1 2021-11-25 Jan Hubicka <jh@suse.cz>
3 * ipa-cp.c (ipa_value_range_from_jfunc): Remove forgotten early return.
5 2021-11-25 Roger Sayle <roger@nextmovesoftware.com>
8 * match.pd (minus @0 @0): Check tree_expr_maybe_infinite_p.
10 2021-11-25 Martin Jambor <mjambor@suse.cz>
13 * ipa-prop.h (ipa_get_param): New overload. Move bits of the existing
15 * ipa-param-manipulation.h (ipa_param_adjustments): New member
16 function get_updated_index_or_split.
17 * ipa-param-manipulation.c
18 (ipa_param_adjustments::get_updated_index_or_split): New function.
19 * ipa-prop.c (adjust_agg_replacement_values): Reimplement, add
20 capability to identify scalarized parameters and perform substitution
22 (ipcp_transform_function): Create descriptors earlier, handle new
23 return values of adjust_agg_replacement_values.
25 2021-11-25 Aldy Hernandez <aldyh@redhat.com>
27 * gimple-range-path.cc (path_range_query::compute_ranges_defined): Remove.
28 (path_range_query::compute_ranges_in_block): Revert to bitmap order.
29 * gimple-range-path.h: Remove compute_ranges_defined.
31 2021-11-25 Andrew Stubbs <ams@codesourcery.com>
34 * config/gcn/gcn.c (move_callee_saved_registers): Ensure that the
35 number of spilled registers is counted correctly.
37 2021-11-25 Jan Hubicka <hubicka@ucw.cz>
39 * ipa-prop.h (ipa_node_params::ipa_node_params): Initialize
42 2021-11-25 Andrew MacLeod <amacleod@redhat.com>
44 PR tree-optimization/103359
45 * gimple-range-fold.cc (fold_using_range::range_of_phi): If arg is
46 equivalent to def, don't initially include it's range.
48 2021-11-25 Jan Hubicka <hubicka@ucw.cz>
50 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Do not check
53 2021-11-25 Richard Biener <rguenther@suse.de>
55 * cgraphunit.c (symbol_table::output_weakrefs): Remove
57 (get_alias_symbol): Remove now unused function.
59 2021-11-25 Richard Biener <rguenther@suse.de>
61 * cfgrtl.c (rtl_verify_fallthru): Do not stop verifying
63 (skip_insns_after_block): Remove unreachable break and continue.
65 2021-11-25 Richard Biener <rguenther@suse.de>
67 * cfgexpand.c (label_rtx_for_bb): Remove dead loop construct.
69 2021-11-25 Richard Biener <rguenther@suse.de>
71 * regset.h (REG_SET_EMPTY_P): New macro.
72 * cfgcleanup.c (thread_jump): Use REG_SET_EMPTY_P.
74 2021-11-25 Martin Liska <mliska@suse.cz>
76 * doc/invoke.texi: Use @option for -Wuninitialized.
78 2021-11-25 Aldy Hernandez <aldyh@redhat.com>
80 PR tree-optimization/103254
81 * gimple-range-path.cc (path_range_query::compute_ranges): Move
82 exported boolean code...
83 (path_range_query::compute_imports): ...here.
85 2021-11-25 Aldy Hernandez <aldyh@redhat.com>
87 PR tree-optimization/103254
88 * gimple-range-path.cc (path_range_query::compute_ranges_defined): New
89 (path_range_query::compute_ranges_in_block): Move to
90 compute_ranges_defined.
91 * gimple-range-path.h (compute_ranges_defined): New.
93 2021-11-25 Jakub Jelinek <jakub@redhat.com>
95 PR tree-optimization/103417
96 * match.pd ((X & Y) CMP 0): Only handle eq and ne. Commonalize
99 2021-11-25 Jakub Jelinek <jakub@redhat.com>
101 PR tree-optimization/103376
102 * gimple-ssa-store-merging.c (perform_symbolic_merge): For
103 BIT_IOR_EXPR, if masked1 && masked2 && masked1 != masked2, don't
104 punt, but set the corresponding result byte to MARKER_BYTE_UNKNOWN.
105 For BIT_XOR_EXPR similarly and if masked1 == masked2 and the
106 byte isn't MARKER_BYTE_UNKNOWN, set the corresponding result byte to
109 2021-11-25 Jakub Jelinek <jakub@redhat.com>
112 * doc/invoke.texi (-Wcomma-subscript): Document that for
113 -std=c++20 the option isn't enabled by default with -Wno-deprecated
114 but for -std=c++23 it is.
116 2021-11-24 Jan Hubicka <hubicka@ucw.cz>
118 * ipa-modref.c (implicit_eaf_flags_for_edge_and_arg): Break out from...
119 (modref_merge_call_site_flags): ... here.
120 (ipa_merge_modref_summary_after_inlining): Use it.
122 2021-11-24 Thomas Schwinge <thomas@codesourcery.com>
124 * cfgloop.c (verify_loop_structure): Reduce scope of
125 'class loop *loop' variable.
126 * ipa-fnsummary.c (analyze_function_body): Likewise.
127 * loop-init.c (fix_loop_structure): Likewise.
128 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
129 * predict.c (predict_loops): Likewise.
130 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
131 * tree-vectorizer.c (pass_vectorize::execute): Likewise.
133 2021-11-24 Andrew MacLeod <amacleod@redhat.com>
135 PR tree-optimization/103231
136 * gimple-range.cc (gimple_ranger::gimple_ranger): Create stmt stack.
137 (gimple_ranger::gimple_ranger): Delete stmt stack.
138 (gimple_ranger::range_of_stmt): Process depenedencies if they have no
140 (gimple_ranger::prefill_name): New.
141 (gimple_ranger::prefill_stmt_dependencies): New.
142 * gimple-range.h (class gimple_ranger): Add prototypes.
144 2021-11-24 Andrew MacLeod <amacleod@redhat.com>
146 * gimple-range-cache.cc (ranger_cache::get_global_range): Always
147 return a range, return if it came from the cache or not.
148 (get_non_stale_global_range): Rename to get_global_range, and return
149 the temporal state in a flag.
150 * gimple-range-cache.h (get_non_stale_global_range): Rename and adjust.
151 * gimple-range.cc (gimple_ranger::range_of_expr): No need to query
153 (gimple_ranger::range_of_stmt): Adjust for global cache temporal state
156 2021-11-24 Andrew MacLeod <amacleod@redhat.com>
158 * gimple-range.cc (gimple_ranger::range_on_edge): Call trailer when
159 a constant is encountered to terminate the trace.
161 2021-11-24 Maciej W. Rozycki <macro@embecosm.com>
164 * reload.c (find_reloads_address_1): Also accept the ASHIFT form
165 of indexed addressing.
166 (find_reloads): Adjust accordingly.
168 2021-11-24 Richard Biener <rguenther@suse.de>
169 Jan Hubicka <jh@suse.cz>
171 PR tree-optimization/103168
172 * ipa-modref.h (struct modref_summary): Add load_accesses.
173 * ipa-modref.c (modref_summary::finalize): Initialize load_accesses.
174 * tree-ssa-sccvn.c (visit_reference_op_call): Use modref
175 info to walk the virtual use->def chain to CSE const/pure
176 function calls possibly reading from memory.
178 2021-11-24 Thomas Schwinge <thomas@codesourcery.com>
181 * gimplify.c (gimplify_adjust_omp_clauses): Restore previous
184 2021-11-24 Richard Biener <rguenther@suse.de>
187 * match.pd: Avoid canonicalizing (le/ge @0 @0) to (eq @0 @0)
188 with NaNs and -ftrapping-math.
190 2021-11-24 Jakub Jelinek <jakub@redhat.com>
193 * omp-general.c (omp_context_selector_matches): For ACCEL_COMPILER,
194 return 0 for kind(host) and continue for kind(nohost).
196 2021-11-24 Jakub Jelinek <jakub@redhat.com>
199 * attribs.h (lookup_attribute): Allow attr_name to start with
200 underscore, as long as canonicalize_attr_name returns false.
201 (lookup_attribute_by_prefix): Don't call get_attribute_name twice.
202 * attribs.c (extract_attribute_substring): Reimplement using
203 canonicalize_attr_name.
204 (register_scoped_attribute): Change gcc_assert into
205 gcc_checking_assert, verify !canonicalize_attr_name rather than
206 that str.str doesn't start with '_'.
208 2021-11-24 Jakub Jelinek <jakub@redhat.com>
210 PR tree-optimization/103376
211 * gimple-ssa-store-merging.c (perform_symbolic_merge): Add CODE
212 argument. If CODE is not BIT_IOR_EXPR, ensure that one of masked1
214 (find_bswap_or_nop_1, find_bswap_or_nop,
215 imm_store_chain_info::try_coalesce_bswap): Adjust
216 perform_symbolic_merge callers.
218 2021-11-24 Richard Biener <rguenther@suse.de>
220 * tree-ssa-loop-ivopts.c (find_givs): Take loop body as
221 argument instead of re-computing it.
222 (find_interesting_uses): Likewise.
223 (find_induction_variables): Pass through loop body.
224 (tree_ssa_iv_optimize_loop): Pass down loop body.
226 2021-11-24 Tamar Christina <tamar.christina@arm.com>
228 * tree-ssa-phiopt.c (spaceship_replacement): Handle new canonical
231 2021-11-24 Tamar Christina <tamar.christina@arm.com>
233 * tree.c (bitmask_inv_cst_vector_p): New.
234 * tree.h (bitmask_inv_cst_vector_p): New.
235 * match.pd: Use it in new bitmask compare pattern.
237 2021-11-24 Jason Merrill <jason@redhat.com>
239 * timevar.h (class auto_cond_timevar): New.
241 2021-11-24 Hongtao Liu <hongtao.liu@intel.com>
242 H.J. Lu <hongjiu.lu@intel.com>
244 PR tree-optimization/103194
245 * match.pd (gimple_nop_atomic_bit_test_and_p): Extended to
247 * tree-ssa-ccp.c (gimple_nop_convert): Declare.
248 (optimize_atomic_bit_test_and): Enhance
249 optimize_atomic_bit_test_and to handle truncation.
251 2021-11-23 Martin Sebor <msebor@redhat.com>
254 * Makefile.in (OBJS): Add gimple-warn-recursion.o.
255 * common.opt: Add -Winfinite-recursion.
256 * doc/invoke.texi (-Winfinite-recursion): Document.
257 * passes.def (pass_warn_recursion): Schedule a new pass.
258 * tree-pass.h (make_pass_warn_recursion): Declare.
259 * gimple-warn-recursion.c: New file.
261 2021-11-23 Bill Schmidt <wschmidt@linux.ibm.com>
263 * config/rs6000/rs6000-call.c (rs6000_invalid_new_builtin): Change
264 error messages for ENB_P8V and ENB_P9V.
266 2021-11-23 Bill Schmidt <wschmidt@linux.ibm.com>
268 * config/rs6000/rs6000-builtin-new.def: Add power6-64 stanza. Move
269 CMPB to power6-64 stanza.
270 * config/rs6000/rs6000-call.c (rs6000_invalid_new_builtin): Handle
272 (rs6000_new_builtin_is_supported): Likewise.
273 (rs6000_expand_new_builtin): Likewise. Clean up formatting.
274 (rs6000_init_builtins): Handle ENB_P6_64 case.
275 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): Add BSTZ_P6_64.
276 (stanza_map): Add entry mapping power6-64 to BSTZ_P6_64.
277 (enable_string): Add "ENB_P6_64".
278 (write_decls): Add ENB_P6_64 to bif_enable enum.
280 2021-11-23 Jan Hubicka <jh@suse.cz>
282 * ipa-modref-tree.h (struct modref_tree): Remove max_bases, max_refs
284 (modref_tree::modref_tree): Remove parametr.
285 (modref_tree::insert_base): Add max_bases parameter.
286 (modref_tree::insert): Add max_bases, max_refs, max_accesses
288 (modref_tree::insert): New member function.
289 (modref_tree::merge): Add max_bases, max_refs, max_accesses
291 (modref_tree::insert): New member function.
292 * ipa-modref-tree.c (test_insert_search_collapse): Update.
293 (test_merge): Update.
294 * ipa-modref.c (dump_records): Don't dump max_refs and max_bases.
295 (dump_lto_records): Likewise.
296 (modref_summary::finalize): Fix whitespace.
297 (get_modref_function_summary): Likewise.
298 (modref_access_analysis::record_access): Update.
299 (modref_access_analysis::record_access_lto): Update.
300 (modref_access_analysis::process_fnspec): Update.
301 (analyze_function): Update.
302 (modref_summaries::duplicate): Update.
303 (modref_summaries_lto::duplicate): Update.
304 (write_modref_records): Update.
305 (read_modref_records): Update.
306 (read_section): Update.
307 (propagate_unknown_call): Update.
308 (modref_propagate_in_scc): Update.
309 (ipa_merge_modref_summary_after_inlining): Update.
311 2021-11-23 Martin Liska <mliska@suse.cz>
313 * doc/invoke.texi: Remove 2 more duplicite param descriptions.
315 2021-11-23 Richard Biener <rguenther@suse.de>
317 PR tree-optimization/103361
318 * gimple-loop-jam.c (adjust_unroll_factor): Use lambda_int
319 for the dependence distance.
320 * tree-data-ref.c (print_lambda_vector): Properly print a lambda_int.
322 2021-11-23 Jakub Jelinek <jakub@redhat.com>
324 * tree-inline.h (struct copy_body_data): Remove
325 transform_lang_insert_block member.
326 * tree-inline.c (remap_block): Don't call
327 id->transform_lang_insert_block.
328 (optimize_inline_calls, copy_gimple_seq_and_replace_locals,
329 tree_function_versioning, maybe_inline_call_in_expr,
330 copy_fn): Don't initialize id.transform_lang_insert_block.
331 * gimplify.c (gimplify_omp_loop): Likewise.
333 2021-11-23 Jan Hubicka <hubicka@ucw.cz>
335 PR tree-optimization/103335
336 * tree-ssa-dse.c (valid_ao_ref_for_dse): Rename to ...
337 (valid_ao_ref_kill_for_dse): ... this; do not check that boundaries
338 are divisible by BITS_PER_UNIT.
339 (get_byte_aligned_range_containing_ref): New function.
340 (get_byte_aligned_range_contained_in_ref): New function.
341 (normalize_ref): Rename to ...
342 (get_byte_range): ... this one; handle accesses not aligned to byte
343 boundary; return range in bytes rater than updating ao_ref.
344 (clear_live_bytes_for_ref): Take write ref by reference; simplify using
346 (setup_live_bytes_from_ref): Likewise.
347 (clear_bytes_written_by): Update.
348 (live_bytes_read): Update.
349 (dse_classify_store): Simplify tech before live_bytes_read checks.
351 2021-11-23 Andrew Pinski <apinski@marvell.com>
353 PR tree-optimization/102216
354 * gimple-fold.c (fold_stmt_1): Add canonicalization
355 of "&MEM[ssa_n, CST]" to "ssa_n p+ CST", note this
356 can only be done if !in_place.
358 2021-11-23 Jakub Jelinek <jakub@redhat.com>
361 * gimplify.c (replace_reduction_placeholders): Remove.
362 (note_no_context_vars): New function.
363 (gimplify_omp_loop): For OMP_PARALLEL's BIND_EXPR create a new
364 BLOCK. Use copy_tree_body_r with walk_tree instead of unshare_expr
365 and replace_reduction_placeholders for duplication of
366 OMP_CLAUSE_REDUCTION_{INIT,MERGE} expressions. Ensure all mentioned
367 automatic vars have DECL_CONTEXT set to non-NULL before doing so
368 and reset it afterwards for those vars and their corresponding
371 2021-11-23 Haochen Gui <guihaoc@gcc.gnu.org>
374 * config/rs6000/altivec.md (altivec_vreve<mode>2 for VEC_K): Use
375 xxbrq for v16qi, xxbrq + xxbrh for v8hi and xxbrq + xxbrw for v4si
376 or v4sf when p9_vector is set.
377 (altivec_vreve<mode>2 for VEC_64): Defined. Implemented by xxswapd.
379 2021-11-23 Navid Rahimi <navidrahimi@microsoft.com>
381 PR tree-optimization/102232
382 * match.pd (x * (1 + y / x) - y) -> (x - y % x): New optimization.
384 2021-11-23 Navid Rahimi <navidrahimi@microsoft.com>
386 PR tree-optimization/96779
387 * match.pd (-x == x) -> (x == 0): New optimization.
389 2021-11-22 Roger Sayle <roger@nextmovesoftware.com>
391 PR tree-optimization/98953
392 PR tree-optimization/103345
393 * gimple-ssa-store-merging.c (find_bswap_or_nop_1): Handle
394 BIT_XOR_EXPR and PLUS_EXPR the same as BIT_IOR_EXPR.
395 (pass_optimize_bswap::execute): Likewise.
397 2021-11-22 Martin Liska <mliska@suse.cz>
399 * doc/invoke.texi: Remove duplicate documentation for 3 params.
401 2021-11-22 Richard Biener <rguenther@suse.de>
403 PR tree-optimization/103351
404 * tree-ssa-dce.c (sort_phi_args): Sort after e->dest_idx as
407 2021-11-22 liuhongt <hongtao.liu@intel.com>
410 * config/i386/constraints.md (Bk): New
411 define_memory_constraint.
412 * config/i386/i386-protos.h (ix86_gpr_tls_address_pattern_p):
414 * config/i386/i386.c (ix86_gpr_tls_address_pattern_p): New
416 * config/i386/i386.md (*movsi_internal): Don't allow
417 mask/sse/mmx move in TLS code sequences.
418 (*movdi_internal): Ditto.
420 2021-11-22 Kewen Lin <linkw@linux.ibm.com>
422 * config/xtensa/xtensa.md (movdi_internal, movdf_internal): Fix split
425 2021-11-21 Jakub Jelinek <jakub@redhat.com>
428 * config/i386/i386-options.c (ix86_valid_target_attribute_p): If
429 fndecl already has DECL_FUNCTION_SPECIFIC_TARGET, use that as base
430 instead of target_option_default_node.
432 2021-11-21 Jan Hubicka <hubicka@ucw.cz>
435 * ipa-modref.c (parm_map_for_arg): Rename to ...
436 (parm_map_for_ptr): .. this one; handle static chain and calls to
438 (modref_access_analysis::get_access): Use parm_map_for_ptr.
439 (modref_access_analysis::process_fnspec): Update.
440 (modref_access_analysis::analyze_load): Update.
441 (modref_access_analysis::analyze_store): Update.
443 2021-11-21 Jan Hubicka <hubicka@ucw.cz>
445 * ipa-modref.c (ignore_nondeterminism_p): Move earlier in source
447 (ignore_retval_p): Likewise.
448 (ignore_stores_p): Likewise.
449 (parm_map_for_arg): Likewise.
450 (class modref_access_analysis): New class.
451 (modref_access_analysis::set_side_effects): New member function.
452 (modref_access_analysis::set_nondeterministic): New member function.
453 (get_access): Turn to ...
454 (modref_access_analysis::get_access): ... this one.
455 (record_access): Turn to ...
456 (modref_access_analysis::record_access): ... this one.
457 (record_access_lto): Turn to ...
458 (modref_access_analysis::record_access_lto): ... This one.
459 (record_access_p): Turn to ...
460 (modref_access_analysis::record_access_p): ... This one
461 (modref_access_analysis::record_unknown_load): New member function.
462 (modref_access_analysis::record_unknown_store): New member function.
463 (get_access_for_fnspec): Turn to ...
464 (modref_access_analysis::get_access_for_fnspec): ... this one.
465 (merge_call_side_effects): Turn to ...
466 (moderf_access_analysis::merge_call_side_effects): Turn to ...
467 (collapse_loads): Move later in source code.
468 (collapse_stores): Move later in source code.
469 (process_fnspec): Turn to ...
470 (modref_access_analysis::process_fnspec): ... this one.
471 (analyze_call): Turn to ...
472 (modref_access_analysis::analyze_call): ... this one.
473 (struct summary_ptrs): Remove.
474 (analyze_load): Turn to ...
475 (modref_access_analysis::analyze_load): ... this one.
476 (analyze_store): Turn to ...
477 (modref_access_analysis::analyze_store): ... this one.
478 (analyze_stmt): Turn to ...
479 (modref_access_analysis::analyze_stmt): ... This one.
480 (remove_summary): Remove.
481 (modref_access_analysis::propagate): Break out from ...
482 (modref_access_analysis::analyze): Break out from ...
483 (analyze_function): ... here.
485 2021-11-21 Roger Sayle <roger@nextmovesoftware.com>
486 Robin Dapp <rdapp@linux.ibm.com>
489 * tree-ssa-math-opts.c (convert_mult_to_widen): Recognize
490 signed WIDEN_MULT_EXPR if the target supports umul_widen_optab.
492 2021-11-20 Jan Hubicka <hubicka@ucw.cz>
495 * ipa-modref.c (ignore_nondeterminism_p): Allow looping pure/cont.
496 (merge_call_side_effects): Improve debug output.
498 2021-11-20 Jan Hubicka <hubicka@ucw.cz>
501 * ipa-pure-const.c (propagate_pure_const): Fix merging of loping flag.
503 2021-11-20 Jeff Law <jeffreyalaw@gmail.com>
505 PR tree-optimization/103226
506 * config/bfin/bfin.md (doloop pattern, splitter and expander): Clobber
509 2021-11-20 Andrew Pinski <apinski@marvell.com>
511 PR tree-optimization/103220
512 * match.pd ((type) X bitop CST): Don't check if CST
513 fits into the type if only the sign changes.
515 2021-11-20 Alexandre Oliva <oliva@adacore.com>
517 PR tree-optimization/102988
518 * gimple-harden-conditionals.cc (detach_value): Copy SSA_NAME
519 without decl sharing.
521 2021-11-19 Iain Sandoe <iain@sandoe.co.uk>
523 * doc/invoke.texi: Remove whitespace after an @option.
525 2021-11-19 Paul A. Clarke <pc@us.ibm.com>
527 * config/rs6000/emmintrin.h (_mm_sad_epu8): Use vec_absd when
528 _ARCH_PWR9, optimize vec_sum2s when LE.
530 2021-11-19 Iain Sandoe <iain@sandoe.co.uk>
533 * config/darwin-driver.c (darwin_driver_init): Handle exported
534 symbols and symbol lists (suppress automatic export of the TLS
536 * config/darwin.c (darwin_rename_builtins): Remove workaround.
537 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
538 (REAL_LIBGCC_SPEC): Handle revised library uses.
539 * config/darwin.opt (nodefaultexport): New.
540 * config/i386/darwin.h (PR80556_WORKAROUND): Remove.
541 * config/i386/darwin32-biarch.h (PR80556_WORKAROUND): Likewise.
542 * config/i386/darwin64-biarch.h (PR80556_WORKAROUND): Likewise.
544 2021-11-19 Martin Jambor <mjambor@suse.cz>
546 * opts.c (default_options_table): Switch off
547 flag_semantic_interposition at Ofast.
548 * doc/invoke.texi (Optimize Options): Document that Ofast switches off
549 -fsemantic-interposition.
551 2021-11-19 Jan Hubicka <hubicka@ucw.cz>
553 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Do not guard modref
554 by !gimple_call_chain.
556 2021-11-19 Martin Sebor <msebor@redhat.com>
560 * doc/invoke.texi (-Waddress): Update.
562 2021-11-19 Andrew MacLeod <amacleod@redhat.com>
564 PR tree-optimization/103254
565 * gimple-range-gori.cc (range_def_chain::get_def_chain): Limit the
566 depth for all statements with multple ssa names.
568 2021-11-19 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
570 * config/s390/s390.md (define_peephole2): Variable insn points
571 to the first matched insn. Use peep2_next_insn(1) to refer to
572 the second matched insn.
574 2021-11-19 Tamar Christina <tamar.christina@arm.com>
576 PR tree-optimization/103311
578 * tree-vect-slp-patterns.c (vect_validate_multiplication): Fix CONJ
580 (complex_mul_pattern::matches): Move check downwards.
582 2021-11-19 Martin Liska <mliska@suse.cz>
585 2021-11-19 Martin Liska <mliska@suse.cz>
587 * cfgexpand.c (pass_expand::execute): Use option directly.
588 * function.c (allocate_struct_function): Likewise.
589 * gimple-low.c (lower_function_body): Likewise.
590 (lower_stmt): Likewise.
591 * gimple-ssa-backprop.c (backprop::prepare_change): Likewise.
592 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call): Likewise.
593 * ipa-split.c (split_function): Likewise.
594 * lto-streamer-in.c (input_function): Likewise.
595 * sese.c (sese_insert_phis_for_liveouts): Likewise.
596 * ssa-iterators.h (num_imm_uses): Likewise.
597 * tree-cfg.c (make_blocks): Likewise.
598 (gimple_merge_blocks): Likewise.
599 * tree-inline.c (tree_function_versioning): Likewise.
600 * tree-loop-distribution.c (generate_loops_for_partition): Likewise.
601 * tree-sra.c (analyze_access_subtree): Likewise.
602 * tree-ssa-dce.c (remove_dead_stmt): Likewise.
603 * tree-ssa-loop-ivopts.c (remove_unused_ivs): Likewise.
604 * tree-ssa-phiopt.c (spaceship_replacement): Likewise.
605 * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
606 * tree-ssa-tail-merge.c (tail_merge_optimize): Likewise.
607 * tree-ssa-threadedge.c (propagate_threaded_block_debug_into): Likewise.
608 * tree-ssa.c (gimple_replace_ssa_lhs): Likewise.
609 (target_for_debug_bind): Likewise.
610 (insert_debug_temp_for_var_def): Likewise.
611 (insert_debug_temps_for_defs): Likewise.
612 (reset_debug_uses): Likewise.
613 * tree-ssanames.c (release_ssa_name_fn): Likewise.
614 * tree-vect-loop-manip.c (adjust_vec_debug_stmts): Likewise.
615 (adjust_debug_stmts): Likewise.
616 (adjust_phi_and_debug_stmts): Likewise.
617 (vect_do_peeling): Likewise.
618 * tree-vect-loop.c (vect_transform_loop_stmt): Likewise.
619 (vect_transform_loop): Likewise.
620 * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Remove
621 (MAY_HAVE_DEBUG_BIND_STMTS): Remove.
622 (MAY_HAVE_DEBUG_STMTS): Use options directly.
624 2021-11-19 Giuliano Belinassi <gbelinassi@suse.de>
626 * gcc.c (process_command): Skip dumpdir override if file is a
628 * doc/invoke.texi: Update -dumpdir documentation.
630 2021-11-19 Andrew Pinski <apinski@marvell.com>
632 PR tree-optimization/103314
633 * match.pd ((type) X op CST): Restrict the equal
634 TYPE_PRECISION case to GIMPLE only.
636 2021-11-19 Martin Liska <mliska@suse.cz>
639 * ipa-modref-tree.h (struct modref_parm_map): Add default
641 * ipa-modref.c (ipa_merge_modref_summary_after_inlining): Use it.
643 2021-11-19 Richard Biener <rguenther@suse.de>
646 * tree-eh.c (operation_could_trap_helper_p): Properly handle
647 fixed-point RDIV_EXPR.
649 2021-11-19 Richard Biener <rguenther@suse.de>
651 PR tree-optimization/102436
652 * tree-ssa-loop-im.c (execute_sm_if_changed): Add mode
653 to just create the if structure and return the then block.
654 (execute_sm): Add flag to indicate the var will re-use
656 (hoist_memory_references): Support a single conditional
657 block with all stores as special case.
659 2021-11-19 Andrew Pinski <apinski@marvell.com>
661 PR tree-optimization/103317
662 * tree-ssa-phiopt.c (minmax_replacement): For the non empty
663 middle bb case, check to make sure it has a single predecessor.
665 2021-11-19 Andrew Pinski <apinski@marvell.com>
667 PR tree-optimization/103257
669 ((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0):
670 Disable until !canonicalize_math_p ().
672 2021-11-19 Marek Polacek <polacek@redhat.com>
676 * doc/invoke.texi: Update documentation for -Wuninitialized.
677 * tree.c (stabilize_reference): Set location.
679 2021-11-19 liuhongt <hongtao.liu@intel.com>
682 * config/i386/x86-tune-costs.h (skylake_cost): Reduce cost of
683 storing 256/512-bit SSE register to be equal to cost of
684 unaligned store to avoid odd alignment peeling.
685 (icelake_cost): Ditto.
687 2021-11-18 David Edelsohn <dje.gcc@gmail.com>
689 * config/rs6000/predicates.md (current_file_function_operand):
690 Add flag_semantic_interposition to call of decl_replaceable_p.
692 2021-11-18 Martin Liska <mliska@suse.cz>
694 * ipa-modref.c (analyze_function): Do not execute the code
695 only if dump_file != NULL.
697 2021-11-18 Martin Liska <mliska@suse.cz>
699 * ipa-modref.c (analyze_function): Use fnode instead of repeated
700 cgraph_node::get (current_function_decl).
702 2021-11-18 Jan Hubicka <hubicka@ucw.cz>
704 * cgraph.c (cgraph_node::get_availability): Update call of
706 (cgraph_node::verify_node): Verify that semantic_interposition flag
708 * cgraph.h: (symtab_node): Add semantic_interposition flag.
709 * cgraphclones.c (set_new_clone_decl_and_node_flags): Clear
710 semantic_interposition flag.
711 * cgraphunit.c (cgraph_node::finalize_function): Set
712 semantic_interposition flag.
713 (cgraph_node::add_new_function): Likewise.
714 (varpool_node::finalize_decl): Likewise.
715 (cgraph_node::create_wrapper): Likewise.
716 * common.opt (fsemantic-interposition): Turn to optimization node.
717 * lto-cgraph.c (lto_output_node): Stream semantic_interposition.
718 (lto_output_varpool_node): Likewise.
719 (input_overwrite_node): Likewise.
720 (input_varpool_node): Likewise.
721 * symtab.c (symtab_node::dump_base): Dump new flag.
722 * varasm.c (decl_replaceable_p): Add semantic_interposition_p
724 * varasm.h (decl_replaceable_p): Update declaration.
725 * varpool.c (varpool_node::ctor_useable_for_folding_p):
726 Use semantic_interposition flag.
727 (varpool_node::get_availability): Likewise.
728 (varpool_node::create_alias): Copy semantic_interposition flag.
730 2021-11-18 Jan Hubicka <hubicka@ucw.cz>
733 * ipa-modref.c (modref_eaf_analysis::merge_call_lhs_flags): Unused
734 parameter may still be returned.
735 (modref_eaf_analysis::analyze_ssa_name): Call merge_call_lhs_flags
736 even for unused function args.
738 2021-11-18 Tamar Christina <tamar.christina@arm.com>
740 PR tree-optimization/103311
741 * tree-vect-slp-patterns.c (complex_mul_pattern::matches,
742 complex_fms_pattern::matches): Check for multiplications.
744 2021-11-18 H.J. Lu <hjl.tools@gmail.com>
747 * config/i386/i386.c (ix86_output_jmp_thunk_or_indirect): Emit
748 CS prefix for -mindirect-branch-cs-prefix.
749 (ix86_output_indirect_branch_via_reg): Likewise.
750 * config/i386/i386.opt: Add -mindirect-branch-cs-prefix.
751 * doc/invoke.texi: Document -mindirect-branch-cs-prefix.
753 2021-11-18 David Edelsohn <dje.gcc@gmail.com>
755 * config/rs6000/driver-rs6000.c (detect_processor_aix): Add
758 2021-11-18 Tamar Christina <tamar.christina@arm.com>
760 PR tree-optimization/103253
761 * tree-ssa-math-opts.c (convert_mult_to_fma): Check for LHS.
763 2021-11-18 Matthias Kretz <m.kretz@gsi.de>
765 * doc/extend.texi: Document __builtin_assoc_barrier.
767 2021-11-18 Martin Liska <mliska@suse.cz>
769 * cfgexpand.c (pass_expand::execute): Use option directly.
770 * function.c (allocate_struct_function): Likewise.
771 * gimple-low.c (lower_function_body): Likewise.
772 (lower_stmt): Likewise.
773 * gimple-ssa-backprop.c (backprop::prepare_change): Likewise.
774 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call): Likewise.
775 * ipa-split.c (split_function): Likewise.
776 * lto-streamer-in.c (input_function): Likewise.
777 * sese.c (sese_insert_phis_for_liveouts): Likewise.
778 * ssa-iterators.h (num_imm_uses): Likewise.
779 * tree-cfg.c (make_blocks): Likewise.
780 (gimple_merge_blocks): Likewise.
781 * tree-inline.c (tree_function_versioning): Likewise.
782 * tree-loop-distribution.c (generate_loops_for_partition): Likewise.
783 * tree-sra.c (analyze_access_subtree): Likewise.
784 * tree-ssa-dce.c (remove_dead_stmt): Likewise.
785 * tree-ssa-loop-ivopts.c (remove_unused_ivs): Likewise.
786 * tree-ssa-phiopt.c (spaceship_replacement): Likewise.
787 * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
788 * tree-ssa-tail-merge.c (tail_merge_optimize): Likewise.
789 * tree-ssa-threadedge.c (propagate_threaded_block_debug_into): Likewise.
790 * tree-ssa.c (gimple_replace_ssa_lhs): Likewise.
791 (target_for_debug_bind): Likewise.
792 (insert_debug_temp_for_var_def): Likewise.
793 (insert_debug_temps_for_defs): Likewise.
794 (reset_debug_uses): Likewise.
795 * tree-ssanames.c (release_ssa_name_fn): Likewise.
796 * tree-vect-loop-manip.c (adjust_vec_debug_stmts): Likewise.
797 (adjust_debug_stmts): Likewise.
798 (adjust_phi_and_debug_stmts): Likewise.
799 (vect_do_peeling): Likewise.
800 * tree-vect-loop.c (vect_transform_loop_stmt): Likewise.
801 (vect_transform_loop): Likewise.
802 * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Remove
803 (MAY_HAVE_DEBUG_BIND_STMTS): Remove.
804 (MAY_HAVE_DEBUG_STMTS): Use options directly.
806 2021-11-18 Richard Biener <rguenther@suse.de>
808 PR tree-optimization/103277
809 * tree-ssa-dse.c (need_ab_cleanup): New.
810 (dse_optimize_redundant_stores): Adjust.
811 (delete_dead_or_redundant_assignment): Get extra
812 need_ab_cleanup argument and set when abnormal cleanup is
814 (dse_optimize_call): Adjust.
815 (dse_optimize_stmt): Likewise.
816 (pass_dse::execute): Allocate and deallocate need_ab_cleanup.
817 Perform abnormal cleanup.
818 * tree-ssa-dse.h (delete_dead_or_redundant_assignment): Adjust.
820 2021-11-18 Hongyu Wang <hongyu.wang@intel.com>
822 * config/i386/i386-expand.c (ix86_expand_atomic_fetch_op_loop):
823 Adjust generated cfg to avoid infinite loop.
825 2021-11-18 konglin1 <lingling.kong@intel.com>
827 * config/i386/avx512fp16intrin.h (_mm512_mul_pch): Add alias for _mm512_fmul_pch.
828 (_mm512_mask_mul_pch): Likewise.
829 (_mm512_maskz_mul_pch): Likewise.
830 (_mm512_mul_round_pch): Likewise.
831 (_mm512_mask_mul_round_pch): Likewise.
832 (_mm512_maskz_mul_round_pch): Likewise.
833 (_mm512_cmul_pch): Likewise.
834 (_mm512_mask_cmul_pch): Likewise.
835 (_mm512_maskz_cmul_pch): Likewise.
836 (_mm512_cmul_round_pch): Likewise.
837 (_mm512_mask_cmul_round_pch): Likewise.
838 (_mm512_maskz_cmul_round_pch): Likewise.
839 (_mm_mul_sch): Likewise.
840 (_mm_mask_mul_sch): Likewise.
841 (_mm_maskz_mul_sch): Likewise.
842 (_mm_mul_round_sch): Likewise.
843 (_mm_mask_mul_round_sch): Likewise.
844 (_mm_maskz_mul_round_sch): Likewise.
845 (_mm_cmul_sch): Likewise.
846 (_mm_mask_cmul_sch): Likewise.
847 (_mm_maskz_cmul_sch): Likewise.
848 (_mm_cmul_round_sch): Likewise.
849 (_mm_mask_cmul_round_sch): Likewise.
850 (_mm_maskz_cmul_round_sch): Likewise.
851 * config/i386/avx512fp16vlintrin.h (_mm_mul_pch): Likewise.
852 (_mm_mask_mul_pch): Likewise.
853 (_mm_maskz_mul_pch): Likewise.
854 (_mm256_mul_pch): Likewise.
855 (_mm256_mask_mul_pch): Likewise.
856 (_mm256_maskz_mul_pch): Likewise.
857 (_mm_cmul_pch): Likewise.
858 (_mm_mask_cmul_pch): Likewise.
859 (_mm_maskz_cmul_pch): Likewise.
860 (_mm256_cmul_pch): Likewise.
861 (_mm256_mask_cmul_pch): Likewise.
862 (_mm256_maskz_cmul_pch): Likewise.
864 2021-11-17 Andrew Pinski <apinski@marvell.com>
866 PR tree-optimization/103228
867 PR tree-optimization/55177
868 * match.pd ((type) X bitop CST): Also do this
869 transformation for nop conversions.
871 2021-11-17 Martin Sebor <msebor@redhat.com>
873 PR tree-optimization/102759
874 * gimple-array-bounds.cc (build_printable_array_type): Move...
875 * gimple-ssa-warn-access.cc (build_printable_array_type): Avoid
876 pathological function redeclarations that remove a previously
878 Improve formatting of function arguments in informational notes.
879 * pointer-query.cc (build_printable_array_type): ...to here.
880 * pointer-query.h (build_printable_array_type): Declared.
882 2021-11-17 H.J. Lu <hjl.tools@gmail.com>
885 * config/i386/i386-opts.h (harden_sls): New enum.
886 * config/i386/i386.c (output_indirect_thunk): Mitigate against
887 SLS for function return.
888 (ix86_output_function_return): Likewise.
889 (ix86_output_jmp_thunk_or_indirect): Mitigate against indirect
891 (ix86_output_indirect_jmp): Likewise.
892 (ix86_output_call_insn): Likewise.
893 * config/i386/i386.opt: Add -mharden-sls=.
894 * doc/invoke.texi: Document -mharden-sls=.
896 2021-11-17 H.J. Lu <hjl.tools@gmail.com>
899 * config/i386/i386.c (ix86_code_end): Remove "%!" before ret.
900 (ix86_output_function_return): Likewise.
901 * config/i386/i386.md (simple_return_pop_internal): Likewise.
903 2021-11-17 Jan Hubicka <hubicka@ucw.cz>
906 * ipa-modref.c (read_modref_records): Fix streaminig in of every_access
909 2021-11-17 Uroš Bizjak <ubizjak@gmail.com>
911 * config/i386/i386.c (indirect_thunks_used): Redefine as HARD_REG_SET.
912 (ix86_code_end): Use TEST_HARD_REG_BIT on indirect_thunks_used.
913 (ix86_output_indirect_branch_via_reg): Use SET_HARD_REG_BIT
914 on indirect_thunks_used.
915 (ix86_output_indirect_function_return): Ditto.
917 2021-11-17 Jan Hubicka <hubicka@ucw.cz>
919 * ipa-modref-tree.c: Include cgraph.h and tree-streamer.h.
920 (modref_access_node::stream_out): New member function.
921 (modref_access_node::stream_in): New member function.
922 * ipa-modref-tree.h (modref_access_node::stream_out,
923 modref_access_node::stream_in): Declare.
924 * ipa-modref.c (modref_summary_lto::useful_p): Free useless kills.
925 (modref_summary_lto::dump): Dump kills.
926 (analyze_store): Record kills for LTO
927 (analyze_stmt): Likewise.
928 (modref_summaries_lto::duplicate): Duplicate kills.
929 (write_modref_records): Use new stream_out member function.
930 (read_modref_records): Likewise.
931 (modref_write): Stream out kills.
932 (read_section): Stream in kills
933 (remap_kills): New function.
934 (update_signature): Use it.
936 2021-11-17 Uroš Bizjak <ubizjak@gmail.com>
938 * config/i386/i386.h (LEGACY_SSE_REGNO_P): New predicate.
939 (SSE_REGNO_P): Use LEGACY_SSE_REGNO_P predicate.
940 * config/i386/i386.c (zero_all_vector_registers):
941 Use LEGACY_SSE_REGNO_P predicate.
942 (ix86_register_priority): Use REX_INT_REGNO_P, REX_SSE_REGNO_P
943 and EXT_REG_SSE_REGNO_P predicates.
944 (ix86_hard_regno_call_part_clobbered): Use REX_SSE_REGNO_P
945 and LEGACY_SSE_REGNO_P predicates.
947 2021-11-17 Jason Merrill <jason@redhat.com>
949 * doc/invoke.texi (C++ Dialect Options): Document
950 -fimplicit-constexpr.
952 2021-11-17 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
954 * config/aarch64/aarch64-modes.def (VECTOR_MODE): New V8DI mode.
955 * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Handle
957 * config/aarch64/iterators.md (define_mode_attr nunits): Add entry
960 2021-11-17 Martin Uecker <uecker@gcc.gnu.org>
964 * gimplify.c (gimplify_var_or_parm_decl): Update comment.
965 (gimplify_compound_lval): Gimplify base expression first.
966 (gimplify_target_expr): Add comment.
968 2021-11-17 Jakub Jelinek <jakub@redhat.com>
970 PR tree-optimization/103192
971 * tree-ssa-loop-im.c (move_computations_worker): Use
972 reset_flow_sensitive_info instead of manually clearing
973 SSA_NAME_RANGE_INFO and do it for all SSA_NAMEs, not just ones
976 2021-11-17 Jakub Jelinek <jakub@redhat.com>
978 PR tree-optimization/103255
979 * gimple-range-fold.cc (fold_using_range::range_of_address): Return
980 range_nonzero rather than unadjusted base's range. Formatting fixes.
982 2021-11-17 Richard Sandiford <richard.sandiford@arm.com>
984 * doc/md.texi (cond_fmin@var{mode}, cond_fmax@var{mode}): Document.
985 * optabs.def (cond_fmin_optab, cond_fmax_optab): New optabs.
986 * internal-fn.def (COND_FMIN, COND_FMAX): New functions.
987 * internal-fn.c (first_commutative_argument): Handle them.
988 (FOR_EACH_COND_FN_PAIR): Likewise.
989 * match.pd (UNCOND_BINARY, COND_BINARY): Likewise.
990 * config/aarch64/aarch64-sve.md (cond_<fmaxmin><mode>): New
993 2021-11-17 Kewen Lin <linkw@linux.ibm.com>
995 * config/i386/i386.md (*add<dwi>3_doubleword, *addv<dwi>4_doubleword,
996 *addv<dwi>4_doubleword_1, *sub<dwi>3_doubleword,
997 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1,
998 *add<dwi>3_doubleword_cc_overflow_1, *divmodsi4_const,
999 *neg<dwi>2_doubleword, *tls_dynamic_gnu2_combine_64_<mode>): Fix split
1002 2021-11-17 Andrew Pinski <apinski@marvell.com>
1004 PR tree-optimization/103288
1005 * tree-ssa-phiopt.c (value_replacement): Return early if middle
1006 block has more than one pred.
1008 2021-11-17 Kewen Lin <linkw@linux.ibm.com>
1010 * config/visium/visium.md (*add<mode>3_insn, *addsi3_insn, *addi3_insn,
1011 *sub<mode>3_insn, *subsi3_insn, *subdi3_insn, *neg<mode>2_insn,
1012 *negdi2_insn, *and<mode>3_insn, *ior<mode>3_insn, *xor<mode>3_insn,
1013 *one_cmpl<mode>2_insn, *ashl<mode>3_insn, *ashr<mode>3_insn,
1014 *lshr<mode>3_insn, *trunchiqi2_insn, *truncsihi2_insn,
1015 *truncdisi2_insn, *extendqihi2_insn, *extendqisi2_insn,
1016 *extendhisi2_insn, *extendsidi2_insn, *zero_extendqihi2_insn,
1017 *zero_extendqisi2_insn, *zero_extendsidi2_insn): Fix split condition.
1019 2021-11-17 Marek Polacek <polacek@redhat.com>
1021 PR preprocessor/103026
1022 * doc/invoke.texi: Document -Wbidi-chars.
1024 2021-11-17 Jan Hubicka <hubicka@ucw.cz>
1027 * ipa-modref.c (ipa_merge_modref_summary_after_inlining): Fix clearing
1030 2021-11-16 Jan Hubicka <jh@suse.cz>
1032 * ipa-modref.c (get_modref_function_summary): Declare.
1033 * ipa-modref.h (get_modref_function_summary): New function.
1034 * tree-ssa-dse.c (clear_live_bytes_for_ref): Break out from ...
1035 (clear_bytes_written_by): ... here; also clear memory killed by
1038 2021-11-16 Iain Sandoe <iain@sandoe.co.uk>
1040 * ggc-common.c (gt_pch_save): If we cannot find a suitable
1041 memory segment for save, then error-out, do not try to
1043 (gt_pch_restore): Save the existing line table, and when
1044 the replacement is being read, use that when constructing
1047 2021-11-16 Peter Bergner <bergner@linux.ibm.com>
1050 * config/rs6000/mma.md (*vsx_assemble_pair): Add early-clobber for
1052 (*mma_assemble_acc): Likewise.
1054 2021-11-16 Martin Sebor <msebor@redhat.com>
1056 PR tree-optimization/102960
1057 * gimple-fold.c (get_range_strlen): Take bitmap as an argument rather
1058 than a pointer to it.
1059 (get_range_strlen_tree): Same. Remove bitmap allocation. Use
1061 (get_maxval_strlen): Use an auto_bitmap.
1062 * tree-ssa-strlen.c (get_range_strlen_dynamic): Factor out PHI
1064 (get_range_strlen_phi): ...into this function.
1065 Avoid assuming maximum string length is constant
1066 (printf_strlen_execute): Dump pointer query cache contents when
1067 details are requisted.
1069 2021-11-16 Jason Merrill <jason@redhat.com>
1071 * langhooks.h (struct lang_hooks): Adjust comment.
1072 * print-tree.c (print_node): Also call print_xnode hook for
1075 2021-11-16 Andrew Pinski <apinski@marvell.com>
1077 PR tree-optimization/103218
1078 * match.pd: New pattern for "((type)(a<0)) << SIGNBITOFA".
1080 2021-11-16 Claudiu Zissulescu <claziss@synopsys.com>
1082 * config/arc/arc.md (maddhisi4): Use a single move to accumulator.
1083 (umaddhisi4): Likewise.
1084 (machi): Update pattern.
1087 2021-11-16 Richard Biener <rguenther@suse.de>
1089 PR tree-optimization/102880
1090 * tree-ssa-dce.c (sort_phi_args): New function.
1091 (make_forwarders_with_degenerate_phis): Likewise.
1092 (perform_tree_ssa_dce): Call
1093 make_forwarders_with_degenerate_phis.
1095 2021-11-16 Richard Biener <rguenther@suse.de>
1097 PR tree-optimization/102880
1098 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Push
1099 single_pred (bb1) condition to places that really need it.
1100 (match_simplify_replacement): Likewise.
1101 (value_replacement): Likewise.
1102 (replace_phi_edge_with_variable): Deal with extra edges
1105 2021-11-16 Martin Jambor <mjambor@suse.cz>
1107 * cfgexpand.c (expand_gimple_basic_block): Use build_debug_expr_decl,
1108 add a fixme note about the mode assignment perhaps being unnecessary.
1109 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
1111 (ipa_param_body_adjustments::mark_dead_statements): Likewise.
1112 (ipa_param_body_adjustments::reset_debug_stmts): Likewise.
1113 * tree-inline.c (remap_ssa_name): Likewise.
1114 (tree_function_versioning): Likewise.
1115 * tree-into-ssa.c (rewrite_debug_stmt_uses): Likewise.
1116 * tree-ssa-loop-ivopts.c (remove_unused_ivs): Likewise.
1117 * tree-ssa.c (insert_debug_temp_for_var_def): Likewise.
1119 2021-11-16 Jakub Jelinek <jakub@redhat.com>
1121 PR tree-optimization/103208
1122 * omp-expand.c (expand_omp_build_cond): New function.
1123 (expand_omp_for_init_counts, expand_omp_for_init_vars,
1124 expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Use it.
1126 2021-11-16 Jakub Jelinek <jakub@redhat.com>
1128 PR tree-optimization/102009
1129 * gimple-ssa-warn-access.cc (pass_waccess::check_alloc_size_call):
1130 Punt if any of alloc_size arguments is out of bounds vs. number of
1133 2021-11-16 Roger Sayle <roger@nextmovesoftware.com>
1135 * config/i386/i386.md (*bmi2_rorx<mode3>_1): Make conditional
1136 on !optimize_function_for_size_p.
1137 (*<any_rotate><mode>3_1): Add preferred_for_size attribute.
1138 (define_splits): Conditionalize on !optimize_function_for_size_p.
1139 (*bmi2_rorxsi3_1_zext): Likewise.
1140 (*<any_rotate>si2_1_zext): Add preferred_for_size attribute.
1141 (define_splits): Conditionalize on !optimize_function_for_size_p.
1143 2021-11-16 Jan Hubicka <jh@suse.cz>
1146 * ipa-modref.c (merge_call_side_effects): Fix uninitialized
1149 2021-11-16 Andrew Pinski <apinski@marvell.com>
1151 PR tree-optimization/103245
1152 * match.pd: Combine the abs pattern matching using multiplication.
1153 Adding optional nop_convert too.
1155 2021-11-16 H.J. Lu <hjl.tools@gmail.com>
1157 PR middle-end/103268
1158 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Add a missing
1161 2021-11-15 Siddhesh Poyarekar <siddhesh@gotplt.org>
1163 * gimple-fold.c (gimple_fold_builtin_strncat): Use ranges to
1164 determine if it is safe to transform to strcat.
1165 (gimple_fold_builtin_snprintf): Likewise.
1167 2021-11-15 Siddhesh Poyarekar <siddhesh@gotplt.org>
1169 * gimple-fold.c (known_lower): New function.
1170 (gimple_fold_builtin_strncat_chk,
1171 gimple_fold_builtin_memory_chk, gimple_fold_builtin_stxcpy_chk,
1172 gimple_fold_builtin_stxncpy_chk,
1173 gimple_fold_builtin_snprintf_chk,
1174 gimple_fold_builtin_sprintf_chk): Use it.
1176 2021-11-15 Siddhesh Poyarekar <siddhesh@gotplt.org>
1178 * gimple-fold.c (dump_transformation): New function.
1179 (gimple_fold_builtin_stxcpy_chk,
1180 gimple_fold_builtin_stxncpy_chk): Use it. Simplify to
1181 BUILT_IN_STRNCPY if return value is not used.
1183 2021-11-15 H.J. Lu <hjl.tools@gmail.com>
1185 PR middle-end/103184
1186 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Check optab
1187 before transforming equivalent, but slighly different cases to
1188 their canonical forms.
1190 2021-11-15 Iain Sandoe <iain@sandoe.co.uk>
1193 * config/darwin.h (TARGET_DTORS_FROM_CXA_ATEXIT): New.
1194 * doc/tm.texi: Regenerated.
1195 * doc/tm.texi.in: Add TARGET_DTORS_FROM_CXA_ATEXIT hook.
1196 * ipa.c (cgraph_build_static_cdtor_1): Return the built
1198 (build_cxa_atexit_decl): New.
1199 (build_dso_handle_decl): New.
1200 (build_cxa_dtor_registrations): New.
1201 (compare_cdtor_tu_order): New.
1202 (build_cxa_atexit_fns): New.
1203 (ipa_cdtor_merge): If dtors_from_cxa_atexit is set,
1204 process the DTORs/CTORs accordingly.
1205 (pass_ipa_cdtor_merge::gate): Also run if
1206 dtors_from_cxa_atexit is set.
1207 * target.def (dtors_from_cxa_atexit): New hook.
1209 2021-11-15 Iain Sandoe <iain@sandoe.co.uk>
1211 * config.in: Regenerate.
1212 * configure: Regenerate.
1213 * configure.ac: Test ld64 for -platform-version support.
1215 2021-11-15 Thomas Schwinge <thomas@codesourcery.com>
1217 * diagnostic-spec.h (typedef xint_hash_t)
1218 (typedef xint_hash_map_t): Replace with...
1219 (typedef nowarn_map_t): ... this.
1220 (nowarn_map): Adjust.
1221 * diagnostic-spec.c (nowarn_map, suppress_warning_at): Likewise.
1223 2021-11-15 Thomas Schwinge <thomas@codesourcery.com>
1225 * profile.c (branch_prob): Use 'location_hash' for
1228 2021-11-15 Aldy Hernandez <aldyh@redhat.com>
1230 PR tree-optimization/103207
1231 * value-range.cc (irange::set): Drop overflow.
1233 2021-11-15 Jakub Jelinek <jakub@redhat.com>
1235 * gimplify.c (optimize_target_teams): Only add OMP_CLAUSE_THREAD_LIMIT
1236 to OMP_TARGET_CLAUSES if it isn't there already.
1238 2021-11-15 Aldy Hernandez <aldyh@redhat.com>
1240 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
1241 Remove useless code.
1242 (path_range_query::ssa_defined_in_bb): New.
1243 (path_range_query::ssa_range_in_phi): Avoid fold_range call that
1244 could trigger additional lookups.
1245 Do not use the cache for ARGs defined in this block.
1246 (path_range_query::compute_ranges_in_block): Use ssa_defined_in_bb.
1247 (path_range_query::maybe_register_phi_relation): Same.
1248 (path_range_query::range_of_stmt): Adjust comment.
1249 * gimple-range-path.h (ssa_defined_in_bb): New.
1251 2021-11-15 Aldy Hernandez <aldyh@redhat.com>
1253 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
1254 Default to global range if nothing found.
1256 2021-11-15 Richard Biener <rguenther@suse.de>
1258 PR tree-optimization/103237
1259 * tree-vect-loop.c (vect_is_simple_reduction): Fail for
1260 double reductions with multiple inner loop LC PHI nodes.
1262 2021-11-15 Hongyu Wang <hongyu.wang@intel.com>
1265 * config/i386/i386-expand.c (ix86_expand_atomic_fetch_op_loop):
1266 New expand function.
1267 * config/i386/i386-options.c (ix86_target_string): Add
1268 -mrelax-cmpxchg-loop flag.
1269 (ix86_valid_target_attribute_inner_p): Likewise.
1270 * config/i386/i386-protos.h (ix86_expand_atomic_fetch_op_loop):
1271 New expand function prototype.
1272 * config/i386/i386.opt: Add -mrelax-cmpxchg-loop.
1273 * config/i386/sync.md (atomic_fetch_<logic><mode>): New expander
1275 (atomic_<logic>_fetch<mode>): Likewise.
1276 (atomic_fetch_nand<mode>): Likewise.
1277 (atomic_nand_fetch<mode>): Likewise.
1278 (atomic_fetch_<logic><mode>): New expander for DI,TI modes.
1279 (atomic_<logic>_fetch<mode>): Likewise.
1280 (atomic_fetch_nand<mode>): Likewise.
1281 (atomic_nand_fetch<mode>): Likewise.
1282 * doc/invoke.texi: Document -mrelax-cmpxchg-loop.
1284 2021-11-15 Richard Biener <rguenther@suse.de>
1286 PR tree-optimization/103219
1287 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Use single_exit
1288 to determine the exit for the VN region.
1290 2021-11-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1292 * tree-ssa-loop.c (pass_vectorize): Move to tree-vectorizer.c.
1293 (pass_data_vectorize): Likewise.
1294 (make_pass_vectorize): Likewise.
1295 * tree-vectorizer.c (vectorize_loops): Merge with
1296 pass_vectorize::execute and replace cfun occurences with fun param.
1297 (adjust_simduid_builtins): Add fun param, replace cfun occurences with
1298 fun, and adjust callers approrpiately.
1299 (note_simd_array_uses): Likewise.
1300 (vect_loop_dist_alias_call): Likewise.
1301 (set_uid_loop_bbs): Likewise.
1302 (vect_transform_loops): Likewise.
1303 (try_vectorize_loop_1): Likewise.
1304 (try_vectorize_loop): Likewise.
1306 2021-11-15 Jakub Jelinek <jakub@redhat.com>
1309 * config/i386/sync.md (atomic_bit_test_and_set<mode>,
1310 atomic_bit_test_and_complement<mode>,
1311 atomic_bit_test_and_reset<mode>): Use OPTAB_WIDEN instead of
1314 2021-11-15 Maciej W. Rozycki <macro@embecosm.com>
1316 * config/vax/vax.c (vax_lra_p): New prototype and function.
1317 (TARGET_LRA_P): Wire it.
1318 * config/vax/vax.opt (mlra): New option.
1319 * doc/invoke.texi (Option Summary, VAX Options): Document the
1322 2021-11-14 Jan Hubicka <jh@suse.cz>
1324 * ipa-modref.h (struct modref_summary): Add nondeterministic
1325 and calls_interposable flags.
1326 * ipa-modref.c (modref_summary::modref_summary): Initialize new flags.
1327 (modref_summary::useful_p): Check new flags.
1328 (struct modref_summary_lto): Add nondeterministic and
1329 calls_interposable flags.
1330 (modref_summary_lto::modref_summary_lto): Initialize new flags.
1331 (modref_summary_lto::useful_p): Check new flags.
1332 (modref_summary::dump): Dump new flags.
1333 (modref_summary_lto::dump): Dump new flags.
1334 (ignore_nondeterminism_p): New function.
1335 (merge_call_side_effects): Merge new flags.
1336 (process_fnspec): Likewise.
1337 (analyze_load): Volatile access is nondeterministic.
1338 (analyze_store): Liekwise.
1339 (analyze_stmt): Volatile ASM is nondeterministic.
1340 (analyze_function): Clear new flags.
1341 (modref_summaries::duplicate): Duplicate new flags.
1342 (modref_summaries_lto::duplicate): Duplicate new flags.
1343 (modref_write): Stream new flags.
1344 (read_section): Stream new flags.
1345 (propagate_unknown_call): Update new flags.
1346 (modref_propagate_in_scc): Propagate new flags.
1347 * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Check
1349 * tree-ssa-structalias.c (determine_global_memory_access):
1352 2021-11-14 Maciej W. Rozycki <macro@embecosm.com>
1354 * config/vax/vax.h (SET_RATIO): New macro.
1355 * config/vax/vax.md (UNSPEC_SETMEM_FILL): New constant.
1356 (setmemhi): New expander.
1357 (setmemhi1): New insn and splitter.
1358 (*setmemhi1): New insn.
1360 2021-11-14 Jan Hubicka <hubicka@ucw.cz>
1362 * ipa-modref-tree.c (modref_access_node::update_for_kills): New
1364 (modref_access_node::merge_for_kills): Likewise.
1365 (modref_access_node::insert_kill): Likewise.
1366 * ipa-modref-tree.h (modref_access_node::update_for_kills,
1367 modref_access_node::merge_for_kills, modref_access_node::insert_kill):
1369 (modref_access_node::useful_for_kill): New member function.
1370 * ipa-modref.c (modref_summary::useful_p): Release useless kills.
1371 (lto_modref_summary): Add kills.
1372 (modref_summary::dump): Dump kills.
1373 (record_access): Add mdoref_access_node parameter.
1374 (record_access_lto): Likewise.
1375 (merge_call_side_effects): Merge kills.
1376 (analyze_call): Add ALWAYS_EXECUTED param and pass it around.
1377 (struct summary_ptrs): Add always_executed filed.
1378 (analyze_load): Update.
1379 (analyze_store): Update; record kills.
1380 (analyze_stmt): Add always_executed; record kills in clobbers.
1381 (analyze_function): Track always_executed.
1382 (modref_summaries::duplicate): Duplicate kills.
1383 (update_signature): Release kills.
1384 * ipa-modref.h (struct modref_summary): Add kills.
1385 * tree-ssa-alias.c (alias_stats): Add kill stats.
1386 (dump_alias_stats): Dump kill stats.
1387 (store_kills_ref_p): Break out from ...
1388 (stmt_kills_ref_p): Use it; handle modref info based kills.
1390 2021-11-14 Aldy Hernandez <aldyh@redhat.com>
1392 PR tree-optimization/103229
1393 * gimple-range-cache.cc (ssa_global_cache::clear): Do not pass
1394 null value to memset.
1396 2021-11-14 Jan Hubicka <hubicka@ucw.cz>
1398 * ipa-modref-tree.c (modref_access_node::get_call_arg): New member
1400 (modref_access_node::get_ao_ref): Likewise.
1401 * ipa-modref-tree.h (modref_access_node::get_call_arg): Declare.
1402 (modref_access_node::get_ao_ref): Declare.
1403 * tree-ssa-alias.c (modref_may_conflict): Use new accessors.
1404 * tree-ssa-dse.c (dse_optimize_call): Use new accessors.
1406 2021-11-13 Jan Hubicka <jh@suse.cz>
1409 * dbgcnt.def (ipa_attr): New counters.
1410 * ipa-pure-const.c: Include dbgcnt.c
1411 (ipa_make_function_const): Use debug counter.
1412 (ipa_make_function_pure): Likewise.
1413 (propagate_pure_const): Fix bug in my previous change.
1415 2021-11-13 Jan Hubicka <hubicka@ucw.cz>
1417 * ipa-modref-tree.c (modref_access_node::range_info_useful_p):
1418 Offline from ipa-modref-tree.h.
1419 (modref_access_node::dump): Move from ipa-modref.c; make member
1421 * ipa-modref-tree.h (modref_access_node::range_info_useful_p.
1422 modref_access_node::dump): Declare.
1423 * ipa-modref.c (dump_access): Remove.
1424 (dump_records): Update.
1425 (dump_lto_records): Update.
1426 (record_access): Update.
1427 (record_access_lto): Update.
1429 2021-11-13 Jan Hubicka <hubicka@ucw.cz>
1431 * ipa-modref.c (modref_summary::modref_summary): Clear new flags.
1432 (modref_summary::dump): Dump try_dse.
1433 (modref_summary::finalize): Add FUN attribute; compute try-dse.
1434 (analyze_function): Update.
1435 (read_section): Update.
1436 (update_signature): Update.
1437 (pass_ipa_modref::execute): Update.
1438 * ipa-modref.h (struct modref_summary):
1439 * tree-ssa-alias.c (ao_ref_init_from_ptr_and_range): Export.
1440 * tree-ssa-alias.h (ao_ref_init_from_ptr_and_range): Declare.
1441 * tree-ssa-dse.c (dse_optimize_call): New function.
1442 (dse_optimize_stmt): Use it.
1444 2021-11-13 Jan Hubicka <hubicka@ucw.cz>
1446 * ipa-modref-tree.c: Move #if CHECKING_P to proper place.
1448 2021-11-13 Jan Hubicka <jh@suse.cz>
1451 (struct modref_access_node): Move longer member functions to
1453 (modref_ref_node::try_merge_with): Turn into modreef_acces_node member
1455 * ipa-modref-tree.c (modref_access_node::contains): Move here
1456 from ipa-modref-tree.h.
1457 (modref_access_node::update): Likewise.
1458 (modref_access_node::merge): Likewise.
1459 (modref_access_node::closer_pair_p): Likewise.
1460 (modref_access_node::forced_merge): Likewise.
1461 (modref_access_node::update2): Likewise.
1462 (modref_access_node::combined_offsets): Likewise.
1463 (modref_access_node::try_merge_with): Likewise.
1464 (modref_access_node::insert): Likewise.
1466 2021-11-13 Jan Hubicka <jh@suse.cz>
1468 * ipa-modref.c (modref_summary::global_memory_read_p): Remove.
1469 (modref_summary::global_memory_written_p): Remove.
1470 (modref_summary::dump): Dump new flags.
1471 (modref_summary::finalize): New member function.
1472 (analyze_function): Call it.
1473 (read_section): Call it.
1474 (update_signature): Call it.
1475 (pass_ipa_modref::execute): Call it.
1476 * ipa-modref.h (struct modref_summary): Remove
1477 global_memory_read_p and global_memory_written_p.
1478 Add global_memory_read, global_memory_written.
1479 * tree-ssa-structalias.c (determine_global_memory_access):
1482 2021-11-13 Jan Hubicka <jh@suse.cz>
1484 * ipa-fnsummary.c (compute_fn_summary): Use type_attribut_allowed_p
1485 * ipa-param-manipulation.c
1486 (ipa_param_adjustments::type_attribute_allowed_p):
1487 New member function.
1488 (drop_type_attribute_if_params_changed_p): New function.
1489 (build_adjusted_function_type): Use it.
1490 * ipa-param-manipulation.h: Add type_attribute_allowed_p.
1492 2021-11-13 David Malcolm <dmalcolm@redhat.com>
1494 * doc/invoke.texi (Static Analyzer Options): Add
1495 -Wno-analyzer-tainted-allocation-size,
1496 -Wno-analyzer-tainted-divisor, -Wno-analyzer-tainted-offset, and
1497 -Wno-analyzer-tainted-size to list. Add
1498 -Wanalyzer-tainted-allocation-size, -Wanalyzer-tainted-divisor,
1499 -Wanalyzer-tainted-offset, and -Wanalyzer-tainted-size to list
1500 of options effectively enabled by -fanalyzer.
1501 (-Wanalyzer-tainted-allocation-size): New.
1502 (-Wanalyzer-tainted-array-index): Tweak wording; add link to CWE.
1503 (-Wanalyzer-tainted-divisor): New.
1504 (-Wanalyzer-tainted-offset): New.
1505 (-Wanalyzer-tainted-size): New.
1507 2021-11-13 Jan Hubicka <jh@suse.cz>
1509 * attr-fnspec.h (attr_fnspec::arg_eaf_flags): Break out from ...
1510 * gimple.c (gimple_call_arg_flags): ... here.
1511 * ipa-modref.c (analyze_parms): Record flags known from fnspec.
1512 (modref_merge_call_site_flags): Use arg_eaf_flags.
1514 2021-11-13 Aldy Hernandez <aldyh@redhat.com>
1516 PR tree-optimization/103222
1517 * gimple-range-path.cc (path_range_query::compute_ranges_in_phis):
1519 (path_range_query::compute_ranges_in_block): Call
1520 compute_ranges_in_phis.
1521 * gimple-range-path.h (path_range_query::compute_ranges_in_phis):
1524 2021-11-13 Jan Hubicka <jh@suse.cz>
1526 * ipa-fnsummary.c (compute_fn_summary): Do not give up on signature
1527 changes on "fn spec" attribute; give up on varadic types.
1528 * ipa-param-manipulation.c: Include attribs.h.
1529 (build_adjusted_function_type): New parameter ARG_MODIFIED; if it is
1530 true remove "fn spec" attribute.
1531 (ipa_param_adjustments::build_new_function_type): Update.
1532 (ipa_param_body_adjustments::modify_formal_parameters): update.
1533 * ipa-sra.c: Include attribs.h.
1534 (ipa_sra_preliminary_function_checks): Do not check for TYPE_ATTRIBUTES.
1536 2021-11-13 Aldy Hernandez <aldyh@redhat.com>
1538 * gimple-range-path.cc (path_range_query::path_range_query): Merge
1540 (path_range_query::import_p): Move from header file.
1541 (path_range_query::~path_range_query): Adjust for combined ctors.
1542 * gimple-range-path.h: Merge ctors.
1543 (path_range_query::import_p): Move to .cc file.
1545 2021-11-13 Jan Hubicka <jh@suse.cz>
1547 PR tree-optimization/103182
1548 * ipa-modref.c (callee_to_caller_flags): Fix merging of flags.
1549 (modref_eaf_analysis::analyze_ssa_name): Fix merging of flags.
1551 2021-11-12 Stafford Horne <shorne@gmail.com>
1553 * config/or1k/or1k-protos.h (or1k_profile_hook): New function.
1554 * config/or1k/or1k.h (PROFILE_HOOK): Change macro to reference
1555 new function or1k_profile_hook.
1556 * config/or1k/or1k.c (struct machine_function): Add new field
1557 set_mcount_arg_insn.
1558 (or1k_profile_hook): New function.
1559 (or1k_init_pic_reg): Update to inject pic rtx after _mcount arg
1561 (or1k_frame_pointer_required): Frame pointer no longer needed
1564 2021-11-12 Jan Hubicka <jh@suse.cz>
1566 PR tree-optimization/103209
1567 * tree-ssa-structalias.c (find_func_aliases_for_call): Fix
1568 use of handle_rhs_call
1570 2021-11-12 Aldy Hernandez <aldyh@redhat.com>
1572 PR tree-optimization/103202
1573 * gimple-range-path.cc
1574 (path_range_query::compute_ranges_in_block): Solve PHI imports first.
1576 2021-11-12 Jan Hubicka <jh@suse.cz>
1578 * ipa-pure-const.c (propagate_pure_const): Remove redundant check;
1579 fix call of ipa_make_function_const and ipa_make_function_pure.
1581 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
1583 * config/aarch64/aarch64.c (aarch64_vector_op::n_advsimd_ops): Delete.
1584 (aarch64_vector_op::m_seen_loads): Likewise.
1585 (aarch64_vector_costs::aarch64_vector_costs): Don't push to
1587 (aarch64_vector_op::count_ops): Remove vectype and factor parameters.
1588 Remove code that tries to predict different vec_flags from the
1590 (aarch64_vector_costs::add_stmt_cost): Update accordingly.
1591 Remove m_advsimd_ops handling.
1593 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
1595 * config/aarch64/aarch64.c (aarch64_vector_costs::m_saw_sve_only_op)
1596 (aarch64_sve_only_stmt_p): Delete.
1597 (aarch64_vector_costs::prefer_unrolled_loop): New function,
1598 extracted from adjust_body_cost.
1599 (aarch64_vector_costs::better_main_loop_than_p): New function,
1600 using heuristics extracted from adjust_body_cost and
1601 adjust_body_cost_sve.
1602 (aarch64_vector_costs::adjust_body_cost_sve): Remove
1603 advsimd_cycles_per_iter and could_use_advsimd parameters.
1604 Update after changes above.
1605 (aarch64_vector_costs::adjust_body_cost): Update after changes above.
1607 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
1609 * config/aarch64/aarch64.c (aarch64_vec_op_count::m_vf_factor):
1610 New member variable.
1611 (aarch64_vec_op_count::aarch64_vec_op_count): Add a parameter for it.
1612 (aarch64_vec_op_count::vf_factor): New function.
1613 (aarch64_vector_costs::aarch64_vector_costs): When costing for
1614 neoverse-512tvb, pass a vf_factor of 2 for the Neoverse V1 version
1616 (aarch64_vector_costs::adjust_body_cost): Read the vf factor
1617 instead of hard-coding 2.
1619 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
1621 * config/aarch64/aarch64.c
1622 (aarch64_vec_op_count::rename_cycles_per_iter): New function.
1623 (aarch64_vec_op_count::min_nonpred_cycles_per_iter): Likewise.
1624 (aarch64_vec_op_count::min_pred_cycles_per_iter): Likewise.
1625 (aarch64_vec_op_count::min_cycles_per_iter): Likewise.
1626 (aarch64_vec_op_count::dump): Move earlier in file. Dump the
1627 above properties too.
1628 (aarch64_estimate_min_cycles_per_iter): Delete.
1629 (adjust_body_cost): Use aarch64_vec_op_count::min_cycles_per_iter
1630 instead of aarch64_estimate_min_cycles_per_iter. Rely on the dump
1631 routine to print CPI estimates.
1632 (adjust_body_cost_sve): Likewise. Use the other functions above
1633 instead of doing the work inline.
1635 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
1637 * config/aarch64/aarch64.c (aarch64_vec_op_count): Allow default
1639 (aarch64_vec_op_count::base_issue_info): Remove handling of null
1641 (aarch64_vec_op_count::simd_issue_info): Likewise.
1642 (aarch64_vec_op_count::sve_issue_info): Likewise.
1643 (aarch64_vector_costs::m_ops): Turn into a vector.
1644 (aarch64_vector_costs::m_advsimd_ops): Likewise.
1645 (aarch64_vector_costs::aarch64_vector_costs): Add entries to
1646 the vectors based on aarch64_tune_params.
1647 (aarch64_vector_costs::analyze_loop_vinfo): Update the pred_ops
1648 of all entries in m_ops.
1649 (aarch64_vector_costs::add_stmt_cost): Call count_ops for all
1651 (aarch64_estimate_min_cycles_per_iter): Remove issue_info
1652 parameter and get the information from the ops instead.
1653 (aarch64_vector_costs::adjust_body_cost_sve): Take a
1654 aarch64_vec_issue_info instead of a aarch64_vec_op_count.
1655 (aarch64_vector_costs::adjust_body_cost): Update call accordingly.
1656 Exit earlier if m_ops is empty for either cost structure.
1658 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
1660 * config/aarch64/aarch64.c (aarch64_vector_costs::m_scalar_ops)
1661 (aarch64_vector_costs::m_sve_ops): Replace with...
1662 (aarch64_vector_costs::m_ops): ...this.
1663 (aarch64_vector_costs::analyze_loop_vinfo): Update accordingly.
1664 (aarch64_vector_costs::adjust_body_cost_sve): Likewise.
1665 (aarch64_vector_costs::aarch64_vector_costs): Likewise.
1666 Initialize m_vec_flags here rather than in add_stmt_cost.
1667 (aarch64_vector_costs::count_ops): Test for scalar reductions too.
1668 Allow vectype to be null.
1669 (aarch64_vector_costs::add_stmt_cost): Call count_ops for scalar
1670 code too. Don't require vectype to be nonnull.
1671 (aarch64_vector_costs::adjust_body_cost): Take the loop_vec_info
1672 and scalar costs as parameters. Use the scalar costs to determine
1673 the cycles per iteration of the scalar loop, then multiply it
1674 by the estimated VF.
1675 (aarch64_vector_costs::finish_cost): Update call accordingly.
1677 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
1679 * config/aarch64/aarch64.c (aarch64_dr_type): New function.
1680 (aarch64_vector_costs::count_ops): Use it rather than the
1681 vectype to determine floatness.
1683 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
1685 * config/aarch64/aarch64.c (aarch64_sve_in_loop_reduction_latency):
1686 Remove vectype parameter and get floatness from the type of the
1688 (arch64_in_loop_reduction_latency): Likewise.
1689 (aarch64_detect_vector_stmt_subtype): Update caller.
1690 (aarch64_vector_costs::count_ops): Likewise.
1692 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
1694 * config/aarch64/aarch64.c (aarch64_sve_op_count): Fold into...
1695 (aarch64_vec_op_count): ...this. Add a constructor.
1696 (aarch64_vec_op_count::vec_flags): New function.
1697 (aarch64_vec_op_count::base_issue_info): Likewise.
1698 (aarch64_vec_op_count::simd_issue_info): Likewise.
1699 (aarch64_vec_op_count::sve_issue_info): Likewise.
1700 (aarch64_vec_op_count::m_issue_info): New member variable.
1701 (aarch64_vec_op_count::m_vec_flags): Likewise.
1702 (aarch64_vector_costs): Add a constructor.
1703 (aarch64_vector_costs::m_sve_ops): Change type to aarch64_vec_op_count.
1704 (aarch64_vector_costs::aarch64_vector_costs): New function.
1705 Initialize m_scalar_ops, m_advsimd_ops and m_sve_ops.
1706 (aarch64_vector_costs::count_ops): Remove vec_flags and
1707 issue_info parameters, using the new aarch64_vec_op_count
1709 (aarch64_vector_costs::add_stmt_cost): Update call accordingly.
1710 (aarch64_sve_op_count::dump): Fold into...
1711 (aarch64_vec_op_count::dump): ..here.
1713 2021-11-12 Richard Sandiford <richard.sandiford@arm.com>
1715 * config/aarch64/aarch64.c: Include tree-dfa.h.
1716 (aarch64_check_consecutive_mems): New function that takes MEM_EXPR
1717 and MEM_OFFSET into account.
1718 (aarch64_swap_ldrstr_operands): Use it.
1719 (aarch64_operands_ok_for_ldpstp): Likewise. Check that the
1720 address of the second memory doesn't depend on the result of
1723 2021-11-12 Jan Hubicka <hubicka@ucw.cz>
1725 * ipa-pure-const.c (ipa_make_function_pure): Fix exit condition.
1727 2021-11-12 Jan Hubicka <jh@suse.cz>
1729 PR tree-optimization/103175
1730 * ipa-modref.c (modref_lattice::merge): Add sanity check.
1731 (callee_to_caller_flags): Make flags adjustment sane.
1732 (modref_eaf_analysis::analyze_ssa_name): Likewise.
1734 2021-11-12 Jan Hubicka <jh@suse.cz>
1737 * ipa-modref.c (analyze_function, modref_propagate_in_scc): Do
1738 not mark pure/const function if there are side-effects.
1740 2021-11-12 Chung-Lin Tang <cltang@codesourcery.com>
1742 * tree.h (OMP_CLAUSE_MAP_RUNTIME_IMPLICIT_P): New access macro for
1743 'implicit' bit, using 'base.deprecated_flag' field of tree_node.
1744 * tree-pretty-print.c (dump_omp_clause): Add support for printing
1745 implicit attribute in tree dumping.
1746 * gimplify.c (gimplify_adjust_omp_clauses_1):
1747 Set OMP_CLAUSE_MAP_RUNTIME_IMPLICIT_P to 1 if map clause is implicitly
1749 (gimplify_adjust_omp_clauses): Adjust place of adding implicitly created
1750 clauses, from simple append, to starting of list, after non-map clauses.
1751 * omp-low.c (lower_omp_target): Add GOMP_MAP_IMPLICIT bits into kind
1752 values passed to libgomp for implicit maps.
1754 2021-11-12 Jakub Jelinek <jakub@redhat.com>
1756 * omp-builtins.def (BUILT_IN_GOMP_TEAMS): Remove.
1757 (BUILT_IN_GOMP_TEAMS4): New.
1758 * builtin-types.def (BT_FN_VOID_UINT_UINT): Remove.
1759 (BT_FN_BOOL_UINT_UINT_UINT_BOOL): New.
1760 * omp-low.c (lower_omp_teams): Use GOMP_teams4 instead of
1761 GOMP_teams, pass to it also num_teams lower-bound expression
1762 or a dup of upper-bound if it is missing and a flag whether
1763 it is the first call or not.
1765 2021-11-12 Martin Liska <mliska@suse.cz>
1767 PR tree-optimization/102497
1768 * gimple-predicate-analysis.cc (add_pred): Remove unused
1771 2021-11-12 Richard Biener <rguenther@suse.de>
1773 PR tree-optimization/103204
1774 * tree-ssa-sccvn.c (valueize_refs_1): Re-valueize the
1775 top operand after folding in an address.
1777 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
1779 * gimple-range-path.cc (path_range_query::path_range_query): New
1780 ctor without a ranger.
1781 (path_range_query::~path_range_query): Free ranger if necessary.
1782 (path_range_query::range_on_path_entry): Adjust m_ranger for pointer.
1783 (path_range_query::ssa_range_in_phi): Same.
1784 (path_range_query::compute_ranges_in_block): Same.
1785 (path_range_query::compute_imports): Same.
1786 (path_range_query::compute_ranges): Same.
1787 (path_range_query::range_of_stmt): Same.
1788 (path_range_query::compute_outgoing_relations): Same.
1789 * gimple-range-path.h (class path_range_query): New ctor.
1790 * tree-ssa-loop-ch.c (ch_base::copy_headers): Remove gimple_ranger
1791 as path_range_query allocates one.
1792 * tree-ssa-threadbackward.c (class back_threader): Remove m_ranger.
1793 (back_threader::~back_threader): Same.
1795 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
1797 * tree-ssa-threadbackward.c
1798 (back_threader_profitability::profitable_path_p): Remove loop
1799 crossing restriction.
1801 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
1803 * ipa-modref.c (analyze_function): Do pure/const discovery, return
1805 (pass_modref::execute): If pure/const is discovered fixup cfg.
1806 (ignore_edge): Do not ignore pure/const edges.
1807 (modref_propagate_in_scc): Do pure/const discovery, return true if
1808 cdtor was promoted pure/const.
1809 (pass_ipa_modref::execute): If needed remove unreachable functions.
1810 * ipa-pure-const.c (warn_function_noreturn): Fix whitespace.
1811 (warn_function_cold): Likewise.
1812 (skip_function_for_local_pure_const): Move earlier.
1813 (ipa_make_function_const): Break out from ...
1814 (ipa_make_function_pure): Break out from ...
1815 (propagate_pure_const): ... here.
1816 (pass_local_pure_const::execute): Use it.
1817 * ipa-utils.h (ipa_make_function_const): Declare.
1818 (ipa_make_function_pure): Declare.
1819 * passes.def: Move early modref after pure-const.
1821 2021-11-11 David Malcolm <dmalcolm@redhat.com>
1824 * diagnostic-show-locus.c (def_policy): Use def_tabstop.
1826 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
1828 * config/aarch64/aarch64-builtins.c (TYPES_COMBINE): Delete.
1829 (TYPES_COMBINEP): Delete.
1830 * config/aarch64/aarch64-simd-builtins.def: Declare type-
1831 qualified builtins for vcombine_* intrinsics.
1832 * config/aarch64/arm_neon.h (vcombine_s8): Remove unnecessary
1834 (vcombine_s16): Likewise.
1835 (vcombine_s32): Likewise.
1836 (vcombine_f32): Likewise.
1837 (vcombine_u8): Use type-qualified builtin and remove casts.
1838 (vcombine_u16): Likewise.
1839 (vcombine_u32): Likewise.
1840 (vcombine_u64): Likewise.
1841 (vcombine_p8): Likewise.
1842 (vcombine_p16): Likewise.
1843 (vcombine_p64): Likewise.
1844 (vcombine_bf16): Remove unnecessary cast.
1845 * config/aarch64/iterators.md (VD_I): New mode iterator.
1846 (VDC_P): New mode iterator.
1848 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
1850 * config/aarch64/aarch64-builtins.c (TYPES_LOAD1_U): Define.
1851 (TYPES_LOAD1_P): Define.
1852 (TYPES_STORE1_U): Define.
1853 (TYPES_STORE1P): Rename to...
1854 (TYPES_STORE1_P): This.
1855 (get_mem_type_for_load_store): Add unsigned and poly types.
1856 (aarch64_general_gimple_fold_builtin): Add unsigned and poly
1857 type-qualified builtin declarations.
1858 * config/aarch64/aarch64-simd-builtins.def: Declare type-
1859 qualified builtins for LD1/ST1.
1860 * config/aarch64/arm_neon.h (vld1_p8): Use type-qualified
1861 builtin and remove cast.
1862 (vld1_p16): Likewise.
1863 (vld1_u8): Likewise.
1864 (vld1_u16): Likewise.
1865 (vld1_u32): Likewise.
1866 (vld1q_p8): Likewise.
1867 (vld1q_p16): Likewise.
1868 (vld1q_p64): Likewise.
1869 (vld1q_u8): Likewise.
1870 (vld1q_u16): Likewise.
1871 (vld1q_u32): Likewise.
1872 (vld1q_u64): Likewise.
1873 (vst1_p8): Likewise.
1874 (vst1_p16): Likewise.
1875 (vst1_u8): Likewise.
1876 (vst1_u16): Likewise.
1877 (vst1_u32): Likewise.
1878 (vst1q_p8): Likewise.
1879 (vst1q_p16): Likewise.
1880 (vst1q_p64): Likewise.
1881 (vst1q_u8): Likewise.
1882 (vst1q_u16): Likewise.
1883 (vst1q_u32): Likewise.
1884 (vst1q_u64): Likewise.
1885 * config/aarch64/iterators.md (VALLP_NO_DI): New iterator.
1887 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
1889 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
1890 builtins for vector reduction.
1891 * config/aarch64/arm_neon.h (vaddv_u8): Use type-qualified
1892 builtin and remove casts.
1893 (vaddv_u16): Likewise.
1894 (vaddv_u32): Likewise.
1895 (vaddvq_u8): Likewise.
1896 (vaddvq_u16): Likewise.
1897 (vaddvq_u32): Likewise.
1898 (vaddvq_u64): Likewise.
1900 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
1902 * config/aarch64/aarch64-simd-builtins.def:
1903 * config/aarch64/arm_neon.h (vpaddq_u8): Use type-qualified
1904 builtin and remove casts.
1905 (vpaddq_u16): Likewise.
1906 (vpaddq_u32): Likewise.
1907 (vpaddq_u64): Likewise.
1908 (vpadd_u8): Likewise.
1909 (vpadd_u16): Likewise.
1910 (vpadd_u32): Likewise.
1911 (vpaddd_u64): Likewise.
1913 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
1915 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
1916 builtins for [r]subhn[2].
1917 * config/aarch64/arm_neon.h (vsubhn_s16): Remove unnecessary
1919 (vsubhn_s32): Likewise.
1920 (vsubhn_s64): Likewise.
1921 (vsubhn_u16): Use type-qualified builtin and remove casts.
1922 (vsubhn_u32): Likewise.
1923 (vsubhn_u64): Likewise.
1924 (vrsubhn_s16): Remove unnecessary cast.
1925 (vrsubhn_s32): Likewise.
1926 (vrsubhn_s64): Likewise.
1927 (vrsubhn_u16): Use type-qualified builtin and remove casts.
1928 (vrsubhn_u32): Likewise.
1929 (vrsubhn_u64): Likewise.
1930 (vrsubhn_high_s16): Remove unnecessary cast.
1931 (vrsubhn_high_s32): Likewise.
1932 (vrsubhn_high_s64): Likewise.
1933 (vrsubhn_high_u16): Use type-qualified builtin and remove
1935 (vrsubhn_high_u32): Likewise.
1936 (vrsubhn_high_u64): Likewise.
1937 (vsubhn_high_s16): Remove unnecessary cast.
1938 (vsubhn_high_s32): Likewise.
1939 (vsubhn_high_s64): Likewise.
1940 (vsubhn_high_u16): Use type-qualified builtin and remove
1942 (vsubhn_high_u32): Likewise.
1943 (vsubhn_high_u64): Likewise.
1945 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
1947 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
1948 builtins for [r]addhn[2].
1949 * config/aarch64/arm_neon.h (vaddhn_s16): Remove unnecessary
1951 (vaddhn_s32): Likewise.
1952 (vaddhn_s64): Likewise.
1953 (vaddhn_u16): Use type-qualified builtin and remove casts.
1954 (vaddhn_u32): Likewise.
1955 (vaddhn_u64): Likewise.
1956 (vraddhn_s16): Remove unnecessary cast.
1957 (vraddhn_s32): Likewise.
1958 (vraddhn_s64): Likewise.
1959 (vraddhn_u16): Use type-qualified builtin and remove casts.
1960 (vraddhn_u32): Likewise.
1961 (vraddhn_u64): Likewise.
1962 (vaddhn_high_s16): Remove unnecessary cast.
1963 (vaddhn_high_s32): Likewise.
1964 (vaddhn_high_s64): Likewise.
1965 (vaddhn_high_u16): Use type-qualified builtin and remove
1967 (vaddhn_high_u32): Likewise.
1968 (vaddhn_high_u64): Likewise.
1969 (vraddhn_high_s16): Remove unnecessary cast.
1970 (vraddhn_high_s32): Likewise.
1971 (vraddhn_high_s64): Likewise.
1972 (vraddhn_high_u16): Use type-qualified builtin and remove
1974 (vraddhn_high_u32): Likewise.
1975 (vraddhn_high_u64): Likewise.
1977 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
1979 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
1980 qualifiers in generator macros for uhsub builtins.
1981 * config/aarch64/arm_neon.h (vhsub_s8): Remove unnecessary
1983 (vhsub_s16): Likewise.
1984 (vhsub_s32): Likewise.
1985 (vhsub_u8): Use type-qualified builtin and remove casts.
1986 (vhsub_u16): Likewise.
1987 (vhsub_u32): Likewise.
1988 (vhsubq_s8): Remove unnecessary cast.
1989 (vhsubq_s16): Likewise.
1990 (vhsubq_s32): Likewise.
1991 (vhsubq_u8): Use type-qualified builtin and remove casts.
1992 (vhsubq_u16): Likewise.
1993 (vhsubq_u32): Likewise.
1995 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
1997 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
1998 qualifiers in generator macros for u[r]hadd builtins.
1999 * config/aarch64/arm_neon.h (vhadd_s8): Remove unnecessary
2001 (vhadd_s16): Likewise.
2002 (vhadd_s32): Likewise.
2003 (vhadd_u8): Use type-qualified builtin and remove casts.
2004 (vhadd_u16): Likewise.
2005 (vhadd_u32): Likewise.
2006 (vhaddq_s8): Remove unnecessary cast.
2007 (vhaddq_s16): Likewise.
2008 (vhaddq_s32): Likewise.
2009 (vhaddq_u8): Use type-qualified builtin and remove casts.
2010 (vhaddq_u16): Likewise.
2011 (vhaddq_u32): Likewise.
2012 (vrhadd_s8): Remove unnecessary cast.
2013 (vrhadd_s16): Likewise.
2014 (vrhadd_s32): Likewise.
2015 (vrhadd_u8): Use type-qualified builtin and remove casts.
2016 (vrhadd_u16): Likewise.
2017 (vrhadd_u32): Likewise.
2018 (vrhaddq_s8): Remove unnecessary cast.
2019 (vrhaddq_s16): Likewise.
2020 (vrhaddq_s32): Likewise.
2021 (vrhaddq_u8): Use type-wualified builtin and remove casts.
2022 (vrhaddq_u16): Likewise.
2023 (vrhaddq_u32): Likewise.
2025 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
2027 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
2028 qualifiers in generator macros for usub[lw][2] builtins.
2029 * config/aarch64/arm_neon.h (vsubl_s8): Remove unnecessary
2031 (vsubl_s16): Likewise.
2032 (vsubl_s32): Likewise.
2033 (vsubl_u8): Use type-qualified builtin and remove casts.
2034 (vsubl_u16): Likewise.
2035 (vsubl_u32): Likewise.
2036 (vsubl_high_s8): Remove unnecessary cast.
2037 (vsubl_high_s16): Likewise.
2038 (vsubl_high_s32): Likewise.
2039 (vsubl_high_u8): Use type-qualified builtin and remove casts.
2040 (vsubl_high_u16): Likewise.
2041 (vsubl_high_u32): Likewise.
2042 (vsubw_s8): Remove unnecessary casts.
2043 (vsubw_s16): Likewise.
2044 (vsubw_s32): Likewise.
2045 (vsubw_u8): Use type-qualified builtin and remove casts.
2046 (vsubw_u16): Likewise.
2047 (vsubw_u32): Likewise.
2048 (vsubw_high_s8): Remove unnecessary cast.
2049 (vsubw_high_s16): Likewise.
2050 (vsubw_high_s32): Likewise.
2051 (vsubw_high_u8): Use type-qualified builtin and remove casts.
2052 (vsubw_high_u16): Likewise.
2053 (vsubw_high_u32): Likewise.
2055 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
2057 * config/aarch64/aarch64-simd-builtins.def: Use BINOPU type
2058 qualifiers in generator macros for uadd[lw][2] builtins.
2059 * config/aarch64/arm_neon.h (vaddl_s8): Remove unnecessary
2061 (vaddl_s16): Likewise.
2062 (vaddl_s32): Likewise.
2063 (vaddl_u8): Use type-qualified builtin and remove casts.
2064 (vaddl_u16): Likewise.
2065 (vaddl_u32): Likewise.
2066 (vaddl_high_s8): Remove unnecessary cast.
2067 (vaddl_high_s16): Likewise.
2068 (vaddl_high_s32): Likewise.
2069 (vaddl_high_u8): Use type-qualified builtin and remove casts.
2070 (vaddl_high_u16): Likewise.
2071 (vaddl_high_u32): Likewise.
2072 (vaddw_s8): Remove unnecessary cast.
2073 (vaddw_s16): Likewise.
2074 (vaddw_s32): Likewise.
2075 (vaddw_u8): Use type-qualified builtin and remove casts.
2076 (vaddw_u16): Likewise.
2077 (vaddw_u32): Likewise.
2078 (vaddw_high_s8): Remove unnecessary cast.
2079 (vaddw_high_s16): Likewise.
2080 (vaddw_high_s32): Likewise.
2081 (vaddw_high_u8): Use type-qualified builtin and remove casts.
2082 (vaddw_high_u16): Likewise.
2083 (vaddw_high_u32): Likewise.
2085 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
2087 * config/aarch64/aarch64-simd-builtins.def: Declare type-
2088 qualified builtins for [R]SHRN[2].
2089 * config/aarch64/arm_neon.h (vshrn_n_u16): Use type-qualified
2090 builtin and remove casts.
2091 (vshrn_n_u32): Likewise.
2092 (vshrn_n_u64): Likewise.
2093 (vrshrn_high_n_u16): Likewise.
2094 (vrshrn_high_n_u32): Likewise.
2095 (vrshrn_high_n_u64): Likewise.
2096 (vrshrn_n_u16): Likewise.
2097 (vrshrn_n_u32): Likewise.
2098 (vrshrn_n_u64): Likewise.
2099 (vshrn_high_n_u16): Likewise.
2100 (vshrn_high_n_u32): Likewise.
2101 (vshrn_high_n_u64): Likewise.
2103 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
2105 * config/aarch64/aarch64-simd-builtins.def: Declare unsigned
2106 type-qualified builtins for XTN[2].
2107 * config/aarch64/arm_neon.h (vmovn_high_u16): Use type-
2108 qualified builtin and remove casts.
2109 (vmovn_high_u32): Likewise.
2110 (vmovn_high_u64): Likewise.
2111 (vmovn_u16): Likewise.
2112 (vmovn_u32): Likewise.
2113 (vmovn_u64): Likewise.
2115 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
2117 * config/aarch64/aarch64-simd-builtins.def: Use poly type
2118 qualifier in builtin generator macros.
2119 * config/aarch64/arm_neon.h (vmul_p8): Use type-qualified
2120 builtin and remove casts.
2121 (vmulq_p8): Likewise.
2122 (vmull_high_p8): Likewise.
2123 (vmull_p8): Likewise.
2125 2021-11-11 Jonathan Wright <jonathan.wright@arm.com>
2127 * config/aarch64/aarch64-simd-builtins.def: Declare type-
2128 qualified builtin generators for unsigned MLA/MLS intrinsics.
2129 * config/aarch64/arm_neon.h (vmla_n_u16): Use type-qualified
2131 (vmla_n_u32): Likewise.
2132 (vmla_u8): Likewise.
2133 (vmla_u16): Likewise.
2134 (vmla_u32): Likewise.
2135 (vmlaq_n_u16): Likewise.
2136 (vmlaq_n_u32): Likewise.
2137 (vmlaq_u8): Likewise.
2138 (vmlaq_u16): Likewise.
2139 (vmlaq_u32): Likewise.
2140 (vmls_n_u16): Likewise.
2141 (vmls_n_u32): Likewise.
2142 (vmls_u8): Likewise.
2143 (vmls_u16): Likewise.
2144 (vmls_u32): Likewise.
2145 (vmlsq_n_u16): Likewise.
2146 (vmlsq_n_u32): Likewise.
2147 (vmlsq_u8): Likewise.
2148 (vmlsq_u16): Likewise.
2149 (vmlsq_u32): Likewise.
2151 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
2153 * ipa-modref.c (modref_summary::useful_p): Check also for side-effects
2154 with looping const/pure.
2155 (modref_summary_lto::useful_p): Likewise.
2156 (merge_call_side_effects): Merge side effects before early exit
2158 (process_fnspec): Also handle pure functions.
2159 (analyze_call): Do not early exit on looping pure const.
2160 (propagate_unknown_call): Also handle nontrivial SCC as side-effect.
2161 (modref_propagate_in_scc): Update.
2163 2021-11-11 Richard Biener <rguenther@suse.de>
2165 PR tree-optimization/103190
2166 * tree-ssa-reassoc.c (insert_stmt_after): Only assert on asm goto.
2168 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
2170 * gimple-range-path.cc (path_range_query::add_copies_to_imports):
2172 (path_range_query::compute_imports): ...this. Adapt it so it can
2173 be passed the imports bitmap instead of working on m_imports.
2174 (path_range_query::compute_ranges): Call compute_imports in all
2175 cases unless an imports bitmap is passed.
2176 * gimple-range-path.h (path_range_query::compute_imports): New.
2177 (path_range_query::add_copies_to_imports): Remove.
2178 * tree-ssa-threadbackward.c (back_threader::resolve_def): Remove.
2179 (back_threader::find_paths_to_names): Inline resolve_def.
2180 (back_threader::find_paths): Call compute_imports.
2181 (back_threader::resolve_phi): Adjust comment.
2183 2021-11-11 Richard Biener <rguenther@suse.de>
2185 PR tree-optimization/103188
2186 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p):
2187 Remove query parameter, split out check for size
2189 (ch_base::m_ranger, cb_base::m_query): Remove.
2190 (ch_base::copy_headers): Split processing loop into
2191 analysis around which we allocate and use ranger and
2192 transform where we do not.
2193 (pass_ch::execute): Do not allocate/free ranger here.
2194 (pass_ch_vect::execute): Likewise.
2196 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
2198 * ipa-pure-const.c (propagate_pure_const): Self recursion is
2201 2021-11-11 Jan Hubicka <hubicka@ucw.cz>
2203 * cgraph.c (set_noreturn_flag_1): New function.
2204 (cgraph_node::set_noreturn_flag): New member function
2205 * cgraph.h (cgraph_node::set_noreturn_flags): Declare.
2206 * ipa-pure-const.c (pass_local_pure_const::execute): Use it.
2208 2021-11-11 Aldy Hernandez <aldyh@redhat.com>
2209 Richard Biener <rguenther@suse.de>
2211 * tree-ssa-loop-ch.c (entry_loop_condition_is_static): Resolve
2212 statically to the edge remaining in the loop.
2214 2021-11-11 Richard Biener <rguenther@suse.de>
2216 PR middle-end/103181
2217 * tree-eh.c (operation_could_trap_helper_p): Properly
2218 check vector constants for a zero element for integer
2219 division. Separate floating point and integer division code.
2221 2021-11-11 Jakub Jelinek <jakub@redhat.com>
2224 * dwarf2out.c (field_byte_offset): Do the PCC_BITFIELD_TYPE_MATTERS
2225 handling only for DECL_BIT_FIELD_TYPE decls.
2227 2021-11-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2230 * config/aarch64/aarch64.c (aarch64_process_target_attr): Check if
2231 token is arch extension without leading '+' and emit appropriate
2232 diagnostic for the same.
2234 2021-11-11 Jakub Jelinek <jakub@redhat.com>
2236 * tree.h (OMP_CLAUSE_NUM_TEAMS_EXPR): Rename to ...
2237 (OMP_CLAUSE_NUM_TEAMS_UPPER_EXPR): ... this.
2238 (OMP_CLAUSE_NUM_TEAMS_LOWER_EXPR): Define.
2239 * tree.c (omp_clause_num_ops): Increase num ops for
2240 OMP_CLAUSE_NUM_TEAMS to 2.
2241 * tree-pretty-print.c (dump_omp_clause): Print optional lower bound
2242 for OMP_CLAUSE_NUM_TEAMS.
2243 * gimplify.c (gimplify_scan_omp_clauses): Gimplify
2244 OMP_CLAUSE_NUM_TEAMS_LOWER_EXPR if non-NULL.
2245 (optimize_target_teams): Use OMP_CLAUSE_NUM_TEAMS_UPPER_EXPR instead
2246 of OMP_CLAUSE_NUM_TEAMS_EXPR. Handle OMP_CLAUSE_NUM_TEAMS_LOWER_EXPR.
2247 * omp-low.c (lower_omp_teams): Use OMP_CLAUSE_NUM_TEAMS_UPPER_EXPR
2248 instead of OMP_CLAUSE_NUM_TEAMS_EXPR.
2249 * omp-expand.c (expand_teams_call, get_target_arguments): Likewise.
2251 2021-11-11 Richard Biener <rguenther@suse.de>
2253 * cfganal.c (find_pdom): Remove.
2254 (control_dependences::find_control_dependence): Remove
2255 special-casing of entry block, call get_immediate_dominator
2257 * gimple-predicate-analysis.cc (find_pdom): Remove.
2258 (find_dom): Likewise.
2259 (find_control_equiv_block): Call get_immediate_dominator
2261 (compute_control_dep_chain): Likewise.
2262 (predicate::init_from_phi_def): Likewise.
2264 2021-11-11 Richard Biener <rguenther@suse.de>
2266 * cfganal.h (control_dependences::control_dependence_map):
2268 (control_dependences::m_bitmaps): New.
2269 * cfganal.c (control_dependences::set_control_dependence_map_bit):
2271 (control_dependences::clear_control_dependence_bitmap):
2273 (control_dependences::find_control_dependence): Do not
2274 find_edge for the abnormal edge test.
2275 (control_dependences::control_dependences): Instead do not
2276 add abnormal edges to the edge list. Adjust.
2277 (control_dependences::~control_dependences): Likewise.
2278 (control_dependences::get_edges_dependent_on): Likewise.
2279 * function-tests.c: Include bitmap.h.
2281 2021-11-11 Kewen Lin <linkw@linux.ibm.com>
2283 * doc/invoke.texi: Change references to "future cpu" to "power10",
2284 "-mcpu=future" to "-mcpu=power10". Adjust words for float128.
2286 2021-11-11 Cui,Lili <lili.cui@intel.com>
2288 * config/i386/i386-options.c (m_CORE_AVX2): Remove Alderlake
2290 (processor_cost_table): Use alderlake_cost for Alderlake.
2291 * config/i386/i386.c (ix86_sched_init_global): Handle Alderlake.
2292 * config/i386/x86-tune-costs.h (struct processor_costs): Add alderlake
2294 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Alderlake
2296 (ix86_adjust_cost): Handle Alderlake.
2297 * config/i386/x86-tune.def (X86_TUNE_SCHEDULE): Enable for Alderlake.
2298 (X86_TUNE_PARTIAL_REG_DEPENDENCY): Likewise.
2299 (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY): Likewise.
2300 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): Likewise.
2301 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
2302 (X86_TUNE_MEMORY_MISMATCH_STALL): Likewise.
2303 (X86_TUNE_USE_LEAVE): Likewise.
2304 (X86_TUNE_PUSH_MEMORY): Likewise.
2305 (X86_TUNE_USE_INCDEC): Likewise.
2306 (X86_TUNE_INTEGER_DFMODE_MOVES): Likewise.
2307 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
2308 (X86_TUNE_USE_SAHF): Likewise.
2309 (X86_TUNE_USE_BT): Likewise.
2310 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
2311 (X86_TUNE_ONE_IF_CONV_INSN): Likewise.
2312 (X86_TUNE_AVOID_MFENCE): Likewise.
2313 (X86_TUNE_USE_SIMODE_FIOP): Likewise.
2314 (X86_TUNE_EXT_80387_CONSTANTS): Likewise.
2315 (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL): Likewise.
2316 (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Likewise.
2317 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
2318 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
2319 (X86_TUNE_AVOID_4BYTE_PREFIXES): Likewise.
2320 (X86_TUNE_USE_GATHER): Disable for Alderlake.
2321 (X86_TUNE_AVX256_MOVE_BY_PIECES): Likewise.
2322 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
2324 2021-11-11 liuhongt <hongtao.liu@intel.com>
2327 * config/i386/sse.md (V_128_256): Extend to V8HF/V16HF.
2328 (avxsizesuffix): Ditto.
2330 2021-11-11 Kito Cheng <kito.cheng@sifive.com>
2332 * common/config/riscv/riscv-common.c (riscv_subset_list::to_string): Fix
2333 wrong marco checking.
2335 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
2337 PR tree-optimization/102906
2338 * tree-ssa-loop-ch.c (entry_loop_condition_is_static): New.
2339 (should_duplicate_loop_header_p): Call entry_loop_condition_is_static.
2340 (class ch_base): Add m_ranger and m_query.
2341 (ch_base::copy_headers): Pass m_query to
2342 entry_loop_condition_is_static.
2343 (pass_ch::execute): Allocate and deallocate m_ranger and
2345 (pass_ch_vect::execute): Same.
2347 2021-11-10 Andrew Pinski <apinski@marvell.com>
2350 * config/aarch64/aarch64-simd.md (aarch64_simd_dup<mode>):
2351 Use vwcore iterator for the r constraint output string.
2353 2021-11-10 qing zhao <qing.zhao@oracle.com>
2355 * internal-fn.c (expand_DEFERRED_INIT): Apply pattern initialization
2356 only when have_insn_for return true for the mode. Fix a memory leak.
2358 2021-11-10 Christophe Lyon <christophe.lyon@foss.st.com>
2360 * config/arm/arm.c (cortexa9_extra_costs, cortexa8_extra_costs,
2361 cortexa5_extra_costs, cortexa7_extra_costs,
2362 cortexa12_extra_costs, cortexa15_extra_costs, v7m_extra_costs):
2363 Initialize movi, dup and extract costing fields.
2365 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
2367 * gimple-range-path.cc (path_range_query::path_range_query): Do
2369 (path_range_query::dump): Change m_path uses to non-pointer.
2370 (path_range_query::defined_outside_path): Same.
2371 (path_range_query::set_path): Same.
2372 (path_range_query::add_copies_to_imports): Same.
2373 (path_range_query::range_of_stmt): Same.
2374 (path_range_query::compute_outgoing_relations): Same.
2375 (path_range_query::compute_ranges): Imports are now optional.
2376 Implement overload that takes an edge.
2377 * gimple-range-path.h (class path_range_query): Make imports
2378 optional for compute_ranges. Add compute_ranges(edge) overload.
2379 Make m_path an auto_vec instead of a pointer and adjust
2382 2021-11-10 Tamar Christina <tamar.christina@arm.com>
2384 * tree-vectorizer.h (struct scalar_cond_masked_key): Add inverted_p.
2385 (default_hash_traits<scalar_conf_masked_key>): Likewise.
2386 * tree-vect-stmts.c (vectorizable_condition): Check if inverse of mask
2388 * tree-vectorizer.c (scalar_cond_masked_key::get_cond_ops_from_tree):
2389 Register mask inverses.
2391 2021-11-10 Tamar Christina <tamar.christina@arm.com>
2393 * tree-vectorizer.c (vectorize_loops): Do local CSE through RPVN upon
2394 successful vectorization.
2396 2021-11-10 Andrew MacLeod <amacleod@redhat.com>
2398 * gimple-range-cache.cc (sbr_vector::grow): New.
2399 (sbr_vector::set_bb_range): Call grow.
2400 (sbr_vector::get_bb_range): Same.
2401 (sbr_vector::bb_range_p): Remove assert.
2403 2021-11-10 Tamar Christina <tamar.christina@arm.com>
2405 * config/aarch64/aarch64-simd.md (*aarch64_topbits_shuffle<mode>_le
2406 ,*aarch64_topbits_shuffle<mode>_be): Remove.
2408 2021-11-10 Jan Hubicka <jh@suse.cz>
2410 * ipa-modref.c: Include tree-eh.h
2411 (modref_summary::modref_summary): Initialize side_effects.
2412 (struct modref_summary_lto): New bool field side_effects.
2413 (modref_summary_lto::modref_summary_lto): Initialize side_effects.
2414 (modref_summary::dump): Dump side_effects.
2415 (modref_summary_lto::dump): Dump side_effects.
2416 (merge_call_side_effects): Merge side effects.
2417 (process_fnspec): Calls to non-const/pure or looping
2418 function is a side effect.
2419 (analyze_call): Self-recursion is a side-effect; handle
2421 (analyze_load): Watch for volatile and throwing memory.
2422 (analyze_store): Likewise.
2423 (analyze_stmt): Watch for volatitle asm.
2424 (analyze_function): Handle side_effects.
2425 (modref_summaries::duplicate): Duplicate side_effects.
2426 (modref_summaries_lto::duplicate): Likewise.
2427 (modref_write): Stream side_effects.
2428 (read_section): Likewise.
2429 (update_signature): Update.
2430 (propagate_unknown_call): Handle side_effects.
2431 (modref_propagate_in_scc): Likewise.
2432 * ipa-modref.h (struct modref_summary): Add side_effects.
2433 * ipa-pure-const.c (special_builtin_state): Rename to ...
2434 (builtin_safe_for_const_function_p): ... this one.
2435 (check_call): Update.
2436 (finite_function_p): Break out from ...
2437 (propagate_pure_const): ... here
2438 * ipa-utils.h (finite_function): Declare.
2440 2021-11-10 Lucas A. M. Magalhães <lamm@linux.ibm.com>
2442 * config.gcc (powerpc*-*-*): Remove -rpath from
2443 --with-advance-toolchain.
2445 2021-11-10 Marek Polacek <polacek@redhat.com>
2448 * attribs.c (struct scoped_attributes): Add a bool member.
2449 (lookup_scoped_attribute_spec): Forward declare.
2450 (register_scoped_attributes): New bool parameter, defaulted to
2452 (handle_ignored_attributes_option): New function.
2453 (free_attr_data): New function.
2454 (init_attributes): Call handle_ignored_attributes_option.
2455 (attr_namespace_ignored_p): New function.
2456 (decl_attributes): Check attr_namespace_ignored_p before
2458 * attribs.h (free_attr_data): Declare.
2459 (register_scoped_attributes): Adjust declaration.
2460 (handle_ignored_attributes_option): Declare.
2461 (canonicalize_attr_name): New function template.
2462 (canonicalize_attr_name): Use it.
2463 * common.opt (Wattributes=): New option with a variable.
2464 * doc/extend.texi: Document #pragma GCC diagnostic ignored_attributes.
2465 * doc/invoke.texi: Document -Wno-attributes=.
2466 * opts.c (common_handle_option) <case OPT_Wattributes_>: Handle.
2467 * plugin.h (register_scoped_attributes): Adjust declaration.
2468 * toplev.c (compile_file): Call free_attr_data.
2470 2021-11-10 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
2472 * config/arm/arm-cpus.in (cortex-a710): New CPU.
2473 * config/arm/arm-tables.opt: Regenerate.
2474 * config/arm/arm-tune.md: Regenerate.
2475 * doc/invoke.texi: Update docs.
2477 2021-11-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2479 * config/aarch64/aarch64-builtins.c
2480 (aarch64_general_gimple_fold_builtin): Mark argument as unused.
2482 2021-11-10 Martin Liska <mliska@suse.cz>
2484 * lto-wrapper.c (merge_and_complain): Make the first argument
2487 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
2489 * config/aarch64/iterators.md (optab): Use fmax_nan instead of
2490 smax_nan and fmin_nan instead of smin_nan.
2491 (maxmin_uns): Rename to...
2492 (fmaxmin): ...this and make the same changes. Remove entries
2493 unrelated to fmax* and fmin*.
2494 * config/aarch64/aarch64.md (<maxmin_uns><mode>3): Rename to...
2495 (<fmaxmin><mode>3): ...this.
2496 * config/aarch64/aarch64-simd.md (aarch64_<maxmin_uns>p<mode>):
2498 (aarch64_<optab>p<mode>): ...this.
2499 (<maxmin_uns><mode>3): Rename to...
2500 (<fmaxmin><mode>3): ...this.
2501 (reduc_<maxmin_uns>_scal_<mode>): Rename to...
2502 (reduc_<optab>_scal_<mode>): ...this and update gen* call.
