:code><VI_32:mode>3): Remove expander.
19109 (<any_logic:code><VI_32:mode>3): Rename from
19110 "*<any_logic:code><VI_32:mode>3".
19111 (uavg<mode>3_ceil): Do not call ix86_fixup_binary_operands_no_copy.
19112 * config/i386/sse.md (div<VF2:mode>3): Do not call
19113 ix86_fixup_binary_operands_no_copy.
19114 (div<VF1:mode>3): Ditto.
19115 (<maxmin:code><VI8_AVX2_AVX512F:mode>3): Ditto.
19116 (smulhrsv4hi3): Ditto.
19117 (smulhrsv2hi3): Ditto.
19119 2021-05-27 Martin Sebor <msebor@redhat.com>
19121 * ggc.h (gt_ggc_mx): Add overloads for all integers.
19123 * hash-map.h (class hash_map): Add pch_nx_helper overloads for all
19125 (hash_map::operator==): New function.
19127 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
19130 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
19131 For TARGET_XOP bypass SSE comparisons for all supported vector modes.
19132 * config/i386/mmx.md (*xop_maskcmp<MMXMODEI:mode>3): New insn pattern.
19133 (*xop_maskcmp<VI_32:mode>3): Ditto.
19134 (*xop_maskcmp_uns<MMXMODEI:mode>3): Ditto.
19135 (*xop_maskcmp_uns<VI_32:mode>3): Ditto.
19137 2021-05-27 Richard Earnshaw <rearnsha@arm.com>
19140 * config/arm/arm.c (arm_configure_build_target): Remove parameter
19141 opts_set, directly check opts parameters for being non-null.
19142 (arm_option_restore): Update call to arm_configure_build_target.
19143 (arm_option_override): Likewise.
19144 (arm_can_inline_p): Likewise.
19145 (arm_valid_target_attribute_tree): Likewise.
19146 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
19147 * config/arm/arm-protos.h (arm_configure_build_target): Adjust
19150 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
19152 * vr-values.c (simplify_conversion_using_ranges): Use
19153 get_range_query instead of get_global_range_query.
19155 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
19157 * gimple-range.cc (get_range_global): Move to value-query.cc.
19158 (gimple_range_global): Same.
19159 (get_global_range_query): Same.
19160 (global_range_query::range_of_expr): Same.
19161 * gimple-range.h (class global_range_query): Move to
19163 (gimple_range_global): Same.
19164 * tree-ssanames.c (get_range_info): Move to value-query.cc.
19165 (get_ptr_nonnull): Same.
19166 * tree-ssanames.h (get_range_info): Remove.
19167 (get_ptr_nonnull): Remove.
19168 * value-query.cc (get_ssa_name_range_info): Move from
19170 (get_ssa_name_ptr_info_nonnull): Same.
19171 (get_range_global): Move from gimple-range.cc.
19172 (gimple_range_global): Same.
19173 (get_global_range_query): Same.
19174 (global_range_query::range_of_expr): Same.
19175 * value-query.h (class global_range_query): Move from
19177 (gimple_range_global): Same.
19179 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
19182 * config/i386/mmx.md (uavgv4qi3_ceil): New insn pattern.
19183 (uavgv2hi3_ceil): Ditto.
19185 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
19188 * doc/extend.texi (scalar_storage_order): Rephrase slightly.
19190 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
19192 * tree-ssanames.c (get_range_info): Merge both copies of
19193 get_range_info into one that works with irange.
19194 * tree-ssanames.h (get_range_info): Remove version that works on
19197 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
19199 * builtins.c (check_nul_terminated_array): Convert to get_range_query.
19200 (expand_builtin_strnlen): Same.
19201 (determine_block_size): Same.
19202 * fold-const.c (expr_not_equal_to): Same.
19203 * gimple-fold.c (size_must_be_zero_p): Same.
19204 * gimple-match-head.c: Include gimple-range.h.
19205 * gimple-pretty-print.c (dump_ssaname_info): Convert to get_range_query.
19206 * gimple-ssa-warn-restrict.c
19207 (builtin_memref::extend_offset_range): Same.
19208 * graphite-sese-to-poly.c (add_param_constraints): Same.
19209 * internal-fn.c (get_min_precision): Same.
19210 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): Same.
19211 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Same.
19213 * tree-data-ref.c (split_constant_offset): Same.
19214 (dr_step_indicator): Same.
19215 * tree-dfa.c (get_ref_base_and_extent): Same.
19216 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
19217 * tree-ssa-loop-niter.c (refine_value_range_using_guard): Same.
19218 (determine_value_range): Same.
19219 (record_nonwrapping_iv): Same.
19220 (infer_loop_bounds_from_signedness): Same.
19221 (scev_var_range_cant_overflow): Same.
19222 * tree-ssa-phiopt.c (two_value_replacement): Same.
19223 * tree-ssa-pre.c (insert_into_preds_of_block): Same.
19224 * tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Same.
19225 * tree-ssa-strlen.c (handle_builtin_stxncpy_strncat): Same.
19227 (dump_strlen_info): Same.
19228 (set_strlen_range): Same.
19229 (maybe_diag_stxncpy_trunc): Same.
19230 (get_len_or_size): Same.
19231 (handle_integral_assign): Same.
19232 * tree-ssa-structalias.c (find_what_p_points_to): Same.
19233 * tree-ssa-uninit.c (find_var_cmp_const): Same.
19234 * tree-switch-conversion.c (bit_test_cluster::emit): Same.
19235 * tree-vect-patterns.c (vect_get_range_info): Same.
19236 (vect_recog_divmod_pattern): Same.
19237 * tree-vrp.c (intersect_range_with_nonzero_bits): Same.
19238 (register_edge_assert_for_2): Same.
19239 (determine_value_range_1): Same.
19240 * tree.c (get_range_pos_neg): Same.
19241 * vr-values.c (vr_values::get_lattice_entry): Same.
19242 (vr_values::update_value_range): Same.
19243 (simplify_conversion_using_ranges): Same.
19245 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
19247 * gimple-ssa-warn-alloca.c (alloca_call_type): Use
19248 get_range_query instead of query argument.
19249 (pass_walloca::execute): Enable and disable global ranger.
19251 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
19253 * gimple-ssa-evrp.c (rvrp_folder::rvrp_folder): Call
19255 (rvrp_folder::~rvrp_folder): Call disable_ranger.
19256 (hybrid_folder::hybrid_folder): Call enable_ranger.
19257 (hybrid_folder::~hybrid_folder): Call disable_ranger.
19259 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
19261 * function.c (allocate_struct_function): Set cfun->x_range_query.
19262 * function.h (struct function): Declare x_range_query.
19263 (get_range_query): New.
19264 (get_global_range_query): New.
19265 * gimple-range-cache.cc (ssa_global_cache::ssa_global_cache):
19266 Remove call to safe_grow_cleared.
19267 * gimple-range.cc (get_range_global): New.
19268 (gimple_range_global): Move from gimple-range.h.
19269 (get_global_range_query): New.
19270 (global_range_query::range_of_expr): New.
19271 (enable_ranger): New.
19272 (disable_ranger): New.
19273 * gimple-range.h (gimple_range_global): Move to gimple-range.cc.
19274 (class global_range_query): New.
19275 (enable_ranger): New.
19276 (disable_ranger): New.
19277 * gimple-ssa-evrp.c (evrp_folder::~evrp_folder): Rename
19278 dump_all_value_ranges to dump.
19279 * tree-vrp.c (vrp_prop::finalize): Same.
19280 * value-query.cc (range_query::dump): New.
19281 * value-query.h (range_query::dump): New.
19282 * vr-values.c (vr_values::dump_all_value_ranges): Rename to...
19283 (vr_values::dump): ...this.
19284 * vr-values.h (class vr_values): Rename dump_all_value_ranges to
19285 dump and make virtual.
19287 2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
19289 * config/i386/i386.c (ix86_autovectorize_vector_modes):
19290 Add V4QImode and V16QImode for TARGET_SSE2.
19291 * doc/sourcebuild.texi (Vector-specific attributes):
19292 Add vect64 and vect32 description.
19294 2021-05-26 Bernd Edlinger <bernd.edlinger@hotmail.de>
19296 * gimple-range-gori.cc (range_def_chain::register_dependency):
19297 Resize m_def_chain when needed.
19299 2021-05-26 Christophe Lyon <christophe.lyon@linaro.org>
19301 * config/arm/mve.md (mve_vaddvq_<supf><mode>): Prefix with '@'.
19302 * config/arm/neon.md (reduc_plus_scal_<mode>): Move to ..
19303 * config/arm/vec-common.md: .. here. Add support for MVE.
19305 2021-05-26 Jakub Jelinek <jakub@redhat.com>
19307 * config/epiphany/epiphany.c (epiphany_print_operand_address): Remove
19309 * config/microblaze/microblaze.c (microblaze_legitimize_address,
19311 microblaze_option_override, print_operand): Likewise.
19312 * config/microblaze/microblaze.md (call_internal_plt,
19313 call_value_intern_plt, call_value_intern): Likewise.
19314 * config/arm/aout.h (ASM_OUTPUT_ALIGN): Likewise.
19315 * config/iq2000/iq2000.md (call_internal1, call_value_internal1,
19316 call_value_multiple_internal1): Likewise.
19317 * config/bfin/bfin.c (symbolic_reference_mentioned_p): Likewise.
19319 2021-05-26 Jan-Benedict Glaw <jbglaw@lug-owl.de>
19321 * config/arc/arc.c (arc_address_cost, arc_print_operand_address,
19322 arc_ccfsm_advance, symbolic_reference_mentioned_p,
19323 arc_raw_symbolic_reference_mentioned_p): Remove register
19326 2021-05-26 Jakub Jelinek <jakub@redhat.com>
19329 * omp-low.c: Include omp-offload.h.
19330 (create_omp_child_function): If current_function_decl has
19331 "omp declare target" attribute and is_gimple_omp_offloaded,
19332 remove that attribute from the copy of attribute list and
19333 add "omp target entrypoint" attribute instead.
19334 (lower_omp_target): Mark .omp_data_sizes.* and .omp_data_kinds.*
19335 variables for offloading if in omp_maybe_offloaded_ctx.
19336 * omp-offload.c (pass_omp_target_link::execute): Nullify second
19337 argument to GOMP_target_data_ext in offloaded code.
19339 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
19341 * config/csky/csky.c (csky_can_change_mode_class): Delete.
19342 For csky, HF/SF mode use the low bits of VREGS.
19344 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
19346 * gimplify.c (gimplify_decl_expr): Do not clear TREE_READONLY on a
19347 DECL which is a reference for OMP.
19349 2021-05-26 Martin Liska <mliska@suse.cz>
19351 PR gcov-profile/100751
19352 * doc/gcov.texi: Document that __gcov_dump can be called just
19353 once and that __gcov_reset resets run-time counters.
19355 2021-05-26 Martin Liska <mliska@suse.cz>
19357 * doc/install.texi: Port relevant part from install-old.texi
19358 and re-generate list of CPUs and systems.
19360 2021-05-26 Martin Liska <mliska@suse.cz>
19362 * Makefile.in: Remove it.
19363 * doc/include/fdl.texi: Update next/previous chapters.
19364 * doc/install.texi: Likewise.
19365 * doc/install-old.texi: Removed.
19367 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
19369 * config/csky/csky.c (ck810_legitimate_index_p): Support
19370 "base + index" with DF mode.
19371 * config/csky/constraints.md ("Y"): New constraint for memory operands
19372 without index register.
19373 * config/csky/csky_insn_fpuv2.md (fpuv3_movdf): Use "Y" instead of "m"
19374 when mov between memory and general registers, and lower their priority.
19375 * config/csky/csky_insn_fpuv3.md (fpuv2_movdf): Likewise.
19377 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
19379 * config/csky/csky.c (TARGET_PROMOTE_PROTOTYPES): Delete.
19381 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
19383 * config/csky/csky.md (untyped_call): Emit clobber for return
19384 registers to mark them used.
19386 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
19388 * config/csky/csky.md (cskyv2_sextend_ldbs): New.
19390 2021-05-26 Andrew Pinski <apinski@marvell.com>
19392 * match.pd (x < 0 ? ~y : y): New patterns.
19394 2021-05-26 Andrew Pinski <apinski@marvell.com>
19396 * match.pd (A?CST1:CST2): Add simplifcations for A?0:+-1, A?+-1:0,
19397 A?POW2:0 and A?0:POW2.
19399 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
19401 * gimple-range-gori.cc (class logical_stmt_cache): Delete
19402 (logical_stmt_cache::logical_stmt_cache ): Delete.
19403 (logical_stmt_cache::~logical_stmt_cache): Delete.
19404 (logical_stmt_cache::cache_entry::dump): Delete.
19405 (logical_stmt_cache::get_range): Delete.
19406 (logical_stmt_cache::cached_name ): Delete.
19407 (logical_stmt_cache::same_cached_name): Delete.
19408 (logical_stmt_cache::cacheable_p): Delete.
19409 (logical_stmt_cache::slot_diagnostics ): Delete.
19410 (logical_stmt_cache::dump): Delete.
19411 (gori_compute_cache::gori_compute_cache): Delete.
19412 (gori_compute_cache::~gori_compute_cache): Delete.
19413 (gori_compute_cache::compute_operand_range): Delete.
19414 (gori_compute_cache::cache_stmt): Delete.
19415 * gimple-range-gori.h (gori_compute::compute_operand_range): Remove
19417 (class gori_compute_cache): Delete.
19419 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
19421 * gimple-range.cc (fold_using_range::range_of_range_op): Use m_gori
19423 (fold_using_range::range_of_address): Adjust.
19424 (fold_using_range::range_of_phi): Adjust.
19425 * gimple-range.h (class fur_source): Adjust.
19426 (fur_source::fur_source): Adjust.
19428 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
19430 * gimple-range-gori.cc (gori_compute::expr_range_at_stmt): Rename
19431 from expr_range_in_bb and adjust.
19432 (gori_compute::compute_name_range_op): Adjust.
19433 (gori_compute::optimize_logical_operands): Adjust.
19434 (gori_compute::compute_logical_operands_in_chain): Adjust.
19435 (gori_compute::compute_operand1_range): Adjust.
19436 (gori_compute::compute_operand2_range): Adjust.
19437 (ori_compute_cache::cache_stmt): Adjust.
19438 * gimple-range-gori.h (gori_compute): Rename prototype.
19440 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
19442 * gimple-range.cc (gimple_ranger::range_of_expr): Non-null should be
19443 checked only after range_of_stmt, not range_on_entry.
19444 (gimple_ranger::range_on_entry): Check for non-null in any
19445 predecessor block, if it is not already non-null.
19446 (gimple_ranger::range_on_exit): DOnt check for non-null after
19447 range on entry call.
19448 (gimple_ranger::dump_bb): New. Split from dump.
19449 (gimple_ranger::dump): Adjust.
19450 * gimple-range.h (class gimple_ranger): Adjust.
19452 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
19454 * gimple-range-cache.cc (struct range_timestamp): Delete.
19455 (class temporal_cache): Adjust.
19456 (temporal_cache::get_timestamp): Delete.
19457 (temporal_cache::set_dependency): Delete.
19458 (temporal_cache::temporal_value): Adjust.
19459 (temporal_cache::current_p): Take dependencies as params.
19460 (temporal_cache::set_timestamp): Adjust.
19461 (temporal_cache::set_always_current): Adjust.
19462 (ranger_cache::get_non_stale_global_range): Adjust.
19463 (ranger_cache::register_dependency): Delete.
19464 * gimple-range-cache.h (class range_cache): Adjust.
19466 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
19468 * gimple-range-gori.cc (range_def_chain::range_def_chain): init
19470 (range_def_chain::~range_def_chain): Dispose of obstack rather than
19471 each individual bitmap.
19472 (range_def_chain::set_import): New.
19473 (range_def_chain::get_imports): New.
19474 (range_def_chain::chain_import_p): New.
19475 (range_def_chain::register_dependency): Rename from build_def_chain
19477 (range_def_chain::def_chain_in_bitmap_p): New.
19478 (range_def_chain::add_def_chain_to_bitmap): New.
19479 (range_def_chain::has_def_chain): Just check first depenedence.
19480 (range_def_chain::get_def_chain): Process imports, use generic
19481 register_dependency routine.
19482 (range_def_chain::dump): New.
19483 (gori_map::gori_map): Allocate import list.
19484 (gori_map::~gori_map): Release imports.
19485 (gori_map::exports): Check for past allocated block size.
19486 (gori_map::imports): New.
19487 (gori_map::def_chain_in_export_p): Delete.
19488 (gori_map::is_import_p): New.
19489 (gori_map::maybe_add_gori): Handle imports.
19490 (gori_map::dump): Adjust output, add imports.
19491 (gori_compute::has_edge_range_p): Remove def_chain_in_export call.
19492 (gori_export_iterator::gori_export_iterator): New.
19493 (gori_export_iterator::next): New.
19494 (gori_export_iterator::get_name): New.
19495 * gimple-range-gori.h (range_def_chain): Add imports and direct
19496 dependecies via struct rdc.
19497 (range_def_chain::depend1): New.
19498 (range_def_chain::depend2): New.
19499 (class gori_map): Adjust.
19500 (FOR_EACH_GORI_IMPORT_NAME): New.
19501 (FOR_EACH_GORI_EXPORT_NAME): New.
19502 (class gori_export_iterator): New.
19504 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
19506 * gimple-range-cache.cc (ranger_cache::ranger_cache): Move initial
19507 export cache filling to here.
19508 * gimple-range-gori.cc (gori_compute::gori_compute) : From Here.
19510 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
19512 * gimple-range-gori.cc (range_def_chain): Move to gimple-range-gori.h.
19513 (gori_map): Move to gimple-range-gori.h.
19514 (gori_compute::gori_compute): Adjust.
19515 (gori_compute::~gori_compute): Delete.
19516 (gori_compute::compute_operand_range_switch): Adjust.
19517 (gori_compute::compute_operand_range): Adjust.
19518 (gori_compute::compute_logical_operands): Adjust.
19519 (gori_compute::has_edge_range_p ): Adjust.
19520 (gori_compute::set_range_invariant): Delete.
19521 (gori_compute::dump): Adjust.
19522 (gori_compute::outgoing_edge_range_p): Adjust.
19523 * gimple-range-gori.h (class range_def_chain): Relocate here.
19524 (class gori_map): Relocate here.
19525 (class gori_compute): Inherit from gori_map, and adjust.
19527 2021-05-25 Aldy Hernandez <aldyh@redhat.com>
19529 * value-range.cc (range_tests_legacy): Use
19530 build_nonstandard_integer_type instead of int and short.
19532 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
19534 * gimplify.c (gimplify_decl_expr): Clear TREE_READONLY on the DECL
19535 when really creating an initialization statement for it.
19537 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
19539 * tree-inline.c (setup_one_parameter): Fix thinko in new condition.
19541 2021-05-25 Kito Cheng <kito.cheng@sifive.com>
19543 * config/riscv/riscv.h (ASM_SPEC): Pass -mno-relax.
19545 2021-05-25 Martin Liska <mliska@suse.cz>
19547 PR tree-optimization/92860
19549 * optc-save-gen.awk: Remove exceptions.
19551 2021-05-25 Martin Liska <mliska@suse.cz>
19553 * asan.h (sanitize_coverage_p): New function.
19554 * doc/extend.texi: Document it.
19555 * fold-const.c (fold_range_test): Use sanitize_flags_p
19556 instead of flag_sanitize_coverage.
19557 (fold_truth_andor): Likewise.
19558 * sancov.c: Likewise.
19559 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
19560 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Handle
19561 -fsanitize-coverage when inlining.
19563 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
19565 * config/csky/csky-modes.def : Fix copyright.
19567 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
19569 * config/csky/csky-modes.def : Amend copyright.
19570 * config/csky/csky_insn_fpuv2.md : Likewise.
19571 * config/csky/csky_insn_fpuv3.md : Likewise.
19573 2021-05-25 Richard Biener <rguenther@suse.de>
19575 PR middle-end/100727
19576 * calls.c (initialize_argument_information): Explicitely test
19577 for WITH_SIZE_EXPR.
19578 * gimple-expr.c (mark_addressable): Skip outer WITH_SIZE_EXPR.
19580 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
19582 * config/csky/csky.h (FRAME_POINTER_REGNUM): Use
19583 HARD_FRAME_POINTER_REGNUM and FRAME_POINTER_REGNUM instead of
19584 the signle definition. The signle definition may not work well
19585 at simplify_subreg_regno().
19586 (HARD_FRAME_POINTER_REGNUM): New.
19587 (ELIMINABLE_REGS): Add for HARD_FRAME_POINTER_REGNUM.
19588 * config/csky/csky.c (get_csky_live_regs, csky_can_eliminate,
19589 csky_initial_elimination_offset, csky_expand_prologue,
19590 csky_expand_epilogue): Add for HARD_FRAME_POINTER_REGNUM.
19592 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
19594 * config/csky/csky.c (csky_option_override):
19595 Init csky_arch_isa_features[] in advance, so TARGET_DSP
19596 and TARGET_DIV can be set well.
19598 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
19600 * config/csky/constraints.md ("l", "h"): Delete.
19601 * config/csky/csky.h (reg_class, REG_CLASS_NAMES,
19602 REG_CLASS_CONTENTS): Delete LO_REGS and HI_REGS.
19603 * config/csky/csky.c (regno_reg_classm,
19604 csky_secondary_reload, csky_register_move_cost):
19605 Use HILO_REGS instead of LO_REGS and HI_REGS.
19607 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
19609 * config/csky/constraints.md ("W"): New constriant for mem operand
19610 with base reg, index register.
19611 ("Q"): Renamed and modified "csky_valid_fpuv2_mem_operand" to
19612 "csky_valid_mem_constraint_operand" to deal with both "Q" and "W"
19614 ("Dv"): New constraint for const double value that can be used at
19616 * config/csky/csky-modes.def (HFmode): New mode.
19617 * config/csky/csky-protos.h (csky_valid_fpuv2_mem_operand): Rename
19618 to "csky_valid_mem_constraint_operand" and support new constraint
19620 (csky_get_movedouble_length): New.
19621 (fpuv3_output_move): New.
19622 (fpuv3_const_double): New.
19623 * config/csky/csky.c (csky_option_override): New arch CK860 with fpv3.
19624 (decompose_csky_address): Refine.
19625 (csky_print_operand): New "CONST_DOUBLE" operand.
19626 (csky_output_move): Support fpv3 instructions.
19627 (csky_get_movedouble_length): New.
19628 (fpuv3_output_move): New.
19629 (fpuv3_const_double): New.
19630 (csky_emit_compare): Cover float comparsion.
19631 (csky_emit_compare_float): Refine.
19632 (csky_vaild_fpuv2_mem_operand): Rename to
19633 "csky_valid_mem_constraint_operand" and support new constraint "W".
19634 (ck860_rtx_costs): New.
19635 (csky_rtx_costs): Add the cost calculation of CK860.
19636 (regno_reg_class): New vregs for fpuv3.
19637 (csky_dbx_regno): Likewise.
19638 (csky_cpu_cpp_builtins): New builtin macro for fpuv3.
19639 (csky_conditional_register_usage): Suporrot fpuv3.
19640 (csky_dwarf_register_span): Suporrot fpuv3.
19641 (csky_init_builtins, csky_mangle_type): Support "__fp16" type.
19642 (ck810_legitimate_index_p): Support fp16.
19643 * config/csky/csky.h (TARGET_TLS): ADD CK860.
19644 (CSKY_VREG_P, CSKY_VREG_LO_P, CSKY_VREG_HI_P): Support fpuv3.
19645 (TARGET_SINGLE_FPU): Support fpuv3.
19646 (TARGET_SUPPORT_FPV3): New.
19647 (FIRST_PSEUDO_REGISTER): Change to 202 to hold the new fpuv3 registers.
19648 (FIXED_REGISTERS, CALL_REALLY_USED_REGISTERS, REGISTER_NAMES,
19649 REG_CLASS_CONTENTS): Support fpuv3.
19650 * config/csky/csky.md (movsf): Move to cksy_insn_fpu.md and refine.
19651 (csky_movsf_fpv2): Likewise.
19652 (ck801_movsf): Likewise.
19653 (csky_movsf): Likewise.
19655 (csky_movdf_fpv2): Likewise.
19656 (ck801_movdf): Likewise.
19657 (csky_movdf): Likewise.
19658 (movsicc): Refine. Use "comparison_operatior" instead of
19659 "ordered_comparison_operatior".
19660 (addsicc): Likewise.
19661 (CSKY_FIRST_VFP3_REGNUM, CSKY_LAST_VFP3_REGNUM): New constant.
19662 (call_value_internal_vh): New.
19663 * config/csky/csky_cores.def (CK860): New arch and cpu.
19668 * config/csky/csky_insn_fpu.md: Refactor. Separate all float patterns
19669 into emit-patterns and match-patterns, remain the emit-patterns here,
19670 and move the match-patterns to csky_insn_fpuv2.md or
19671 csky_insn_fpuv3.md.
19672 * config/csky/csky_insn_fpuv2.md: New file for fpuv2 instructions.
19673 * config/csky/csky_insn_fpuv3.md: New file and new patterns for fpuv3
19675 * config/csky/csky_isa.def (fcr): New.
19680 (CK860): New definition for ck860.
19681 * config/csky/csky_tables.opt (ck860): New processors ck860,
19682 ck860f. And new arch ck860.
19687 * config/csky/predicates.md (csky_float_comparsion_operator): Delete
19688 "geu", "gtu", "leu", "ltu", which will never appear at float comparison.
19689 * config/csky/t-csky-elf: Support 860.
19690 * config/csky/t-csky-linux: Likewise.
19691 * doc/md.texi: Add "Q" and "W" constraints for C-SKY.
19693 2021-05-24 Aaron Sawdey <acsawdey@linux.ibm.com>
19695 * config/rs6000/genfusion.pl (gen_logical_addsubf): Refactor to
19696 add generation of logical-add and add-logical fusion pairs.
19697 * config/rs6000/rs6000-cpus.def: Add new fusion to ISA 3.1 mask
19699 * config/rs6000/rs6000.c (rs6000_option_override_internal): Turn on
19700 logical-add and add-logical fusion by default.
19701 * config/rs6000/rs6000.opt: Add -mpower10-fusion-logical-add and
19702 -mpower10-fusion-add-logical options.
19703 * config/rs6000/fusion.md: Regenerate file.
19705 2021-05-24 Aldy Hernandez <aldyh@redhat.com>
19707 * value-range.cc (irange::legacy_equal_p): Check type when
19708 comparing VR_VARYING types.
19709 (range_tests_legacy): Test comparing VARYING ranges of different
19712 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
19714 * config/aarch64/aarch64.c (neoversen1_tunings):
19715 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
19717 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
19719 * config/aarch64/aarch64.c (aarch64_classify_symbol): Use GOT for
19720 extern weak symbols. Limit symbol offsets for non-GOT symbols with
19723 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
19725 * config/arm/neon.md (vec_load_lanesxi<mode>)
19726 (vec_store_lanexoi<mode>): Move ...
19727 * config/arm/vec-common.md: here.
19729 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
19731 * config/arm/neon.md (vec_load_lanesoi<mode>)
19732 (vec_store_lanesoi<mode>): Move ...
19733 * config/arm/vec-common.md: here.
19735 2021-05-24 liuhongt <hongtao.liu@intel.com>
19738 * config/i386/i386.c (ix86_gimple_fold_builtin): Replacing
19739 stmt with GIMPLE_NOP when lhs doesn't exist.
19741 2021-05-23 Uroš Bizjak <ubizjak@gmail.com>
19744 * config/i386/mmx.md (*push<VI_32:mode>2_rex64):
19745 New instruction pattern.
19746 (*push<VI_32:mode>2): Ditto.
19747 (push splitter for SSE registers): New splitter.
19749 2021-05-23 Andrew Pinski <apinski@marvell.com>
19751 * match.pd ((A & C) != 0 ? D : 0): Limit to non pointer types.
19753 2021-05-22 Aaron Sawdey <acsawdey@linux.ibm.com>
19755 * config/rs6000/genfusion.pl (gen_addadd): Fix incorrect attr types.
19756 * config/rs6000/fusion.md: Regenerate file.
19758 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
19760 * config/rs6000/genfusion.pl (gen_addadd): New function.
19761 * config/rs6000/fusion.md: Regenerate file.
19762 * config/rs6000/rs6000-cpus.def: Add
19763 OPTION_MASK_P10_FUSION_2ADD to masks.
19764 * config/rs6000/rs6000.c (rs6000_option_override_internal):
19765 Handle default value of OPTION_MASK_P10_FUSION_2ADD.
19766 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2add.
19768 2021-05-21 Jakub Jelinek <jakub@redhat.com>
19770 PR middle-end/99928
19771 * tree.h (OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET): Define.
19772 * gimplify.c (enum gimplify_omp_var_data): Fix up
19773 GOVD_MAP_HAS_ATTACHMENTS value, add GOVD_FIRSTPRIVATE_IMPLICIT.
19774 (omp_lastprivate_for_combined_outer_constructs): If combined target
19775 has GOVD_FIRSTPRIVATE_IMPLICIT set for the decl, change it to
19776 GOVD_MAP | GOVD_SEEN.
19777 (gimplify_scan_omp_clauses): Set GOVD_FIRSTPRIVATE_IMPLICIT for
19778 firstprivate clauses with OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT.
19779 (gimplify_adjust_omp_clauses): For firstprivate clauses with
19780 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT either clear that bit and
19781 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET too, or remove it and
19782 let it be replaced by implicit map clause.
19784 2021-05-21 Jakub Jelinek <jakub@redhat.com>
19786 PR middle-end/99928
19787 * gimplify.c (omp_lastprivate_for_combined_outer_constructs): New
19789 (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LASTPRIVATE>: Use it.
19790 (gimplify_omp_for): Likewise.
19792 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19794 PR middle-end/90115
19795 * omp-low.c (oacc_privatization_candidate_p): Reject 'static',
19796 'external' in blocks.
19798 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19800 PR middle-end/90115
19801 * flag-types.h (enum openacc_privatization): New.
19802 * params.opt (-param=openacc-privatization): New.
19803 * doc/invoke.texi (openacc-privatization): Document it.
19804 * omp-general.h (get_openacc_privatization_dump_flags): New
19806 * omp-low.c (oacc_privatization_candidate_p): Add diagnostics.
19807 * omp-offload.c (execute_oacc_device_lower)
19808 <IFN_UNIQUE_OACC_PRIVATE>: Re-work diagnostics.
19809 * target.def (goacc.adjust_private_decl): Add 'location_t'
19811 * doc/tm.texi: Regenerate.
19812 * config/gcn/gcn-protos.h (gcn_goacc_adjust_private_decl): Adjust.
19813 * config/gcn/gcn-tree.c (gcn_goacc_adjust_private_decl): Likewise.
19814 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl):
19815 Likewise. Preserve it for...
19816 (nvptx_goacc_expand_var_decl): ... use here.
19818 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19820 * doc/sourcebuild.texi (Other attributes): Document '__OPTIMIZE__'
19823 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19825 PR middle-end/90115
19826 * omp-low.c (oacc_privatization_candidate_p): New function.
19827 (oacc_privatization_scan_clause_chain)
19828 (oacc_privatization_scan_decl_chain): Use it. Also
19829 'gcc_checking_assert' that we're not seeing duplicates.
19831 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19833 PR middle-end/90115
19834 * omp-offload.c (execute_oacc_device_lower): Skip processing if no
19837 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19839 PR middle-end/90115
19840 * omp-offload.c (execute_oacc_device_lower): Explain.
19842 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19844 PR middle-end/90115
19845 * omp-offload.c (execute_oacc_device_lower)
19846 <IFN_UNIQUE_OACC_PRIVATE>: Diagnose and handle for 'level == -1'
19848 * internal-fn.c (expand_UNIQUE): Don't expect
19849 'IFN_UNIQUE_OACC_PRIVATE'.
19851 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19853 PR middle-end/90115
19854 * omp-low.c (lower_omp_for): Don't evaluate OpenMP 'for' clauses.
19856 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
19858 PR middle-end/90115
19859 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl)
19860 (nvptx_goacc_expand_var_decl): Tighten.
19862 2021-05-21 Julian Brown <julian@codesourcery.com>
19863 Chung-Lin Tang <cltang@codesourcery.com>
19864 Thomas Schwinge <thomas@codesourcery.com>
19866 PR middle-end/90115
19867 * doc/tm.texi.in (TARGET_GOACC_EXPAND_VAR_DECL)
19868 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Add documentation hooks.
19869 * doc/tm.texi: Regenerate.
19870 * expr.c (expand_expr_real_1): Expand decls using the
19871 expand_var_decl OpenACC hook if defined.
19872 * internal-fn.c (expand_UNIQUE): Handle IFN_UNIQUE_OACC_PRIVATE.
19873 * internal-fn.h (IFN_UNIQUE_CODES): Add OACC_PRIVATE.
19874 * omp-low.c (omp_context): Add oacc_privatization_candidates
19876 (lower_oacc_reductions): Add PRIVATE_MARKER parameter. Insert
19878 (lower_oacc_head_tail): Add PRIVATE_MARKER parameter. Modify
19879 private marker's gimple call arguments, and pass it to
19880 lower_oacc_reductions.
19881 (oacc_privatization_scan_clause_chain)
19882 (oacc_privatization_scan_decl_chain, lower_oacc_private_marker):
19884 (lower_omp_for, lower_omp_target, lower_omp_1): Use these.
19885 * omp-offload.c (convert.h): Include.
19886 (oacc_loop_xform_head_tail): Treat private-variable markers like
19887 fork/join when transforming head/tail sequences.
19888 (struct var_decl_rewrite_info): Add struct.
19889 (oacc_rewrite_var_decl, is_sync_builtin_call): New functions.
19890 (execute_oacc_device_lower): Support rewriting gang-private
19891 variables using target hook, and fix up addr_expr and var_decl
19893 * target.def (adjust_private_decl, expand_var_decl): New hooks.
19894 * config/gcn/gcn-protos.h (gcn_goacc_adjust_gangprivate_decl):
19896 (gcn_goacc_adjust_private_decl): ...this.
19897 * config/gcn/gcn-tree.c (gcn_goacc_adjust_gangprivate_decl):
19899 (gcn_goacc_adjust_private_decl): ...this. Add LEVEL parameter.
19900 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_GANGPRIVATE_DECL): Rename
19901 definition using gcn_goacc_adjust_gangprivate_decl...
19902 (TARGET_GOACC_ADJUST_PRIVATE_DECL): ...to this, using
19903 gcn_goacc_adjust_private_decl.
19904 * config/nvptx/nvptx.c (tree-pretty-print.h): Include.
19905 (gang_private_shared_size): New global variable.
19906 (gang_private_shared_align): Likewise.
19907 (gang_private_shared_sym): Likewise.
19908 (gang_private_shared_hmap): Likewise.
19909 (nvptx_option_override): Initialize these.
19910 (nvptx_file_end): Output gang_private_shared_sym.
19911 (nvptx_goacc_adjust_private_decl, nvptx_goacc_expand_var_decl):
19913 (nvptx_set_current_function): Clear gang_private_shared_hmap.
19914 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Define hook.
19915 (TARGET_GOACC_EXPAND_VAR_DECL): Likewise.
19917 2021-05-21 H.J. Lu <hjl.tools@gmail.com>
19919 * config/i386/i386-modes.def (MAX_BITSIZE_MODE_ANY_INT): Removed.
19921 2021-05-21 Richard Biener <rguenther@suse.de>
19922 H.J. Lu <hjl.tools@gmail.com>
19924 PR middle-end/90773
19925 * expr.c (expand_constructor): Elide expand_constructor if
19926 move by pieces is preferred.
19928 2021-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19930 * config/aarch64/aarch64-builtins.c (aarch64_call_properties):
19931 Take a flag and mode value as arguments.
19932 (aarch64_modifies_global_state_p): Likewise.
19933 (aarch64_reads_global_state_p): Likewise.
19934 (aarch64_could_trap_p): Likewise.
19935 (aarch64_get_attributes): Likewise.
19936 (aarch64_init_simd_builtins): Adjust callsite of above.
19937 (aarch64_init_fcmla_laneq_builtins): Use aarch64_get_attributes to get
19938 function attributes to apply to builtins.
19939 (aarch64_init_crc32_builtins): Likewise.
19940 (aarch64_init_builtin_rsqrt): Likewise.
19942 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
19944 * config/rs6000/rs6000.md (define_attr "type"): Add types for fusion.
19945 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10): Use new fusion types.
19946 (gen_2logical): Use new fusion types.
19947 * config/rs6000/fusion.md: Regenerate.
19949 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
19952 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
19953 Handle V4QI and V2HI modes.
19954 (ix86_expand_sse_movcc): Ditto.
19955 * config/i386/mmx.md (*<sat_plusminus:insn><VI_32:mode>3):
19956 New instruction pattern.
19957 (*eq<VI_32:mode>3): Ditto.
19958 (*gt<VI_32:mode>3): Ditto.
19959 (*xop_pcmov_<VI_32:mode>): Ditto.
19960 (mmx_pblendvb32): Ditto.
19961 (mmx_pblendvb64): Rename from mmx_pblendvb.
19962 (vec_cmp<VI_32:mode><VI_32:mode>): New expander.
19963 (vec_cmpu<VI_32:mode><VI_32:mode>): Ditto.
19964 (vcond<VI_32:mode><VI_32:mode>): Ditto.
19965 (vcondu<VI_32:mode><VI_32:mode>): Ditto.
19966 (vcond_mask_<VI_32:mode><VI_32:mode>): Ditto.
19968 2021-05-21 Jakub Jelinek <jakub@redhat.com>
19970 PR tree-optimization/94589
19971 * tree-ssa-phiopt.c (spaceship_replacement): For integral rhs1 and
19972 rhs2, treat x <= 4 equivalently to x < 5 etc. In cmp1 and cmp2 (if
19973 not the same as cmp3) treat <= the same as < and >= the same as >.
19974 Don't require that cond2_phi_edge is true edge, instead take
19975 false/true edges into account based on cmp1/cmp2 comparison kinds.
19977 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
19980 * config/i386/mmx.md (SMAXMIN_MMXMODEI): New mode iterator.
19981 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Macroize expander
19982 from <smaxmin:code>v4hi3> and <smaxmin:code><MMXMODE14:mode>3
19983 using SMAXMIN_MMXMODEI mode iterator.
19984 (*<smaxmin:code>v4qi3): New insn pattern.
19985 (*<smaxmin:code>v2hi3): Ditto.
19986 (SMAXMIN_VI_32): New mode iterator.
19987 (<smaxmin:code><SMAXMIN_VI_32>mode3): New expander.
19988 (UMAXMIN_MMXMODEI): New mode iterator.
19989 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Macroize expander
19990 from <umaxmin:code>v8qi3> and <umaxmin:code><MMXMODE24:mode>3
19991 using UMAXMIN_MMXMODEI mode iterator.
19992 (*<umaxmin:code>v4qi3): New insn pattern.
19993 (*<umaxmin:code>v2hi3): Ditto.
19994 (UMAXMIN_VI_32): New mode iterator.
19995 (<umaxmin:code><UMAXMIN_VI_32>mode3): New expander.
19996 (abs<VI_32:mode>2): New insn pattern.
19997 (ssse3_abs<MMXMODEI:mode>2, abs<MMXMODEI:mode>2): Move from ...
19998 * config/i386/sse.md: ... here.
20000 2021-05-20 Clement Chigot <clement.chigot@atos.net>
20001 David Edelsohn <dje.gcc@gmail.com>
20003 * collect2.c (scan_prog_file): Issue non-fatal warning for
20006 2021-05-20 Jonathan Wakely <jwakely@redhat.com>
20008 * doc/invoke.texi (-Wno-c++11-extensions)
20009 (-Wno-c++14-extensions, -Wno-c++17-extensions)
20010 (-Wno-c++20-extensions, -Wno-c++23-extensions): Document
20013 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
20015 * config/c6x/c6x.c (c6x_output_file_unwind): Use dwarf_debuginfo_p.
20016 * config/darwin.c (darwin_override_options): Likewise.
20017 * config/i386/cygming.h (DBX_REGISTER_NUMBER): Likewise.
20018 * config/i386/darwin.h (DBX_REGISTER_NUMBER): Likewise.
20019 (DWARF2_FRAME_REG_OUT): Likewise.
20020 * config/mips/mips.c (mips_output_filename): Likewise.
20021 * config/rs6000/rs6000.c (rs6000_xcoff_declare_function_name):
20023 (rs6000_dbx_register_number): Likewise.
20024 * dbxout.c: Include flags.h.
20025 * dwarf2cfi.c (cfi_label_required_p): Likewise.
20026 (dwarf2out_do_frame): Likewise.
20027 * except.c: Include flags.h.
20028 * final.c (dwarf2_debug_info_emitted_p): Likewise.
20029 (final_scan_insn_1): Likewise.
20030 * flags.h (dwarf_debuginfo_p): New function declaration.
20031 * opts.c (dwarf_debuginfo_p): New function definition.
20032 * targhooks.c (default_debug_unwind_info): Use dwarf_debuginfo_p.
20033 * toplev.c (process_options): Likewise.
20035 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
20037 * common.opt: Change type to support bitmasks.
20038 * flag-types.h (enum debug_info_type): Rename enumerator constants.
20039 (NO_DEBUG): New bitmask.
20040 (DBX_DEBUG): Likewise.
20041 (DWARF2_DEBUG): Likewise.
20042 (XCOFF_DEBUG): Likewise.
20043 (VMS_DEBUG): Likewise.
20044 (VMS_AND_DWARF2_DEBUG): Likewise.
20045 * flags.h (debug_set_to_format): New function declaration.
20046 (debug_set_count): Likewise.
20047 (debug_set_names): Likewise.
20048 * opts.c (debug_type_masks): Array of bitmasks for debug formats.
20049 (debug_set_to_format): New function definition.
20050 (debug_set_count): Likewise.
20051 (debug_set_names): Likewise.
20052 (set_debug_level): Update access to debug_type_names.
20053 * toplev.c: Likewise.
20055 2021-05-20 Martin Sebor <msebor@redhat.com>
20057 PR middle-end/100684
20058 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Handle C++ lambda.
20060 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
20063 * config/i386/i386.md (isa): Remove x64_bmi.
20064 (enabled): Remove x64_bmi.
20065 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
20066 Remove general register alternative.
20067 (*andnot<VI_32:mode>3): Ditto.
20068 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
20069 (*<any_logic:code><VI_32:mode>3): Ditto.
20071 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
20073 * config/arm/arm.c: Include head files tree-vectorizer.h and
20076 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
20079 * config/i386/mmx.md (Yv_Yw): Revert adding V4QI and V2HI modes.
20080 (*<plusminus:insn><VI_32:mode>3): Use Yw instad of <Yv_Yw> constrint.
20081 (<s>mulv4hi3_highpart): New expander.
20082 (*<s>mulv2hi3_highpart): New insn pattern.
20083 (<s>mulv2hi3_higpart): New expander.
20084 (*<any_shift:insn>v2hi3): New insn pattern.
20085 (<any_shift:insn>v2hi3): New expander.
20086 * config/i386/sse.md (smulhrsv2hi3): New expander.
20087 (*smulhrsv2hi3): New insn pattern.
20089 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
20091 * doc/invoke.texi (vect-inner-loop-cost-factor): Document new
20093 * params.opt (vect-inner-loop-cost-factor): New.
20094 * targhooks.c (default_add_stmt_cost): Replace hardcoded factor
20095 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR, include head file
20096 tree-vectorizer.h and its required ones.
20097 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Replace
20098 hardcoded factor 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR.
20099 * config/arm/arm.c (arm_add_stmt_cost): Likewise.
20100 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
20101 * config/rs6000/rs6000.c (rs6000_add_stmt_cost): Likewise.
20102 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
20104 (_loop_vec_info::_loop_vec_info): Init inner_loop_cost_factor.
20105 * tree-vectorizer.h (_loop_vec_info): Add inner_loop_cost_factor.
20106 (LOOP_VINFO_INNER_LOOP_COST_FACTOR): New macro.
20108 2021-05-20 Christophe Lyon <christophe.lyon@linaro.org>
20109 Torbjörn Svensson <torbjorn.svensson@st.com>
20112 * doc/cpp.texi (Common Predefined Macros): Document __FILE_NAME__.
20114 2021-05-20 Jakub Jelinek <jakub@redhat.com>
20116 PR middle-end/99928
20117 * gimplify.c (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LINEAR>: For
20118 explicit linear clause when combined with target, make it map(tofrom:)
20119 instead of no clause or firstprivate.
20121 2021-05-20 Jakub Jelinek <jakub@redhat.com>
20123 PR tree-optimization/94589
20124 * match.pd ((X & Y) == X -> (X & ~Y) == 0): Simplify even in presence
20125 of integral conversions.
20127 2021-05-19 Andrew MacLeod <amacleod@redhat.com>
20129 * gimple-range.cc (fur_source::get_operand): New.
20130 (gimple_range_fold): Delete.
20131 (fold_using_range::fold_stmt): Move from gimple_ranger::calc_stmt.
20132 (fold_using_range::range_of_range_op): Move from gimple_ranger.
20133 (fold_using_range::range_of_address): Ditto.
20134 (fold_using_range::range_of_phi): Ditto.
20135 (fold_using_range::range_of_call): Ditto.
20136 (fold_using_range::range_of_builtin_ubsan_call): Move from
20137 range_of_builtin_ubsan_call.
20138 (fold_using_range::range_of_builtin_call): Move from
20139 range_of_builtin_call.
20140 (gimple_ranger::range_of_builtin_call): Delete.
20141 (fold_using_range::range_of_cond_expr): Move from gimple_ranger.
20142 (gimple_ranger::fold_range_internal): New.
20143 (gimple_ranger::range_of_stmt): Use new fold_using_range API.
20144 (fold_using_range::range_of_ssa_name_with_loop_info): Move from
20145 gimple_ranger. Improve ranges of SSA_NAMES when possible.
20146 * gimple-range.h (gimple_ranger): Remove various range_of routines.
20147 (class fur_source): New.
20148 (class fold_using_range): New.
20149 (fur_source::fur_source): New.
20151 * vr-values.c (vr_values::extract_range_basic): Use fold_using_range
20152 instead of range_of_builtin_call.
20154 2021-05-19 Jonathan Wakely <jwakely@redhat.com>
20156 * doc/cpp.texi (Common Predefined Macros): Update documentation
20157 for the __GXX_EXPERIMENTAL_CXX0X__ macro.
20159 2021-05-19 Alex Coplan <alex.coplan@arm.com>
20162 * config/arm/arm.md (nonsecure_call_internal): Always ensure
20163 callee's address is in a register.
20165 2021-05-19 Geng Qi <gengqi@linux.alibaba.com>
20167 * common/config/riscv/riscv-common.c
20168 (riscv_subset_list::parsing_subset_version): Properly parse the letter
20170 (riscv_subset_list::parse_std_ext,
20171 riscv_subset_list::parse_multiletter_ext): To handle errors generated
20172 in riscv_subset_list::parsing_subset_version.
20174 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
20176 * config/aarch64/aarch64-simd.md: Use "neon_move_narrow_q"
20177 type attribute in patterns generating XTN(2).
20179 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
20181 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_pack_trunc_<mode>):
20182 Remove as duplicate of...
20183 (aarch64_xtn<mode>): This.
20184 (aarch64_xtn2<mode>_le): Move position in file.
20185 (aarch64_xtn2<mode>_be): Move position in file.
20186 (aarch64_xtn2<mode>): Move position in file.
20187 (vec_pack_trunc_<mode>): Define as an expander.
20189 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
20191 * config/aarch64/aarch64-simd-builtins.def: Split builtin
20192 generation for aarch64_<sur>q<r>shr<u>n_n<mode> pattern into
20193 separate scalar and vector generators.
20194 * config/aarch64/aarch64-simd.md
20195 (aarch64_<sur>q<r>shr<u>n_n<mode>): Define as an expander and
20197 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_le): This and...
20198 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_be): This.
20199 * config/aarch64/iterators.md: Define SD_HSDI iterator.
20201 2021-05-19 Jonathn Wright <jonathan.wright@arm.com>
20203 * config/aarch64/aarch64-simd.md: Use UNSPEC_SQXTUN instead
20205 * config/aarch64/iterators.md: Remove UNSPEC_SQXTUN2.
20207 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
20209 * config/aarch64/aarch64-simd.md (aarch64_<sur>q<r>shr<u>n2_n<mode>):
20210 Implement as an expand emitting a big/little endian
20211 instruction pattern.
20212 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_le): Define.
20213 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_be): Define.
20215 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
20217 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn2<mode>):
20218 Implement as an expand emitting a big/little endian
20219 instruction pattern.
20220 (aarch64_<sur><addsub>hn2<mode>_insn_le): Define.
20221 (aarch64_<sur><addsub>hn2<mode>_insn_be): Define.
20222 * config/aarch64/iterators.md: Remove UNSPEC_[R]ADDHN2 and
20223 UNSPEC_[R]SUBHN2 unspecs and ADDSUBHN2 iterator.
20225 2021-05-19 Richard Biener <rguenther@suse.de>
20227 PR middle-end/100672
20228 * fold-const.c (fold_negate_expr_1): Use element_precision.
20229 (negate_expr_p): Likewise.
20231 2021-05-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
20233 * config/aarch64/iterators.md (SVE_PRED_LOAD): New iterator.
20234 (pred_load): New int attribute.
20235 * config/aarch64/aarch64-sve.md
20236 (aarch64_load_<ANY_EXTEND:optab><SVE_HSDI:mode><SVE_PARTIAL_I:mode>): Use
20237 SVE_PRED_LOAD enum iterator and corresponding pred_load attribute.
20238 * config/aarch64/aarch64-sve-builtins-base.cc (expand): Update call to
20239 code_for_aarch64_load.
20241 2021-05-19 Richard Biener <rguenther@suse.de>
20243 * cfgexpand.c (discover_nonconstant_array_refs_r): Make
20244 sure TARGET_MEM_REF bases are expanded as memory.
20245 * tree-ssa-operands.c (operands_scanner::get_tmr_operands):
20246 Do not mark TARGET_MEM_REF bases addressable.
20247 * tree-ssa.c (non_rewritable_mem_ref_base): Handle
20248 TARGET_MEM_REF bases as never rewritable.
20249 * gimple-walk.c (walk_stmt_load_store_addr_ops): Do not
20250 walk TARGET_MEM_REF bases as address-takens.
20251 * tree-ssa-dce.c (ref_may_be_aliased): Handle TARGET_MEM_REF.
20253 2021-05-19 Richard Biener <rguenther@suse.de>
20255 * builtins.c (get_object_alignment_1): Strip outer
20257 * tree-dfa.c (get_ref_base_and_extent): Handle outer
20258 WITH_SIZE_EXPR for size processing and process the
20260 * tree-ssa-alias.c (ao_ref_base_alias_set): Strip
20261 outer WITH_SIZE_EXPR.
20262 (ao_ref_base_alias_ptr_type): Likewise.
20263 (refs_may_alias_p_2): Allow WITH_SIZE_EXPR in ref->ref
20264 and handle that accordingly, stripping it for the
20265 core alias workers.
20266 * tree.c (get_base_address): Handle WITH_SIZE_EXPR by
20267 looking through it instead of returning NULL.
20269 2021-05-19 Jakub Jelinek <jakub@redhat.com>
20271 PR middle-end/100576
20272 * builtins.c (check_read_access): Convert bound to size_type_node if
20275 2021-05-19 Richard Biener <rguenther@suse.de>
20277 * tree-cfg.c (verify_types_in_gimple_min_lval): Inline...
20278 (verify_types_in_gimple_reference): ... here. Sanitize.
20279 (verify_gimple_call): Verify references in LHS and arguments.
20280 (verify_gimple_assign_single): Reject WITH_SIZE_EXPR.
20282 2021-05-19 Uroš Bizjak <ubizjak@gmail.com>
20284 * config/i386/i386.h (VALID_INT_MODE_P):
20285 Add V8QI, V4HI and V2SI modes for TARGET_64BIT.
20286 * config/i386/i386.md (isa): Add x64_bmi.
20287 (enabled): Handle x64_bmi.
20288 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
20289 Add alternative using 64bit general registers.
20290 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
20292 2021-05-19 Jakub Jelinek <jakub@redhat.com>
20294 PR middle-end/99928
20295 * tree.h (OMP_MASTER_COMBINED): Define.
20296 * gimplify.c (gimplify_scan_omp_clauses): Rewrite lastprivate
20297 handling for outer combined/composite constructs to a loop.
20298 Handle lastprivate on combined target.
20299 (gimplify_expr): Formatting fix.
20301 2021-05-19 Xionghu Luo <luoxhu@linux.ibm.com>
20303 * passes.def: Add sink_code pass before store_merging.
20304 * tree-ssa-sink.c (pass_sink_code:clone): New.
20306 2021-05-18 Bill Schmidt <wschmidt@linux.ibm.com>
20308 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Remove call to
20309 rs6000_special_adjust_field_align_p.
20310 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
20311 * config/rs6000/rs6000-call.c (rs6000_function_arg_boundary):
20312 Remove ABI warning.
20313 (rs6000_function_arg): Likewise.
20314 * config/rs6000/rs6000-protos.h
20315 (rs6000_special_adjust_field_align_p): Remove prototype.
20316 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
20318 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Remove call to
20319 rs6000_special_adjust_field_align_p.
20321 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
20324 * config/i386/i386.h (VALID_SSE2_REG_MODE):
20325 Add V4QI and V2HI modes.
20326 (VALID_INT_MODE_P): Ditto.
20327 * config/i386/mmx.md (VI_32): New mode iterator.
20328 (mmxvecsize): Handle V4QI and V2HI.
20330 (mov<VI_32:mode>): New expander.
20331 (*mov<mode>_internal): New insn pattern.
20332 (movmisalign<VI_32:mode>): New expander.
20333 (neg<VI_32:mode>): New expander.
20334 (<plusminus:insn><VI_32:mode>3): New expander.
20335 (*<plusminus:insn><VI_32:mode>3): New insn pattern.
20336 (mulv2hi3): New expander.
20337 (*mulv2hi3): New insn pattern.
20338 (one_cmpl<VI_32:mode>2): New expander.
20339 (*andnot<VI_32:mode>3): New insn pattern.
20340 (<any_logic:code><VI_32:mode>3): New expander.
20341 (*<any_logic:code><VI_32:mode>3): New insn pattern.
20343 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
20345 * config/i386/sse.md (<any_extend:insn>v4qiv4di2):
20346 Fix a mode mismatch with operand 1.
20348 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
20351 * config/i386/i386-expand.c (split_double_mode): Return
20352 temporary register when simplify_gen_subreg fails with
20353 the high half od the paradoxical subreg.
20355 2021-05-18 Richard Biener <rguenther@suse.de>
20357 * cfgexpand.c (expand_one_var): Pass in forced_stack_var
20358 and honor it when expanding.
20359 (expand_used_vars_for_block): Pass through forced_stack_var.
20360 (expand_used_vars): Likewise.
20361 (discover_nonconstant_array_refs_r): Set bits in
20362 forced_stack_vars instead of marking vars TREE_ADDRESSABLE.
20363 (avoid_type_punning_on_regs): Likewise.
20364 (discover_nonconstant_array_refs): Likewise.
20365 (pass_expand::execute): Create and pass down forced_stack_var
20366 bitmap. For parameters and returns temporarily set
20367 TREE_ADDRESSABLE when expand_function_start.
20369 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
20371 * doc/sourcebuild.texi: Document 'dg-note'.
20373 2021-05-18 Tobias Burnus <tobias@codesourcery.com>
20376 * configure: Regenerate.
20377 * configure.ac (BUILD_CFLAG, BUILD_CXXFLAGS): Add $(CFLAGS-$@).
20379 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
20381 * gimple.h (is_gimple_omp_oacc): Tighten.
20382 * omp-low.c (check_omp_nesting_restrictions): Adjust.
20384 2021-05-18 Richard Biener <rguenther@suse.de>
20386 * tree-ssa-operands.c (mark_address_taken): Simplify.
20388 2021-05-18 Martin Liska <mliska@suse.cz>
20390 * config/gcn/mkoffload.c (STR): Redefine.
20391 * config/i386/intelmic-mkoffload.c (STR): Likewise.
20392 * config/nvptx/mkoffload.c (STR): Likewise.
20394 2021-05-18 Martin Liska <mliska@suse.cz>
20396 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
20397 Use startswith function instead of strncmp.
20398 * common/config/bfin/bfin-common.c (bfin_handle_option): Likewise.
20399 * common/config/riscv/riscv-common.c (riscv_subset_list::parse): Likewise.
20400 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): Likewise.
20401 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Likewise.
20402 * config/alpha/alpha.c (alpha_elf_section_type_flags): Likewise.
20403 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
20404 * config/arm/arm.c (arm_file_start): Likewise.
20405 (arm_valid_target_attribute_rec): Likewise.
20406 (thumb1_md_asm_adjust): Likewise.
20407 * config/arm/driver-arm.c (host_detect_local_cpu): Likewise.
20408 * config/avr/avr.c (STR_PREFIX_P): Likewise.
20409 (avr_set_current_function): Likewise.
20410 (avr_handle_addr_attribute): Likewise.
20411 (avr_asm_output_aligned_decl_common): Likewise.
20412 (avr_asm_named_section): Likewise.
20413 (avr_section_type_flags): Likewise.
20414 (avr_asm_select_section): Likewise.
20415 * config/c6x/c6x.c (c6x_in_small_data_p): Likewise.
20416 (c6x_section_type_flags): Likewise.
20417 * config/darwin-c.c (darwin_cfstring_ref_p): Likewise.
20418 (darwin_objc_declare_unresolved_class_reference): Likewise.
20419 (darwin_objc_declare_class_definition): Likewise.
20420 * config/darwin.c (indirect_data): Likewise.
20421 (darwin_encode_section_info): Likewise.
20422 (darwin_objc2_section): Likewise.
20423 (darwin_objc1_section): Likewise.
20424 (machopic_select_section): Likewise.
20425 (darwin_globalize_label): Likewise.
20426 (darwin_label_is_anonymous_local_objc_name): Likewise.
20427 (darwin_asm_named_section): Likewise.