2503 (aarch64_reduc_<maxmin_uns>_internal<mode>): Rename to...
2504 (aarch64_reduc_<optab>_internal<mode>): ...this.
2505 (aarch64_reduc_<maxmin_uns>_internalv2si): Rename to...
2506 (aarch64_reduc_<optab>_internalv2si): ...this.
2507 * config/aarch64/aarch64-sve.md (<maxmin_uns><mode>3): Rename to...
2508 (<fmaxmin><mode>3): ...this.
2509 * config/aarch64/aarch64-simd-builtins.def (smax_nan, smin_nan)
2511 (fmax_nan, fmin_nan): ...this.
2512 * config/aarch64/arm_neon.h (vmax_f32, vmax_f64, vmaxq_f32, vmaxq_f64)
2513 (vmin_f32, vmin_f64, vminq_f32, vminq_f64, vmax_f16, vmaxq_f16)
2514 (vmin_f16, vminq_f16): Update accordingly.
2516 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
2518 * tree-vectorizer.h (vector_costs::finish_cost): Take the
2519 corresponding scalar costs as a parameter.
2520 (finish_cost): Likewise.
2521 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
2522 (vect_estimate_min_profitable_iters): Update accordingly.
2523 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Likewise.
2524 * tree-vectorizer.c (vector_costs::finish_cost): Likewise.
2525 * config/aarch64/aarch64.c (aarch64_vector_costs::finish_cost):
2527 * config/rs6000/rs6000.c (rs6000_cost_data::finish_cost): Likewise.
2529 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
2531 * tree-vectorizer.h (_loop_vec_info::scalar_costs): New member
2533 (_loop_vec_info::single_scalar_iteration_cost): Delete.
2534 (LOOP_VINFO_SINGLE_SCALAR_ITERATION_COST): Delete.
2535 (vector_costs::total_cost): New function.
2536 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
2537 after above changes.
2538 (_loop_vec_info::~_loop_vec_info): Delete scalar_costs.
2539 (vect_compute_single_scalar_iteration_cost): Store the costs
2540 in loop_vinfo->scalar_costs.
2541 (vect_estimate_min_profitable_iters): Get the scalar cost from
2542 loop_vinfo->scalar_costs.
2544 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
2546 * tree-vectorizer.h (vector_costs::better_main_loop_than_p)
2547 (vector_costs::better_epilogue_loop_than_p)
2548 (vector_costs::compare_inside_loop_cost)
2549 (vector_costs::compare_outside_loop_cost): Likewise.
2550 * tree-vectorizer.c (vector_costs::better_main_loop_than_p)
2551 (vector_costs::better_epilogue_loop_than_p)
2552 (vector_costs::compare_inside_loop_cost)
2553 (vector_costs::compare_outside_loop_cost): New functions,
2554 containing code moved from...
2555 * tree-vect-loop.c (vect_better_loop_vinfo_p): ...here.
2557 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
2559 * tree-vectorizer.h (_loop_vec_info): Remove vec_outside_cost
2560 and vec_inside_cost.
2561 (vector_costs::outside_cost): New function.
2562 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
2564 (vect_estimate_min_profitable_iters): Likewise.
2565 (vect_better_loop_vinfo_p): Get the inside and outside costs
2566 from the loop_vec_infos' vector_costs.
2568 2021-11-10 Richard Sandiford <richard.sandiford@arm.com>
2570 * tree-vectorizer.h (vec_info::target_cost_data): Replace with...
2571 (_loop_vec_info::vector_costs): ...this.
2572 (LOOP_VINFO_TARGET_COST_DATA): Delete.
2573 * tree-vectorizer.c (vec_info::vec_info): Remove target_cost_data
2575 (vec_info::~vec_info): Remove corresponding delete.
2576 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2577 vector_costs to null.
2578 (_loop_vec_info::~_loop_vec_info): Delete vector_costs.
2579 (vect_analyze_loop_operations): Update after above changes.
2580 (vect_analyze_loop_2): Likewise.
2581 (vect_estimate_min_profitable_iters): Likewise.
2582 * tree-vect-slp.c (vect_slp_analyze_operations): Likewise.
2584 2021-11-10 Jan Hubicka <hubicka@ucw.cz>
2586 * tree-core.h (EAF_DIRECT): Remove.
2587 (EAF_NOCLOBBER): Remove.
2588 (EAF_UNUSED): Remove.
2589 (EAF_NOESCAPE): Remove.
2590 (EAF_NO_DIRECT_CLOBBER): New.
2591 (EAF_NO_INDIRECT_CLOBBER): New.
2592 (EAF_NODIRECTESCAPE): Remove.
2593 (EAF_NO_DIRECT_ESCAPE): New.
2594 (EAF_NO_INDIRECT_ESCAPE): New.
2595 (EAF_NOT_RETURNED): Remove.
2596 (EAF_NOT_RETURNED_INDIRECTLY): New.
2597 (EAF_NOREAD): Remove.
2598 (EAF_NO_DIRECT_READ): New.
2599 (EAF_NO_INDIRECT_READ): New.
2600 * gimple.c (gimple_call_arg_flags): Update for new flags.
2601 (gimple_call_retslot_flags): Update for new flags.
2602 * ipa-modref.c (dump_eaf_flags): Likewise.
2603 (remove_useless_eaf_flags): Likewise.
2604 (deref_flags): Likewise.
2605 (modref_lattice::init): Likewise.
2606 (modref_lattice::merge): Likewise.
2607 (modref_lattice::merge_direct_load): Likewise.
2608 (modref_lattice::merge_direct_store): Likewise.
2609 (modref_eaf_analysis::merge_call_lhs_flags): Likewise.
2610 (callee_to_caller_flags): Likewise.
2611 (modref_eaf_analysis::analyze_ssa_name): Likewise.
2612 (modref_eaf_analysis::propagate): Likewise.
2613 (modref_merge_call_site_flags): Likewise.
2614 * ipa-modref.h (interposable_eaf_flags): Likewise.
2615 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1) Likewise.
2616 * tree-ssa-structalias.c (handle_call_arg): Likewise.
2617 (handle_rhs_call): Likewise.
2618 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Likewise.
2620 2021-11-10 Tamar Christina <tamar.christina@arm.com>
2622 * tree-vect-slp-patterns.c (complex_mul_pattern::matches): Remove l1node.
2624 2021-11-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2626 * config/aarch64/aarch64-builtins.c
2627 (aarch64_general_gimple_fold_builtin): Change pointer alignment and
2630 2021-11-10 Jan Hubicka <jh@suse.cz>
2632 * ipa-modref-tree.h (modref_tree::remap_params): Fix off-by-one error.
2634 2021-11-10 H.J. Lu <hongjiu.lu@intel.com>
2635 Hongtao Liu <hongtao.liu@intel.com>
2637 PR middle-end/102566
2638 * match.pd (nop_atomic_bit_test_and_p): New match.
2639 * tree-ssa-ccp.c (convert_atomic_bit_not): New function.
2640 (gimple_nop_atomic_bit_test_and_p): New prototype.
2641 (optimize_atomic_bit_test_and): Transform equivalent, but slighly
2642 different cases to their canonical forms.
2644 2021-11-10 liuhongt <hongtao.liu@intel.com>
2646 PR tree-optimization/103126
2647 * tree-vect-loop.c (neutral_op_for_reduction): Remove static.
2648 * tree-vectorizer.h (neutral_op_for_reduction): Declare.
2649 * tree-if-conv.c : Include tree-vectorizer.h.
2650 (is_cond_scalar_reduction): Handle
2651 BIT_XOR_EXPR/BIT_IOR_EXPR/BIT_AND_EXPR.
2652 (convert_scalar_cond_reduction): Ditto.
2654 2021-11-10 konglin1 <lingling.kong@intel.com>
2656 * config/i386/sse.md (cmul<conj_op><mode>3): add new define_expand.
2657 (cmla<conj_op><mode>4): Likewise
2659 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
2661 * tree-ssa-threadedge.c: Do not include
2662 gimple-ssa-evrp-analyze.h.
2663 * value-pointer-equiv.cc: Same.
2665 2021-11-10 Aldy Hernandez <aldyh@redhat.com>
2667 * tree-ssa-threadbackward.c
2668 (back_threader::maybe_register_path_dump): Abstract path dumping...
2669 (dump_path): ...here.
2670 (back_threader::resolve_phi): Call dump_path.
2673 2021-11-10 konglin1 <lingling.kong@intel.com>
2675 * config/i386/sse.md (fma_<complexpairopname>_<mode>_pair):
2676 Add new define_insn.
2677 (fma_<mode>_fmaddc_bcst): Add new define_insn_and_split.
2678 (fma_<mode>_fcmaddc_bcst): Likewise
2680 2021-11-10 liuhongt <hongtao.liu@intel.com>
2683 * match.pd: Simplify (trunc)fmax/fmin((extend)a, (extend)b) to
2686 2021-11-10 Andrew Pinski <apinski@marvell.com>
2689 * config/aarch64/aarch64.c (aarch64_evpc_ins): Don't use target
2690 as an input, use original one.
2692 2021-11-10 Sandra Loosemore <sandra@codesourcery.com>
2694 * config/nios2/nios2.c (nios2_can_inline_p): New.
2695 (TARGET_CAN_INLINE_P): Define.
2697 2021-11-09 Bill Schmidt <wschmidt@linux.ibm.com>
2699 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin):
2700 Disable gimple fold for RS6000_BIF_{XVMINDP,XVMINSP,VMINFP} and
2701 RS6000_BIF_{XVMAXDP,XVMAXSP,VMAXFP} when fast-math is not set.
2702 (lxvrse_expand_builtin): Modify the expansion for sign extension.
2703 All extensions are done within VSX registers.
2705 2021-11-09 Andrew Pinski <apinski@marvell.com>
2707 PR tree-optimization/10352
2708 * match.pd: Remove check of TYPE_PRECISION for
2711 2021-11-09 Andrew MacLeod <amacleod@redhat.com>
2713 * function.c (allocate_struct_function): Don't set x_range_query.
2714 * function.h (get_range_query): Move to value-query.h.
2715 * gimple-range.cc (enable_ranger): Check that query is currently NULL.
2716 (disable_ranger): Clear function current query field.
2717 * value-query.cc (get_global_range_query): Relocate to:
2718 * value-query.h (get_global_range_query): Here and inline.
2719 (get_range_query): Relocate here from function.h.
2721 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
2723 * tree-ssa-threadbackward.c
2724 (back_threader::maybe_register_path_dump): New.
2725 (back_threader::maybe_register_path): Call maybe_register_path_dump.
2727 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
2729 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
2730 Return NULL when unprofitable.
2732 2021-11-09 Martin Jambor <mjambor@suse.cz>
2734 * tree.h (build_debug_expr_decl): Declare.
2735 * tree.c (build_debug_expr_decl): New function.
2736 * cfgexpand.c (avoid_deep_ter_for_debug): Use build_debug_expr_decl
2737 instead of building a DEBUG_EXPR_DECL.
2738 * ipa-param-manipulation.c
2739 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
2740 * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Likewise.
2741 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
2742 * tree-ssa-phiopt.c (spaceship_replacement): Likewise.
2743 * tree-ssa-reassoc.c (make_new_ssa_for_def): Likewise.
2745 2021-11-09 Bill Schmidt <wschmidt@linux.ibm.com>
2747 * config/rs6000/rs6000-call.c (rs6000_type_string): New function.
2748 (def_builtin): Change debug formatting for easier parsing and
2749 include more information.
2750 (rs6000_init_builtins): Add dump of autogenerated builtins.
2751 (altivec_init_builtins): Dump __builtin_altivec_mask_for_load for
2754 2021-11-09 Marek Polacek <polacek@redhat.com>
2756 * ipa-modref.c (analyze_parms): Fix a typo.
2758 2021-11-09 Kito Cheng <kito.cheng@sifive.com>
2761 * common/config/riscv/riscv-common.c (multi_letter_subset_rank): Remove
2762 assertion for Z*-ext.
2764 2021-11-09 Jan Hubicka <hubicka@ucw.cz>
2766 * ipa-modref.c (analyze_parms): Add past_flags, past_retslot_flags
2767 and past_static_chain; merge past summary with current one.
2768 (analyze_function): Update.
2770 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
2772 * doc/invoke.texi (Invoking GCC): Document --param=threader-debug.
2774 2021-11-09 Martin Liska <mliska@suse.cz>
2776 * print-rtl.c (rtx_writer::rtx_writer): Make the compilation
2777 conditional based on
2778 * print-rtl.h (class rtx_writer): Likewise.
2780 2021-11-09 Thomas Schwinge <thomas@codesourcery.com>
2782 * input.h (location_hash): Use 'BUILTINS_LOCATION' as spare value
2783 for 'Deleted'. Turn into a '#define'.
2785 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
2787 * gimple-range-path.cc (path_range_query::dump): Clean up.
2788 (path_range_query::compute_ranges): Same.
2789 * value-relation.cc (path_oracle::dump): Same.
2791 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
2793 * dumpfile.c (dump_options): Remove TDF_THREADING entry.
2794 * dumpfile.h (enum dump_flag): Remove TDF_THREADING and adjust
2796 * flag-types.h (enum threader_debug): New.
2797 * gimple-range-path.cc (DEBUG_SOLVER): Use param_threader_debug.
2798 * params.opt: Add entry for --param=threader-debug=.
2800 2021-11-09 Thomas Schwinge <thomas@codesourcery.com>
2802 PR middle-end/101204
2804 * diagnostic-spec.h (typedef xint_hash_t): Turn into...
2805 (struct xint_hash_t): ... this.
2806 * doc/gty.texi: Update.
2808 2021-11-09 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
2810 * config/arm/arm-cpus.in (armv9): New define.
2811 (ARMv9a): New group.
2812 (armv9-a): New arch definition.
2813 * config/arm/arm-tables.opt: Regenerate.
2814 * config/arm/arm.h (BASE_ARCH_9A): New arch enum value.
2815 * config/arm/t-aprofile: Added armv9-a and armv9+simd.
2816 * config/arm/t-arm-elf: Added arm9-a, v9_fps and all_v9_archs
2817 to MULTILIB_MATCHES.
2818 * config/arm/t-multilib: Added v9_a_nosimd_variants and
2819 v9_a_simd_variants to MULTILIB_MATCHES.
2820 * doc/invoke.texi: Update docs.
2822 2021-11-09 Martin Jambor <mjambor@suse.cz>
2825 * ipa-param-manipulation.c (replace_with_mapped_expr): Early
2826 return with error_mark_mode when part of expression is mapped to
2828 (ipa_param_body_adjustments::remap_with_debug_expressions): Set
2829 mapped value to NULL if walk_tree returns error_mark_mode.
2831 2021-11-09 Aldy Hernandez <aldyh@redhat.com>
2833 * tree-ssa-strlen.c (compare_nonzero_chars): Pass statement
2835 (get_addr_stridx): Same.
2837 (get_range_strlen_dynamic): Same.
2838 (handle_builtin_strlen): Same.
2839 (handle_builtin_strchr): Same.
2840 (handle_builtin_strcpy): Same.
2841 (maybe_diag_stxncpy_trunc): Same.
2842 (handle_builtin_stxncpy_strncat): Same.
2843 (handle_builtin_memcpy): Same.
2844 (handle_builtin_strcat): Same.
2845 (handle_alloc_call): Same.
2846 (handle_builtin_memset): Same.
2847 (handle_builtin_string_cmp): Same.
2848 (handle_pointer_plus): Same.
2849 (count_nonzero_bytes_addr): Same.
2850 (count_nonzero_bytes): Same.
2851 (handle_store): Same.
2852 (fold_strstr_to_strncmp): Same.
2853 (handle_integral_assign): Same.
2854 (check_and_optimize_stmt): Same.
2855 (class strlen_dom_walker): Replace evrp with ranger.
2856 (strlen_dom_walker::before_dom_children): Remove evrp.
2857 (strlen_dom_walker::after_dom_children): Remove evrp.
2858 * gimple-ssa-warn-access.cc (maybe_check_access_sizes):
2859 Restrict sprintf output.
2861 2021-11-09 Martin Liska <mliska@suse.cz>
2863 * genconditions.c (write_one_condition): Add const qualifier
2866 2021-11-09 Jeff Law <jeffreyalaw@gmail.com>
2868 * match.pd: New pattern to simplify (1 << n) & M ==/!= 0 for M
2871 2021-11-08 Bill Schmidt <wschmidt@linux.ibm.com>
2873 * config/rs6000/rs6000-builtin-new.def (VEC_INIT_V16QI): Use
2874 escape-newline support.
2875 (VEC_INIT_V4SI): Likewise.
2876 (VEC_INIT_V8HI): Likewise.
2877 (PACK_V1TI): Likewise.
2879 (VFIRSTMISMATCHOREOSINDEX_V16QI): Likewise.
2880 (VFIRSTMISMATCHOREOSINDEX_V8HI): Likewise.
2881 (VFIRSTMISMATCHOREOSINDEX_V4SI): Likewise.
2885 (FMAF128_ODD): Likewise.
2886 (VSCEQPUO): Likewise.
2888 (VSIEQPF): Likewise.
2889 (VSTDCQP): Likewise.
2890 (PACK_TD): Likewise.
2891 (TABORTDC): Likewise.
2892 (TABORTDCI): Likewise.
2893 (SE_LXVRBX): Likewise.
2894 (SE_LXVRHX): Likewise.
2895 (SE_LXVRWX): Likewise.
2896 (SE_LXVRDX): Likewise.
2897 (VREPLACE_UN_UV2DI): Likewise.
2898 (VREPLACE_UN_UV4SI): Likewise.
2899 (VREPLACE_UN_V2DI): Likewise.
2900 (VREPLACE_ELT_UV2DI): Likewise.
2901 (VREPLACE_ELT_V2DI): Likewise.
2902 (ZE_LXVRBX): Likewise.
2903 (ZE_LXVRHX): Likewise.
2904 (ZE_LXVRWX): Likewise.
2905 (ZE_LXVRDX): Likewise.
2907 (CNTLZDM): Likewise.
2908 (CNTTZDM): Likewise.
2911 (PMXVBF16GER2): Likewise.
2912 (PMXVBF16GER2_INTERNAL): Likewise.
2913 (PMXVBF16GER2NN): Likewise.
2914 (PMXVBF16GER2NN_INTERNAL): Likewise.
2915 (PMXVBF16GER2NP): Likewise.
2916 (PMXVBF16GER2NP_INTERNAL): Likewise.
2917 (PMXVBF16GER2PN): Likewise.
2918 (PMXVBF16GER2PN_INTERNAL): Likewise.
2919 (PMXVBF16GER2PP): Likewise.
2920 (PMXVBF16GER2PP_INTERNAL): Likewise.
2921 (PMXVF16GER2): Likewise.
2922 (PMXVF16GER2_INTERNAL): Likewise.
2923 (PMXVF16GER2NN): Likewise.
2924 (PMXVF16GER2NN_INTERNAL): Likewise.
2925 (PMXVF16GER2NP): Likewise.
2926 (PMXVF16GER2NP_INTERNAL): Likewise.
2927 (PMXVF16GER2PN): Likewise.
2928 (PMXVF16GER2PN_INTERNAL): Likewise.
2929 (PMXVF16GER2PP): Likewise.
2930 (PMXVF16GER2PP_INTERNAL): Likewise.
2931 (PMXVF32GER_INTERNAL): Likewise.
2932 (PMXVF32GERNN): Likewise.
2933 (PMXVF32GERNN_INTERNAL): Likewise.
2934 (PMXVF32GERNP): Likewise.
2935 (PMXVF32GERNP_INTERNAL): Likewise.
2936 (PMXVF32GERPN): Likewise.
2937 (PMXVF32GERPN_INTERNAL): Likewise.
2938 (PMXVF32GERPP): Likewise.
2939 (PMXVF32GERPP_INTERNAL): Likewise.
2940 (PMXVF64GER): Likewise.
2941 (PMXVF64GER_INTERNAL): Likewise.
2942 (PMXVF64GERNN): Likewise.
2943 (PMXVF64GERNN_INTERNAL): Likewise.
2944 (PMXVF64GERNP): Likewise.
2945 (PMXVF64GERNP_INTERNAL): Likewise.
2946 (PMXVF64GERPN): Likewise.
2947 (PMXVF64GERPN_INTERNAL): Likewise.
2948 (PMXVF64GERPP): Likewise.
2949 (PMXVF64GERPP_INTERNAL): Likewise.
2950 (PMXVI16GER2): Likewise.
2951 (PMXVI16GER2_INTERNAL): Likewise.
2952 (PMXVI16GER2PP): Likewise.
2953 (PMXVI16GER2PP_INTERNAL): Likewise.
2954 (PMXVI16GER2S): Likewise.
2955 (PMXVI16GER2S_INTERNAL): Likewise.
2956 (PMXVI16GER2SPP): Likewise.
2957 (PMXVI16GER2SPP_INTERNAL): Likewise.
2958 (PMXVI4GER8): Likewise.
2959 (PMXVI4GER8_INTERNAL): Likewise.
2960 (PMXVI4GER8PP): Likewise.
2961 (PMXVI4GER8PP_INTERNAL): Likewise.
2962 (PMXVI8GER4): Likewise.
2963 (PMXVI8GER4_INTERNAL): Likewise.
2964 (PMXVI8GER4PP): Likewise.
2965 (PMXVI8GER4PP_INTERNAL): Likewise.
2966 (PMXVI8GER4SPP): Likewise.
2967 (PMXVI8GER4SPP_INTERNAL): Likewise.
2968 * config/rs6000/rs6000-gen-builtins.c (MAXLINES): New macro.
2969 (linebuf): Increase size.
2970 (lines): New variable.
2971 (lastline): Likewise.
2972 (real_line_pos): New function.
2973 (diag): Change signature.
2974 (bif_diag): Change signature; support escape-newline handling.
2975 (ovld_diag): Likewise.
2976 (fatal): Move earlier.
2977 (consume_whitespace): Adjust diag call.
2978 (advance_line): Add escape-newline handling; call fatal.
2979 (safe_inc_pos): Adjust diag call.
2980 (match_identifier): Likewise.
2981 (match_integer): Likewise.
2982 (match_to_right_bracket): Call fatal instead of diag; adjust diag
2984 (match_basetype): Adjust diag calls.
2985 (match_bracketed_pair): Likewise.
2986 (match_const_restriction): Likewise.
2987 (match_type): Likewise.
2988 (parse_args): Likewise.
2989 (parse_bif_attrs): Likewise.
2990 (complete_vector_type): Likewise.
2991 (complete_base_type): Likewise.
2992 (parse_prototype): Likewise.
2993 (parse_bif_entry): Likewise.
2994 (parse_bif_stanza): Likewise.
2995 (parse_ovld_entry): Likewise.
2996 (parse_ovld_stanza): Likewise.
2997 (main): Allocate buffers for lines[].
2999 2021-11-08 Eric Botcazou <ebotcazou@adacore.com>
3001 * config/i386/i386.c (legitimize_pic_address): Adjust comment and
3002 use the REG argument on the CM_LARGE_PIC code path as well.
3003 * config/i386/predicates.md (gotoff_operand): Do not treat VxWorks
3004 specially with the large code models.
3006 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
3008 * gimple.c (gimple_call_static_chain_flags): Revert the workaround
3009 allowing interposition since issues with binds_to_local_def were
3012 2021-11-08 Andrew MacLeod <amacleod@redhat.com>
3014 PR tree-optimization/103122
3015 * gimple-range.cc (gimple_ranger::range_of_expr): Request the cache
3016 entry with "calulate new values" set to false.
3018 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
3020 * builtins.c (is_simple_builtin): Add builitin_dwarf_cfa
3021 and builtin_return_address.
3022 (builtin_fnspec): Annotate builtin_return,
3023 bulitin_eh_pointer, builtin_eh_filter, builtin_unwind_resume,
3024 builtin_cxa_end_cleanup, builtin_eh_copy_values,
3025 builtin_frame_address, builtin_apply_args,
3026 builtin_asan_before_dynamic_init, builtin_asan_after_dynamic_init,
3027 builtin_prefetch, builtin_dwarf_cfa, builtin_return_addrss
3029 * ipa-pure-const.c (special_builtin_state): Add builtin_dwarf_cfa
3030 and builtin_return_address.
3032 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
3034 PR tree-optimization/103177
3035 * passes.def: Move uncprop after pure/const and modref.
3037 2021-11-08 Martin Jambor <mjambor@suse.cz>
3041 * tree-inline.c (remap_gimple_stmt): Unshare the expression without
3042 location before invoking remap_with_debug_expressions on it.
3043 * ipa-param-manipulation.c
3044 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
3046 2021-11-08 David Edelsohn <dje.gcc@gmail.com>
3048 * config/rs6000/vsx.md (vsx_splat_v4si_di): Revert "wa"
3051 2021-11-08 Richard Biener <rguenther@suse.de>
3053 * cfgloop.c (verify_loop_structure): Use a temporary BB flag
3054 instead of an sbitmap to cache irreducible state.
3056 2021-11-08 Aldy Hernandez <aldyh@redhat.com>
3058 PR tree-optimization/103120
3059 * gimple-range-path.cc (path_range_query::range_defined_in_block):
3060 Bail if there's a cache entry.
3062 2021-11-08 Bill Schmidt <wschmidt@linux.ibm.com>
3064 * config/rs6000/rs6000.c (rs6000_builtin_reciprocal): Use
3065 rs6000_builtin_decls_x when appropriate.
3066 (add_condition_to_bb): Likewise.
3067 (rs6000_atomic_assign_expand_fenv): Likewise.
3069 2021-11-08 Bill Schmidt <wschmidt@linux.ibm.com>
3071 * config/rs6000/rs6000-call.c (rs6000_new_builtin_decl): New function.
3072 (rs6000_builtin_decl): Call it.
3074 2021-11-08 Martin Liska <mliska@suse.cz>
3076 * cgraph.c (cgraph_node::dump): Dump it from decl.
3078 2021-11-08 Martin Liska <mliska@suse.cz>
3080 PR gcov-profile/100520
3081 * coverage.c (coverage_compute_profile_id): Strip .gk when
3082 compare debug is used.
3083 * system.h (endswith): New function.
3085 2021-11-08 Martin Liska <mliska@suse.cz>
3087 * cgraph.c (cgraph_node::dump): Dump static_chain_decl.
3089 2021-11-08 Thomas Schwinge <thomas@codesourcery.com>
3091 * config/rs6000/rbtree.c: Fix 'Copyright (C) 2020-21' into '2020-2021'
3092 * config/rs6000/rbtree.h: Likewise.
3093 * config/rs6000/rs6000-builtin-new.def: Likewise.
3094 * config/rs6000/rs6000-gen-builtins.c: Likewise.
3095 * config/rs6000/rs6000-overload.def: Likewise.
3097 2021-11-08 Richard Sandiford <richard.sandiford@arm.com>
3099 * config/aarch64/aarch64.c (aarch64_ldn_stn_vectors): New function.
3100 (aarch64_address_cost): Use it instead of testing for CImode and
3103 2021-11-08 Richard Sandiford <richard.sandiford@arm.com>
3105 * genmodes.c (emit_insn_modes_h): Define NUM_MODE_* macros.
3106 * expmed.h (NUM_MODE_INT): Delete in favor of genmodes definitions.
3107 (NUM_MODE_PARTIAL_INT, NUM_MODE_VECTOR_INT): Likewise.
3108 * real.h (real_format_for_mode): Use NUM_MODE_FLOAT and
3109 NUM_MODE_DECIMAL_FLOAT.
3110 (REAL_MODE_FORMAT): Likewise.
3112 2021-11-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
3114 * tree-vect-loop.c (vect_better_loop_vinfo_p): Change how epilogue loop
3117 2021-11-08 Richard Biener <rguenther@suse.de>
3119 * tree-vectorizer.h (vect_create_loop_vinfo): Add main_loop_info
3121 * tree-vect-loop.c (vect_create_loop_vinfo): Likewise. Set
3122 LOOP_VINFO_ORIG_LOOP_INFO and conditionalize set of
3123 LOOP_VINFO_NITERS_ASSUMPTIONS.
3124 (vect_analyze_loop_1): Adjust.
3125 (vect_analyze_loop): Move loop constraint setting and
3126 SCEV/niter reset here from vect_create_loop_vinfo to perform
3128 (vect_analyze_loop_form): Move dumping of symbolic niters
3129 here from vect_create_loop_vinfo.
3131 2021-11-08 Jan Hubicka <hubicka@ucw.cz>
3133 * ipa-modref-tree.h (enum modref_special_parms): New enum.
3134 (struct modref_access_node): update for special parms.
3135 (struct modref_ref_node): Likewise.
3136 (struct modref_parm_map): Likewise.
3137 (struct modref_tree): Likewise.
3138 * ipa-modref.c (dump_access): Likewise.
3139 (get_access): Detect static chain.
3140 (parm_map_for_arg): Take tree as arg instead of
3142 (merge_call_side_effects): Compute map for static chain.
3143 (process_fnspec): Update.
3144 (struct escape_point): Remove retslot_arg and static_chain_arg.
3145 (analyze_parms): Update.
3146 (compute_parm_map): Update.
3147 (propagate_unknown_call): Update.
3148 (modref_propagate_in_scc): Update.
3149 (modref_merge_call_site_flags): Update.
3150 (ipa_merge_modref_summary_after_inlining): Update.
3151 * tree-ssa-alias.c (modref_may_conflict): Handle static chain.
3152 * ipa-modref-tree.c (test_merge): Update.
3154 2021-11-08 Haochen Gui <guihaoc@gcc.gnu.org>
3156 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin): Disable
3157 gimple fold for VSX_BUILTIN_XVMINDP, ALTIVEC_BUILTIN_VMINFP,
3158 VSX_BUILTIN_XVMAXDP, ALTIVEC_BUILTIN_VMAXFP when fast-math is not
3161 2021-11-08 liuhongt <hongtao.liu@intel.com>
3163 PR tree-optimization/103077
3164 * doc/invoke.texi (Options That Control Optimization):
3165 Update documentation for -ftree-loop-vectorize and
3166 -ftree-slp-vectorize which are enabled by default at -02.
3168 2021-11-08 liuhongt <hongtao.liu@intel.com>
3171 * match.pd (Simplifcation (trunc)copysign((extend)a, (extend)b)
3172 to .COPYSIGN (a, b)): Add !HONOR_SNANS.
3174 2021-11-08 liuhongt <hongtao.liu@intel.com>
3177 * match.pd: Simplify
3178 (trunc)fma ((extend)a, (extend)b, (extend)c) to IFN_FMA (a, b,
3179 c) under flag_unsafe_math_optimizations.
3181 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
3184 * params.opt (modref-max-depth): Add range.
3185 (modref-max-adjustments): Fix range.
3187 2021-11-07 Aldy Hernandez <aldyh@redhat.com>
3189 * tree-pass.h (make_pass_vrp_threader): Remove.
3190 * tree-ssa-threadbackward.c
3191 (back_threader_profitability::profitable_path_p): Remove
3192 ASSERT_EXPR references.
3193 * tree-ssa-threadedge.c (jt_state::register_equivs_stmt): Same.
3194 * tree-vrp.c (vrp_folder::simplify_casted_conds): Same.
3195 (execute_vrp): Same.
3196 (class hybrid_threader): Remove.
3197 (hybrid_threader::hybrid_threader): Remove.
3198 (hybrid_threader::~hybrid_threader): Remove.
3199 (hybrid_threader::before_dom_children): Remove.
3200 (hybrid_threader::after_dom_children): Remove.
3201 (execute_vrp_threader): Remove.
3202 (class pass_vrp_threader): Remove.
3203 (make_pass_vrp_threader): Remove.
3205 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
3207 * gimple.c (gimple_call_arg_flags): Use interposable_eaf_flags.
3208 (gimple_call_retslot_flags): Likewise.
3209 (gimple_call_static_chain_flags): Likewise.
3210 * ipa-modref.c (remove_useless_eaf_flags): Do not remove everything for
3212 (modref_summary::useful_p): Likewise.
3213 (modref_summary_lto::useful_p): Likewise.
3214 (analyze_parms): Do not give up on NOVOPS.
3215 (analyze_function): When dumping report chnages in EAF flags
3216 between IPA and local pass.
3217 (modref_merge_call_site_flags): Compute implicit eaf flags
3218 based on callee ecf_flags and fnspec; if the function does not
3219 bind to current defs use interposable_eaf_flags.
3220 (modref_propagate_flags_in_scc): Update.
3221 * ipa-modref.h (interposable_eaf_flags): New function.
3223 2021-11-07 Bill Schmidt <wschmidt@linux.ibm.com>
3225 * config/rs6000/rs6000-call.c (rs6000_expand_new_builtin): New
3227 (rs6000_invalid_new_builtin): New function.
3228 (rs6000_expand_builtin): Call rs6000_expand_new_builtin.
3229 (rs6000_expand_ldst_mask): New function.
3230 (new_cpu_expand_builtin): Likewise.
3231 (elemrev_icode): Likewise.
3232 (ldv_expand_builtin): Likewise.
3233 (lxvrse_expand_builtin): Likewise.
3234 (lxvrze_expand_builtin): Likewise.
3235 (stv_expand_builtin): Likewise.
3236 (new_mma_expand_builtin): Likewise.
3237 (new_htm_spr_num): Likewise.
3238 (new_htm_expand_builtin): Likewise.
3239 (rs6000_expand_new_builtin): Likewise.
3240 (rs6000_init_builtins): Initialize altivec_builtin_mask_for_load.
3242 2021-11-07 Jan Hubicka <hubicka@ucw.cz>
3244 * ipa-modref.c (modref_lattice): Add do_dataflow,
3245 changed and propagate_to fields.
3246 (modref_lattice::release): Free propagate_to
3247 (modref_lattice::merge): Do not give up early on unknown
3249 (modref_lattice::merge_deref): Likewise.
3250 (modref_eaf_analysis): Update toplevel comment.
3251 (modref_eaf_analysis::analyze_ssa_name): Record postponned ssa names;
3252 do optimistic dataflow initialization.
3253 (modref_eaf_analysis::merge_with_ssa_name): Build dataflow graph.
3254 (modref_eaf_analysis::propagate): New member function.
3255 (analyze_parms): Update to new API of modref_eaf_analysis.
3257 2021-11-06 Jan Hubicka <hubicka@ucw.cz>
3259 * cgraph.h (cgraph_node::can_be_discarded_p): Do not
3260 return true on functions from other partition.
3262 2021-11-06 David Edelsohn <dje.gcc@gmail.com>
3264 * config/rs6000/vsx.md (vsx_splat_v4si): Change constraints to "wa".
3265 (vsx_splat_v4si_di): Change constraint to "wa".
3267 2021-11-06 Aldy Hernandez <aldyh@redhat.com>
3269 PR tree-optimization/103061
3270 * value-relation.cc (path_oracle::path_oracle): Initialize
3272 (path_oracle::killing_def): Set m_killed_defs.
3273 (path_oracle::query_relation): Do not look at the root oracle for
3275 * value-relation.h (class path_oracle): Add m_killed_defs.
3277 2021-11-06 Aldy Hernandez <aldyh@redhat.com>
3279 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
3280 Remove gotos and other cleanups.
3282 2021-11-05 Jan Hubicka <hubicka@ucw.cz>
3285 * ipa-modref-tree.h (modref_tree::insert): Do nothing for
3286 paradoxical and zero sized accesses.
3288 2021-11-05 Jan Hubicka <hubicka@ucw.cz>
3291 * ipa-modref-tree.h (struct modref_access_node): Avoid left shift
3294 2021-11-05 Iain Sandoe <iain@sandoe.co.uk>
3296 * config/darwin.h (ASM_GENERATE_INTERNAL_LABEL): Add LTRAMP
3297 to the list of symbol prefixes that must be made linker-
3300 2021-11-05 Iain Sandoe <iain@sandoe.co.uk>
3302 * config.host: Add support for aarch64-*-darwin.
3303 * config/aarch64/host-aarch64-darwin.c: New file.
3304 * config/aarch64/x-darwin: New file.
3306 2021-11-05 Andrew MacLeod <amacleod@redhat.com>
3308 PR tree-optimization/103093
3309 * gimple-range-gori.cc (range_def_chain::get_imports): Remove assert.
3311 2021-11-05 Andrew MacLeod <amacleod@redhat.com>
3313 PR tree-optimization/102943
3314 * gimple-range-cache.cc (class update_list): New.
3315 (update_list::add): Replace add_to_update.
3316 (update_list::pop): New.
3317 (ranger_cache::ranger_cache): Adjust.
3318 (ranger_cache::~ranger_cache): Adjust.
3319 (ranger_cache::add_to_update): Delete.
3320 (ranger_cache::propagate_cache): Adjust to new class.
3321 (ranger_cache::propagate_updated_value): Ditto.
3322 (ranger_cache::fill_block_cache): Ditto.
3323 * gimple-range-cache.h (class ranger_cache): Adjust to update class.
3325 2021-11-05 Richard Biener <rguenther@suse.de>
3327 * tree-vect-loop.c (vect_analyze_loop): Remove obsolete
3328 comment and expand on another one. Combine nested if.
3330 2021-11-05 John David Anglin <danglin@gcc.gnu.org>
3333 * config/pa/pa-modes.def: Add OImode integer type.
3334 * config/pa/pa.c (pa_scalar_mode_supported_p): Allow TImode
3336 * config/pa/pa.h (MIN_UNITS_PER_WORD) Define to MIN_UNITS_PER_WORD
3337 to UNITS_PER_WORD if IN_LIBGCC2.
3338 * config/pa/pa.md (addti3, addvti3, subti3, subvti3, negti2,
3339 negvti2, ashlti3, shrpd_internal): New patterns.
3340 Change some multi instruction types to multi.
3342 2021-11-05 Jakub Jelinek <jakub@redhat.com>
3345 * config/i386/i386.h
3346 (stringop_algs::stringop_strategy::stringop_strategy): Make the ctor
3349 2021-11-05 Wilco Dijkstra <wdijkstr@arm.com>
3352 * config/aarch64/aarch64.c (aarch64_mov_operand_p): Strip the salt
3354 * config/aarch64/constraints.md: Support const in Usw.
3356 2021-11-05 John David Anglin <danglin@gcc.gnu.org>
3358 * config/pa/pa.h (PREFERRED_DEBUGGING_TYPE): Define to DWARF2_DEBUG.
3359 * config/pa/pa64-hpux.h (PREFERRED_DEBUGGING_TYPE): Remove define.
3361 2021-11-05 Richard Biener <rguenther@suse.de>
3363 * tree-vectorizer.h (vec_info_shared::n_stmts): Add.
3364 (LOOP_VINFO_N_STMTS): Likewise.
3365 (vec_info_for_bb): Remove unused function.
3366 * tree-vectorizer.c (vec_info_shared::vec_info_shared):
3367 Initialize n_stmts member.
3368 * tree-vect-loop.c: Remove INCLUDE_FUNCTIONAL.
3369 (vect_create_loop_vinfo): Do not set loop->aux.
3370 (vect_analyze_loop_2): Do not get n_stmts as argument,
3371 instead use LOOP_VINFO_N_STMTS. Set LOOP_VINFO_VECTORIZABLE_P
3373 (vect_analyze_loop_1): Remove callback, get the mode iterator
3374 and autodetected_vector_mode as argument, advancing the
3375 iterator and initializing autodetected_vector_mode here.
3376 (vect_analyze_loop): Split analysis loop into two, first
3377 processing main loops only and then epilogues.
3379 2021-11-05 Martin Jambor <mjambor@suse.cz>
3381 * ipa-prop.c (compute_complex_assign_jump_func): Remove
3382 unnecessary check for RECORD_TYPE.
3384 2021-11-05 Gerald Pfeifer <gerald@pfeifer.com>
3386 * doc/install.texi2html: Do not generate old.html any longer.
3388 2021-11-05 Martin Liska <mliska@suse.cz>
3391 * opts.c (finish_options): Reset flag_gtoggle when it is used.
3393 2021-11-05 Jakub Jelinek <jakub@redhat.com>
3396 * dwarf2out.c (add_const_value_attribute): Add MODE argument, use it
3397 in CONST_WIDE_INT handling. Adjust recursive calls.
3398 (add_location_or_const_value_attribute): Pass DECL_MODE (decl) to
3399 new add_const_value_attribute argument.
3400 (tree_add_const_value_attribute): Pass TYPE_MODE (type) to new
3401 add_const_value_attribute argument.
3403 2021-11-05 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3405 * config/vx-common.h: Test value of TARGET_VXWORKS7 rather
3408 2021-11-05 Richard Biener <rguenther@suse.de>
3410 * tree-vectorizer.h (struct vect_loop_form_info): New.
3411 (vect_analyze_loop_form): Adjust.
3412 (vect_create_loop_vinfo): New.
3413 * tree-parloops.c (gather_scalar_reductions): Adjust for
3414 vect_analyze_loop_form API change.
3415 * tree-vect-loop.c: Include <functional>.
3416 (vect_analyze_loop_form_1): Rename to vect_analyze_loop_form,
3417 take struct vect_loop_form_info as output parameter and adjust.
3418 (vect_analyze_loop_form): Rename to vect_create_loop_vinfo and
3419 split out call to the original vect_analyze_loop_form_1.
3420 (vect_reanalyze_as_main_loop): Rename to...
3421 (vect_analyze_loop_1): ... this, factor out the call to
3422 vect_analyze_loop_form and generalize to be able to use it twice ...
3423 (vect_analyze_loop): ... here. Perform vect_analyze_loop_form
3426 2021-11-05 Xionghu Luo <luoxhu@linux.ibm.com>
3429 * config/rs6000/fusion.md: Regenerate.
3430 * config/rs6000/genfusion.pl: Fix incorrect clobber constraint.
3432 2021-11-04 Andreas Krebbel <krebbel@linux.ibm.com>
3434 * config/s390/s390.h (STACK_CHECK_MOVING_SP): New macro
3437 2021-11-04 Tamar Christina <tamar.christina@arm.com>
3439 * config/aarch64/aarch64-builtins.c
3440 (aarch64_general_gimple_fold_builtin): Add ashl, sshl, ushl, ashr,
3441 ashr_simd, lshr, lshr_simd.
3442 * config/aarch64/aarch64-simd-builtins.def (lshr): Use USHIFTIMM.
3443 * config/aarch64/arm_neon.h (vshr_n_u8, vshr_n_u16, vshr_n_u32,
3444 vshrq_n_u8, vshrq_n_u16, vshrq_n_u32, vshrq_n_u64): Fix type hack.
3446 2021-11-04 Tamar Christina <tamar.christina@arm.com>
3448 * match.pd: New negate+shift pattern.
3450 2021-11-04 Andrew MacLeod <amacleod@redhat.com>
3452 PR tree-optimization/103079
3453 * gimple-range-gori.cc (gimple_range_calc_op1): Treat undefined as
3455 (gimple_range_calc_op2): Ditto.
3457 2021-11-04 Martin Jambor <mjambor@suse.cz>
3460 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
3461 members remap_with_debug_expressions, m_dead_ssa_debug_equiv,
3462 m_dead_stmt_debug_equiv and prepare_debug_expressions. Added
3463 parameter to mark_dead_statements.
3464 * ipa-param-manipulation.c: Include tree-phinodes.h and cfgexpand.h.
3465 (ipa_param_body_adjustments::mark_dead_statements): New parameter
3466 debugstack, push into it all SSA names used in debug statements,
3467 produce m_dead_ssa_debug_equiv mapping for the removed param.
3468 (replace_with_mapped_expr): New function.
3469 (ipa_param_body_adjustments::remap_with_debug_expressions): Likewise.
3470 (ipa_param_body_adjustments::prepare_debug_expressions): Likewise.
3471 (ipa_param_body_adjustments::common_initialization): Gather and
3472 procecc SSA which will be removed but are in debug statements. Simplify.
3473 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
3475 * tree-inline.c (remap_gimple_stmt): Create a debug bind when possible
3476 when avoiding a copy of an unnecessary statement. Remap removed SSA
3477 names in existing debug statements.
3478 (tree_function_versioning): Do not create DEBUG_EXPR_DECL for removed
3479 parameters if we have already done so.
3481 2021-11-04 Jan Hubicka <hubicka@ucw.cz>
3484 * gimple.c (gimple_call_static_chain_flags): Handle case when
3485 nested function does not bind locally.
3487 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
3489 * config/aarch64/aarch64.c (aarch64_function_value): Generate
3490 a register rtx for Neon vector-tuple modes.
3491 (aarch64_layout_arg): Likewise.
3493 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
3495 * lower-subreg.c (simple_move): Prevent decomposition if
3496 modes are not tieable.
3498 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
3499 Richard Sandiford <richard.sandiford@arm.com>
3501 * config/aarch64/aarch64-builtins.c (v2x8qi_UP): Define.
3502 (v2x4hi_UP): Likewise.
3503 (v2x4hf_UP): Likewise.
3504 (v2x4bf_UP): Likewise.
3505 (v2x2si_UP): Likewise.
3506 (v2x2sf_UP): Likewise.
3507 (v2x1di_UP): Likewise.
3508 (v2x1df_UP): Likewise.
3509 (v2x16qi_UP): Likewise.
3510 (v2x8hi_UP): Likewise.
3511 (v2x8hf_UP): Likewise.
3512 (v2x8bf_UP): Likewise.
3513 (v2x4si_UP): Likewise.
3514 (v2x4sf_UP): Likewise.
3515 (v2x2di_UP): Likewise.
3516 (v2x2df_UP): Likewise.
3517 (v3x8qi_UP): Likewise.
3518 (v3x4hi_UP): Likewise.
3519 (v3x4hf_UP): Likewise.
3520 (v3x4bf_UP): Likewise.
3521 (v3x2si_UP): Likewise.
3522 (v3x2sf_UP): Likewise.
3523 (v3x1di_UP): Likewise.
3524 (v3x1df_UP): Likewise.
3525 (v3x16qi_UP): Likewise.
3526 (v3x8hi_UP): Likewise.
3527 (v3x8hf_UP): Likewise.
3528 (v3x8bf_UP): Likewise.
3529 (v3x4si_UP): Likewise.
3530 (v3x4sf_UP): Likewise.
3531 (v3x2di_UP): Likewise.
3532 (v3x2df_UP): Likewise.
3533 (v4x8qi_UP): Likewise.
3534 (v4x4hi_UP): Likewise.
3535 (v4x4hf_UP): Likewise.
3536 (v4x4bf_UP): Likewise.
3537 (v4x2si_UP): Likewise.
3538 (v4x2sf_UP): Likewise.
3539 (v4x1di_UP): Likewise.
3540 (v4x1df_UP): Likewise.
3541 (v4x16qi_UP): Likewise.
3542 (v4x8hi_UP): Likewise.
3543 (v4x8hf_UP): Likewise.
3544 (v4x8bf_UP): Likewise.
3545 (v4x4si_UP): Likewise.
3546 (v4x4sf_UP): Likewise.
3547 (v4x2di_UP): Likewise.
3548 (v4x2df_UP): Likewise.
3549 (TYPES_GETREGP): Delete.
3550 (TYPES_SETREGP): Likewise.
3551 (TYPES_LOADSTRUCT_U): Define.
3552 (TYPES_LOADSTRUCT_P): Likewise.
3553 (TYPES_LOADSTRUCT_LANE_U): Likewise.
3554 (TYPES_LOADSTRUCT_LANE_P): Likewise.
3555 (TYPES_STORE1P): Move for consistency.
3556 (TYPES_STORESTRUCT_U): Define.
3557 (TYPES_STORESTRUCT_P): Likewise.
3558 (TYPES_STORESTRUCT_LANE_U): Likewise.
3559 (TYPES_STORESTRUCT_LANE_P): Likewise.
3560 (aarch64_simd_tuple_types): Define.
3561 (aarch64_lookup_simd_builtin_type): Handle tuple type lookup.
3562 (aarch64_init_simd_builtin_functions): Update frontend lookup
3563 for builtin functions after handling arm_neon.h pragma.
3564 (register_tuple_type): Manually set modes of single-integer
3565 tuple types. Record tuple types.
3566 * config/aarch64/aarch64-modes.def
3567 (ADV_SIMD_D_REG_STRUCT_MODES): Define D-register tuple modes.
3568 (ADV_SIMD_Q_REG_STRUCT_MODES): Define Q-register tuple modes.
3569 (SVE_MODES): Give single-vector modes priority over vector-
3571 (VECTOR_MODES_WITH_PREFIX): Set partial-vector mode order to
3572 be after all single-vector modes.
3573 * config/aarch64/aarch64-simd-builtins.def: Update builtin
3574 generator macros to reflect modifications to the backend
3576 * config/aarch64/aarch64-simd.md (aarch64_simd_ld2<mode>):
3577 Use vector-tuple mode iterator and rename to...
3578 (aarch64_simd_ld2<vstruct_elt>): This.
3579 (aarch64_simd_ld2r<mode>): Use vector-tuple mode iterator and
3581 (aarch64_simd_ld2r<vstruct_elt>): This.
3582 (aarch64_vec_load_lanesoi_lane<mode>): Use vector-tuple mode
3583 iterator and rename to...
3584 (aarch64_vec_load_lanes<mode>_lane<vstruct_elt>): This.
3585 (vec_load_lanesoi<mode>): Use vector-tuple mode iterator and
3587 (vec_load_lanes<mode><vstruct_elt>): This.
3588 (aarch64_simd_st2<mode>): Use vector-tuple mode iterator and
3590 (aarch64_simd_st2<vstruct_elt>): This.
3591 (aarch64_vec_store_lanesoi_lane<mode>): Use vector-tuple mode
3592 iterator and rename to...
3593 (aarch64_vec_store_lanes<mode>_lane<vstruct_elt>): This.
3594 (vec_store_lanesoi<mode>): Use vector-tuple mode iterator and
3596 (vec_store_lanes<mode><vstruct_elt>): This.
3597 (aarch64_simd_ld3<mode>): Use vector-tuple mode iterator and
3599 (aarch64_simd_ld3<vstruct_elt>): This.
3600 (aarch64_simd_ld3r<mode>): Use vector-tuple mode iterator and
3602 (aarch64_simd_ld3r<vstruct_elt>): This.
3603 (aarch64_vec_load_lanesci_lane<mode>): Use vector-tuple mode
3604 iterator and rename to...
3605 (vec_load_lanesci<mode>): This.
3606 (aarch64_simd_st3<mode>): Use vector-tuple mode iterator and
3608 (aarch64_simd_st3<vstruct_elt>): This.
3609 (aarch64_vec_store_lanesci_lane<mode>): Use vector-tuple mode
3610 iterator and rename to...
3611 (vec_store_lanesci<mode>): This.
3612 (aarch64_simd_ld4<mode>): Use vector-tuple mode iterator and
3614 (aarch64_simd_ld4<vstruct_elt>): This.
3615 (aarch64_simd_ld4r<mode>): Use vector-tuple mode iterator and
3617 (aarch64_simd_ld4r<vstruct_elt>): This.
3618 (aarch64_vec_load_lanesxi_lane<mode>): Use vector-tuple mode
3619 iterator and rename to...
3620 (vec_load_lanesxi<mode>): This.
3621 (aarch64_simd_st4<mode>): Use vector-tuple mode iterator and
3623 (aarch64_simd_st4<vstruct_elt>): This.
3624 (aarch64_vec_store_lanesxi_lane<mode>): Use vector-tuple mode
3625 iterator and rename to...
3626 (vec_store_lanesxi<mode>): This.
3627 (mov<mode>): Define for Neon vector-tuple modes.
3628 (aarch64_ld1x3<VALLDIF:mode>): Use vector-tuple mode iterator
3630 (aarch64_ld1x3<vstruct_elt>): This.
3631 (aarch64_ld1_x3_<mode>): Use vector-tuple mode iterator and
3633 (aarch64_ld1_x3_<vstruct_elt>): This.
3634 (aarch64_ld1x4<VALLDIF:mode>): Use vector-tuple mode iterator
3636 (aarch64_ld1x4<vstruct_elt>): This.
3637 (aarch64_ld1_x4_<mode>): Use vector-tuple mode iterator and
3639 (aarch64_ld1_x4_<vstruct_elt>): This.
3640 (aarch64_st1x2<VALLDIF:mode>): Use vector-tuple mode iterator
3642 (aarch64_st1x2<vstruct_elt>): This.
3643 (aarch64_st1_x2_<mode>): Use vector-tuple mode iterator and
3645 (aarch64_st1_x2_<vstruct_elt>): This.
3646 (aarch64_st1x3<VALLDIF:mode>): Use vector-tuple mode iterator
3648 (aarch64_st1x3<vstruct_elt>): This.
3649 (aarch64_st1_x3_<mode>): Use vector-tuple mode iterator and
3651 (aarch64_st1_x3_<vstruct_elt>): This.
3652 (aarch64_st1x4<VALLDIF:mode>): Use vector-tuple mode iterator
3654 (aarch64_st1x4<vstruct_elt>): This.
3655 (aarch64_st1_x4_<mode>): Use vector-tuple mode iterator and
3657 (aarch64_st1_x4_<vstruct_elt>): This.
3658 (*aarch64_mov<mode>): Define for vector-tuple modes.
3659 (*aarch64_be_mov<mode>): Likewise.
3660 (aarch64_ld<VSTRUCT:nregs>r<VALLDIF:mode>): Use vector-tuple
3661 mode iterator and rename to...
3662 (aarch64_ld<nregs>r<vstruct_elt>): This.
3663 (aarch64_ld2<mode>_dreg): Use vector-tuple mode iterator and
3665 (aarch64_ld2<vstruct_elt>_dreg): This.
3666 (aarch64_ld3<mode>_dreg): Use vector-tuple mode iterator and
3668 (aarch64_ld3<vstruct_elt>_dreg): This.
3669 (aarch64_ld4<mode>_dreg): Use vector-tuple mode iterator and
3671 (aarch64_ld4<vstruct_elt>_dreg): This.
3672 (aarch64_ld<VSTRUCT:nregs><VDC:mode>): Use vector-tuple mode
3673 iterator and rename to...
3674 (aarch64_ld<nregs><vstruct_elt>): Use vector-tuple mode
3675 iterator and rename to...
3676 (aarch64_ld<VSTRUCT:nregs><VQ:mode>): Use vector-tuple mode
3677 (aarch64_ld1x2<VQ:mode>): Delete.
3678 (aarch64_ld1x2<VDC:mode>): Use vector-tuple mode iterator and
3680 (aarch64_ld1x2<vstruct_elt>): This.
3681 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use vector-
3682 tuple mode iterator and rename to...
3683 (aarch64_ld<nregs>_lane<vstruct_elt>): This.
3684 (aarch64_get_dreg<VSTRUCT:mode><VDC:mode>): Delete.
3685 (aarch64_get_qreg<VSTRUCT:mode><VQ:mode>): Likewise.
3686 (aarch64_st2<mode>_dreg): Use vector-tuple mode iterator and
3688 (aarch64_st2<vstruct_elt>_dreg): This.
3689 (aarch64_st3<mode>_dreg): Use vector-tuple mode iterator and
3691 (aarch64_st3<vstruct_elt>_dreg): This.
3692 (aarch64_st4<mode>_dreg): Use vector-tuple mode iterator and
3694 (aarch64_st4<vstruct_elt>_dreg): This.
3695 (aarch64_st<VSTRUCT:nregs><VDC:mode>): Use vector-tuple mode
3696 iterator and rename to...
3697 (aarch64_st<nregs><vstruct_elt>): This.
3698 (aarch64_st<VSTRUCT:nregs><VQ:mode>): Use vector-tuple mode
3699 iterator and rename to aarch64_st<nregs><vstruct_elt>.
3700 (aarch64_st<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use vector-
3701 tuple mode iterator and rename to...
3702 (aarch64_st<nregs>_lane<vstruct_elt>): This.
3703 (aarch64_set_qreg<VSTRUCT:mode><VQ:mode>): Delete.
3704 (aarch64_simd_ld1<mode>_x2): Use vector-tuple mode iterator
3706 (aarch64_simd_ld1<vstruct_elt>_x2): This.
3707 * config/aarch64/aarch64.c (aarch64_advsimd_struct_mode_p):
3708 Refactor to include new vector-tuple modes.
3709 (aarch64_classify_vector_mode): Add cases for new vector-
3711 (aarch64_advsimd_partial_struct_mode_p): Define.
3712 (aarch64_advsimd_full_struct_mode_p): Likewise.
3713 (aarch64_advsimd_vector_array_mode): Likewise.
3714 (aarch64_sve_data_mode): Change location in file.
3715 (aarch64_array_mode): Handle case of Neon vector-tuple modes.
3716 (aarch64_hard_regno_nregs): Handle case of partial Neon
3718 (aarch64_classify_address): Refactor to include handling of
3719 Neon vector-tuple modes.
3720 (aarch64_print_operand): Print "d" for "%R" for a partial
3721 Neon vector structure.
3722 (aarch64_expand_vec_perm_1): Use new vector-tuple mode.
3723 (aarch64_modes_tieable_p): Prevent tieing Neon partial struct
3724 modes with scalar machines modes larger than 8 bytes.
3725 (aarch64_can_change_mode_class): Don't allow changes between
3726 partial and full Neon vector-structure modes.
3727 * config/aarch64/arm_neon.h (vst2_lane_f16): Use updated
3728 builtin and remove boiler-plate code for opaque mode.
3729 (vst2_lane_f32): Likewise.
3730 (vst2_lane_f64): Likewise.
3731 (vst2_lane_p8): Likewise.
3732 (vst2_lane_p16): Likewise.
3733 (vst2_lane_p64): Likewise.
3734 (vst2_lane_s8): Likewise.
3735 (vst2_lane_s16): Likewise.
3736 (vst2_lane_s32): Likewise.
3737 (vst2_lane_s64): Likewise.
3738 (vst2_lane_u8): Likewise.
3739 (vst2_lane_u16): Likewise.
3740 (vst2_lane_u32): Likewise.
3741 (vst2_lane_u64): Likewise.
3742 (vst2q_lane_f16): Likewise.
3743 (vst2q_lane_f32): Likewise.
3744 (vst2q_lane_f64): Likewise.
3745 (vst2q_lane_p8): Likewise.
3746 (vst2q_lane_p16): Likewise.
3747 (vst2q_lane_p64): Likewise.
3748 (vst2q_lane_s8): Likewise.
3749 (vst2q_lane_s16): Likewise.
3750 (vst2q_lane_s32): Likewise.
3751 (vst2q_lane_s64): Likewise.
3752 (vst2q_lane_u8): Likewise.
3753 (vst2q_lane_u16): Likewise.
3754 (vst2q_lane_u32): Likewise.
3755 (vst2q_lane_u64): Likewise.
3756 (vst3_lane_f16): Likewise.
3757 (vst3_lane_f32): Likewise.
3758 (vst3_lane_f64): Likewise.
3759 (vst3_lane_p8): Likewise.
3760 (vst3_lane_p16): Likewise.
3761 (vst3_lane_p64): Likewise.
3762 (vst3_lane_s8): Likewise.
3763 (vst3_lane_s16): Likewise.
3764 (vst3_lane_s32): Likewise.
3765 (vst3_lane_s64): Likewise.
3766 (vst3_lane_u8): Likewise.
3767 (vst3_lane_u16): Likewise.
3768 (vst3_lane_u32): Likewise.
3769 (vst3_lane_u64): Likewise.
3770 (vst3q_lane_f16): Likewise.
3771 (vst3q_lane_f32): Likewise.
3772 (vst3q_lane_f64): Likewise.
3773 (vst3q_lane_p8): Likewise.
3774 (vst3q_lane_p16): Likewise.
3775 (vst3q_lane_p64): Likewise.
3776 (vst3q_lane_s8): Likewise.
3777 (vst3q_lane_s16): Likewise.
3778 (vst3q_lane_s32): Likewise.
3779 (vst3q_lane_s64): Likewise.
3780 (vst3q_lane_u8): Likewise.
3781 (vst3q_lane_u16): Likewise.
3782 (vst3q_lane_u32): Likewise.
3783 (vst3q_lane_u64): Likewise.
3784 (vst4_lane_f16): Likewise.
3785 (vst4_lane_f32): Likewise.
3786 (vst4_lane_f64): Likewise.
3787 (vst4_lane_p8): Likewise.
3788 (vst4_lane_p16): Likewise.
3789 (vst4_lane_p64): Likewise.
3790 (vst4_lane_s8): Likewise.
3791 (vst4_lane_s16): Likewise.
3792 (vst4_lane_s32): Likewise.
3793 (vst4_lane_s64): Likewise.
3794 (vst4_lane_u8): Likewise.
3795 (vst4_lane_u16): Likewise.
3796 (vst4_lane_u32): Likewise.
3797 (vst4_lane_u64): Likewise.
3798 (vst4q_lane_f16): Likewise.
3799 (vst4q_lane_f32): Likewise.
3800 (vst4q_lane_f64): Likewise.
3801 (vst4q_lane_p8): Likewise.
3802 (vst4q_lane_p16): Likewise.
3803 (vst4q_lane_p64): Likewise.
3804 (vst4q_lane_s8): Likewise.
3805 (vst4q_lane_s16): Likewise.
3806 (vst4q_lane_s32): Likewise.
3807 (vst4q_lane_s64): Likewise.
3808 (vst4q_lane_u8): Likewise.
3809 (vst4q_lane_u16): Likewise.
3810 (vst4q_lane_u32): Likewise.
3811 (vst4q_lane_u64): Likewise.
3812 (vtbl3_s8): Likewise.
3813 (vtbl3_u8): Likewise.
3814 (vtbl3_p8): Likewise.
3815 (vtbl4_s8): Likewise.
3816 (vtbl4_u8): Likewise.
3817 (vtbl4_p8): Likewise.
3818 (vld1_u8_x3): Likewise.
3819 (vld1_s8_x3): Likewise.
3820 (vld1_u16_x3): Likewise.
3821 (vld1_s16_x3): Likewise.
3822 (vld1_u32_x3): Likewise.
3823 (vld1_s32_x3): Likewise.
3824 (vld1_u64_x3): Likewise.
3825 (vld1_s64_x3): Likewise.
3826 (vld1_f16_x3): Likewise.
3827 (vld1_f32_x3): Likewise.
3828 (vld1_f64_x3): Likewise.
3829 (vld1_p8_x3): Likewise.
3830 (vld1_p16_x3): Likewise.
3831 (vld1_p64_x3): Likewise.
3832 (vld1q_u8_x3): Likewise.
3833 (vld1q_s8_x3): Likewise.
3834 (vld1q_u16_x3): Likewise.
3835 (vld1q_s16_x3): Likewise.
3836 (vld1q_u32_x3): Likewise.
3837 (vld1q_s32_x3): Likewise.
3838 (vld1q_u64_x3): Likewise.
3839 (vld1q_s64_x3): Likewise.
3840 (vld1q_f16_x3): Likewise.
3841 (vld1q_f32_x3): Likewise.
3842 (vld1q_f64_x3): Likewise.
3843 (vld1q_p8_x3): Likewise.
3844 (vld1q_p16_x3): Likewise.
3845 (vld1q_p64_x3): Likewise.
3846 (vld1_u8_x2): Likewise.
3847 (vld1_s8_x2): Likewise.
3848 (vld1_u16_x2): Likewise.
3849 (vld1_s16_x2): Likewise.
3850 (vld1_u32_x2): Likewise.
3851 (vld1_s32_x2): Likewise.
3852 (vld1_u64_x2): Likewise.
3853 (vld1_s64_x2): Likewise.
3854 (vld1_f16_x2): Likewise.
3855 (vld1_f32_x2): Likewise.
3856 (vld1_f64_x2): Likewise.
3857 (vld1_p8_x2): Likewise.
3858 (vld1_p16_x2): Likewise.
3859 (vld1_p64_x2): Likewise.
3860 (vld1q_u8_x2): Likewise.
3861 (vld1q_s8_x2): Likewise.
3862 (vld1q_u16_x2): Likewise.
3863 (vld1q_s16_x2): Likewise.
3864 (vld1q_u32_x2): Likewise.
3865 (vld1q_s32_x2): Likewise.
3866 (vld1q_u64_x2): Likewise.
3867 (vld1q_s64_x2): Likewise.
3868 (vld1q_f16_x2): Likewise.
3869 (vld1q_f32_x2): Likewise.
3870 (vld1q_f64_x2): Likewise.
3871 (vld1q_p8_x2): Likewise.
3872 (vld1q_p16_x2): Likewise.
3873 (vld1q_p64_x2): Likewise.
3874 (vld1_s8_x4): Likewise.
3875 (vld1q_s8_x4): Likewise.
3876 (vld1_s16_x4): Likewise.
3877 (vld1q_s16_x4): Likewise.
3878 (vld1_s32_x4): Likewise.
3879 (vld1q_s32_x4): Likewise.
3880 (vld1_u8_x4): Likewise.
3881 (vld1q_u8_x4): Likewise.
3882 (vld1_u16_x4): Likewise.
3883 (vld1q_u16_x4): Likewise.
3884 (vld1_u32_x4): Likewise.
3885 (vld1q_u32_x4): Likewise.
3886 (vld1_f16_x4): Likewise.
3887 (vld1q_f16_x4): Likewise.
3888 (vld1_f32_x4): Likewise.
3889 (vld1q_f32_x4): Likewise.
3890 (vld1_p8_x4): Likewise.
3891 (vld1q_p8_x4): Likewise.
3892 (vld1_p16_x4): Likewise.
3893 (vld1q_p16_x4): Likewise.
3894 (vld1_s64_x4): Likewise.
3895 (vld1_u64_x4): Likewise.
3896 (vld1_p64_x4): Likewise.
3897 (vld1q_s64_x4): Likewise.
3898 (vld1q_u64_x4): Likewise.
3899 (vld1q_p64_x4): Likewise.
3900 (vld1_f64_x4): Likewise.
3901 (vld1q_f64_x4): Likewise.
3902 (vld2_s64): Likewise.
3903 (vld2_u64): Likewise.
3904 (vld2_f64): Likewise.
3905 (vld2_s8): Likewise.
3906 (vld2_p8): Likewise.
3907 (vld2_p64): Likewise.
3908 (vld2_s16): Likewise.
3909 (vld2_p16): Likewise.
3910 (vld2_s32): Likewise.
3911 (vld2_u8): Likewise.
3912 (vld2_u16): Likewise.
3913 (vld2_u32): Likewise.
3914 (vld2_f16): Likewise.
3915 (vld2_f32): Likewise.
3916 (vld2q_s8): Likewise.
3917 (vld2q_p8): Likewise.
3918 (vld2q_s16): Likewise.
3919 (vld2q_p16): Likewise.
3920 (vld2q_p64): Likewise.
3921 (vld2q_s32): Likewise.
3922 (vld2q_s64): Likewise.
3923 (vld2q_u8): Likewise.
3924 (vld2q_u16): Likewise.
3925 (vld2q_u32): Likewise.
3926 (vld2q_u64): Likewise.
3927 (vld2q_f16): Likewise.
3928 (vld2q_f32): Likewise.
3929 (vld2q_f64): Likewise.
3930 (vld3_s64): Likewise.
3931 (vld3_u64): Likewise.
3932 (vld3_f64): Likewise.
3933 (vld3_s8): Likewise.
3934 (vld3_p8): Likewise.
3935 (vld3_s16): Likewise.
3936 (vld3_p16): Likewise.
3937 (vld3_s32): Likewise.
3938 (vld3_u8): Likewise.
3939 (vld3_u16): Likewise.
3940 (vld3_u32): Likewise.
3941 (vld3_f16): Likewise.
3942 (vld3_f32): Likewise.
3943 (vld3_p64): Likewise.
3944 (vld3q_s8): Likewise.
3945 (vld3q_p8): Likewise.
3946 (vld3q_s16): Likewise.
3947 (vld3q_p16): Likewise.
3948 (vld3q_s32): Likewise.
3949 (vld3q_s64): Likewise.
3950 (vld3q_u8): Likewise.
3951 (vld3q_u16): Likewise.
3952 (vld3q_u32): Likewise.
3953 (vld3q_u64): Likewise.
3954 (vld3q_f16): Likewise.
3955 (vld3q_f32): Likewise.
3956 (vld3q_f64): Likewise.
3957 (vld3q_p64): Likewise.
3958 (vld4_s64): Likewise.
3959 (vld4_u64): Likewise.
3960 (vld4_f64): Likewise.
3961 (vld4_s8): Likewise.
3962 (vld4_p8): Likewise.
3963 (vld4_s16): Likewise.
3964 (vld4_p16): Likewise.
3965 (vld4_s32): Likewise.
3966 (vld4_u8): Likewise.
3967 (vld4_u16): Likewise.
3968 (vld4_u32): Likewise.
3969 (vld4_f16): Likewise.
3970 (vld4_f32): Likewise.
3971 (vld4_p64): Likewise.
3972 (vld4q_s8): Likewise.
3973 (vld4q_p8): Likewise.
3974 (vld4q_s16): Likewise.
3975 (vld4q_p16): Likewise.
3976 (vld4q_s32): Likewise.
3977 (vld4q_s64): Likewise.
3978 (vld4q_u8): Likewise.
3979 (vld4q_u16): Likewise.
3980 (vld4q_u32): Likewise.
3981 (vld4q_u64): Likewise.
3982 (vld4q_f16): Likewise.
3983 (vld4q_f32): Likewise.
3984 (vld4q_f64): Likewise.
3985 (vld4q_p64): Likewise.
3986 (vld2_dup_s8): Likewise.
3987 (vld2_dup_s16): Likewise.
3988 (vld2_dup_s32): Likewise.
3989 (vld2_dup_f16): Likewise.
3990 (vld2_dup_f32): Likewise.
3991 (vld2_dup_f64): Likewise.
3992 (vld2_dup_u8): Likewise.
3993 (vld2_dup_u16): Likewise.
3994 (vld2_dup_u32): Likewise.
3995 (vld2_dup_p8): Likewise.
3996 (vld2_dup_p16): Likewise.
3997 (vld2_dup_p64): Likewise.
3998 (vld2_dup_s64): Likewise.
3999 (vld2_dup_u64): Likewise.
4000 (vld2q_dup_s8): Likewise.
4001 (vld2q_dup_p8): Likewise.
4002 (vld2q_dup_s16): Likewise.
4003 (vld2q_dup_p16): Likewise.
4004 (vld2q_dup_s32): Likewise.
4005 (vld2q_dup_s64): Likewise.
4006 (vld2q_dup_u8): Likewise.
4007 (vld2q_dup_u16): Likewise.
4008 (vld2q_dup_u32): Likewise.
4009 (vld2q_dup_u64): Likewise.
4010 (vld2q_dup_f16): Likewise.
4011 (vld2q_dup_f32): Likewise.
4012 (vld2q_dup_f64): Likewise.
4013 (vld2q_dup_p64): Likewise.
4014 (vld3_dup_s64): Likewise.
4015 (vld3_dup_u64): Likewise.
4016 (vld3_dup_f64): Likewise.
4017 (vld3_dup_s8): Likewise.
4018 (vld3_dup_p8): Likewise.
4019 (vld3_dup_s16): Likewise.
4020 (vld3_dup_p16): Likewise.
4021 (vld3_dup_s32): Likewise.
4022 (vld3_dup_u8): Likewise.
4023 (vld3_dup_u16): Likewise.
4024 (vld3_dup_u32): Likewise.
4025 (vld3_dup_f16): Likewise.
4026 (vld3_dup_f32): Likewise.
4027 (vld3_dup_p64): Likewise.
4028 (vld3q_dup_s8): Likewise.
4029 (vld3q_dup_p8): Likewise.
4030 (vld3q_dup_s16): Likewise.
4031 (vld3q_dup_p16): Likewise.
4032 (vld3q_dup_s32): Likewise.
4033 (vld3q_dup_s64): Likewise.
4034 (vld3q_dup_u8): Likewise.
4035 (vld3q_dup_u16): Likewise.
4036 (vld3q_dup_u32): Likewise.
4037 (vld3q_dup_u64): Likewise.
4038 (vld3q_dup_f16): Likewise.
4039 (vld3q_dup_f32): Likewise.
4040 (vld3q_dup_f64): Likewise.
4041 (vld3q_dup_p64): Likewise.
4042 (vld4_dup_s64): Likewise.
4043 (vld4_dup_u64): Likewise.
4044 (vld4_dup_f64): Likewise.
4045 (vld4_dup_s8): Likewise.
4046 (vld4_dup_p8): Likewise.
4047 (vld4_dup_s16): Likewise.
4048 (vld4_dup_p16): Likewise.
4049 (vld4_dup_s32): Likewise.
4050 (vld4_dup_u8): Likewise.
4051 (vld4_dup_u16): Likewise.
4052 (vld4_dup_u32): Likewise.
4053 (vld4_dup_f16): Likewise.
4054 (vld4_dup_f32): Likewise.
4055 (vld4_dup_p64): Likewise.
4056 (vld4q_dup_s8): Likewise.
4057 (vld4q_dup_p8): Likewise.
4058 (vld4q_dup_s16): Likewise.
4059 (vld4q_dup_p16): Likewise.
4060 (vld4q_dup_s32): Likewise.
4061 (vld4q_dup_s64): Likewise.
4062 (vld4q_dup_u8): Likewise.
4063 (vld4q_dup_u16): Likewise.
4064 (vld4q_dup_u32): Likewise.
4065 (vld4q_dup_u64): Likewise.
4066 (vld4q_dup_f16): Likewise.
4067 (vld4q_dup_f32): Likewise.
4068 (vld4q_dup_f64): Likewise.
4069 (vld4q_dup_p64): Likewise.
4070 (vld2_lane_u8): Likewise.
4071 (vld2_lane_u16): Likewise.
4072 (vld2_lane_u32): Likewise.
4073 (vld2_lane_u64): Likewise.
4074 (vld2_lane_s8): Likewise.
4075 (vld2_lane_s16): Likewise.
4076 (vld2_lane_s32): Likewise.
4077 (vld2_lane_s64): Likewise.
4078 (vld2_lane_f16): Likewise.
4079 (vld2_lane_f32): Likewise.
4080 (vld2_lane_f64): Likewise.
4081 (vld2_lane_p8): Likewise.
4082 (vld2_lane_p16): Likewise.
4083 (vld2_lane_p64): Likewise.
4084 (vld2q_lane_u8): Likewise.
4085 (vld2q_lane_u16): Likewise.
4086 (vld2q_lane_u32): Likewise.
4087 (vld2q_lane_u64): Likewise.
4088 (vld2q_lane_s8): Likewise.
4089 (vld2q_lane_s16): Likewise.
4090 (vld2q_lane_s32): Likewise.
4091 (vld2q_lane_s64): Likewise.
4092 (vld2q_lane_f16): Likewise.
4093 (vld2q_lane_f32): Likewise.
4094 (vld2q_lane_f64): Likewise.
4095 (vld2q_lane_p8): Likewise.
4096 (vld2q_lane_p16): Likewise.
4097 (vld2q_lane_p64): Likewise.
4098 (vld3_lane_u8): Likewise.
4099 (vld3_lane_u16): Likewise.
4100 (vld3_lane_u32): Likewise.
4101 (vld3_lane_u64): Likewise.
4102 (vld3_lane_s8): Likewise.
4103 (vld3_lane_s16): Likewise.
4104 (vld3_lane_s32): Likewise.
4105 (vld3_lane_s64): Likewise.
4106 (vld3_lane_f16): Likewise.
4107 (vld3_lane_f32): Likewise.
4108 (vld3_lane_f64): Likewise.
4109 (vld3_lane_p8): Likewise.
4110 (vld3_lane_p16): Likewise.
4111 (vld3_lane_p64): Likewise.
4112 (vld3q_lane_u8): Likewise.
4113 (vld3q_lane_u16): Likewise.
4114 (vld3q_lane_u32): Likewise.
4115 (vld3q_lane_u64): Likewise.
4116 (vld3q_lane_s8): Likewise.
4117 (vld3q_lane_s16): Likewise.
4118 (vld3q_lane_s32): Likewise.
4119 (vld3q_lane_s64): Likewise.
4120 (vld3q_lane_f16): Likewise.
4121 (vld3q_lane_f32): Likewise.
4122 (vld3q_lane_f64): Likewise.
4123 (vld3q_lane_p8): Likewise.
4124 (vld3q_lane_p16): Likewise.
4125 (vld3q_lane_p64): Likewise.
4126 (vld4_lane_u8): Likewise.
4127 (vld4_lane_u16): Likewise.
4128 (vld4_lane_u32): Likewise.
4129 (vld4_lane_u64): Likewise.
4130 (vld4_lane_s8): Likewise.
4131 (vld4_lane_s16): Likewise.
4132 (vld4_lane_s32): Likewise.
4133 (vld4_lane_s64): Likewise.
4134 (vld4_lane_f16): Likewise.
4135 (vld4_lane_f32): Likewise.
4136 (vld4_lane_f64): Likewise.
4137 (vld4_lane_p8): Likewise.
4138 (vld4_lane_p16): Likewise.
4139 (vld4_lane_p64): Likewise.
4140 (vld4q_lane_u8): Likewise.
4141 (vld4q_lane_u16): Likewise.
4142 (vld4q_lane_u32): Likewise.
4143 (vld4q_lane_u64): Likewise.
4144 (vld4q_lane_s8): Likewise.
4145 (vld4q_lane_s16): Likewise.
4146 (vld4q_lane_s32): Likewise.
4147 (vld4q_lane_s64): Likewise.
4148 (vld4q_lane_f16): Likewise.
4149 (vld4q_lane_f32): Likewise.
4150 (vld4q_lane_f64): Likewise.
4151 (vld4q_lane_p8): Likewise.
4152 (vld4q_lane_p16): Likewise.
4153 (vld4q_lane_p64): Likewise.
4154 (vqtbl2_s8): Likewise.
4155 (vqtbl2_u8): Likewise.
4156 (vqtbl2_p8): Likewise.
4157 (vqtbl2q_s8): Likewise.
4158 (vqtbl2q_u8): Likewise.
4159 (vqtbl2q_p8): Likewise.
4160 (vqtbl3_s8): Likewise.
4161 (vqtbl3_u8): Likewise.
4162 (vqtbl3_p8): Likewise.
4163 (vqtbl3q_s8): Likewise.
4164 (vqtbl3q_u8): Likewise.
4165 (vqtbl3q_p8): Likewise.
4166 (vqtbl4_s8): Likewise.
4167 (vqtbl4_u8): Likewise.
4168 (vqtbl4_p8): Likewise.
4169 (vqtbl4q_s8): Likewise.
4170 (vqtbl4q_u8): Likewise.
4171 (vqtbl4q_p8): Likewise.
4172 (vqtbx2_s8): Likewise.
4173 (vqtbx2_u8): Likewise.
4174 (vqtbx2_p8): Likewise.
4175 (vqtbx2q_s8): Likewise.
4176 (vqtbx2q_u8): Likewise.
4177 (vqtbx2q_p8): Likewise.
4178 (vqtbx3_s8): Likewise.
4179 (vqtbx3_u8): Likewise.
4180 (vqtbx3_p8): Likewise.
4181 (vqtbx3q_s8): Likewise.
4182 (vqtbx3q_u8): Likewise.
4183 (vqtbx3q_p8): Likewise.
4184 (vqtbx4_s8): Likewise.
4185 (vqtbx4_u8): Likewise.
4186 (vqtbx4_p8): Likewise.
4187 (vqtbx4q_s8): Likewise.
4188 (vqtbx4q_u8): Likewise.
4189 (vqtbx4q_p8): Likewise.
4190 (vst1_s64_x2): Likewise.
4191 (vst1_u64_x2): Likewise.
4192 (vst1_f64_x2): Likewise.
4193 (vst1_s8_x2): Likewise.
4194 (vst1_p8_x2): Likewise.
4195 (vst1_s16_x2): Likewise.
4196 (vst1_p16_x2): Likewise.
4197 (vst1_s32_x2): Likewise.
4198 (vst1_u8_x2): Likewise.
4199 (vst1_u16_x2): Likewise.
4200 (vst1_u32_x2): Likewise.
4201 (vst1_f16_x2): Likewise.
4202 (vst1_f32_x2): Likewise.
4203 (vst1_p64_x2): Likewise.
4204 (vst1q_s8_x2): Likewise.
4205 (vst1q_p8_x2): Likewise.
4206 (vst1q_s16_x2): Likewise.
4207 (vst1q_p16_x2): Likewise.
4208 (vst1q_s32_x2): Likewise.
4209 (vst1q_s64_x2): Likewise.
4210 (vst1q_u8_x2): Likewise.
4211 (vst1q_u16_x2): Likewise.
4212 (vst1q_u32_x2): Likewise.
4213 (vst1q_u64_x2): Likewise.
4214 (vst1q_f16_x2): Likewise.
4215 (vst1q_f32_x2): Likewise.
4216 (vst1q_f64_x2): Likewise.
4217 (vst1q_p64_x2): Likewise.
4218 (vst1_s64_x3): Likewise.
4219 (vst1_u64_x3): Likewise.
4220 (vst1_f64_x3): Likewise.
4221 (vst1_s8_x3): Likewise.
4222 (vst1_p8_x3): Likewise.
4223 (vst1_s16_x3): Likewise.
4224 (vst1_p16_x3): Likewise.
4225 (vst1_s32_x3): Likewise.
4226 (vst1_u8_x3): Likewise.
4227 (vst1_u16_x3): Likewise.
4228 (vst1_u32_x3): Likewise.
4229 (vst1_f16_x3): Likewise.
4230 (vst1_f32_x3): Likewise.
4231 (vst1_p64_x3): Likewise.
4232 (vst1q_s8_x3): Likewise.
4233 (vst1q_p8_x3): Likewise.
4234 (vst1q_s16_x3): Likewise.
4235 (vst1q_p16_x3): Likewise.
4236 (vst1q_s32_x3): Likewise.
4237 (vst1q_s64_x3): Likewise.
4238 (vst1q_u8_x3): Likewise.
4239 (vst1q_u16_x3): Likewise.
4240 (vst1q_u32_x3): Likewise.
4241 (vst1q_u64_x3): Likewise.
4242 (vst1q_f16_x3): Likewise.
4243 (vst1q_f32_x3): Likewise.
4244 (vst1q_f64_x3): Likewise.
4245 (vst1q_p64_x3): Likewise.
4246 (vst1_s8_x4): Likewise.
4247 (vst1q_s8_x4): Likewise.
4248 (vst1_s16_x4): Likewise.
4249 (vst1q_s16_x4): Likewise.
4250 (vst1_s32_x4): Likewise.
4251 (vst1q_s32_x4): Likewise.
4252 (vst1_u8_x4): Likewise.
4253 (vst1q_u8_x4): Likewise.
4254 (vst1_u16_x4): Likewise.
4255 (vst1q_u16_x4): Likewise.
4256 (vst1_u32_x4): Likewise.
4257 (vst1q_u32_x4): Likewise.
4258 (vst1_f16_x4): Likewise.
4259 (vst1q_f16_x4): Likewise.
4260 (vst1_f32_x4): Likewise.
4261 (vst1q_f32_x4): Likewise.
4262 (vst1_p8_x4): Likewise.
4263 (vst1q_p8_x4): Likewise.
4264 (vst1_p16_x4): Likewise.
4265 (vst1q_p16_x4): Likewise.
4266 (vst1_s64_x4): Likewise.
4267 (vst1_u64_x4): Likewise.
4268 (vst1_p64_x4): Likewise.
4269 (vst1q_s64_x4): Likewise.
4270 (vst1q_u64_x4): Likewise.
4271 (vst1q_p64_x4): Likewise.
4272 (vst1_f64_x4): Likewise.
4273 (vst1q_f64_x4): Likewise.
4274 (vst2_s64): Likewise.
4275 (vst2_u64): Likewise.
4276 (vst2_f64): Likewise.
4277 (vst2_s8): Likewise.
4278 (vst2_p8): Likewise.
4279 (vst2_s16): Likewise.
4280 (vst2_p16): Likewise.
4281 (vst2_s32): Likewise.
4282 (vst2_u8): Likewise.
4283 (vst2_u16): Likewise.
4284 (vst2_u32): Likewise.
4285 (vst2_f16): Likewise.
4286 (vst2_f32): Likewise.
4287 (vst2_p64): Likewise.
4288 (vst2q_s8): Likewise.
4289 (vst2q_p8): Likewise.
4290 (vst2q_s16): Likewise.
4291 (vst2q_p16): Likewise.
4292 (vst2q_s32): Likewise.
4293 (vst2q_s64): Likewise.
4294 (vst2q_u8): Likewise.
4295 (vst2q_u16): Likewise.
4296 (vst2q_u32): Likewise.
4297 (vst2q_u64): Likewise.
4298 (vst2q_f16): Likewise.
4299 (vst2q_f32): Likewise.
4300 (vst2q_f64): Likewise.
4301 (vst2q_p64): Likewise.
4302 (vst3_s64): Likewise.
4303 (vst3_u64): Likewise.
4304 (vst3_f64): Likewise.
4305 (vst3_s8): Likewise.
4306 (vst3_p8): Likewise.
4307 (vst3_s16): Likewise.
4308 (vst3_p16): Likewise.
4309 (vst3_s32): Likewise.
4310 (vst3_u8): Likewise.
4311 (vst3_u16): Likewise.
4312 (vst3_u32): Likewise.
4313 (vst3_f16): Likewise.
4314 (vst3_f32): Likewise.
4315 (vst3_p64): Likewise.
4316 (vst3q_s8): Likewise.
4317 (vst3q_p8): Likewise.
4318 (vst3q_s16): Likewise.
4319 (vst3q_p16): Likewise.
4320 (vst3q_s32): Likewise.
4321 (vst3q_s64): Likewise.
4322 (vst3q_u8): Likewise.
4323 (vst3q_u16): Likewise.
4324 (vst3q_u32): Likewise.
4325 (vst3q_u64): Likewise.
4326 (vst3q_f16): Likewise.
4327 (vst3q_f32): Likewise.
4328 (vst3q_f64): Likewise.
4329 (vst3q_p64): Likewise.
4330 (vst4_s64): Likewise.
4331 (vst4_u64): Likewise.
4332 (vst4_f64): Likewise.
4333 (vst4_s8): Likewise.
4334 (vst4_p8): Likewise.
4335 (vst4_s16): Likewise.
4336 (vst4_p16): Likewise.
4337 (vst4_s32): Likewise.
4338 (vst4_u8): Likewise.
4339 (vst4_u16): Likewise.
4340 (vst4_u32): Likewise.
4341 (vst4_f16): Likewise.
4342 (vst4_f32): Likewise.
4343 (vst4_p64): Likewise.
4344 (vst4q_s8): Likewise.
4345 (vst4q_p8): Likewise.
4346 (vst4q_s16): Likewise.
4347 (vst4q_p16): Likewise.
4348 (vst4q_s32): Likewise.
4349 (vst4q_s64): Likewise.
4350 (vst4q_u8): Likewise.
4351 (vst4q_u16): Likewise.
4352 (vst4q_u32): Likewise.
4353 (vst4q_u64): Likewise.
4354 (vst4q_f16): Likewise.
4355 (vst4q_f32): Likewise.
4356 (vst4q_f64): Likewise.
4357 (vst4q_p64): Likewise.
4358 (vtbx4_s8): Likewise.
4359 (vtbx4_u8): Likewise.
4360 (vtbx4_p8): Likewise.
4361 (vld1_bf16_x2): Likewise.
4362 (vld1q_bf16_x2): Likewise.
4363 (vld1_bf16_x3): Likewise.
4364 (vld1q_bf16_x3): Likewise.
4365 (vld1_bf16_x4): Likewise.
4366 (vld1q_bf16_x4): Likewise.
4367 (vld2_bf16): Likewise.
4368 (vld2q_bf16): Likewise.
4369 (vld2_dup_bf16): Likewise.
4370 (vld2q_dup_bf16): Likewise.
4371 (vld3_bf16): Likewise.
4372 (vld3q_bf16): Likewise.
4373 (vld3_dup_bf16): Likewise.
4374 (vld3q_dup_bf16): Likewise.
4375 (vld4_bf16): Likewise.
4376 (vld4q_bf16): Likewise.
4377 (vld4_dup_bf16): Likewise.
4378 (vld4q_dup_bf16): Likewise.
4379 (vst1_bf16_x2): Likewise.
4380 (vst1q_bf16_x2): Likewise.
4381 (vst1_bf16_x3): Likewise.
4382 (vst1q_bf16_x3): Likewise.
4383 (vst1_bf16_x4): Likewise.
4384 (vst1q_bf16_x4): Likewise.
4385 (vst2_bf16): Likewise.
4386 (vst2q_bf16): Likewise.
4387 (vst3_bf16): Likewise.
4388 (vst3q_bf16): Likewise.
4389 (vst4_bf16): Likewise.
4390 (vst4q_bf16): Likewise.
4391 (vld2_lane_bf16): Likewise.
4392 (vld2q_lane_bf16): Likewise.
4393 (vld3_lane_bf16): Likewise.
4394 (vld3q_lane_bf16): Likewise.
4395 (vld4_lane_bf16): Likewise.
4396 (vld4q_lane_bf16): Likewise.
4397 (vst2_lane_bf16): Likewise.
4398 (vst2q_lane_bf16): Likewise.
4399 (vst3_lane_bf16): Likewise.
4400 (vst3q_lane_bf16): Likewise.
4401 (vst4_lane_bf16): Likewise.
4402 (vst4q_lane_bf16): Likewise.
4403 * config/aarch64/geniterators.sh: Modify iterator regex to
4404 match new vector-tuple modes.
4405 * config/aarch64/iterators.md (insn_count): Extend mode
4406 attribute with vector-tuple type information.
4408 (Vendreg): Likewise.
4411 (VSTRUCT_2D): New mode iterator.
4412 (VSTRUCT_2DNX): Likewise.
4413 (VSTRUCT_2DX): Likewise.
4414 (VSTRUCT_2Q): Likewise.
4415 (VSTRUCT_2QD): Likewise.
4416 (VSTRUCT_3D): Likewise.
4417 (VSTRUCT_3DNX): Likewise.
4418 (VSTRUCT_3DX): Likewise.
4419 (VSTRUCT_3Q): Likewise.
4420 (VSTRUCT_3QD): Likewise.
4421 (VSTRUCT_4D): Likewise.
4422 (VSTRUCT_4DNX): Likewise.
4423 (VSTRUCT_4DX): Likewise.
4424 (VSTRUCT_4Q): Likewise.
4425 (VSTRUCT_4QD): Likewise.
4426 (VSTRUCT_D): Likewise.
4427 (VSTRUCT_Q): Likewise.
4428 (VSTRUCT_QD): Likewise.
4429 (VSTRUCT_ELT): New mode attribute.
4430 (vstruct_elt): Likewise.
4431 * genmodes.c (VECTOR_MODE): Add default prefix and order
4433 (VECTOR_MODE_WITH_PREFIX): Define.
4434 (make_vector_mode): Add mode prefix and order parameters.
4436 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
4438 * expmed.c (extract_bit_field_1): Ensure modes are tieable.
4440 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
4442 * expr.c (emit_group_load_1): Remove historic workaround.
4444 2021-11-04 Jonathan Wright <jonathan.wright@arm.com>
4446 * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
4447 Factor out main loop to...
4448 (aarch64_init_simd_builtin_functions): This new function.
4449 (register_tuple_type): Define.
4450 (aarch64_scalar_builtin_type_p): Define.
4451 (handle_arm_neon_h): Define.
4452 * config/aarch64/aarch64-c.c (aarch64_pragma_aarch64): Handle
4453 pragma for arm_neon.h.
4454 * config/aarch64/aarch64-protos.h (aarch64_advsimd_struct_mode_p):
4456 (handle_arm_neon_h): Likewise.
4457 * config/aarch64/aarch64.c (aarch64_advsimd_struct_mode_p):
4458 Remove static modifier.
4459 * config/aarch64/arm_neon.h (target): Remove Neon vector
4460 structure type definitions.
4462 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
4464 PR tree-optimization/102943
4465 * gimple-range-path.cc (path_range_query::range_on_path_entry):
4466 Prefer range_of_expr unless there are no statements in the BB.
4468 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
4470 PR tree-optimization/102943
4471 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
4472 Avoid duplicate calculation of paths.
4474 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
4476 PR tree-optimization/102943
4477 * gimple-range-path.cc (path_range_query::compute_phi_relations):
4478 Only compute relations for SSA names in the import list.
4479 (path_range_query::compute_outgoing_relations): Same.
4480 * gimple-range-path.h (path_range_query::import_p): New.
4482 2021-11-04 Richard Biener <rguenther@suse.de>
4484 PR rtl-optimization/103075
4485 * simplify-rtx.c (exact_int_to_float_conversion_p): Return
4486 false for a VOIDmode operand.
4488 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
4490 * config/aarch64/aarch64.c (aarch64_vector_costs): Make member
4491 variables private and add "m_" to their names. Remove is_loop.
4492 (aarch64_record_potential_advsimd_unrolling): Replace with...
4493 (aarch64_vector_costs::record_potential_advsimd_unrolling): ...this.
4494 (aarch64_analyze_loop_vinfo): Replace with...
4495 (aarch64_vector_costs::analyze_loop_vinfo): ...this.
4496 Move initialization of (m_)vec_flags to add_stmt_cost.
4497 (aarch64_analyze_bb_vinfo): Delete.
4498 (aarch64_count_ops): Replace with...
4499 (aarch64_vector_costs::count_ops): ...this.
4500 (aarch64_vector_costs::add_stmt_cost): Set m_vec_flags,
4501 using m_costing_for_scalar to test whether we're costing
4502 scalar or vector code.
4503 (aarch64_adjust_body_cost_sve): Replace with...
4504 (aarch64_vector_costs::adjust_body_cost_sve): ...this.
4505 (aarch64_adjust_body_cost): Replace with...
4506 (aarch64_vector_costs::adjust_body_cost): ...this.
4507 (aarch64_vector_costs::finish_cost): Use m_vinfo instead of is_loop.
4509 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
4511 * target.def (targetm.vectorize.init_cost): Replace with...
4512 (targetm.vectorize.create_costs): ...this.
4513 (targetm.vectorize.add_stmt_cost): Delete.
4514 (targetm.vectorize.finish_cost): Likewise.
4515 (targetm.vectorize.destroy_cost_data): Likewise.
4516 * doc/tm.texi.in (TARGET_VECTORIZE_INIT_COST): Replace with...
4517 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
4518 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
4519 (TARGET_VECTORIZE_FINISH_COST): Likewise.
4520 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
4521 * doc/tm.texi: Regenerate.
4522 * tree-vectorizer.h (vec_info::vec_info): Remove target_cost_data
4524 (vec_info::target_cost_data): Change from a void * to a vector_costs *.
4525 (vector_costs): New class.
4526 (init_cost): Take a vec_info and return a vector_costs.
4527 (dump_stmt_cost): Remove data parameter.
4528 (add_stmt_cost): Replace vinfo and data parameters with a vector_costs.
4529 (add_stmt_costs): Likewise.
4530 (finish_cost): Replace data parameter with a vector_costs.
4531 (destroy_cost_data): Delete.
4532 * tree-vectorizer.c (dump_stmt_cost): Remove data argument and
4534 (vec_info::vec_info): Remove the target_cost_data parameter and
4535 initialize the member variable to null instead.
4536 (vec_info::~vec_info): Delete target_cost_data instead of calling
4538 (vector_costs::add_stmt_cost): New function.
4539 (vector_costs::finish_cost): Likewise.
4540 (vector_costs::record_stmt_cost): Likewise.
4541 (vector_costs::adjust_cost_for_freq): Likewise.
4542 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Update
4543 call to vec_info::vec_info.
4544 (vect_compute_single_scalar_iteration_cost): Update after above
4545 changes to costing interface.
4546 (vect_analyze_loop_operations): Likewise.
4547 (vect_estimate_min_profitable_iters): Likewise.
4548 (vect_analyze_loop_2): Initialize LOOP_VINFO_TARGET_COST_DATA
4549 at the start_over point, where it needs to be recreated after
4550 trying without slp. Update retry code accordingly.
4551 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Update call
4552 to vec_info::vec_info.
4553 (vect_slp_analyze_operation): Update after above changes to costing
4555 (vect_bb_vectorization_profitable_p): Likewise.
4556 * targhooks.h (default_init_cost): Replace with...
4557 (default_vectorize_create_costs): ...this.
4558 (default_add_stmt_cost): Delete.
4559 (default_finish_cost, default_destroy_cost_data): Likewise.
4560 * targhooks.c (default_init_cost): Replace with...
4561 (default_vectorize_create_costs): ...this.
4562 (default_add_stmt_cost): Delete, moving logic to vector_costs instead.
4563 (default_finish_cost, default_destroy_cost_data): Delete.
4564 * config/aarch64/aarch64.c (aarch64_vector_costs): Inherit from
4565 vector_costs. Add a constructor.
4566 (aarch64_init_cost): Replace with...
4567 (aarch64_vectorize_create_costs): ...this.
4568 (aarch64_add_stmt_cost): Replace with...
4569 (aarch64_vector_costs::add_stmt_cost): ...this. Use record_stmt_cost
4570 to adjust the cost for inner loops.
4571 (aarch64_finish_cost): Replace with...
4572 (aarch64_vector_costs::finish_cost): ...this.
4573 (aarch64_destroy_cost_data): Delete.
4574 (TARGET_VECTORIZE_INIT_COST): Replace with...
4575 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
4576 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
4577 (TARGET_VECTORIZE_FINISH_COST): Likewise.
4578 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
4579 * config/i386/i386.c (ix86_vector_costs): New structure.
4580 (ix86_init_cost): Replace with...
4581 (ix86_vectorize_create_costs): ...this.
4582 (ix86_add_stmt_cost): Replace with...
4583 (ix86_vector_costs::add_stmt_cost): ...this. Use adjust_cost_for_freq
4584 to adjust the cost for inner loops.
4585 (ix86_finish_cost, ix86_destroy_cost_data): Delete.
4586 (TARGET_VECTORIZE_INIT_COST): Replace with...
4587 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
4588 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
4589 (TARGET_VECTORIZE_FINISH_COST): Likewise.
4590 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
4591 * config/rs6000/rs6000.c (TARGET_VECTORIZE_INIT_COST): Replace with...
4592 (TARGET_VECTORIZE_CREATE_COSTS): ...this.
4593 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
4594 (TARGET_VECTORIZE_FINISH_COST): Likewise.
4595 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
4596 (rs6000_cost_data): Inherit from vector_costs.
4597 Add a constructor. Drop loop_info, cost and costing_for_scalar
4598 in favor of the corresponding vector_costs member variables.
4599 Add "m_" to the names of the remaining member variables and
4601 (rs6000_density_test): Replace with...
4602 (rs6000_cost_data::density_test): ...this.
4603 (rs6000_init_cost): Replace with...
4604 (rs6000_vectorize_create_costs): ...this.
4605 (rs6000_update_target_cost_per_stmt): Replace with...
4606 (rs6000_cost_data::update_target_cost_per_stmt): ...this.
4607 (rs6000_add_stmt_cost): Replace with...
4608 (rs6000_cost_data::add_stmt_cost): ...this. Use adjust_cost_for_freq
4609 to adjust the cost for inner loops.
4610 (rs6000_adjust_vect_cost_per_loop): Replace with...
4611 (rs6000_cost_data::adjust_vect_cost_per_loop): ...this.
4612 (rs6000_finish_cost): Replace with...
4613 (rs6000_cost_data::finish_cost): ...this. Group loop code
4614 into a single if statement and pass the loop_vinfo down to
4616 (rs6000_destroy_cost_data): Delete.
4618 2021-11-04 Aldy Hernandez <aldyh@redhat.com>
4620 PR tree-optimization/103062
4621 PR tree-optimization/103062
4622 * value-pointer-equiv.cc (ssa_equiv_stack::ssa_equiv_stack):
4623 Increase size of allocation by 1.
4624 (ssa_equiv_stack::push_replacement): Grow as needed.
4625 (ssa_equiv_stack::get_replacement): Same.
4626 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
4627 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Remove delete.
4628 (pointer_equiv_analyzer::set_global_equiv): Grow as needed.
4629 (pointer_equiv_analyzer::get_equiv): Same.
4630 (pointer_equiv_analyzer::get_equiv_expr): Remove const.
4631 * value-pointer-equiv.h (class pointer_equiv_analyzer): Remove
4632 const markers. Use auto_vec instead of tree *.
4634 2021-11-04 Richard Biener <rguenther@suse.de>
4636 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Remove always
4637 true parameter and inline valueization.
4638 (vn_nary_op_lookup_1): Inline valueization from ...
4639 (vn_nary_op_compute_hash): ... here and remove it here.
4640 * tree-ssa-pre.c (phi_translate_1): Do not valueize
4641 before vn_nary_lookup_pieces.
4642 (get_representative_for): Mark created SSA representatives
4645 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
4647 * simplify-rtx.c (simplify_context::simplify_gen_vec_select): Assert
4648 that the operand has a vector mode. Use subreg_lowpart_offset
4649 to test whether an index corresponds to the low part.
4651 2021-11-04 Richard Sandiford <richard.sandiford@arm.com>
4653 * read-rtl.c: Remove dead !GENERATOR_FILE block.
4654 * read-rtl-function.c (function_reader::consolidate_singletons):
4655 Generate canonical CONST_VECTORs.
4657 2021-11-04 liuhongt <hongtao.liu@intel.com>
4660 * config/i386/predicates.md (reg_or_notreg_operand): Rename to ..
4661 (regmem_or_bitnot_regmem_operand): .. and extend to handle
4663 * config/i386/sse.md (*<avx512>_vpternlog<mode>_1): Force_reg
4664 the operands which are required to be register_operand.
4665 (*<avx512>_vpternlog<mode>_2): Ditto.
4666 (*<avx512>_vpternlog<mode>_3): Ditto.
4667 (*<avx512>_vternlog<mode>_all): Disallow embeded broadcast for
4668 vector HFmodes since it's not a real AVX512FP16 instruction.
4670 2021-11-04 liuhongt <hongtao.liu@intel.com>
4673 * match.pd: simplify (trunc)copysign((extend)a, (extend)b) to
4674 .COPYSIGN (a,b) when a and b are same type as the truncation
4675 type and has less precision than extend type.
4677 2021-11-04 Richard Biener <rguenther@suse.de>
4679 * doc/generic.texi: Update TARGET_MEM_REF and MEM_REF
4682 2021-11-04 Hongyu Wang <hongyu.wang@intel.com>
4684 * config/i386/sse.md (VI2_AVX512VNNIBW): New mode iterator.
4685 (VI1_AVX512VNNI): Likewise.
4686 (SDOT_VPDP_SUF): New mode_attr.
4689 (sdot_prod<mode>): Use VI2_AVX512F iterator, expand to
4690 vpdpwssd when VNNI targets available.
4691 (usdot_prod<mode>): New expander for vector QImode.
4693 2021-11-04 Hongyu Wang <hongyu.wang@intel.com>
4695 * config/i386/amxtileintrin.h (_tile_loadd_internal): Add
4696 parentheses to base and stride.
4697 (_tile_stream_loadd_internal): Likewise.
4698 (_tile_stored_internal): Likewise.
4700 2021-11-03 Maciej W. Rozycki <macro@embecosm.com>
4702 * config/riscv/riscv.c (riscv_class_max_nregs): Swap the
4703 arguments to `reg_class_subset_p'.
4705 2021-11-03 Joseph Myers <joseph@codesourcery.com>
4708 * fold-const.c (fold_init): New function.
4709 * fold-const.h (fold_init): New prototype.
4711 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
4713 * params.opt (param_vrp2_mode): Make ranger the default for VRP2.
4715 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
4717 * gimple-range-fold.cc (fold_using_range::range_of_phi): Don't import
4718 a range from edge if arg == phidef.
4720 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
4722 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Test
4723 for constant before any other processing.
4725 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
4727 * flag-types.h (RANGER_DEBUG_ALL): Fix values.
4729 2021-11-03 Andrew MacLeod <amacleod@redhat.com>
4731 * gimple-range.cc (gimple_ranger::gimple_ranger): Initialize current_bb.
4732 (gimple_ranger::range_of_expr): Pick up range_on_entry when there is
4733 no explcit context and current_bb is set.
4734 (gimple_ranger::fold_stmt): New.
4735 * gimple-range.h (current_bb, fold_stmt): New.
4736 * tree-vrp.c (rvrp_folder::fold_stmt): Call ranger's fold_stmt.
4738 2021-11-03 Richard Biener <rguenther@suse.de>
4740 PR tree-optimization/102970
4741 * tree-ssa-pre.c (phi_translate_1): Drop clique and base
4742 when translating a MEM_REF over a backedge.
4744 2021-11-03 Philipp Tomsich <philipp.tomsich@vrull.eu>
4746 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New Ampere-1 core.
4747 * config/aarch64/aarch64-tune.md: Regenerate.
4748 * config/aarch64/aarch64-cost-tables.h: Add extra costs for Ampere-1.
4749 * config/aarch64/aarch64.c: Add tuning structures for Ampere-1.
4750 * doc/invoke.texi: Add documentation for Ampere-1 core.
4752 2021-11-03 Wilco Dijkstra <wdijkstr@arm.com>
4754 * config/aarch64/aarch64.md (movsi): Add alternative for GOT accesses.
4756 (ldr_got_small_<mode>): Remove pattern.
4757 (ldr_got_small_sidi): Likewise.
4758 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately): Keep
4759 GOT accesses as moves.
4760 (aarch64_print_operand): Correctly print got_lo12 in L specifier.
4761 (aarch64_mov_operand_p): Make GOT accesses valid move operands.
4762 * config/aarch64/constraints.md: Add new constraint Usw for GOT access.
4764 2021-11-03 Martin Liska <mliska@suse.cz>
4766 * gcov.c (read_line): Remove dead variable.
4768 2021-11-03 Martin Liska <mliska@suse.cz>
4771 * ipa-fnsummary.c (edge_predicate_pool): Rename predicate class to ipa_predicate.
4772 (ipa_fn_summary::account_size_time): Likewise.
4773 (edge_set_predicate): Likewise.
4774 (set_hint_predicate): Likewise.
4775 (add_freqcounting_predicate): Likewise.
4776 (evaluate_conditions_for_known_args): Likewise.
4777 (evaluate_properties_for_edge): Likewise.
4778 (remap_freqcounting_preds_after_dup): Likewise.
4779 (ipa_fn_summary_t::duplicate): Likewise.
4780 (set_cond_stmt_execution_predicate): Likewise.
4781 (set_switch_stmt_execution_predicate): Likewise.
4782 (compute_bb_predicates): Likewise.
4783 (will_be_nonconstant_expr_predicate): Likewise.
4784 (will_be_nonconstant_predicate): Likewise.
4785 (phi_result_unknown_predicate): Likewise.
4786 (predicate_for_phi_result): Likewise.
4787 (analyze_function_body): Likewise.
4788 (compute_fn_summary): Likewise.
4789 (summarize_calls_size_and_time): Likewise.
4790 (estimate_calls_size_and_time): Likewise.
4791 (ipa_call_context::estimate_size_and_time): Likewise.
4792 (remap_edge_summaries): Likewise.
4793 (remap_freqcounting_predicate): Likewise.
4794 (ipa_merge_fn_summary_after_inlining): Likewise.
4795 (ipa_update_overall_fn_summary): Likewise.
4796 (read_ipa_call_summary): Likewise.
4797 (inline_read_section): Likewise.
4798 * ipa-fnsummary.h (struct ipa_freqcounting_predicate): Likewise.
4799 * ipa-predicate.c (predicate::add_clause): Likewise.
4800 (ipa_predicate::add_clause): Likewise.
4801 (predicate::or_with): Likewise.
4802 (ipa_predicate::or_with): Likewise.
4803 (predicate::evaluate): Likewise.
4804 (ipa_predicate::evaluate): Likewise.
4805 (predicate::probability): Likewise.
4806 (ipa_predicate::probability): Likewise.
4807 (dump_condition): Likewise.
4808 (dump_clause): Likewise.
4809 (predicate::dump): Likewise.
4810 (ipa_predicate::dump): Likewise.
4811 (predicate::debug): Likewise.
4812 (ipa_predicate::debug): Likewise.
4813 (predicate::remap_after_duplication): Likewise.
4814 (ipa_predicate::remap_after_duplication): Likewise.
4815 (predicate::remap_after_inlining): Likewise.
4816 (ipa_predicate::remap_after_inlining): Likewise.
4817 (predicate::stream_in): Likewise.
4818 (ipa_predicate::stream_in): Likewise.
4819 (predicate::stream_out): Likewise.
4820 (ipa_predicate::stream_out): Likewise.
4821 (add_condition): Likewise.
4822 * ipa-predicate.h (class predicate): Likewise.
4823 (class ipa_predicate): Likewise.
4824 (add_condition): Likewise.
4826 2021-11-03 Richard Biener <rguenther@suse.de>
4828 * bitmap.h (bitmap_bit_p): Change the return type to bool.
4829 * bitmap.c (bitmap_bit_p): Likewise.
4830 * sbitmap.h (bitmap_bit_p): Likewise.
4831 (bitmap_set_bit): Return whether the bit changed.
4832 (bitmap_clear_bit): Likewise.
4833 * tree-ssa.c (verify_vssa): Make use of the changed state
4834 from bitmap_set_bit.
4836 2021-11-03 Richard Biener <rguenther@suse.de>
4838 PR middle-end/103033
4839 * internal-fn.c (expand_DEFERRED_INIT): Elide the
4840 native_interpret_expr path in favor of folding the
4841 VIEW_CONVERT_EXPR generated when punning the RHS.
4843 2021-11-03 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
4845 * config/s390/s390.c (s390_loop_unroll_adjust): In case of early
4848 2021-11-03 Jan Hubicka <hubicka@ucw.cz>
4851 * ipa-modref.c (callee_to_caller_flags): New function.
4852 (modref_eaf_analysis::analyze_ssa_name): Use it.
4853 (ipa_merge_modref_summary_after_inlining): Fix whitespace.
4855 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
4857 * ipa-modref.c (modref_eaf_analysis::analyze_ssa_name): Revert
4860 2021-11-02 Roger Sayle <roger@nextmovesoftware.com>
4861 Uroš Bizjak <ubizjak@gmail.com>
4863 * config/i386/i386.md (<any_rotate>ti3): Provide expansion for
4864 rotations by non-constant amounts.
4866 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
4868 * ipa-modref.c: Fix anonymous namespace placement.
4869 (class modref_eaf_analysis): New class.
4870 (analyze_ssa_name_flags): Turn to ...
4871 (modref_eaf_analysis::analyze_ssa_name): ... this one.
4872 (merge_call_lhs_flags): Turn to ...
4873 (modref_eaf_analysis::merge_call_lhs_flags): .. this one
4874 (modref_eaf_analysis::merge_with_ssa_name): New member function.
4875 (record_escape_points): Turn to ...
4876 (modref_eaf_analysis::record_escape_points): ... this one.
4877 (analyze_parms): Updat
4878 (ipa_merge_modref_summary_after_inlining): Move to the end of file.
4880 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
4882 * gimple.c (gimple_call_static_chain_flags): New function.
4883 * gimple.h (gimple_call_static_chain_flags): Declare
4884 * ipa-modref.c (modref_summary::modref_summary): Initialize
4886 (modref_summary_lto::modref_summary_lto): Likewise.
4887 (modref_summary::useful_p): Test static_chain_flags.
4888 (modref_summary_lto::useful_p): Likewise.
4889 (struct modref_summary_lto): Add static_chain_flags.
4890 (modref_summary::dump): Dump static_chain_flags.
4891 (modref_summary_lto::dump): Likewise.
4892 (struct escape_point): Add static_cahin_arg.
4893 (analyze_ssa_name_flags): Use gimple_call_static_chain_flags.
4894 (analyze_parms): Handle static chains.
4895 (modref_summaries::duplicate): Duplicate static_chain_flags.
4896 (modref_summaries_lto::duplicate): Likewise.
4897 (modref_write): Stream static_chain_flags.
4898 (read_section): Likewise.
4899 (modref_merge_call_site_flags): Handle static_chain_flags.
4900 * ipa-modref.h (struct modref_summary): Add static_chain_flags.
4901 * tree-ssa-structalias.c (handle_rhs_call): Use
4902 gimple_static_chain_flags.
4904 2021-11-02 Richard Biener <rguenther@suse.de>
4906 PR tree-optimization/103029
4907 * tree-vect-loop-manip.c (vect_loop_versioning): Ensure
4908 the PHI nodes in the loop maintain their original operand
4911 2021-11-02 Jan Hubicka <hubicka@ucw.cz>
4913 * tree-core.h (EAF_NOT_RETURNED_DIRECTLY): New flag.
4914 (EAF_NOREAD): Renumber.
4915 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOT_RETURNED_DIRECTLY.
4916 (remove_useless_eaf_flags): Handle EAF_NOT_RETURNED_DIRECTLY
4917 (deref_flags): Likewise.
4918 (modref_lattice::init): Likewise.
4919 (modref_lattice::merge): Likewise.
4920 (merge_call_lhs_flags): Likewise.
4921 (analyze_ssa_name_flags): Likewise.
4922 (modref_merge_call_site_flags): Likewise.
4923 * tree-ssa-structalias.c (handle_call_arg): Likewise.
4925 2021-11-02 Maciej W. Rozycki <macro@embecosm.com>
4927 * config/riscv/riscv.c (riscv_rtx_costs): Correct a CONST_INT_P
4928 check and remove an unused local variable with shNadd/shNadd.uw
4931 2021-11-02 Tamar Christina <tamar.christina@arm.com>
4933 PR tree-optimization/103007
4934 * tree-vect-slp-patterns.c (complex_fms_pattern::matches): Add elem
4937 2021-11-02 Richard Biener <rguenther@suse.de>
4939 PR middle-end/103038
4940 * fold-const.c (native_interpret_expr): Handle OFFSET_TYPE.
4941 (can_native_interpret_type_p): Likewise.
4942 * internal-fn.c (expand_DEFERRED_INIT): View-convert the
4943 RHS if the LHS is an SSA name.
4945 2021-11-02 Richard Sandiford <richard.sandiford@arm.com>
4947 * langhooks.h (lang_hooks_for_types::simulate_record_decl): New hook.
4948 * langhooks-def.h (lhd_simulate_record_decl): Declare.
4949 (LANG_HOOKS_SIMULATE_RECORD_DECL): Define.
4950 (LANG_HOOKS_FOR_TYPES_INITIALIZER): Include it.
4951 * langhooks.c (lhd_simulate_record_decl): New function.
4953 2021-11-02 Jakub Jelinek <jakub@redhat.com>
4956 * config/i386/i386.c (ix86_vector_mode_supported_p): Reject vector
4957 modes with TImode inner mode if 32-bit.
4959 2021-11-02 liuhongt <hongtao.liu@intel.com>
4961 * doc/sourcebuild.texi (vect_slp_v4qi_store_unalign,
4962 vect_slp_v2hi_store_unalign, vect_slp_v4hi_store_unalign,
4963 vect_slp_v4si_store_unalign): Document efficient target.
4964 (vect_slp_v4qi_store_unalign_1, vect_slp_v8qi_store_unalign_1,
4965 vect_slp_v16qi_store_unalign_1): Ditto.
4966 (vect_slp_v2hi_store_align,vect_slp_v2qi_store_align,
4967 vect_slp_v2si_store_align, vect_slp_v4qi_store_align): Ditto.
4968 (struct_4char_block_move, struct_8char_block_move,
4969 struct_16char_block_move): Ditto.
4971 2021-11-02 Roger Sayle <roger@nextmovesoftware.com>
4972 Jakub Jelinek <jakub@redhat.com>
4975 * config/i386/i386-expand.c (ix86_expand_v1ti_to_ti,
4976 ix86_expand_ti_to_v1ti): New helper functions.
4977 (ix86_expand_v1ti_shift): Check if the amount operand is an
4978 integer constant, and expand as a TImode shift if it isn't.
4979 (ix86_expand_v1ti_rotate): Check if the amount operand is an
4980 integer constant, and expand as a TImode rotate if it isn't.
4981 (ix86_expand_v1ti_ashiftrt): New function to expand arithmetic
4982 right shifts of V1TImode quantities.
4983 * config/i386/i386-protos.h (ix86_expand_v1ti_ashift): Prototype.
4984 * config/i386/sse.md (ashlv1ti3, lshrv1ti3): Change constraints
4985 to QImode general_operand, and let the helper functions lower
4986 shifts by non-constant operands, as TImode shifts. Make
4987 conditional on TARGET_64BIT.
4988 (ashrv1ti3): New expander calling ix86_expand_v1ti_ashiftrt.
4989 (rotlv1ti3, rotrv1ti3): Change shift operand to QImode.
4990 Make conditional on TARGET_64BIT.
4992 2021-11-02 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
4994 * config/s390/s390.md ("*cc_to_int", "tabort", "*tabort_1",
4995 "*tabort_1_plus"): Remove operands non-null check.
4997 2021-11-01 Martin Liska <mliska@suse.cz>
4999 * opt-functions.awk: Add new sanity checking.
5000 * optc-gen.awk: Add new argument to integer_range_info.
5001 * params.opt: Update 2 params which have negative IntegerRange.
5003 2021-11-01 qing zhao <qing.zhao@oracle.com>
5005 * gimplify.c (gimplify_decl_expr): Do not add call to
5006 __builtin_clear_padding when a variable is a gimple register
5007 or it might not have padding.
5008 (gimplify_init_constructor): Likewise.
5010 2021-11-01 Tamar Christina <tamar.christina@arm.com>
5012 * config/arm/aarch-common-protos.h (struct vector_cost_table): Add
5013 movi, dup and extract costing fields.
5014 * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs,
5015 thunderx_extra_costs, thunderx2t99_extra_costs,
5016 thunderx3t110_extra_costs, tsv110_extra_costs, a64fx_extra_costs): Use
5018 * config/arm/aarch-cost-tables.h (generic_extra_costs,
5019 cortexa53_extra_costs, cortexa57_extra_costs, cortexa76_extra_costs,
5020 exynosm1_extra_costs, xgene1_extra_costs): Likewise
5021 * config/aarch64/aarch64-simd.md (aarch64_simd_dup<mode>): Add r->w dup.
5022 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add extra costs.
5024 2021-11-01 Tamar Christina <tamar.christina@arm.com>
5026 * cse.c (add_to_set): New.
5027 (find_sets_in_insn): Register constants in sets.
5028 (canonicalize_insn): Use auto_vec instead.
5029 (cse_insn): Try materializing using vec_dup.
5030 * rtl.h (simplify_context::simplify_gen_vec_select,
5031 simplify_gen_vec_select): New.
5032 * simplify-rtx.c (simplify_context::simplify_gen_vec_select): New.
5034 2021-11-01 David Malcolm <dmalcolm@redhat.com>
5036 * common.opt (fdiagnostics-escape-format=): New.
5037 (diagnostics_escape_format): New enum.
5038 (DIAGNOSTICS_ESCAPE_FORMAT_UNICODE): New enum value.
5039 (DIAGNOSTICS_ESCAPE_FORMAT_BYTES): Likewise.
5040 * diagnostic-format-json.cc (json_end_diagnostic): Add
5041 "escape-source" attribute.
5042 * diagnostic-show-locus.c
5043 (exploc_with_display_col::exploc_with_display_col): Replace
5044 "tabstop" param with a cpp_char_column_policy and add an "aspect"
5045 param. Use these to compute m_display_col accordingly.
5046 (struct char_display_policy): New struct.
5047 (layout::m_policy): New field.
5048 (layout::m_escape_on_output): New field.
5049 (def_policy): New function.
5050 (make_range): Update for changes to exploc_with_display_col ctor.
5051 (default_print_decoded_ch): New.
5052 (width_per_escaped_byte): New.
5053 (escape_as_bytes_width): New.
5054 (escape_as_bytes_print): New.
5055 (escape_as_unicode_width): New.
5056 (escape_as_unicode_print): New.
5058 (layout::layout): Initialize new fields. Update m_exploc ctor
5059 call for above change to ctor.
5060 (layout::maybe_add_location_range): Update for changes to
5061 exploc_with_display_col ctor.
5062 (layout::calculate_x_offset_display): Update for change to
5064 (layout::print_source_line): Pass policy
5065 to cpp_display_width_computation. Capture cpp_decoded_char when
5066 calling process_next_codepoint. Move printing of source code to
5067 m_policy.m_print_cb.
5068 (line_label::line_label): Pass in policy rather than context.
5069 (layout::print_any_labels): Update for change to line_label ctor.
5070 (get_affected_range): Pass in policy rather than context, updating
5071 calls to location_compute_display_column accordingly.
5072 (get_printed_columns): Likewise, also for cpp_display_width.
5073 (correction::correction): Pass in policy rather than tabstop.
5074 (correction::compute_display_cols): Pass m_policy rather than
5075 m_tabstop to cpp_display_width.
5076 (correction::m_tabstop): Replace with...
5077 (correction::m_policy): ...this.
5078 (line_corrections::line_corrections): Pass in policy rather than
5080 (line_corrections::m_context): Replace with...
5081 (line_corrections::m_policy): ...this.
5082 (line_corrections::add_hint): Update to use m_policy rather than
5084 (line_corrections::add_hint): Likewise.
5085 (layout::print_trailing_fixits): Likewise.
5086 (selftest::test_display_widths): New.
5087 (selftest::test_layout_x_offset_display_utf8): Update to use
5088 policy rather than tabstop.
5089 (selftest::test_one_liner_labels_utf8): Add test of escaping
5091 (selftest::test_diagnostic_show_locus_one_liner_utf8): Update to
5092 use policy rather than tabstop.
5093 (selftest::test_overlapped_fixit_printing): Likewise.
5094 (selftest::test_overlapped_fixit_printing_utf8): Likewise.
5095 (selftest::test_overlapped_fixit_printing_2): Likewise.
5096 (selftest::test_tab_expansion): Likewise.
5097 (selftest::test_escaping_bytes_1): New.
5098 (selftest::test_escaping_bytes_2): New.
5099 (selftest::diagnostic_show_locus_c_tests): Call the new tests.
5100 * diagnostic.c (diagnostic_initialize): Initialize
5101 context->escape_format.
5102 (convert_column_unit): Update to use default character width policy.
5103 (selftest::test_diagnostic_get_location_text): Likewise.
5104 * diagnostic.h (enum diagnostics_escape_format): New enum.
5105 (diagnostic_context::escape_format): New field.
5106 * doc/invoke.texi (-fdiagnostics-escape-format=): New option.
5107 (-fdiagnostics-format=): Add "escape-source" attribute to examples
5108 of JSON output, and document it.
5109 * input.c (location_compute_display_column): Pass in "policy"
5110 rather than "tabstop", passing to
5111 cpp_byte_column_to_display_column.
5112 (selftest::test_cpp_utf8): Update to use cpp_char_column_policy.
5113 * input.h (class cpp_char_column_policy): New forward decl.
5114 (location_compute_display_column): Pass in "policy" rather than
5116 * opts.c (common_handle_option): Handle
5117 OPT_fdiagnostics_escape_format_.
5118 * selftest.c (temp_source_file::temp_source_file): New ctor
5119 overload taking a size_t.
5120 * selftest.h (temp_source_file::temp_source_file): Likewise.
5122 2021-11-01 Aldy Hernandez <aldyh@redhat.com>
5124 * dbgcnt.def: Add debug counter for back_thread[12] and
5125 back_threadfull[12].
5126 * passes.def: Pass "first" argument to each back threading pass.
5127 * tree-ssa-threadbackward.c (back_threader::back_threader): Add
5129 (back_threader::debug_counter): New.
5130 (back_threader::maybe_register_path): Call debug_counter.
5132 2021-11-01 Aldy Hernandez <aldyh@redhat.com>
5134 * tree-ssa-threadbackward.c (BT_NONE): New.
5137 (back_threader::back_threader): Add flags.
5138 Move loop initialization here.
5139 (back_threader::~back_threader): New.
5140 (back_threader::find_taken_edge_switch): Change solver and ranger
5142 (back_threader::find_taken_edge_cond): Same.
5143 (back_threader::find_paths_to_names): Same.
5144 (back_threader::find_paths): Same.
5145 (back_threader::dump): Same.
5146 (try_thread_blocks): Merge into thread_blocks.
5147 (back_threader::thread_blocks): New.
5148 (do_early_thread_jumps): Merge into thread_blocks.
5149 (do_thread_jumps): Merge into thread_blocks.
5150 (back_threader::thread_through_all_blocks): Remove.
5152 2021-11-01 Andrew MacLeod <amacleod@redhat.com>
5154 PR tree-optimization/103003
5155 * value-relation.cc (dom_oracle::register_relation): If the 2
5156 ssa names are the same, don't register any relation.
5158 2021-11-01 Dan Li <ashimida@linux.alibaba.com>
5160 * config/aarch64/aarch64.c (aarch64_expand_epilogue): Remove
5161 redundant check for calls_eh_return.
5162 * config/aarch64/aarch64.md (*do_return): Likewise.
5164 2021-11-01 Xionghu Luo <luoxhu@linux.ibm.com>
5166 * cfghooks.c (cfg_hook_duplicate_loop_to_header_edge): Rename
5167 duplicate_loop_to_header_edge to
5168 duplicate_loop_body_to_header_edge.
5169 (cfg_hook_duplicate_loop_body_to_header_edge): Likewise.
5170 * cfghooks.h (struct cfg_hooks): Likewise.
5171 (cfg_hook_duplicate_loop_body_to_header_edge): Likewise.
5172 * cfgloopmanip.c (duplicate_loop_body_to_header_edge): Likewise.
5173 (clone_loop_to_header_edge): Likewise.
5174 * cfgloopmanip.h (duplicate_loop_body_to_header_edge): Likewise.
5175 * cfgrtl.c (struct cfg_hooks): Likewise.
5176 * doc/loop.texi: Likewise.
5177 * loop-unroll.c (unroll_loop_constant_iterations): Likewise.
5178 (unroll_loop_runtime_iterations): Likewise.
5179 (unroll_loop_stupid): Likewise.
5180 (apply_opt_in_copies): Likewise.
5181 * tree-cfg.c (struct cfg_hooks): Likewise.
5182 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
5183 (try_peel_loop): Likewise.
5184 * tree-ssa-loop-manip.c (copy_phi_node_args): Likewise.
5185 (gimple_duplicate_loop_body_to_header_edge): Likewise.
5186 (tree_transform_and_unroll_loop): Likewise.
5187 * tree-ssa-loop-manip.h (gimple_duplicate_loop_body_to_header_edge):
5190 2021-11-01 Xionghu Luo <luoxhu@linux.ibm.com>
5192 * cfgloopmanip.c (loop_version): Refactor loopify to
5193 loop_version. Move condition generation after loopify.
5195 * cfgloopmanip.h (loopify): Delete.
5197 2021-10-31 Jan Hubicka <hubicka@ucw.cz>
5199 * ipa-fnsummary.c: Include tree-dfa.h.
5200 (points_to_local_or_readonly_memory_p): Return true on return
5202 * ipa-modref.c (analyze_ssa_name_flags): Fix handling of copy
5205 2021-10-30 Tobias Burnus <tobias@codesourcery.com>
5207 PR middle-end/102972
5208 * omp-low.c (omp_runtime_api_call): Use DECL_ASSEMBLER_NAME to get
5209 internal Fortran name; new permit_num_teams arg to permit
5210 omp_get_num_teams and omp_get_team_num.
5211 (scan_omp_1_stmt): Update call to it, add missing call for
5212 reverse offload, and check for strictly nested API calls in teams.
5214 2021-10-30 Jakub Jelinek <jakub@redhat.com>
5216 * gimplify.c (gimplify_omp_for): Diagnose threadprivate iterators.
5218 2021-10-29 David Malcolm <dmalcolm@redhat.com>
5220 * selftest.c (assert_streq): Add newlines when emitting non-equal
5223 2021-10-29 David Malcolm <dmalcolm@redhat.com>
5225 * Makefile.in: Fix syntax for reference to LIBDEPS in
5228 2021-10-29 Bernhard Reutner-Fischer <aldot@gcc.gnu.org>
5230 * doc/install.texi: Bump required minimum DejaGnu version.
5232 2021-10-29 Aldy Hernandez <aldyh@redhat.com>
5233 Andrew MacLeod <amacleod@redhat.com>
5235 * value-relation.cc (path_oracle::killing_def): Add a
5236 self-equivalence so we don't look to the root oracle.
5238 2021-10-29 Aldy Hernandez <aldyh@redhat.com>
5240 * passes.def: Replace the pass_thread_jumps before VRP* with
5241 pass_thread_jumps_full. Remove all pass_vrp_threader instances.
5242 * tree-ssa-threadbackward.c (pass_data_thread_jumps_full):
5243 Remove hyphen from "thread-full" name.
5245 2021-10-29 Andrew MacLeod <amacleod@redhat.com>
5247 PR tree-optimization/102983
5248 * gimple-range-cache.h (propagate_updated_value): Make public.
5249 * gimple-range.cc (gimple_ranger::range_of_stmt): Propagate exports
5250 when processing gcond stmts.
5252 2021-10-29 Jan Hubicka <hubicka@ucw.cz>
5254 * gimple.c (gimple_call_retslot_flags): New function.
5255 * gimple.h (gimple_call_retslot_flags): Declare.
5256 * ipa-modref.c: Include tree-cfg.h.
5257 (struct escape_entry): Turn parm_index to signed.
5258 (modref_summary_lto::modref_summary_lto): Add retslot_flags.
5259 (modref_summary::modref_summary): Initialize retslot_flags.
5260 (struct modref_summary_lto): Likewise.
5261 (modref_summary::useful_p): Check retslot_flags.
5262 (modref_summary_lto::useful_p): Likewise.
5263 (modref_summary::dump): Dump retslot_flags.
5264 (modref_summary_lto::dump): Likewise.
5265 (struct escape_point): Add hidden_args enum.
5266 (analyze_ssa_name_flags): Ignore return slot return;
5267 use gimple_call_retslot_flags.
5268 (record_escape_points): Break out from ...
5269 (analyze_parms): ... here; handle retslot_flags.
5270 (modref_summaries::duplicate): Duplicate retslot_flags.
5271 (modref_summaries_lto::duplicate): Likewise.
5272 (modref_write_escape_summary): Stream parm_index as signed.
5273 (modref_read_escape_summary): Likewise.
5274 (modref_write): Stream retslot_flags.
5275 (read_section): Likewise.
5276 (struct escape_map): Fix typo in comment.
5277 (update_escape_summary_1): Fix whitespace.
5278 (ipa_merge_modref_summary_after_inlining): Drop retslot_flags.
5279 (modref_merge_call_site_flags): Merge retslot_flags.
5280 * ipa-modref.h (struct modref_summary): Add retslot_flags.
5281 * tree-ssa-structalias.c (handle_rhs_call): Handle retslot_flags.
5283 2021-10-29 Tamar Christina <tamar.christina@arm.com>
5285 PR tree-optimization/102977
5286 * tree-vect-slp-patterns.c (vect_match_call_p): Remove.
5287 (vect_detect_pair_op): Add crosslane check.
5288 (vect_match_call_complex_mla): Remove.
5289 (class complex_mul_pattern): Update comment.
5290 (complex_mul_pattern::matches): Update detection.
5291 (class complex_fma_pattern): Remove.
5292 (complex_fma_pattern::matches): Remove.
5293 (complex_fma_pattern::recognize): Remove.
5294 (complex_fma_pattern::build): Remove.
5295 (class complex_fms_pattern): Update comment.
5296 (complex_fms_pattern::matches): Remove.
5297 (complex_operations_pattern::recognize): Remove complex_fma_pattern
5299 2021-10-29 Jakub Jelinek <jakub@redhat.com>
5301 * gimple-fold.c (gimple_fold_builtin_memset): Copy over location from
5304 2021-10-29 Xionghu Luo <luoxhu@linux.ibm.com>
5307 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Add
5308 patterns match and emit for VSX xxpermdi.
5310 2021-10-29 liuhongt <hongtao.liu@intel.com>
5313 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF): New
5315 (V16HF_FTYPE_V16HF): Ditto.
5316 (V32HF_FTYPE_V32HF): Ditto.
5317 (V8HF_FTYPE_V8HF_ROUND): Ditto.
5318 (V16HF_FTYPE_V16HF_ROUND): Ditto.
5319 (V32HF_FTYPE_V32HF_ROUND): Ditto.
5320 * config/i386/i386-builtin.def ( IX86_BUILTIN_FLOORPH,
5321 IX86_BUILTIN_CEILPH, IX86_BUILTIN_TRUNCPH,
5322 IX86_BUILTIN_FLOORPH256, IX86_BUILTIN_CEILPH256,
5323 IX86_BUILTIN_TRUNCPH256, IX86_BUILTIN_FLOORPH512,
5324 IX86_BUILTIN_CEILPH512, IX86_BUILTIN_TRUNCPH512): New builtin.
5325 * config/i386/i386-builtins.c
5326 (ix86_builtin_vectorized_function): Enable vectorization for
5327 HFmode FLOOR/CEIL/TRUNC operation.
5328 * config/i386/i386-expand.c (ix86_expand_args_builtin): Handle
5330 * config/i386/sse.md (rint<mode>2, nearbyint<mode>2): Extend
5333 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
5334 Andrew MacLeod <amacleod@redhat.com>
5336 * value-relation.cc (path_oracle::killing_def): Walk the
5337 equivalency list and remove SSA from any equivalencies.
5339 2021-10-28 Stafford Horne <shorne@gmail.com>
5341 * config/or1k/or1k.h (PROFILE_HOOK): Add return address argument
5344 2021-10-28 Jakub Jelinek <jakub@redhat.com>
5346 PR tree-optimization/102951
5347 * fold-const.h (address_compare): Declare.
5348 * fold-const.c (address_compare): New function.
5349 * match.pd (cmp (convert1?@2 addr@0) (convert2? addr@1)): Use
5350 address_compare helper.
5351 (minmax cmp (convert1?@2 addr@0) (convert2?@3 addr@1)): New
5354 2021-10-28 Andrew MacLeod <amacleod@redhat.com>
5356 * vr-values.c (simplify_using_ranges::fold_cond): Change fold message.
5358 2021-10-28 Andrew MacLeod <amacleod@redhat.com>
5360 PR tree-optimization/102940
5361 * tree-vrp.c (execute_ranger_vrp): Reset scev.
5363 2021-10-28 Richard Purdie <richard.purdie@linuxfoundation.org>
5365 * config/nios2/linux.h (MUSL_DYNAMIC_LINKER): Add musl linker
5367 2021-10-28 Richard Purdie <richard.purdie@linuxfoundation.org>
5369 * configure: Regenerate.
5370 * configure.ac: Use CPPFLAGS_FOR_BUILD for GMPINC
5372 2021-10-28 Eric Botcazou <ebotcazou@adacore.com>
5374 * doc/invoke.texi (%X): Remove obsolete reference to -Wl.
5376 2021-10-28 Richard Biener <rguenther@suse.de>
5379 * fold-const.c (fold_convert_const): Avoid int to float
5380 constant folding with -frounding-math and inexact result.
5381 * simplify-rtx.c (simplify_const_unary_operation): Likewise
5382 for both float and unsigned_float.
5384 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
5386 * tree-ssa-threadbackward.c
5387 (back_threader::find_taken_edge_switch): Use find_case_label_range
5388 instead of find_taken_edge.
5390 2021-10-28 Aldy Hernandez <aldyh@redhat.com>
5392 * tree-ssa-threadbackward.c (class back_threader_registry):
5393 Inherit from back_jt_path_registry.
5394 (back_threader_registry::thread_through_all_blocks): Remove.
5395 (back_threader_registry::register_path): Remove
5396 m_lowlevel_registry prefix.
5398 2021-10-28 Richard Biener <rguenther@suse.de>
5401 * fold-const.c (fold_convert_const_real_from_real): Honor
5402 -frounding-math if the conversion is not exact.
5403 * simplify-rtx.c (simplify_const_unary_operation): Do not
5404 simplify FLOAT_TRUNCATE with sign dependent rounding.
5406 2021-10-28 Richard Biener <rguenther@suse.de>
5408 PR tree-optimization/102949
5409 * tree-vect-stmts.c (ensure_base_align): Look at the
5410 dr_info of a group leader and assert we are looking at
5411 one with analyzed alignment.
5413 2021-10-28 Kewen Lin <linkw@linux.ibm.com>
5416 * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Consider
5417 V1T1 mode for unaligned load and store.
5419 2021-10-28 Kito Cheng <kito.cheng@sifive.com>
5421 * config/riscv/riscv.md (zero_extendsidi2_internal): Allow ZBB
5424 2021-10-28 Kito Cheng <kito.cheng@sifive.com>
5426 * config/riscv/arch-canonicalize (CANONICAL_ORDER): Add `i` to
5429 2021-10-28 Alexandre Oliva <oliva@adacore.com>
5431 * common.opt (fharden-compares): New.
5432 (fharden-conditional-branches): New.
5433 * doc/invoke.texi: Document new options.
5434 * gimple-harden-conditionals.cc: New.
5435 * Makefile.in (OBJS): Build it.
5436 * passes.def: Add new passes.
5437 * tree-pass.h (make_pass_harden_compares): Declare.
5438 (make_pass_harden_conditional_branches): Declare.
5440 2021-10-28 Xionghu Luo <luoxhu@linux.ibm.com>
5443 * config/rs6000/altivec.md: Add vsx register constraints.
5444 * config/rs6000/vsx.md (vsx_xxsel<mode>): Delete.
5445 (vsx_xxsel<mode>2): Likewise.
5446 (vsx_xxsel<mode>3): Likewise.
5447 (vsx_xxsel<mode>4): Likewise.
5449 2021-10-28 Xionghu Luo <luoxhu@linux.ibm.com>
5452 * config/rs6000/altivec.md (*altivec_vsel<mode>): Change to ...
5453 (altivec_vsel<mode>): ... this and update define.
5454 (*altivec_vsel<mode>_uns): Delete.
5455 (altivec_vsel<mode>2): New define_insn.
5456 (altivec_vsel<mode>3): Likewise.
5457 (altivec_vsel<mode>4): Likewise.
5458 * config/rs6000/rs6000-call.c (altivec_expand_vec_sel_builtin): New.
5459 (altivec_expand_builtin): Call altivec_expand_vec_sel_builtin to expand
5461 * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Use bit-wise
5462 selection instead of per element.
5463 * config/rs6000/vector.md:
5464 * config/rs6000/vsx.md (*vsx_xxsel<mode>): Change to ...
5465 (vsx_xxsel<mode>): ... this and update define.
5466 (*vsx_xxsel<mode>_uns): Delete.
5467 (vsx_xxsel<mode>2): New define_insn.
5468 (vsx_xxsel<mode>3): Likewise.
5469 (vsx_xxsel<mode>4): Likewise.
5471 2021-10-28 Hongyu Wang <hongyu.wang@intel.com>
5473 * config/i386/i386.c (use_rsqrt_p): Add mode parameter, enable
5474 HFmode rsqrt without TARGET_SSE_MATH.
5475 (ix86_optab_supported_p): Refactor rint, adjust floor, ceil,
5476 btrunc condition to be restricted by -ftrapping-math, adjust
5477 use_rsqrt_p function call.
5478 * config/i386/i386.md (rcphf2): New define_insn.
5479 (rsqrthf2): Likewise.
5480 * config/i386/sse.md (div<mode>3): Change VF2H to VF2.
5481 (div<mode>3): New expander for HF mode.
5482 (rsqrt<mode>2): Likewise.
5483 (*avx512fp16_vmrcpv8hf2): New define_insn for rpad pass.
5484 (*avx512fp16_vmrsqrtv8hf2): Likewise.
5486 2021-10-27 Saagar Jha <saagar@saagarjha.com>
5488 * config.gcc: Adjust for Darwin21.
5489 * config/darwin-c.c (macosx_version_as_macro): Likewise.
5490 * config/darwin-driver.c (validate_macosx_version_min):
5492 (darwin_find_version_from_kernel): Likewise.
5494 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
5496 * gimple-range-path.cc
5497 (path_range_query::range_defined_in_block): Call killing_def.
5499 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
5501 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Dump
5502 operands as well as relation.
5503 * gimple-range-path.cc
5504 (path_range_query::compute_ranges_in_block): Compute PHI relations
5505 first. Compute outgoing relations at the end.
5506 (path_range_query::compute_ranges): Remove call to compute_relations.
5507 (path_range_query::compute_relations): Remove.
5508 (path_range_query::maybe_register_phi_relation): New.
5509 (path_range_query::compute_phi_relations): Abstract out
5510 registering one PHI relation to...
5511 (path_range_query::compute_outgoing_relations): ...here.
5512 * gimple-range-path.h (class path_range_query): Remove
5514 Add maybe_register_phi_relation.
5516 2021-10-27 Aldy Hernandez <aldyh@redhat.com>
5517 Andrew MacLeod <amacleod@redhat.com>
5519 * value-relation.cc (path_oracle::killing_def): Kill second
5522 2021-10-27 Martin Jambor <mjambor@suse.cz>
5524 * ipa-cp.c (good_cloning_opportunity_p): Decide whether to use
5525 profile feedback depending on their local availability.
5527 2021-10-27 Martin Jambor <mjambor@suse.cz>
5529 * params.opt (param_ipa_cp_profile_count_base): New parameter.
5530 * doc/invoke.texi (Optimize Options): Add entry for
5531 ipa-cp-profile-count-base.
5532 * ipa-cp.c (max_count): Replace with base_count, replace all
5533 occurrences too, unless otherwise stated.
5534 (ipcp_cloning_candidate_p): identify mostly-directly called
5535 functions based on their counts, not max_count.
5536 (compare_edge_profile_counts): New function.
5537 (ipcp_propagate_stage): Instead of setting max_count, find the
5538 appropriate edge count in a sorted vector of counts of eligible
5539 edges and make it the base_count.
5541 2021-10-27 Martin Jambor <mjambor@suse.cz>
5543 * ipa-cp.c (struct caller_statistics): New fields rec_count_sum,
5544 n_nonrec_calls and itself, document all fields.
5545 (init_caller_stats): Initialize the above new fields.
5546 (gather_caller_stats): Gather self-recursive counts and calls number.
5547 (get_info_about_necessary_edges): Gather counts of self-recursive and
5548 other edges bringing in the requested value separately.
5549 (dump_profile_updates): Rework to dump info about a single node only.
5550 (lenient_count_portion_handling): New function.
5551 (struct gather_other_count_struct): New type.
5552 (gather_count_of_non_rec_edges): New function.
5553 (struct desc_incoming_count_struct): New type.
5554 (analyze_clone_icoming_counts): New function.
5555 (adjust_clone_incoming_counts): Likewise.
5556 (update_counts_for_self_gen_clones): Likewise.
5557 (update_profiling_info): Rewritten.
5558 (update_specialized_profile): Adjust call to dump_profile_updates.
5559 (create_specialized_node): Do not update profiling info.
5560 (decide_about_value): New parameter self_gen_clones, either push new
5561 clones into it or updat their profile counts. For self-recursively
5562 generated values, use a portion of the node count instead of count
5563 from self-recursive edges to estimate goodness.
5564 (decide_whether_version_node): Gather clones for self-generated values
5565 in a new vector, update their profiles at once at the end.
5567 2021-10-27 Richard Biener <rguenther@suse.de>
5569 * tree-vectorizer.c (vect_transform_loops): New function,
5571 (try_vectorize_loop_1): ... here. Simplify as epilogues
5572 are now fully handled in the split part.
5574 2021-10-27 Jakub Jelinek <jakub@redhat.com>
5576 * omp-expand.c (expand_omp_for_init_counts): Handle non-rectangular
5577 iterators with pointer types.
5578 (expand_omp_for_init_vars, extract_omp_for_update_vars): Likewise.
5580 2021-10-26 Martin Sebor <msebor@redhat.com>
5582 PR tree-optimization/102238
5583 PR tree-optimization/102919
5584 * gimple-ssa-sprintf.c (get_string_length): Add an argument.
5585 (array_elt_at_offset): Move to pointer-query.
5586 (set_aggregate_size_and_offset): New function.
5587 (field_at_offset): Move to pointer-query.
5588 (get_origin_and_offset): Rename...
5589 (get_origin_and_offset_r): this. Add an argument. Make aggregate
5590 handling more robust.
5591 (get_origin_and_offset): New.
5592 (alias_offset): Add an argument.
5593 (format_string): Use subobject size determined by get_origin_and_offset.
5594 * pointer-query.cc (field_at_offset): Move from gimple-ssa-sprintf.c.
5595 Improve/correct handling of aggregates.
5596 (array_elt_at_offset): Same.
5597 * pointer-query.h (field_at_offset): Declare.
5598 (array_elt_at_offset): Declare.
5600 2021-10-26 Martin Sebor <msebor@redhat.com>
5602 * builtins.c (check_strncat_sizes): Pass access_data ctor additional
5604 (expand_builtin_memcmp): Move code to gimple-ssa-warn-access.cc.
5605 (expand_builtin_fork_or_exec): Same.
5606 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Pass
5607 compute_objsize additional arguments.
5608 (inbounds_memaccess_p): Same.
5609 (array_bounds_checker::check_array_bounds): Add an assert. Stash
5610 statement in a member.
5611 (check_array_bounds_dom_walker::before_dom_children): Same.
5612 * gimple-array-bounds.h (array_bounds_checker::m_stmt): New member.
5613 * gimple-ssa-sprintf.c (get_destination_size): Add an argument.
5614 (handle_printf_call): Pass a new argument.
5615 * gimple-ssa-warn-access.cc (get_size_range): Add an argument.
5616 (check_access): Add an argument and pass it along to callees.
5617 (check_read_access): Make a member function.
5618 (pass_waccess::check_strcat): Pass access_data ctor additional
5620 (pass_waccess::check_strncat): Same.
5621 (pass_waccess::check_stxcpy): Same.
5622 (pass_waccess::check_stxncpy): Same.
5623 (pass_waccess::check_strncmp): Same.
5624 (pass_waccess::check_read_access): Same.
5625 (pass_waccess::check_builtin): Same.
5626 (pass_waccess::maybe_check_access_sizes): Same.
5627 (pass_waccess::maybe_check_dealloc_call): Same.
5628 * gimple-ssa-warn-access.h (check_read_access): Declare a new
5630 * pointer-query.cc (compute_objsize_r): Add an argument.
5631 (gimple_call_return_array): Same.
5632 (gimple_call_alloc_size): Same.
5633 (access_ref::access_ref): Same.
5634 (access_ref::get_ref): Same.
5635 (pointer_query::get_ref): Same.
5636 (handle_min_max_size): Pass an arguments to callees.
5637 (handle_array_ref): Add an argument.
5638 (handle_mem_ref): Same.
5639 (compute_objsize): Same.
5640 * pointer-query.h (struct access_ref): Adjust signatures.
5641 (struct access_data): Same.
5642 (gimple_call_alloc_size): Add an argument.
5643 (gimple_parm_array_size): Same.
5644 (compute_objsize): Same.
5645 * tree-ssa-strlen.c (strlen_pass::adjust_last_stmt): Pass an additional
5646 argument to compute_objsize.
5647 (strlen_pass::maybe_warn_overflow): Same.
5648 (maybe_diag_stxncpy_trunc): Same.
5650 2021-10-26 Martin Sebor <msebor@redhat.com>
5652 PR middle-end/102453
5653 * gimple-ssa-warn-access.cc (pass_waccess::check_atomic_builtin): New.
5654 (pass_waccess::check_atomic_builtin): Call it.
5656 2021-10-26 Vladimir N. Makarov <vmakarov@redhat.com>
5658 PR rtl-optimization/102842
5659 * lra-constraints.c (match_reload): Ignore out in checking values
5661 (curr_insn_transform): Collect outputs before doing reloads of operands.
5663 2021-10-26 Paul A. Clarke <pc@us.ibm.com>
5666 * config/rs6000/x86intrin.h: Move some included headers to new
5667 headers. Include new immintrin.h instead of those headers.
5668 * config/rs6000/immintrin.h: New.
5669 * config/rs6000/x86gprintrin.h: New.
5670 * config.gcc (powerpc*-*-*): Add new headers to extra_headers.
5672 2021-10-26 Richard Biener <rguenther@suse.de>
5674 * tree-vectorizer.h (vect_create_addr_base_for_vector_ref):
5675 Remove byte_offset parameter.
5676 (vect_create_data_ref_ptr): Likewise.
5677 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
5679 (vect_create_data_ref_ptr): Likewise.
5680 * tree-vect-stmts.c (vectorizable_store): Adjust.
5681 (vectorizable_load): Likewise.
5683 2021-10-26 Richard Biener <rguenther@suse.de>
5685 PR tree-optimization/96109
5686 * tree-vectorizer.h (dr_misalignment): Add optional offset
5688 * tree-vect-data-refs.c (dr_misalignment): Likewise. Remove
5689 offset applied for negative stride accesses.
5690 (vect_enhance_data_refs_alignment): Compute negative stride
5691 access offset and pass it to dr_misalignment.
5692 * tree-vect-stmts.c (get_negative_load_store_type): Pass
5693 negative offset to dr_misalignment.
5694 (get_group_load_store_type): Likewise.
5695 (get_load_store_type): Likewise.
5696 (vectorizable_store): Remove asserts about alignment.
5697 (vectorizable_load): Likewise.
5699 2021-10-26 Kewen Lin <linkw@linux.ibm.com>
5701 PR tree-optimization/102897
5702 * tree-ssa-forwprop.c (simplify_permutation): Remove a wrong assertion.
5704 2021-10-26 Richard Biener <rguenther@suse.de>
5706 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
5707 Take offset in bytes.
5708 (vect_create_data_ref_ptr): Likewise.
5709 * tree-vect-loop-manip.c (get_misalign_in_elems): Multiply
5710 offset by element size.
5711 (vect_create_cond_for_align_checks): Likewise.
5712 * tree-vect-stmts.c (get_negative_load_store_type): Likewise.
5713 (vectorizable_load): Remove duplicate leftover from merge
5716 2021-10-26 Roger Sayle <roger@nextmovesoftware.com>
5718 * config/i386/i386-expand.c (ix86_expand_v1ti_shift): New helper
5719 function to expand V1TI mode logical shifts by integer constants.
5720 (ix86_expand_v1ti_rotate): New helper function to expand V1TI
5721 mode rotations by integer constants.
5722 * config/i386/i386-protos.h (ix86_expand_v1ti_shift,
5723 ix86_expand_v1ti_rotate): Prototype new functions here.
5724 * config/i386/sse.md (ashlv1ti3, lshrv1ti3, rotlv1ti3, rotrv1ti3):
5725 New TARGET_SSE2 expanders to implement V1TI shifts and rotations.
5727 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
5729 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
5730 Avoid threading circular paths.
5732 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
5734 * tree-ssa-threadbackward.c (back_threader::resolve_phi):
5735 Attempt to resolve all incoming paths to a PHI.
5736 (back_threader::resolve_def): Always return true for PHIs.
5738 2021-10-26 Aldy Hernandez <aldyh@redhat.com>
5740 * tree-ssa-threadbackward.c (back_threader::find_paths_to_names):
5741 Always try to resolve path without looking back.
5742 * tree-ssa-threadupdate.c (dump_jump_thread): Indidicate whether
5743 edge is a back edge.
5745 2021-10-26 Kewen Lin <linkw@linux.ibm.com>
5747 PR tree-optimization/102789
5748 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
5749 update inits of simd_lane_access.
5751 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
5753 * tree-vrp.c (vrp_simplify_cond_using_ranges): Add return type and
5754 move to vr-values.c.
5755 (simplify_casted_conds): Move to vrp_folder class.
5756 (execute_vrp): Call via vrp_folder now.
5757 * vr-values.c (simplify_cond_using_ranges_1): Call simplify_casted_cond.
5758 (simplify_using_ranges::simplify_casted_cond): Relocate from tree-vrp.c.
5759 * vr-values.h (simplify_casted_cond): Add prototype.
5761 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
5763 * tree-vrp.c (rvrp_folder::fold_stmt): If simplification fails, try
5766 2021-10-25 Paul A. Clarke <pc@us.ibm.com>
5768 * config/rs6000/smmintrin.h (_mm_testz_si128): Add "extern" to
5770 (_mm_testc_si128): Likewise.
5771 (_mm_testnzc_si128): Likewise.
5772 (_mm_blend_ps): Likewise.
5773 (_mm_blendv_ps): Likewise.
5774 (_mm_blend_pd): Likewise.
5775 (_mm_blendv_pd): Likewise.
5776 (_mm_ceil_pd): Likewise.
5777 (_mm_ceil_sd): Likewise.
5778 (_mm_ceil_ps): Likewise.
5779 (_mm_ceil_ss): Likewise.
5780 (_mm_floor_pd): Likewise.
5781 (_mm_floor_sd): Likewise.
5782 (_mm_floor_ps): Likewise.
5783 (_mm_floor_ss): Likewise.
5784 (_mm_minpos_epu16): Likewise.
5785 (_mm_mul_epi32): Likewise.
5786 (_mm_cvtepi8_epi16): Likewise.
5787 (_mm_packus_epi32): Likewise.
5788 (_mm_cmpgt_epi64): Likewise.
5790 2021-10-25 Roger Sayle <roger@nextmovesoftware.com>
5792 * simplify-rtx.c (simplify_binary_operation_1) [SS_ASHIFT]: Simplify
5793 shifts of the mode's smin_value and smax_value when the bit count
5794 operand doesn't have side-effects.
5795 [US_ASHIFT]: Likewise, simplify shifts of the mode's umax_value
5796 when the bit count operand doesn't have side-effects.
5797 (simplify_const_binary_operation) [SS_ASHIFT, US_ASHIFT]: Perform
5798 compile-time evaluation of saturating left shifts with constant
5801 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
5803 * gimple-range.cc (gimple_ranger::export_global_ranges): Remove check
5806 2021-10-25 Andrew MacLeod <amacleod@redhat.com>
5808 * flag-types.h (enum ranger_debug): Adjust values.
5809 * params.opt (ranger_debug): Ditto.
5811 2021-10-25 Martin Jambor <mjambor@suse.cz>
5813 PR tree-optimization/102886
5814 * tree-sra.c (totally_scalarize_subtree): Fix the out of
5817 2021-10-25 Andrew Pinski <apinski@marvell.com>
5819 * tree-ssa-dce.c (simple_dce_from_worklist):
5820 Check stmt_unremovable_because_of_non_call_eh_p also
5821 before removing the statement.
5823 2021-10-25 Richard Biener <rguenther@suse.de>
5825 PR tree-optimization/102905
5826 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
5827 Use vect_supportable_dr_alignment again to determine whether
5828 an access is supported when not aligned.
5830 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
5832 * config/riscv/riscv.c (riscv_rtx_costs): Handle cost model
5835 2021-10-25 Jim Wilson <jimw@sifive.com>
5836 Kito Cheng <kito.cheng@sifive.com>
5837 Jia-Wei Chen <jiawei@iscas.ac.cn>
5838 Shi-Hua Liao <shihua@iscas.ac.cn>
5840 * config/riscv/bitmanip.md (shiftm1): New.
5841 (*bset<mode>): Ditto.
5842 (*bset<mode>_mask): Ditto.
5843 (*bset<mode>_1): Ditto.
5844 (*bset<mode>_1_mask): Ditto.
5845 (*bseti<mode>): Ditto.
5846 (*bclr<mode>): Ditto.
5847 (*bclri<mode>): Ditto.
5848 (*binv<mode>): Ditto.
5849 (*binvi<mode>): Ditto.
5850 (*bext<mode>): Ditto.
5852 * config/riscv/predicates.md (splittable_const_int_operand):
5854 (single_bit_mask_operand): New.
5855 (not_single_bit_mask_operand): Ditto.
5856 (const31_operand): Ditto.
5857 (const63_operand): Ditto.
5858 * config/riscv/riscv.c (riscv_build_integer_1): Handle bseti.
5859 (riscv_output_move): Ditto.
5860 (riscv_print_operand): Handle new operand type: T and S.
5861 * config/riscv/riscv.h (SINGLE_BIT_MASK_OPERAND): New.
5863 2021-10-25 Jim Wilson <jimw@sifive.com>
5865 * config/riscv/riscv.c (riscv_build_integer_1): Build integer
5868 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
5870 * config/riscv/riscv.c (riscv_extend_cost): Handle cost model
5872 (riscv_rtx_costs): Ditto.
5874 2021-10-25 Jim Wilson <jimw@sifive.com>
5875 Kito Cheng <kito.cheng@sifive.com>
5876 Jia-Wei Chen <jiawei@iscas.ac.cn>
5878 * config/riscv/bitmanip.md (bitmanip_bitwise): New.
5879 (bitmanip_minmax): New.
5880 (clz_ctz_pcnt): New.
5881 (bitmanip_optab): New.
5882 (bitmanip_insn): New.
5883 (*<optab>_not<mode>): New.
5884 (*xor_not<mode>): New.
5885 (<bitmanip_optab>si2): New.
5886 (*<bitmanip_optab>disi2): New.
5887 (<bitmanip_optab>di2): New.
5888 (*zero_extendhi<GPR:mode>2_bitmanip): New.
5889 (*extend<SHORT:mode><SUPERQI:mode>2_zbb): New.
5890 (*zero_extendhi<GPR:mode>2_zbb): New.
5893 (rotrsi3_sext): New.
5896 (rotlsi3_sext): New.
5897 (bswap<mode>2): New.
5898 (<bitmanip_optab><mode>3): New.
5899 * config/riscv/riscv.md (type): Add rotate.
5900 (zero_extendhi<GPR:mode>2): Change to define_expand pattern.
5901 (*zero_extendhi<GPR:mode>2): New.
5902 (extend<SHORT:mode><SUPERQI:mode>2): Change to define_expand pattern.
5903 (*extend<SHORT:mode><SUPERQI:mode>2): New.
5905 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
5907 * config/riscv/riscv.c (riscv_extend_cost): Handle cost model
5909 (riscv_rtx_costs): Ditto.
5911 2021-10-25 Jim Wilson <jimw@sifive.com>
5912 Kito Cheng <kito.cheng@sifive.com>
5913 Jia-Wei Chen <jiawei@iscas.ac.cn>
5915 * config/riscv/bitmanip.md (*zero_extendsidi2_bitmanip): New.
5920 (riscv_rtx_costs): Ditto.
5921 * config/riscv/riscv.md: Include bitmanip.md
5922 (type): Add bitmanip bype.
5923 (zero_extendsidi2): Change to define_expand pattern.
5924 (*zero_extendsidi2_internal): New.
5925 (zero_extendsidi2_shifted): Disable for ZBA.
5927 2021-10-25 Kito Cheng <kito.cheng@sifive.com>
5929 * common/config/riscv/riscv-common.c (riscv_ext_version_table):
5930 Add zba, zbb, zbc and zbs.
5931 (riscv_ext_flag_table): Ditto.
5932 * config/riscv/riscv-opts.h (MASK_ZBA): New.
5936 (TARGET_ZBA): Ditto.
5937 (TARGET_ZBB): Ditto.
5938 (TARGET_ZBC): Ditto.
5939 (TARGET_ZBS): Ditto.
5940 * config/riscv/riscv.opt (riscv_zb_subext): New.
5942 2021-10-25 liuhongt <hongtao.liu@intel.com>
5945 * match.pd: Simplify (_Float16) sqrtf((float) a) to .SQRT(a)
5946 when direct_internal_fn_supported_p, similar for sqrt/sqrtl.
5948 2021-10-25 Richard Biener <rguenther@suse.de>
5950 PR tree-optimization/102920
5951 * tree-ssa-sccvn.h (expressions_equal_p): Add argument
5952 controlling VN_TOP matching behavior.
5953 * tree-ssa-sccvn.c (expressions_equal_p): Likewise.
5954 (vn_phi_eq): Do not optimistically match VN_TOP.
5956 2021-10-25 konglin1 <lingling.kong@intel.com>
5958 * config/i386/sse.md (fma_<mode>_fadd_fmul): Add new
5959 define_insn_and_split.
5960 (fma_<mode>_fadd_fcmul):Likewise
5961 (fma_<complexopname>_<mode>_fma_zero):Likewise
5963 2021-10-24 John David Anglin <danglin@gcc.gnu.org>
5965 * config/pa/pa-d.c (pa_d_handle_target_float_abi): Don't check
5966 TARGET_DISABLE_FPREGS.
5967 * config/pa/pa.c (fix_range): Use MASK_SOFT_FLOAT instead of
5968 MASK_DISABLE_FPREGS.
5969 (hppa_rtx_costs): Don't check TARGET_DISABLE_FPREGS. Adjust
5970 cost of hardware integer multiplication.
5971 (pa_conditional_register_usage): Don't check TARGET_DISABLE_FPREGS.
5972 * config/pa/pa.h (INT14_OK_STRICT): Likewise.
5973 * config/pa/pa.md: Don't check TARGET_DISABLE_FPREGS. Check
5974 TARGET_SOFT_FLOAT in patterns that use xmpyu instruction.
5975 * config/pa/pa.opt (mdisable-fpregs): Change target mask to
5976 SOFT_FLOAT. Revise comment.
5977 (msoft-float): New option.
5979 2021-10-24 John David Anglin <danglin@gcc.gnu.org>
5981 * config/pa/pa.md: Don't use 'G' constraint in integer move patterns.
5983 2021-10-24 Gerald Pfeifer <gerald@pfeifer.com>
5985 * doc/install.texi (Specific): Remove obsolete details
5986 around GNU/Linux on Itanium.
5987 (Specific): Remove reference to Windows for Itanium.
5989 2021-10-23 Bernhard Reutner-Fischer <aldot@gcc.gnu.org>
5991 * config/i386/x86-tune-sched-bd.c (dispatch_group): Commentary
5994 2021-10-23 Jan Hubicka <hubicka@ucw.cz>
5996 * tree-ssa-structalias.c (compute_points_to_sets): Cleanup.
5998 2021-10-23 Roger Sayle <roger@nextmovesoftware.com>
6000 * config/i386/sse.md (<any_logic>v1ti3): New define_insn to
6001 implement V1TImode AND, IOR and XOR on TARGET_SSE2 (and above).
6002 (one_cmplv1ti2): New define expand.
6004 2021-10-22 Eric Gallager <egallager@gcc.gnu.org>
6007 * Makefile.in: Handle dvidir and install-dvi target.
6008 * configure: Regenerate.
6009 * configure.ac: Add install-dvi to target_list.
6011 2021-10-22 Gerald Pfeifer <gerald@pfeifer.com>
6013 * doc/install.texi (Binaries): Convert mingw-w64.org to https.
6016 2021-10-22 Richard Biener <rguenther@suse.de>
6018 PR tree-optimization/102893
6019 * tree-ssa-dce.c (find_obviously_necessary_stmts): Fix the
6020 test for an exit edge.
6022 2021-10-22 Aldy Hernandez <aldyh@redhat.com>
6023 Andrew MacLeod <amacleod@redhat.com>
6025 * gimple-range-path.cc (path_range_query::compute_phi_relations):
6026 Kill any global relations we may know before registering a new
6028 * value-relation.cc (path_oracle::killing_def): New.
6029 * value-relation.h (path_oracle::killing_def): New.
6031 2021-10-22 Richard Biener <rguenther@suse.de>
6034 * tree-ssa-sccvn.c (vn_phi_insert): For undefined SSA args
6036 (vn_phi_lookup): Likewise.
6038 2021-10-21 H.J. Lu <hjl.tools@gmail.com>
6041 * doc/invoke.texi: Document -fcf-protection requires i686 or
6044 2021-10-21 Eric Botcazou <ebotcazou@adacore.com>
6046 PR middle-end/102764
6047 * cfgexpand.c (expand_gimple_basic_block): Robustify latest change.
6049 2021-10-21 Jonathan Wright <jonathan.wright@arm.com>
6051 * config/aarch64/arm_neon.h (__STRUCTN): Delete function
6052 macro and all invocations.
6054 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
6056 * doc/invoke.texi (ranger-debug): Document.
6057 * flag-types.h (enum ranger_debug): New.
6058 (enum evrp_mode): Remove debug values.
6059 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Use new debug flag.
6060 * gimple-range-gori.cc (gori_compute::gori_compute): Ditto.
6061 * gimple-range.cc (gimple_ranger::gimple_ranger): Ditto.
6062 * gimple-ssa-evrp.c (hybrid_folder::choose_value): Ditto.
6063 (execute_early_vrp): Use evrp-mode directly.
6064 * params.opt (enum evrp_mode): Remove debug values.
6065 (ranger-debug): New.
6066 (ranger-logical-depth): Relocate to be in alphabetical order.
6068 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
6070 * doc/invoke.texi: (vrp1-mode, vrp2-mode): Document.
6071 * flag-types.h: (enum vrp_mode): New.
6072 * params.opt: (vrp1-mode, vrp2-mode): New.
6073 * tree-vrp.c (vrp_pass_num): New.
6074 (pass_vrp::pass_vrp): Set pass number.
6075 (pass_vrp::execute): Choose which VRP mode to execute.
6077 2021-10-21 Andrew MacLeod <amacleod@redhat.com>
6079 * gimple-ssa-evrp.c (class rvrp_folder): Move to tree-vrp.c.
6080 (execute_early_vrp): For ranger only mode, invoke ranger_vrp.
6081 * tree-vrp.c (class rvrp_folder): Relocate here.
6082 (execute_ranger_vrp): New.
6083 * tree-vrp.h (execute_ranger_vrp): Export.
6085 2021-10-21 Martin Liska <mliska@suse.cz>
6089 * opts.c (finish_options): Process flag_var_tracking* options
6090 here as they can be adjusted by optimize attribute.
6091 Process also flag_syntax_only and flag_gtoggle.
6092 * toplev.c (process_options): Remove it here.
6093 * common.opt: Make debug_nonbind_markers_p as PerFunction
6094 attribute as it depends on optimization level.
6096 2021-10-21 Martin Jambor <mjambor@suse.cz>
6098 PR tree-optimization/102505
6099 * tree-sra.c (totally_scalarize_subtree): Check that the
6100 encountered field fits within the acces we would like to put it
6103 2021-10-21 Aldy Hernandez <aldyh@redhat.com>
6105 * tree-ssa-threadbackward.c
6106 (back_threader::maybe_register_path): Remove circular paths check.
6108 2021-10-21 Richard Biener <rguenther@suse.de>
6110 * toplev.c (process_options): Move the initial debug_hooks
6112 (toplev::main): ... before the call of the post_options
6115 2021-10-21 Richard Biener <rguenther@suse.de>
6117 PR tree-optimization/102847
6118 * tree-vect-stmts.c (vect_model_load_cost): Add the scalar
6119 load cost in the prologue for VMAT_INVARIANT.
6121 2021-10-21 Richard Biener <rguenther@suse.de>
6123 PR tree-optimization/102847
6124 * tree-vect-stmts.c (vect_model_load_cost): Explicitely
6125 handle VMAT_INVARIANT as a splat in the prologue.
6127 2021-10-21 Hongyu Wang <hongyu.wang@intel.com>
6130 * config/i386/i386.c (ix86_get_ssemov): Adjust HFmode vector
6131 move to use the same logic as HImode.
6133 2021-10-21 Richard Biener <rguenther@suse.de>
6135 * tree-vect-slp.c (vect_build_slp_tree_1): Remove
6136 superfluous gimple_call_nothrow_p check.
6138 2021-10-21 Andrew Pinski <apinski@marvell.com>
6140 * tree-cfg.c (maybe_remove_writeonly_store): Add dce_ssa_names argument.
6141 Mark the ssa-name of the rhs as one to be removed.
6142 (execute_fixup_cfg): Update call to maybe_remove_writeonly_store.
6143 Call simple_dce_from_worklist at the end to a simple dce.
6145 2021-10-21 Andrew Pinski <apinski@marvell.com>
6147 * tree-cfg.c (maybe_remove_writeonly_store): New function
6148 factored out from ...
6149 (execute_fixup_cfg): Here. Call maybe_remove_writeonly_store.
6151 2021-10-21 Andrew Pinski <apinski@marvell.com>
6153 * tree-cfg.c (execute_fixup_cfg): Remove comment
6154 about standalone pass.
6156 2021-10-21 Andrew Pinski <apinski@marvell.com>
6158 * tree-cfg.c (execute_fixup_cfg): Output when the statement
6159 is removed when it is a write only var.
6161 2021-10-21 Aldy Hernandez <aldyh@redhat.com>
6163 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
6164 Avoid threading circular paths.
6166 2021-10-20 Alex Coplan <alex.coplan@arm.com>
6168 * calls.c (initialize_argument_information): Remove some dead
6169 code, remove handling for function_arg returning const_int.
6170 * doc/tm.texi: Delete documentation for unused target hooks.
6171 * doc/tm.texi.in: Likewise.
6172 * target.def (load_bounds_for_arg): Delete.
6173 (store_bounds_for_arg): Delete.
6174 (load_returned_bounds): Delete.
6175 (store_returned_bounds): Delete.
6176 * targhooks.c (default_load_bounds_for_arg): Delete.
6177 (default_store_bounds_for_arg): Delete.
6178 (default_load_returned_bounds): Delete.
6179 (default_store_returned_bounds): Delete.
6180 * targhooks.h (default_load_bounds_for_arg): Delete.
6181 (default_store_bounds_for_arg): Delete.
6182 (default_load_returned_bounds): Delete.
6183 (default_store_returned_bounds): Delete.
6185 2021-10-20 Jonathan Wakely <jwakely@redhat.com>
6187 * doc/extend.texi (Basic Asm): Clarify that asm is not an
6189 * doc/invoke.texi (-fno-asm): Fix description for C++.
6191 2021-10-20 Jonathan Wakely <jwakely@redhat.com>
6193 * doc/install.texi: Remove link to old.html
6195 2021-10-20 Tamar Christina <tamar.christina@arm.com>
6197 * config/aarch64/aarch64-simd.md (*aarch64_cmtst_same_<mode>): New.
6199 2021-10-20 Tamar Christina <tamar.christina@arm.com>
6201 * config/aarch64/aarch64-simd.md (*aarch64_narrow_trunc<mode>): New.
6203 2021-10-20 Tamar Christina <tamar.christina@arm.com>
6205 * config/aarch64/aarch64-simd.md (aarch64_simd_ashr<mode>): Add case cmp
6207 * config/aarch64/constraints.md (D1): New.
6209 2021-10-20 Tamar Christina <tamar.christina@arm.com>
6211 * config/aarch64/aarch64-simd.md
6212 (*aarch64_<srn_op>topbits_shuffle<mode>_le): New.
6213 (*aarch64_topbits_shuffle<mode>_le): New.
6214 (*aarch64_<srn_op>topbits_shuffle<mode>_be): New.
6215 (*aarch64_topbits_shuffle<mode>_be): New.
6216 * config/aarch64/predicates.md
6217 (aarch64_simd_shift_imm_vec_exact_top): New.
6219 2021-10-20 Tamar Christina <tamar.christina@arm.com>
6221 * config/aarch64/aarch64-simd.md (*aarch64_<srn_op>shrn<mode>_vect,
6222 *aarch64_<srn_op>shrn<mode>2_vect_le,
6223 *aarch64_<srn_op>shrn<mode>2_vect_be): New.
6224 * config/aarch64/iterators.md (srn_op): New.
6226 2021-10-20 Chung-Lin Tang <cltang@codesourcery.com>
6228 * omp-low.c (omp_copy_decl_2): For !ctx, use record_vars to add new copy
6230 (scan_sharing_clauses): Place copy of OMP_CLAUSE_IN_REDUCTION decl in
6231 ctx->outer instead of ctx.
6233 2021-10-20 Martin Liska <mliska@suse.cz>
6236 2021-10-19 Martin Liska <mliska@suse.cz>
6239 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
6240 * system.h (strip_whilespaces): New function.
6242 2021-10-20 Martin Liska <mliska@suse.cz>
6245 2021-10-19 Martin Liska <mliska@suse.cz>
6248 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
6251 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
6253 * config/aarch64/aarch64.c (aarch64_case_values_threshold):
6254 Change to 8 with -Os, 11 otherwise.
6256 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
6258 * config/aarch64/aarch64.c (neoversev1_tunings):
6259 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
6260 (neoversen2_tunings): Likewise.
6262 2021-10-20 Wilco Dijkstra <wdijkstr@arm.com>
6265 * config/aarch64/aarch64.md (frint_pattern): Update comment.
6266 * config/aarch64/aarch64-simd-builtins.def: Change frintn to roundeven.
6267 * config/aarch64/arm_fp16.h: Change frintn to roundeven.
6268 * config/aarch64/arm_neon.h: Likewise.
6269 * config/aarch64/iterators.md (frint_pattern): Use roundeven for FRINTN.
6271 2021-10-20 Martin Liska <mliska@suse.cz>
6273 * config/arm/arm.c (arm_unwind_emit_sequence): Do not declare
6274 already declared global variable.
6275 (arm_unwind_emit_set): Use out_file as function argument.
6276 (arm_unwind_emit): Likewise.
6277 * config/darwin.c (machopic_output_data_section_indirection): Likewise.
6278 (machopic_output_stub_indirection): Likewise.
6279 (machopic_output_indirection): Likewise.
6280 (machopic_finish): Likewise.
6281 * config/i386/i386.c (ix86_asm_output_function_label): Likewise.
6282 * config/i386/winnt.c (i386_pe_seh_unwind_emit): Likewise.
6283 * config/ia64/ia64.c (process_epilogue): Likewise.
6284 (process_cfa_adjust_cfa): Likewise.
6285 (process_cfa_register): Likewise.
6286 (process_cfa_offset): Likewise.
6287 (ia64_asm_unwind_emit): Likewise.
6288 * config/s390/s390.c (s390_asm_output_function_label): Likewise.
6290 2021-10-20 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
6292 * config/aarch64/aarch64-builtins.c
6293 (aarch64_general_gimple_fold_builtin):
6294 lower vld1 and vst1 variants of the neon builtins
6295 * config/aarch64/aarch64-protos.h:
6296 (aarch64_general_gimple_fold_builtin): Add gsi parameter.
6297 * config/aarch64/aarch64.c (aarch64_general_gimple_fold_builtin):
6300 2021-10-20 Andre Simoes Dias Vieira <andre.simoesdiasvieira@arm.com>
6302 * match.pd: Generate IFN_TRUNC.
6304 2021-10-20 Richard Biener <rguenther@suse.de>
6306 PR tree-optimization/102853
6307 * tree-data-ref.c (split_constant_offset_1): Bail out
6308 immediately if the expression traps on overflow.
6310 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
6312 * tree-ssa-threadbackward.c (back_threader::~back_threader): Remove.
6314 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
6316 * tree-ssa-threadupdate.c (back_jt_path_registry::adjust_paths_after_duplication):
6317 Remove superflous debugging message.
6318 (back_jt_path_registry::duplicate_thread_path): Same.
6320 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
6322 * tree-ssa-threadbackward.c (back_threader_registry::back_threader_registry):
6324 (back_threader_registry::register_path): Remove m_threaded_paths.
6326 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
6328 PR tree-optimization/102814
6329 * doc/invoke.texi: Document --param=max-fsm-thread-length.
6330 * params.opt: Add --param=max-fsm-thread-length.
6331 * tree-ssa-threadbackward.c
6332 (back_threader_profitability::profitable_path_p): Fail on paths
6333 longer than max-fsm-thread-length.
6335 2021-10-20 Eric Botcazou <ebotcazou@adacore.com>
6337 PR middle-end/102764
6338 * cfgexpand.c (expand_gimple_basic_block): Disregard a final debug
6339 statement to reset the current location for the outgoing edges.
6341 2021-10-20 Aldy Hernandez <aldyh@redhat.com>
6342 Richard Biener <rguenther@suse.de>
6344 * tree-ssa-threadupdate.c (cancel_thread): Dump threading reason
6345 on the same line as the threading cancellation.
6346 (jt_path_registry::cancel_invalid_paths): Avoid rotating loops.
6347 Avoid threading through loop headers where the path remains in the
6350 2021-10-20 Siddhesh Poyarekar <siddhesh@gotplt.org>
6352 * tree-object-size.c (unknown): Make into a function. Adjust
6354 (unknown_object_size): Simplify implementation.
6356 2021-10-20 Hongtao Liu <hongtao.liu@intel.com>
6357 Kewen Lin <linkw@linux.ibm.com>
6359 * doc/sourcebuild.texi (Effective-Target Keywords): Document
6360 vect_slp_v2qi_store, vect_slp_v4qi_store, vect_slp_v8qi_store,
6361 vect_slp_v16qi_store, vect_slp_v2hi_store,
6362 vect_slp_v4hi_store, vect_slp_v2si_store, vect_slp_v4si_store.
6364 2021-10-19 Jonathan Wakely <jwakely@redhat.com>
6366 * doc/extend.texi (Basic PowerPC Built-in Functions): Fix typo.
6368 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
6372 * config/rs6000/emmintrin.h: Guard POWER8 intrinsics.
6373 * config/rs6000/pmmintrin.h: Same.
6374 * config/rs6000/smmintrin.h: Same.
6375 * config/rs6000/tmmintrin.h: Same.
6377 2021-10-19 Paul A. Clarke <pc@us.ibm.com>
6379 * config.gcc (extra_headers): Add nmmintrin.h.
6381 2021-10-19 Richard Biener <rguenther@suse.de>
6383 * tree-vectorizer.h (vect_supportable_dr_alignment): Add
6384 misalignment parameter.
6385 * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
6386 Do not change DR_MISALIGNMENT in place, instead pass the
6387 adjusted misalignment to vect_supportable_dr_alignment.
6388 (vect_peeling_supportable): Likewise.
6389 (vect_peeling_hash_get_lowest_cost): Adjust.
6390 (vect_enhance_data_refs_alignment): Likewise.
6391 (vect_vfa_access_size): Likewise.
6392 (vect_supportable_dr_alignment): Add misalignment
6393 parameter and simplify.
6394 * tree-vect-stmts.c (get_negative_load_store_type): Adjust.
6395 (get_group_load_store_type): Likewise.
6396 (get_load_store_type): Likewise.
6398 2021-10-19 Clément Chigot <clement.chigot@atos.net>
6400 * config/rs6000/rs6000.c (rs6000_xcoff_file_end): Move
6401 __tls_get_addr reference to .text csect.
6403 2021-10-19 Martin Liska <mliska@suse.cz>
6406 * config/aarch64/aarch64.c (aarch64_process_one_target_attr):
6409 2021-10-19 Richard Biener <rguenther@suse.de>
6411 * tree-vectorizer.h (vect_get_store_cost): Adjust signature.
6412 (vect_get_load_cost): Likewise.
6413 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
6414 alignment support scheme and misalignment as arguments
6416 (vect_get_peeling_costs_all_drs): Compute that info here
6417 and note that we shouldn't need to.
6418 * tree-vect-stmts.c (vect_model_store_cost): Get
6419 alignment support scheme and misalignment as arguments.
6420 (vect_get_store_cost): Likewise.
6421 (vect_model_load_cost): Likewise.
6422 (vect_get_load_cost): Likewise.
6423 (vectorizable_store): Pass down alignment support scheme
6424 and misalignment to costing.
6425 (vectorizable_load): Likewise.
6427 2021-10-19 Richard Biener <rguenther@suse.de>
6429 * tree-vect-stmts.c (get_negative_load_store_type): Add
6430 offset output parameter and initialize it.
6431 (get_group_load_store_type): Likewise.
6432 (get_load_store_type): Likewise.
6433 (vectorizable_store): Use offset as computed by
6434 get_load_store_type.
6435 (vectorizable_load): Likewise.
6437 2021-10-19 Richard Biener <rguenther@suse.de>
6439 PR tree-optimization/102827
6440 * tree-if-conv.c (predicate_statements): Add pe parameter
6441 and use that edge to insert invariant stmts on.
6442 (combine_blocks): Pass through pe.
6443 (tree_if_conversion): Compute the edge to insert invariant
6444 stmts on and pass it along.
6446 2021-10-19 Roger Sayle <roger@nextmovesoftware.com>
6449 * config/bfin/bfin.md (addsubv2hi3, subaddv2hi3, ssaddsubv2hi3,
6450 sssubaddv2hi3): Swap the order of operators in vec_concat.
6452 2021-10-19 Xionghu Luo <luoxhu@linux.ibm.com>
6454 * config/rs6000/altivec.md (*altivec_vmrghb_internal): Delete.
6455 (altivec_vmrghb_direct): New.
6456 (*altivec_vmrghh_internal): Delete.
6457 (altivec_vmrghh_direct): New.
6458 (*altivec_vmrghw_internal): Delete.
6459 (altivec_vmrghw_direct_<mode>): New.
6460 (altivec_vmrghw_direct): Delete.
6461 (*altivec_vmrglb_internal): Delete.
6462 (altivec_vmrglb_direct): New.
6463 (*altivec_vmrglh_internal): Delete.
6464 (altivec_vmrglh_direct): New.
6465 (*altivec_vmrglw_internal): Delete.
6466 (altivec_vmrglw_direct_<mode>): New.
6467 (altivec_vmrglw_direct): Delete.
6468 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
6469 * config/rs6000/rs6000.c (altivec_expand_vec_perm_const):
6471 * config/rs6000/vsx.md (vsx_xxmrghw_<mode>): Adjust.
6472 (vsx_xxmrglw_<mode>): Adjust.
6474 2021-10-19 Aldy Hernandez <aldyh@redhat.com>
6476 * passes.def: Change threading comment before pass_ccp pass.
6478 2021-10-19 Haochen Gui <guihaoc@gcc.gnu.org>
6480 * config/rs6000/rs6000-call.c (altivec_expand_lxvr_builtin):
6481 Modify the expansion for sign extension. All extensions are done
6482 within VSX registers.
6484 2021-10-19 Richard Biener <rguenther@suse.de>
6486 * tree-vect-stmts.c (get_group_load_store_type): Add
6487 misalignment output parameter and initialize it.
6488 (get_group_load_store_type): Likewise.
6489 (vectorizable_store): Remove now redundant queries.
6490 (vectorizable_load): Likewise.
6492 2021-10-19 Richard Biener <rguenther@suse.de>
6494 * tree-vectorizer.h (vect_supportable_dr_alignment): Remove
6495 check_aligned argument.
6496 * tree-vect-data-refs.c (vect_supportable_dr_alignment):
6498 (vect_peeling_hash_insert): Add supportable_if_not_aligned
6499 argument and do not call vect_supportable_dr_alignment here.
6500 (vect_peeling_supportable): Adjust.
6501 (vect_enhance_data_refs_alignment): Compute whether the
6502 access is supported with different alignment here and
6503 pass that down to vect_peeling_hash_insert.
6504 (vect_vfa_access_size): Adjust.
6505 * tree-vect-stmts.c (vect_get_store_cost): Likewise.
6506 (vect_get_load_cost): Likewise.
6507 (get_negative_load_store_type): Likewise.
6508 (get_group_load_store_type): Likewise.
6509 (get_load_store_type): Likewise.
6511 2021-10-19 Martin Liska <mliska@suse.cz>
6514 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p): Strip whitespaces.
6515 * system.h (strip_whilespaces): New function.
6517 2021-10-19 dianhong xu <dianhong.xu@intel.com>
6519 * config/i386/avx512fp16intrin.h:
6520 (_mm512_set1_pch): New intrinsic.
6521 * config/i386/avx512fp16vlintrin.h:
6522 (_mm256_set1_pch): New intrinsic.
6523 (_mm_set1_pch): Ditto.
6525 2021-10-18 Andrew MacLeod <amacleod@redhat.com>
6527 PR tree-optimization/102796
6528 * gimple-range.cc (gimple_ranger::range_on_edge): Process EH edges
6529 normally. Return get_tree_range for non gimple_range_ssa_p names.
6530 (gimple_ranger::range_of_stmt): Use get_tree_range for non
6531 gimple_range_ssa_p names.
6533 2021-10-18 Uroš Bizjak <ubizjak@gmail.com>
6536 * config/i386/i386.c (ix86_print_operand_address):
6537 Error out for non-address_operand asm operands.
6539 2021-10-18 Richard Biener <rguenther@suse.de>
6541 * tree-vect-data-refs.c (vect_peeling_hash_insert): Do
6542 not auto-convert dr_alignment_support to bool.
6543 (vect_peeling_supportable): Likewise.
6544 (vect_enhance_data_refs_alignment): Likewise.
6545 (vect_supportable_dr_alignment): Commonize read/write case.
6546 * tree-vect-stmts.c (vect_get_store_cost): Use
6547 dr_alignment_support, not int, for the vect_supportable_dr_alignment
6549 (vect_get_load_cost): Likewise.
6551 2021-10-18 Siddhesh Poyarekar <siddhesh@gotplt.org>
6553 * tree-object-size.c (object_sizes_execute): Consolidate LHS
6554 null check and do it early.
6556 2021-10-18 Richard Biener <rguenther@suse.de>
6558 * tree-vect-stmts.c (vectorizable_store): Use the
6559 computed alignment scheme instead of querying
6562 2021-10-18 Richard Biener <rguenther@suse.de>
6564 * tree-vect-stmts.c (vectorizable_store): Do not recompute
6565 alignment scheme already determined by get_load_store_type.
6567 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
6569 * tree-ssa-threadbackward.c (class pass_thread_jumps_full):
6570 Clone corresponding pass.
6572 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
6574 * combine.c (recog_for_combine): For an unrecognized move/set of
6575 a constant, try force_const_mem to place it in the constant pool.
6576 * cse.c (constant_pool_entries_cost, constant_pool_entries_regcost):
6577 Delete global variables (that are no longer assigned a cost value).
6578 (cse_insn): Simplify logic for deciding whether to place a folded
6579 constant in the constant pool using force_const_mem.
6580 (cse_main): Remove zero initialization of constant_pool_entries_cost
6581 and constant_pool_entries_regcost.
6582 * config/i386/i386.c (ix86_rtx_costs): Make memory accesses
6583 fractionally more expensive, when optimizing for speed.
6585 2021-10-18 Martin Liska <mliska@suse.cz>
6587 PR gcov-profile/102746
6588 PR gcov-profile/102747
6589 * gcov.c (main): Return return_code.
6590 (output_gcov_file): Mark return_code when error happens.
6591 (generate_results): Likewise.
6592 (read_graph_file): Likewise.
6593 (read_count_file): Likewise.
6595 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
6597 * config/bfin/bfin.md (define_constants): Remove UNSPEC_ONES.
6598 (define_insn "ones"): Replace UNSPEC_ONES with a truncate of
6599 a popcount, allowing compile-time evaluation/simplification.
6600 (popcountsi2, popcounthi2): New expanders using a "ones" insn.
6602 2021-10-18 Richard Biener <rguenther@suse.de>
6604 PR tree-optimization/102788
6605 * tree-vect-patterns.c (vect_init_pattern_stmt): Allow
6607 (vect_pattern_recog_1): Likewise.
6608 (vect_recog_bool_pattern): Continue matching the pattern
6609 even if we do not have a vector type for a conversion
6612 2021-10-18 Roger Sayle <roger@nextmovesoftware.com>
6614 * simplify-rtx.c (simplify_const_unary_operation) [SS_NEG, SS_ABS]:
6615 Evalute SS_NEG and SS_ABS of a constant argument.
6617 2021-10-18 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
6620 * gimple-match-head.c (try_conditional_simplification): Add case for single operand.
6621 * internal-fn.def: Add entry for COND_NEG internal function.
6622 * internal-fn.c (FOR_EACH_CODE_MAPPING): Add entry for
6623 NEGATE_EXPR, COND_NEG mapping.
6624 * optabs.def: Add entry for cond_neg_optab.
6625 * match.pd (UNCOND_UNARY, COND_UNARY): New operator lists.
6626 (vec_cond COND (foo A) B) -> (IFN_COND_FOO COND A B): New pattern.
6627 (vec_cond COND B (foo A)) -> (IFN_COND_FOO ~COND A B): Likewise.
6629 2021-10-18 Aldy Hernandez <aldyh@redhat.com>
6631 * tree-ssa-strlen.c (class strlen_pass): Rename from
6633 (handle_builtin_stxncpy_strncat): Move to strlen_pass.
6634 (handle_assign): Same.
6635 (adjust_last_stmt): Same.
6636 (maybe_warn_overflow): Same.
6637 (handle_builtin_strlen): Same.
6638 (handle_builtin_strchr): Same.
6639 (handle_builtin_strcpy): Same.
6640 (handle_builtin_strncat): Same.
6641 (handle_builtin_stxncpy_strncat): Same.
6642 (handle_builtin_memcpy): Same.
6643 (handle_builtin_strcat): Same.
6644 (handle_alloc_call): Same.
6645 (handle_builtin_memset): Same.
6646 (handle_builtin_memcmp): Same.
6647 (get_len_or_size): Same.
6648 (strxcmp_eqz_result): Same.
6649 (handle_builtin_string_cmp): Same.
6650 (handle_pointer_plus): Same.
6651 (count_nonzero_bytes_addr): Same.
6652 (count_nonzero_bytes): Same.
6653 (handle_store): Same.
6654 (strlen_check_and_optimize_call): Same.
6655 (handle_integral_assign): Same.
6656 (check_and_optimize_stmt): Same.
6657 (printf_strlen_execute): Rename strlen_dom_walker to strlen_pass.
6659 2021-10-18 Richard Biener <rguenther@suse.de>
6661 PR tree-optimization/102798
6662 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref):
6663 Only copy points-to info to newly generated SSA names.
6665 2021-10-18 Martin Liska <mliska@suse.cz>
6667 * dbgcnt.c (dbg_cnt_process_opt): Remove unused but set variable.
6668 * gcov.c (get_cycles_count): Likewise.
6669 * lto-compress.c (lto_compression_zlib): Likewise.
6670 (lto_uncompression_zlib): Likewise.
6671 * targhooks.c (default_pch_valid_p): Likewise.
6673 2021-10-17 Aldy Hernandez <aldyh@redhat.com>
6675 * tree-pass.h (make_pass_thread_jumps_full): New.
6676 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Inline.
6677 (try_thread_blocks): Add resolve and speed arguments.
6678 (pass_thread_jumps::execute): Inline.
6679 (do_early_thread_jumps): New.
6680 (do_thread_jumps): New.
6681 (make_pass_thread_jumps): Move.
6682 (pass_early_thread_jumps::gate): Inline.
6683 (pass_early_thread_jumps::execute): Inline.
6684 (class pass_thread_jumps_full): New.
6686 2021-10-16 Piotr Kubaj <pkubaj@FreeBSD.org>
6688 * configure.ac: Treat powerpc64*-*-freebsd* the same as
6689 powerpc64-*-freebsd*.
6690 * configure: Regenerate.
6692 2021-10-16 H.J. Lu <hjl.tools@gmail.com>
6694 * value-query.cc (get_ssa_name_ptr_info_nonnull): Change
6695 set_ptr_nonull to set_ptr_nonnull in comments.
6697 2021-10-16 Jan Hubicka <hubicka@ucw.cz>
6699 PR tree-optimization/102720
6700 * tree-ssa-structalias.c (compute_points_to_sets): Fix producing
6701 of call used and clobbered sets.
6703 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
6705 * config/darwin.h (LINK_COMMAND_SPEC_A): Update 'r' handling to
6706 skip gomp and itm when r or nodefaultlibs is given.
6707 (DSYMUTIL_SPEC): Do not call dsymutil for '-r' link lines.
6708 Update ordering of exclusions, remove duplicate 'v' addition
6709 (collect2 will add this from the main command line).
6711 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
6713 * config/darwin-driver.c (darwin_driver_init): Revise comments, handle
6714 filelist and framework options in specs instead of code.
6715 * config/darwin.h (SUBTARGET_DRIVER_SELF_SPECS): Update to handle link
6716 specs that are really driver ones.
6717 (DARWIN_CC1_SPEC): Likewise.
6718 (CPP_SPEC): Likewise.
6719 (SYSROOT_SPEC): Append space.
6720 (LINK_SYSROOT_SPEC): Remove most driver link specs.
6721 (STANDARD_STARTFILE_PREFIX_2): Update link-related specs.
6722 (STARTFILE_SPEC): Likewise.
6723 (ASM_MMACOSX_VERSION_MIN_SPEC): Fix line wrap.
6724 (ASM_SPEC): Update driver-related specs.
6725 (ASM_FINAL_SPEC): Likewise.
6726 * config/darwin.opt: Remove now unused option aliases.
6727 * config/i386/darwin.h (EXTRA_ASM_OPTS): Ensure space after opt.
6728 (ASM_SPEC): Update driver-related specs.
6730 2021-10-15 Roger Sayle <roger@nextmovesoftware.com>
6732 * config/i386/i386.c (ix86_hardreg_mov_ok): For vector modes,
6733 allow standard_sse_constant_p immediate constants.
6735 2021-10-15 Iain Sandoe <iain@sandoe.co.uk>
6737 * config.gcc: Add tm-dwarf2.h to tm_d-file.
6739 2021-10-15 Andrew MacLeod <amacleod@redhat.com>
6741 * gimple-range-fold.h (gimple_range_ssa_p): Don't process names
6742 that occur in abnormal phis.
6743 * gimple-range.cc (gimple_ranger::range_on_edge): Return false for
6744 abnormal and EH edges.
6745 * gimple-ssa-evrp.c (rvrp_folder::value_of_expr): Ditto.
6746 (rvrp_folder::value_on_edge): Ditto.
6747 (rvrp_folder::value_of_stmt): Ditto.
6748 (hybrid_folder::value_of_expr): Ditto for ranger queries.
6749 (hybrid_folder::value_on_edge): Ditto.
6750 (hybrid_folder::value_of_stmt): Ditto.
6751 * value-query.cc (gimple_range_global): Always return a range if
6752 the type is supported.
6754 2021-10-15 John David Anglin <danglin@gcc.gnu.org>
6756 * config/pa/pa.md: Consistently use "rG" constraint for copy
6757 instruction in move patterns.
6759 2021-10-15 Andrew Stubbs <ams@codesourcery.com>
6761 * config/gcn/gcn-hsa.h (S_FIJI): Set unconditionally.
6764 * config/gcn/gcn.c: Hard code SRAM ECC settings for old architectures.
6765 * config/gcn/mkoffload.c (ELFABIVERSION_AMDGPU_HSA): Rename to ...
6766 (ELFABIVERSION_AMDGPU_HSA_V3): ... this.
6767 (ELFABIVERSION_AMDGPU_HSA_V4): New.
6768 (SET_SRAM_ECC_UNSUPPORTED): New.
6769 (copy_early_debug_info): Create elf flags to match the other objects.
6770 (main): Just let the attribute flags pass through.
6772 2021-10-15 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
6774 * tree-loop-distribution.c (reduction_var_overflows_first):
6775 Pass the type of reduction_var as first argument as it is also
6776 done for the load type.
6777 (loop_distribution::transform_reduction_loop): Add missing
6778 TREE_TYPE while determining precission of reduction_var.
6780 2021-10-15 Richard Biener <rguenther@suse.de>
6782 * defaults.h (PREFERRED_DEBUGGING_TYPE): Choose DWARF2_DEBUG
6784 * toplev.c (process_options): Warn when STABS debugging is
6785 enabled but not the preferred format.
6786 * config/pa/som.h (PREFERRED_DEBUGGING_TYPE): Define to
6788 * config/pdp11/pdp11.h (PREFERRED_DEBUGGING_TYPE): Likewise.
6790 2021-10-15 Richard Biener <rguenther@suse.de>
6793 * tree-inline.c (copy_bb): Avoid underflowing nargs.
6795 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
6797 * config/i386/i386-expand.c (ix86_expand_vec_perm): Convert
6798 HFmode input operand to HImode.
6799 (ix86_vectorize_vec_perm_const): Likewise.
6800 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1_hf):
6802 (*avx512f_permvar_truncv8siv8hi_1_hf):
6805 2021-10-15 Richard Biener <rguenther@suse.de>
6807 PR middle-end/102682
6808 * expmed.c (store_bit_field_1): Ensure a LHS subreg would
6809 not create a paradoxical subreg.
6811 2021-10-15 Hongyu Wang <hongyu.wang@intel.com>
6813 * config/i386/i386-expand.c (ix86_expand_vector_init):
6814 For half_vector concat for HFmode, handle them like HImode.
6816 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
6818 * tree-ssa-threadbackward.c (class back_threader): Add m_resolve.
6819 (back_threader::back_threader): Same.
6820 (back_threader::resolve_phi): Try to solve without looking back if
6822 (back_threader::find_paths_to_names): Same.
6823 (try_thread_blocks): Pass resolve argument to back threader.
6824 (pass_early_thread_jumps::execute): Same.
6826 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
6828 * doc/invoke.texi: Remove max-fsm-thread-length,
6829 max-fsm-thread-paths, and fsm-maximum-phi-arguments.
6831 * tree-ssa-threadbackward.c (back_threader::back_threader): Remove
6833 (back_threader_registry::back_threader_registry): Same.
6834 (back_threader_profitability::profitable_path_p): Remove
6835 param_max_fsm_thread-length.
6836 (back_threader_registry::register_path): Remove
6837 m_max_allowable_paths.
6839 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
6841 * tree-ssa-threadbackward.c (class back_threader): Make m_imports
6843 (back_threader::~back_threader): Do not release m_path.
6845 2021-10-14 Andrew MacLeod <amacleod@redhat.com>
6847 PR tree-optimization/102738
6848 * vr-values.c (simplify_using_ranges::simplify): Handle RSHIFT_EXPR.
6850 2021-10-14 Kwok Cheung Yeung <kcy@codesourcery.com>
6852 * omp-general.c (omp_check_context_selector): Move from c-omp.c.
6853 (omp_mark_declare_variant): Move from c-omp.c.
6854 (omp_context_name_list_prop): Update for Fortran strings.
6855 * omp-general.h (omp_check_context_selector): New prototype.
6856 (omp_mark_declare_variant): New prototype.
6858 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
6860 * config/rs6000/rs6000.c (rs6000_density_test): Move early
6861 exit test further up the function.
6863 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
6865 * config/arm/arm.c (arm_add_stmt_cost): Delete.
6866 (TARGET_VECTORIZE_ADD_STMT_COST): Delete.
6868 2021-10-14 Martin Jambor <mjambor@suse.cz>
6870 * doc/invoke.texi (Optimize Options): Add entry for
6871 ipa-cp-recursive-freq-factor.
6873 2021-10-14 Tamar Christina <tamar.christina@arm.com>
6875 * match.pd: New rule.
6877 2021-10-14 Jan Hubicka <hubicka@ucw.cz>
6880 * ipa-modref-tree.h (modref_access_node::update2):
6881 Also check that parm_offset is unchanged.
6882 (modref_ref_node::insert_access): Fix updating of
6885 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
6887 * tree-ssa-threadbackward.c (back_threader::resolve_phi): Add
6890 2021-10-14 Richard Biener <rguenther@suse.de>
6892 PR tree-optimization/102659
6893 * tree-if-conv.c (if_convertible_gimple_assign_stmt_p): Also
6894 rewrite pointer typed undefined overflow operations.
6895 (predicate_statements): Likewise. Make sure to emit invariant
6896 conversions in the preheader.
6897 * tree-vectorizer.c (vect_loop_vectorized_call): Look through
6898 non-empty preheaders.
6899 * tree-data-ref.c (dr_analyze_indices): Strip useless
6900 conversions to the MEM_REF base type.
6902 2021-10-14 Martin Liska <mliska@suse.cz>
6904 * common.opt: Stop using AUTODETECT_VALUE
6905 and use EnabledBy where possible.
6906 * opts.c: Enable OPT_fvar_tracking with optimize >= 1.
6907 * toplev.c (AUTODETECT_VALUE): Remove macro.
6908 (process_options): Simplify by using EnabledBy and
6909 OPT_fvar_tracking. Use OPTION_SET_P macro instead of
6912 2021-10-14 Jonathan Wright <jonathan.wright@arm.com>
6914 * config/aarch64/arm_neon.h (vld1_s8_x3): Use signed type for
6916 (vld1_s32_x3): Likewise.
6918 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
6920 PR tree-optimization/102736
6921 PR tree-optimization/102736
6922 * gimple-range-path.cc (path_range_query::range_on_path_entry):
6923 Assert that the requested range is defined outside the path.
6924 (path_range_query::ssa_range_in_phi): Do not call
6925 range_on_path_entry for SSA names that are defined within the
6928 2021-10-14 Iain Sandoe <iain@sandoe.co.uk>
6930 * config/darwin-driver.c (darwin_find_version_from_kernel):
6931 Quote internal identifiers and avoid contractions in
6933 (darwin_default_min_version): Likewise.
6934 (darwin_driver_init): Likewise.
6936 2021-10-14 Martin Jambor <mjambor@suse.cz>
6938 * params.opt (ipa-cp-recursive-freq-factor): New.
6939 * ipa-cp.c (ipcp_value): Switch to inline initialization. New members
6940 scc_no, self_recursion_generated_level, same_scc and
6941 self_recursion_generated_p.
6942 (ipcp_lattice::add_value): Replaced parameter unlimited with
6943 same_lat_gen_level, usit it determine limit of values and store it to
6945 (ipcp_lattice<valtype>::print): Dump the new fileds.
6946 (allocate_and_init_ipcp_value): Take same_lat_gen_level as a new
6947 parameter and store it to the new value.
6948 (self_recursively_generated_p): Removed.
6949 (propagate_vals_across_arith_jfunc): Use self_recursion_generated_p
6950 instead of self_recursively_generated_p, store self generation level
6952 (value_topo_info<valtype>::add_val): Set scc_no.
6953 (value_topo_info<valtype>::propagate_effects): Multiply frequencies of
6954 recursively feeding values and self generated values by appropriate
6957 2021-10-14 Richard Sandiford <richard.sandiford@arm.com>
6959 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Remove
6960 redundant test for flag_vect_cost_model.
6962 2021-10-14 Aldy Hernandez <aldyh@redhat.com>
6964 * bitmap.c (debug): New overloaded function for auto_bitmaps.
6965 * bitmap.h (debug): Same.
6967 2021-10-14 Hongyu Wang <hongyu.wang@intel.com>
6969 * config/i386/avx512fp16intrin.h (_mm512_mask_fcmadd_pch):
6970 Adjust builtin call.
6971 (_mm512_mask3_fcmadd_pch): Likewise.
6972 (_mm512_mask_fmadd_pch): Likewise
6973 (_mm512_mask3_fmadd_pch): Likewise
6974 (_mm512_mask_fcmadd_round_pch): Likewise
6975 (_mm512_mask3_fcmadd_round_pch): Likewise
6976 (_mm512_mask_fmadd_round_pch): Likewise
6977 (_mm512_mask3_fmadd_round_pch): Likewise
6978 (_mm_mask_fcmadd_sch): Likewise
6979 (_mm_mask3_fcmadd_sch): Likewise
6980 (_mm_mask_fmadd_sch): Likewise
6981 (_mm_mask3_fmadd_sch): Likewise
6982 (_mm_mask_fcmadd_round_sch): Likewise
6983 (_mm_mask3_fcmadd_round_sch): Likewise
6984 (_mm_mask_fmadd_round_sch): Likewise
6985 (_mm_mask3_fmadd_round_sch): Likewise
6986 (_mm_fcmadd_round_sch): Likewise
6987 * config/i386/avx512fp16vlintrin.h (_mm_mask_fmadd_pch):
6988 Adjust builtin call.
6989 (_mm_mask3_fmadd_pch): Likewise
6990 (_mm256_mask_fmadd_pch): Likewise
6991 (_mm256_mask3_fmadd_pch): Likewise
6992 (_mm_mask_fcmadd_pch): Likewise
6993 (_mm_mask3_fcmadd_pch): Likewise
6994 (_mm256_mask_fcmadd_pch): Likewise
6995 (_mm256_mask3_fcmadd_pch): Likewise
6996 * config/i386/i386-builtin.def: Add mask3 builtin for complex
6997 fma, and adjust mask_builtin to corresponding expander.
6998 * config/i386/i386-expand.c (ix86_expand_round_builtin):
6999 Skip eraseing embedded rounding for expanders that emits
7001 * config/i386/sse.md (complexmove): New mode_attr.
7002 (<avx512>_fmaddc_<mode>_mask1<round_expand_name>): New expander.
7003 (<avx512>_fcmaddc_<mode>_mask1<round_expand_name>): Likewise.
7004 (avx512fp16_fmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
7005 (avx512fp16_fcmaddcsh_v8hf_mask1<round_expand_name>): Likewise.
7006 (avx512fp16_fcmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
7007 (avx512fp16_fmaddcsh_v8hf_mask3<round_expand_name>): Likewise.
7008 * config/i386/subst.md (round_embedded_complex): New subst.
7010 2021-10-14 John David Anglin <danglin@gcc.gnu.org>
7012 * config/pa/pa.md (cbranchsf4): Disable if TARGET_SOFT_FLOAT.
7013 (cbranchdf4): Likewise.
7014 Add missing move patterns for TARGET_SOFT_FLOAT.
7016 2021-10-13 Roger Sayle <roger@nextmovesoftware.com>
7018 * config/i386/i386-expand.c (ix86_expand_vector_move): Use a
7019 pseudo intermediate when moving a SUBREG into a hard register,
7020 by checking ix86_hardreg_mov_ok.
7021 (ix86_expand_vector_extract): Store zero-extended SImode
7022 intermediate in a pseudo, then set target using a SUBREG_PROMOTED
7024 * config/i386/sse.md (mov<VMOVE>_internal): Prevent CSE creating
7025 complex (SUBREG) sets of (vector) hard registers before reload, by
7026 checking ix86_hardreg_mov_ok.
7028 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
7030 * ctfc.h (enum ctf_dtu_d_union_enum): Remove redundant comma.
7032 2021-10-13 Indu Bhagat <indu.bhagat@oracle.com>
7034 * dwarf2ctf.c (gen_ctf_array_type): Fix typo in comment.
7036 2021-10-13 Martin Sebor <msebor@redhat.com>
7038 PR middle-end/102630
7039 * pointer-query.cc (compute_objsize_r): Handle named address spaces.
7041 2021-10-13 Iain Sandoe <iain@sandoe.co.uk>
7043 * collect2.c (is_lto_object_file): Release simple-object
7044 resources, close files.
7046 2021-10-13 Andre Vieira <andre.simoesdiasvieira@arm.com>
7048 * config/arm/arm.c (thumb2_legitimate_address_p): Use VALID_MVE_MODE
7049 when checking mve addressing modes.
7050 (mve_vector_mem_operand): Fix the way we handle pre, post and offset
7052 (arm_print_operand): Fix printing of POST_ and PRE_MODIFY.
7053 * config/arm/mve.md: Use mve_memory_operand predicate everywhere where
7054 there is a single Ux constraint.
7056 2021-10-13 John David Anglin <danglin@gcc.gnu.org>
7058 * config/pa/pa.md (muldi3): Add support for inlining 64-bit
7059 multiplication on 32-bit PA 1.1 and 2.0 targets.
7061 2021-10-13 Martin Liska <mliska@suse.cz>
7063 PR gcov-profile/90364
7064 * coverage.c (build_info): Emit checksum to the global variable.
7065 (build_info_type): Add new field for checksum.
7066 (coverage_obj_finish): Pass object_checksum.
7067 (coverage_init): Use 0 as checksum for .gcno files.
7068 * gcov-dump.c (dump_gcov_file): Dump also new checksum field.
7069 * gcov.c (read_graph_file): Read also checksum.
7070 * doc/invoke.texi: Document the behaviour change.
7072 2021-10-13 Richard Biener <rguenther@suse.de>
7074 * gimple-iterator.h (gsi_iterator_update): Add GSI_LAST_NEW_STMT,
7075 start at integer value 2.
7076 * gimple-iterator.c (gsi_insert_seq_nodes_before): Update
7077 the iterator for GSI_LAST_NEW_STMT.
7078 (gsi_insert_seq_nodes_after): Likewise.
7079 * tree-if-conv.c (predicate_statements): Use GSI_LAST_NEW_STMT.
7080 * tree-ssa.c (execute_update_addresses_taken): Correct bogus
7081 arguments to gsi_replace.
7083 2021-10-13 Martin Liska <mliska@suse.cz>
7086 * common.opt: Use EnabledBy instead of detection in
7087 finish_options and process_options.
7088 * opts.c (finish_options): Remove handling of
7089 x_flag_unroll_all_loops.
7090 * toplev.c (process_options): Likewise for flag_web and
7091 flag_rename_registers.
7093 2021-10-13 Richard Biener <rguenther@suse.de>
7095 PR tree-optimization/102659
7096 * tree-if-conv.c (need_to_rewrite_undefined): New flag.
7097 (if_convertible_gimple_assign_stmt_p): Mark the loop for
7098 rewrite when stmts with undefined behavior on integer
7100 (combine_blocks): Predicate also when we need to rewrite stmts.
7101 (predicate_statements): Rewrite affected stmts to something
7102 with well-defined behavior on overflow.
7103 (tree_if_conversion): Initialize need_to_rewrite_undefined.
7105 2021-10-13 Richard Biener <rguenther@suse.de>
7108 * ipa-sra.c (ptr_parm_has_nonarg_uses): Fix volatileness
7111 2021-10-13 Jose E. Marchesi <jose.marchesi@oracle.com>
7113 * dwarf2ctf.c: Fix typo in comment.
7115 2021-10-12 Bill Schmidt <wschmidt@linux.ibm.com>
7118 * config/rs6000/altivec.h (vec_cpsgn): Swap operand order.
7119 * config/rs6000/rs6000-overload.def (VEC_COPYSIGN): Use SKIP to
7120 avoid generating an automatic #define of vec_cpsgn. Use the
7121 correct built-in for V4SFmode that doesn't depend on VSX.
7123 2021-10-12 Uroš Bizjak <ubizjak@gmail.com>
7127 * config/i386/i386.md (*add<mode>_1_slp): Rewrite as
7128 define_insn_and_split pattern. Add alternative 1 and split it
7129 post reload to insert operand 1 into the low part of operand 0.
7130 (*sub<mode>_1_slp): Ditto.
7131 (*and<mode>_1_slp): Ditto.
7132 (*<any_or:code><mode>_1_slp): Ditto.
7133 (*ashl<mode>3_1_slp): Ditto.
7134 (*<any_shiftrt:insn><mode>3_1_slp): Ditto.
7135 (*<any_rotate:insn><mode>3_1_slp): Ditto.
7136 (*neg<mode>_1_slp): New insn_and_split pattern.
7137 (*one_cmpl<mode>_1_slp): Ditto.
7139 2021-10-12 David Edelsohn <dje.gcc@gmail.com>
7141 * doc/install.texi: Update MinGW and mingw-64 Binaries
7144 2021-10-12 Daniel Le Duc Khoi Nguyen <greenrecyclebin@gmail.com>
7146 * doc/extend.texi (Common Variable Attributes): Fix typos in
7147 alloc_size documentation.
7149 2021-10-12 Richard Biener <rguenther@suse.de>
7151 PR tree-optimization/102696
7152 * tree-vect-slp.c (vect_build_slp_tree_2): Properly mark
7153 the tree fatally failed when we reject a BIT_FIELD_REF.
7155 2021-10-12 Richard Biener <rguenther@suse.de>
7157 PR tree-optimization/102572
7158 * tree-vect-stmts.c (vect_build_gather_load_calls): When
7159 gathering the vectorized defs for the mask pass in the
7160 desired mask vector type so invariants will be handled
7163 2021-10-12 Tamar Christina <tamar.christina@arm.com>
7165 * config/aarch64/aarch64-sve.md (*fcm<cmp_op><mode>_bic_combine,
7166 *fcm<cmp_op><mode>_nor_combine, *fcmuo<mode>_bic_combine,
7167 *fcmuo<mode>_nor_combine): New.
7169 2021-10-12 Eric Botcazou <ebotcazou@adacore.com>
7172 * config/sparc/sparc-modes.def (OI): New integer mode.
7174 2021-10-12 Jakub Jelinek <jakub@redhat.com>
7176 * gimple-fold.h (clear_padding_type_may_have_padding_p): Declare.
7177 * gimple-fold.c (clear_padding_type_may_have_padding_p): No longer
7180 2021-10-12 Jakub Jelinek <jakub@redhat.com>
7182 * tree-vectorizer.h (loop_cost_model): New function.
7183 (unlimited_cost_model): Use it.
7184 * tree-vect-loop.c (vect_analyze_loop_costing): Use loop_cost_model
7185 call instead of flag_vect_cost_model.
7186 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7187 (vect_prune_runtime_alias_test_list): Likewise. Also use it instead
7188 of flag_simd_cost_model.
7190 2021-10-12 liuhongt <hongtao.liu@intel.com>
7193 * config/i386/i386-expand.c (emit_reduc_half): Handle
7195 * config/i386/mmx.md (reduc_<code>_scal_v4qi): New expander.
7196 (reduc_plus_scal_v4qi): Ditto.
7198 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
7200 * config/rs6000/smmintrin.h (_mm_cmpeq_epi64, _mm_cmpgt_epi64,
7201 _mm_mullo_epi32, _mm_mul_epi32, _mm_packus_epi32): New.
7202 * config/rs6000/nmmintrin.h: Copy from i386, tweak to suit.
7204 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
7206 * config/rs6000/smmintrin.h (_mm_cvtepi8_epi16, _mm_cvtepi8_epi32,
7207 _mm_cvtepi8_epi64, _mm_cvtepi16_epi32, _mm_cvtepi16_epi64,
7208 _mm_cvtepi32_epi64, _mm_cvtepu8_epi16, _mm_cvtepu8_epi32,
7209 _mm_cvtepu8_epi64, _mm_cvtepu16_epi32, _mm_cvtepu16_epi64,
7210 _mm_cvtepu32_epi64): New.
7212 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
7214 * config/rs6000/smmintrin.h (_mm_test_all_zeros,
7215 _mm_test_all_ones, _mm_test_mix_ones_zeros): Rewrite as macro.
7217 2021-10-12 Paul A. Clarke <pc@us.ibm.com>
7219 * config/rs6000/smmintrin.h (_mm_min_epi8, _mm_min_epu16,
7220 _mm_min_epi32, _mm_min_epu32, _mm_max_epi8, _mm_max_epu16,
7221 _mm_max_epi32, _mm_max_epu32): New.
7223 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
7225 * ipa-modref-tree.h (struct modref_access_node): Revert
7227 (struct modref_ref_node): Likewise.
7229 2021-10-11 Jan Hubicka <hubicka@ucw.cz>
7231 * ipa-modref-tree.h (modref_tree::global_access_p): New member
7234 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
7235 ignore_stores_eaf_flags): Move to ipa-modref.h
7236 (remove_useless_eaf_flags): Remove early exit on NOCLOBBER.
7237 (modref_summary::global_memory_read_p): New member function.
7238 (modref_summary::global_memory_written_p): New member function.
7239 * ipa-modref.h (modref_summary::global_memory_read_p,
7240 modref_summary::global_memory_written_p): Declare.
7241 (implicint_const_eaf_flags,implicit_pure_eaf_flags,
7242 ignore_stores_eaf_flags): move here.
7243 * tree-ssa-structalias.c: Include ipa-modref-tree.h, ipa-modref.h
7245 (handle_rhs_call): Rewrite.
7246 (handle_call_arg): New function.
7247 (determine_global_memory_access): New function.
7248 (handle_const_call): Remove
7249 (handle_pure_call): Remove
7250 (find_func_aliases_for_call): Update use of handle_rhs_call.
7251 (compute_points_to_sets): Handle global memory acccesses
7254 2021-10-11 Diane Meirowitz <diane.meirowitz@oracle.com>
7256 * doc/invoke.texi: Add link to UndefinedBehaviorSanitizer
7257 documentation, mention UBSAN_OPTIONS, similar to what is done
7258 for AddressSanitizer.
7260 2021-10-11 Richard Biener <rguenther@suse.de>
7262 PR middle-end/102683
7263 * internal-fn.c (expand_DEFERRED_INIT): Check for mode
7264 availability before building an integer type for storage
7267 2021-10-11 Richard Biener <rguenther@suse.de>
7269 PR middle-end/101480
7270 * gimple.c (gimple_call_fnspec): Do not mark operator new/delete
7273 2021-10-11 Martin Liska <mliska@suse.cz>
7275 * common.opt: Remove Init(2) for some options.
7276 * toplev.c (process_options): Do not use AUTODETECT_VALUE, but
7277 use rather OPTION_SET_P.
7279 2021-10-11 Martin Liska <mliska@suse.cz>
7281 * common.opt: Remove usage of IRA_REGION_AUTODETECT.
7282 * flag-types.h (enum ira_region): Likewise.
7283 * toplev.c (process_options): Use OPTION_SET_P instead of
7284 IRA_REGION_AUTODETECT.
7286 2021-10-11 Jakub Jelinek <jakub@redhat.com>
7288 * omp-low.c (omp_runtime_api_call): Handle omp_get_max_teams,
7289 omp_[sg]et_teams_thread_limit and omp_set_num_teams.
7291 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
7293 * config/s390/s390-protos.h (s390_rawmemchr): Add prototype.
7294 * config/s390/s390.c (s390_rawmemchr): New function.
7295 * config/s390/s390.md (rawmemchr<SINT:mode>): New expander.
7296 * config/s390/vector.md (@vec_vfees<mode>): Basically a copy of
7297 the pattern vfees<mode> from vx-builtins.md.
7298 * config/s390/vx-builtins.md (*vfees<mode>): Remove.
7300 2021-10-11 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
7302 * builtins.c (get_memory_rtx): Change to external linkage.
7303 * builtins.h (get_memory_rtx): Add function prototype.
7304 * doc/md.texi (rawmemchr<mode>): Document.
7305 * internal-fn.c (expand_RAWMEMCHR): Define.
7306 * internal-fn.def (RAWMEMCHR): Add.
7307 * optabs.def (rawmemchr_optab): Add.
7308 * tree-loop-distribution.c (find_single_drs): Change return code
7309 behaviour by also returning true if no single store was found
7311 (loop_distribution::classify_partition): Respect the new return
7312 code behaviour of function find_single_drs.
7313 (loop_distribution::execute): Call new function
7314 transform_reduction_loop in order to replace rawmemchr or strlen
7315 like loops by calls into builtins.
7316 (generate_reduction_builtin_1): New function.
7317 (generate_rawmemchr_builtin): New function.
7318 (generate_strlen_builtin_1): New function.
7319 (generate_strlen_builtin): New function.
7320 (generate_strlen_builtin_using_rawmemchr): New function.
7321 (reduction_var_overflows_first): New function.
7322 (determine_reduction_stmt_1): New function.
7323 (determine_reduction_stmt): New function.
7324 (loop_distribution::transform_reduction_loop): New function.
7326 2021-10-11 Martin Liska <mliska@suse.cz>
7328 * tree.c (cl_option_hasher::hash): Use cl_optimization_hash
7329 and remove legacy hashing code.
7331 2021-10-11 Kito Cheng <kito.cheng@sifive.com>
7334 * builtins.c (maybe_emit_call_builtin___clear_cache): Allow
7335 CONST_INT for BEGIN and END, and use gcc_assert rather than
7338 2021-10-10 Jakub Jelinek <jakub@redhat.com>
7341 * var-tracking.c (add_stores): For cselib_sp_derived_value_p values
7342 use MO_VAL_SET if loc is not sp.
7344 2021-10-10 Andrew Pinski <apinski@marvell.com>
7346 PR tree-optimization/102622
7347 * match.pd: Swap the order of a?pow2cst:0 and a?-1:0 transformations.
7348 Swap the order of a?0:pow2cst and a?0:-1 transformations.
7350 2021-10-09 liuhongt <hongtao.liu@intel.com>
7353 * config/i386/i386-expand.c (ix86_valid_mask_cmp_mode): Handle
7355 (ix86_use_mask_cmp_p): Ditto.
7356 (ix86_expand_sse_movcc): Ditto.
7357 * config/i386/i386.md (setcc_hf_mask): New define_insn.
7358 (movhf_mask): Ditto.
7359 (UNSPEC_MOVCC_MASK): New unspec.
7360 * config/i386/sse.md (UNSPEC_PCMP): Move to i386.md.
7362 2021-10-08 Vladimir N. Makarov <vmakarov@redhat.com>
7364 PR rtl-optimization/102627
7365 * lra-constraints.c (split_reg): Use at least natural mode of hard reg.
7367 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
7369 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Grow
7372 2021-10-08 Aldy Hernandez <aldyh@redhat.com>
7374 * value-range.cc (irange::debug): New.
7375 * value-range.h (irange::debug): New.
7377 2021-10-08 Richard Sandiford <richard.sandiford@arm.com>
7379 PR tree-optimization/102385
7380 * predict.h (change_edge_frequency): Declare.
7381 * predict.c (change_edge_frequency): New function.
7382 * tree-ssa-loop-manip.h (tree_transform_and_unroll_loop): Remove
7384 (tree_unroll_loop): Likewise.
7385 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Update accordingly.
7386 * tree-predcom.c (pcom_worker::tree_predictive_commoning_loop):
7388 * tree-ssa-loop-prefetch.c (loop_prefetch_arrays): Likewise.
7389 * tree-ssa-loop-manip.c (tree_unroll_loop): Likewise.
7390 (tree_transform_and_unroll_loop): Likewise. Use single_dom_exit
7391 to retrieve the exit edges. Make all the old profile update code
7392 conditional on !single_loop_p -- the case it was written for --
7393 and use a different approach for the single-loop case.
7395 2021-10-08 Martin Liska <mliska@suse.cz>
7397 * config/alpha/alpha.c (alpha_option_override): Use new macro
7399 * config/arc/arc.c (arc_override_options): Likewise.
7400 * config/arm/arm.c (arm_option_override): Likewise.
7401 * config/bfin/bfin.c (bfin_load_pic_reg): Likewise.
7402 * config/c6x/c6x.c (c6x_option_override): Likewise.
7403 * config/csky/csky.c: Likewise.
7404 * config/darwin.c (darwin_override_options): Likewise.
7405 * config/frv/frv.c (frv_option_override): Likewise.
7406 * config/i386/djgpp.h: Likewise.
7407 * config/i386/i386.c (ix86_stack_protect_guard): Likewise.
7408 (ix86_max_noce_ifcvt_seq_cost): Likewise.
7409 * config/ia64/ia64.c (ia64_option_override): Likewise.
7410 (ia64_override_options_after_change): Likewise.
7411 * config/m32c/m32c.c (m32c_option_override): Likewise.
7412 * config/m32r/m32r.c (m32r_init): Likewise.
7413 * config/m68k/m68k.c (m68k_option_override): Likewise.
7414 * config/microblaze/microblaze.c (microblaze_option_override): Likewise.
7415 * config/mips/mips.c (mips_option_override): Likewise.
7416 * config/nios2/nios2.c (nios2_option_override): Likewise.