20428 (darwin_asm_output_dwarf_offset): Likewise.
20429 * config/frv/frv.c (frv_string_begins_with): Likewise.
20430 (frv_in_small_data_p): Likewise.
20431 * config/gcn/mkoffload.c (STR): Likewise.
20433 * config/i386/i386-builtins.c (get_builtin_code_for_version): Likewise.
20434 * config/i386/i386-options.c (ix86_option_override_internal): Likewise.
20435 * config/i386/i386.c (x86_64_elf_section_type_flags): Likewise.
20436 (ix86_md_asm_adjust): Likewise.
20437 * config/i386/intelmic-mkoffload.c (STR): Likewise.
20438 * config/i386/winnt.c (i386_pe_asm_named_section): Likewise.
20439 (i386_pe_file_end): Likewise.
20440 * config/ia64/ia64.c (ia64_in_small_data_p): Likewise.
20441 (ia64_section_type_flags): Likewise.
20442 * config/mips/driver-native.c (host_detect_local_cpu): Likewise.
20443 * config/mips/mips.c (mips_handle_interrupt_attr): Likewise.
20444 (mips16_stub_function_p): Likewise.
20445 (mips_function_rodata_section): Likewise.
20446 * config/msp430/msp430.c (msp430_mcu_name): Likewise.
20447 (msp430_function_section): Likewise.
20448 (msp430_section_type_flags): Likewise.
20449 (msp430_expand_helper): Likewise.
20450 * config/nios2/nios2.c (nios2_small_section_name_p): Likewise.
20451 (nios2_valid_target_attribute_rec): Likewise.
20452 * config/nvptx/mkoffload.c (process): Likewise.
20454 * config/pa/som.h: Likewise.
20455 * config/pdp11/pdp11.c (pdp11_output_ident): Likewise.
20456 * config/riscv/riscv.c (riscv_elf_select_rtx_section): Likewise.
20457 * config/rs6000/rs6000.c (VTABLE_NAME_P): Likewise.
20458 (rs6000_inner_target_options): Likewise.
20459 * config/s390/driver-native.c (s390_host_detect_local_cpu): Likewise.
20460 * config/sparc/driver-sparc.c (host_detect_local_cpu): Likewise.
20461 * config/vax/vax.c (vax_output_int_move): Likewise.
20462 * config/vms/vms-ld.c (startswith): Likewise.
20463 (process_args): Likewise.
20465 * config/vms/vms.c: Likewise.
20467 2021-05-18 Jakub Jelinek <jakub@redhat.com>
20469 PR rtl-optimization/100590
20470 * regcprop.c (copyprop_hardreg_forward_1): Only DCE dead sets if
20471 they are NONJUMP_INSN_P.
20473 2021-05-18 Jakub Jelinek <jakub@redhat.com>
20476 * function.c (push_dummy_function): Set DECL_ARTIFICIAL and
20477 DECL_ASSEMBLER_NAME on the fn_decl.
20479 2021-05-18 Jakub Jelinek <jakub@redhat.com>
20481 PR tree-optimization/94589
20482 * tree-ssa-phiopt.c (spaceship_replacement): Pattern match
20483 phi result used in (res & ~1) == 0 comparison as res >= 0 as
20484 res == 2 would be UB with -ffinite-math-only.
20486 2021-05-18 Martin Liska <mliska@suse.cz>
20488 * Makefile.in: genversion.o should depend on DATESTAMP.
20490 2021-05-18 Claudiu Zissulescu <claziss@synopsys.com>
20492 * config/arc/simdext.md (negv2si2): Remove round bracket.
20494 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
20496 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Define
20497 _Bool as macro expanding to _Bool.
20499 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
20502 * tree.c (build_reference_type_for_mode)
20503 (build_pointer_type_for_mode): Pick pointer mode if MODE argument
20505 (build_reference_type, build_pointer_type): Invoke
20506 build_*_type_for_mode with VOIDmode.
20508 2021-05-17 Andrew MacLeod <amacleod@redhat.com>
20510 PR tree-optimization/100512
20511 * gimple-range-cache.cc (ranger_cache::set_global_range): Mark const
20512 and non-zero pointer ranges as invariant.
20513 * gimple-range.cc (gimple_ranger::range_of_stmt): Remove pointer
20514 processing from here.
20516 2021-05-17 Tom de Vries <tdevries@suse.de>
20519 * config/nvptx/nvptx-protos.h (nvptx_output_atomic_insn): Declare
20520 * config/nvptx/nvptx.c (nvptx_output_barrier)
20521 (nvptx_output_atomic_insn): New function.
20522 (nvptx_print_operand): Add support for 'B'.
20523 * config/nvptx/nvptx.md: Use nvptx_output_atomic_insn for atomic
20526 2021-05-17 Aldy Hernandez <aldyh@redhat.com>
20528 PR tree-optimization/100349
20529 * vr-values.c (bounds_of_var_in_loop): Bail if scev returns
20532 2021-05-17 Tamar Christina <tamar.christina@arm.com>
20534 * config/aarch64/driver-aarch64.c (DEFAULT_ARCH): New.
20535 (host_detect_local_cpu): Use it.
20537 2021-05-17 Martin Liska <mliska@suse.cz>
20539 * doc/invoke.texi: Add 2 missing dots.
20541 2021-05-17 Marius Hillenbrand <mhillen@linux.ibm.com>
20543 PR bootstrap/100552
20544 * configure.ac: Replace pattern substitution with call to sed.
20545 * configure: Regenerate.
20547 2021-05-17 Richard Biener <rguenther@suse.de>
20549 PR middle-end/100582
20550 * tree.c (array_at_struct_end_p): Get to the base of the
20551 reference before looking for the underlying decl.
20553 2021-05-17 Joern Rennecke <joern.rennecke@embecosm.com>
20555 * genoutput.c (validate_insn_alternatives) Make "wrong number of
20556 alternatives" message more specific, and remove assumption on where
20559 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
20561 * config/arm/iterators.md (V16): New iterator.
20562 (VH_cvtto): New iterator.
20563 (v_cmp_result): Added V4HF and V8HF support.
20564 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result>): Use VDQWH.
20565 (vcond<mode><mode>): Likewise.
20566 (vcond_mask_<mode><v_cmp_result>): Likewise.
20567 (vcond<VH_cvtto><mode>): New expander.
20569 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
20571 * config/arm/arm-protos.h (arm_expand_vector_compare): Update
20573 * config/arm/arm.c (arm_expand_vector_compare): Add support for
20575 (arm_expand_vcond): Likewise.
20576 * config/arm/iterators.md (supf): Remove VCMPNEQ_S, VCMPEQQ_S,
20577 VCMPEQQ_N_S, VCMPNEQ_N_S.
20578 (VCMPNEQ, VCMPEQQ, VCMPEQQ_N, VCMPNEQ_N): Remove.
20579 * config/arm/mve.md (@mve_vcmp<mve_cmp_op>q_<mode>): Add '@' prefix.
20580 (@mve_vcmp<mve_cmp_op>q_f<mode>): Likewise.
20581 (@mve_vcmp<mve_cmp_op>q_n_f<mode>): Likewise.
20582 (@mve_vpselq_<supf><mode>): Likewise.
20583 (@mve_vpselq_f<mode>"): Likewise.
20584 * config/arm/neon.md (vec_cmp<mode><v_cmp_result): Enable for MVE
20585 and move to vec-common.md.
20586 (vec_cmpu<mode><mode>): Likewise.
20587 (vcond<mode><mode>): Likewise.
20588 (vcond<V_cvtto><mode>): Likewise.
20589 (vcondu<mode><v_cmp_result>): Likewise.
20590 (vcond_mask_<mode><v_cmp_result>): Likewise.
20591 * config/arm/unspecs.md (VCMPNEQ_U, VCMPNEQ_S, VCMPEQQ_S)
20592 (VCMPEQQ_N_S, VCMPNEQ_N_S, VCMPEQQ_U, CMPEQQ_N_U, VCMPNEQ_N_U)
20593 (VCMPGEQ_N_S, VCMPGEQ_S, VCMPGTQ_N_S, VCMPGTQ_S, VCMPLEQ_N_S)
20594 (VCMPLEQ_S, VCMPLTQ_N_S, VCMPLTQ_S, VCMPCSQ_N_U, VCMPCSQ_U)
20595 (VCMPHIQ_N_U, VCMPHIQ_U): Remove.
20596 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result): Moved
20598 (vec_cmpu<mode><mode>): Likewise.
20599 (vcond<mode><mode>): Likewise.
20600 (vcond<V_cvtto><mode>): Likewise.
20601 (vcondu<mode><v_cmp_result>): Likewise.
20602 (vcond_mask_<mode><v_cmp_result>): Likewise. Added unsafe math
20605 2021-05-17 liuhongt <hongtao.liu@intel.com>
20608 * config/i386/i386.c (ix86_gimple_fold_builtin): Use
20609 gsi_insert_seq_before instead.
20611 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
20613 * doc/sourcebuild.texi (arm_qbit_ok): Rename into...
20614 (arm_sat_ok): ...this.
20616 2021-05-17 Martin Liska <mliska@suse.cz>
20618 * lto-wrapper.c (merge_flto_options): Factor out a new function.
20619 (merge_and_complain): Use it.
20620 (run_gcc): Merge also linker command line -flto=foo argument
20623 2021-05-16 Christophe Lyon <christophe.lyon@linaro.org>
20625 * config/arm/arm.h (CPP_SPEC): Remove error message about
20626 -mlittle-endian/-mbig-endian conflict.
20628 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
20630 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
20631 __ROP_PROTECT__ if -mrop-protect is selected.
20633 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
20635 * config/rs6000/rs6000-internal.h (rs6000_stack): Add
20636 rop_hash_save_offset and rop_hash_size.
20637 * config/rs6000/rs6000-logue.c (rs6000_stack_info): Compute
20638 rop_hash_size and rop_hash_save_offset.
20639 (debug_stack_info): Dump rop_hash_save_offset and rop_hash_size.
20640 (rs6000_emit_prologue): Emit hashst[p] in prologue.
20641 (rs6000_emit_epilogue): Emit hashchk[p] in epilogue.
20642 * config/rs6000/rs6000.md (unspec): Add UNSPEC_HASHST and
20644 (hashst): New define_insn.
20645 (hashchk): Likewise.
20647 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
20649 * config/rs6000/rs6000.c (rs6000_option_override_internal):
20650 Disable shrink wrap when inserting ROP-protect instructions.
20651 * config/rs6000/rs6000.opt (mrop-protect): New option.
20652 (mprivileged): Likewise.
20653 * doc/invoke.texi: Document mrop-protect and mprivileged.
20655 2021-05-15 Hans-Peter Nilsson <hp@axis.com>
20657 * reorg.c (fill_slots_from_thread): Reinstate code typoed out in
20660 2021-05-15 Martin Jambor <mjambor@suse.cz>
20663 2021-05-13 Martin Jambor <mjambor@suse.cz>
20665 PR tree-optimization/100453
20666 * tree-sra.c (sra_modify_assign): All const base accesses do not
20667 need refreshing, not just those from decl_pool.
20668 (sra_modify_assign): Do not refresh into a const base decl.
20670 2021-05-15 Jakub Jelinek <jakub@redhat.com>
20672 PR rtl-optimization/100342
20673 * regcprop.c (copy_value): When copying a source reg in a wider
20674 mode than it has recorded for the value, adjust recorded destination
20675 mode too or punt if !REG_CAN_CHANGE_MODE_P.
20677 2021-05-14 Jason Merrill <jason@redhat.com>
20679 * intl.h: Add comments.
20681 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20683 * config/aarch64/aarch64-simd.md
20684 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Split into...
20685 (aarch64_sqdmlsl2_lane<mode>_internal): ... This...
20686 (aarch64_sqdmlal2_lane<mode>_internal): ... And this.
20687 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Split into ...
20688 (aarch64_sqdmlsl2_laneq<mode>_internal): ... This...
20689 (aarch64_sqdmlal2_laneq<mode>_internal): ... And this.
20690 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>_internal): Split into...
20691 (aarch64_sqdmlsl2_n<mode>_internal): ... This...
20692 (aarch64_sqdmlal2_n<mode>_internal): ... And this.
20694 2021-05-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
20697 * config/arm/arm_neon.h (vtst_s8): Replace call to vtst builtin with it's
20698 boolean logic equivalent.
20699 (vtst_s16): Likewise.
20700 (vtst_s32): Likewise.
20701 (vtst_u8): Likewise.
20702 (vtst_u16): Likewise.
20703 (vtst_u32): Likewise.
20704 (vtst_p8): Likewise.
20705 (vtst_p16): Likewise.
20706 (vtstq_s8): Likewise.
20707 (vtstq_s16): Likewise.
20708 (vtstq_s32): Likewise.
20709 (vtstq_u8): Likewise.
20710 (vtstq_u16): Likewise.
20711 (vtstq_u32): Likewise.
20712 (vtstq_p8): Likewise.
20713 (vtstq_p16): Likewise.
20714 * config/arm/arm_neon_builtins.def: Remove entry for vtst.
20715 * config/arm/neon.md (neon_vtst<mode>): Remove pattern.
20717 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20719 * config/aarch64/aarch64-simd.md (aarch64_sqdmlal2<mode>): Merge into...
20720 (aarch64_sqdml<SBINQOPS:as>l2<mode>): ... This.
20721 (aarch64_sqdmlsl2<mode>): Delete.
20722 (aarch64_sqdmlal2_lane<mode>): Merge this...
20723 (aarch64_sqdmlsl2_lane<mode>): ... And this...
20724 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>): ... Into this.
20725 (aarch64_sqdmlal2_laneq<mode>): Merge this...
20726 (aarch64_sqdmlsl2_laneq<mode>): ... And this...
20727 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>): ... Into this.
20728 (aarch64_sqdmlal2_n<mode>): Merge this...
20729 (aarch64_sqdmlsl2_n<mode>): ... And this...
20730 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>): ... Into this.
20732 2021-05-13 Martin Sebor <msebor@redhat.com>
20734 PR middle-end/100574
20735 * builtins.c (access_ref::get_ref): Improve detection of PHIs with
20736 all null arguments.
20738 2021-05-13 Martin Sebor <msebor@redhat.com>
20740 PR tree-optimization/93100
20741 PR middle-end/98583
20742 * tree-ssa-uninit.c (check_defs): Exclude intrinsic functions that
20743 don't modify referenced objects.
20745 2021-05-13 Martin Jambor <mjambor@suse.cz>
20747 PR tree-optimization/100453
20748 * tree-sra.c (sra_modify_assign): All const base accesses do not
20749 need refreshing, not just those from decl_pool.
20750 (sra_modify_assign): Do not refresh into a const base decl.
20752 2021-05-13 Martin Liska <mliska@suse.cz>
20754 * tree-ssa-dom.c: Remove m_simplifier.
20756 2021-05-13 Richard Earnshaw <rearnsha@arm.com>
20759 * config/arm/arm.c (arm_canonicalize_comparison): Correctly
20760 canonicalize DImode inequality comparisons against the
20761 maximum integral value.
20763 2021-05-13 Jakub Jelinek <jakub@redhat.com>
20765 PR tree-optimization/98856
20766 * config/i386/i386.c (ix86_shift_rotate_cost): Add CODE argument.
20767 Expect V2DI and V4DI arithmetic right shifts to be emulated.
20768 (ix86_rtx_costs, ix86_add_stmt_cost): Adjust ix86_shift_rotate_cost
20770 * config/i386/i386-expand.c (expand_vec_perm_2perm_interleave,
20771 expand_vec_perm_2perm_pblendv): New functions.
20772 (ix86_expand_vec_perm_const_1): Use them.
20773 * config/i386/sse.md (ashr<mode>3<mask_name>): Rename to ...
20774 (<mask_codefor>ashr<mode>3<mask_name>): ... this.
20775 (ashr<mode>3): New define_expand with VI248_AVX512BW iterator.
20776 (ashrv4di3): New define_expand.
20777 (ashrv2di3): Change condition to TARGET_SSE2, handle !TARGET_XOP
20778 and !TARGET_AVX512VL expansion.
20780 2021-05-13 Uroš Bizjak <ubizjak@gmail.com>
20783 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Force mode
20784 sizes < 16 to a register when constructing vpcmov pattern.
20785 * config/i386/mmx.md (*xop_pcmov_<mode>): Use MMXMODE124 mode.
20787 2021-05-13 Martin Liska <mliska@suse.cz>
20789 * gcov-io.c (gcov_write_block): Remove.
20790 (gcov_write_words): Likewise.
20791 (gcov_read_words): Re-implement using gcov_read_bytes.
20792 (gcov_allocate): Remove.
20793 (GCOV_BLOCK_SIZE): Likewise.
20794 (struct gcov_var): Remove most of the fields.
20795 (gcov_position): Implement with ftell.
20796 (gcov_rewrite): Remove setting of start and offset fields.
20797 (from_file): Re-format.
20798 (gcov_open): Remove setbuf call. It should not be needed.
20799 (gcov_close): Remove internal buffer handling.
20800 (gcov_magic): Use __builtin_bswap32.
20801 (gcov_write_counter): Use directly gcov_write_unsigned.
20802 (gcov_write_string): Use direct fwrite and do not round
20804 (gcov_seek): Use directly fseek.
20805 (gcov_write_tag): Use gcov_write_unsigned directly.
20806 (gcov_write_length): Likewise.
20807 (gcov_write_tag_length): Likewise.
20808 (gcov_read_bytes): Use directly fread.
20809 (gcov_read_unsigned): Use gcov_read_words.
20810 (gcov_read_counter): Likewise.
20811 (gcov_read_string): Use gcov_read_bytes.
20812 * gcov-io.h (GCOV_WORD_SIZE): Adjust to reflect
20813 that size is not in bytes, but words (4B).
20814 (GCOV_TAG_FUNCTION_LENGTH): Likewise.
20815 (GCOV_TAG_ARCS_LENGTH): Likewise.
20816 (GCOV_TAG_ARCS_NUM): Likewise.
20817 (GCOV_TAG_COUNTER_LENGTH): Likewise.
20818 (GCOV_TAG_COUNTER_NUM): Likewise.
20819 (GCOV_TAG_SUMMARY_LENGTH): Likewise.
20821 2021-05-13 liuhongt <hongtao.liu@intel.com>
20824 * config/i386/sse.md (ssedoublevecmode): Add attribute for
20825 V64QI/V32HI/V16SI/V4DI.
20826 (ssehalfvecmode): Add attribute for V2DI/V2DF.
20827 (*vec_concatv4si_0): Extend to VI124_128.
20828 (*vec_concat<mode>_0): New pre-reload splitter.
20829 * config/i386/predicates.md (movq_parallel): New predicate.
20831 2021-05-13 Alexandre Oliva <oliva@adacore.com>
20833 * targhooks.c (default_zero_call_used_regs): Retry using
20834 successfully-zeroed registers as sources.
20836 2021-05-12 Tobias Burnus <tobias@codesourcery.com>
20838 * omp-low.c (finish_taskreg_scan): Use the proper detach decl.
20840 2021-05-12 Aldy Hernandez <aldyh@redhat.com>
20843 * gimple-range.cc (range_of_builtin_call): Skip out on
20844 processing __builtin_clz when varying.
20846 2021-05-12 Tom de Vries <tdevries@suse.de>
20849 * config/nvptx/nvptx-opts.h (enum ptx_version): New enum.
20850 * config/nvptx/nvptx.c (nvptx_file_start): Print .version according
20851 to ptx_version_option.
20852 * config/nvptx/nvptx.h (TARGET_PTX_6_3): Define.
20853 * config/nvptx/nvptx.md (define_insn "nvptx_shuffle<mode>")
20854 (define_insn "nvptx_vote_ballot"): Use sync variant for
20856 * config/nvptx/nvptx.opt (ptx_version): Add enum.
20857 (mptx): Add option.
20858 * doc/invoke.texi (Nvidia PTX Options): Add mptx item.
20860 2021-05-12 Richard Biener <rguenther@suse.de>
20862 PR tree-optimization/100566
20863 * tree-ssa-sccvn.c (dominated_by_p_w_unex): Properly handle
20864 allow_back for all edge queries.
20866 2021-05-12 liuhongt <hongtao.liu@intel.com>
20869 * config/i386/sse.md (<sse4_1_avx2>_pblendvb): Add
20870 splitters for pblendvb of NOT mask register.
20872 2021-05-12 Richard Biener <rguenther@suse.de>
20874 PR tree-optimization/100519
20875 * tree-ssa-reassoc.c (can_associate_p): Split into...
20876 (can_associate_op_p): ... this
20877 (can_associate_type_p): ... and this.
20878 (is_reassociable_op): Call can_associate_op_p.
20879 (break_up_subtract_bb): Call the appropriate predicates.
20880 (reassociate_bb): Likewise.
20882 2021-05-12 Martin Liska <mliska@suse.cz>
20884 * lto-wrapper.c (merge_and_complain): Merge -flto=arg options.
20885 (run_gcc): Use -flto argument detection for merged
20888 2021-05-12 Martin Liska <mliska@suse.cz>
20890 * lto-wrapper.c (print_lto_docs_link): New function.
20891 (run_gcc): Print warning about missing job server detection
20892 after we know NR of partitions. Do the same for -flto{,=1}.
20893 * opts.c (get_option_html_page): Support -flto option.
20895 2021-05-12 Martin Liska <mliska@suse.cz>
20897 * lto-wrapper.c (get_options_from_collect_gcc_options): Change
20899 (append_option): Remove.
20900 (find_option): Rework to use the vector type.
20901 (remove_option): Remove.
20902 (merge_and_complain): Use vectors for cl_decoded_option data
20904 (append_compiler_options): Likewise.
20905 (append_diag_options): Likewise.
20906 (append_linker_options): Likewise.
20907 (append_offload_options): Likewise.
20908 (compile_offload_image): Likewise.
20909 (compile_images_for_offload_targets): Likewise.
20910 (find_and_merge_options): Likewise.
20911 (run_gcc): Likewise.
20913 2021-05-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
20916 * dwarf2out.c (dwarf2out_finish): Set
20917 have_multiple_function_sections with multi-range text_section.
20919 2021-05-12 Martin Liska <mliska@suse.cz>
20921 PR bootstrap/100560
20922 * Makefile.in: Remove version.h from linker command line.
20924 2021-05-12 Richard Biener <rguenther@suse.de>
20926 PR middle-end/100547
20927 * rtl.h (rtvec_alloc): Make argument size_t.
20928 * rtl.c (rtvec_alloc): Verify the count is less than INT_MAX.
20930 2021-05-12 Jakub Jelinek <jakub@redhat.com>
20932 PR middle-end/100508
20933 * cfgexpand.c (expand_debug_expr): For DEBUG_EXPR_DECL with vector
20934 type, don't reuse DECL_RTL if it has different mode, instead force
20935 creation of a new DEBUG_EXPR.
20937 2021-05-12 Jakub Jelinek <jakub@redhat.com>
20938 Marc Glisse <marc.glisse@inria.fr>
20940 PR tree-optimization/94589
20941 * match.pd ((X & Y) == X -> (X & ~Y) == 0,
20942 (X | Y) == Y -> (X & ~Y) == 0): New GIMPLE simplifications.
20944 2021-05-12 Uroš Bizjak <ubizjak@gmail.com>
20947 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Handle V2SF mode.
20948 * config/i386/mmx.md (MMXMODE124): New mode iterator.
20950 (mmxintvecmode): New mode attribute.
20951 (mmxintvecmodelower): Ditto.
20952 (*mmx_maskcmpv2sf3_comm): New insn pattern.
20953 (*mmx_maskcmpv2sf3): Ditto.
20954 (vec_cmpv2sfv2si): New expander.
20955 (vcond<V2FI:mode>v2si): Ditto.
20956 (mmx_vlendvps): New insn pattern.
20957 (vcond<MMXMODE124:mode><MMXMODEI:mode>): Also handle V2SFmode.
20958 (vcondu<MMXMODE124:mode><MMXMODEI:mode>): Ditto.
20959 (vcond_mask_<mode><mmxintvecmodelower>): Ditto.
20961 2021-05-11 Martin Sebor <msebor@redhat.com>
20963 PR middle-end/21433
20964 * expr.c (expand_expr_real_1): Replace unreachable code with an assert.
20966 2021-05-11 Richard Biener <rguenther@suse.de>
20968 * gimple-fold.c (gimple_fold_call): Do not call
20969 maybe_fold_reference on call arguments or the static chain.
20970 (fold_stmt_1): Do not call maybe_fold_reference on GIMPLE_ASM
20973 2021-05-11 Martin Liska <mliska@suse.cz>
20975 * builtins.def (DEF_HSAIL_BUILTIN): Remove.
20976 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
20977 (DEF_HSAIL_SAT_BUILTIN): Likewise.
20978 (DEF_HSAIL_INTR_BUILTIN): Likewise.
20979 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
20980 * doc/frontends.texi: Remove BRIG.
20981 * doc/install.texi: Likewise.
20982 * doc/invoke.texi: Likewise.
20983 * doc/standards.texi: Likewise.
20984 * brig-builtins.def: Removed.
20985 * brig/ChangeLog: Removed.
20986 * brig/Make-lang.in: Removed.
20987 * brig/brig-builtins.h: Removed.
20988 * brig/brig-c.h: Removed.
20989 * brig/brig-lang.c: Removed.
20990 * brig/brigfrontend/brig-arg-block-handler.cc: Removed.
20991 * brig/brigfrontend/brig-atomic-inst-handler.cc: Removed.
20992 * brig/brigfrontend/brig-basic-inst-handler.cc: Removed.
20993 * brig/brigfrontend/brig-branch-inst-handler.cc: Removed.
20994 * brig/brigfrontend/brig-cmp-inst-handler.cc: Removed.
20995 * brig/brigfrontend/brig-code-entry-handler.cc: Removed.
20996 * brig/brigfrontend/brig-code-entry-handler.h: Removed.
20997 * brig/brigfrontend/brig-comment-handler.cc: Removed.
20998 * brig/brigfrontend/brig-control-handler.cc: Removed.
20999 * brig/brigfrontend/brig-copy-move-inst-handler.cc: Removed.
21000 * brig/brigfrontend/brig-cvt-inst-handler.cc: Removed.
21001 * brig/brigfrontend/brig-fbarrier-handler.cc: Removed.
21002 * brig/brigfrontend/brig-function-handler.cc: Removed.
21003 * brig/brigfrontend/brig-function.cc: Removed.
21004 * brig/brigfrontend/brig-function.h: Removed.
21005 * brig/brigfrontend/brig-inst-mod-handler.cc: Removed.
21006 * brig/brigfrontend/brig-label-handler.cc: Removed.
21007 * brig/brigfrontend/brig-lane-inst-handler.cc: Removed.
21008 * brig/brigfrontend/brig-machine.c: Removed.
21009 * brig/brigfrontend/brig-machine.h: Removed.
21010 * brig/brigfrontend/brig-mem-inst-handler.cc: Removed.
21011 * brig/brigfrontend/brig-module-handler.cc: Removed.
21012 * brig/brigfrontend/brig-queue-inst-handler.cc: Removed.
21013 * brig/brigfrontend/brig-seg-inst-handler.cc: Removed.
21014 * brig/brigfrontend/brig-signal-inst-handler.cc: Removed.
21015 * brig/brigfrontend/brig-to-generic.cc: Removed.
21016 * brig/brigfrontend/brig-to-generic.h: Removed.
21017 * brig/brigfrontend/brig-util.cc: Removed.
21018 * brig/brigfrontend/brig-util.h: Removed.
21019 * brig/brigfrontend/brig-variable-handler.cc: Removed.
21020 * brig/brigfrontend/hsa-brig-format.h: Removed.
21021 * brig/brigfrontend/phsa.h: Removed.
21022 * brig/brigspec.c: Removed.
21023 * brig/config-lang.in: Removed.
21024 * brig/gccbrig.texi: Removed.
21025 * brig/lang-specs.h: Removed.
21026 * brig/lang.opt: Removed.
21028 2021-05-11 Richard Biener <rguenther@suse.de>
21031 * ipa-param-manipulation.c
21032 (ipa_param_body_adjustments::modify_call_stmt): Avoid
21033 altering SSA_NAME_DEF_STMT by adjusting the calls LHS
21034 via gimple_call_lhs_ptr.
21036 2021-05-11 Alex Coplan <alex.coplan@arm.com>
21039 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear):
21040 Avoid emitting CFA adjusts on the sp if we have the fp.
21042 2021-05-11 Richard Sandiford <richard.sandiford@arm.com>
21044 * config/aarch64/iterators.md (VMUL_CHANGE_NLANES): Delete.
21045 (VMULD): New iterator.
21046 (VCOND): Handle V4HF and V8HF.
21047 (VCONQ): Fix entry for V2SF.
21048 * config/aarch64/aarch64-simd.md (mul_lane<mode>3): Use VMULD
21049 instead of VMUL. Use a 64-bit vector mode for the indexed operand.
21050 (*aarch64_mul3_elt_<vswap_width_name><mode>): Merge with...
21051 (mul_laneq<mode>3): ...this define_insn. Use VMUL instead of VDQSF.
21052 Use a 128-bit vector mode for the indexed operand. Use stype for
21053 the scheduling type.
21055 2021-05-11 Richard Biener <rguenther@suse.de>
21057 * gimple-fold.c (maybe_fold_reference): Only return
21058 is_gimple_min_invariant values.
21060 2021-05-11 Richard Biener <rguenther@suse.de>
21062 PR middle-end/100509
21063 * gimple-fold.c (fold_gimple_assign): Only call
21064 get_symbol_constant_value on register type symbols.
21066 2021-05-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
21067 Joe Ramsay <joe.ramsay@arm.com>
21070 * config/arm/arm_mve.h (__arm_vstrwq_scatter_offset): Fix wrong arguments.
21071 (__arm_vcmpneq): Remove duplicate definition.
21072 (__arm_vstrwq_scatter_offset_p): Likewise.
21073 (__arm_vmaxq_x): Likewise.
21074 (__arm_vmlsdavaq): Likewise.
21075 (__arm_vmlsdavaxq): Likewise.
21076 (__arm_vmlsdavq_p): Likewise.
21077 (__arm_vmlsdavxq_p): Likewise.
21078 (__arm_vrmlaldavhaq): Likewise.
21079 (__arm_vstrbq_p): Likewise.
21080 (__arm_vstrbq_scatter_offset): Likewise.
21081 (__arm_vstrbq_scatter_offset_p): Likewise.
21082 (__arm_vstrdq_scatter_offset): Likewise.
21083 (__arm_vstrdq_scatter_offset_p): Likewise.
21084 (__arm_vstrdq_scatter_shifted_offset): Likewise.
21085 (__arm_vstrdq_scatter_shifted_offset_p): Likewise.
21087 2021-05-11 Jakub Jelinek <jakub@redhat.com>
21089 PR middle-end/100471
21090 * omp-low.c (lower_omp_task_reductions): For OMP_TASKLOOP, if data
21091 is 0, bypass the reduction loop including
21092 GOMP_taskgroup_reduction_unregister call.
21094 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
21096 * config/rs6000/rs6000.c (struct rs6000_cost_data): New member
21097 costing_for_scalar.
21098 (rs6000_density_test): Early return if costing_for_scalar is true.
21099 (rs6000_init_cost): Init costing_for_scalar of rs6000_cost_data.
21101 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
21103 * doc/tm.texi: Regenerated.
21104 * target.def (init_cost): Add new parameter costing_for_scalar.
21105 * targhooks.c (default_init_cost): Adjust for new parameter.
21106 * targhooks.h (default_init_cost): Likewise.
21107 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
21108 (vect_compute_single_scalar_iteration_cost): Likewise.
21109 (vect_analyze_loop_2): Likewise.
21110 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
21111 (vect_bb_vectorization_profitable_p): Likewise.
21112 * tree-vectorizer.h (init_cost): Likewise.
21113 * config/aarch64/aarch64.c (aarch64_init_cost): Likewise.
21114 * config/i386/i386.c (ix86_init_cost): Likewise.
21115 * config/rs6000/rs6000.c (rs6000_init_cost): Likewise.
21117 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
21119 * config/rs6000/rs6000.c (rs6000_vect_nonmem): Renamed to
21120 vect_nonmem and moved into...
21121 (struct rs6000_cost_data): ...here.
21122 (rs6000_init_cost): Use vect_nonmem of cost_data instead.
21123 (rs6000_add_stmt_cost): Likewise.
21124 (rs6000_finish_cost): Likewise.
21126 2021-05-10 Eric Botcazou <ebotcazou@adacore.com>
21128 * range-op.cc (get_bool_state): Adjust head comment.
21129 (operator_not_equal::op1_range): Fix comment.
21130 (operator_bitwise_xor::op1_range): Remove call to gcc_unreachable.
21132 2021-05-10 Martin Sebor <msebor@redhat.com>
21134 PR middle-end/100425
21135 PR middle-end/100510
21136 * gimple-ssa-warn-alloca.c (pass_walloca::firast_time_p): Rename...
21137 (pass_walloca::xlimit_certain_p): ...to this.
21138 (pass_walloca::gate): Execute for any kind of handled warning.
21139 (pass_walloca::execute): Avoid issuing "maybe" and "unbounded"
21140 warnings when xlimit_certain_p is set.
21142 2021-05-10 Pat Haugen <pthaugen@linux.ibm.com>
21144 * config/rs6000/rs6000.c (rs6000_ira_change_pseudo_allocno_class):
21145 Return ALTIVEC_REGS if that is best_class.
21146 (rs6000_compute_pressure_classes): Add ALTIVEC_REGS.
21148 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
21150 * config/arm/arm.h (CPP_SPEC): Remove error message about
21153 2021-05-10 Martin Jambor <mjambor@suse.cz>
21155 * ipa-prop.h (IPA_NODE_REF): Removed.
21156 (IPA_NODE_REF_GET_CREATE): Likewise.
21157 (IPA_EDGE_REF): Likewise.
21158 (IPA_EDGE_REF_GET_CREATE): Likewise.
21159 (IS_VALID_JUMP_FUNC_INDEX): Likewise.
21160 * ipa-cp.c (print_all_lattices): Replaced IPA_NODE_REF with a direct
21161 use of ipa_node_params_sum.
21162 (ipcp_versionable_function_p): Likewise.
21163 (push_node_to_stack): Likewise.
21164 (pop_node_from_stack): Likewise.
21165 (set_single_call_flag): Replaced two IPA_NODE_REF with one single
21166 direct use of ipa_node_params_sum.
21167 (initialize_node_lattices): Replaced IPA_NODE_REF with a direct use of
21168 ipa_node_params_sum.
21169 (ipa_context_from_jfunc): Replaced IPA_EDGE_REF with a direct use of
21171 (ipcp_verify_propagated_values): Replaced IPA_NODE_REF with a direct
21172 use of ipa_node_params_sum.
21173 (self_recursively_generated_p): Likewise.
21174 (propagate_scalar_across_jump_function): Likewise.
21175 (propagate_context_across_jump_function): Replaced IPA_EDGE_REF with a
21176 direct use of ipa_edge_args_sum, moved the lookup after the early
21177 exit. Replaced IPA_NODE_REF with a direct use of ipa_node_params_sum.
21178 (propagate_bits_across_jump_function): Replaced IPA_NODE_REF with
21179 direct uses of ipa_node_params_sum.
21180 (propagate_vr_across_jump_function): Likewise.
21181 (propagate_aggregate_lattice): Likewise.
21182 (propagate_aggs_across_jump_function): Likewise.
21183 (propagate_constants_across_call): Likewise, also replaced
21184 IPA_EDGE_REF with a direct use of ipa_edge_args_sum.
21185 (good_cloning_opportunity_p): Replaced IPA_NODE_REF with a direct use
21186 of ipa_node_params_sum.
21187 (estimate_local_effects): Likewise.
21188 (add_all_node_vals_to_toposort): Likewise.
21189 (propagate_constants_topo): Likewise.
21190 (ipcp_propagate_stage): Likewise.
21191 (ipcp_discover_new_direct_edges): Likewise.
21192 (calls_same_node_or_its_all_contexts_clone_p): Likewise.
21193 (cgraph_edge_brings_value_p): Likewise (in both overloaded functions).
21194 (get_info_about_necessary_edges): Likewise.
21195 (want_remove_some_param_p): Likewise.
21196 (create_specialized_node): Likewise.
21197 (self_recursive_pass_through_p): Likewise.
21198 (self_recursive_agg_pass_through_p): Likewise.
21199 (find_more_scalar_values_for_callers_subset): Likewise and also
21200 replaced IPA_EDGE_REF with direct uses of ipa_edge_args_sum, in one
21201 case replacing two of those with a single query.
21202 (find_more_contexts_for_caller_subset): Likewise for the
21203 ipa_polymorphic_call_context overload.
21204 (intersect_aggregates_with_edge): Replaced IPA_EDGE_REF with a direct
21205 use of ipa_edge_args_sum. Replaced IPA_NODE_REF with direct uses of
21206 ipa_node_params_sum.
21207 (find_aggregate_values_for_callers_subset): Likewise, also reusing
21208 results of ipa_edge_args_sum->get.
21209 (cgraph_edge_brings_all_scalars_for_node): Replaced IPA_NODE_REF with
21210 direct uses of ipa_node_params_sum, replaced IPA_EDGE_REF with a
21211 direct use of ipa_edge_args_sum.
21212 (cgraph_edge_brings_all_agg_vals_for_node): Likewise, moved node
21213 summary query after the early exit and reused the result later.
21214 (decide_about_value): Replaced IPA_NODE_REF with a direct use of
21215 ipa_node_params_sum.
21216 (decide_whether_version_node): Likewise. Removed re-querying for
21217 summaries after cloning.
21218 (spread_undeadness): Replaced IPA_NODE_REF with a direct use of
21219 ipa_node_params_sum.
21220 (has_undead_caller_from_outside_scc_p): Likewise, reusing results of
21222 (identify_dead_nodes): Likewise.
21223 (ipcp_store_bits_results): Replaced IPA_NODE_REF with direct uses of
21224 ipa_node_params_sum.
21225 (ipcp_store_vr_results): Likewise.
21226 * ipa-fnsummary.c (evaluate_properties_for_edge): Likewise.
21227 (ipa_fn_summary_t::duplicate): Likewise.
21228 (analyze_function_body): Likewise.
21229 (estimate_calls_size_and_time): Likewise.
21230 (ipa_cached_call_context::duplicate_from): Likewise.
21231 (ipa_call_context::equal_to): Likewise.
21232 (remap_edge_params): Likewise.
21233 (ipa_merge_fn_summary_after_inlining): Likewise.
21234 (inline_read_section): Likewise.
21235 * ipa-icf.c (sem_function::param_used_p): Likewise.
21236 * ipa-modref.c (compute_parm_map): Likewise.
21237 (compute_parm_map): Replaced IPA_EDGE_REF with a direct use of
21239 (get_access_for_fnspec): Replaced IPA_NODE_REF with a direct use of
21240 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
21242 * ipa-profile.c (check_argument_count): Likewise.
21243 * ipa-prop.c (ipa_alloc_node_params): Replaced IPA_NODE_REF_GET_CREATE
21244 with a direct use of ipa_node_params_sum.
21245 (ipa_initialize_node_params): Likewise.
21246 (ipa_print_node_jump_functions_for_edge): Replaced IPA_EDGE_REF with a
21247 direct use of ipa_edge_args_sum and reused the query result.
21248 (ipa_compute_jump_functions_for_edge): Replaced IPA_NODE_REF with a
21249 direct use of ipa_node_params_sum and replaced IPA_EDGE_REF with a
21250 direct use of ipa_edge_args_sum.
21251 (ipa_note_param_call): Replaced IPA_NODE_REF with a direct use of
21252 ipa_node_params_sum and reused the result of the query.
21253 (ipa_analyze_node): Likewise.
21254 (ipa_analyze_controlled_uses): Replaced IPA_NODE_REF with a direct use
21255 of ipa_node_params_sum.
21256 (update_jump_functions_after_inlining): Replaced IPA_EDGE_REF with
21257 direct uses of ipa_edge_args_sum.
21258 (update_indirect_edges_after_inlining): Replaced IPA_NODE_REF with
21259 direct uses of ipa_node_params_sum and replaced IPA_EDGE_REF with a
21260 direct use of ipa_edge_args_sum. Removed superficial re-querying the
21262 (propagate_controlled_uses): Replaced IPA_NODE_REF with direct uses of
21263 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
21265 (ipa_propagate_indirect_call_infos): Replaced IPA_EDGE_REF with a
21266 direct use of ipa_edge_args_sum.
21267 (ipa_edge_args_sum_t::duplicate): Replaced IPA_NODE_REF with a direct
21268 use of ipa_node_params_sum.
21269 (ipa_print_node_params): Likewise.
21270 (ipa_write_node_info): Likewise and also replaced IPA_EDGE_REF with
21271 direct uses of ipa_edge_args_sum.
21272 (ipa_read_edge_info): Replaced IPA_EDGE_REF with a direct use of
21274 (ipa_read_node_info): Replaced IPA_NODE_REF with a direct use of
21275 ipa_node_params_sum.
21276 (ipa_prop_write_jump_functions): Likewise. Move variable node to the
21277 scopes where it is used.
21279 2021-05-10 Uroš Bizjak <ubizjak@gmail.com>
21281 * config/i386/i386-expand.c (ix86_expand_sse_movcc)
21282 <case E_V2SImode>: Force op_true to register.
21284 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
21286 * config/arm/iterators.md (MVE_FP_COMPARISONS): New.
21287 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_f<mode>)
21288 (mve_vcmp<mve_cmp_op>q_n_f<mode>): New, merge all vcmp_*f*
21290 (mve_vcmpeqq_f<mode>, mve_vcmpeqq_n_f<mode>, mve_vcmpgeq_f<mode>)
21291 (mve_vcmpgeq_n_f<mode>, mve_vcmpgtq_f<mode>)
21292 (mve_vcmpgtq_n_f<mode>, mve_vcmpleq_f<mode>)
21293 (mve_vcmpleq_n_f<mode>, mve_vcmpltq_f<mode>)
21294 (mve_vcmpltq_n_f<mode>, mve_vcmpneq_f<mode>)
21295 (mve_vcmpneq_n_f<mode>): Remove.
21296 * config/arm/unspecs.md (VCMPEQQ_F, VCMPEQQ_N_F, VCMPGEQ_F)
21297 (VCMPGEQ_N_F, VCMPGTQ_F, VCMPGTQ_N_F, VCMPLEQ_F, VCMPLEQ_N_F)
21298 (VCMPLTQ_F, VCMPLTQ_N_F, VCMPNEQ_F, VCMPNEQ_N_F): Remove.
21300 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
21302 * config/arm/iterators.md (MVE_COMPARISONS): New.
21304 (mve_cmp_type): New.
21305 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_<mode>): New, merge all
21307 (mve_vcmpneq_<mode>, mve_vcmpcsq_n_<mode>, mve_vcmpcsq_<mode>)
21308 (mve_vcmpeqq_n_<mode>, mve_vcmpeqq_<mode>, mve_vcmpgeq_n_<mode>)
21309 (mve_vcmpgeq_<mode>, mve_vcmpgtq_n_<mode>, mve_vcmpgtq_<mode>)
21310 (mve_vcmphiq_n_<mode>, mve_vcmphiq_<mode>, mve_vcmpleq_n_<mode>)
21311 (mve_vcmpleq_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
21312 (mve_vcmpneq_n_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
21313 (mve_vcmpneq_n_<mode>): Remove.
21315 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
21317 * config/arm/arm_mve.h (__arm_vcmp*): Remove 's' suffix.
21318 * config/arm/arm_mve_builtins.def (vcmp*): Remove 's' suffix.
21319 * config/arm/mve.md (mve_vcmp*): Remove 's' suffix in pattern
21322 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
21324 * config/arm/arm_mve_builtins.def (vcmpneq_u): Remove.
21325 (vcmpneq_n_u): Likewise.
21326 (vcmpeqq_u,): Likewise.
21327 (vcmpeqq_n_u): Likewise.
21328 * config/arm/iterators.md (supf): Remove VCMPNEQ_U, VCMPEQQ_U,
21329 VCMPEQQ_N_U and VCMPNEQ_N_U.
21330 * config/arm/mve.md (mve_vcmpneq): Remove <supf> iteration.
21331 (mve_vcmpeqq_n): Likewise.
21332 (mve_vcmpeqq): Likewise.
21333 (mve_vcmpneq_n): Likewise.
21335 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
21337 * config/arm/arm_mve.h (__arm_vcmpeq*u*, __arm_vcmpne*u*): Call
21338 the 's' version of the builtin.
21340 2021-05-10 Richard Biener <rguenther@suse.de>
21342 PR tree-optimization/100492
21343 * tree-loop-distribution.c (find_seed_stmts_for_distribution):
21344 Find nothing when the loop contains an irreducible region.
21346 2021-05-10 Richard Biener <rguenther@suse.de>
21348 PR middle-end/100464
21350 * gimple-fold.c (canonicalize_constructor_val): Do not set
21353 2021-05-10 Richard Biener <rguenther@suse.de>
21355 PR tree-optimization/100434
21356 * tree-ssa-dse.c (initialize_ao_ref_for_dse): Handle
21358 (dse_optimize_stmt): Handle call LHS by dropping the
21359 LHS or the whole call if it doesn't have other
21361 (pass_dse::execute): Adjust.
21363 2021-05-10 Martin Liska <mliska@suse.cz>
21365 * Makefile.in: Add missing genversion rule.
21367 2021-05-10 Alex Coplan <alex.coplan@arm.com>
21370 * config/arm/mve.md (*mve_mov<mode>): Simplify output code. Use
21371 vldrw.u32 and vstrw.32 for V2D[IF]mode loads and stores.
21373 2021-05-10 Martin Liska <mliska@suse.cz>
21375 * builtins.c (is_builtin_name): Use startswith
21376 function instead of strncmp.
21377 * collect2.c (main): Likewise.
21378 (has_lto_section): Likewise.
21379 (scan_libraries): Likewise.
21380 * coverage.c (coverage_checksum_string): Likewise.
21381 (coverage_init): Likewise.
21382 * dwarf2out.c (is_cxx): Likewise.
21383 (gen_compile_unit_die): Likewise.
21384 * gcc-ar.c (main): Likewise.
21385 * gcc.c (init_spec): Likewise.
21386 (read_specs): Likewise.
21387 (execute): Likewise.
21388 (check_live_switch): Likewise.
21389 * genattrtab.c (write_attr_case): Likewise.
21390 (IS_ATTR_GROUP): Likewise.
21391 * gencfn-macros.c (main): Likewise.
21392 * gengtype.c (type_for_name): Likewise.
21393 (gen_rtx_next): Likewise.
21394 (get_file_langdir): Likewise.
21395 (write_local): Likewise.
21396 * genmatch.c (get_operator): Likewise.
21397 (get_operand_type): Likewise.
21398 (expr::gen_transform): Likewise.
21399 * genoutput.c (validate_optab_operands): Likewise.
21400 * incpath.c (add_sysroot_to_chain): Likewise.
21401 * langhooks.c (lang_GNU_C): Likewise.
21402 (lang_GNU_CXX): Likewise.
21403 (lang_GNU_Fortran): Likewise.
21404 (lang_GNU_OBJC): Likewise.
21405 * lto-wrapper.c (run_gcc): Likewise.
21406 * omp-general.c (omp_max_simt_vf): Likewise.
21407 * omp-low.c (omp_runtime_api_call): Likewise.
21408 * opts-common.c (parse_options_from_collect_gcc_options): Likewise.
21409 * read-rtl-function.c (function_reader::read_rtx_operand_r): Likewise.
21410 * real.c (real_from_string): Likewise.
21411 * selftest.c (assert_str_startswith): Likewise.
21412 * timevar.c (timer::validate_phases): Likewise.
21413 * tree.c (get_file_function_name): Likewise.
21414 * ubsan.c (ubsan_use_new_style_p): Likewise.
21415 * varasm.c (default_function_rodata_section): Likewise.
21416 (incorporeal_function_p): Likewise.
21417 (default_section_type_flags): Likewise.
21418 * system.h (startswith): Define startswith.
21420 2021-05-10 Martin Liska <mliska@suse.cz>
21422 * bitmap.h (class auto_bitmap): Remove
21423 __cplusplus >= 201103.
21424 * config/aarch64/aarch64.c: Likewise.
21425 * gimple-ssa-store-merging.c (store_immediate_info::store_immediate_info):
21427 * sbitmap.h: Likewise.
21429 2021-05-10 Martin Liska <mliska@suse.cz>
21431 * Makefile.in: Rename gcov-iov to genversion and depend
21432 on version.h (instead of gcov-iov.h).
21433 * gcov-io.h: Include version.h instread of gcov-iov.h.
21434 * gengtype-state.c (read_state_version): Likewise.
21435 * gcov-iov.c: Moved to...
21436 * genversion.c: ...here.
21437 * lto-streamer.h (LTO_major_version): Define it with
21439 * version.c: Removed.
21440 * version.h: Removed.
21442 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
21444 * config/arc/arc.md (UNSPEC_ARC_DMPYWH): Define.
21445 * config/arc/simdext.md (VCT): Add predicates for iterator
21448 (voptab): Likewise.
21449 (vec_widen_<V_US>mult_hi_v4hi): Change pattern predicate.
21450 (<voptab>v2si3): New patterns.
21452 (reduc_plus_scal_v4hi): Likewise.
21453 (reduc_plus_scal_v2si): Likewise.
21454 (vec_duplicatev2si): Likewise.
21455 (vec_duplicatev4hi): Likewise.
21457 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
21459 * config/arc/simdext.md: Format and cleanup file.
21461 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
21463 * config/arc/simdext.md (movmisalignv2hi): Allow misaligned access
21464 only when munaligned-access option is on.
21465 (movmisalign<mode>): Likewise.
21467 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
21469 * common/config/arc/arc-common.c (arc_handle_option): Remove dot
21471 * config/arc/arc.c (arc_reorg): Remove underscore from string.
21473 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
21475 * config/arc/arc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
21476 (CTZ_DEFINED_VALUE_AT_ZERO): Likewise.
21477 * config/arc/arc.md (clrsbsi2): Cleanup pattern.
21478 (norm_f): Likewise.
21481 (clzsi2): Use fls instruction when available.
21482 (arc_clzsi2): Likewise.
21484 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
21486 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add r26 and r27.
21488 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
21490 * doc/extend.texi (__builtin_arc_sr): Swap arguments.
21492 2021-05-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
21494 PR middle-end/100467
21495 * toplev.c (compile_file): Call insn_locations_init before
21496 targetm.asm_out.code_end.
21498 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
21501 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
21503 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
21505 2021-05-07 Jakub Jelinek <jakub@redhat.com>
21506 Andrew Stubbs <amd@codesourcery.com>
21509 * builtins.c (try_store_by_multiple_pieces): Use force_operand for
21510 emit_move_insn operands.
21512 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
21514 * cfgexpand.c (expand_gimple_basic_block): Do not inherit a current
21515 location for the outgoing edges of an empty block.
21516 * dwarf2out.c (add_subscript_info): Retrieve the bounds and index
21517 type by means of the get_array_descr_info langhook, if it is set and
21518 returns true. Remove obsolete code dealing with unnamed subtypes.
21520 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
21522 * gimple-range-cache.cc (ssa_block_ranges): Virtualize.
21523 (sbr_vector): Renamed from ssa_block_cache.
21524 (sbr_vector::sbr_vector): Allocate from obstack abd initialize.
21525 (ssa_block_ranges::~ssa_block_ranges): Remove.
21526 (sbr_vector::set_bb_range): Use varying and undefined cached values.
21527 (ssa_block_ranges::set_bb_varying): Remove.
21528 (sbr_vector::get_bb_range): Adjust assert.
21529 (sbr_vector::bb_range_p): Adjust assert.
21530 (~block_range_cache): No freeing loop required.
21531 (block_range_cache::get_block_ranges): Remove.
21532 (block_range_cache::set_bb_range): Inline get_block_ranges.
21533 (block_range_cache::set_bb_varying): Remove.
21534 * gimple-range-cache.h (set_bb_varying): Remove prototype.
21535 * value-range.h (irange_allocator::get_memory): New.
21537 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
21539 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Search
21540 dominator tree is available and requested.
21541 (ranger_cache::ssa_range_in_bb): Don't search dom tree here.
21542 (ranger_cache::fill_block_cache): Don't search dom tree here either.
21543 * gimple-range-cache.h (non_null_deref_p): Add dom_search param.
21545 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
21547 * gimple-range.cc (gimple_ranger::range_on_exit): Handle block with
21548 only PHI nodes better.
21550 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
21552 * gimple-range-edge.h (gimple_outgoing_range): Rename from
21554 (gcond_edge_range): Export prototype.
21555 * gimple-range-edge.cc (gcond_edge_range): New.
21556 (gimple_outgoing_range::edge_range_p): Use gcond_edge_range.
21557 * gimple-range-gori.h (gori_compute): Use gimple_outgoing_range.
21559 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
21561 * gimple-range-edge.cc (outgoing_range::calc_switch_ranges): Compute
21562 default range into a temp and allocate only what is needed.
21564 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
21566 * range-op.cc (operator_trunc_mod::wi_fold): x % 0 is UNDEFINED.
21568 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
21570 * gimple-range.h (gimple_range_global): Pick up parameter initial
21571 values, and use-before defined locals are UNDEFINED.
21573 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
21575 * doc/extend.texi (scalar_storage_order): Mention effect on pointer
21577 * tree.h (reverse_storage_order_for_component_p): Return false if
21578 the type is a pointer.
21580 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
21582 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
21584 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
21587 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
21588 Handle V8QI, V4HI and V2SI modes.
21589 * config/i386/mmx.md (mmx_pblendvb): New insn pattern.
21590 * config/i386/sse.md (unspec): Move UNSPEC_BLENDV ...
21591 * config/i386/i386.md (unspec): ... here.
21593 2021-05-07 Tobias Burnus <tobias@codesourcery.com>
21594 Tom de Vries <tdevries@suse.de>
21596 * omp-low.c (lower_rec_simd_input_clauses): Set max_vf = 1 if
21597 a truth_value_p reduction variable is nonintegral.
21599 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
21602 * config/i386/i386-expand.c (ix86_use_mask_cmp_p):
21603 Return false for mode sizes < 16.
21605 2021-05-07 Jakub Jelinek <jakub@redhat.com>
21608 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn.
21610 2021-05-06 Martin Jambor <mjambor@suse.cz>
21612 * ipa-sra.c (ipa_sra_dump_all_summaries): Dump edge summaries even
21613 when there is no function summary.
21614 (ipa_sra_summarize_function): produce edge summaries even when
21617 2021-05-06 Tom Tromey <tom@tromey.com>
21619 * godump.c (string_hash_eq): Remove.
21620 (go_finish): Use htab_eq_string.
21622 2021-05-06 Tom Tromey <tom@tromey.com>
21624 * gengtype-state.c (read_state): Use htab_eq_string.
21625 (string_eq): Remove.
21627 2021-05-06 Tom Tromey <tom@tromey.com>
21629 * gensupport.c (htab_eq_string): Remove.
21631 2021-05-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
21634 * debug.h (gcc_debug_hooks): Add set_ignored_loc function pointer.
21635 * dwarf2out.h (dw_fde_node::ignored_debug): New data item.
21636 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Add dummy
21637 set_ignored_loc callbacks.
21638 * debug.c (do_nothing_debug_hooks): Likewise.
21639 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
21640 * dwarf2out.c (text_section_used, cold_text_section_used): Remove.
21641 (in_text_section_p, last_text_label, last_cold_label,
21642 switch_text_ranges, switch_cold_ranges): New data items.
21643 (dwarf2out_note_section_used): Remove.
21644 (dwarf2out_begin_prologue): Set fde->ignored_debug and
21646 (mark_ignored_debug_section): New helper function.
21647 (dwarf2out_end_epilogue, dwarf2out_switch_text_section): Call
21648 mark_ignored_debug_section.
21649 (dwarf2_debug_hooks): Use dwarf2out_set_ignored_loc.
21650 (dwarf2_lineno_debug_hooks): Use dummy for set_ignored_loc.
21651 (size_of_aranges): Adjust formula for multi-part text ranges size.
21652 (output_aranges): Output multi-part text ranges.
21653 (dwarf2out_set_ignored_loc): New callback function.
21654 (dwarf2out_finish): Output multi-part text ranges.
21655 (dwarf2out_c_finalize): Clear new data items.
21656 * final.c (final_start_function_1): Call set_ignored_loc callback.
21657 (final_scan_insn_1): Likewise.
21658 * ggc-page.c (gt_ggc_mx): New helper function.
21659 * stringpool.c (gt_pch_nx): Likewise.
21661 2021-05-06 Richard Biener <rguenther@suse.de>
21663 * timevar.def (TV_TREE_INSERT_PHI_NODES): Remove.
21664 (TV_TREE_SSA_REWRITE_BLOCKS): Likewise.
21665 (TV_TREE_INTO_SSA): New.
21666 * tree-into-ssa.c (insert_phi_nodes): Do not account separately.
21667 (rewrite_blocks): Likewise.
21668 (pass_data_build_ssa): Account to TV_TREE_INTO_SSA.
21670 2021-05-06 Jakub Jelinek <jakub@redhat.com>
21672 * tree-ssa-phiopt.c (value_replacement, minmax_replacement,
21673 abs_replacement, xor_replacement,
21674 cond_removal_in_popcount_clz_ctz_pattern,
21675 replace_phi_edge_with_variable): Change type of phi argument from
21676 gimple * to gphi *.
21678 2021-05-06 Richard Biener <rguenther@suse.de>
21680 * tree-ssa-loop-split.c (split_loop): Delay updating SSA form.
21681 Output an opt-info message.
21682 (do_split_loop_on_cond): Likewise.
21683 (tree_ssa_split_loops): Update SSA form here.
21685 2021-05-06 Richard Biener <rguenther@suse.de>
21687 * tree-inline.c (tree_function_versioning): Fix DECL_BY_REFERENCE
21688 return variable removal.
21690 2021-05-06 Marius Hillenbrand <mhillen@linux.ibm.com>
21692 * config/s390/s390-builtins.def (O_M5, O1_M5, ...): Remove unused macros.