7417 * config/nvptx/nvptx.c (nvptx_option_override): Likewise.
7418 * config/pa/pa.c (pa_option_override): Likewise.
7419 * config/riscv/riscv.c (riscv_option_override): Likewise.
7420 * config/rs6000/aix71.h: Likewise.
7421 * config/rs6000/aix72.h: Likewise.
7422 * config/rs6000/aix73.h: Likewise.
7423 * config/rs6000/rs6000.c (darwin_rs6000_override_options): Likewise.
7424 (rs6000_override_options_after_change): Likewise.
7425 (rs6000_linux64_override_options): Likewise.
7426 (glibc_supports_ieee_128bit): Likewise.
7427 (rs6000_option_override_internal): Likewise.
7428 (rs6000_file_start): Likewise.
7429 (rs6000_darwin_file_start): Likewise.
7430 * config/rs6000/rtems.h: Likewise.
7431 * config/rs6000/sysv4.h: Likewise.
7432 * config/rs6000/vxworks.h (SUB3TARGET_OVERRIDE_OPTIONS): Likewise.
7433 * config/s390/s390.c (s390_option_override): Likewise.
7434 * config/sh/linux.h: Likewise.
7435 * config/sh/netbsd-elf.h (while): Likewise.
7436 * config/sh/sh.c (sh_option_override): Likewise.
7437 * config/sol2.c (solaris_override_options): Likewise.
7438 * config/sparc/sparc.c (sparc_option_override): Likewise.
7439 * config/tilegx/tilegx.c (tilegx_option_override): Likewise.
7440 * config/visium/visium.c (visium_option_override): Likewise.
7441 * config/vxworks.c (vxworks_override_options): Likewise.
7442 * lto-opts.c (lto_write_options): Likewise.
7443 * omp-expand.c (expand_omp_simd): Likewise.
7444 * omp-general.c (omp_max_vf): Likewise.
7445 * omp-offload.c (oacc_xform_loop): Likewise.
7446 * opts.h (OPTION_SET_P): Likewise.
7447 * targhooks.c (default_max_noce_ifcvt_seq_cost): Likewise.
7448 * toplev.c (process_options): Likewise.
7449 * tree-predcom.c: Likewise.
7450 * tree-sra.c (analyze_all_variable_accesses): Likewise.
7452 2021-10-08 liuhongt <hongtao.liu@intel.com>
7455 * config/i386/i386.c (ix86_optab_supported_p):
7456 Return true for HFmode.
7457 * match.pd: Simplify (_Float16) ceil ((double) x) to
7458 __builtin_ceilf16 (a) when a is _Float16 type and
7459 direct_internal_fn_supported_p.
7461 2021-10-08 liuhongt <hongtao.liu@intel.com>
7464 * config/i386/i386-expand.c (emit_reduc_half): Hanlde V4HImode.
7465 * config/i386/mmx.md (reduc_plus_scal_v4hi): New.
7466 (reduc_<code>_scal_v4hi): New.
7468 2021-10-08 liuhongt <hongtao.liu@intel.com>
7470 * common.opt (ftree-vectorize): Add Var(flag_tree_vectorize).
7471 * doc/invoke.texi (Options That Control Optimization): Update
7473 * opts.c (default_options_table): Enable auto-vectorization at
7474 O2 with very-cheap cost model.
7475 (finish_options): Use cheap cost model for
7476 explicit -ftree{,-loop}-vectorize.
7478 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
7480 * ctfc.c (ctfc_delete_container): Free hash table contents.
7482 2021-10-07 Indu Bhagat <indu.bhagat@oracle.com>
7484 * toplev.c (process_options): Do not warn for GNU GIMPLE.
7486 2021-10-07 Siddhesh Poyarekar <siddhesh@gotplt.org>
7488 * tree-object-size.c (addr_object_size,
7489 compute_builtin_object_size): Drop PDECL and POFF arguments.
7490 (addr_object_size): Adjust calls.
7491 * tree-object-size.h (compute_builtin_object_size): Drop PDECL
7494 2021-10-07 Roger Sayle <roger@nextmovesoftware.com>
7496 * rtl.def (SMUL_HIGHPART, UMUL_HIGHPART): New RTX codes for
7497 representing signed and unsigned high-part multiplication resp.
7498 * simplify-rtx.c (simplify_binary_operation_1) [SMUL_HIGHPART,
7499 UMUL_HIGHPART]: Simplify high-part multiplications by zero.
7500 [SS_PLUS, US_PLUS, SS_MINUS, US_MINUS, SS_MULT, US_MULT,
7501 SS_DIV, US_DIV]: Similar simplifications for saturating
7503 (simplify_const_binary_operation) [SS_PLUS, US_PLUS, SS_MINUS,
7504 US_MINUS, SS_MULT, US_MULT, SMUL_HIGHPART, UMUL_HIGHPART]:
7505 Implement compile-time evaluation for constant operands.
7506 * dwarf2out.c (mem_loc_descriptor): Skip SMUL_HIGHPART and
7508 * doc/rtl.texi (smul_highpart, umul_highpart): Document RTX codes.
7509 * doc/md.texi (smul@var{m}3_highpart, umul@var{m3}_highpart):
7510 Mention the new smul_highpart and umul_highpart RTX codes.
7511 * doc/invoke.texi: Silence @xref "compilation" warnings.
7513 2021-10-07 Martin Jambor <mjambor@suse.cz>
7516 * ipa-prop.c (ipa_edge_args_sum_t::duplicate): Also handle the
7517 case when the source reference description corresponds to a
7518 referance taken in a function src->caller is inlined to.
7520 2021-10-07 Jan Hubicka <hubicka@ucw.cz>
7523 * ipa-modref-tree.h (modref_access_node::contains_p): Handle offsets
7525 (modref_access_node::try_merge_with): Add sanity check that there
7526 are no redundant entries in the list.
7528 2021-10-07 Richard Biener <rguenther@suse.de>
7530 PR tree-optimization/102608
7531 * tree-ssa-sccvn.c (visit_stmt): Drop .DEFERRED_INIT to
7534 2021-10-07 Martin Liska <mliska@suse.cz>
7536 * toplev.c (toplev::main): Make
7537 save_opt_decoded_options a pointer type
7538 * toplev.h: Likewise.
7540 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
7542 * config/gcn/gcn-valu.md (gather<mode>_insn_2offsets<exec>): Apply
7543 HAVE_GCN_ASM_GLOBAL_LOAD_FIXED.
7544 (scatter<mode>_insn_2offsets<exec_scatter>): Likewise.
7546 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
7548 * config/gcn/gcn-hsa.h (SRAMOPT): Include the whole option string.
7549 Adjust for new -msram-ecc=any behaviour.
7550 (ASM_SPEC): Adjust -mxnack and -msram-ecc usage.
7551 * config/gcn/gcn.c (output_file_start): Implement -msram-ecc=any.
7552 * config/gcn/mkoffload.c (EF_AMDGPU_XNACK): Rename to ...
7553 (EF_AMDGPU_XNACK_V3): ... this.
7554 (EF_AMDGPU_SRAM_ECC): Rename to ...
7555 (EF_AMDGPU_SRAM_ECC_V3): ... this.
7556 (EF_AMDGPU_FEATURE_XNACK_V4): New.
7557 (EF_AMDGPU_FEATURE_XNACK_UNSUPPORTED_V4): New.
7558 (EF_AMDGPU_FEATURE_XNACK_ANY_V4): New.
7559 (EF_AMDGPU_FEATURE_XNACK_OFF_V4): New.
7560 (EF_AMDGPU_FEATURE_XNACK_ON_V4): New.
7561 (EF_AMDGPU_FEATURE_SRAMECC_V4): New.
7562 (EF_AMDGPU_FEATURE_SRAMECC_UNSUPPORTED_V4): New.
7563 (EF_AMDGPU_FEATURE_SRAMECC_ANY_V4): New.
7564 (EF_AMDGPU_FEATURE_SRAMECC_OFF_V4): New.
7565 (EF_AMDGPU_FEATURE_SRAMECC_ON_V4): New.
7566 (SET_XNACK_ON): New.
7567 (SET_XNACK_OFF): New.
7569 (SET_SRAM_ECC_ON): New.
7570 (SET_SRAM_ECC_ANY): New.
7571 (SET_SRAM_ECC_OFF): New.
7572 (TEST_SRAM_ECC_ANY): New.
7573 (TEST_SRAM_ECC_ON): New.
7574 (main): Implement HSACOv4 and -msram-ecc=any.
7576 2021-10-07 Andrew Stubbs <ams@codesourcery.com>
7578 * config.in: Regenerate.
7579 * config/gcn/gcn-hsa.h (X_FIJI): New macro.
7583 (A_FIJI): Rename to ...
7585 (A_900): Rename to ...
7587 (A_906): Rename to ...
7589 (A_908): Rename to ...
7591 (SRAMOPT): New macro.
7592 (ASM_SPEC): Adjust xnack option usage.
7593 * config/gcn/gcn.c (output_file_start): Adjust amdgcn_target usage.
7594 * configure: Regenerate.
7595 * configure.ac: Detect LLVM assembler dialect.
7597 2021-10-07 Richard Biener <rguenther@suse.de>
7599 * tree-pretty-print.c (dump_generic_node): Do not elide
7600 printing '&' when dumping with -gimple.
7602 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
7604 * gimple-range-cache.cc (non_null_ref::adjust_range): Call new
7606 * gimple-range-fold.cc (adjust_pointer_diff_expr): Ditto.
7607 (adjust_imagpart_expr): Ditto.
7608 * value-range.cc (irange::irange_intersect): Call new routine if
7609 RHS is a single pair.
7610 (irange::intersect): New wide_int version.
7611 * value-range.h (class irange): New prototype.
7613 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
7615 * gimple-range-edge.cc (gimple_outgoing_range::gimple_outgoing_range):
7616 Add parameter to limit size when recognizing switches.
7617 (gimple_outgoing_range::edge_range_p): Check size limit.
7618 * gimple-range-edge.h (gimple_outgoing_range): Add size field.
7619 * gimple-range-gori.cc (gori_map::calculate_gori): Ignore switches
7620 that exceed the size limit.
7621 (gori_compute::gori_compute): Add initializer.
7622 * params.opt (evrp-switch-limit): New.
7623 * doc/invoke.texi: Update docs.
7625 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
7627 * value-range.h (irange::set_varying): Use TYPE_MIN_VALUE and
7628 TYPE_MAX_VALUE instead of creating new trees when possible.
7630 2021-10-06 Andrew MacLeod <amacleod@redhat.com>
7632 * gimple-range-cache.cc (non_null_ref::adjust_range): Check for
7633 zero and non-zero more efficently.
7635 2021-10-06 Richard Biener <rguenther@suse.de>
7638 * dumpfile.h (TDF_GIMPLE_VAL): New.
7639 (dump_flag): Re-order and adjust TDF_* flags. Make
7640 the enum uint32_t. Use std::underlying_type in the
7642 (optgroup_flag): Likewise for the operator overloads.
7643 * tree-pretty-print.c (dump_generic_node): Wrap ADDR_EXPR
7644 in _Literal if TDF_GIMPLE_VAL.
7645 * gimple-pretty-print.c (dump_gimple_assign): Add
7646 TDF_GIMPLE_VAL to flags when dumping operands where only
7647 is_gimple_val are allowed.
7648 (dump_gimple_cond): Likewise.
7650 2021-10-06 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
7652 * gimple-isel.cc (gimple_expand_vec_cond_expr): Remove redundant if
7655 2021-10-05 qing zhao <qing.zhao@oracle.com>
7657 PR middle-end/102359
7658 * gimplify.c (gimplify_decl_expr): Not add initialization for an
7659 auto variable when it has been initialized by frontend.
7661 2021-10-05 Aldy Hernandez <aldyh@redhat.com>
7663 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
7666 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
7668 * common/config/avr/avr-common.c (avr_handle_option): Mark
7669 argument as ATTRIBUTE_UNUSED.
7671 2021-10-05 Jan-Benedict Glaw <jbglaw@lug-owl.de>
7673 * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC):
7674 Undefine before redefinition.
7676 2021-10-05 Richard Biener <rguenther@suse.de>
7678 * toplev.c (no_backend): Remove global var.
7679 (process_options): Pass in no_backend, move post_options
7680 langhook call to toplev::main.
7681 (do_compile): Pass in no_backend, move process_options call
7683 (toplev::run_self_tests): Check no_backend at the caller.
7684 (toplev::main): Call post_options and process_options
7685 split out from do_compile, do self-tests only if
7686 no_backend is initialized.
7688 2021-10-05 Richard Biener <rguenther@suse.de>
7690 * tree-cfg.c (dump_function_to_file): Dump the UID of the
7691 function as part of the name when requested.
7692 * tree-pretty-print.c (dump_function_name): Dump the UID when
7693 requested and the langhook produced the actual name.
7695 2021-10-05 Richard Biener <rguenther@suse.de>
7697 PR middle-end/102587
7698 PR middle-end/102285
7699 * internal-fn.c (expand_DEFERRED_INIT): Fall back to
7700 zero-initialization as last resort, use the constant
7701 size as given by the DEFERRED_INIT argument to build
7704 2021-10-04 Marek Polacek <polacek@redhat.com>
7707 * doc/invoke.texi: Document -Warray-compare.
7709 2021-10-04 Richard Biener <rguenther@suse.de>
7711 * gimplify.c (is_var_need_auto_init): DECL_HARD_REGISTER
7712 variables are not to be initialized.
7714 2021-10-04 Richard Biener <rguenther@suse.de>
7716 * expr.h (non_mem_decl_p): Declare.
7717 (mem_ref_refers_to_non_mem_p): Likewise.
7718 * expr.c (non_mem_decl_p): Export.
7719 (mem_ref_refers_to_non_mem_p): Likewise.
7720 * internal-fn.c (expand_DEFERRED_INIT): Do not expand the LHS
7721 but check the base with mem_ref_refers_to_non_mem_p
7724 2021-10-04 Richard Biener <rguenther@suse.de>
7726 PR tree-optimization/102570
7727 * tree-ssa-sccvn.h (vn_reference_op_struct): Document
7728 we are using clique for the internal function code.
7729 * tree-ssa-sccvn.c (vn_reference_op_eq): Compare the
7730 internal function code.
7731 (print_vn_reference_ops): Print the internal function code.
7732 (vn_reference_op_compute_hash): Hash it.
7733 (copy_reference_ops_from_call): Record it.
7734 (visit_stmt): Remove the restriction around internal function
7736 (fully_constant_vn_reference_p): Use fold_const_call and handle
7738 (vn_reference_eq): Compare call return types.
7739 * tree-ssa-pre.c (create_expression_by_pieces): Handle
7740 generating calls to internal functions.
7741 (compute_avail): Remove the restriction around internal function
7744 2021-10-04 Aldy Hernandez <aldyh@redhat.com>
7746 PR tree-optimization/102560
7747 * gimple-ssa-warn-alloca.c (alloca_call_type): Remove static
7748 marker for invalid_range.
7750 2021-10-04 Richard Biener <rguenther@suse.de>
7752 PR middle-end/102587
7753 * internal-fn.c (expand_DEFERRED_INIT): Guard register
7754 initialization path an avoid initializing VLA registers
7757 2021-10-04 Eric Botcazou <ebotcazou@adacore.com>
7759 * config/rs6000/vxworks.h (TARGET_INIT_LIBFUNCS): Delete.
7761 2021-10-03 Martin Liska <mliska@suse.cz>
7763 * toplev.c (toplev::main): Check opt_index if it is a part
7766 2021-10-02 Aldy Hernandez <aldyh@redhat.com>
7768 PR tree-optimization/102563
7769 * range-op.cc (operator_lshift::op1_range): Do not clobber
7772 2021-10-02 Martin Liska <mliska@suse.cz>
7774 * toplev.c (toplev::main): save_decoded_options[0] is program
7775 name and so it should be skipped.
7777 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
7779 PR tree-optimization/102546
7780 * range-op.cc (operator_lshift::op1_range): Teach range-ops that
7781 X << Y is non-zero implies X is also non-zero.
7783 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
7785 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
7787 * config/aarch64/aarch64-tune.md: Regenerate.
7788 * doc/invoke.texi: Update docs.
7790 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
7792 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
7794 * config/aarch64/aarch64-tune.md: Regenerate.
7795 * doc/invoke.texi: Update docs.
7797 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
7799 * config/aarch64/aarch64-cores.def (AARCH64_CORE): New
7801 * config/aarch64/aarch64-tune.md: Regenerate.
7802 * doc/invoke.texi: Update docs.
7804 2021-10-01 Martin Sebor <msebor@redhat.com>
7807 * doc/invoke.texi (-Waddress): Update.
7808 * gengtype.c (write_types): Avoid -Waddress.
7809 * poly-int.h (POLY_SET_COEFF): Avoid using null.
7811 2021-10-01 John David Anglin <danglin@gcc.gnu.org>
7814 * config/pa/pa.c (pa_option_override): Default to dwarf version 4
7817 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
7819 * config/aarch64/aarch64.h (AARCH64_FL_V9): Update value.
7821 2021-10-01 Aldy Hernandez <aldyh@redhat.com>
7823 * gimple-range-path.cc (path_range_query::compute_ranges): Use
7825 * gimple-range-path.h (class path_range_query): Remove shadowed
7827 (path_range_query::get_path_oracle): New.
7829 2021-10-01 Jakub Jelinek <jakub@redhat.com>
7830 Richard Biener <rguenther@suse.de>
7833 * doc/invoke.texi (-fsanitize=integer-divide-by-zero): Remove
7834 INT_MIN / -1 division detection from here ...
7835 (-fsanitize=signed-integer-overflow): ... and add it here.
7837 2021-10-01 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
7839 * config/aarch64/aarch64-arches.def (AARCH64_ARCH): Added
7841 * config/aarch64/aarch64.h (AARCH64_FL_V9): New.
7842 (AARCH64_FL_FOR_ARCH9): New flags for Armv9-A.
7843 (AARCH64_ISA_V9): New ISA flag.
7844 * doc/invoke.texi: Update docs.
7846 2021-10-01 Martin Liska <mliska@suse.cz>
7848 * toplev.c (toplev::main): Save decoded optimization options.
7849 * toplev.h (save_opt_decoded_options): New.
7850 * doc/extend.texi: Be more clear about optimize and target
7853 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
7855 * explow.c: Include langhooks.h.
7856 (set_stack_check_libfunc): Build a proper function type.
7858 2021-10-01 Eric Botcazou <ebotcazou@adacore.com>
7861 * config/i386/i386.c (legitimate_pic_address_disp_p): For PE-COFF do
7862 not return true for external weak function symbols in medium model.
7864 2021-10-01 Jakub Jelinek <jakub@redhat.com>
7866 * tree.h (OMP_CLAUSE_ORDER_REPRODUCIBLE): Define.
7867 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_ORDER>: Print
7868 reproducible: for OMP_CLAUSE_ORDER_REPRODUCIBLE.
7869 * omp-general.c (omp_extract_for_data): If OMP_CLAUSE_ORDER is seen
7870 without OMP_CLAUSE_ORDER_UNCONSTRAINED, overwrite sched_kind to
7871 OMP_CLAUSE_SCHEDULE_STATIC.
7873 2021-10-01 Richard Biener <rguenther@suse.de>
7875 PR middle-end/102518
7876 * tree-inline.c (setup_one_parameter): Avoid substituting
7877 an invariant into contexts where a GIMPLE register is not valid.
7879 2021-09-30 Przemyslaw Wirkus <przemyslaw.wirkus@arm.com>
7881 * config/arm/arm-cpus.in: Add Cortex-R52+ CPU.
7882 * config/arm/arm-tables.opt: Regenerate.
7883 * config/arm/arm-tune.md: Regenerate.
7884 * doc/invoke.texi: Update docs.
7886 2021-09-30 Uroš Bizjak <ubizjak@gmail.com>
7889 * config/i386/i386.md
7890 (sign_extend:WIDE (any_logic:NARROW (memory, immediate)) splitters):
7893 2021-09-30 Tobias Burnus <tobias@codesourcery.com>
7895 * omp-low.c (omp_runtime_api_call): Add omp_aligned_{,c}alloc and
7896 omp_{c,re}alloc, fix omp_alloc/omp_free.
7898 2021-09-30 Martin Liska <mliska@suse.cz>
7900 * defaults.h (ASM_OUTPUT_ASCII): Do not hide global variable
7901 asm_out_file and stream directly to MYFILE.
7903 2021-09-30 Richard Biener <rguenther@suse.de>
7905 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
7906 Restore and fix condition under which we apply npeel to
7907 the DRs misalignment value.
7909 2021-09-30 Richard Biener <rguenther@suse.de>
7911 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
7912 Fix npeel check for variable amount of peeling.
7914 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
7916 * lto-wrapper.c (run_gcc): Plug snprintf overflow.
7918 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
7920 * gimple-range.cc (gimple_ranger::debug): New.
7921 * gimple-range.h (class gimple_ranger): Add debug.
7923 2021-09-30 Aldy Hernandez <aldyh@redhat.com>
7925 PR middle-end/102519
7926 * tree-vrp.c (hybrid_threader::~hybrid_threader): Free m_query.
7928 2021-09-29 Indu Bhagat <indu.bhagat@oracle.com>
7931 * btfout.c (GTY): Add GTY (()) albeit for cosmetic only purpose.
7932 (btf_finalize): Empty the hash_map btf_var_ids.
7934 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
7936 * tree-vrp.c (thread_through_all_blocks): Return bool.
7937 (execute_vrp_threader): Return TODO_* flags.
7938 (pass_data_vrp_threader): Set todo_flags_finish to 0.
7940 2021-09-29 Aldy Hernandez <aldyh@redhat.com>
7942 * timevar.def (TV_TREE_VRP_THREADER): New.
7943 * tree-vrp.c: Use TV_TREE_VRP_THREADER for VRP threader pass.
7945 2021-09-29 David Faust <david.faust@oracle.com>
7947 * config.gcc (bpf-*-*): Do not overwrite extra_headers.
7949 2021-09-29 Jonathan Wright <jonathan.wright@arm.com>
7951 * config/aarch64/aarch64-builtins.c (TYPES_BINOP_PPU): Define
7952 new type qualifier enum.
7953 (TYPES_TERNOP_SSSU): Likewise.
7954 (TYPES_TERNOP_PPPU): Likewise.
7955 * config/aarch64/aarch64-simd-builtins.def: Define PPU, SSU,
7956 PPPU and SSSU builtin generator macros for qtbl1 and qtbx1
7958 * config/aarch64/arm_neon.h (vqtbl1_p8): Use type-qualified
7959 builtin and remove casts.
7960 (vqtbl1_s8): Likewise.
7961 (vqtbl1q_p8): Likewise.
7962 (vqtbl1q_s8): Likewise.
7963 (vqtbx1_s8): Likewise.
7964 (vqtbx1_p8): Likewise.
7965 (vqtbx1q_s8): Likewise.
7966 (vqtbx1q_p8): Likewise.
7967 (vtbl1_p8): Likewise.
7968 (vtbl2_p8): Likewise.
7969 (vtbx2_p8): Likewise.
7971 2021-09-29 Richard Biener <rguenther@suse.de>
7973 * tree-vect-data-refs.c (vect_dr_misalign_for_aligned_access):
7975 (vect_update_misalignment_for_peel): Use it to update
7976 misaligned to the value necessary for an aligned access.
7977 (vect_get_peeling_costs_all_drs): Likewise.
7978 (vect_enhance_data_refs_alignment): Likewise.
7980 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7982 * config/aarch64/aarch64.c (aarch64_expand_cpymem): Count number of
7983 emitted operations and adjust heuristic for code size.
7985 2021-09-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7987 * config/aarch64/aarch64.c (aarch64_expand_setmem): Count number of
7988 emitted operations and adjust heuristic for code size.
7990 2021-09-29 Jakub Jelinek <jakub@redhat.com>
7992 PR middle-end/102504
7993 * gimplify.c (gimplify_scan_omp_clauses): Use omp_check_private even
7994 in OMP_SCOPE clauses, not just on worksharing construct clauses.
7996 2021-09-28 Geng Qi <gengqi@linux.alibaba.com>
7998 * config/riscv/riscv.md (mulv<mode>4): Call gen_smul<mode>3_highpart.
7999 (<u>mulditi3): Call <su>muldi3_highpart.
8000 (<u>muldi3_highpart): Rename to <su>muldi3_highpart.
8001 (<u>mulsidi3): Call <su>mulsi3_highpart.
8002 (<u>mulsi3_highpart): Rename to <su>mulsi3_highpart.
8004 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
8006 * config/darwin.h (DSYMUTIL_SPEC): Recognize D sources.
8008 2021-09-28 Iain Sandoe <iain@sandoe.co.uk>
8010 * config/rs6000/darwin.h (FIXED_R13): Add for PPC64.
8011 (FIRST_SAVED_GP_REGNO): Save from R13 even when it is one
8014 2021-09-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8016 * config/aarch64/aarch64.h (AARCH64_FL_LS64): Define
8017 (AARCH64_FL_V8_7): Likewise.
8018 (AARCH64_FL_FOR_ARCH8_7): Likewise.
8019 * config/aarch64/aarch64-arches.def (armv8.7-a): Define.
8020 * config/aarch64/aarch64-option-extensions.def (ls64): Define.
8021 * doc/invoke.texi: Document the above.
8023 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
8025 * dbgcnt.c (dbg_cnt_counter): New.
8026 * dbgcnt.h (dbg_cnt_counter): New.
8027 * dumpfile.c (dump_options): Add entry for TDF_THREADING.
8028 * dumpfile.h (enum dump_flag): Add TDF_THREADING.
8029 * gimple-range-path.cc (DEBUG_SOLVER): Use TDF_THREADING.
8030 * tree-ssa-threadupdate.c (dump_jump_thread_path): Dump out
8033 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
8035 * cfgcleanup.c (pass_jump::execute): Check
8036 flag_expensive_optimizations.
8037 (pass_jump_after_combine::gate): Same.
8038 * doc/invoke.texi (-fthread-jumps): Enable for -O1.
8039 * opts.c (default_options_table): Enable -fthread-jumps at -O1.
8040 * tree-ssa-threadupdate.c
8041 (fwd_jt_path_registry::remove_jump_threads_including): Bail unless
8044 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
8046 * tree-ssa-reassoc.c (biased_names): New global.
8047 (propagate_bias_p): New function.
8048 (loop_carried_phi): Remove.
8049 (propagate_rank): Propagate bias along single uses.
8050 (get_rank): Update biased_names when needed.
8052 2021-09-28 Ilya Leoshkevich <iii@linux.ibm.com>
8054 * passes.def (pass_reassoc): Rename parameter to early_p.
8055 * tree-ssa-reassoc.c (reassoc_bias_loop_carried_phi_ranks_p):
8057 (phi_rank): Don't bias loop-carried phi ranks
8058 before vectorization pass.
8059 (execute_reassoc): Add bias_loop_carried_phi_ranks_p parameter.
8060 (pass_reassoc::pass_reassoc): Add bias_loop_carried_phi_ranks_p
8062 (pass_reassoc::set_param): Set bias_loop_carried_phi_ranks_p
8064 (pass_reassoc::execute): Pass bias_loop_carried_phi_ranks_p to
8066 (pass_reassoc::bias_loop_carried_phi_ranks_p): New member.
8068 2021-09-28 Jakub Jelinek <jakub@redhat.com>
8071 * config/i386/i386.c (standard_80387_constant_p): Don't recognize
8072 special 80387 instruction XFmode constants if flag_rounding_math.
8074 2021-09-28 Richard Biener <rguenther@suse.de>
8076 PR tree-optimization/100112
8077 * tree-ssa-sccvn.c (visit_reference_op_load): Record the
8078 referece into the hashtable twice in case last_vuse is
8079 different from the original vuse on the stmt.
8081 2021-09-28 Jakub Jelinek <jakub@redhat.com>
8083 PR middle-end/102492
8084 * gimplify.c (gimplify_adjust_omp_clauses_1): Don't call the
8085 omp_finish_clause langhook on implicitly added OMP_CLAUSE_PRIVATE
8086 clauses on SIMD constructs.
8088 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
8090 PR tree-optimization/102511
8091 * gimple-range-path.cc (path_range_query::range_on_path_entry):
8092 Return VARYING when nothing found.
8094 2021-09-28 Hongyu Wang <hongyu.wang@intel.com>
8097 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): Add
8099 (VALID_SSE2_REG_VHF_MODE): Add V4HFmode and V2HFmode.
8100 (VALID_MMX_REG_MODE): Add V4HFmode.
8101 (SSE_REG_MODE_P): Replace VALID_AVX512FP16_REG_MODE with
8102 vector mode condition.
8103 * config/i386/i386.c (classify_argument): Parse V4HF/V2HF
8105 (function_arg_32): Add V4HFmode.
8106 (function_arg_advance_32): Likewise.
8107 * config/i386/i386.md (mode): Add V4HF/V2HF.
8108 (MODE_SIZE): Likewise.
8109 * config/i386/mmx.md (MMXMODE): Add V4HF mode.
8110 (V_32): Add V2HF mode.
8111 (VHF_32_64): New mode iterator.
8112 (*mov<mode>_internal): Adjust sse alternatives to support
8114 (*mov<mode>_internal): Adjust sse alternatives to support
8116 (<insn><mode>3): New define_insn for add/sub/mul/div.
8118 2021-09-28 Aldy Hernandez <aldyh@redhat.com>
8120 * tree-ssa-threadbackward.c (pass_thread_jumps::gate): Check
8122 (pass_early_thread_jumps::gate): Same.
8123 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
8124 Return if !flag_thread_jumps.
8125 * tree-ssa-threadupdate.c
8126 (jt_path_registry::register_jump_thread): Assert that
8127 flag_thread_jumps is true.
8129 2021-09-28 liuhongt <hongtao.liu@intel.com>
8132 (simplify_context::simplify_binary_operation_1): Relax
8133 condition of simplifying (vec_concat:M (vec_select op0
8134 index0)(vec_select op1 index1)) to allow different modes
8135 between op0 and M, but have same inner mode.
8137 2021-09-28 liuhongt <hongtao.liu@intel.com>
8139 * config/i386/i386-expand.c (emit_reduc_half): Handle
8140 V8HF/V16HF/V32HFmode.
8141 * config/i386/sse.md (REDUC_SSE_PLUS_MODE): Add V8HF.
8142 (REDUC_SSE_SMINMAX_MODE): Ditto.
8143 (REDUC_PLUS_MODE): Add V16HF and V32HF.
8144 (REDUC_SMINMAX_MODE): Ditto.
8146 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
8148 * gimple-range-path.cc
8149 (path_range_query::precompute_ranges_in_block): Rename to...
8150 (path_range_query::compute_ranges_in_block): ...this.
8151 (path_range_query::precompute_ranges): Rename to...
8152 (path_range_query::compute_ranges): ...this.
8153 (path_range_query::precompute_relations): Rename to...
8154 (path_range_query::compute_relations): ...this.
8155 (path_range_query::precompute_phi_relations): Rename to...
8156 (path_range_query::compute_phi_relations): ...this.
8157 * gimple-range-path.h: Rename precompute* to compute*.
8158 * tree-ssa-threadbackward.c
8159 (back_threader::find_taken_edge_switch): Same.
8160 (back_threader::find_taken_edge_cond): Same.
8161 * tree-ssa-threadedge.c
8162 (hybrid_jt_simplifier::compute_ranges_from_state): Same.
8163 (hybrid_jt_state::register_equivs_stmt): Inline...
8164 * tree-ssa-threadedge.h: ...here.
8166 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
8168 * tree-vrp.c (lhs_of_dominating_assert): Remove.
8169 (class vrp_jt_state): Remove.
8170 (class vrp_jt_simplifier): Remove.
8171 (vrp_jt_simplifier::simplify): Remove.
8172 (class vrp_jump_threader): Remove.
8173 (vrp_jump_threader::vrp_jump_threader): Remove.
8174 (vrp_jump_threader::~vrp_jump_threader): Remove.
8175 (vrp_jump_threader::before_dom_children): Remove.
8176 (vrp_jump_threader::after_dom_children): Remove.
8178 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
8180 * passes.def (pass_vrp_threader): New.
8181 * tree-pass.h (make_pass_vrp_threader): Add make_pass_vrp_threader.
8182 * tree-ssa-threadedge.c (hybrid_jt_state::register_equivs_stmt): New.
8183 (hybrid_jt_simplifier::hybrid_jt_simplifier): New.
8184 (hybrid_jt_simplifier::simplify): New.
8185 (hybrid_jt_simplifier::compute_ranges_from_state): New.
8186 * tree-ssa-threadedge.h (class hybrid_jt_state): New.
8187 (class hybrid_jt_simplifier): New.
8188 * tree-vrp.c (execute_vrp): Remove ASSERT_EXPR based jump
8190 (class hybrid_threader): New.
8191 (hybrid_threader::hybrid_threader): New.
8192 (hybrid_threader::~hybrid_threader): New.
8193 (hybrid_threader::before_dom_children): New.
8194 (hybrid_threader::after_dom_children): New.
8195 (execute_vrp_threader): New.
8196 (class pass_vrp_threader): New.
8197 (make_pass_vrp_threader): New.
8199 2021-09-27 Martin Liska <mliska@suse.cz>
8201 * output.h (enum section_flag): New.
8202 (SECTION_FORGET): Remove.
8203 (SECTION_ENTSIZE): Make it (1UL << 8) - 1.
8204 (SECTION_STYLE_MASK): Define it based on other enum
8206 * varasm.c (switch_to_section): Remove unused handling of
8209 2021-09-27 Martin Liska <mliska@suse.cz>
8211 * common.opt: Add new variable flag_default_complex_method.
8212 * opts.c (finish_options): Handle flags related to
8213 x_flag_complex_method.
8214 * toplev.c (process_options): Remove option handling related
8215 to flag_complex_method.
8217 2021-09-27 Richard Biener <rguenther@suse.de>
8219 PR middle-end/102450
8220 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid using
8221 type_for_size, instead use int_mode_for_size.
8223 2021-09-27 Andrew Pinski <apinski@marvell.com>
8226 * gimplify.c (gimplify_save_expr): Return early
8227 if the type of val is error_mark_node.
8229 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
8231 * tree-ssanames.c (ssa_name_has_boolean_range): Use
8234 2021-09-27 Aldy Hernandez <aldyh@redhat.com>
8236 * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
8237 vrp_visit_cond_stmt.
8238 * tree-ssa-dom.c (cprop_operand): Convert to range_query API.
8239 (cprop_into_stmt): Same.
8240 (dom_opt_dom_walker::optimize_stmt): Same.
8242 2021-09-27 Richard Biener <rguenther@suse.de>
8244 PR tree-optimization/97351
8245 PR tree-optimization/97352
8246 PR tree-optimization/82426
8247 * tree-vectorizer.h (dr_misalignment): Add vector type
8249 (aligned_access_p): Likewise.
8250 (known_alignment_for_access_p): Likewise.
8251 (vect_supportable_dr_alignment): Likewise.
8252 (vect_known_alignment_in_bytes): Likewise. Refactor.
8253 (DR_MISALIGNMENT): Remove.
8254 (vect_update_shared_vectype): Likewise.
8255 * tree-vect-data-refs.c (dr_misalignment): Refactor, handle
8256 a vector type with larger alignment requirement and apply
8257 the negative step adjustment here.
8258 (vect_calculate_target_alignment): Remove.
8259 (vect_compute_data_ref_alignment): Get explicit vector type
8260 argument, do not apply a negative step alignment adjustment
8262 (vect_slp_analyze_node_alignment): Re-analyze alignment
8263 when we re-visit the DR with a bigger desired alignment but
8264 keep more precise results from smaller alignments.
8265 * tree-vect-slp.c (vect_update_shared_vectype): Remove.
8266 (vect_slp_analyze_node_operations_1): Do not update the
8267 shared vector type on stmts.
8268 * tree-vect-stmts.c (vect_analyze_stmt): Push/pop the
8269 vector type of an SLP node to the representative stmt-info.
8270 (vect_transform_stmt): Likewise.
8272 2021-09-27 liuhongt <hongtao.liu@intel.com>
8275 2021-09-09 liuhongt <hongtao.liu@intel.com>
8278 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
8279 (reduc_plus_scal_v4sf): .. this, New define_expand.
8280 (reduc_plus_scal_v2df): .. and this, New define_expand.
8282 2021-09-26 liuhongt <hongtao.liu@intel.com>
8284 * doc/extend.texi (Half-Precision): Remove storage only
8285 description for _Float16 w/o avx512fp16.
8287 2021-09-25 Dimitar Dimitrov <dimitar@dinux.eu>
8289 * config/pru/constraints.md (Rrio): New constraint.
8290 * config/pru/predicates.md (regio_operand): New predicate.
8291 * config/pru/pru-pragma.c (pru_register_pragmas): Register
8292 the __regio_symbol address space.
8293 * config/pru/pru-protos.h (pru_symref2ioregno): Declaration.
8294 * config/pru/pru.c (pru_symref2ioregno): New helper function.
8295 (pru_legitimate_address_p): Remove.
8296 (pru_addr_space_legitimate_address_p): Use the address space
8298 (pru_nongeneric_pointer_addrspace): New helper function.
8299 (pru_insert_attributes): New function to validate __regio_symbol
8301 (TARGET_INSERT_ATTRIBUTES): New macro.
8302 (TARGET_LEGITIMATE_ADDRESS_P): Remove.
8303 (TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P): New macro.
8304 * config/pru/pru.h (enum reg_class): Add REGIO_REGS class.
8305 * config/pru/pru.md (*regio_readsi): New pattern to read I/O
8307 (*regio_nozext_writesi): New pattern to write to I/O registers.
8308 (*regio_zext_write_r30<EQS0:mode>): Ditto.
8309 * doc/extend.texi: Document the new PRU Named Address Space.
8311 2021-09-24 Patrick Palka <ppalka@redhat.com>
8315 * real.c (encode_ieee_double): Avoid unwanted sign extension.
8316 (encode_ieee_quad): Likewise.
8318 2021-09-24 Vladimir Makarov <vmakarov@redhat.com>
8320 PR rtl-optimization/102147
8321 * ira-build.c (ira_conflict_vector_profitable_p): Make
8322 profitability calculation independent of host compiler pointer and
8325 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
8327 * gimple-range-path.cc (path_range_query::path_range_query):
8328 Move debugging header...
8329 (path_range_query::precompute_ranges): ...here.
8330 (path_range_query::internal_range_of_expr): Do not call
8331 range_on_path_entry if NAME is defined in the current block.
8333 2021-09-24 Richard Biener <rguenther@suse.de>
8335 * cfghooks.c (verify_flow_info): Verify unallocated BB and
8336 edge flags are not set.
8338 2021-09-24 Aldy Hernandez <aldyh@redhat.com>
8340 * tree-ssa-threadupdate.c (jt_path_registry::cancel_invalid_paths):
8342 (jt_path_registry::register_jump_thread): Call
8343 cancel_invalid_paths.
8344 * tree-ssa-threadupdate.h (class jt_path_registry): Add
8345 cancel_invalid_paths.
8347 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
8349 PR tree-optimization/102400
8350 * tree-ssa-sccvn.c (vn_reference_insert_pieces): Initialize
8351 result_vdef to zero value.
8353 2021-09-24 Feng Xue <fxue@os.amperecomputing.com>
8355 PR tree-optimization/102451
8356 * tree-ssa-dse.c (delete_dead_or_redundant_call): Record bb of stmt
8359 2021-09-24 Hongyu Wang <hongyu.wang@intel.com>
8361 * config/i386/sse.md (cond_<insn><mode>): Extend to support
8363 (cond_mul<mode>): Likewise.
8364 (cond_div<mode>): Likewise.
8365 (cond_<code><mode>): Likewise.
8366 (cond_fma<mode>): Likewise.
8367 (cond_fms<mode>): Likewise.
8368 (cond_fnma<mode>): Likewise.
8369 (cond_fnms<mode>): Likewise.
8371 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
8373 PR tree-optimization/102463
8374 * gimple-range-fold.cc (fold_using_range::relation_fold_and_or): If
8375 there is no range-ops handler, don't look for a relation.
8377 2021-09-23 Andrew MacLeod <amacleod@redhat.com>
8379 * gimple-range-cache.cc (ranger_cache::ranger_cache): Take
8380 non-executable_edge flag as parameter.
8381 * gimple-range-cache.h (ranger_cache): Adjust prototype.
8382 * gimple-range-gori.cc (gori_compute::gori_compute): Take
8383 non-executable_edge flag as parameter.
8384 (gori_compute::outgoing_edge_range_p): Check new flag.
8385 * gimple-range-gori.h (gori_compute): Adjust prototype.
8386 * gimple-range.cc (gimple_ranger::gimple_ranger): Create new flag.
8387 (gimple_ranger::range_on_edge): Check new flag.
8388 * gimple-range.h (gimple_ranger::non_executable_edge_flag): New.
8389 * gimple-ssa-evrp.c (rvrp_folder): Pass ranger flag to simplifer.
8390 (hybrid_folder::hybrid_folder): Set ranger non-executable flag value.
8391 (hybrid_folder::fold_stmt): Set flag value in the simplifer.
8392 * vr-values.c (simplify_using_ranges::set_and_propagate_unexecutable):
8393 Use not_executable flag if provided inmstead of EDGE_EXECUTABLE.
8394 (simplify_using_ranges::simplify_switch_using_ranges): Clear
8395 EDGE_EXECUTABLE like it originally did.
8396 (simplify_using_ranges::cleanup_edges_and_switches): Clear any
8397 NON_EXECUTABLE flags.
8398 (simplify_using_ranges::simplify_using_ranges): Adjust.
8399 * vr-values.h (class simplify_using_ranges): Adjust.
8400 (simplify_using_ranges::set_range_query): Add non-executable flag param.
8402 2021-09-23 Bill Schmidt <wschmidt@linux.ibm.com>
8405 * config/rs6000/rs6000-call.c (rs6000_aggregate_candidate): Detect
8406 zero-width bit fields and return indicator.
8407 (rs6000_discover_homogeneous_aggregate): Diagnose when the
8408 presence of a zero-width bit field changes parameter passing in
8411 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
8413 * gimple-range-fold.cc (fold_using_range::range_of_phi):
8414 Remove dominator check.
8416 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
8418 * gimple-range-path.cc (path_range_query::precompute_relations):
8419 Hoist edge calculations before using EDGE_SUCC.
8421 2021-09-23 Jonathan Wakely <jwakely@redhat.com>
8423 * configure.ac: Fix --with-multilib-list description.
8424 * configure: Regenerate.
8426 2021-09-23 Richard Biener <rguenther@suse.de>
8428 PR tree-optimization/102448
8429 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
8430 Clear alignment info copied from DR_PTR_INFO.
8432 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
8434 * config/i386/i386-expand.c (ix86_use_mask_cmp_p): Enable
8436 * config/i386/sse.md (sseintvecmodelower): Add HF vector modes.
8437 (<avx512>_store<mode>_mask): Extend to support HF vector modes.
8438 (vec_cmp<mode><avx512fmaskmodelower>): Likewise.
8439 (vcond_mask_<mode><avx512fmaskmodelower>): Likewise.
8440 (vcond<mode><mode>): New expander.
8441 (vcond<mode><sseintvecmodelower>): Likewise.
8442 (vcond<sseintvecmodelower><mode>): Likewise.
8443 (vcondu<mode><sseintvecmodelower>): Likewise.
8445 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
8447 * config/i386/sse.md (extend<ssePHmodelower><mode>2):
8449 (extendv4hf<mode>2): Likewise.
8450 (extendv2hfv2df2): Likewise.
8451 (trunc<mode><ssePHmodelower>2): Likewise.
8452 (avx512fp16_vcvt<castmode>2ph_<mode>): Rename to ...
8453 (trunc<mode>v4hf2): ... this, and drop constraints.
8454 (avx512fp16_vcvtpd2ph_v2df): Rename to ...
8455 (truncv2dfv2hf2): ... this, and likewise.
8457 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
8459 * config/i386/sse.md (float<floatunssuffix><mode><ssePHmodelower>2):
8461 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>):
8463 (float<floatunssuffix><mode>v4hf2): ... this, and drop constraints.
8464 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Rename to ...
8465 (float<floatunssuffix>v2div2hf2): ... this, and likewise.
8467 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
8469 * config/i386/i386.md (fix<fixunssuffix>_trunchf<mode>2): New expander.
8470 (fixuns_trunchfhi2): Likewise.
8471 (*fixuns_trunchfsi2zext): New define_insn.
8472 * config/i386/sse.md (ssePHmodelower): New mode_attr.
8473 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
8474 New expander for same element vector fix_truncate.
8475 (fix<fixunssuffix>_trunc<ssePHmodelower><mode>2):
8476 Likewise for V4HF to V4SI/V4DI fix_truncate.
8477 (fix<fixunssuffix>_truncv2hfv2di2):
8478 Likeise for V2HF to V2DI fix_truncate.
8480 2021-09-23 Hongyu Wang <hongyu.wang@intel.com>
8482 * config/i386/i386.md (<code>hf3): New expander.
8484 2021-09-23 liuhongt <hongtao.liu@intel.com>
8486 * config/i386/sse.md (FMAMODEM): extend to handle FP16.
8487 (VFH_SF_AVX512VL): Extend to handle HFmode.
8488 (VF_SF_AVX512VL): Deleted.
8490 2021-09-23 liuhongt <hongtao.liu@intel.com>
8492 * config/i386/i386.md (rinthf2): New expander.
8493 (nearbyinthf2): New expander.
8495 2021-09-23 Aldy Hernandez <aldyh@redhat.com>
8497 * tree-ssa-dom.c (class dom_jump_threader_simplifier): Rename...
8498 (class dom_jt_state): ...this and provide virtual overrides.
8499 (dom_jt_state::register_equiv): New.
8500 (class dom_jt_simplifier): Rename from
8501 dom_jump_threader_simplifier.
8502 (dom_jump_threader_simplifier::simplify): Rename...
8503 (dom_jt_simplifier::simplify): ...to this.
8504 (pass_dominator::execute): Use dom_jt_simplifier and
8506 * tree-ssa-threadedge.c (jump_threader::jump_threader):
8508 (jt_state::register_equivs_stmt): Abstract out...
8509 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
8511 (jump_threader::thread_around_empty_blocks): Update state.
8512 (jump_threader::thread_through_normal_block): Same.
8513 (jt_state::jt_state): Remove.
8514 (jt_state::push): Remove pass specific bits. Keep block vector
8516 (jt_state::append_path): New.
8517 (jt_state::pop): Remove pass specific bits.
8518 (jt_state::register_equiv): Same.
8519 (jt_state::record_ranges_from_stmt): Same.
8520 (jt_state::register_equivs_on_edge): Same. Rename...
8521 (jt_state::register_equivs_edge): ...to this.
8522 (jt_state::dump): New.
8523 (jt_state::debug): New.
8524 (jump_threader_simplifier::simplify): Remove.
8525 (jt_state::get_path): New.
8526 * tree-ssa-threadedge.h (class jt_simplifier): Make into a base
8527 class. Expose common functionality as virtual methods.
8528 (class jump_threader_simplifier): Same. Rename...
8529 (class jt_simplifier): ...to this.
8530 * tree-vrp.c (class vrp_jump_threader_simplifier): Rename...
8531 (class vrp_jt_simplifier): ...to this. Provide pass specific
8533 (class vrp_jt_state): New.
8534 (vrp_jump_threader_simplifier::simplify): Rename...
8535 (vrp_jt_simplifier::simplify): ...to this. Inline code from
8536 what used to be the base class.
8537 (vrp_jump_threader::vrp_jump_threader): Use vrp_jt_state and
8540 2021-09-22 Tobias Burnus <tobias@codesourcery.com>
8543 * doc/invoke.texi (-Wno-missing-include-dirs.): Document Fortran
8546 2021-09-22 Roger Sayle <roger@nextmovesoftware.com>
8547 Richard Biener <rguenther@suse.de>
8549 * match.pd (negation simplifications): Implement some negation
8550 folding transformations from fold-const.c's fold_negate_expr.
8551 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Add a SIMPLIFY
8552 argument, to control whether the op should be simplified prior
8553 to looking up/assigning a value number.
8554 (vn_nary_build_or_lookup): Update call to vn_nary_build_or_lookup_1.
8555 (vn_nary_simplify): Likewise.
8556 (visit_nary_op): Likewise, but when constructing a NEGATE_EXPR
8557 now call vn_nary_build_or_lookup_1 disabling simplification.
8559 2021-09-22 Jiufu Guo <guojiufu@linux.ibm.com>
8561 PR tree-optimization/102087
8562 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
8563 Update bound/cmp/control for niter.
8565 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
8567 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
8568 Move check for non-empty BB here.
8569 (fur_source::register_outgoing_edges): ...from here.
8571 2021-09-22 Aldy Hernandez <aldyh@redhat.com>
8573 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
8574 Remove call to improve_range_with_equivs.
8575 (path_range_query::improve_range_with_equivs): Remove
8576 * gimple-range-path.h: Remove improve_range_with_equivs.
8578 2021-09-22 dianhong xu <dianhong.xu@intel.com>
8580 * config/i386/avx512fp16intrin.h:
8581 (_mm512_mask_blend_ph): New intrinsic.
8582 (_mm512_permutex2var_ph): Ditto.
8583 (_mm512_permutexvar_ph): Ditto.
8584 * config/i386/avx512fp16vlintrin.h:
8585 (_mm256_mask_blend_ph): New intrinsic.
8586 (_mm256_permutex2var_ph): Ditto.
8587 (_mm256_permutexvar_ph): Ditto.
8588 (_mm_mask_blend_ph): Ditto.
8589 (_mm_permutex2var_ph): Ditto.
8590 (_mm_permutexvar_ph): Ditto.
8592 2021-09-22 dianhong xu <dianhong.xu@intel.com>
8594 * config/i386/avx512fp16intrin.h: Add new intrinsics.
8595 (_mm512_conj_pch): New intrinsic.
8596 (_mm512_mask_conj_pch): Ditto.
8597 (_mm512_maskz_conj_pch): Ditto.
8598 * config/i386/avx512fp16vlintrin.h: Add new intrinsics.
8599 (_mm256_conj_pch): New intrinsic.
8600 (_mm256_mask_conj_pch): Ditto.
8601 (_mm256_maskz_conj_pch): Ditto.
8602 (_mm_conj_pch): Ditto.
8603 (_mm_mask_conj_pch): Ditto.
8604 (_mm_maskz_conj_pch): Ditto.
8606 2021-09-22 dianhong xu <dianhong.xu@intel.com>
8608 * config/i386/avx512fp16intrin.h (_MM512_REDUCE_OP): New macro
8609 (_mm512_reduce_add_ph): New intrinsic.
8610 (_mm512_reduce_mul_ph): Ditto.
8611 (_mm512_reduce_min_ph): Ditto.
8612 (_mm512_reduce_max_ph): Ditto.
8613 * config/i386/avx512fp16vlintrin.h
8614 (_MM256_REDUCE_OP/_MM_REDUCE_OP): New macro.
8615 (_mm256_reduce_add_ph): New intrinsic.
8616 (_mm256_reduce_mul_ph): Ditto.
8617 (_mm256_reduce_min_ph): Ditto.
8618 (_mm256_reduce_max_ph): Ditto.
8619 (_mm_reduce_add_ph): Ditto.
8620 (_mm_reduce_mul_ph): Ditto.
8621 (_mm_reduce_min_ph): Ditto.
8622 (_mm_reduce_max_ph): Ditto.
8624 2021-09-22 dianhong xu <dianhong.xu@intel.com>
8626 * config/i386/avx512fp16intrin.h (__m512h_u, __m256h_u,
8627 __m128h_u): New typedef.
8628 (_mm512_load_ph): New intrinsic.
8629 (_mm256_load_ph): Ditto.
8630 (_mm_load_ph): Ditto.
8631 (_mm512_loadu_ph): Ditto.
8632 (_mm256_loadu_ph): Ditto.
8633 (_mm_loadu_ph): Ditto.
8634 (_mm512_store_ph): Ditto.
8635 (_mm256_store_ph): Ditto.
8636 (_mm_store_ph): Ditto.
8637 (_mm512_storeu_ph): Ditto.
8638 (_mm256_storeu_ph): Ditto.
8639 (_mm_storeu_ph): Ditto.
8640 (_mm512_abs_ph): Ditto.
8641 * config/i386/avx512fp16vlintrin.h
8642 (_mm_abs_ph): Ditto.
8643 (_mm256_abs_ph): Ditto.
8645 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
8647 * config/s390/tpf.md (prologue_tpf, epilogue_tpf): Add cc clobber.
8649 2021-09-22 Andreas Krebbel <krebbel@linux.ibm.com>
8652 * config/s390/s390.c (s390_expand_insv): Emit a normal move if it
8653 is actually a full copy of the source operand into the target.
8654 Don't emit a strict low part move if source and target mode match.
8656 2021-09-22 Jakub Jelinek <jakub@redhat.com>
8658 PR middle-end/102415
8659 * omp-expand.c (expand_omp_single): If region->exit is NULL,
8660 assert region->entry is GIMPLE_OMP_SCOPE region and return.
8662 2021-09-22 Jakub Jelinek <jakub@redhat.com>
8664 * tree.h (OMP_CLAUSE_ALLOCATE_ALIGN): Define.
8665 * tree.c (omp_clause_num_ops): Change number of OMP_CLAUSE_ALLOCATE
8666 arguments from 2 to 3.
8667 * tree-pretty-print.c (dump_omp_clause): Print allocator() around
8668 allocate clause allocator and print align if present.
8669 * omp-low.c (scan_sharing_clauses): Force allocate_map entry even
8670 for omp_default_mem_alloc if align modifier is present. If align
8671 modifier is present, use TREE_LIST to encode both allocator and
8673 (lower_private_allocate, lower_rec_input_clauses, create_task_copyfn):
8674 Handle align modifier on allocator clause if present.
8676 2021-09-22 liuhongt <hongtao.liu@intel.com>
8678 * config/i386/i386.md (define_attr "isa"): Add
8680 (define_attr "enabled"): Correspond fma_or_avx512vl to
8681 TARGET_FMA || TARGET_AVX512VL.
8682 * config/i386/mmx.md (fmav2sf4): Extend to AVX512 fma.
8687 2021-09-22 liuhongt <hongtao.liu@intel.com>
8689 * config/i386/i386.md (cstorehf3): New define_expand.
8691 2021-09-22 liuhongt <hongtao.liu@intel.com>
8693 * config/i386/i386.md (<rounding_insn>hf2): New expander.
8694 (sse4_1_round<mode>2): Extend from MODEF to MODEFH.
8695 * config/i386/sse.md (*sse4_1_round<ssescalarmodesuffix>):
8696 Extend from VF_128 to VFH_128.
8698 2021-09-22 liuhongt <hongtao.liu@intel.com>
8700 * config/i386/i386-features.c (i386-features.c): Handle
8702 * config/i386/i386.md (sqrthf2): New expander.
8703 (*sqrthf2): New define_insn.
8704 * config/i386/sse.md
8705 (*<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
8708 2021-09-22 liuhongt <hongtao.liu@intel.com>
8710 * config/i386/avx512fp16intrin.h (_mm_mask_fcmadd_sch):
8712 (_mm_mask3_fcmadd_sch): Likewise.
8713 (_mm_maskz_fcmadd_sch): Likewise.
8714 (_mm_fcmadd_sch): Likewise.
8715 (_mm_mask_fmadd_sch): Likewise.
8716 (_mm_mask3_fmadd_sch): Likewise.
8717 (_mm_maskz_fmadd_sch): Likewise.
8718 (_mm_fmadd_sch): Likewise.
8719 (_mm_mask_fcmadd_round_sch): Likewise.
8720 (_mm_mask3_fcmadd_round_sch): Likewise.
8721 (_mm_maskz_fcmadd_round_sch): Likewise.
8722 (_mm_fcmadd_round_sch): Likewise.
8723 (_mm_mask_fmadd_round_sch): Likewise.
8724 (_mm_mask3_fmadd_round_sch): Likewise.
8725 (_mm_maskz_fmadd_round_sch): Likewise.
8726 (_mm_fmadd_round_sch): Likewise.
8727 (_mm_fcmul_sch): Likewise.
8728 (_mm_mask_fcmul_sch): Likewise.
8729 (_mm_maskz_fcmul_sch): Likewise.
8730 (_mm_fmul_sch): Likewise.
8731 (_mm_mask_fmul_sch): Likewise.
8732 (_mm_maskz_fmul_sch): Likewise.
8733 (_mm_fcmul_round_sch): Likewise.
8734 (_mm_mask_fcmul_round_sch): Likewise.
8735 (_mm_maskz_fcmul_round_sch): Likewise.
8736 (_mm_fmul_round_sch): Likewise.
8737 (_mm_mask_fmul_round_sch): Likewise.
8738 (_mm_maskz_fmul_round_sch): Likewise.
8739 * config/i386/i386-builtin.def: Add corresponding new builtins.
8740 * config/i386/sse.md
8741 (avx512fp16_fmaddcsh_v8hf_maskz<round_expand_name>): New expander.
8742 (avx512fp16_fcmaddcsh_v8hf_maskz<round_expand_name>): Ditto.
8743 (avx512fp16_fma_<complexopname>sh_v8hf<mask_scalarcz_name><round_scalarcz_name>):
8745 (avx512fp16_<complexopname>sh_v8hf_mask<round_name>): Ditto.
8746 (avx512fp16_<complexopname>sh_v8hf<mask_scalarc_name><round_scalarcz_name>):
8748 * config/i386/subst.md (mask_scalarcz_name): New.
8749 (mask_scalarc_name): Ditto.
8750 (mask_scalarc_operand3): Ditto.
8751 (mask_scalarcz_operand4): Ditto.
8752 (round_scalarcz_name): Ditto.
8753 (round_scalarc_mask_operand3): Ditto.
8754 (round_scalarcz_mask_operand4): Ditto.
8755 (round_scalarc_mask_op3): Ditto.
8756 (round_scalarcz_mask_op4): Ditto.
8757 (round_scalarcz_constraint): Ditto.
8758 (round_scalarcz_nimm_predicate): Ditto.
8759 (mask_scalarcz): Ditto.
8760 (mask_scalarc): Ditto.
8761 (round_scalarcz): Ditto.
8763 2021-09-22 liuhongt <hongtao.liu@intel.com>
8765 * config/i386/avx512fp16intrin.h (_mm512_fcmadd_pch):
8767 (_mm512_mask_fcmadd_pch): Likewise.
8768 (_mm512_mask3_fcmadd_pch): Likewise.
8769 (_mm512_maskz_fcmadd_pch): Likewise.
8770 (_mm512_fmadd_pch): Likewise.
8771 (_mm512_mask_fmadd_pch): Likewise.
8772 (_mm512_mask3_fmadd_pch): Likewise.
8773 (_mm512_maskz_fmadd_pch): Likewise.
8774 (_mm512_fcmadd_round_pch): Likewise.
8775 (_mm512_mask_fcmadd_round_pch): Likewise.
8776 (_mm512_mask3_fcmadd_round_pch): Likewise.
8777 (_mm512_maskz_fcmadd_round_pch): Likewise.
8778 (_mm512_fmadd_round_pch): Likewise.
8779 (_mm512_mask_fmadd_round_pch): Likewise.
8780 (_mm512_mask3_fmadd_round_pch): Likewise.
8781 (_mm512_maskz_fmadd_round_pch): Likewise.
8782 (_mm512_fcmul_pch): Likewise.
8783 (_mm512_mask_fcmul_pch): Likewise.
8784 (_mm512_maskz_fcmul_pch): Likewise.
8785 (_mm512_fmul_pch): Likewise.
8786 (_mm512_mask_fmul_pch): Likewise.
8787 (_mm512_maskz_fmul_pch): Likewise.
8788 (_mm512_fcmul_round_pch): Likewise.
8789 (_mm512_mask_fcmul_round_pch): Likewise.
8790 (_mm512_maskz_fcmul_round_pch): Likewise.
8791 (_mm512_fmul_round_pch): Likewise.
8792 (_mm512_mask_fmul_round_pch): Likewise.
8793 (_mm512_maskz_fmul_round_pch): Likewise.
8794 * config/i386/avx512fp16vlintrin.h (_mm_fmadd_pch):
8796 (_mm_mask_fmadd_pch): Likewise.
8797 (_mm_mask3_fmadd_pch): Likewise.
8798 (_mm_maskz_fmadd_pch): Likewise.
8799 (_mm256_fmadd_pch): Likewise.
8800 (_mm256_mask_fmadd_pch): Likewise.
8801 (_mm256_mask3_fmadd_pch): Likewise.
8802 (_mm256_maskz_fmadd_pch): Likewise.
8803 (_mm_fcmadd_pch): Likewise.
8804 (_mm_mask_fcmadd_pch): Likewise.
8805 (_mm_mask3_fcmadd_pch): Likewise.
8806 (_mm_maskz_fcmadd_pch): Likewise.
8807 (_mm256_fcmadd_pch): Likewise.
8808 (_mm256_mask_fcmadd_pch): Likewise.
8809 (_mm256_mask3_fcmadd_pch): Likewise.
8810 (_mm256_maskz_fcmadd_pch): Likewise.
8811 (_mm_fmul_pch): Likewise.
8812 (_mm_mask_fmul_pch): Likewise.
8813 (_mm_maskz_fmul_pch): Likewise.
8814 (_mm256_fmul_pch): Likewise.
8815 (_mm256_mask_fmul_pch): Likewise.
8816 (_mm256_maskz_fmul_pch): Likewise.
8817 (_mm_fcmul_pch): Likewise.
8818 (_mm_mask_fcmul_pch): Likewise.
8819 (_mm_maskz_fcmul_pch): Likewise.
8820 (_mm256_fcmul_pch): Likewise.
8821 (_mm256_mask_fcmul_pch): Likewise.
8822 (_mm256_maskz_fcmul_pch): Likewise.
8823 * config/i386/i386-builtin-types.def (V8HF_FTYPE_V8HF_V8HF_V8HF,
8824 V8HF_FTYPE_V16HF_V16HF_V16HF, V16HF_FTYPE_V16HF_V16HF_V16HF_UQI,
8825 V32HF_FTYPE_V32HF_V32HF_V32HF_INT,
8826 V32HF_FTYPE_V32HF_V32HF_V32HF_UHI_INT): Add new builtin types.
8827 * config/i386/i386-builtin.def: Add new builtins.
8828 * config/i386/i386-expand.c: Handle new builtin types.
8829 * config/i386/subst.md (SUBST_CV): New.
8830 (maskc_name): Ditto.
8831 (maskc_operand3): Ditto.
8833 (sdc_maskz_name): Ditto.
8834 (sdc_mask_op4): Ditto.
8835 (sdc_mask_op5): Ditto.
8836 (sdc_mask_mode512bit_condition): Ditto.
8838 (round_maskc_operand3): Ditto.
8839 (round_sdc_mask_operand4): Ditto.
8840 (round_maskc_op3): Ditto.
8841 (round_sdc_mask_op4): Ditto.
8842 (round_saeonly_sdc_mask_operand5): Ditto.
8843 * config/i386/sse.md (unspec): Add complex fma unspecs.
8844 (avx512fmaskcmode): New.
8845 (UNSPEC_COMPLEX_F_C_MA): Ditto.
8846 (UNSPEC_COMPLEX_F_C_MUL): Ditto.
8847 (complexopname): Ditto.
8848 (<avx512>_fmaddc_<mode>_maskz<round_expand_name>): New expander.
8849 (<avx512>_fcmaddc_<mode>_maskz<round_expand_name>): Ditto.
8850 (fma_<complexopname>_<mode><sdc_maskz_name><round_name>): New
8852 (<avx512>_<complexopname>_<mode>_mask<round_name>): Ditto.
8853 (<avx512>_<complexopname>_<mode><maskc_name><round_name>): Ditto.
8855 2021-09-22 Kewen Lin <linkw@linux.ibm.com>
8857 * config/rs6000/rs6000.opt (rs6000-density-pct-threshold,
8858 rs6000-density-size-threshold, rs6000-density-penalty,
8859 rs6000-density-load-pct-threshold,
8860 rs6000-density-load-num-threshold): New parameter.
8861 * config/rs6000/rs6000.c (rs6000_density_test): Adjust with
8862 corresponding parameters.
8864 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
8866 * gimple-range-path.cc (path_range_query::defined_outside_path):
8868 (path_range_query::range_on_path_entry): New.
8869 (path_range_query::internal_range_of_expr): Resolve unknowns
8871 (path_range_query::improve_range_with_equivs): New.
8872 (path_range_query::ssa_range_in_phi): Resolve unknowns with
8874 * gimple-range-path.h (class path_range_query): Add
8875 defined_outside_path, range_on_path_entry, and
8876 improve_range_with_equivs.
8878 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
8880 * gimple-range-path.cc (path_range_query::add_to_imports): New.
8881 (path_range_query::add_copies_to_imports): New.
8882 (path_range_query::precompute_ranges): Call
8883 add_copies_to_imports.
8884 * gimple-range-path.h (class path_range_query): Add prototypes
8885 for add_copies_to_imports and add_to_imports.
8887 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
8889 * gimple-range-path.cc (path_range_query::range_defined_in_block):
8890 Remove useless code.
8892 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
8894 * gimple-range-fold.h (class fur_source): Make oracle protected.
8895 * gimple-range-path.cc (path_range_query::path_range_query): Add
8896 resolve argument. Initialize oracle.
8897 (path_range_query::~path_range_query): Delete oracle.
8898 (path_range_query::range_of_stmt): Adapt to use relations.
8899 (path_range_query::precompute_ranges): Pre-compute relations.
8900 (class jt_fur_source): New
8901 (jt_fur_source::jt_fur_source): New.
8902 (jt_fur_source::register_relation): New.
8903 (jt_fur_source::query_relation): New.
8904 (path_range_query::precompute_relations): New.
8905 (path_range_query::precompute_phi_relations): New.
8906 * gimple-range-path.h (path_range_query): Add resolve argument.
8907 Add oracle, precompute_relations, precompute_phi_relations.
8908 * tree-ssa-threadbackward.c (back_threader::back_threader): Pass
8909 resolve argument to solver.
8911 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
8913 * gimple-range-fold.cc (fold_using_range::range_of_range_op):
8914 Rename postfold_gcond_edges to register_outgoing_edges and
8916 (fold_using_range::postfold_gcond_edges): Rename...
8917 (fur_source::register_outgoing_edges): ...to this.
8918 * gimple-range-fold.h (postfold_gcond_edges): Rename to
8919 register_outgoing_edges and move to fur_source.
8921 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
8923 * gimple-range-fold.cc (fold_using_range::range_of_phi): Check
8924 dom_info_available_p.
8926 2021-09-21 Aldy Hernandez <aldyh@redhat.com>
8928 * gimple-range-cache.cc (non_null_ref::non_null_ref): Use create
8929 and quick_grow_cleared instead of safe_grow_cleared.
8931 2021-09-21 Thomas Schwinge <thomas@codesourcery.com>
8934 * omp-oacc-neuter-broadcast.cc (oacc_do_neutering): Evaluate
8937 2021-09-21 Richard Earnshaw <rearnsha@arm.com>
8939 * configure.ac: Detect when the assembler supports new-style
8940 architecture extensions.
8941 * common/config/arm/arm-common.c (arm_rewrite_mcpu): Return
8942 the full CPU string if the assembler can grok it.
8943 (arm_rewrite_march): Likewise but for the architecture.
8944 * config.in: Regenerate.
8945 * configure: Regenerate.
8947 2021-09-21 Richard Biener <rguenther@suse.de>
8949 PR tree-optimization/102421
8950 * tree-vect-loop.c (vect_dissolve_slp_only_groups): Copy and
8951 adjust alignment info.
8953 2021-09-21 Kewen Lin <linkw@linux.ibm.com>
8955 * ipa-fnsummary.c (ipa_fn_summary_write): Remove inconsistent
8956 bitfield stream out.
8958 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
8960 * gimple-range-fold.cc (fold_using_range::range_of_phi): Ignore
8961 undefined edges, apply an equivalence if appropriate.
8962 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Return
8963 UNDEFINED if EDGE_EXECUTABLE is not set.
8964 * gimple-range.cc (gimple_ranger::gimple_ranger): Set all edges
8965 as EXECUTABLE upon startup.
8966 (gimple_ranger::range_on_edge): Return UNDEFINED for edges without
8967 EDGE_EXECUTABLE set.
8968 * vr-values.c (set_and_propagate_unexecutable): New.
8969 (simplify_using_ranges::fold_cond): Call set_and_propagate.
8970 (simplify_using_ranges::simplify_switch_using_ranges): Ditto.
8971 * vr-values.h: Add prototype.
8973 2021-09-20 Andrew MacLeod <amacleod@redhat.com>
8975 * value-relation.cc (equiv_oracle::register_initial_def): New.
8976 (equiv_oracle::register_relation): Call register_initial_def.
8977 (equiv_oracle::add_equiv_to_block): New. Split register_relation.
8978 (relation_oracle::register_stmt): Check def block of PHI arguments.
8979 * value-relation.h (equiv_oracle): Add new prototypes.
8981 2021-09-20 Matthias Kretz <m.kretz@gsi.de>
8983 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
8984 Define __RECIPROCAL_MATH__, __NO_SIGNED_ZEROS__,
8985 __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__, and
8986 __ROUNDING_MATH__ according to their corresponding flags.
8987 * doc/cpp.texi: Document __RECIPROCAL_MATH__,
8988 __NO_SIGNED_ZEROS__, __NO_TRAPPING_MATH__, __ASSOCIATIVE_MATH__,
8989 and __ROUNDING_MATH__.
8991 2021-09-20 Richard Biener <rguenther@suse.de>
8993 * tree-vect-stmts.c (vectorizable_load): Use the vectype
8996 2021-09-20 Richard Biener <rguenther@suse.de>
8998 * tree-vect-data-refs.c (vect_duplicate_ssa_name_ptr_info):
8999 Do not compute alignment of the vectorized access here.
9001 2021-09-20 Richard Biener <rguenther@suse.de>
9003 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
9004 Store -1 for runtime alias peeling iterations.
9006 2021-09-20 Richard Biener <rguenther@suse.de>
9008 * config.gcc: Obsolete hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11*.
9010 2021-09-20 Thomas Schwinge <thomas@codesourcery.com>
9012 * input.c (string_concat_db::record_string_concatenation)
9013 (string_concat_db::get_string_concatenation): Skip for
9014 'RESERVED_LOCATION_P'.
9016 2021-09-20 Richard Biener <rguenther@suse.de>
9018 PR tree-optimization/65206
9019 * tree-data-ref.h (struct data_reference): Add alt_indices,
9021 * tree-data-ref.c (free_data_ref): Release alt_indices.
9022 (dr_analyze_indices): Work on struct indices and get DR_REF as tree.
9023 (create_data_ref): Adjust.
9024 (initialize_data_dependence_relation): Split into head
9025 and tail. When the base objects fail to match up try
9026 again with pointer-based analysis of indices.
9027 * tree-vectorizer.c (vec_info_shared::check_datarefs): Do
9028 not compare the lazily computed alternate set of indices.
9030 2021-09-20 Iain Sandoe <iain@sandoe.co.uk>
9032 * gcc.c: Test for execute OK when we find the
9033 programs for assembler linker and dsymutil and those
9034 were specified at configure-time.
9036 2021-09-19 Martin Sebor <msebor@redhat.com>
9038 PR middle-end/102403
9039 * gimple-predicate-analysis.cc (predicate::init_from_control_deps):
9040 Correct a function pre/postcondition.
9042 2021-09-19 Martin Sebor <msebor@redhat.com>
9044 PR middle-end/102243
9045 * tree-ssa-strlen.c (get_range): Handle null cfun.
9047 2021-09-19 Iain Sandoe <iain@sandoe.co.uk>
9049 * config/darwin.h (LINK_COMMAND_SPEC_A): Use Darwin10
9050 unwinder shim as a convenience library.
9052 2021-09-19 Andrew Pinski <apinski@marvell.com>
9054 * doc/install.texi: Add note about
9055 binutils 2.35 is required for LTO usage.
9057 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
9059 * tree-ssa-threadbackward.c
9060 (back_threader_registry::register_path): Use push_edge.
9061 * tree-ssa-threadedge.c
9062 (jump_threader::thread_around_empty_blocks): Same.
9063 (jump_threader::thread_through_normal_block): Same.
9064 (jump_threader::thread_across_edge): Same. Also, use auto_bitmap.
9066 * tree-ssa-threadupdate.c
9067 (jt_path_registry::allocate_thread_edge): Remove.
9068 (jt_path_registry::push_edge): New.
9069 (dump_jump_thread_path): Make static.
9070 * tree-ssa-threadupdate.h (allocate_thread_edge): Remove.
9073 2021-09-19 Aldy Hernandez <aldyh@redhat.com>
9075 * gimple-range-path.cc (path_range_query::path_range_query): Add
9077 (path_range_query::dump): Remove extern declaration of dump_ranger.
9078 * gimple-range-trace.cc (dump_ranger): Add DEBUG_FUNCTION marker.
9079 * gimple-range-trace.h (dump_ranger): Add prototype.
9081 2021-09-19 John Ericson <git@JohnEricson.me>
9083 * gcc.c (find_a_program): New function, factored out of...
9084 (find_a_file): Here.
9085 (execute): Use find_a_program when looking for programs rather
9088 2021-09-19 Matwey V. Kornilov <matwey.kornilov@gmail.com>
9090 * config/avr/avr-mcus.def: Add atmega324pb.
9091 * doc/avr-mmcu.texi: Corresponding changes.
9093 2021-09-19 Roger Sayle <roger@nextmovesoftware.com>
9096 * match.pd (cmp @0 REAL_CST@1): When @0 is also REAL_CST, apply
9097 the same transformations as to @1. For comparisons against NaN,
9098 don't check HONOR_SNANS but confirm that neither operand is a
9101 2021-09-19 Benjamin Peterson <benjamin@locrian.net>
9103 * attribs.c (make_unique_name): Delete.
9104 * attribs.h (make_unique_name): Delete.
9106 2021-09-19 Andrew Pinski <apinski@marvell.com>
9108 * lra-constraints.c (check_and_process_move): Assert
9109 that dclass and sclass are greater than or equal to NO_REGS.
9111 2021-09-18 Jakub Jelinek <jakub@redhat.com>
9113 * tree.h (OMP_CLAUSE_ORDER_UNCONSTRAINED): Define.
9114 * tree-pretty-print.c (dump_omp_clause): Print unconstrained:
9115 for OMP_CLAUSE_ORDER_UNCONSTRAINED.
9117 2021-09-18 liuhongt <hongtao.liu@intel.com>
9119 * config/i386/i386-features.c (remove_partial_avx_dependency):
9120 Restrict TARGET_USE_VECTOR_FP_CONVERTS and
9121 TARGET_USE_VECTOR_CONVERTS to conversion instructions only.
9123 2021-09-18 Jakub Jelinek <jakub@redhat.com>
9125 * gimplify.c (omp_default_clause): For C/C++ default({,first}private),
9126 if file/namespace scope variable doesn't have predetermined sharing,
9127 treat it as if there was default(none).
9129 2021-09-18 liuhongt <hongtao.liu@intel.com>
9131 * config/i386/avx512fp16intrin.h (_mm_fmadd_sh):
9133 (_mm_mask_fmadd_sh): Likewise.
9134 (_mm_mask3_fmadd_sh): Likewise.
9135 (_mm_maskz_fmadd_sh): Likewise.
9136 (_mm_fmadd_round_sh): Likewise.
9137 (_mm_mask_fmadd_round_sh): Likewise.
9138 (_mm_mask3_fmadd_round_sh): Likewise.
9139 (_mm_maskz_fmadd_round_sh): Likewise.
9140 (_mm_fnmadd_sh): Likewise.
9141 (_mm_mask_fnmadd_sh): Likewise.
9142 (_mm_mask3_fnmadd_sh): Likewise.
9143 (_mm_maskz_fnmadd_sh): Likewise.
9144 (_mm_fnmadd_round_sh): Likewise.
9145 (_mm_mask_fnmadd_round_sh): Likewise.
9146 (_mm_mask3_fnmadd_round_sh): Likewise.
9147 (_mm_maskz_fnmadd_round_sh): Likewise.
9148 (_mm_fmsub_sh): Likewise.
9149 (_mm_mask_fmsub_sh): Likewise.
9150 (_mm_mask3_fmsub_sh): Likewise.
9151 (_mm_maskz_fmsub_sh): Likewise.
9152 (_mm_fmsub_round_sh): Likewise.
9153 (_mm_mask_fmsub_round_sh): Likewise.
9154 (_mm_mask3_fmsub_round_sh): Likewise.
9155 (_mm_maskz_fmsub_round_sh): Likewise.
9156 (_mm_fnmsub_sh): Likewise.
9157 (_mm_mask_fnmsub_sh): Likewise.
9158 (_mm_mask3_fnmsub_sh): Likewise.
9159 (_mm_maskz_fnmsub_sh): Likewise.
9160 (_mm_fnmsub_round_sh): Likewise.
9161 (_mm_mask_fnmsub_round_sh): Likewise.
9162 (_mm_mask3_fnmsub_round_sh): Likewise.
9163 (_mm_maskz_fnmsub_round_sh): Likewise.
9164 * config/i386/i386-builtin-types.def
9165 (V8HF_FTYPE_V8HF_V8HF_V8HF_UQI_INT): New builtin type.
9166 * config/i386/i386-builtin.def: Add new builtins.
9167 * config/i386/i386-expand.c: Handle new builtin type.
9168 * config/i386/sse.md (fmai_vmfmadd_<mode><round_name>):
9169 Ajdust to support FP16.
9170 (fmai_vmfmsub_<mode><round_name>): Ditto.
9171 (fmai_vmfnmadd_<mode><round_name>): Ditto.
9172 (fmai_vmfnmsub_<mode><round_name>): Ditto.
9173 (*fmai_fmadd_<mode>): Ditto.
9174 (*fmai_fmsub_<mode>): Ditto.
9175 (*fmai_fnmadd_<mode><round_name>): Ditto.
9176 (*fmai_fnmsub_<mode><round_name>): Ditto.
9177 (avx512f_vmfmadd_<mode>_mask<round_name>): Ditto.
9178 (avx512f_vmfmadd_<mode>_mask3<round_name>): Ditto.
9179 (avx512f_vmfmadd_<mode>_maskz<round_expand_name>): Ditto.
9180 (avx512f_vmfmadd_<mode>_maskz_1<round_name>): Ditto.
9181 (*avx512f_vmfmsub_<mode>_mask<round_name>): Ditto.
9182 (avx512f_vmfmsub_<mode>_mask3<round_name>): Ditto.
9183 (*avx512f_vmfmsub_<mode>_maskz_1<round_name>): Ditto.
9184 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
9185 (*avx512f_vmfnmsub_<mode>_mask3<round_name>): Ditto.
9186 (*avx512f_vmfnmsub_<mode>_mask<round_name>): Ditto.
9187 (*avx512f_vmfnmadd_<mode>_mask<round_name>): Renamed to ...
9188 (avx512f_vmfnmadd_<mode>_mask<round_name>) ... this, and
9189 adjust to support FP16.
9190 (avx512f_vmfnmadd_<mode>_mask3<round_name>): Ditto.
9191 (avx512f_vmfnmadd_<mode>_maskz_1<round_name>): Ditto.
9192 (avx512f_vmfnmadd_<mode>_maskz<round_expand_name>): New
9195 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
9197 * config/i386/sse.md (avx512fmaskmodelower): Extend to support
9199 (maskload<mode><avx512fmaskmodelower>): Ditto.
9200 (maskstore<mode><avx512fmaskmodelower>): Ditto.
9202 2021-09-18 H.J. Lu <hjl.tools@gmail.com>
9204 * config/i386/i386-expand.c (ix86_expand_fp_absneg_operator):
9206 (ix86_expand_copysign): Ditto.
9207 (ix86_expand_xorsign): Ditto.
9208 * config/i386/i386.c (ix86_build_const_vector): Handle HF vector
9210 (ix86_build_signbit_mask): Ditto.
9211 (ix86_can_change_mode_class): Ditto.
9212 * config/i386/i386.md
9213 (SSEMODEF): Add HFmode.
9214 (ssevecmodef): Ditto.
9215 (<code>hf2): New define_expand.
9216 (*<code>hf2_1): New define_insn_and_split.
9217 (copysign<mode>): Extend to support HFmode under AVX512FP16.
9218 (xorsign<mode>): Ditto.
9219 * config/i386/sse.md (VFB): New mode iterator.
9220 (VFB_128_256): Ditto.
9222 (sseintvecmode2): Support HF vector mode.
9223 (<code><mode>2): Use new mode iterator.
9224 (*<code><mode>2): Ditto.
9225 (copysign<mode>3): Ditto.
9226 (xorsign<mode>3): Ditto.
9227 (<code><mode>3<mask_name>): Ditto.
9228 (<code><mode>3<mask_name>): Ditto.
9229 (<sse>_andnot<mode>3<mask_name>): Adjust for HF vector mode.
9230 (<sse>_andnot<mode>3<mask_name>): Ditto.
9231 (*<code><mode>3<mask_name>): Ditto.
9232 (*<code><mode>3<mask_name>): Ditto.
9234 2021-09-18 liuhongt <hongtao.liu@intel.com>
9236 * config/i386/avx512fp16intrin.h (_mm512_mask_fmadd_ph):
9238 (_mm512_mask3_fmadd_ph): Likewise.
9239 (_mm512_maskz_fmadd_ph): Likewise.
9240 (_mm512_fmadd_round_ph): Likewise.
9241 (_mm512_mask_fmadd_round_ph): Likewise.
9242 (_mm512_mask3_fmadd_round_ph): Likewise.
9243 (_mm512_maskz_fmadd_round_ph): Likewise.
9244 (_mm512_fnmadd_ph): Likewise.
9245 (_mm512_mask_fnmadd_ph): Likewise.
9246 (_mm512_mask3_fnmadd_ph): Likewise.
9247 (_mm512_maskz_fnmadd_ph): Likewise.
9248 (_mm512_fnmadd_round_ph): Likewise.
9249 (_mm512_mask_fnmadd_round_ph): Likewise.
9250 (_mm512_mask3_fnmadd_round_ph): Likewise.
9251 (_mm512_maskz_fnmadd_round_ph): Likewise.
9252 (_mm512_fmsub_ph): Likewise.
9253 (_mm512_mask_fmsub_ph): Likewise.
9254 (_mm512_mask3_fmsub_ph): Likewise.
9255 (_mm512_maskz_fmsub_ph): Likewise.
9256 (_mm512_fmsub_round_ph): Likewise.
9257 (_mm512_mask_fmsub_round_ph): Likewise.
9258 (_mm512_mask3_fmsub_round_ph): Likewise.
9259 (_mm512_maskz_fmsub_round_ph): Likewise.
9260 (_mm512_fnmsub_ph): Likewise.
9261 (_mm512_mask_fnmsub_ph): Likewise.
9262 (_mm512_mask3_fnmsub_ph): Likewise.
9263 (_mm512_maskz_fnmsub_ph): Likewise.
9264 (_mm512_fnmsub_round_ph): Likewise.
9265 (_mm512_mask_fnmsub_round_ph): Likewise.
9266 (_mm512_mask3_fnmsub_round_ph): Likewise.
9267 (_mm512_maskz_fnmsub_round_ph): Likewise.
9268 * config/i386/avx512fp16vlintrin.h (_mm256_fmadd_ph):
9270 (_mm256_mask_fmadd_ph): Likewise.
9271 (_mm256_mask3_fmadd_ph): Likewise.
9272 (_mm256_maskz_fmadd_ph): Likewise.
9273 (_mm_fmadd_ph): Likewise.
9274 (_mm_mask_fmadd_ph): Likewise.
9275 (_mm_mask3_fmadd_ph): Likewise.
9276 (_mm_maskz_fmadd_ph): Likewise.
9277 (_mm256_fnmadd_ph): Likewise.
9278 (_mm256_mask_fnmadd_ph): Likewise.
9279 (_mm256_mask3_fnmadd_ph): Likewise.
9280 (_mm256_maskz_fnmadd_ph): Likewise.
9281 (_mm_fnmadd_ph): Likewise.
9282 (_mm_mask_fnmadd_ph): Likewise.
9283 (_mm_mask3_fnmadd_ph): Likewise.
9284 (_mm_maskz_fnmadd_ph): Likewise.
9285 (_mm256_fmsub_ph): Likewise.
9286 (_mm256_mask_fmsub_ph): Likewise.
9287 (_mm256_mask3_fmsub_ph): Likewise.
9288 (_mm256_maskz_fmsub_ph): Likewise.
9289 (_mm_fmsub_ph): Likewise.
9290 (_mm_mask_fmsub_ph): Likewise.
9291 (_mm_mask3_fmsub_ph): Likewise.
9292 (_mm_maskz_fmsub_ph): Likewise.
9293 (_mm256_fnmsub_ph): Likewise.
9294 (_mm256_mask_fnmsub_ph): Likewise.
9295 (_mm256_mask3_fnmsub_ph): Likewise.
9296 (_mm256_maskz_fnmsub_ph): Likewise.
9297 (_mm_fnmsub_ph): Likewise.
9298 (_mm_mask_fnmsub_ph): Likewise.
9299 (_mm_mask3_fnmsub_ph): Likewise.
9300 (_mm_maskz_fnmsub_ph): Likewise.
9301 * config/i386/i386-builtin.def: Add corresponding new builtins.
9302 * config/i386/sse.md
9303 (<avx512>_fmadd_<mode>_maskz<round_expand_name>): Adjust to
9304 support HF vector modes.
9305 (<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name><round_name>):
9307 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
9308 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
9309 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
9310 (<avx512>_fmadd_<mode>_mask<round_name>): Ditto.
9311 (<avx512>_fmadd_<mode>_mask3<round_name>): Ditto.
9312 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Ditto.
9313 (<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name><round_name>):
9315 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
9316 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
9317 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
9318 (<avx512>_fmsub_<mode>_mask<round_name>): Ditto.
9319 (<avx512>_fmsub_<mode>_mask3<round_name>): Ditto.
9320 (<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name><round_name>):
9322 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1): Ditto.
9323 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2): Ditto.
9324 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3): Ditto.
9325 (<avx512>_fnmadd_<mode>_mask<round_name>): Ditto.
9326 (<avx512>_fnmadd_<mode>_mask3<round_name>): Ditto.
9327 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Ditto.
9328 (<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name><round_name>):
9330 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Ditto.
9331 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2): Ditto.
9332 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3): Ditto.
9333 (<avx512>_fnmsub_<mode>_mask<round_name>): Ditto.
9334 (<avx512>_fnmsub_<mode>_mask3<round_name>): Ditto.
9336 2021-09-18 liuhongt <hongtao.liu@intel.com>
9338 * config/i386/avx512fp16intrin.h (_mm512_fmaddsub_ph):
9340 (_mm512_mask_fmaddsub_ph): Likewise.
9341 (_mm512_mask3_fmaddsub_ph): Likewise.
9342 (_mm512_maskz_fmaddsub_ph): Likewise.
9343 (_mm512_fmaddsub_round_ph): Likewise.
9344 (_mm512_mask_fmaddsub_round_ph): Likewise.
9345 (_mm512_mask3_fmaddsub_round_ph): Likewise.
9346 (_mm512_maskz_fmaddsub_round_ph): Likewise.
9347 (_mm512_mask_fmsubadd_ph): Likewise.
9348 (_mm512_mask3_fmsubadd_ph): Likewise.
9349 (_mm512_maskz_fmsubadd_ph): Likewise.
9350 (_mm512_fmsubadd_round_ph): Likewise.
9351 (_mm512_mask_fmsubadd_round_ph): Likewise.
9352 (_mm512_mask3_fmsubadd_round_ph): Likewise.
9353 (_mm512_maskz_fmsubadd_round_ph): Likewise.
9354 * config/i386/avx512fp16vlintrin.h (_mm256_fmaddsub_ph):
9356 (_mm256_mask_fmaddsub_ph): Likewise.
9357 (_mm256_mask3_fmaddsub_ph): Likewise.
9358 (_mm256_maskz_fmaddsub_ph): Likewise.
9359 (_mm_fmaddsub_ph): Likewise.
9360 (_mm_mask_fmaddsub_ph): Likewise.
9361 (_mm_mask3_fmaddsub_ph): Likewise.
9362 (_mm_maskz_fmaddsub_ph): Likewise.
9363 (_mm256_fmsubadd_ph): Likewise.
9364 (_mm256_mask_fmsubadd_ph): Likewise.
9365 (_mm256_mask3_fmsubadd_ph): Likewise.
9366 (_mm256_maskz_fmsubadd_ph): Likewise.
9367 (_mm_fmsubadd_ph): Likewise.
9368 (_mm_mask_fmsubadd_ph): Likewise.
9369 (_mm_mask3_fmsubadd_ph): Likewise.
9370 (_mm_maskz_fmsubadd_ph): Likewise.
9371 * config/i386/i386-builtin.def: Add corresponding new builtins.
9372 * config/i386/sse.md (VFH_SF_AVX512VL): New mode iterator.
9373 * (<avx512>_fmsubadd_<mode>_maskz<round_expand_name>): New expander.
9374 * (<avx512>_fmaddsub_<mode>_maskz<round_expand_name>): Use
9376 * (<sd_mask_codefor>fma_fmaddsub_<mode><sd_maskz_name><round_name>):
9378 * (<avx512>_fmaddsub_<mode>_mask<round_name>): Ditto.
9379 * (<avx512>_fmaddsub_<mode>_mask3<round_name>): Ditto.
9380 * (<sd_mask_codefor>fma_fmsubadd_<mode><sd_maskz_name><round_name>):
9382 * (<avx512>_fmsubadd_<mode>_mask<round_name>): Ditto.
9383 * (<avx512>_fmsubadd_<mode>_mask3<round_name>): Ditto.
9385 2021-09-18 liuhongt <hongtao.liu@intel.com>
9388 * config/i386/i386.c (ix86_print_operand): Handle
9389 V8HF/V16HF/V32HFmode.
9390 * config/i386/i386.h (VALID_BCST_MODE_P): Add HFmode.
9391 * config/i386/sse.md (avx512bcst): Remove.
9393 2021-09-17 Martin Sebor <msebor@redhat.com>
9395 * Makefile.in (OBJS): Add gimple-predicate-analysis.o.
9396 * tree-ssa-uninit.c (max_phi_args): Move to gimple-predicate-analysis.
9397 (MASK_SET_BIT, MASK_TEST_BIT, MASK_EMPTY): Same.
9398 (check_defs): Add comment.
9399 (can_skip_redundant_opnd): Update comment.
9400 (compute_uninit_opnds_pos): Adjust to namespace change.
9401 (find_pdom): Move to gimple-predicate-analysis.cc.
9403 (struct uninit_undef_val_t): New.
9404 (is_non_loop_exit_postdominating): Move to gimple-predicate-analysis.cc.
9405 (find_control_equiv_block): Same.
9406 (MAX_NUM_CHAINS, MAX_CHAIN_LEN, MAX_POSTDOM_CHECK): Same.
9407 (MAX_SWITCH_CASES): Same.
9408 (compute_control_dep_chain): Same.
9409 (find_uninit_use): Use predicate analyzer.
9410 (struct pred_info): Move to gimple-predicate-analysis.
9411 (convert_control_dep_chain_into_preds): Same.
9412 (find_predicates): Same.
9413 (collect_phi_def_edges): Same.
9414 (warn_uninitialized_phi): Use predicate analyzer.
9415 (find_def_preds): Move to gimple-predicate-analysis.
9416 (dump_pred_info): Same.
9417 (dump_pred_chain): Same.
9418 (dump_predicates): Same.
9419 (destroy_predicate_vecs): Remove.
9420 (execute_late_warn_uninitialized): New.
9421 (get_cmp_code): Move to gimple-predicate-analysis.
9422 (is_value_included_in): Same.
9423 (value_sat_pred_p): Same.
9424 (find_matching_predicate_in_rest_chains): Same.
9425 (is_use_properly_guarded): Same.
9426 (prune_uninit_phi_opnds): Same.
9427 (find_var_cmp_const): Same.
9428 (use_pred_not_overlap_with_undef_path_pred): Same.
9429 (pred_equal_p): Same.
9430 (is_neq_relop_p): Same.
9431 (is_neq_zero_form_p): Same.
9432 (pred_expr_equal_p): Same.
9433 (is_pred_expr_subset_of): Same.
9434 (is_pred_chain_subset_of): Same.
9435 (is_included_in): Same.
9436 (is_superset_of): Same.
9438 (simplify_pred): Same.
9439 (simplify_preds_2): Same.
9440 (simplify_preds_3): Same.
9441 (simplify_preds_4): Same.
9442 (simplify_preds): Same.
9444 (push_to_worklist): Same.
9445 (get_pred_info_from_cmp): Same.
9446 (is_degenerated_phi): Same.
9447 (normalize_one_pred_1): Same.
9448 (normalize_one_pred): Same.
9449 (normalize_one_pred_chain): Same.
9450 (normalize_preds): Same.
9451 (can_one_predicate_be_invalidated_p): Same.
9452 (can_chain_union_be_invalidated_p): Same.
9453 (uninit_uses_cannot_happen): Same.
9454 (pass_late_warn_uninitialized::execute): Define.
9455 * gimple-predicate-analysis.cc: New file.
9456 * gimple-predicate-analysis.h: New file.
9458 2021-09-17 Julian Brown <julian@codesourcery.com>
9460 * config/gcn/gcn.c (gimple.h): Include.
9461 (gcn_fork_join): Emit barrier for worker-level joins.
9462 * omp-oacc-neuter-broadcast.cc (find_local_vars_to_propagate): Add
9463 writes_gang_private bitmap parameter. Set bit for blocks
9464 containing gang-private variable writes.
9465 (worker_single_simple): Don't emit barrier after predicated block.
9466 (worker_single_copy): Don't emit barrier if we're not broadcasting
9467 anything and the block contains no gang-private writes.
9468 (neuter_worker_single): Don't predicate blocks that only contain
9469 NOPs or internal marker functions. Pass has_gang_private_write
9470 argument to worker_single_copy.
9471 (oacc_do_neutering): Add writes_gang_private bitmap handling.
9473 2021-09-17 Julian Brown <julian@codesourcery.com>
9475 * config/gcn/gcn-protos.h
9476 (gcn_goacc_create_worker_broadcast_record): Update prototype.
9477 * config/gcn/gcn-tree.c (gcn_goacc_get_worker_red_decl): Use
9478 preallocated block of LDS memory. Do not cache/share decls for
9479 reduction temporaries between invocations.
9480 (gcn_goacc_reduction_teardown): Unshare VAR on second use.
9481 (gcn_goacc_create_worker_broadcast_record): Add OFFSET parameter
9482 and return temporary LDS space at that offset. Return pointer in
9484 * config/gcn/gcn.c (acc_lds_size, gang_private_hwm, lds_allocs):
9486 (ACC_LDS_SIZE): Define as acc_lds_size.
9487 (gcn_init_machine_status): Don't initialise lds_allocated,
9488 lds_allocs, reduc_decls fields of machine function struct.
9489 (gcn_option_override): Handle default size for gang-private
9490 variables and -mgang-private-size option.
9491 (gcn_expand_prologue): Use LDS_SIZE instead of LDS_SIZE-1 when
9492 initialising M0_REG.
9493 (gcn_shared_mem_layout): New function.
9494 (gcn_print_lds_decl): Update comment. Use global lds_allocs map and
9495 gang_private_hwm variable.
9496 (TARGET_GOACC_SHARED_MEM_LAYOUT): Define target hook.
9497 * config/gcn/gcn.h (machine_function): Remove lds_allocated,
9498 lds_allocs, reduc_decls. Add reduction_base, reduction_limit.
9499 * config/gcn/gcn.opt (gang_private_size_opt): New global.
9500 (mgang-private-size=): New option.
9501 * doc/tm.texi.in (TARGET_GOACC_SHARED_MEM_LAYOUT): Place
9503 * doc/tm.texi: Regenerate.
9504 * omp-oacc-neuter-broadcast.cc (targhooks.h, diagnostic-core.h):
9506 (build_sender_ref): Handle sender_decl being pointer.
9507 (worker_single_copy): Add PLACEMENT and ISOLATE_BROADCASTS
9508 parameters. Pass placement argument to
9509 create_worker_broadcast_record hook invocations. Handle
9510 sender_decl being pointer and isolate_broadcasts inserting extra
9512 (blk_offset_map_t): Add typedef.
9513 (neuter_worker_single): Add BLK_OFFSET_MAP parameter. Pass
9514 preallocated range to worker_single_copy call.
9515 (dfs_broadcast_reachable_1): New function.
9516 (idx_decl_pair_t, used_range_vec_t): New typedefs.
9517 (sort_size_descending): New function.
9518 (addr_range): New class.
9519 (splay_tree_compare_addr_range, splay_tree_free_key)
9520 (first_fit_range, merge_ranges_1, merge_ranges): New functions.
9521 (execute_omp_oacc_neuter_broadcast): Rename to...
9522 (oacc_do_neutering): ... this. Add BOUNDS_LO, BOUNDS_HI
9523 parameters. Arrange layout of shared memory for broadcast
9525 (execute_omp_oacc_neuter_broadcast): New function.
9526 (pass_omp_oacc_neuter_broadcast::gate): Remove num_workers==1
9527 handling from here. Enable pass for all OpenACC routines in order
9528 to call shared memory-layout hook.
9529 * target.def (create_worker_broadcast_record): Add OFFSET
9531 (shared_mem_layout): New hook.
9533 2021-09-17 Julian Brown <julian@codesourcery.com>
9534 Thomas Schwinge <thomas@codesourcery.com>
9536 * omp-oacc-neuter-broadcast.cc
9537 (pass_omp_oacc_neuter_broadcast::gate): Disable if num_workers is
9539 (execute_omp_oacc_neuter_broadcast): Adjust.
9541 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
9543 * value-relation.cc (class equiv_chain): Move to header file.
9544 (path_oracle::path_oracle): New.
9545 (path_oracle::~path_oracle): New.
9546 (path_oracle::register_relation): New.
9547 (path_oracle::query_relation): New.
9548 (path_oracle::reset_path): New.
9549 (path_oracle::dump): New.
9550 * value-relation.h (class equiv_chain): Move to here.
9551 (class path_oracle): New.
9553 2021-09-17 Andrew MacLeod <amacleod@redhat.com>
9555 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a DOM
9557 * gimple-range-fold.cc (fur_depend::register_relation): Use
9558 register_stmt/edge routines.
9559 * value-relation.cc (equiv_chain::find): Relocate from equiv_oracle.
9560 (equiv_oracle::equiv_oracle): Create self equivalence cache.
9561 (equiv_oracle::~equiv_oracle): Release same.
9562 (equiv_oracle::equiv_set): Return entry from self equiv cache if there
9563 are no equivalences.
9564 (equiv_oracle::find_equiv_block): Move list find to equiv_chain.
9565 (equiv_oracle::register_relation): Rename from register_equiv.
9566 (relation_chain_head::find_relation): Relocate from dom_oracle.
9567 (relation_oracle::register_stmt): New.
9568 (relation_oracle::register_edge): New.
9569 (dom_oracle::*): Rename from relation_oracle.
9570 (dom_oracle::register_relation): Adjust to call equiv_oracle.
9571 (dom_oracle::set_one_relation): Split from register_relation.
9572 (dom_oracle::register_transitives): Consolidate 2 methods.
9573 (dom_oracle::find_relation_block): Move core to relation_chain.
9574 (dom_oracle::query_relation): Rename from find_relation_dom and adjust.
9575 * value-relation.h (class relation_oracle): New pure virtual base.
9576 (class equiv_oracle): Inherit from relation_oracle and adjust.
9577 (class dom_oracle): Rename from old relation_oracle and adjust.
9579 2021-09-17 Martin Sebor <msebor@redhat.com>
9581 PR middle-end/102200
9582 * pointer-query.cc (access_ref::inform_access): Handle MIN/MAX_EXPR.
9583 (handle_min_max_size): Change argument. Store original SSA_NAME for
9584 operands to potentially distinct (sub)objects.
9585 (compute_objsize_r): Adjust call to the above.
9587 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
9589 * config/rs6000/rs6000.c (rs6000-builtins.h): New include.
9590 (rs6000_new_builtin_vectorized_function): New function.
9591 (rs6000_new_builtin_md_vectorized_function): Likewise.
9592 (rs6000_builtin_vectorized_function): Call
9593 rs6000_new_builtin_vectorized_function.
9594 (rs6000_builtin_md_vectorized_function): Call
9595 rs6000_new_builtin_md_vectorized_function.
9597 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
9599 * config/rs6000/rs6000-builtin-new.def (ASSEMBLE_ACC): Add mmaint flag.
9600 (ASSEMBLE_PAIR): Likewise.
9601 (BUILD_ACC): Likewise.
9602 (DISASSEMBLE_ACC): Likewise.
9603 (DISASSEMBLE_PAIR): Likewise.
9604 (PMXVBF16GER2): Likewise.
9605 (PMXVBF16GER2NN): Likewise.
9606 (PMXVBF16GER2NP): Likewise.
9607 (PMXVBF16GER2PN): Likewise.
9608 (PMXVBF16GER2PP): Likewise.
9609 (PMXVF16GER2): Likewise.
9610 (PMXVF16GER2NN): Likewise.
9611 (PMXVF16GER2NP): Likewise.
9612 (PMXVF16GER2PN): Likewise.
9613 (PMXVF16GER2PP): Likewise.
9614 (PMXVF32GER): Likewise.
9615 (PMXVF32GERNN): Likewise.
9616 (PMXVF32GERNP): Likewise.
9617 (PMXVF32GERPN): Likewise.
9618 (PMXVF32GERPP): Likewise.
9619 (PMXVF64GER): Likewise.
9620 (PMXVF64GERNN): Likewise.
9621 (PMXVF64GERNP): Likewise.
9622 (PMXVF64GERPN): Likewise.
9623 (PMXVF64GERPP): Likewise.
9624 (PMXVI16GER2): Likewise.
9625 (PMXVI16GER2PP): Likewise.
9626 (PMXVI16GER2S): Likewise.
9627 (PMXVI16GER2SPP): Likewise.
9628 (PMXVI4GER8): Likewise.
9629 (PMXVI4GER8PP): Likewise.
9630 (PMXVI8GER4): Likewise.
9631 (PMXVI8GER4PP): Likewise.
9632 (PMXVI8GER4SPP): Likewise.
9633 (XVBF16GER2): Likewise.
9634 (XVBF16GER2NN): Likewise.
9635 (XVBF16GER2NP): Likewise.
9636 (XVBF16GER2PN): Likewise.
9637 (XVBF16GER2PP): Likewise.
9638 (XVF16GER2): Likewise.
9639 (XVF16GER2NN): Likewise.
9640 (XVF16GER2NP): Likewise.
9641 (XVF16GER2PN): Likewise.
9642 (XVF16GER2PP): Likewise.
9643 (XVF32GER): Likewise.
9644 (XVF32GERNN): Likewise.
9645 (XVF32GERNP): Likewise.
9646 (XVF32GERPN): Likewise.
9647 (XVF32GERPP): Likewise.
9648 (XVF64GER): Likewise.
9649 (XVF64GERNN): Likewise.
9650 (XVF64GERNP): Likewise.
9651 (XVF64GERPN): Likewise.
9652 (XVF64GERPP): Likewise.
9653 (XVI16GER2): Likewise.
9654 (XVI16GER2PP): Likewise.
9655 (XVI16GER2S): Likewise.
9656 (XVI16GER2SPP): Likewise.
9657 (XVI4GER8): Likewise.
9658 (XVI4GER8PP): Likewise.
9659 (XVI8GER4): Likewise.
9660 (XVI8GER4PP): Likewise.
9661 (XVI8GER4SPP): Likewise.
9662 (XXMFACC): Likewise.
9663 (XXMTACC): Likewise.
9664 (XXSETACCZ): Likewise.
9665 (ASSEMBLE_PAIR_V): Likewise.
9666 (BUILD_PAIR): Likewise.
9667 (DISASSEMBLE_PAIR_V): Likewise.
9670 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_mma_builtin):
9671 Handle RS6000_BIF_LXVP and RS6000_BIF_STXVP.
9672 * config/rs6000/rs6000-gen-builtins.c (attrinfo): Add ismmaint.
9673 (parse_bif_attrs): Handle ismmaint.
9674 (write_decls): Add bif_mmaint_bit and bif_is_mmaint.
9675 (write_bif_static_init): Handle ismmaint.
9677 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
9679 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_new_builtin): New
9681 (rs6000_gimple_fold_builtin): Call rs6000_gimple_fold_new_builtin.
9682 (rs6000_new_builtin_valid_without_lhs): New function.
9683 (rs6000_gimple_fold_new_mma_builtin): Likewise.
9684 (rs6000_gimple_fold_new_builtin): Likewise.
9686 2021-09-17 Thomas Schwinge <thomas@codesourcery.com>
9688 * hash-table.h (hash_table<Descriptor, Lazy, Allocator>::expand):
9689 Destruct stale Value objects.
9690 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
9693 2021-09-17 Roger Sayle <roger@nextmovesoftware.com>
9696 * match.pd (shift optimizations): Disable recent sign-changing
9697 optimization for shifts by zero, these will be folded later.
9699 2021-09-17 Bill Schmidt <wschmidt@linux.ibm.com>
9701 * config/rs6000/rs6000-builtin-new.def (__builtin_mffsl): Move from
9702 [power9] to [always].
9704 2021-09-17 Richard Biener <rguenther@suse.de>
9706 * tree-vect-stmts.c (vectorizable_load): Do not frob
9709 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
9711 * config/i386/i386-features.c (remove_partial_avx_dependency):
9712 Also check TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY and
9713 and TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY before generating
9715 * config/i386/i386.h (TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY):
9717 (TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
9718 * config/i386/i386.md (SSE FP to FP splitters): Replace
9719 TARGET_SSE_PARTIAL_REG_DEPENDENCY with
9720 TARGET_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY.
9721 (SSE INT to FP splitter): Replace TARGET_SSE_PARTIAL_REG_DEPENDENCY
9722 with TARGET_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY.
9723 * config/i386/x86-tune.def
9724 (X86_TUNE_SSE_PARTIAL_REG_FP_CONVERTS_DEPENDENCY): New.
9725 (X86_TUNE_SSE_PARTIAL_REG_CONVERTS_DEPENDENCY): Likewise.
9727 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
9730 * config/i386/i386-features.c (remove_partial_avx_dependency):
9731 Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
9732 before generating vxorps.
9734 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
9736 * config/i386/i386-options.c (processor_cost_table): Use
9737 tremont_cost for Tremont.
9738 * config/i386/x86-tune-costs.h (tremont_memcpy): New.
9739 (tremont_memset): Likewise.
9740 (tremont_cost): Likewise.
9741 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
9744 2021-09-17 H.J. Lu <hjl.tools@gmail.com>
9746 * common/config/i386/i386-common.c: Use Haswell scheduling model
9748 * config/i386/i386.c (ix86_sched_init_global): Prepare for Tremont
9750 * config/i386/x86-tune-sched.c (ix86_issue_rate): Change Tremont
9752 (ix86_adjust_cost): Handle Tremont.
9753 * config/i386/x86-tune.def (X86_TUNE_SSE_PARTIAL_REG_DEPENDENCY):
9755 (X86_TUNE_USE_LEAVE): Likewise.
9756 (X86_TUNE_PUSH_MEMORY): Likewise.
9757 (X86_TUNE_MISALIGNED_MOVE_STRING_PRO_EPILOGUES): Likewise.
9758 (X86_TUNE_USE_CLTD): Likewise.
9759 (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI): Likewise.
9760 (X86_TUNE_AVOID_MFENCE): Likewise.
9761 (X86_TUNE_SSE_TYPELESS_STORES): Likewise.
9762 (X86_TUNE_SSE_LOAD0_BY_PXOR): Likewise.
9763 (X86_TUNE_ACCUMULATE_OUTGOING_ARGS): Disable for Tremont.
9764 (X86_TUNE_FOUR_JUMP_LIMIT): Likewise.
9765 (X86_TUNE_OPT_AGU): Likewise.
9766 (X86_TUNE_AVOID_LEA_FOR_ADDR): Likewise.
9767 (X86_TUNE_AVOID_MEM_OPND_FOR_CMOVE): Likewise.
9768 (X86_TUNE_EXPAND_ABS): Likewise.
9769 (X86_TUNE_SPLIT_MEM_OPND_FOR_FP_CONVERTS): Likewise.
9770 (X86_TUNE_SLOW_PSHUFB): Likewise.
9772 2021-09-17 Eric Botcazou <ebotcazou@adacore.com>
9774 PR rtl-optimization/102306
9775 * combine.c (try_combine): Abort the combination if we are about to
9776 duplicate volatile references.
9778 2021-09-17 liuhongt <hongtao.liu@intel.com>
9780 * config/i386/avx512fp16intrin.h (_mm_undefined_ph):
9782 (_mm256_undefined_ph): Likewise.
9783 (_mm512_undefined_ph): Likewise.
9784 (_mm_cvtsh_h): Likewise.
9785 (_mm256_cvtsh_h): Likewise.
9786 (_mm512_cvtsh_h): Likewise.
9787 (_mm512_castph_ps): Likewise.
9788 (_mm512_castph_pd): Likewise.
9789 (_mm512_castph_si512): Likewise.
9790 (_mm512_castph512_ph128): Likewise.
9791 (_mm512_castph512_ph256): Likewise.
9792 (_mm512_castph128_ph512): Likewise.
9793 (_mm512_castph256_ph512): Likewise.
9794 (_mm512_zextph128_ph512): Likewise.
9795 (_mm512_zextph256_ph512): Likewise.
9796 (_mm512_castps_ph): Likewise.
9797 (_mm512_castpd_ph): Likewise.
9798 (_mm512_castsi512_ph): Likewise.
9799 * config/i386/avx512fp16vlintrin.h (_mm_castph_ps):
9801 (_mm256_castph_ps): Likewise.
9802 (_mm_castph_pd): Likewise.
9803 (_mm256_castph_pd): Likewise.
9804 (_mm_castph_si128): Likewise.
9805 (_mm256_castph_si256): Likewise.
9806 (_mm_castps_ph): Likewise.
9807 (_mm256_castps_ph): Likewise.
9808 (_mm_castpd_ph): Likewise.
9809 (_mm256_castpd_ph): Likewise.
9810 (_mm_castsi128_ph): Likewise.
9811 (_mm256_castsi256_ph): Likewise.
9812 (_mm256_castph256_ph128): Likewise.
9813 (_mm256_castph128_ph256): Likewise.
9814 (_mm256_zextph128_ph256): Likewise.
9816 2021-09-17 liuhongt <hongtao.liu@intel.com>
9818 * config/i386/avx512fp16intrin.h (_mm_cvtsh_ss):
9820 (_mm_mask_cvtsh_ss): Likewise.
9821 (_mm_maskz_cvtsh_ss): Likewise.
9822 (_mm_cvtsh_sd): Likewise.
9823 (_mm_mask_cvtsh_sd): Likewise.
9824 (_mm_maskz_cvtsh_sd): Likewise.
9825 (_mm_cvt_roundsh_ss): Likewise.
9826 (_mm_mask_cvt_roundsh_ss): Likewise.