21693 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
21694 (s390_vec_permi_dbl, s390_vpdi): Use the O3_U2 type for the immediate
21696 * config/s390/s390.c (s390_const_operand_ok): Remove unused
21699 2021-05-06 Jakub Jelinek <jakub@redhat.com>
21701 PR tree-optimization/94589
21702 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Call
21703 spaceship_replacement.
21704 (cond_only_block_p, spaceship_replacement): New functions.
21706 2021-05-06 Richard Biener <rguenther@suse.de>
21709 * tree-emutls.c (gen_emutls_addr): Pass in whether we're
21710 dealing with a debug use and only query existing addresses
21712 (lower_emutls_1): Avoid splitting out addresses for debug
21713 stmts, reset the debug stmt when we fail to find existing
21715 (lower_emutls_phi_arg): Set wi.stmt.
21717 2021-05-06 Christoph Muellner <cmuellner@gcc.gnu.org>
21720 * config/riscv/riscv.c (riscv_block_move_loop): Use cbranch helper.
21721 * config/riscv/riscv.md (cbranch<mode>4): Generate helpers.
21722 (stack_protect_test): Use cbranch helper.
21724 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
21727 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
21728 always return the establisher frame for __builtin_frame_address (0).
21730 2021-05-05 Ivan Sorokin <vanyacpp@gmail.com>
21733 * config/i386/i386-builtins.c (ix86_cpu_model_type_node): New.
21734 (ix86_cpu_model_var): Likewise.
21735 (ix86_cpu_features2_type_node): Likewise.
21736 (ix86_cpu_features2_var): Likewise.
21737 (fold_builtin_cpu): Cache __cpu_model and __cpu_features2 with
21740 2021-05-05 Martin Sebor <msebor@redhat.com>
21742 * passes.def (pass_warn_printf): Run after SSA.
21744 2021-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
21746 * config/arm/neon.md (neon_vtst_combine<mode>): New pattern.
21747 * config/arm/predicates.md (minus_one_operand): New predicate.
21749 2021-05-05 Jeff Law <jlaw@tachyum.com>
21751 * config/avr/avr.md: Remove references to CC_STATUS_INIT.
21753 2021-05-05 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
21755 PR rtl-optimization/100263
21756 * postreload.c (move2add_valid_value_p): Ensure register can
21759 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
21761 PR rtl-optimization/100411
21762 * cfgcleanup.c (try_crossjump_to_edge): Also skip end of prologue
21763 and beginning of function markers.
21765 2021-05-05 Jeff Law <jlaw@tachyum.com>
21767 * config/cr16/cr16.h (NOTICE_UPDATE_CC): Remove.
21768 * config/cr16/cr16.c (notice_update_cc): Remove.
21769 * config/cr16/cr16-protos.h (notice_update_cc): Remove.
21771 2021-05-05 Uroš Bizjak <ubizjak@gmail.com>
21774 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
21775 Handle V8QI, V4HI and V2SI modes.
21776 * config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
21777 (ix86_build_signbit_mask): Ditto.
21778 * config/i386/mmx.md (MMXMODE14): New mode iterator.
21779 (<smaxmin:code><MMXMODE14:mode>3): New expander.
21780 (*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
21781 (<umaxmin:code><MMXMODE24:mode>3): New expander.
21782 (*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
21783 (vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
21784 (vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
21785 (vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
21786 (vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
21787 (vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
21789 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
21791 * dwarf2out.c (loc_list_from_tree_1) <DECL>: During early DWARF, do
21792 not expand the VALUE_EXPR of variables put in the non-local frame.
21793 * gimplify.c (gimplify_type_sizes) <RECORD_TYPE>: If the type is not
21794 to be ignored for debug info, ensure its variable offsets are not.
21796 2021-05-05 Richard Biener <rguenther@suse.de>
21798 PR tree-optimization/79333
21799 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
21800 Fold stmt following SSA edges.
21802 2021-05-05 Richard Biener <rguenther@suse.de>
21804 PR middle-end/100394
21805 * calls.c (expand_call): Preserve possibly throwing calls.
21806 * cfgexpand.c (expand_call_stmt): When a call can throw signal
21807 RTL expansion there are side-effects.
21808 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Simplify,
21809 mark all possibly throwing stmts necessary unless we can elide
21811 * tree-ssa-dse.c (pass_dse::execute): Preserve exceptions unless
21812 -fdelete-dead-exceptions.
21813 * tree.h (DECL_PURE_P): Add note about exceptions.
21815 2021-05-05 Alexandre Oliva <oliva@adacore.com>
21817 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Make it
21820 2021-05-04 David Edelsohn <dje.gcc@gmail.com>
21822 * config/rs6000/rs6000-call.c (rs6000_output_mi_thunk): Use
21823 get_fnname_from_decl for name of thunk.
21824 * config/rs6000/rs6000.c (rs6000_declare_alias): Use assemble_name
21825 and ASM_OUTPUT_LABEL.
21826 (rs6000_xcoff_declare_function_name): Use assemble_name and
21828 (rs6000_xcoff_declare_object_name): Use ASM_OUTPUT_LABEL.
21829 (rs6000_xcoff_encode_section_info): Don't add mapping class
21830 for aliases. Always add [DS] mapping class to primary
21832 (rs6000_asm_weaken_decl): Don't explicitly add [DS].
21834 2021-05-04 Martin Sebor <msebor@redhat.com>
21836 PR middle-end/100307
21837 * builtins.c (compute_objsize_r): Clear base0 for pointers.
21839 2021-05-04 Jeff Law <jlaw@tachyum.com>
21841 * config/bfin/bfin.h (NOTICE_UPDATE_CC): Remove.
21843 2021-05-04 Segher Boessenkool <segher@kernel.crashing.org>
21845 * caller-save.c: Remove CC0.
21846 * cfgcleanup.c: Remove CC0.
21847 * cfgrtl.c: Remove CC0.
21848 * combine.c: Remove CC0.
21849 * compare-elim.c: Remove CC0.
21850 * conditions.h: Remove CC0.
21851 * config/h8300/h8300.h: Remove CC0.
21852 * config/h8300/h8300-protos.h: Remove CC0.
21853 * config/h8300/peepholes.md: Remove CC0.
21854 * config/i386/x86-tune-sched.c: Remove CC0.
21855 * config/m68k/m68k.c: Remove CC0.
21856 * config/rl78/rl78.c: Remove CC0.
21857 * config/sparc/sparc.c: Remove CC0.
21858 * config/xtensa/xtensa.c: Remove CC0.
21859 (gen_conditional_move): Use pc_rtx instead of cc0_rtx in a piece of
21860 RTL where that is used as a placeholder only.
21861 * cprop.c: Remove CC0.
21862 * cse.c: Remove CC0.
21863 * cselib.c: Remove CC0.
21864 * df-problems.c: Remove CC0.
21865 * df-scan.c: Remove CC0.
21866 * doc/md.texi: Remove CC0. Adjust an example.
21867 * doc/rtl.texi: Remove CC0. Adjust an example.
21868 * doc/tm.texi: Regenerate.
21869 * doc/tm.texi.in: Remove CC0.
21870 * emit-rtl.c: Remove CC0.
21871 * final.c: Remove CC0.
21872 * fwprop.c: Remove CC0.
21873 * gcse-common.c: Remove CC0.
21874 * gcse.c: Remove CC0.
21875 * genattrtab.c: Remove CC0.
21876 * genconfig.c: Remove CC0.
21877 * genemit.c: Remove CC0.
21878 * genextract.c: Remove CC0.
21879 * gengenrtl.c: Remove CC0.
21880 * genrecog.c: Remove CC0.
21881 * haifa-sched.c: Remove CC0.
21882 * ifcvt.c: Remove CC0.
21883 * ira-costs.c: Remove CC0.
21884 * ira.c: Remove CC0.
21885 * jump.c: Remove CC0.
21886 * loop-invariant.c: Remove CC0.
21887 * lra-constraints.c: Remove CC0.
21888 * lra-eliminations.c: Remove CC0.
21889 * optabs.c: Remove CC0.
21890 * postreload-gcse.c: Remove CC0.
21891 * postreload.c: Remove CC0.
21892 * print-rtl.c: Remove CC0.
21893 * read-rtl-function.c: Remove CC0.
21894 * reg-notes.def: Remove CC0.
21895 * reg-stack.c: Remove CC0.
21896 * reginfo.c: Remove CC0.
21897 * regrename.c: Remove CC0.
21898 * reload.c: Remove CC0.
21899 * reload1.c: Remove CC0.
21900 * reorg.c: Remove CC0.
21901 * resource.c: Remove CC0.
21902 * rtl.c: Remove CC0.
21903 * rtl.def: Remove CC0.
21904 * rtl.h: Remove CC0.
21905 * rtlanal.c: Remove CC0.
21906 * sched-deps.c: Remove CC0.
21907 * sched-rgn.c: Remove CC0.
21908 * shrink-wrap.c: Remove CC0.
21909 * simplify-rtx.c: Remove CC0.
21910 * system.h: Remove CC0. Poison NOTICE_UPDATE_CC, CC_STATUS_MDEP_INIT,
21911 CC_STATUS_MDEP, and CC_STATUS.
21912 * target.def: Remove CC0.
21913 * valtrack.c: Remove CC0.
21914 * var-tracking.c: Remove CC0.
21916 2021-05-04 Richard Biener <rguenther@suse.de>
21918 PR tree-optimization/100414
21919 * tree-ssa-phiopt.c (get_non_trapping): Do not compute dominance
21921 (tree_ssa_phiopt_worker): But unconditionally here.
21923 2021-05-04 Tobias Burnus <tobias@codesourcery.com>
21925 * omp-low.c (lower_rec_input_clauses, lower_reduction_clauses): Handle
21926 && and || with floating-point and complex arguments.
21928 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
21930 * tree-inline.c (insert_debug_decl_map): Delete.
21931 (copy_debug_stmt): Minor tweak.
21932 (setup_one_parameter): Do not use a variable if the value is either
21933 a read-only DECL or a non-addressable local variable in the caller.
21934 In this case, insert the debug-only variable in the map manually.
21935 (expand_call_inline): Do not generate a CLOBBER for these values.
21936 * tree-inline.h (debug_map): Minor tweak.
21938 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
21940 * builtins.c (builtin_with_linkage_p): Return true for stp[n]cpy.
21941 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Tidy up.
21943 2021-05-04 Richard Biener <rguenther@suse.de>
21945 PR tree-optimization/100329
21946 * tree-ssa-reassoc.c (can_reassociate_p): Do not reassociate
21948 (insert_stmt_after): Assert we're not running into asm goto.
21950 2021-05-04 Richard Biener <rguenther@suse.de>
21952 PR tree-optimization/100398
21953 * tree-ssa-dse.c (pass_dse::execute): Preserve control
21956 2021-05-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
21958 * builtins.c (try_store_by_multiple_pieces): Fix constfun's prototype.
21960 2021-05-04 Alexandre Oliva <oliva@adacore.com>
21962 * builtins.c (try_store_by_multiple_pieces): New.
21963 (expand_builtin_memset_args): Use it. If target_char_cast
21964 fails, proceed as for non-constant val. Pass len's ctz to...
21965 * expr.c (clear_storage_hints): ... this. Try store by
21966 multiple pieces after setmem.
21967 (clear_storage): Adjust.
21968 * expr.h (clear_storage_hints): Likewise.
21969 (try_store_by_multiple_pieces): Declare.
21970 * passes.def: Replace the last copy_prop with ccp.
21972 2021-05-03 Tom de Vries <tdevries@suse.de>
21975 * omp-low.c (lower_rec_input_clauses): Disable SIMT for user-defined
21978 2021-05-03 Richard Biener <rguenther@suse.de>
21980 * tree-ssa-dse.c (dse_classify_store): Track two PHI defs.
21982 2021-05-03 Richard Biener <rguenther@suse.de>
21984 * tree-ssa-dse.c: Do not include domwalk.h but cfganal.h.
21985 (dse_dom_walker): Remove.
21986 (dse_dom_walker::dse_optimize_stmt): Rename...
21987 (dse_optimize_stmt): ... to this, pass in live_bytes sbitmap.
21988 (dse_dom_walker::before_dom_children): Inline ...
21989 (pass_dse::execute): ... here. Perform a reverse program
21992 2021-05-03 H.J. Lu <hjl.tools@gmail.com>
21995 * configure: Regenerated.
21997 2021-05-03 Ilya Leoshkevich <iii@linux.ibm.com>
22000 * config/s390/s390.c (s390_hard_fp_reg_p): New function.
22001 (s390_md_asm_adjust): Handle hard registers.
22003 2021-05-03 Jakub Jelinek <jakub@redhat.com>
22005 PR tree-optimization/100382
22006 * tree-ssa-dse.c: Include tree-eh.h.
22007 (dse_dom_walker::before_dom_children): Don't remove stmts if
22008 stmt_unremovable_because_of_non_call_eh_p is true.
22010 2021-05-02 David Edelsohn <dje.gcc@gmail.com>
22012 * varasm.c (compute_reloc_for_var): Split out from...
22013 (get_variable_section): Use it.
22014 * output.h (compute_reloc_for_var): Declare.
22015 * config/rs6000/rs6000-protos.h
22016 (rs6000_xcoff_asm_output_aligned_decl_common): Change alignment to
22018 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address_aix):
22019 Don't append storage mapping class to symbol.
22020 (rs6000_xcoff_asm_named_section): Add BS and UL mapping classes.
22021 Don't convert TLS BSS to common.
22022 (rs6000_xcoff_unique_section): Don't fall back to select_secton.
22023 (rs6000_xcoff_section_type_flags): Add SECTION_BSS if DECL is
22025 (rs6000_xcoff_asm_globalize_decl_name): Don't strip storage
22027 (rs6000_xcoff_asm_output_aligned_decl_common): Align is unsigned int.
22028 If align is 0 from TLS class, use the same rules as varasm.c
22029 If not common, switch to BSS section manually.
22030 If common, emit appropriate comm or lcomm directive.
22031 (rs6000_xcoff_encode_section_info): Add logic to append all
22032 storage mapping classes.
22033 (rs6000_asm_weaken_decl): Adjust for qualname symbols.
22034 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
22035 rs6000_xcoff_asm_output_aligned_decl_common.
22036 (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
22037 rs6000_xcoff_asm_output_aligned_decl_common.
22038 (ASM_OUTPUT_TLS_COMMON): Use
22039 rs6000_xcoff_asm_output_aligned_decl_common.
22041 2021-05-02 Jakub Jelinek <jakub@redhat.com>
22044 * config/nvptx/nvptx.c (nvptx_sese_pseudo): Use nullptr instead of 0
22045 as first argument of pseudo_node_t constructors.
22047 2021-05-02 Jakub Jelinek <jakub@redhat.com>
22050 * config/i386/t-i386 (TM_H): Add $(srcdir)/config/i386/i386-isa.def.
22052 2021-05-01 Aldy Hernandez <aldyh@redhat.com>
22054 * value-range.cc (DEFINE_INT_RANGE_GC_STUBS): Remove.
22055 (gt_pch_nx (int_range<1> *&)): New.
22056 (gt_ggc_mx (int_range<1> *&)): New.
22057 * value-range.h (class irange): Add GTY support for
22060 2021-05-01 Geng Qi <gengqi@linux.alibaba.com>
22062 * doc/options.texi (Negative): Change either or to both and.
22064 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
22066 * config/aarch64/aarch64-simd-builtins.def: Add
22067 float_ml[as][q]_laneq builtin generator macros.
22068 * config/aarch64/aarch64-simd.md (mul_laneq<mode>3): Define.
22069 (aarch64_float_mla_laneq<mode>): Define.
22070 (aarch64_float_mls_laneq<mode>): Define.
22071 * config/aarch64/arm_neon.h (vmla_laneq_f32): Use RTL builtin
22072 instead of GCC vector extensions.
22073 (vmlaq_laneq_f32): Likewise.
22074 (vmls_laneq_f32): Likewise.
22075 (vmlsq_laneq_f32): Likewise.
22077 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
22079 * config/aarch64/aarch64-simd-builtins.def: Add
22080 float_ml[as]_lane builtin generator macros.
22081 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt<mode>):
22083 (mul_lane<mode>3): This, and re-order arguments.
22084 (aarch64_float_mla_lane<mode>): Define.
22085 (aarch64_float_mls_lane<mode>): Define.
22086 * config/aarch64/arm_neon.h (vmla_lane_f32): Use RTL builtin
22087 instead of GCC vector extensions.
22088 (vmlaq_lane_f32): Likewise.
22089 (vmls_lane_f32): Likewise.
22090 (vmlsq_lane_f32): Likewise.
22092 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
22094 * config/aarch64/aarch64-simd-builtins.def: Add float_ml[as]
22095 builtin generator macros.
22096 * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>):
22098 (aarch64_float_mls<mode>): Define.
22099 * config/aarch64/arm_neon.h (vmla_f32): Use RTL builtin
22100 instead of relying on GCC vector extensions.
22101 (vmla_f64): Likewise.
22102 (vmlaq_f32): Likewise.
22103 (vmlaq_f64): Likewise.
22104 (vmls_f32): Likewise.
22105 (vmls_f64): Likewise.
22106 (vmlsq_f32): Likewise.
22107 (vmlsq_f64): Likewise.
22108 * config/aarch64/iterators.md: Define VDQF_DF mode iterator.
22110 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
22112 * config/aarch64/aarch64-simd-builtins.def: Add
22113 float_ml[as]_n_builtin generator macros.
22114 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt_from_dup<mode>):
22116 (mul_n<mode>3): This, and re-order arguments.
22117 (aarch64_float_mla_n<mode>): Define.
22118 (aarch64_float_mls_n<mode>): Define.
22119 * config/aarch64/arm_neon.h (vmla_n_f32): Use RTL builtin
22120 instead of inline asm.
22121 (vmlaq_n_f32): Likewise.
22122 (vmls_n_f32): Likewise.
22123 (vmlsq_n_f32): Likewise.
22125 2021-04-30 Jonathan Wright <joanthan.wright@arm.com>
22127 * config/aarch64/aarch64-simd-builtins.def: Add pmull[2]
22128 builtin generator macros.
22129 * config/aarch64/aarch64-simd.md (aarch64_pmullv8qi): Define.
22130 (aarch64_pmull_hiv16qi_insn): Define.
22131 (aarch64_pmull_hiv16qi): Define.
22132 * config/aarch64/arm_neon.h (vmull_high_p8): Use RTL builtin
22133 instead of inline asm.
22134 (vmull_p8): Likewise.
22136 2021-04-30 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
22138 * config/avr/avr.md: Adjust peepholes to match and
22139 generate parallels with clobber of REG_CC.
22140 (mov<mode>_insn): Rename to mov<mode>_insn_split.
22141 (*mov<mode>_insn): Rename to mov<mode>_insn.
22143 2021-04-30 David Edelsohn <dje.gcc@gmail.com>
22145 * varasm.c (use_blocks_for_decl_p): Don't use section anchors
22146 for VAR_DECLs if -fdata-sections enabled.
22148 2021-04-30 Michael Meissner <meissner@linux.ibm.com>
22150 PR bootstrap/100327
22151 * config/rs6000/rs6000.c
22152 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Define.
22153 (rs6000_libgcc_floating_mode_supported_p): New target hook.
22155 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
22157 * tree-ssa-threadbackward.c (class thread_jumps): Split out code
22159 (class back_threader_registry): ...to here...
22160 (class back_threader_profitability): ...and here...
22161 (thread_jumps::thread_through_all_blocks): Remove argument.
22162 (back_threader_registry::back_threader_registry): New.
22163 (back_threader_registry::~back_threader_registry): New.
22164 (back_threader_registry::thread_through_all_blocks): New.
22165 (thread_jumps::profitable_jump_thread_path): Move from here...
22166 (back_threader_profitability::profitable_path_p): ...to here.
22167 (thread_jumps::find_taken_edge): New.
22168 (thread_jumps::convert_and_register_current_path): Move...
22169 (back_threader_registry::register_path): ...to here.
22170 (thread_jumps::register_jump_thread_path_if_profitable): Move...
22171 (thread_jumps::maybe_register_path): ...to here.
22172 (thread_jumps::handle_phi): Call find_taken_edge and
22173 maybe_register_path.
22174 (thread_jumps::handle_assignment): Same.
22175 (thread_jumps::fsm_find_control_statement_thread_paths): Remove
22176 tree argument to handle_phi and handle_assignment.
22177 (thread_jumps::find_jump_threads_backwards): Set m_name. Remove
22178 set of m_speed_p and m_max_threaded_paths.
22179 (pass_thread_jumps::execute): Remove second argument from
22180 find_jump_threads_backwards.
22181 (pass_early_thread_jumps::execute): Same.
22183 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
22185 * tree-ssa-dom.c (class dom_jump_threader_simplifier): New.
22186 (class dom_opt_dom_walker): Initialize some class variables.
22187 (pass_dominator::execute): Pass evrp_range_analyzer and
22188 dom_jump_threader_simplifier to dom_opt_dom_walker.
22189 Adjust for some functions moving into classes.
22190 (simplify_stmt_for_jump_threading): Adjust and move to...
22191 (jump_threader_simplifier::simplify): ...here.
22192 (dom_opt_dom_walker::before_dom_children): Adjust for
22193 m_evrp_range_analyzer.
22194 (dom_opt_dom_walker::after_dom_children): Remove x_vr_values hack.
22195 (test_for_singularity): Place in dom_opt_dom_walker class.
22196 (dom_opt_dom_walker::optimize_stmt): The argument
22197 evrp_range_analyzer is now a class field.
22198 * tree-ssa-threadbackward.c (class thread_jumps): Add m_registry.
22199 (thread_jumps::thread_through_all_blocks): New.
22200 (thread_jumps::convert_and_register_current_path): Use m_registry.
22201 (pass_thread_jumps::execute): Adjust for thread_through_all_blocks
22202 being in the threader class.
22203 (pass_early_thread_jumps::execute): Same.
22204 * tree-ssa-threadedge.c (threadedge_initialize_values): Move...
22205 (jump_threader::jump_threader): ...here.
22206 (threadedge_finalize_values): Move...
22207 (jump_threader::~jump_threader): ...here.
22208 (jump_threader::remove_jump_threads_including): New.
22209 (jump_threader::thread_through_all_blocks): New.
22210 (record_temporary_equivalences_from_phis): Move...
22211 (jump_threader::record_temporary_equivalences_from_phis): ...here.
22212 (record_temporary_equivalences_from_stmts_at_dest): Move...
22213 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
22215 (simplify_control_stmt_condition_1): Move to jump_threader class.
22216 (simplify_control_stmt_condition): Move...
22217 (jump_threader::simplify_control_stmt_condition): ...here.
22218 (thread_around_empty_blocks): Move...
22219 (jump_threader::thread_around_empty_blocks): ...here.
22220 (thread_through_normal_block): Move...
22221 (jump_threader::thread_through_normal_block): ...here.
22222 (thread_across_edge): Move...
22223 (jump_threader::thread_across_edge): ...here.
22224 (thread_outgoing_edges): Move...
22225 (jump_threader::thread_outgoing_edges): ...here.
22226 * tree-ssa-threadedge.h: Move externally facing functings...
22227 (class jump_threader): ...here...
22228 (class jump_threader_simplifier): ...and here.
22229 * tree-ssa-threadupdate.c (struct redirection_data): Remove comment.
22230 (jump_thread_path_allocator::jump_thread_path_allocator): New.
22231 (jump_thread_path_allocator::~jump_thread_path_allocator): New.
22232 (jump_thread_path_allocator::allocate_thread_edge): New.
22233 (jump_thread_path_allocator::allocate_thread_path): New.
22234 (jump_thread_path_registry::jump_thread_path_registry): New.
22235 (jump_thread_path_registry::~jump_thread_path_registry): New.
22236 (jump_thread_path_registry::allocate_thread_edge): New.
22237 (jump_thread_path_registry::allocate_thread_path): New.
22238 (dump_jump_thread_path): Make extern.
22239 (debug (const vec<jump_thread_edge *> &path)): New.
22240 (struct removed_edges): Move to tree-ssa-threadupdate.h.
22241 (struct thread_stats_d): Remove.
22242 (remove_ctrl_stmt_and_useless_edges): Make static.
22243 (lookup_redirection_data): Move...
22244 (jump_thread_path_registry::lookup_redirection_data): ...here.
22245 (ssa_redirect_edges): Make static.
22246 (thread_block_1): Move...
22247 (jump_thread_path_registry::thread_block_1): ...here.
22248 (thread_block): Move...
22249 (jump_thread_path_registry::thread_block): ...here.
22250 (thread_through_loop_header): Move...
22251 (jump_thread_path_registry::thread_through_loop_header): ...here.
22252 (mark_threaded_blocks): Move...
22253 (jump_thread_path_registry::mark_threaded_blocks): ...here.
22254 (debug_path): Move...
22255 (jump_thread_path_registry::debug_path): ...here.
22256 (debug_all_paths): Move...
22257 (jump_thread_path_registry::dump): ..here.
22258 (rewire_first_differing_edge): Move...
22259 (jump_thread_path_registry::rewire_first_differing_edge): ...here.
22260 (adjust_paths_after_duplication): Move...
22261 (jump_thread_path_registry::adjust_paths_after_duplication): ...here.
22262 (duplicate_thread_path): Move...
22263 (jump_thread_path_registry::duplicate_thread_path): ..here.
22264 (remove_jump_threads_including): Move...
22265 (jump_thread_path_registry::remove_jump_threads_including): ...here.
22266 (thread_through_all_blocks): Move to...
22267 (jump_thread_path_registry::thread_through_all_blocks): ...here.
22268 (delete_jump_thread_path): Remove.
22269 (register_jump_thread): Move...
22270 (jump_thread_path_registry::register_jump_thread): ...here.
22271 * tree-ssa-threadupdate.h: Move externally facing functions...
22272 (class jump_thread_path_allocator): ...here...
22273 (class jump_thread_path_registry): ...and here.
22274 (thread_through_all_blocks): Remove.
22275 (struct removed_edges): New.
22276 (register_jump_thread): Remove.
22277 (remove_jump_threads_including): Remove.
22278 (delete_jump_thread_path): Remove.
22279 (remove_ctrl_stmt_and_useless_edges): Remove.
22280 (free_dom_edge_info): New prototype.
22281 * tree-vrp.c: Remove x_vr_values hack.
22282 (class vrp_jump_threader_simplifier): New.
22283 (vrp_jump_threader_simplifier::simplify): New.
22284 (vrp_jump_threader::vrp_jump_threader): Adjust method signature.
22285 Remove m_dummy_cond.
22286 Instantiate m_simplifier and m_threader.
22287 (vrp_jump_threader::thread_through_all_blocks): New.
22288 (vrp_jump_threader::simplify_stmt): Remove.
22289 (vrp_jump_threader::after_dom_children): Do not set m_dummy_cond.
22290 Remove x_vr_values hack.
22291 (execute_vrp): Adjust for thread_through_all_blocks being in a
22294 2021-04-30 Christophe Lyon <christophe.lyon@linaro.org>
22296 * genflags.c (gen_insn): Print failed expansion string.
22298 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
22300 * expr.c (alignment_for_piecewise_move): Call mode_for_size
22301 without limit to MAX_FIXED_MODE_SIZE.
22303 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
22305 PR middle-end/90773
22306 * builtins.c (builtin_memset_gen_str): Don't use return from
22307 simplify_gen_subreg.
22309 2021-04-30 Uroš Bizjak <ubizjak@gmail.com>
22312 * config/i386/i386.md (*add<mode>3_carry_0r): New insn pattern.
22313 (*addsi3_carry_zext_0r): Ditto.
22314 (*sub<mode>3_carry_0): Ditto.
22315 (*subsi3_carry_zext_0r): Ditto.
22316 * config/i386/predicates.md (ix86_carry_flag_unset_operator):
22318 * config/i386/i386.c (ix86_rtx_costs) <case PLUS, case MINUS>:
22319 Also consider ix86_carry_flag_unset_operator to calculate
22320 the cost of adc/sbb insn.
22322 2021-04-30 Roman Zhuykov <zhroma@ispras.ru>
22324 PR rtl-optimization/100225
22325 PR rtl-optimization/84878
22326 * modulo-sched.c (sms_schedule): Use note_stores to skip loops
22327 where we have an instruction which touches (writes) any hard
22328 register from df->regular_block_artificial_uses set.
22329 Allow not-single-set instruction only right before basic block
22332 2021-04-30 Geng Qi <gengqi@linux.alibaba.com>
22334 * config/riscv/riscv.opt (march=,mabi=): Negative itself.
22336 2021-04-30 LevyHsu <admin@levyhsu.com>
22338 * config/riscv/riscv.c (riscv_min_arithmetic_precision): New.
22339 * config/riscv/riscv.h (TARGET_MIN_ARITHMETIC_PRECISION): New.
22340 * config/riscv/riscv.md (addv<mode>4, uaddv<mode>4): New.
22341 (subv<mode>4, usubv<mode>4, mulv<mode>4, umulv<mode>4): New.
22343 2021-04-29 Alexandre Oliva <oliva@adacore.com>
22345 * config.gcc: Merged x86 and x86_64 cpu_type-setting cases.
22347 2021-04-29 Alexandre Oliva <oliva@adacore.com>
22349 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Rename to...
22350 (ASM_OUTPUT_MAX_SKIP_ALIGN): ... this. Enclose in do/while(0).
22351 * config/i386/i386.c: Adjust.
22352 * config/i386/i386.md: Adjust.
22353 * config/i386/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Drop.
22354 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
22355 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
22356 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
22357 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
22358 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
22359 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
22360 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
22361 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
22362 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
22363 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
22365 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
22367 * config/i386/i386-expand.c (ix86_expand_int_compare):
22368 Swap operands of GTU and LEU comparison to emit carry flag comparison.
22369 * config/i386/i386.md (*add<mode>3_carry_0): Change insn
22370 predicate to allow more combine opportunities with memory operands.
22371 (*sub<mode>3_carry_0): Ditto.
22373 2021-04-29 Richard Sandiford <richard.sandiford@arm.com>
22375 PR rtl-optimization/100303
22376 * rtl-ssa/accesses.cc (function_info::make_use_available): Take a
22377 boolean that indicates whether the use will only be used in
22378 debug instructions. Treat it in the same way that existing
22379 cross-EBB debug references would be handled if so.
22380 (function_info::make_uses_available): Likewise.
22381 * rtl-ssa/functions.h (function_info::make_uses_available): Update
22382 prototype accordingly.
22383 (function_info::make_uses_available): Likewise.
22384 * fwprop.c (try_fwprop_subst): Update call accordingly.
22386 2021-04-29 Jeff Law <jlaw@tachyum.com>
22388 * config/nios2/nios2-protos.h (nios2_fpu_insn_enabled): Move outside
22391 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
22392 Richard Biener <rguenther@suse.de>
22395 * config/i386/i386-builtin.def (IX86_BUILTIN_MASKLOADPD)
22396 (IX86_BUILTIN_MASKLOADPS, IX86_BUILTIN_MASKLOADPD256)
22397 (IX86_BUILTIN_MASKLOADPS256, IX86_BUILTIN_MASKLOADD)
22398 (IX86_BUILTIN_MASKLOADQ, IX86_BUILTIN_MASKLOADD256)
22399 (IX86_BUILTIN_MASKLOADQ256): Move from SPECIAL_ARGS
22400 to PURE_ARGS category.
22401 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
22402 Handle PURE_ARGS category.
22403 * config/i386/i386-expand.c (ix86_expand_builtin): Ditto.
22405 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
22407 * configure.ac: Check for the presence of sys/locking.h header and
22408 for whether _LK_LOCK is supported by _locking.
22409 * configure: Regenerate.
22410 * config.in: Likewise.
22411 * gcov-io.h: Define GCOV_LOCKED_WITH_LOCKING if HOST_HAS_LK_LOCK.
22412 * gcov-io.c (gcov_open): Add support for GCOV_LOCKED_WITH_LOCKING.
22413 * system.h: Include <sys/locking.h> if HAVE_SYS_LOCKING_H.
22415 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
22417 * config/i386/predicates.md (fcmov_comparison_operator):
22418 Do not check for trivial FP comparison operator.
22419 <case GEU, case LTU>: Allow CCGZmode.
22420 <case GTU, case LEU>: Do not allow CCCmode.
22421 (ix86_comparison_operator) <case GTU, case LEU>: Allow only CCmode.
22422 (ix86_carry_flag_operator): Match only LTU and UNLT code.
22423 Do not check for trivial FP comparison operator. Allow CCGZmode.
22425 2021-04-29 Tom de Vries <tdevries@suse.de>
22427 * omp-expand.c (expand_omp_simd): Add step_orig, and replace uses of
22428 fd->loop.step by either step or orig_step.
22430 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
22432 * config/sparc/sparc.c (gen_load_pcrel_sym): Delete.
22433 (load_got_register): Do the PIC dance here.
22434 (sparc_legitimize_tls_address): Simplify.
22435 (sparc_emit_probe_stack_range): Likewise.
22436 (sparc32_initialize_trampoline): Likewise.
22437 (sparc64_initialize_trampoline): Likewise.
22438 * config/sparc/sparc.md (load_pcrel_sym<P:mode>): Add @ marker.
22439 (probe_stack_range<P:mode>): Likewise.
22440 (flush<P:mode>): Likewise.
22441 (tgd_hi22<P:mode>): Likewise.
22442 (tgd_lo10<P:mode>): Likewise.
22443 (tgd_add<P:mode>): Likewise.
22444 (tgd_call<P:mode>): Likewise.
22445 (tldm_hi22<P:mode>): Likewise.
22446 (tldm_lo10<P:mode>): Likewise.
22447 (tldm_add<P:mode>): Likewise.
22448 (tldm_call<P:mode>): Likewise.
22449 (tldo_hix22<P:mode>): Likewise.
22450 (tldo_lox10<P:mode>): Likewise.
22451 (tldo_add<P:mode>): Likewise.
22452 (tie_hi22<P:mode>): Likewise.
22453 (tie_lo10<P:mode>): Likewise.
22454 (tie_add<P:mode>): Likewise.
22455 (tle_hix22<P:mode>): Likewise.
22456 (tle_lox10<P:mode>): Likewise.
22457 (stack_protect_setsi): Rename to...
22458 (stack_protect_set32): ...this.
22459 (stack_protect_setdi): Rename to...
22460 (stack_protect_set64): ...this.
22461 (stack_protect_set): Adjust calls to above.
22462 (stack_protect_testsi): Rename to...
22463 (stack_protect_test32): ...this.
22464 (stack_protect_testdi): Rename to...
22465 (stack_protect_test64): ...this.
22466 (stack_protect_test): Adjust calls to above.
22468 2021-04-29 H.J. Lu <hjl.tools@gmail.com>
22470 PR middle-end/90773
22471 * builtins.c (builtin_memcpy_read_str): Add a dummy argument.
22472 (builtin_strncpy_read_str): Likewise.
22473 (builtin_memset_read_str): Add an argument for the previous RTL
22474 information and generate the new RTL from the previous RTL info.
22475 (builtin_memset_gen_str): Likewise.
22476 * builtins.h (builtin_strncpy_read_str): Update the prototype.
22477 (builtin_memset_read_str): Likewise.
22478 * expr.c (by_pieces_ninsns): If targetm.overlap_op_by_pieces_p()
22479 returns true, round up size and alignment to the widest integer
22480 mode for maximum size.
22481 (pieces_addr::adjust): Add a pointer to by_pieces_prev argument
22482 and pass it to m_constfn.
22483 (op_by_pieces_d): Add m_push and m_overlap_op_by_pieces.
22484 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
22485 initialize m_push. Initialize m_overlap_op_by_pieces with
22486 targetm.overlap_op_by_pieces_p ().
22487 (op_by_pieces_d::run): Pass the previous RTL information to
22488 pieces_addr::adjust and generate overlapping operations if
22489 m_overlap_op_by_pieces is true.
22491 (move_by_pieces_d::move_by_pieces_d): Updated for op_by_pieces_d
22493 (store_by_pieces_d::store_by_pieces_d): Updated for op_by_pieces_d
22495 (can_store_by_pieces): Use by_pieces_constfn on constfun.
22496 (store_by_pieces): Use by_pieces_constfn on constfun. Updated
22497 for op_by_pieces_d change.
22498 (clear_by_pieces_1): Add a dummy argument.
22499 (clear_by_pieces): Updated for op_by_pieces_d change.
22500 (compare_by_pieces_d::compare_by_pieces_d): Likewise.
22501 (string_cst_read_str): Add a dummy argument.
22502 * expr.h (by_pieces_constfn): Add a dummy argument.
22503 (by_pieces_prev): New.
22504 * target.def (overlap_op_by_pieces_p): New target hook.
22505 * config/i386/i386.c (TARGET_OVERLAP_OP_BY_PIECES_P): New.
22506 * doc/tm.texi.in: Add TARGET_OVERLAP_OP_BY_PIECES_P.
22507 * doc/tm.texi: Regenerated.
22509 2021-04-29 Richard Biener <rguenther@suse.de>
22511 PR tree-optimization/100253
22512 * tree-vect-stmts.c (vectorizable_load): Do not assume
22513 element alignment when DR_MISALIGNMENT is -1.
22514 (vectorizable_store): Likewise.
22516 2021-04-29 Jakub Jelinek <jakub@redhat.com>
22519 * config/aarch64/aarch64.c (aarch64_add_offset_1_temporaries): Use
22520 absu_hwi instead of abs_hwi.
22522 2021-04-29 Richard Biener <rguenther@suse.de>
22524 PR middle-end/38474
22525 * tree-ssa-structalias.c (add_graph_edge): Avoid direct
22526 forwarding when indirect forwarding through ESCAPED
22529 2021-04-29 Tom de Vries <tdevries@suse.de>
22532 * internal-fn.c (expand_GOMP_SIMT_ENTER_ALLOC)
22533 (expand_GOMP_SIMT_LAST_LANE, expand_GOMP_SIMT_ORDERED_PRED)
22534 (expand_GOMP_SIMT_VOTE_ANY, expand_GOMP_SIMT_XCHG_BFLY)
22535 (expand_GOMP_SIMT_XCHG_IDX): Ensure target is assigned to.
22537 2021-04-29 Richard Biener <rguenther@suse.de>
22539 PR tree-optimization/99912
22540 * tree-ssa-dse.c (dse_dom_walker::m_need_cfg_cleanup): New.
22541 (dse_dom_walker::todo): Likewise.
22542 (dse_dom_walker::dse_optimize_stmt): Move VDEF check to the
22544 (dse_dom_walker::before_dom_children): Remove trivially
22545 dead SSA defs and schedule CFG cleanup if we removed all
22547 (pass_dse::execute): Get TODO as computed by the DOM walker
22548 and return it. Wipe dominator info earlier.
22550 2021-04-29 Richard Biener <rguenther@suse.de>
22553 * ipa-prop.c (ipcp_modif_dom_walker::before_dom_children):
22554 Track blocks to cleanup EH in new m_need_eh_cleanup.
22555 (ipcp_modif_dom_walker::cleanup_eh): New.
22556 (ipcp_transform_function): Release dominator info before
22559 2021-04-29 Martin Sebor <msebor@redhat.com>
22561 PR middle-end/100250
22562 * attribs.c (attr_access::array_as_string): Avoid dereferencing
22563 a pointer when it's null.
22565 2021-04-29 Martin Sebor <msebor@redhat.com>
22567 * Makefile.in (OBJS): Add ipa-free-lang-data.o.
22568 * ipa-free-lang-data.cc: New file.
22569 * tree.c: Move pass free_lang_data to file above.
22570 (build_array_type_1): Declare extern.
22571 * tree.h (build_array_type_1): Declare.
22573 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22575 * config/aarch64/aarch64-simd-builtins.def: Modify comment to
22576 make consistent with updated RTL pattern.
22577 * config/aarch64/aarch64-simd.md (aarch64_<sur>qmovn<mode>):
22578 Implement using ss_truncate and us_truncate rather than
22580 * config/aarch64/iterators.md: Remove redundant unspecs and
22581 iterator: UNSPEC_[SU]QXTN and SUQMOVN respectively.
22583 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22585 * config/aarch64/arm_acle.h (__attribute__): Make intrinsic
22586 attributes consistent with those defined in arm_neon.h.
22588 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22590 * config/aarch64/arm_fp16.h (__attribute__): Make intrinsic
22591 attributes consistent with those defined in arm_neon.h.
22593 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22595 * config/aarch64/aarch64-simd-builtins.def: Add
22596 float_trunc_rodd builtin generator macros.
22597 * config/aarch64/aarch64-simd.md (aarch64_float_trunc_rodd_df):
22599 (aarch64_float_trunc_rodd_lo_v2sf): Define.
22600 (aarch64_float_trunc_rodd_hi_v4sf_le): Define.
22601 (aarch64_float_trunc_rodd_hi_v4sf_be): Define.
22602 (aarch64_float_trunc_rodd_hi_v4sf): Define.
22603 * config/aarch64/arm_neon.h (vcvtx_f32_f64): Use RTL builtin
22604 instead of inline asm.
22605 (vcvtx_high_f32_f64): Likewise.
22606 (vcvtxd_f32_f64): Likewise.
22607 * config/aarch64/iterators.md: Add FCVTXN unspec.
22609 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22611 * config/aarch64/aarch64-simd-builtins.def: Add tbx1 builtin
22613 * config/aarch64/aarch64-simd.md (aarch64_tbx1<mode>):
22615 * config/aarch64/arm_neon.h (vqtbx1_s8): USE RTL builtin
22616 instead of inline asm.
22617 (vqtbx1_u8): Likewise.
22618 (vqtbx1_p8): Likewise.
22619 (vqtbx1q_s8): Likewise.
22620 (vqtbx1q_u8): Likewise.
22621 (vqtbx1q_p8): Likewise.
22622 (vtbx2_s8): Likewise.
22623 (vtbx2_u8): Likewise.
22624 (vtbx2_p8): Likewise.
22626 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22628 * config/aarch64/aarch64-simd-builtins.def: Add tbl1 builtin
22630 * config/aarch64/arm_neon.h (vqtbl1_p8): Use RTL builtin
22631 instead of inline asm.
22632 (vqtbl1_s8): Likewise.
22633 (vqtbl1_u8): Likewise.
22634 (vqtbl1q_p8): Likewise.
22635 (vqtbl1q_s8): Likewise.
22636 (vqtbl1q_u8): Likewise.
22637 (vtbl1_s8): Likewise.
22638 (vtbl1_u8): Likewise.
22639 (vtbl1_p8): Likewise.
22640 (vtbl2_s8): Likewise.
22641 (vtbl2_u8): Likewise.
22642 (vtbl2_p8): Likewise.
22644 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22646 * config/aarch64/aarch64-simd-builtins.def: Add polynomial
22647 ssri_n buitin generator macro.
22648 * config/aarch64/arm_neon.h (vsri_n_p8): Use RTL builtin
22649 instead of inline asm.
22650 (vsri_n_p16): Likewise.
22651 (vsri_n_p64): Likewise.
22652 (vsriq_n_p8): Likewise.
22653 (vsriq_n_p16): Likewise.
22654 (vsriq_n_p64): Likewise.
22656 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22658 * config/aarch64/aarch64-simd-builtins.def: Use VALLP mode
22659 iterator for polynomial ssli_n builtin generator macro.
22660 * config/aarch64/arm_neon.h (vsli_n_p8): Use RTL builtin
22661 instead of inline asm.
22662 (vsli_n_p16): Likewise.
22663 (vsliq_n_p8): Likewise.
22664 (vsliq_n_p16): Likewise.
22665 * config/aarch64/iterators.md: Define VALLP mode iterator.
22667 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22669 * config/aarch64/aarch64-simd-builtins.def: Use VDQV_L
22670 iterator to generate [su]adalp RTL builtins.
22671 * config/aarch64/aarch64-simd.md: Use VDQV_L iterator in
22672 [su]adalp RTL pattern.
22673 * config/aarch64/arm_neon.h (vpadal_s32): Use RTL builtin
22674 instead of inline asm.
22675 (vpadal_u32): Likewise.
22677 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22679 * config/aarch64/aarch64-simd-builtins.def: Add [su]addlp
22680 builtin generator macros.
22681 * config/aarch64/aarch64-simd.md (aarch64_<su>addlp<mode>):
22683 * config/aarch64/arm_neon.h (vpaddl_s8): Use RTL builtin
22684 instead of inline asm.
22685 (vpaddl_s16): Likewise.
22686 (vpaddl_s32): Likewise.
22687 (vpaddl_u8): Likewise.
22688 (vpaddl_u16): Likewise.
22689 (vpaddl_u32): Likewise.
22690 (vpaddlq_s8): Likewise.
22691 (vpaddlq_s16): Likewise.
22692 (vpaddlq_s32): Likewise.
22693 (vpaddlq_u8): Likewise.
22694 (vpaddlq_u16): Likewise.
22695 (vpaddlq_u32): Liwewise.
22696 * config/aarch64/iterators.md: Define [SU]ADDLP unspecs with
22697 appropriate attributes.
22699 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22701 * config/aarch64/aarch64-simd-builtins.def: Use VDQ_I iterator
22702 for aarch64_addp<mode> builtin macro generator.
22703 * config/aarch64/aarch64-simd.md: Use VDQ_I iterator in
22704 aarch64_addp<mode> RTL pattern.
22705 * config/aarch64/arm_neon.h (vpaddq_s8): Use RTL builtin
22706 instead of inline asm.
22707 (vpaddq_s16): Likewise.
22708 (vpaddq_s32): Likewise.
22709 (vpaddq_s64): Likewise.
22710 (vpaddq_u8): Likewise.
22711 (vpaddq_u16): Likewise.
22712 (vpaddq_u32): Likewise.
22713 (vpaddq_u64): Likewise.
22715 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
22717 * config/aarch64/aarch64-simd-builtins.def: Add sq[r]dmulh_n
22718 builtin generator macros.
22719 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_n<mode>):
22721 * config/aarch64/arm_neon.h (vqdmulh_n_s16): Use RTL builtin
22722 instead of inline asm.
22723 (vqdmulh_n_s32): Likewise.
22724 (vqdmulhq_n_s16): Likewise.
22725 (vqdmulhq_n_s32): Likewise.
22726 (vqrdmulh_n_s16): Likewise.
22727 (vqrdmulh_n_s32): Likewise.
22728 (vqrdmulhq_n_s16): Likewise.
22729 (vqrdmulhq_n_s32): Likewise.
22731 2021-04-28 Tobias Burnus <tobias@codesourcery.com>
22733 * doc/install.texi (--enable-offload-defaulted): Document.
22735 2021-04-28 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
22737 * config/avr/avr-dimode.md: Turn existing patterns into
22738 define_insn_and_split style patterns where the splitter
22739 adds a clobber of the condition code register. Drop "cc"
22740 attribute. Add new patterns to match output of
22742 * config/avr/avr-fixed.md: Likewise.
22743 * config/avr/avr.c (cc_reg_rtx): New.
22744 (avr_parallel_insn_from_insns): Adjust insn count
22745 for removal of set of cc0.
22746 (avr_is_casesi_sequence): Likewise.
22747 (avr_casei_sequence_check_operands): Likewise.
22748 (avr_optimize_casesi): Likewise. Also insert
22749 new insns after jump_insn.
22750 (avr_pass_casesi::avr_rest_of_handle_casesi): Adjust
22751 for removal of set of cc0.
22752 (avr_init_expanders): Initialize cc_reg_rtx.
22753 (avr_regno_reg_class): Handle REG_CC.
22754 (cond_string): Remove usage of CC_OVERFLOW_UNUSABLE.
22755 (avr_notice_update_cc): Remove function.
22756 (ret_cond_branch): Remove usage of CC_OVERFLOW_UNUSABLE.
22757 (compare_condition): Adjust for PARALLEL with
22759 (out_shift_with_cnt): Likewise.
22760 (ashlhi3_out): Likewise.
22761 (ashrhi3_out): Likewise.
22762 (lshrhi3_out): Likewise.
22763 (avr_class_max_nregs): Return single reg for REG_CC.
22764 (avr_compare_pattern): Check for REG_CC instead
22766 (avr_reorg_remove_redundant_compare): Likewise.
22767 (avr_reorg):Adjust for PARALLEL with REG_CC clobber.
22768 (avr_hard_regno_nregs): Return single reg for REG_CC.
22769 (avr_hard_regno_mode_ok): Allow only CCmode for REG_CC.
22770 (avr_md_asm_adjust): Clobber REG_CC.
22771 (TARGET_HARD_REGNO_NREGS): Define.
22772 (TARGET_CLASS_MAX_NREGS): Define.
22773 (TARGET_MD_ASM_ADJUST): Define.
22774 * config/avr/avr.h (FIRST_PSEUDO_REGISTER): Adjust
22776 (enum reg_class): Add CC_REG class.
22777 (NOTICE_UPDATE_CC): Remove.
22778 (CC_OVERFLOW_UNUSABLE): Remove.
22779 (CC_NO_CARRY): Remove.
22780 * config/avr/avr.md: Turn existing patterns into
22781 define_insn_and_split style patterns where the splitter
22782 adds a clobber of the condition code register. Drop "cc"
22783 attribute. Add new patterns to match output of
22785 (sez): Remove unused pattern.
22787 2021-04-28 Richard Earnshaw <rearnsha@arm.com>
22790 * config/arm/arm.c (arm_hard_regno_mode_ok): Only allow VPR to be
22793 2021-04-28 Richard Sandiford <richard.sandiford@arm.com>
22796 * config/aarch64/constraints.md (Utq): Require the address to
22797 be valid for both the element mode and for V2DImode.
22799 2021-04-28 Jakub Jelinek <jakub@redhat.com>
22800 Tobias Burnus <tobias@codesourcery.com>
22802 * configure.ac (OFFLOAD_DEFAULTED): AC_DEFINE if offload-defaulted.
22803 * gcc.c (process_command): New variable.
22804 (driver::maybe_putenv_OFFLOAD_TARGETS): If OFFLOAD_DEFAULTED,
22805 set it if -foffload is defaulted.
22806 * lto-wrapper.c (OFFLOAD_TARGET_DEFAULT_ENV): Define.
22807 (compile_offload_image): If OFFLOAD_DEFAULTED and
22808 OFFLOAD_TARGET_DEFAULT is in the environment, don't fail
22809 if corresponding mkoffload can't be found.
22810 (compile_images_for_offload_targets): Likewise. Free and clear
22811 offload_names if no valid offload is found.
22812 * config.in: Regenerate.
22813 * configure: Regenerate.
22815 2021-04-28 Richard Biener <rguenther@suse.de>
22817 PR tree-optimization/100292
22818 * tree-vect-generic.c (expand_vector_condition): Do not fold
22821 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
22823 * config/rs6000/aix.h (SUBTARGET_DRIVER_SELF_SPECS): New.
22824 * config/rs6000/aix64.opt (m64): New.
22827 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
22829 * config/vax/vax.c (print_operand_address, vax_address_cost_1)
22830 (index_term_p): Handle ASHIFT too.
22832 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
22834 * config/vax/builtins.md (jbb<ccss>i<mode>): Remove operand #3.
22835 (sync_lock_test_and_set<mode>): Adjust accordingly.
22836 (sync_lock_release<mode>): Likewise.
22838 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
22840 * config/vax/vax-protos.h (adjacent_operands_p): Remove
22842 * config/vax/vax.c (adjacent_operands_p): Remove.
22844 2021-04-27 Maciej W. Rozycki <macro@linux-mips.org>
22846 * ifcvt.c (dead_or_predicable) [!IFCVT_MODIFY_TESTS]: Fall
22847 through to the non-conditional execution case if getting the
22848 condition for conditional execution has failed.
22850 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
22852 PR middle-end/100284
22853 * gimple.c (gimple_could_trap_p_1): Remove VEC_COND_EXPR test.
22854 * tree-eh.c (operation_could_trap_p): Handle VEC_COND_EXPR rather
22855 than asserting on it.
22857 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
22859 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): Protect
22860 with TARGET_AIX_OS.
22862 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
22865 * calls.c (precompute_register_parameters): Additionally test
22866 targetm.precompute_tls_p to pre-compute argument.
22867 * config/rs6000/aix.h (TARGET_PRECOMPUTE_TLS_P): Define.
22868 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): New.
22869 * target.def (precompute_tls_p): New.
22870 * doc/tm.texi.in (TARGET_PRECOMPUTE_TLS_P): Add hook documentation.
22871 * doc/tm.texi: Regenerated.
22873 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22876 * config/aarch64/aarch64.c (aarch64_print_operand): Cast -UINTVAL
22877 back to HOST_WIDE_INT.
22879 2021-04-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
22882 * simplify-rtx.c (simplify_context::simplify_subreg): Check the
22883 memory alignment for the outer mode.
22885 2021-04-27 H.J. Lu <hjl.tools@gmail.com>
22887 PR middle-end/90773
22888 * expr.c (op_by_pieces_d::get_usable_mode): New member function.
22889 (op_by_pieces_d::run): Cange a while loop to a do-while loop.
22891 2021-04-27 Alex Coplan <alex.coplan@arm.com>
22894 * config/arm/arm.c (arm_split_compare_and_swap): Fix up codegen
22895 with negative immediates: ensure we expand cbranchsi4_scratch
22896 correctly and ensure we satisfy its constraints.
22897 * config/arm/sync.md
22898 (@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): Don't
22899 attempt to tie two output operands together with constraints;
22900 collapse two alternatives.
22901 (@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
22902 * config/arm/thumb1.md (cbranchsi4_neg_late): New.
22904 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22907 * config/aarch64/predicates.md (aarch64_sub_immediate,
22908 aarch64_plus_immediate): Use -UINTVAL instead of -INTVAL.
22909 * config/aarch64/aarch64.md (casesi, rotl<mode>3): Likewise.
22910 * config/aarch64/aarch64.c (aarch64_print_operand,
22911 aarch64_split_atomic_op, aarch64_expand_subvti): Likewise.
22913 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22915 PR tree-optimization/100239
22916 * tree-vect-generic.c (lower_vec_perm): Don't accept constant
22917 permutations with all indices from the first zero element as vec_shl.
22919 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22921 PR rtl-optimization/100254
22922 * cfgcleanup.c (outgoing_edges_match): Check REG_EH_REGION on
22923 last1 and last2 insns rather than BB_END (bb1) and BB_END (bb2) insns.
22925 2021-04-27 Richard Biener <rguenther@suse.de>
22927 PR tree-optimization/99912
22928 * passes.def: Add comment about new TODO_remove_unused_locals.
22929 * tree-stdarg.c (pass_data_stdarg): Run TODO_remove_unused_locals
22932 2021-04-27 Richard Biener <rguenther@suse.de>
22934 PR tree-optimization/99912
22935 * passes.def (pass_all_optimizations): Add pass_dse before
22936 the first pass_dce, move the first pass_dse before the
22937 pass_dce following pass_pre.
22939 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22941 PR tree-optimization/95527
22942 * generic-match-head.c: Include tm.h.
22943 * gimple-match-head.c: Include tm.h.
22944 * match.pd (CLZ == INTEGER_CST): Don't use
22945 #ifdef CLZ_DEFINED_VALUE_AT_ZERO, only test CLZ_DEFINED_VALUE_AT_ZERO
22946 if clz == CFN_CLZ. Add missing val declaration.
22947 (CTZ cmp CST): New simplifications.
22949 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22951 PR tree-optimization/96696
22952 * expr.c (expand_expr_divmod): New function.
22953 (expand_expr_real_2) <case TRUNC_DIV_EXPR>: Use it for truncations and
22954 divisions. Formatting fixes.
22955 <case MULT_EXPR>: Optimize x / y * y as x - x % y if the latter is
22958 2021-04-27 Martin Jambor <mjambor@suse.cz>
22961 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
22962 If removing a call statement LHS SSA name, release it.
22964 2021-04-27 Richard Earnshaw <rearnsha@arm.com>
22967 * config/arm/arm.c (THUMB2_WORK_REGS): Check PIC_OFFSET_TABLE_REGNUM
22968 is valid before including it in the mask.
22970 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
22973 * config/aarch64/aarch64.c (aarch64_comp_type_attributes): Handle
22976 2021-04-27 Richard Biener <rguenther@suse.de>
22978 PR tree-optimization/100051
22979 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Add
22980 disambiguator based on access size vs. decl size.
22982 2021-04-27 Richard Biener <rguenther@suse.de>
22984 PR tree-optimization/100278
22985 * tree-ssa-pre.c (compute_avail): Give up when we cannot
22986 adjust TBAA beacuse of mismatching bases.
22988 2021-04-27 Jakub Jelinek <jakub@redhat.com>
22991 * config/i386/i386.md (*<insn><mode>3_mask, *<insn><mode>3_mask_1):
22992 For any_rotate define_insn_split and following splitters, use
22993 SWI iterator instead of SWI48.
22995 2021-04-27 Richard Biener <rguenther@suse.de>
22997 PR tree-optimization/99776
22998 * match.pd (bit_field_ref (ctor)): Relax element extract
22999 type compatibility checks.
23001 2021-04-27 Cui,Lili <lili.cui@intel.com>
23003 * common/config/i386/i386-common.c (processor_names):
23004 Sync processor_names with processor_type.
23005 * config/i386/i386-options.c (processor_cost_table):
23006 Sync processor_cost_table with processor_type.
23008 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
23010 * value-range.cc (irange::irange_set_1bit_anti_range): Add assert.
23011 (irange::set): Call irange_set_1bit_anti_range for handling all
23012 1-bit ranges. Fall through on ~[MIN,MAX].
23014 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
23016 * value-range.cc (irange::legacy_num_pairs): Remove.
23017 (irange::invert): Change gcc_assert to gcc_checking_assert.
23018 * value-range.h (irange::num_pairs): Adjust for a cached
23019 num_pairs(). Also, rename all gcc_assert's to
23020 gcc_checking_assert's.
23022 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
23024 * value-range.cc (irange::operator=): Set m_kind.
23025 (irange::copy_to_legacy): Handle varying and undefined sources
23026 as a legacy copy since they can be easily copied.
23027 (irange::irange_set): Set m_kind.
23028 (irange::irange_set_anti_range): Same.
23029 (irange::set): Rename normalize_min_max to normalize_kind.
23030 (irange::verify_range): Adjust for multi-ranges having the
23032 (irange::irange_union): Set m_kind.
23033 (irange::irange_intersect): Same.
23034 (irange::invert): Same.
23035 * value-range.h (irange::kind): Always return m_kind.
23036 (irange::varying_p): Rename to...
23037 (irange::varying_comptaible_p): ...this.
23038 (irange::undefined_p): Only look at m_kind.
23039 (irange::irange): Always set VR_UNDEFINED if applicable.
23040 (irange::set_undefined): Always set VR_UNDEFINED.
23041 (irange::set_varying): Always set m_kind to VR_VARYING.
23042 (irange::normalize_min_max): Rename to...
23043 (irange::normalize_kind): ...this.
23045 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
23047 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
23048 Adjust for constant_p including varying_p.
23049 * tree-vrp.c (vrp_prop::finalize): Same.
23050 (determine_value_range): Same.
23051 * vr-values.c (vr_values::range_of_expr): Same.
23052 * value-range.cc (irange::symbolic_p): Do not check varying_p.
23053 (irange::constant_p): Same.
23055 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
23057 * value-range.cc (irange::legacy_lower_bound): Replace
23058 !undefined_p check with num_ranges > 0.
23059 (irange::legacy_upper_bound): Same.
23060 * value-range.h (irange::type): Same.
23061 (irange::lower_bound): Same.
23062 (irange::upper_bound): Same.
23064 2021-04-26 Richard Biener <rguenther@suse.de>
23066 PR tree-optimization/99956
23067 * gimple-loop-interchange.cc (compute_access_stride):
23068 Try instantiating the access in a shallower loop nest
23069 if instantiating failed.
23070 (compute_access_strides): Pass adjustable loop_nest
23071 to compute_access_stride.
23073 2021-04-26 Christophe Lyon <christophe.lyon@linaro.org>
23075 * doc/sourcebuild.texi (arm_cmse_hw): Document.
23077 2021-04-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23079 * config/aarch64/iterators.md (vwcore): Handle V4BF, V8BF.
23081 2021-04-26 Thomas Schwinge <thomas@codesourcery.com>
23082 Nathan Sidwell <nathan@codesourcery.com>
23083 Tom de Vries <vries@codesourcery.com>
23084 Julian Brown <julian@codesourcery.com>
23085 Kwok Cheung Yeung <kcy@codesourcery.com>
23087 * omp-offload.c (oacc_validate_dims): Implement
23088 '-Wopenacc-parallelism'.
23089 * doc/invoke.texi (-Wopenacc-parallelism): Document.
23091 2021-04-26 Richard Biener <rguenther@suse.de>
23093 * tree-cfg.h (gimplify_build1): Remove.
23094 (gimplify_build2): Likewise.
23095 (gimplify_build3): Likewise.
23096 * tree-cfg.c (gimplify_build1): Move to tree-vect-generic.c.
23097 (gimplify_build2): Likewise.
23098 (gimplify_build3): Likewise.
23099 * tree-vect-generic.c (gimplify_build1): Move from tree-cfg.c.
23101 (gimplify_build2): Likewise.
23102 (gimplify_build3): Likewise.
23103 (tree_vec_extract): Use resimplify with following SSA edges.
23104 (expand_vector_parallel): Avoid passing NULL size/bitpos
23105 to tree_vec_extract.
23106 * expr.c (store_constructor): Deal with zero-element CTORs.
23107 * match.pd (bit_field_ref <vector CTOR>): Make sure to
23108 produce vector constants when possible.
23110 2021-04-26 Richard Biener <rguenther@suse.de>
23112 * tree-complex.c: Include gimple-fold.h.
23113 (expand_complex_addition): Use gimple_build.
23114 (expand_complex_multiplication_components): Likewise.
23115 (expand_complex_multiplication): Likewise.
23116 (expand_complex_div_straight): Likewise.
23117 (expand_complex_div_wide): Likewise.
23118 (expand_complex_division): Likewise.
23119 (expand_complex_conjugate): Likewise.
23120 (expand_complex_comparison): Likewise.
23122 2021-04-26 Richard Biener <rguenther@suse.de>
23124 * tree-ssa-phiopt.c (two_value_replacement): Remove use
23125 of legacy gimplify_buildN API.
23127 2021-04-26 Richard Biener <rguenther@suse.de>
23129 PR tree-optimization/99473
23130 * tree-ssa-phiopt.c (cond_store_replacement): Handle all
23133 2021-04-26 Richard Biener <rguenther@suse.de>
23135 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin):
23136 Use replace_call_with_value.
23138 2021-04-26 Richard Biener <rguenther@suse.de>
23140 * tree-ssa-propagate.h (valid_gimple_rhs_p): Remove.
23141 (update_gimple_call): Likewise.
23142 (update_call_from_tree): Likewise.
23143 * tree-ssa-propagate.c (valid_gimple_rhs_p): Remove.
23144 (valid_gimple_call_p): Likewise.
23145 (move_ssa_defining_stmt_for_defs): Likewise.
23146 (finish_update_gimple_call): Likewise.
23147 (update_gimple_call): Likewise.
23148 (update_call_from_tree): Likewise.
23149 (propagate_tree_value_into_stmt): Use replace_call_with_value.
23150 * gimple-fold.h (update_gimple_call): Declare.
23151 * gimple-fold.c (valid_gimple_rhs_p): Move here from
23152 tree-ssa-propagate.c.
23153 (update_gimple_call): Likewise.
23154 (valid_gimple_call_p): Likewise.
23155 (finish_update_gimple_call): Likewise, and simplify.
23156 (gimplify_and_update_call_from_tree): Implement
23157 update_call_from_tree functionality, avoid excessive
23158 push/pop_gimplify_context.
23159 (gimple_fold_builtin): Use only gimplify_and_update_call_from_tree.
23160 (gimple_fold_call): Likewise.
23161 * gimple-ssa-sprintf.c (try_substitute_return_value): Likewise.
23162 * tree-ssa-ccp.c (ccp_folder::fold_stmt): Likewise.
23163 (pass_fold_builtins::execute): Likewise.
23164 (optimize_stack_restore): Use replace_call_with_value.
23165 * tree-cfg.c (fold_loop_internal_call): Likewise.
23166 * tree-ssa-dce.c (maybe_optimize_arith_overflow): Use
23167 only gimplify_and_update_call_from_tree.
23168 * tree-ssa-strlen.c (handle_builtin_strlen): Likewise.
23169 (handle_builtin_strchr): Likewise.
23170 * tsan.c: Include gimple-fold.h instead of tree-ssa-propagate.h.
23172 2021-04-26 Jakub Jelinek <jakub@redhat.com>
23175 * vmsdbgout.c (ASM_OUTPUT_DEBUG_STRING, vmsdbgout_begin_block,
23176 vmsdbgout_end_block, lookup_filename, vmsdbgout_source_line): Remove
23179 2021-04-25 liuhongt <hongtao.liu@intel.com>
23182 * config/i386/i386-builtin.def (BDESC): Change the icode of
23183 the following builtins to CODE_FOR_nothing.
23184 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
23185 IX86_BUILTIN_PCMPEQB128, IX86_BUILTIN_PCMPEQW128,
23186 IX86_BUILTIN_PCMPEQD128, IX86_BUILTIN_PCMPEQQ,
23187 IX86_BUILTIN_PCMPEQB256, IX86_BUILTIN_PCMPEQW256,
23188 IX86_BUILTIN_PCMPEQD256, IX86_BUILTIN_PCMPEQQ256,
23189 IX86_BUILTIN_PCMPGTB128, IX86_BUILTIN_PCMPGTW128,
23190 IX86_BUILTIN_PCMPGTD128, IX86_BUILTIN_PCMPGTQ,
23191 IX86_BUILTIN_PCMPGTB256, IX86_BUILTIN_PCMPGTW256,
23192 IX86_BUILTIN_PCMPGTD256, IX86_BUILTIN_PCMPGTQ256.
23193 * config/i386/sse.md (avx2_eq<mode>3): Deleted.
23194 (sse2_eq<mode>3): Ditto.
23195 (sse4_1_eqv2di3): Ditto.
23196 (sse2_gt<mode>3): Rename to ..
23197 (*sse2_gt<mode>3): .. this.
23199 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
23202 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
23205 * config/darwin.c (darwin_binds_local_p): Assume that any
23206 public symbol might be interposed for PIC code. Update function
23207 header comment to reflect current Darwin capability.
23209 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
23212 * config/darwin.c (darwin_binds_local_p): Assume that any
23213 public symbol might be interposed for PIC code. Update function
23214 header comment to reflect current Darwin capability.
23216 2021-04-24 Richard Sandiford <richard.sandiford@arm.com>
23218 * doc/sourcebuild.texi: Document no-opts and any-opts target
23221 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
23223 * config/bpf/bpf.h (ASM_OUTPUT_ALIGNED_BSS): Use .type and .lcomm.
23225 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
23227 * config/bpf/bpf.h (FUNCTION_BOUNDARY): Set to 64.
23229 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
23232 * config/i386/i386-options.c (ix86_option_override_internal):
23233 Error out when -m96bit-long-double is used with 64bit targets.
23234 * config/i386/i386.md (*pushxf_rounded): Remove pattern.
23236 2021-04-23 Martin Liska <mliska@suse.cz>
23238 * lto-wrapper.c: Remove FIXME about usage of
23239 hardware_concurrency. The function is not on par with
23242 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
23245 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
23246 Copy operand 3 to operand 4. Use sse_reg_operand
23247 as operand 3 predicate.
23248 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): Ditto.
23249 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
23250 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
23251 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2):
23252 Copy operand 1 to operand 0.
23253 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): Ditto.
23254 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
23255 (LDX_ATOMIC/LDX_ATOMIC FP store peephole2 with mem blockage): Ditto.
23257 2021-04-23 Alex Coplan <alex.coplan@arm.com>
23259 PR rtl-optimization/100230
23260 * early-remat.c (early_remat::sort_candidates): Use delete[]
23261 instead of delete for array allocated with new[].
23263 2021-04-23 Richard Biener <rguenther@suse.de>
23265 * genmatch.c (lower_cond): Remove VEC_COND_EXPR special-casing.
23266 (capture_info::capture_info): Likewise.
23267 (capture_info::walk_match): Likewise.
23268 (expr::gen_transform): Likewise.
23269 (dt_simplify::gen_1): Likewise.
23270 * gimple-match-head.c (maybe_resimplify_conditional_op):
23271 Remove VEC_COND_EXPR special-casing.
23272 (gimple_simplify): Likewise.
23273 * gimple.c (gimple_could_trap_p_1): Adjust.
23274 * tree-ssa-pre.c (compute_avail): Allow VEC_COND_EXPR
23275 to participate in PRE.
23277 2021-04-23 Richard Biener <rguenther@suse.de>
23279 * cfganal.c (connect_infinite_loops_to_exit): First call
23280 add_noreturn_fake_exit_edges.
23281 * ipa-sra.c (process_scan_results): Do not call the now redundant
23282 add_noreturn_fake_exit_edges.
23283 * predict.c (tree_estimate_probability): Likewise.
23284 (rebuild_frequencies): Likewise.
23285 * store-motion.c (one_store_motion_pass): Likewise.
23287 2021-04-23 Richard Biener <rguenther@suse.de>
23289 PR tree-optimization/100222
23290 * predict.c (pass_profile::execute): Remove redundant call to
23291 mark_irreducible_loops.
23292 (report_predictor_hitrates): Likewise.
23294 2021-04-23 Richard Biener <rguenther@suse.de>
23296 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Avoid
23297 valid_gimple_rhs_p by instead gimplifying to one.
23299 2021-04-23 Richard Biener <rguenther@suse.de>
23301 PR tree-optimization/99971
23302 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
23303 Always use TBAA for loads.
23305 2021-04-23 liuhongt <hongtao.liu@intel.com>
23308 * config/i386/i386-options.c (ix86_option_override_internal):
23309 Clear MASK_AVX256_SPLIT_UNALIGNED_LOAD/STORE in x_target_flags
23310 when X86_TUNE_AVX256_UNALIGNED_LOAD/STORE_OPTIMAL is enabled
23311 by target attribute.
23313 2021-04-23 David Edelsohn <dje.gcc@gmail.com>
23315 * config/rs6000/aix71.h (PREFERRED_DEBUGGING_TYPE): Change to
23317 * config/rs6000/aix72.h (PREFERRED_DEBUGGING_TYPE): Same.
23319 2021-04-22 David Edelsohn <dje.gcc@gmail.com>
23321 * config.gcc (powerpc-ibm-aix6.*): Remove.
23322 * config/rs6000/aix61.h: Delete.
23324 2021-04-22 Martin Liska <mliska@suse.cz>
23326 PR testsuite/100159
23327 PR testsuite/100192
23328 * builtins.c (expand_builtin): Fix typos and missing comments.
23329 * dwarf2out.c (gen_subprogram_die): Likewise.
23330 (gen_struct_or_union_type_die): Likewise.
23332 2021-04-22 Uroš Bizjak <ubizjak@gmail.com>
23335 * config/i386/i386-expand.c (ix86_expand_convert_uns_sidf_sse):
23336 Remove the sign with FE_DOWNWARD, where x - x = -0.0.
23338 2021-04-21 Iain Sandoe <iain@sandoe.co.uk>
23340 * config/i386/darwin.h (TARGET_64BIT): Remove definition
23341 based on TARGET_ISA_64BIT.
23342 (TARGET_64BIT_P): Remove definition based on
23343 TARGET_ISA_64BIT_P().
23345 2021-04-21 Martin Liska <mliska@suse.cz>
23348 2021-04-21 Martin Liska <mliska@suse.cz>
23350 * lto-wrapper.c (cpuset_popcount): Remove.
23351 (init_num_threads): Remove and use hardware_concurrency.
23353 2021-04-21 Martin Liska <mliska@suse.cz>
23356 * main.c (main): Call toplev::finalize in CHECKING_P mode.
23357 * ipa-modref.c (ipa_modref_c_finalize): summaries are NULL
23358 when incremental LTO linking happens.
23360 2021-04-21 Martin Liska <mliska@suse.cz>
23362 * lto-wrapper.c (run_gcc): When -flto=jobserver is used, but the
23363 makeserver cannot be detected, then use -flto=N fallback.
23365 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
23367 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): When cross-compiling,
23368 default to yes for aarch64-linux-gnu.
23369 * configure: Regenerate.
23371 2021-04-21 Martin Liska <mliska@suse.cz>
23373 * lto-wrapper.c (cpuset_popcount): Remove.
23374 (init_num_threads): Remove and use hardware_concurrency.
23376 2021-04-21 Martin Liska <mliska@suse.cz>
23378 * config/i386/i386.c: Remove superfluous || TARGET_MACHO
23379 which remains to be '(... || 0)' and clang complains about it.
23380 * dwarf2out.c (AT_vms_delta): Declare conditionally.
23381 (add_AT_vms_delta): Likewise.
23382 * tree.c (fld_simplified_type): Use rather more common pattern
23383 for disabling of something (#if 0).
23384 (get_tree_code_name): Likewise.
23385 (verify_type_variant): Likewise.
23387 2021-04-21 Martin Liska <mliska@suse.cz>
23389 * config/i386/i386-expand.c (decide_alignment): Use newly named
23390 macro TARGET_CPU_P.
23391 * config/i386/i386.c (ix86_decompose_address): Likewise.
23392 (ix86_address_cost): Likewise.
23393 (ix86_lea_outperforms): Likewise.
23394 (ix86_avoid_lea_for_addr): Likewise.
23395 (ix86_add_stmt_cost): Likewise.
23396 * config/i386/i386.h (TARGET_*): Remove.
23397 (TARGET_CPU_P): New macro.
23398 * config/i386/i386.md: Use newly named macro TARGET_CPU_P.
23399 * config/i386/x86-tune-sched-atom.c (do_reorder_for_imul): Likewise.
23400 (swap_top_of_ready_list): Likewise.
23401 (ix86_atom_sched_reorder): Likewise.
23402 * config/i386/x86-tune-sched-bd.c (ix86_bd_has_dispatch): Likewise.
23403 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Likewise.
23405 2021-04-21 Martin Liska <mliska@suse.cz>
23407 * config/i386/i386-options.c (TARGET_EXPLICIT_NO_SAHF_P):
23409 (SET_TARGET_NO_SAHF): Likewise.
23410 (TARGET_EXPLICIT_PREFETCH_SSE_P): Likewise.
23411 (SET_TARGET_PREFETCH_SSE): Likewise.
23412 (TARGET_EXPLICIT_NO_TUNE_P): Likewise.
23413 (SET_TARGET_NO_TUNE): Likewise.
23414 (TARGET_EXPLICIT_NO_80387_P): Likewise.
23415 (SET_TARGET_NO_80387): Likewise.
23417 * config/i386/i386.h (TARGET_*): Remove.
23418 * opth-gen.awk: Generate new used macros.
23420 2021-04-21 Martin Liska <mliska@suse.cz>
23422 * config/i386/i386.h (PTA_*): Remove.
23423 (enum pta_flag): New.
23424 (DEF_PTA): Generate PTA_* values from i386-isa.def.
23425 * config/i386/i386-isa.def: New file.
23427 2021-04-21 Alex Coplan <alex.coplan@arm.com>
23430 * config/aarch64/aarch64-bti-insert.c (aarch64_bti_j_insn_p): New.
23431 (rest_of_insert_bti): Avoid inserting duplicate bti j insns for
23432 jump table targets.
23434 2021-04-21 H.J. Lu <hjl.tools@gmail.com>
23436 * config.gcc: Install mwaitintrin.h for i[34567]86-*-* and
23437 x86_64-*-* targets.
23438 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_MWAIT_SET):
23440 (OPTION_MASK_ISA2_MWAIT_UNSET): Likewise.
23441 (ix86_handle_option): Handle -mmwait.
23442 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
23443 Replace OPTION_MASK_ISA_SSE3 with OPTION_MASK_ISA2_MWAIT on
23444 __builtin_ia32_monitor and __builtin_ia32_mwait.
23445 * config/i386/i386-options.c (isa2_opts): Add -mmwait.
23446 (ix86_valid_target_attribute_inner_p): Likewise.
23447 (ix86_option_override_internal): Enable mwait/monitor
23448 instructions for -msse3.
23449 * config/i386/i386.h (TARGET_MWAIT): New.
23450 (TARGET_MWAIT_P): Likewise.
23451 * config/i386/i386.opt: Add -mmwait.
23452 * config/i386/mwaitintrin.h: New file.
23453 * config/i386/pmmintrin.h: Include <mwaitintrin.h>.
23454 * config/i386/sse.md (sse3_mwait): Replace TARGET_SSE3 with
23456 (@sse3_monitor_<mode>): Likewise.
23457 * config/i386/x86gprintrin.h: Include <mwaitintrin.h>.
23458 * doc/extend.texi: Document mwait target attribute.
23459 * doc/invoke.texi: Document -mmwait.
23461 2021-04-21 Martin Liska <mliska@suse.cz>
23463 * config/i386/i386-options.c (DEF_ENUM): Remove it.
23464 * config/i386/i386-opts.h (DEF_ENUM): Likewise.
23465 * config/i386/stringop.def (DEF_ENUM): Likewise.
23467 2021-04-21 Martin Liska <mliska@suse.cz>
23469 * tree-cfg.c (gimple_verify_flow_info): Use qD instead
23470 of print_generic_expr.
23472 2021-04-21 Jakub Jelinek <jakub@redhat.com>
23474 PR rtl-optimization/100148
23475 * cprop.c (constprop_register): Use next_nondebug_insn instead of
23478 2021-04-21 Martin Liska <mliska@suse.cz>
23481 * cgraphunit.c (cgraph_node::analyze): Remove duplicate
23482 free_dominance_info calls.
23484 2021-04-21 Richard Biener <rguenther@suse.de>
23486 * gimple-fold.c (maybe_fold_reference): Remove is_lhs
23487 parameter (and assume it to be false).
23488 (fold_gimple_assign): Adjust, remove all callers of
23489 maybe_fold_reference calling it with is_lhs true.
23490 (gimple_fold_call): Likewise.
23491 (fold_stmt_1): Likewise.
23493 2021-04-21 Richard Biener <rguenther@suse.de>
23495 * fold-const.c (pedantic_non_lvalue_loc): Remove.
23496 (fold_binary_loc): Adjust.
23497 (fold_ternary_loc): Likewise.
23499 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
23501 PR middle-end/100130
23502 * varasm.c (get_block_for_decl): Make sure that any use of the
23503 retain attribute matches the section's retain flag.
23504 (switch_to_section): Check for retain mismatches even when
23505 changing sections, but do not warn if the given decl is the
23506 section's named.decl.
23507 (output_object_block): Pass the first decl in the block (if any)
23508 to switch_to_section.
23510 2021-04-20 H.J. Lu <hjl.tools@gmail.com>
23512 * config/i386/i386-c.c (ix86_target_macros_internal): Define
23513 __CRC32__ for -mcrc32.
23514 * config/i386/i386-options.c (ix86_option_override_internal):
23515 Enable crc32 instruction for -msse4.2.
23516 * config/i386/i386.md (sse4_2_crc32<mode>): Remove TARGET_SSE4_2
23518 (sse4_2_crc32di): Likewise.
23519 * config/i386/ia32intrin.h: Use crc32 target option for CRC32
23522 2021-04-20 Segher Boessenkool <segher@kernel.crashing.org>
23525 * config/rs6000/rs6000.c (rs6000_machine_from_flags): Do not consider
23528 2021-04-20 Martin Liska <mliska@suse.cz>
23530 * doc/invoke.texi: Fix typo.
23531 * params.opt: Likewise.
23533 2021-04-20 Martin Liska <mliska@suse.cz>
23535 * doc/invoke.texi: Document new param.
23537 2021-04-19 Andrew MacLeod <amacleod@redhat.com>
23539 PR tree-optimization/100081
23540 * gimple-range-cache.h (ranger_cache): Inherit from gori_compute
23541 rather than gori_compute_cache.
23542 * gimple-range-gori.cc (is_gimple_logical_p): Move to top of file.
23543 (range_def_chain::m_logical_depth): New member.
23544 (range_def_chain::range_def_chain): Initialize m_logical_depth.
23545 (range_def_chain::get_def_chain): Don't build defchains through more
23546 than LOGICAL_LIMIT logical expressions.
23547 * params.opt (param_ranger_logical_depth): New.
23549 2021-04-19 Richard Earnshaw <rearnsha@arm.com>
23552 * config/arm/arm.c (arm_configure_build_target): Do not strip
23553 extended FPU/SIMD feature bits from the target ISA when -mfpu
23554 is specified (partial revert of r11-8168).
23556 2021-04-19 Thomas Schwinge <thomas@codesourcery.com>
23558 * params.opt (-param=openacc-kernels=): Add.
23559 * omp-oacc-kernels-decompose.cc
23560 (pass_omp_oacc_kernels_decompose::gate): Use it.
23561 * doc/invoke.texi (-fopenacc-kernels=@var{mode}): Move...
23562 (--param): ... here, 'openacc-kernels'.
23564 2021-04-19 Martin Liska <mliska@suse.cz>
23567 * gengtype.c (finish_root_table): Align function arguments
23568 in between declaration and definition.
23570 2021-04-19 Eric Botcazou <ebotcazou@adacore.com>
23572 * config/i386/winnt.c (i386_pe_seh_cold_init): Properly deal with
23573 frames larger than the SEH maximum frame size.
23575 2021-04-18 Segher Boessenkool <segher@kernel.crashing.org>
23577 PR rtl-optimization/99927
23578 * combine.c (distribute_notes) [REG_UNUSED]: If the register already
23579 is dead, just drop it.
23581 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
23584 * config/i386/winnt-d.c (TARGET_D_TEMPLATES_ALWAYS_COMDAT): Define.
23585 * doc/tm.texi: Regenerate.
23586 * doc/tm.texi.in (D language and ABI): Add @hook for
23587 TARGET_D_TEMPLATES_ALWAYS_COMDAT.
23589 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
23591 * config/darwin-d.c (darwin_d_handle_target_object_format): New
23593 (darwin_d_register_target_info): New function.
23594 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23595 * config/dragonfly-d.c (dragonfly_d_handle_target_object_format): New
23597 (dragonfly_d_register_target_info): New function.
23598 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23599 * config/freebsd-d.c (freebsd_d_handle_target_object_format): New
23601 (freebsd_d_register_target_info): New function.
23602 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23603 * config/glibc-d.c (glibc_d_handle_target_object_format): New
23605 (glibc_d_register_target_info): New function.
23606 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23607 * config/i386/i386-d.c (ix86_d_handle_target_object_format): New
23609 (ix86_d_register_target_info): Add ix86_d_handle_target_object_format
23610 as handler for objectFormat key.
23611 * config/i386/winnt-d.c (winnt_d_handle_target_object_format): New
23613 (winnt_d_register_target_info): New function.
23614 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23615 * config/netbsd-d.c (netbsd_d_handle_target_object_format): New
23617 (netbsd_d_register_target_info): New function.
23618 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23619 * config/openbsd-d.c (openbsd_d_handle_target_object_format): New
23621 (openbsd_d_register_target_info): New function.
23622 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23623 * config/pa/pa-d.c (pa_d_handle_target_object_format): New function.
23624 (pa_d_register_target_info): Add pa_d_handle_target_object_format as
23625 handler for objectFormat key.
23626 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_object_format): New
23628 (rs6000_d_register_target_info): Add
23629 rs6000_d_handle_target_object_format as handler for objectFormat key.
23630 * config/sol2-d.c (solaris_d_handle_target_object_format): New
23632 (solaris_d_register_target_info): New function.
23633 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
23635 2021-04-16 Jakub Jelinek <jakub@redhat.com>
23638 * config/aarch64/aarch64.c (aarch64_function_arg_alignment): Change
23639 abi_break argument from bool * to unsigned *, store there the pre-GCC 9
23641 (aarch64_layout_arg, aarch64_gimplify_va_arg_expr): Adjust callers.
23642 (aarch64_function_arg_regno_p): Likewise. Only emit -Wpsabi note if
23643 the old and new alignment after applying MIN/MAX to it is different.
23645 2021-04-16 Tamar Christina <tamar.christina@arm.com>
23648 * config/aarch64/aarch64-sve.md (@aarch64_sve_trn1_conv<mode>): New.
23649 * config/aarch64/aarch64.c (aarch64_expand_sve_const_pred_trn): Use new
23651 * config/aarch64/iterators.md (UNSPEC_TRN1_CONV): New.
23653 2021-04-16 Bill Schmidt <wschmidt@linux.ibm.com>
23655 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Revise
23656 this section and its subsections.
23658 2021-04-16 Jakub Jelinek <jakub@redhat.com>
23661 * config/aarch64/aarch64.md (*neg_asr_si2_extr, *extrsi5_insn_di): New
23662 define_insn patterns.
23664 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
23666 PR rtl-optimization/98689
23667 * reg-notes.def (UNTYPED_CALL): New note.
23668 * combine.c (distribute_notes): Handle it.
23669 * emit-rtl.c (try_split): Likewise.
23670 * rtlanal.c (rtx_properties::try_to_add_insn): Likewise. Assume
23671 that calls with the note implicitly set all return value registers.
23672 * builtins.c (expand_builtin_apply): Add a REG_UNTYPED_CALL
23675 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
23677 PR rtl-optimization/99596
23678 * rtlanal.c (rtx_properties::try_to_add_insn): Don't add global
23679 register accesses for const calls. Assume that pure functions
23680 can only read from global registers. Ignore cases in which
23681 the stack pointer has been marked global.
23683 2021-04-16 Jakub Jelinek <jakub@redhat.com>
23686 * tree-vect-loop.c (vect_transform_loop): Don't remove just
23687 dead scalar .MASK_LOAD calls, but also dead .COND_* calls - replace
23688 them by their last argument.
23690 2021-04-15 Martin Liska <mliska@suse.cz>
23692 * doc/invoke.texi: Other params don't use it, remove it.
23694 2021-04-15 Richard Biener <rguenther@suse.de>
23696 * gimple-builder.h: Add deprecation note.
23698 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
23701 * attribs.h (restrict_type_identity_attributes_to): Declare.
23702 * attribs.c (restrict_type_identity_attributes_to): New function.
23704 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
23707 * attribs.h (affects_type_identity_attributes): Declare.
23708 * attribs.c (remove_attributes_matching): New function.
23709 (affects_type_identity_attributes): Likewise.
23711 2021-04-15 Jakub Jelinek <jakub@redhat.com>
23714 * config/aarch64/aarch64.md (*<LOGICAL:optab>_<SHIFT:optab><mode>3):
23715 Add combine splitters for *<LOGICAL:optab>_ashl<mode>3 with
23716 ZERO_EXTEND, SIGN_EXTEND or AND.
23718 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
23720 PR rtl-optimization/99929
23721 * rtl.h (same_vector_encodings_p): New function.
23722 * cse.c (exp_equiv_p): Check that CONST_VECTORs have the same encoding.
23723 * cselib.c (rtx_equal_for_cselib_1): Likewise.
23724 * jump.c (rtx_renumbered_equal_p): Likewise.
23725 * lra-constraints.c (operands_match_p): Likewise.
23726 * reload.c (operands_match_p): Likewise.
23727 * rtl.c (rtx_equal_p_cb, rtx_equal_p): Likewise.
23729 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
23731 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Print
23732 more information about variable-length CONST_VECTORs.
23734 2021-04-14 Vladimir N. Makarov <vmakarov@redhat.com>
23736 PR rtl-optimization/100066
23737 * lra-constraints.c (split_reg): Check paradoxical_subreg_p for
23738 ordered modes when choosing splitting mode for hard reg.
23740 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
23743 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector_sel):
23745 (aarch64_expand_sve_const_vector): Use it for nelts_per_pattern==2.
23747 2021-04-14 Andreas Krebbel <krebbel@linux.ibm.com>
23749 * config/s390/s390-builtins.def (O_M5, O_M12, ...): Add new macros
23750 for mask operand types.
23751 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
23752 (s390_vec_permi_dbl, s390_vpdi): Use the M5 type for the immediate
23754 (s390_vec_msum_u128, s390_vmslg): Use the M12 type for the
23756 * config/s390/s390.c (s390_const_operand_ok): Check the new
23757 operand types and generate a list of valid values.
23759 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
23761 * doc/tm.texi: Regenerate.
23762 * doc/tm.texi.in (D language and ABI): Add @hook for
23763 TARGET_D_REGISTER_OS_TARGET_INFO.
23765 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
23767 * config/aarch64/aarch64-d.c (aarch64_d_handle_target_float_abi): New
23769 (aarch64_d_register_target_info): New function.
23770 * config/aarch64/aarch64-protos.h (aarch64_d_register_target_info):
23772 * config/aarch64/aarch64.h (TARGET_D_REGISTER_CPU_TARGET_INFO):
23774 * config/arm/arm-d.c (arm_d_handle_target_float_abi): New function.
23775 (arm_d_register_target_info): New function.
23776 * config/arm/arm-protos.h (arm_d_register_target_info): Declare.
23777 * config/arm/arm.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23778 * config/i386/i386-d.c (ix86_d_handle_target_float_abi): New function.
23779 (ix86_d_register_target_info): New function.
23780 * config/i386/i386-protos.h (ix86_d_register_target_info): Declare.
23781 * config/i386/i386.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23782 * config/mips/mips-d.c (mips_d_handle_target_float_abi): New function.
23783 (mips_d_register_target_info): New function.
23784 * config/mips/mips-protos.h (mips_d_register_target_info): Declare.
23785 * config/mips/mips.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23786 * config/pa/pa-d.c (pa_d_handle_target_float_abi): New function.
23787 (pa_d_register_target_info): New function.
23788 * config/pa/pa-protos.h (pa_d_register_target_info): Declare.
23789 * config/pa/pa.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23790 * config/riscv/riscv-d.c (riscv_d_handle_target_float_abi): New
23792 (riscv_d_register_target_info): New function.
23793 * config/riscv/riscv-protos.h (riscv_d_register_target_info): Declare.
23794 * config/riscv/riscv.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23795 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_float_abi): New
23797 (rs6000_d_register_target_info): New function.
23798 * config/rs6000/rs6000-protos.h (rs6000_d_register_target_info):
23800 * config/rs6000/rs6000.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23801 * config/s390/s390-d.c (s390_d_handle_target_float_abi): New function.
23802 (s390_d_register_target_info): New function.
23803 * config/s390/s390-protos.h (s390_d_register_target_info): Declare.
23804 * config/s390/s390.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23805 * config/sparc/sparc-d.c (sparc_d_handle_target_float_abi): New
23807 (sparc_d_register_target_info): New function.
23808 * config/sparc/sparc-protos.h (sparc_d_register_target_info): Declare.
23809 * config/sparc/sparc.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
23810 * doc/tm.texi: Regenerate.
23811 * doc/tm.texi.in (D language and ABI): Add @hook for
23812 TARGET_D_REGISTER_CPU_TARGET_INFO.
23814 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
23816 * config/i386/i386-d.c (ix86_d_has_stdcall_convention): New function.
23817 * config/i386/i386-protos.h (ix86_d_has_stdcall_convention): Declare.
23818 * config/i386/i386.h (TARGET_D_HAS_STDCALL_CONVENTION): Define.
23819 * doc/tm.texi: Regenerate.
23820 * doc/tm.texi.in (D language and ABI): Add @hook for
23821 TARGET_D_HAS_STDCALL_CONVENTION.
23823 2021-04-14 Richard Biener <rguenther@suse.de>
23825 * tree-cfg.c (verify_gimple_assign_ternary): Verify that
23826 VEC_COND_EXPRs have a gimple_val condition.
23827 * tree-ssa-propagate.c (valid_gimple_rhs_p): VEC_COND_EXPR
23828 can no longer have a GENERIC condition.
23830 2021-04-14 Richard Earnshaw <rearnsha@arm.com>
23833 * config/arm/arm.c (arm_configure_build_target): Strip isa_all_fpbits
23834 from the isa_delta when -mfpu has been used.
23835 (arm_options_perform_arch_sanity_checks): It's the architecture that
23836 lacks an FPU not the processor.
23838 2021-04-13 Richard Biener <rguenther@suse.de>
23840 PR tree-optimization/100053
23841 * tree-ssa-sccvn.c (vn_nary_op_get_predicated_value): Do
23842 not use optimistic dominance queries for backedges to validate
23844 (dominated_by_p_w_unex): Add parameter to ignore executable
23845 state on backedges.
23846 (rpo_elim::eliminate_avail): Adjust.
23848 2021-04-13 Jakub Jelinek <jakub@redhat.com>
23851 * config/aarch64/aarch64.md (*aarch64_bfxil<mode>_extr,
23852 *aarch64_bfxilsi_extrdi): New define_insn patterns.
23854 2021-04-13 Jakub Jelinek <jakub@redhat.com>
23857 * simplify-rtx.c (simplify_immed_subreg): For MODE_COMPOSITE_P
23858 outermode, return NULL if the result doesn't encode back to the
23859 original byte sequence.
23860 (simplify_gen_subreg): Don't create SUBREGs from constants to
23861 MODE_COMPOSITE_P outermode.
23863 2021-04-12 Jakub Jelinek <jakub@redhat.com>
23865 PR rtl-optimization/99905
23866 * combine.c (expand_compound_operation): If pos + len > modewidth,
23867 perform the right shift by pos in inner_mode and then convert to mode,
23868 instead of trying to simplify a shift of rtx with inner_mode by pos
23869 as if it was a shift in mode.
23871 2021-04-12 Jakub Jelinek <jakub@redhat.com>
23874 * combine.c (simplify_and_const_int_1): Don't optimize varop
23875 away if it has side-effects.
23877 2021-04-12 Martin Liska <mliska@suse.cz>
23879 * doc/extend.texi: Escape @smallexample content.
23881 2021-04-12 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
23883 * config/s390/s390.md ("*movdi_31", "*movdi_64"): Add
23884 alternative in order to load a DFP zero.
23886 2021-04-12 Martin Liska <mliska@suse.cz>
23888 * doc/extend.texi: Be more precise in documentation
23889 of symver attribute.
23891 2021-04-12 Martin Liska <mliska@suse.cz>
23894 * gimplify.c (gimplify_expr): Right now, we unpoison all
23895 variables before a goto <dest>. We should not do it if we are
23898 2021-04-12 Cui,Lili <lili.cui@intel.com>
23900 * common/config/i386/cpuinfo.h (get_intel_cpu): Handle
23902 * common/config/i386/i386-common.c (processor_names): Add
23904 (processor_alias_table): Add rocketlake.
23905 * common/config/i386/i386-cpuinfo.h (processor_subtypes): Add
23906 INTEL_COREI7_ROCKETLAKE.
23907 * config.gcc: Add -march=rocketlake.
23908 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
23910 * config/i386/i386-options.c (m_ROCKETLAKE) : Define.
23911 (processor_cost_table): Add rocketlake cost.
23912 * config/i386/i386.h (ix86_size_cost) : Define
23914 (processor_type) : Add PROCESSOR_ROCKETLAKE.
23915 (PTA_ROCKETLAKE): Ditto.
23916 * doc/extend.texi: Add rocketlake.
23917 * doc/invoke.texi: Add rocketlake.
23919 2021-04-12 Cui,Lili <lili.cui@intel.com>
23921 * config/i386/i386.h (PTA_ALDERLAKE): Change alderlake ISA list.
23922 * config/i386/i386-options.c (m_CORE_AVX2): Add m_ALDERLAKE.
23923 * common/config/i386/cpuinfo.h (get_intel_cpu): Add AlderLake model.
23924 * doc/invoke.texi: Change alderlake ISA list.
23926 2021-04-11 Hafiz Abid Qadeer <abidh@codesourcery.com>
23928 PR middle-end/98088
23929 * omp-expand.c (expand_oacc_collapse_init): Update condition in
23932 2021-04-10 H.J. Lu <hjl.tools@gmail.com>
23935 * config/i386/serializeintrin.h (_serialize): Defined as macro.
23937 2021-04-10 Jakub Jelinek <jakub@redhat.com>
23940 * expr.c (expand_expr_addr_expr_1): Test is_global_var rather than
23941 just TREE_STATIC on COMPOUND_LITERAL_EXPR_DECLs.
23943 2021-04-10 Jakub Jelinek <jakub@redhat.com>
23945 PR middle-end/99989
23946 * gimple-ssa-warn-alloca.c
23947 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit to
23948 0 with integer precision unconditionally.
23950 2021-04-10 Jakub Jelinek <jakub@redhat.com>
23952 PR rtl-optimization/98601
23953 * rtlanal.c (rtx_addr_can_trap_p_1): Allow in assert unknown size
23954 not just for BLKmode, but also for VOIDmode. For STRICT_ALIGNMENT
23955 unaligned_mems handle VOIDmode like BLKmode.
23957 2021-04-10 Jan Hubicka <hubicka@ucw.cz>
23960 * tree.c (free_lang_data_in_decl): Do not release body of
23961 declare_variant_alt.
23963 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
23965 * config/aarch64/aarch64.c (aarch64_option_restore): If the
23966 architecture was specified explicitly and the tuning wasn't,
23967 tune for the architecture rather than the configured default CPU.
23969 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
23971 * config/aarch64/aarch64.md (tlsdesc_small_sve_<mode>): Use X30
23972 as the temporary register.
23974 2021-04-09 Martin Liska <mliska@suse.cz>
23976 * doc/extend.texi: Move non-target attributes on the top level.
23978 2021-04-09 Martin Liska <mliska@suse.cz>
23980 * doc/invoke.texi: Document minimum and maximum value of the
23981 argument for both supported compression algorithms.
23983 2021-04-08 David Edelsohn <dje.gcc@gmail.com>
23985 * config/rs6000/rs6000.c (rs6000_xcoff_select_section): Select
23986 TLS BSS before TLS data.
23987 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .comm.
23989 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
23991 * doc/sourcebuild.texi (stdint_types_mbig_endian): Document.
23993 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
23995 * match.pd: Extend vec_cond folds to handle shifts.
23997 2021-04-08 Maciej W. Rozycki <macro@orcam.me.uk>
23999 * config/vax/vax.md: Fix comment for `*bit<mode>' pattern's
24002 2021-04-08 Alex Coplan <alex.coplan@arm.com>
24005 * config/arm/iterators.md (MVE_vecs): New.
24006 (V_elem): Also handle V2DF.
24007 * config/arm/mve.md (*mve_mov<mode>): Rename to ...
24008 (*mve_vdup<mode>): ... this. Remove second alternative since
24009 vec_duplicate of const_int is not canonical RTL, and we don't
24010 want to match symbol_refs.
24011 (*mve_vec_duplicate<mode>): Delete (pattern is redundant).
24013 2021-04-08 Xionghu Luo <luoxhu@linux.ibm.com>
24015 * fold-const.c (fold_single_bit_test): Fix typo.
24016 * print-rtl.c (print_rtx_insn_vec): Call print_rtl_single
24019 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
24021 PR tree-optimization/97513
24022 * tree-vect-slp.c (vect_add_slp_permutation): New function,
24024 (vectorizable_slp_permutation): ...here. Detect cases in which
24025 all VEC_PERM_EXPRs are guaranteed to have the same stepped
24026 permute vector and only generate one permute vector for that case.
24027 Extend that case to handle variable-length vectors.
24029 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
24031 PR tree-optimization/99873
24032 * tree-vect-slp.c (vect_slp_prefer_store_lanes_p): New function.
24033 (vect_build_slp_instance): Don't split store groups that could
24034 use IFN_STORE_LANES.
24036 2021-04-07 Jakub Jelinek <jakub@redhat.com>
24039 * varasm.c (output_constant_pool_contents): Don't strip name encoding
24040 from XSTR (desc->sym, 0) or from label before passing those to
24043 2021-04-07 Richard Biener <rguenther@suse.de>
24045 PR tree-optimization/99954
24046 * tree-loop-distribution.c: Include tree-affine.h.
24047 (generate_memcpy_builtin): Try using tree-affine to prove
24049 (loop_distribution::classify_builtin_ldst): Always classify
24052 2021-04-07 Richard Biener <rguenther@suse.de>
24054 PR tree-optimization/99947
24055 * tree-vect-loop.c (vectorizable_induction): Pre-allocate
24056 steps vector to avoid pushing elements from the reallocated
24059 2021-04-07 Richard Biener <rguenther@suse.de>
24061 * tree-ssa-sccvn.h (print_vn_reference_ops): Declare.
24062 * tree-ssa-pre.c (print_pre_expr): Factor out VN reference operand
24064 * tree-ssa-sccvn.c (print_vn_reference_ops): ... into this new
24066 (debug_vn_reference_ops): New.
24068 2021-04-07 Bin Cheng <bin.cheng@linux.alibaba.com>
24070 PR tree-optimization/98736
24071 * tree-loop-distribution.c
24072 * (loop_distribution::bb_top_order_init):
24073 Compute RPO with programing order preserved by calling function
24074 rev_post_order_and_mark_dfs_back_seme.
24076 2021-04-06 Vladimir N. Makarov <vmakarov@redhat.com>
24079 * lra-constraints.c (split_reg): Don't check paradoxical_subreg_p.
24080 * lra-lives.c (clear_sparseset_regnos, regnos_in_sparseset_p): New
24082 (process_bb_lives): Don't update biggest mode of hard reg for
24083 implicit in multi-register group. Use the new functions for
24084 updating dead_set and unused_set by register notes.
24086 2021-04-06 Xianmiao Qu <xianmiao_qu@c-sky.com>
24088 * config/csky/csky_pipeline_ck802.md : Use insn reservation name
24091 2021-04-06 H.J. Lu <hjl.tools@gmail.com>
24093 * config/i386/x86-tune-costs.h (skylake_memcpy): Updated.
24094 (skylake_memset): Likewise.
24095 (skylake_cost): Change CLEAR_RATIO to 17.
24096 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
24097 Replace m_CANNONLAKE, m_ICELAKE_CLIENT, m_ICELAKE_SERVER,
24098 m_TIGERLAKE and m_SAPPHIRERAPIDS with m_SKYLAKE and m_CORE_AVX512.
24100 2021-04-06 Richard Biener <rguenther@suse.de>
24102 PR tree-optimization/99880
24103 * tree-vect-loop.c (maybe_set_vectorized_backedge_value): Only
24104 set vectorized defs of relevant PHIs.
24106 2021-04-06 Richard Biener <rguenther@suse.de>
24108 PR tree-optimization/99924
24109 * tree-vect-slp.c (vect_bb_partition_graph_r): Do not mark
24110 nodes w/o scalar stmts as visited.
24112 2021-04-06 Alex Coplan <alex.coplan@arm.com>
24115 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Also use base
24116 PCS for [su]fix_optab.
24118 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
24120 * config/darwin.c (machopic_legitimize_pic_address): Check
24121 that the current pic register is one of the hard reg set
24122 before setting liveness.
24124 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
24126 * config/darwin.c (machopic_legitimize_pic_address): Fix
24127 whitespace, remove unused code.
24129 2021-04-03 Jakub Jelinek <jakub@redhat.com>
24131 PR tree-optimization/99882
24132 * gimple-ssa-store-merging.c (bswap_view_convert): Handle val with
24135 2021-04-03 Jakub Jelinek <jakub@redhat.com>
24137 PR rtl-optimization/99863
24138 * dse.c (replace_read): Drop regs_live argument. Instead of
24139 regs_live, use store_insn->fixed_regs_live if non-NULL,
24140 otherwise punt if insns sequence clobbers or sets any hard
24143 2021-04-03 Jakub Jelinek <jakub@redhat.com>
24146 * targhooks.h (default_print_patchable_function_entry_1): Declare.
24147 * targhooks.c (default_print_patchable_function_entry_1): New function,
24148 copied from default_print_patchable_function_entry with an added flags
24150 (default_print_patchable_function_entry): Rewritten into a small
24151 wrapper around default_print_patchable_function_entry_1.
24152 * config/rs6000/rs6000.c (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
24154 (rs6000_print_patchable_function_entry): New function.
24156 2021-04-02 Eric Botcazou <ebotcazou@adacore.com>
24158 * doc/invoke.texi (fdelete-dead-exceptions): Minor tweak.
24160 2021-04-01 Jason Merrill <jason@redhat.com>
24163 * common.opt: Document v15 and v16.
24165 2021-04-01 Richard Biener <rguenther@suse.de>
24167 PR tree-optimization/99863
24168 * gimplify.c (gimplify_init_constructor): Recompute vector
24171 2021-04-01 Jakub Jelinek <jakub@redhat.com>
24173 * doc/extend.texi (symver attribute): Fix up syntax errors
24176 2021-04-01 Jakub Jelinek <jakub@redhat.com>
24178 PR tree-optimization/96573
24179 * gimple-ssa-store-merging.c (init_symbolic_number): Handle
24180 also pointer types.
24182 2021-04-01 Richard Biener <rguenther@suse.de>
24184 PR tree-optimization/99856
24185 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Promote
24186 precision to vector element precision.
24188 2021-04-01 Martin Jambor <mjambor@suse.cz>
24190 PR tree-optimization/97009
24191 * tree-sra.c (access_or_its_child_written): New function.
24192 (propagate_subaccesses_from_rhs): Use it instead of a simple grp_write
24195 2021-03-31 Jan Hubicka <hubicka@ucw.cz>
24198 * cif-code.def (USES_COMDAT_LOCAL): Make CIF_FINAL_NORMAL.
24200 2021-03-31 Pat Haugen <pthaugen@linux.ibm.com>
24203 * config/rs6000/altivec.md (xxspltiw_v4si, xxspltiw_v4sf_inst,
24204 xxspltidp_v2df_inst, xxsplti32dx_v4si_inst, xxsplti32dx_v4sf_inst,
24205 xxblend_<mode>, xxpermx_inst, xxeval): Mark prefixed.
24206 * config/rs6000/mma.md (mma_<vvi4i4i8>, mma_<avvi4i4i8>,
24207 mma_<vvi4i4i2>, mma_<avvi4i4i2>, mma_<vvi4i4>, mma_<avvi4i4>,
24208 mma_<pvi4i2>, mma_<apvi4i2>, mma_<vvi4i4i4>, mma_<avvi4i4i4>):
24210 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Adjust test.
24211 * config/rs6000/rs6000.md (define_attr "maybe_prefixed"): New.
24212 (define_attr "prefixed"): Update initializer.
24214 2021-03-31 Jakub Jelinek <jakub@redhat.com>
24217 * dwarf2out.c (debug_ranges_dwo_section): New variable.
24218 (DW_RANGES_IDX_SKELETON): Define.
24219 (struct dw_ranges): Add begin_entry and end_entry members.
24220 (DEBUG_DWO_RNGLISTS_SECTION): Define.
24221 (add_ranges_num): Adjust r initializer for addition of *_entry
24223 (add_ranges_by_labels): For -gsplit-dwarf and force_direct,
24224 set idx to DW_RANGES_IDX_SKELETON.
24225 (use_distinct_base_address_for_range): New function.
24226 (index_rnglists): Don't set r->idx if it is equal to
24227 DW_RANGES_IDX_SKELETON. Initialize r->begin_entry and
24228 r->end_entry for -gsplit-dwarf if those will be needed by
24230 (output_rnglists): Add DWO argument. If true, switch to
24231 debug_ranges_dwo_section rather than debug_ranges_section.
24232 Adjust l1/l2 label indexes. Only output the offset table when
24233 dwo is true and don't include in there the skeleton range
24234 entry if present. For -gsplit-dwarf, skip ranges that belong
24235 to the other rnglists section. Change return type from void
24236 to bool and return true if there are any range entries for
24237 the other section. For dwarf_split_debug_info use
24238 DW_RLE_startx_endx, DW_RLE_startx_length and DW_RLE_base_addressx
24239 entries instead of DW_RLE_start_end, DW_RLE_start_length and
24240 DW_RLE_base_address. Use use_distinct_base_address_for_range.
24241 (init_sections_and_labels): Initialize debug_ranges_dwo_section
24242 if -gsplit-dwarf and DWARF >= 5. Adjust ranges_section_label
24243 and range_base_label indexes.
24244 (dwarf2out_finish): Call index_rnglists earlier before finalizing
24245 .debug_addr. Never emit DW_AT_rnglists_base attribute. For
24246 -gsplit-dwarf and DWARF >= 5 call output_rnglists up to twice
24247 with different dwo arguments.
24248 (dwarf2out_c_finalize): Clear debug_ranges_dwo_section.
24250 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
24252 PR tree-optimization/98268
24253 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Call
24254 recompute_tree_invariant_for_addr_expr after successfully
24255 folding a TARGET_MEM_REF that occurs inside an ADDR_EXPR.
24257 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
24259 PR tree-optimization/99726
24260 * tree-data-ref.c (create_intersect_range_checks_index): Bail
24261 out if there is more than one access function SCEV for the loop
24264 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
24266 PR rtl-optimization/97141
24267 PR rtl-optimization/98726
24268 * emit-rtl.c (valid_for_const_vector_p): Return true for
24270 * rtx-vector-builder.h (rtx_vector_builder::step): Return a
24271 poly_wide_int instead of a wide_int.
24272 (rtx_vector_builder::apply_set): Take a poly_wide_int instead
24274 * rtx-vector-builder.c (rtx_vector_builder::apply_set): Likewise.
24275 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Return
24276 false for CONST_VECTORs that cannot be forced to memory.
24277 * config/aarch64/aarch64-simd.md (mov<mode>): If a CONST_VECTOR
24278 is too complex to force to memory, build it up from individual
24281 2021-03-31 Jan Hubicka <jh@suse.cz>
24284 * cgraph.c (cgraph_node::release_body): Fix overactive check.
24286 2021-03-31 Christophe Lyon <christophe.lyon@linaro.org>
24289 * config/arm/vec-common.md (mul<mode>3): Disable on iwMMXT, expect
24292 2021-03-31 H.J. Lu <hjl.tools@gmail.com>
24294 * config/i386/i386-expand.c (expand_set_or_cpymem_via_rep):
24295 For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, don't convert QImode
24297 (decide_alg): For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, use
24298 "rep movsb/stosb" only for known sizes.
24299 * config/i386/i386-options.c (processor_cost_table): Use Ice
24300 Lake cost for Cannon Lake, Ice Lake, Tiger Lake, Sapphire
24301 Rapids and Alder Lake.
24302 * config/i386/i386.h (TARGET_PREFER_KNOWN_REP_MOVSB_STOSB): New.
24303 * config/i386/x86-tune-costs.h (icelake_memcpy): New.
24304 (icelake_memset): Likewise.
24305 (icelake_cost): Likewise.
24306 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
24309 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
24312 * config/aarch64/aarch64.c
24313 (aarch64_vectorize_preferred_vector_alignment): Query the size
24314 of the provided SVE vector; do not assume that all SVE vectors
24315 have the same size.
24317 2021-03-31 Jan Hubicka <jh@suse.cz>
24320 * cgraph.c (cgraph_node::release_body): Remove all callers and
24322 * cgraphclones.c (cgraph_node::materialize_clone): Do not do it here.
24323 * cgraphunit.c (cgraph_node::expand): And here.
24325 2021-03-31 Martin Liska <mliska@suse.cz>
24327 * ipa-modref.c (analyze_ssa_name_flags): Fix coding style
24328 and one negated condition.
24330 2021-03-31 Jakub Jelinek <jakub@redhat.com>
24331 Richard Sandiford <richard.sandiford@arm.com>
24334 * config/aarch64/aarch64.md (*add<mode>3_poly_1): Swap Uai and Uav
24335 constraints on operands[2] and similarly 0 and rk constraints
24336 on operands[1] corresponding to that.
24338 2021-03-31 Jakub Jelinek <jakub@redhat.com>
24341 * configure.ac (HAVE_LD_BROKEN_PE_DWARF5): New AC_DEFINE if PECOFF
24342 linker doesn't support DWARF sections new in DWARF5.
24343 * config/i386/i386-options.c (ix86_option_override_internal): Default
24344 to dwarf_version 4 if HAVE_LD_BROKEN_PE_DWARF5 for TARGET_PECOFF
24346 * config.in: Regenerated.
24347 * configure: Regenerated.
24349 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24352 * config/aarch64/aarch64.c (aarch64_analyze_loop_vinfo): Check for
24353 available issue_info before using it.
24355 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24358 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): Do not allow zero
24361 2021-03-30 Xionghu Luo <luoxhu@linux.ibm.com>
24364 * config/rs6000/altivec.md (altivec_lvsl_reg): Change to ...
24365 (altivec_lvsl_reg_<mode>): ... this.
24366 (altivec_lvsr_reg): Change to ...
24367 (altivec_lvsr_reg_<mode>): ... this.
24368 * config/rs6000/predicates.md (vec_set_index_operand): New.
24369 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
24370 Enable 32bit variable vec_insert for all TARGET_VSX.
24371 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
24372 Enable 32bit variable vec_insert for p9 and above.
24373 (rs6000_expand_vector_set_var_p8): Rename to ...
24374 (rs6000_expand_vector_set_var_p7): ... this.
24375 (rs6000_expand_vector_set): Use TARGET_VSX and adjust assert
24377 * config/rs6000/vector.md (vec_set<mode>): Use vec_set_index_operand.
24378 * config/rs6000/vsx.md (xl_len_r): Use gen_altivec_lvsl_reg_di and
24379 gen_altivec_lvsr_reg_di.
24381 2021-03-30 H.J. Lu <hjl.tools@gmail.com>
24384 * config/i386/ia32intrin.h (__rdtsc): Defined as macro.
24385 (__rdtscp): Likewise.
24387 2021-03-30 Tamar Christina <tamar.christina@arm.com>
24389 PR tree-optimization/99825
24390 * tree-vect-slp-patterns.c (vect_check_evenodd_blend):
24391 Reject non-mult 2 lanes.
24393 2021-03-30 Richard Earnshaw <rearnsha@arm.com>
24396 * config/arm/arm.c (arm_file_start): Fix emission of
24397 Tag_ABI_VFP_args attribute.
24399 2021-03-30 Richard Biener <rguenther@suse.de>
24401 PR tree-optimization/99824
24402 * stor-layout.c (set_min_and_max_values_for_integral_type):
24403 Assert the precision is within the bounds of
24404 WIDE_INT_MAX_PRECISION.
24405 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
24406 the outermost component ref only to lower the access size
24407 and initialize that from the access type.
24409 2021-03-30 Richard Sandiford <richard.sandiford@arm.com>
24412 * config/aarch64/aarch64.md (mov<mode>): Pass multi-instruction
24413 CONST_INTs to aarch64_expand_mov_immediate when called after RA.
24415 2021-03-30 Mihailo Stojanovic <mihailo.stojanovic@typhoon-hil.com>
24417 * config/aarch64/aarch64.md
24418 (<optab>_trunc<fcvt_target><GPI:mode>2): Set the "arch"
24419 attribute to disambiguate between SIMD and FP variants of the
24422 2021-03-29 Jan Hubicka <hubicka@ucw.cz>
24424 * ipa-modref.c (merge_call_lhs_flags): Correct handling of deref.
24425 (analyze_ssa_name_flags): Fix typo in comment.
24427 2021-03-29 Alex Coplan <alex.coplan@arm.com>
24430 * config/aarch64/aarch64-sve-builtins.cc
24431 (function_builder::add_function): Add placeholder_p argument, use
24432 placeholder decls if this is set.
24433 (function_builder::add_unique_function): Instead of conditionally adding
24434 direct overloads, unconditionally add either a direct overload or a
24436 (function_builder::add_overloaded_function): Set placeholder_p if we're
24437 using C++ overloads. Use the obstack for string storage instead
24438 of relying on the tree nodes.
24439 (function_builder::add_overloaded_functions): Don't return early for
24440 m_direct_overloads: we need to add placeholders.
24441 * config/aarch64/aarch64-sve-builtins.h
24442 (function_builder::add_function): Add placeholder_p argument.
24444 2021-03-29 Richard Biener <rguenther@suse.de>
24446 PR tree-optimization/99807
24447 * tree-vect-slp.c (vect_slp_analyze_node_operations_1): Move
24448 assert below VEC_PERM handling.
24450 2021-03-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24453 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>): Use
24454 aarch64_simd_or_scalar_imm_zero to match zeroes. Remove pattern
24455 matching const_int 0.
24456 (move_lo_quad_internal_be_<mode>): Likewise.
24457 (move_lo_quad_<mode>): Update for the above.
24458 * config/aarch64/iterators.md (VQ_2E): Delete.
24460 2021-03-29 Jakub Jelinek <jakub@redhat.com>
24462 PR tree-optimization/99777
24463 * fold-const.c (extract_muldiv_1): For conversions, punt on casts from
24464 types other than scalar integral types.