9827 (_mm_maskz_cvt_roundsh_ss): Likewise.
9828 (_mm_cvt_roundsh_sd): Likewise.
9829 (_mm_mask_cvt_roundsh_sd): Likewise.
9830 (_mm_maskz_cvt_roundsh_sd): Likewise.
9831 (_mm_cvtss_sh): Likewise.
9832 (_mm_mask_cvtss_sh): Likewise.
9833 (_mm_maskz_cvtss_sh): Likewise.
9834 (_mm_cvtsd_sh): Likewise.
9835 (_mm_mask_cvtsd_sh): Likewise.
9836 (_mm_maskz_cvtsd_sh): Likewise.
9837 (_mm_cvt_roundss_sh): Likewise.
9838 (_mm_mask_cvt_roundss_sh): Likewise.
9839 (_mm_maskz_cvt_roundss_sh): Likewise.
9840 (_mm_cvt_roundsd_sh): Likewise.
9841 (_mm_mask_cvt_roundsd_sh): Likewise.
9842 (_mm_maskz_cvt_roundsd_sh): Likewise.
9843 * config/i386/i386-builtin-types.def
9844 (V8HF_FTYPE_V2DF_V8HF_V8HF_UQI_INT,
9845 V8HF_FTYPE_V4SF_V8HF_V8HF_UQI_INT,
9846 V2DF_FTYPE_V8HF_V2DF_V2DF_UQI_INT,
9847 V4SF_FTYPE_V8HF_V4SF_V4SF_UQI_INT): Add new builtin types.
9848 * config/i386/i386-builtin.def: Add corrresponding new builtins.
9849 * config/i386/i386-expand.c: Handle new builtin types.
9850 * config/i386/sse.md (VF48_128): New mode iterator.
9851 (avx512fp16_vcvtsh2<ssescalarmodesuffix><mask_scalar_name><round_saeonly_scalar_name>):
9853 (avx512fp16_vcvt<ssescalarmodesuffix>2sh<mask_scalar_name><round_scalar_name>):
9856 2021-09-17 liuhongt <hongtao.liu@intel.com>
9858 * config/i386/avx512fp16intrin.h (_mm512_cvtph_pd):
9860 (_mm512_mask_cvtph_pd): Likewise.
9861 (_mm512_maskz_cvtph_pd): Likewise.
9862 (_mm512_cvt_roundph_pd): Likewise.
9863 (_mm512_mask_cvt_roundph_pd): Likewise.
9864 (_mm512_maskz_cvt_roundph_pd): Likewise.
9865 (_mm512_cvtxph_ps): Likewise.
9866 (_mm512_mask_cvtxph_ps): Likewise.
9867 (_mm512_maskz_cvtxph_ps): Likewise.
9868 (_mm512_cvtx_roundph_ps): Likewise.
9869 (_mm512_mask_cvtx_roundph_ps): Likewise.
9870 (_mm512_maskz_cvtx_roundph_ps): Likewise.
9871 (_mm512_cvtxps_ph): Likewise.
9872 (_mm512_mask_cvtxps_ph): Likewise.
9873 (_mm512_maskz_cvtxps_ph): Likewise.
9874 (_mm512_cvtx_roundps_ph): Likewise.
9875 (_mm512_mask_cvtx_roundps_ph): Likewise.
9876 (_mm512_maskz_cvtx_roundps_ph): Likewise.
9877 (_mm512_cvtpd_ph): Likewise.
9878 (_mm512_mask_cvtpd_ph): Likewise.
9879 (_mm512_maskz_cvtpd_ph): Likewise.
9880 (_mm512_cvt_roundpd_ph): Likewise.
9881 (_mm512_mask_cvt_roundpd_ph): Likewise.
9882 (_mm512_maskz_cvt_roundpd_ph): Likewise.
9883 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_pd):
9885 (_mm_mask_cvtph_pd): Likewise.
9886 (_mm_maskz_cvtph_pd): Likewise.
9887 (_mm256_cvtph_pd): Likewise.
9888 (_mm256_mask_cvtph_pd): Likewise.
9889 (_mm256_maskz_cvtph_pd): Likewise.
9890 (_mm_cvtxph_ps): Likewise.
9891 (_mm_mask_cvtxph_ps): Likewise.
9892 (_mm_maskz_cvtxph_ps): Likewise.
9893 (_mm256_cvtxph_ps): Likewise.
9894 (_mm256_mask_cvtxph_ps): Likewise.
9895 (_mm256_maskz_cvtxph_ps): Likewise.
9896 (_mm_cvtxps_ph): Likewise.
9897 (_mm_mask_cvtxps_ph): Likewise.
9898 (_mm_maskz_cvtxps_ph): Likewise.
9899 (_mm256_cvtxps_ph): Likewise.
9900 (_mm256_mask_cvtxps_ph): Likewise.
9901 (_mm256_maskz_cvtxps_ph): Likewise.
9902 (_mm_cvtpd_ph): Likewise.
9903 (_mm_mask_cvtpd_ph): Likewise.
9904 (_mm_maskz_cvtpd_ph): Likewise.
9905 (_mm256_cvtpd_ph): Likewise.
9906 (_mm256_mask_cvtpd_ph): Likewise.
9907 (_mm256_maskz_cvtpd_ph): Likewise.
9908 * config/i386/i386-builtin.def: Add corresponding new builtins.
9909 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
9910 * config/i386/i386-expand.c: Handle new builtin types.
9911 * config/i386/sse.md
9912 (VF4_128_8_256): New.
9913 (VF48H_AVX512VL): Ditto.
9914 (ssePHmode): Add HF vector modes.
9915 (castmode): Add new convertable modes.
9918 (avx512fp16_vcvt<castmode>2ph_<mode><mask_name><round_name>): Ditto.
9919 (avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
9920 (*avx512fp16_vcvt<castmode>2ph_<mode>): Ditto.
9921 (avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
9922 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask): Ditto.
9923 (*avx512fp16_vcvt<castmode>2ph_<mode>_mask_1): Ditto.
9924 (avx512fp16_float_extend_ph<mode>2<mask_name><round_saeonly_name>):
9926 (avx512fp16_float_extend_ph<mode>2<mask_name>): Ditto.
9927 (*avx512fp16_float_extend_ph<mode>2_load<mask_name>): Ditto.
9928 (avx512fp16_float_extend_phv2df2<mask_name>): Ditto.
9929 (*avx512fp16_float_extend_phv2df2_load<mask_name>): Ditto.
9931 2021-09-17 liuhongt <hongtao.liu@intel.com>
9933 * config/i386/avx512fp16intrin.h (_mm_cvttsh_i32):
9935 (_mm_cvttsh_u32): Likewise.
9936 (_mm_cvtt_roundsh_i32): Likewise.
9937 (_mm_cvtt_roundsh_u32): Likewise.
9938 (_mm_cvttsh_i64): Likewise.
9939 (_mm_cvttsh_u64): Likewise.
9940 (_mm_cvtt_roundsh_i64): Likewise.
9941 (_mm_cvtt_roundsh_u64): Likewise.
9942 * config/i386/i386-builtin.def: Add corresponding new builtins.
9943 * config/i386/sse.md
9944 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<round_saeonly_name>):
9947 2021-09-17 liuhongt <hongtao.liu@intel.com>
9949 * config/i386/avx512fp16intrin.h (_mm512_cvttph_epi32):
9951 (_mm512_mask_cvttph_epi32): Likewise.
9952 (_mm512_maskz_cvttph_epi32): Likewise.
9953 (_mm512_cvtt_roundph_epi32): Likewise.
9954 (_mm512_mask_cvtt_roundph_epi32): Likewise.
9955 (_mm512_maskz_cvtt_roundph_epi32): Likewise.
9956 (_mm512_cvttph_epu32): Likewise.
9957 (_mm512_mask_cvttph_epu32): Likewise.
9958 (_mm512_maskz_cvttph_epu32): Likewise.
9959 (_mm512_cvtt_roundph_epu32): Likewise.
9960 (_mm512_mask_cvtt_roundph_epu32): Likewise.
9961 (_mm512_maskz_cvtt_roundph_epu32): Likewise.
9962 (_mm512_cvttph_epi64): Likewise.
9963 (_mm512_mask_cvttph_epi64): Likewise.
9964 (_mm512_maskz_cvttph_epi64): Likewise.
9965 (_mm512_cvtt_roundph_epi64): Likewise.
9966 (_mm512_mask_cvtt_roundph_epi64): Likewise.
9967 (_mm512_maskz_cvtt_roundph_epi64): Likewise.
9968 (_mm512_cvttph_epu64): Likewise.
9969 (_mm512_mask_cvttph_epu64): Likewise.
9970 (_mm512_maskz_cvttph_epu64): Likewise.
9971 (_mm512_cvtt_roundph_epu64): Likewise.
9972 (_mm512_mask_cvtt_roundph_epu64): Likewise.
9973 (_mm512_maskz_cvtt_roundph_epu64): Likewise.
9974 (_mm512_cvttph_epi16): Likewise.
9975 (_mm512_mask_cvttph_epi16): Likewise.
9976 (_mm512_maskz_cvttph_epi16): Likewise.
9977 (_mm512_cvtt_roundph_epi16): Likewise.
9978 (_mm512_mask_cvtt_roundph_epi16): Likewise.
9979 (_mm512_maskz_cvtt_roundph_epi16): Likewise.
9980 (_mm512_cvttph_epu16): Likewise.
9981 (_mm512_mask_cvttph_epu16): Likewise.
9982 (_mm512_maskz_cvttph_epu16): Likewise.
9983 (_mm512_cvtt_roundph_epu16): Likewise.
9984 (_mm512_mask_cvtt_roundph_epu16): Likewise.
9985 (_mm512_maskz_cvtt_roundph_epu16): Likewise.
9986 * config/i386/avx512fp16vlintrin.h (_mm_cvttph_epi32):
9988 (_mm_mask_cvttph_epi32): Likewise.
9989 (_mm_maskz_cvttph_epi32): Likewise.
9990 (_mm256_cvttph_epi32): Likewise.
9991 (_mm256_mask_cvttph_epi32): Likewise.
9992 (_mm256_maskz_cvttph_epi32): Likewise.
9993 (_mm_cvttph_epu32): Likewise.
9994 (_mm_mask_cvttph_epu32): Likewise.
9995 (_mm_maskz_cvttph_epu32): Likewise.
9996 (_mm256_cvttph_epu32): Likewise.
9997 (_mm256_mask_cvttph_epu32): Likewise.
9998 (_mm256_maskz_cvttph_epu32): Likewise.
9999 (_mm_cvttph_epi64): Likewise.
10000 (_mm_mask_cvttph_epi64): Likewise.
10001 (_mm_maskz_cvttph_epi64): Likewise.
10002 (_mm256_cvttph_epi64): Likewise.
10003 (_mm256_mask_cvttph_epi64): Likewise.
10004 (_mm256_maskz_cvttph_epi64): Likewise.
10005 (_mm_cvttph_epu64): Likewise.
10006 (_mm_mask_cvttph_epu64): Likewise.
10007 (_mm_maskz_cvttph_epu64): Likewise.
10008 (_mm256_cvttph_epu64): Likewise.
10009 (_mm256_mask_cvttph_epu64): Likewise.
10010 (_mm256_maskz_cvttph_epu64): Likewise.
10011 (_mm_cvttph_epi16): Likewise.
10012 (_mm_mask_cvttph_epi16): Likewise.
10013 (_mm_maskz_cvttph_epi16): Likewise.
10014 (_mm256_cvttph_epi16): Likewise.
10015 (_mm256_mask_cvttph_epi16): Likewise.
10016 (_mm256_maskz_cvttph_epi16): Likewise.
10017 (_mm_cvttph_epu16): Likewise.
10018 (_mm_mask_cvttph_epu16): Likewise.
10019 (_mm_maskz_cvttph_epu16): Likewise.
10020 (_mm256_cvttph_epu16): Likewise.
10021 (_mm256_mask_cvttph_epu16): Likewise.
10022 (_mm256_maskz_cvttph_epu16): Likewise.
10023 * config/i386/i386-builtin.def: Add new builtins.
10024 * config/i386/sse.md
10025 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name><round_saeonly_name>):
10027 (avx512fp16_fix<fixunssuffix>_trunc<mode>2<mask_name>): Ditto.
10028 (*avx512fp16_fix<fixunssuffix>_trunc<mode>2_load<mask_name>): Ditto.
10029 (avx512fp16_fix<fixunssuffix>_truncv2di2<mask_name>): Ditto.
10030 (avx512fp16_fix<fixunssuffix>_truncv2di2_load<mask_name>): Ditto.
10032 2021-09-17 liuhongt <hongtao.liu@intel.com>
10034 * config/i386/avx512fp16intrin.h (_mm_cvtsh_i32): New intrinsic.
10035 (_mm_cvtsh_u32): Likewise.
10036 (_mm_cvt_roundsh_i32): Likewise.
10037 (_mm_cvt_roundsh_u32): Likewise.
10038 (_mm_cvtsh_i64): Likewise.
10039 (_mm_cvtsh_u64): Likewise.
10040 (_mm_cvt_roundsh_i64): Likewise.
10041 (_mm_cvt_roundsh_u64): Likewise.
10042 (_mm_cvti32_sh): Likewise.
10043 (_mm_cvtu32_sh): Likewise.
10044 (_mm_cvt_roundi32_sh): Likewise.
10045 (_mm_cvt_roundu32_sh): Likewise.
10046 (_mm_cvti64_sh): Likewise.
10047 (_mm_cvtu64_sh): Likewise.
10048 (_mm_cvt_roundi64_sh): Likewise.
10049 (_mm_cvt_roundu64_sh): Likewise.
10050 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
10051 * config/i386/i386-builtin.def: Add corresponding new builtins.
10052 * config/i386/i386-expand.c (ix86_expand_round_builtin):
10053 Handle new builtin types.
10054 * config/i386/sse.md
10055 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix><round_name>):
10057 (avx512fp16_vcvtsh2<sseintconvertsignprefix>si<rex64namesuffix>_2): Likewise.
10058 (avx512fp16_vcvt<floatsuffix>si2sh<rex64namesuffix><round_name>): Likewise.
10060 2021-09-16 Bill Schmidt <wschmidt@linux.ibm.com>
10062 * config/rs6000/rs6000-c.c (rs6000-builtins.h): New include.
10063 (altivec_resolve_new_overloaded_builtin): New forward decl.
10064 (rs6000_new_builtin_type_compatible): New function.
10065 (altivec_resolve_overloaded_builtin): Call
10066 altivec_resolve_new_overloaded_builtin.
10067 (altivec_build_new_resolved_builtin): New function.
10068 (altivec_resolve_new_overloaded_builtin): Likewise.
10069 * config/rs6000/rs6000-call.c (rs6000_new_builtin_is_supported):
10071 * config/rs6000/rs6000-gen-builtins.c (write_decls): Remove _p from
10072 name of rs6000_new_builtin_is_supported.
10074 2021-09-16 Uroš Bizjak <ubizjak@gmail.com>
10076 * config/i386/i386-protos.h (ix86_decompose_address):
10077 Change return type to bool.
10078 * config/i386/i386.c (ix86_decompose_address): Ditto.
10080 2021-09-16 Tobias Burnus <tobias@codesourcery.com>
10083 * config/rs6000/t-rs6000 (build/rs6000-gen-builtins.o, build/rbtree.o):
10084 Added 'build/' to target, use build/%.o rule.
10085 (build/rs6000-gen-builtins$(build_exeext)): Add 'build/' and
10086 '$(build_exeext)' to target and 'build/' for the *.o files.
10087 (rs6000-builtins.c): Update for those changes; run rs6000-gen-builtins
10090 2021-09-16 Martin Jambor <mjambor@suse.cz>
10092 * cgraph.c (cgraph_node::dump): Do not check caller count sums if
10093 the body has been removed. Remove trailing whitespace.
10095 2021-09-16 Richard Biener <rguenther@suse.de>
10097 PR middle-end/102360
10098 * internal-fn.c (expand_DEFERRED_INIT): Make pattern-init
10099 of non-memory more robust.
10101 2021-09-16 Daniel Cederman <cederman@gaisler.com>
10103 * config/sparc/sparc-opts.h (enum sparc_processor_type): Add LEON5
10104 * config/sparc/sparc.c (struct processor_costs): Add LEON5 costs
10105 (leon5_adjust_cost): Increase cost of store with data dependency
10106 on ALU instruction and FPU anti-dependencies.
10107 (sparc_option_override): Add LEON5 costs
10108 (sparc_adjust_cost): Add LEON5 cost adjustments
10109 * config/sparc/sparc.h: Add LEON5
10110 * config/sparc/sparc.md: Include LEON5 scheduling information
10111 * config/sparc/sparc.opt: Add LEON5
10112 * doc/invoke.texi: Add LEON5
10113 * config/sparc/leon5.md: New file.
10115 2021-09-16 Daniel Cederman <cederman@gaisler.com>
10117 * config/sparc/sparc.md (stack_protect_set32): Add NOP to prevent
10118 sensitive sequence for B2BST errata workaround.
10120 2021-09-16 Daniel Cederman <cederman@gaisler.com>
10122 * config/sparc/sparc.c (sparc_do_work_around_errata): Do not begin
10123 functions with atomic instruction in the UT700 errata workaround.
10125 2021-09-16 Daniel Cederman <cederman@gaisler.com>
10127 * config/sparc/sparc.c (next_active_non_empty_insn): New function
10128 that returns next active non empty assembly instruction.
10129 (sparc_do_work_around_errata): Use new function.
10131 2021-09-16 Daniel Cederman <cederman@gaisler.com>
10133 * config/sparc/sparc.c (store_insn_p): Add predicate for store
10135 (load_insn_p): Add predicate for load attributes.
10136 (sparc_do_work_around_errata): Use new predicates.
10138 2021-09-16 Andreas Larsson <andreas@gaisler.com>
10140 * config/sparc/sparc.c (dump_target_flag_bits): Print bit names for
10143 2021-09-16 Martin Liska <mliska@suse.cz>
10145 * config/mips/netbsd.h: Fix typo in name of a macro.
10147 2021-09-16 liuhongt <hongtao.liu@intel.com>
10149 PR middle-end/102080
10150 * match.pd: Check mask type when doing cond_op related gimple
10152 * tree.c (is_truth_type_for): New function.
10153 * tree.h (is_truth_type_for): New declaration.
10155 2021-09-16 liuhongt <hongtao.liu@intel.com>
10157 * config/i386/avx512fp16intrin.h (_mm512_cvtepi32_ph): New
10159 (_mm512_mask_cvtepi32_ph): Likewise.
10160 (_mm512_maskz_cvtepi32_ph): Likewise.
10161 (_mm512_cvt_roundepi32_ph): Likewise.
10162 (_mm512_mask_cvt_roundepi32_ph): Likewise.
10163 (_mm512_maskz_cvt_roundepi32_ph): Likewise.
10164 (_mm512_cvtepu32_ph): Likewise.
10165 (_mm512_mask_cvtepu32_ph): Likewise.
10166 (_mm512_maskz_cvtepu32_ph): Likewise.
10167 (_mm512_cvt_roundepu32_ph): Likewise.
10168 (_mm512_mask_cvt_roundepu32_ph): Likewise.
10169 (_mm512_maskz_cvt_roundepu32_ph): Likewise.
10170 (_mm512_cvtepi64_ph): Likewise.
10171 (_mm512_mask_cvtepi64_ph): Likewise.
10172 (_mm512_maskz_cvtepi64_ph): Likewise.
10173 (_mm512_cvt_roundepi64_ph): Likewise.
10174 (_mm512_mask_cvt_roundepi64_ph): Likewise.
10175 (_mm512_maskz_cvt_roundepi64_ph): Likewise.
10176 (_mm512_cvtepu64_ph): Likewise.
10177 (_mm512_mask_cvtepu64_ph): Likewise.
10178 (_mm512_maskz_cvtepu64_ph): Likewise.
10179 (_mm512_cvt_roundepu64_ph): Likewise.
10180 (_mm512_mask_cvt_roundepu64_ph): Likewise.
10181 (_mm512_maskz_cvt_roundepu64_ph): Likewise.
10182 (_mm512_cvtepi16_ph): Likewise.
10183 (_mm512_mask_cvtepi16_ph): Likewise.
10184 (_mm512_maskz_cvtepi16_ph): Likewise.
10185 (_mm512_cvt_roundepi16_ph): Likewise.
10186 (_mm512_mask_cvt_roundepi16_ph): Likewise.
10187 (_mm512_maskz_cvt_roundepi16_ph): Likewise.
10188 (_mm512_cvtepu16_ph): Likewise.
10189 (_mm512_mask_cvtepu16_ph): Likewise.
10190 (_mm512_maskz_cvtepu16_ph): Likewise.
10191 (_mm512_cvt_roundepu16_ph): Likewise.
10192 (_mm512_mask_cvt_roundepu16_ph): Likewise.
10193 (_mm512_maskz_cvt_roundepu16_ph): Likewise.
10194 * config/i386/avx512fp16vlintrin.h (_mm_cvtepi32_ph): New
10196 (_mm_mask_cvtepi32_ph): Likewise.
10197 (_mm_maskz_cvtepi32_ph): Likewise.
10198 (_mm256_cvtepi32_ph): Likewise.
10199 (_mm256_mask_cvtepi32_ph): Likewise.
10200 (_mm256_maskz_cvtepi32_ph): Likewise.
10201 (_mm_cvtepu32_ph): Likewise.
10202 (_mm_mask_cvtepu32_ph): Likewise.
10203 (_mm_maskz_cvtepu32_ph): Likewise.
10204 (_mm256_cvtepu32_ph): Likewise.
10205 (_mm256_mask_cvtepu32_ph): Likewise.
10206 (_mm256_maskz_cvtepu32_ph): Likewise.
10207 (_mm_cvtepi64_ph): Likewise.
10208 (_mm_mask_cvtepi64_ph): Likewise.
10209 (_mm_maskz_cvtepi64_ph): Likewise.
10210 (_mm256_cvtepi64_ph): Likewise.
10211 (_mm256_mask_cvtepi64_ph): Likewise.
10212 (_mm256_maskz_cvtepi64_ph): Likewise.
10213 (_mm_cvtepu64_ph): Likewise.
10214 (_mm_mask_cvtepu64_ph): Likewise.
10215 (_mm_maskz_cvtepu64_ph): Likewise.
10216 (_mm256_cvtepu64_ph): Likewise.
10217 (_mm256_mask_cvtepu64_ph): Likewise.
10218 (_mm256_maskz_cvtepu64_ph): Likewise.
10219 (_mm_cvtepi16_ph): Likewise.
10220 (_mm_mask_cvtepi16_ph): Likewise.
10221 (_mm_maskz_cvtepi16_ph): Likewise.
10222 (_mm256_cvtepi16_ph): Likewise.
10223 (_mm256_mask_cvtepi16_ph): Likewise.
10224 (_mm256_maskz_cvtepi16_ph): Likewise.
10225 (_mm_cvtepu16_ph): Likewise.
10226 (_mm_mask_cvtepu16_ph): Likewise.
10227 (_mm_maskz_cvtepu16_ph): Likewise.
10228 (_mm256_cvtepu16_ph): Likewise.
10229 (_mm256_mask_cvtepu16_ph): Likewise.
10230 (_mm256_maskz_cvtepu16_ph): Likewise.
10231 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
10232 * config/i386/i386-builtin.def: Add corresponding new builtins.
10233 * config/i386/i386-expand.c
10234 (ix86_expand_args_builtin): Handle new builtin types.
10235 (ix86_expand_round_builtin): Ditto.
10236 * config/i386/i386-modes.def: Declare V2HF and V6HF.
10237 * config/i386/sse.md (VI2H_AVX512VL): New.
10238 (qq2phsuff): Ditto.
10239 (sseintvecmode): Add HF vector modes.
10240 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode><mask_name><round_name>):
10242 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
10243 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>): Ditto.
10244 (avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
10245 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask): Ditto.
10246 (*avx512fp16_vcvt<floatsuffix><sseintconvert>2ph_<mode>_mask_1): Ditto.
10247 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
10248 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di): Ditto.
10249 (avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
10250 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask): Ditto.
10251 (*avx512fp16_vcvt<floatsuffix>qq2ph_v2di_mask_1): Ditto.
10252 * config/i386/subst.md (round_qq2phsuff): New subst_attr.
10254 2021-09-16 liuhongt <hongtao.liu@intel.com>
10256 * config/i386/avx512fp16intrin.h (_mm512_cvtph_epi32):
10258 (_mm512_mask_cvtph_epi32): Likewise.
10259 (_mm512_maskz_cvtph_epi32): Likewise.
10260 (_mm512_cvt_roundph_epi32): Likewise.
10261 (_mm512_mask_cvt_roundph_epi32): Likewise.
10262 (_mm512_maskz_cvt_roundph_epi32): Likewise.
10263 (_mm512_cvtph_epu32): Likewise.
10264 (_mm512_mask_cvtph_epu32): Likewise.
10265 (_mm512_maskz_cvtph_epu32): Likewise.
10266 (_mm512_cvt_roundph_epu32): Likewise.
10267 (_mm512_mask_cvt_roundph_epu32): Likewise.
10268 (_mm512_maskz_cvt_roundph_epu32): Likewise.
10269 (_mm512_cvtph_epi64): Likewise.
10270 (_mm512_mask_cvtph_epi64): Likewise.
10271 (_mm512_maskz_cvtph_epi64): Likewise.
10272 (_mm512_cvt_roundph_epi64): Likewise.
10273 (_mm512_mask_cvt_roundph_epi64): Likewise.
10274 (_mm512_maskz_cvt_roundph_epi64): Likewise.
10275 (_mm512_cvtph_epu64): Likewise.
10276 (_mm512_mask_cvtph_epu64): Likewise.
10277 (_mm512_maskz_cvtph_epu64): Likewise.
10278 (_mm512_cvt_roundph_epu64): Likewise.
10279 (_mm512_mask_cvt_roundph_epu64): Likewise.
10280 (_mm512_maskz_cvt_roundph_epu64): Likewise.
10281 (_mm512_cvtph_epi16): Likewise.
10282 (_mm512_mask_cvtph_epi16): Likewise.
10283 (_mm512_maskz_cvtph_epi16): Likewise.
10284 (_mm512_cvt_roundph_epi16): Likewise.
10285 (_mm512_mask_cvt_roundph_epi16): Likewise.
10286 (_mm512_maskz_cvt_roundph_epi16): Likewise.
10287 (_mm512_cvtph_epu16): Likewise.
10288 (_mm512_mask_cvtph_epu16): Likewise.
10289 (_mm512_maskz_cvtph_epu16): Likewise.
10290 (_mm512_cvt_roundph_epu16): Likewise.
10291 (_mm512_mask_cvt_roundph_epu16): Likewise.
10292 (_mm512_maskz_cvt_roundph_epu16): Likewise.
10293 * config/i386/avx512fp16vlintrin.h (_mm_cvtph_epi32):
10295 (_mm_mask_cvtph_epi32): Likewise.
10296 (_mm_maskz_cvtph_epi32): Likewise.
10297 (_mm256_cvtph_epi32): Likewise.
10298 (_mm256_mask_cvtph_epi32): Likewise.
10299 (_mm256_maskz_cvtph_epi32): Likewise.
10300 (_mm_cvtph_epu32): Likewise.
10301 (_mm_mask_cvtph_epu32): Likewise.
10302 (_mm_maskz_cvtph_epu32): Likewise.
10303 (_mm256_cvtph_epu32): Likewise.
10304 (_mm256_mask_cvtph_epu32): Likewise.
10305 (_mm256_maskz_cvtph_epu32): Likewise.
10306 (_mm_cvtph_epi64): Likewise.
10307 (_mm_mask_cvtph_epi64): Likewise.
10308 (_mm_maskz_cvtph_epi64): Likewise.
10309 (_mm256_cvtph_epi64): Likewise.
10310 (_mm256_mask_cvtph_epi64): Likewise.
10311 (_mm256_maskz_cvtph_epi64): Likewise.
10312 (_mm_cvtph_epu64): Likewise.
10313 (_mm_mask_cvtph_epu64): Likewise.
10314 (_mm_maskz_cvtph_epu64): Likewise.
10315 (_mm256_cvtph_epu64): Likewise.
10316 (_mm256_mask_cvtph_epu64): Likewise.
10317 (_mm256_maskz_cvtph_epu64): Likewise.
10318 (_mm_cvtph_epi16): Likewise.
10319 (_mm_mask_cvtph_epi16): Likewise.
10320 (_mm_maskz_cvtph_epi16): Likewise.
10321 (_mm256_cvtph_epi16): Likewise.
10322 (_mm256_mask_cvtph_epi16): Likewise.
10323 (_mm256_maskz_cvtph_epi16): Likewise.
10324 (_mm_cvtph_epu16): Likewise.
10325 (_mm_mask_cvtph_epu16): Likewise.
10326 (_mm_maskz_cvtph_epu16): Likewise.
10327 (_mm256_cvtph_epu16): Likewise.
10328 (_mm256_mask_cvtph_epu16): Likewise.
10329 (_mm256_maskz_cvtph_epu16): Likewise.
10330 * config/i386/i386-builtin-types.def: Add new builtin types.
10331 * config/i386/i386-builtin.def: Add new builtins.
10332 * config/i386/i386-expand.c
10333 (ix86_expand_args_builtin): Handle new builtin types.
10334 (ix86_expand_round_builtin): Ditto.
10335 * config/i386/sse.md (sseintconvert): New.
10336 (ssePHmode): Ditto.
10337 (UNSPEC_US_FIX_NOTRUNC): Ditto.
10338 (sseintconvertsignprefix): Ditto.
10339 (avx512fp16_vcvtph2<sseintconvertsignprefix><sseintconvert>_<mode><mask_name><round_name>):
10342 2021-09-16 liuhongt <hongtao.liu@intel.com>
10344 * config/i386/avx512fp16intrin.h: (_mm_cvtsi16_si128):
10346 (_mm_cvtsi128_si16): Likewise.
10347 (_mm_mask_load_sh): Likewise.
10348 (_mm_maskz_load_sh): Likewise.
10349 (_mm_mask_store_sh): Likewise.
10350 (_mm_move_sh): Likewise.
10351 (_mm_mask_move_sh): Likewise.
10352 (_mm_maskz_move_sh): Likewise.
10353 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
10354 * config/i386/i386-builtin.def: Add corresponding new builtins.
10355 * config/i386/i386-expand.c
10356 (ix86_expand_special_args_builtin): Handle new builtin types.
10357 (ix86_expand_vector_init_one_nonzero): Adjust for FP16 target.
10358 * config/i386/sse.md (VI2F): New mode iterator.
10359 (vec_set<mode>_0): Use new mode iterator.
10360 (avx512f_mov<ssescalarmodelower>_mask): Adjust for HF vector mode.
10361 (avx512f_store<mode>_mask): Ditto.
10363 2021-09-16 Kewen Lin <linkw@linux.ibm.com>
10365 * config/rs6000/rs6000.opt (-mtoc-fusion): Remove.
10367 2021-09-15 David Edelsohn <dje.gcc@gmail.com>
10369 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
10370 Proceed if no symbol summary or the symbol alias flag is false.
10372 2021-09-15 Jakub Jelinek <jakub@redhat.com>
10376 * varasm.c (output_constructor_regular_field): Instead of assertion
10377 that array_size_for_constructor result is equal to size of
10378 TREE_TYPE (local->val) in bytes, assert that the type size is greater
10379 or equal to array_size_for_constructor result and use type size as
10382 2021-09-15 Martin Liska <mliska@suse.cz>
10385 * config/i386/vxworks.h: Use new macro TARGET_CPU_P.
10387 2021-09-15 Martin Liska <mliska@suse.cz>
10390 * config/rs6000/rs6000.c (rs6000_xcoff_encode_section_info):
10391 Check that we have a symbol summary for a symbol.
10393 2021-09-15 Richard Biener <rguenther@suse.de>
10396 * config/rs6000/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
10397 to inherit from elfos.h
10399 2021-09-15 liuhongt <hongtao.liu@intel.com>
10402 * config/i386/i386-expand.c
10403 (ix86_expand_vector_init_interleave): Use puncklwd to pack 2
10405 (ix86_expand_vector_set): Use blendw instead of pinsrw.
10406 * config/i386/i386.c (ix86_can_change_mode_class): Adjust for
10407 AVX512FP16 which supports 16bit vector load.
10408 * config/i386/sse.md (avx512bw_interleave_highv32hi<mask_name>):
10410 (avx512bw_interleave_high<mode><mask_name>): .. this, and
10411 extend to V32HFmode.
10412 (avx2_interleave_highv16hi<mask_name>): Rename to ..
10413 (avx2_interleave_high<mode><mask_name>): .. this, and extend
10415 (vec_interleave_highv8hi<mask_name>): Rename to ..
10416 (vec_interleave_high<mode><mask_name>): .. this, and extend to V8HFmode.
10417 (<mask_codefor>avx512bw_interleave_lowv32hi<mask_name>):
10419 (<mask_codefor>avx512bw_interleave_low<mode><mask_name>):
10420 this, and extend to V32HFmode.
10421 (avx2_interleave_lowv16hi<mask_name>): Rename to ..
10422 (avx2_interleave_low<mode><mask_name>): .. this, and extend to V16HFmode.
10423 (vec_interleave_lowv8hi<mask_name>): Rename to ..
10424 (vec_interleave_low<mode><mask_name>): .. this, and extend to V8HFmode.
10425 (sse4_1_pblendw): Rename to ..
10426 (sse4_1_pblend<blendsuf>): .. this, and extend to V8HFmode.
10427 (avx2_pblendph): New define_expand.
10428 (<sse2p4_1>_pinsr<ssemodesuffix>): Refactor, use
10429 sseintmodesuffix instead of ssemodesuffix.
10430 (blendsuf): New mode attr.
10432 2021-09-15 Richard Biener <rguenther@suse.de>
10434 * tree-vectorizer.h (dr_misalignment): Move out of line.
10435 (dr_target_alignment): New.
10436 (DR_TARGET_ALIGNMENT): Wrap dr_target_alignment.
10437 (set_dr_target_alignment): New.
10438 (SET_DR_TARGET_ALIGNMENT): Wrap set_dr_target_alignment.
10439 * tree-vect-data-refs.c (dr_misalignment): Compute and
10440 return the group members misalignment.
10441 (vect_compute_data_ref_alignment): Use SET_DR_TARGET_ALIGNMENT.
10442 (vect_analyze_data_refs_alignment): Compute alignment only
10443 for the first element of a DR group.
10444 (vect_slp_analyze_node_alignment): Likewise.
10446 2021-09-15 Hongyu Wang <hongyu.wang@intel.com>
10448 * config/i386/avx512fp16intrin.h: Adjust all builtin calls.
10449 * config/i386/avx512fp16vlintrin.h: Likewise.
10450 * config/i386/i386-builtin.def: Adjust builtin name and
10451 enumeration to match AVX512F style.
10453 2021-09-15 Richard Biener <rguenther@suse.de>
10455 PR tree-optimization/102318
10456 * tree-vect-loop.c (vect_transform_cycle_phi): Revert
10457 previous change and do the mode conversion separately from
10458 the sign conversion.
10460 2021-09-15 Hongtao Liu <hongtao.liu@intel.com>
10461 Peter Cordes <peter@cordes.ca>
10464 * config/i386/sse.md (extract_suf): Add V8SF/V8SI/V4DF/V4DI.
10465 (*vec_extract<mode><ssescalarmodelower>_valign): Output
10466 vextract{i,f}{32x4,64x2} instruction when byte_offset % 16 ==
10469 2021-09-15 Richard Biener <rguenther@suse.de>
10471 * config.gcc: Remove vax-*-openbsd* configuration.
10473 2021-09-15 Richard Biener <rguenther@suse.de>
10475 * config.gcc: Remove m68k-openbsd.
10477 2021-09-15 Max Filippov <jcmvbkbc@gmail.com>
10480 * config/xtensa/t-xtensa (TM_H): Add include/xtensa-config.h.
10482 2021-09-14 Peter Bergner <bergner@linux.ibm.com>
10484 * config/rs6000/mma.md (unspec): Delete UNSPEC_MMA_XXSETACCZ.
10485 (unspecv): Add UNSPECV_MMA_XXSETACCZ.
10486 (*mma_xxsetaccz): Delete.
10487 (mma_xxsetaccz): Change to define_insn. Remove operand 1.
10488 Use UNSPECV_MMA_XXSETACCZ. Update comment.
10489 * config/rs6000/rs6000.c (rs6000_rtx_costs): Use UNSPECV_MMA_XXSETACCZ.
10491 2021-09-14 Iain Sandoe <iain@sandoe.co.uk>
10493 * Makefile.in: Remove variables related to applying no-PIE
10494 to the exes on $build.
10495 * configure: Regenerate.
10496 * configure.ac: Remove configuration related to applying
10497 no-PIE to the exes on $build.
10499 2021-09-14 Claudiu Zissulescu <claziss@synopsys.com>
10501 * config/arc/arc.md (doloop_end): Add missing mode.
10502 (loop_end): Likewise.
10504 2021-09-14 Jakub Jelinek <jakub@redhat.com>
10506 * gimplify.c (goa_stabilize_expr): Add depth argument, propagate
10507 it to recursive calls, for depth above 7 just gimplify or return.
10508 Perform a test even for MODIFY_EXPR, ADDR_EXPR, COMPOUND_EXPR with
10509 __builtin_clear_padding and TARGET_EXPR.
10510 (gimplify_omp_atomic): Adjust goa_stabilize_expr callers.
10512 2021-09-14 liuhongt <hongtao.liu@intel.com>
10514 * config/i386/avx512fp16intrin.h (_mm_fpclass_sh_mask):
10516 (_mm_mask_fpclass_sh_mask): Likewise.
10517 (_mm512_mask_fpclass_ph_mask): Likewise.
10518 (_mm512_fpclass_ph_mask): Likewise.
10519 (_mm_getexp_sh): Likewise.
10520 (_mm_mask_getexp_sh): Likewise.
10521 (_mm_maskz_getexp_sh): Likewise.
10522 (_mm512_getexp_ph): Likewise.
10523 (_mm512_mask_getexp_ph): Likewise.
10524 (_mm512_maskz_getexp_ph): Likewise.
10525 (_mm_getexp_round_sh): Likewise.
10526 (_mm_mask_getexp_round_sh): Likewise.
10527 (_mm_maskz_getexp_round_sh): Likewise.
10528 (_mm512_getexp_round_ph): Likewise.
10529 (_mm512_mask_getexp_round_ph): Likewise.
10530 (_mm512_maskz_getexp_round_ph): Likewise.
10531 (_mm_getmant_sh): Likewise.
10532 (_mm_mask_getmant_sh): Likewise.
10533 (_mm_maskz_getmant_sh): Likewise.
10534 (_mm512_getmant_ph): Likewise.
10535 (_mm512_mask_getmant_ph): Likewise.
10536 (_mm512_maskz_getmant_ph): Likewise.
10537 (_mm_getmant_round_sh): Likewise.
10538 (_mm_mask_getmant_round_sh): Likewise.
10539 (_mm_maskz_getmant_round_sh): Likewise.
10540 (_mm512_getmant_round_ph): Likewise.
10541 (_mm512_mask_getmant_round_ph): Likewise.
10542 (_mm512_maskz_getmant_round_ph): Likewise.
10543 * config/i386/avx512fp16vlintrin.h (_mm_mask_fpclass_ph_mask):
10545 (_mm_fpclass_ph_mask): Likewise.
10546 (_mm256_mask_fpclass_ph_mask): Likewise.
10547 (_mm256_fpclass_ph_mask): Likewise.
10548 (_mm256_getexp_ph): Likewise.
10549 (_mm256_mask_getexp_ph): Likewise.
10550 (_mm256_maskz_getexp_ph): Likewise.
10551 (_mm_getexp_ph): Likewise.
10552 (_mm_mask_getexp_ph): Likewise.
10553 (_mm_maskz_getexp_ph): Likewise.
10554 (_mm256_getmant_ph): Likewise.
10555 (_mm256_mask_getmant_ph): Likewise.
10556 (_mm256_maskz_getmant_ph): Likewise.
10557 (_mm_getmant_ph): Likewise.
10558 (_mm_mask_getmant_ph): Likewise.
10559 (_mm_maskz_getmant_ph): Likewise.
10560 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
10561 * config/i386/i386-builtin.def: Add corresponding new builtins.
10562 * config/i386/i386-expand.c
10563 (ix86_expand_args_builtin): Handle new builtin types.
10564 (ix86_expand_round_builtin): Ditto.
10565 * config/i386/sse.md (vecmemsuffix): Add HF vector modes.
10566 (<avx512>_getexp<mode><mask_name><round_saeonly_name>): Adjust
10567 to support HF vector modes.
10568 (avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name):
10570 (avx512dq_fpclass<mode><mask_scalar_merge_name>): Ditto.
10571 (avx512dq_vmfpclass<mode><mask_scalar_merge_name>): Ditto.
10572 (<avx512>_getmant<mode><mask_name><round_saeonly_name>): Ditto.
10573 (avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
10576 2021-09-14 liuhongt <hongtao.liu@intel.com>
10578 * config/i386/avx512fp16intrin.h (_mm512_reduce_ph):
10580 (_mm512_mask_reduce_ph): Likewise.
10581 (_mm512_maskz_reduce_ph): Likewise.
10582 (_mm512_reduce_round_ph): Likewise.
10583 (_mm512_mask_reduce_round_ph): Likewise.
10584 (_mm512_maskz_reduce_round_ph): Likewise.
10585 (_mm_reduce_sh): Likewise.
10586 (_mm_mask_reduce_sh): Likewise.
10587 (_mm_maskz_reduce_sh): Likewise.
10588 (_mm_reduce_round_sh): Likewise.
10589 (_mm_mask_reduce_round_sh): Likewise.
10590 (_mm_maskz_reduce_round_sh): Likewise.
10591 (_mm512_roundscale_ph): Likewise.
10592 (_mm512_mask_roundscale_ph): Likewise.
10593 (_mm512_maskz_roundscale_ph): Likewise.
10594 (_mm512_roundscale_round_ph): Likewise.
10595 (_mm512_mask_roundscale_round_ph): Likewise.
10596 (_mm512_maskz_roundscale_round_ph): Likewise.
10597 (_mm_roundscale_sh): Likewise.
10598 (_mm_mask_roundscale_sh): Likewise.
10599 (_mm_maskz_roundscale_sh): Likewise.
10600 (_mm_roundscale_round_sh): Likewise.
10601 (_mm_mask_roundscale_round_sh): Likewise.
10602 (_mm_maskz_roundscale_round_sh): Likewise.
10603 * config/i386/avx512fp16vlintrin.h: (_mm_reduce_ph):
10605 (_mm_mask_reduce_ph): Likewise.
10606 (_mm_maskz_reduce_ph): Likewise.
10607 (_mm256_reduce_ph): Likewise.
10608 (_mm256_mask_reduce_ph): Likewise.
10609 (_mm256_maskz_reduce_ph): Likewise.
10610 (_mm_roundscale_ph): Likewise.
10611 (_mm_mask_roundscale_ph): Likewise.
10612 (_mm_maskz_roundscale_ph): Likewise.
10613 (_mm256_roundscale_ph): Likewise.
10614 (_mm256_mask_roundscale_ph): Likewise.
10615 (_mm256_maskz_roundscale_ph): Likewise.
10616 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
10617 * config/i386/i386-builtin.def: Add corresponding new builtins.
10618 * config/i386/i386-expand.c
10619 (ix86_expand_args_builtin): Handle new builtin types.
10620 (ix86_expand_round_builtin): Ditto.
10621 * config/i386/sse.md (<mask_codefor>reducep<mode><mask_name>):
10623 (<mask_codefor>reducep<mode><mask_name><round_saeonly_name>):
10624 ... this, and adjust for round operands.
10625 (reduces<mode><mask_scalar_name>): Likewise, with ...
10626 (reduces<mode><mask_scalar_name><round_saeonly_scalar_name):
10628 (<avx512>_rndscale<mode><mask_name><round_saeonly_name>):
10629 Adjust for HF vector modes.
10630 (avx512f_rndscale<mode><mask_scalar_name><round_saeonly_scalar_name>):
10632 (*avx512f_rndscale<mode><round_saeonly_name>): Ditto.
10634 2021-09-14 liuhongt <hongtao.liu@intel.com>
10636 * config/i386/avx512fp16intrin.h: (_mm512_rcp_ph):
10638 (_mm512_mask_rcp_ph): Likewise.
10639 (_mm512_maskz_rcp_ph): Likewise.
10640 (_mm_rcp_sh): Likewise.
10641 (_mm_mask_rcp_sh): Likewise.
10642 (_mm_maskz_rcp_sh): Likewise.
10643 (_mm512_scalef_ph): Likewise.
10644 (_mm512_mask_scalef_ph): Likewise.
10645 (_mm512_maskz_scalef_ph): Likewise.
10646 (_mm512_scalef_round_ph): Likewise.
10647 (_mm512_mask_scalef_round_ph): Likewise.
10648 (_mm512_maskz_scalef_round_ph): Likewise.
10649 (_mm_scalef_sh): Likewise.
10650 (_mm_mask_scalef_sh): Likewise.
10651 (_mm_maskz_scalef_sh): Likewise.
10652 (_mm_scalef_round_sh): Likewise.
10653 (_mm_mask_scalef_round_sh): Likewise.
10654 (_mm_maskz_scalef_round_sh): Likewise.
10655 * config/i386/avx512fp16vlintrin.h (_mm_rcp_ph):
10657 (_mm256_rcp_ph): Likewise.
10658 (_mm_mask_rcp_ph): Likewise.
10659 (_mm256_mask_rcp_ph): Likewise.
10660 (_mm_maskz_rcp_ph): Likewise.
10661 (_mm256_maskz_rcp_ph): Likewise.
10662 (_mm_scalef_ph): Likewise.
10663 (_mm256_scalef_ph): Likewise.
10664 (_mm_mask_scalef_ph): Likewise.
10665 (_mm256_mask_scalef_ph): Likewise.
10666 (_mm_maskz_scalef_ph): Likewise.
10667 (_mm256_maskz_scalef_ph): Likewise.
10668 * config/i386/i386-builtin.def: Add new builtins.
10669 * config/i386/sse.md (VFH_AVX512VL): New.
10670 (avx512fp16_rcp<mode>2<mask_name>): Ditto.
10671 (avx512fp16_vmrcpv8hf2<mask_scalar_name>): Ditto.
10672 (avx512f_vmscalef<mode><mask_scalar_name><round_scalar_name>):
10673 Adjust to support HF vector modes.
10674 (<avx512>_scalef<mode><mask_name><round_name>): Ditto.
10676 2021-09-14 liuhongt <hongtao.liu@intel.com>
10678 * config/i386/avx512fp16intrin.h: (_mm512_sqrt_ph):
10680 (_mm512_mask_sqrt_ph): Likewise.
10681 (_mm512_maskz_sqrt_ph): Likewise.
10682 (_mm512_sqrt_round_ph): Likewise.
10683 (_mm512_mask_sqrt_round_ph): Likewise.
10684 (_mm512_maskz_sqrt_round_ph): Likewise.
10685 (_mm512_rsqrt_ph): Likewise.
10686 (_mm512_mask_rsqrt_ph): Likewise.
10687 (_mm512_maskz_rsqrt_ph): Likewise.
10688 (_mm_rsqrt_sh): Likewise.
10689 (_mm_mask_rsqrt_sh): Likewise.
10690 (_mm_maskz_rsqrt_sh): Likewise.
10691 (_mm_sqrt_sh): Likewise.
10692 (_mm_mask_sqrt_sh): Likewise.
10693 (_mm_maskz_sqrt_sh): Likewise.
10694 (_mm_sqrt_round_sh): Likewise.
10695 (_mm_mask_sqrt_round_sh): Likewise.
10696 (_mm_maskz_sqrt_round_sh): Likewise.
10697 * config/i386/avx512fp16vlintrin.h (_mm_sqrt_ph): New intrinsic.
10698 (_mm256_sqrt_ph): Likewise.
10699 (_mm_mask_sqrt_ph): Likewise.
10700 (_mm256_mask_sqrt_ph): Likewise.
10701 (_mm_maskz_sqrt_ph): Likewise.
10702 (_mm256_maskz_sqrt_ph): Likewise.
10703 (_mm_rsqrt_ph): Likewise.
10704 (_mm256_rsqrt_ph): Likewise.
10705 (_mm_mask_rsqrt_ph): Likewise.
10706 (_mm256_mask_rsqrt_ph): Likewise.
10707 (_mm_maskz_rsqrt_ph): Likewise.
10708 (_mm256_maskz_rsqrt_ph): Likewise.
10709 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
10710 * config/i386/i386-builtin.def: Add corresponding new builtins.
10711 * config/i386/i386-expand.c
10712 (ix86_expand_args_builtin): Handle new builtins.
10713 (ix86_expand_round_builtin): Ditto.
10714 * config/i386/sse.md (VF_AVX512FP16VL): New.
10715 (sqrt<mode>2): Adjust for HF vector modes.
10716 (<sse>_sqrt<mode>2<mask_name><round_name>): Likewise.
10717 (<sse>_vmsqrt<mode>2<mask_scalar_name><round_scalar_name>):
10719 (<sse>_rsqrt<mode>2<mask_name>): New.
10720 (avx512fp16_vmrsqrtv8hf2<mask_scalar_name>): Likewise.
10722 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
10724 PR bootstrap/101574
10725 * diagnostic-spec.c (warning_suppressed_at, copy_warning): Handle
10726 'RESERVED_LOCATION_P' locations.
10727 * warning-control.cc (get_nowarn_spec, suppress_warning)
10728 (copy_warning): Likewise.
10730 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
10732 * diagnostic-spec.h (typedef xint_hash_t): Use 'location_t' instead of...
10733 (typedef key_type_t): ... this. Remove.
10734 (nowarn_map): Document.
10735 * diagnostic-spec.c (nowarn_map): Likewise.
10736 * warning-control.cc (convert_to_key): Evolve functions into...
10737 (get_location): ... these. Adjust all users.
10739 2021-09-13 Thomas Schwinge <thomas@codesourcery.com>
10741 * warning-control.cc (copy_warning): Remove 'nowarn_map' setup.
10743 2021-09-13 Jason Merrill <jason@redhat.com>
10745 * params.opt: Add destructive-interference-size and
10746 constructive-interference-size.
10747 * doc/invoke.texi: Document them.
10748 * config/aarch64/aarch64.c (aarch64_override_options_internal):
10750 * config/arm/arm.c (arm_option_override): Set them.
10751 * config/i386/i386-options.c (ix86_option_override_internal):
10754 2021-09-13 Martin Liska <mliska@suse.cz>
10755 H.J. Lu <hjl.tools@gmail.com>
10758 * common/config/i386/cpuinfo.h (cpu_indicator_init): Add support
10759 for x86-64 micro levels for __builtin_cpu_supports.
10760 * common/config/i386/i386-cpuinfo.h (enum feature_priority):
10761 Add priorities for the micro-arch levels.
10762 (enum processor_features): Add new features.
10763 * common/config/i386/i386-isas.h: Add micro-arch features.
10764 * config/i386/i386-builtins.c (get_builtin_code_for_version):
10765 Support the micro-arch levels by callsing
10766 __builtin_cpu_supports.
10767 * doc/extend.texi: Document that the levels are support by
10768 __builtin_cpu_supports.
10770 2021-09-13 Andrew Pinski <apinski@marvell.com>
10773 * config/aarch64/aarch64-builtins.c (aarch64_fold_builtin_lane_check):
10775 (aarch64_general_fold_builtin): Handle AARCH64_SIMD_BUILTIN_LANE_CHECK.
10776 (aarch64_general_gimple_fold_builtin): Likewise.
10778 2021-09-13 Andrew Pinski <apinski@marvell.com>
10780 * config.gcc: Add m32r-*-linux* and m32rle-*-linux*
10781 to the Unsupported targets list.
10782 Remove support for m32r-*-linux* and m32rle-*-linux*.
10783 * config/m32r/linux.h: Removed.
10784 * config/m32r/t-linux: Removed.
10786 2021-09-13 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10789 * config/aarch64/aarch64.c (aarch64_classify_address): Don't allow
10790 register index for SVE predicate modes.
10792 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
10794 * tree-ssa-threadbackward.c
10795 (back_threader_profitability::profitable_path_p): Remove FSM
10797 (back_threader_registry::register_path): Same.
10798 * tree-ssa-threadedge.c
10799 (jump_threader::simplify_control_stmt_condition): Same.
10800 * tree-ssa-threadupdate.c (jt_path_registry::jt_path_registry):
10801 Add backedge_threads argument.
10802 (fwd_jt_path_registry::fwd_jt_path_registry): Pass
10803 backedge_threads argument.
10804 (back_jt_path_registry::back_jt_path_registry): Same.
10805 (dump_jump_thread_path): Adjust for FSM removal.
10806 (back_jt_path_registry::rewire_first_differing_edge): Same.
10807 (back_jt_path_registry::adjust_paths_after_duplication): Same.
10808 (back_jt_path_registry::update_cfg): Same.
10809 (jt_path_registry::register_jump_thread): Same.
10810 * tree-ssa-threadupdate.h (enum jump_thread_edge_type): Remove
10812 (class back_jt_path_registry): Add backedge_threads to
10815 2021-09-13 Martin Liska <mliska@suse.cz>
10818 * asan.h (sanitize_coverage_p): Handle when fn == NULL.
10820 2021-09-13 H.J. Lu <hjl.tools@gmail.com>
10823 * config/i386/i386.h (TARGET_AVX256_MOVE_BY_PIECES): New.
10824 (TARGET_AVX256_STORE_BY_PIECES): Likewise.
10825 (MOVE_MAX): Check TARGET_AVX256_MOVE_BY_PIECES and
10826 TARGET_AVX256_STORE_BY_PIECES instead of
10827 TARGET_AVX256_SPLIT_UNALIGNED_LOAD and
10828 TARGET_AVX256_SPLIT_UNALIGNED_STORE.
10829 (STORE_MAX_PIECES): Check TARGET_AVX256_STORE_BY_PIECES instead
10830 of TARGET_AVX256_SPLIT_UNALIGNED_STORE.
10831 * config/i386/x86-tune.def (X86_TUNE_AVX256_MOVE_BY_PIECES): New.
10832 (X86_TUNE_AVX256_STORE_BY_PIECES): Likewise.
10834 2021-09-13 liuhongt <hongtao.liu@intel.com>
10836 PR bootstrap/102302
10837 * expmed.c (extract_bit_field_using_extv): Use
10838 gen_lowpart_if_possible instead of gen_lowpart to avoid ICE.
10840 2021-09-13 Aldy Hernandez <aldyh@redhat.com>
10842 * Makefile.in (OBJS): Add value-pointer-equiv.o.
10843 * gimple-ssa-evrp.c (class ssa_equiv_stack): Move to
10844 value-pointer-equiv.*.
10845 (ssa_equiv_stack::ssa_equiv_stack): Same.
10846 (ssa_equiv_stack::enter): Same.
10847 (ssa_equiv_stack::leave): Same.
10848 (ssa_equiv_stack::push_replacement): Same.
10849 (ssa_equiv_stack::get_replacement): Same.
10850 (is_pointer_ssa): Same.
10851 (class pointer_equiv_analyzer): Same.
10852 (pointer_equiv_analyzer::pointer_equiv_analyzer): Same.
10853 (pointer_equiv_analyzer::~pointer_equiv_analyzer): Same.
10854 (pointer_equiv_analyzer::set_global_equiv): Same.
10855 (pointer_equiv_analyzer::set_cond_equiv): Same.
10856 (pointer_equiv_analyzer::get_equiv): Same.
10857 (pointer_equiv_analyzer::enter): Same.
10858 (pointer_equiv_analyzer::leave): Same.
10859 (pointer_equiv_analyzer::get_equiv_expr): Same.
10860 (pta_valueize): Same.
10861 (pointer_equiv_analyzer::visit_stmt): Same.
10862 (pointer_equiv_analyzer::visit_edge): Same.
10863 (hybrid_folder::value_of_expr): Same.
10864 (hybrid_folder::value_on_edge): Same.
10865 * value-pointer-equiv.cc: New file.
10866 * value-pointer-equiv.h: New file.
10868 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
10871 * gimple-fold.c (gimple_fold_builtin_memory_op): Allow folding
10872 memcpy if the size is not more than MOVE_MAX * MOVE_RATIO.
10874 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
10877 * config/arm/arm.md (movmisaligndi): New define_expand.
10878 * config/arm/vec-common.md (movmisalign<mode>): Iterate over VDQ mode.
10880 2021-09-13 Richard Earnshaw <rearnsha@arm.com>
10883 * emit-rtl.c (gen_highpart): Use adjust_address to handle
10884 MEM rather than calling simplify_gen_subreg.
10886 2021-09-13 Jan-Benedict Glaw <jbglaw@ług-owl.de>
10888 * config/alpha/vms.h (INIT_CUMULATIVE_ARGS): Wrap multi-statment
10889 define into a block.
10891 2021-09-13 Richard Biener <rguenther@suse.de>
10893 * config/darwin.h (DARWIN_PREFER_DWARF): Do not define.
10894 * config/i386/darwin.h (PREFERRED_DEBUGGING_TYPE): Do not
10895 change based on DARWIN_PREFER_DWARF not being defined.
10897 2021-09-13 Richard Biener <rguenther@suse.de>
10899 * config/i386/lynx.h: Remove undef of PREFERRED_DEBUGGING_TYPE
10900 to inherit from elfos.h
10902 2021-09-13 Richard Biener <rguenther@suse.de>
10904 * config.gcc: Add cr16-*-* to the list of obsoleted targets.
10906 2021-09-13 Richard Biener <rguenther@suse.de>
10908 * config/avr/elf.h (PREFERRED_DEBUGGING_TYPE): Remove
10909 override, pick up DWARF2_DEBUG define from elfos.h
10911 2021-09-13 Richard Biener <rguenther@suse.de>
10913 * config/rx/rx.h (PREFERRED_DEBUGGING_TYPE): Always define to
10916 2021-09-13 Richard Biener <rguenther@suse.de>
10918 * config/alpha/vms.h (PREFERRED_DEBUGGING_TYPE): Define to
10921 2021-09-13 Richard Biener <rguenther@suse.de>
10923 * config/i386/cygming.h: Always default to DWARF2 debugging.
10924 Do not define DBX_DEBUGGING_INFO, that's done via dbxcoff.h
10926 * doc/install.texi: Document binutils 2.16 as minimum
10927 requirement for mingw.
10929 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
10931 * config/rs6000/rs6000.c (struct rs6000_cost_data): New members
10932 nstmts, nloads and extra_ctor_cost.
10933 (rs6000_density_test): Add load density related heuristics. Do
10934 extra costing on vector construction statements if need.
10935 (rs6000_init_cost): Init new members.
10936 (rs6000_update_target_cost_per_stmt): New function.
10937 (rs6000_add_stmt_cost): Factor vect_nonmem hunk out to function
10938 rs6000_update_target_cost_per_stmt and call it.
10940 2021-09-13 Kewen Lin <linkw@linux.ibm.com>
10942 * config/rs6000/rs6000.c (struct rs6000_cost_data): Remove typedef.
10943 (rs6000_init_cost): Adjust.
10945 2021-09-13 liuhongt <hongtao.liu@intel.com>
10947 * config/i386/i386.md: (UNSPEC_COPYSIGN): Remove.
10948 (UNSPEC_XORSIGN): Ditto.
10950 2021-09-12 Roger Sayle <roger@nextmovesoftware.com>
10952 * expr.c (convert_move): Preserve SUBREG_PROMOTED_VAR_P when
10953 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
10956 2021-09-11 Aldy Hernandez <aldyh@redhat.com>
10958 * tree-ssa-threadbackward.c (class back_threader_registry): Use
10959 back_jt_path_registry.
10960 * tree-ssa-threadedge.c (jump_threader::jump_threader): Use
10961 fwd_jt_path_registry.
10962 * tree-ssa-threadedge.h (class jump_threader): Same..
10963 * tree-ssa-threadupdate.c
10964 (jump_thread_path_registry::jump_thread_path_registry): Rename...
10965 (jt_path_registry::jt_path_registry): ...to this.
10966 (jump_thread_path_registry::~jump_thread_path_registry): Rename...
10967 (jt_path_registry::~jt_path_registry): ...this.
10968 (fwd_jt_path_registry::fwd_jt_path_registry): New.
10969 (fwd_jt_path_registry::~fwd_jt_path_registry): New.
10970 (jump_thread_path_registry::allocate_thread_edge): Rename...
10971 (jt_path_registry::allocate_thread_edge): ...to this.
10972 (jump_thread_path_registry::allocate_thread_path): Rename...
10973 (jt_path_registry::allocate_thread_path): ...to this.
10974 (jump_thread_path_registry::lookup_redirection_data): Rename...
10975 (fwd_jt_path_registry::lookup_redirection_data): ...to this.
10976 (jump_thread_path_registry::thread_block_1): Rename...
10977 (fwd_jt_path_registry::thread_block_1): ...to this.
10978 (jump_thread_path_registry::thread_block): Rename...
10979 (fwd_jt_path_registry::thread_block): ...to this.
10980 (jt_path_registry::thread_through_loop_header): Rename...
10981 (fwd_jt_path_registry::thread_through_loop_header): ...to this.
10982 (jump_thread_path_registry::mark_threaded_blocks): Rename...
10983 (fwd_jt_path_registry::mark_threaded_blocks): ...to this.
10984 (jump_thread_path_registry::debug_path): Rename...
10985 (jt_path_registry::debug_path): ...to this.
10986 (jump_thread_path_registry::dump): Rename...
10987 (jt_path_registry::debug): ...to this.
10988 (jump_thread_path_registry::rewire_first_differing_edge): Rename...
10989 (back_jt_path_registry::rewire_first_differing_edge): ...to this.
10990 (jump_thread_path_registry::adjust_paths_after_duplication): Rename...
10991 (back_jt_path_registry::adjust_paths_after_duplication): ...to this.
10992 (jump_thread_path_registry::duplicate_thread_path): Rename...
10993 (back_jt_path_registry::duplicate_thread_path): ...to this. Also,
10994 drop ill-formed candidates.
10995 (jump_thread_path_registry::remove_jump_threads_including): Rename...
10996 (fwd_jt_path_registry::remove_jump_threads_including): ...to this.
10997 (jt_path_registry::thread_through_all_blocks): New.
10998 (back_jt_path_registry::update_cfg): New.
10999 (fwd_jt_path_registry::update_cfg): New.
11000 (jump_thread_path_registry::register_jump_thread): Rename...
11001 (jt_path_registry::register_jump_thread): ...to this.
11002 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
11004 (class jt_path_registry): ...here.
11005 (class fwd_jt_path_registry): New.
11006 (class back_jt_path_registry): New.
11008 2021-09-10 liuhongt <hongtao.liu@intel.com>
11011 2021-09-01 liuhongt <hongtao.liu@intel.com>
11013 * emit-rtl.c (validate_subreg): Get rid of all float-int
11016 2021-09-10 Jakub Jelinek <jakub@redhat.com>
11018 * tree-core.h (enum omp_memory_order): Add OMP_MEMORY_ORDER_MASK,
11019 OMP_FAIL_MEMORY_ORDER_UNSPECIFIED, OMP_FAIL_MEMORY_ORDER_RELAXED,
11020 OMP_FAIL_MEMORY_ORDER_ACQUIRE, OMP_FAIL_MEMORY_ORDER_RELEASE,
11021 OMP_FAIL_MEMORY_ORDER_ACQ_REL, OMP_FAIL_MEMORY_ORDER_SEQ_CST and
11022 OMP_FAIL_MEMORY_ORDER_MASK enumerators.
11023 (OMP_FAIL_MEMORY_ORDER_SHIFT): Define.
11024 * gimple-pretty-print.c (dump_gimple_omp_atomic_load,
11025 dump_gimple_omp_atomic_store): Print [weak] for weak atomic
11027 * gimple.h (enum gf_mask): Change GF_OMP_ATOMIC_MEMORY_ORDER
11028 to 6-bit mask, adjust GF_OMP_ATOMIC_NEED_VALUE value and add
11029 GF_OMP_ATOMIC_WEAK.
11030 (gimple_omp_atomic_weak_p, gimple_omp_atomic_set_weak): New inline
11032 * tree.h (OMP_ATOMIC_WEAK): Define.
11033 * tree-pretty-print.c (dump_omp_atomic_memory_order): Adjust for
11034 fail memory order being encoded in the same enum and also print
11035 fail clause if present.
11036 (dump_generic_node): Print weak clause if OMP_ATOMIC_WEAK.
11037 * gimplify.c (goa_stabilize_expr): Add target_expr and rhs arguments,
11038 handle pre_p == NULL case as a test mode that only returns value
11039 but doesn't change gimplify nor change anything otherwise, adjust
11040 recursive calls, add MODIFY_EXPR, ADDR_EXPR, COND_EXPR, TARGET_EXPR
11041 and CALL_EXPR handling, adjust COMPOUND_EXPR handling for
11042 __builtin_clear_padding calls, for !rhs gimplify as lvalue rather
11044 (gimplify_omp_atomic): Adjust goa_stabilize_expr caller. Handle
11045 COND_EXPR rhs. Set weak flag on gimple load/store for
11047 * omp-expand.c (omp_memory_order_to_fail_memmodel): New function.
11048 (omp_memory_order_to_memmodel): Adjust for fail clause encoded
11050 (expand_omp_atomic_cas): New function.
11051 (expand_omp_atomic_pipeline): Use omp_memory_order_to_fail_memmodel
11053 (expand_omp_atomic): Attempt to optimize atomic compare and exchange
11054 using expand_omp_atomic_cas.
11056 2021-09-10 Aldy Hernandez <aldyh@redhat.com>
11057 Michael Matz <matz@suse.de>
11059 * tree-pass.h (PROP_loop_opts_done): New.
11060 * gimple-range-path.cc (path_range_query::internal_range_of_expr):
11061 Intersect with global range.
11062 * tree-ssa-loop.c (tree_ssa_loop_done): Set PROP_loop_opts_done.
11063 * tree-ssa-threadbackward.c
11064 (back_threader_profitability::profitable_path_p): Disable
11065 threading through latches until after loop optimizations have run.
11067 2021-09-10 David Faust <david.faust@oracle.com>
11069 * doc/invoke.texi: Document BPF -mcpu, -mjmpext, -mjmp32 and -malu32
11072 2021-09-10 David Faust <david.faust@oracle.com>
11074 * config/bpf/bpf-opts.h (bpf_isa_version): New enum.
11075 * config/bpf/bpf-protos.h (bpf_expand_cbranch): New.
11076 * config/bpf/bpf.c (bpf_option_override): Handle -mcpu option.
11077 (bpf_expand_cbranch): New function.
11078 * config/bpf/bpf.md (AM mode iterator): Conditionalize support for SI
11080 (zero_extendsidi2): Only use mov32 instruction if it is available.
11081 (SIM mode iterator): Conditionalize support for SI mode.
11082 (JM mode iterator): New.
11083 (cbranchdi4): Update name, use new JM iterator. Use bpf_expand_cbranch.
11084 (*branch_on_di): Update name, use new JM iterator.
11085 * config/bpf/bpf.opt: (mjmpext): New option.
11086 (malu32): Likewise.
11087 (mjmp32): Likewise.
11089 (bpf_isa): New enum.
11091 2021-09-10 David Faust <david.faust@oracle.com>
11093 * config/bpf/bpf.md (zero_extendhidi2): Add new output template
11094 for register-to-register extensions.
11095 (zero_extendqidi2): Likewise.
11097 2021-09-10 Richard Biener <rguenther@suse.de>
11099 PR middle-end/102273
11100 * internal-fn.c (expand_DEFERRED_INIT): Always expand non-SSA vars.
11102 2021-09-10 Richard Biener <rguenther@suse.de>
11104 PR middle-end/102269
11105 * gimplify.c (is_var_need_auto_init): Empty types do not need
11108 2021-09-10 Richard Biener <rguenther@suse.de>
11110 * configure.ac (--with-stabs): Remove.
11111 * configure: Regenerate.
11112 * doc/install.texi: Remove --with-stabs documentation.
11114 2021-09-10 liuhongt <hongtao.liu@intel.com>
11116 * config/i386/avx512fp16intrin.h: (_mm512_cmp_ph_mask):
11118 (_mm512_mask_cmp_ph_mask): Likewise.
11119 (_mm512_cmp_round_ph_mask): Likewise.
11120 (_mm512_mask_cmp_round_ph_mask): Likewise.
11121 (_mm_cmp_sh_mask): Likewise.
11122 (_mm_mask_cmp_sh_mask): Likewise.
11123 (_mm_cmp_round_sh_mask): Likewise.
11124 (_mm_mask_cmp_round_sh_mask): Likewise.
11125 (_mm_comieq_sh): Likewise.
11126 (_mm_comilt_sh): Likewise.
11127 (_mm_comile_sh): Likewise.
11128 (_mm_comigt_sh): Likewise.
11129 (_mm_comige_sh): Likewise.
11130 (_mm_comineq_sh): Likewise.
11131 (_mm_ucomieq_sh): Likewise.
11132 (_mm_ucomilt_sh): Likewise.
11133 (_mm_ucomile_sh): Likewise.
11134 (_mm_ucomigt_sh): Likewise.
11135 (_mm_ucomige_sh): Likewise.
11136 (_mm_ucomineq_sh): Likewise.
11137 (_mm_comi_round_sh): Likewise.
11138 (_mm_comi_sh): Likewise.
11139 * config/i386/avx512fp16vlintrin.h (_mm_cmp_ph_mask): New intrinsic.
11140 (_mm_mask_cmp_ph_mask): Likewise.
11141 (_mm256_cmp_ph_mask): Likewise.
11142 (_mm256_mask_cmp_ph_mask): Likewise.
11143 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
11144 * config/i386/i386-builtin.def: Add corresponding new builtins.
11145 * config/i386/i386-expand.c
11146 (ix86_expand_args_builtin): Handle new builtin types.
11147 (ix86_expand_round_builtin): Ditto.
11148 * config/i386/i386.md (ssevecmode): Add HF mode.
11149 (MODEFH): New mode iterator.
11150 * config/i386/sse.md
11151 (V48H_AVX512VL): New mode iterator to support HF vector modes.
11152 Ajdust corresponding description.
11153 (ssecmpintprefix): New.
11154 (VI12_AVX512VL): Adjust to support HF vector modes.
11155 (cmp_imm_predicate): Likewise.
11156 (<avx512>_cmp<mode>3<mask_scalar_merge_name><round_saeonly_name>):
11158 (avx512f_vmcmp<mode>3<round_saeonly_name>): Likewise.
11159 (avx512f_vmcmp<mode>3_mask<round_saeonly_name>): Likewise.
11160 (<sse>_<unord>comi<round_saeonly_name>): Likewise.
11162 2021-09-10 liuhongt <hongtao.liu@intel.com>
11164 * config/i386/avx512fp16intrin.h: (_mm512_max_ph): New intrinsic.
11165 (_mm512_mask_max_ph): Likewise.
11166 (_mm512_maskz_max_ph): Likewise.
11167 (_mm512_min_ph): Likewise.
11168 (_mm512_mask_min_ph): Likewise.
11169 (_mm512_maskz_min_ph): Likewise.
11170 (_mm512_max_round_ph): Likewise.
11171 (_mm512_mask_max_round_ph): Likewise.
11172 (_mm512_maskz_max_round_ph): Likewise.
11173 (_mm512_min_round_ph): Likewise.
11174 (_mm512_mask_min_round_ph): Likewise.
11175 (_mm512_maskz_min_round_ph): Likewise.
11176 (_mm_max_sh): Likewise.
11177 (_mm_mask_max_sh): Likewise.
11178 (_mm_maskz_max_sh): Likewise.
11179 (_mm_min_sh): Likewise.
11180 (_mm_mask_min_sh): Likewise.
11181 (_mm_maskz_min_sh): Likewise.
11182 (_mm_max_round_sh): Likewise.
11183 (_mm_mask_max_round_sh): Likewise.
11184 (_mm_maskz_max_round_sh): Likewise.
11185 (_mm_min_round_sh): Likewise.
11186 (_mm_mask_min_round_sh): Likewise.
11187 (_mm_maskz_min_round_sh): Likewise.
11188 * config/i386/avx512fp16vlintrin.h (_mm_max_ph): New intrinsic.
11189 (_mm256_max_ph): Likewise.
11190 (_mm_mask_max_ph): Likewise.
11191 (_mm256_mask_max_ph): Likewise.
11192 (_mm_maskz_max_ph): Likewise.
11193 (_mm256_maskz_max_ph): Likewise.
11194 (_mm_min_ph): Likewise.
11195 (_mm256_min_ph): Likewise.
11196 (_mm_mask_min_ph): Likewise.
11197 (_mm256_mask_min_ph): Likewise.
11198 (_mm_maskz_min_ph): Likewise.
11199 (_mm256_maskz_min_ph): Likewise.
11200 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
11201 * config/i386/i386-builtin.def: Add corresponding new builtins.
11202 * config/i386/i386-expand.c
11203 (ix86_expand_args_builtin): Handle new builtin types.
11204 * config/i386/sse.md
11205 (<code><mode>3<mask_name><round_saeonly_name>): Adjust to
11206 support HF vector modes.
11207 (*<code><mode>3<mask_name><round_saeonly_name>): Likewise.
11208 (ieee_<ieee_maxmin><mode>3<mask_name><round_saeonly_name>):
11210 (<sse>_vm<code><mode>3<mask_scalar_name><round_saeonly_scalar_name>):
11212 * config/i386/subst.md (round_saeonly_mode512bit_condition):
11213 Adjust for HF vector modes.
11215 2021-09-10 Liu, Hongtao <hongtao.liu@intel.com>
11217 * config/i386/avx512fp16intrin.h (_mm_add_sh): New intrinsic.
11218 (_mm_mask_add_sh): Likewise.
11219 (_mm_maskz_add_sh): Likewise.
11220 (_mm_sub_sh): Likewise.
11221 (_mm_mask_sub_sh): Likewise.
11222 (_mm_maskz_sub_sh): Likewise.
11223 (_mm_mul_sh): Likewise.
11224 (_mm_mask_mul_sh): Likewise.
11225 (_mm_maskz_mul_sh): Likewise.
11226 (_mm_div_sh): Likewise.
11227 (_mm_mask_div_sh): Likewise.
11228 (_mm_maskz_div_sh): Likewise.
11229 (_mm_add_round_sh): Likewise.
11230 (_mm_mask_add_round_sh): Likewise.
11231 (_mm_maskz_add_round_sh): Likewise.
11232 (_mm_sub_round_sh): Likewise.
11233 (_mm_mask_sub_round_sh): Likewise.
11234 (_mm_maskz_sub_round_sh): Likewise.
11235 (_mm_mul_round_sh): Likewise.
11236 (_mm_mask_mul_round_sh): Likewise.
11237 (_mm_maskz_mul_round_sh): Likewise.
11238 (_mm_div_round_sh): Likewise.
11239 (_mm_mask_div_round_sh): Likewise.
11240 (_mm_maskz_div_round_sh): Likewise.
11241 * config/i386/i386-builtin-types.def: Add corresponding builtin types.
11242 * config/i386/i386-builtin.def: Add corresponding new builtins.
11243 * config/i386/i386-expand.c
11244 (ix86_expand_round_builtin): Handle new builtins.
11245 * config/i386/sse.md (VF_128): Change description.
11246 (<sse>_vm<plusminus_insn><mode>3<mask_scalar_name><round_scalar_name>):
11247 Adjust to support HF vector modes.
11248 (<sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name><round_scalar_name>):
11251 2021-09-10 H.J. Lu <hjl.tools@gmail.com>
11253 * config/i386/i386-expand.c
11254 (ix86_avx256_split_vector_move_misalign): Handle V16HF mode.
11255 * config/i386/i386.c
11256 (ix86_preferred_simd_mode): Handle HF mode.
11257 * config/i386/sse.md (V_256H): New mode iterator.
11258 (avx_vextractf128<mode>): Use it.
11259 (VEC_INIT_MODE): Align vector HFmode condition to vector
11260 HImodes since there're no real HF instruction used.
11261 (VEC_INIT_HALF_MODE): Ditto.
11263 (VIHF_AVX512BW): Ditto.
11264 (*vec_extracthf): Ditto.
11265 (VEC_EXTRACT_MODE): Ditto.
11267 2021-09-10 Richard Biener <rguenther@suse.de>
11270 * config/dbx.h: Remove.
11271 * config/dbxcoff.h: Do not define PREFERRED_DEBUGGING_TYPE.
11272 * config/lynx.h: Likewise.
11274 2021-09-10 liuhongt <hongtao.liu@intel.com>
11276 * config/i386/i386-expand.c (ix86_expand_copysign): Expand
11277 right into ANDNOT + AND + IOR, using paradoxical subregs.
11278 (ix86_split_copysign_const): Remove.
11279 (ix86_split_copysign_var): Ditto.
11280 * config/i386/i386-protos.h (ix86_split_copysign_const): Dotto.
11281 (ix86_split_copysign_var): Ditto.
11282 * config/i386/i386.md (@copysign<mode>3_const): Ditto.
11283 (@copysign<mode>3_var): Ditto.
11285 2021-09-09 qing zhao <qing.zhao@oracle.com>
11287 * builtins.c (expand_builtin_memset): Make external visible.
11288 * builtins.h (expand_builtin_memset): Declare extern.
11289 * common.opt (ftrivial-auto-var-init=): New option.
11290 * doc/extend.texi: Document the uninitialized attribute.
11291 * doc/invoke.texi: Document -ftrivial-auto-var-init.
11292 * flag-types.h (enum auto_init_type): New enumerated type
11294 * gimple-fold.c (clear_padding_type): Add one new parameter.
11295 (clear_padding_union): Likewise.
11296 (clear_padding_emit_loop): Likewise.
11297 (clear_type_padding_in_mask): Likewise.
11298 (gimple_fold_builtin_clear_padding): Handle this new parameter.
11299 * gimplify.c (gimple_add_init_for_auto_var): New function.
11300 (gimple_add_padding_init_for_auto_var): New function.
11301 (is_var_need_auto_init): New function.
11302 (gimplify_decl_expr): Add initialization to automatic variables per
11304 (gimplify_call_expr): Add one new parameter for call to
11305 __builtin_clear_padding.
11306 (gimplify_init_constructor): Add padding initialization in the end.
11307 * internal-fn.c (INIT_PATTERN_VALUE): New macro.
11308 (expand_DEFERRED_INIT): New function.
11309 * internal-fn.def (DEFERRED_INIT): New internal function.
11310 * tree-cfg.c (verify_gimple_call): Verify calls to .DEFERRED_INIT.
11311 * tree-sra.c (generate_subtree_deferred_init): New function.
11312 (scan_function): Avoid setting cannot_scalarize_away_bitmap for
11313 calls to .DEFERRED_INIT.
11314 (sra_modify_deferred_init): New function.
11315 (sra_modify_function_body): Handle calls to DEFERRED_INIT specially.
11316 * tree-ssa-structalias.c (find_func_aliases_for_call): Likewise.
11317 * tree-ssa-uninit.c (warn_uninit): Handle calls to DEFERRED_INIT
11319 (check_defs): Likewise.
11320 (warn_uninitialized_vars): Likewise.
11321 * tree-ssa.c (ssa_undefined_value_p): Likewise.
11322 * tree.c (build_common_builtin_nodes): Build tree node for
11323 BUILT_IN_CLEAR_PADDING when needed.
11325 2021-09-09 Richard Biener <rguenther@suse.de>
11327 * tree-ssa-loop-im.c (fill_always_executed_in_1): Walk
11330 2021-09-09 Richard Biener <rguenther@suse.de>
11332 * tree-ssa-loop-im.c (fill_always_executed_in_1): Integrate
11333 DOM walk from get_loop_body_in_dom_order using a worklist
11336 2021-09-09 liuhongt <hongtao.liu@intel.com>
11338 * config.gcc: Add avx512fp16vlintrin.h.
11339 * config/i386/avx512fp16intrin.h: (_mm512_add_ph): New intrinsic.
11340 (_mm512_mask_add_ph): Likewise.
11341 (_mm512_maskz_add_ph): Likewise.
11342 (_mm512_sub_ph): Likewise.
11343 (_mm512_mask_sub_ph): Likewise.
11344 (_mm512_maskz_sub_ph): Likewise.
11345 (_mm512_mul_ph): Likewise.
11346 (_mm512_mask_mul_ph): Likewise.
11347 (_mm512_maskz_mul_ph): Likewise.
11348 (_mm512_div_ph): Likewise.
11349 (_mm512_mask_div_ph): Likewise.
11350 (_mm512_maskz_div_ph): Likewise.
11351 (_mm512_add_round_ph): Likewise.
11352 (_mm512_mask_add_round_ph): Likewise.
11353 (_mm512_maskz_add_round_ph): Likewise.
11354 (_mm512_sub_round_ph): Likewise.
11355 (_mm512_mask_sub_round_ph): Likewise.
11356 (_mm512_maskz_sub_round_ph): Likewise.
11357 (_mm512_mul_round_ph): Likewise.
11358 (_mm512_mask_mul_round_ph): Likewise.
11359 (_mm512_maskz_mul_round_ph): Likewise.
11360 (_mm512_div_round_ph): Likewise.
11361 (_mm512_mask_div_round_ph): Likewise.
11362 (_mm512_maskz_div_round_ph): Likewise.
11363 * config/i386/avx512fp16vlintrin.h: New header.
11364 * config/i386/i386-builtin-types.def (V16HF, V8HF, V32HF):
11365 Add new builtin types.
11366 * config/i386/i386-builtin.def: Add corresponding builtins.
11367 * config/i386/i386-expand.c
11368 (ix86_expand_args_builtin): Handle new builtin types.
11369 (ix86_expand_round_builtin): Likewise.
11370 * config/i386/immintrin.h: Include avx512fp16vlintrin.h
11371 * config/i386/sse.md (VFH): New mode_iterator.
11373 (avx512fmaskmode): Add HF vector modes.
11374 (avx512fmaskhalfmode): Likewise.
11375 (<plusminus_insn><mode>3<mask_name><round_name>): Adjust to for
11377 (*<plusminus_insn><mode>3<mask_name><round_name>): Likewise.
11378 (mul<mode>3<mask_name><round_name>): Likewise.
11379 (*mul<mode>3<mask_name><round_name>): Likewise.
11380 (div<mode>3): Likewise.
11381 (<sse>_div<mode>3<mask_name><round_name>): Likewise.
11382 * config/i386/subst.md (SUBST_V): Add HF vector modes.
11383 (SUBST_A): Likewise.
11384 (round_mode512bit_condition): Adjust for V32HFmode.
11386 2021-09-09 liuhongt <hongtao.liu@intel.com>
11389 * config/i386/sse.md (reduc_plus_scal_<mode>): Split to ..
11390 (reduc_plus_scal_v4sf): .. this, New define_expand.
11391 (reduc_plus_scal_v2df): .. and this, New define_expand.
11393 2021-09-09 liuhongt <hongtao.liu@intel.com>
11396 * config/i386/sse.md (*vec_extract<mode><ssescalarmodelower>_valign):
11399 2021-09-08 Jonathan Wakely <jwakely@redhat.com>
11402 * doc/trouble.texi (Copy Assignment): Fix description of
11403 behaviour and fix code in example.
11405 2021-09-08 Segher Boessenkool <segher@kernel.crashing.org>
11408 * config/rs6000/rs6000-logue.c (rs6000_emit_epilogue): For ELFv2 use
11409 r11 instead of r12 for restoring CR.
11411 2021-09-08 Jakub Jelinek <jakub@redhat.com>
11412 liuhongt <hongtao.liu@intel.com>
11415 * config/i386/i386.md (@xorsign<mode>3_1): Remove.
11416 * config/i386/i386-expand.c (ix86_expand_xorsign): Expand right away
11417 into AND with mask and XOR, using paradoxical subregs.
11418 (ix86_split_xorsign): Remove.
11419 * config/i386/i386-protos.h (ix86_split_xorsign): Remove.
11421 2021-09-08 Di Zhao <dizhao@os.amperecomputing.com>
11423 * tree-ssa-sccvn.c (vn_nary_op_insert_into): fix result compare
11425 2021-09-08 Jakub Jelinek <jakub@redhat.com>
11428 * config/i386/i386.md (xorsign<mode>3): If operands[1] is equal to
11429 operands[2], emit abs<mode>2 instead.
11430 (@xorsign<mode>3_1): Add early-clobbers for output operand, enable
11431 first alternative even for avx, add another alternative with
11432 =&Yv <- 0, Yv, Yvm constraints.
11433 * config/i386/i386-expand.c (ix86_split_xorsign): If op0 is equal
11434 to op1, emit vpandn instead.
11436 2021-09-08 liuhongt <hongtao.liu@intel.com>
11438 * config/i386/avx512fp16intrin.h (_mm_set_ph): New intrinsic.
11439 (_mm256_set_ph): Likewise.
11440 (_mm512_set_ph): Likewise.
11441 (_mm_setr_ph): Likewise.
11442 (_mm256_setr_ph): Likewise.
11443 (_mm512_setr_ph): Likewise.
11444 (_mm_set1_ph): Likewise.
11445 (_mm256_set1_ph): Likewise.
11446 (_mm512_set1_ph): Likewise.
11447 (_mm_setzero_ph): Likewise.
11448 (_mm256_setzero_ph): Likewise.
11449 (_mm512_setzero_ph): Likewise.
11450 (_mm_set_sh): Likewise.
11451 (_mm_load_sh): Likewise.
11452 (_mm_store_sh): Likewise.
11453 * config/i386/i386-builtin-types.def (V8HF): New type.
11454 (DEF_FUNCTION_TYPE (V8HF, V8HI)): New builtin function type
11455 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
11456 Support vector HFmodes.
11457 (ix86_expand_vector_init_one_nonzero): Likewise.
11458 (ix86_expand_vector_init_one_var): Likewise.
11459 (ix86_expand_vector_init_interleave): Likewise.
11460 (ix86_expand_vector_init_general): Likewise.
11461 (ix86_expand_vector_set): Likewise.
11462 (ix86_expand_vector_extract): Likewise.
11463 (ix86_expand_vector_init_concat): Likewise.
11464 (ix86_expand_sse_movcc): Handle vector HFmodes.
11465 (ix86_expand_vector_set_var): Ditto.
11466 * config/i386/i386-modes.def: Add HF vector modes in comment.
11467 * config/i386/i386.c (classify_argument): Add HF vector modes.
11468 (ix86_hard_regno_mode_ok): Allow HF vector modes for AVX512FP16.
11469 (ix86_vector_mode_supported_p): Likewise.
11470 (ix86_set_reg_reg_cost): Handle vector HFmode.
11471 (ix86_get_ssemov): Handle vector HFmode.
11472 (function_arg_advance_64): Pass unamed V16HFmode and V32HFmode
11474 (function_arg_advance_32): Pass V8HF/V16HF/V32HF by sse reg for 32bit
11476 (function_arg_advance_32): Ditto.
11477 * config/i386/i386.h (VALID_AVX512FP16_REG_MODE): New.
11478 (VALID_AVX256_REG_OR_OI_MODE): Rename to ..
11479 (VALID_AVX256_REG_OR_OI_VHF_MODE): .. this, and add V16HF.
11480 (VALID_SSE2_REG_VHF_MODE): New.
11481 (VALID_AVX512VL_128_REG_MODE): Add V8HF and TImode.
11482 (SSE_REG_MODE_P): Add vector HFmode.
11483 * config/i386/i386.md (mode): Add HF vector modes.
11484 (MODE_SIZE): Likewise.
11485 (ssemodesuffix): Add ph suffix for HF vector modes.
11486 * config/i386/sse.md (VFH_128): New mode iterator.
11487 (VMOVE): Adjust for HF vector modes.
11489 (V_256_512): Likewise.
11490 (avx512): Likewise.
11491 (avx512fmaskmode): Likewise.
11492 (shuffletype): Likewise.
11493 (sseinsnmode): Likewise.
11494 (ssedoublevecmode): Likewise.
11495 (ssehalfvecmode): Likewise.
11496 (ssehalfvecmodelower): Likewise.
11497 (ssePScmode): Likewise.
11498 (ssescalarmode): Likewise.
11499 (ssescalarmodelower): Likewise.
11500 (sseintprefix): Likewise.
11502 (bcstscalarsuff): Likewise.
11503 (xtg_mode): Likewise.
11504 (VI12HF_AVX512VL): New mode_iterator.
11505 (VF_AVX512FP16): Likewise.
11507 (VIHF_256): Likewise.
11508 (VIHF_AVX512BW): Likewise.
11509 (V16_256): Likewise.
11510 (V32_512): Likewise.
11511 (sseintmodesuffix): New mode_attr.
11512 (sse): Add scalar and vector HFmodes.
11513 (ssescalarmode): Add vector HFmode mapping.
11514 (ssescalarmodesuffix): Add sh suffix for HFmode.
11515 (*<sse>_vm<insn><mode>3): Use VFH_128.
11516 (*<sse>_vm<multdiv_mnemonic><mode>3): Likewise.
11517 (*ieee_<ieee_maxmin><mode>3): Likewise.
11518 (<avx512>_blendm<mode>): New define_insn.
11519 (vec_setv8hf): New define_expand.
11520 (vec_set<mode>_0): New define_insn for HF vector set.
11521 (*avx512fp16_movsh): Likewise.
11522 (avx512fp16_movsh): Likewise.
11523 (vec_extract_lo_v32hi): Rename to ...
11524 (vec_extract_lo_<mode>): ... this, and adjust to allow HF
11526 (vec_extract_hi_v32hi): Likewise.
11527 (vec_extract_hi_<mode>): Likewise.
11528 (vec_extract_lo_v16hi): Likewise.
11529 (vec_extract_lo_<mode>): Likewise.
11530 (vec_extract_hi_v16hi): Likewise.
11531 (vec_extract_hi_<mode>): Likewise.
11532 (vec_set_hi_v16hi): Likewise.
11533 (vec_set_hi_<mode>): Likewise.
11534 (vec_set_lo_v16hi): Likewise.
11535 (vec_set_lo_<mode>): Likewise.
11536 (*vec_extract<mode>_0): New define_insn_and_split for HF
11538 (*vec_extracthf): New define_insn.
11539 (VEC_EXTRACT_MODE): Add HF vector modes.
11540 (PINSR_MODE): Add V8HF.
11541 (sse2p4_1): Likewise.
11542 (pinsr_evex_isa): Likewise.
11543 (<sse2p4_1>_pinsr<ssemodesuffix>): Adjust to support
11544 insert for V8HFmode.
11545 (pbroadcast_evex_isa): Add HF vector modes.
11546 (AVX2_VEC_DUP_MODE): Likewise.
11547 (VEC_INIT_MODE): Likewise.
11548 (VEC_INIT_HALF_MODE): Likewise.
11549 (avx2_pbroadcast<mode>): Adjust to support HF vector mode
11551 (avx2_pbroadcast<mode>_1): Likewise.
11552 (<avx512>_vec_dup<mode>_1): Likewise.
11553 (<avx512>_vec_dup<mode><mask_name>): Likewise.
11554 (<mask_codefor><avx512>_vec_dup_gpr<mode><mask_name>):
11557 2021-09-08 Guo, Xuepeng <xuepeng.guo@intel.com>
11558 H.J. Lu <hongjiu.lu@intel.com>
11559 Liu Hongtao <hongtao.liu@intel.com>
11560 Wang Hongyu <hongyu.wang@intel.com>
11561 Xu Dianhong <dianhong.xu@intel.com>
11563 * common/config/i386/cpuinfo.h (get_available_features):
11564 Detect FEATURE_AVX512FP16.
11565 * common/config/i386/i386-common.c
11566 (OPTION_MASK_ISA_AVX512FP16_SET,
11567 OPTION_MASK_ISA_AVX512FP16_UNSET,
11568 OPTION_MASK_ISA2_AVX512FP16_SET,
11569 OPTION_MASK_ISA2_AVX512FP16_UNSET): New.
11570 (OPTION_MASK_ISA2_AVX512BW_UNSET,
11571 OPTION_MASK_ISA2_AVX512BF16_UNSET): Add AVX512FP16.
11572 (ix86_handle_option): Handle -mavx512fp16.
11573 * common/config/i386/i386-cpuinfo.h (enum processor_features):
11574 Add FEATURE_AVX512FP16.
11575 * common/config/i386/i386-isas.h: Add entry for AVX512FP16.
11576 * config.gcc: Add avx512fp16intrin.h.
11577 * config/i386/avx512fp16intrin.h: New intrinsic header.
11578 * config/i386/cpuid.h: Add bit_AVX512FP16.
11579 * config/i386/i386-builtin-types.def: (FLOAT16): New primitive type.
11580 * config/i386/i386-builtins.c: Support _Float16 type for i386
11582 (ix86_register_float16_builtin_type): New function.
11583 (ix86_float16_type_node): New.
11584 * config/i386/i386-c.c (ix86_target_macros_internal): Define
11586 * config/i386/i386-expand.c (ix86_expand_branch): Support
11588 (ix86_prepare_fp_compare_args): Adjust TARGET_SSE_MATH &&
11589 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
11590 (ix86_expand_fp_movcc): Ditto.
11591 * config/i386/i386-isa.def: Add PTA define for AVX512FP16.
11592 * config/i386/i386-options.c (isa2_opts): Add -mavx512fp16.
11593 (ix86_valid_target_attribute_inner_p): Add avx512fp16 attribute.
11594 * config/i386/i386.c (ix86_get_ssemov): Use
11595 vmovdqu16/vmovw/vmovsh for HFmode/HImode scalar or vector.
11596 (ix86_get_excess_precision): Use
11597 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when TARGET_AVX512FP16
11599 (sse_store_index): Use SFmode cost for HFmode cost.
11600 (inline_memory_move_cost): Add HFmode, and perfer SSE cost over
11601 GPR cost for HFmode.
11602 (ix86_hard_regno_mode_ok): Allow HImode in sse register.
11603 (ix86_mangle_type): Add manlging for _Float16 type.
11604 (inline_secondary_memory_needed): No memory is needed for
11605 16bit movement between gpr and sse reg under
11607 (ix86_multiplication_cost): Adjust TARGET_SSE_MATH &&
11608 SSE_FLOAT_MODE_P to SSE_FLOAT_MODE_SSEMATH_OR_HF_P.
11609 (ix86_division_cost): Ditto.
11610 (ix86_rtx_costs): Ditto.
11611 (ix86_add_stmt_cost): Ditto.
11612 (ix86_optab_supported_p): Ditto.
11613 * config/i386/i386.h (VALID_AVX512F_SCALAR_MODE): Add HFmode.
11614 (SSE_FLOAT_MODE_SSEMATH_OR_HF_P): Add HFmode.
11615 (PTA_SAPPHIRERAPIDS): Add PTA_AVX512FP16.
11616 * config/i386/i386.md (mode): Add HFmode.
11617 (MODE_SIZE): Add HFmode.
11618 (isa): Add avx512fp16.
11619 (enabled): Handle avx512fp16.
11620 (ssemodesuffix): Add sh suffix for HFmode.
11621 (comm): Add mult, div.
11622 (plusminusmultdiv): New code iterator.
11623 (insn): Add mult, div.
11624 (*movhf_internal): Adjust for avx512fp16 instruction.
11625 (*movhi_internal): Ditto.
11626 (*cmpi<unord>hf): New define_insn for HFmode.
11627 (*ieee_s<ieee_maxmin>hf3): Likewise.
11628 (extendhf<mode>2): Likewise.
11629 (trunc<mode>hf2): Likewise.
11630 (float<floatunssuffix><mode>hf2): Likewise.
11631 (*<insn>hf): Likewise.
11632 (cbranchhf4): New expander.
11633 (movhfcc): Likewise.
11634 (<insn>hf3): Likewise.
11635 (mulhf3): Likewise.
11636 (divhf3): Likewise.
11637 * config/i386/i386.opt: Add mavx512fp16.
11638 * config/i386/immintrin.h: Include avx512fp16intrin.h.
11639 * doc/invoke.texi: Add mavx512fp16.
11640 * doc/extend.texi: Add avx512fp16 Usage Notes.
11642 2021-09-08 liuhongt <hongtao.liu@intel.com>
11644 * common.opt: Support -fexcess-precision=16.
11645 * config/aarch64/aarch64.c (aarch64_excess_precision): Return
11646 FLT_EVAL_METHOD_PROMOTE_TO_FLOAT16 when
11647 EXCESS_PRECISION_TYPE_FLOAT16.
11648 * config/arm/arm.c (arm_excess_precision): Ditto.
11649 * config/i386/i386.c (ix86_get_excess_precision): Ditto.
11650 * config/m68k/m68k.c (m68k_excess_precision): Issue an error
11651 when EXCESS_PRECISION_TYPE_FLOAT16.
11652 * config/s390/s390.c (s390_excess_precision): Ditto.
11653 * coretypes.h (enum excess_precision_type): Add
11654 EXCESS_PRECISION_TYPE_FLOAT16.
11655 * doc/tm.texi (TARGET_C_EXCESS_PRECISION): Update documents.
11656 * doc/tm.texi.in (TARGET_C_EXCESS_PRECISION): Ditto.
11657 * doc/extend.texi (Half-Precision): Document
11658 -fexcess-precision=16.
11659 * flag-types.h (enum excess_precision): Add
11660 EXCESS_PRECISION_FLOAT16.
11661 * target.def (excess_precision): Update document.
11662 * tree.c (excess_precision_type): Set excess_precision_type to
11663 EXCESS_PRECISION_FLOAT16 when -fexcess-precision=16.
11665 2021-09-08 liuhongt <hongtao.liu@intel.com>
11667 * doc/extend.texi: (@node Floating Types): Adjust the wording.
11668 (@node Half-Precision): Ditto.
11670 2021-09-07 Takayuki 'January June' Suwa <jjsuwa_sys3175@yahoo.co.jp>
11673 * config/xtensa/xtensa.c (xtensa_emit_move_sequence): Add
11674 'CONST_INT_P (src)' to the condition of the block that tries to
11675 eliminate literal when loading integer contant.
11677 2021-09-07 David Faust <david.faust@oracle.com>
11679 * doc/extend.texi (BPF Type Attributes) New node.
11680 Document new preserve_access_index attribute.
11681 Document new preserve_access_index builtin.
11682 * doc/invoke.texi: Document -mco-re and -mno-co-re options.
11684 2021-09-07 David Faust <david.faust@oracle.com>
11686 * config/bpf/bpf.c: Adjust includes.
11687 (bpf_handle_preserve_access_index_attribute): New function.
11688 (bpf_attribute_table): Use it here.
11689 (bpf_builtins): Add BPF_BUILTIN_PRESERVE_ACCESS_INDEX.
11690 (bpf_option_override): Handle "-mco-re" option.
11691 (bpf_asm_init_sections): New.
11692 (TARGET_ASM_INIT_SECTIONS): Redefine.
11693 (bpf_file_end): New.
11694 (TARGET_ASM_FILE_END): Redefine.
11695 (bpf_init_builtins): Add "__builtin_preserve_access_index".
11696 (bpf_core_compute, bpf_core_get_index): New.
11697 (is_attr_preserve_access): New.
11698 (bpf_expand_builtin): Handle new builtins.
11699 (bpf_core_newdecl, bpf_core_is_maybe_aggregate_access): New.
11700 (bpf_core_walk): New.
11701 (bpf_resolve_overloaded_builtin): New.
11702 (TARGET_RESOLVE_OVERLOADED_BUILTIN): Redefine.
11703 (handle_attr): New.
11704 (pass_bpf_core_attr): New RTL pass.
11705 * config/bpf/bpf-passes.def: New file.
11706 * config/bpf/bpf-protos.h (make_pass_bpf_core_attr): New.
11707 * config/bpf/coreout.c: New file.
11708 * config/bpf/coreout.h: Likewise.
11709 * config/bpf/t-bpf (TM_H): Add $(srcdir)/config/bpf/coreout.h.
11710 (coreout.o): New rule.
11711 (PASSES_EXTRA): Add $(srcdir)/config/bpf/bpf-passes.def.
11712 * config.gcc (bpf): Add coreout.h to extra_headers.
11713 Add coreout.o to extra_objs.
11714 Add $(srcdir)/config/bpf/coreout.c to target_gtfiles.
11716 2021-09-07 David Faust <david.faust@oracle.com>
11718 * btfout.c (get_btf_id): Function is no longer static.
11719 * ctfc.h: Expose it here.
11721 2021-09-07 David Faust <david.faust@oracle.com>
11723 * ctfc.c (ctf_lookup_tree_type): New function.
11724 * ctfc.h: Likewise.
11726 2021-09-07 David Faust <david.faust@oracle.com>
11728 * ctfc.c (ctf_dtd_lookup): Function is no longer static.
11729 * ctfc.h: Analogous change.
11731 2021-09-07 David Faust <david.faust@oracle.com>
11733 * dwarf2out.c (lookup_type_die): Function is no longer static.
11734 * dwarf2out.h: Expose it here.
11736 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
11738 * dwarf2ctf.c (ctf_debug_finalize): Make it static.
11739 (ctf_debug_early_finish): New definition.
11740 (ctf_debug_finish): Likewise.
11741 * dwarf2ctf.h (ctf_debug_finalize): Remove declaration.
11742 (ctf_debug_early_finish): New declaration.
11743 (ctf_debug_finish): Likewise.
11744 * dwarf2out.c (dwarf2out_finish): Invoke ctf_debug_finish.
11745 (dwarf2out_early_finish): Invoke ctf_debug_early_finish.
11747 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
11749 * config/bpf/bpf.c (bpf_option_override): For BPF backend, disable LTO
11750 support when compiling for CO-RE.
11751 * config/bpf/bpf.opt: Add new command line option -mco-re.
11753 2021-09-07 Indu Bhagat <indu.bhagat@oracle.com>
11755 * flag-types.h (enum debug_info_type): Add new enum
11756 DINFO_TYPE_BTF_WITH_CORE.
11757 (BTF_WITH_CORE_DEBUG): New bitmask.
11758 * flags.h (btf_with_core_debuginfo_p): New declaration.
11759 * opts.c (btf_with_core_debuginfo_p): New definition.
11761 2021-09-07 Jason Merrill <jason@redhat.com>
11763 * tree.h (error_operand_p): Change to inline function.
11765 2021-09-07 Aldy Hernandez <aldyh@redhat.com>
11767 * tree-ssa-threadedge.c (forwarder_block_p): Rename to...
11768 (empty_block_with_phis_p): ...this.
11769 (potentially_threadable_block): Same.
11770 (jump_threader::thread_through_normal_block): Same.
11772 2021-09-07 Eric Botcazou <ebotcazou@adacore.com>
11775 * dwarf2out.c (mark_base_types): New overloaded function.
11776 (dwarf2out_early_finish): Invoke it on the COMDAT type list as well
11777 as the compilation unit, and call move_marked_base_types afterward.
11779 2021-09-07 H.J. Lu <hjl.tools@gmail.com>
11782 * config/i386/i386-expand.c (ix86_expand_convert_uns_sisf_sse):
11784 (ix86_expand_vector_convert_uns_vsivsf): Likewise.
11786 2021-09-07 Richard Biener <rguenther@suse.de>
11788 PR tree-optimization/102226
11789 * tree-vect-loop.c (vect_transform_cycle_phi): Record
11790 the converted value for the epilogue PHI use.
11792 2021-09-07 Martin Liska <mliska@suse.cz>
11794 PR gcov-profile/80223
11795 * ipa-inline.c (can_inline_edge_p): Similarly to sanitizer
11796 options, do not inline when no_profile_instrument_function
11797 attributes are different in early inliner. It's fine to inline
11798 it after PGO instrumentation.
11800 2021-09-07 Richard Biener <rguenther@suse.de>
11802 PR tree-optimization/101555
11803 * tree-ssa-pre.c (translate_vuse_through_block): Do not
11804 perform an alias walk to determine the validity of the
11805 mem at the start of the block which is already guaranteed
11806 by means of prune_clobbered_mems.
11807 (phi_translate_1): Pass edge to translate_vuse_through_block.
11809 2021-09-07 Xionghu Luo <luoxhu@linux.ibm.com>
11812 * config/rs6000/rs6000.md (fmod<mode>3): New define_expand.
11813 (remainder<mode>3): Likewise.
11815 2021-09-07 YunQiang Su <yunqiang.su@cipunited.com>
11817 * config/mips/mips.c (mips_file_start): add .module for
11820 2021-09-06 Roger Sayle <roger@nextmovesoftware.com>
11822 * wide-int.cc (wi::clz): Reorder tests to ensure the result
11823 is zero for all negative values.
11825 2021-09-06 Tobias Burnus <tobias@codesourcery.com>
11827 * doc/invoke.texi (-foffload-options): Fix @opindex.
11829 2021-09-06 H.J. Lu <hjl.tools@gmail.com>
11832 * config/i386/i386-expand.c (ix86_split_xorsign): Use operands[2].
11833 * config/i386/i386.md (@xorsign<mode>3_1): Add non-destructive
11834 source alternative for AVX.
11836 2021-09-06 liuhongt <hongtao.liu@intel.com>
11838 PR middle-end/102182
11839 * optabs.c (expand_fix): Add from1 to avoid from being
11842 2021-09-06 Eric Botcazou <ebotcazou@adacore.com>
11844 * dwarf2out.c (modified_type_die): Deal with all array types earlier
11845 and use local variable consistently throughout the function.
11847 2021-09-06 Jakub Jelinek <jakub@redhat.com>
11849 PR tree-optimization/102207
11850 * match.pd: Don't demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
11851 were promoted from signed to wider unsigned type.
11853 2021-09-06 Andrew Pinski <apinski@marvell.com>
11855 PR tree-optimization/63184
11856 * match.pd: Add simplification of pointer_diff of two pointer_plus
11857 with addr_expr in the first operand of each pointer_plus.
11858 Add simplificatoin of ne/eq of two pointer_plus with addr_expr
11859 in the first operand of each pointer_plus.
11861 2021-09-06 Richard Biener <rguenther@suse.de>
11863 PR tree-optimization/102176
11864 * tree-vect-slp.c (vect_slp_gather_vectorized_scalar_stmts):
11866 (vect_bb_slp_scalar_cost): Use the computed set of
11867 vectorized scalar stmts instead of relying on the out-of-date
11868 and not accurate PURE_SLP_STMT.
11869 (vect_bb_vectorization_profitable_p): Compute the set
11870 of vectorized scalar stmts.
11872 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
11874 * gimple-range-path.cc (path_range_query::range_of_stmt): Remove
11875 GIMPLE_COND special casing.
11876 (path_range_query::range_defined_in_block): Use range_of_stmt
11877 instead of calling fold_range directly.
11879 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
11881 * gimple-range-path.cc (path_range_query::range_of_expr): Set
11882 m_undefined_path when appropriate.
11883 (path_range_query::internal_range_of_expr): Copy from range_of_expr.
11884 (path_range_query::unreachable_path_p): New.
11885 (path_range_query::precompute_ranges): Set m_undefined_path.
11886 * gimple-range-path.h (path_range_query::unreachable_path_p): New.
11887 (path_range_query::internal_range_of_expr): New.
11888 * tree-ssa-threadbackward.c (back_threader::find_taken_edge_cond):
11889 Use unreachable_path_p.
11891 2021-09-05 Aldy Hernandez <aldyh@redhat.com>
11893 * tree-ssa-threadbackward.c (back_threader::maybe_register_path):
11894 Remove argument and call find_taken_edge.
11895 (back_threader::resolve_phi): Do not calculate taken edge before
11896 calling maybe_register_path.
11897 (back_threader::find_paths_to_names): Same.
11899 2021-09-05 Jeff Law <jlaw@localhost.localdomain>
11901 * config/h8300/h8300.md (QHSI2 mode iterator): New mode iterator.
11902 * config/h8300/testcompare.md (store_c): Update name, use new
11904 (store_neg_c, store_shifted_c): New patterns.
11906 2021-09-03 Segher Boessenkool <segher@kernel.crashing.org>
11909 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): On ELFv2 use r11
11910 instead of r12 for CR save, in all cases.
11912 2021-09-03 Andrew Pinski <apinski@marvell.com>
11914 * config/aarch64/aarch64-sve-builtins.cc (register_vector_type):
11915 Handle error_mark_node as the type of the type_decl.
11917 2021-09-03 Andrew Pinski <apinski@marvell.com>
11919 * config/aarch64/aarch64-builtins.c (struct aarch64_simd_type_info):
11921 (aarch64_simd_types): Likewise.
11922 (aarch64_simd_intOI_type_node): Likewise.
11923 (aarch64_simd_intCI_type_node): Likewise.
11924 (aarch64_simd_intXI_type_node): Likewise.
11925 * config/aarch64/aarch64.h (aarch64_fp16_type_node): Likewise.
11926 (aarch64_fp16_ptr_type_node): Likewise.
11927 (aarch64_bf16_type_node): Likewise.
11928 (aarch64_bf16_ptr_type_node): Likewise.
11930 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11932 * range-op.cc (operator_minus::op1_op2_relation_effect): Abstract
11934 (minus_op1_op2_relation_effect): ...here.
11935 (class operator_pointer_diff): New.
11936 (operator_pointer_diff::op1_op2_relation_effect): Call
11937 minus_op1_op2_relation_effect.
11938 (integral_table::integral_table): Add entry for POINTER_DIFF_EXPR.
11940 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11942 * tree-ssa-threadbackward.c (back_threader::thread_through_all_blocks):
11943 Add may_peel_loop_headers.
11944 (back_threader_registry::thread_through_all_blocks): Same.
11945 (try_thread_blocks): Pass may_peel_loop_headers argument.
11946 (pass_early_thread_jumps::execute): Same.
11948 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11950 * tree-ssa-threadedge.c (has_phis_p): New.
11951 (forwarder_block_p): New.
11952 (potentially_threadable_block): Call forwarder_block_p.
11953 (jump_threader::thread_around_empty_blocks): Call has_phis_p.
11954 (jump_threader::thread_through_normal_block): Call
11957 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11959 * tree-ssa-threadbackward.c (back_threader::dump): New.
11960 (back_threader::debug): New.
11961 (back_threader_profitability::profitable_path_p): Dump blocks
11962 even if we are bailing early.
11964 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11966 * tree-ssa-threadupdate.c (cancel_thread): New.
11967 (jump_thread_path_registry::thread_block_1): Use cancel_thread.
11968 (jump_thread_path_registry::mark_threaded_blocks): Same.
11969 (jump_thread_path_registry::register_jump_thread): Same.
11971 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11973 * tree-ssa-threadedge.c (jt_state::push): Only call methods for
11974 which objects are available.
11975 (jt_state::pop): Same.
11976 (jt_state::register_equiv): Same.
11977 (jt_state::register_equivs_on_edge): Same.
11979 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11981 * tree-ssa-threadedge.c (jump_threader::thread_across_edge):
11982 Move pop until after a thread is registered.
11984 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11986 * tree-ssa-threadupdate.c (debug): New.
11988 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11990 * gimple-range-trace.cc (push_dump_file::push_dump_file): New.
11991 (push_dump_file::~push_dump_file): New.
11992 (dump_ranger): Change dump_file temporarily while dumping
11994 * gimple-range-trace.h (class push_dump_file): New.