24466 2021-03-28 David Edelsohn <dje.gcc@gmail.com>
24468 * config/rs6000/rs6000.c (rs6000_output_dwarf_dtprel): Do not add
24469 XCOFF TLS reloc decorations.
24471 2021-03-28 Gerald Pfeifer <gerald@pfeifer.com>
24473 * doc/analyzer.texi (Analyzer Internals): Update link to
24474 "A Memory Model for Static Analysis of C Programs".
24476 2021-03-26 David Edelsohn <dje.gcc@gmail.com>
24478 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Call function.
24479 * config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align):
24481 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align): New.
24482 (rs6000_special_round_type_align): Recursively check innermost first
24485 2021-03-26 Jakub Jelinek <jakub@redhat.com>
24488 * dwarf2out.h (struct dw_fde_node): Add rule18 member.
24489 * dwarf2cfi.c (dwarf2out_frame_debug_expr): When handling (set hfp sp)
24490 assignment with drap_reg active, queue reg save for hfp with offset 0
24491 and flush queued reg saves. When handling a push with rule18,
24492 defer queueing reg save for hfp and just assert the offset is 0.
24493 (scan_trace): Assert that fde->rule18 is false.
24495 2021-03-26 Vladimir Makarov <vmakarov@redhat.com>
24498 * ira-costs.c (record_reg_classes): Put case with
24499 CT_RELAXED_MEMORY adjacent to one with CT_MEMORY.
24500 * ira.c (ira_setup_alts): Ditto.
24501 * lra-constraints.c (process_alt_operands): Ditto.
24502 * recog.c (asm_operand_ok): Ditto.
24503 * reload.c (find_reloads): Ditto.
24505 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24507 * config/aarch64/aarch64-protos.h
24508 (cpu_addrcost_table::post_modify_ld3_st3): New member variable.
24509 (cpu_addrcost_table::post_modify_ld4_st4): Likewise.
24510 * config/aarch64/aarch64.c (generic_addrcost_table): Update
24511 accordingly, using the same costs as for post_modify.
24512 (exynosm1_addrcost_table, xgene1_addrcost_table): Likewise.
24513 (thunderx2t99_addrcost_table, thunderx3t110_addrcost_table):
24514 (tsv110_addrcost_table, qdf24xx_addrcost_table): Likewise.
24515 (a64fx_addrcost_table): Likewise.
24516 (neoversev1_addrcost_table): New.
24517 (neoversev1_tunings): Use neoversev1_addrcost_table.
24518 (aarch64_address_cost): Use the new post_modify costs for CImode
24521 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24523 * config/aarch64/aarch64.opt
24524 (-param=aarch64-loop-vect-issue-rate-niters=): New parameter.
24525 * doc/invoke.texi: Document it.
24526 * config/aarch64/aarch64-protos.h (aarch64_base_vec_issue_info)
24527 (aarch64_scalar_vec_issue_info, aarch64_simd_vec_issue_info)
24528 (aarch64_advsimd_vec_issue_info, aarch64_sve_vec_issue_info)
24529 (aarch64_vec_issue_info): New structures.
24530 (cpu_vector_cost): Write comments above the variables rather
24532 (cpu_vector_cost::issue_info): New member variable.
24533 * config/aarch64/aarch64.c: Include gimple-pretty-print.h
24534 and tree-ssa-loop-niter.h.
24535 (generic_vector_cost, a64fx_vector_cost, qdf24xx_vector_cost)
24536 (thunderx_vector_cost, tsv110_vector_cost, cortexa57_vector_cost)
24537 (exynosm1_vector_cost, xgene1_vector_cost, thunderx2t99_vector_cost)
24538 (thunderx3t110_vector_cost): Initialize issue_info to null.
24539 (neoversev1_scalar_issue_info, neoversev1_advsimd_issue_info)
24540 (neoversev1_sve_issue_info, neoversev1_vec_issue_info): New structures.
24541 (neoversev1_vector_cost): Use them.
24542 (aarch64_vec_op_count, aarch64_sve_op_count): New structures.
24543 (aarch64_vector_costs::saw_sve_only_op): New member variable.
24544 (aarch64_vector_costs::num_vector_iterations): Likewise.
24545 (aarch64_vector_costs::scalar_ops): Likewise.
24546 (aarch64_vector_costs::advsimd_ops): Likewise.
24547 (aarch64_vector_costs::sve_ops): Likewise.
24548 (aarch64_vector_costs::seen_loads): Likewise.
24549 (aarch64_simd_vec_costs_for_flags): New function.
24550 (aarch64_analyze_loop_vinfo): Initialize num_vector_iterations.
24551 Count the number of predicate operations required by SVE WHILE
24553 (aarch64_comparison_type, aarch64_multiply_add_p): New functions.
24554 (aarch64_sve_only_stmt_p, aarch64_in_loop_reduction_latency): Likewise.
24555 (aarch64_count_ops): Likewise.
24556 (aarch64_add_stmt_cost): Record whether see an SVE operation
24557 that cannot currently be implementing using Advanced SIMD.
24558 Record issue information about the scalar, Advanced SIMD
24559 and (where relevant) SVE versions of a loop.
24560 (aarch64_vec_op_count::dump): New function.
24561 (aarch64_sve_op_count::dump): Likewise.
24562 (aarch64_estimate_min_cycles_per_iter): Likewise.
24563 (aarch64_adjust_body_cost): If issue information is available,
24564 try to compare the issue rates of the various loop implementations
24565 and increase or decrease the vector body cost accordingly.
24567 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24569 * config/aarch64/aarch64.c (aarch64_detect_vector_stmt_subtype):
24570 Assume a zero cost for induction phis.
24572 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24574 * config/aarch64/aarch64.c (aarch64_embedded_comparison_type): New
24576 (aarch64_adjust_stmt_cost): Add the costs of embedded scalar and
24577 vector comparisons.
24579 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24581 * config/aarch64/aarch64.c (aarch64_detect_scalar_stmt_subtype):
24583 (aarch64_add_stmt_cost): Call it.
24585 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24587 * config/aarch64/aarch64-tuning-flags.def (matched_vector_throughput):
24588 New tuning parameter.
24589 * config/aarch64/aarch64.c (neoversev1_tunings): Use it.
24590 (aarch64_estimated_sve_vq): New function.
24591 (aarch64_vector_costs::analyzed_vinfo): New member variable.
24592 (aarch64_vector_costs::is_loop): Likewise.
24593 (aarch64_vector_costs::unrolled_advsimd_niters): Likewise.
24594 (aarch64_vector_costs::unrolled_advsimd_stmts): Likewise.
24595 (aarch64_record_potential_advsimd_unrolling): New function.
24596 (aarch64_analyze_loop_vinfo, aarch64_analyze_bb_vinfo): Likewise.
24597 (aarch64_add_stmt_cost): Call aarch64_analyze_loop_vinfo or
24598 aarch64_analyze_bb_vinfo on the first use of a costs structure.
24599 Detect whether we're vectorizing a loop for SVE that might be
24600 completely unrolled if it used Advanced SIMD instead.
24601 (aarch64_adjust_body_cost_for_latency): New function.
24602 (aarch64_finish_cost): Call it.
24604 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24606 * config/aarch64/aarch64.c (aarch64_vector_costs): New structure.
24607 (aarch64_init_cost): New function.
24608 (aarch64_add_stmt_cost): Use aarch64_vector_costs instead of
24609 the default unsigned[3].
24610 (aarch64_finish_cost, aarch64_destroy_cost_data): New functions.
24611 (TARGET_VECTORIZE_INIT_COST): Override.
24612 (TARGET_VECTORIZE_FINISH_COST): Likewise.
24613 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
24615 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24617 * config/aarch64/aarch64.c (neoversev1_advsimd_vector_cost)
24618 (neoversev1_sve_vector_cost): New cost structures.
24619 (neoversev1_vector_cost): Likewise.
24620 (neoversev1_tunings): Use them. Enable use_new_vector_costs.
24622 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24624 * config/aarch64/aarch64-protos.h
24625 (sve_vec_cost::scatter_store_elt_cost): New member variable.
24626 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
24627 accordingly, taking the cost from the cost of a scalar_store.
24628 (a64fx_sve_vector_cost): Likewise.
24629 (aarch64_detect_vector_stmt_subtype): Detect scatter stores.
24631 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24633 * config/aarch64/aarch64-protos.h
24634 (simd_vec_cost::store_elt_extra_cost): New member variable.
24635 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
24636 accordingly, using the vec_to_scalar cost for the new field.
24637 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
24638 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
24639 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
24640 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
24641 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
24642 (thunderx3t110_advsimd_vector_cost): Likewise.
24643 (aarch64_detect_vector_stmt_subtype): Detect single-element stores.
24645 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24647 * config/aarch64/aarch64-protos.h (simd_vec_cost::ld2_st2_permute_cost)
24648 (simd_vec_cost::ld3_st3_permute_cost): New member variables.
24649 (simd_vec_cost::ld4_st4_permute_cost): Likewise.
24650 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
24651 accordingly, using zero for the new costs.
24652 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
24653 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
24654 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
24655 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
24656 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
24657 (thunderx3t110_advsimd_vector_cost): Likewise.
24658 (aarch64_ld234_st234_vectors): New function.
24659 (aarch64_adjust_stmt_cost): Likewise.
24660 (aarch64_add_stmt_cost): Call aarch64_adjust_stmt_cost if using
24661 the new vector costs.
24663 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24665 * config/aarch64/aarch64-protos.h (sve_vec_cost): Turn into a
24666 derived class of simd_vec_cost. Add information about CLAST[AB]
24667 and FADDA instructions.
24668 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
24669 accordingly, using the vec_to_scalar costs for the new fields.
24670 (a64fx_sve_vector_cost): Likewise.
24671 (aarch64_reduc_type): New function.
24672 (aarch64_sve_in_loop_reduction_latency): Likewise.
24673 (aarch64_detect_vector_stmt_subtype): Take a vinfo parameter.
24674 Use aarch64_sve_in_loop_reduction_latency to handle SVE reductions
24675 that occur in the loop body.
24676 (aarch64_add_stmt_cost): Update call accordingly.
24678 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
24680 * config/aarch64/aarch64-tuning-flags.def (use_new_vector_costs):
24682 * config/aarch64/aarch64-protos.h (simd_vec_cost): Put comments
24683 above the fields rather than to the right.
24684 (simd_vec_cost::reduc_i8_cost): New member variable.
24685 (simd_vec_cost::reduc_i16_cost): Likewise.
24686 (simd_vec_cost::reduc_i32_cost): Likewise.
24687 (simd_vec_cost::reduc_i64_cost): Likewise.
24688 (simd_vec_cost::reduc_f16_cost): Likewise.
24689 (simd_vec_cost::reduc_f32_cost): Likewise.
24690 (simd_vec_cost::reduc_f64_cost): Likewise.
24691 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
24692 accordingly, using the vec_to_scalar_cost for the new fields.
24693 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
24694 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
24695 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
24696 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
24697 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
24698 (thunderx3t110_advsimd_vector_cost): Likewise.
24699 (aarch64_use_new_vector_costs_p): New function.
24700 (aarch64_simd_vec_costs): New function, split out from...
24701 (aarch64_builtin_vectorization_cost): ...here.
24702 (aarch64_is_reduction): New function.
24703 (aarch64_detect_vector_stmt_subtype): Likewise.
24704 (aarch64_add_stmt_cost): Call aarch64_detect_vector_stmt_subtype if
24705 using the new vector costs.
24707 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
24710 * tree-emutls.c (get_emutls_init_templ_addr): Mark initializer of weak
24711 TLS declarations as public.
24713 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
24715 * config/aarch64/aarch64-d.c (IN_TARGET_CODE): Define.
24716 * config/arm/arm-d.c (IN_TARGET_CODE): Likewise.
24717 * config/i386/i386-d.c (IN_TARGET_CODE): Likewise.
24718 * config/mips/mips-d.c (IN_TARGET_CODE): Likewise.
24719 * config/pa/pa-d.c (IN_TARGET_CODE): Likewise.
24720 * config/riscv/riscv-d.c (IN_TARGET_CODE): Likewise.
24721 * config/rs6000/rs6000-d.c (IN_TARGET_CODE): Likewise.
24722 * config/s390/s390-d.c (IN_TARGET_CODE): Likewise.
24723 * config/sparc/sparc-d.c (IN_TARGET_CODE): Likewise.
24725 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
24728 * config.gcc (*-*-cygwin*): Add winnt-d.o
24729 (*-*-mingw*): Likewise.
24730 * config/i386/cygwin.h (EXTRA_TARGET_D_OS_VERSIONS): New macro.
24731 * config/i386/mingw32.h (EXTRA_TARGET_D_OS_VERSIONS): Likewise.
24732 * config/i386/t-cygming: Add winnt-d.o.
24733 * config/i386/winnt-d.c: New file.
24735 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
24737 * config/freebsd-d.c: Include memmodel.h.
24739 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
24742 * config.gcc (*-*-openbsd*): Add openbsd-d.o.
24743 * config/t-openbsd: Add openbsd-d.o.
24744 * config/openbsd-d.c: New file.
24746 2021-03-25 Stam Markianos-Wright <stam.markianos-wright@arm.com>
24748 PR tree-optimization/96974
24749 * tree-vect-stmts.c (vect_get_vector_types_for_stmt): Replace assert
24750 with graceful exit.
24752 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
24755 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
24759 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
24760 always_inline in system headers.
24762 2021-03-25 Kewen Lin <linkw@linux.ibm.com>
24764 * tree-vect-loop.c (vect_model_reduction_cost): Init inside_cost.
24766 2021-03-25 Jakub Jelinek <jakub@redhat.com>
24769 * tree-core.h (enum operand_equal_flag): Add OEP_ADDRESS_OF_SAME_FIELD.
24770 * fold-const.c (operand_compare::operand_equal_p): Don't compare
24771 field offsets if OEP_ADDRESS_OF_SAME_FIELD.
24773 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
24777 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
24778 always_inline in system headers.
24780 2021-03-25 Richard Biener <rguenther@suse.de>
24782 PR tree-optimization/99746
24783 * tree-vect-slp-patterns.c (complex_pattern::build): Do not mark
24784 the scalar stmt as patterned. Instead set up required things
24787 2021-03-25 Xionghu Luo <luoxhu@linux.ibm.com>
24789 * config/rs6000/rs6000.c (power8_costs): Change l2 cache
24792 2021-03-24 Martin Liska <mliska@suse.cz>
24795 * common/config/i386/i386-common.c (ARRAY_SIZE): Fix off-by-one
24797 * config/i386/i386-options.c (ix86_option_override_internal):
24798 Add run-time assert.
24800 2021-03-24 Martin Jambor <mjambor@suse.cz>
24803 * ipa-cp.c (initialize_node_lattices): Mark as bottom all
24804 parameters with unknown type.
24805 (ipacp_value_safe_for_type): New function.
24806 (propagate_vals_across_arith_jfunc): Verify that the constant type
24807 can be used for a type of the formal parameter.
24808 (propagate_vals_across_ancestor): Likewise.
24809 (propagate_scalar_across_jump_function): Likewise. Pass the type
24810 also to propagate_vals_across_ancestor.
24812 2021-03-24 Christophe Lyon <christophe.lyon@linaro.org>
24815 * config/arm/mve.md (movmisalign<mode>_mve_store): Use Ux
24817 (movmisalign<mode>_mve_load): Likewise.
24819 2021-03-24 Jakub Jelinek <jakub@redhat.com>
24822 * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
24823 movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.
24825 2021-03-24 Alexandre Oliva <oliva@adacore.com>
24827 * doc/sourcebuild.texi (sysconf): New effective target.
24829 2021-03-24 Alexandre Oliva <oliva@adacore.com>
24831 * config/i386/predicates.md (reg_or_const_vec_operand): New.
24832 * config/i386/sse.md (ssse3_pshufbv8qi3): Add an expander for
24833 the now *-prefixed insn_and_split, turn the splitter const vec
24834 into an input for the insn, making it an ignored immediate for
24835 non-split cases, and loaded into the scratch register
24838 2021-03-23 Vladimir N. Makarov <vmakarov@redhat.com>
24841 * config/aarch64/constraints.md (Utq, UOb, UOh, UOw, UOd, UOty):
24842 Use define_relaxed_memory_constraint for them.
24844 2021-03-23 Iain Sandoe <iain@sandoe.co.uk>
24847 * config/host-darwin.c (darwin_gt_pch_use_address): Add a
24848 colon to the diagnostic message.
24850 2021-03-23 Ilya Leoshkevich <iii@linux.ibm.com>
24852 * fwprop.c (fwprop_propagation::fwprop_propagation): Look at
24854 (try_fwprop_subst_note): Use set_info instead of insn_info.
24855 (try_fwprop_subst_pattern): Likewise.
24856 (try_fwprop_subst_notes): Likewise.
24857 (try_fwprop_subst): Likewise.
24858 (forward_propagate_subreg): Likewise.
24859 (forward_propagate_and_simplify): Likewise.
24860 (forward_propagate_into): Likewise.
24861 * rtl-ssa/accesses.h (set_info::single_nondebug_use) New
24863 (set_info::single_nondebug_insn_use): Likewise.
24864 (set_info::single_phi_use): Likewise.
24865 * rtl-ssa/member-fns.inl (set_info::single_nondebug_use) New
24867 (set_info::single_nondebug_insn_use): Likewise.
24868 (set_info::single_phi_use): Likewise.
24870 2021-03-23 Christophe Lyon <christophe.lyon@linaro.org>
24872 * doc/sourcebuild.texi (arm_dsp_ok, arm_dsp): Document.
24874 2021-03-23 Jakub Jelinek <jakub@redhat.com>
24877 * config/aarch64/aarch64.c (aarch64_add_offset): Tell
24878 expand_mult to perform an unsigned rather than a signed
24881 2021-03-23 H.J. Lu <hjl.tools@gmail.com>
24884 * config/i386/cpuid.h (__cpuid): Add __volatile__.
24885 (__cpuid_count): Likewise.
24887 2021-03-23 Richard Biener <rguenther@suse.de>
24889 PR tree-optimization/99721
24890 * tree-vect-slp.c (vect_slp_analyze_node_operations):
24891 Make sure we can schedule the node.
24893 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
24895 * config/riscv/riscv.c (riscv_subword): Take endianness into
24896 account when calculating the byte offset.
24898 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
24900 * config/riscv/predicates.md (subreg_lowpart_operator): New predicate
24901 * config/riscv/riscv.md (*addsi3_extended2, *subsi3_extended2)
24902 (*negsi2_extended2, *mulsi3_extended2, *<optab>si3_mask)
24903 (*<optab>si3_mask_1, *<optab>di3_mask, *<optab>di3_mask_1)
24904 (*<optab>si3_extend_mask, *<optab>si3_extend_mask_1): Use
24905 new predicate "subreg_lowpart_operator"
24907 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
24909 * config/riscv/riscv.c (riscv_swap_instruction): New function
24910 to byteswap an SImode rtx containing an instruction.
24911 (riscv_trampoline_init): Byteswap the generated instructions
24914 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
24916 * common/config/riscv/riscv-common.c
24917 (TARGET_DEFAULT_TARGET_FLAGS): Set default endianness.
24918 * config.gcc (riscv32be-*, riscv64be-*): Set
24919 TARGET_BIG_ENDIAN_DEFAULT to 1.
24920 * config/riscv/elf.h (LINK_SPEC): Change -melf* value
24921 depending on default endianness.
24922 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
24923 * config/riscv/linux.h (LINK_SPEC): Likewise.
24924 * config/riscv/riscv.c (TARGET_DEFAULT_TARGET_FLAGS): Set
24925 default endianness.
24926 * config/riscv/riscv.h (DEFAULT_ENDIAN_SPEC): New macro.
24928 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
24930 * config/riscv/elf.h (LINK_SPEC): Pass linker endianness flag.
24931 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
24932 * config/riscv/linux.h (LINK_SPEC): Likewise.
24933 * config/riscv/riscv.h (ASM_SPEC): Pass -mbig-endian and
24935 (BYTES_BIG_ENDIAN): Handle big endian.
24936 (WORDS_BIG_ENDIAN): Define to BYTES_BIG_ENDIAN.
24937 * config/riscv/riscv.opt (-mbig-endian, -mlittle-endian): New
24939 * doc/invoke.texi (-mbig-endian, -mlittle-endian): Document.
24941 2021-03-23 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
24943 * regcprop.c (find_oldest_value_reg): Ask target whether
24944 different mode is fine for replacement register.
24946 2021-03-23 Aldy Hernandez <aldyh@redhat.com>
24948 PR tree-optimization/99296
24949 * value-range.cc (irange::irange_set_1bit_anti_range): New.
24950 (irange::irange_set_anti_range): Call irange_set_1bit_anti_range
24951 * value-range.h (irange::irange_set_1bit_anti_range): New.
24953 2021-03-22 Vladimir N. Makarov <vmakarov@redhat.com>
24956 * config/aarch64/constraints.md (UtQ): Use
24957 define_relaxed_memory_constraint for it.
24958 * doc/md.texi (define_relaxed_memory_constraint): Describe it.
24959 * genoutput.c (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
24960 * genpreds.c (constraint_data): Add bitfield is_relaxed_memory.
24961 (have_relaxed_memory_constraints): New static var.
24962 (relaxed_memory_start, relaxed_memory_end): Ditto.
24963 (add_constraint): Add arg is_relaxed_memory. Check name for
24964 relaxed memory. Set up is_relaxed_memory in constraint_data and
24965 have_relaxed_memory_constraints. Adjust calls.
24966 (choose_enum_order): Process relaxed memory.
24967 (write_tm_preds_h): Ditto.
24968 (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
24969 * gensupport.c (process_rtx): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
24970 * ira-costs.c (record_reg_classes): Process CT_RELAXED_MEMORY.
24971 * ira-lives.c (single_reg_class): Use
24972 insn_extra_relaxed_memory_constraint.
24973 * ira.c (ira_setup_alts): CT_RELAXED_MEMORY.
24974 * lra-constraints.c (valid_address_p): Use
24975 insn_extra_relaxed_memory_constraint instead of other memory
24977 (process_alt_operands): Process CT_RELAXED_MEMORY.
24978 (curr_insn_transform): Use insn_extra_relaxed_memory_constraint.
24979 * recog.c (asm_operand_ok, preprocess_constraints): Process
24981 * reload.c (find_reloads): Ditto.
24982 * rtl.def (DEFINE_RELAXED_MEMORY_CONSTRAINT): New.
24983 * stmt.c (parse_input_constraint): Use
24984 insn_extra_relaxed_memory_constraint.
24986 2021-03-22 Segher Boessenkool <segher@kernel.crashing.org>
24989 * ubsan.c (ubsan_instrument_float_cast): Don't test for unordered if
24992 2021-03-22 Alex Coplan <alex.coplan@arm.com>
24995 * config/arm/arm-protos.h (neon_make_constant): Add generate
24996 argument to guard emitting insns, default to true.
24997 * config/arm/arm.c (arm_legitimate_constant_p_1): Reject
24998 CONST_VECTORs which neon_make_constant can't handle.
24999 (neon_vdup_constant): Add generate argument, avoid emitting
25000 insns if it's not set.
25001 (neon_make_constant): Plumb new generate argument through.
25002 * config/arm/constraints.md (Ui): New. Use it...
25003 * config/arm/mve.md (*mve_mov<mode>): ... here.
25004 * config/arm/vec-common.md (movv8hf): Use neon_make_constant to
25005 synthesize constants.
25007 2021-03-22 Richard Biener <rguenther@suse.de>
25009 * debug.h: Add deprecation warning.
25011 2021-03-22 Richard Biener <rguenther@suse.de>
25013 PR tree-optimization/99694
25014 * tree-ssa-sccvn.c (visit_phi): Ignore edges with the
25017 2021-03-22 Kito Cheng <kito.cheng@sifive.com>
25020 * config/riscv/riscv.c (riscv_expand_block_move): Get RTL value
25021 after type checking.
25023 2021-03-22 Jakub Jelinek <jakub@redhat.com>
25027 * dwarf2out.c (get_full_len): Use get_precision rather than
25029 (add_const_value_attribute): Make sure add_AT_wide argument has
25030 precision prec rather than some very wide one.
25032 2021-03-22 Kewen Lin <linkw@linux.ibm.com>
25034 * config/rs6000/rs6000.md (*rotldi3_insert_sf,
25035 *mov<SFDF:mode><SFDF2:mode>cc_p9, floatsi<mode>2_lfiwax,
25036 floatsi<mode>2_lfiwax_mem, floatunssi<mode>2_lfiwzx,
25037 floatunssi<mode>2_lfiwzx_mem, *floatsidf2_internal,
25038 *floatunssidf2_internal, fix_trunc<mode>si2_stfiwx,
25039 fix_trunc<mode>si2_internal, fixuns_trunc<mode>si2_stfiwx,
25040 *round32<mode>2_fprs, *roundu32<mode>2_fprs,
25041 *fix_trunc<mode>si2_internal): Fix empty split condition.
25042 * config/rs6000/vsx.md (*vsx_le_undo_permute_<mode>,
25043 vsx_reduc_<VEC_reduc_name>_v2df, vsx_reduc_<VEC_reduc_name>_v4sf,
25044 *vsx_reduc_<VEC_reduc_name>_v2df_scalar,
25045 *vsx_reduc_<VEC_reduc_name>_v4sf_scalar): Likewise.
25047 2021-03-22 Xionghu Luo <luoxhu@linux.ibm.com>
25050 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
25051 Convert idx to DImode.
25052 (rs6000_expand_vector_set_var_p8): Likewise.
25054 2021-03-21 Jakub Jelinek <jakub@redhat.com>
25057 * dwarf2out.c (insert_float): Change return type from void to
25058 unsigned, handle GET_MODE_SIZE (mode) == 2 and return element size.
25059 (mem_loc_descriptor, loc_descriptor, add_const_value_attribute):
25062 2021-03-20 H.J. Lu <hjl.tools@gmail.com>
25065 * config/i386/i386.c (construct_container): Check cfun != NULL
25066 before accessing silent_p.
25068 2021-03-20 Ahamed Husni <ahamedhusni73@gmail.com>
25070 * asan.c: Fix typos in comments.
25072 2021-03-20 Vladimir N. Makarov <vmakarov@redhat.com>
25074 PR rtl-optimization/99680
25075 * lra-constraints.c (skip_contraint_modifiers): Rename to skip_constraint_modifiers.
25076 (process_address_1): Check empty constraint before using
25079 2021-03-19 Pat Haugen <pthaugen@linux.ibm.com>
25081 * config/rs6000/rs6000.c (power10_cost): New.
25082 (rs6000_option_override_internal): Set Power10 costs.
25083 (rs6000_issue_rate): Set Power10 issue rate.
25084 * config/rs6000/power10.md: Rewrite for Power10.
25086 2021-03-19 Vladimir N. Makarov <vmakarov@redhat.com>
25089 * lra-constraints.c (process_address_1): Don't use unknown
25090 constraint for address constraint.
25092 2021-03-19 Iain Sandoe <iain@sandoe.co.uk>
25095 * config.gcc (powerpc-*-darwin8): Delete the reference to
25096 the now removed darwin8.h.
25098 2021-03-19 Olivier Hainque <hainque@adacore.com>
25101 * config/vxworksae.h (VX_CPU_PREFIX): Define.
25103 2021-03-19 John David Anglin <danglin@gcc.gnu.org>
25105 * config/pa/pa.c (import_milli): Use memcpy instead of strncpy.
25107 2021-03-19 Tamar Christina <tamar.christina@arm.com>
25109 PR tree-optimization/99656
25110 * tree-vect-slp-patterns.c (linear_loads_p,
25111 complex_add_pattern::matches, is_eq_or_top,
25112 vect_validate_multiplication, complex_mul_pattern::matches,
25113 complex_fms_pattern::matches): Remove complex_perm_kinds_t.
25114 * tree-vectorizer.h: (complex_load_perm_t): Removed.
25115 (slp_tree_to_load_perm_map_t): Use complex_perm_kinds_t instead of
25116 complex_load_perm_t.
25118 2021-03-19 H.J. Lu <hjl.tools@gmail.com>
25121 * config/i386/i386-options.c (ix86_init_machine_status): Set
25123 * config/i386/i386.c (init_cumulative_args): Set silent_p to
25125 (construct_container): Return early for return and argument
25126 errors if silent_p is true.
25127 * config/i386/i386.h (machine_function): Add silent_p.
25129 2021-03-19 Jakub Jelinek <jakub@redhat.com>
25132 * config/arm/constraints.md (Ds): New constraint.
25133 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): Use w,Ds
25134 constraint instead of w,Dm.
25136 2021-03-19 Andrew Stubbs <ams@codesourcery.com>
25138 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Fix quotes
25141 2021-03-19 Eric Botcazou <ebotcazou@adacore.com>
25143 PR middle-end/99641
25144 * fold-const.c (native_encode_initializer) <CONSTRUCTOR>: For an
25145 array type, do the computation of the current position in sizetype.
25147 2021-03-18 Vladimir N. Makarov <vmakarov@redhat.com>
25150 * lra-constraints.c (process_address_1): Use lookup_constraint
25151 only for a single constraint.
25153 2021-03-18 Martin Sebor <msebor@redhat.com>
25155 PR middle-end/99502
25156 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): Rename...
25157 (inbounds_memaccess_p): ...to this. Check the ending offset of
25158 the accessed member.
25160 2021-03-18 Andrew Stubbs <ams@codesourcery.com>
25162 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Add %< and
25163 %> quote markers to error messages.
25164 (gcn_goacc_validate_dims): Likewise.
25165 (gcn_conditional_register_usage): Remove exclaimation mark from error
25167 (gcn_vectorize_vec_perm_const): Ensure perm is fully uninitialized.
25169 2021-03-18 Jan Hubicka <hubicka@ucw.cz>
25171 * config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
25174 2021-03-18 Sinan Lin <sinan@isrc.iscas.ac.cn>
25175 Kito Cheng <kito.cheng@sifive.com>
25177 * config/riscv/riscv.c (riscv_block_move_straight): Change type
25178 to unsigned HOST_WIDE_INT for parameter and local variable with
25179 HOST_WIDE_INT type.
25180 (riscv_adjust_block_mem): Ditto.
25181 (riscv_block_move_loop): Ditto.
25182 (riscv_expand_block_move): Ditto.
25184 2021-03-18 Nick Clifton <nickc@redhat.com>
25186 * config/v850/v850.c (construct_restore_jr): Increase static
25188 (construct_save_jarl): Likewise.
25189 * config/v850/v850.h (DWARF2_DEBUGGING_INFO): Define.
25191 2021-03-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25193 * config/aarch64/aarch64.c (aarch64_adjust_generic_arch_tuning): Define.
25194 (aarch64_override_options_internal): Use it.
25195 (generic_tunings): Add AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS to
25198 2021-03-17 Sandra Loosemore <sandra@codesourcery.com>
25200 * config/nios2/nios2.c (nios2_custom_check_insns): Clean up
25201 error message format issues.
25202 (nios2_option_override): Likewise.
25203 (nios2_expand_fpu_builtin): Likewise.
25204 (nios2_init_custom_builtins): Adjust to avoid bogus strncpy
25205 truncation warning.
25206 (nios2_expand_custom_builtin): More error message format fixes.
25207 (nios2_expand_rdwrctl_builtin): Likewise.
25208 (nios2_expand_rdprs_builtin): Likewise.
25209 (nios2_expand_eni_builtin): Likewise.
25210 (nios2_expand_builtin): Likewise.
25211 (nios2_register_custom_code): Likewise.
25212 (nios2_valid_target_attribute_rec): Likewise.
25213 (nios2_add_insn_asm): Fix uninitialized variable warning.
25215 2021-03-17 Jan Hubicka <jh@suse.cz>
25217 * config/i386/x86-tune-costs.h (struct processor_costs): Update costs
25218 of gather to match reality.
25219 * config/i386/x86-tune.def (X86_TUNE_USE_GATHER): Enable for znver3.
25221 2021-03-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25223 * config/aarch64/aarch64-builtins.c (aarch64_expand_rng_builtin): Use EQ
25224 to compare against CC_REG rather than NE.
25226 2021-03-17 H.J. Lu <hjl.tools@gmail.com>
25229 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support
25230 inline assembly statements.
25231 (ix86_print_operand): Update 'P' handling for -fno-plt.
25233 2021-03-17 Tamar Christina <tamar.christina@arm.com>
25236 * config/aarch64/aarch64.c
25237 (aarch64_simd_clone_compute_vecsize_and_simdlen): Remove unused var.
25239 2021-03-16 Segher Boessenkool <segher@kernel.crashing.org>
25242 * config/rs6000/predicates.md (branch_comparison_operator): Allow
25243 ordered and unordered for CCFPmode, if flag_finite_math_only.
25245 2021-03-16 Jakub Jelinek <jakub@redhat.com>
25248 * config/i386/i386-expand.c (ix86_split_lea_for_addr): Emit a MULT
25249 rather than ASHIFT.
25250 * config/i386/i386.md (mult by 1248 into ashift): New splitter.
25252 2021-03-16 Martin Liska <mliska@suse.cz>
25255 * optc-save-gen.awk: Add flag_ipa_ra to exceptions for
25256 cl_optimization_compare function.
25258 2021-03-16 Ilya Leoshkevich <iii@linux.ibm.com>
25260 * config/s390/s390.c (f_constraint_p): Treat "fv" constraints
25263 2021-03-16 Jakub Jelinek <jakub@redhat.com>
25266 * config/i386/i386.h (struct machine_function): Add
25267 has_explicit_vzeroupper bitfield.
25268 * config/i386/i386-expand.c (ix86_expand_builtin): Set
25269 cfun->machine->has_explicit_vzeroupper when expanding
25270 IX86_BUILTIN_VZEROUPPER.
25271 * config/i386/i386-features.c (rest_of_handle_insert_vzeroupper):
25272 Do the mode switching only when TARGET_VZEROUPPER, expensive
25273 optimizations turned on and not optimizing for size.
25274 (pass_insert_vzeroupper::gate): Enable even when
25275 cfun->machine->has_explicit_vzeroupper is set.
25277 2021-03-16 Jakub Jelinek <jakub@redhat.com>
25280 * config/aarch64/aarch64.c
25281 (aarch64_simd_clone_compute_vecsize_and_simdlen): If not a function
25282 definition, walk TYPE_ARG_TYPES list if non-NULL for argument types
25283 instead of DECL_ARGUMENTS. Ignore types for uniform arguments.
25285 2021-03-15 Richard Biener <rguenther@suse.de>
25287 PR tree-optimization/98834
25288 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle missing
25289 subsetting by truncating the access size.
25291 2021-03-15 Jan Hubicka <hubicka@ucw.cz>
25293 * config/i386/i386-options.c (processor_cost_table): Add znver3_cost.
25294 * config/i386/x86-tune-costs.h (znver3_cost): New gobal variable; copy
25297 2021-03-15 Martin Liska <mliska@suse.cz>
25299 * spellcheck.c: Add missing comma in initialization.
25301 2021-03-14 Uroš Bizjak <ubizjak@gmail.com>
25303 * config/i386/sse.md (*vec_extract<mode>): Merge alternative 0 with
25304 alternative 2 and alternative 1 with alternative 3 using
25305 YW register constraint.
25306 (*vec_extract<PEXTR_MODE12:mode>_zext): Merge alternatives
25307 using YW register constraint.
25308 (*vec_extractv16qi_zext): Ditto.
25309 (*vec_extractv4si): Merge alternatives 4 and 5
25310 using Yw register constraint.
25311 (*ssse3_palignr<mode>_perm): Use Yw instead of v for alternative 3.
25313 2021-03-13 Martin Sebor <msebor@redhat.com>
25315 PR tree-optimization/99489
25316 * builtins.c (gimple_call_alloc_size): Fail gracefully when argument
25317 is not a call statement.
25319 2021-03-13 Jakub Jelinek <jakub@redhat.com>
25321 PR tree-optimization/99544
25322 * match.pd (X + (X << C) -> X * (1 + (1 << C))): Don't simplify
25323 if for vector types multiplication can't be done in type's mode.
25325 2021-03-12 Eric Botcazou <ebotcazou@adacore.com>
25328 * config/sparc/constraints.md (w): Rename to...
25329 (W): ... this and ditch previous implementation.
25330 * config/sparc/sparc.md (*movdi_insn_sp64): Replace W with m.
25331 (*movdf_insn_sp64): Likewise.
25332 (*mov<VM64:mode>_insn_sp64): Likewise.
25333 * config/sparc/sync.md (*atomic_compare_and_swap<mode>_1): Replace
25335 (atomic_compare_and_swap_leon3_1): Likewise.
25336 (*atomic_compare_and_swapdi_v8plus): Likewise.
25337 * config/sparc/sparc.c (memory_ok_for_ldd): Remove useless test on
25338 architecture and add missing address validity check during LRA.
25340 2021-03-12 Tobias Burnus <tobias@codesourcery.com>
25343 * gimplify.c (omp_add_variable): Handle NULL_TREE as size
25344 occuring for assumed-size arrays in use_device_{ptr,addr}.
25346 2021-03-12 Jakub Jelinek <jakub@redhat.com>
25349 * config/i386/constraints.md (YW): New internal constraint.
25350 * config/i386/sse.md (v_Yw): Add V4TI, V2TI, V1TI and TI cases.
25351 (*<sse2_avx2>_<insn><mode>3<mask_name>,
25352 *<sse2_avx2>_uavg<mode>3<mask_name>, *abs<mode>2,
25353 *<s>mul<mode>3_highpart<mask_name>): Use <v_Yw> instead of v in
25355 (<sse2_avx2>_psadbw): Use YW instead of v in constraints.
25356 (*avx2_pmaddwd, *sse2_pmaddwd, *<code>v8hi3, *<code>v16qi3,
25357 avx2_pmaddubsw256, ssse3_pmaddubsw128): Merge last two alternatives
25358 into one, use Yw instead of former x,v.
25359 (ashr<mode>3, <insn><mode>3): Use <v_Yw> instead of x in constraints of
25360 the last alternative.
25361 (<sse2_avx2>_packsswb<mask_name>, <sse2_avx2>_packssdw<mask_name>,
25362 <sse2_avx2>_packuswb<mask_name>, <sse4_1_avx2>_packusdw<mask_name>,
25363 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, <ssse3_avx2>_palignr<mode>,
25364 <ssse3_avx2>_pshufb<mode>3<mask_name>): Merge last two alternatives
25365 into one, use <v_Yw> instead of former x,v.
25366 (avx2_interleave_highv32qi<mask_name>,
25367 vec_interleave_highv16qi<mask_name>): Use Yw instead of v in
25368 constraints. Add && <mask_avx512bw_condition> to condition.
25369 (avx2_interleave_lowv32qi<mask_name>,
25370 vec_interleave_lowv16qi<mask_name>,
25371 avx2_interleave_highv16hi<mask_name>,
25372 vec_interleave_highv8hi<mask_name>,
25373 avx2_interleave_lowv16hi<mask_name>, vec_interleave_lowv8hi<mask_name>,
25374 avx2_pshuflw_1<mask_name>, sse2_pshuflw_1<mask_name>,
25375 avx2_pshufhw_1<mask_name>, sse2_pshufhw_1<mask_name>,
25376 avx2_<code>v16qiv16hi2<mask_name>, sse4_1_<code>v8qiv8hi2<mask_name>,
25377 *sse4_1_<code>v8qiv8hi2<mask_name>_1, <sse2_avx2>_<insn><mode>3): Use
25378 Yw instead of v in constraints.
25379 * config/i386/mmx.md (Yv_Yw): New define_mode_attr.
25380 (*mmx_<insn><mode>3, mmx_ashr<mode>3, mmx_<insn><mode>3): Use <Yv_Yw>
25381 instead of Yv in constraints.
25382 (*mmx_<insn><mode>3, *mmx_mulv4hi3, *mmx_smulv4hi3_highpart,
25383 *mmx_umulv4hi3_highpart, *mmx_pmaddwd, *mmx_<code>v4hi3,
25384 *mmx_<code>v8qi3, mmx_pack<s_trunsuffix>swb, mmx_packssdw,
25385 mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
25386 *mmx_uavgv8qi3, *mmx_uavgv4hi3, mmx_psadbw): Use Yw instead of Yv in
25388 (*mmx_pinsrw, *mmx_pinsrb, *mmx_pextrw, *mmx_pextrw_zext, *mmx_pextrb,
25389 *mmx_pextrb_zext): Use YW instead of Yv in constraints.
25390 (*mmx_eq<mode>3, mmx_gt<mode>3): Use x instead of Yv in constraints.
25391 (mmx_andnot<mode>3, *mmx_<code><mode>3): Split last alternative into
25392 two, one with just x, another isa avx512vl with v.
25394 2021-03-12 Martin Liska <mliska@suse.cz>
25396 * doc/invoke.texi: Add missing param documentation.
25398 2021-03-11 David Malcolm <dmalcolm@redhat.com>
25401 * Makefile.in (ANALYZER_OBJS): Add analyzer/feasible-graph.o and
25402 analyzer/trimmed-graph.o.
25403 * doc/analyzer.texi (Analyzer Paths): Rewrite description of
25404 feasibility checking to reflect new implementation.
25405 * doc/invoke.texi (-fdump-analyzer-feasibility): Document new
25407 * shortest-paths.h (shortest_paths::get_shortest_distance): New.
25409 2021-03-11 David Malcolm <dmalcolm@redhat.com>
25411 * digraph.cc (selftest::test_shortest_paths): Update
25412 shortest_paths init for new param. Add test of
25413 SPS_TO_GIVEN_TARGET.
25414 * shortest-paths.h (enum shortest_path_sense): New.
25415 (shortest_paths::shortest_paths): Add "sense" param.
25416 Update for renamings. Generalize to use "sense" param.
25417 (shortest_paths::get_shortest_path): Rename param.
25418 (shortest_paths::m_sense): New field.
25419 (shortest_paths::m_prev): Rename...
25420 (shortest_paths::m_best_edge): ...to this.
25421 (shortest_paths::get_shortest_path): Update for renamings.
25422 Conditionalize flipping of path on sense of traversal.
25424 2021-03-11 David Malcolm <dmalcolm@redhat.com>
25426 * digraph.cc (selftest::test_shortest_paths): Add test coverage
25427 for paths from B and C.
25428 * shortest-paths.h (shortest_paths::shortest_paths): Handle
25429 unreachable nodes, rather than asserting.
25431 2021-03-11 David Edelsohn <dje.gcc@gmail.com>
25434 * config/rs6000/rs6000.c (rs6000_xcoff_file_start): Don't create
25435 xcoff_tbss_section_name.
25436 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .lcomm.
25437 * xcoffout.c (xcoff_tbss_section_name): Delete.
25438 * xcoffout.h (xcoff_tbss_section_name): Delete.
25440 2021-03-11 Richard Biener <rguenther@suse.de>
25442 PR tree-optimization/99523
25443 * tree-cfg.c (dump_function_to_file): Dump SSA names
25444 w/o identifier to the decls section as well, not only those
25445 without a VAR_DECL.
25447 2021-03-11 Jakub Jelinek <jakub@redhat.com>
25450 * ipa-icf-gimple.c (func_checker::compare_gimple_call): For internal
25451 function calls with lhs fail if the lhs don't have compatible types.
25453 2021-03-11 Hans-Peter Nilsson <hp@axis.com>
25455 * config/cris/cris.h (HARD_FRAME_POINTER_REGNUM): Define.
25456 Change FRAME_POINTER_REGNUM to correspond to a new faked
25457 register faked_fp, part of GENNONACR_REGS like faked_ap.
25458 (CRIS_FAKED_REGS_CONTENTS): New helper macro.
25459 (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS, CALL_USED_REGISTERS):
25460 (REG_ALLOC_ORDER, REG_CLASS_CONTENTS, REGNO_OK_FOR_BASE_P)
25461 (ELIMINABLE_REGS, REGISTER_NAMES): Adjust accordingly.
25462 * config/cris/cris.md (CRIS_FP_REGNUM): Renumber to new faked
25464 (CRIS_REAL_FP_REGNUM): New constant.
25465 * config/cris/cris.c (cris_reg_saved_in_regsave_area): Check
25466 for HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM.
25467 (cris_initial_elimination_offset): Handle elimination changes
25468 to HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM
25469 and add one from FRAME_POINTER_REGNUM to
25470 HARD_FRAME_POINTER_REGNUM.
25471 (cris_expand_prologue, cris_expand_epilogue): Emit code for
25472 hard_frame_pointer_rtx instead of frame_pointer_rtx.
25474 2021-03-10 David Edelsohn <dje.gcc@gmail.com>
25477 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Add check for DCmode.
25478 * config/rs6000/rs6000.c (rs6000_special_round_type_align): Same.
25480 2021-03-10 Vladimir N. Makarov <vmakarov@redhat.com>
25483 * lra-constraints.c (process_address_1): Don't check unknown
25484 constraint, use X for empty constraint.
25486 2021-03-10 Alex Coplan <alex.coplan@arm.com>
25488 * config/aarch64/aarch64.c (aarch64_vfp_is_call_or_return_candidate):
25489 Fix typo in comment describing "is_ha" argument.
25491 2021-03-10 John David Anglin <danglin@gcc.gnu.org>
25493 * doc/sourcebuild.texi: Document LRA target selector.
25495 2021-03-10 David Malcolm <dmalcolm@redhat.com>
25497 * doc/ux.texi: Add subsection contrasting interactive versus
25498 batch usage of GCC.
25500 2021-03-10 Joel Hutton <joel.hutton@arm.com>
25503 * tree-vect-stmts.c (vectorizable_store): Fix scatter store mask
25505 (vectorizable_load): Fix gather load mask check condition.
25507 2021-03-10 Richard Biener <rguenther@suse.de>
25509 PR tree-optimization/99510
25510 * tree.c (check_aligned_type): Check that the candidate
25511 has TYPE_USER_ALIGN set instead of matching with the
25514 2021-03-10 Eric Botcazou <ebotcazou@adacore.com>
25516 * config/sparc/sparc.c (sparc_regmode_natural_size): Return 4 for
25517 float and vector integer modes only if the mode is not larger.
25519 2021-03-10 Hans-Peter Nilsson <hp@axis.com>
25521 * config/cris/cris.h (DWARF_FRAME_REGISTERS): Define.
25523 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
25525 * ira.c (ira_setup_alts, ira_get_dup_out_num): Process digital
25527 * ira-lives.c (single_reg_class): Ditto.
25529 2021-03-09 Sebastian Huber <sebastian.huber@embedded-brains.de>
25531 * config.gcc (aarch64-*-rtems*): Include general rtems.h after
25532 the architecture-specific rtems.h.
25533 (aarch64-*-rtems*): Likewise.
25534 (arm*-*-rtems*): Likewise.
25535 (epiphany-*-rtems*): Likewise.
25536 (riscv*-*-rtems*): Likewise.
25538 2021-03-09 Jakub Jelinek <jakub@redhat.com>
25540 PR tree-optimization/99305
25541 * tree-ssa-phiopt.c (conditional_replacement): Test integer_pow2p
25542 before integer_all_onesp instead of vice versa.
25544 2021-03-09 Richard Earnshaw <rearnsha@arm.com>
25546 * common/config/arm/arm-common.c (arm_config_default): Change type
25547 of 'i' to unsigned.
25549 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
25552 * lra-constraints.c (process_address_1): Process constraint 'g'
25553 separately and digital constraints containing more one digit.
25555 2021-03-09 Nick Clifton <nickc@redhat.com>
25557 * config/rx/rx.h (DBX_DEBUGGING_INFO): Define.
25558 (DWARF"_DEBUGGING_INFO): Define.
25560 2021-03-09 Eric Botcazou <ebotcazou@adacore.com>
25563 * calls.c (initialize_argument_information): When the argument
25564 is passed by reference, do not make a copy in a thunk only if
25565 the argument is already in memory. Remove redundant test for
25566 the case of callee copy.
25568 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
25571 * lra-constraints.c (process_address_1): Process 0..9 constraints
25572 in process_address_1.
25574 2021-03-09 Andreas Krebbel <krebbel@linux.ibm.com>
25576 * config/s390/s390.c (struct s390_processor processor_table):
25577 Binutils name string must not be empty.
25579 2021-03-09 Claudiu Zissulescu <claziss@synopsys.com>
25581 * config/arc/arc.c (arc_attr_type): Remove function.
25583 2021-03-09 Martin Liska <mliska@suse.cz>
25586 * config/i386/i386-options.c (ix86_option_override_internal):
25587 Set isa_flags for OPTS argument and not for the global
25590 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
25592 * config/rs6000/predicates.md (ds_form_mem_operand): Check
25595 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
25598 * config/rs6000/predicates.md (ds_form_mem_operand) New
25600 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10) Use
25601 ds_form_mem_operand in ld/lwa patterns.
25602 * config/rs6000/fusion.md: Regenerate file.
25604 2021-03-08 Martin Sebor <msebor@redhat.com>
25606 PR middle-end/98266
25607 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): New function.
25608 (array_bounds_checker::check_array_bounds): Call it.
25610 2021-03-08 Martin Sebor <msebor@redhat.com>
25612 PR middle-end/97631
25613 * tree-ssa-strlen.c (maybe_warn_overflow): Test rawmem.
25614 (handle_builtin_stxncpy_strncat): Rename locals. Determine
25615 destination size from allocation calls. Issue a more appropriate
25617 (handle_builtin_memcpy): Pass true as rawmem to maybe_warn_overflow.
25618 (handle_builtin_memset): Same.
25620 2021-03-08 Peter Bergner <bergner@linux.ibm.com>
25623 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_permute): Add an assert
25624 to ensure we do not have an Altivec style address.
25625 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): Disable if passed
25626 an Altivec style address.
25627 (*vsx_le_perm_store_<mode>): Likewise.
25628 (splitters after *vsx_le_perm_store_<mode>): Likewise.
25629 (vsx_load_<mode>): Disable special expander if passed an Altivec
25631 (vsx_store_<mode>): Likewise.
25633 2021-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25636 * config/aarch64/predicates.md (aarch64_simd_shift_imm_vec_qi): Define.
25637 (aarch64_simd_shift_imm_vec_hi): Likewise.
25638 (aarch64_simd_shift_imm_vec_si): Likewise.
25639 (aarch64_simd_shift_imm_vec_di): Likewise.
25640 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le): Use
25641 predicate from above.
25642 (aarch64_shrn<mode>_insn_be): Likewise.
25643 (aarch64_rshrn<mode>_insn_le): Likewise.
25644 (aarch64_rshrn<mode>_insn_be): Likewise.
25645 (aarch64_shrn2<mode>_insn_le): Likewise.
25646 (aarch64_shrn2<mode>_insn_be): Likewise.
25647 (aarch64_rshrn2<mode>_insn_le): Likewise.
25648 (aarch64_rshrn2<mode>_insn_be): Likewise.
25650 2021-03-08 Vladimir N. Makarov <vmakarov@redhat.com>
25653 * lra-constraints.c (skip_contraint_modifiers): New function.
25654 (process_address_1): Use it before lookup_constraint call.
25656 2021-03-08 Martin Liska <mliska@suse.cz>
25659 * config/i386/i386-options.c (ix86_option_override_internal):
25660 Enable UINTR and HRESET for -march that supports it.
25662 2021-03-08 Ilya Leoshkevich <iii@linux.ibm.com>
25664 * config/s390/s390.c (f_constraint_p): New function.
25665 (s390_md_asm_adjust): Implement TARGET_MD_ASM_ADJUST.
25666 (TARGET_MD_ASM_ADJUST): Likewise.
25668 2021-03-08 Tobias Burnus <tobias@codesourcery.com>
25671 * tree-nested.c (convert_local_reference_stmt): Avoid calling
25672 lookup_field_for_decl for Fortran module (= namespace context).
25674 2021-03-08 Andreas Krebbel <krebbel@linux.ibm.com>
25676 * config/s390/s390.c (s390_expand_vec_compare): Implement <0
25677 comparison with arithmetic right shift.
25678 (s390_expand_vcond): No need for a force_reg anymore.
25679 s390_vec_compare will do it.
25680 * config/s390/vector.md ("vec_cmp<mode><tointvec>"): Accept also
25681 immediate operands.
25683 2021-03-07 Jakub Jelinek <jakub@redhat.com>
25686 * config/i386/constraints.md (Yw): Use SSE_REGS if TARGET_SSE
25687 but TARGET_AVX512BW or TARGET_AVX512VL is not set. Adjust description
25689 * config/i386/sse.md (v_Yw): New define_mode_attr.
25690 (*<insn><mode>3, *mul<mode>3<mask_name>, *avx2_<code><mode>3,
25691 *sse4_1_<code><mode>3<mask_name>): Use <v_Yw> instead of v
25693 * config/i386/mmx.md (mmx_pshufw_1, *vec_dupv4hi): Use Yw instead of
25694 xYw in constraints.
25696 2021-03-06 Julian Brown <julian@codesourcery.com>
25698 * tree-pretty-print.c (dump_generic_node): Emit non-generic
25699 address space info for aggregates.
25701 2021-03-06 Hans-Peter Nilsson <hp@axis.com>
25703 * config/cris/cris.h (MAX_FIXED_MODE_SIZE): Don't define.
25705 2021-03-05 Jakub Jelinek <jakub@redhat.com>
25707 PR middle-end/99322
25708 * tree-cfg.c (bb_to_omp_idx): New variable.
25709 (execute_build_cfg): Release the bb_to_omp_idx vector after
25710 cleanup_tree_cfg returns.
25711 (handle_abnormal_edges): Remove bb_to_omp_idx argument, adjust
25712 for bb_to_omp_idx being a vec<int> instead of pointer to array
25714 (make_edges): Remove bb_to_omp_idx local variable, don't pass
25715 it to handle_abnormal_edges, adjust for bb_to_omp_idx being a
25716 vec<int> instead of pointer to array of ints and don't free/release
25718 (remove_bb): When removing a bb and placing forced label somewhere
25719 else, ensure it is put into the same OpenMP region during cfg
25720 pass if possible or to entry successor as fallback. Unregister
25721 bb from bb_to_omp_idx.
25723 2021-03-05 Vladimir N. Makarov <vmakarov@redhat.com>
25726 * lra-constraints.c (process_address_1): Skip decomposing address
25727 for asm insn operand with unknown constraint.
25729 2021-03-05 Martin Jambor <mjambor@suse.cz>
25732 * cgraph.c (cgraph_edge::set_call_stmt): Do not update all
25733 corresponding speculative edges if we are about to resolve
25734 sepculation. Make edge direct (and so resolve speculations) before
25735 removing it from call_site_hash.
25736 (cgraph_edge::make_direct): Relax the initial assert to allow calling
25737 the function on speculative direct edges.
25739 2021-03-05 Eric Botcazou <ebotcazou@adacore.com>
25741 PR rtl-optimization/99376
25742 * rtlanal.c (nonzero_bits1) <arithmetic operators>: If the number
25743 of low-order zero bits is too large, set the result to 0 directly.
25745 2021-03-04 Jakub Jelinek <jakub@redhat.com>
25747 PR middle-end/93235
25748 * expmed.c (store_bit_field_using_insv): Return false of xop0 is a
25749 SUBREG and a SUBREG to op_mode can't be created.
25751 2021-03-04 Alex Coplan <alex.coplan@arm.com>
25754 * config/aarch64/aarch64-sve-builtins.cc
25755 (function_resolver::require_vector_type): Handle error_mark_node.
25757 2021-03-04 Ilya Leoshkevich <iii@linux.ibm.com>
25759 * cfgexpand.c (expand_asm_loc): Pass new parameter.
25760 (expand_asm_stmt): Likewise.
25761 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Add new
25763 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
25764 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
25765 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
25766 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
25767 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
25768 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
25769 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
25770 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
25771 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
25772 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
25773 * doc/tm.texi (md_asm_adjust): Likewise.
25774 * target.def (md_asm_adjust): Likewise.
25776 2021-03-04 Richard Biener <rguenther@suse.de>
25778 PR middle-end/97855
25779 * tree-pretty-print.c: Poison pp_printf.
25780 (dump_decl_name): Avoid use of pp_printf.
25781 (dump_block_node): Likewise.
25782 (dump_generic_node): Likewise.
25784 2021-03-04 Martin Sebor <msebor@redhat.com>
25786 PR middle-end/96963
25787 PR middle-end/94655
25788 * builtins.c (handle_array_ref): New helper.
25789 (handle_mem_ref): New helper.
25790 (compute_objsize_r): Factor out ARRAY_REF and MEM_REF handling
25791 into new helper functions. Correct a workaround for vectorized
25794 2021-03-03 Pat Haugen <pthaugen@linux.ibm.com>
25796 * config/rs6000/dfp.md (extendddtd2, trunctddd2, *cmp<mode>_internal1,
25797 floatditd2, ftrunc<mode>2, fix<mode>di2, dfp_ddedpd_<mode>,
25798 dfp_denbcd_<mode>, dfp_dxex_<mode>, dfp_diex_<mode>,
25799 *dfp_sgnfcnc_<mode>, dfp_dscli_<mode>, dfp_dscri_<mode>): Update size
25800 attribute for Power10.
25801 * config/rs6000/mma.md (*movoo): Likewise.
25802 * config/rs6000/rs6000.md (define_attr "size"): Add 256.
25803 (define_mode_attr bits): Add DD/TD modes.
25804 * config/rs6000/sync.md (load_quadpti, store_quadpti, load_lockedpti,
25805 store_conditionalpti): Update size attribute for Power10.
25807 2021-03-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
25810 * config/sparc/t-sparc (tree-ssanames.o-warn): Don't error for
25811 -Wuninitialized, -Wmaybe-uninitialized.
25812 (wide-int.o-warn): Likewise.
25814 2021-03-03 Richard Earnshaw <rearnsha@arm.com>
25816 * common/config/arm/arm-common.c: Include configargs.h.
25817 (arm_config_default): New function.
25818 (arm_target_mode): Renamed from arm_target_thumb_only. Handle
25819 processors that do not support Thumb. Take into account the
25820 --with-mode configuration setting for selecting the default.
25821 * config/arm/arm.h (OPTION_DEFAULT_SPECS): Remove entry for 'mode'.
25822 (TARGET_MODE_SPEC_FUNCTIONS): Update for function name change.