11996 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
11998 * gimple-range-trace.cc (debug_seed_ranger): Remove static.
11999 (dump_ranger): Dump function name.
12001 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
12003 * gimple-range-path.cc (path_range_query::range_defined_in_block):
12004 Adjust for non-null.
12005 (path_range_query::adjust_for_non_null_uses): New.
12006 (path_range_query::precompute_ranges): Call
12007 adjust_for_non_null_uses.
12008 * gimple-range-path.h: Add m_non_null and
12009 adjust_for_non_null_uses.
12011 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
12013 * gimple-range-path.cc (path_range_query::dump): Dump path
12015 (path_range_query::precompute_ranges): Dump entire path.
12017 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
12019 * value-relation.cc (relation_oracle::debug): New.
12020 * value-relation.h (relation_oracle::debug): New.
12022 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
12024 * tree-ssa-loop-ch.c: Remove unnecessary include file.
12026 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
12028 * gimple-range-fold.cc (fold_using_range::postfold_gcond_edges):
12029 Skip statements with no defining BB.
12030 * gimple-range-path.cc (path_range_query::range_defined_in_block):
12031 Do not get confused by statements with no defining BB.
12033 2021-09-03 Aldy Hernandez <aldyh@redhat.com>
12035 * gimple-range-fold.cc (adjust_imagpart_expr): Move from
12036 gimple_range_adjustment. Add support for constants.
12037 (adjust_realpart_expr): New.
12038 (gimple_range_adjustment): Move IMAGPART_EXPR code to
12039 adjust_imagpart_expr.
12040 * range-op.cc (integral_table::integral_table): Add entry for
12043 2021-09-03 Jakub Jelinek <jakub@redhat.com>
12045 * omp-expand.c (expand_omp_atomic_pipeline): Use
12046 IFN_ATOMIC_COMPARE_EXCHANGE instead of
12047 BUILT_IN_SYNC_VAL_COMPARE_AND_SWAP_? so that memory order
12050 2021-09-03 Jakub Jelinek <jakub@redhat.com>
12053 * tree.h (DECL_FIELD_ABI_IGNORED): Changed into rvalue only macro
12054 that is false if DECL_BIT_FIELD.
12055 (SET_DECL_FIELD_ABI_IGNORED, DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD,
12056 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD): Define.
12057 * tree-streamer-out.c (pack_ts_decl_common_value_fields): For
12058 DECL_BIT_FIELD stream DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead
12059 of DECL_FIELD_ABI_IGNORED.
12060 * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Use
12061 SET_DECL_FIELD_ABI_IGNORED instead of writing to
12062 DECL_FIELD_ABI_IGNORED and for DECL_BIT_FIELD use
12063 SET_DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead.
12064 * lto-streamer-out.c (hash_tree): For DECL_BIT_FIELD hash
12065 DECL_FIELD_CXX_ZERO_WIDTH_BIT_FIELD instead of DECL_FIELD_ABI_IGNORED.
12067 2021-09-03 liuhongt <hongtao.liu@intel.com>
12070 * config/i386/amxbf16intrin.h : Remove macro check for __AMX_BF16__.
12071 * config/i386/amxint8intrin.h : Remove macro check for __AMX_INT8__.
12072 * config/i386/amxtileintrin.h : Remove macro check for __AMX_TILE__.
12074 2021-09-02 Martin Sebor <msebor@redhat.com>
12076 PR tree-optimization/17506
12078 * tree-ssa-uninit.c (warn_uninit): Remove conditional guarding note.
12080 2021-09-02 Richard Biener <rguenther@suse.de>
12082 * tree-ssa-loop-im.c (fill_always_executed_in_1): Refine
12083 fix for PR78185 and continue processing when leaving
12084 finite inner loops.
12086 2021-09-02 Jakub Jelinek <jakub@redhat.com>
12088 PR tree-optimization/99591
12089 * match.pd: Demote operands of IFN_{ADD,SUB,MUL}_OVERFLOW if they
12092 2021-09-02 Richard Biener <rguenther@suse.de>
12095 2021-09-02 Richard Biener <rguenther@suse.de>
12097 PR tree-optimization/102155
12098 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
12099 over a part of the RPO array and do not recurse here.
12100 Dump blocks marked as always executed.
12101 (fill_always_executed_in): Walk over the RPO array and
12102 process loops whose header we run into.
12103 (loop_invariant_motion_in_fun): Compute the first RPO
12104 using rev_post_order_and_mark_dfs_back_seme in iteration
12105 order and pass that to fill_always_executed_in.
12107 2021-09-02 liuhongt <hongtao.liu@intel.com>
12109 * config/i386/i386-modes.def (FLOAT_MODE): Define ieee HFmode.
12110 * config/i386/i386.c (enum x86_64_reg_class): Add
12111 X86_64_SSEHF_CLASS.
12112 (merge_classes): Handle X86_64_SSEHF_CLASS.
12113 (examine_argument): Ditto.
12114 (construct_container): Ditto.
12115 (classify_argument): Ditto, and set HFmode/HCmode to
12116 X86_64_SSEHF_CLASS.
12117 (function_value_32): Return _FLoat16/Complex Float16 by
12119 (function_value_64): Return _Float16/Complex Float16 by SSE
12121 (ix86_print_operand): Handle CONST_DOUBLE HFmode.
12122 (ix86_secondary_reload): Require gpr as intermediate register
12123 to store _Float16 from sse register when sse4 is not
12125 (ix86_libgcc_floating_mode_supported_p): Enable _FLoat16 under
12127 (ix86_scalar_mode_supported_p): Ditto.
12128 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Defined.
12129 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add HFmode.
12130 (VALID_INT_MODE_P): Add HFmode and HCmode.
12131 * config/i386/i386.md (*pushhf_rex64): New define_insn.
12133 (*movhf_internal): Ditto.
12134 * doc/extend.texi (Half-Precision Floating Point): Documemt
12137 2021-09-02 Richard Biener <rguenther@suse.de>
12139 PR tree-optimization/102155
12140 * tree-ssa-loop-im.c (fill_always_executed_in_1): Iterate
12141 over a part of the RPO array and do not recurse here.
12142 Dump blocks marked as always executed.
12143 (fill_always_executed_in): Walk over the RPO array and
12144 process loops whose header we run into.
12145 (loop_invariant_motion_in_fun): Compute the first RPO
12146 using rev_post_order_and_mark_dfs_back_seme in iteration
12147 order and pass that to fill_always_executed_in.
12149 2021-09-02 YunQiang Su <syq@debian.org>
12152 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
12154 * config/mips/mips.c (mips_module_isa_name): New.
12155 mips_file_start: add .module mipsREV to all asm output
12157 2021-09-01 Jeff Law <jlaw@localhost.localdomain>
12159 PR tree-optimization/102152
12160 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Reduce a vector
12161 comparison to a scalar comparison before calling
12162 update_stmt_if_modified.
12164 2021-09-01 Andrew Pinski <apinski@marvell.com>
12167 * config/aarch64/aarch64.c (aarch64_expand_setmem):
12168 Check STRICT_ALIGNMENT before creating an overlapping
12171 2021-09-01 Martin Sebor <msebor@redhat.com>
12173 * gimple-ssa-warn-access.cc (get_size_range): Add argument.
12174 (check_access): Pass additional argument.
12175 (check_memop_access): Remove template and make a member function.
12176 (maybe_check_dealloc_call): Make a pass_waccess member function.
12177 (class pass_waccess): Add, rename, and remove members.
12178 (pass_waccess::pass_waccess): Adjust to name change.
12179 (pass_waccess::~pass_waccess): Same.
12180 (check_alloca): Make a member function.
12181 (check_alloc_size_call): Same.
12182 (check_strcat): Same.
12183 (check_strncat): Same.
12184 (check_stxcpy): Same.
12185 (check_stxncpy): Same.
12186 (check_strncmp): Same.
12187 (maybe_warn_rdwr_sizes): Rename...
12188 (pass_waccess::maybe_check_access_sizes): ...to this.
12189 (pass_waccess::check_call): Adjust to name changes.
12190 (pass_waccess::maybe_check_dealloc_call): Make a pass_waccess member
12192 (pass_waccess::execute): Adjust to name changes.
12193 * gimple-ssa-warn-access.h (check_memop_access): Remove.
12194 * pointer-query.cc (access_ref::phi): Handle null pointer.
12195 (access_ref::inform_access): Same.
12196 (pointer_query::put_ref): Modify a cached value, not a copy of it.
12197 (pointer_query::dump): New function.
12198 (compute_objsize_r): Avoid overwriting access_ref::bndrng. Cache
12200 * pointer-query.h (pointer_query::dump): Declare.
12201 * tree-ssa-strlen.c (get_range): Simplify. Use function query.
12202 (dump_strlen_info): Use function query.
12203 (printf_strlen_execute): Factor code out into pointer_query::put_ref.
12205 2021-09-01 Thomas Schwinge <thomas@codesourcery.com>
12207 * tree.c (walk_tree_1) <OMP_CLAUSE>: Simplify.
12209 2021-09-01 Iain Sandoe <iain@sandoe.co.uk>
12211 * doc/extend.texi: Document unavailable attribute.
12212 * print-tree.c (print_node): Handle unavailable attribute.
12213 * tree-core.h (struct tree_base): Add a bit to carry unavailability.
12214 * tree.c (error_unavailable_use): New.
12215 * tree.h (TREE_UNAVAILABLE): New.
12216 (error_unavailable_use): New.
12218 2021-09-01 Jakub Jelinek <jakub@redhat.com>
12220 PR tree-optimization/102124
12221 * tree-vect-patterns.c (vect_recog_widen_op_pattern): For ORIG_CODE
12222 MINUS_EXPR, if itype is unsigned with smaller precision than type,
12223 add an extra cast to signed variant of itype to ensure sign-extension.
12225 2021-09-01 Martin Liska <mliska@suse.cz>
12227 * graph.c (draw_cfg_node_succ_edges): Do not color fallthru
12228 edges and rather use colors for TRUE and FALSE edges.
12230 2021-09-01 Richard Biener <rguenther@suse.de>
12232 PR tree-optimization/93491
12233 * tree-ssa-pre.c (compute_avail): Set BB_MAY_NOTRETURN
12234 after processing the stmt itself. Do not consider
12235 pure functions possibly not returning. Properly avoid
12236 adding possibly trapping calls to EXP_GEN when there's
12237 a preceeding possibly not returning call.
12238 * tree-ssa-sccvn.c (vn_reference_may_trap): Conservatively
12241 2021-09-01 Richard Biener <rguenther@suse.de>
12243 PR tree-optimization/102139
12244 * tree-vectorizer.h (vec_base_alignments): Adjust hash-map
12245 type to record a std::pair of the stmt-info and the innermost
12247 (dr_vec_info::group): New member.
12248 * tree-vect-data-refs.c (vect_record_base_alignment): Adjust.
12249 (vect_compute_data_ref_alignment): Verify the recorded
12250 base alignment can be used.
12251 (data_ref_pair): Remove.
12252 (dr_group_sort_cmp): Adjust.
12253 (vect_analyze_data_ref_accesses): Store the group-ID in the
12254 dr_vec_info and operate on a vector of dr_vec_infos.
12256 2021-09-01 YunQiang Su <yunqiang.su@cipunited.com>
12258 * read-md.c (md_reader::handle_enum): support value assignation.
12259 * doc/md.texi: record define_c_enum value assignation support.
12261 2021-09-01 Jakub Jelinek <jakub@redhat.com>
12263 PR tree-optimization/102141
12264 * gimple-ssa-store-merging.c (bswap_view_convert): Add BEFORE
12265 argument. If false, emit stmts after gsi instead of before, and
12267 (bswap_replace): Adjust callers. When converting output of bswap,
12268 emit VIEW_CONVERT prepratation stmts after a copy of gsi instead
12271 2021-09-01 liuhongt <hongtao.liu@intel.com>
12273 * emit-rtl.c (validate_subreg): Get rid of all float-int
12276 2021-09-01 liuhongt <hongtao.liu@intel.com>
12279 2021-08-30 liuhongt <hongtao.liu@intel.com>
12281 * expmed.c (extract_bit_field_1): Make sure we're playing with
12282 integral modes before call extract_integral_bit_field.
12283 (extract_integral_bit_field): Add a parameter of type
12284 scalar_int_mode which corresponds to of tmode.
12285 And call extract_and_convert_fixed_bit_field instead of
12286 extract_fixed_bit_field and convert_extracted_bit_field.
12287 (extract_and_convert_fixed_bit_field): New function, it's a
12288 combination of extract_fixed_bit_field and
12289 convert_extracted_bit_field.
12291 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
12293 * tree.c (walk_tree_1) <OMP_CLAUSE_TILE>: Handle three operands.
12295 2021-08-31 Thomas Schwinge <thomas@codesourcery.com>
12297 * omp-general.h (omp_is_reference): Rename to...
12298 (omp_privatize_by_reference): ... this. Adjust all users...
12299 * omp-general.c: ... here, ...
12300 * gimplify.c: ... here, ...
12301 * omp-expand.c: ... here, ...
12302 * omp-low.c: ... here.
12304 2021-08-31 Martin Sebor <msebor@redhat.com>
12306 * gimple-ssa-warn-access.cc (maybe_warn_alloc_args_overflow): Test
12307 pointer element for equality to zero, not that of the cotaining
12310 2021-08-31 Martin Sebor <msebor@redhat.com>
12312 * gcc-rich-location.h (gcc_rich_location): Make ctor explicit.
12314 2021-08-31 Martin Sebor <msebor@redhat.com>
12316 * function.h (function): Add comments.
12317 (get_range_query): Same. Add attribute returns nonnull.
12319 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
12321 * expr.c (convert_modes): Don't use subreg_promoted_mode on a
12322 SUBREG if it can't be guaranteed to a SUBREG_PROMOTED_VAR_P set.
12323 Instead use the standard (safer) is_a <scalar_int_mode> idiom.
12325 2021-08-31 Jeff Law <jlaw@localhost.localdomain>
12327 * config.gcc (cris-*-elf, cris-*-none): Remove dbxelf.h from
12329 (m32r-*-elf, m32rle-*-elf, m32r-*-linux): Likewise.
12330 (mn10300-*-*, am33_2.0-*-linux*): Likewise.
12331 (xtensa*-*-elf, xtensa*-*-linux, xtensa*-*-uclinux): Likewise.
12332 (m32c-*-elf*, m32c-*-rtems*): Likewise.
12333 * config/cris/cris.h (DBX_NO_XREFS): Remove.
12334 (DBX_CONTIN_LENGTH, DBX_CONTIN_CHAR): Likewise.
12335 * config/m32r/m32r.h (DBXOUT_SOURCE_LINE): Likewise.
12336 (DBX_DEBUGGING_INFO, DBX_CONTIN_LENGTH): Likewise.
12337 * config/mn10300/mn10300.h (DEFAULT_GDB_EXTENSIONS): Likewise.
12338 * config/mn10300/linux.h (DBX_REGISTER_NAMES): Likewise.
12340 2021-08-31 Marcel Vollweiler <marcel@codesourcery.com>
12342 * gimplify.c (gimplify_scan_omp_clauses): Error handling. 'ancestor' only
12343 allowed on target constructs and only with particular other clauses.
12344 * omp-expand.c (expand_omp_target): Output of 'sorry, not supported' if
12345 'ancestor' is used.
12346 * omp-low.c (check_omp_nesting_restrictions): Error handling. No nested OpenMP
12347 structs when 'ancestor' is used.
12348 (scan_omp_1_stmt): No usage of OpenMP runtime routines in a target region when
12349 'ancestor' is used.
12350 * tree-pretty-print.c (dump_omp_clause): Append 'ancestor'.
12351 * tree.h (OMP_CLAUSE_DEVICE_ANCESTOR): Define macro.
12353 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
12355 * expr.c (convert_modes): Preserve SUBREG_PROMOTED_VAR_P when
12356 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
12358 * simplify-rtx.c (simplify_unary_operation_1) [SIGN_EXTEND]:
12359 Likewise, preserve SUBREG_PROMOTED_VAR_P when creating a (wider)
12360 partial subreg from a SUBREG_PROMOTED_VAR_P subreg. Generate
12361 SIGN_EXTEND of the SUBREG_REG when a subreg would be paradoxical.
12362 [ZERO_EXTEND]: Likewise, preserve SUBREG_PROMOTED_VAR_P when
12363 creating a (wider) partial subreg from a SUBREG_PROMOTED_VAR_P
12364 subreg. Generate ZERO_EXTEND of the SUBREG_REG when a subreg
12365 would be paradoxical.
12367 2021-08-31 Roger Sayle <roger@nextmovesoftware.com>
12369 * combine.c (combine_simplify_rtx): Avoid converting an explicit
12370 TRUNCATE into a lowpart SUBREG on !TRULY_NOOP_TRUNCATION targets.
12371 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12373 2021-08-31 Richard Biener <rguenther@suse.de>
12375 PR tree-optimization/102142
12376 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Fix
12377 condition under which to unset the visited flag.
12379 2021-08-31 Richard Biener <rguenther@suse.de>
12381 PR middle-end/102129
12382 * tree-ssa-ter.c (find_replaceable_in_bb): Do not move
12383 possibly trapping expressions across calls.
12385 2021-08-31 Jakub Jelinek <jakub@redhat.com>
12387 PR tree-optimization/102134
12388 * tree-ssa-ccp.c (bit_value_binop) <case RSHIFT_EXPR>: If sgn is
12389 UNSIGNED and r1val | r1mask has MSB set, ensure lzcount doesn't
12392 2021-08-31 Andrew Pinski <apinski@marvell.com>
12395 * collect-utils.c (setup_signals): New declaration.
12396 * collect-utils.h (setup_signals): New function.
12397 * collect2.c (handler): Delete.
12398 (main): Instead of manually setting up the signals,
12399 just call setup_signals.
12400 * lto-wrapper.c (main): Likewise.
12402 2021-08-31 Andrew Pinski <apinski@marvell.com>
12405 * config/i386/i386-protos.h (x86_output_aligned_bss):
12406 Change align argument to unsigned type.
12407 (x86_elf_aligned_decl_common): Likewise.
12408 * config/i386/i386.c (x86_elf_aligned_decl_common): Likewise.
12409 (x86_output_aligned_bss): Likewise.
12411 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
12413 * config/mips/mips.c (mips_module_isa_name): New.
12414 mips_file_start: add .module mipsREV to all asm output
12416 2021-08-31 YunQiang Su <yunqiang.su@cipunited.com>
12418 * config/mips/mips.h (struct mips_cpu_info): define enum mips_isa;
12419 use enum instead of int for 'isa' member.
12420 * config.gcc, config/mips/mips.c, config/mips/mips-cpus.def,
12421 config/mips/netbsd.h: replace hardcoded numbers with enum.
12423 2021-08-31 liuhongt <hongtao.liu@intel.com>
12425 * config/i386/sse.md (*<avx512>_ucmp<mode>3_1): Change from
12426 define_split to define_insn_and_split.
12427 (*avx2_eq<mode>3): Removed.
12428 (<avx512>_eq<mode>3<mask_scalar_merge_name>): Adjust pattern
12429 (<avx512>_eq<mode>3<mask_scalar_merge_name>_1): Rename to ..
12430 (*<avx512>_eq<mode>3<mask_scalar_merge_name>_1): .. this, and
12432 (*avx2_gt<mode>3): Removed.
12433 (<avx512>_gt<mode>3<mask_scalar_merge_name>): Change from
12434 define_insn to define_expand, and adjust pattern.
12435 (UNSPEC_MASKED_EQ, UNSPEC_MASKED_GT): Removed.
12437 2021-08-30 David Malcolm <dmalcolm@redhat.com>
12440 * Makefile.in (ANALYZER_OBJS): Add analyzer/call-info.o.
12442 2021-08-30 Jason Merrill <jason@redhat.com>
12444 * doc/invoke.texi: Document -Wmissing-requires.
12446 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
12448 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Remove
12449 TARGET_EXTRA_BUILTINS guard.
12451 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
12453 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Change
12454 initialization of V2DI_type_node and unsigned_V2DI_type_node.
12456 2021-08-30 Bill Schmidt <wschmidt@linux.ibm.com>
12458 * config/rs6000/darwin.h (SUBTARGET_INIT_BUILTINS): Use the new
12459 decl when new_builtins_are_live.
12460 * config/rs6000/rs6000-builtin-new.def (__builtin_cfstring): New
12463 2021-08-30 Pat Haugen <pthaugen@linux.ibm.com>
12465 * config/rs6000/rs6000-cpus.def (ISA_3_1_MASKS_SERVER): Add
12466 OPTION_MASK_P10_FUSION_2STORE.
12467 (POWERPC_MASKS): Likewise.
12468 * config/rs6000/rs6000.c (rs6000_option_override_internal): Enable
12469 store fusion for Power10.
12470 (is_fusable_store): New.
12471 (power10_sched_reorder): Likewise.
12472 (rs6000_sched_reorder): Do Power10 specific reordering.
12473 (rs6000_sched_reorder2): Likewise.
12474 * config/rs6000/rs6000.opt: Add new option.
12476 2021-08-30 Richard Biener <rguenther@suse.de>
12478 PR tree-optimization/102128
12479 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
12480 Move scanning for if-converted scalar code to the caller
12481 and instead delay clearing the visited flag for profitable
12483 (vect_slp_region): Cost all subgraphs before scheduling.
12484 For if-converted BB vectorization scan for scalar COND_EXPRs
12485 and do not vectorize if any found and the cost model is
12488 2021-08-30 Richard Biener <rguenther@suse.de>
12490 * common.opt (fexceptions): Mark
12491 EnabledBy(fnon-call-exceptions).
12492 * doc/invoke.texi (fnon-call-exceptions): Document this
12493 enables -fexceptions.
12495 2021-08-30 Sebastian Huber <sebastian.huber@embedded-brains.de>
12497 * tsystem.h (abort): Define abort() if inhibit_libc is defined and it
12498 is not already defined.
12500 2021-08-30 liuhongt <hongtao.liu@intel.com>
12502 * expmed.c (extract_bit_field_1): Make sure we're playing with
12503 integral modes before call extract_integral_bit_field.
12504 (extract_integral_bit_field): Add a parameter of type
12505 scalar_int_mode which corresponds to of tmode.
12506 And call extract_and_convert_fixed_bit_field instead of
12507 extract_fixed_bit_field and convert_extracted_bit_field.
12508 (extract_and_convert_fixed_bit_field): New function, it's a
12509 combination of extract_fixed_bit_field and
12510 convert_extracted_bit_field.
12512 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
12514 * config/darwin.c (darwin_libc_has_function): Do not run
12515 the checks for x86 or modern Darwin. Make sure that there
12516 is a value set for darwin_macosx_version_min before testing.
12518 2021-08-29 Iain Sandoe <iain@sandoe.co.uk>
12520 * config/i386/darwin.h (CLEAR_INSN_CACHE): New.
12522 2021-08-28 Jan Hubicka <hubicka@ucw.cz>
12524 * ipa-modref-tree.h (modref_access_node::merge): Break out
12525 logic combining offsets and logic merging ranges to ...
12526 (modref_access_node::combined_offsets): ... here
12527 (modref_access_node::update2): ... here
12528 (modref_access_node::closer_pair_p): New member function.
12529 (modref_access_node::forced_merge): New member function.
12530 (modre_ref_node::insert): Do merging when table is full.
12532 2021-08-28 YunQiang Su <yunqiang.su@cipunited.com>
12535 * config.gcc: MIPS: use N64 ABI by default if the triple end
12536 with -gnuabi64, which is used by Debian since 2013.
12538 2021-08-28 Alexandre Oliva <oliva@adacore.com>
12540 * ipa-modref.c (analyze_function): Skip debug stmts.
12541 * tree-inline.c (estimate_num_insn): Consider builtins even
12542 without a cgraph_node.
12544 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
12546 * config/h8300/bitfield.md (cstore<mode>4): Remove expander.
12547 * config/h8300/h8300.c (h8300_expand_branch): Remove function.
12548 * config/h8300/h8300-protos.h (h8300_expadn_branch): Remove prototype.
12549 * config/h8300/h8300.md (eqne): New code iterator.
12550 (geultu, geultu_to_c): Similarly.
12551 * config/h8300/testcompare.md (cstore<mode>4): Dummy expander.
12552 (store_c_<mode>, store_c_i_<mode>): New define_insn_and_splits
12553 (cmp<mode>_c): New pattern
12555 2021-08-27 Jeff Law <jlaw@localhost.localdomain>
12557 * tree-ssa-dom.c (reduce_vector_comparison_to_scalar_comparison): New
12559 (dom_opt_dom_walker::optimize_stmt): Use it.
12561 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
12563 * config/darwin.c (finalize_ctors): Add a section-start linker-
12565 (finalize_dtors): Likewise.
12566 * config/darwin.h (MIN_LD64_INIT_TERM_START_LABELS): New.
12568 2021-08-27 Bill Schmidt <wschmidt@linux.ibm.com>
12570 * config/rs6000/rs6000-call.c (rs6000-builtins.h): New #include.
12571 (rs6000_init_builtins): Call rs6000_init_generated_builtins. Skip the
12572 old initialization logic when new builtins are enabled.
12573 * config/rs6000/rs6000-gen-builtins.c (write_decls): Rename
12574 rs6000_autoinit_builtins to rs6000_init_generated_builtins.
12575 (write_init_file): Likewise.
12577 2021-08-27 Iain Sandoe <iain@sandoe.co.uk>
12579 * configure.ac (darwin2[[0-9]]* | darwin19*): Alter use of
12580 gcc_GAS_CHECK_FEATURE to remove an extraneous parameter.
12581 (amdgcn-* | gcn-*) Likewise.
12583 2021-08-27 Anthony Sharp <anthonysharp15@gmail.com>
12585 * symbol-summary.h: Added missing template keyword.
12587 2021-08-27 Richard Biener <rguenther@suse.de>
12589 PR tree-optimization/45178
12590 * tree-ssa-dce.c (find_obviously_necessary_stmts): For
12591 infinite loops without exit do not mark control dependent
12592 edges of the latch necessary.
12594 2021-08-27 konglin1 <lingling.kong@intel.com>
12597 * config/i386/sse.md: (<avx512>scattersi<mode>): Add mask operand to
12599 (<avx512>scattersi<mode>): Likewise.
12600 (*avx512f_scattersi<VI48F:mode>): Merge mask operand to set_dest.
12601 (*avx512f_scatterdi<VI48F:mode>): Likewise
12603 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
12605 * config/rs6000/rs6000.c (rs6000_builtin_md_vectorized_function): Add
12606 support for built-in functions MISC_BUILTIN_DIVWE, MISC_BUILTIN_DIVWEU,
12607 MISC_BUILTIN_DIVDE, MISC_BUILTIN_DIVDEU, P10_BUILTIN_CFUGED,
12608 P10_BUILTIN_CNTLZDM, P10_BUILTIN_CNTTZDM, P10_BUILTIN_PDEPD and
12609 P10_BUILTIN_PEXTD on Power10.
12611 2021-08-27 Kewen Lin <linkw@linux.ibm.com>
12613 * config/rs6000/rs6000-call.c (builtin_function_type): Add unsigned
12614 signedness for some Power10 bifs.
12616 2021-08-27 David Edelsohn <dje.gcc@gmail.com>
12619 * config/rs6000/rs6000.c (rs6000_adjust_field_align): Use
12620 computed alignment if the entire struct has attribute packed.
12622 2021-08-27 liuhongt <hongtao.liu@intel.com>
12626 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
12627 IX86_BUILTIN_SHUFPD512, IX86_BUILTIN_SHUFPS512,
12628 IX86_BUILTIN_SHUFPD256, IX86_BUILTIN_SHUFPS,
12629 IX86_BUILTIN_SHUFPS256.
12630 (ix86_masked_all_ones): New function.
12632 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
12634 * config/i386/i386.md (*btr<mode>_1): Call force_reg unconditionally.
12635 (conditional moves with memory inputs splitters): Ditto.
12636 * config/i386/sse.md (one_cmpl<mode>2): Simplify.
12638 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
12640 * ipa-modref-tree.h (modref_access_node::try_merge_with): Restart
12641 search after merging.
12643 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
12645 * config/rs6000/rs6000-overload.def: Add remaining overloads.
12647 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
12649 * config/rs6000/rs6000-builtin-new.def: Add cell stanza.
12651 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
12653 * config/rs6000/rs6000-builtin-new.def: Add ieee128-hw, dfp,
12654 crypto, and htm stanzas.
12656 2021-08-26 Bill Schmidt <wschmidt@linux.ibm.com>
12658 * config/rs6000/rs6000-builtin-new.def: Add mma stanza.
12660 2021-08-26 Martin Sebor <msebor@redhat.com>
12662 * tree-ssa-uninit.c (warn_uninit): Refactor and simplify.
12663 (warn_uninit_phi_uses): Remove argument from calls to warn_uninit.
12664 (warn_uninitialized_vars): Same. Reduce visibility of locals.
12665 (warn_uninitialized_phi): Same.
12667 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
12669 * tree-ssa-ccp.c (get_individual_bits): Helper function to
12670 extract the individual bits from a widest_int constant (mask).
12671 (gray_code_bit_flips): New read-only table for effiently
12672 enumerating permutations/combinations of bits.
12673 (bit_value_binop) [LROTATE_EXPR, RROTATE_EXPR]: Handle rotates
12674 by unknown counts that are guaranteed less than the target
12675 precision and four or fewer unknown bits by enumeration.
12676 [LSHIFT_EXPR, RSHIFT_EXPR]: Likewise, also handle shifts by
12677 enumeration under the same conditions. Handle remaining
12678 shifts as a mask based upon the minimum possible shift value.
12680 2021-08-26 Roger Sayle <roger@nextmovesoftware.com>
12681 Richard Biener <rguenther@suse.de>
12683 * match.pd (shift transformations): Remove a redundant
12684 !POINTER_TYPE_P check.
12686 2021-08-26 Uroš Bizjak <ubizjak@gmail.com>
12689 * config/i386/i386.md (cmove reg-reg move elimination peephole2s):
12690 Set all_regs to true in the call to replace_rtx.
12692 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
12694 * ipa-modref-tree.c (test_insert_search_collapse): Update test.
12695 * ipa-modref-tree.h (modref_base_node::insert): Be smarter when
12696 hiting --param modref-max-refs limit.
12697 (modref_tree:insert_base): Be smarter when hitting
12698 --param modref-max-bases limit. Add new parameter REF.
12699 (modref_tree:insert): Update.
12700 (modref_tree:merge): Update.
12701 * ipa-modref.c (read_modref_records): Update.
12703 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
12705 * params.opt: (modref-max-adjustments): Add full stop.
12707 2021-08-26 Jan Hubicka <hubicka@ucw.cz>
12709 * ipa-modref-tree.h (modref_ref_node::verify): New member
12711 (modref_ref_node::insert): Use it.
12712 (modref_ref_node::try_mere_with): Fix off by one error.
12714 2021-08-26 Martin Liska <mliska@suse.cz>
12715 Stefan Kneifel <stefan.kneifel@bluewin.ch>
12717 * cgraph.h (create_version_clone_with_body): Add new parameter.
12718 * cgraphclones.c: Likewise.
12719 * multiple_target.c (create_dispatcher_calls): Do not use
12721 (create_target_clone): Likewise here.
12723 2021-08-26 Jonathan Yong <10walls@gmail.com>
12725 * doc/extend.texi: Add note about reserved priorities
12726 to the constructor attribute.
12728 2021-08-25 Martin Sebor <msebor@redhat.com>
12730 * gimple-range-cache.cc (ssa_global_cache::dump): Avoid printing
12731 range table header alone.
12732 * gimple-range.cc (gimple_ranger::export_global_ranges): Same.
12734 2021-08-25 Jan Hubicka <hubicka@ucw.cz>
12736 * doc/invoke.texi: Document --param modref-max-adjustments.
12737 * ipa-modref-tree.c (test_insert_search_collapse): Update.
12738 (test_merge): Update.
12739 * ipa-modref-tree.h (struct modref_access_node): Add adjustments;
12740 (modref_access_node::operator==): Fix handling of access ranges.
12741 (modref_access_node::contains): Constify parameter; handle also
12742 mismatched parm offsets.
12743 (modref_access_node::update): New function.
12744 (modref_access_node::merge): New function.
12745 (unspecified_modref_access_node): Update constructor.
12746 (modref_ref_node::insert_access): Add record_adjustments parameter;
12748 (modref_ref_node::try_merge_with): New private function.
12749 (modref_tree::insert): New record_adjustments parameter.
12750 (modref_tree::merge): New record_adjustments parameter.
12751 (modref_tree::copy_from): Update.
12752 * ipa-modref.c (dump_access): Dump adjustments field.
12753 (get_access): Update constructor.
12754 (record_access): Update call of insert.
12755 (record_access_lto): Update call of insert.
12756 (merge_call_side_effects): Add record_adjustments parameter.
12757 (get_access_for_fnspec): Update.
12758 (process_fnspec): Update.
12759 (analyze_call): Update.
12760 (analyze_function): Update.
12761 (read_modref_records): Update.
12762 (ipa_merge_modref_summary_after_inlining): Update.
12763 (propagate_unknown_call): Update.
12764 (modref_propagate_in_scc): Update.
12765 * params.opt (param-max-modref-adjustments=): New.
12767 2021-08-25 Michael Meissner <meissner@linux.ibm.com>
12769 * config/rs6000/vsx.md (UNSPEC_XXSPLTIDP): Rename from
12771 (xxspltiw_v4si): Use vecperm type attribute.
12772 (xxspltiw_v4si_inst): Use vecperm type attribute.
12773 (xxspltiw_v4sf_inst): Likewise.
12774 (xxspltidp_v2df): Use vecperm type attribute. Use
12775 UNSPEC_XXSPLTIDP instead of UNSPEC_XXSPLTID.
12776 (xxspltidp_v2df_inst): Likewise.
12777 (xxsplti32dx_v4si): Use vecperm type attribute.
12778 (xxsplti32dx_v4si_inst): Likewise.
12779 (xxsplti32dx_v4sf_inst): Likewise.
12780 (xxblend_<mode>): Likewise.
12781 (xxpermx): Likewise.
12782 (xxpermx_inst): Likewise.
12783 (xxeval): Likewise.
12785 2021-08-25 Lewis Hyatt <lhyatt@gmail.com>
12788 * coretypes.h (typedef diagnostic_input_charset_callback): Declare.
12789 * diagnostic.c (diagnostic_initialize_input_context): New function.
12790 * diagnostic.h (diagnostic_initialize_input_context): Declare.
12791 * input.c (default_charset_callback): New function.
12792 (file_cache::initialize_input_context): New function.
12793 (file_cache_slot::create): Added ability to convert the input
12794 according to the input context.
12795 (file_cache::file_cache): Initialize the new input context.
12796 (class file_cache_slot): Added new m_alloc_offset member.
12797 (file_cache_slot::file_cache_slot): Initialize the new member.
12798 (file_cache_slot::~file_cache_slot): Handle potentially offset buffer.
12799 (file_cache_slot::maybe_grow): Likewise.
12800 (file_cache_slot::needs_read_p): Handle NULL fp, which is now possible.
12801 (file_cache_slot::get_next_line): Likewise.
12802 * input.h (class file_cache): Added input context member.
12804 2021-08-25 Richard Biener <rguenther@suse.de>
12806 PR tree-optimization/102046
12807 * tree-vect-slp.c (vect_build_slp_tree_2): Conservatively
12808 update ->any_pattern when swapping operands.
12810 2021-08-25 Hongyu Wang <hongyu.wang@intel.com>
12813 * config/i386/i386.c (ix86_live_on_entry): Adjust comment.
12814 (ix86_decompose_address): Remove retval check for ASHIFT,
12815 allow non-canonical zero extend if AND mask covers ASHIFT
12817 (ix86_legitimate_address_p): Adjust condition for decompose.
12818 (ix86_rtx_costs): Adjust cost for lea with non-canonical
12820 Co-Authored by: Uros Bizjak <ubizjak@gmail.com>
12822 2021-08-25 Jiufu Guo <guojiufu@linux.ibm.com>
12824 PR tree-optimization/101145
12825 * tree-ssa-loop-niter.c (number_of_iterations_until_wrap):
12827 (number_of_iterations_lt): Invoke above function.
12828 (adjust_cond_for_loop_until_wrap):
12829 Merge to number_of_iterations_until_wrap.
12830 (number_of_iterations_cond): Update invokes for
12831 adjust_cond_for_loop_until_wrap and number_of_iterations_lt.
12833 2021-08-25 konglin1 <lingling.kong@intel.com>
12836 * config/i386/avx512dqintrin.h (_mm512_fpclass_ps_mask): Fix
12837 macro define in O0.
12838 (_mm512_mask_fpclass_ps_mask): Ditto.
12840 2021-08-25 Kewen Lin <linkw@linux.ibm.com>
12842 * config/rs6000/altivec.md (vec_unpacku_hi_v16qi): Remove.
12843 (vec_unpacku_hi_v8hi): Likewise.
12844 (vec_unpacku_lo_v16qi): Likewise.
12845 (vec_unpacku_lo_v8hi): Likewise.
12846 (vec_unpacku_hi_<VP_small_lc>): New define_expand.
12847 (vec_unpacku_lo_<VP_small_lc>): Likewise.
12849 2021-08-24 David Edelsohn <dje.gcc@gmail.com>
12851 * config/rs6000/aix.h (SYSTEM_IMPLICIT_EXTERN_C): Delete.
12852 * config/rs6000/aix71.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
12853 * config/rs6000/aix72.h (SYSTEM_IMPLICIT_EXTERN_C): Define.
12854 * config/rs6000/aix73.h (TARGET_AIX_VERSION): Increase to 73.
12856 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
12858 PR middle-end/102031
12859 * simplify-rtx.c (simplify_truncation): When comparing precisions
12860 use "subreg_prec" variable, not "subreg_mode".
12862 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
12864 * config/rs6000/rs6000-builtin-new.def: Add power10 and power10-64
12867 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
12869 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Initialize
12870 various pointer type nodes.
12871 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add enum
12872 values for various pointer types.
12873 (ptr_V16QI_type_node): New macro.
12874 (ptr_V1TI_type_node): New macro.
12875 (ptr_V2DI_type_node): New macro.
12876 (ptr_V2DF_type_node): New macro.
12877 (ptr_V4SI_type_node): New macro.
12878 (ptr_V4SF_type_node): New macro.
12879 (ptr_V8HI_type_node): New macro.
12880 (ptr_unsigned_V16QI_type_node): New macro.
12881 (ptr_unsigned_V1TI_type_node): New macro.
12882 (ptr_unsigned_V8HI_type_node): New macro.
12883 (ptr_unsigned_V4SI_type_node): New macro.
12884 (ptr_unsigned_V2DI_type_node): New macro.
12885 (ptr_bool_V16QI_type_node): New macro.
12886 (ptr_bool_V8HI_type_node): New macro.
12887 (ptr_bool_V4SI_type_node): New macro.
12888 (ptr_bool_V2DI_type_node): New macro.
12889 (ptr_bool_V1TI_type_node): New macro.
12890 (ptr_pixel_type_node): New macro.
12891 (ptr_intQI_type_node): New macro.
12892 (ptr_uintQI_type_node): New macro.
12893 (ptr_intHI_type_node): New macro.
12894 (ptr_uintHI_type_node): New macro.
12895 (ptr_intSI_type_node): New macro.
12896 (ptr_uintSI_type_node): New macro.
12897 (ptr_intDI_type_node): New macro.
12898 (ptr_uintDI_type_node): New macro.
12899 (ptr_intTI_type_node): New macro.
12900 (ptr_uintTI_type_node): New macro.
12901 (ptr_long_integer_type_node): New macro.
12902 (ptr_long_unsigned_type_node): New macro.
12903 (ptr_float_type_node): New macro.
12904 (ptr_double_type_node): New macro.
12905 (ptr_long_double_type_node): New macro.
12906 (ptr_dfloat64_type_node): New macro.
12907 (ptr_dfloat128_type_node): New macro.
12908 (ptr_ieee128_type_node): New macro.
12909 (ptr_ibm128_type_node): New macro.
12910 (ptr_vector_pair_type_node): New macro.
12911 (ptr_vector_quad_type_node): New macro.
12912 (ptr_long_long_integer_type_node): New macro.
12913 (ptr_long_long_unsigned_type_node): New macro.
12915 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
12917 * config/rs6000/rs6000-builtin-new.def: Add power9-vector, power9,
12918 and power9-64 stanzas.
12920 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
12921 Tom de Vries <tdevries@suse.de>
12923 * config.gcc (nvptx-*-*): Define {c,c++}_target_objs.
12924 * config/nvptx/nvptx-protos.h (nvptx_cpu_cpp_builtins): Prototype.
12925 * config/nvptx/nvptx.h (TARGET_CPU_CPP_BUILTINS): Implement with
12926 a call to the new nvptx_cpu_cpp_builtins function in nvptx-c.c.
12927 * config/nvptx/t-nvptx (nvptx-c.o): New rule.
12928 * config/nvptx/nvptx-c.c: New source file.
12929 (nvptx_cpu_cpp_builtins): Move implementation here.
12931 2021-08-24 Martin Sebor <msebor@redhat.com>
12933 PR middle-end/101600
12934 PR middle-end/101977
12935 * gimple-ssa-warn-access.cc (maybe_warn_for_bound): Tighten up
12936 the phrasing of a warning.
12937 (check_access): Use the remaining size after subtracting any offset
12938 rather than the whole object size.
12939 * pointer-query.cc (access_ref::get_ref): Clear BASE0 flag if it's
12940 clear for any nonnull PHI argument.
12941 (compute_objsize): Clear argument.
12943 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
12945 * config/rs6000/rs6000-builtin-new.def: Add power8-vector stanza.
12947 2021-08-24 Bill Schmidt <wschmidt@linux.ibm.com>
12949 * config/rs6000/rs6000-builtin-new.def: Add power7 and power7-64
12952 2021-08-24 Andrew MacLeod <amacleod@redhat.com>
12954 * value-relation.cc (rr_transitive_table): New.
12955 (relation_transitive): New.
12956 (value_relation::swap): Remove.
12957 (value_relation::apply_transitive): New.
12958 (relation_oracle::relation_oracle): Allocate a new tmp bitmap.
12959 (relation_oracle::register_relation): Call register_transitives.
12960 (relation_oracle::register_transitives): New.
12961 * value-relation.h (relation_oracle): Add new temporary bitmap and
12964 2021-08-24 H.J. Lu <hjl.tools@gmail.com>
12967 * config/i386/i386-expand.c (ix86_expand_vector_move): Broadcast
12968 from integer to a pseudo vector register.
12970 2021-08-24 Richard Biener <rguenther@suse.de>
12972 PR tree-optimization/100089
12973 * tree-vectorizer.h (vect_slp_bb): Rename to ...
12974 (vect_slp_if_converted_bb): ... this and get the original
12975 loop as new argument.
12976 * tree-vectorizer.c (try_vectorize_loop_1): Revert previous fix,
12977 pass original loop to vect_slp_if_converted_bb.
12978 * tree-vect-slp.c (vect_bb_vectorization_profitable_p):
12979 If orig_loop was passed scan the not vectorized stmts
12980 for COND_EXPRs and force not profitable if found.
12981 (vect_slp_region): Pass down all SLP instances to costing
12982 if orig_loop was specified.
12983 (vect_slp_bbs): Pass through orig_loop.
12984 (vect_slp_bb): Rename to ...
12985 (vect_slp_if_converted_bb): ... this and get the original
12986 loop as new argument.
12987 (vect_slp_function): Adjust.
12989 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
12992 * config/arm/arm.md (attribute arch): Add fix_vlldm.
12993 (arch_enabled): Use it.
12994 * config/arm/vfp.md (lazy_store_multiple_insn): Add alternative to
12995 use when erratum mitigation is needed.
12997 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
13000 * config/arm/arm.opt (mfix-cmse-cve-2021-35465): New option.
13001 * doc/invoke.texi (Arm Options): Document it.
13002 * config/arm/arm-cpus.in (quirk_vlldm): New feature bit.
13003 (ALL_QUIRKS): Add quirk_vlldm.
13004 (cortex-m33): Add quirk_vlldm.
13005 (cortex-m35p, cortex-m55): Likewise.
13006 * config/arm/arm.c (arm_option_override): Enable fix_vlldm if
13007 targetting an affected CPU and not explicitly controlled on
13010 2021-08-24 Richard Earnshaw <rearnsha@arm.com>
13012 * config/arm/vfp.md (lazy_store_multiple_insn): Rewrite as valid RTL.
13013 (lazy_load_multiple_insn): Likewise.
13015 2021-08-24 liuhongt <hongtao.liu@intel.com>
13018 * config/i386/sse.md (<avx512>_vternlog<mode><sd_maskz_name>):
13019 Enable avx512 embedded broadcast.
13020 (*<avx512>_vternlog<mode>_all): Ditto.
13021 (<avx512>_vternlog<mode>_mask): Ditto.
13023 2021-08-24 liuhongt <hongtao.liu@intel.com>
13026 * config/i386/i386.c (ix86_rtx_costs): Define cost for
13028 * config/i386/i386.h (STRIP_UNARY): New macro.
13029 * config/i386/predicates.md (reg_or_notreg_operand): New
13031 * config/i386/sse.md (*<avx512>_vternlog<mode>_all): New define_insn.
13032 (*<avx512>_vternlog<mode>_1): New pre_reload
13033 define_insn_and_split.
13034 (*<avx512>_vternlog<mode>_2): Ditto.
13035 (*<avx512>_vternlog<mode>_3): Ditto.
13036 (any_logic1,any_logic2): New code iterator.
13037 (logic_op): New code attribute.
13038 (ternlogsuffix): Extend to VNxDF and VNxSF.
13040 2021-08-24 Richard Biener <rguenther@suse.de>
13042 * doc/invoke.texi (vect-inner-loop-cost-factor): Adjust.
13043 * params.opt (--param vect-inner-loop-cost-factor): Adjust
13045 * tree-vect-loop.c (vect_analyze_loop_form): Initialize
13046 inner_loop_cost_factor to the minimum of the estimated number
13047 of iterations of the inner loop and vect-inner-loop-cost-factor.
13049 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
13050 Richard Biener <rguenther@suse.de>
13052 * config/i386/i386-features.c (compute_convert_gain): Provide
13053 more accurate values for CONST_INT, when optimizing for size.
13054 * config/i386/i386.c (COSTS_N_BYTES): Move definition from here...
13055 * config/i386/i386.h (COSTS_N_BYTES): to here.
13057 2021-08-24 Roger Sayle <roger@nextmovesoftware.com>
13058 Jakub Jelinek <jakub@redhat.com>
13060 PR middle-end/102029
13061 * match.pd (shift transformations): Add an additional check for
13062 !POINTER_TYPE_P in the recently added left shift transformation.
13064 2021-08-24 liuhongt <hongtao.liu@intel.com>
13066 PR tree-optimization/100089
13067 * tree-vectorizer.c (try_vectorize_loop_1): Disable slp in
13068 loop vectorizer when cost model is very-cheap.
13070 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
13072 * config/rs6000/rs6000-gen-builtins.c (parse_bif_entry): Don't call
13073 asprintf, which is not available on AIX.
13075 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
13077 * config.gcc (target_gtfiles): Add ./rs6000-builtins.h.
13078 * config/rs6000/t-rs6000 (EXTRA_GTYPE_DEPS): Set.
13080 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
13082 * config.gcc (powerpc*-*-*): Add rs6000-builtins.o to extra_objs.
13083 * config/rs6000/rs6000-gen-builtins.c (main): Close init_file
13085 * config/rs6000/t-rs6000 (rs6000-gen-builtins.o): New target.
13086 (rbtree.o): Likewise.
13087 (rs6000-gen-builtins): Likewise.
13088 (rs6000-builtins.c): Likewise.
13089 (rs6000-builtins.h): Likewise.
13090 (rs6000.o): Add dependency.
13091 (EXTRA_HEADERS): Add rs6000-vecdefines.h.
13092 (rs6000-vecdefines.h): New target.
13093 (rs6000-builtins.o): Likewise.
13094 (rs6000-call.o): Add rs6000-builtins.h as a dependency.
13095 (rs6000-c.o): Likewise.
13097 2021-08-23 Bill Schmidt <wschmidt@linux.ibm.com>
13100 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace):
13101 Diagnose buffer overrun.
13102 (safe_inc_pos): Fix overrun detection.
13103 (match_identifier): Diagnose buffer overrun.
13104 (match_integer): Likewise.
13105 (match_to_right_bracket): Likewise.
13107 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
13109 * ipa-modref-tree.h (modref_access_node::range_info_useful_p):
13110 Improve range compare.
13111 (modref_access_node::contains): New member function.
13112 (modref_access_node::search): Remove.
13113 (modref_access_node::insert): Be smarter about subaccesses.
13115 2021-08-23 Thomas Schwinge <thomas@codesourcery.com>
13117 * config/i386/i386-options.c (ix86_omp_device_kind_arch_isa)
13118 <omp_device_arch> [ACCEL_COMPILER]: Match "intel_mic".
13119 * config/i386/t-omp-device (omp-device-properties-i386) <arch>:
13122 2021-08-23 Jeff Law <jlaw@localhost.localdomain>
13124 * config/h8300/h8300-protos.h (h8300_expand_epilogue): Add new
13126 * config/h8300/jumpcall.md (call, call_value): Restrict to
13127 !SIBLING_CALL_P cases.
13128 (subcall, sibcall_value): New patterns & expanders.
13129 * config/h8300/proepi.md (epilogue): Pass new argument to
13130 h8300_expand_epilogue.
13131 (sibcall_epilogue): New expander.
13132 * config/h8300/h8300.c (h8300_expand_epilogue): Handle sibcall
13134 (h8300_ok_for_sibcall_p): New function.
13135 (TARGET_FUNCTION_OK_FOR_SIBCALL): define.
13137 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
13139 * simplify-rtx.c (simplify_unary_operation_1): [TRUNCATE]:
13140 Handle case where the operand is already the desired mode.
13142 2021-08-23 Richard Biener <rguenther@suse.de>
13145 * tree-ssa-structalias.c (ipa_pta_execute): Check in_other_partition
13146 in addition to has_gimple_body.
13148 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
13150 PR middle-end/101949
13151 * ipa-modref.c (analyze_ssa_name_flags): Fix merging of
13154 2021-08-23 Martin Liska <mliska@suse.cz>
13156 * doc/invoke.texi: Put the option out of -mxl-mode-app-model
13159 2021-08-23 Richard Biener <rguenther@suse.de>
13161 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
13162 Properly scale the inner loop cost only once.
13164 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
13166 * tree-ssa-ccp.c (bit_value_binop) [TRUNC_MOD_EXPR, TRUNC_DIV_EXPR]:
13167 Provide bounds for unsigned (and signed with non-negative operands)
13168 division and modulus.
13170 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
13172 * simplify-rtx.c (simplify_truncation): Generalize simplification
13173 of (truncate:A (subreg:B X)).
13174 (simplify_unary_operation_1) [FLOAT_TRUNCATE, FLOAT_EXTEND,
13175 SIGN_EXTEND, ZERO_EXTEND]: Handle cases where the operand
13176 already has the desired machine mode.
13177 (test_scalar_int_ops): Add tests that useless extensions and
13178 truncations are optimized away.
13179 (test_scalar_int_ext_ops): New self-test function to confirm
13180 that truncations of extensions are correctly simplified.
13181 (test_scalar_int_ext_ops2): New self-test function to check
13182 truncations of truncations, extensions of extensions, and
13183 truncations of extensions.
13184 (test_scalar_ops): Call the above two functions with a
13185 representative sampling of integer machine modes.
13187 2021-08-23 Roger Sayle <roger@nextmovesoftware.com>
13189 * match.pd (shift transformations): Change the sign of an
13190 LSHIFT_EXPR if it reduces the number of explicit conversions.
13192 2021-08-23 Jakub Jelinek <jakub@redhat.com>
13194 PR tree-optimization/86723
13195 * gimple-ssa-store-merging.c (find_bswap_or_nop_finalize): Add
13196 cast64_to_32 argument, set *cast64_to_32 to false, unless n is
13197 non-memory permutation of 64-bit src which only has bytes of
13198 0 or [5..8] and n->range is 4.
13199 (find_bswap_or_nop): Add cast64_to_32 and mask arguments, adjust
13200 find_bswap_or_nop_finalize caller, support bswap with some bytes
13201 zeroed, as long as at least two bytes are not zeroed.
13202 (bswap_replace): Add mask argument and handle masking of bswap
13204 (maybe_optimize_vector_constructor): Adjust find_bswap_or_nop
13205 caller, punt if cast64_to_32 or mask is not all ones.
13206 (pass_optimize_bswap::execute): Adjust find_bswap_or_nop_finalize
13207 caller, for now punt if cast64_to_32.
13209 2021-08-23 Richard Biener <rguenther@suse.de>
13211 PR tree-optimization/79334
13212 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Record
13213 a type also for COMPONENT_REFs.
13214 (vn_reference_may_trap): Check ARRAY_REF with constant index
13215 against the array domain.
13217 2021-08-23 liuhongt <hongtao.liu@intel.com>
13220 * config/i386/sse.md (*avx512f_pshufb_truncv8hiv8qi_1): Add
13221 TARGET_AVX512BW to condition.
13223 2021-08-23 Jakub Jelinek <jakub@redhat.com>
13226 * dwarf2out.c (gen_variable_die): Add DW_AT_location for global
13227 register variables already during early_dwarf if possible.
13229 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
13231 * config/arm/arm_mve.h: Fix __arm_vctp16q return type.
13233 2021-08-23 Christophe Lyon <christophe.lyon@foss.st.com>
13236 * config/arm/arm.opt: Fix typo.
13237 * config/arm/t-rmprofile: Fix typo.
13239 2021-08-23 Jakub Jelinek <jakub@redhat.com>
13241 * tree.h (OMP_CLAUSE_GRAINSIZE_STRICT): Define.
13242 (OMP_CLAUSE_NUM_TASKS_STRICT): Define.
13243 * tree-pretty-print.c (dump_omp_clause) <case OMP_CLAUSE_GRAINSIZE,
13244 case OMP_CLAUSE_NUM_TASKS>: Print strict: modifier.
13245 * omp-expand.c (expand_task_call): Use GOMP_TASK_FLAG_STRICT in iflags
13246 if either grainsize or num_tasks clause has the strict modifier.
13248 2021-08-23 Martin Liska <mliska@suse.cz>
13250 * dbgcnt.def (DEBUG_COUNTER): New counter.
13251 * gimple.c (gimple_call_arg_flags): Use it in IPA PTA.
13253 2021-08-23 Jan Hubicka <hubicka@ucw.cz>
13255 * ipa-modref.c (analyze_ssa_name_flags): Improve handling of return slot.
13257 2021-08-23 Xi Ruoyao <xry111@mengyan1223.wang>
13260 * config/mips/mips-protos.h (mips_msa_output_shift_immediate):
13262 * config/mips/mips.c (mips_msa_output_shift_immediate): New
13264 * config/mips/mips-msa.md (vashl<mode>3, vashr<mode>3,
13265 vlshr<mode>3): Call it.
13267 2021-08-22 Jan Hubicka <hubicka@ucw.cz>
13268 Martin Liska <mliska@suse.cz>
13270 PR middle-end/101949
13271 * ipa-modref.c (analyze_ssa_name_flags): Indirect call implies
13274 2021-08-21 Dragan Mladjenovic <OT_Dragan.Mladjenovic@mediatek.com>
13276 * config/mips/mips.c (mips_function_rodata_section,
13277 TARGET_ASM_FUNCTION_RODATA_SECTION): Removed.
13279 2021-08-21 John David Anglin <danglin@gcc.gnu.org>
13281 * config/pa/pa.c (pa_asm_output_aligned_common): Remove warning.
13283 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
13285 * configure.ac (thread-local storage support): Remove tls_first_major
13286 and tls_first_minor. Use "$conftest_s" to check support.
13287 * configure: Regenerate.
13289 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
13291 * configure.ac: Fixup formatting.
13293 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
13295 * acinclude.m4 (gcc_GAS_CHECK_FEATURE): Remove third argument and ...
13296 * configure.ac: ... update all callers.
13298 2021-08-20 Serge Belyshev <belyshev@depni.sinp.msu.ru>
13301 * acinclude.m4 (_gcc_COMPUTE_GAS_VERSION, _gcc_GAS_VERSION_GTE_IFELSE)
13302 (gcc_GAS_VERSION_GTE_IFELSE): Remove.
13303 (gcc_GAS_CHECK_FEATURE): Do not handle in-tree case specially.
13304 * configure.ac: Remove gcc_cv_gas_major_version, gcc_cv_gas_minor_version.
13305 Remove remaining checks for in-tree assembler.
13306 * configure: Regenerate.
13308 2021-08-20 Jeff Law <jlaw@localhost.localdomain>
13310 * config/h8300/h8300.c (shift_alg_hi): Improve arithmetic shift right
13311 by 15 bits for H8/300H and H8/S. Improve logical shifts by 12
13313 (shift_alg_si): Improve arithmetic right shift by 28-30 bits for
13314 H8/300H. Improve arithmetic shift right by 15 bits for H8/S.
13315 Improve logical shifts by 27 bits for H8/S.
13316 (get_shift_alg): Corresponding changes.
13317 (h8300_option_override): Revert to loops for -Os when profitable.
13319 2021-08-20 Richard Biener <rguenther@suse.de>
13321 * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
13323 (vect_analyze_data_ref_accesses): Likewise. Assign the BB
13324 index as group_id when dataref_groups were not computed.
13325 * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
13326 we advace to the next BB.
13328 2021-08-20 Jakub Jelinek <jakub@redhat.com>
13330 * omp-builtins.def (BUILT_IN_GOMP_WARNING, BUILT_IN_GOMP_ERROR): New
13333 2021-08-20 Martin Liska <mliska@suse.cz>
13335 PR gcov-profile/89961
13336 * gcov.c (make_gcov_file_name): Rewrite using std::string.
13337 (mangle_name): Simplify, do not used the second argument.
13338 (strip_extention): New function.
13339 (get_md5sum): Likewise.
13340 (get_gcov_intermediate_filename): Handle properly -p and -x
13342 (output_gcov_file): Use string type.
13343 (generate_results): Likewise.
13344 (md5sum_to_hex): Remove.
13346 2021-08-20 Michael Meissner <meissner@linux.ibm.com>
13348 * config/rs6000/altivec.md (UNSPEC_XXEVAL): Move to vsx.md.
13349 (UNSPEC_XXSPLTIW): Move to vsx.md.
13350 (UNSPEC_XXSPLTID): Move to vsx.md.
13351 (UNSPEC_XXSPLTI32DX): Move to vsx.md.
13352 (UNSPEC_XXBLEND): Move to vsx.md.
13353 (UNSPEC_XXPERMX): Move to vsx.md.
13354 (VM3): Move to vsx.md.
13355 (VM3_char): Move to vsx.md.
13356 (xxspltiw_v4si): Move to vsx.md.
13357 (xxspltiw_v4sf): Move to vsx.md.
13358 (xxspltiw_v4sf_inst): Move to vsx.md.
13359 (xxspltidp_v2df): Move to vsx.md.
13360 (xxspltidp_v2df_inst): Move to vsx.md.
13361 (xxsplti32dx_v4si_inst): Move to vsx.md.
13362 (xxsplti32dx_v4sf): Move to vsx.md.
13363 (xxsplti32dx_v4sf_inst): Move to vsx.md.
13364 (xxblend_<mode>): Move to vsx.md.
13365 (xxpermx): Move to vsx.md.
13366 (xxpermx_inst): Move to vsx.md.
13367 * config/rs6000/vsx.md (UNSPEC_XXEVAL): Move from altivec.md.
13368 (UNSPEC_XXSPLTIW): Move from altivec.md.
13369 (UNSPEC_XXSPLTID): Move from altivec.md.
13370 (UNSPEC_XXSPLTI32DX): Move from altivec.md.
13371 (UNSPEC_XXBLEND): Move from altivec.md.
13372 (UNSPEC_XXPERMX): Move from altivec.md.
13373 (VM3): Move from altivec.md.
13374 (VM3_char): Move from altivec.md.
13375 (xxspltiw_v4si): Move from altivec.md.
13376 (xxspltiw_v4sf): Move from altivec.md.
13377 (xxspltiw_v4sf_inst): Move from altivec.md.
13378 (xxspltidp_v2df): Move from altivec.md.
13379 (xxspltidp_v2df_inst): Move from altivec.md.
13380 (xxsplti32dx_v4si_inst): Move from altivec.md.
13381 (xxsplti32dx_v4sf): Move from altivec.md.
13382 (xxsplti32dx_v4sf_inst): Move from altivec.md.
13383 (xxblend_<mode>): Move from altivec.md.
13384 (xxpermx): Move from altivec.md.
13385 (xxpermx_inst): Move from altivec.md.
13387 2021-08-19 Roger Sayle <roger@nextmovesoftware.com>
13389 * tree-vect-generic.c (expand_vector_operations_1): Use either
13390 gimplify_build1 or gimplify_build2 instead of gimple_build_assign
13391 when constructing scalar splat expressions.
13393 2021-08-19 Peter Bergner <bergner@linux.ibm.com>
13396 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Cast
13397 pointer to __vector_pair *.
13399 2021-08-19 Martin Sebor <msebor@redhat.com>
13401 * gimple-range.cc: Add comments.
13402 * gimple-range.h: Same.
13404 2021-08-19 Martin Sebor <msebor@redhat.com>
13406 PR middle-end/101984
13407 * gimple-ssa-warn-access.cc (pass_waccess::execute): Also call
13410 2021-08-19 Jeff Law <jlaw@localhost.localdomain>
13412 * config.gcc (h8300-*-elf*): Do not include dbxelf.h.
13413 (h8300-*-linux*, v850-*-rtems*, v850*-elf*): Likewise.
13414 * config/v850/v850.h (DEFAULT_GDB_EXTENSIONS): Remove.
13416 2021-08-19 Jakub Jelinek <jakub@redhat.com>
13418 PR middle-end/101950
13419 * optabs.c (expand_clrsb_using_clz): New function.
13420 (expand_unop): Use it as another clrsb expansion fallback.
13422 2021-08-19 liuhongt <hongtao.liu@intel.com>
13425 2021-07-28 liuhongt <hongtao.liu@intel.com>
13428 * config/i386/i386.h (processor_costs): Add new member
13430 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
13431 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
13432 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
13433 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
13434 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
13435 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
13436 generic_cost, core_cost): Initialize integer_to_sse same value
13438 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
13439 * config/i386/i386.c (ix86_builtin_vectorization_cost):
13440 Use integer_to_sse instead of sse_op to calculate the cost of
13443 2021-08-18 Iain Sandoe <iain@sandoe.co.uk>
13445 * config.gcc: Include rpath.opt for Darwin.
13446 * config/darwin.h (DRIVER_SELF_SPECS): Handle -rpath.
13448 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
13450 PR bootstrap/101959
13451 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor_expand):
13454 2021-08-18 Jonathan Wright <jonathan.wright@arm.com>
13456 * config/aarch64/arm_neon.h (vld3_lane_f64): Use float RTL
13457 pattern and type cast.
13458 (vld4_lane_f32): Use float RTL pattern.
13459 (vld4q_lane_f64): Use float type cast.
13461 2021-08-18 Jan Hubicka <hubicka@ucw.cz>
13463 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Check also
13466 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
13468 * hash-map-tests.c (test_map_of_type_with_ctor_and_dtor): Extend.
13469 (test_map_of_type_with_ctor_and_dtor_expand): Add function.
13470 (hash_map_tests_c_tests): Call it.
13472 2021-08-18 Thomas Schwinge <thomas@codesourcery.com>
13474 * ggc.h (enum ggc_collect): New.
13475 (ggc_collect): Use it.
13476 * ggc-page.c: Adjust.
13477 * ggc-common.c: Likewise.
13478 * ggc-tests.c: Likewise.
13479 * read-rtl-function.c: Likewise.
13480 * selftest-run-tests.c: Likewise.
13481 * doc/gty.texi (Invoking the garbage collector): Likewise.
13483 2021-08-18 liuhongt <hongtao.liu@intel.com>
13486 * config/i386/i386.h (TARGET_V2DF_REDUCTION_PREFER_HADDPD):
13488 * config/i386/sse.md (*sse3_haddv2df3_low): Add
13489 TARGET_V2DF_REDUCTION_PREFER_HADDPD.
13490 (*sse3_hsubv2df3_low): Ditto.
13491 * config/i386/x86-tune.def
13492 (X86_TUNE_V2DF_REDUCTION_PREFER_HADDPD): New tune.
13494 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
13496 * gimple-range-gori.cc (gori_compute::gori_compute): Enable tracing.
13497 (gori_compute::compute_operand_range): Add tracing.
13498 (gori_compute::logical_combine): Ditto.
13499 (gori_compute::compute_logical_operands): Ditto.
13500 (gori_compute::compute_operand1_range): Ditto.
13501 (gori_compute::compute_operand2_range): Ditto.
13502 (gori_compute::outgoing_edge_range_p): Ditto.
13503 * gimple-range-gori.h (class gori_compute): Add range_tracer.
13505 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
13507 * flag-types.h (enum evrp_mode): Adjust evrp-mode values.
13508 * gimple-range-cache.cc (DEBUG_RANGE_CACHE): Relocate from.
13509 * gimple-range-trace.h (DEBUG_RANGE_CACHE): Here.
13510 * params.opt (--param=evrp-mode): Adjust options.
13512 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
13514 * Makefile.in (OBJS): Add gimple-range-trace.o.
13515 * gimple-range-cache.h (enable_new_values): Remove unused prototype.
13516 * gimple-range-fold.cc: Adjust headers.
13517 * gimple-range-trace.cc: New.
13518 * gimple-range-trace.h: New.
13519 * gimple-range.cc (gimple_ranger::gimple_ranger): Enable tracer.
13520 (gimple_ranger::range_of_expr): Add tracing.
13521 (gimple_ranger::range_on_entry): Ditto.
13522 (gimple_ranger::range_on_exit): Ditto.
13523 (gimple_ranger::range_on_edge): Ditto.
13524 (gimple_ranger::fold_range_internal): Ditto.
13525 (gimple_ranger::dump_bb): Do not calculate edge range twice.
13526 (trace_ranger::*): Remove.
13527 (enable_ranger): Never create a trace_ranger.
13528 (debug_seed_ranger): Move to gimple-range-trace.cc.
13529 (dump_ranger): Ditto.
13530 (debug_ranger): Ditto.
13531 * gimple-range.h: Include gimple-range-trace.h.
13532 (range_on_entry, range_on_exit): No longer virtual.
13533 (class trace_ranger): Remove.
13534 (DEBUG_RANGE_CACHE): Move to gimple-range-trace.h.
13536 2021-08-17 Martin Sebor <msebor@redhat.com>
13538 PR middle-end/101854
13539 * builtins.c (expand_builtin_alloca): Move warning code to check_alloca
13540 in gimple-ssa-warn-access.cc.
13541 * calls.c (alloc_max_size): Move code to check_alloca.
13542 (get_size_range): Move to pointer-query.cc.
13543 (maybe_warn_alloc_args_overflow): Move to gimple-ssa-warn-access.cc.
13544 (get_attr_nonstring_decl): Move to tree.c.
13545 (fntype_argno_type): Move to gimple-ssa-warn-access.cc.
13546 (append_attrname): Same.
13547 (maybe_warn_rdwr_sizes): Same.
13548 (initialize_argument_information): Move code to
13549 gimple-ssa-warn-access.cc.
13550 * calls.h (maybe_warn_alloc_args_overflow): Move to
13551 gimple-ssa-warn-access.h.
13552 (get_attr_nonstring_decl): Move to tree.h.
13553 (maybe_warn_nonstring_arg): Move to gimple-ssa-warn-access.h.
13554 (enum size_range_flags): Move to pointer-query.h.
13555 (get_size_range): Same.
13556 * gimple-ssa-warn-access.cc (has_location): Remove unused overload
13557 to avoid Clang -Wunused-function.
13558 (get_size_range): Declare static.
13559 (maybe_emit_free_warning): Rename...
13560 (maybe_check_dealloc_call): ...to this for consistency.
13561 (class pass_waccess): Add members.
13562 (pass_waccess::~pass_waccess): Defined.
13563 (alloc_max_size): Move here from calls.c.
13564 (maybe_warn_alloc_args_overflow): Same.
13565 (check_alloca): New function.
13566 (check_alloc_size_call): New function.
13567 (check_strncat): Handle another warning flag.
13568 (pass_waccess::check_builtin): Handle alloca.
13569 (fntype_argno_type): Move here from calls.c.
13570 (append_attrname): Same.
13571 (maybe_warn_rdwr_sizes): Same.
13572 (pass_waccess::check_call): Define.
13573 (check_nonstring_args): New function.
13574 (pass_waccess::check): Call new member functions.
13575 (pass_waccess::execute): Enable ranger.
13576 * gimple-ssa-warn-access.h (get_size_range): Move here from calls.h.
13577 (maybe_warn_nonstring_arg): Same.
13578 * gimple-ssa-warn-restrict.c: Remove #include.
13579 * pointer-query.cc (get_size_range): Move here from calls.c.
13580 * pointer-query.h (enum size_range_flags): Same.
13581 (get_size_range): Same.
13582 * tree.c (get_attr_nonstring_decl): Move here from calls.c.
13583 * tree.h (get_attr_nonstring_decl): Move here from calls.h.
13585 2021-08-17 Thomas Schwinge <thomas@codesourcery.com>
13587 * ggc.h (ggc_collect): Add 'force_collect' parameter.
13588 * ggc-page.c (ggc_collect): Use that one instead of global
13589 'ggc_force_collect'. Adjust all users.
13590 * doc/gty.texi (Invoking the garbage collector): Update.
13591 * ggc-internal.h (ggc_force_collect): Remove.
13592 * ggc-common.c (ggc_force_collect): Likewise.
13593 * selftest.h (forcibly_ggc_collect): Remove.
13594 * ggc-tests.c (selftest::forcibly_ggc_collect): Likewise.
13595 * read-rtl-function.c (test_loading_labels): Adjust.
13596 * selftest-run-tests.c (run_tests): Likewise.
13598 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
13600 * config/darwin.c (darwin_file_end): Reset and reclaim the
13601 section names table at the end of compile.
13603 2021-08-17 Iain Sandoe <iain@sandoe.co.uk>
13606 * config.in: Regenerate.
13607 * config/i386/darwin.h (EXTRA_ASM_OPTS): New
13608 (ASM_SPEC): Pass options to disable branch shortening where
13610 * configure: Regenerate.
13611 * configure.ac: Detect versions of 'as' that support the
13612 optimisation which has the bug.
13614 2021-08-17 Richard Biener <rguenther@suse.de>
13616 * optabs-query.c (supports_vec_gather_load_p): Also check
13618 (supports_vec_scatter_store_p): Likewise.
13619 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Fall
13620 back to masked variants if non-masked are not supported.
13621 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
13622 When we need to use masked gather/scatter but do not have
13623 a mask set up a constant true one.
13624 * tree-vect-stmts.c (vect_check_scalar_mask): Also allow
13625 non-SSA_NAME masks.
13627 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
13629 * tree-ssa-ccp.c (bit_value_binop) [MINUS_EXPR]: Use same
13630 algorithm as PLUS_EXPR to improve subtraction bit bounds.
13631 [POINTER_DIFF_EXPR]: Treat as synonymous with MINUS_EXPR.
13633 2021-08-17 Roger Sayle <roger@nextmovesoftware.com>
13635 * tree-ssa-ccp.c (bit_value_mult_const): New helper function to
13636 calculate the mask-value pair result of a multiplication by an
13638 (bit_value_binop) [MULT_EXPR]: Call it from here for
13639 multiplications by (sparse) non-negative constants.
13641 2021-08-17 Christophe Lyon <christophe.lyon@foss.st.com>
13644 * config.gcc (gcc_cv_initfini_array): Leave undefined for
13645 uclinuxfdpiceabi targets.
13647 2021-08-17 Alexandre Oliva <oliva@adacore.com>
13649 * tree-inline.c (maybe_move_debug_stmts_to_successors): Don't
13650 reverse debug stmts.
13652 2021-08-17 Alexandre Oliva <oliva@adacore.com>
13654 * tree-cfg.c (dump_function_to_file): Use fun, not cfun.
13656 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
13658 * config/aarch64/arm_neon.h (__LD4_LANE_FUNC): Delete.
13659 (__LD4Q_LANE_FUNC): Likewise.
13660 (vld4_lane_u8): Define without macro.
13661 (vld4_lane_u16): Likewise.
13662 (vld4_lane_u32): Likewise.
13663 (vld4_lane_u64): Likewise.
13664 (vld4_lane_s8): Likewise.
13665 (vld4_lane_s16): Likewise.
13666 (vld4_lane_s32): Likewise.
13667 (vld4_lane_s64): Likewise.
13668 (vld4_lane_f16): Likewise.
13669 (vld4_lane_f32): Likewise.
13670 (vld4_lane_f64): Likewise.
13671 (vld4_lane_p8): Likewise.
13672 (vld4_lane_p16): Likewise.
13673 (vld4_lane_p64): Likewise.
13674 (vld4q_lane_u8): Likewise.
13675 (vld4q_lane_u16): Likewise.
13676 (vld4q_lane_u32): Likewise.
13677 (vld4q_lane_u64): Likewise.
13678 (vld4q_lane_s8): Likewise.
13679 (vld4q_lane_s16): Likewise.
13680 (vld4q_lane_s32): Likewise.
13681 (vld4q_lane_s64): Likewise.
13682 (vld4q_lane_f16): Likewise.
13683 (vld4q_lane_f32): Likewise.
13684 (vld4q_lane_f64): Likewise.
13685 (vld4q_lane_p8): Likewise.
13686 (vld4q_lane_p16): Likewise.
13687 (vld4q_lane_p64): Likewise.
13688 (vld4_lane_bf16): Likewise.
13689 (vld4q_lane_bf16): Likewise.
13691 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
13693 * config/aarch64/arm_neon.h (__LD3_LANE_FUNC): Delete.
13694 (__LD3Q_LANE_FUNC): Delete.
13695 (vld3_lane_u8): Define without macro.
13696 (vld3_lane_u16): Likewise.
13697 (vld3_lane_u32): Likewise.
13698 (vld3_lane_u64): Likewise.
13699 (vld3_lane_s8): Likewise.
13700 (vld3_lane_s16): Likewise.
13701 (vld3_lane_s32): Likewise.
13702 (vld3_lane_s64): Likewise.
13703 (vld3_lane_f16): Likewise.
13704 (vld3_lane_f32): Likewise.
13705 (vld3_lane_f64): Likewise.
13706 (vld3_lane_p8): Likewise.
13707 (vld3_lane_p16): Likewise.
13708 (vld3_lane_p64): Likewise.
13709 (vld3q_lane_u8): Likewise.
13710 (vld3q_lane_u16): Likewise.
13711 (vld3q_lane_u32): Likewise.
13712 (vld3q_lane_u64): Likewise.
13713 (vld3q_lane_s8): Likewise.
13714 (vld3q_lane_s16): Likewise.
13715 (vld3q_lane_s32): Likewise.
13716 (vld3q_lane_s64): Likewise.
13717 (vld3q_lane_f16): Likewise.
13718 (vld3q_lane_f32): Likewise.
13719 (vld3q_lane_f64): Likewise.
13720 (vld3q_lane_p8): Likewise.
13721 (vld3q_lane_p16): Likewise.
13722 (vld3q_lane_p64): Likewise.
13723 (vld3_lane_bf16): Likewise.
13724 (vld3q_lane_bf16): Likewise.
13726 2021-08-17 Jonathan Wright <jonathan.wright@arm.com>
13728 * config/aarch64/arm_neon.h (__LD2_LANE_FUNC): Delete.
13729 (__LD2Q_LANE_FUNC): Likewise.
13730 (vld2_lane_u8): Define without macro.
13731 (vld2_lane_u16): Likewise.
13732 (vld2_lane_u32): Likewise.
13733 (vld2_lane_u64): Likewise.
13734 (vld2_lane_s8): Likewise.
13735 (vld2_lane_s16): Likewise.
13736 (vld2_lane_s32): Likewise.
13737 (vld2_lane_s64): Likewise.
13738 (vld2_lane_f16): Likewise.
13739 (vld2_lane_f32): Likewise.
13740 (vld2_lane_f64): Likewise.
13741 (vld2_lane_p8): Likewise.
13742 (vld2_lane_p16): Likewise.
13743 (vld2_lane_p64): Likewise.
13744 (vld2q_lane_u8): Likewise.
13745 (vld2q_lane_u16): Likewise.
13746 (vld2q_lane_u32): Likewise.
13747 (vld2q_lane_u64): Likewise.
13748 (vld2q_lane_s8): Likewise.
13749 (vld2q_lane_s16): Likewise.
13750 (vld2q_lane_s32): Likewise.
13751 (vld2q_lane_s64): Likewise.
13752 (vld2q_lane_f16): Likewise.
13753 (vld2q_lane_f32): Likewise.
13754 (vld2q_lane_f64): Likewise.
13755 (vld2q_lane_p8): Likewise.
13756 (vld2q_lane_p16): Likewise.
13757 (vld2q_lane_p64): Likewise.
13758 (vld2_lane_bf16): Likewise.
13759 (vld2q_lane_bf16): Likewise.
13761 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
13763 * haifa-sched.c (advance_one_cycle): Output more context-synchronization
13766 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
13768 * haifa-sched.c (enum rfs_decision, rfs_str): Add RFS_AUTOPREF.
13769 (rank_for_schedule): Use it.
13771 2021-08-17 Maxim Kuvyrkov <maxim.kuvyrkov@linaro.org>
13773 PR rtl-optimization/91598
13774 * haifa-sched.c (autopref_rank_for_schedule): Prioritize "irrelevant"
13775 insns after memory reads and before memory writes.
13777 2021-08-17 Alistair_Lee <alistair.lee@arm.com>
13779 * rtl.h (CONST_VECTOR_P): New macro.
13780 * config/aarch64/aarch64.c (aarch64_get_sve_pred_bits): Use RTL
13781 code testing macros.
13782 (aarch64_ptrue_all_mode): Likewise.
13783 (aarch64_expand_mov_immediate): Likewise.
13784 (aarch64_const_vec_all_in_range_p): Likewise.
13785 (aarch64_rtx_costs): Likewise.
13786 (aarch64_legitimate_constant_p): Likewise.
13787 (aarch64_simd_valid_immediate): Likewise.
13788 (aarch64_simd_make_constant): Likewise.
13789 (aarch64_convert_mult_to_shift): Likewise.
13790 (aarch64_expand_sve_vec_perm): Likewise.
13791 (aarch64_vec_fpconst_pow_of_2): Likewise.
13793 2021-08-17 Andrew MacLeod <amacleod@redhat.com>
13795 PR tree-optimization/101938
13796 * range-op.cc (operator_abs::op1_range): Special case
13797 -TYPE_MIN_VALUE for flag_wrapv.
13799 2021-08-17 Kewen Lin <linkw@linux.ibm.com>
13801 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Add the cost for
13804 2021-08-17 Jakub Jelinek <jakub@redhat.com>
13806 * tree.def (OMP_SCOPE): New tree code.
13807 * tree.h (OMP_SCOPE_BODY, OMP_SCOPE_CLAUSES): Define.
13808 * tree-nested.c (convert_nonlocal_reference_stmt,
13809 convert_local_reference_stmt, convert_gimple_call): Handle
13811 * tree-pretty-print.c (dump_generic_node): Handle OMP_SCOPE.
13812 * gimple.def (GIMPLE_OMP_SCOPE): New gimple code.
13813 * gimple.c (gimple_build_omp_scope): New function.
13814 (gimple_copy): Handle GIMPLE_OMP_SCOPE.
13815 * gimple.h (gimple_build_omp_scope): Declare.
13816 (gimple_has_substatements): Handle GIMPLE_OMP_SCOPE.
13817 (gimple_omp_scope_clauses, gimple_omp_scope_clauses_ptr,
13818 gimple_omp_scope_set_clauses): New inline functions.
13819 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_SCOPE.
13820 * gimple-pretty-print.c (dump_gimple_omp_scope): New function.
13821 (pp_gimple_stmt_1): Handle GIMPLE_OMP_SCOPE.
13822 * gimple-walk.c (walk_gimple_stmt): Likewise.
13823 * gimple-low.c (lower_stmt): Likewise.
13824 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
13825 (gimplify_scan_omp_clauses): For task reductions, handle OMP_SCOPE
13826 like ORT_WORKSHARE constructs. Adjust diagnostics for %<scope%>
13827 allowing task reductions. Reject inscan reductions on scope.
13828 (omp_find_stores_stmt): Handle GIMPLE_OMP_SCOPE.
13829 (gimplify_omp_workshare, gimplify_expr): Handle OMP_SCOPE.
13830 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_SCOPE.
13831 (estimate_num_insns): Likewise.
13832 * omp-low.c (build_outer_var_ref): Look through GIMPLE_OMP_SCOPE
13833 contexts if var isn't privatized there.
13834 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_SCOPE.
13835 (scan_omp_1_stmt): Likewise.
13836 (maybe_add_implicit_barrier_cancel): Look through outer
13838 (lower_omp_scope): New function.
13839 (lower_omp_task_reductions): Handle OMP_SCOPE.
13840 (lower_omp_1): Handle GIMPLE_OMP_SCOPE.
13841 (diagnose_sb_1, diagnose_sb_2): Likewise.
13842 * omp-expand.c (expand_omp_single): Support also GIMPLE_OMP_SCOPE.
13843 (expand_omp): Handle GIMPLE_OMP_SCOPE.
13844 (omp_make_gimple_edges): Likewise.
13845 * omp-builtins.def (BUILT_IN_GOMP_SCOPE_START): New built-in.
13847 2021-08-17 Richard Biener <rguenther@suse.de>
13849 PR tree-optimization/101925
13850 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Set
13851 reverse on COMPONENT_REF and ARRAY_REF according to
13852 what reverse_storage_order_for_component_p does.
13853 (vn_reference_eq): Compare reversed on reference ops.
13854 (reverse_storage_order_for_component_p): New overload.
13855 (vn_reference_lookup_3): Check reverse_storage_order_for_component_p
13856 on the reference looked up.
13858 2021-08-17 Jeff Law <jlaw@localhost.localdomain>
13860 * config/h8300/h8300.c (shift_alg_si): Avoid loops for most SImode
13861 shifts on the H8/S.
13862 (h8300_option_override): Use loops on H8/S more often when optimizing
13864 (get_shift_alg): Handle new "special" cases on H8/S. Simplify
13865 accordingly. Handle various arithmetic right shifts with special
13866 sequences that we couldn't handle before.
13868 2021-08-16 Jeff Law <jlaw@localhost.localdomain>
13870 * config.gcc (rl78-*-elf*): Do not include dbxelf.h.
13872 2021-08-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
13874 * config/sparc/rtemself.h (SPARC_GCOV_TYPE_SIZE): Define.
13875 * config/sparc/sparc.c (sparc_gcov_type_size): New.
13876 (TARGET_GCOV_TYPE_SIZE): Redefine if SPARC_GCOV_TYPE_SIZE is defined.
13877 * coverage.c (get_gcov_type): Use targetm.gcov_type_size().
13878 * doc/tm.texi (TARGET_GCOV_TYPE_SIZE): Add hook under "Misc".
13879 * doc/tm.texi.in: Regenerate.
13880 * target.def (gcov_type_size): New target hook.
13881 * targhooks.c (default_gcov_type_size): New.
13882 * targhooks.h (default_gcov_type_size): Declare.
13883 * tree-profile.c (gimple_gen_edge_profiler): Use precision of
13885 (gimple_gen_time_profiler): Likewise.
13887 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
13889 * dwarf2out.c (add_scalar_info): Deal with DW_AT_data_bit_offset.
13891 2021-08-16 Tobias Burnus <tobias@codesourcery.com>
13893 PR middle-end/101931
13894 * omp-low.c (omp_runtime_api_call): Update for routines
13895 added in the meanwhile.
13897 2021-08-16 Martin Liska <mliska@suse.cz>
13899 PR tree-optimization/100393
13900 * tree-switch-conversion.c (group_cluster::dump): Use
13901 get_comparison_count.
13902 (jump_table_cluster::find_jump_tables): Pre-compute number of
13903 comparisons and then decrement it. Cache also max_ratio.
13904 (jump_table_cluster::can_be_handled): Change signature.
13905 * tree-switch-conversion.h (get_comparison_count): New.
13907 2021-08-16 Eric Botcazou <ebotcazou@gcc.gnu.org>
13909 * dwarf2out.c (add_data_member_location_attribute): Use GNAT
13910 encodings only when -fgnat-encodings=all is specified.
13911 (add_bound_info): Likewise.
13912 (add_byte_size_attribute): Likewise.
13913 (gen_member_die): Likewise.
13915 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
13917 * omp-oacc-neuter-broadcast.cc
13918 (execute_omp_oacc_neuter_broadcast): Plug 'par' memory leak.
13920 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
13922 * omp-oacc-neuter-broadcast.cc
13923 (execute_omp_oacc_neuter_broadcast): Clarify memory management for
13926 2021-08-16 Thomas Schwinge <thomas@codesourcery.com>
13928 * omp-oacc-neuter-broadcast.cc (field_map): Move variable into...
13929 (execute_omp_oacc_neuter_broadcast): ... here.
13930 (install_var_field, build_receiver_ref, build_sender_ref): Take
13931 'field_map_t *' parameter. Adjust all users.
13932 (worker_single_copy, neuter_worker_single): Take a
13933 'record_field_map_t *' parameter. Adjust all users.
13935 2021-08-16 liuhongt <hongtao.liu@intel.com>
13938 * config/i386/i386.md (ldexp<mode>3): Force operands[1] to
13941 2021-08-16 Martin Liska <mliska@suse.cz>
13944 * multiple_target.c (create_dispatcher_calls): Make default
13945 function local only if it is a definition.
13947 2021-08-16 Martin Liska <mliska@suse.cz>
13950 * ipa-icf-gimple.c (func_checker::compare_ssa_name): Do not
13951 consider equal SSA_NAMEs when one is a param.
13953 2021-08-16 liuhongt <hongtao.liu@intel.com>
13956 * config/i386/i386-expand.c (ix86_expand_vec_perm_vpermt2):
13957 Support vpermi2b for V32QI/V16QImode.
13958 (ix86_extract_perm_from_pool_constant): New function.
13959 (ix86_expand_vec_one_operand_perm_avx512): Support
13960 vpermw/vpermb under TARGET_AVX512BW/TARGET_AVX512VBMI.
13961 (expand_vec_perm_1): Adjust comments for upper.
13962 * config/i386/i386-protos.h (ix86_extract_perm_from_pool_constant):
13964 * config/i386/predicates.md (permvar_truncate_operand): New predicate.
13965 (pshufb_truncv4siv4hi_operand): Ditto.
13966 (pshufb_truncv8hiv8qi_operand): Ditto.
13967 * config/i386/sse.md (*avx512bw_permvar_truncv16siv16hi_1):
13968 New pre_reload define_insn_and_split.
13969 (*avx512f_permvar_truncv8siv8hi_1): Ditto.
13970 (*avx512f_vpermvar_truncv8div8si_1): Ditto.
13971 (*avx512f_permvar_truncv32hiv32qi_1): Ditto.
13972 (*avx512f_permvar_truncv16hiv16qi_1): Ditto.
13973 (*avx512f_permvar_truncv4div4si_1): Ditto.
13974 (*avx512f_pshufb_truncv8hiv8qi_1): Ditto.
13975 (*avx512f_pshufb_truncv4siv4hi_1): Ditto.
13976 (*avx512f_pshufd_truncv2div2si_1): Ditto.
13978 2021-08-16 Kito Cheng <kito.cheng@sifive.com>
13980 * config/riscv/multilib-generator: Support code model option for
13982 * doc/install.texi: Add document of new option for
13983 --with-multilib-generator.
13985 2021-08-15 Clément Chigot <clement.chigot@atos.net>
13987 * config/rs6000/rs6000.c (xcoff_tls_exec_model_detected): New.
13988 (rs6000_legitimize_tls_address_aix): Use it.
13989 (rs6000_xcoff_file_end): Add ".ref __tls_get_addr" when
13990 xcoff_tls_exec_model_detected is true.
13992 2021-08-15 Jeff Law <jlaw@localhost.localdomain>
13994 * config/h8300/h8300.c (shift_alg_si): Retune H8/300H shifts
13995 to allow a bit more code growth, saving many dozens of cycles.
13996 (h8300_option_override): Adjus shift_alg_si if optimizing for
13998 (get_shift_alg): Use special + inline shifts for residuals
14001 2021-08-14 Stafford Horne <shorne@gmail.com>
14004 * config/or1k/or1k-opts.h: New file.
14005 * config/or1k/or1k.c (or1k_legitimize_address_1, print_reloc):
14006 Support generating gotha relocations if -mcmodel=large is
14008 * config/or1k/or1k.h (TARGET_CMODEL_SMALL, TARGET_CMODEL_LARGE):
14010 * config/or1k/or1k.opt (mcmodel=): New option.
14011 * doc/invoke.texi (OpenRISC Options): Document mcmodel.
14013 2021-08-14 Martin Sebor <msebor@redhat.com>
14015 PR middle-end/101791
14016 * gimple-ssa-warn-access.cc (new_delete_mismatch_p): Use new argument
14017 to valid_new_delete_pair_p.
14018 * tree.c (valid_new_delete_pair_p): Add argument.
14019 * tree.h (valid_new_delete_pair_p): Same.
14021 2021-08-14 Jakub Jelinek <jakub@redhat.com>
14024 * config/i386/i386-expand.c (expand_vec_perm_broadcast_1)
14025 <case E_V64QImode>: For this mode assert
14026 !TARGET_AVX512BW || d->perm[0] rather than !TARGET_AVX2 || d->perm[0].
14028 2021-08-13 Michael Meissner <meissner@linux.ibm.com>
14031 * config/rs6000/altivec.md (xxeval): Use register_predicate
14032 instead of altivec_register_predicate.
14034 2021-08-13 Martin Sebor <msebor@redhat.com>
14036 PR middle-end/101734
14037 * tree-ssa-uninit.c (maybe_warn_read_write_only): New function.
14038 (maybe_warn_operand): Call it.
14040 2021-08-13 Martin Liska <mliska@suse.cz>
14043 * attribs.c (decl_attributes): Make naked functions "noipa"
14046 2021-08-13 Martin Liska <mliska@suse.cz>
14049 * symtab.c (symtab_node::noninterposable_alias): Do not create
14050 local aliases for target_clone functions as the clonning pass
14053 2021-08-13 Martin Liska <mliska@suse.cz>
14055 * opts.c (LIVE_PATCHING_OPTION): Define.
14056 (control_options_for_live_patching): Use it in error messages.
14058 2021-08-13 Jan Hubicka <hubicka@ucw.cz>
14060 * ipa-modref.c (dump_eaf_flags): Dump EAF_NOREAD.
14061 (implicit_const_eaf_flags, implicit_pure_eaf_flags,
14062 ignore_stores_eaf_flags): New constants.
14063 (remove_useless_eaf_flags): New function.
14064 (eaf_flags_useful_p): Use it.
14065 (deref_flags): Add EAF_NOT_RETURNED if flag is unused;
14067 (modref_lattice::init): Add EAF_NOREAD.
14068 (modref_lattice::add_escape_point): Do not reacord escape point if
14070 (modref_lattice::merge): EAF_NOESCAPE implies EAF_NODIRECTESCAPE;
14071 use remove_useless_eaf_flags.
14072 (modref_lattice::merge_deref): Use ignore_stores_eaf_flags.
14073 (modref_lattice::merge_direct_load): Add EAF_NOREAD
14074 (analyze_ssa_name_flags): Fix handling EAF_NOT_RETURNED
14075 (analyze_parms): Use remove_useless_eaf_flags.
14076 (ipa_merge_modref_summary_after_inlining): Use ignore_stores_eaf_flags.
14077 (modref_merge_call_site_flags): Add caller and ecf_flags parameter;
14078 use remove_useless_eaf_flags.
14079 (modref_propagate_flags_in_scc): Update.
14080 * ipa-modref.h: Turn eaf_flags_t back to char.
14081 * tree-core.h (EAF_NOT_RETURNED): Fix.
14082 (EAF_NOREAD): New constant
14083 * tree-ssa-alias.c: (ref_maybe_used_by_call_p_1): Check for
14085 * tree-ssa-structalias.c (handle_rhs_call): Handle new flags.
14086 (handle_pure_call): Likewise.
14088 2021-08-12 Jakub Jelinek <jakub@redhat.com>
14090 * tree.def (OMP_MASKED): New tree code.
14091 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_FILTER.
14092 * tree.h (OMP_MASKED_BODY, OMP_MASKED_CLAUSES, OMP_MASKED_COMBINED,
14093 OMP_CLAUSE_FILTER_EXPR): Define.
14094 * tree.c (omp_clause_num_ops): Add OMP_CLAUSE_FILTER entry.
14095 (omp_clause_code_name): Likewise.
14096 (walk_tree_1): Handle OMP_CLAUSE_FILTER.
14097 * tree-nested.c (convert_nonlocal_omp_clauses,
14098 convert_local_omp_clauses): Handle OMP_CLAUSE_FILTER.
14099 (convert_nonlocal_reference_stmt, convert_local_reference_stmt,
14100 convert_gimple_call): Handle GIMPLE_OMP_MASTER.
14101 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_FILTER.
14102 (dump_generic_node): Handle OMP_MASTER.
14103 * gimple.def (GIMPLE_OMP_MASKED): New gimple code.
14104 * gimple.c (gimple_build_omp_masked): New function.
14105 (gimple_copy): Handle GIMPLE_OMP_MASKED.
14106 * gimple.h (gimple_build_omp_masked): Declare.
14107 (gimple_has_substatements): Handle GIMPLE_OMP_MASKED.
14108 (gimple_omp_masked_clauses, gimple_omp_masked_clauses_ptr,
14109 gimple_omp_masked_set_clauses): New inline functions.
14110 (CASE_GIMPLE_OMP): Add GIMPLE_OMP_MASKED.
14111 * gimple-pretty-print.c (dump_gimple_omp_masked): New function.
14112 (pp_gimple_stmt_1): Handle GIMPLE_OMP_MASKED.
14113 * gimple-walk.c (walk_gimple_stmt): Likewise.
14114 * gimple-low.c (lower_stmt): Likewise.
14115 * gimplify.c (is_gimple_stmt): Handle OMP_MASTER.
14116 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_FILTER. For clauses
14117 that take one expression rather than decl or constant, force
14118 gimplification of that into a SSA_NAME or temporary unless min
14120 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_FILTER.
14121 (gimplify_expr): Handle OMP_MASKED.
14122 * tree-inline.c (remap_gimple_stmt): Handle GIMPLE_OMP_MASKED.
14123 (estimate_num_insns): Likewise.
14124 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_FILTER.
14125 (check_omp_nesting_restrictions): Handle GIMPLE_OMP_MASKED. Adjust
14126 diagnostics for existence of masked construct.
14127 (scan_omp_1_stmt, lower_omp_master, lower_omp_1, diagnose_sb_1,
14128 diagnose_sb_2): Handle GIMPLE_OMP_MASKED.
14129 * omp-expand.c (expand_omp_synch, expand_omp, omp_make_gimple_edges):
14132 2021-08-12 Uroš Bizjak <ubizjak@gmail.com>
14135 * config/i386/i386.md (avx512f_scalef<mode>2): New insn pattern.
14136 (ldexp<mode>3): Use avx512f_scalef<mode>2.
14137 (UNSPEC_SCALEF): Move from ...
14138 * config/i386/sse.md (UNSPEC_SCALEF): ... here.
14140 2021-08-12 Jan Hubicka <hubicka@ucw.cz>
14142 * ipa-split.c (consider_split): Fix condition testing void functions.
14144 2021-08-12 Aldy Hernandez <aldyh@redhat.com>
14146 * doc/invoke.texi: Remove docs for threader-mode param.
14147 * flag-types.h (enum threader_mode): Remove.
14148 * params.opt: Remove threader-mode param.
14149 * tree-ssa-threadbackward.c (class back_threader): Remove
14150 path_is_unreachable_p.
14151 Make find_paths private.
14152 Add maybe_thread and thread_through_all_blocks.
14153 Remove reference marker for m_registry.
14154 Remove reference marker for m_profit.
14155 (back_threader::back_threader): Adjust for registry and profit not
14157 (dump_path): Move down.
14158 (debug): Move down.
14159 (class thread_jumps): Remove.
14160 (class back_threader_registry): Remove m_all_paths.
14162 (thread_jumps::thread_through_all_blocks): Move to back_threader
14164 (fsm_find_thread_path): Remove
14165 (back_threader::maybe_thread): New.
14166 (back_threader::thread_through_all_blocks): Move from
14168 (back_threader_registry::back_threader_registry): Remove
14170 (back_threader_registry::~back_threader_registry): Remove.
14171 (thread_jumps::find_taken_edge): Remove.
14172 (thread_jumps::check_subpath_and_update_thread_path): Remove.
14173 (thread_jumps::maybe_register_path): Remove.
14174 (thread_jumps::handle_phi): Remove.
14175 (handle_assignment_p): Remove.
14176 (thread_jumps::handle_assignment): Remove.
14177 (thread_jumps::fsm_find_control_statement_thread_paths): Remove.
14178 (thread_jumps::find_jump_threads_backwards): Remove.
14179 (thread_jumps::find_jump_threads_backwards_with_ranger): Remove.
14180 (try_thread_blocks): Rename find_jump_threads_backwards to
14182 (pass_early_thread_jumps::execute): Same.
14184 2021-08-12 Tobias Burnus <tobias@codesourcery.com>
14186 * tree-core.h (omp_clause_proc_bind_kind): Add
14187 OMP_CLAUSE_PROC_BIND_PRIMARY.
14188 * tree-pretty-print.c (dump_omp_clause): Add TODO comment to
14189 change 'master' to 'primary' in proc_bind for OpenMP 5.1.
14191 2021-08-12 Claudiu Zissulescu <claziss@synopsys.com>
14193 * common/config/arc/arc-common.c (arc_option_init_struct): Remove
14194 fno-common reference.
14195 * config/arc/arc.c (arc_override_options): Remove overriding of
14198 2021-08-12 Jakub Jelinek <jakub@redhat.com>
14201 * config/i386/i386-expand.c (ix86_expand_vec_one_operand_perm_avx512):
14202 If d->testing_p, return true after performing checks instead of
14203 actually expanding the insn.
14204 (expand_vec_perm_broadcast_1): Handle V32HImode - assert
14205 !TARGET_AVX512BW and return false.
14207 2021-08-12 Eric Botcazou <ebotcazou@gcc.gnu.org>
14209 * configure.ac (PE linker --disable-dynamicbase support): New check.
14210 * configure: Regenerate.
14211 * config.in: Likewise.
14212 * config/i386/mingw32.h (LINK_SPEC_DISABLE_DYNAMICBASE): New define.
14213 (LINK_SPEC): Use it.
14214 * config/i386/mingw-w64.h (LINK_SPEC_DISABLE_DYNAMICBASE): Likewise.
14215 (LINK_SPEC): Likewise.
14217 2021-08-12 liuhongt <hongtao.liu@intel.com>
14220 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_2): New
14221 post_reload define_insn_and_split.
14222 (*avx512bw_zero_extendv32qiv32hi2_2): Ditto.
14223 (*sse4_1_zero_extendv8qiv8hi2_4): Ditto.
14224 (*avx512f_zero_extendv16hiv16si2_2): Ditto.
14225 (*avx2_zero_extendv8hiv8si2_2): Ditto.
14226 (*sse4_1_zero_extendv4hiv4si2_4): Ditto.
14227 (*avx512f_zero_extendv8siv8di2_2): Ditto.
14228 (*avx2_zero_extendv4siv4di2_2): Ditto.
14229 (*sse4_1_zero_extendv2siv2di2_4): Ditto.
14230 (VI248_256, VI248_512, VI148_512, VI148_256, VI148_128): New
14233 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
14235 * config/rs6000/rs6000-builtin-new.def: Add always, power5, and
14238 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
14240 * config/rs6000/rs6000-builtin-new.def: Add vsx stanza.
14242 2021-08-11 Bill Schmidt <wschmidt@linux.ibm.com>
14244 * config/rs6000/rs6000-builtin-new.def: Finish altivec stanza.
14245 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Move
14246 initialization of pcvoid_type_node here...
14247 (altivec_init_builtins): ...from here.
14248 * config/rs6000/rs6000.h (rs6000_builtin_type_index): Add
14249 RS6000_BTI_const_ptr_void.
14250 (pcvoid_type_node): New macro.
14252 2021-08-11 Richard Biener <rguenther@suse.de>
14255 * tree-ssa-forwprop.c (pass_forwprop::execute): Do not decompose
14256 hard-register accesses.
14258 2021-08-11 Richard Biener <rguenther@suse.de>
14260 * tree-ssa-operands.c (operands_scanner::get_expr_operands):
14261 Do not look at COMPONENT_REF FIELD_DECLs TREE_THIS_VOLATILE
14262 to determine has_volatile_ops.
14264 2021-08-11 Eric Botcazou <ebotcazou@gcc.gnu.org>
14266 * cfgexpand.c (expand_used_vars): Reuse attribs local variable.
14268 2021-08-11 Jan Hubicka <hubicka@ucw.cz>
14269 Alexandre Oliva <oliva@adacore.com>
14271 * ipa-modref.c (modref_lattice::dump): Fix escape_point's min_flags
14273 (modref_lattice::merge_deref): Fix handling of indirect scape points.
14274 (update_escape_summary_1): Likewise.
14275 (update_escape_summary): Likewise.
14276 (ipa_merge_modref_summary_after_inlining): Likewise.
14278 2021-08-11 Richard Biener <rguenther@suse.de>
14280 PR middle-end/101858
14281 * fold-const.c (fold_binary_loc): Guard simplification
14282 of X < (cast) (1 << Y) to integer types.
14284 2021-08-11 Richard Biener <rguenther@suse.de>
14286 PR tree-optimization/101861
14287 * tree-vect-stmts.c (vectorizable_load): Fix error in
14288 previous change with regard to gather vectorization.
14290 2021-08-11 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
14293 * config/arm/arm_neon.h (vdup_n_s8): Replace call to builtin
14295 (vdup_n_s16): Likewise.
14296 (vdup_n_s32): Likewise.
14297 (vdup_n_s64): Likewise.
14298 (vdup_n_u8): Likewise.
14299 (vdup_n_u16): Likewise.
14300 (vdup_n_u32): Likewise.
14301 (vdup_n_u64): Likewise.
14302 (vdup_n_p8): Likewise.
14303 (vdup_n_p16): Likewise.
14304 (vdup_n_p64): Likewise.
14305 (vdup_n_f16): Likewise.
14306 (vdup_n_f32): Likewise.
14307 (vdupq_n_s8): Likewise.
14308 (vdupq_n_s16): Likewise.
14309 (vdupq_n_s32): Likewise.
14310 (vdupq_n_s64): Likewise.
14311 (vdupq_n_u8): Likewise.
14312 (vdupq_n_u16): Likewise.
14313 (vdupq_n_u32): Likewise.
14314 (vdupq_n_u64): Likewise.
14315 (vdupq_n_p8): Likewise.
14316 (vdupq_n_p16): Likewise.
14317 (vdupq_n_p64): Likewise.
14318 (vdupq_n_f16): Likewise.
14319 (vdupq_n_f32): Likewise.
14320 (vmov_n_s8): Replace call to builtin with call to corresponding
14322 (vmov_n_s16): Likewise.
14323 (vmov_n_s32): Likewise.
14324 (vmov_n_s64): Likewise.
14325 (vmov_n_u8): Likewise.
14326 (vmov_n_u16): Likewise.
14327 (vmov_n_u32): Likewise.
14328 (vmov_n_u64): Likewise.
14329 (vmov_n_p8): Likewise.
14330 (vmov_n_p16): Likewise.
14331 (vmov_n_f16): Likewise.
14332 (vmov_n_f32): Likewise.
14333 (vmovq_n_s8): Likewise.
14334 (vmovq_n_s16): Likewise.
14335 (vmovq_n_s32): Likewise.
14336 (vmovq_n_s64): Likewise.
14337 (vmovq_n_u8): Likewise.
14338 (vmovq_n_u16): Likewise.
14339 (vmovq_n_u32): Likewise.
14340 (vmovq_n_u64): Likewise.
14341 (vmovq_n_p8): Likewise.
14342 (vmovq_n_p16): Likewise.
14343 (vmovq_n_f16): Likewise.
14344 (vmovq_n_f32): Likewise.
14345 * config/arm/arm_neon_builtins.def: Remove entries for vdup_n.
14347 2021-08-11 liuhongt <hongtao.liu@intel.com>
14350 * config/i386/i386.md (ldexp<mode>3): Extend to vscalefs[sd]
14351 when TARGET_AVX512F and TARGET_SSE_MATH.
14353 2021-08-10 Jakub Jelinek <jakub@redhat.com>
14356 * config/i386/i386-expand.c (expand_vec_perm_even_odd): Return false
14357 for V32HImode if !TARGET_AVX512BW.
14358 (ix86_vectorize_vec_perm_const) <case E_V32HImode, case E_V64QImode>:
14359 If !TARGET_AVX512BW and TARGET_AVX512F and d.testing_p, don't fail
14360 early, but actually check the permutation.
14362 2021-08-10 Richard Biener <rguenther@suse.de>
14364 PR tree-optimization/101809
14365 * tree-vect-stmts.c (get_load_store_type): Allow emulated
14366 gathers with offset vector nunits being a constant multiple
14367 of the data vector nunits.
14368 (vect_get_gather_scatter_ops): Use the appropriate nunits
14369 for the offset vector defs.
14370 (vectorizable_store): Adjust call to
14371 vect_get_gather_scatter_ops.
14372 (vectorizable_load): Likewise. Handle the case of less
14373 offset vectors than data vectors.
14375 2021-08-10 Jakub Jelinek <jakub@redhat.com>
14378 * config/i386/sse.md (*avx512f_shuf_<shuffletype>64x2_1<mask_name>_1,
14379 *avx512f_shuf_<shuffletype>32x4_1<mask_name>_1): New define_insn
14382 2021-08-10 Richard Biener <rguenther@suse.de>
14384 PR tree-optimization/101801
14385 PR tree-optimization/101819
14386 * tree-vectorizer.h (vect_emulated_vector_p): Declare.
14387 * tree-vect-loop.c (vect_emulated_vector_p): New function.
14388 (vectorizable_reduction): Re-instantiate a check for emulated
14390 * tree-vect-stmts.c (vectorizable_shift): Likewise.
14391 (vectorizable_operation): Likewise. Cost emulated vector
14392 operations according to the scalar sequence synthesized by
14395 2021-08-10 Richard Biener <rguenther@suse.de>
14397 PR middle-end/101824
14398 * tree-nested.c (get_frame_field): Mark the COMPONENT_REF as
14399 volatile in case the variable was.
14401 2021-08-10 H.J. Lu <hjl.tools@gmail.com>
14404 * config/i386/constraints.md (BC): Document for integer SSE
14405 constant all bits set operand.
14406 (BF): New constraint for const floating-point all bits set
14408 * config/i386/i386.c (standard_sse_constant_p): Likewise.
14409 (standard_sse_constant_opcode): Likewise.
14410 * config/i386/sse.md (sseconstm1): New mode attribute.
14411 (mov<mode>_internal): Replace BC with <sseconstm1>.
14413 2021-08-10 liuhongt <hongtao.liu@intel.com>
14415 * config/i386/sse.md (cond_<insn><mode>): New expander.
14416 (VI248_AVX512VLBW): New mode iterator.
14417 * config/i386/predicates.md
14418 (nonimmediate_or_const_vec_dup_operand): New predicate.
14420 2021-08-09 Andrew MacLeod <amacleod@redhat.com>
14422 PR tree-optimization/101741
14423 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Check
14424 type of parameter for toupper/tolower.
14426 2021-08-09 Martin Jambor <mjambor@suse.cz>
14428 PR testsuite/101654
14429 * ipa-prop.c (propagate_controlled_uses): Removed a spurious space.
14431 2021-08-09 Pat Haugen <pthaugen@linux.ibm.com>
14433 * config/rs6000/rs6000.c (is_load_insn1): Verify destination is a
14435 (is_store_insn1): Verify source is a register.
14437 2021-08-09 Uroš Bizjak <ubizjak@gmail.com>
14440 * config/i386/mmx.md (<any_logic:code>v2sf3):
14441 Rename from *mmx_<any_logic:code>v2sf3
14443 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
14445 * config/nvptx/nvptx.c: Cross-reference parts adapted in
14446 'gcc/omp-oacc-neuter-broadcast.cc'.
14447 * omp-low.c: Likewise.
14448 * omp-oacc-neuter-broadcast.cc: Cross-reference parts adapted from
14451 2021-08-09 Julian Brown <julian@codesourcery.com>
14452 Kwok Cheung Yeung <kcy@codesourcery.com>
14453 Thomas Schwinge <thomas@codesourcery.com>
14455 * config/gcn/gcn.c (gcn_init_builtins): Override decls for
14456 BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START,
14457 BUILT_IN_GOACC_SINGLE_COPY_END and BUILT_IN_GOACC_BARRIER.
14458 (gcn_goacc_validate_dims): Turn on worker partitioning unconditionally.
14459 (gcn_fork_join): Update comment.
14460 * config/gcn/gcn.opt (flag_worker_partitioning): Remove.
14461 (macc_experimental_workers): Remove unused option.
14463 2021-08-09 Julian Brown <julian@codesourcery.com>
14464 Nathan Sidwell <nathan@codesourcery.com> (via 'gcc/config/nvptx/nvptx.c' master)
14465 Kwok Cheung Yeung <kcy@codesourcery.com>
14466 Thomas Schwinge <thomas@codesourcery.com>
14468 * Makefile.in (OBJS): Add omp-oacc-neuter-broadcast.o.
14469 * doc/tm.texi.in (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD):
14470 Add documentation hook.
14471 * doc/tm.texi: Regenerate.
14472 * omp-oacc-neuter-broadcast.cc: New file.
14473 * omp-builtins.def (BUILT_IN_GOACC_BARRIER)
14474 (BUILT_IN_GOACC_SINGLE_START, BUILT_IN_GOACC_SINGLE_COPY_START)
14475 (BUILT_IN_GOACC_SINGLE_COPY_END): New builtins.
14476 * passes.def (pass_omp_oacc_neuter_broadcast): Add pass.
14477 * target.def (goacc.create_worker_broadcast_record): Add target
14479 * tree-pass.h (make_pass_omp_oacc_neuter_broadcast): Add
14481 * config/gcn/gcn-protos.h (gcn_goacc_adjust_propagation_record):
14482 Rename prototype to...
14483 (gcn_goacc_create_worker_broadcast_record): ... this.
14484 * config/gcn/gcn-tree.c (gcn_goacc_adjust_propagation_record): Rename
14486 (gcn_goacc_create_worker_broadcast_record): ... this.
14487 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_PROPAGATION_RECORD):
14489 (TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD): ... this.