25824 2021-03-03 Martin Liska <mliska@suse.cz>
25826 PR gcov-profile/97461
25827 * gcov-io.h (GCOV_PREALLOCATED_KVP): Remove.
25829 2021-03-03 Eric Botcazou <ebotcazou@adacore.com>
25832 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
25833 point back the hard frame pointer to its default location when the
25834 frame is larger than SEH_MAX_FRAME_SIZE.
25836 2021-03-03 Jakub Jelinek <jakub@redhat.com>
25839 * config/i386/predicates.md (logic_operator): New define_predicate.
25840 * config/i386/i386.md (mov + mem using comm arith peephole2):
25841 Punt if operands[1] is EXT_REX_SSE_REGNO_P, AVX512BW is not enabled
25842 and the inner mode is [QH]Imode.
25844 2021-03-03 Jakub Jelinek <jakub@redhat.com>
25847 * dwarf2out.c (dw_loc_list_struct): Add end_entry member.
25848 (new_loc_list): Clear end_entry.
25849 (output_loc_list): Only use DW_LLE_startx_length for -gsplit-dwarf
25850 if HAVE_AS_LEB128, otherwise use DW_LLE_startx_endx. Fix comment
25852 (index_location_lists): For dwarf_version >= 5 without HAVE_AS_LEB128,
25853 initialize also end_entry.
25855 2021-03-03 Jakub Jelinek <jakub@redhat.com>
25858 * cfgrtl.c (fixup_partitions): When changing some bbs from hot to cold
25859 partitions, if in non-layout mode after reorder_blocks also move
25860 affected blocks to ensure a single partition transition.
25862 2021-03-03 Jason Merrill <jason@redhat.com>
25865 * cgraphunit.c (process_function_and_variable_attributes): Don't
25866 warn about flatten on an alias if the target also has it.
25867 * cgraph.h (symtab_node::get_alias_target_tree): New.
25869 2021-03-02 David Edelsohn <dje.gcc@gmail.com>
25871 * config/rs6000/rs6000.md (tls_get_tpointer_internal): Prepend
25872 period to symbol name.
25873 (tls_get_addr_internal<mode>): Same.
25875 2021-03-02 David Malcolm <dmalcolm@redhat.com>
25878 * diagnostic-show-locus.c
25879 (selftest::test_one_liner_many_fixits_2): Fix accidental usage of
25882 2021-03-02 Martin Sebor <msebor@redhat.com>
25884 PR middle-end/99276
25885 * builtins.c (warn_for_access): Remove stray warning text.
25887 2021-03-02 Martin Sebor <msebor@redhat.com>
25889 PR middle-end/99295
25890 * doc/extend.texi (attribute malloc): Reword and clarify nonaliasing
25893 2021-03-02 Jakub Jelinek <jakub@redhat.com>
25896 * dwarf2out.c (output_macinfo_op): Use DW_MACRO_*_str* even with
25897 -gdwarf-5 -gstrict-dwarf. For -gsplit-dwarf -gdwarf-5 use
25898 DW_MACRO_*_strx instead of DW_MACRO_*_strp. Handle
25899 DW_MACRO_define_strx and DW_MACRO_undef_strx.
25900 (save_macinfo_strings): Use DW_MACRO_*_str* even with
25901 -gdwarf-5 -gstrict-dwarf. Handle DW_MACRO_define_strx and
25902 DW_MACRO_undef_strx.
25904 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
25906 * config/s390/s390-builtin-types.def (BT_FN_V4SF_V8HI_UINT): New
25908 (BT_FN_V8HI_V8HI_UINT): Likewise.
25909 (BT_FN_V8HI_V4SF_V4SF_UINT): Likewise.
25910 * config/s390/s390-builtins.def (B_NNPA): New macro definition.
25911 (s390_vclfnhs, s390_vclfnls, s390_vcrnfs, s390_vcfn, s390_vcnf):
25912 New builtin definitions.
25913 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Bump
25914 vector extension version.
25915 * config/s390/s390.c (s390_expand_builtin): Check if builtins are
25916 available with current -march level.
25917 * config/s390/s390.md (UNSPEC_NNPA_VCLFNHS_V8HI)
25918 (UNSPEC_NNPA_VCLFNLS_V8HI, UNSPEC_NNPA_VCRNFS_V8HI)
25919 (UNSPEC_NNPA_VCFN_V8HI, UNSPEC_NNPA_VCNF_V8HI): New constants.
25920 * config/s390/vecintrin.h (vec_extend_to_fp32_hi): New macro.
25921 (vec_extend_to_fp32_lo): Likewise.
25922 (vec_round_from_fp32): Likewise.
25923 (vec_convert_to_fp16): Likewise.
25924 (vec_convert_from_fp16): Likewise.
25925 * config/s390/vx-builtins.md (vclfnhs_v8hi): New insn pattern.
25926 (vclfnls_v8hi): Likewise.
25927 (vcrnfs_v8hi): Likewise.
25928 (vcfn_v8hi): Likewise.
25929 (vcnf_v8hi): Likewise.
25931 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
25933 * common/config/s390/s390-common.c (processor_flags_table): New entry.
25934 * config.gcc: Enable arch14 for --with-arch and --with-tune.
25935 * config/s390/driver-native.c (s390_host_detect_local_cpu): Pick
25936 arch14 for unknown CPU models.
25937 * config/s390/s390-opts.h (enum processor_type): Add PROCESSOR_ARCH14.
25938 * config/s390/s390.c (s390_issue_rate): Add case for PROCESSOR_ARCH14.
25939 (s390_get_sched_attrmask): Likewise.
25940 (s390_get_unit_mask): Likewise.
25941 * config/s390/s390.h (enum processor_flags): Add PF_NNPA and PF_ARCH14.
25942 (TARGET_CPU_ARCH14, TARGET_CPU_ARCH14_P, TARGET_CPU_NNPA)
25943 (TARGET_CPU_NNPA_P, TARGET_ARCH14, TARGET_ARCH14_P, TARGET_NNPA)
25944 (TARGET_NNPA_P): New macro definitions.
25945 * config/s390/s390.md ("cpu_facility", "enabled"): Add arch14 and nnpa.
25946 * config/s390/s390.opt: Add PROCESSOR_ARCH14.
25948 2021-03-02 Jakub Jelinek <jakub@redhat.com>
25950 PR middle-end/95757
25951 * tree-vrp.c (register_edge_assert_for): Remove superfluous ()s around
25952 condition. Call register_edge_assert_for_1 for == 0, != 0, == 1 and
25953 != 1 comparisons if name is lhs of a comparison.
25955 2021-03-01 Iain Sandoe <iain@sandoe.co.uk>
25959 * config/darwin-protos.h (darwin_should_restore_cfa_state): New.
25960 * config/darwin.c (darwin_should_restore_cfa_state): New.
25961 * config/darwin.h (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New.
25962 * doc/tm.texi: Regenerated.
25963 * doc/tm.texi.in: Document TARGET_ASM_SHOULD_RESTORE_CFA_STATE.
25964 * dwarf2cfi.c (connect_traces): If the target requests, restore
25965 the CFA expression after a DW_CFA_restore.
25966 * target.def (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New hook.
25968 2021-03-01 Martin Liska <mliska@suse.cz>
25971 * optc-save-gen.awk: Add 4 more exceptions.
25973 2021-03-01 Nathan Sidwell <nathan@acm.org>
25976 * tree.h (TYPE_ALIGN_RAW): New accessor.
25977 (TYPE_ALIGN): Use it.
25979 2021-03-01 Jan Hubicka <jh@suse.cz>
25982 * ipa-fnsummary.c (compute_fn_summary): Fix sanity check.
25984 2021-03-01 Eric Botcazou <ebotcazou@adacore.com>
25987 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
25988 point the hard frame pointer to the SSE register save area instead
25989 of the general register save area. Perform only minimal adjustment
25990 for small frames if it is initially not correctly aligned.
25991 (ix86_expand_prologue): Remove early saves for a SEH target.
25992 * config/i386/winnt.c (struct seh_frame_state): Document constraint.
25994 2021-02-28 Jakub Jelinek <jakub@redhat.com>
25997 * ipa.c (symbol_table::remove_unreachable_nodes): Fix a comment
25998 typo - referneced -> referenced.
25999 * tree.c (component_ref_size): Fix comment typo -
26000 refernce -> reference.
26001 * tree-ssa-alias.c (access_path_may_continue_p): Fix comment typo -
26002 traling -> trailing.
26003 (aliasing_component_refs_p): Fix comment typos -
26004 refernce -> reference and refernece -> reference and
26005 traling -> trailing.
26006 (nonoverlapping_refs_since_match_p): Fix comment typo -
26007 referneces -> references.
26008 * doc/invoke.texi (--param modref-max-bases): Fix a typo -
26009 referneces -> references.
26011 2021-02-27 Iain Sandoe <iain@sandoe.co.uk>
26013 * config/host-darwin.c (darwin_gt_pch_use_address): Modify
26014 diagnostic message to avoid use of a contraction and format
26017 2021-02-27 Jakub Jelinek <jakub@redhat.com>
26020 * gcse.c (gcse_or_cprop_is_too_expensive): Use %wu instead of
26021 HOST_WIDE_INT_PRINT_UNSIGNED in warning format string.
26022 * ipa-devirt.c (ipa_odr_read_section): Use %wd instead of
26023 HOST_WIDE_INT_PRINT_DEC in inform format string. Fix comment
26026 2021-02-26 Richard Biener <rguenther@suse.de>
26028 PR middle-end/99281
26029 * expr.c (store_field): For calls with return-slot optimization
26030 and addressable return type expand the store directly.
26032 2021-02-26 Richard Biener <rguenther@suse.de>
26035 * builtins.c (warn_string_no_nul): Fix diagnostic formatting.
26037 2021-02-26 Peter Bergner <bergner@linux.ibm.com>
26040 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Replace assert
26043 2021-02-26 Aaron Sawdey <acsawdey@linux.ibm.com>
26045 * config.gcc: Add rs6000-pcrel-opt.o.
26046 * config/rs6000/rs6000-pcrel-opt.c: New file.
26047 * config/rs6000/pcrel-opt.md: New file.
26048 * config/rs6000/predicates.md: Add d_form_memory predicate.
26049 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_PCREL_OPT.
26050 * config/rs6000/rs6000-passes.def: Add pass_pcrel_opt.
26051 * config/rs6000/rs6000-protos.h: Add reg_to_non_prefixed(),
26052 pcrel_opt_valid_mem_p(), output_pcrel_opt_reloc(),
26053 and make_pass_pcrel_opt().
26054 * config/rs6000/rs6000.c (reg_to_non_prefixed): Make global.
26055 (rs6000_option_override_internal): Add pcrel-opt.
26056 (rs6000_delegitimize_address): Support pcrel-opt.
26057 (rs6000_opt_masks): Add pcrel-opt.
26058 (pcrel_opt_valid_mem_p): New function.
26059 (reg_to_non_prefixed): Make global.
26060 (rs6000_asm_output_opcode): Reset prepend_p_to_next_insn.
26061 (output_pcrel_opt_reloc): New function.
26062 * config/rs6000/rs6000.md (loads_extern_addr): New attr.
26063 (pcrel_extern_addr): Set loads_extern_addr.
26064 Add include for pcrel-opt.md.
26065 * config/rs6000/rs6000.opt: Add -mpcrel-opt.
26066 * config/rs6000/t-rs6000: Add rules for pcrel-opt.c and
26069 2021-02-26 YunQiang Su <yunqiang.su@cipunited.com>
26072 * config/mips/mips.c (mips_expand_ext_as_unaligned_load):
26073 If TARGET_64BIT and dest is SUBREG, we check the width, if it
26074 equal to SImode, we use SImode operation, just like what we are
26077 2021-02-26 Marek Polacek <polacek@redhat.com>
26079 * builtins.c (warn_for_access): Fix typos.
26081 2021-02-25 Iain Sandoe <iain@sandoe.co.uk>
26083 * config/aarch64/aarch64.md (<optab>_rol<mode>3): Add a '#'
26084 mark in front of the immediate quantity.
26085 (<optab>_rolsi3_uxtw): Likewise.
26087 2021-02-25 Richard Earnshaw <rearnsha@arm.com>
26090 * config/arm/thumb2.md (nonsecure_call_reg_thumb2_fpcxt): New pattern.
26091 (nonsecure_call_value_reg_thumb2_fpcxt): Likewise.
26092 (nonsecure_call_reg_thumb2): Restrict to using r4 for the callee
26093 address and disable when the FPCXT is not available.
26094 (nonsecure_call_value_reg_thumb2): Likewise.
26096 2021-02-25 Nathan Sidwell <nathan@acm.org>
26099 * doc/invoke.texi (flang-info-module-cmi): Renamed option.
26101 2021-02-25 Tamar Christina <tamar.christina@arm.com>
26103 * tree-vect-slp.c (optimize_load_redistribution_1): Abort on NULL nodes.
26105 2021-02-25 Richard Biener <rguenther@suse.de>
26107 PR tree-optimization/99253
26108 * tree-vect-loop.c (check_reduction_path): First compute
26109 code, then verify out-of-loop uses.
26111 2021-02-25 Jakub Jelinek <jakub@redhat.com>
26114 * match.pd ((T)(A) + CST -> (T)(A + CST)): Add :s to convert.
26116 2021-02-25 Jakub Jelinek <jakub@redhat.com>
26118 PR tree-optimization/80635
26119 * tree-vrp.c (vrp_simplify_cond_using_ranges): Also handle
26120 VIEW_CONVERT_EXPR if modes are the same, innerop is integral and
26121 has mode precision.
26123 2021-02-25 Richard Biener <rguenther@suse.de>
26125 * tree-vect-slp.c (optimize_load_redistribution_1): Delay
26126 load_map population.
26127 (vect_match_slp_patterns_2): Revert part of last change.
26128 (vect_analyze_slp): Do not interleave optimize_load_redistribution
26129 with pattern detection but do it afterwards. Dump the
26130 whole SLP graph after pattern recognition and load
26131 redistribution optimization finished.
26133 2021-02-24 Jakub Jelinek <jakub@redhat.com>
26136 * omp-low.c (struct omp_context): Add teams_nested_p and
26137 nonteams_nested_p members.
26138 (scan_omp_target): Diagnose teams nested inside of target with other
26139 directives strictly nested inside of the same target.
26140 (check_omp_nesting_restrictions): Set ctx->teams_nested_p or
26141 ctx->nonteams_nested_p as needed.
26143 2021-02-24 Vladimir N. Makarov <vmakarov@redhat.com>
26145 PR inline-asm/99123
26146 * lra-constraints.c (uses_hard_regs_p): Don't use decompose_mem_address.
26148 2021-02-24 Hans-Peter Nilsson <hp@axis.com>
26150 * config/cris/cris.c (cris_expand_prologue): Set
26151 current_function_static_stack_size, if flag_stack_usage_info.
26153 2021-02-24 Pat Haugen <pthaugen@linux.ibm.com>
26155 * config/rs6000/rs6000.c (next_insn_prefixed_p): Rename.
26156 (rs6000_final_prescan_insn): Adjust.
26157 (rs6000_asm_output_opcode): Likewise.
26159 2021-02-24 Martin Sebor <msebor@redhat.com>
26161 PR middle-end/97172
26162 * attribs.c (attr_access::free_lang_data): Clear attribute arg spec
26163 from function arguments.
26165 2021-02-24 Tamar Christina <tamar.christina@arm.com>
26167 PR tree-optimization/99220
26168 * tree-vect-slp.c (optimize_load_redistribution_1): Remove
26169 node from cache when it's about to be deleted.
26171 2021-02-24 Jakub Jelinek <jakub@redhat.com>
26173 PR tree-optimization/99225
26174 * fold-const.c (fold_binary_loc) <case NE_EXPR>: In (x & (1 << y)) != 0
26175 to ((x >> y) & 1) != 0 simplifications use build_one_cst instead of
26176 build_int_cst (..., 1). Formatting fixes.
26178 2021-02-24 Tamar Christina <tamar.christina@arm.com>
26180 PR tree-optimization/99149
26181 * tree-vect-slp-patterns.c (vect_detect_pair_op): Don't recreate the
26183 (vect_slp_reset_pattern): Remove.
26184 (complex_fma_pattern::matches): Remove call to vect_slp_reset_pattern.
26185 (complex_mul_pattern::build, complex_fma_pattern::build,
26186 complex_fms_pattern::build): Fix ref counts.
26187 * tree-vect-slp.c (vect_free_slp_tree): Undo SLP only pattern relevancy
26188 when node is being deleted.
26189 (vect_match_slp_patterns_2): Correct result of cache hit on patterns.
26190 (vect_schedule_slp): Invalidate SLP_TREE_REPRESENTATIVE of removed
26192 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize value.
26194 2021-02-24 Matthias Klose <doko@ubuntu.com>
26197 2020-12-07 Matthias Klose <doko@ubuntu.com>
26199 * genextract.c (print_header): Undefine ENABLE_RTL_CHECKING
26200 and ENABLE_RTL_FLAG_CHECKING.
26202 2021-02-24 Richard Biener <rguenther@suse.de>
26205 * builtins.c (fold_builtin_next_arg): Avoid NULL arg.
26207 2021-02-23 Peter Bergner <bergner@linux.ibm.com>
26209 * config/rs6000/mma.md (mma_assemble_pair): Rename from this...
26210 (vsx_assemble_pair): ...to this.
26211 (*mma_assemble_pair): Rename from this...
26212 (*vsx_assemble_pair): ...to this.
26213 (mma_disassemble_pair): Rename from this...
26214 (vsx_disassemble_pair): ...to this.
26215 (*mma_disassemble_pair): Rename from this...
26216 (*vsx_disassemble_pair): ...to this.
26217 * config/rs6000/rs6000-builtin.def (BU_MMA_V2, BU_MMA_V3,
26218 BU_COMPAT): New macros.
26219 (mma_assemble_pair): Rename from this...
26220 (vsx_assemble_pair): ...to this.
26221 (mma_disassemble_pair): Rename from this...
26222 (vsx_disassemble_pair): ...to this.
26223 (mma_assemble_pair): New compatibility built-in.
26224 (mma_disassemble_pair): Likewise.
26225 * config/rs6000/rs6000-call.c (struct builtin_compatibility): New.
26226 (RS6000_BUILTIN_COMPAT): Define.
26227 (bdesc_compat): New.
26228 (mma_expand_builtin): Use VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
26229 (rs6000_gimple_fold_mma_builtin): Use MMA_BUILTIN_DISASSEMBLE_PAIR
26230 and VSX_BUILTIN_ASSEMBLE_PAIR.
26231 (rs6000_init_builtins): Register compatibility built-ins.
26232 (mma_init_builtins): Use VSX_BUILTIN_ASSEMBLE_PAIR,
26233 VSX_BUILTIN_ASSEMBLE_PAIR_INTERNAL, VSX_BUILTIN_DISASSEMBLE_PAIR and
26234 VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
26235 * doc/extend.texi (__builtin_mma_assemble_pair): Rename from this...
26236 (__builtin_vsx_assemble_pair): ...to this.
26237 (__builtin_mma_disassemble_pair): Rename from this...
26238 (__builtin_vsx_disassemble_pair): ...to this.
26240 2021-02-23 Martin Liska <mliska@suse.cz>
26243 * ipa-icf.c (sem_variable::merge): Do not merge 2 variables
26244 with different alignment. That leads to an invalid red zone
26245 size allocated in runtime.
26247 2021-02-23 Jakub Jelinek <jakub@redhat.com>
26249 PR tree-optimization/99204
26250 * fold-const.c (fold_read_from_constant_string): Check that
26251 tree_fits_uhwi_p (index) rather than just that index is INTEGER_CST.
26253 2021-02-23 Segher Boessenkool <segher@kernel.crashing.org>
26254 Kewen Lin <linkw@gcc.gnu.org>
26256 * config/rs6000/rs6000.md (*rotl<mode>3_insert_3): Renamed to...
26257 (rotl<mode>3_insert_3): ...this.
26258 (plus_ior_xor): New code_iterator.
26259 (define_split for GPR rl*imi): New splitter.
26260 * config/rs6000/vsx.md (vsx_init_v4si): Use gen_rotldi3_insert_3
26261 for integer merging.
26263 2021-02-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26265 * config/aarch64/aarch64-tuning-flags.def (cse_sve_vl_constants):
26267 * config/aarch64/aarch64.md (add<mode>3): Force CONST_POLY_INT immediates
26268 into a register when the above is enabled.
26269 * config/aarch64/aarch64.c (neoversev1_tunings):
26270 AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
26271 (aarch64_rtx_costs): Use AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
26273 2021-02-22 Hans-Peter Nilsson <hp@axis.com>
26275 * config/cris/cris.c (cris_print_operand) <'T'>: Change
26276 valid operand from is now an addi mult-value to shift-value.
26277 * config/cris/cris.md (*addi): Change expression of scaled
26278 operand from mult to ashift.
26279 * config/cris/cris.md (*addi_reload): New insn_and_split.
26281 2021-02-22 John David Anglin <danglin@gcc.gnu.org>
26284 * config/pa/pa.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Define as
26285 hook_bool_const_tree_hwi_hwi_const_tree_true.
26286 (pa_asm_output_mi_thunk): Add support for nonzero vcall_offset.
26288 2021-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
26290 PR rtl-optimization/98791
26291 * ira-conflicts.c (process_regs_for_copy): Don't create allocno copies
26292 for unordered modes.
26294 2021-02-22 Martin Liska <mliska@suse.cz>
26296 * tree-inline.c (inline_forbidden_p): Set
26297 inline_forbidden_reason.
26299 2021-02-22 Richard Biener <rguenther@suse.de>
26301 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Dump
26304 2021-02-22 Richard Biener <rguenther@suse.de>
26306 PR tree-optimization/99165
26307 * gimple-ssa-store-merging.c (pass_store_merging::process_store):
26308 Accumulate changed to ret.
26310 2021-02-21 Uros Bizjak <ubizjak@gmail.com>
26313 2020-12-09 Uroš Bizjak <ubizjak@gmail.com>
26315 * config/i386/i386.h (REG_ALLOC_ORDER): Remove
26317 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
26320 * config/s390/vector.md (trunctf<DFP_ALL:mode>2_vr): New
26322 (trunctf<DFP_ALL:mode>2): Likewise.
26323 (trunctdtf2_vr): Likewise.
26324 (trunctdtf2): Likewise.
26325 (extend<DFP_ALL:mode>tf2_vr): Likewise.
26326 (extend<DFP_ALL:mode>tf2): Likewise.
26327 (extendtftd2_vr): Likewise.
26328 (extendtftd2): Likewise.
26330 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
26332 * config/s390/vector.md (*fprx2_to_tf): Rename to fprx2_to_tf,
26333 add memory alternative.
26334 (tf_to_fprx2): New pattern.
26336 2021-02-19 Martin Sebor <msebor@redhat.com>
26339 * attribs.c (init_attr_rdwr_indices): Guard vblist use.
26340 (attr_access::free_lang_data): Remove a spurious test.
26342 2021-02-19 Nathan Sidwell <nathan@acm.org>
26344 * doc/invoke.texi (flang-info-module-read): Document.
26346 2021-02-19 Martin Liska <mliska@suse.cz>
26348 PR translation/99167
26349 * params.opt: Fix typo.
26351 2021-02-19 Richard Biener <rguenther@suse.de>
26353 PR middle-end/99122
26354 * tree-inline.c (inline_forbidden_p): Do not inline functions
26355 with VLA arguments or return value.
26357 2021-02-19 Jakub Jelinek <jakub@redhat.com>
26360 * config/arm/arm.md (*stack_protect_combined_set_insn,
26361 *stack_protect_combined_test_insn): If force_const_mem result
26362 is not valid general operand, force its address into the destination
26365 2021-02-19 Jakub Jelinek <jakub@redhat.com>
26368 * tree-cfg.c (gimple_merge_blocks): If bb a starts with eh landing
26369 pad or non-local label, put FORCED_LABELs from bb b after that label
26370 rather than before it.
26372 2021-02-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
26375 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3): Use
26376 expand_vector_broadcast' to emit the vec_duplicate operand.
26378 2021-02-18 Vladimir N. Makarov <vmakarov@redhat.com>
26380 PR rtl-optimization/96264
26381 * lra-remat.c (reg_overlap_for_remat_p): Check also output insn
26384 2021-02-18 H.J. Lu <hjl.tools@gmail.com>
26387 * varasm.c (get_section): Replace SUPPORTS_SHF_GNU_RETAIN with
26388 looking up the retain attribute.
26389 (resolve_unique_section): Likewise.
26390 (get_variable_section): Likewise.
26391 (switch_to_section): Likewise. Warn when a symbol without the
26392 retain attribute and a symbol with the retain attribute are
26393 placed in the section with the same name, instead of the used
26395 * doc/extend.texi: Document the "retain" attribute.
26397 2021-02-18 Nathan Sidwell <nathan@acm.org>
26400 * doc/invoke.texi (flang-info-include-translate): Document header
26403 2021-02-18 Richard Biener <rguenther@suse.de>
26405 PR middle-end/99122
26406 * ipa-fnsummary.c (analyze_function_body): Set
26407 CIF_FUNCTION_NOT_INLINABLE for VLA parameter calls.
26408 * tree-inline.c (insert_init_debug_bind): Pass NULL for
26409 error_mark_node values.
26410 (force_value_to_type): Do not build V_C_Es for WITH_SIZE_EXPR
26412 (setup_one_parameter): Delay force_value_to_type until when
26415 2021-02-18 Hans-Peter Nilsson <hp@axis.com>
26417 PR tree-optimization/99142
26418 * match.pd (clz cmp 0): Gate replacement on single_use of clz result.
26420 2021-02-18 Jakub Jelinek <jakub@redhat.com>
26422 * wide-int-bitmask.h (wide_int_bitmask::wide_int_bitmask (),
26423 wide_int_bitmask::wide_int_bitmask (uint64_t),
26424 wide_int_bitmask::wide_int_bitmask (uint64_t, uint64_t),
26425 wide_int_bitmask::operator ~ () const,
26426 wide_int_bitmask::operator | (wide_int_bitmask) const,
26427 wide_int_bitmask::operator & (wide_int_bitmask) const): Use constexpr
26429 * config/i386/i386.h (PTA_3DNOW, PTA_3DNOW_A, PTA_64BIT, PTA_ABM,
26430 PTA_AES, PTA_AVX, PTA_BMI, PTA_CX16, PTA_F16C, PTA_FMA, PTA_FMA4,
26431 PTA_FSGSBASE, PTA_LWP, PTA_LZCNT, PTA_MMX, PTA_MOVBE, PTA_NO_SAHF,
26432 PTA_PCLMUL, PTA_POPCNT, PTA_PREFETCH_SSE, PTA_RDRND, PTA_SSE, PTA_SSE2,
26433 PTA_SSE3, PTA_SSE4_1, PTA_SSE4_2, PTA_SSE4A, PTA_SSSE3, PTA_TBM,
26434 PTA_XOP, PTA_AVX2, PTA_BMI2, PTA_RTM, PTA_HLE, PTA_PRFCHW, PTA_RDSEED,
26435 PTA_ADX, PTA_FXSR, PTA_XSAVE, PTA_XSAVEOPT, PTA_AVX512F, PTA_AVX512ER,
26436 PTA_AVX512PF, PTA_AVX512CD, PTA_NO_TUNE, PTA_SHA, PTA_PREFETCHWT1,
26437 PTA_CLFLUSHOPT, PTA_XSAVEC, PTA_XSAVES, PTA_AVX512DQ, PTA_AVX512BW,
26438 PTA_AVX512VL, PTA_AVX512IFMA, PTA_AVX512VBMI, PTA_CLWB, PTA_MWAITX,
26439 PTA_CLZERO, PTA_NO_80387, PTA_PKU, PTA_AVX5124VNNIW, PTA_AVX5124FMAPS,
26440 PTA_AVX512VPOPCNTDQ, PTA_SGX, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES,
26441 PTA_AVX512VBMI2, PTA_VPCLMULQDQ, PTA_AVX512BITALG, PTA_RDPID,
26442 PTA_PCONFIG, PTA_WBNOINVD, PTA_AVX512VP2INTERSECT, PTA_PTWRITE,
26443 PTA_AVX512BF16, PTA_WAITPKG, PTA_MOVDIRI, PTA_MOVDIR64B, PTA_ENQCMD,
26444 PTA_CLDEMOTE, PTA_SERIALIZE, PTA_TSXLDTRK, PTA_AMX_TILE, PTA_AMX_INT8,
26445 PTA_AMX_BF16, PTA_UINTR, PTA_HRESET, PTA_KL, PTA_WIDEKL, PTA_AVXVNNI,
26446 PTA_X86_64_BASELINE, PTA_X86_64_V2, PTA_X86_64_V3, PTA_X86_64_V4,
26447 PTA_CORE2, PTA_NEHALEM, PTA_WESTMERE, PTA_SANDYBRIDGE, PTA_IVYBRIDGE,
26448 PTA_HASWELL, PTA_BROADWELL, PTA_SKYLAKE, PTA_SKYLAKE_AVX512,
26449 PTA_CASCADELAKE, PTA_COOPERLAKE, PTA_CANNONLAKE, PTA_ICELAKE_CLIENT,
26450 PTA_ICELAKE_SERVER, PTA_TIGERLAKE, PTA_SAPPHIRERAPIDS, PTA_ALDERLAKE,
26451 PTA_KNL, PTA_BONNELL, PTA_SILVERMONT, PTA_GOLDMONT, PTA_GOLDMONT_PLUS,
26452 PTA_TREMONT, PTA_KNM): Use constexpr instead of const.
26454 2021-02-18 Jakub Jelinek <jakub@redhat.com>
26456 PR middle-end/99109
26457 * gimple-array-bounds.cc (build_zero_elt_array_type): Rename to ...
26458 (build_printable_array_type): ... this. Add nelts argument. For
26459 overaligned eltype, use TYPE_MAIN_VARIANT (eltype) instead. If
26460 nelts, call build_array_type_nelts.
26461 (array_bounds_checker::check_mem_ref): Use build_printable_array_type
26462 instead of build_zero_elt_array_type and build_array_type_nelts.
26464 2021-02-18 Jakub Jelinek <jakub@redhat.com>
26467 * config/i386/i386.c (distance_non_agu_define): Don't call
26468 extract_insn_cached here.
26469 (ix86_lea_outperforms): Save and restore recog_data around call
26470 to distance_non_agu_define and distance_agu_use.
26471 (ix86_ok_to_clobber_flags): Remove.
26472 (ix86_avoid_lea_for_add): Don't call ix86_ok_to_clobber_flags.
26473 (ix86_avoid_lea_for_addr): Likewise. Adjust function comment.
26474 * config/i386/i386.md (*lea<mode>): Change from define_insn_and_split
26475 into define_insn. Move the splitting to define_peephole2 and
26476 check there using peep2_regno_dead_p if FLAGS_REG is dead.
26478 2021-02-17 Julian Brown <julian@codesourcery.com>
26480 * gimplify.c (gimplify_scan_omp_clauses): Handle ATTACH_DETACH
26483 2021-02-17 Xi Ruoyao <xry111@mengyan1223.wang>
26486 * config/mips/mips.c (mips_symbol_insns): Do not use
26487 MSA_SUPPORTED_MODE_P if mode is MAX_MACHINE_MODE.
26489 2021-02-16 Vladimir N. Makarov <vmakarov@redhat.com>
26491 PR inline-asm/98096
26492 * stmt.c (resolve_operand_name_1): Take inout operands into account
26493 for access to labels by names.
26494 * doc/extend.texi: Describe counting operands for accessing labels.
26496 2021-02-16 Richard Biener <rguenther@suse.de>
26498 PR tree-optimization/38474
26499 * tree-ssa-structalias.c (variable_info::address_taken): New.
26500 (new_var_info): Initialize address_taken.
26501 (process_constraint): Set address_taken.
26502 (solve_constraints): Use the new address_taken flag rather
26503 than is_reg_var for sorting variables.
26504 (dump_constraint): Dump the variable number if the name
26507 2021-02-16 Jakub Jelinek <jakub@redhat.com>
26510 * tree-vect-stmts.c (vectorizable_simd_clone_call): For num_calls != 1
26511 multiply by 4096 and for inbranch by 8192.
26512 * config/i386/i386.c (ix86_simd_clone_usable): For TARGET_AVX512F,
26513 return 3, 2 or 1 for mangle letters 'b', 'c' or 'd'.
26515 2021-02-15 Maya Rashish <coypu@sdf.org>
26517 * config/aarch64/aarch64.c (aarch64_init_builtins):
26518 Call SUBTARGET_INIT_BUILTINS.
26520 2021-02-15 Peter Bergner <bergner@linux.ibm.com>
26522 PR rtl-optimization/98872
26523 * init-regs.c (initialize_uninitialized_regs): Skip initialization
26524 if CONST0_RTX is NULL.
26526 2021-02-15 Richard Sandiford <richard.sandiford@arm.com>
26528 PR rtl-optimization/98863
26529 * rtl-ssa/functions.h (function_info::bb_live_out_info): Delete.
26530 (function_info::build_info): Turn into a declaration, moving the
26531 definition to internals.h.
26532 (function_info::bb_walker): Declare.
26533 (function_info::create_reg_use): Likewise.
26534 (function_info::calculate_potential_phi_regs): Take a build_info
26536 (function_info::place_phis, function_info::create_ebbs): Declare.
26537 (function_info::calculate_ebb_live_in_for_debug): Likewise.
26538 (function_info::populate_backedge_phis): Delete.
26539 (function_info::start_block, function_info::end_block): Declare.
26540 (function_info::populate_phi_inputs): Delete.
26541 (function_info::m_potential_phi_regs): Move information to build_info.
26542 * rtl-ssa/internals.h: New file.
26543 (function_info::bb_phi_info): New class.
26544 (function_info::build_info): Moved from functions.h.
26545 Add a constructor and destructor.
26546 (function_info::build_info::ebb_use): Delete.
26547 (function_info::build_info::ebb_def): Likewise.
26548 (function_info::build_info::bb_live_out): Likewise.
26549 (function_info::build_info::tmp_ebb_live_in_for_debug): New variable.
26550 (function_info::build_info::potential_phi_regs): Likewise.
26551 (function_info::build_info::potential_phi_regs_for_debug): Likewise.
26552 (function_info::build_info::ebb_def_regs): Likewise.
26553 (function_info::build_info::bb_phis): Likewise.
26554 (function_info::build_info::bb_mem_live_out): Likewise.
26555 (function_info::build_info::bb_to_rpo): Likewise.
26556 (function_info::build_info::def_stack): Likewise.
26557 (function_info::build_info::old_def_stack_limit): Likewise.
26558 * rtl-ssa/internals.inl (function_info::build_info::record_reg_def):
26559 Remove the regno argument. Push the previous definition onto the
26560 definition stack where necessary.
26561 * rtl-ssa/accesses.cc: Include internals.h.
26562 * rtl-ssa/changes.cc: Likewise.
26563 * rtl-ssa/blocks.cc: Likewise.
26564 (function_info::build_info::build_info): Define.
26565 (function_info::build_info::~build_info): Likewise.
26566 (function_info::bb_walker): New class.
26567 (function_info::bb_walker::bb_walker): Define.
26568 (function_info::add_live_out_use): Convert a logarithmic-complexity
26569 test into a linear one. Allow the same definition to be passed
26571 (function_info::calculate_potential_phi_regs): Moved from
26572 functions.cc. Take a build_info parameter and store the
26573 information there instead.
26574 (function_info::place_phis): New function.
26575 (function_info::add_entry_block_defs): Update call to record_reg_def.
26576 (function_info::calculate_ebb_live_in_for_debug): New function.
26577 (function_info::add_phi_nodes): Use bb_phis to decide which
26578 registers need phi nodes and initialize ebb_def_regs accordingly.
26579 Do not add degenerate phis here.
26580 (function_info::add_artificial_accesses): Use create_reg_use.
26581 Assert that all definitions are listed in the DF LR sets.
26582 Update call to record_reg_def.
26583 (function_info::record_block_live_out): Record live-out register
26584 values in the phis of successor blocks. Use the live-out set
26585 when processing the last block in an EBB, instead of always
26586 using the live-in sets of successor blocks. AND the live sets
26587 with the set of registers that have been defined in the EBB,
26588 rather than with all potential phi registers. Cope correctly
26589 with branches back to the start of the current EBB.
26590 (function_info::start_block): New function.
26591 (function_info::end_block): Likewise.
26592 (function_info::populate_phi_inputs): Likewise.
26593 (function_info::create_ebbs): Likewise.
26594 (function_info::process_all_blocks): Rewrite into a multi-phase
26596 * rtl-ssa/functions.cc: Include internals.h.
26597 (function_info::calculate_potential_phi_regs): Move to blocks.cc.
26598 (function_info::init_function_data): Remove caller.
26599 * rtl-ssa/insns.cc: Include internals.h
26600 (function_info::create_reg_use): New function. Lazily any
26601 degenerate phis needed by the linear RPO view.
26602 (function_info::record_use): Use create_reg_use. When processing
26603 debug uses, use potential_phi_regs and test it before checking
26604 whether the register is live on entry to the current EBB. Lazily
26605 calculate ebb_live_in_for_debug.
26606 (function_info::record_call_clobbers): Update call to record_reg_def.
26607 (function_info::record_def): Likewise.
26609 2021-02-15 Martin Liska <mliska@suse.cz>
26611 * toplev.c (init_asm_output): Free output of
26612 gen_command_line_string function.
26613 (process_options): Likewise.
26615 2021-02-15 Martin Liska <mliska@suse.cz>
26617 * params.opt: Add 2 missing Param keywords.
26619 2021-02-15 Eric Botcazou <ebotcazou@adacore.com>
26621 * df-core.c (df_worklist_dataflow_doublequeue): Use proper cast.
26623 2021-02-15 Jakub Jelinek <jakub@redhat.com>
26625 PR tree-optimization/99079
26626 * match.pd (A % (pow2pcst << N) -> A & ((pow2pcst << N) - 1)): Remove
26627 useless tree_nop_conversion_p (type, TREE_TYPE (@3)) check. Instead
26628 require both type and TREE_TYPE (@1) to be integral types and either
26629 type having smaller or equal precision, or TREE_TYPE (@1) being
26630 unsigned type, or type being signed type. If TREE_TYPE (@1)
26631 doesn't have wrapping overflow, perform the subtraction of one in
26634 2021-02-14 Jan Hubicka <hubicka@ucw.cz>
26635 Richard Biener <rguether@suse.de>
26638 * ipa-reference.c (ipa_init): Only conditinally initialize
26639 reference_vars_to_consider.
26640 (propagate): Conditionally deninitialize reference_vars_to_consider.
26641 (ipa_reference_write_optimization_summary): Sanity check that
26642 reference_vars_to_consider is not allocated.
26644 2021-02-13 Levy Hsu <admin@levyhsu.com>
26647 * config/riscv/riscv-shorten-memrefs.c (pass_shorten_memrefs): Add
26648 extend parameter to get_si_mem_base_reg declaration.
26649 (get_si_mem_base_reg): Add extend parameter. Set it.
26650 (analyze): Pass extend arg to get_si_mem_base_reg.
26651 (transform): Likewise. Use it when rewriting mems.
26652 * config/riscv/riscv.c (riscv_legitimize_move): Check for subword
26653 loads and emit sign/zero extending load followed by subreg move.
26655 2021-02-13 Jim Wilson <jimw@sifive.com>
26658 * config/riscv/riscv.c (riscv_compressed_lw_address_p): Drop early
26659 exit when !reload_completed. Only perform check for compressed reg
26660 if reload_completed.
26661 (riscv_rtx_costs): In MEM case, when optimizing for size and
26662 shorten memrefs, if not compressible, then increase cost.
26664 2021-02-13 Jakub Jelinek <jakub@redhat.com>
26666 PR rtl-optimization/98439
26667 * recog.c (pass_split_before_regstack::gate): Enable even when
26668 pass_split_before_sched2 is enabled if -fselective-scheduling2 is
26671 2021-02-13 Jakub Jelinek <jakub@redhat.com>
26674 * config/i386/mmx.md (*mmx_pshufd_1): Add a combine splitter for
26675 swap of V2SImode elements in memory into DImode memory rotate by 32.
26677 2021-02-12 Martin Sebor <msebor@redhat.com>
26679 * tree-pretty-print.c (print_generic_expr_to_str): Update comment.
26681 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
26683 * rtl-ssa/accesses.cc (function_info::make_use_available): Use
26684 m_temp_obstack rather than m_obstack to allocate the temporary use.
26686 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
26688 * df-problems.c (df_lr_bb_local_compute): Treat partial definitions
26689 as read-modify operations.
26691 2021-02-12 Richard Biener <rguenther@suse.de>
26693 PR middle-end/38474
26694 * ipa-fnsummary.c (unmodified_parm_1): Only walk when
26695 fbi->aa_walk_budget is bigger than zero. Update
26696 fbi->aa_walk_budget.
26697 (param_change_prob): Likewise.
26698 * ipa-prop.c (detect_type_change_from_memory_writes):
26699 Properly account walk_aliased_vdefs.
26700 (parm_preserved_before_stmt_p): Canonicalize updates.
26701 (parm_ref_data_preserved_p): Likewise.
26702 (parm_ref_data_pass_through_p): Likewise.
26703 (determine_known_aggregate_parts): Account own alias queries.
26705 2021-02-12 Martin Liska <mliska@suse.cz>
26707 * opts-common.c (decode_cmdline_option): Release werror_arg.
26708 * opts.c (gen_producer_string): Release output of
26709 gen_command_line_string.
26711 2021-02-12 Richard Biener <rguenther@suse.de>
26713 PR tree-optimization/38474
26714 * params.opt (-param=max-store-chains-to-track=): New param.
26715 (-param=max-stores-to-track=): Likewise.
26716 * doc/invoke.texi (max-store-chains-to-track): Document.
26717 (max-stores-to-track): Likewise.
26718 * gimple-ssa-store-merging.c (pass_store_merging::m_n_chains):
26720 (pass_store_merging::m_n_stores): Likewise.
26721 (pass_store_merging::terminate_and_process_chain): Update
26722 m_n_stores and m_n_chains.
26723 (pass_store_merging::process_store): Likewise. Terminate
26724 oldest chains if the number of stores or chains get too large.
26725 (imm_store_chain_info::terminate_and_process_chain): Dump
26728 2021-02-11 Eric Botcazou <ebotcazou@adacore.com>
26730 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
26731 the cold section, emit a nop before the directive if the previous
26732 active instruction can throw.
26734 2021-02-11 Peter Bergner <bergner@linux.ibm.com>
26737 * config/rs6000/predicates.md (mma_assemble_input_operand): Restrict
26738 memory addresses that are legal for quad word accesses.
26740 2021-02-11 Andrea Corallo <andrea.corallo@arm.com>
26743 * config/arm/thumb2.md (*doloop_end_internal): Generate
26744 alternative sequence to handle long range branches.
26746 2021-02-11 Joel Hutton <joel.hutton@arm.com>
26748 PR tree-optimization/98772
26749 * optabs-tree.c (supportable_half_widening_operation): New function
26750 to check for supportable V8QI->V8HI widening patterns.
26751 * optabs-tree.h (supportable_half_widening_operation): New function.
26752 * tree-vect-stmts.c (vect_create_half_widening_stmts): New function
26753 to create promotion stmts for V8QI->V8HI widening patterns.
26754 (vectorizable_conversion): Add case for V8QI->V8HI.
26756 2021-02-11 Richard Biener <rguenther@suse.de>
26758 * sparseset.h (SPARSESET_ELT_BITS): Remove.
26759 (SPARSESET_ELT_TYPE): Use unsigned int.
26760 * fwprop.c: Do not include sparseset.h.
26762 2021-02-10 Jakub Jelinek <jakub@redhat.com>
26765 * varasm.c (declare_weak): For -fsyntax-only, allow even
26766 TREE_ASM_WRITTEN function decls.
26768 2021-02-10 Jakub Jelinek <jakub@redhat.com>
26771 * config/i386/sse.md (fix<fixunssuffix>_truncv2sfv2di2,
26772 <insn>v8qiv8hi2, <insn>v8qiv8si2, <insn>v4qiv4si2, <insn>v4hiv4si2,
26773 <insn>v8qiv8di2, <insn>v4qiv4di2, <insn>v2qiv2di2, <insn>v4hiv4di2,
26774 <insn>v2hiv2di2, <insn>v2siv2di2): Force operands[1] into REG before
26775 calling simplify_gen_subreg on it.
26777 2021-02-10 Martin Liska <mliska@suse.cz>
26779 * config/nvptx/nvptx.c (nvptx_option_override): Use
26780 flag_patchable_function_entry instead of the removed
26781 function_entry_patch_area_size.
26783 2021-02-10 Martin Liska <mliska@suse.cz>
26785 PR tree-optimization/99002
26786 PR tree-optimization/99026
26787 * gimple-if-to-switch.cc (if_chain::is_beneficial): Fix memory
26788 leak when adjacent cases are merged.
26789 * tree-switch-conversion.c (switch_decision_tree::analyze_switch_statement): Use
26791 (make_pass_lower_switch): Remove trailing whitespace.
26792 * tree-switch-conversion.h (release_clusters): New.
26794 2021-02-10 Richard Biener <rguenther@suse.de>
26796 PR rtl-optimization/99054
26797 * cfgrtl.c (rtl-optimization/99054): Return an auto_vec.
26798 (fixup_partitions): Adjust.
26799 (rtl_verify_edges): Likewise.
26801 2021-02-10 Jakub Jelinek <jakub@redhat.com>
26803 PR middle-end/99007
26804 * gimplify.c (gimplify_scan_omp_clauses): For MEM_REF on reductions,
26805 temporarily disable gimplify_ctxp->into_ssa around gimplify_expr
26808 2021-02-10 Richard Biener <rguenther@suse.de>
26811 * ipa-pure-const.c (propagate_malloc): Use an auto_vec<>
26814 2021-02-10 Richard Biener <rguenther@suse.de>
26816 PR tree-optimization/99024
26817 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Only
26818 clear loop->aux if it is associated with the destroyed loop_vinfo.
26820 2021-02-10 Martin Liska <mliska@suse.cz>
26822 PR tree-optimization/99002
26823 * gimple-if-to-switch.cc (find_conditions): Fix memory leak
26826 2021-02-10 Martin Liska <mliska@suse.cz>
26829 * ipa-icf.c (sem_item::add_reference): Fix memory leak when
26830 a reference exists.
26832 2021-02-10 Jakub Jelinek <jakub@redhat.com>
26835 * dwarf2out.c (prune_unused_types_walk): Mark DW_TAG_variable DIEs
26836 at class scope for DWARF5+.
26838 2021-02-09 Eric Botcazou <ebotcazou@adacore.com>
26840 PR rtl-optimization/96015
26841 * reorg.c (skip_consecutive_labels): Minor comment tweaks.
26842 (relax_delay_slots): When deleting a jump to the next active
26843 instruction over a barrier, first delete the barrier if the
26844 jump is the only way to reach the target label.
26846 2021-02-09 Andre Vieira <andre.simoesdiasvieira@arm.com>
26848 * config/aarch64/aarch64-cost-tables.h: Add entries for vect.mul.
26849 * config/aarch64/aarch64.c (aarch64_rtx_mult_cost): Use vect.mul for
26850 vector multiplies and vect.alu for SSRA.
26851 * config/arm/aarch-common-protos.h (struct vector_cost_table): Define
26852 vect.mul cost field.
26853 * config/arm/aarch-cost-tables.h: Add entries for vect.mul.
26854 * config/arm/arm.c: Likewise.
26856 2021-02-09 Richard Biener <rguenther@suse.de>
26858 PR tree-optimization/98863
26859 * tree-ssa-sccvn.h (vn_avail::next_undo): Add.
26860 * tree-ssa-sccvn.c (last_pushed_avail): New global.
26861 (rpo_elim::eliminate_push_avail): Chain pushed avails.
26862 (unwind_state::avail_top): Add.
26863 (do_unwind): Rewrite unwinding of avail entries.
26864 (do_rpo_vn): Initialize last_pushed_avail and
26865 avail_top of the undo state.
26867 2021-02-09 Jakub Jelinek <jakub@redhat.com>
26869 PR middle-end/99004
26870 * calls.c (maybe_warn_rdwr_sizes): Change s0 and s1 type from
26871 const char * to char * and free those pointers after use.
26873 2021-02-09 Richard Biener <rguenther@suse.de>
26875 PR tree-optimization/99017
26876 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Allow
26877 zero vector cost entries.
26879 2021-02-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
26881 PR middle-end/98974
26882 * tree-vect-stmts.c (vectorizable_condition): Remove shadow vec_num
26883 parameter in vectorizable_condition.
26885 2021-02-08 Richard Biener <rguenther@suse.de>
26888 * tree.c (walk_tree_1): Walk VECTOR_CST elements.
26890 2021-02-08 Martin Liska <mliska@suse.cz>
26893 * cfgexpand.c (pass_expand::execute): Parse per-function option
26894 flag_patchable_function_entry and use it.
26895 * common.opt: Remove function_entry_patch_area_size and
26896 function_entry_patch_area_start global variables.
26897 * opts.c (parse_and_check_patch_area): New function.
26898 (common_handle_option): Use it.
26899 * opts.h (parse_and_check_patch_area): New function.
26900 * toplev.c (process_options): Parse and use
26901 function_entry_patch_area_size.
26903 2021-02-08 Martin Sebor <msebor@redhat.com>
26905 * doc/extend.texi (attribute malloc): Correct typos.
26907 2021-02-05 Nathan Sidwell <nathan@acm.org>
26910 * gcc.c (driver::maybe_run_linker): Check for input file
26911 accessibility if not linking.
26913 2021-02-05 Richard Biener <rguenther@suse.de>
26915 PR tree-optimization/98855
26916 * tree-vectorizer.h (add_stmt_cost): New overload.
26917 * tree-vect-slp.c (li_cost_vec_cmp): New.
26918 (vect_bb_slp_scalar_cost): Cost individual loop regions
26919 separately. Account for the scalar instance root stmt.
26921 2021-02-05 Tom de Vries <tdevries@suse.de>
26924 * tree-switch-conversion.c (jump_table_cluster::emit): Add loc
26926 (bit_test_cluster::emit): Reuse location_t for newly created
26928 (switch_decision_tree::try_switch_expansion): Preserve
26930 * tree-switch-conversion.h: Change function signatures.
26932 2021-02-05 Jakub Jelinek <jakub@redhat.com>
26935 * config/i386/i386-options.c (m_NONE, m_ALL): Define.
26936 * config/i386/x86-tune.def (X86_TUNE_BRANCH_PREDICTION_HINTS,
26937 X86_TUNE_PROMOTE_QI_REGS): Use m_NONE instead of 0U.
26938 (X86_TUNE_QIMODE_MATH): Use m_ALL instead of ~0U.
26940 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26942 * config/aarch64/aarch64-simd-builtins.def (get_high): Define builtin.
26943 * config/aarch64/aarch64-simd.md (aarch64_get_high<mode>): Define.
26944 * config/aarch64/arm_neon.h (__GET_HIGH): Delete.
26945 (vget_high_f16): Reimplement using new builtin.
26946 (vget_high_f32): Likewise.
26947 (vget_high_f64): Likewise.
26948 (vget_high_p8): Likewise.
26949 (vget_high_p16): Likewise.
26950 (vget_high_p64): Likewise.
26951 (vget_high_s8): Likewise.
26952 (vget_high_s16): Likewise.
26953 (vget_high_s32): Likewise.
26954 (vget_high_s64): Likewise.
26955 (vget_high_u8): Likewise.
26956 (vget_high_u16): Likewise.
26957 (vget_high_u32): Likewise.
26958 (vget_high_u64): Likewise.
26960 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26962 * config/aarch64/aarch64-simd-builtins.def (get_low): Define builtin.
26963 * config/aarch64/aarch64-simd.md (aarch64_get_low<mode>): Define.
26964 * config/aarch64/arm_neon.h (__GET_LOW): Delete.
26965 (vget_low_f16): Reimplement using new builtin.
26966 (vget_low_f32): Likewise.
26967 (vget_low_f64): Likewise.
26968 (vget_low_p8): Likewise.
26969 (vget_low_p16): Likewise.
26970 (vget_low_p64): Likewise.
26971 (vget_low_s8): Likewise.
26972 (vget_low_s16): Likewise.
26973 (vget_low_s32): Likewise.
26974 (vget_low_s64): Likewise.
26975 (vget_low_u8): Likewise.
26976 (vget_low_u16): Likewise.
26977 (vget_low_u32): Likewise.
26978 (vget_low_u64): Likewise.
26980 2021-02-05 Kito Cheng <kito.cheng@sifive.com>
26982 * gcc.c (print_multilib_info): Check all required argument is provided
26985 2021-02-05 liuhongt <hongtao.liu@intel.com>
26988 * config/i386/i386-expand.c (ix86_expand_sse_cmp): Don't
26989 generate integer mask comparison for 128/256-bits vector when
26990 op_true/op_false is NULL_RTX or CONSTM1_RTX/CONST0_RTX. Also
26991 delete redundant !maskcmp condition.
26992 (ix86_expand_int_vec_cmp): Ditto but no redundant deletion
26994 (ix86_expand_sse_movcc): Delete definition of maskcmp, add the
26995 condition directly to if (maskcmp), add extra check for
26996 cmpmode, it should be MODE_INT.
26997 (ix86_expand_fp_vec_cmp): Pass NULL to ix86_expand_sse_cmp's
26998 parameters op_true/op_false.
26999 (ix86_use_mask_cmp_p): New.
27001 2021-02-05 liuhongt <hongtao.liu@intel.com>
27004 * config/i386/x86-tune.def (X86_TUNE_AVX256_UNALIGNED_LOAD_OPTIMAL):
27005 Remove m_GENERIC from ~list.
27006 (X86_TUNE_AVX256_UNALIGNED_STORE_OPTIMAL): Ditto.
27008 2021-02-04 David Malcolm <dmalcolm@redhat.com>
27011 * diagnostic-show-locus.c (compatible_locations_p): Require
27012 locations in the same macro map to be either both from the
27013 macro definition, or both from the macro arguments.
27015 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
27017 * config/aarch64/aarch64-simd-builtins.def: Add
27018 [su]mull_hi_lane[q] builtin generator macros.
27019 * config/aarch64/aarch64-simd.md
27020 (aarch64_<su>mull_hi_lane<mode>_insn): Define.
27021 (aarch64_<su>mull_hi_lane<mode>): Define.
27022 (aarch64_<su>mull_hi_laneq<mode>_insn): Define.
27023 (aarch64_<su>mull_hi_laneq<mode>): Define.
27024 * config/aarch64/arm_neon.h (vmull_high_lane_s16): Use RTL
27025 builtin instead of inline asm.
27026 (vmull_high_lane_s32): Likewise.
27027 (vmull_high_lane_u16): Likewise.
27028 (vmull_high_lane_u32): Likewise.
27029 (vmull_high_laneq_s16): Likewise.
27030 (vmull_high_laneq_s32): Likewise.
27031 (vmull_high_laneq_u16): Likewise.
27032 (vmull_high_laneq_u32): Liekwise.
27034 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
27036 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_hi_n
27037 builtin generator macros.
27038 * config/aarch64/aarch64-simd.md
27039 (aarch64_<su>mull_hi_n<mode>_insn): Define.
27040 (aarch64_<su>mull_hi_n<mode>): Define.
27041 * config/aarch64/arm_neon.h (vmull_high_n_s16): Use RTL builtin
27042 instead of inline asm.
27043 (vmull_high_n_s32): Likewise.
27044 (vmull_high_n_u16): Likewise.
27045 (vmull_high_n_u32): Likewise.
27047 2021-02-04 Richard Biener <rguenther@suse.de>
27049 PR tree-optimization/98855
27050 * tree-vect-loop.c (vectorizable_phi): Do not cost
27051 single-argument PHIs.
27052 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Likewise.
27053 * tree-vect-stmts.c (vectorizable_bswap): Also perform
27054 costing for SLP operation.
27056 2021-02-04 Martin Liska <mliska@suse.cz>
27058 * doc/extend.texi: Mention -mprefer-vector-width in target
27061 2021-02-03 Martin Sebor <msebor@redhat.com>
27063 PR tree-optimization/98937
27064 * tree-ssa-strlen.c (strlen_dom_walker::~strlen_dom_walker): Define.
27065 Flush pointer_query cache.
27067 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
27069 * config/rs6000/genfusion.pl (gen_2logical): Add missing
27070 fixes based on patch review.
27071 * config/rs6000/fusion.md: Regenerate file.
27073 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
27075 * config/rs6000/t-rs6000: Comment out auto generation of
27078 2021-02-03 Andrew Stubbs <ams@codesourcery.com>
27080 * config/gcn/gcn-opts.h (enum processor_type): Add PROCESSOR_GFX908.
27081 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Add gfx908.
27082 (output_file_start): Add gfx908.
27083 * config/gcn/gcn.opt (gpu_type): Add gfx908.
27084 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Add march=gfx908.
27085 (MULTILIB_DIRNAMES): Add gfx908.
27086 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): New define.
27087 (main): Recognize gfx908.
27088 * config/gcn/t-omp-device: Add gfx908.
27090 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
27092 * config/aarch64/aarch64-simd-builtins.def: Add
27093 [su]mlsl_hi_lane[q] builtin macro generators.
27094 * config/aarch64/aarch64-simd.md
27095 (aarch64_<su>mlsl_hi_lane<mode>_insn): Define.
27096 (aarch64_<su>mlsl_hi_lane<mode>): Define.
27097 (aarch64_<su>mlsl_hi_laneq<mode>_insn): Define.
27098 (aarch64_<su>mlsl_hi_laneq<mode>): Define.
27099 * config/aarch64/arm_neon.h (vmlsl_high_lane_s16): Use RTL
27100 builtin instead of inline asm.
27101 (vmlsl_high_lane_s32): Likewise.
27102 (vmlsl_high_lane_u16): Likewise.
27103 (vmlsl_high_lane_u32): Likewise.
27104 (vmlsl_high_laneq_s16): Likewise.
27105 (vmlsl_high_laneq_s32): Likewise.
27106 (vmlsl_high_laneq_u16): Likewise.
27107 (vmlsl_high_laneq_u32): Likewise.
27108 (vmlal_high_laneq_u32): Likewise.
27110 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
27112 * config/aarch64/aarch64-simd-builtins.def: Add
27113 [su]mlal_hi_lane[q] builtin generator macros.
27114 * config/aarch64/aarch64-simd.md
27115 (aarch64_<su>mlal_hi_lane<mode>_insn): Define.
27116 (aarch64_<su>mlal_hi_lane<mode>): Define.
27117 (aarch64_<su>mlal_hi_laneq<mode>_insn): Define.
27118 (aarch64_<su>mlal_hi_laneq<mode>): Define.
27119 * config/aarch64/arm_neon.h (vmlal_high_lane_s16): Use RTL
27120 builtin instead of inline asm.
27121 (vmlal_high_lane_s32): Likewise.
27122 (vmlal_high_lane_u16): Likewise.
27123 (vmlal_high_lane_u32): Likewise.
27124 (vmlal_high_laneq_s16): Likewise.
27125 (vmlal_high_laneq_s32): Likewise.
27126 (vmlal_high_laneq_u16): Likewise.
27127 (vmlal_high_laneq_u32): Likewise.
27129 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
27131 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_hi_n
27132 builtin generator macros.
27133 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_hi_n<mode>_insn):
27135 (aarch64_<su>mlsl_hi_n<mode>): Define.
27136 * config/aarch64/arm_neon.h (vmlsl_high_n_s16): Use RTL builtin
27137 instead of inline asm.
27138 (vmlsl_high_n_s32): Likewise.
27139 (vmlsl_high_n_u16): Likewise.
27140 (vmlsl_high_n_u32): Likewise.
27142 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
27144 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_hi_n
27145 builtin generator macros.
27146 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_hi_n<mode>_insn):
27148 (aarch64_<su>mlal_hi_n<mode>): Define.
27149 * config/aarch64/arm_neon.h (vmlal_high_n_s16): Use RTL builtin
27150 instead of inline asm.
27151 (vmlal_high_n_s32): Likewise.
27152 (vmlal_high_n_u16): Likewise.
27153 (vmlal_high_n_u32): Likewise.
27155 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
27157 * config/aarch64/aarch64-simd-builtins.def: Add RTL builtin
27159 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal_hi<mode>):
27161 (aarch64_<su>mlal_hi<mode>_insn): This.
27162 (aarch64_<su>mlal_hi<mode>): Define.
27163 * config/aarch64/arm_neon.h (vmlal_high_s8): Use RTL builtin
27164 instead of inline asm.
27165 (vmlal_high_s16): Likewise.
27166 (vmlal_high_s32): Likewise.
27167 (vmlal_high_u8): Likewise.
27168 (vmlal_high_u16): Likewise.
27169 (vmlal_high_u32): Likewise.
27171 2021-02-03 Ilya Leoshkevich <iii@linux.ibm.com>
27173 * lra-spills.c (remove_pseudos): Call lra_update_insn_recog_data()
27174 after calling alter_subreg() on a (mem).
27176 2021-02-03 Martin Liska <mliska@suse.cz>
27179 * lto-streamer-out.c (produce_lto_section): Fill up missing
27181 * lto-streamer.h (struct lto_section): Add _padding field.
27183 2021-02-03 Richard Biener <rguenther@suse.de>
27185 * lto-streamer.c (lto_get_section_name): Free temporary
27187 * tree-loop-distribution.c
27188 (loop_distribution::merge_dep_scc_partitions): Free edge data.
27190 2021-02-03 Jakub Jelinek <jakub@redhat.com>
27192 PR middle-end/97487
27193 * ifcvt.c (noce_can_force_operand): New function.
27194 (noce_emit_move_insn): Use it.
27195 (noce_try_sign_mask): Likewise. Formatting fix.
27197 2021-02-03 Jakub Jelinek <jakub@redhat.com>
27199 PR middle-end/97971
27200 * lra-constraints.c (process_alt_operands): For inline asm, don't call
27201 fatal_insn, but instead return false.
27203 2021-02-03 Jakub Jelinek <jakub@redhat.com>
27205 PR tree-optimization/98287
27206 * config/i386/mmx.md (<insn><mode>3): For shifts don't enable expander
27209 2021-02-03 Tamar Christina <tamar.christina@arm.com>
27211 PR tree-optimization/98928
27212 * tree-vect-loop.c (vect_analyze_loop_2): Change
27213 STMT_VINFO_SLP_VECT_ONLY to STMT_VINFO_SLP_VECT_ONLY_PATTERN.
27214 * tree-vect-slp-patterns.c (complex_pattern::build): Likewise.
27215 * tree-vectorizer.h (STMT_VINFO_SLP_VECT_ONLY_PATTERN): New.
27216 (class _stmt_vec_info): Add slp_vect_pattern_only_p.
27218 2021-02-02 Richard Biener <rguenther@suse.de>
27220 * gimple-loop-interchange.cc (prepare_data_references):
27222 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
27223 * tree-ssa-loop-im.c (hoist_memory_references): Likewise.
27224 * tree-vect-stmts.c (vectorizable_condition): Do not
27226 (vectorizable_comparison): Likewise.
27228 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27230 * config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
27231 * config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
27232 * config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
27233 (vrsqrteq_u32): Likewise.
27235 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27237 * config/aarch64/aarch64-simd-builtins.def (sqxtun2): Define builtin.
27238 * config/aarch64/aarch64-simd.md (aarch64_sqxtun2<mode>_le): Define.
27239 (aarch64_sqxtun2<mode>_be): Likewise.
27240 (aarch64_sqxtun2<mode>): Likewise.
27241 * config/aarch64/arm_neon.h (vqmovun_high_s16): Reimplement using builtin.
27242 (vqmovun_high_s32): Likewise.
27243 (vqmovun_high_s64): Likewise.
27244 * config/aarch64/iterators.md (UNSPEC_SQXTUN2): Define.
27246 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27248 * config/aarch64/aarch64-simd-builtins.def (bfdot_lane, bfdot_laneq): Use
27250 (bfmlalb_lane, bfmlalt_lane, bfmlalb_lane_q, bfmlalt_lane_q): Use FP flags.
27252 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27254 * config/aarch64/aarch64-simd-builtins.def (fcmla_lane0, fcmla_lane90,
27255 fcmla_lane180, fcmla_lane270, fcmlaq_lane0, fcmlaq_lane90, fcmlaq_lane180,
27256 fcmlaq_lane270, scvtf, ucvtf, fcvtzs, fcvtzu, scvtfsi, scvtfdi, ucvtfsi,
27257 ucvtfdi, fcvtzshf, fcvtzuhf, fmlal_lane_low, fmlsl_lane_low,
27258 fmlal_laneq_low, fmlsl_laneq_low, fmlalq_lane_low, fmlslq_lane_low,
27259 fmlalq_laneq_low, fmlslq_laneq_low, fmlal_lane_high, fmlsl_lane_high,
27260 fmlal_laneq_high, fmlsl_laneq_high, fmlalq_lane_high, fmlslq_lane_high,
27261 fmlalq_laneq_high, fmlslq_laneq_high): Use FP flags.
27263 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27265 * config/aarch64/aarch64-builtins.c (FLAG_LOAD): Define.
27266 * config/aarch64/aarch64-simd-builtins.def (ld1x2, ld2, ld3, ld4, ld2r,
27267 ld3r, ld4r, ld1, ld1x3, ld1x4): Use LOAD flags.
27269 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27271 * config/aarch64/aarch64-simd-builtins.def (combine, zip1, zip2,
27272 uzp1, uzp2, trn1, trn2, simd_bsl): Use AUTO_FP flags.
27274 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27276 * config/aarch64/aarch64-simd-builtins.def (clrsb, clz, ctz, popcount,
27277 vec_smult_lane_, vec_smlal_lane_, vec_smult_laneq_, vec_smlal_laneq_,
27278 vec_umult_lane_, vec_umlal_lane_, vec_umult_laneq_, vec_umlal_laneq_,
27279 ashl, sshl, ushl, srshl, urshl, sdot_lane, udot_lane, sdot_laneq,
27280 udot_laneq, usdot_lane, usdot_laneq, sudot_lane, sudot_laneq, ashr,
27281 ashr_simd, lshr, lshr_simd, srshr_n, urshr_n, ssra_n, usra_n, srsra_n,
27282 ursra_n, sshll_n, ushll_n, sshll2_n, ushll2_n, ssri_n, usri_n, ssli_n,
27283 ssli_n, usli_n, bswap, rbit, simd_bsl, eor3q, rax1q, xarq, bcaxq): Use
27284 NONE builtin flags.
27286 2021-02-02 Jakub Jelinek <jakub@redhat.com>
27288 PR tree-optimization/98848
27289 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
27290 STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
27292 2021-02-02 Kito Cheng <kito.cheng@sifive.com>
27295 * expr.c: Check mode before calling store_expr.
27297 2021-02-02 Christophe Lyon <christophe.lyon@linaro.org>
27299 * config/arm/iterators.md (supf): Remove VORNQ_S and VORNQ_U.
27301 * config/arm/mve.md (mve_vornq_s<mode>): New entry for vorn
27302 instruction using expression ior.
27303 (mve_vornq_u<mode>): New expander.
27304 (mve_vornq_f<mode>): Use ior code instead of unspec.
27305 * config/arm/unspecs.md (VORNQ_S, VORNQ_U, VORNQ_F): Remove.
27307 2021-02-02 Alexandre Oliva <oliva@adacore.com>
27309 * tree-nested.c (convert_nonlocal_reference_op): Move
27310 current_function_decl restore after re-gimplification.
27311 (convert_local_reference_op): Likewise.
27313 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27315 * config/aarch64/aarch64-simd-builtins.def (rshrn, rshrn2):
27317 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>_insn_le):
27319 (aarch64_rshrn<mode>_insn_be): Likewise.
27320 (aarch64_rshrn<mode>): Likewise.
27321 (aarch64_rshrn2<mode>_insn_le): Likewise.
27322 (aarch64_rshrn2<mode>_insn_be): Likewise.
27323 (aarch64_rshrn2<mode>): Likewise.
27324 * config/aarch64/aarch64.md (unspec): Add UNSPEC_RSHRN.
27325 * config/aarch64/arm_neon.h (vrshrn_high_n_s16): Reimplement
27327 (vrshrn_high_n_s32): Likewise.
27328 (vrshrn_high_n_s64): Likewise.
27329 (vrshrn_high_n_u16): Likewise.
27330 (vrshrn_high_n_u32): Likewise.
27331 (vrshrn_high_n_u64): Likewise.
27332 (vrshrn_n_s16): Likewise.
27333 (vrshrn_n_s32): Likewise.
27334 (vrshrn_n_s64): Likewise.
27335 (vrshrn_n_u16): Likewise.
27336 (vrshrn_n_u32): Likewise.
27337 (vrshrn_n_u64): Likewise.
27339 2021-02-01 Sergei Trofimovich <siarheit@google.com>
27341 PR tree-optimization/98499
27342 * ipa-modref.c (analyze_ssa_name_flags): treat RVO
27343 conservatively and assume all possible side-effects.
27345 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27347 * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi,
27348 vec_unpacku_hi_): Define builtins.
27349 * config/aarch64/arm_neon.h (vmovl_high_s8): Reimplement using
27351 (vmovl_high_s16): Likewise.
27352 (vmovl_high_s32): Likewise.
27353 (vmovl_high_u8): Likewise.
27354 (vmovl_high_u16): Likewise.
27355 (vmovl_high_u32): Likewise.
27357 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27359 * config/aarch64/aarch64-simd-builtins.def (sabdl, uabdl):
27361 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl<mode>): New
27363 * config/aarch64/aarch64.md (unspec): Define UNSPEC_SABDL,
27365 * config/aarch64/arm_neon.h (vabdl_s8): Reimplemet using
27367 (vabdl_s16): Likewise.
27368 (vabdl_s32): Likewise.
27369 (vabdl_u8): Likewise.
27370 (vabdl_u16): Likewise.
27371 (vabdl_u32): Likewise.
27372 * config/aarch64/iterators.md (ABDL): New int iterator.
27373 (sur): Handle UNSPEC_SABDL, UNSPEC_UABDL.
27375 2021-02-01 Martin Sebor <msebor@redhat.com>
27377 * tree.h (BLOCK_VARS): Add comment.
27378 (BLOCK_SUBBLOCKS): Same.
27379 (BLOCK_SUPERCONTEXT): Same.
27380 (BLOCK_ABSTRACT_ORIGIN): Same.
27381 (inlined_function_outer_scope_p): Same.
27383 2021-02-01 Martin Sebor <msebor@redhat.com>
27385 PR middle-end/97172
27386 * attribs.c (attr_access::free_lang_data): Define new function.
27387 * attribs.h (attr_access::free_lang_data): Declare new function.
27389 2021-02-01 Richard Biener <rguenther@suse.de>
27391 * vec.h (auto_vec::auto_vec): Add memory stat parameters
27393 * bitmap.h (auto_bitmap::auto_bitmap): Likewise.
27395 2021-02-01 Tamar Christina <tamar.christina@arm.com>
27397 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>,
27398 aarch64_<su>mlsl<mode>, aarch64_<su>mlsl_n<mode>): Flip mult operands.
27400 2021-02-01 Richard Biener <rguenther@suse.de>
27402 PR rtl-optimization/98863
27403 * config/i386/i386-features.c (convert_scalars_to_vector):
27404 Set DF_RD_PRUNE_DEAD_DEFS.
27406 2021-01-31 Eric Botcazou <ebotcazou@adacore.com>
27408 * system.h (SIZE_MAX): Define if not already defined.
27410 2021-01-30 Aaron Sawdey <acsawdey@linux.ibm.com>
27412 * config/rs6000/genfusion.pl (gen_2logical): New function to
27413 generate patterns for logical-logical fusion.
27414 * config/rs6000/fusion.md: Regenerated patterns.
27415 * config/rs6000/rs6000-cpus.def: Add
27416 OPTION_MASK_P10_FUSION_2LOGICAL.
27417 * config/rs6000/rs6000.c (rs6000_option_override_internal):
27418 Enable logical-logical fusion for p10.
27419 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2logical.
27421 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
27423 * config/rs6000/rs6000.opt: Add periods to new AIX options.
27425 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
27427 * config/rs6000/rs6000.opt (mabi=vec-extabi): New.
27428 (mabi=vec-default): New.
27429 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
27430 __EXTABI__ for AIX Vector extended ABI.
27431 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print AIX Vector
27433 (conditional_register_usage): If AIX vec_extabi enabled, vs20-vs31
27435 * doc/invoke.texi (PowerPC mabi): Add AIX vec-extabi and vec-default.
27437 2021-01-30 Jakub Jelinek <jakub@redhat.com>
27439 * config/i386/i386-features.c (remove_partial_avx_dependency): Clear
27440 DF_DEFER_INSN_RESCAN after calling df_process_deferred_rescans.
27442 2021-01-29 Vladimir N. Makarov <vmakarov@redhat.com>
27445 * lra-constraints.c (in_class_p): Don't narrow class only for REG
27448 2021-01-29 Will Schmidt <will_schmidt@vnet.ibm.com>
27450 * config/rs6000/rs6000-call.c (rs6000_expand_binup_builtin): Add
27451 clauses for CODE_FOR_vsx_xvcvuxddp_scale and
27452 CODE_FOR_vsx_xvcvsxddp_scale to the parameter checking code.
27454 2021-01-29 Andrew MacLeod <amacleod@redhat.com>
27456 PR tree-optimization/98866
27457 * gimple-range-gori.h (gori_compute:set_range_invariant): New.
27458 * gimple-range-gori.cc (gori_map::set_range_invariant): New.
27459 (gori_map::m_maybe_invariant): Rename from all_outgoing.
27460 (gori_map::gori_map): Rename all_outgoing to m_maybe_invariant.
27461 (gori_map::is_export_p): Ditto.
27462 (gori_map::calculate_gori): Ditto.
27463 (gori_compute::set_range_invariant): New.
27464 * gimple-range.cc (gimple_ranger::range_of_stmt): Set range
27465 invariant for pointers evaluating to [1, +INF].
27467 2021-01-29 Richard Biener <rguenther@suse.de>
27469 PR rtl-optimization/98863
27470 * config/i386/i386-features.c (remove_partial_avx_dependency):
27471 Do not perform DF analysis.
27472 (pass_data_remove_partial_avx_dependency): Remove
27475 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
27477 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_n
27478 builtin generator macros.
27479 * config/aarch64/aarch64-simd.md (aarch64_<su>mull_n<mode>):
27481 * config/aarch64/arm_neon.h (vmull_n_s16): Use RTL builtin
27482 instead of inline asm.
27483 (vmull_n_s32): Likewise.
27484 (vmull_n_u16): Likewise.
27485 (vmull_n_u32): Likewise.
27487 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27489 * config/aarch64/aarch64-simd-builtins.def (sabdl2, uabdl2):
27491 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
27493 (aarch64_<sur>abdl2<mode>): ... This.
27494 (<sur>sadv16qi): Adjust use of above.
27495 * config/aarch64/arm_neon.h (vabdl_high_s8): Reimplement using
27497 (vabdl_high_s16): Likewise.
27498 (vabdl_high_s32): Likewise.
27499 (vabdl_high_u8): Likewise.
27500 (vabdl_high_u16): Likewise.
27501 (vabdl_high_u32): Likewise.
27503 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27505 * config/aarch64/aarch64-simd-builtins.def (sabal2): Define
27507 (uabal2): Likewise.
27508 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal2<mode>): New
27510 * config/aarch64/aarch64.md (unspec): Add UNSPEC_SABAL2 and
27512 * config/aarch64/arm_neon.h (vabal_high_s8): Reimplement using
27514 (vabal_high_s16): Likewise.
27515 (vabal_high_s32): Likewise.
27516 (vabal_high_u8): Likewise.
27517 (vabal_high_u16): Likewise.
27518 (vabal_high_u32): Likewise.
27519 * config/aarch64/iterators.md (ABAL2): New mode iterator.
27520 (sur): Handle UNSPEC_SABAL2, UNSPEC_UABAL2.
27522 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27524 * config/aarch64/aarch64-simd-builtins.def (sabal): Define
27527 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal<mode>_4):
27529 (aarch64_<sur>abal<mode>): ... This
27530 (<sur>sadv16qi): Adust use of the above.
27531 * config/aarch64/arm_neon.h (vabal_s8): Reimplement using
27533 (vabal_s16): Likewise.
27534 (vabal_s32): Likewise.
27535 (vabal_u8): Likewise.
27536 (vabal_u16): Likewise.
27537 (vabal_u32): Likewise.
27539 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27541 * config/aarch64/aarch64-simd-builtins.def (saddlv, uaddlv):
27543 * config/aarch64/aarch64-simd.md (aarch64_<su>addlv<mode>):
27545 * config/aarch64/arm_neon.h (vaddlv_s8): Reimplement using
27547 (vaddlv_s16): Likewise.
27548 (vaddlv_u8): Likewise.
27549 (vaddlv_u16): Likewise.
27550 (vaddlvq_s8): Likewise.
27551 (vaddlvq_s16): Likewise.
27552 (vaddlvq_s32): Likewise.
27553 (vaddlvq_u8): Likewise.
27554 (vaddlvq_u16): Likewise.
27555 (vaddlvq_u32): Likewise.
27556 (vaddlv_s32): Likewise.
27557 (vaddlv_u32): Likewise.
27558 * config/aarch64/iterators.md (VDQV_L): New mode iterator.
27559 (unspec): Add UNSPEC_SADDLV, UNSPEC_UADDLV.
27560 (Vwstype): New mode attribute.
27562 (VWIDE_S): Likewise.
27563 (USADDLV): New int iterator.
27564 (su): Handle UNSPEC_SADDLV, UNSPEC_UADDLV.
27566 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
27568 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_lane[q]
27569 builtin generator macros.
27570 * config/aarch64/aarch64-simd.md (aarch64_vec_<su>mlsl_lane<Qlane>):
27572 * config/aarch64/arm_neon.h (vmlsl_lane_s16): Use RTL builtin
27573 instead of inline asm.
27574 (vmlsl_lane_s32): Likewise.
27575 (vmlsl_lane_u16): Likewise.
27576 (vmlsl_lane_u32): Likewise.
27577 (vmlsl_laneq_s16): Likewise.
27578 (vmlsl_laneq_s32): Likewise.
27579 (vmlsl_laneq_u16): Likewise.
27580 (vmlsl_laneq_u32): Likewise.
27582 2021-01-29 Richard Biener <rguenther@suse.de>
27584 * doc/invoke.texi (--param max-gcse-memory): Document unit
27586 * gcse.c (gcse_or_cprop_is_too_expensive): Adjust.
27587 * params.opt (--param max-gcse-memory): Adjust default and
27588 document unit of size.
27590 2021-01-29 Richard Biener <rguenther@suse.de>
27592 PR rtl-optimization/98863
27593 * gcse.c (gcse_or_cprop_is_too_expensive): Use unsigned
27594 HOST_WIDE_INT for the memory estimate.
27596 2021-01-29 Bin Cheng <bin.cheng@linux.alibaba.com>
27597 Richard Biener <rguenther@suse.de>
27599 PR tree-optimization/97627
27600 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
27601 Do not analyze fake edges.
27603 2021-01-29 Richard Biener <rguenther@suse.de>
27605 PR rtl-optimization/98144
27606 * df.h (df_mir_bb_info): Add con_visited member.
27607 * df-problems.c (df_mir_alloc): Initialize con_visited,
27608 do not fully populate IN and OUT.
27609 (df_mir_reset): Likewise.
27610 (df_mir_confluence_0): Set con_visited.
27611 (df_mir_confluence_n): Properly handle implicitely
27612 fully populated IN and OUT as designated by con_visited
27613 and update con_visited accordingly.
27615 2021-01-29 Jakub Jelinek <jakub@redhat.com>
27618 * config/arm/vec-common.md (mve_vshlq_<supf><mode>,
27619 vashl<mode>3, vashr<mode>3, vlshr<mode>3): Add
27620 && !TARGET_REALLY_IWMMXT to conditions.
27622 2021-01-29 Jakub Jelinek <jakub@redhat.com>
27625 * cfgbuild.c (find_bb_boundaries): Reset debug_insn when seeing
27628 2021-01-28 Marek Polacek <polacek@redhat.com>
27631 * stor-layout.c (finalize_type_size): If we reset TYPE_USER_ALIGN in
27632 the main variant, maybe reset it in its variants too.
27633 * tree.c (check_base_type): Return true only if TYPE_USER_ALIGN match.
27634 (check_aligned_type): Check if TYPE_USER_ALIGN match.
27636 2021-01-28 Christophe Lyon <christophe.lyon@linaro.org>
27639 * config/arm/arm.c (arm_rtx_costs_internal): Adjust cost of vector
27640 of constant zero for comparisons.
27642 2021-01-28 Michael Meissner <meissner@linux.ibm.com>
27644 * config/rs6000/rs6000.c (rs6000_mangle_decl_assembler_name): Add
27645 support for mapping built-in function names for long double
27646 built-in functions if long double is IEEE 128-bit.
27648 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
27650 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_n
27651 builtin generator macros.
27652 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_n<mode>):
27654 * config/aarch64/arm_neon.h (vmlsl_n_s16): Use RTL builtin
27655 instead of inline asm.
27656 (vmlsl_n_s32): Likewise.
27657 (vmlsl_n_u16): Likewise.
27658 (vmlsl_n_u32): Likewise.
27660 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
27662 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_n
27663 builtin generator macros.
27664 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>):
27666 * config/aarch64/arm_neon.h (vmlal_n_s16): Use RTL builtin
27667 instead of inline asm.
27668 (vmlal_n_s32): Likewise.
27669 (vmlal_n_u16): Likewise.
27670 (vmlal_n_u32): Likewise.
27672 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27674 * config/aarch64/aarch64-simd-builtins.def (shrn2): Define
27676 * config/aarch64/aarch64-simd.md (aarch64_shrn2<mode>_insn_le):
27678 (aarch64_shrn2<mode>_insn_be): Likewise.
27679 (aarch64_shrn2<mode>): Likewise.
27680 * config/aarch64/arm_neon.h (vshrn_high_n_s16): Reimlplement
27682 (vshrn_high_n_s32): Likewise.
27683 (vshrn_high_n_s64): Likewise.
27684 (vshrn_high_n_u16): Likewise.
27685 (vshrn_high_n_u32): Likewise.
27686 (vshrn_high_n_u64): Likewise.
27688 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27690 * config/aarch64/aarch64-simd-builtins.def (shrn): Define
27692 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le):
27694 (aarch64_shrn<mode>_insn_be): Likewise.
27695 (aarch64_shrn<mode>): Likewise.
27696 * config/aarch64/arm_neon.h (vshrn_n_s16): Reimplement using
27698 (vshrn_n_s32): Likewise.
27699 (vshrn_n_s64): Likewise.
27700 (vshrn_n_u16): Likewise.
27701 (vshrn_n_u32): Likewise.
27702 (vshrn_n_u64): Likewise.
27703 * config/aarch64/iterators.md (vn_mode): New mode attribute.
27705 2021-01-28 Richard Biener <rguenther@suse.de>
27707 PR rtl-optimization/80960
27708 * dse.c (check_mem_read_rtx): Call get_addr on the
27711 2021-01-28 Xionghu Luo <luoxhu@linux.ibm.com>
27712 David Edelsohn <dje.gcc@gmail.com>
27715 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
27716 Don't generate VIEW_CONVERT_EXPR for fcode ALTIVEC_BUILTIN_VEC_INSERT
27718 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
27720 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Remove the
27721 wrapper call rs6000_expand_vector_set_var for cleanup. Call
27722 rs6000_expand_vector_set_var_p9 and rs6000_expand_vector_set_var_p8
27724 (rs6000_expand_vector_set_var): Delete.
27725 (rs6000_expand_vector_set_var_p9): Make static.
27726 (rs6000_expand_vector_set_var_p8): Make static.
27728 2021-01-28 Xing GUO <higuoxing@gmail.com>
27730 * common/config/riscv/riscv-common.c
27731 (riscv_subset_list::parsing_subset_version): Fix -march option parsing
27732 when `p` extension exists.
27734 2021-01-27 Vladimir N. Makarov <vmakarov@redhat.com>
27736 PR rtl-optimization/97684
27737 * ira.c (ira): Call ira_set_pseudo_classes before
27738 update_equiv_regs when it is necessary.
27740 2021-01-27 Jakub Jelinek <jakub@redhat.com>
27743 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Use
27744 %w0, %w1 and %2 instead of %0, %1 and %2.
27746 2021-01-27 Aaron Sawdey <acsawdey@linux.ibm.com>
27748 * config/rs6000/genfusion.pl: New script to generate
27749 define_insn_and_split patterns so combine can arrange fused
27750 instructions next to each other.
27751 * config/rs6000/fusion.md: New file, generated fused instruction
27752 patterns for combine.
27753 * config/rs6000/predicates.md (const_m1_to_1_operand): New predicate.
27754 (non_update_memory_operand): New predicate.
27755 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_P10_FUSION and
27756 OPTION_MASK_P10_FUSION_LD_CMPI to ISA_3_1_MASKS_SERVER and
27758 * config/rs6000/rs6000-protos.h (address_is_non_pfx_d_or_x): Add
27760 * config/rs6000/rs6000.c (rs6000_option_override_internal):
27761 Automatically set OPTION_MASK_P10_FUSION and
27762 OPTION_MASK_P10_FUSION_LD_CMPI if target is power10.
27763 (rs600_opt_masks): Allow -mpower10-fusion
27764 in function attributes.
27765 (address_is_non_pfx_d_or_x): New function.
27766 * config/rs6000/rs6000.h: Add MASK_P10_FUSION.
27767 * config/rs6000/rs6000.md: Include fusion.md.
27768 * config/rs6000/rs6000.opt: Add -mpower10-fusion
27769 and -mpower10-fusion-ld-cmpi.
27770 * config/rs6000/t-rs6000: Add dependencies involving fusion.md.
27772 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
27774 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal
27775 builtin generator macros.
27776 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal<mode>):
27778 (aarch64_<su>mlal<mode>): This.
27779 * config/aarch64/arm_neon.h (vmlal_s8): Use RTL builtin
27780 instead of inline asm.
27781 (vmlal_s16): Likewise.
27782 (vmlal_s32): Likewise.
27783 (vmlal_u8): Likewise.
27784 (vmlal_u16): Likewise.
27785 (vmlal_u32): Likewise.
27787 2021-01-27 Richard Biener <rguenther@suse.de>
27789 PR tree-optimization/98854
27790 * tree-vect-slp.c (vect_build_slp_tree_2): Also build
27791 PHIs from scalars when the number of CTORs matches the
27792 number of children.
27794 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
27796 * config/aarch64/aarch64-simd-builtins.def: Add mls_n builtin
27798 * config/aarch64/aarch64-simd.md (*aarch64_mls_elt_merge<mode>):
27800 (aarch64_mls_n<mode>): This.
27801 * config/aarch64/arm_neon.h (vmls_n_s16): Use RTL builtin
27803 (vmls_n_s32): Likewise.
27804 (vmls_n_u16): Likewise.
27805 (vmls_n_u32): Likewise.
27806 (vmlsq_n_s16): Likewise.
27807 (vmlsq_n_s32): Likewise.
27808 (vmlsq_n_u16): Likewise.
27809 (vmlsq_n_u32): Likewise.
27811 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
27813 * config/aarch64/aarch64-simd-builtins.def: Add mls builtin
27815 * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather
27817 (vmls_s16): Likewise.
27818 (vmls_s32): Likewise.
27819 (vmls_u8): Likewise.
27820 (vmls_u16): Likewise.
27821 (vmls_u32): Likewise.
27822 (vmlsq_s8): Likewise.
27823 (vmlsq_s16): Likewise.
27824 (vmlsq_s32): Likewise.
27825 (vmlsq_u8): Likewise.
27826 (vmlsq_u16): Likewise.
27827 (vmlsq_u32): Likewise.
27829 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
27831 * config/aarch64/aarch64-simd-builtins.def: Add mla_n builtin
27833 * config/aarch64/aarch64-simd.md (*aarch64_mla_elt_merge<mode>):
27835 (aarch64_mla_n<mode>): This.
27836 * config/aarch64/arm_neon.h (vmla_n_s16): Use RTL builtin
27838 (vmla_n_s32): Likewise.
27839 (vmla_n_u16): Likewise.
27840 (vmla_n_u32): Likewise.
27841 (vmlaq_n_s16): Likewise.
27842 (vmlaq_n_s32): Likewise.
27843 (vmlaq_n_u16): Likewise.
27844 (vmlaq_n_u32): Likewise.
27846 2021-01-27 liuhongt <hongtao.liu@intel.com>
27849 * config/i386/sse.md (sse2_gt<mode>3): Drop !TARGET_XOP in condition.
27850 (*sse2_eq<mode>3): Ditto.
27852 2021-01-27 Jakub Jelinek <jakub@redhat.com>
27854 * tree-pass.h (PROP_trees): Rename to ...
27855 (PROP_gimple): ... this.
27856 * cfgexpand.c (pass_data_expand): Replace PROP_trees with PROP_gimple.
27857 * passes.c (execute_function_dump, execute_function_todo,
27858 execute_one_ipa_transform_pass, execute_one_pass): Likewise.
27859 * varpool.c (ctor_for_folding): Likewise.
27861 2021-01-27 Jakub Jelinek <jakub@redhat.com>
27863 PR tree-optimization/97260
27864 * varpool.c: Include tree-pass.h.
27865 (ctor_for_folding): In GENERIC return DECL_INITIAL for TREE_READONLY
27866 non-TREE_SIDE_EFFECTS automatic variables.
27868 2021-01-26 Paul Fee <paul.f.fee@gmail.com>
27870 * doc/cpp.texi (__cplusplus): Document value for -std=c++23
27872 * doc/invoke.texi: Document -std=c++23 and -std=gnu++23.
27873 * dwarf2out.c (highest_c_language): Recognise C++20 and C++23.
27874 (gen_compile_unit_die): Recognise C++23.
27876 2021-01-26 Jakub Jelinek <jakub@redhat.com>
27879 * dwarf2asm.c (dw2_assemble_integer): Cast DWARF2_ADDR_SIZE to int
27882 2021-01-26 Jakub Jelinek <jakub@redhat.com>
27885 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
27886 Use UINTVAL (shft_amnt) and UINTVAL (mask) instead of INTVAL (shft_amnt)
27887 and INTVAL (mask). Add && INTVAL (mask) > 0 condition.
27889 2021-01-26 Richard Biener <rguenther@suse.de>
27891 * gimple-pretty-print.c (dump_binary_rhs): Handle
27892 VEC_WIDEN_{PLUS,MINUS}_{LO,HI}_EXPR.
27894 2021-01-26 Richard Biener <rguenther@suse.de>
27896 PR middle-end/98726
27897 * tree.h (vector_cst_int_elt): Remove.
27898 * tree.c (vector_cst_int_elt): Use poly_wide_int for computations,
27901 2021-01-26 Andrew Stubbs <ams@codesourcery.com>
27903 * config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
27904 for V64DFmode min/max reductions.
27906 2021-01-26 Jakub Jelinek <jakub@redhat.com>
27908 * dwarf2asm.c (dw2_assemble_integer): Handle size twice as large
27909 as DWARF2_ADDR_SIZE if x is not a scalar int by emitting it as
27910 two halves, one with x and the other with const0_rtx, ordered
27911 depending on endianity.
27913 2021-01-26 Alexandre Oliva <oliva@adacore.com>
27915 * gimplify.c (gimplify_decl_expr): Skip asan marking calls for
27916 temporaries not seen in binding block, and not about to be
27917 added as gimple variables.
27919 2021-01-25 Martin Sebor <msebor@redhat.com>
27922 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust warning text.
27924 2021-01-25 Martin Liska <mliska@suse.cz>
27926 * value-prof.c (get_nth_most_common_value): Use %s instead
27929 2021-01-25 Jakub Jelinek <jakub@redhat.com>
27932 * configure.ac (HAVE_AS_GDWARF_5_DEBUG_FLAG): Only define if
27933 readelf -wi is able to read the emitted .debug_info back.
27934 * configure: Regenerated.
27936 2021-01-25 Martin Liska <mliska@suse.cz>
27938 PR gcov-profile/98739
27939 * common.opt: Add missing sign symbol.
27940 * value-prof.c (get_nth_most_common_value): Restore handling
27941 of PROFILE_REPRODUCIBILITY_PARALLEL_RUNS and
27942 PROFILE_REPRODUCIBILITY_MULTITHREADED.
27944 2021-01-25 Richard Biener <rguenther@suse.de>
27946 PR middle-end/98807
27947 * tree.c (vector_element_bits): Always use precision of
27948 the element type for boolean vectors.
27950 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
27952 * config/rtems.h (STARTFILE_SPEC): Remove qnolinkcmds.
27953 (ENDFILE_SPEC): Evaluate qnolinkcmds.
27955 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
27957 * config/rtems.h (STARTFILE_SPEC): Remove nostdlib and
27958 nostartfiles handling since this is already done by
27959 LINK_COMMAND_SPEC. Evaluate qnolinkcmds.
27960 (ENDFILE_SPEC): Remove nostdlib and nostartfiles handling since this
27961 is already done by LINK_COMMAND_SPEC.
27962 (LIB_SPECS): Remove nostdlib and nodefaultlibs handling since
27963 this is already done by LINK_COMMAND_SPEC. Remove qnolinkcmds
27966 2021-01-25 Jakub Jelinek <jakub@redhat.com>
27969 * fold-const-call.c (host_size_t_cst_p): Renamed to ...
27970 (size_t_cst_p): ... this. Check and store unsigned HOST_WIDE_INT
27971 value rather than host size_t.
27972 (fold_const_call): Change type of s2 from size_t to
27973 unsigned HOST_WIDE_INT. Use size_t_cst_p instead of
27974 host_size_t_cst_p. For strncmp calls, pass MIN (s2, SIZE_MAX)
27975 instead of s2 as last argument.
27977 2021-01-25 Tamar Christina <tamar.christina@arm.com>
27979 * config/arm/iterators.md (rotsplit1, rotsplit2, conj_op, fcmac1,
27980 VCMLA_OP, VCMUL_OP): New.
27981 * config/arm/mve.md (mve_vcmlaq<mve_rot><mode>): Support vec_dup 0.
27982 * config/arm/neon.md (cmul<conj_op><mode>3): New.
27983 * config/arm/unspecs.md (UNSPEC_VCMLA_CONJ, UNSPEC_VCMLA180_CONJ,
27984 UNSPEC_VCMUL_CONJ): New.
27985 * config/arm/vec-common.md (cmul<conj_op><mode>3, arm_vcmla<rot><mode>,
27986 cml<fcmac1><conj_op><mode>4): New.
27988 2021-01-23 Jakub Jelinek <jakub@redhat.com>
27991 * config/rs6000/mmintrin.h (__m64): Add __may_alias__ attribute.
27993 2021-01-22 Jonathan Wright <jonathan.wright@arm.com>
27995 * config/aarch64/aarch64-simd-builtins.def: Add mla builtin
27997 * config/aarch64/arm_neon.h (vmla_s8): Use RTL builtin rather
27999 (vmla_s16): Likewise.
28000 (vmla_s32): Likewise.
28001 (vmla_u8): Likewise.
28002 (vmla_u16): Likewise.
28003 (vmla_u32): Likewise.
28004 (vmlaq_s8): Likewise.
28005 (vmlaq_s16): Likewise.
28006 (vmlaq_s32): Likewise.
28007 (vmlaq_u8): Likewise.
28008 (vmlaq_u16): Likewise.
28009 (vmlaq_u32): Likewise.
28011 2021-01-22 David Malcolm <dmalcolm@redhat.com>
28013 * doc/invoke.texi (GCC_EXTRA_DIAGNOSTIC_OUTPUT): Add @findex
28016 2021-01-22 Jakub Jelinek <jakub@redhat.com>
28019 * dwarf2out.c (output_file_names): For -gdwarf-5, if there are no
28020 filenames to emit, still emit the required 0 index directory and
28021 filename entries that match DW_AT_comp_dir and DW_AT_name of the
28024 2021-01-22 Marek Polacek <polacek@redhat.com>
28027 * doc/invoke.texi: Update C++ ABI Version 15 description.
28029 2021-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28031 PR tree-optimization/98766
28032 * tree-ssa-math-opts.c (convert_mult_to_fma): Use maybe_le when
28033 comparing against type size with param_avoid_fma_max_bits.
28035 2021-01-22 Richard Biener <rguenther@suse.de>
28037 PR middle-end/98793
28038 * tree.c (vector_element_bits): Key single-bit bool vector on
28039 integer mode rather than not vector mode.
28041 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
28044 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
28045 Generate ARRAY_REF(VIEW_CONVERT_EXPR) for P8 and later
28047 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): Update
28048 to call different path for P8 and P9.
28049 (rs6000_expand_vector_set_var_p9): New function.
28050 (rs6000_expand_vector_set_var_p8): New function.
28052 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
28056 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
28057 Ajdust variable index vec_insert from address dereference to
28058 ARRAY_REF(VIEW_CONVERT_EXPR) tree expression.
28059 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
28061 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): New function.
28063 2021-01-22 Martin Liska <mliska@suse.cz>
28065 PR gcov-profile/98739
28066 * profile.c (compute_value_histograms): Drop time profile for
28067 -fprofile-reproducible=multithreaded.
28069 2021-01-22 Nathan Sidwell <nathan@acm.org>
28071 * gcc.c (process_command): Don't check OPT_SPECIAL_input_file
28074 2021-01-22 Richard Biener <rguenther@suse.de>
28076 PR middle-end/98773
28077 * tree-data-ref.c (initalize_matrix_A): Revert previous
28078 change, retaining failing on HOST_WIDE_INT_MIN CHREC_RIGHT.
28080 2021-01-22 Jakub Jelinek <jakub@redhat.com>
28082 PR tree-optimization/90248
28083 * match.pd (X cmp 0.0 ? 1.0 : -1.0 -> copysign(1, +-X),
28084 X cmp 0.0 ? -1.0 : +1.0 -> copysign(1, -+X)): Remove
28086 (X * (X cmp 0.0 ? 1.0 : -1.0) -> +-abs(X),
28087 X * (X cmp 0.0 ? -1.0 : 1.0) -> +-abs(X)): New simplifications.
28089 2021-01-22 Jakub Jelinek <jakub@redhat.com>
28091 PR tree-optimization/98255
28092 * tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
28093 extend index - low_bound from sizetype's precision rather than index
28095 (get_addr_base_and_unit_offset_1): Likewise.
28096 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
28097 * gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
28099 2021-01-22 Richard Biener <rguenther@suse.de>
28101 PR tree-optimization/98786
28102 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Avoid
28103 adding new uses of abnormals. Verify we deal with a conditional
28106 2021-01-22 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
28109 * optc-save-gen.awk: Add arm_fp16_format to checked_options.
28111 2021-01-22 liuhongt <hongtao.liu@intel.com>
28115 * config/i386/sse.md (VI_128_256): New mode iterator.
28116 (*avx_cmp<mode>3_1, *avx_cmp<mode>3_2, *avx_cmp<mode>3_3,
28117 *avx_cmp<mode>3_4, *avx2_eq<mode>3, *avx2_pcmp<mode>3_1,
28118 *avx2_pcmp<mode>3_2, *avx2_gt<mode>3): New
28119 define_insn_and_split to lower avx512 vector comparison to avx
28120 version when dest is vector.
28121 (*<avx512>_cmp<mode>3,*<avx512>_cmp<mode>3,*<avx512>_ucmp<mode>3):
28122 define_insn_and_split for negating the comparison result.
28123 * config/i386/predicates.md (float_vector_all_ones_operand):
28125 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Use
28126 general NOT operator without UNSPEC_MASKOP.
28128 2021-01-21 Vladimir N. Makarov <vmakarov@redhat.com>
28130 PR rtl-optimization/98777
28131 * lra-int.h (lra_pmode_pseudo): New extern.
28132 * lra.c (lra_pmode_pseudo): New global.
28134 * lra-eliminations.c (eliminate_regs_in_insn): Use it.
28136 2021-01-21 Ilya Leoshkevich <iii@linux.ibm.com>
28138 * fwprop.c (fwprop_propagation::classify_result): Allow
28139 (subreg (mem)) simplifications.
28141 2021-01-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28143 * config/aarch64/aarch64-simd.md (aarch64_sqdml<SBINQOPS:as>l<mode>):
28145 (aarch64_sqdmlal<mode>): ... This...
28146 (aarch64_sqdmlsl<mode>): ... And this.
28147 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Split into...
28148 (aarch64_sqdmlal_lane<mode>): ... This...
28149 (aarch64_sqdmlsl_lane<mode>): ... And this.
28150 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Split into...
28151 (aarch64_sqdmlsl_laneq<mode>): ... This...
28152 (aarch64_sqdmlal_laneq<mode>): ... And this.
28153 (aarch64_sqdml<SBINQOPS:as>l_n<mode>): Split into...
28154 (aarch64_sqdmlsl_n<mode>): ... This...
28155 (aarch64_sqdmlal_n<mode>): ... And this.
28156 (aarch64_sqdml<SBINQOPS:as>l2<mode>_internal): Split into...
28157 (aarch64_sqdmlal2<mode>_internal): ... This...
28158 (aarch64_sqdmlsl2<mode>_internal): ... And this.
28160 2021-01-21 Christophe Lyon <christophe.lyon@linaro.org>
28162 * config/arm/arm_mve.h (__arm_vcmpneq_s8): Fix return type.
28164 2021-01-21 Andrea Corallo <andrea.corallo@arm.com>
28167 * doc/sourcebuild.texi (arm_thumb2_no_arm_v8_1_lob): Document.
28169 2021-01-21 liuhongt <hongtao.liu@intel.com>
28171 PR rtl-optimization/98694
28172 * regcprop.c (copy_value): If SRC had been assigned a mode
28173 narrower than the copy, we can't link DEST into the chain even
28174 they have same hard_regno_nregs(i.e. HImode/SImode in i386
28177 2021-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28179 * config/aarch64/aarch64-simd.md (aarch64_get_lane<mode>):
28180 Convert to define_insn_and_split. Split into simple move when moving
28183 2021-01-20 Segher Boessenkool <segher@kernel.crashing.org>
28185 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Change assert.
28186 Adjust comment. Simplify code.
28188 2021-01-20 Jakub Jelinek <jakub@redhat.com>
28191 * dwarf2out.c (reset_indirect_string): Also reset indirect strings
28192 with DW_FORM_line_strp form.
28193 (prune_unused_types_update_strings): Don't add into debug_str_hash
28194 indirect strings with DW_FORM_line_strp form.
28195 (adjust_name_comp_dir): New function.
28196 (dwarf2out_finish): Call it on CU DIEs after resetting
28197 debug_line_str_hash.
28199 2021-01-20 Vladimir N. Makarov <vmakarov@redhat.com>
28201 PR rtl-optimization/98722
28202 * lra-eliminations.c (eliminate_regs_in_insn): Check that target
28203 has no 3-op add insn to transform insns containing two pluses.
28205 2021-01-20 Richard Biener <rguenther@suse.de>
28207 * hwint.h (add_hwi): New function.
28208 (mul_hwi): Likewise.
28209 * tree-data-ref.c (initialize_matrix_A): Properly translate
28210 tree constants and avoid HOST_WIDE_INT_MIN.
28211 (lambda_matrix_row_add): Avoid undefined integer overflow
28212 and return true on such overflow.
28213 (lambda_matrix_right_hermite): Handle overflow from
28214 lambda_matrix_row_add gracefully. Simplify previous fix.
28215 (analyze_subscript_affine_affine): Likewise.
28217 2021-01-20 Eugene Rozenfeld <erozen@microsoft.com>
28219 PR tree-optimization/96674
28220 * match.pd: New patterns: x < y || y == XXX_MIN --> x <= y - 1
28221 x >= y && y != XXX_MIN --> x > y - 1
28223 2021-01-20 Richard Sandiford <richard.sandiford@arm.com>
28225 PR tree-optimization/98535
28226 * tree-vect-slp.c (duplicate_and_interleave): Use quick_grow_cleared.
28227 If the high and low permutes are the same, remove the high permutes
28228 from the working set and only continue with the low ones.
28230 2021-01-20 Jakub Jelinek <jakub@redhat.com>
28232 PR tree-optimization/98721
28233 * builtins.c (access_ref::inform_access): Don't assume
28234 SSA_NAME_IDENTIFIER must be non-NULL. Print messages about
28235 object whenever allocfn is NULL, rather than only when DECL_P
28236 is true. Use %qE instead of %qD for that. Formatting fixes.
28238 2021-01-20 Richard Biener <rguenther@suse.de>
28240 PR tree-optimization/98758
28241 * tree-data-ref.c (int_divides_p): Use lambda_int arguments.
28242 (lambda_matrix_right_hermite): Avoid undefinedness with
28243 signed integer abs and multiplication.
28244 (analyze_subscript_affine_affine): Use lambda_int.
28246 2021-01-20 David Malcolm <dmalcolm@redhat.com>
28249 * dwarf2out.c (output_line_info): Rename static variable
28250 "generation", moving it out of the function to...
28251 (output_line_info_generation): New.
28252 (init_sections_and_labels): Likewise, renaming the variable to...
28253 (init_sections_and_labels_generation): New.
28254 (dwarf2out_c_finalize): Reset the new variables.
28256 2021-01-19 Martin Sebor <msebor@redhat.com>
28258 PR middle-end/98664
28259 * tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
28260 all functions, even if they're not declared artificial or inline.
28261 * tree.c (tree_inlined_location): Use macro expansion location
28262 only if scope traversal fails to expose one.
28264 2021-01-19 Richard Sandiford <richard.sandiford@arm.com>
28266 PR rtl-optimization/92294
28267 * alias.c (compare_base_symbol_refs): Take an extra parameter
28268 and add the distance between two symbols to it. Enshrine in
28269 comments that -1 means "either 0 or 1, but we can't tell
28270 which at compile time".
28271 (memrefs_conflict_p): Update call accordingly.
28272 (rtx_equal_for_memref_p): Likewise. Take the distance between symbols
28275 2021-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28277 * config/aarch64/aarch64-simd-builtins.def (sqshl, uqshl,
28278 sqrshl, uqrshl, sqadd, uqadd, sqsub, uqsub, suqadd, usqadd, sqmovn,
28279 uqmovn, sqxtn2, uqxtn2, sqabs, sqneg, sqdmlal, sqdmlsl, sqdmlal_lane,
28280 sqdmlsl_lane, sqdmlal_laneq, sqdmlsl_laneq, sqdmlal_n, sqdmlsl_n,
28281 sqdmlal2, sqdmlsl2, sqdmlal2_lane, sqdmlsl2_lane, sqdmlal2_laneq,
28282 sqdmlsl2_laneq, sqdmlal2_n, sqdmlsl2_n, sqdmull, sqdmull_lane,
28283 sqdmull_laneq, sqdmull_n, sqdmull2, sqdmull2_lane, sqdmull2_laneq,
28284 sqdmull2_n, sqdmulh, sqrdmulh, sqdmulh_lane, sqdmulh_laneq,
28285 sqrdmulh_lane, sqrdmulh_laneq, sqshrun_n, sqrshrun_n, sqshrn_n,
28286 uqshrn_n, sqrshrn_n, uqrshrn_n, sqshlu_n, sqshl_n, uqshl_n, sqrdmlah,
28287 sqrdmlsh, sqrdmlah_lane, sqrdmlsh_lane, sqrdmlah_laneq, sqrdmlsh_laneq,
28288 sqmovun): Use NONE flags.
28290 2021-01-19 Richard Biener <rguenther@suse.de>
28293 * ipa-modref.c (analyze_stmt): Only record a summary for a
28296 2021-01-19 Richard Biener <rguenther@suse.de>
28298 PR middle-end/98638
28299 * tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
28301 2021-01-19 Daniel Hellstrom <daniel@gaisler.com>
28303 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
28304 built-in define __FIX_LEON3FT_TN0018.
28306 2021-01-19 Richard Biener <rguenther@suse.de>
28309 * tree-inline.c (tree_function_versioning): Set input_location
28310 to UNKNOWN_LOCATION throughout the function.
28312 2021-01-19 Tobias Burnus <tobias@codesourcery.com>
28315 * omp-low.c (lower_omp_target): Handle nonpointer is_device_ptr.
28317 2021-01-19 Martin Jambor <mjambor@suse.cz>
28320 * ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
28321 whether non-call exceptions allow removal of a statement.
28322 (isra_analyze_call): Pass the appropriate function to
28323 ssa_name_only_returned_p.
28325 2021-01-19 Geng Qi <gengqi@linux.alibaba.com>
28327 * config/riscv/arch-canonicalize (longext_sort): New function for
28328 sorting 'multi-letter'.
28329 * config/riscv/multilib-generator: Adjusting the loop of 'alt' in
28330 'alts'. The 'arch' may not be the first of 'alts'.
28331 (_expand_combination): Add underline for the 'ext' without '*'.
28332 This is because, a single-letter extension can always be treated well
28333 with a '_' prefix, but it cannot be separated out if it is appended
28336 2021-01-18 Vladimir N. Makarov <vmakarov@redhat.com>
28339 * ira.c (ira): Skip abnormal critical edge splitting.
28341 2021-01-18 Jakub Jelinek <jakub@redhat.com>
28343 PR tree-optimization/98727
28344 * tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
28345 second .MUL_OVERFLOW operand for signed multiplication with overflow
28346 checking if the second operand of multiplication is not constant.
28348 2021-01-18 David Edelsohn <dje.gcc@gmail.com>
28350 * doc/invoke.texi (-gdwarf): TPF defaults to version 2 and AIX
28351 defaults to version 4.
28353 2021-01-18 David Malcolm <dmalcolm@redhat.com>
28355 * attribs.h (fndecl_dealloc_argno): New decl.
28356 * builtins.c (call_dealloc_argno): Split out second half of
28358 (fndecl_dealloc_argno): New.
28359 * doc/extend.texi (Common Function Attributes): Document the
28360 interaction between the analyzer and the malloc attribute.
28361 * doc/invoke.texi (Static Analyzer Options): Likewise.
28363 2021-01-17 David Edelsohn <dje.gcc@gmail.com>
28365 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Override
28366 dwarf_version to 4.
28367 * config/rs6000/aix72.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
28369 2021-01-17 Martin Jambor <mjambor@suse.cz>
28372 * cgraph.c (clone_of_p): Check also former_clone_of as we climb
28375 2021-01-17 Mark Wielaard <mark@klomp.org>
28377 * common.opt (gdwarf-): Init(5).
28378 * doc/invoke.texi (-gdwarf): Document default to 5.
28380 2021-01-16 Kwok Cheung Yeung <kcy@codesourcery.com>
28382 * builtin-types.def
28383 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
28385 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
28386 ...this. Add extra argument.
28387 * gimplify.c (omp_default_clause): Ensure that event handle is
28388 firstprivate in a task region.
28389 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_DETACH.
28390 (gimplify_adjust_omp_clauses): Likewise.
28391 * omp-builtins.def (BUILT_IN_GOMP_TASK): Change function type to
28392 BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR.
28393 * omp-expand.c (expand_task_call): Add GOMP_TASK_FLAG_DETACH to flags
28394 if detach clause specified. Add detach argument when generating
28396 * omp-low.c (scan_sharing_clauses): Setup data environment for detach
28398 (finish_taskreg_scan): Move field for variable containing the event
28399 handle to the front of the struct.
28400 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_DETACH. Fix
28402 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
28403 OMP_CLAUSE_DETACH clause.
28404 (convert_local_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
28405 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_DETACH.
28406 * tree.c (omp_clause_num_ops): Add entry for OMP_CLAUSE_DETACH.
28408 (omp_clause_code_name): Add entry for OMP_CLAUSE_DETACH. Fix
28410 (walk_tree_1): Handle OMP_CLAUSE_DETACH.
28412 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
28414 * config/nios2/t-rtems: Reset all MULTILIB_* variables. Shorten
28415 multilib directory names. Use MULTILIB_REQUIRED instead of
28416 MULTILIB_EXCEPTIONS. Add -mhw-mul -mhw-mulx -mhw-div
28417 -mcustom-fpu-cfg=fph2 multilib.
28419 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
28421 * config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
28422 (nios2_init_fpu_configs): Provide register values for new
28423 -mcustom-fpu-cfg=fph2 option variant.
28424 * doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
28427 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
28429 * config/nios2/nios2.c (nios2_custom_check_insns): Remove
28430 custom instruction warnings.
28432 2021-01-16 Jakub Jelinek <jakub@redhat.com>
28434 PR tree-optimization/96669
28435 * match.pd ((CST << x) & 1 -> x == 0): New simplification.
28437 2021-01-16 Jakub Jelinek <jakub@redhat.com>
28439 PR tree-optimization/96271
28440 * passes.def: Pass false argument to first two pass_cd_dce
28441 instances and true to last instance. Add comment that
28442 last instance rewrites no longer addressed locals.
28443 * tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
28445 (pass_cd_dce::set_pass_param): New method.
28446 (pass_cd_dce::execute): Return TODO_update_address_taken from
28447 last cd_dce instance.
28449 2021-01-15 Carl Love <cel@us.ibm.com>
28451 * config/rs6000/altivec.h (vec_mulh, vec_div, vec_dive, vec_mod):
28453 * config/rs6000/altivec.md (VIlong): Move define to file vsx.md.
28454 * config/rs6000/rs6000-builtin.def (DIVES_V4SI, DIVES_V2DI,
28455 DIVEU_V4SI, DIVEU_V2DI, DIVS_V4SI, DIVS_V2DI, DIVU_V4SI,
28456 DIVU_V2DI, MODS_V2DI, MODS_V4SI, MODU_V2DI, MODU_V4SI,
28457 MULHS_V2DI, MULHS_V4SI, MULHU_V2DI, MULHU_V4SI, MULLD_V2DI):
28458 Add builtin define.
28459 (MULH, DIVE, MOD): Add new BU_P10_OVERLOAD_2 definitions.
28460 * config/rs6000/rs6000-call.c (VSX_BUILTIN_VEC_DIV,
28461 VSX_BUILTIN_VEC_DIVE, P10_BUILTIN_VEC_MOD, P10_BUILTIN_VEC_MULH):
28462 New overloaded definitions.
28463 (builtin_function_type) [P10V_BUILTIN_DIVEU_V4SI,
28464 P10V_BUILTIN_DIVEU_V2DI, P10V_BUILTIN_DIVU_V4SI,
28465 P10V_BUILTIN_DIVU_V2DI, P10V_BUILTIN_MODU_V2DI,
28466 P10V_BUILTIN_MODU_V4SI, P10V_BUILTIN_MULHU_V2DI,
28467 P10V_BUILTIN_MULHU_V4SI]: Add case
28468 statement for builtins.
28469 * config/rs6000/rs6000.md (bits): Add new attribute sizes V4SI, V2DI.
28470 * config/rs6000/vsx.md (VIlong): Moved from config/rs6000/altivec.md.
28471 (UNSPEC_VDIVES, UNSPEC_VDIVEU): New unspec definitions.
28472 (vsx_mul_v2di): Add if TARGET_POWER10 statement.
28473 (vsx_udiv_v2di): Add if TARGET_POWER10 statement.
28474 (dives_<mode>, diveu_<mode>, div<mode>3, uvdiv<mode>3,
28475 mods_<mode>, modu_<mode>, mulhs_<mode>, mulhu_<mode>, mulv2di3):
28476 Add define_insn, mode is VIlong.
28477 * doc/extend.texi (vec_mulh, vec_mul, vec_div, vec_dive, vec_mod):
28478 Add builtin descriptions.
28480 2021-01-15 Eric Botcazou <ebotcazou@adacore.com>
28482 * final.c (final_start_function_1): Reset force_source_line.
28484 2021-01-15 Jakub Jelinek <jakub@redhat.com>
28486 PR tree-optimization/96669
28487 * match.pd (((1 << A) & 1) != 0 -> A == 0,
28488 ((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
28489 possibly different power of two constants and to right shift too.
28491 2021-01-15 Jakub Jelinek <jakub@redhat.com>
28493 PR tree-optimization/96681
28494 * match.pd ((x < 0) ^ (y < 0) to (x ^ y) < 0): New simplification.
28495 ((x >= 0) ^ (y >= 0) to (x ^ y) < 0): Likewise.