14491 2021-08-09 Tejas Belagod <tejas.belagod@arm.com>
14494 * config/aarch64/aarch64-simd.md (vlshr<mode>3, vashr<mode>3): Use
14495 the right iterator.
14497 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
14499 * Makefile.in (GTFILES): Remove '$(srcdir)/omp-offload.c'.
14501 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
14503 * builtins.def (DEF_GOACC_BUILTIN, DEF_GOMP_BUILTIN): Don't
14504 consider '-foffload-abi'.
14505 * common.opt (-foffload-abi): Remove 'Var', 'Init'.
14506 * opts.c (common_handle_option) <-foffload-abi> [ACCEL_COMPILER]:
14509 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
14511 * optc-gen.awk: Sanity check that 'Init' doesn't appear without
14514 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
14516 * omp-builtins.def (BUILT_IN_ACC_GET_DEVICE_TYPE): Remove.
14518 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
14520 * doc/gty.texi (Files): Update.
14522 2021-08-09 Thomas Schwinge <thomas@codesourcery.com>
14524 * doc/gty.texi (Files): Fix GTY header file example.
14526 2021-08-09 Roger Sayle <roger@nextmovesoftware.com>
14528 * tree-ssa-ccp.c (value_mask_to_min_max): Helper function to
14529 determine the upper and lower bounds from a mask-value pair.
14530 (bit_value_unop) [ABS_EXPR, ABSU_EXPR]: Add support for
14531 absolute value and unsigned absolute value expressions.
14532 (bit_value_binop): Initialize *VAL's precision.
14533 [LT_EXPR, LE_EXPR]: Use value_mask_to_min_max to determine
14534 upper and lower bounds of operands. Add LE_EXPR/GE_EXPR
14535 support when the operands are unknown but potentially equal.
14536 [MIN_EXPR, MAX_EXPR]: Support minimum/maximum expressions.
14538 2021-08-09 Bin Cheng <bin.cheng@linux.alibaba.com>
14540 * config/aarch64/aarch64.md
14541 (*extend<SHORT:mode><GPI:mode>2_aarch64): Use %<GPI:w>0.
14543 2021-08-08 Sergei Trofimovich <siarheit@google.com>
14545 * lra-constraints.c: Fix s/otput/output/ typo.
14547 2021-08-06 Martin Sebor <msebor@redhat.com>
14549 * builtins.c (expand_builtin_memchr): Move to gimple-ssa-warn-access.cc.
14550 (expand_builtin_strcat): Same.
14551 (expand_builtin_stpncpy): Same.
14552 (expand_builtin_strncat): Same.
14553 (check_read_access): Same.
14554 (check_memop_access): Same.
14555 (expand_builtin_strlen): Move checks to gimple-ssa-warn-access.cc.
14556 (expand_builtin_strnlen): Same.
14557 (expand_builtin_memcpy): Same.
14558 (expand_builtin_memmove): Same.
14559 (expand_builtin_mempcpy): Same.
14560 (expand_builtin_strcpy): Same.
14561 (expand_builtin_strcpy_args): Same.
14562 (expand_builtin_stpcpy_1): Same.
14563 (expand_builtin_strncpy): Same.
14564 (expand_builtin_memset): Same.
14565 (expand_builtin_bzero): Same.
14566 (expand_builtin_strcmp): Same.
14567 (expand_builtin_strncmp): Same.
14568 (expand_builtin): Remove handlers.
14569 (fold_builtin_strlen): Add a comment.
14570 * builtins.h (check_access): Move to gimple-ssa-warn-access.cc.
14571 * calls.c (maybe_warn_nonstring_arg): Same.
14572 * diagnostic-spec.c (nowarn_spec_t::nowarn_spec_t): Add warning option.
14573 * gimple-fold.c (gimple_fold_builtin_strcpy): Pass argument to callee.
14574 (gimple_fold_builtin_stpcpy): Same.
14575 * gimple-ssa-warn-access.cc (has_location): New function.
14576 (get_location): Same.
14577 (get_callee_fndecl): Same.
14578 (call_nargs): Same.
14580 (warn_string_no_nul): Define.
14581 (unterminated_array): Same.
14582 (check_nul_terminated_array): Same.
14583 (maybe_warn_nonstring_arg): Same.
14584 (maybe_warn_for_bound): Same.
14585 (warn_for_access): Same.
14586 (check_access): Same.
14587 (check_memop_access): Same.
14588 (check_read_access): Same.
14589 (warn_dealloc_offset): Use helper functions.
14590 (maybe_emit_free_warning): Same.
14591 (class pass_waccess): Add members.
14592 (check_strcat): New function.
14593 (check_strncat): New function.
14594 (check_stxcpy): New function.
14595 (check_stxncpy): New function.
14596 (check_strncmp): New function.
14597 (pass_waccess::check_builtin): New function.
14598 (pass_waccess::check): Call it.
14599 * gimple-ssa-warn-access.h (warn_string_no_nul): Move here from
14601 (maybe_warn_for_bound): Same.
14602 (check_access): Same.
14603 (check_memop_access): Same.
14604 (check_read_access): Same.
14605 * pointer-query.h (struct access_data): Define a ctor overload.
14607 2021-08-06 Richard Biener <rguenther@suse.de>
14609 PR tree-optimization/101801
14610 * tree-vectorizer.h (vect_worthwhile_without_simd_p): Rename...
14611 (vect_can_vectorize_without_simd_p): ... to this.
14612 * tree-vect-loop.c (vect_worthwhile_without_simd_p): Rename...
14613 (vect_can_vectorize_without_simd_p): ... to this and fold
14614 in vect_min_worthwhile_factor.
14615 (vect_min_worthwhile_factor): Remove.
14616 (vectorizable_reduction): Adjust and remove the cost part.
14617 * tree-vect-stmts.c (vectorizable_shift): Likewise.
14618 (vectorizable_operation): Likewise.
14620 2021-08-06 Uroš Bizjak <ubizjak@gmail.com>
14623 * config/i386/i386.md (cmove reg-to-reg move elimination peephole2s):
14624 Add general_gr_operand predicate to operand 3.
14626 2021-08-06 Roger Sayle <roger@nextmovesoftware.com>
14628 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern): Use
14629 CFN_BUILT_IN_CLRSB* instead of BUILT_IN_CLRSB* for consistency.
14631 2021-08-06 Tamar Christina <tamar.christina@arm.com>
14633 * config/aarch64/aarch64-sve-builtins.cc (register_svpattern,
14634 register_svprfop): Pass vec<> by pointer.
14635 * langhooks-def.h (lhd_simulate_enum_decl): Likewise.
14636 * langhooks.c (lhd_simulate_enum_decl): Likewise.
14637 * langhooks.h (struct lang_hooks_for_types): Likewise.
14639 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
14641 * config/aarch64/arm_neon.h (vst1_bf16_x2): Use
14642 __builtin_memcpy instead of constructing an additional
14643 __builtin_aarch64_simd_oi one vector at a time.
14644 (vst1q_bf16_x2): Likewise.
14645 (vst1_bf16_x3): Use __builtin_memcpy instead of constructing
14646 an additional __builtin_aarch64_simd_ci one vector at a time.
14647 (vst1q_bf16_x3): Likewise.
14648 (vst1_bf16_x4): Use __builtin_memcpy instead of a union.
14649 (vst1q_bf16_x4): Likewise.
14650 (vst2_bf16): Use __builtin_memcpy instead of constructing an
14651 additional __builtin_aarch64_simd_oi one vector at a time.
14652 (vst2q_bf16): Likewise.
14653 (vst3_bf16): Use __builtin_memcpy instead of constructing an
14654 additional __builtin_aarch64_simd_ci mode one vector at a
14656 (vst3q_bf16): Likewise.
14657 (vst4_bf16): Use __builtin_memcpy instead of constructing an
14658 additional __builtin_aarch64_simd_xi one vector at a time.
14659 (vst4q_bf16): Likewise.
14661 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
14663 * config/aarch64/arm_neon.h (__ST2_LANE_FUNC): Delete.
14664 (__ST2Q_LANE_FUNC): Delete.
14665 (vst2_lane_f16): Use __builtin_memcpy to copy vector
14666 structure instead of constructing __builtin_aarch64_simd_oi
14667 one vector at a time.
14668 (vst2_lane_f32): Likewise.
14669 (vst2_lane_f64): Likewise.
14670 (vst2_lane_p8): Likewise.
14671 (vst2_lane_p16): Likewise.
14672 (vst2_lane_p64): Likewise.
14673 (vst2_lane_s8): Likewise.
14674 (vst2_lane_s16): Likewise.
14675 (vst2_lane_s32): Likewise.
14676 (vst2_lane_s64): Likewise.
14677 (vst2_lane_u8): Likewise.
14678 (vst2_lane_u16): Likewise.
14679 (vst2_lane_u32): Likewise.
14680 (vst2_lane_u64): Likewise.
14681 (vst2_lane_bf16): Likewise.
14682 (vst2q_lane_f16): Use __builtin_memcpy to copy vector
14683 structure instead of using a union.
14684 (vst2q_lane_f32): Likewise.
14685 (vst2q_lane_f64): Likewise.
14686 (vst2q_lane_p8): Likewise.
14687 (vst2q_lane_p16): Likewise.
14688 (vst2q_lane_p64): Likewise.
14689 (vst2q_lane_s8): Likewise.
14690 (vst2q_lane_s16): Likewise.
14691 (vst2q_lane_s32): Likewise.
14692 (vst2q_lane_s64): Likewise.
14693 (vst2q_lane_u8): Likewise.
14694 (vst2q_lane_u16): Likewise.
14695 (vst2q_lane_u32): Likewise.
14696 (vst2q_lane_u64): Likewise.
14697 (vst2q_lane_bf16): Likewise.
14699 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
14701 * config/aarch64/arm_neon.h (__ST3_LANE_FUNC): Delete.
14702 (__ST3Q_LANE_FUNC): Delete.
14703 (vst3_lane_f16): Use __builtin_memcpy to copy vector
14704 structure instead of constructing __builtin_aarch64_simd_ci
14705 one vector at a time.
14706 (vst3_lane_f32): Likewise.
14707 (vst3_lane_f64): Likewise.
14708 (vst3_lane_p8): Likewise.
14709 (vst3_lane_p16): Likewise.
14710 (vst3_lane_p64): Likewise.
14711 (vst3_lane_s8): Likewise.
14712 (vst3_lane_s16): Likewise.
14713 (vst3_lane_s32): Likewise.
14714 (vst3_lane_s64): Likewise.
14715 (vst3_lane_u8): Likewise.
14716 (vst3_lane_u16): Likewise.
14717 (vst3_lane_u32): Likewise.
14718 (vst3_lane_u64): Likewise.
14719 (vst3_lane_bf16): Likewise.
14720 (vst3q_lane_f16): Use __builtin_memcpy to copy vector
14721 structure instead of using a union.
14722 (vst3q_lane_f32): Likewise.
14723 (vst3q_lane_f64): Likewise.
14724 (vst3q_lane_p8): Likewise.
14725 (vst3q_lane_p16): Likewise.
14726 (vst3q_lane_p64): Likewise.
14727 (vst3q_lane_s8): Likewise.
14728 (vst3q_lane_s16): Likewise.
14729 (vst3q_lane_s32): Likewise.
14730 (vst3q_lane_s64): Likewise.
14731 (vst3q_lane_u8): Likewise.
14732 (vst3q_lane_u16): Likewise.
14733 (vst3q_lane_u32): Likewise.
14734 (vst3q_lane_u64): Likewise.
14735 (vst3q_lane_bf16): Likewise.
14737 2021-08-06 Jonathan Wright <jonathan.wright@arm.com>
14739 * config/aarch64/arm_neon.h (__ST4_LANE_FUNC): Delete.
14740 (__ST4Q_LANE_FUNC): Delete.
14741 (vst4_lane_f16): Use __builtin_memcpy to copy vector
14742 structure instead of constructing __builtin_aarch64_simd_xi
14743 one vector at a time.
14744 (vst4_lane_f32): Likewise.
14745 (vst4_lane_f64): Likewise.
14746 (vst4_lane_p8): Likewise.
14747 (vst4_lane_p16): Likewise.
14748 (vst4_lane_p64): Likewise.
14749 (vst4_lane_s8): Likewise.
14750 (vst4_lane_s16): Likewise.
14751 (vst4_lane_s32): Likewise.
14752 (vst4_lane_s64): Likewise.
14753 (vst4_lane_u8): Likewise.
14754 (vst4_lane_u16): Likewise.
14755 (vst4_lane_u32): Likewise.
14756 (vst4_lane_u64): Likewise.
14757 (vst4_lane_bf16): Likewise.
14758 (vst4q_lane_f16): Use __builtin_memcpy to copy vector
14759 structure instead of using a union.
14760 (vst4q_lane_f32): Likewise.
14761 (vst4q_lane_f64): Likewise.
14762 (vst4q_lane_p8): Likewise.
14763 (vst4q_lane_p16): Likewise.
14764 (vst4q_lane_p64): Likewise.
14765 (vst4q_lane_s8): Likewise.
14766 (vst4q_lane_s16): Likewise.
14767 (vst4q_lane_s32): Likewise.
14768 (vst4q_lane_s64): Likewise.
14769 (vst4q_lane_u8): Likewise.
14770 (vst4q_lane_u16): Likewise.
14771 (vst4q_lane_u32): Likewise.
14772 (vst4q_lane_u64): Likewise.
14773 (vst4q_lane_bf16): Likewise.
14775 2021-08-06 Martin Liska <mliska@suse.cz>
14777 * config/rs6000/rs6000.c (rs6000_option_override_internal): When
14778 a target option is restored, it can have
14779 rs6000_long_double_type_size set to FLOAT_PRECISION_TFmode
14780 and error should not be emitted.
14782 2021-08-06 Sebastian Huber <sebastian.huber@embedded-brains.de>
14784 * gcov-io.h (gcov_write): Declare.
14785 * gcov-io.c (gcov_write): New.
14786 (gcov_write_counter): Remove.
14787 (gcov_write_tag_length): Likewise.
14788 (gcov_write_summary): Replace gcov_write_tag_length() with calls to
14789 gcov_write_unsigned().
14790 * doc/invoke.texi (fprofile-info-section): Mention
14791 __gcov_info_to_gdca().
14793 2021-08-06 Martin Sebor <msebor@redhat.com>
14795 * dominance.c (prune_bbs_to_update_dominators): Adjust by-value vec
14796 arguments to by-reference.
14797 (iterate_fix_dominators): Same.
14798 * dominance.h (iterate_fix_dominators): Same.
14799 * ipa-prop.h: Call auto_vec::to_vec_legacy.
14800 * tree-data-ref.c (dump_data_dependence_relation): Adjust by-value vec
14801 arguments to by-reference.
14802 (debug_data_dependence_relation): Same.
14803 (dump_data_dependence_relations): Same.
14804 * tree-data-ref.h (debug_data_dependence_relation): Same.
14805 (dump_data_dependence_relations): Same.
14806 * tree-predcom.c (dump_chains): Same.
14807 (initialize_root_vars_lm): Same.
14808 (determine_unroll_factor): Same.
14809 (replace_phis_by_defined_names): Same.
14810 (insert_init_seqs): Same.
14811 (pcom_worker::tree_predictive_commoning_loop): Call
14812 auto_vec::to_vec_legacy.
14813 * tree-ssa-pre.c (insert_into_preds_of_block): Adjust by-value vec
14814 arguments to by-reference.
14815 * tree-ssa-threadbackward.c (populate_worklist): Same.
14816 (back_threader::resolve_def): Same.
14817 * tree-vect-data-refs.c (vect_check_nonzero_value): Same.
14818 (vect_enhance_data_refs_alignment): Same.
14819 (vect_check_lower_bound): Same.
14820 (vect_prune_runtime_alias_test_list): Same.
14821 (vect_permute_store_chain): Same.
14822 * tree-vect-slp-patterns.c (vect_normalize_conj_loc): Same.
14823 * tree-vect-stmts.c (vect_create_vectorized_demotion_stmts): Same.
14824 * tree-vectorizer.h (vect_permute_store_chain): Same.
14825 * vec.c (test_init): New function.
14826 (vec_c_tests): Call new function.
14827 * vec.h (vec): Declare ctors, dtor, and assignment.
14828 (auto_vec::vec_to_legacy): New function.
14829 (vec::copy): Adjust initialization.
14831 2021-08-05 H.J. Lu <hjl.tools@gmail.com>
14834 * config/i386/i386.c (ix86_can_inline_p): Ignore MASK_80387 if
14835 callee only uses GPRs.
14836 * config/i386/ia32intrin.h: Revert commit 5463cee2770.
14837 * config/i386/serializeintrin.h: Revert commit 71958f740f1.
14838 * config/i386/x86gprintrin.h: Add
14839 #pragma GCC target("general-regs-only") and #pragma GCC pop_options
14840 to disable non-GPR ISAs.
14842 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
14844 PR middle-end/101787
14845 * doc/md.texi (cond_ashl, cond_ashr, cond_lshr): Document.
14847 2021-08-05 Richard Sandiford <richard.sandiford@arm.com>
14849 * tree-vectorizer.h (vect_is_store_elt_extraction, vect_is_reduction)
14850 (vect_reduc_type, vect_embedded_comparison_type, vect_comparison_type)
14851 (vect_is_extending_load, vect_is_integer_truncation): New functions,
14852 moved from aarch64.c but given different names.
14853 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction)
14854 (aarch64_is_reduction, aarch64_reduc_type)
14855 (aarch64_embedded_comparison_type, aarch64_comparison_type)
14856 (aarch64_extending_load_p, aarch64_integer_truncation_p): Delete
14857 in favor of the above. Update callers accordingly.
14859 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
14862 * config/arm/arm-cpus.in (generic-armv7-a): Add quirk to suppress
14863 writing .cpu directive in asm output.
14864 * config/arm/arm.c (arm_identify_fpu_from_isa): New variable.
14865 (arm_last_printed_arch_string): Delete.
14866 (arm_last-printed_fpu_string): Delete.
14867 (arm_configure_build_target): If use of floating-point/SIMD is
14868 disabled, remove all fp/simd related features from the target ISA.
14869 (last_arm_targ_options): New variable.
14870 (arm_print_asm_arch_directives): Add new parameters. Change order
14871 of emitted directives and handle all cases here.
14872 (arm_file_start): Always call arm_print_asm_arch_directives, move
14873 all generation of .arch/.arch_extension here.
14874 (arm_file_end): Call arm_print_asm_arch.
14875 (arm_declare_function_name): Call arm_print_asm_arch_directives
14876 instead of printing .arch/.fpu directives directly.
14878 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
14880 * config/arm/arm.c (arm_configure_build_target): Don't call
14881 arm_option_reconfigure_globals.
14882 (arm_option_restore): Call arm_option_reconfigure_globals after
14883 reconfiguring the target.
14884 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
14886 2021-08-05 Richard Earnshaw <rearnsha@arm.com>
14888 * config/arm/arm.c (arm_configure_build_target): Ensure the target's
14889 arch_name is always set.
14891 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
14893 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
14894 of vec_select high-half from being added into Neon subtract
14897 2021-08-05 Jonathan Wright <jonathan.wright@arm.com>
14899 * config/aarch64/aarch64.c: Traverse RTL tree to prevent cost
14900 of vec_select high-half from being added into Neon add cost.
14902 2021-08-05 Kewen Lin <linkw@linux.ibm.com>
14904 * cfgloop.h (loops_list::loops_list): Add one optional argument
14905 root and adjust accordingly, update loop tree walking and factor
14907 * cfgloop.c (loops_list::walk_loop_tree): ... this. New function.
14909 2021-08-05 Eric Botcazou <ebotcazou@gcc.gnu.org>
14911 PR tree-optimization/101626
14912 * tree-sra.c (propagate_subaccesses_from_rhs): Do not set the
14913 reverse scalar storage order on a pointer or vector component.
14915 2021-08-05 liuhongt <hongtao.liu@intel.com>
14917 * config/i386/sse.md (cond_<code><mode>): New expander.
14919 2021-08-05 liuhongt <hongtao.liu@intel.com>
14921 * config/i386/sse.md (cond_<code><mode>): New expander.
14923 2021-08-05 liuhongt <hongtao.liu@intel.com>
14925 * config/i386/sse.md (cond_<code><mode>): New expander.
14927 2021-08-04 David Malcolm <dmalcolm@redhat.com>
14930 * Makefile.in (ANALYZER_OBJS): Add analyzer/region-model-asm.o.
14932 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
14935 * config/i386/i386.h (STORE_MAX_PIECES): Allow 16/32/64 bytes
14936 only if TARGET_INTER_UNIT_MOVES_TO_VEC is true.
14938 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
14941 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
14942 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
14943 data with SSE register from one memory location to another.
14945 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
14947 * config/s390/s390.c (expand_perm_with_vpdi): New function.
14948 (vectorize_vec_perm_const_1): Call expand_perm_with_vpdi.
14949 * config/s390/vector.md (*vpdi1<mode>, @vpdi1<mode>): Enable a
14950 parameterized expander.
14951 (*vpdi4<mode>, @vpdi4<mode>): Likewise.
14953 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
14955 * config/s390/s390.c (MAX_VECT_LEN): Define macro.
14956 (struct expand_vec_perm_d): Define struct.
14957 (expand_perm_with_merge): New function.
14958 (vectorize_vec_perm_const_1): New function.
14959 (s390_vectorize_vec_perm_const): New function.
14960 (TARGET_VECTORIZE_VEC_PERM_CONST): Define target macro.
14962 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
14964 * config/s390/vector.md (V_HW_64): Remove mode iterator.
14965 (*vec_load_pair<mode>): Use V_HW_2 instead of V_HW_64.
14966 * config/s390/vx-builtins.md
14967 (vec_scatter_element<V_HW_2:mode>_SI): Use V_HW_2 instead of
14970 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
14972 * config/s390/s390.md (UNSPEC_VEC_PERMI): Remove constant
14974 * config/s390/vector.md (*vpdi1<mode>, *vpdi4<mode>): New pattern
14976 * config/s390/vx-builtins.md (*vec_permi<mode>): Emit generic rtx
14977 instead of an unspec.
14979 2021-08-04 Andreas Krebbel <krebbel@linux.ibm.com>
14981 * config/s390/s390-modes.def: Add more vector modes to support
14982 concatenation of two vectors.
14983 * config/s390/s390-protos.h (s390_expand_merge_perm_const): Add
14985 (s390_expand_merge): Likewise.
14986 * config/s390/s390.c (s390_expand_merge_perm_const): New function.
14987 (s390_expand_merge): New function.
14988 * config/s390/s390.md (UNSPEC_VEC_MERGEH, UNSPEC_VEC_MERGEL):
14989 Remove constant definitions.
14990 * config/s390/vector.md (V_HW_2): Add mode iterators.
14991 (VI_HW_4, V_HW_4): Rename VI_HW_4 to V_HW_4.
14992 (vec_2x_nelts, vec_2x_wide): New mode attributes.
14993 (*vmrhb, *vmrlb, *vmrhh, *vmrlh, *vmrhf, *vmrlf, *vmrhg, *vmrlg):
14994 New pattern definitions.
14995 (vec_widen_umult_lo_<mode>, vec_widen_umult_hi_<mode>)
14996 (vec_widen_smult_lo_<mode>, vec_widen_smult_hi_<mode>)
14997 (vec_unpacks_lo_v4sf, vec_unpacks_hi_v4sf, vec_unpacks_lo_v2df)
14998 (vec_unpacks_hi_v2df): Adjust expanders to emit non-unspec RTX for
15000 * config/s390/vx-builtins.md (V_HW_4): Remove mode iterator. Now
15002 (vec_mergeh<mode>, vec_mergel<mode>): Use s390_expand_merge to
15003 emit vec merge pattern.
15005 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
15007 * config/aarch64/aarch64.c (aarch64_strip_extend_vec_half):
15009 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent cost of
15010 vec_select high-half from being added into Neon multiply
15012 * rtlanal.c (vec_series_highpart_p): Define.
15013 * rtlanal.h (vec_series_highpart_p): Declare.
15015 2021-08-04 Jonathan Wright <jonathan.wright@arm.com>
15017 * config/aarch64/aarch64.c (aarch64_strip_duplicate_vec_elt):
15019 (aarch64_rtx_mult_cost): Traverse RTL tree to prevent
15020 vec_select cost from being added into Neon multiply cost.
15022 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
15024 * tree-vect-loop.c (vect_better_loop_vinfo_p): Detect cases in
15025 which old_loop_vinfo is an epilogue loop that handles a constant
15026 number of iterations.
15028 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
15030 * tree-vect-loop.c (vect_analyze_loop): Print a dump message
15031 when a reanalyzed loop fails to be cheaper than the current
15034 2021-08-04 Richard Sandiford <richard.sandiford@arm.com>
15036 * config/aarch64/aarch64.c: Fix a typo.
15038 2021-08-04 Vincent Lefèvre <vincent-gcc@vinc17.net>
15040 PR gcov-profile/101773
15041 * gcov-io.c (gcov_close): Check return code of a fclose.
15043 2021-08-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
15046 * dwarf2out.c (dwarf2out_assembly_start): Emit a dummy
15047 .file statement when needed.
15049 2021-08-04 Richard Biener <rguenther@suse.de>
15051 * tree-vect-data-refs.c (vect_check_gather_scatter):
15052 Include widening conversions only when the result is
15053 still handed by native gather or the current offset
15054 size not already matches the data size.
15055 Also succeed analysis in case there's no native support,
15056 noted by a IFN_LAST ifn and a NULL decl.
15057 (vect_analyze_data_refs): Always consider gathers.
15058 * tree-vect-patterns.c (vect_recog_gather_scatter_pattern):
15059 Test for no IFN gather rather than decl gather.
15060 * tree-vect-stmts.c (vect_model_load_cost): Pass in the
15061 gather-scatter info and cost emulated gathers accordingly.
15062 (vect_truncate_gather_scatter_offset): Properly test for
15064 (vect_use_strided_gather_scatters_p): Likewise.
15065 (get_load_store_type): Handle emulated gathers and its
15067 (vectorizable_load): Likewise. Emulate them by extracting
15068 scalar offsets, doing scalar loads and a vector construct.
15070 2021-08-04 H.J. Lu <hjl.tools@gmail.com>
15073 * expr.c (op_by_pieces_d::op_by_pieces_d): Add a max_pieces
15074 argument to set m_max_size.
15075 (move_by_pieces_d): Pass MOVE_MAX_PIECES to op_by_pieces_d.
15076 (store_by_pieces_d): Pass STORE_MAX_PIECES to op_by_pieces_d.
15077 (compare_by_pieces_d): Pass COMPARE_MAX_PIECES to op_by_pieces_d.
15079 2021-08-04 Roger Sayle <roger@nextmovesoftware.com>
15080 Marc Glisse <marc.glisse@inria.fr>
15082 * match.pd (bit_ior, bit_xor): Canonicalize (X*C1)|(X*C2) and
15083 (X*C1)^(X*C2) as X*(C1+C2), and related variants, using
15084 tree_nonzero_bits to ensure that operands are bit-wise disjoint.
15086 2021-08-04 Richard Biener <rguenther@suse.de>
15088 * tree-ssa-forwprop.c (pass_forwprop::execute): Split
15089 out code to decompose vector loads ...
15090 (optimize_vector_load): ... here. Generalize it to
15091 handle intermediate widening and TARGET_MEM_REF loads
15092 and apply it to loads with a supported vector mode as well.
15094 2021-08-04 Richard Biener <rguenther@suse.de>
15096 PR tree-optimization/101756
15097 * tree-vect-slp.c (vectorizable_bb_reduc_epilogue): Make sure
15098 the result of the reduction epilogue is compatible to the original
15101 2021-08-04 liuhongt <hongtao.liu@intel.com>
15104 * config/i386/i386.md (peephole2): Refine predicate from
15105 register_operand to general_reg_operand.
15107 2021-08-04 Aldy Hernandez <aldyh@redhat.com>
15109 * gimple-range-path.h (path_range_query::dump): Mark override.
15111 2021-08-04 Richard Biener <rguenther@suse.de>
15113 PR tree-optimization/101769
15114 * tree-tailcall.c (eliminate_tail_call): Add the created loop
15115 for the first recursion and return it via the new output parameter.
15116 (optimize_tail_call): Pass through new output param.
15117 (tree_optimize_tail_calls_1): After creating all latches,
15118 add the created loop to the loop tree. Do not mark loops for fixup.
15120 2021-08-04 Martin Liska <mliska@suse.cz>
15122 * doc/invoke.texi: Document threader-mode param.
15124 2021-08-04 liuhongt <hongtao.liu@intel.com>
15126 * config/i386/sse.md (cond_fma<mode>): New expander.
15127 (cond_fms<mode>): Ditto.
15128 (cond_fnma<mode>): Ditto.
15129 (cond_fnms<mode>): Ditto.
15131 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
15133 * config/rs6000/vsx.md (*vsx_le_perm_store_<mode>): Use && instead of &.
15135 2021-08-03 Segher Boessenkool <segher@kernel.crashing.org>
15137 * config/rs6000/constraints.md: Remove "e" from the list of available
15138 constraint characters.
15140 2021-08-03 Eugene Rozenfeld <erozen@microsoft.com>
15142 PR gcov-profile/71672
15143 * auto-profile.c (afdo_indirect_call): Fix setup of the historgram value for indirect calls.
15145 2021-08-03 Paul A. Clarke <pc@us.ibm.com>
15147 * config/rs6000/smmintrin.h (_mm_minpos_epu16): New.
15149 2021-08-03 H.J. Lu <hjl.tools@gmail.com>
15151 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): In 64-bit mode,
15152 try XMM31 to avoid vzeroupper.
15154 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
15156 * doc/invoke.texi: Document -mtune=neoverse-512tvb and
15157 -mcpu=neoverse-512tvb.
15158 * config/aarch64/aarch64-cores.def (neoverse-512tvb): New entry.
15159 * config/aarch64/aarch64-tune.md: Regenerate.
15160 * config/aarch64/aarch64.c (neoverse512tvb_sve_vector_cost)
15161 (neoverse512tvb_sve_issue_info, neoverse512tvb_vec_issue_info)
15162 (neoverse512tvb_vector_cost, neoverse512tvb_tunings): New structures.
15163 (aarch64_adjust_body_cost_sve): Handle -mtune=neoverse-512tvb.
15164 (aarch64_adjust_body_cost): Likewise.
15166 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
15168 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Only
15169 record issue information for operations that occur in the
15172 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
15174 * config/aarch64/aarch64.c (aarch64_multiply_add_p): Add a vec_flags
15175 parameter. Detect cases in which an Advanced SIMD MLA would almost
15176 certainly require a MOV.
15177 (aarch64_count_ops): Update accordingly.
15179 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
15181 * config/aarch64/aarch64.c (aarch64_is_store_elt_extraction): New
15182 function, split out from...
15183 (aarch64_detect_vector_stmt_subtype): ...here.
15184 (aarch64_add_stmt_cost): Treat extracting element 0 as free.
15186 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
15188 * config/aarch64/aarch64-protos.h (sve_vec_cost):
15189 Add gather_load_x32_cost and gather_load_x64_cost.
15190 * config/aarch64/aarch64.c (generic_sve_vector_cost)
15191 (a64fx_sve_vector_cost, neoversev1_sve_vector_cost): Update
15192 accordingly, using the values given by the scalar_load * number
15193 of elements calculation that we used previously.
15194 (aarch64_detect_vector_stmt_subtype): Use the new fields.
15196 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
15198 * config/aarch64/aarch64.c (aarch64_adjust_body_cost_sve): New
15199 function, split out from...
15200 (aarch64_adjust_body_cost): ...here.
15202 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
15204 * config/aarch64/fractional-cost.h: New file.
15205 * config/aarch64/aarch64.c: Include <algorithm> (indirectly)
15206 and cost_fraction.h.
15207 (vec_cost_fraction): New typedef.
15208 (aarch64_detect_scalar_stmt_subtype): Use it for statement costs.
15209 (aarch64_detect_vector_stmt_subtype): Likewise.
15210 (aarch64_sve_adjust_stmt_cost, aarch64_adjust_stmt_cost): Likewise.
15211 (aarch64_estimate_min_cycles_per_iter): Use vec_cost_fraction
15213 (aarch64_adjust_body_cost): Likewise.
15214 (aarch64_test_cost_fraction): New function.
15215 (aarch64_run_selftests): Call it.
15217 2021-08-03 Richard Sandiford <richard.sandiford@arm.com>
15219 * config/aarch64/aarch64-protos.h (tune_params::sve_width): Turn
15221 * config/aarch64/aarch64.c (aarch64_cmp_autovec_modes): Update
15223 (aarch64_estimated_poly_value): Likewise. Use the least significant
15224 set bit for the minimum and likely values. Use the most significant
15225 set bit for the maximum value.
15227 2021-08-03 liuhongt <hongtao.liu@intel.com>
15229 * config/i386/sse.md (cond_<insn><mode>): New expander.
15230 (cond_mul<mode>): Ditto.
15232 2021-08-03 Kewen Lin <linkw@linux.ibm.com>
15234 * tree-cfg.c (move_sese_region_to_fn): Fix typos on dloop.
15236 2021-08-03 liuhongt <hongtao.liu@intel.com>
15238 * config/i386/sse.md (cond_<insn><mode>):New expander.
15239 (cond_mul<mode>): Ditto.
15240 (cond_div<mode>): Ditto.
15242 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
15244 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Also
15245 check stack_realign_needed for stack realignment.
15246 (ix86_legitimate_constant_p): Always allow CONST_WIDE_INT smaller
15247 than the largest integer supported by vector register.
15248 * config/i386/i386.h (MAX_MOVE_MAX): New. Set to 64.
15249 (MOVE_MAX): Set to bytes of the largest integer supported by
15251 (STORE_MAX_PIECES): New.
15253 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
15255 * config/i386/i386-expand.c (ix86_expand_vector_move): Call
15256 ix86_gen_scratch_sse_rtx to get a scratch SSE register to copy
15257 data from one memory location to another.
15259 2021-08-02 H.J. Lu <hjl.tools@gmail.com>
15261 PR middle-end/90773
15262 * config/i386/i386.c (TARGET_GEN_MEMSET_SCRATCH_RTX): New.
15264 2021-08-02 Aldy Hernandez <aldyh@redhat.com>
15266 PR tree-optimization/101724
15267 * params.opt: Remove --param=threader-iterative.
15268 * tree-ssa-threadbackward.c (pass_thread_jumps::execute): Remove
15271 2021-08-02 Tom de Vries <tdevries@suse.de>
15273 PR middle-end/101665
15274 * doc/extend.texi (nonnull attribute): Improve documentation.
15276 2021-08-02 Andrew Pinski <apinski@marvell.com>
15278 PR rtl-optimization/101683
15279 * rtlanal.c (may_trap_p_1): Handle UNSIGNED_FIX.
15281 2021-08-02 Roger Sayle <roger@nextmovesoftware.com>
15283 * tree-ssa-phiopt.c (cond_removal_in_builtin_zero_pattern):
15284 Renamed from cond_removal_in_popcount_clz_ctz_pattern.
15285 Add support for BSWAP, FFS, PARITY and CLRSB builtins.
15286 (tree_ssa_phiop_worker): Update call to function above.
15288 2021-08-01 H.J. Lu <hjl.tools@gmail.com>
15291 * config/i386/i386.md (bsr_rex64_1_zext): New.
15292 (combine splitter for constant - clzll): Replace gen_bsr_rex64_1
15293 with gen_bsr_rex64_1_zext.
15295 2021-07-31 Jakub Jelinek <jakub@redhat.com>
15298 * config/i386/i386.md (bsr_rex64_1, bsr_1, bsr_zext_1): New
15299 define_insn patterns.
15300 (*bsr_rex64_2, *bsr_2): New define_insn_and_split patterns.
15301 Add combine splitters for constant - clz.
15302 (clz<mode>2): Use a temporary pseudo for bsr result.
15304 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
15306 * config/rs6000/smmintrin.h (_mm_floor_pd, _mm_floor_ps,
15307 _mm_floor_sd, _mm_floor_ss): New.
15309 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
15311 * config/rs6000/smmintrin.h (_mm_ceil_pd, _mm_ceil_ps,
15312 _mm_ceil_sd, _mm_ceil_ss): New.
15314 2021-07-30 Paul A. Clarke <pc@us.ibm.com>
15316 * config/rs6000/smmintrin.h (_mm_blend_pd, _mm_blendv_pd,
15317 _mm_blend_ps, _mm_blendv_ps): New.
15319 2021-07-30 Roger Sayle <roger@nextmovesoftware.com>
15320 Uroš Bizjak <ubizjak@gmail.com>
15322 * config/i386/i386.md (*dec_cmov<mode>): New define_insn_and_split
15323 to generate a conditional move using the carry flag after sub $1.
15324 (peephole2): Eliminate a register-to-register move by inverting
15325 the condition of a conditional move.
15327 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
15329 * config/mmix/mmix.md ("call", "call_value", "*call_real")
15330 ("*call_value_real"): Don't generate rtx mentioning the generic
15331 operands 1 and 2 to "call", and similarly for "call_value".
15332 * config/mmix/mmix.c (mmix_print_operand_punct_valid_p)
15333 (mmix_print_operand): Use '!' instead of 'p'.
15335 2021-07-30 Hans-Peter Nilsson <hp@bitrange.com>
15337 * doc/md.texi (call): Correct information about operand 2.
15338 * config/mmix/mmix.md ("call", "call_value"): Remove fixed FIXMEs.
15340 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
15342 * range-op.cc (operator_trunc_mod::wi_fold): Fold constants.
15344 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
15346 * range-op.cc (operator_div::wi_fold): Return UNDEFINED for [0, 0] divisor.
15348 2021-07-30 Andrew MacLeod <amacleod@redhat.com>
15350 * gimple-range-cache.cc (*::set_bb_range): Change const basic_block to
15351 const_basic_block..
15352 (*::get_bb_range): Ditto.
15353 (*::bb_range_p): Ditto.
15354 * gimple-range-cache.h: Change prototypes.
15356 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
15358 PR middle-end/90773
15359 * builtins.c (builtin_memcpy_read_str): Change the mode argument
15360 from scalar_int_mode to fixed_size_mode.
15361 (builtin_strncpy_read_str): Likewise.
15362 (gen_memset_value_from_prev): New function.
15363 (builtin_memset_read_str): Change the mode argument from
15364 scalar_int_mode to fixed_size_mode. Use gen_memset_value_from_prev
15365 and support CONST_VECTOR.
15366 (builtin_memset_gen_str): Likewise.
15367 (try_store_by_multiple_pieces): Use by_pieces_constfn to declare
15369 * builtins.h (builtin_strncpy_read_str): Replace scalar_int_mode
15370 with fixed_size_mode.
15371 (builtin_memset_read_str): Likewise.
15372 * expr.c (widest_int_mode_for_size): Renamed to ...
15373 (widest_fixed_size_mode_for_size): Add a bool argument to
15374 indicate if QI vector mode can be used.
15375 (by_pieces_ninsns): Call widest_fixed_size_mode_for_size
15376 instead of widest_int_mode_for_size.
15377 (pieces_addr::adjust): Change the mode argument from
15378 scalar_int_mode to fixed_size_mode.
15379 (op_by_pieces_d): Make m_len read-only. Add a bool member,
15380 m_qi_vector_mode, to indicate that QI vector mode can be used.
15381 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
15382 initialize m_qi_vector_mode. Call widest_fixed_size_mode_for_size
15383 instead of widest_int_mode_for_size.
15384 (op_by_pieces_d::get_usable_mode): Change the mode argument from
15385 scalar_int_mode to fixed_size_mode. Call
15386 widest_fixed_size_mode_for_size instead of
15387 widest_int_mode_for_size.
15388 (op_by_pieces_d::smallest_fixed_size_mode_for_size): New member
15389 function to return the smallest integer or QI vector mode.
15390 (op_by_pieces_d::run): Call widest_fixed_size_mode_for_size
15391 instead of widest_int_mode_for_size. Call
15392 smallest_fixed_size_mode_for_size instead of
15393 smallest_int_mode_for_size.
15394 (store_by_pieces_d::store_by_pieces_d): Add a bool argument to
15395 indicate that QI vector mode can be used and pass it to
15396 op_by_pieces_d::op_by_pieces_d.
15397 (can_store_by_pieces): Call widest_fixed_size_mode_for_size
15398 instead of widest_int_mode_for_size. Pass memsetp to
15399 widest_fixed_size_mode_for_size to support QI vector mode.
15400 Allow all CONST_VECTORs for memset if vec_duplicate is supported.
15401 (store_by_pieces): Pass memsetp to
15402 store_by_pieces_d::store_by_pieces_d.
15403 (clear_by_pieces_1): Removed.
15404 (clear_by_pieces): Replace clear_by_pieces_1 with
15405 builtin_memset_read_str and pass true to store_by_pieces_d to
15406 support vector mode broadcast.
15407 (string_cst_read_str): Change the mode argument from
15408 scalar_int_mode to fixed_size_mode.
15409 * expr.h (by_pieces_constfn): Change scalar_int_mode to
15411 (by_pieces_prev): Likewise.
15412 * rtl.h (lowpart_subreg_regno): New.
15413 * rtlanal.c (lowpart_subreg_regno): New. A wrapper around
15414 simplify_subreg_regno.
15415 * target.def (gen_memset_scratch_rtx): New hook.
15416 * doc/tm.texi.in: Add TARGET_GEN_MEMSET_SCRATCH_RTX.
15417 * doc/tm.texi: Regenerated.
15419 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
15422 * config/mips/mips.c (mips_atomic_assign_expand_fenv): Use
15423 TARGET_EXPR instead of MODIFY_EXPR.
15425 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
15428 * config/mips/mips-protos.h (mips_expand_vec_cmp_expr): Declare.
15429 * config/mips/mips.c (mips_expand_vec_cmp_expr): New function.
15430 * config/mips/mips-msa.md (vec_cmp<MSA:mode><mode_i>): New
15432 (vec_cmpu<IMSA:mode><mode_i>): New expander.
15434 2021-07-30 H.J. Lu <hjl.tools@gmail.com>
15437 * config/i386/i386-options.c (ix86_option_override_internal):
15438 Don't enable LZCNT/POPCNT if they have been disabled explicitly.
15440 2021-07-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
15443 * config/arm/arm_neon.h (vld1_p64): Replace call to builtin by
15444 explicitly dereferencing __a.
15445 (vld1_s64): Likewise.
15446 (vld1_u64): Likewise.
15447 * config/arm/arm_neon_builtins.def (vld1): Remove entry for di
15448 and change to VAR13.
15450 2021-07-30 Aldy Hernandez <aldyh@redhat.com>
15452 * gimple-loop-versioning.cc (lv_dom_walker::lv_dom_walker): Remove
15453 use of m_range_analyzer.
15454 (loop_versioning::lv_dom_walker::before_dom_children): Same.
15455 (loop_versioning::lv_dom_walker::after_dom_children): Remove.
15456 (loop_versioning::prune_loop_conditions): Replace vr_values use
15457 with range_query interface.
15458 (pass_loop_versioning::execute): Use ranger.
15460 2021-07-30 Xi Ruoyao <xry111@mengyan1223.wang>
15463 * ipa-devirt.c (ipa_odr_read_section): Compare the precision of
15464 enum values, and emit a warning if they mismatch.
15466 2021-07-30 Kewen Lin <linkw@linux.ibm.com>
15468 * cfgloop.h (as_const): New function.
15469 (class loop_iterator): Rename to ...
15470 (class loops_list): ... this.
15471 (loop_iterator::next): Rename to ...
15472 (loops_list::Iter::fill_curr_loop): ... this and adjust.
15473 (loop_iterator::loop_iterator): Rename to ...
15474 (loops_list::loops_list): ... this and adjust.
15475 (loops_list::Iter): New class.
15476 (loops_list::iterator): New type.
15477 (loops_list::const_iterator): New type.
15478 (loops_list::begin): New function.
15479 (loops_list::end): Likewise.
15480 (loops_list::begin const): Likewise.
15481 (loops_list::end const): Likewise.
15482 (FOR_EACH_LOOP): Remove.
15483 (FOR_EACH_LOOP_FN): Remove.
15484 * cfgloop.c (flow_loops_dump): Adjust FOR_EACH_LOOP* with range-based
15485 for loop with loops_list instance.
15486 (sort_sibling_loops): Likewise.
15487 (disambiguate_loops_with_multiple_latches): Likewise.
15488 (verify_loop_structure): Likewise.
15489 * cfgloopmanip.c (create_preheaders): Likewise.
15490 (force_single_succ_latches): Likewise.
15491 * config/aarch64/falkor-tag-collision-avoidance.c
15492 (execute_tag_collision_avoidance): Likewise.
15493 * config/mn10300/mn10300.c (mn10300_scan_for_setlb_lcc): Likewise.
15494 * config/s390/s390.c (s390_adjust_loops): Likewise.
15495 * doc/loop.texi: Likewise.
15496 * gimple-loop-interchange.cc (pass_linterchange::execute): Likewise.
15497 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
15498 * gimple-loop-versioning.cc (loop_versioning::analyze_blocks): Likewise.
15499 (loop_versioning::make_versioning_decisions): Likewise.
15500 * gimple-ssa-split-paths.c (split_paths): Likewise.
15501 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
15502 * graphite.c (canonicalize_loop_form): Likewise.
15503 (graphite_transform_loops): Likewise.
15504 * ipa-fnsummary.c (analyze_function_body): Likewise.
15505 * ipa-pure-const.c (analyze_function): Likewise.
15506 * loop-doloop.c (doloop_optimize_loops): Likewise.
15507 * loop-init.c (loop_optimizer_finalize): Likewise.
15508 (fix_loop_structure): Likewise.
15509 * loop-invariant.c (calculate_loop_reg_pressure): Likewise.
15510 (move_loop_invariants): Likewise.
15511 * loop-unroll.c (decide_unrolling): Likewise.
15512 (unroll_loops): Likewise.
15513 * modulo-sched.c (sms_schedule): Likewise.
15514 * predict.c (predict_loops): Likewise.
15515 (pass_profile::execute): Likewise.
15516 * profile.c (branch_prob): Likewise.
15517 * sel-sched-ir.c (sel_finish_pipelining): Likewise.
15518 (sel_find_rgns): Likewise.
15519 * tree-cfg.c (replace_loop_annotate): Likewise.
15520 (replace_uses_by): Likewise.
15521 (move_sese_region_to_fn): Likewise.
15522 * tree-if-conv.c (pass_if_conversion::execute): Likewise.
15523 * tree-loop-distribution.c (loop_distribution::execute): Likewise.
15524 * tree-parloops.c (parallelize_loops): Likewise.
15525 * tree-predcom.c (tree_predictive_commoning): Likewise.
15526 * tree-scalar-evolution.c (scev_initialize): Likewise.
15527 (scev_reset): Likewise.
15528 * tree-ssa-dce.c (find_obviously_necessary_stmts): Likewise.
15529 * tree-ssa-live.c (remove_unused_locals): Likewise.
15530 * tree-ssa-loop-ch.c (ch_base::copy_headers): Likewise.
15531 * tree-ssa-loop-im.c (analyze_memory_references): Likewise.
15532 (tree_ssa_lim_initialize): Likewise.
15533 * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Likewise.
15534 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize): Likewise.
15535 * tree-ssa-loop-manip.c (get_loops_exits): Likewise.
15536 * tree-ssa-loop-niter.c (estimate_numbers_of_iterations): Likewise.
15537 (free_numbers_of_iterations_estimates): Likewise.
15538 * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Likewise.
15539 * tree-ssa-loop-split.c (tree_ssa_split_loops): Likewise.
15540 * tree-ssa-loop-unswitch.c (tree_ssa_unswitch_loops): Likewise.
15541 * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
15542 (pass_scev_cprop::execute): Likewise.
15543 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
15544 * tree-ssa-sccvn.c (do_rpo_vn): Likewise.
15545 * tree-ssa-threadupdate.c
15546 (jump_thread_path_registry::thread_through_all_blocks): Likewise.
15547 * tree-vectorizer.c (vectorize_loops): Likewise.
15548 * tree-vrp.c (vrp_asserts::find_assert_locations): Likewise.
15550 2021-07-29 Hans-Peter Nilsson <hp@bitrange.com>
15552 * config/mmix/mmix.c (mmix_function_arg_1): Avoid
15553 generating a VOIDmode register for e.g the
15554 function_arg_info::end_marker.
15556 2021-07-29 Jeff Law <jeffreyalaw@gmail.com>
15558 * config/h8300/h8300-modes.def: Add CCZ, CCV and CCC, drop CCZNV.
15559 * config/h8300/h8300.md (H8cc mode iterator): Add CCZ.
15560 (cc mode_attr): Similarly.
15561 (ccz subst_attr): Similarly.
15562 * config/h8300/jumpcall.md: Add new patterns for branch-on-bit.
15563 * config/h8300/testcompare.md: Remove various cc0 based patterns
15564 that had been commented out. Add pattern to set CCZ from a bit
15567 2021-07-29 Thomas Schwinge <thomas@codesourcery.com>
15568 Julian Brown <julian@codesourcery.com>
15569 Kwok Cheung Yeung <kcy@codesourcery.com>
15571 * omp-offload.c (oacc_loop_xform_head_tail, oacc_loop_process):
15572 'update_stmt' after modification.
15573 (pass_oacc_loop_designation): New function, extracted out of...
15574 (pass_oacc_device_lower): ... this.
15575 (pass_data_oacc_loop_designation, pass_oacc_loop_designation)
15576 (make_pass_oacc_loop_designation): New
15577 * passes.def: Add it.
15578 * tree-parloops.c (create_parallel_loop): Adjust.
15579 * tree-pass.h (make_pass_oacc_loop_designation): New.
15581 2021-07-29 Aldy Hernandez <aldyh@redhat.com>
15583 * flag-types.h (enum threader_mode): New.
15584 * params.opt: Add entry for --param=threader-mode.
15585 * tree-ssa-threadbackward.c (THREADER_ITERATIVE_MODE): New.
15586 (class back_threader): New.
15587 (back_threader::back_threader): New.
15588 (back_threader::~back_threader): New.
15589 (back_threader::maybe_register_path): New.
15590 (back_threader::find_taken_edge): New.
15591 (back_threader::find_taken_edge_switch): New.
15592 (back_threader::find_taken_edge_cond): New.
15593 (back_threader::resolve_def): New.
15594 (back_threader::resolve_phi): New.
15595 (back_threader::find_paths_to_names): New.
15596 (back_threader::find_paths): New.
15599 (thread_jumps::find_jump_threads_backwards): Call ranger threader.
15600 (thread_jumps::find_jump_threads_backwards_with_ranger): New.
15601 (pass_thread_jumps::execute): Abstract out code...
15602 (try_thread_blocks): ...here.
15603 * tree-ssa-threadedge.c (jump_threader::thread_outgoing_edges):
15604 Abstract out threading candidate code to...
15605 (single_succ_to_potentially_threadable_block): ...here.
15606 * tree-ssa-threadedge.h (single_succ_to_potentially_threadable_block):
15608 * tree-ssa-threadupdate.c (register_jump_thread): Return boolean.
15609 * tree-ssa-threadupdate.h (class jump_thread_path_registry):
15610 Return bool from register_jump_thread.
15612 2021-07-29 Andreas Krebbel <krebbel@linux.ibm.com>
15614 * target.def: in0 and in1 do not need to be registers.
15615 * doc/tm.texi: Regenerate.
15617 2021-07-29 liuhongt <hongtao.liu@intel.com>
15620 * config/i386/i386.c (ix86_widen_mult_cost): New function.
15621 (ix86_add_stmt_cost): Use ix86_widen_mult_cost for
15624 2021-07-29 Jiufu Guo <guojiufu@linux.ibm.com>
15627 * config/rs6000/rs6000.c (TARGET_PREFERRED_DOLOOP_MODE): New hook.
15628 (rs6000_preferred_doloop_mode): New hook.
15629 * doc/tm.texi: Regenerate.
15630 * doc/tm.texi.in: Add hook preferred_doloop_mode.
15631 * target.def (preferred_doloop_mode): New hook.
15632 * targhooks.c (default_preferred_doloop_mode): New hook.
15633 * targhooks.h (default_preferred_doloop_mode): New hook.
15634 * tree-ssa-loop-ivopts.c (compute_doloop_base_on_mode): New function.
15635 (add_iv_candidate_for_doloop): Call targetm.preferred_doloop_mode
15636 and compute_doloop_base_on_mode.
15638 2021-07-28 Martin Sebor <msebor@redhat.com>
15640 PR middle-end/101494
15641 * tree-ssa-uninit.c (maybe_warn_operand): Correct object offset
15642 and size computation.
15644 2021-07-28 Martin Sebor <msebor@redhat.com>
15646 PR middle-end/101601
15647 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Remove
15649 Handle pointers to functions.
15651 2021-07-28 Martin Sebor <msebor@redhat.com>
15653 * Makefile.in (OBJS): Add gimple-ssa-warn-access.o and pointer-query.o.
15654 * attribs.h (fndecl_dealloc_argno): Move fndecl_dealloc_argno to tree.h.
15655 * builtins.c (compute_objsize_r): Move to pointer-query.cc.
15656 (access_ref::access_ref): Same.
15657 (access_ref::phi): Same.
15658 (access_ref::get_ref): Same.
15659 (access_ref::size_remaining): Same.
15660 (access_ref::offset_in_range): Same.
15661 (access_ref::add_offset): Same.
15662 (access_ref::inform_access): Same.
15663 (ssa_name_limit_t::visit_phi): Same.
15664 (ssa_name_limit_t::leave_phi): Same.
15665 (ssa_name_limit_t::next): Same.
15666 (ssa_name_limit_t::next_phi): Same.
15667 (ssa_name_limit_t::~ssa_name_limit_t): Same.
15668 (pointer_query::pointer_query): Same.
15669 (pointer_query::get_ref): Same.
15670 (pointer_query::put_ref): Same.
15671 (pointer_query::flush_cache): Same.
15672 (warn_string_no_nul): Move to gimple-ssa-warn-access.cc.
15673 (check_nul_terminated_array): Same.
15674 (unterminated_array): Same.
15675 (maybe_warn_for_bound): Same.
15676 (check_read_access): Same.
15677 (warn_for_access): Same.
15678 (get_size_range): Same.
15679 (check_access): Same.
15680 (gimple_call_alloc_size): Move to tree.c.
15681 (gimple_parm_array_size): Move to pointer-query.cc.
15682 (get_offset_range): Same.
15683 (gimple_call_return_array): Same.
15684 (handle_min_max_size): Same.
15685 (handle_array_ref): Same.
15686 (handle_mem_ref): Same.
15687 (compute_objsize): Same.
15688 (gimple_call_alloc_p): Move to gimple-ssa-warn-access.cc.
15689 (call_dealloc_argno): Same.
15690 (fndecl_dealloc_argno): Same.
15691 (new_delete_mismatch_p): Same.
15692 (matching_alloc_calls_p): Same.
15693 (warn_dealloc_offset): Same.
15694 (maybe_emit_free_warning): Same.
15695 * builtins.h (check_nul_terminated_array): Move to
15696 gimple-ssa-warn-access.h.
15697 (check_nul_terminated_array): Same.
15698 (warn_string_no_nul): Same.
15699 (unterminated_array): Same.
15700 (class ssa_name_limit_t): Same.
15701 (class pointer_query): Same.
15702 (struct access_ref): Same.
15703 (class range_query): Same.
15704 (struct access_data): Same.
15705 (gimple_call_alloc_size): Same.
15706 (gimple_parm_array_size): Same.
15707 (compute_objsize): Same.
15708 (class access_data): Same.
15709 (maybe_emit_free_warning): Same.
15710 * calls.c (initialize_argument_information): Remove call to
15711 maybe_emit_free_warning.
15712 * gimple-array-bounds.cc: Include new header..
15713 * gimple-fold.c: Same.
15714 * gimple-ssa-sprintf.c: Same.
15715 * gimple-ssa-warn-restrict.c: Same.
15716 * passes.def: Add pass_warn_access.
15717 * tree-pass.h (make_pass_warn_access): Declare.
15718 * tree-ssa-strlen.c: Include new headers.
15719 * tree.c (fndecl_dealloc_argno): Move here from builtins.c.
15720 * tree.h (fndecl_dealloc_argno): Move here from attribs.h.
15721 * gimple-ssa-warn-access.cc: New file.
15722 * gimple-ssa-warn-access.h: New file.
15723 * pointer-query.cc: New file.
15724 * pointer-query.h: New file.
15726 2021-07-28 Jakub Jelinek <jakub@redhat.com>
15728 PR middle-end/101624
15729 * ubsan.c (maybe_instrument_pointer_overflow,
15730 instrument_object_size): Only test DECL_REGISTER on VAR_DECLs,
15731 PARM_DECLs or RESULT_DECLs.
15732 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
15734 2021-07-28 Jakub Jelinek <jakub@redhat.com>
15736 PR middle-end/101642
15737 * match.pd (bswap16 (x) == bswap16 (y)): Cast both operands
15738 to type of bswap16 for comparison.
15739 (bswap16 (x) == cst): Cast bswap16 operand to type of cst.
15741 2021-07-28 Richard Biener <rguenther@suse.de>
15743 PR tree-optimization/101615
15744 * tree-vect-slp.c (vect_optimize_slp): Materialize permutes
15745 at CTOR SLP graph entries.
15747 2021-07-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15749 * config/aarch64/aarch64.md (*extend<SHORT:mode><GPI:mode>2_aarch64):
15750 Add "r,w" alternative.
15752 2021-07-28 H.J. Lu <hjl.tools@gmail.com>
15755 * config/i386/i386.c (ix86_avx_u128_mode_needed): Don't set
15756 AVX_U128_DIRTY when all bits are zero.
15758 2021-07-28 Richard Biener <rguenther@suse.de>
15760 PR tree-optimization/101615
15761 * tree-vect-slp.c (vect_optimize_slp): Pre-existing vector
15762 external nodes cannot be permuted so make them perm_out 0.
15764 2021-07-28 Andrew Stubbs <ams@codesourcery.com>
15767 * config.in: Regenerate.
15768 * config/gcn/gcn-hsa.h (A_FIJI): New define.
15769 (A_900): New define.
15770 (A_906): New define.
15771 (A_908): New define.
15772 (ASM_SPEC): Use A_FIJI, A_900, A_906 and A_908.
15773 * config/gcn/gcn.c (output_file_start): Adjust attributes according
15774 to the assembler capabilities.
15775 * config/gcn/mkoffload.c (main): Likewise.
15776 * configure: Regenerate.
15777 * configure.ac: Add tests for LLVM assembler attribute features.
15779 2021-07-28 Andrew MacLeod <amacleod@redhat.com>
15781 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Check for
15782 cond_false and cond_true on branches.
15784 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
15786 * config/aarch64/aarch64.c (aarch64_gen_adjusted_ldpstp): use
15789 2021-07-28 Bin Cheng <bin.cheng@linux.alibaba.com>
15791 * alias.c (init_alias_analysis): Don't skip prologue/epilogue.
15793 2021-07-28 Jakub Jelinek <jakub@redhat.com>
15796 * config/i386/sse.md (vashr<mode>3): Split into vashrv8di3 expander
15797 and vashrv4di3 expander, where the latter requires just TARGET_AVX2
15798 and has special !TARGET_AVX512VL expansion.
15799 (vashrv2di3<mask_name>): Rename to ...
15800 (vashrv2di3): ... this. Change condition to TARGET_XOP || TARGET_AVX2
15801 and add special !TARGET_XOP && !TARGET_AVX512VL expansion.
15803 2021-07-28 Martin Uecker <muecker@gwdg.de>
15805 * calls.c (maybe_warn_rdwr_sizes): Correct argument
15806 numbers in warning that were switched.
15808 2021-07-28 Kewen Lin <linkw@linux.ibm.com>
15810 PR tree-optimization/101596
15811 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Fix wrong check
15812 by using new_type's precision instead.
15814 2021-07-28 liuhongt <hongtao.liu@intel.com>
15817 * config/i386/i386.h (processor_costs): Add new member
15819 * config/i386/x86-tune-costs.h (ix86_size_cost, i386_cost,
15820 i486_cost, pentium_cost, lakemont_cost, pentiumpro_cost,
15821 geode_cost, k6_cost, athlon_cost, k8_cost, amdfam10_cost,
15822 bdver_cost, znver1_cost, znver2_cost, znver3_cost,
15823 btver1_cost, btver2_cost, btver3_cost, pentium4_cost,
15824 nocona_cost, atom_cost, atom_cost, slm_cost, intel_cost,
15825 generic_cost, core_cost): Initialize integer_to_sse same value
15827 (skylake_cost): Initialize integer_to_sse twice as much as sse_op.
15828 * config/i386/i386.c (ix86_builtin_vectorization_cost):
15829 Use integer_to_sse instead of sse_op to calculate the cost of
15832 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
15834 * config/rs6000/rs6000-gen-builtins.c (write_ovld_static_init): New
15836 (write_init_file): Call write_ovld_static_init.
15838 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
15840 * config/rs6000/rs6000-gen-builtins.c (write_bif_static_init): New
15842 (write_init_file): Call write_bif_static_init.
15844 2021-07-27 Bill Schmidt <wschmidt@linux.ibm.com>
15846 * config/rs6000/rs6000-gen-builtins.c (typemap): New struct.
15847 (TYPE_MAP_SIZE): New macro.
15848 (type_map): New initialized variable.
15849 (typemap_cmp): New function.
15850 (write_type_node): Likewise.
15851 (write_fntype_init): Implement.
15853 2021-07-27 Martin Sebor <msebor@redhat.com>
15855 PR tree-optimization/101584
15856 * tree-ssa-uninit.c (builtin_call_nomodifying_p): New function.
15857 (check_defs): Call it.
15859 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
15861 * tree-ssa-dom.c (dom_jump_threader_simplifier):
15862 Put avail_exprs_stack in the class, instead of passing it to
15863 jump_threader_simplifier.
15864 (dom_jump_threader_simplifier::simplify): Add state argument.
15865 (dom_opt_dom_walker): Add state.
15866 (pass_dominator::execute): Pass state to threader.
15867 (dom_opt_dom_walker::before_dom_children): Use state.
15868 * tree-ssa-threadedge.c (jump_threader::jump_threader): Replace
15869 arguments by state.
15870 (jump_threader::record_temporary_equivalences_from_phis):
15871 Register equivalences through the state variable.
15872 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
15873 Record ranges in a statement through the state variable.
15874 (jump_threader::simplify_control_stmt_condition): Pass state to
15876 (jump_threader::simplify_control_stmt_condition_1): Same.
15877 (jump_threader::thread_around_empty_blocks): Remove obsolete
15879 (jump_threader::thread_through_normal_block): Record equivalences
15880 on edge through the state variable.
15881 (jump_threader::thread_across_edge): Abstract state pushing.
15882 (jt_state::jt_state): New.
15883 (jt_state::push): New.
15884 (jt_state::pop): New.
15885 (jt_state::register_equiv): New.
15886 (jt_state::record_ranges_from_stmt): New.
15887 (jt_state::register_equivs_on_edge): New.
15888 (jump_threader_simplifier::jump_threader_simplifier): Move from
15890 (jump_threader_simplifier::simplify): Add state argument.
15891 * tree-ssa-threadedge.h (class jt_state): New.
15892 (class jump_threader): Add state to constructor.
15893 (class jump_threader_simplifier): Add state to simplify. Remove
15894 avail_exprs_stack from class.
15895 * tree-vrp.c (vrp_jump_threader_simplifier::simplify): Add state
15897 (vrp_jump_threader::vrp_jump_threader): Add state.
15898 (vrp_jump_threader::~vrp_jump_threader): Cleanup state.
15900 2021-07-27 Aldy Hernandez <aldyh@redhat.com>
15902 * Makefile.in (OBJS): Add gimple-range-path.o.
15903 * gimple-range-path.cc: New file.
15904 * gimple-range-path.h: New file.
15906 2021-07-27 Jonathan Wright <jonathan.wright@arm.com>
15908 * config/aarch64/aarch64-simd.md: Push sign/zero-extension
15909 inside vec_duplicate for all patterns.
15910 * simplify-rtx.c (simplify_context::simplify_unary_operation_1):
15911 Push sign/zero-extension inside vec_duplicate.
15913 2021-07-27 Richard Biener <rguenther@suse.de>
15915 PR tree-optimization/101573
15916 * tree-ssa-uninit.c (warn_uninit_phi_uses): New function
15917 looking at uninitialized PHI arg defs in some constrained cases.
15918 (warn_uninitialized_vars): Call it.
15919 (execute_early_warn_uninitialized): Calculate dominators.
15921 2021-07-27 Richard Biener <rguenther@suse.de>
15923 PR tree-optimization/39821
15924 * tree-vect-stmts.c (vect_model_promotion_demotion_cost): Use
15925 vector_stmt for widening arithmetic.
15926 (vectorizable_conversion): Adjust.
15928 2021-07-27 Martin Jambor <mjambor@suse.cz>
15930 * cgraph.h (ipa_replace_map): New field force_load_ref.
15931 * ipa-prop.h (ipa_param_descriptor): Reduce precision of move_cost,
15932 aded new flag load_dereferenced, adjusted comments.
15933 (ipa_get_param_dereferenced): New function.
15934 (ipa_set_param_dereferenced): Likewise.
15935 * cgraphclones.c (cgraph_node::create_virtual_clone): Follow it.
15936 * ipa-cp.c: Include gimple.h.
15937 (ipcp_discover_new_direct_edges): Take into account dereferenced flag.
15938 (get_replacement_map): New parameter force_load_ref, set the
15939 appropriate flag in ipa_replace_map if set.
15940 (struct symbol_and_index_together): New type.
15941 (adjust_refs_in_act_callers): New function.
15942 (adjust_references_in_caller): Likewise.
15943 (create_specialized_node): When appropriate, call
15944 adjust_references_in_caller and force only load references.
15945 * ipa-prop.c (load_from_dereferenced_name): New function.
15946 (ipa_analyze_controlled_uses): Also detect loads from a
15947 dereference, harden testing of call statements.
15948 (ipa_write_node_info): Stream the dereferenced flag.
15949 (ipa_read_node_info): Likewise.
15950 (ipa_set_jf_constant): Also create refdesc when jump function
15951 references a variable.
15952 (cgraph_node_for_jfunc): Rename to symtab_node_for_jfunc, work
15953 also on references of variables and return a symtab_node. Adjust
15955 (propagate_controlled_uses): Also remove references to VAR_DECLs.
15957 2021-07-27 Jakub Jelinek <jakub@redhat.com>
15959 PR middle-end/101586
15960 * gimple-fold.c (clear_padding_type): Ignore FIELD_DECLs with byte
15961 positions above or equal to sz except for diagnostics of flexible
15964 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
15966 PR tree-optimization/78888
15967 * gimple-range-fold.cc (get_letter_range): New.
15968 (fold_using_range::range_of_builtin_call): Call get_letter_range.
15970 2021-07-26 Andrew MacLeod <amacleod@redhat.com>
15972 PR tree-optimization/78888
15973 * gimple-range-fold.cc (fold_using_range::range_of_builtin_call): Add cases
15974 for CFN_BUILT_IN_TOUPPER and CFN_BUILT_IN_TOLOWER.
15976 2021-07-26 Roger Sayle <roger@nextmovesoftware.com>
15977 Marc Glisse <marc.glisse@inria.fr>
15979 * match.pd (rotate): Simplify equality/inequality of rotations.
15980 (bswap): Simplify equality/inequality tests of byte swapping.
15982 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
15984 * range-op.cc (operator_bitwise_xor::op1_op2_relation_effect):
15987 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
15989 * range-op.cc (operator_lshift::fold_range): Pass rel to
15990 base class fold_range.
15991 (operator_rshift::fold_range): Same.
15993 2021-07-26 Ashimida <ashimida@linux.alibaba.com>
15996 * toplev.h (min_align_loops_log): Remove declaration.
15997 (min_align_jumps_log, min_align_labels_log): Likewise.
15998 (min_align_functions_log): Likewise.
16000 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
16002 * tree-vrp.c (vrp_simplify_cond_using_ranges): Rename vr_values
16004 (execute_vrp): Abstract out simplification of conditionals...
16005 (simplify_casted_conds): ...here.
16007 2021-07-26 Aldy Hernandez <aldyh@redhat.com>
16009 * gimple-array-bounds.cc (array_bounds_checker::get_value_range):
16010 Add gimple argument.
16011 (array_bounds_checker::check_array_ref): Same.
16012 (array_bounds_checker::check_addr_expr): Same.
16013 (array_bounds_checker::check_array_bounds): Pass statement to
16014 check_array_bounds and check_addr_expr.
16015 * gimple-array-bounds.h (check_array_bounds): Add gimple argument.
16016 (check_addr_expr): Same.
16017 (get_value_range): Same.
16019 2021-07-26 Tamar Christina <tamar.christina@arm.com>
16021 * config/aarch64/aarch64-simd-builtins.def (sdot, udot): Rename to..
16022 (sdot_prod, udot_prod): ... This.
16023 * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): Merged
16025 (<sur>dot_prod<vsi2qi>): ... this.
16026 (aarch64_<sur>dot_lane<vsi2qi>, aarch64_<sur>dot_laneq<vsi2qi>):
16027 Change operands order.
16028 (<sur>sadv16qi): Use new operands order.
16029 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32,
16030 vdotq_s32): Use new RTL ordering.
16032 2021-07-26 Tamar Christina <tamar.christina@arm.com>
16034 * config/aarch64/aarch64-builtins.c (TYPES_TERNOP_SUSS,
16035 aarch64_types_ternop_suss_qualifiers): New.
16036 * config/aarch64/aarch64-simd-builtins.def (usdot_prod): Use it.
16037 * config/aarch64/aarch64-simd.md (usdot_prod<vsi2qi>): Re-organize RTL.
16038 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Use it.
16040 2021-07-23 Jakub Jelinek <jakub@redhat.com>
16042 PR rtl-optimization/101562
16043 * expmed.c (store_integral_bit_field): Only use movstrict_optab
16044 if the operand isn't paradoxical.
16046 2021-07-23 Aldy Hernandez <aldyh@redhat.com>
16048 * gimple-array-bounds.h (class array_bounds_checker): Change
16049 ranges type to range_query.
16051 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
16053 * config/aarch64/arm_neon.h (vst1_s64_x2): Use
16054 __builtin_memcpy instead of constructing
16055 __builtin_aarch64_simd_oi one vector at a time.
16056 (vst1_u64_x2): Likewise.
16057 (vst1_f64_x2): Likewise.
16058 (vst1_s8_x2): Likewise.
16059 (vst1_p8_x2): Likewise.
16060 (vst1_s16_x2): Likewise.
16061 (vst1_p16_x2): Likewise.
16062 (vst1_s32_x2): Likewise.
16063 (vst1_u8_x2): Likewise.
16064 (vst1_u16_x2): Likewise.
16065 (vst1_u32_x2): Likewise.
16066 (vst1_f16_x2): Likewise.
16067 (vst1_f32_x2): Likewise.
16068 (vst1_p64_x2): Likewise.
16069 (vst1q_s8_x2): Likewise.
16070 (vst1q_p8_x2): Likewise.
16071 (vst1q_s16_x2): Likewise.
16072 (vst1q_p16_x2): Likewise.
16073 (vst1q_s32_x2): Likewise.
16074 (vst1q_s64_x2): Likewise.
16075 (vst1q_u8_x2): Likewise.
16076 (vst1q_u16_x2): Likewise.
16077 (vst1q_u32_x2): Likewise.
16078 (vst1q_u64_x2): Likewise.
16079 (vst1q_f16_x2): Likewise.
16080 (vst1q_f32_x2): Likewise.
16081 (vst1q_f64_x2): Likewise.
16082 (vst1q_p64_x2): Likewise.
16084 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
16086 * config/aarch64/arm_neon.h (vst1_s64_x3): Use
16087 __builtin_memcpy instead of constructing
16088 __builtin_aarch64_simd_ci one vector at a time.
16089 (vst1_u64_x3): Likewise.
16090 (vst1_f64_x3): Likewise.
16091 (vst1_s8_x3): Likewise.
16092 (vst1_p8_x3): Likewise.
16093 (vst1_s16_x3): Likewise.
16094 (vst1_p16_x3): Likewise.
16095 (vst1_s32_x3): Likewise.
16096 (vst1_u8_x3): Likewise.
16097 (vst1_u16_x3): Likewise.
16098 (vst1_u32_x3): Likewise.
16099 (vst1_f16_x3): Likewise.
16100 (vst1_f32_x3): Likewise.
16101 (vst1_p64_x3): Likewise.
16102 (vst1q_s8_x3): Likewise.
16103 (vst1q_p8_x3): Likewise.
16104 (vst1q_s16_x3): Likewise.
16105 (vst1q_p16_x3): Likewise.
16106 (vst1q_s32_x3): Likewise.
16107 (vst1q_s64_x3): Likewise.
16108 (vst1q_u8_x3): Likewise.
16109 (vst1q_u16_x3): Likewise.
16110 (vst1q_u32_x3): Likewise.
16111 (vst1q_u64_x3): Likewise.
16112 (vst1q_f16_x3): Likewise.
16113 (vst1q_f32_x3): Likewise.
16114 (vst1q_f64_x3): Likewise.
16115 (vst1q_p64_x3): Likewise.
16117 2021-07-23 H.J. Lu <hjl.tools@gmail.com>
16120 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): Don't return
16121 hard register when LRA is in progress.
16123 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
16125 * config/aarch64/arm_neon.h (vst1_s8_x4): Use
16126 __builtin_memcpy instead of using a union.
16127 (vst1q_s8_x4): Likewise.
16128 (vst1_s16_x4): Likewise.
16129 (vst1q_s16_x4): Likewise.
16130 (vst1_s32_x4): Likewise.
16131 (vst1q_s32_x4): Likewise.
16132 (vst1_u8_x4): Likewise.
16133 (vst1q_u8_x4): Likewise.
16134 (vst1_u16_x4): Likewise.
16135 (vst1q_u16_x4): Likewise.
16136 (vst1_u32_x4): Likewise.
16137 (vst1q_u32_x4): Likewise.
16138 (vst1_f16_x4): Likewise.
16139 (vst1q_f16_x4): Likewise.
16140 (vst1_f32_x4): Likewise.
16141 (vst1q_f32_x4): Likewise.
16142 (vst1_p8_x4): Likewise.
16143 (vst1q_p8_x4): Likewise.
16144 (vst1_p16_x4): Likewise.
16145 (vst1q_p16_x4): Likewise.
16146 (vst1_s64_x4): Likewise.
16147 (vst1_u64_x4): Likewise.
16148 (vst1_p64_x4): Likewise.
16149 (vst1q_s64_x4): Likewise.
16150 (vst1q_u64_x4): Likewise.
16151 (vst1q_p64_x4): Likewise.
16152 (vst1_f64_x4): Likewise.
16153 (vst1q_f64_x4): Likewise.
16155 2021-07-23 Jonathan Wrightt <jonathan.wright@arm.com>
16157 * config/aarch64/arm_neon.h (vst2_s64): Use __builtin_memcpy
16158 instead of constructing __builtin_aarch64_simd_oi one vector
16160 (vst2_u64): Likewise.
16161 (vst2_f64): Likewise.
16162 (vst2_s8): Likewise.
16163 (vst2_p8): Likewise.
16164 (vst2_s16): Likewise.
16165 (vst2_p16): Likewise.
16166 (vst2_s32): Likewise.
16167 (vst2_u8): Likewise.
16168 (vst2_u16): Likewise.
16169 (vst2_u32): Likewise.
16170 (vst2_f16): Likewise.
16171 (vst2_f32): Likewise.
16172 (vst2_p64): Likewise.
16173 (vst2q_s8): Likewise.
16174 (vst2q_p8): Likewise.
16175 (vst2q_s16): Likewise.
16176 (vst2q_p16): Likewise.
16177 (vst2q_s32): Likewise.
16178 (vst2q_s64): Likewise.
16179 (vst2q_u8): Likewise.
16180 (vst2q_u16): Likewise.
16181 (vst2q_u32): Likewise.
16182 (vst2q_u64): Likewise.
16183 (vst2q_f16): Likewise.
16184 (vst2q_f32): Likewise.
16185 (vst2q_f64): Likewise.
16186 (vst2q_p64): Likewise.
16188 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
16190 * config/aarch64/arm_neon.h (vst3_s64): Use __builtin_memcpy
16191 instead of constructing __builtin_aarch64_simd_ci one vector
16193 (vst3_u64): Likewise.
16194 (vst3_f64): Likewise.
16195 (vst3_s8): Likewise.
16196 (vst3_p8): Likewise.
16197 (vst3_s16): Likewise.
16198 (vst3_p16): Likewise.
16199 (vst3_s32): Likewise.
16200 (vst3_u8): Likewise.
16201 (vst3_u16): Likewise.
16202 (vst3_u32): Likewise.
16203 (vst3_f16): Likewise.
16204 (vst3_f32): Likewise.
16205 (vst3_p64): Likewise.
16206 (vst3q_s8): Likewise.
16207 (vst3q_p8): Likewise.
16208 (vst3q_s16): Likewise.
16209 (vst3q_p16): Likewise.
16210 (vst3q_s32): Likewise.
16211 (vst3q_s64): Likewise.
16212 (vst3q_u8): Likewise.
16213 (vst3q_u16): Likewise.
16214 (vst3q_u32): Likewise.
16215 (vst3q_u64): Likewise.
16216 (vst3q_f16): Likewise.
16217 (vst3q_f32): Likewise.
16218 (vst3q_f64): Likewise.
16219 (vst3q_p64): Likewise.
16221 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
16223 * config/aarch64/arm_neon.h (vst4_s64): Use __builtin_memcpy
16224 instead of constructing __builtin_aarch64_simd_xi one vector
16226 (vst4_u64): Likewise.
16227 (vst4_f64): Likewise.
16228 (vst4_s8): Likewise.
16229 (vst4_p8): Likewise.
16230 (vst4_s16): Likewise.
16231 (vst4_p16): Likewise.
16232 (vst4_s32): Likewise.
16233 (vst4_u8): Likewise.
16234 (vst4_u16): Likewise.
16235 (vst4_u32): Likewise.
16236 (vst4_f16): Likewise.
16237 (vst4_f32): Likewise.
16238 (vst4_p64): Likewise.
16239 (vst4q_s8): Likewise.
16240 (vst4q_p8): Likewise.
16241 (vst4q_s16): Likewise.
16242 (vst4q_p16): Likewise.
16243 (vst4q_s32): Likewise.
16244 (vst4q_s64): Likewise.
16245 (vst4q_u8): Likewise.
16246 (vst4q_u16): Likewise.
16247 (vst4q_u32): Likewise.
16248 (vst4q_u64): Likewise.
16249 (vst4q_f16): Likewise.
16250 (vst4q_f32): Likewise.
16251 (vst4q_f64): Likewise.
16252 (vst4q_p64): Likewise.
16254 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
16256 * config/aarch64/arm_neon.h (vtbx4_s8): Use __builtin_memcpy
16257 instead of constructing __builtin_aarch64_simd_oi one vector
16259 (vtbx4_u8): Likewise.
16260 (vtbx4_p8): Likewise.
16262 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
16264 * config/aarch64/arm_neon.h (vtbl3_s8): Use __builtin_memcpy
16265 instead of constructing __builtin_aarch64_simd_oi one vector
16267 (vtbl3_u8): Likewise.
16268 (vtbl3_p8): Likewise.
16269 (vtbl4_s8): Likewise.
16270 (vtbl4_u8): Likewise.
16271 (vtbl4_p8): Likewise.
16273 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
16275 * config/aarch64/arm_neon.h (vqtbx2_s8): Use __builtin_memcpy
16276 instead of constructing __builtin_aarch64_simd_oi one vector
16278 (vqtbx2_u8): Likewise.
16279 (vqtbx2_p8): Likewise.
16280 (vqtbx2q_s8): Likewise.
16281 (vqtbx2q_u8): Likewise.
16282 (vqtbx2q_p8): Likewise.
16283 (vqtbx3_s8): Use __builtin_memcpy instead of constructing
16284 __builtin_aarch64_simd_ci one vector at a time.
16285 (vqtbx3_u8): Likewise.
16286 (vqtbx3_p8): Likewise.
16287 (vqtbx3q_s8): Likewise.
16288 (vqtbx3q_u8): Likewise.
16289 (vqtbx3q_p8): Likewise.
16290 (vqtbx4_s8): Use __builtin_memcpy instead of constructing
16291 __builtin_aarch64_simd_xi one vector at a time.
16292 (vqtbx4_u8): Likewise.
16293 (vqtbx4_p8): Likewise.
16294 (vqtbx4q_s8): Likewise.
16295 (vqtbx4q_u8): Likewise.
16296 (vqtbx4q_p8): Likewise.
16298 2021-07-23 Jonathan Wright <jonathan.wright@arm.com>
16300 * config/aarch64/arm_neon.h (vqtbl2_s8): Use __builtin_memcpy
16301 instead of constructing __builtin_aarch64_simd_oi one vector
16303 (vqtbl2_u8): Likewise.
16304 (vqtbl2_p8): Likewise.
16305 (vqtbl2q_s8): Likewise.
16306 (vqtbl2q_u8): Likewise.
16307 (vqtbl2q_p8): Likewise.
16308 (vqtbl3_s8): Use __builtin_memcpy instead of constructing
16309 __builtin_aarch64_simd_ci one vector at a time.
16310 (vqtbl3_u8): Likewise.
16311 (vqtbl3_p8): Likewise.
16312 (vqtbl3q_s8): Likewise.
16313 (vqtbl3q_u8): Likewise.
16314 (vqtbl3q_p8): Likewise.
16315 (vqtbl4_s8): Use __builtin_memcpy instead of constructing
16316 __builtin_aarch64_simd_xi one vector at a time.
16317 (vqtbl4_u8): Likewise.
16318 (vqtbl4_p8): Likewise.
16319 (vqtbl4q_s8): Likewise.
16320 (vqtbl4q_u8): Likewise.
16321 (vqtbl4q_p8): Likewise.
16323 2021-07-23 Haochen Gui <guihaoc@gcc.gnu.org>
16326 * config/rs6000/rs6000.md (cstore<mode>4): Fix wrong fall through.
16328 2021-07-22 Andrew Pinski <apinski@marvell.com>
16330 PR tree-optimization/10153
16331 * tree-tailcall.c (create_tailcall_accumulator):
16332 Don't call fold_convert as the type should be correct already.
16333 (tree_optimize_tail_calls_1): Use build_{one,zero}_cst instead
16334 of integer_{one,zero}_node for the call of create_tailcall_accumulator.
16336 2021-07-22 Aldy Hernandez <aldyh@redhat.com>
16338 * gimple-range-cache.cc (non_null_ref::adjust_range): Replace
16339 varying_p check for null/non-null check.
16341 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
16343 PR tree-optimization/101511
16344 * value-relation.cc (relation_oracle::query_relation): Check if ssa1
16345 is in ssa2's equiv set, and don't trap if so.
16347 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
16349 PR tree-optimization/101497
16350 * gimple-range-fold.cc (fold_using_range::range_of_cond_expr): Check
16353 2021-07-22 Andrew MacLeod <amacleod@redhat.com>
16355 PR tree-optimization/101496
16356 * vr-values.c (simplify_using_ranges::fold_cond): Call range_of_stmt
16357 first, then vrp_visit_cond_Stmt.
16359 2021-07-22 liuhongt <hongtao.liu@intel.com>
16361 * config/i386/i386-expand.c
16362 (ix86_broadcast_from_integer_constant): Rename to ..
16363 (ix86_broadcast_from_constant): .. this, and extend it to
16365 (ix86_expand_vector_move): Extend to float mode.
16366 * config/i386/i386-features.c
16367 (replace_constant_pool_with_broadcast): Remove.
16368 (remove_partial_avx_dependency_gate): Ditto.
16369 (constant_pool_broadcast): Ditto.
16370 (class pass_constant_pool_broadcast): Ditto.
16371 (make_pass_constant_pool_broadcast): Ditto.
16372 (remove_partial_avx_dependency): Adjust gate.
16373 * config/i386/i386-passes.def: Remove pass_constant_pool_broadcast.
16374 * config/i386/i386-protos.h
16375 (make_pass_constant_pool_broadcast): Remove.
16377 2021-07-22 liuhongt <hongtao.liu@intel.com>
16379 * config/i386/constraints.md (Wb): New constraint.
16381 * config/i386/i386.md (*ashlhi3_1): Extend to avx512 mask
16383 (*ashlqi3_1): Ditto.
16384 (*<insn><mode>3_1): Split to ..
16385 (*ashr<mode>3_1): this, ...
16386 (*lshr<mode>3_1): and this, also extend this pattern to avx512
16388 (*<insn><mode>3_1): Split to ..
16389 (*ashr<mode>3_1): this, ...
16390 (*lshrqi3_1): and this, also extend this pattern to avx512
16392 (*lshrhi3_1): And this, also extend this pattern to avx512
16394 * config/i386/sse.md (k<code><mode>): New define_split after
16395 it to convert generic shift pattern to mask shift ones.
16397 2021-07-21 Thomas Schwinge <thomas@codesourcery.com>
16398 Joseph Myers <joseph@codesourcery.com>
16399 Cesar Philippidis <cesar@codesourcery.com>
16401 * tree-core.h (omp_clause_code): Add 'OMP_CLAUSE_NOHOST'.
16402 * tree.c (omp_clause_num_ops, omp_clause_code_name, walk_tree_1):
16404 * tree-pretty-print.c (dump_omp_clause): Likewise.
16405 * omp-general.c (oacc_verify_routine_clauses): Likewise.
16406 * gimplify.c (gimplify_scan_omp_clauses)
16407 (gimplify_adjust_omp_clauses): Likewise.
16408 * tree-nested.c (convert_nonlocal_omp_clauses)
16409 (convert_local_omp_clauses): Likewise.
16410 * omp-low.c (scan_sharing_clauses): Likewise.
16411 * omp-offload.c (execute_oacc_device_lower): Update.
16413 2021-07-21 Martin Sebor <msebor@redhat.com>
16415 * tree-ssa-alias.c (walk_aliased_vdefs_1): Fix typos in a comment.
16417 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
16419 * config/rs6000/rs6000-gen-builtins.c (write_init_bif_table):
16422 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
16424 * config/rs6000/rs6000-gen-builtins.c (write_fntype): New
16426 (write_fntype_init): New stub function.
16427 (write_init_bif_table): Likewise.
16428 (write_init_ovld_table): New function.
16429 (write_init_file): Implement.
16431 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
16433 * config/rs6000/rs6000-gen-builtins.c
16434 (write_autogenerated_header): New function.
16435 (write_decls): Likewise.
16436 (write_extern_fntype): New callback function.
16437 (write_header_file): Implement.
16439 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
16441 * config/rs6000/rs6000-gen-builtins.c (write_defines_file):
16444 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
16446 * config/rs6000/rs6000-gen-builtins.c (complete_vector_type): New
16448 (complete_base_type): Likewise.
16449 (construct_fntype_id): Likewise.
16450 (parse_bif_entry): Call contruct_fntype_id.
16451 (parse_ovld_entry): Likewise.
16453 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
16455 * config/rs6000/rs6000-gen-builtins.c (ovld_stanza): New struct.
16456 (MAXOVLDSTANZAS): New macro.
16457 (ovld_stanzas): New variable.
16458 (curr_ovld_stanza): Likewise.
16459 (MAXOVLDS): New macro.
16460 (ovlddata): New struct.
16461 (ovlds): New variable.
16462 (curr_ovld): Likewise.
16463 (max_ovld_args): Likewise.
16464 (parse_ovld_entry): New function.
16465 (parse_ovld_stanza): Likewise.
16466 (parse_ovld): Implement.
16468 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
16470 * config/rs6000/rs6000-gen-builtins.c (parse_bif_attrs):
16473 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
16475 * config/rs6000/rs6000-gen-builtins.c (parse_args): New function.
16476 (parse_prototype): Implement.
16478 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
16480 * config/rs6000/rs6000-gen-builtins.c (bif_stanza): New enum.
16481 (curr_bif_stanza): New variable.
16482 (stanza_entry): New struct.
16483 (stanza_map): New initialized variable.
16484 (enable_string): Likewise.
16485 (fnkinds): New enum.
16486 (typelist): New struct.
16487 (attrinfo): Likewise.
16488 (MAXRESTROPNDS): New macro.
16489 (prototype): New struct.
16490 (MAXBIFS): New macro.
16491 (bifdata): New struct.
16492 (bifs): New variable.
16493 (curr_bif): Likewise.
16494 (bif_order): Likewise.
16495 (bif_index): Likewise.
16496 (fatal): New function.
16497 (stanza_name_to_stanza): Likewise.
16498 (parse_bif_attrs): New stub function.
16499 (parse_prototype): Likewise.
16500 (parse_bif_entry): New function.
16501 (parse_bif_stanza): Likewise.
16502 (parse_bif): Implement.
16503 (set_bif_order): New function.
16504 (create_bif_order): Implement.
16506 2021-07-21 Bill Schmidt <wschmidt@linux.ibm.com>
16508 * config/rs6000/rs6000-gen-builtins.c (rbtree.h): New #include.
16509 (num_bifs): New variable.
16510 (num_ovld_stanzas): Likewise.
16511 (num_ovlds): Likewise.
16512 (parse_codes): New enum.
16513 (bif_rbt): New variable.
16514 (ovld_rbt): Likewise.
16515 (fntype_rbt): Likewise.
16516 (bifo_rbt): Likewise.
16517 (parse_bif): New stub function.
16518 (create_bif_order): Likewise.
16519 (parse_ovld): Likewise.
16520 (write_header_file): Likewise.
16521 (write_init_file): Likewise.
16522 (write_defines_file): Likewise.
16523 (delete_output_files): New function.
16526 2021-07-21 H.J. Lu <hjl.tools@gmail.com>
16529 * config/i386/i386-builtin.def: Remove OPTION_MASK_ISA_SSE4_2
16530 from CRC32 _builtin functions.
16532 2021-07-21 Sebastian Huber <sebastian.huber@embedded-brains.de>
16534 * coverage.c (build_gcov_info_var_registration): Mark the object placed
16535 in the linker set as referenced so that it does not get optimized away.
16537 2021-07-21 Kito Cheng <kito.cheng@sifive.com>
16540 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
16542 * config.gcc (riscv*-*-*): Detect which python is available.
16544 2021-07-21 Jakub Jelinek <jakub@redhat.com>
16546 PR middle-end/101535
16547 * gimplify.c (omp_check_private): Properly skip ORT_TARGET_DATA
16548 contexts in which decl isn't privatized and for ORT_TARGET return
16549 false if decl is mapped.
16551 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
16553 * gimple-loop-jam.c: Include tree-ssa-sccvn.h.
16554 (tree_loop_unroll_and_jam): Run value-numbering on a loop that
16555 has been successfully unrolled.
16557 2021-07-21 Richard Sandiford <richard.sandiford@arm.com>
16559 * tree-ssa-loop-manip.c (determine_exit_conditions): Return a null
16560 exit condition if no tail loop is needed, and if the original exit
16561 condition should therefore be kept as-is.
16562 (tree_transform_and_unroll_loop): Handle that case here too.
16564 2021-07-21 Kewen Lin <linkw@linux.ibm.com>
16566 * tree-data-ref.c (free_dependence_relations): Adjust to pass vec
16568 (free_data_refs): Likewise.
16569 * tree-data-ref.h (free_dependence_relations): Likewise.
16570 (free_data_refs): Likewise.
16571 * tree-predcom.c (struct chain): Use auto_vec instead of vec for
16573 (struct component): Likewise.
16574 (pcom_worker::pcom_worker): Adjust for auto_vec and renaming changes.
16575 (pcom_worker::~pcom_worker): Likewise.
16576 (pcom_worker::release_chain): Adjust as auto_vec changes.
16577 (pcom_worker::loop): Rename to ...
16578 (pcom_worker::m_loop): ... this.
16579 (pcom_worker::datarefs): Rename to ...
16580 (pcom_worker::m_datarefs): ... this. Use auto_vec instead of vec.
16581 (pcom_worker::dependences): Rename to ...
16582 (pcom_worker::m_dependences): ... this. Use auto_vec instead of vec.
16583 (pcom_worker::chains): Rename to ...
16584 (pcom_worker::m_chains): ... this. Use auto_vec instead of vec.
16585 (pcom_worker::looparound_phis): Rename to ...
16586 (pcom_worker::m_looparound_phis): ... this. Use auto_vec instead of
16588 (pcom_worker::cache): Rename to ...
16589 (pcom_worker::m_cache): ... this. Use auto_vec instead of vec.
16590 (pcom_worker::release_chain): Adjust for auto_vec changes.
16591 (pcom_worker::release_chains): Adjust for auto_vec and renaming
16593 (release_component): Remove.
16594 (release_components): Adjust for release_component removal.
16595 (component_of): Adjust to use vec.
16596 (merge_comps): Likewise.
16597 (pcom_worker::aff_combination_dr_offset): Adjust for renaming changes.
16598 (pcom_worker::determine_offset): Likewise.
16599 (class comp_ptrs): Remove.
16600 (pcom_worker::split_data_refs_to_components): Adjust for renaming
16601 changes, for comp_ptrs removal with auto_vec.
16602 (pcom_worker::suitable_component_p): Adjust for renaming changes.
16603 (pcom_worker::filter_suitable_components): Adjust for release_component
16605 (pcom_worker::valid_initializer_p): Adjust for renaming changes.
16606 (pcom_worker::find_looparound_phi): Likewise.
16607 (pcom_worker::add_looparound_copies): Likewise.
16608 (pcom_worker::determine_roots_comp): Likewise.
16609 (pcom_worker::single_nonlooparound_use): Likewise.
16610 (pcom_worker::execute_pred_commoning_chain): Likewise.
16611 (pcom_worker::execute_pred_commoning): Likewise.
16612 (pcom_worker::try_combine_chains): Likewise.
16613 (pcom_worker::prepare_initializers_chain): Likewise.
16614 (pcom_worker::prepare_initializers): Likewise.
16615 (pcom_worker::prepare_finalizers_chain): Likewise.
16616 (pcom_worker::prepare_finalizers): Likewise.
16617 (pcom_worker::tree_predictive_commoning_loop): Likewise.
16619 2021-07-20 Martin Sebor <msebor@redhat.com>
16621 PR middle-end/101397
16622 * builtins.c (gimple_call_return_array): Add argument. Correct
16623 offsets for memchr, mempcpy, stpcpy, and stpncpy.
16624 (compute_objsize_r): Adjust offset computation for argument returning
16627 2021-07-20 Martin Sebor <msebor@redhat.com>
16629 PR middle-end/101300
16630 * tree-ssa-uninit.c (check_defs): Handle UBSAN built-ins.
16632 2021-07-20 Jeff Law <jlaw@localhost.localdomain>
16634 * function.c (assign_parm_setup_block): Use adjust_address instead
16635 of change_address to preserve MEM_EXPR and friends.
16637 2021-07-20 Martin Sebor <msebor@redhat.com>
16639 * cfgloop.h (single_likely_exit): Adjust by-value argument to
16640 by-const-reference.
16641 * cfgloopanal.c (single_likely_exit): Same.
16642 * cgraph.h (struct cgraph_node): Same.
16643 * cgraphclones.c (cgraph_node::create_virtual_clone): Same.
16644 * genautomata.c (merge_states): Same.
16645 * genextract.c (VEC_char_to_string): Same.
16646 * genmatch.c (dt_node::gen_kids_1): Same.
16647 (walk_captures): Adjust by-value argument to by-reference.
16648 * gimple-ssa-store-merging.c (check_no_overlap): Adjust by-value argument
16649 to by-const-reference.
16650 * gimple.c (gimple_build_call_vec): Same.
16651 (gimple_build_call_internal_vec): Same.
16652 (gimple_build_switch): Same.
16653 (sort_case_labels): Same.
16654 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
16656 * gimple.h (gimple_build_call_vec): Adjust by-value argument to
16657 by-const-reference.
16658 (gimple_build_call_internal_vec): Same.
16659 (gimple_build_switch): Same.
16660 (sort_case_labels): Same.
16661 (preprocess_case_label_vec_for_gimple): Adjust by-value argument to
16663 * haifa-sched.c (calc_priorities): Adjust by-value argument to
16664 by-const-reference.
16665 (sched_init_luids): Same.
16666 (haifa_init_h_i_d): Same.
16667 * ipa-cp.c (ipa_get_indirect_edge_target_1): Same.
16668 (adjust_callers_for_value_intersection): Adjust by-value argument to
16670 (find_more_scalar_values_for_callers_subset): Adjust by-value argument to
16671 by-const-reference.
16672 (find_more_contexts_for_caller_subset): Same.
16673 (find_aggregate_values_for_callers_subset): Same.
16674 (copy_useful_known_contexts): Same.
16675 * ipa-fnsummary.c (remap_edge_summaries): Same.
16676 (remap_freqcounting_predicate): Same.
16677 * ipa-inline.c (add_new_edges_to_heap): Adjust by-value argument to
16679 * ipa-predicate.c (predicate::remap_after_inlining): Adjust by-value argument
16680 to by-const-reference.
16681 * ipa-predicate.h (predicate::remap_after_inlining): Same.
16682 * ipa-prop.c (ipa_find_agg_cst_for_param): Same.
16683 * ipa-prop.h (ipa_find_agg_cst_for_param): Same.
16684 * ira-build.c (ira_loop_tree_body_rev_postorder): Same.
16685 * read-rtl.c (add_overload_instance): Same.
16686 * rtl.h (native_decode_rtx): Same.
16687 (native_decode_vector_rtx): Same.
16688 * sched-int.h (sched_init_luids): Same.
16689 (haifa_init_h_i_d): Same.
16690 * simplify-rtx.c (native_decode_vector_rtx): Same.
16691 (native_decode_rtx): Same.
16692 * tree-call-cdce.c (gen_shrink_wrap_conditions): Same.
16693 (shrink_wrap_one_built_in_call_with_conds): Same.
16694 (shrink_wrap_conditional_dead_built_in_calls): Same.
16695 * tree-data-ref.c (create_runtime_alias_checks): Same.
16696 (compute_all_dependences): Same.
16697 * tree-data-ref.h (compute_all_dependences): Same.
16698 (create_runtime_alias_checks): Same.
16699 (index_in_loop_nest): Same.
16700 * tree-if-conv.c (mask_exists): Same.
16701 * tree-loop-distribution.c (class loop_distribution): Same.
16702 (loop_distribution::create_rdg_vertices): Same.
16703 (dump_rdg_partitions): Same.
16704 (debug_rdg_partitions): Same.
16705 (partition_contains_all_rw): Same.
16706 (loop_distribution::distribute_loop): Same.
16707 * tree-parloops.c (oacc_entry_exit_ok_1): Same.
16708 (oacc_entry_exit_single_gang): Same.
16709 * tree-ssa-loop-im.c (hoist_memory_references): Same.
16710 (loop_suitable_for_sm): Same.
16711 * tree-ssa-loop-niter.c (bound_index): Same.
16712 * tree-ssa-reassoc.c (update_ops): Same.
16713 (swap_ops_for_binary_stmt): Same.
16714 (rewrite_expr_tree): Same.
16715 (rewrite_expr_tree_parallel): Same.
16716 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Same.
16717 * tree-ssa-sccvn.h (ao_ref_init_from_vn_reference): Same.
16718 * tree-ssa-structalias.c (process_all_all_constraints): Same.
16719 (make_constraints_to): Same.
16720 (handle_lhs_call): Same.
16721 (find_func_aliases_for_builtin_call): Same.
16722 (sort_fieldstack): Same.
16723 (check_for_overlaps): Same.
16724 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Same.
16725 (vect_create_cond_for_unequal_addrs): Same.
16726 (vect_create_cond_for_lower_bounds): Same.
16727 (vect_create_cond_for_alias_checks): Same.
16728 * tree-vect-slp-patterns.c (vect_validate_multiplication): Same.
16729 * tree-vect-slp.c (vect_analyze_slp_instance): Same.
16730 (vect_make_slp_decision): Same.
16731 (vect_slp_bbs): Same.
16732 (duplicate_and_interleave): Same.
16733 (vect_transform_slp_perm_load): Same.
16734 (vect_schedule_slp): Same.
16735 * tree-vectorizer.h (vect_transform_slp_perm_load): Same.
16736 (vect_schedule_slp): Same.
16737 (duplicate_and_interleave): Same.
16738 * tree.c (build_vector_from_ctor): Same.
16739 (build_vector): Same.
16740 (check_vector_cst): Same.
16741 (check_vector_cst_duplicate): Same.
16742 (check_vector_cst_fill): Same.
16743 (check_vector_cst_stepped): Same.
16744 * tree.h (build_vector_from_ctor): Same.
16746 2021-07-20 Jakub Jelinek <jakub@redhat.com>
16749 * config/rs6000/rs6000-protos.h (easy_altivec_constant): Change return
16750 type from bool to int.
16751 * config/rs6000/rs6000.c (vspltis_constant): Fix up handling the
16752 EASY_VECTOR_MSB case if either step or copies is not 1.
16753 (vspltis_shifted): Fix comment typo.
16754 (easy_altivec_constant): Change return type from bool to int, instead
16755 of returning true return byte size of the element mode that should be
16756 used to synthetize the constant.
16757 * config/rs6000/predicates.md (easy_vector_constant_msb): Require
16758 that vspltis_shifted is 0, handle the case where easy_altivec_constant
16759 assumes using different vector mode from CONST_VECTOR's mode.
16760 * config/rs6000/altivec.md (easy_vector_constant_msb splitter): Use
16761 easy_altivec_constant to determine mode in which -1 >> -1 should be
16762 performed, use rs6000_expand_vector_init instead of gen_vec_initv4sisi.
16764 2021-07-20 Richard Biener <rguenther@suse.de>
16767 * dwarf2out.h (dwarf_file_data): Add key member.
16768 * dwarf2out.c (dwarf_file_hasher::equal): Compare key.
16769 (dwarf_file_hasher::hash): Hash key.
16770 (lookup_filename): Remap the filename and store it in the
16771 filename member of dwarf_file_data when creating a new
16773 (file_name_acquire): Do not remap the filename again.
16774 (maybe_emit_file): Likewise.
16776 2021-07-20 Jonathan Wright <jonathan.wright@arm.com>
16778 * config/aarch64/aarch64-simd-builtins.def: Use two variant
16779 generators for all TBL/TBX intrinsics and rename to
16780 consistent forms: qtbl[1234] or qtbx[1234].
16781 * config/aarch64/aarch64-simd.md (aarch64_tbl1<mode>):
16783 (aarch64_qtbl1<mode>): This.
16784 (aarch64_tbx1<mode>): Rename to...
16785 (aarch64_qtbx1<mode>): This.
16786 (aarch64_tbl2v16qi): Delete.
16787 (aarch64_tbl3<mode>): Rename to...
16788 (aarch64_qtbl2<mode>): This.
16789 (aarch64_tbx4<mode>): Rename to...
16790 (aarch64_qtbx2<mode>): This.
16791 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_1): Use
16792 renamed qtbl1 and qtbl2 RTL patterns.
16793 * config/aarch64/arm_neon.h (vqtbl1_p8): Use renamed qtbl1
16795 (vqtbl1_s8): Likewise.
16796 (vqtbl1_u8): Likewise.
16797 (vqtbl1q_p8): Likewise.
16798 (vqtbl1q_s8): Likewise.
16799 (vqtbl1q_u8): Likewise.
16800 (vqtbx1_s8): Use renamed qtbx1 RTL pattern.
16801 (vqtbx1_u8): Likewise.
16802 (vqtbx1_p8): Likewise.
16803 (vqtbx1q_s8): Likewise.
16804 (vqtbx1q_u8): Likewise.
16805 (vqtbx1q_p8): Likewise.
16806 (vtbl1_s8): Use renamed qtbl1 RTL pattern.
16807 (vtbl1_u8): Likewise.
16808 (vtbl1_p8): Likewise.
16809 (vtbl2_s8): Likewise
16810 (vtbl2_u8): Likewise.
16811 (vtbl2_p8): Likewise.
16812 (vtbl3_s8): Use renamed qtbl2 RTL pattern.
16813 (vtbl3_u8): Likewise.
16814 (vtbl3_p8): Likewise.
16815 (vtbl4_s8): Likewise.
16816 (vtbl4_u8): Likewise.
16817 (vtbl4_p8): Likewise.
16818 (vtbx2_s8): Use renamed qtbx2 RTL pattern.
16819 (vtbx2_u8): Likewise.
16820 (vtbx2_p8): Likewise.
16821 (vqtbl2_s8): Use renamed qtbl2 RTL pattern.
16822 (vqtbl2_u8): Likewise.
16823 (vqtbl2_p8): Likewise.
16824 (vqtbl2q_s8): Likewise.
16825 (vqtbl2q_u8): Likewise.
16826 (vqtbl2q_p8): Likewise.
16827 (vqtbx2_s8): Use renamed qtbx2 RTL pattern.
16828 (vqtbx2_u8): Likewise.
16829 (vqtbx2_p8): Likewise.
16830 (vqtbx2q_s8): Likewise.
16831 (vqtbx2q_u8): Likewise.
16832 (vqtbx2q_p8): Likewise.
16833 (vtbx4_s8): Likewise.
16834 (vtbx4_u8): Likewise.
16835 (vtbx4_p8): Likewise.
16837 2021-07-20 Uroš Bizjak <ubizjak@gmail.com>
16840 * config/i386/sync.md (define_peephole2 atomic_storedi_fpu):
16842 (define_peephole2 atomic_loaddi_fpu): Ditto.
16844 2021-07-20 Kito Cheng <kito.cheng@sifive.com>
16846 * config.gcc (riscv*-*-*): Detect which python is available.
16848 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
16850 * config/rs6000/vsx.md (mulhs_<mode>): Rename to...
16851 (smul<mode>3_highpart): ... this.
16852 (mulhu_<mode>): Rename to...
16853 (umul<mode>3_highpart): ... this.
16854 * config/rs6000/rs6000-builtin.def (MULHS_V2DI, MULHS_V4SI,
16855 MULHU_V2DI, MULHU_V4SI): Adjust.
16857 2021-07-20 Kewen Lin <linkw@linux.ibm.com>
16859 PR tree-optimization/100696
16860 * internal-fn.c (first_commutative_argument): Add info for IFN_MULH.
16861 * internal-fn.def (IFN_MULH): New internal function.
16862 * tree-vect-patterns.c (vect_recog_mulhs_pattern): Add support to
16863 recog normal multiply highpart as IFN_MULH.
16864 * config/i386/i386.c (ix86_add_stmt_cost): Adjust for combined
16867 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
16869 * config/elfos.h (CTF_DEBUGGING_INFO): New definition.
16870 (BTF_DEBUGGING_INFO): Likewise.
16871 * doc/tm.texi.in: Document the new macros.
16872 * doc/tm.texi: Regenerated.
16873 * toplev.c: Guard initialization of debug hooks.
16875 2021-07-19 Indu Bhagat <indu.bhagat@oracle.com>
16877 * flags.h (ctf_debuginfo_p): New function declaration.
16878 * opts.c (ctf_debuginfo_p): New function definition.
16880 2021-07-19 Andrew Stubbs <ams@codesourcery.com>
16883 * config/gcn/gcn-hsa.h (DRIVER_SELF_SPECS): New.
16884 (ASM_SPEC): Set -mattr for xnack and sram-ecc.
16885 * config/gcn/gcn-opts.h (enum sram_ecc_type): New.
16886 * config/gcn/gcn-valu.md: Add a warning comment.
16887 * config/gcn/gcn.c (gcn_option_override): Add "sorry" for -mxnack.
16888 (output_file_start): Add xnack and sram-ecc state to ".amdgcn_target".
16889 * config/gcn/gcn.md: Add a warning comment.
16890 * config/gcn/gcn.opt: Add -mxnack and -msram-ecc.
16891 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): Remove
16893 (EF_AMDGPU_XNACK): New.
16894 (EF_AMDGPU_SRAM_ECC): New.
16896 (copy_early_debug_info): Use elf_flags.
16897 (main): Handle -mxnack and -msram-ecc options.
16898 * doc/invoke.texi: Document -mxnack and -msram-ecc.
16900 2021-07-19 Andrew Pinski <apinski@marvell.com>
16903 * config/aarch64/aarch64.md (csneg3_uxtw_insn): Rename to ...
16904 (*cs<neg_not_cs>3_uxtw_insn4): and extend to NEG_NOT.
16906 2021-07-19 Richard Biener <rguenther@suse.de>
16908 PR tree-optimization/101505
16909 * tree-vect-patterns.c (vect_determine_precisions): Walk
16910 PHIs also for loop vectorization.
16912 2021-07-19 Richard Biener <rguenther@suse.de>
16914 * gimple.h (gimple_expr_type): Remove.
16915 * doc/gimple.texi: Remove gimple_expr_type documentation.
16917 2021-07-19 Richard Biener <rguenther@suse.de>
16919 * tree-ssa-sccvn.c (vn_reference_eq): Handle NULL vr->type.
16920 (ao_ref_init_from_vn_reference): Likewise.
16921 (fully_constant_reference): Likewise.
16922 (vn_reference_lookup_call): Do not set vr->type to random
16924 * tree-ssa-pre.c (compute_avail): Do not try to PRE calls
16926 * tree-vect-generic.c (expand_vector_piecewise): Pass in
16927 whether we expanded parallel.
16928 (expand_vector_parallel): Adjust.
16929 (expand_vector_addition): Likewise.
16930 (expand_vector_comparison): Likewise.
16931 (expand_vector_operation): Likewise.
16932 (expand_vector_scalar_condition): Likewise.
16933 (expand_vector_conversion): Likewise.
16935 2021-07-19 Richard Biener <rguenther@suse.de>
16937 * tree-vrp.c (register_edge_assert_for_2): Use the
16939 (vrp_folder::fold_predicate_in): Likewise.
16940 * vr-values.c (gimple_assign_nonzero_p): Likewise.
16941 (vr_values::extract_range_from_comparison): Likewise.
16942 (vr_values::extract_range_from_ubsan_builtin): Use the
16943 type of the first operand.
16944 (vr_values::extract_range_basic): Push down type
16945 computation, use the appropriate LHS.
16946 (vr_values::extract_range_from_assignment): Use the
16949 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
16952 * common/config/i386/i386-common.c (ix86_handle_option): For
16953 -mgeneral-regs-only, enable the GPR only instructions which are
16954 enabled implicitly by SSE ISAs unless they have been disabled
16957 2021-07-18 H.J. Lu <hjl.tools@gmail.com>
16960 * config/i386/i386.c (ix86_check_avx_upper_stores): Moved before
16961 ix86_avx_u128_mode_needed.
16962 (ix86_avx_u128_mode_needed): Return AVX_U128_DIRTY if callee
16963 returns AVX register.
16965 2021-07-17 Jan Hubicka <hubicka@ucw.cz>
16967 * tree-ssa-structalias.c (handle_rhs_call): Support EAF_NOT_RETURNED.
16968 (handle_const_call): Liekise
16969 (handle_pure_call): Liekise
16971 2021-07-17 Andrew MacLeod <amacleod@redhat.com>
16973 PR tree-optimization/96542
16974 * range-op.cc (range_operator::wi_fold_in_parts): New.
16975 (range_operator::fold_range): Call wi_fold_in_parts.