28496 ((x < 0) ^ (y >= 0) to (x ^ y) >= 0): Likewise.
28497 ((x >= 0) ^ (y < 0) to (x ^ y) >= 0): Likewise.
28499 2021-01-15 Alexandre Oliva <oliva@adacore.com>
28501 * opts.c (gen_command_line_string): Exclude -dumpbase-ext.
28503 2021-01-15 Tamar Christina <tamar.christina@arm.com>
28505 * config/aarch64/aarch64-simd.md (cml<fcmac1><conj_op><mode>4,
28506 cmul<conj_op><mode>3): New.
28507 * config/aarch64/iterators.md (UNSPEC_FCMUL,
28508 UNSPEC_FCMUL180, UNSPEC_FCMLA_CONJ, UNSPEC_FCMLA180_CONJ,
28509 UNSPEC_CMLA_CONJ, UNSPEC_CMLA180_CONJ, UNSPEC_CMUL, UNSPEC_CMUL180,
28510 FCMLA_OP, FCMUL_OP, conj_op, rotsplit1, rotsplit2, fcmac1, sve_rot1,
28511 sve_rot2, SVE2_INT_CMLA_OP, SVE2_INT_CMUL_OP, SVE2_INT_CADD_OP): New.
28512 (rot): Add UNSPEC_FCMUL, UNSPEC_FCMUL180.
28513 (rot_op): Renamed to conj_op.
28514 * config/aarch64/aarch64-sve.md (cml<fcmac1><conj_op><mode>4,
28515 cmul<conj_op><mode>3): New.
28516 * config/aarch64/aarch64-sve2.md (cml<fcmac1><conj_op><mode>4,
28517 cmul<conj_op><mode>3): New.
28519 2021-01-15 David Malcolm <dmalcolm@redhat.com>
28523 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
28524 Escape the tempfile name when constructing the expected output.
28526 2021-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28528 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlsl_hi<mode>):
28530 (aarch64_<su>mlsl_hi<mode>): ... This.
28531 (aarch64_<su>mlsl_hi<mode>): Define.
28532 (*aarch64_<su>mlsl<mode): Rename to...
28533 (aarch64_<su>mlsl<mode): ... This.
28534 * config/aarch64/aarch64-simd-builtins.def (smlsl, umlsl,
28535 smlsl_hi, umlsl_hi): Define builtins.
28536 * config/aarch64/arm_neon.h (vmlsl_high_s8, vmlsl_high_s16,
28537 vmlsl_high_s32, vmlsl_high_u8, vmlsl_high_u16, vmlsl_high_u32,
28538 vmlsl_s8, vmlsl_s16, vmlsl_s32, vmlsl_u8,
28539 vmlsl_u16, vmlsl_u32): Reimplement with builtins.
28541 2021-01-15 Uroš Bizjak <ubizjak@gmail.com>
28543 * config/i386/i386-c.c (ix86_target_macros):
28544 Use cpp_define_formatted for __SIZEOF_FLOAT80__ definition.
28546 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
28549 * config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
28550 * Makefile.in (RTL_SSA_H): New variable.
28551 * config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
28552 * config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
28553 * config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
28555 * config/aarch64/aarch64-cc-fusion.cc: New file.
28557 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
28559 * recog.h (insn_change_watermark::~insn_change_watermark): Avoid
28560 calling cancel_changes for changes that no longer exist.
28562 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
28564 * rtl-ssa/functions.h (function_info::ref_defs): Rename to...
28565 (function_info::reg_defs): ...this.
28566 * rtl-ssa/member-fns.inl (function_info::ref_defs): Rename to...
28567 (function_info::reg_defs): ...this.
28569 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
28572 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
28574 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
28577 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
28580 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
28582 2021-01-15 Richard Biener <rguenther@suse.de>
28584 PR tree-optimization/96376
28585 * tree-vect-stmts.c (get_load_store_type): Disregard alignment
28586 for VMAT_INVARIANT.
28588 2021-01-15 Martin Liska <mliska@suse.cz>
28590 * doc/install.texi: Document that some tests need pytest module.
28591 * doc/sourcebuild.texi: Likewise.
28593 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
28596 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
28598 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
28600 * config/arm/mve.md (mve_vshrq_n_s<mode>_imm): New entry.
28601 (mve_vshrq_n_u<mode>_imm): Likewise.
28602 * config/arm/neon.md (vashr<mode>3, vlshr<mode>3): Move to ...
28603 * config/arm/vec-common.md: ... here.
28605 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
28607 * config/arm/mve.md (mve_vshlq_<supf><mode>): Move to
28609 * config/arm/neon.md (vashl<mode>3): Delete.
28610 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): New.
28611 (vasl<mode>3): New expander.
28613 2021-01-15 Richard Biener <rguenther@suse.de>
28615 PR tree-optimization/98685
28616 * tree-vect-slp.c (vect_schedule_slp_node): Refactor handling
28617 of vector extern defs.
28619 2021-01-14 David Malcolm <dmalcolm@redhat.com>
28622 * diagnostic.c (diagnostic_kind_text): Break out this array
28624 (diagnostic_build_prefix): ...here.
28625 (fancy_abort): Detect when diagnostic_initialize has not yet been
28626 called and fall back to a minimal implementation of printing the
28627 ICE, rather than segfaulting in internal_error.
28629 2021-01-14 David Malcolm <dmalcolm@redhat.com>
28631 * diagnostic.c (diagnostic_initialize): Eliminate
28632 parseable_fixits_p in favor of initializing extra_output_kind from
28633 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
28634 (convert_column_unit): New function, split out from...
28635 (diagnostic_converted_column): ...this.
28636 (print_parseable_fixits): Add "column_unit" and "tabstop" params.
28637 Use them to call convert_column_unit on the column values.
28638 (diagnostic_report_diagnostic): Eliminate conditional on
28639 parseable_fixits_p in favor of a switch statement on
28640 extra_output_kind, passing the appropriate values to the new
28641 params of print_parseable_fixits.
28642 (selftest::test_print_parseable_fixits_none): Update for new
28643 params of print_parseable_fixits.
28644 (selftest::test_print_parseable_fixits_insert): Likewise.
28645 (selftest::test_print_parseable_fixits_remove): Likewise.
28646 (selftest::test_print_parseable_fixits_replace): Likewise.
28647 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
28649 (selftest::diagnostic_c_tests): Call it.
28650 * diagnostic.h (enum diagnostics_extra_output_kind): New.
28651 (diagnostic_context::parseable_fixits_p): Delete field in favor
28653 (diagnostic_context::extra_output_kind): ...this new field.
28654 * doc/invoke.texi (Environment Variables): Add
28655 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
28656 * opts.c (common_handle_option): Update handling of
28657 OPT_fdiagnostics_parseable_fixits for change to diagnostic_context
28660 2021-01-14 Tamar Christina <tamar.christina@arm.com>
28662 * tree-vect-slp-patterns.c (class complex_operations_pattern,
28663 complex_operations_pattern::matches,
28664 complex_operations_pattern::recognize,
28665 complex_operations_pattern::build): New.
28666 (slp_patterns): Use it.
28668 2021-01-14 Tamar Christina <tamar.christina@arm.com>
28670 * internal-fn.def (COMPLEX_FMS, COMPLEX_FMS_CONJ): New.
28671 * optabs.def (cmls_optab, cmls_conj_optab): New.
28672 * doc/md.texi: Document them.
28673 * tree-vect-slp-patterns.c (class complex_fms_pattern,
28674 complex_fms_pattern::matches, complex_fms_pattern::recognize,
28675 complex_fms_pattern::build): New.
28677 2021-01-14 Tamar Christina <tamar.christina@arm.com>
28679 * internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
28680 * optabs.def (cmla_optab, cmla_conj_optab): New.
28681 * doc/md.texi: Document them.
28682 * tree-vect-slp-patterns.c (vect_match_call_p,
28683 class complex_fma_pattern, vect_slp_reset_pattern,
28684 complex_fma_pattern::matches, complex_fma_pattern::recognize,
28685 complex_fma_pattern::build): New.
28687 2021-01-14 Tamar Christina <tamar.christina@arm.com>
28689 * internal-fn.def (COMPLEX_MUL, COMPLEX_MUL_CONJ): New.
28690 * optabs.def (cmul_optab, cmul_conj_optab): New.
28691 * doc/md.texi: Document them.
28692 * tree-vect-slp-patterns.c (vect_match_call_complex_mla,
28693 vect_normalize_conj_loc, is_eq_or_top, vect_validate_multiplication,
28694 vect_build_combine_node, class complex_mul_pattern,
28695 complex_mul_pattern::matches, complex_mul_pattern::recognize,
28696 complex_mul_pattern::build): New.
28698 2021-01-14 Tamar Christina <tamar.christina@arm.com>
28700 * tree-vect-slp.c (optimize_load_redistribution_1): New.
28701 (optimize_load_redistribution, vect_is_slp_load_node): New.
28702 (vect_match_slp_patterns): Use it.
28704 2021-01-14 Tamar Christina <tamar.christina@arm.com>
28706 * tree-vect-slp-patterns.c (complex_add_pattern::build):
28709 2021-01-14 Thomas Schwinge <thomas@codesourcery.com>
28711 * config/gcn/mkoffload.c (main): Create an offload image only in
28712 64-bit configurations.
28714 2021-01-14 H.J. Lu <hjl.tools@gmail.com>
28717 * config/i386/i386-options.c (ix86_option_override_internal):
28718 Issue an error for -fcf-protection with CF_BRANCH when compiling
28719 for 32-bit non-TARGET_CMOV targets.
28721 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
28724 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p):
28725 Remove declaration and initialization of shadow variable "ret".
28726 (ix86_option_override_internal): Remove delcaration of
28727 shadow variable "i". Redeclare shadowed variable to unsigned.
28728 * common/config/i386/i386-common.c (pta_size): Redeclare to unsigned.
28729 * config/i386/i386-builtins.c (get_builtin_code_for_version):
28730 Update for redeclaration.
28731 * config/i386/i386.h (pta_size): Ditto.
28733 2021-01-14 Richard Biener <rguenther@suse.de>
28735 PR tree-optimization/98674
28736 * tree-data-ref.c (base_supports_access_fn_components_p): New.
28737 (initialize_data_dependence_relation): For two bases without
28738 possible access fns resort to type size equality when determining
28739 shape compatibility.
28741 2021-01-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
28744 * config/arm/arm_neon.h: Replace calls to __builtin_vcge* by
28745 <=, >= operators in vcle and vcge intrinsics respectively.
28746 * config/arm/arm_neon_builtins.def: Remove entry for
28749 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
28752 * config/i386/i386-options.c (ix86_function_specific_save):
28753 Remove redundant assignment to opts->x_ix86_branch_cost.
28754 * config/i386/i386.c (ix86_prefetch_sse):
28755 Rename from x86_prefetch_sse. Update all uses.
28756 * config/i386/i386.h: Update for rename.
28757 * config/i386/i386-options.h: Ditto.
28759 2021-01-14 Jakub Jelinek <jakub@redhat.com>
28762 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3,
28763 *sse4_1_zero_extendv4hiv4si2_3, *sse4_1_zero_extendv2siv2di2_3):
28764 Use Bm instead of m for non-avx. Add isa attribute.
28766 2021-01-14 Jakub Jelinek <jakub@redhat.com>
28768 PR tree-optimization/96688
28769 * match.pd (~(X >> Y) -> ~X >> Y): New simplification if
28770 ~X can be simplified.
28772 2021-01-14 Richard Sandiford <richard.sandiford@arm.com>
28774 * tree-vect-stmts.c (vect_model_load_cost): Account for unused
28775 IFN_LOAD_LANES results.
28777 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28779 * config/aarch64/aarch64-simd.md (aarch64_<su>xtl<mode>):
28781 (aarch64_xtn<mode>): Likewise.
28782 * config/aarch64/aarch64-simd-builtins.def (sxtl, uxtl, xtn):
28785 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
28787 (vmovl_s16): Likewise.
28788 (vmovl_s32): Likewise.
28789 (vmovl_u8): Likewise.
28790 (vmovl_u16): Likewise.
28791 (vmovl_u32): Likewise.
28792 (vmovn_s16): Likewise.
28793 (vmovn_s32): Likewise.
28794 (vmovn_s64): Likewise.
28795 (vmovn_u16): Likewise.
28796 (vmovn_u32): Likewise.
28797 (vmovn_u64): Likewise.
28799 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28801 * config/aarch64/aarch64-simd.md (aarch64_<su>qxtn2<mode>_le):
28803 (aarch64_<su>qxtn2<mode>_be): Likewise.
28804 (aarch64_<su>qxtn2<mode>): Likewise.
28805 * config/aarch64/aarch64-simd-builtins.def (sqxtn2, uqxtn2):
28807 * config/aarch64/iterators.md (SAT_TRUNC): Define code_iterator.
28808 (su): Handle ss_truncate and us_truncate.
28809 * config/aarch64/arm_neon.h (vqmovn_high_s16): Reimplement using
28811 (vqmovn_high_s32): Likewise.
28812 (vqmovn_high_s64): Likewise.
28813 (vqmovn_high_u16): Likewise.
28814 (vqmovn_high_u32): Likewise.
28815 (vqmovn_high_u64): Likewise.
28817 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28819 * config/aarch64/aarch64-simd.md (aarch64_xtn2<mode>_le):
28821 (aarch64_xtn2<mode>_be): Likewise.
28822 (aarch64_xtn2<mode>): Likewise.
28823 * config/aarch64/aarch64-simd-builtins.def (xtn2): Define
28825 * config/aarch64/arm_neon.h (vmovn_high_s16): Reimplement using
28827 (vmovn_high_s32): Likewise.
28828 (vmovn_high_s64): Likewise.
28829 (vmovn_high_u16): Likewise.
28830 (vmovn_high_u32): Likewise.
28831 (vmovn_high_u64): Likewise.
28833 2021-01-13 Stafford Horne <shorne@gmail.com>
28835 * config/or1k/or1k.h (ASM_PREFERRED_EH_DATA_FORMAT): New macro.
28837 2021-01-13 Stafford Horne <shorne@gmail.com>
28839 * config/or1k/linux.h (TARGET_ASM_FILE_END): Define macro.
28841 2021-01-13 Stafford Horne <shorne@gmail.com>
28843 * config/or1k/or1k.h (TARGET_CPU_CPP_BUILTINS): Add builtin
28844 define for __or1k_hard_float__.
28846 2021-01-13 Stafford Horne <shorne@gmail.com>
28848 * config/or1k/or1k.h (NO_PROFILE_COUNTERS): Define as 1.
28849 (PROFILE_HOOK): Define to call _mcount.
28850 (FUNCTION_PROFILER): Change from abort to no-op.
28852 2021-01-13 Jakub Jelinek <jakub@redhat.com>
28854 PR tree-optimization/96691
28855 * match.pd ((~X | C) ^ D -> (X | C) ^ (~D ^ C),
28856 (~X & C) ^ D -> (X & C) ^ (D ^ C)): New simplifications if
28857 (~D ^ C) or (D ^ C) can be simplified.
28859 2021-01-13 Richard Biener <rguenther@suse.de>
28861 PR tree-optimization/92645
28862 * match.pd (BIT_FIELD_REF to conversion): Delay canonicalization
28863 until after vector lowering.
28865 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
28867 * config/aarch64/aarch64-sve.md (fnma<mode>4): Extend from SVE_FULL_I
28869 (@aarch64_pred_fnma<mode>, cond_fnma<mode>, *cond_fnma<mode>_2)
28870 (*cond_fnma<mode>_4, *cond_fnma<mode>_any): Likewise.
28872 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
28874 * config/aarch64/aarch64-sve.md (fma<mode>4): Extend from SVE_FULL_I
28876 (@aarch64_pred_fma<mode>, cond_fma<mode>, *cond_fma<mode>_2)
28877 (*cond_fma<mode>_4, *cond_fma<mode>_any): Likewise.
28879 2021-01-13 Richard Biener <rguenther@suse.de>
28881 PR tree-optimization/92645
28882 * tree-vect-slp.c (vect_build_slp_tree_1): Relax supported
28883 BIT_FIELD_REF argument.
28884 (vect_build_slp_tree_2): Record the desired vector type
28885 on the external vector def.
28886 (vectorizable_slp_permutation): Handle required punning
28887 of existing vector defs.
28889 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
28891 * rtl-ssa/accesses.h (def_lookup): Fix order of comparison results.
28893 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
28895 * config/sh/sh.md (movsf_ie): Remove operands[2] test.
28897 2021-01-13 Samuel Thibault <samuel.thibault@ens-lyon.org>
28899 * config.gcc [$target == *-*-gnu*]: Enable
28900 'default_gnu_indirect_function'.
28902 2021-01-13 Jakub Jelinek <jakub@redhat.com>
28905 * optabs.c (expand_vec_perm_const): Don't force v0 and v1 into
28906 registers before calling targetm.vectorize.vec_perm_const, only after
28908 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const): Handle
28909 two argument permutation when one operand is zero vector and only
28910 after that force operands into registers.
28911 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_1): New
28912 define_insn_and_split pattern.
28913 (*avx512bw_zero_extendv32qiv32hi2_1): Likewise.
28914 (*avx512f_zero_extendv16hiv16si2_1): Likewise.
28915 (*avx2_zero_extendv8hiv8si2_1): Likewise.
28916 (*avx512f_zero_extendv8siv8di2_1): Likewise.
28917 (*avx2_zero_extendv4siv4di2_1): Likewise.
28918 * config/mips/mips.c (mips_vectorize_vec_perm_const): Force operands
28920 * config/arm/arm.c (arm_vectorize_vec_perm_const): Likewise.
28921 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Likewise.
28922 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const): Likewise.
28923 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const): Likewise.
28924 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const): Likewise.
28925 * config/gcn/gcn.c (gcn_vectorize_vec_perm_const): Likewise. Use std::swap.
28927 2021-01-13 Martin Liska <mliska@suse.cz>
28929 PR tree-optimization/98455
28930 * gimple-if-to-switch.cc (condition_info::record_phi_mapping):
28931 Record also virtual PHIs.
28932 (pass_if_to_switch::execute): Return TODO_cleanup_cfg only
28935 2021-01-13 Jonathan Wakely <jwakely@redhat.com>
28937 * doc/invoke.texi (C++ Modules): Fix typos.
28939 2021-01-13 Richard Biener <rguenther@suse.de>
28941 PR tree-optimization/98640
28942 * tree-ssa-sccvn.c (visit_nary_op): Do not try to
28943 handle plus or minus from a truncated operand to be
28946 2021-01-13 Jakub Jelinek <jakub@redhat.com>
28949 * config/i386/i386.md (*btr<mode>_1, *btr<mode>_2): New
28950 define_insn_and_split patterns.
28951 (splitter after *btr<mode>_2): New splitter.
28953 2021-01-13 Martin Liska <mliska@suse.cz>
28956 * cgraphunit.c (analyze_functions): Remove dead code.
28958 2021-01-13 Qian Jianhua <qianjh@cn.fujitsu.com>
28960 * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
28961 * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
28962 (a64fx_regmove_cost, a64fx_vector_cost): New.
28963 (a64fx_tunings): Use the new added cost tables.
28965 2021-01-13 Jakub Jelinek <jakub@redhat.com>
28968 * config/i386/predicates.md (pmovzx_parallel): New predicate.
28969 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): New
28970 define_insn_and_split pattern.
28971 (*sse4_1_zero_extendv4hiv4si2_3): Likewise.
28972 (*sse4_1_zero_extendv2siv2di2_3): Likewise.
28974 2021-01-13 Julian Brown <julian@codesourcery.com>
28976 * config/gcn/gcn.c (gcn_conditional_register_usage): Remove dead code
28977 to fix v0 register.
28979 2021-01-13 Julian Brown <julian@codesourcery.com>
28981 * config/gcn/gcn.c (gcn_md_reorg): Fix case where EXEC reg is live
28984 2021-01-13 Julian Brown <julian@codesourcery.com>
28986 * config/gcn/gcn-valu.md (recip<mode>2<exec>, recip<mode>2): Use unspec
28987 for reciprocal-approximation instructions.
28988 (div<mode>3): Use fused multiply-accumulate operations for reciprocal
28989 refinement and division result.
28990 * config/gcn/gcn.md (UNSPEC_RCP): New unspec constant.
28992 2021-01-13 Julian Brown <julian@codesourcery.com>
28994 * config/gcn/gcn-valu.md (subdf): Rename to...
28997 2021-01-12 Martin Liska <mliska@suse.cz>
28999 * gcov.c (source_info::debug): Fix printf format for 32-bit hosts.
29001 2021-01-12 Andrea Corallo <andrea.corallo@arm.com>
29003 * function-abi.h: Fix typo.
29005 2021-01-12 Christophe Lyon <christophe.lyon@linaro.org>
29009 * config/arm/arm.h (ARM_HAVE_NEON_V8QI_LDST): New macro.
29010 (ARM_HAVE_NEON_V16QI_LDST, ARM_HAVE_NEON_V4HI_LDST): Likewise.
29011 (ARM_HAVE_NEON_V8HI_LDST, ARM_HAVE_NEON_V2SI_LDST): Likewise.
29012 (ARM_HAVE_NEON_V4SI_LDST, ARM_HAVE_NEON_V4HF_LDST): Likewise.
29013 (ARM_HAVE_NEON_V8HF_LDST, ARM_HAVE_NEON_V4BF_LDST): Likewise.
29014 (ARM_HAVE_NEON_V8BF_LDST, ARM_HAVE_NEON_V2SF_LDST): Likewise.
29015 (ARM_HAVE_NEON_V4SF_LDST, ARM_HAVE_NEON_DI_LDST): Likewise.
29016 (ARM_HAVE_NEON_V2DI_LDST): Likewise.
29017 (ARM_HAVE_V8QI_LDST, ARM_HAVE_V16QI_LDST): Likewise.
29018 (ARM_HAVE_V4HI_LDST, ARM_HAVE_V8HI_LDST): Likewise.
29019 (ARM_HAVE_V2SI_LDST, ARM_HAVE_V4SI_LDST, ARM_HAVE_V4HF_LDST): Likewise.
29020 (ARM_HAVE_V8HF_LDST, ARM_HAVE_V4BF_LDST, ARM_HAVE_V8BF_LDST): Likewise.
29021 (ARM_HAVE_V2SF_LDST, ARM_HAVE_V4SF_LDST, ARM_HAVE_DI_LDST): Likewise.
29022 (ARM_HAVE_V2DI_LDST): Likewise.
29023 * config/arm/mve.md (*movmisalign<mode>_mve_store): New pattern.
29024 (*movmisalign<mode>_mve_load): New pattern.
29025 * config/arm/neon.md (movmisalign<mode>): Move to ...
29026 * config/arm/vec-common.md: ... here.
29028 2021-01-12 Vladimir N. Makarov <vmakarov@redhat.com>
29031 * lra-eliminations.c (eliminate_regs_in_insn): Add transformation
29032 of pattern 'plus (plus (hard reg, const), pseudo)'.
29034 2021-01-12 Richard Biener <rguenther@suse.de>
29036 PR tree-optimization/98550
29037 * tree-vect-slp.c (vect_record_max_nunits): Check whether
29038 the group size is a multiple of the vector element count.
29039 (vect_build_slp_tree_1): When we need to fail because
29040 the vector type choosen causes unrolling do so lazily
29041 without affecting matches only at the end to guide group splitting.
29043 2021-01-12 Martin Liska <mliska@suse.cz>
29046 * optc-save-gen.awk: Compare also n_target_save vars with
29049 2021-01-12 Martin Liska <mliska@suse.cz>
29051 * gcov.c (source_info::debug): New.
29052 (print_usage): Add --debug (-D) option.
29053 (process_args): Likewise.
29054 (generate_results): Call src->debug after
29055 accumulate_line_counts.
29056 (read_graph_file): Properly assign id for EXIT_BLOCK.
29057 * profile.c (branch_prob): Dump function body before it is
29060 2021-01-12 Jakub Jelinek <jakub@redhat.com>
29062 PR tree-optimization/98629
29063 * tree-ssa-math-opts.c (arith_overflow_check_p): Don't update use_stmt
29064 unless returning non-zero.
29066 2021-01-12 Jakub Jelinek <jakub@redhat.com>
29068 PR tree-optimization/95731
29069 * tree-ssa-reassoc.c (optimize_range_tests_cmp_bitwise): Also optimize
29070 x < 0 && y < 0 && z < 0 into (x | y | z) < 0 for signed x, y, z.
29071 (optimize_range_tests): Call optimize_range_tests_cmp_bitwise
29072 only after optimize_range_tests_var_bound.
29074 2021-01-12 Jakub Jelinek <jakub@redhat.com>
29076 * configure.ac: Ensure c/Make-lang.in comes first in @all_lang_makefrags@.
29077 * configure: Regenerated.
29079 2021-01-12 liuhongt <hongtao.liu@intel.com>
29082 * config/i386/i386-builtins.h (BUILTIN_DESC_SWAP_OPERANDS):
29084 * config/i386/i386-expand.c (ix86_expand_sse_comi): Delete
29087 2021-01-12 Alexandre Oliva <oliva@adacore.com>
29089 * ssa-iterators.h (end_imm_use_stmt_traverse): Forward
29091 (auto_end_imm_use_stmt_traverse): New struct.
29092 (FOR_EACH_IMM_USE_STMT): Use it.
29093 (BREAK_FROM_IMM_USE_STMT, RETURN_FROM_IMM_USE_STMT): Remove,
29095 * gimple-ssa-strength-reduction.c: ... here, ...
29096 * graphite-scop-detection.c: ... here, ...
29097 * ipa-modref.c, ipa-pure-const.c, ipa-sra.c: ... here, ...
29098 * tree-predcom.c, tree-ssa-ccp.c: ... here, ...
29099 * tree-ssa-dce.c, tree-ssa-dse.c: ... here, ...
29100 * tree-ssa-loop-ivopts.c, tree-ssa-math-opts.c: ... here, ...
29101 * tree-ssa-phiprop.c, tree-ssa.c: ... here, ...
29102 * tree-vect-slp.c: ... and here, ...
29103 * doc/tree-ssa.texi: ... and the example here.
29105 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
29107 * config/aarch64/aarch64-sve.md (sdiv_pow2<mode>3): Extend from
29108 SVE_FULL_I to SVE_I. Generate an UNSPEC_PRED_X.
29109 (*sdiv_pow2<mode>3): New pattern.
29110 (@cond_<sve_int_op><mode>): Extend from SVE_FULL_I to SVE_I.
29111 Wrap the ASRD in an UNSPEC_PRED_X.
29112 (*cond_<sve_int_op><mode>_2): Likewise. Replace the UNSPEC_PRED_X
29113 predicate with a constant PTRUE, if it isn't already.
29114 (*cond_<sve_int_op><mode>_z): Replace with...
29115 (*cond_<sve_int_op><mode>_any): ...this new pattern.
29117 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
29119 * config/aarch64/aarch64-sve.md (*cond_bic<mode>_2): Extend from
29120 SVE_FULL_I to SVE_I.
29121 (*cond_bic<mode>_any): Likewise.
29123 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
29125 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart)
29126 (@aarch64_pred_<MUL_HIGHPART:optab><mode>): Extend from SVE_FULL_I
29129 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
29131 * config/aarch64/aarch64-sve.md (<su>abd<mode>_3): Extend from
29132 SVE_FULL_I to SVE_I.
29133 (*aarch64_cond_<su>abd<mode>_2): Likewise.
29134 (*aarch64_cond_<su>abd<mode>_any): Likewise.
29135 (@aarch64_pred_<su>abd<mode>): Likewise. Use UNSPEC_PRED_X
29136 for the max and min but not for the minus.
29137 (*aarch64_cond_<su>abd<mode>_3): New pattern.
29139 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
29141 * config/aarch64/iterators.md (SVE_24I): New iterator.
29142 * config/aarch64/aarch64-sve.md (*aarch64_adr<mode>_shift): Extend from
29143 SVE_FULL_SDI to SVE_24I. Use containers rather than elements.
29145 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
29147 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_BINARY:optab><mode>)
29148 (*cond_<SVE_INT_BINARY:optab><mode>_2): Extend from SVE_FULL_I
29150 (*cond_<SVE_INT_BINARY:optab><mode>_3): Likewise.
29151 (*cond_<SVE_INT_BINARY:optab><mode>_any): Likewise.
29152 (*cond_<SVE_INT_BINARY:optab><mode>_2_const): Likewise.
29153 (*cond_<SVE_INT_BINARY:optab><mode>_any_const): Likewise.
29155 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
29157 * config/aarch64/aarch64-sve.md (<SVE_INT_BINARY_IMM:optab><mode>3)
29158 (@aarch64_pred_<SVE_INT_BINARY_IMM:optab><mode>)
29159 (*post_ra_<SVE_INT_BINARY_IMM:optab><mode>3): Extend from SVE_FULL_I
29162 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
29164 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3)
29165 (v<ASHIFT:optab><mode>3, @aarch64_pred_<optab><mode>)
29166 (*post_ra_v<ASHIFT:optab><mode>3): Extend from SVE_FULL_I to SVE_I.
29168 2021-01-11 Martin Liska <mliska@suse.cz>
29171 * symtab-clones.h (clone_info::release): Release
29172 symtab::m_clones with ggc_delete as it's a GGC memory.
29174 2021-01-11 Matthias Klose <doko@ubuntu.com>
29176 * Makefile.in (LINK_PROGRESS): Show the link target.
29178 2021-01-11 Richard Biener <rguenther@suse.de>
29180 PR tree-optimization/91403
29181 * tree-vect-data-refs.c (vect_analyze_group_access_1): Cap
29182 single-element interleaving group size at 4096 elements.
29184 2021-01-11 Richard Biener <rguenther@suse.de>
29186 PR tree-optimization/98526
29187 * tree-vect-loop.c (vect_model_reduction_cost): Remove costing
29188 of the actual reduction op for the regular case.
29189 (vectorizable_reduction): Cost the stmts
29190 vect_transform_reduction produces here.
29192 2021-01-11 Andreas Krebbel <krebbel@linux.ibm.com>
29194 * tree-ssa-forwprop.c (simplify_vector_constructor): For
29195 big-endian, use UNPACK[_FLOAT]_HI.
29197 2021-01-11 Tamar Christina <tamar.christina@arm.com>
29199 * tree-vect-slp-patterns.c (class complex_pattern,
29200 class complex_add_pattern): Add parameters to matches.
29201 (complex_add_pattern::build): Free memory.
29202 (complex_add_pattern::matches): Move validation end of match.
29203 (complex_add_pattern::recognize): Likewise.
29205 2021-01-11 Tamar Christina <tamar.christina@arm.com>
29207 * tree-vect-slp-patterns.c (linear_loads_p): Fix externals.
29209 2021-01-11 Tamar Christina <tamar.christina@arm.com>
29211 * tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
29213 2021-01-11 Jakub Jelinek <jakub@redhat.com>
29215 PR tree-optimization/95867
29216 * tree-ssa-math-opts.h: New header.
29217 * tree-ssa-math-opts.c: Include tree-ssa-math-opts.h.
29218 (powi_as_mults): No longer static. Use build_one_cst instead of
29219 build_real. Formatting fix.
29220 * tree-ssa-reassoc.c: Include tree-ssa-math-opts.h.
29221 (attempt_builtin_powi): Handle multiplication reassociation without
29222 powi_fndecl using powi_as_mults.
29223 (reassociate_bb): For integral types don't require
29224 -funsafe-math-optimizations to call attempt_builtin_powi.
29226 2021-01-11 Jakub Jelinek <jakub@redhat.com>
29228 PR tree-optimization/95852
29229 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): Change
29230 mul_stmts parameter type to vec<gimple *> &. Before cond_stmt
29231 allow in the bb any of the stmts in that vector, div_stmt and
29232 up to 3 cast stmts.
29233 (arith_cast_equal_p): New function.
29234 (arith_overflow_check_p): Add cast_stmt argument, handle signed
29235 multiply overflow checks.
29236 (match_arith_overflow): Adjust caller. Handle signed multiply
29239 2021-01-11 Jakub Jelinek <jakub@redhat.com>
29241 PR tree-optimization/95852
29242 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): New function.
29243 (uaddsub_overflow_check_p): Renamed to ...
29244 (arith_overflow_check_p): ... this. Handle also multiplication
29245 with overflow check.
29246 (match_uaddsub_overflow): Renamed to ...
29247 (match_arith_overflow): ... this. Add cfg_changed argument. Handle
29248 also multiplication with overflow check. Adjust function comment.
29249 (math_opts_dom_walker::after_dom_children): Adjust callers. Call
29250 match_arith_overflow also for MULT_EXPR.
29252 2021-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29254 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
29255 __builtin_convertvector.
29256 (vmovl_s16): Likewise.
29257 (vmovl_s32): Likewise.
29258 (vmovl_u8): Likewise.
29259 (vmovl_u16): Likewise.
29260 (vmovl_u32): Likewise.
29261 (vmovn_s16): Likewise.
29262 (vmovn_s32): Likewise.
29263 (vmovn_s64): Likewise.
29264 (vmovn_u16): Likewise.
29265 (vmovn_u32): Likewise.
29266 (vmovn_u64): Likewise.
29268 2021-01-11 Martin Liska <mliska@suse.cz>
29270 * gimple-if-to-switch.cc (struct condition_info): Use auto_var.
29271 (if_chain::is_beneficial): Delete clusters
29272 (find_conditions): Make second argument of conditions_in_bbs a
29273 pointer so that we control over it's lifetime.
29274 (pass_if_to_switch::execute): Delete them.
29276 2021-01-11 Kewen Lin <linkw@linux.ibm.com>
29278 * ira.c (move_unallocated_pseudos): Check other_reg and skip if
29281 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
29283 * config/vax/vax.md (cc): Remove mode attribute.
29284 (subst_<cc>, subst_f<cc>): Rename to...
29285 (subst_<mode>, subst_f<VAXccnz:mode>): ... these respectively.
29286 (*cbranch<VAXint:mode>4_<VAXcc:mode>): Update for `cc' removal.
29287 (*cbranch<VAXfp:mode>4_<VAXccnz:mode>): Likewise.
29288 (*branch_<mode>, *branch_<mode>_reversed): Likewise.
29290 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
29292 * config/vax/vax.md (subst_f<cc>): Add mode to operands and
29293 `const_double_zero'.
29295 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
29297 * config/pdp11/pdp11.md (PDPfp): New mode iterator.
29298 (fcc_cc, fcc_ccnz): Use it. Add mode to `const_double_zero' and
29301 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
29303 * genemit.c (gen_exp) <CONST_DOUBLE>: Handle `const_double_zero'
29305 * read-rtl.c (rtx_reader::read_rtx_code): Handle machine mode
29306 with `const_double_zero'.
29307 * doc/rtl.texi (Constant Expression Types): Document it.
29309 2021-01-09 Jakub Jelinek <jakub@redhat.com>
29312 * tree-cfg.c (verify_gimple_assign_binary): Allow lhs of
29313 POINTER_DIFF_EXPR to be any integral type.
29315 2021-01-09 Jakub Jelinek <jakub@redhat.com>
29317 PR rtl-optimization/98603
29318 * function.c (instantiate_virtual_regs_in_insn): For asm goto
29319 with impossible constraints, drop all SETs, CLOBBERs, drop PARALLEL
29320 if any, set ASM_OPERANDS mode to VOIDmode and change
29321 ASM_OPERANDS_OUTPUT_CONSTRAINT and ASM_OPERANDS_OUTPUT_IDX.
29323 2021-01-09 Alexandre Oliva <oliva@gnu.org>
29326 * final.c (notice_source_line): Narrow down the condition to
29327 skip a line-0 marker.
29329 2021-01-08 Sergei Trofimovich <siarheit@google.com>
29331 * ipa-modref.c (merge_call_side_effects): Fix
29332 linebreak split by reordering two print calls.
29334 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
29336 * config/s390/vector.md (*tf_to_fprx2_0): Rename from
29337 "*mov_tf_to_fprx2_0" for consistency, fix constraint.
29338 (*tf_to_fprx2_1): Rename from "*mov_tf_to_fprx2_1" for
29339 consistency, fix constraint.
29341 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
29343 * config/s390/s390-c.c (s390_def_or_undef_macro): Accept
29344 callables instead of mask values.
29345 (struct target_flag_set_p): New predicate.
29346 (s390_cpu_cpp_builtins_internal): Define or undefine
29347 __LONG_DOUBLE_VX__ macro.
29349 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
29352 * config/i386/i386.c (x86_function_profiler): Use R10 and R11
29353 to call mcount in large model with PIC for NO_PROFILE_COUNTERS
29356 2021-01-08 Richard Biener <rguenther@suse.de>
29358 * tree-ssa-sccvn.c (pass_fre::execute): Reset the SCEV hash table.
29360 2021-01-08 Richard Biener <rguenther@suse.de>
29362 * tree-vect-slp.c (scalar_stmts_to_slp_tree_map_t): Fix.
29363 (vect_build_slp_tree): On cache hit release the matched
29364 scalar stmts vector.
29365 * tree-vect-stmts.c (vectorizable_store): Properly free
29366 vec_oprnds before possibly gathering them again.
29368 2021-01-08 Richard Biener <rguenther@suse.de>
29370 PR tree-optimization/98544
29371 * tree-vect-slp.c (vect_optimize_slp): Always materialize
29372 permutes at a permute node.
29374 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
29377 * config/i386/i386.c (x86_function_profiler): Use R10 to call
29378 mcount in large model. Sorry for large model with PIC.
29380 2021-01-08 Jakub Jelinek <jakub@redhat.com>
29383 * config/i386/i386.opt (ix86_cmodel, ix86_incoming_stack_boundary_arg,
29384 ix86_pmode, ix86_preferred_stack_boundary_arg, ix86_regparm,
29385 ix86_veclibabi_type): Remove x_ prefix, use TargetVariable instead of
29386 TargetSave and initialize for variables with enum types.
29387 (mfentry, mstack-protector-guard-reg=, mstack-protector-guard-offset=,
29388 mstack-protector-guard-symbol=): Add Save.
29389 * config/i386/i386-options.c (ix86_function_specific_save,
29390 ix86_function_specific_restore): Don't save or restore x_ix86_cmodel,
29391 x_ix86_incoming_stack_boundary_arg, x_ix86_pmode,
29392 x_ix86_preferred_stack_boundary_arg, x_ix86_regparm,
29393 x_ix86_veclibabi_type.
29395 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
29397 * config/aarch64/aarch64-sve.md (*cnot<mode>): Extend from
29398 SVE_FULL_I to SVE_I.
29399 (*cond_cnot<mode>_2, *cond_cnot<mode>_any): Likewise.
29401 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
29403 * config/aarch64/aarch64-sve.md (*cond_uxt<mode>_2): Extend from
29404 SVE_FULL_I to SVE_I.
29405 (*cond_uxt<mode>_any): Likewise.
29407 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29409 * config/aarch64/iterators.md (Vwhalf): New iterator.
29410 * config/aarch64/aarch64-simd.md (aarch64_<sur>adalp<mode>_3):
29412 (aarch64_<sur>adalp<mode>): ... This. Make more
29414 (<sur>sadv16qi): Adjust callsite of the above.
29415 * config/aarch64/aarch64-simd-builtins.def (sadalp, uadalp): New
29417 * config/aarch64/arm_neon.h (vpadal_s8): Reimplement using
29419 (vpadal_s16): Likewise.
29420 (vpadal_u8): Likewise.
29421 (vpadal_u16): Likewise.
29422 (vpadalq_s8): Likewise.
29423 (vpadalq_s16): Likewise.
29424 (vpadalq_s32): Likewise.
29425 (vpadalq_u8): Likewise.
29426 (vpadalq_u16): Likewise.
29427 (vpadalq_u32): Likewise.
29429 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29431 * config/aarch64/aarch64-simd.md (aarch64_<su>abd<mode>_3):
29433 (aarch64_<su>abd<mode>): ... This.
29434 (<sur>sadv16qi): Adjust callsite of the above.
29435 * config/aarch64/aarch64-simd-builtins.def (sabd, uabd): Define
29437 * config/aarch64/arm_neon.h (vabd_s8): Reimplement using
29439 (vabd_s16): Likewise.
29440 (vabd_s32): Likewise.
29441 (vabd_u8): Likewise.
29442 (vabd_u16): Likewise.
29443 (vabd_u32): Likewise.
29444 (vabdq_s8): Likewise.
29445 (vabdq_s16): Likewise.
29446 (vabdq_s32): Likewise.
29447 (vabdq_u8): Likewise.
29448 (vabdq_u16): Likewise.
29449 (vabdq_u32): Likewise.
29451 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29453 * config/aarch64/aarch64-simd-builtins.def (saba, uaba): Define
29455 * config/aarch64/arm_neon.h (vaba_s8): Implement using builtin.
29456 (vaba_s16): Likewise.
29457 (vaba_s32): Likewise.
29458 (vaba_u8): Likewise.
29459 (vaba_u16): Likewise.
29460 (vaba_u32): Likewise.
29461 (vabaq_s8): Likewise.
29462 (vabaq_s16): Likewise.
29463 (vabaq_s32): Likewise.
29464 (vabaq_u8): Likewise.
29465 (vabaq_u16): Likewise.
29466 (vabaq_u32): Likewise.
29468 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29470 * config/aarch64/aarch64-simd.md (aba<mode>_3): Rename to...
29471 (aarch64_<su>aba<mode>): ... This. Handle uaba as well.
29472 Change RTL pattern to match.
29474 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
29476 * common/config/riscv/riscv-common.c (riscv_current_subset_list): New.
29477 * config/riscv/riscv-c.c (riscv-subset.h): New.
29478 (INCLUDE_STRING): Define.
29479 (riscv_cpu_cpp_builtins): Add new style architecture extension
29481 * config/riscv/riscv-subset.h (riscv_subset_list::begin): New.
29482 (riscv_subset_list::end): New.
29483 (riscv_current_subset_list): New.
29485 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
29487 * common/config/riscv/riscv-common.c (RISCV_DONT_CARE_VERSION):
29488 Move to riscv-subset.h.
29489 (struct riscv_subset_t): Ditto.
29490 (class riscv_subset_list): Ditto.
29491 * config/riscv/riscv-subset.h (RISCV_DONT_CARE_VERSION): Move
29492 from riscv-common.c.
29493 (struct riscv_subset_t): Ditto.
29494 (class riscv_subset_list): Ditto.
29495 * config/riscv/t-riscv ($(common_out_file)): Add file
29498 2021-01-07 Jakub Jelinek <jakub@redhat.com>
29501 * config/i386/i386.md (*bmi_blsi_<mode>_cmp, *bmi_blsi_<mode>_ccno):
29502 New define_insn patterns.
29504 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
29506 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_UNARY:optab><mode>)
29507 (*cond_<SVE_INT_UNARY:optab><mode>_2): Extend from SVE_FULL_I to SVE_I.
29508 (*cond_<SVE_INT_UNARY:optab><mode>_any): Likewise.
29510 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
29512 PR tree-optimization/98560
29513 * internal-fn.def (IFN_VCONDU, IFN_VCONDEQ): Use type vec_cond.
29514 * internal-fn.c (vec_cond_mask_direct): Get the data mode from
29516 (vec_cond_direct): Likewise argument 2.
29517 (vec_condu_direct, vec_condeq_direct): Delete.
29518 (expand_vect_cond_optab_fn): Rename to...
29519 (expand_vec_cond_optab_fn): ...this, replacing old macro.
29520 (expand_vec_condu_optab_fn, expand_vec_condeq_optab_fn): Delete.
29521 (expand_vect_cond_mask_optab_fn): Rename to...
29522 (expand_vec_cond_mask_optab_fn): ...this, replacing old macro.
29523 (direct_vec_cond_mask_optab_supported_p): Treat the optab as a
29525 (direct_vec_cond_optab_supported_p): Likewise.
29526 (direct_vec_condu_optab_supported_p): Delete.
29527 (direct_vec_condeq_optab_supported_p): Delete.
29528 * gimple-isel.cc: Include internal-fn.h.
29529 (gimple_expand_vec_cond_expr): Check that IFN_VCONDEQ is supported
29532 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
29534 PR tree-optimization/98560
29535 * gimple-isel.cc (gimple_expand_vec_cond_expr): If we fail to use
29536 IFN_VCOND{,U,EQ}, fall back on IFN_VCOND_MASK.
29538 2021-01-07 Uroš Bizjak <ubizjak@gmail.com>
29540 * config/i386/i386.md (insn): Merge from plusminus_insn, shift_insn,
29541 rotate_insn and optab code attributes.
29542 Update all uses to merged code attribute.
29543 * config/i386/sse.md: Update all uses to merged code attribute.
29544 * config/i386/mmx.md: Update all uses to merged code attribute.
29546 2021-01-07 Jakub Jelinek <jakub@redhat.com>
29548 PR tree-optimization/98568
29549 * gimple-ssa-store-merging.c (bswap_view_convert): New function.
29550 (bswap_replace): Use it.
29552 2021-01-06 Vladimir N. Makarov <vmakarov@redhat.com>
29554 PR rtl-optimization/97978
29555 * lra-int.h (lra_hard_reg_split_p): New external.
29556 * lra.c (lra_hard_reg_split_p): New global.
29557 (lra): Set up lra_hard_reg_split_p after splitting a hard reg.
29558 * lra-assigns.c (lra_assign): Don't check allocation correctness
29559 after hard reg splitting.
29561 2021-01-06 Martin Sebor <msebor@redhat.com>
29564 * builtins.c (new_delete_mismatch_p): New overload.
29565 (new_delete_mismatch_p (tree, tree)): Call it.
29567 2021-01-06 Alexandre Oliva <oliva@adacore.com>
29569 * Makefile.in (T_GLIMITS_H): New.
29570 (stmp-int-hdrs): Depend on it, use it.
29571 * config/t-vxworks (T_GLIMITS_H): Override it.
29572 (vxw-glimits.h): New.
29574 2021-01-06 Richard Biener <rguenther@suse.de>
29576 PR tree-optimization/98513
29577 * value-range.cc (intersect_ranges): Compare the upper bounds
29578 for the expected relation.
29580 2021-01-06 Gerald Pfeifer <gerald@pfeifer.com>
29583 2020-12-28 Gerald Pfeifer <gerald@pfeifer.com>
29585 * doc/standards.texi (HSAIL): Remove section.
29587 2021-01-05 Samuel Thibault <samuel.thibault@ens-lyon.org>
29589 * configure: Re-generate.
29591 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29593 * doc/invoke.texi (-std=c++20): Adjust for the publication of
29594 ISO 14882:2020 standard.
29595 * doc/standards.texi: Likewise.
29597 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29599 PR tree-optimization/94802
29600 * expr.h (maybe_optimize_sub_cmp_0): Declare.
29601 * expr.c: Include tree-pretty-print.h and flags.h.
29602 (maybe_optimize_sub_cmp_0): New function.
29603 (do_store_flag): Use it.
29604 * cfgexpand.c (expand_gimple_cond): Likewise.
29606 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
29608 * mux-utils.h (pointer_mux::m_ptr): Tweak description of contents.
29609 * rtlanal.c (simple_regno_set): Tweak description to clarify the
29612 2021-01-05 Richard Biener <rguenther@suse.de>
29614 PR tree-optimization/98516
29615 * tree-vect-slp.c (vect_optimize_slp): Permute the incoming
29616 lanes when materializing on a VEC_PERM node.
29617 (vectorizable_slp_permutation): Dump the permute properly.
29619 2021-01-05 Richard Biener <rguenther@suse.de>
29621 * tree-vect-slp.c (vect_slp_region): Move debug counter
29622 to cover individual subgraphs.
29624 2021-01-05 Richard Biener <rguenther@suse.de>
29626 PR tree-optimization/98428
29627 * tree-vect-slp.c (vect_build_slp_tree_1): Properly reject
29628 vector lane extracts for loop vectorization.
29630 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29632 PR tree-optimization/98514
29633 * tree-ssa-reassoc.c (bb_rank): Change type from long * to
29635 (operand_rank): Change type from hash_map<tree, long> to
29636 hash_map<tree, int64_t>.
29637 (phi_rank): Change return type from long to int64_t.
29638 (loop_carried_phi): Change block_rank variable type from long to
29640 (propagate_rank): Change return type, rank parameter type and
29641 op_rank variable type from long to int64_t.
29642 (find_operand_rank): Change return type from long to int64_t
29643 and change slot variable type from long * to int64_t *.
29644 (insert_operand_rank): Change rank parameter type from long to
29646 (get_rank): Change return type and rank variable type from long to
29647 int64_t. Use PRId64 instead of ld to print the rank.
29648 (init_reassoc): Change rank variable type from long to int64_t
29649 and adjust correspondingly bb_rank and operand_rank initialization.
29651 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29653 PR tree-optimization/96928
29654 * tree-ssa-phiopt.c (xor_replacement): New function.
29655 (tree_ssa_phiopt_worker): Call it.
29657 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29659 PR tree-optimization/96930
29660 * match.pd ((A / (1 << B)) -> (A >> B)): If A is extended
29661 from narrower value which has the same type as 1 << B, perform
29662 the right shift on the narrower value followed by extension.
29664 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29666 PR tree-optimization/96239
29667 * gimple-ssa-store-merging.c (maybe_optimize_vector_constructor): New
29669 (get_status_for_store_merging): Don't return BB_INVALID for blocks
29670 with potential bswap optimizable CONSTRUCTORs.
29671 (pass_store_merging::execute): Optimize vector CONSTRUCTORs with bswap
29674 2021-01-05 Richard Biener <rguenther@suse.de>
29676 PR tree-optimization/98381
29677 * tree.c (vector_element_bits): Properly compute bool vector
29679 * tree-vect-loop.c (vectorizable_live_operation): Properly
29680 compute the last lane bit offset.
29682 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
29685 * config/i386/sse.md (sse_cvtps2pi): Redefine as define_insn_and_split.
29686 Clear the top 64 bytes of the input XMM register.
29687 (sse_cvttps2pi): Ditto.
29689 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
29692 * config/i386/xopintrin.h (_mm256_cmov_si256): New.
29694 2021-01-05 H.J. Lu <hjl.tools@gmail.com>
29697 * config/i386/xmmintrin.h (_mm_extract_pi16): Cast to unsigned
29700 2021-01-05 Claudiu Zissulescu <claziss@synopsys.com>
29702 * config/arc/arc.md (maddsidi4_split): Use ACC_REG_FIRST.
29703 (umaddsidi4_split): Likewise.
29705 2021-01-05 liuhongt <hongtao.liu@intel.com>
29708 * config/i386/sse.md (*sse2_pmovskb_zexthisi): New
29709 define_insn_and_split for zero_extend of subreg HI of pmovskb
29711 (*sse2_pmovskb_zexthisi): Add new combine splitters for
29712 zero_extend of not of subreg HI of pmovskb result.
29714 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
29717 * explow.c (convert_memory_address_addr_space_1): Handle UNSPECs
29719 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate): Use
29720 convert_memory_address to convert symbolic immediates to ptr_mode
29721 before forcing them to memory.
29723 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
29725 PR rtl-optimization/97144
29726 * recog.c (constrain_operands): Initialize matching_operand
29727 for each alternative, rather than only doing it once.
29729 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
29731 PR rtl-optimization/98403
29732 * rtl-ssa/changes.cc (function_info::finalize_new_accesses): Explain
29733 why we don't remove call clobbers.
29734 (function_info::apply_changes_to_insn): Don't attempt to add
29735 call clobbers here.
29737 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
29739 PR tree-optimization/98371
29740 * tree-vect-loop.c (vect_reanalyze_as_main_loop): New function.
29741 (vect_analyze_loop): If an epilogue loop appears to be cheaper
29742 than the main loop, re-analyze it as a main loop before adopting
29745 2021-01-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
29748 * configure.ac (NETLIBS): Determine using AX_LIB_SOCKET_NSL.
29749 * aclocal.m4, configure: Regenerate.
29750 * Makefile.in (NETLIBS): Define.
29751 (BACKEND): Remove $(CODYLIB).
29753 2021-01-05 Jakub Jelinek <jakub@redhat.com>
29755 PR rtl-optimization/98334
29756 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
29757 Optimize (X - 1) * Y + Y to X * Y or (X + 1) * Y - Y to X * Y.
29759 2021-01-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
29761 * tree-inline.c (expand_call_inline): Restore input_location.
29762 Return result from recursive call.
29764 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
29766 PR tree-optimization/95401
29767 * config/aarch64/aarch64-sve-builtins.cc
29768 (gimple_folder::load_store_cookie): Use bits rather than bytes
29769 for the alignment argument to IFN_MASK_LOAD and IFN_MASK_STORE.
29770 * gimple-fold.c (gimple_fold_mask_load_store_mem_ref): Likewise.
29771 * tree-vect-stmts.c (vectorizable_store): Likewise.
29772 (vectorizable_load): Likewise.
29774 2021-01-04 Richard Biener <rguenther@suse.de>
29776 PR tree-optimization/98308
29777 * tree-vect-stmts.c (vectorizable_load): Set invariant mask
29780 2021-01-04 Jakub Jelinek <jakub@redhat.com>
29782 PR tree-optimization/95771
29783 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Handle types
29784 with precision smaller than int's precision and types with precision
29785 twice as large as long long. Formatting fixes.
29787 2021-01-04 Richard Biener <rguenther@suse.de>
29789 PR tree-optimization/98464
29790 * tree-ssa-sccvn.c (vn_valueize_for_srt): Rename from ...
29791 (vn_valueize_wrapper): ... this. Temporarily adjust vn_context_bb.
29792 (process_bb): Adjust.
29794 2021-01-04 Matthew Malcomson <matthew.malcomson@arm.com>
29797 * doc/invoke.texi (-fsanitize=address): Fix wording describing
29798 clash with -fsanitize=hwaddress.
29800 2021-01-04 Richard Biener <rguenther@suse.de>
29802 PR tree-optimization/98282
29803 * tree-ssa-sccvn.c (vn_get_stmt_kind): Classify tcc_reference on
29804 invariants as VN_NARY.
29806 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
29809 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Accept
29810 aarch64_simd_reg_or_zero for operand 2. Use the combinez patterns
29811 to handle zero operands.
29813 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
29815 * config/aarch64/aarch64.c (offset_6bit_signed_scaled_p): New function.
29816 (offset_6bit_unsigned_scaled_p): Fix typo in comment.
29817 (aarch64_sve_prefetch_operand_p): Accept MUL VLs in the range
29820 2021-01-04 Richard Biener <rguenther@suse.de>
29822 PR tree-optimization/98393
29823 * tree-vect-slp.c (vect_build_slp_tree): Properly zero matches
29824 when hitting the limit.
29826 2021-01-04 Richard Biener <rguenther@suse.de>
29828 PR tree-optimization/98291
29829 * tree-vect-loop.c (vectorizable_reduction): Bypass
29830 associativity check for SLP reductions with VF 1.
29832 2021-01-04 Jakub Jelinek <jakub@redhat.com>
29834 PR tree-optimization/96782
29835 * match.pd (x == ~x -> false, x != ~x -> true): New simplifications.
29837 2021-01-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
29839 * collect-utils.c (collect_execute): Check dumppfx.
29840 * collect2.c (maybe_run_lto_and_relink, do_link): Pass atsuffix
29841 to collect_execute.
29842 (do_link): Add new parameter atsuffix.
29843 (main): Handle -dumpdir option. Skip one argument for
29844 -o, -isystem and -B options.
29845 * gcc.c (make_at_file): New helper function.
29846 (close_at_file): Use it.
29848 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29850 * config/darwin.h (MIN_LD64_NO_COAL_SECTS): Adjust.
29851 Amend handling for LD64_VERSION fallback defaults.
29853 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29855 * config.gcc: Compute default version information
29856 from the configured target. Likewise defaults for
29858 * config/darwin10.h: Removed.
29859 * config/darwin12.h: Removed.
29860 * config/darwin9.h: Removed.
29861 * config/rs6000/darwin8.h: Removed.
29863 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29865 * config/darwin9.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
29867 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29869 * config/darwin9.h (STACK_CHECK_STATIC_BUILTIN): Move from here..
29870 * config/darwin.h (STACK_CHECK_STATIC_BUILTIN): .. to here.
29872 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29874 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move from
29876 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): ... to here.
29878 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29880 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move the spec
29881 for the Darwin10 unwinder stub from here ...
29882 * config/darwin.h (LINK_COMMAND_SPEC_A): ... to here.
29884 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
29886 * config/darwin.h (DSYMUTIL_SPEC): Default to DWARF
29887 (ASM_DEBUG_SPEC):Only define if the assembler supports
29889 (PREFERRED_DEBUGGING_TYPE): Default to DWARF.
29890 (DARWIN_PREFER_DWARF): Define.
29891 * config/darwin9.h (PREFERRED_DEBUGGING_TYPE): Remove.
29892 (DARWIN_PREFER_DWARF): Likewise
29893 (DSYMUTIL_SPEC): Likewise.
29894 (COLLECT_RUN_DSYMUTIL): Likewise.
29895 (ASM_DEBUG_SPEC): Likewise.
29896 (ASM_DEBUG_OPTION_SPEC): Likewise.
29898 2021-01-02 Jan Hubicka <jh@suse.cz>
29900 * cfg.c (free_block): ggc_free bb.
29902 2021-01-01 Jakub Jelinek <jakub@redhat.com>
29904 * gcc.c (process_command): Update copyright notice dates.
29905 * gcov-dump.c (print_version): Ditto.
29906 * gcov.c (print_version): Ditto.
29907 * gcov-tool.c (print_version): Ditto.
29908 * gengtype.c (create_file): Ditto.
29909 * doc/cpp.texi: Bump @copying's copyright year.
29910 * doc/cppinternals.texi: Ditto.
29911 * doc/gcc.texi: Ditto.
29912 * doc/gccint.texi: Ditto.
29913 * doc/gcov.texi: Ditto.
29914 * doc/install.texi: Ditto.
29915 * doc/invoke.texi: Ditto.
29917 2021-01-01 Jakub Jelinek <jakub@redhat.com>
29919 * ChangeLog-2020: Rotate ChangeLog. New file.
29922 Copyright (C) 2021 Free Software Foundation, Inc.
29924 Copying and distribution of this file, with or without modification,
29925 are permitted in any medium without royalty provided the copyright
29926 notice and this notice are preserved.