16976 (operator_lshift::wi_fold): Fix broken lshift by [0,0].
16977 * range-op.h (wi_fold_in_parts): Add prototype.
16979 2021-07-16 David Malcolm <dmalcolm@redhat.com>
16981 * doc/analyzer.texi: Add __analyzer_dump_state.
16983 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
16985 * config/rs6000/rbtree.c: New file.
16986 * config/rs6000/rbtree.h: New file.
16988 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
16990 * config/rs6000/rs6000-gen-builtins.c (restriction): New enum.
16991 (typeinfo): Add restr field.
16992 (match_bracketed_pair): New function.
16993 (match_const_restriction): Implement.
16995 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
16997 * config/rs6000/rs6000-gen-builtins.c (match_basetype): Implement.
16999 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
17001 * config/rs6000/rs6000-gen-builtins.c (void_status): New enum.
17002 (basetype): Likewise.
17003 (typeinfo): Likewise.
17004 (handle_pointer): New function.
17005 (match_basetype): New stub function.
17006 (match_const_restriction): Likewise.
17007 (match_type): New function.
17009 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
17011 * config/rs6000/rs6000-gen-builtins.c (consume_whitespace): New
17013 (advance_line): Likewise.
17014 (safe_inc_pos): Likewise.
17015 (match_identifier): Likewise.
17016 (match_integer): Likewise.
17017 (match_to_right_bracket): Likewise.
17019 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
17021 * config/rs6000/rs6000-gen-builtins.c (bif_file): New variable.
17022 (ovld_file): Likewise.
17023 (header_file): Likewise.
17024 (init_file): Likewise.
17025 (defines_file): Likewise.
17026 (pgm_path): Likewise.
17027 (bif_path): Likewise.
17028 (ovld_path): Likewise.
17029 (header_path): Likewise.
17030 (init_path): Likewise.
17031 (defines_path): Likewise.
17032 (LINELEN): New macro.
17033 (linebuf): New variable.
17037 (bif_diag): New function.
17038 (ovld_diag): Likewise.
17040 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
17042 * config/rs6000/rs6000-builtin-new.def: New.
17043 * config/rs6000/rs6000-overload.def: New.
17045 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
17047 * config/rs6000/rs6000-gen-builtins.c: New.
17049 2021-07-16 Bill Schmidt <wschmidt@linux.ibm.com>
17051 * Makefile.in (EXTRA_GTYPE_DEPS): New variable.
17052 (s-gtype): Depend on EXTRA_GTYPE_DEPS.
17053 * gengtype-state.c (state_writer::write_state_file_list): Add a
17054 parameter to the fileslist expression for the number of build
17056 (read_state_files_list): Detect build headers and strip the
17057 initial "./" or ".\" from their names.
17058 * gengtype.c (build_headers): New global variable.
17059 (num_build_headers): Likewise.
17060 (open_base_files): Emit #include for each build header.
17061 (main): Detect and count build headers.
17062 * gengtype.h (build_headers): New extern variable.
17063 (num_build_headers): Likewise.
17065 2021-07-16 Richard Biener <rguenther@suse.de>
17067 * gimple-ssa-store-merging.c (verify_symbolic_number_p): Use
17068 the type of the LHS.
17069 (find_bswap_or_nop_1): Likewise.
17070 (find_bswap_or_nop): Likewise.
17071 * tree-vectorizer.h (vect_get_smallest_scalar_type): Adjust
17073 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
17074 Remove unused parameters, pass in the scalar type. Fix
17075 internal store function handling.
17076 * tree-vect-stmts.c (vect_analyze_stmt): Remove assert.
17077 (vect_get_vector_types_for_stmt): Move down check for
17078 existing vector stmt after we've determined a scalar type.
17079 Pass down the used scalar type to vect_get_smallest_scalar_type.
17080 * tree-vect-generic.c (expand_vector_condition): Use
17081 the type of the LHS.
17082 (expand_vector_scalar_condition): Likewise.
17083 (expand_vector_operations_1): Likewise.
17084 * tree-vect-patterns.c (vect_widened_op_tree): Likewise.
17085 (vect_recog_dot_prod_pattern): Likewise.
17086 (vect_recog_sad_pattern): Likewise.
17087 (vect_recog_widen_op_pattern): Likewise.
17088 (vect_recog_widen_sum_pattern): Likewise.
17089 (vect_recog_mixed_size_cond_pattern): Likewise.
17091 2021-07-16 Jan Hubicka <hubicka@ucw.cz>
17093 * ipa-modref.c (struct escape_entry): Use eaf_fleags_t.
17094 (dump_eaf_flags): Dump EAF_NOT_RETURNED
17095 (eaf_flags_useful_p): Use eaf_fleags_t; handle const functions
17096 and EAF_NOT_RETURNED.
17097 (modref_summary::useful_p): Likewise.
17098 (modref_summary_lto::useful_p): Likewise.
17099 (struct) modref_summary_lto: Use eaf_fleags_t.
17100 (deref_flags): Handle EAF_NOT_RETURNED.
17101 (struct escape_point): Use min_flags.
17102 (modref_lattice::init): Add EAF_NOT_RETURNED.
17103 (merge_call_lhs_flags): Ignore EAF_NOT_RETURNED functions
17104 (analyze_ssa_name_flags): Clear EAF_NOT_RETURNED on return;
17106 (analyze_parms): Also analyze const functions; update conition on
17108 (modref_write): Update streaming.
17109 (read_section): Update streaming.
17110 (remap_arg_flags): Use eaf_flags_t.
17111 (modref_merge_call_site_flags): Hanlde EAF_NOT_RETURNED.
17112 * ipa-modref.h: (eaf_flags_t): New typedef.
17113 (struct modref_summary): Use eaf_flags_t.
17114 * tree-core.h (EAF_NOT_RETURNED): New constant.
17116 2021-07-16 Richard Biener <rguenther@suse.de>
17118 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Use
17119 the type of the LHS.
17120 (gimple_assign_nonnegative_warnv_p): Likewise.
17121 (gimple_call_nonnegative_warnv_p): Likewise. Return false
17122 if the call has no LHS.
17123 * gimple.c (gimple_could_trap_p_1): Use the type of the LHS.
17124 * tree-eh.c (stmt_could_throw_1_p): Likewise.
17125 * tree-inline.c (insert_init_stmt): Likewise.
17126 * tree-ssa-loop-niter.c (get_val_for): Likewise.
17127 * tree-outof-ssa.c (ssa_is_replaceable_p): Use the type of
17129 * tree-ssa-sccvn.c (init_vn_nary_op_from_stmt): Take a
17130 gassign *. Use the type of the lhs.
17131 (vn_nary_op_lookup_stmt): Adjust.
17132 (vn_nary_op_insert_stmt): Likewise.
17134 2021-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
17136 * config/s390/predicates.md (bras_sym_operand): Accept all
17137 functions in 64-bit mode, use UNSPEC_PLT31.
17138 (larl_operand): Use UNSPEC_PLT31.
17139 * config/s390/s390.c (s390_loadrelative_operand_p): Likewise.
17140 (legitimize_pic_address): Likewise.
17141 (s390_emit_tls_call_insn): Mark __tls_get_offset as function,
17143 (s390_delegitimize_address): Use UNSPEC_PLT31.
17144 (s390_output_addr_const_extra): Likewise.
17145 (print_operand): Add @PLT to TLS calls, handle %K.
17146 (s390_function_profiler): Mark __fentry__/_mcount as function,
17147 use %K, use UNSPEC_PLT31.
17148 (s390_output_mi_thunk): Use only UNSPEC_GOT, use %K.
17149 (s390_emit_call): Use UNSPEC_PLT31.
17150 (s390_emit_tpf_eh_return): Mark __tpf_eh_return as function.
17151 * config/s390/s390.md (UNSPEC_PLT31): Rename from UNSPEC_PLT.
17152 (*movdi_64): Use %K.
17153 (reload_base_64): Likewise.
17154 (*sibcall_brc): Likewise.
17155 (*sibcall_brcl): Likewise.
17156 (*sibcall_value_brc): Likewise.
17157 (*sibcall_value_brcl): Likewise.
17159 (*brasl): Likewise.
17160 (*bras_r): Likewise.
17161 (*brasl_r): Likewise.
17162 (*bras_tls): Likewise.
17163 (*brasl_tls): Likewise.
17164 (main_base_64): Likewise.
17165 (reload_base_64): Likewise.
17166 (@split_stack_call<mode>): Likewise.
17168 2021-07-16 Richard Biener <rguenther@suse.de>
17170 PR tree-optimization/101467
17171 * tree-vect-stmts.c (vect_gen_while): Properly guard
17172 make_temp_ssa_name usage.
17174 2021-07-16 Cooper Qu <cooper.qu@linux.alibaba.com>
17176 * config.gcc: Don't use forked print-sysroot-suffix.sh and
17177 t-sysroot-suffix for C-SKY.
17178 * config/csky/print-sysroot-suffix.sh: Delete.
17179 * config/csky/t-csky-linux: Delete.
17180 * config/csky/t-sysroot-suffix: Define MULTILIB_DIRNAMES
17181 instead of CSKY_MULTILIB_DIRNAMES.
17183 2021-07-16 Richard Biener <rguenther@suse.de>
17185 * tree-vect-loop.c (vect_transform_cycle_phi): Correct sign
17186 conversion issues with the partial reduction of the reused
17187 vector accumulator.
17189 2021-07-16 Richard Biener <rguenther@suse.de>
17191 * config/i386/i386-options.c (ix86_option_override_internal): Set
17192 param_vect_partial_vector_usage to zero if not set.
17194 2021-07-15 Uroš Bizjak <ubizjak@gmail.com>
17197 * config/i386/i386.h (VALID_SSE_REG_MODE): Add TDmode.
17198 (VALID_INT_MODE_P): Add SDmode and DDmode.
17199 Add TDmode for TARGET_64BIT.
17200 (VALID_DFP_MODE_P): Remove.
17201 * config/i386/i386.c (ix86_hard_regno_mode_ok):
17202 Do not use VALID_DFP_MODE_P.
17204 2021-07-15 Andrew MacLeod <amacleod@redhat.com>
17206 * gimple-range-fold.cc (adjust_pointer_diff_expr): Use
17208 (fold_using_range::fold_stmt): Ditto.
17209 (fold_using_range::range_of_range_op): Ditto.
17210 (fold_using_range::range_of_phi): Ditto.
17211 (fold_using_range::range_of_call): Ditto.
17212 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
17213 (fold_using_range::range_of_builtin_call): Ditto.
17214 (fold_using_range::range_of_cond_expr): Ditto.
17215 * gimple-range-fold.h (gimple_range_type): New.
17217 2021-07-15 Martin Sebor <msebor@redhat.com>
17219 PR middle-end/97027
17220 * tree-ssa-strlen.c (handle_assign): New function.
17221 (maybe_warn_overflow): Add argument.
17222 (nonzero_bytes_for_type): New function.
17223 (count_nonzero_bytes): Handle more tree types. Call
17224 nonzero_bytes_for_tye.
17225 (count_nonzero_bytes): Handle types.
17226 (handle_store): Handle stores from function calls.
17227 (strlen_check_and_optimize_call): Move code to handle_assign. Call
17228 it for assignments from function calls.
17230 2021-07-15 David Malcolm <dmalcolm@redhat.com>
17235 * doc/invoke.texi: Add -Wanalyzer-use-of-uninitialized-value.
17237 2021-07-15 David Malcolm <dmalcolm@redhat.com>
17239 * doc/invoke.texi (-fdump-analyzer-exploded-paths): New.
17241 2021-07-15 Martin Sebor <msebor@redhat.com>
17245 * fold-const.c (operand_compare::operand_equal_p): Handle OEP_DECL_NAME.
17246 (operand_compare::verify_hash_value): Same.
17247 * tree-core.h (OEP_DECL_NAME): New.
17249 2021-07-15 Martin Jambor <mjambor@suse.cz>
17251 * profile-count.h (profile_count::value): Change the return type to
17253 * gimple-pretty-print.c (dump_gimple_bb_header): Adjust print
17255 * tree-cfg.c (dump_function_to_file): Likewise.
17257 2021-07-15 Bill Schmidt <wschmidt@linux.ibm.com>
17260 * config/rs6000/rs6000-p8swap.c (has_part_mult): New.
17261 (rs6000_analyze_swaps): Insns containing a subreg of a mult are
17264 2021-07-15 Richard Biener <rguenther@suse.de>
17266 * tree-vectorizer.h (vect_gen_while): Match up with
17267 vect_gen_while_not.
17268 * tree-vect-stmts.c (vect_gen_while): Adjust API to that
17269 of vect_gen_while_not.
17270 (vect_gen_while_not): Adjust.
17271 * tree-vect-loop-manip.c (vect_set_loop_controls_directly): Likewise.
17273 2021-07-15 Aldy Hernandez <aldyh@redhat.com>
17275 * gimple-range-cache.cc (non_null_ref::adjust_range): New.
17276 (ranger_cache::range_of_def): Call adjust_range.
17277 (ranger_cache::entry_range): Same.
17278 * gimple-range-cache.h (non_null_ref::adjust_range): New.
17279 * gimple-range.cc (gimple_ranger::range_of_expr): Call
17281 (gimple_ranger::range_on_entry): Same.
17283 2021-07-15 Tamar Christina <tamar.christina@arm.com>
17286 2021-07-14 Tamar Christina <tamar.christina@arm.com>
17288 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
17290 2021-07-15 Tamar Christina <tamar.christina@arm.com>
17293 2021-07-14 Tamar Christina <tamar.christina@arm.com>
17295 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
17296 (sdot_prod, udot_prod): ...These.
17297 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
17298 (aarch64_<sur>dot<vsi2qi>): Rename to...
17299 (<sur>dot_prod<vsi2qi>): ...This.
17300 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
17303 2021-07-15 Jakub Jelinek <jakub@redhat.com>
17305 PR middle-end/101437
17306 * gimplify.c (gimplify_expr): Throw away volatile reads from empty
17307 types even if they have non-BLKmode TYPE_MODE.
17309 2021-07-15 Richard Biener <rguenther@suse.de>
17312 * gcc.c (process_command): Process -gtoggle like process_options
17313 would after parsing options.
17315 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
17317 * cfgexpand.c (expand_asm_loc): Adjust.
17318 (expand_asm_stmt): Likewise.
17319 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Likewise.
17320 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
17321 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
17322 * config/avr/avr.c (avr_md_asm_adjust): Likewise.
17323 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
17324 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
17325 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
17326 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
17327 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
17328 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
17329 * config/s390/s390.c (s390_md_asm_adjust): Likewise.
17330 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
17331 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
17332 * doc/tm.texi: Regenerate.
17333 * target.def: Add location argument to md_asm_adjust.
17335 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
17337 * tree-diagnostic.c (diagnostic_report_current_function): Use the
17338 diagnostic's location, not input_location.
17340 2021-07-15 Trevor Saunders <tbsaunde@tbsaunde.org>
17342 * cfgexpand.c (tree_conflicts_with_clobbers_p): Pass location to
17344 (expand_asm_stmt): Likewise.
17346 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
17348 * config/rs6000/rs6000.c (adjacent_mem_locations): Return the lower
17349 addressed memory rtx, if any.
17350 (rs6000_split_multireg_move): Fix code formatting.
17351 Handle MMA build built-ins with operands in adjacent memory locations.
17353 2021-07-14 Peter Bergner <bergner@linux.ibm.com>
17355 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Move to later
17358 2021-07-14 Jason Merrill <jason@redhat.com>
17360 * sel-sched-ir.h (get_all_loop_exits): Use auto_vec.
17362 2021-07-14 Jason Merrill <jason@redhat.com>
17364 * doc/invoke.texi: -fdelete-dead-exceptions is on by default for
17367 2021-07-14 Tamar Christina <tamar.christina@arm.com>
17369 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
17370 Remove erroneous line.
17372 2021-07-14 Andrew MacLeod <amacleod@redhat.com>
17374 * params.opt (param_evrp_mode): Change default.
17376 2021-07-14 Tamar Christina <tamar.christina@arm.com>
17378 * config/aarch64/aarch64-simd-builtins.def (udot, sdot): Rename to...
17379 (sdot_prod, udot_prod): ...These.
17380 * config/aarch64/aarch64-simd.md (<sur>dot_prod<vsi2qi>): Remove.
17381 (aarch64_<sur>dot<vsi2qi>): Rename to...
17382 (<sur>dot_prod<vsi2qi>): ...This.
17383 * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
17386 2021-07-14 Tamar Christina <tamar.christina@arm.com>
17388 * config/arm/neon.md (<sup>dot_prod<vsi2qi>): Drop statements.
17390 2021-07-14 Tamar Christina <tamar.christina@arm.com>
17392 * doc/sourcebuild.texi (arm_v8_2a_i8mm_neon_hw): Document.
17394 2021-07-14 Tamar Christina <tamar.christina@arm.com>
17396 * config/arm/neon.md (usdot_prod<vsi2qi>): New.
17398 2021-07-14 Tamar Christina <tamar.christina@arm.com>
17400 * config/aarch64/aarch64-simd.md (aarch64_usdot<vsi2qi>): Rename to...
17401 (usdot_prod<vsi2qi>): ... This.
17402 * config/aarch64/aarch64-simd-builtins.def (usdot): Rename to...
17403 (usdot_prod): ...This.
17404 * config/aarch64/arm_neon.h (vusdot_s32, vusdotq_s32): Likewise.
17405 * config/aarch64/aarch64-sve.md (@aarch64_<sur>dot_prod<vsi2qi>):
17407 (@<sur>dot_prod<vsi2qi>): ...This.
17408 * config/aarch64/aarch64-sve-builtins-base.cc
17409 (svusdot_impl::expand): Use it.
17411 2021-07-14 Tamar Christina <tamar.christina@arm.com>
17413 * optabs.def (usdot_prod_optab): New.
17414 * doc/md.texi: Document it and clarify other dot prod optabs.
17415 * optabs-tree.h (enum optab_subtype): Add optab_vector_mixed_sign.
17416 * optabs-tree.c (optab_for_tree_code): Support usdot_prod_optab.
17417 * optabs.c (expand_widen_pattern_expr): Likewise.
17418 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
17419 * tree-vect-loop.c (vectorizable_reduction): Query dot-product kind.
17420 * tree-vect-patterns.c (vect_supportable_direct_optab_p): Take optional
17422 (vect_widened_op_tree): Optionally ignore
17424 (vect_recog_dot_prod_pattern): Support usdot_prod_optab.
17426 2021-07-14 H.J. Lu <hjl.tools@gmail.com>
17429 * config/i386/driver-i386.c (host_detect_local_cpu): Check
17430 "arch [32|64]" and "tune [32|64]" for 32-bit and 64-bit codegen.
17431 Enable UINTR only for 64-bit codegen.
17432 * config/i386/i386-options.c
17433 (ix86_option_override_internal::DEF_PTA): Skip PTA_UINTR if not
17435 * config/i386/i386.h (ARCH_ARG): New.
17436 (CC1_CPU_SPEC): Pass "[arch|tune] 32" for 32-bit codegen and
17437 "[arch|tune] 64" for 64-bit codegen.
17439 2021-07-14 Richard Biener <rguenther@suse.de>
17441 PR tree-optimization/101445
17442 * tree-vect-stmts.c (vectorizable_load): Do the gap adjustment
17443 of the IV in the correct direction for negative stride
17446 2021-07-14 Jakub Jelinek <jakub@redhat.com>
17449 * godump.c (godump_str_hash): New type.
17450 (godump_container::pot_dummy_types): Use string_hash instead of
17451 ptr_hash in the hash_set.
17453 2021-07-14 Richard Biener <rguenther@suse.de>
17455 * tree-vect-loop.c (vect_find_reusable_accumulator): Handle
17456 vector types where the old vector type has a multiple of
17457 the new vector type elements.
17458 (vect_create_partial_epilog): New function, split out from...
17459 (vect_create_epilog_for_reduction): ... here.
17460 (vect_transform_cycle_phi): Reduce the re-used accumulator
17461 to the new vector type.
17463 2021-07-14 Alexandre Oliva <oliva@adacore.com>
17465 * tree-ssa-alias.c (attr_fnspec::verify): Fix index in
17466 non-'t'-sized arg check.
17468 2021-07-14 Alexandre Oliva <oliva@adacore.com>
17470 * tree-cfg.c (cleanup_dead_labels_eh): Update
17471 post_landing_pad label upon change of landing pad block's
17473 (cleanup_dead_labels): Check that a removed label is not that
17476 2021-07-13 Jonathan Wright <jonathan.wright@arm.com>
17478 * combine.c (combine_simplify_rtx): Add vec_select -> subreg
17480 * config/aarch64/aarch64.md (*zero_extend<SHORT:mode><GPI:mode>2_aarch64):
17481 Add Neon to general purpose register case for zero-extend
17483 * config/arm/vfp.md (*arm_movsi_vfp): Remove "*" from *t -> r
17484 case to prevent some cases opting to go through memory.
17485 * cse.c (fold_rtx): Add vec_select -> subreg simplification.
17486 * rtl.c (rtvec_series_p): Define predicate to determine
17487 whether a vector contains a linear series of integers.
17488 * rtl.h (rtvec_series_p): Define.
17489 * rtlanal.c (vec_series_lowpart_p): Define predicate to
17490 determine if a vector selection is equivalent to the low part
17492 * rtlanal.h (vec_series_lowpart_p): Define.
17493 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
17494 Add vec_select -> subreg simplification.
17496 2021-07-13 Paul A. Clarke <pc@us.ibm.com>
17498 * config/rs6000/smmintrin.h (_mm_testz_si128, _mm_testc_si128,
17499 _mm_testnzc_si128, _mm_test_all_ones, _mm_test_all_zeros,
17500 _mm_test_mix_ones_zeros): New.
17502 2021-07-13 Roger Sayle <roger@nextmovesoftware.com>
17503 Richard Biener <rguenther@suse.de>
17505 * gimple.c (gimple_could_trap_p_1): Make S argument a
17506 "const gimple*". Preserve constness in call to
17507 gimple_asm_volatile_p.
17508 (gimple_could_trap_p): Make S argument a "const gimple*".
17509 * gimple.h (gimple_could_trap_p_1, gimple_could_trap_p):
17510 Update function prototypes.
17512 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
17514 * tree-vectorizer.h (vect_reusable_accumulator): New structure.
17515 (_loop_vec_info::main_loop_edge): New field.
17516 (_loop_vec_info::skip_main_loop_edge): Likewise.
17517 (_loop_vec_info::skip_this_loop_edge): Likewise.
17518 (_loop_vec_info::reusable_accumulators): Likewise.
17519 (_stmt_vec_info::reduc_scalar_results): Likewise.
17520 (_stmt_vec_info::reused_accumulator): Likewise.
17521 (vect_get_main_loop_result): Declare.
17522 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize
17523 reduc_scalar_inputs.
17524 (vec_info::free_stmt_vec_info): Free reduc_scalar_inputs.
17525 * tree-vect-loop-manip.c (vect_get_main_loop_result): New function.
17526 (vect_do_peeling): Fill an epilogue loop's main_loop_edge,
17527 skip_main_loop_edge and skip_this_loop_edge fields.
17528 * tree-vect-loop.c (INCLUDE_ALGORITHM): Define.
17529 (vect_emit_reduction_init_stmts): New function.
17530 (get_initial_def_for_reduction): Use it.
17531 (get_initial_defs_for_reduction): Likewise. Change the vinfo
17532 parameter to a loop_vec_info.
17533 (vect_create_epilog_for_reduction): Store the scalar results
17534 in the reduc_info. If an epilogue loop is reusing an accumulator
17535 from the main loop, and if the epilogue loop can also be skipped,
17536 try to place the reduction code in the join block. Record
17537 accumulators that could potentially be reused by epilogue loops.
17538 (vect_transform_cycle_phi): When vectorizing epilogue loops,
17539 try to reuse accumulators from the main loop. Record the initial
17540 value in reduc_info for non-SLP reductions too.
17542 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
17544 * tree-vect-loop.c (get_initial_def_for_reduction): Remove
17545 adjustment handling. Take the neutral value as an argument,
17546 in place of the code argument.
17547 (vect_transform_cycle_phi): Update accordingly. Handle the
17548 initial values of cond reductions separately from code reductions.
17549 Choose the adjustment here rather than in
17550 get_initial_def_for_reduction. Sink the splat of vec_initial_def.
17552 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
17554 * tree-vect-loop.c (neutral_op_for_slp_reduction): Replace with...
17555 (neutral_op_for_reduction): ...this, providing a more general
17557 (vect_create_epilog_for_reduction): Update accordingly.
17558 (vectorizable_reduction): Likewise.
17559 (vect_transform_cycle_phi): Likewise.
17561 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
17563 * tree-vect-loop.c (get_initial_def_for_reduction): Take the
17564 reduc_info instead of the original stmt_vec_info.
17565 (vect_transform_cycle_phi): Update accordingly.
17567 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
17569 * tree-vect-loop.c (get_initial_defs_for_reduction): Take the
17570 reduc_info as an additional parameter.
17571 (vect_transform_cycle_phi): Update accordingly.
17573 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
17575 * tree-vectorizer.h: Include tree-ssa-operands.h.
17576 (vect_phi_initial_value): New function.
17577 * tree-vect-loop.c (neutral_op_for_slp_reduction): Use it.
17578 (get_initial_defs_for_reduction, info_for_reduction): Likewise.
17579 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
17580 (vect_transform_cycle_phi, vectorizable_induction): Likewise.
17582 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
17584 * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
17585 the phi results to vectype after creating them. Remove later
17586 conversion code that thus becomes redundant.
17588 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
17590 * tree-vect-loop.c (vect_create_epilog_for_reduction): Replace
17591 the new_phis vector with a reduc_inputs vector. Combine handling
17592 of reduction chains and ncopies > 1.
17594 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
17596 * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
17597 scalar_results to group_size elements after reducing down from
17598 N*group_size elements. Construct an array_slice of the live-out
17599 stmts and assert that there is one stmt per scalar result.
17601 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
17603 * tree-vect-loop.c (vect_create_epilog_for_reduction): Remove
17604 nested_in_vect_loop and use double_reduc everywhere. Remove dead
17605 assignment to "loop".
17607 2021-07-13 Richard Sandiford <richard.sandiford@arm.com>
17609 * internal-fn.c (vectorized_internal_fn_supported_p): Handle
17610 vector types first. For scalar types, consider both the preferred
17611 vector mode and the alternative vector modes.
17612 * optabs-query.c (can_vec_mask_load_store_p): Use the same
17613 structure as above, in particular using related_vector_mode
17614 for modes provided by autovectorize_vector_modes.
17616 2021-07-13 Jakub Jelinek <jakub@redhat.com>
17617 Richard Biener <rguenther@suse.de>
17619 PR tree-optimization/101419
17620 * tree-pass.h (PROP_objsz): Define.
17621 (make_pass_early_object_sizes): Declare.
17622 * passes.def (pass_all_early_optimizations): Rename pass_object_sizes
17623 there to pass_early_object_sizes, drop parameter.
17624 (pass_all_optimizations): Move pass_object_sizes right after pass_ccp,
17625 drop parameter, move pass_post_ipa_warn right after that.
17626 * tree-object-size.c (pass_object_sizes::execute): Rename to...
17627 (object_sizes_execute): ... this. Add insert_min_max_p argument.
17628 (pass_data_object_sizes): Move after object_sizes_execute.
17629 (pass_object_sizes): Likewise. In execute method call
17630 object_sizes_execute, drop set_pass_param method and insert_min_max_p
17631 non-static data member and its initializer in the ctor.
17632 (pass_data_early_object_sizes, pass_early_object_sizes,
17633 make_pass_early_object_sizes): New.
17634 * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Use
17635 (cfun->curr_properties & PROP_objsz) instead of cfun->after_inlining.
17637 2021-07-13 Kito Cheng <kito.cheng@sifive.com>
17640 * config/riscv/constraints.md ("S"): Update description and remove
17642 * doc/md.texi (Machine Constraints): Document the 'S' constraints
17645 2021-07-13 Richard Biener <rguenther@suse.de>
17648 2021-07-12 Richard Biener <rguenther@suse.de>
17650 * tree-vect-slp.c (vect_slp_region): Show the number of
17651 SLP graph entries in the optimization message.
17653 2021-07-13 Michael Meissner <meissner@linux.ibm.com>
17655 * config/rs6000/altivec.md (xxspltiw_v4sf): Change local variable
17657 * config/rs6000/rs6000-protos.h (rs6000_const_f32_to_i32): Change
17658 return type to long.
17659 * config/rs6000/rs6000.c (rs6000_const_f32_to_i32): Change return
17662 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
17664 * gimple-range-fold.cc (fold_using_range::range_of_builtin_ubsan_call):
17665 Query relation between the 2 operands and use it.
17667 2021-07-12 Sergei Trofimovich <siarheit@google.com>
17669 * doc/cfg.texi: Fix s/ei_safe_safe/ei_safe_edge/ typo.
17671 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
17674 * config/i386/predicates.md (vec_setm_sse41_operand):
17675 Rename from vec_setm_operand.
17676 (vec_setm_avx2_operand): New predicate.
17677 * config/i386/sse.md (vec_set<V_128:mode>): Use V_128 mode iterator.
17678 Use vec_setm_sse41_operand as operand 2 predicate.
17679 (vec_set<V_256_512:mode): New expander.
17680 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_sse41_operand
17681 as operand 2 predicate.
17683 2021-07-12 Andrew MacLeod <amacleod@redhat.com>
17685 PR tree-optimization/101335
17686 * range-op.cc (operator_cast::lhs_op1_relation): Delete.
17688 2021-07-12 Andrew Pinski <apinski@marvell.com>
17690 * tree-ssa-phiopt.c (match_simplify_replacement): Move
17691 insert of the sequence before the movement of the
17692 statement. Check if to see if the statement is used
17693 outside of the original phi to see if we should move it.
17695 2021-07-12 Richard Biener <rguenther@suse.de>
17697 * dump-context.h (debug_dump_context::debug_dump_context):
17698 Add FILE * parameter defaulted to stderr.
17699 * dumpfile.c (debug_dump_context::debug_dump_context): Adjust.
17700 * tree-vect-slp.c (dot_slp_tree): New functions.
17702 2021-07-12 Richard Biener <rguenther@suse.de>
17704 PR tree-optimization/101373
17705 * tree-ssa-pre.c (prune_clobbered_mems): Also prune trapping
17706 references when the BB may not return.
17707 (compute_avail): Pass in the function we're working on and
17708 replace cfun references with it. Externally throwing
17709 const calls also possibly terminate the function.
17710 (pass_pre::execute): Pass down the function we're working on.
17711 * gcse.c (compute_hash_table_work): Externally throwing
17712 const/pure calls also need record_last_mem_set_info.
17713 * postreload-gcse.c (record_opr_changes): Looping or externally
17714 throwing const/pure calls also need record_last_mem_set_info.
17716 2021-07-12 Uroš Bizjak <ubizjak@gmail.com>
17718 * recog.c (memory_address_addr_space_p): Change the type to bool.
17719 Return true/false instead of 1/0.
17720 (offsettable_memref_p): Ditto.
17721 (offsettable_nonstrict_memref_p): Ditto.
17722 (offsettable_address_addr_space_p): Ditto.
17723 Change the type of addressp indirect function to bool.
17724 * recog.h (memory_address_addr_space_p): Change the type to bool.
17725 (strict_memory_address_addr_space_p): Ditto.
17726 (offsettable_memref_p): Ditto.
17727 (offsettable_nonstrict_memref_p): Ditto.
17728 (offsettable_address_addr_space_p): Ditto.
17729 * reload.c (maybe_memory_address_addr_space_p): Ditto.
17730 (strict_memory_address_addr_space_p): Change the type to bool.
17731 Return true/false instead of 1/0.
17732 (maybe_memory_address_addr_space_p): Change the type to bool.
17734 2021-07-12 Richard Biener <rguenther@suse.de>
17736 * tree-vect-slp.c (vect_slp_region): Show the number of
17737 SLP graph entries in the optimization message.
17739 2021-07-12 Richard Biener <rguenther@suse.de>
17741 PR tree-optimization/101394
17742 * tree-ssa-pre.c (do_pre_regular_insertion): Avoid inserting
17743 copies from abnormals for a full redundancy.
17745 2021-07-12 Richard Biener <rguenther@suse.de>
17747 PR middle-end/101423
17748 * gimple.c (gimple_could_trap_p_1): Internal function calls
17750 * tree-eh.c (tree_could_trap_p): Likewise.
17752 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
17755 * config/arm/arm_neon.h (vmul_n_u32): Replace call to builtin with
17757 (vmulq_n_u32): Likewise.
17758 (vmul_n_f32): Gate __a * __b on __FAST_MATH__.
17759 (vmulq_n_f32): Likewise.
17760 (vmul_n_f16): Likewise.
17761 (vmulq_n_f16): Likewise.
17763 2021-07-12 Martin Liska <mliska@suse.cz>
17765 PR sanitizer/101425
17766 * gcc.c (check_offload_target_name): Call
17767 candidates_list_and_hint only if we have a candidate.
17769 2021-07-12 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
17772 * config/arm/neon.md (vec_init): Move to ...
17773 * config/arm/vec-common.md (vec_init): ... here.
17774 Change the pattern's mode to VDQX and gate it on VALID_MVE_MODE.
17776 2021-07-12 Roger Sayle <roger@nextmovesoftware.com>
17778 PR tree-optimization/101403
17779 * match.pd ((T)bswap(X)>>C): Correctly handle cases where
17780 signedness of the shift is not the same as the signedness of
17781 the type extension.
17783 2021-07-09 Roger Sayle <roger@nextmovesoftware.com>
17784 Uroš Bizjak <ubizjak@gmail.com>
17786 * config/i386/i386.md (*divmodsi4_const): Optimize SImode
17787 divmod of a constant numerator with new define_insn_and_split.
17789 2021-07-09 Iain Sandoe <iain@sandoe.co.uk>
17792 * config/i386/i386-expand.c (ix86_expand_call): If a call is
17793 to a non-local-binding, or local but to a public symbol, then
17794 assume that it might be indirected via the lazy symbol binder.
17795 Mark R10 and R10 as clobbered in that case.
17797 2021-07-09 Eric Botcazou <ebotcazou@adacore.com>
17800 * gcc.c (ASM_DEBUG_DWARF_OPTION): Set again to --gdwarf2 in
17801 the case where HAVE_AS_WORKING_DWARF_N_FLAG is not defined
17802 and HAVE_LD_BROKEN_PE_DWARF5 is defined.
17804 2021-07-09 Uroš Bizjak <ubizjak@gmail.com>
17806 * config/i386/i386.md (*udivmodsi4_pow2_zext_1): Limit the
17807 log2 range of operands[3] to [1,31].
17808 (*udivmodsi4_pow2_zext_2): Ditto. Correct insn RTX pattern.
17810 2021-07-09 Sergei Trofimovich <siarheit@google.com>
17812 * doc/md.texi: Don't split @smallexample in multiple @groups.
17814 2021-07-09 Sergei Trofimovich <siarheit@google.com>
17816 * doc/md.texi: Add missing 'see' word.
17818 2021-07-09 Andrew Pinski <apinski@marvell.com>
17820 * tree-ssa-phiopt.c (phiopt_early_allow): Change arguments
17821 to take sequence and gimple_match_op. Accept the case where
17822 op is a SSA_NAME and one statement in the sequence.
17823 Also allow constants.
17824 (gimple_simplify_phiopt): Always pass a sequence to resimplify.
17825 Update call to phiopt_early_allow. Discard the sequence if not
17828 2021-07-09 Xi Ruoyao <xry111@mengyan1223.wang>
17833 * config/mips/mips.c (mips_const_insns): Use MSA_SUPPORTED_MODE_P
17834 instead of ISA_HAS_MSA.
17835 (mips_expand_vec_unpack): Likewise.
17836 (mips_expand_vector_init): Likewise.
17838 2021-07-09 Kewen Lin <linkw@linux.ibm.com>
17840 * config/rs6000/vsx.md (mods_<mode>): Rename to...
17841 (mod<mode>3): ... this.
17842 (modu_<mode>): Rename to...
17843 (umod<mode>3): ... this.
17844 * config/rs6000/rs6000-builtin.def (MODS_V2DI, MODS_V4SI, MODU_V2DI,
17845 MODU_V4SI): Adjust.
17847 2021-07-08 Jeff Law <jeffreyalaw@gmail.com>
17849 * config/h8300/shiftrotate.md (variable shifts): Expose condition
17850 code handling for the test before the loop.
17852 2021-07-08 Martin Jambor <mjambor@suse.cz>
17855 * ipa-sra.c (class isra_call_summary): New member
17856 m_before_any_store, initialize it in the constructor.
17857 (isra_call_summary::dump): Dump the new field.
17858 (ipa_sra_call_summaries::duplicate): Copy it.
17859 (process_scan_results): Set it.
17860 (isra_write_edge_summary): Stream it.
17861 (isra_read_edge_summary): Likewise.
17862 (param_splitting_across_edge): Only override
17863 safe_to_import_accesses if m_before_any_store is set.
17865 2021-07-08 Martin Sebor <msebor@redhat.com>
17867 PR bootstrap/101374
17868 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref):
17869 Use Object Size Type 0 instead of 1.
17871 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
17873 * tree-vect-loop.c (vectorizable_reduction): Remove always-true
17876 2021-07-08 Richard Sandiford <richard.sandiford@arm.com>
17878 * match.pd: Simplify an extend-operate-truncate sequence involving
17881 2021-07-08 Roger Sayle <roger@nextmovesoftware.com>
17882 Richard Biener <rguenther@suse.de>
17884 PR tree-optimization/40210
17885 * match.pd (bswap optimizations): Simplify (bswap(x)>>C1)&C2 as
17886 (x>>C3)&C2 when possible. Simplify bswap(x)>>C1 as ((T)x)>>C2
17887 when possible. Simplify bswap(x)&C1 as (x>>C2)&C1 when 0<=C1<=255.
17889 2021-07-08 Uroš Bizjak <ubizjak@gmail.com>
17892 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
17894 * config/i386/mmx.md (V_32): New mode iterator.
17895 (mov<V_32:mode>): Use V_32 mode iterator.
17896 (*mov<V_32:mode>_internal): Ditto.
17897 (*push<V_32:mode>2_rex64): Ditto.
17898 (*push<V_32:mode>2): Ditto.
17899 (movmisalign<V_32:mode>): Ditto.
17900 (mmx_<any_shiftrt:insn>v1si3): New insn pattern.
17901 (sse4_1_<any_extend:code>v2qiv2hi2): Ditto.
17902 (vec_unpacks_lo_v4qi): New expander.
17903 (vec_unpacks_hi_v4qi): Ditto.
17904 (vec_unpacku_lo_v4qi): Ditto.
17905 (vec_unpacku_hi_v4qi): Ditto.
17906 * config/i386/i386.h (VALID_SSE2_REG_MODE): Add V1SImode.
17907 (VALID_INT_MODE_P): Ditto.
17909 2021-07-08 Michael Meissner <meissner@linux.ibm.com>
17912 * config/rs6000/rs6000.md (udivti3): New insn.
17913 (divti3): New insn.
17914 (umodti3): New insn.
17915 (modti3): New insn.
17917 2021-07-07 Martin Sebor <msebor@redhat.com>
17919 PR tree-optimization/100137
17920 PR tree-optimization/99121
17921 PR tree-optimization/97027
17922 * builtins.c (access_ref::access_ref): Also set offmax.
17923 (access_ref::offset_in_range): Define new function.
17924 (access_ref::add_offset): Set offmax.
17925 (access_ref::inform_access): Handle access_none.
17926 (handle_mem_ref): Clear ostype.
17927 (compute_objsize_r): Handle ASSERT_EXPR.
17928 * builtins.h (struct access_ref): Add offmax member.
17929 * gimple-array-bounds.cc (array_bounds_checker::check_mem_ref): Use
17930 compute_objsize() and simplify.
17932 2021-07-07 Peter Bergner <bergner@linux.ibm.com>
17934 * config/rs6000/rs6000-call.c (mma_init_builtins): Use VSX_BUILTIN_LXVP
17935 and VSX_BUILTIN_STXVP.
17937 2021-07-07 Martin Sebor <msebor@redhat.com>
17940 * config/aarch64/aarch64.c (aarch64_simd_lane_bounds): Remove
17941 a stray %K from error_at() missed in r12-2088.
17943 2021-07-07 Richard Biener <rguenther@suse.de>
17945 PR tree-optimization/99728
17946 * tree-ssa-loop-im.c (gather_mem_refs_stmt): Record
17948 (mem_refs_may_alias_p): Add assert we handled aggregate
17950 (sm_seq_valid_bb): Give up when running into aggregate copies.
17951 (ref_indep_loop_p): Handle aggregate copies as never
17952 being invariant themselves but allow other refs to be
17953 disambiguated against them.
17954 (can_sm_ref_p): Do not try to apply store-motion to aggregate
17957 2021-07-06 Indu Bhagat <indu.bhagat@oracle.com>
17960 * dwarf2ctf.c (ctf_get_AT_data_member_location): Multiply by 8 to get
17963 2021-07-06 Martin Sebor <msebor@redhat.com>
17965 * gimple-pretty-print.c (percent_G_format): Remove.
17966 * tree-diagnostic.c (default_tree_printer): Remove calls.
17967 * tree-pretty-print.c (percent_K_format): Remove.
17968 * tree-pretty-print.h (percent_K_format): Remove.
17970 2021-07-06 Martin Sebor <msebor@redhat.com>
17972 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_builtin):
17973 Remove %K and use error_at.
17974 (aarch64_expand_fcmla_builtin): Same.
17975 (aarch64_expand_builtin_tme): Same.
17976 (aarch64_expand_builtin_memtag): Same.
17977 * config/arm/arm-builtins.c (arm_expand_acle_builtin): Same.
17978 (arm_expand_builtin): Same.
17979 * config/arm/arm.c (bounds_check): Same.
17981 2021-07-06 Martin Sebor <msebor@redhat.com>
17983 * builtins.c (warn_string_no_nul): Remove %G.
17984 (maybe_warn_for_bound): Same.
17985 (warn_for_access): Same.
17986 (check_access): Same.
17987 (check_strncat_sizes): Same.
17988 (expand_builtin_strncat): Same.
17989 (expand_builtin_strncmp): Same.
17990 (expand_builtin): Same.
17991 (expand_builtin_object_size): Same.
17992 (warn_dealloc_offset): Same.
17993 (maybe_emit_free_warning): Same.
17994 * calls.c (maybe_warn_alloc_args_overflow): Same.
17995 (maybe_warn_nonstring_arg): Same.
17996 (maybe_warn_rdwr_sizes): Same.
17997 * expr.c (expand_expr_real_1): Remove %K.
17998 * gimple-fold.c (gimple_fold_builtin_strncpy): Remove %G.
17999 (gimple_fold_builtin_strncat): Same.
18000 * gimple-ssa-sprintf.c (format_directive): Same.
18001 (handle_printf_call): Same.
18002 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Same.
18003 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
18004 (maybe_diag_access_bounds): Same. Call gimple_location.
18005 (check_bounds_or_overlap): Same.
18006 * trans-mem.c (ipa_tm_scan_irr_block): Remove %K. Simplify.
18007 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Remove %G.
18008 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
18009 (maybe_diag_stxncpy_trunc): Same.
18010 (handle_builtin_stxncpy_strncat): Same.
18011 (maybe_warn_pointless_strcmp): Same.
18012 * tree-ssa-uninit.c (maybe_warn_operand): Same.
18014 2021-07-06 Uroš Bizjak <ubizjak@gmail.com>
18017 * config/i386/predicates.md (vec_setm_operand): Enable
18018 register_operand for TARGET_SSE4_1.
18019 * config/i386/mmx.md (vec_setv2hi): Use vec_setm_operand
18020 as operand 2 predicate. Call ix86_expand_vector_set_var
18021 for non-constant index operand.
18022 (vec_setv4qi): Use vec_setm_mmx_operand as operand 2 predicate.
18023 Call ix86_expand_vector_set_var for non-constant index operand.
18025 2021-07-06 Jeff Law <jeffreyalaw@gmail.com>
18027 * config/h8300/jumpcall.md (*branch): When possible, generate
18028 the comparison in CCZN mode.
18029 * config/h8300/predicates.md (simple_memory_operand): Reject all
18030 auto-increment addressing modes.
18032 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
18034 PR bootstrap/100246
18035 * config/i386/i386.h (struct stringop_algs): Define a CTOR for
18038 2021-07-06 Richard Biener <rguenther@suse.de>
18040 * doc/md.texi (vec_fmaddsub<mode>4): Document.
18041 (vec_fmsubadd<mode>4): Likewise.
18042 * optabs.def (vec_fmaddsub$a4): Add.
18043 (vec_fmsubadd$a4): Likewise.
18044 * internal-fn.def (IFN_VEC_FMADDSUB): Add.
18045 (IFN_VEC_FMSUBADD): Likewise.
18046 * tree-vect-slp-patterns.c (addsub_pattern::recognize):
18047 Refactor to handle IFN_VEC_FMADDSUB and IFN_VEC_FMSUBADD.
18048 (addsub_pattern::build): Likewise.
18049 * tree-vect-slp.c (vect_optimize_slp): CFN_VEC_FMADDSUB
18050 and CFN_VEC_FMSUBADD are not transparent for permutes.
18051 * config/i386/sse.md (vec_fmaddsub<mode>4): New expander.
18052 (vec_fmsubadd<mode>4): Likewise.
18054 2021-07-06 Richard Biener <rguenther@suse.de>
18056 * doc/invoke.texi (fmove-loop-stores): Document.
18057 * common.opt (fmove-loop-stores): New option.
18058 * opts.c (default_options_table): Enable -fmove-loop-stores
18059 at -O1 but not -Og.
18060 * tree-ssa-loop-im.c (pass_lim::execute): Pass
18061 flag_move_loop_stores instead of true to
18062 loop_invariant_motion_in_fun.
18064 2021-07-06 Iain Sandoe <iain@sandoe.co.uk>
18066 * doc/install.texi: Document --with-dsymutil.
18068 2021-07-06 Andrew Pinski <apinski@marvell.com>
18070 PR tree-optimization/101256
18071 * dbgcnt.def (phiopt_edge_range): New counter.
18072 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
18073 Check to make sure the new name is defined in the same
18074 bb as the conditional before duplicating range info.
18075 Also add debug counter.
18077 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
18079 PR rtl-optimization/100328
18080 * config/i386/i386-options.c (ix86_option_override_internal):
18081 Set param_ira_consider_dup_in_all_alts to 0.
18083 2021-07-06 Kewen Lin <linkw@linux.ibm.com>
18085 PR rtl-optimization/100328
18086 * doc/invoke.texi (ira-consider-dup-in-all-alts): Document new
18088 * ira.c (ira_get_dup_out_num): Adjust as parameter
18089 param_ira_consider_dup_in_all_alts.
18090 * params.opt (ira-consider-dup-in-all-alts): New.
18091 * ira-conflicts.c (process_regs_for_copy): Add one parameter
18092 single_input_op_has_cstr_p.
18093 (get_freq_for_shuffle_copy): New function.
18094 (add_insn_allocno_copies): Adjust as single_input_op_has_cstr_p.
18095 * ira-int.h (ira_get_dup_out_num): Add one bool parameter.
18097 2021-07-05 Jeff Law <jeffreyalaw@gmail.com>
18099 * config/h8300/shiftrotate.md (shift-by-variable patterns): Update to
18100 generate condition code aware RTL directly.
18102 2021-07-05 Andrew Pinski <apinski@marvell.com>
18104 PR tree-optimization/101039
18105 * match.pd (A CMP 0 ? A : -A): New patterns.
18106 * tree-ssa-phiopt.c (abs_replacement): Delete function.
18107 (tree_ssa_phiopt_worker): Don't call abs_replacement.
18108 Update comment about abs_replacement.
18110 2021-07-05 Andrew Pinski <apinski@marvell.com>
18112 * tree-ssa-phiopt.c (gimple_simplify_phiopt):
18113 If "A ? B : C" fails to simplify, try "(!A) ? C : B".
18115 2021-07-05 Andrew Pinski <apinski@marvell.com>
18117 * tree-ssa-phiopt.c (match_simplify_replacement):
18118 Add early_p argument. Call gimple_simplify_phiopt
18119 instead of gimple_simplify.
18120 (tree_ssa_phiopt_worker): Update call to
18121 match_simplify_replacement and allow unconditionally.
18122 (phiopt_early_allow): New function.
18123 (gimple_simplify_phiopt): New function.
18125 2021-07-05 Andrew Pinski <apinski@marvell.com>
18127 PR middle-end/101237
18128 * fold-const.c (negate_expr_p): Remove call to element_mode
18129 and TREE_MODE/TREE_TYPE when calling HONOR_SIGNED_ZEROS,
18130 HONOR_SIGN_DEPENDENT_ROUNDING, and HONOR_SNANS.
18131 (fold_negate_expr_1): Likewise.
18132 (const_unop): Likewise.
18133 (fold_cond_expr_with_comparison): Likewise.
18134 (fold_binary_loc): Likewise.
18135 (fold_ternary_loc): Likewise.
18136 (tree_call_nonnegative_warnv_p): Likewise.
18137 * match.pd (-(A + B) -> (-B) - A): Likewise.
18139 2021-07-05 Iain Sandoe <iain@sandoe.co.uk>
18141 * configure.ac: Handle --with-dsymutil in the same way as we
18142 do for the assembler and linker. (DEFAULT_DSYMUTIL): New.
18143 Extract the type and version for the dsymutil configured or
18144 found by the default searches.
18145 * config.in: Regenerated.
18146 * configure: Regenerated.
18147 * collect2.c (do_dsymutil): Handle locating dsymutil in the
18148 same way as for the assembler and linker.
18149 * config/darwin.h (DSYMUTIL): Delete.
18150 * gcc.c: Report a configured dsymutil correctly.
18151 * exec-tool.in: Allow for dsymutil.
18153 2021-07-05 Uroš Bizjak <ubizjak@gmail.com>
18155 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
18156 Handle V4QI and V2HI modes.
18157 (expand_vec_perm_blend): Allow 4-byte vector modes with TARGET_SSE4_1.
18158 Handle V4QI mode. Emit mmx_pblendvb32 for 4-byte modes.
18159 (expand_vec_perm_pshufb): Rewrite to use switch statemets.
18160 Handle 4-byte dual operands with TARGET_XOP and single operands
18161 with TARGET_SSSE3. Emit mmx_ppermv32 for TARGET_XOP and
18162 mmx_pshufbv4qi3 for TARGET_SSSE3.
18163 (expand_vec_perm_pblendv): Allow 4-byte vector modes with TARGET_SSE4_1.
18164 (expand_vec_perm_interleave2): Allow 4-byte vector modes.
18165 (expand_vec_perm_pshufb2): Allow 4-byte vector modes with TARGET_SSSE3.
18166 (expand_vec_perm_even_odd_1): Handle V4QI mode.
18167 (expand_vec_perm_broadcast_1): Handle V4QI mode.
18168 (ix86_vectorize_vec_perm_const): Handle V4QI mode.
18169 * config/i386/mmx.md (mmx_ppermv32): New insn pattern.
18170 (mmx_pshufbv4qi3): Ditto.
18171 (*mmx_pblendw32): Ditto.
18172 (*mmx_pblendw64): Rename from *mmx_pblendw.
18173 (mmx_punpckhbw_low): New insn_and_split pattern.
18174 (mmx_punpcklbw_low): Ditto.
18176 2021-07-05 Richard Biener <rguenther@suse.de>
18178 * tree-vect-loop-manip.c (vect_loop_versioning): Do not
18179 set LOOP_C_INFINITE on the vectorized loop.
18181 2021-07-05 Richard Biener <rguenther@suse.de>
18183 PR middle-end/101291
18184 * cfgloopmanip.c (loop_version): Set the loop copy of the
18185 versioned loop to the new loop.
18187 2021-07-04 Iain Sandoe <iain@sandoe.co.uk>
18190 * config.gcc: Ensure that Darwin biarch definitions are
18191 added before i386.h.
18192 * config/i386/darwin.h (TARGET_64BIT): Remove.
18193 (PR80556_WORKAROUND): New.
18194 (REAL_LIBGCC_SPEC): Amend to use PR80556_WORKAROUND.
18195 (DARWIN_SUBARCH_SPEC): New.
18196 * config/i386/darwin32-biarch.h (TARGET_64BIT_DEFAULT,
18197 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
18198 (REAL_LIBGCC_SPEC): Remove.
18199 * config/i386/darwin64-biarch.h (TARGET_64BIT_DEFAULT,
18200 TARGET_BI_ARCH, PR80556_WORKAROUND): New.
18201 (REAL_LIBGCC_SPEC): Remove.
18203 2021-07-03 H.J. Lu <hjl.tools@gmail.com>
18205 PR middle-end/101294
18206 * expr.c (store_constructor): Don't use vec_duplicate on vector.
18208 2021-07-02 Martin Sebor <msebor@redhat.com>
18210 PR middle-end/98871
18211 PR middle-end/98512
18212 * diagnostic.c (get_any_inlining_info): New.
18213 (update_effective_level_from_pragmas): Handle inlining context.
18214 (diagnostic_enabled): Same.
18215 (diagnostic_report_diagnostic): Same.
18216 * diagnostic.h (struct diagnostic_info): Add ctor.
18217 (struct diagnostic_context): Add new member.
18218 * tree-diagnostic.c (set_inlining_locations): New.
18219 (tree_diagnostics_defaults): Set new callback pointer.
18221 2021-07-02 Peter Bergner <bergner@linux.ibm.com>
18223 * config/rs6000/rs6000-builtin.def (BU_MMA_PAIR_LD, BU_MMA_PAIR_ST):
18225 (__builtin_vsx_lxvp, __builtin_vsx_stxvp): New built-ins.
18226 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Expand
18227 lxvp and stxvp built-ins.
18228 (mma_init_builtins): Handle lxvp and stxvp built-ins.
18229 (builtin_function_type): Likewise.
18230 * doc/extend.texi (__builtin_vsx_lxvp, __builtin_mma_stxvp): Document.
18232 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
18234 * config/h8300/h8300-protos.h (compute_a_shift_cc): Accept
18235 additional argument for the code.
18236 * config/h8300/h8300.c (compute_a_shift_cc): Accept additional
18237 argument for the code. Just return if the ZN bits are useful or
18238 not rather than the old style CC_* enums.
18239 * config/h8300/shiftrotate.md (shiftqi_noscratch): Move before
18240 more generic shiftqi patterns.
18241 (shifthi_noscratch, shiftsi_noscratch): Similarly.
18242 (shiftqi_noscratch_set_flags): New pattern.
18243 (shifthi_noscratch_set_flags, shiftsi_noscratch_set_flags): Likewise.
18245 2021-07-02 Andrew MacLeod <amacleod@redhat.com>
18247 PR tree-optimization/101223
18248 * range-op.cc (build_lt): Add -1 for signed values.
18249 (built_gt): Subtract -1 for signed values.
18251 2021-07-02 David Faust <david.faust@oracle.com>
18253 * btfout.c (get_btf_kind): Support BTF_KIND_FLOAT.
18254 (btf_asm_type): Likewise.
18256 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
18258 * config/h8300/h8300-protos.h (output_a_shift): Make first argument
18259 an array of rtx rather than a pointer to rtx. Add code argument.
18260 (compute_a_shift_length): Similarly.
18261 * config/h8300/h8300.c (h8300_shift_costs): Adjust now that the
18262 shift itself isn't an operand. Create dummy operand[0] to carry
18263 a mode and pass a suitable rtx code to compute_a_shift_length.
18264 (get_shift_alg): Adjust operand number of clobber in output templates.
18265 (output_a_shift): Make first argument an array of rtx rather than
18266 a pointer to rtx. Add code argument for the type of shift.
18267 Adjust now that the shift itself is no longer an operand.
18268 (compute_a_shift_length): Similarly.
18269 * config/h8300/shiftrotate.md (shiftqi, shifthi, shiftsi): Use an
18270 iterator rather than nshift_operator.
18271 (shiftqi_noscratch, shifthi_noscratch, shiftsi_noscratch): Likewise.
18272 (shiftqi_clobber_flags): Adjust to API changes in output_a_shift
18273 and compute_a_shift_length.
18274 (shiftqi_noscratch_clobber_flags): Likewise.
18275 (shifthi_noscratch_clobber_flags): Likewise.
18276 (shiftsi_noscratch_clobber_flags): Likewise.
18278 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
18281 * config/darwin.h (DSYMUTIL_SPEC): Do not try to run
18282 dsymutil for BTF/CTF.
18284 2021-07-02 Iain Sandoe <iain@sandoe.co.uk>
18287 * config/darwin.h (CTF_INFO_SECTION_NAME): Update the
18288 segment to include BTF.
18289 (BTF_INFO_SECTION_NAME): New.
18291 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
18293 * config/m32r/m32r-protos.h (call_operand): Adjust return type.
18294 (small_data_operand, memreg_operand, small_insn_p): Likewise.
18295 * config/m32r/m32r.c (call_operand): Adjust return type.
18296 (small_data_operand, memreg_operand): Likewise.
18298 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
18300 * config/frv/frv-protos.h (integer_register_operand): Adjust return
18302 (frv_load_operand, gpr_or_fpr_operand, gpr_no_subreg_operand): Likewise.
18303 (fpr_or_int6_operand, gpr_or_int_operand); Likewise.
18304 (gpr_or_int12_operand, gpr_or_int10_operand); Likewise.
18305 (move_source_operand, move_destination_operand): Likewise.
18306 (condexec_source_operand, condexec_dest_operand): Likewise.
18307 (lr_operand, gpr_or_memory_operand, fpr_or_memory_operand): Likewise.
18308 (reg_or_0_operand, fcc_operand, icc_operand, cc_operand): Likewise.
18309 (fcr_operand, icr_operand, cr_operand, call_operand): Likewise.
18310 (fpr_operand, even_reg_operand, odd_reg_operand): Likewise.
18311 (even_gpr_operand, odd_gpr_operand, quad_fpr_operand): Likewise.
18312 (even_fpr_operand, odd_fpr_operand): Likewise.
18313 (dbl_memory_one_insn_operand, dbl_memory_two_insn_operand): Likewise.
18314 (int12_operand, int6_operand, int5_operand, uint5_operand): Likewise.
18315 (uint4_operand, uint1_operand, int_2word_operand): Likewise
18316 (upper_int16_operand, uint16_operand, symbolic_operand): Likewise.
18317 (relational_operator, float_relational_operator): Likewise.
18318 (ccr_eqne_operator, minmax_operator): Likewise.
18319 (condexec_si_binary_operator, condexec_si_media_operator): Likewise.
18320 (condexec_si_divide_operator, condexec_si_unary_operator): Likewise.
18321 (condexec_sf_conv_operator, condexec_sf_add_operator): Likewise.
18322 (intop_compare_operator, acc_operand, even_acc_operand): Likewise.
18323 (quad_acc_operand, accg_operand): Likewise.
18325 2021-07-02 Jeff Law <jeffreyalaw@gmail.com>
18327 * config/stormy16/stormy16-protos.h (xstormy16_below_100_symbol): Change
18328 return type to a bool.
18329 (nonimmediate_nonstack_operand): Likewise.
18330 (xstormy16_splittable_below100_operand): Likewise.
18331 * config/stormy16/stormy16.c (xstormy16_below_100_symbol): Fix
18333 (xstormy16_splittable_below100_operand): Likewise.
18335 2021-07-02 Richard Biener <rguenther@suse.de>
18337 PR tree-optimization/101293
18338 * tree-ssa-loop-im.c (mem_ref_hasher::equal): Compare MEM_REF bases
18339 with combined offsets.
18340 (gather_mem_refs_stmt): Hash MEM_REFs as if their offset were
18341 combined with the rest of the offset.
18343 2021-07-02 Eric Botcazou <ebotcazou@adacore.com>
18345 * config/i386/i386.c (asm_preferred_eh_data_format): Always use the
18346 PIC encodings for PE-COFF targets.
18348 2021-07-02 Jakub Jelinek <jakub@redhat.com>
18351 * config/i386/i386-expand.c (ix86_broadcast_from_integer_constant):
18352 Return nullptr for TImode inner mode.
18354 2021-07-02 Richard Biener <rguenther@suse.de>
18356 PR tree-optimization/101280
18357 PR tree-optimization/101173
18358 * gimple-loop-interchange.cc
18359 (tree_loop_interchange::valid_data_dependences): Properly
18360 guard all dependence checks with DDR_REVERSED_P or its
18363 2021-07-02 Hongyu Wang <hongyu.wang@intel.com>
18365 * config/i386/i386-expand.c (ix86_expand_builtin):
18366 Add branch to clear odata when ZF is set for asedecenc_expand
18367 and wideaesdecenc_expand.
18369 2021-07-02 Eugene Rozenfeld <erozen@microsoft.com>
18371 * config/i386/gcc-auto-profile: regenerate
18373 2021-07-02 liuhongt <hongtao.liu@intel.com>
18375 * config/i386/sse.md (trunc<mode><pmov_dst_4>2): Refined to ..
18376 (trunc<mode><pmov_dst_4_lower>2): this.
18378 2021-07-01 David Malcolm <dmalcolm@redhat.com>
18380 * diagnostic.h (diagnostic_context::m_file_cache): New field.
18381 * input.c (class fcache): Rename to...
18382 (class file_cache_slot): ...this, making most members private and
18383 prefixing fields with "m_".
18384 (file_cache_slot::get_file_path): New accessor.
18385 (file_cache_slot::get_use_count): New accessor.
18386 (file_cache_slot::missing_trailing_newline_p): New accessor.
18387 (file_cache_slot::inc_use_count): New.
18388 (fcache_buffer_size): Move to...
18389 (file_cache_slot::buffer_size): ...here.
18390 (fcache_line_record_size): Move to...
18391 (file_cache_slot::line_record_size): ...here.
18392 (fcache_tab): Delete, in favor of global_dc->m_file_cache.
18393 (fcache_tab_size): Move to file_cache::num_file_slots.
18394 (diagnostic_file_cache_init): Update for move of fcache_tab
18395 to global_dc->m_file_cache.
18396 (diagnostic_file_cache_fini): Likewise.
18397 (lookup_file_in_cache_tab): Convert to...
18398 (file_cache::lookup_file): ...this.
18399 (diagnostics_file_cache_forcibly_evict_file): Update for move of
18400 fcache_tab to global_dc->m_file_cache, moving most of
18401 implementation to...
18402 (file_cache::forcibly_evict_file): ...this new function and...
18403 (file_cache_slot::evict): ...this new function.
18404 (evicted_cache_tab_entry): Convert to...
18405 (file_cache::evicted_cache_tab_entry): ...this.
18406 (add_file_to_cache_tab): Convert to...
18407 (file_cache::add_file): ...this, moving bulk of implementation
18409 (file_cache_slot::create): ..this new function.
18410 (file_cache::file_cache): New.
18411 (file_cache::~file_cache): New.
18412 (lookup_or_add_file_to_cache_tab): Convert to...
18413 (file_cache::lookup_or_add_file): ..this new function.
18414 (fcache::fcache): Rename to...
18415 (file_cache_slot::file_cache_slot): ...this, adding "m_" prefixes
18417 (fcache::~fcache): Rename to...
18418 (file_cache_slot::~file_cache_slot): ...this, adding "m_" prefixes
18420 (needs_read): Convert to...
18421 (file_cache_slot::needs_read_p): ...this.
18422 (needs_grow): Convert to...
18423 (file_cache_slot::needs_grow_p): ...this.
18424 (maybe_grow): Convert to...
18425 (file_cache_slot::maybe_grow): ...this.
18426 (read_data): Convert to...
18427 (file_cache_slot::read_data): ...this.
18428 (maybe_read_data): Convert to...
18429 (file_cache_slot::maybe_read_data): ...this.
18430 (get_next_line): Convert to...
18431 (file_cache_slot::get_next_line): ...this.
18432 (goto_next_line): Convert to...
18433 (file_cache_slot::goto_next_line): ...this.
18434 (read_line_num): Convert to...
18435 (file_cache_slot::read_line_num): ...this.
18436 (location_get_source_line): Update for moving of globals to
18437 global_dc->m_file_cache.
18438 (location_missing_trailing_newline): Likewise.
18439 * input.h (class file_cache_slot): New forward decl.
18440 (class file_cache): New.
18442 2021-07-01 Michael Meissner <meissner@linux.ibm.com>
18444 * config/rs6000/rs6000.c (rs6000_maybe_emit_fp_cmove): Add IEEE
18445 128-bit floating point conditional move support.
18446 (have_compare_and_set_mask): Add IEEE 128-bit floating point
18448 * config/rs6000/rs6000.md (mov<mode>cc, IEEE128 iterator): New insn.
18449 (mov<mode>cc_p10, IEEE128 iterator): New insn.
18450 (mov<mode>cc_invert_p10, IEEE128 iterator): New insn.
18451 (fpmask<mode>, IEEE128 iterator): New insn.
18452 (xxsel<mode>, IEEE128 iterator): New insn.
18454 2021-07-01 Iain Sandoe <iain@sandoe.co.uk>
18457 * config/darwin.h (CTF_INFO_SECTION_NAME): New.
18459 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
18461 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
18463 * config/i386/i386-protos.h (ix86_expand_vector_init_duplicate):
18465 * config/i386/sse.md (INT_BROADCAST_MODE): New mode iterator.
18466 (vec_duplicate<mode>): New expander.
18468 2021-07-01 H.J. Lu <hjl.tools@gmail.com>
18471 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
18473 (ix86_byte_broadcast): New function.
18474 (ix86_convert_const_wide_int_to_broadcast): Likewise.
18475 (ix86_expand_move): Convert CONST_WIDE_INT to broadcast if mode
18476 size is 16 bytes or bigger.
18477 (ix86_broadcast_from_integer_constant): New function.
18478 (ix86_expand_vector_move): Convert CONST_WIDE_INT and CONST_VECTOR
18479 to broadcast if mode size is 16 bytes or bigger.
18480 * config/i386/i386-protos.h (ix86_gen_scratch_sse_rtx): New
18482 * config/i386/i386.c (ix86_gen_scratch_sse_rtx): New function.
18484 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
18486 * config/i386/predicates.md (ix86_endbr_immediate_operand):
18487 Return true/false instead of 1/0.
18488 (movq_parallel): Ditto.
18490 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
18492 * recog.c (general_operand): Return true/false instead of 1/0.
18493 (register_operand): Ditto.
18494 (immediate_operand): Ditto.
18495 (const_int_operand): Ditto.
18496 (const_scalar_int_operand): Ditto.
18497 (const_double_operand): Ditto.
18498 (push_operand): Ditto.
18499 (pop_operand): Ditto.
18500 (memory_operand): Ditto.
18501 (indirect_operand): Ditto.
18503 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
18505 * genpreds.c (write_predicate_subfunction):
18506 Change the type of written subfunction to bool.
18507 (write_one_predicate_function):
18508 Change the type of written function to bool.
18509 (write_tm_preds_h): Ditto.
18510 * recog.h (*insn_operand_predicate_fn): Change the type to bool.
18511 * recog.c (general_operand): Change the type to bool.
18512 (address_operand): Ditto.
18513 (register_operand): Ditto.
18514 (pmode_register_operand): Ditto.
18515 (scratch_operand): Ditto.
18516 (immediate_operand): Ditto.
18517 (const_int_operand): Ditto.
18518 (const_scalar_int_operand): Ditto.
18519 (const_double_operand): Ditto.
18520 (nonimmediate_operand): Ditto.
18521 (nonmemory_operand): Ditto.
18522 (push_operand): Ditto.
18523 (pop_operand): Ditto.
18524 (memory_operand): Ditto.
18525 (indirect_operand): Ditto.
18526 (ordered_comparison_operator): Ditto.
18527 (comparison_operator): Ditto.
18528 * config/i386/i386-expand.c (ix86_expand_sse_cmp):
18529 Change the type of indirect predicate function to bool.
18530 * config/rs6000/rs6000.c (easy_vector_constant):
18531 Change the type to bool.
18532 * config/mips/mips-protos.h (m16_based_address_p):
18533 Change the type of operand 3 to bool.
18535 2021-07-01 Richard Biener <rguenther@suse.de>
18537 PR tree-optimization/101280
18538 PR tree-optimization/101173
18539 * gimple-loop-interchange.cc
18540 (tree_loop_interchange::valid_data_dependences): Revert
18541 previous change and instead correctly handle DDR_REVERSED_P
18544 2021-07-01 Richard Biener <rguenther@suse.de>
18546 PR tree-optimization/101278
18547 * tree-ssa-dse.c (dse_classify_store): First check for
18548 uses, then ignore stmt for chaining purposes.
18550 2021-07-01 Richard Biener <rguenther@suse.de>
18552 PR tree-optimization/100778
18553 * tree-vect-slp.c (vect_schedule_slp_node): Do not place trapping
18554 vectorized ops ahead of their scalar BB.
18556 2021-07-01 Uroš Bizjak <ubizjak@gmail.com>
18559 * config/i386/i386.md (*nabs<dwi>2_doubleword):
18560 New insn_and_split pattern.
18561 (*nabs<dwi>2_1): Ditto.
18562 * config/i386/i386-features.c
18563 (general_scalar_chain::compute_convert_gain):
18564 Handle (NEG (ABS (...))) RTX. Rewrite src code
18565 scanner as switch statement.
18566 (general_scalar_chain::convert_insn):
18567 Handle (NEG (ABS (...))) RTX.
18568 (general_scalar_to_vector_candidate_p):
18569 Detect (NEG (ABS (...))) RTX. Reorder case statements
18570 for (AND (NOT (...) ...)) fallthrough.
18572 2021-07-01 Richard Biener <rguenther@suse.de>
18574 PR tree-optimization/101178
18575 * tree-vect-slp.c (slpg_vertex::materialize): Remove.
18576 (slpg::perm_in): Add.
18577 (slpg::get_perm_in): Remove.
18578 (slpg::get_perm_materialized): Add.
18579 (vect_optimize_slp): Handle VEC_PERM nodes more optimally
18580 during permute propagation and materialization.
18582 2021-07-01 Jakub Jelinek <jakub@redhat.com>
18585 * dwarf2out.c (loc_list_from_tree_1): Handle COMPOUND_LITERAL_EXPR.
18587 2021-07-01 Jakub Jelinek <jakub@redhat.com>
18589 PR middle-end/94366
18590 * omp-low.c (lower_rec_input_clauses): Rename is_fp_and_or to
18591 is_truth_op, set it for TRUTH_*IF_EXPR regardless of new_var's type,
18592 use boolean_type_node instead of integer_type_node as NE_EXPR type.
18593 (lower_reduction_clauses): Likewise.
18595 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
18597 * config/gcn/gcn.c: Include dwarf2.h.
18598 (gcn_addr_space_debug): New function.
18599 (TARGET_ADDR_SPACE_DEBUG): New hook.
18601 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
18603 * common/config/gcn/gcn-common.c
18604 (gcn_option_optimization_table): Change OPT_fomit_frame_pointer to -O3.
18605 * config/gcn/gcn.c (gcn_expand_prologue): Prefer the frame pointer
18607 (gcn_expand_prologue): Prefer the frame pointer when emitting CFI.
18608 (gcn_frame_pointer_rqd): New function.
18609 (TARGET_FRAME_POINTER_REQUIRED): New hook.
18611 2021-06-30 Hafiz Abid Qadeer <abidh@codesourcery.com>
18613 * config/gcn/gcn.c (move_callee_saved_registers): Emit CFI notes for
18614 prologue register saves.
18615 (gcn_debug_unwind_info): Use UI_DWARF2.
18616 (gcn_dwarf_register_number): Map DWARF_LINK_REGISTER to DWARF PC.
18617 (gcn_dwarf_register_span): DWARF_LINK_REGISTER doesn't span.
18618 * config/gcn/gcn.h: (DWARF_FRAME_RETURN_COLUMN): New define.
18619 (DWARF_LINK_REGISTER): New define.
18620 (FIRST_PSEUDO_REGISTER): Increment.
18621 (FIXED_REGISTERS): Add entry for DWARF_LINK_REGISTER.
18622 (CALL_USED_REGISTERS): Likewise.
18623 (REGISTER_NAMES): Likewise.
18625 2021-06-30 Richard Biener <rguenther@suse.de>
18627 PR tree-optimization/101267
18628 * tree-vect-stmts.c (vect_check_scalar_mask): Adjust
18629 API and use SLP compatible interface of vect_is_simple_use.
18630 Reject not vectorized SLP defs for callers that do not support
18632 (vect_check_store_rhs): Handle masked stores and pass down
18633 the appropriate operator index.
18634 (vectorizable_call): Adjust.
18635 (vectorizable_store): Likewise.
18636 (vectorizable_load): Likewise. Handle SLP pecularity of
18638 (vect_is_simple_use): Remove special-casing of masked stores.
18640 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
18642 * common.opt (foffload): Remove help as Driver only.
18643 * gcc.c (display_help): Add -foffload.
18645 2021-06-30 Tobias Burnus <tobias@codesourcery.com>
18647 * gcc.c (close_at_file, execute): Replace alloca by XALLOCAVEC.
18648 (check_offload_target_name): Fix splitting OFFLOAD_TARGETS into
18649 a candidate list; better inform no offload target is configured
18650 and fix hint extraction when passed target is not '\0' at [len].
18651 * common.opt (foffload): Add tailing '.'.
18652 (foffload-options): Likewise; fix flag name in the help string.
18654 2021-06-30 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
18657 * config/arm/arm_neon.h: Move vabs intrinsics before vcage_f32.
18658 (vcage_f32): Gate comparison on __FAST_MATH__.
18659 (vcageq_f32): Likewise.
18660 (vcale_f32): Likewise.
18661 (vcaleq_f32): Likewise.
18662 (vcagt_f32): Likewise.
18663 (vcagtq_f32): Likewise.
18664 (vcalt_f32): Likewise.
18665 (vcaltq_f32): Likewise.
18666 (vcage_f16): Likewise.
18667 (vcageq_f16): Likewise.
18668 (vcale_f16): Likewise.
18669 (vcaleq_f16): Likewise.
18670 (vcagt_f16): Likewise.
18671 (vcagtq_f16): Likewise.
18672 (vcalt_f16): Likewise.
18673 (vcaltq_f16): Likewise.
18675 2021-06-30 Richard Biener <rguenther@suse.de>
18677 PR tree-optimization/101264
18678 * tree-vect-slp.c (vect_optimize_slp): Propagate the
18679 computed perm_in to all "any" permute successors
18680 we cannot de-duplicate immediately.
18682 2021-06-30 liuhongt <hongtao.liu@intel.com>
18685 * config/i386/sse.md
18686 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>):
18688 (avx512f_sfixupimm<mode><maskz_scalar_name><round_saeonly_name>):
18690 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>"): Refined.
18691 * config/i386/subst.md (maskz_scalar): New define_subst.
18692 (maskz_scalar_name): New subst_attr.
18693 (maskz_scalar_op5): Ditto.
18694 (round_saeonly_maskz_scalar_op5): Ditto.
18695 (round_saeonly_maskz_scalar_operand5): Ditto.
18697 2021-06-30 David Edelsohn <dje.gcc@gmail.com>
18699 * config/rs6000/rs6000.c (rs6000_xcoff_section_type_flags):
18700 Increase code CSECT alignment to at least 32 bytes.
18701 * config/rs6000/xcoff.h (TEXT_SECTION_ASM_OP): Add 32 byte
18702 alignment designation.
18704 2021-06-29 Sergei Trofimovich <siarheit@google.com>
18706 * doc/generic.texi: Fix s/net yet/not yet/ typo.
18708 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
18710 PR tree-optimization/101254
18711 * range-op.cc (operator_minus::op1_op2_relation_effect): Check for
18712 wrapping/non-wrapping when setting the result range.
18714 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
18716 * value-query.cc (gimple_range_global): Allow phis.
18718 2021-06-29 Andrew MacLeod <amacleod@redhat.com>
18720 * vr-values.c (vr_values::vrp_stmt_computes_nonzero): Use stmt.
18721 (simplify_using_ranges::op_with_boolean_value_range_p): Add a
18722 statement for location context.
18723 (check_for_binary_op_overflow): Ditto.
18724 (simplify_using_ranges::get_vr_for_comparison): Ditto.
18725 (simplify_using_ranges::compare_name_with_value): Ditto.
18726 (simplify_using_ranges::compare_names): Ditto.
18727 (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Ditto.
18728 (simplify_using_ranges::simplify_truth_ops_using_ranges): Ditto.
18729 (simplify_using_ranges::simplify_min_or_max_using_ranges): Ditto.
18730 (simplify_using_ranges::simplify_internal_call_using_ranges): Ditto.
18731 (simplify_using_ranges::two_valued_val_range_p): Ditto.
18732 (simplify_using_ranges::simplify): Ditto.
18733 * vr-values.h: Adjust prototypes.
18735 2021-06-29 Uroš Bizjak <ubizjak@gmail.com>
18738 * config/i386/mmx.md (vec_addsubv2sf3): New insn pattern.
18740 2021-06-29 Julian Brown <julian@codesourcery.com>
18742 * config/gcn/gcn.c (gcn_init_libfuncs): New function.
18743 (TARGET_INIT_LIBFUNCS): Define target hook using above function.
18744 * config/gcn/gcn.h (UNITS_PER_WORD): Define to 8 for IN_LIBGCC2, 4
18746 (LIBGCC2_UNITS_PER_WORD, BITS_PER_WORD): Remove definitions.
18747 (MAX_FIXED_MODE_SIZE): Change to 128.
18749 2021-06-29 Julian Brown <julian@codesourcery.com>
18751 * config/gcn/gcn.md (UNSPEC_FLBIT_INT): New unspec constant.
18752 (s_mnemonic): Add clrsb.
18753 (gcn_flbit<mode>_int): Add insn pattern for SImode/DImode.
18754 (clrsb<mode>2): Add expander for SImode/DImode.
18756 2021-06-29 Julian Brown <julian@codesourcery.com>
18758 * config/gcn/gcn.md (<su>mulsidi3, <su>mulsidi3_reg, <su>mulsidi3_imm,
18759 muldi3): Add patterns.
18761 2021-06-29 Julian Brown <julian@codesourcery.com>
18763 * config/gcn/gcn.md (<su>mulsi3_highpart): Change to expander.
18764 (<su>mulsi3_highpart_reg, <su>mulsi3_highpart_imm): New patterns.
18766 2021-06-29 Julian Brown <julian@codesourcery.com>
18768 * config/gcn/gcn.md (mulsi3): Make s_mulk_i32 variant clobber SCC.
18770 2021-06-29 Joseph Myers <joseph@codesourcery.com>
18772 * btfout.c, ctfout.c: Include "memmodel.h".
18774 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
18776 * gcc.c (check_offload_target_name): Cast len argument to
18777 %q.*s to 'int'; avoid -Wstringop-truncation warning.
18779 2021-06-29 Richard Biener <rguenther@suse.de>
18781 * tree-vect-slp.c (vect_optimize_slp): Forward propagate
18782 to "any" permute nodes and relax "any" permute proapgation
18783 during iterative backward propagation.
18785 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
18788 * common.opt (-foffload=): Update description.
18789 (-foffload-options=): New.
18790 * doc/invoke.texi (C Language Options): Document
18791 -foffload and -foffload-options.
18792 * gcc.c (check_offload_target_name): New, split off from
18793 handle_foffload_option.
18794 (check_foffload_target_names): New.
18795 (handle_foffload_option): Handle -foffload=default.
18796 (driver_handle_option): Update for -foffload-options.
18797 * lto-opts.c (lto_write_options): Use -foffload-options
18798 instead of -foffload.
18799 * lto-wrapper.c (merge_and_complain, append_offload_options):
18801 * opts.c (common_handle_option): Likewise.
18803 2021-06-29 Tobias Burnus <tobias@codesourcery.com>
18805 * doc/invoke.texi (C Language Options): Sort options
18806 alphabetically in optlist and also the description itself.
18807 Remove leftover -fallow-single-precision from and add missing
18808 -fgnu-tm to the optlist.
18810 2021-06-29 Richard Biener <rguenther@suse.de>
18812 * tree-vect-slp.c (slpg_vertex::visited): Remove.
18813 (vect_slp_perms_eq): Handle -1 permutes.
18814 (vect_optimize_slp): Rewrite permute propagation.
18816 2021-06-29 Jakub Jelinek <jakub@redhat.com>
18819 * match.pd ((intptr_t)x eq/ne CST to x eq/ne (typeof x) CST): Don't
18820 perform the optimization in GENERIC when sanitizing and x has a
18823 2021-06-29 Richard Biener <rguenther@suse.de>
18825 PR tree-optimization/101242
18826 * tree-vect-slp.c (vect_slp_build_vertices): Force-add
18827 PHIs with not represented initial values as leafs.
18829 2021-06-29 Jan-Benedict Glaw <jbglaw@getslash.de>
18831 * config/pdp11/pdp11.h (ASM_OUTPUT_SKIP): Fix signedness warning.
18832 * config/pdp11/pdp11.c (pdp11_asm_print_operand_punct_valid_p): Remove
18833 "register" keyword.
18834 (pdp11_initial_elimination_offset) Remove unused variable.
18835 (pdp11_cmp_length) Ditto.
18836 (pdp11_insn_cost): Ditto, and fix signedness warning.
18838 2021-06-29 David Edelsohn <dje.gcc@gmail.com>
18840 * btfout.c: Include tm_p.h.
18843 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
18845 * config/bpf/bpf.c (bpf_expand_prologue): Do not mark insns as
18847 (bpf_expand_epilogue): Likewise.
18848 * config/bpf/bpf.h (DWARF2_FRAME_INFO): Define to 0.
18849 Do not define DBX_DEBUGGING_INFO.
18851 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
18853 * doc/invoke.texi: Document the CTF and BTF debug info options.
18855 2021-06-28 Indu Bhagat <indu.bhagat@oracle.com>
18856 David Faust <david.faust@oracle.com>
18857 Jose E. Marchesi <jose.marchesi@oracle.com>
18858 Weimin Pan <weimin.pan@oracle.com>
18860 * Makefile.in: Add ctfc.*, ctfout.c and btfout.c files to
18861 GTFILES. Add new object files.
18862 * common.opt: Add CTF and BTF debug info options.
18863 * btfout.c: New file.
18864 * ctfc.c: Likewise.
18865 * ctfc.h: Likewise.
18866 * ctfout.c: Likewise.
18867 * dwarf2ctf.c: Likewise.
18868 * dwarf2ctf.h: Likewise.
18869 * dwarf2cfi.c (dwarf2out_do_frame): Acknowledge CTF_DEBUG and
18871 * dwarf2out.c (dwarf2out_source_line): Likewise.
18872 (dwarf2out_finish): Skip emitting DWARF if CTF or BTF are to
18874 (debug_format_do_cu): New function.
18875 (dwarf2out_early_finish): Traverse DIEs and emit CTF/BTF for
18877 Include dwarf2ctf.c.
18878 * final.c (dwarf2_debug_info_emitted_p): Acknowledge DWARF-based debug
18880 * flag-types.h (enum debug_info_type): Add CTF_DEBUG and BTF_DEBUG.
18881 (CTF_DEBUG): New bitmask.
18882 (BTF_DEBUG): Likewise.
18883 (enum ctf_debug_info_levels): New enum.
18884 * gengtype.c (open_base_files): Handle ctfc.h.
18885 (main): Handle uint32_t type.
18886 * flags.h (btf_debuginfo_p): New definition.
18887 (dwarf_based_debuginfo_p): Likewise.
18888 * opts.c (debug_type_names): Add entries for CTF and BTF.
18889 (btf_debuginfo_p): New function.
18890 (dwarf_based_debuginfo_p): Likewise.
18891 (common_handle_option): Handle -gctfN and -gbtf options.
18892 (set_debug_level): Set CTF_DEBUG, BTF_DEBUG whenever appropriate.
18893 * toplev.c (process_options): Inform the user and ignore -gctfLEVEL if
18896 2021-06-28 Jose E. Marchesi <jose.marchesi@oracle.com>
18898 * dwarf2out.c (AT_class): Function is no longer static.
18899 (AT_int): Likewise.
18900 (AT_unsigned): Likewise.
18901 (AT_loc): Likewise.
18902 (get_AT): Likewise.
18903 (get_AT_string): Likewise.
18904 (get_AT_flag): Likewise.
18905 (get_AT_unsigned): Likewise.
18906 (get_AT_ref): Likewise.
18907 (new_die_raw): Likewise.
18908 (lookup_decl_die): Likewise.
18909 (base_type_die): Likewise.
18910 (add_name_attribute): Likewise.
18911 (add_AT_int): Likewise.
18912 (add_AT_unsigned): Likewise.
18913 (add_AT_loc): Likewise.
18914 (dw_get_die_tag): New function.
18915 (dw_get_die_child): Likewise.
18916 (dw_get_die_sib): Likewise.
18917 (struct dwarf_file_data): Move from here to dwarf2out.h
18918 (struct dw_attr_struct): Likewise.
18919 * dwarf2out.h: Analogous changes.
18921 2021-06-28 Martin Jambor <mjambor@suse.cz>
18924 * ipa-param-manipulation.h (class ipa_param_body_adjustments): New
18925 members m_dead_stmts and m_dead_ssas.
18926 * ipa-param-manipulation.c
18927 (ipa_param_body_adjustments::mark_dead_statements): New function.
18928 (ipa_param_body_adjustments::common_initialization): Call it on
18929 all removed but not split parameters.
18930 (ipa_param_body_adjustments::ipa_param_body_adjustments): Initialize
18932 (ipa_param_body_adjustments::modify_call_stmt): Remove arguments that
18934 * tree-inline.c (remap_gimple_stmt): Do not copy dead statements, reset
18935 dead debug statements.
18936 (copy_phis_for_bb): Do not copy dead PHI nodes.
18938 2021-06-28 Martin Jambor <mjambor@suse.cz>
18941 * symtab-clones.h (clone_info): Removed member param_adjustments.
18942 * ipa-param-manipulation.h: Adjust initial comment to reflect how we
18943 deal with pass-through splits now.
18944 (ipa_param_performed_split): Removed.
18945 (ipa_param_adjustments::modify_call): Adjusted parameters.
18946 (class ipa_param_body_adjustments): Adjusted parameters of
18947 register_replacement, modify_gimple_stmt and modify_call_stmt.
18948 (ipa_verify_edge_has_no_modifications): Declare.
18949 (ipa_edge_modifications_finalize): Declare.
18950 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Remove
18951 performed_splits processing, pas only edge to padjs->modify_call,
18952 check that call arguments were not modified if they should not have
18954 * cgraphclones.c (cgraph_node::create_clone): Do not copy performed
18956 * ipa-param-manipulation.c (struct pass_through_split_map): New type.
18957 (ipa_edge_modification_info): Likewise.
18958 (ipa_edge_modification_sum): Likewise.
18959 (ipa_edge_modifications): New edge summary.
18960 (ipa_verify_edge_has_no_modifications): New function.
18961 (transitive_split_p): Removed.
18962 (transitive_split_map): Likewise.
18963 (init_transitive_splits): Likewise.
18964 (ipa_param_adjustments::modify_call): Adjusted to use the new edge
18965 summary instead of performed_splits.
18966 (ipa_param_body_adjustments::register_replacement): Drop dummy
18967 parameter, set base_index of the created ipa_param_body_replacement.
18968 (phi_arg_will_live_p): New function.
18969 (ipa_param_body_adjustments::common_initialization): Do not create
18970 IPA_SRA dummy decls.
18971 (simple_tree_swap_info): Removed.
18972 (remap_split_decl_to_dummy): Likewise.
18973 (record_argument_state_1): New function.
18974 (record_argument_state): Likewise.
18975 (ipa_param_body_adjustments::modify_call_stmt): New parameter
18976 orig_stmt. Do not work with dummy decls, save necessary info about
18977 changes to ipa_edge_modifications.
18978 (ipa_param_body_adjustments::modify_gimple_stmt): New parameter
18979 orig_stmt, pass it to modify_call_stmt.
18980 (ipa_param_body_adjustments::modify_cfun_body): Adjust call to
18981 modify_gimple_stmt.
18982 (ipa_edge_modifications_finalize): New function.
18983 * tree-inline.c (remap_gimple_stmt): Pass original statement to
18984 modify_gimple_stmt.
18985 (copy_phis_for_bb): Do not copy dead PHI nodes.
18986 (expand_call_inline): Do not remap performed_splits.
18987 (update_clone_info): Likewise.
18988 * toplev.c: Include ipa-param-manipulation.h.
18989 (toplev::finalize): Call ipa_edge_modifications_finalize.
18991 2021-06-28 Andrew Pinski <apinski@marvell.com>
18993 * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Duplicate range
18994 info if we're the only things setting the target PHI.
18995 (value_replacement): Don't duplicate range here.
18996 (minmax_replacement): Likewise.
18998 2021-06-28 Richard Biener <rguenther@suse.de>
19000 PR tree-optimization/101229
19001 * gimple-walk.c (gimple_walk_op): Handle PHIs.
19003 2021-06-28 Martin Liska <mliska@suse.cz>
19005 * config/v850/v850.c (construct_dispose_instruction): Allocate
19007 (construct_prepare_instruction): Likewise.
19009 2021-06-28 Martin Liska <mliska@suse.cz>
19011 * config/v850/v850.c (v850_option_override): Build default
19013 (v850_can_inline_p): New. Allow MASK_PROLOG_FUNCTION to be
19014 ignored for inlining.
19015 (TARGET_CAN_INLINE_P): New.
19017 2021-06-28 Richard Biener <rguenther@suse.de>
19019 PR tree-optimization/101207
19020 * tree-vect-slp.c (vect_optimize_slp): Do BB reduction
19021 permute eliding for load permutations properly.
19023 2021-06-28 Richard Biener <rguenther@suse.de>
19025 PR tree-optimization/101173
19026 * gimple-loop-interchange.cc
19027 (tree_loop_interchange::valid_data_dependences): Disallow outer
19028 loop dependence distance of zero.
19030 2021-06-28 liuhongt <hongtao.liu@intel.com>
19033 * config/i386/sse.md (*avx_cmp<mode>3_lt): New
19034 define_insn_and_split.
19035 (*avx_cmp<mode>3_ltint): Ditto.
19036 (*avx2_pcmp<mode>3_3): Ditto.
19037 (*avx2_pcmp<mode>3_4): Ditto.
19038 (*avx2_pcmp<mode>3_5): Ditto.
19040 2021-06-28 liuhongt <hongtao.liu@intel.com>
19042 * config/i386/i386-builtin.def (IX86_BUILTIN_BLENDVPD256,
19043 IX86_BUILTIN_BLENDVPS256, IX86_BUILTIN_PBLENDVB256,
19044 IX86_BUILTIN_BLENDVPD, IX86_BUILTIN_BLENDVPS,
19045 IX86_BUILTIN_PBLENDVB128): Replace icode with
19047 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold blendv
19049 * config/i386/sse.md (*<sse4_1_avx2>_pblendvb_lt_subreg_not):
19050 New pre_reload splitter.
19052 2021-06-27 Andrew Pinski <apinski@marvell.com>
19054 PR middle-end/101230
19055 * fold-const.c (fold_ternary_loc): Check
19056 the return value of invert_tree_comparison.
19058 2021-06-27 David Edelsohn <dje.gcc@gmail.com>
19060 * config.gcc: Add SPDX License Identifier.
19061 (powerpc-ibm-aix789): Default to aix73.h.
19062 (powerpc-ibm-aix7.2.*.*): New stanza.
19063 * config/rs6000/aix72.h: Add SPDX License Identifier.
19064 * config/rs6000/aix73.h: New file.
19066 2021-06-26 Jason Merrill <jason@redhat.com>
19068 * except.c: #include "dwarf2.h" instead of "dwarf2out.h".
19070 2021-06-26 Andrew Pinski <apinski@marvell.com>
19072 * genmatch.c (lower_cond): Copy for_subst_vec
19073 for the simplify also.
19074 (lower): Swap the order for lower_for and lower_cond.
19076 2021-06-26 Andrew Pinski <apinski@marvell.com>
19078 * tree-ssa-phiopt.c (match_simplify_replacement): Reset
19079 flow senatitive info on the moved ssa set.
19081 2021-06-26 Andrew Pinski <apinski@marvell.com>
19083 * fold-const.c (fold_cond_expr_with_comparison):
19084 Exand arg0 into comp_code, arg00, and arg01.
19085 (fold_ternary_loc): Use invert_tree_comparison
19086 instead of fold_invert_truthvalue for the case
19087 where we have A CMP B ? C : A.
19089 2021-06-25 Martin Sebor <msebor@redhat.com>
19091 PR middle-end/101216
19092 * calls.c (maybe_warn_rdwr_sizes): Use the no_warning constant.
19094 2021-06-25 Jeff Law <jeffreyalaw@gmail.com>
19096 * config/h8300/h8300.c (select_cc_mode): Handle ASHIFTRT and LSHIFTRT.
19098 2021-06-25 Richard Biener <rguenther@suse.de>
19100 PR tree-optimization/101202
19101 * tree-vect-slp.c (vect_optimize_slp): Explicitely handle
19104 2021-06-25 Richard Biener <rguenther@suse.de>
19106 * tree-vect-slp-patterns.c (addsub_pattern::build): Copy
19107 STMT_VINFO_REDUC_DEF from the original representative.
19109 2021-06-25 Martin Sebor <msebor@redhat.com>
19111 * builtins.c (warn_string_no_nul): Replace uses of TREE_NO_WARNING,
19112 gimple_no_warning_p and gimple_set_no_warning with
19113 warning_suppressed_p, and suppress_warning.
19115 (maybe_warn_for_bound): Same.
19116 (warn_for_access): Same.
19117 (check_access): Same.
19118 (expand_builtin_strncmp): Same.
19119 (fold_builtin_varargs): Same.
19120 * calls.c (maybe_warn_nonstring_arg): Same.
19121 (maybe_warn_rdwr_sizes): Same.
19122 * cfgexpand.c (expand_call_stmt): Same.
19123 * cgraphunit.c (check_global_declaration): Same.
19124 * fold-const.c (fold_undefer_overflow_warnings): Same.
19125 (fold_truth_not_expr): Same.
19126 (fold_unary_loc): Same.
19127 (fold_checksum_tree): Same.
19128 * gimple-array-bounds.cc (array_bounds_checker::check_array_ref): Same.
19129 (array_bounds_checker::check_mem_ref): Same.
19130 (array_bounds_checker::check_addr_expr): Same.
19131 (array_bounds_checker::check_array_bounds): Same.
19132 * gimple-expr.c (copy_var_decl): Same.
19133 * gimple-fold.c (gimple_fold_builtin_strcpy): Same.
19134 (gimple_fold_builtin_strncat): Same.
19135 (gimple_fold_builtin_stxcpy_chk): Same.
19136 (gimple_fold_builtin_stpcpy): Same.
19137 (gimple_fold_builtin_sprintf): Same.
19138 (fold_stmt_1): Same.
19139 * gimple-ssa-isolate-paths.c (diag_returned_locals): Same.
19140 * gimple-ssa-nonnull-compare.c (do_warn_nonnull_compare): Same.
19141 * gimple-ssa-sprintf.c (handle_printf_call): Same.
19142 * gimple-ssa-store-merging.c (imm_store_chain_info::output_merged_store): Same.
19143 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Same.
19144 * gimple-ssa-warn-restrict.h: Adjust declarations.
19145 (maybe_diag_access_bounds): Replace uses of TREE_NO_WARNING,
19146 gimple_no_warning_p and gimple_set_no_warning with
19147 warning_suppressed_p, and suppress_warning.
19148 (check_call): Same.
19149 (check_bounds_or_overlap): Same.
19150 * gimple.c (gimple_build_call_from_tree): Same.
19151 * gimplify.c (gimplify_return_expr): Same.
19152 (gimplify_cond_expr): Same.
19153 (gimplify_modify_expr_complex_part): Same.
19154 (gimplify_modify_expr): Same.
19155 (gimple_push_cleanup): Same.
19156 (gimplify_expr): Same.
19157 * omp-expand.c (expand_omp_for_generic): Same.
19158 (expand_omp_taskloop_for_outer): Same.
19159 * omp-low.c (lower_rec_input_clauses): Same.
19160 (lower_lastprivate_clauses): Same.
19161 (lower_send_clauses): Same.
19162 (lower_omp_target): Same.
19163 * tree-cfg.c (pass_warn_function_return::execute): Same.
19164 * tree-complex.c (create_one_component_var): Same.
19165 * tree-inline.c (remap_gimple_op_r): Same.
19166 (copy_tree_body_r): Same.
19167 (declare_return_variable): Same.
19168 (expand_call_inline): Same.
19169 * tree-nested.c (lookup_field_for_decl): Same.
19170 * tree-sra.c (create_access_replacement): Same.
19171 (generate_subtree_copies): Same.
19172 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Same.
19173 * tree-ssa-forwprop.c (combine_cond_expr_cond): Same.
19174 * tree-ssa-loop-ch.c (ch_base::copy_headers): Same.
19175 * tree-ssa-loop-im.c (execute_sm): Same.
19176 * tree-ssa-phiopt.c (cond_store_replacement): Same.
19177 * tree-ssa-strlen.c (maybe_warn_overflow): Same.
19178 (handle_builtin_strcpy): Same.
19179 (maybe_diag_stxncpy_trunc): Same.
19180 (handle_builtin_stxncpy_strncat): Same.
19181 (handle_builtin_strcat): Same.
19182 * tree-ssa-uninit.c (get_no_uninit_warning): Same.
19183 (set_no_uninit_warning): Same.
19184 (uninit_undefined_value_p): Same.
19185 (warn_uninit): Same.
19186 (maybe_warn_operand): Same.
19187 * tree-vrp.c (compare_values_warnv): Same.
19188 * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr): Same.
19189 (test_for_singularity): Same.
19190 * gimple.h (warning_suppressed_p): New function.
19191 (suppress_warning): Same.
19192 (copy_no_warning): Same.
19193 (gimple_set_block): Call gimple_set_location.
19194 (gimple_set_location): Call copy_warning.
19196 2021-06-25 Martin Sebor <msebor@redhat.com>
19198 * tree.h (warning_suppressed_at, copy_warning,
19199 warning_suppressed_p, suppress_warning): New functions.
19201 2021-06-25 Martin Sebor <msebor@redhat.com>
19203 * Makefile.in (OBJS-libcommon): Add diagnostic-spec.o.
19204 * gengtype.c (open_base_files): Add diagnostic-spec.h.
19205 * diagnostic-spec.c: New file.
19206 * diagnostic-spec.h: New file.
19207 * tree.h (no_warning, all_warnings, suppress_warning_at): New
19209 * warning-control.cc: New file.
19211 2021-06-25 liuhongt <hongtao.liu@intel.com>
19214 * config/i386/i386.c (x86_order_regs_for_local_alloc):
19217 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
19219 PR tree-optimization/101189
19220 * gimple-range-fold.cc (fold_using_range::range_of_range_op): Pass
19221 LHS range of condition to postfold routine.
19222 (fold_using_range::postfold_gcond_edges): Only process the TRUE or
19223 FALSE edge if the LHS range supports it being taken.
19224 * gimple-range-fold.h (postfold_gcond_edges): Add range parameter.
19226 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
19228 * value-relation.cc (equiv_oracle::dump): Do not dump NULL blocks.
19229 (relation_oracle::find_relation_block): Check correct bitmap.
19230 (relation_oracle::dump): Do not dump NULL blocks.
19232 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
19234 * gimple-range-cache.cc (ranger_cache::propagate_cache): Call
19235 range_on_edge instead of manually calculating.
19237 2021-06-24 Andrew MacLeod <amacleod@redhat.com>
19239 * range-op.cc: Fix comment.
19241 2021-06-24 Uroš Bizjak <ubizjak@gmail.com>
19244 * config/i386/i386-expand.c (ix86_expand_sse_unpack):
19245 Handle V8QI and V4HI modes.
19246 * config/i386/mmx.md (sse4_1_<any_extend:code>v4qiv4hi2):
19248 (sse4_1_<any_extend:code>v4qiv4hi2): Ditto.
19249 (mmxpackmode): New mode attribute.
19250 (vec_pack_trunc_<mmxpackmode:mode>): New expander.
19251 (mmxunpackmode): New mode attribute.
19252 (vec_unpacks_lo_<mmxunpackmode:mode>): New expander.
19253 (vec_unpacks_hi_<mmxunpackmode:mode>): Ditto.
19254 (vec_unpacku_lo_<mmxunpackmode:mode>): Ditto.
19255 (vec_unpacku_hi_<mmxunpackmode:mode>): Ditto.
19256 * config/i386/i386.md (extsuffix): Move from ...
19257 * config/i386/sse.md: ... here.
19259 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
19261 * dwarf2out.c (dwarf2out_assembly_start): Emit .file 0 marker here..
19262 (dwarf2out_finish): ...instead of here.
19264 2021-06-24 Eric Botcazou <ebotcazou@adacore.com>
19266 * configure.ac (--gdwarf-5 option): Use objdump instead of readelf.
19267 (working --gdwarf-4/--gdwarf-5 for all sources): Likewise.
19268 (--gdwarf-4 not refusing generated .debug_line): Adjust for Windows.
19269 * configure: Regenerate.
19271 2021-06-24 Richard Biener <rguenther@suse.de>
19273 * config/i386/sse.md (vec_addsubv4df3, vec_addsubv2df3,
19274 vec_addsubv8sf3, vec_addsubv4sf3): Merge into ...
19275 (vec_addsub<mode>3): ... using a new addsub_cst mode attribute.
19277 2021-06-24 Richard Biener <rguenther@suse.de>
19279 * config/i386/sse.md (avx_addsubv4df3): Rename to
19281 (avx_addsubv8sf3): Rename to vec_addsubv8sf3.
19282 (sse3_addsubv2df3): Rename to vec_addsubv2df3.
19283 (sse3_addsubv4sf3): Rename to vec_addsubv4sf3.
19284 * config/i386/i386-builtin.def: Adjust.
19285 * internal-fn.def (VEC_ADDSUB): New internal optab fn.
19286 * optabs.def (vec_addsub_optab): New optab.
19287 * tree-vect-slp-patterns.c (class addsub_pattern): New.
19288 (slp_patterns): Add addsub_pattern.
19289 * tree-vect-slp.c (vect_optimize_slp): Disable propagation
19290 across CFN_VEC_ADDSUB.
19291 * tree-vectorizer.h (vect_pattern::vect_pattern): Make
19293 * doc/md.texi (vec_addsub<mode>3): Document.
19295 2021-06-24 Jakub Jelinek <jakub@redhat.com>
19297 PR middle-end/101170
19298 * df-scan.c (df_ref_record): For paradoxical big-endian SUBREGs
19299 where regno + subreg_regno_offset wraps around use 0 as starting
19302 2021-06-24 Jakub Jelinek <jakub@redhat.com>
19304 PR middle-end/101172
19305 * stor-layout.c (finish_bitfield_representative): If nextf has
19306 error_mark_node type, set repr type to error_mark_node too.
19308 2021-06-24 Ilya Leoshkevich <iii@linux.ibm.com>
19310 * config/s390/s390.c (s390_function_profiler): Ignore labelno
19312 * config/s390/s390.h (NO_PROFILE_COUNTERS): Define.
19314 2021-06-24 Richard Biener <rguenther@suse.de>
19316 * tree-vect-slp.c (vect_optimize_slp): Do not propagate
19317 across operations that have different semantics on different
19320 2021-06-24 Jakub Jelinek <jakub@redhat.com>
19322 * tree.h (OMP_CLAUSE_MAP_IN_REDUCTION): Document meaning for OpenMP.
19323 * gimplify.c (gimplify_scan_omp_clauses): For OpenMP map clauses
19324 with OMP_CLAUSE_MAP_IN_REDUCTION flag partially defer gimplification
19325 of non-decl OMP_CLAUSE_DECL. For OMP_CLAUSE_IN_REDUCTION on
19326 OMP_TARGET user outer_ctx instead of ctx for placeholders and
19327 initializer/combiner gimplification.
19328 * omp-low.c (scan_sharing_clauses): Handle OMP_CLAUSE_MAP_IN_REDUCTION
19329 on target constructs.
19330 (lower_rec_input_clauses): Likewise.
19331 (lower_omp_target): Likewise.
19332 * omp-expand.c (expand_omp_target): Temporarily ignore nowait clause
19333 on target if in_reduction is present.
19335 2021-06-24 Kewen Lin <linkw@linux.ibm.com>
19337 * tree-predcom.c (class pcom_worker): New class.
19338 (release_chain): Renamed to...
19339 (pcom_worker::release_chain): ...this.
19340 (release_chains): Renamed to...
19341 (pcom_worker::release_chains): ...this.
19342 (aff_combination_dr_offset): Renamed to...
19343 (pcom_worker::aff_combination_dr_offset): ...this.
19344 (determine_offset): Renamed to...
19345 (pcom_worker::determine_offset): ...this.
19346 (class comp_ptrs): New class.
19347 (split_data_refs_to_components): Renamed to...
19348 (pcom_worker::split_data_refs_to_components): ...this,
19349 and update with class comp_ptrs.
19350 (suitable_component_p): Renamed to...
19351 (pcom_worker::suitable_component_p): ...this.
19352 (filter_suitable_components): Renamed to...
19353 (pcom_worker::filter_suitable_components): ...this.
19354 (valid_initializer_p): Renamed to...
19355 (pcom_worker::valid_initializer_p): ...this.
19356 (find_looparound_phi): Renamed to...
19357 (pcom_worker::find_looparound_phi): ...this.
19358 (add_looparound_copies): Renamed to...
19359 (pcom_worker::add_looparound_copies): ...this.
19360 (determine_roots_comp): Renamed to...
19361 (pcom_worker::determine_roots_comp): ...this.
19362 (determine_roots): Renamed to...
19363 (pcom_worker::determine_roots): ...this.
19364 (single_nonlooparound_use): Renamed to...
19365 (pcom_worker::single_nonlooparound_use): ...this.
19366 (remove_stmt): Renamed to...
19367 (pcom_worker::remove_stmt): ...this.
19368 (execute_pred_commoning_chain): Renamed to...
19369 (pcom_worker::execute_pred_commoning_chain): ...this.
19370 (execute_pred_commoning): Renamed to...
19371 (pcom_worker::execute_pred_commoning): ...this.
19372 (struct epcc_data): New member worker.
19373 (execute_pred_commoning_cbck): Call execute_pred_commoning
19374 with pcom_worker pointer.
19375 (find_use_stmt): Renamed to...
19376 (pcom_worker::find_use_stmt): ...this.
19377 (find_associative_operation_root): Renamed to...
19378 (pcom_worker::find_associative_operation_root): ...this.
19379 (find_common_use_stmt): Renamed to...
19380 (pcom_worker::find_common_use_stmt): ...this.
19381 (combinable_refs_p): Renamed to...
19382 (pcom_worker::combinable_refs_p): ...this.
19383 (reassociate_to_the_same_stmt): Renamed to...
19384 (pcom_worker::reassociate_to_the_same_stmt): ...this.
19385 (stmt_combining_refs): Renamed to...
19386 (pcom_worker::stmt_combining_refs): ...this.
19387 (combine_chains): Renamed to...
19388 (pcom_worker::combine_chains): ...this.
19389 (try_combine_chains): Renamed to...
19390 (pcom_worker::try_combine_chains): ...this.
19391 (prepare_initializers_chain): Renamed to...
19392 (pcom_worker::prepare_initializers_chain): ...this.
19393 (prepare_initializers): Renamed to...
19394 (pcom_worker::prepare_initializers): ...this.
19395 (prepare_finalizers_chain): Renamed to...
19396 (pcom_worker::prepare_finalizers_chain): ...this.
19397 (prepare_finalizers): Renamed to...
19398 (pcom_worker::prepare_finalizers): ...this.
19399 (tree_predictive_commoning_loop): Renamed to...
19400 (pcom_worker::tree_predictive_commoning_loop): ...this, adjust
19401 some calls and remove some cleanup code.
19402 (tree_predictive_commoning): Adjusted to use pcom_worker instance.
19403 (static variable looparound_phis): Remove.
19404 (static variable name_expansions): Remove.
19406 2021-06-24 Richard Biener <rguenther@suse.de>
19408 * tree-vect-slp.c (slpg_vertex): New struct.
19409 (vect_slp_build_vertices): Adjust.
19410 (vect_optimize_slp): Likewise. Maintain an outgoing permute
19411 and a materialized one.
19413 2021-06-24 Richard Biener <rguenther@suse.de>
19415 PR tree-optimization/101105
19416 * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list):
19417 Only ignore steps when they are equal or scalar order is preserved.
19419 2021-06-24 liuhongt <hongtao.liu@intel.com>
19422 * config/i386/i386-expand.c (ix86_expand_vec_interleave):
19423 Adjust comments for ix86_expand_vecop_qihi2.
19424 (ix86_expand_vecmul_qihi): Renamed to ..
19425 (ix86_expand_vecop_qihi2): Adjust function prototype to
19426 support shift operation, add static to definition.
19427 (ix86_expand_vec_shift_qihi_constant): Add static to definition.
19428 (ix86_expand_vecop_qihi): Call ix86_expand_vecop_qihi2 and
19429 ix86_expand_vec_shift_qihi_constant.
19430 * config/i386/i386-protos.h (ix86_expand_vecmul_qihi): Deleted.
19431 (ix86_expand_vec_shift_qihi_constant): Deleted.
19432 * config/i386/sse.md (VI12_256_512_AVX512VL): New mode
19434 (mulv8qi3): Call ix86_expand_vecop_qihi directly, add
19435 condition TARGET_64BIT.
19436 (mul<mode>3): Ditto.
19437 (<insn><mode>3): Ditto.
19438 (vlshr<mode>3): Extend to support avx512 vlshr.
19439 (v<insn><mode>3): New expander for
19441 (v<insn>v8qi3): Ditto.
19442 (vashrv8hi3<mask_name>): Renamed to ..
19443 (vashr<mode>3): And extend to support V16QImode for avx512.
19444 (vashrv16qi3): Deleted.
19445 (vashrv2di3<mask_name>): Extend expander to support avx512
19448 2021-06-23 Dimitar Dimitrov <dimitar@dinux.eu>
19450 * doc/lto.texi (Design Overview): Update that slim objects are
19453 2021-06-23 Aaron Sawdey <acsawdey@linux.ibm.com>
19455 * config/rs6000/rs6000-cpus.def: Take OPTION_MASK_PCREL_OPT out
19456 of OTHER_POWER10_MASKS so it will not be enabled by default.
19458 2021-06-23 Richard Biener <rguenther@suse.de>
19459 Martin Jambor <mjambor@suse.cz>
19461 * tree-inline.c (setup_one_parameter): Set TREE_READONLY of the
19462 param replacement unconditionally. Adjust comment.
19464 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
19466 * Makefile.in (OBJS): Add gimple-range-fold.o
19467 * gimple-range-fold.cc: New.
19468 * gimple-range-fold.h: New.
19469 * gimple-range-gori.cc (gimple_range_calc_op1): Move to here.
19470 (gimple_range_calc_op2): Ditto.
19471 * gimple-range-gori.h: Move prototypes to here.
19472 * gimple-range.cc: Adjust include files.
19473 (fur_source:fur_source): Relocate to gimple-range-fold.cc.
19474 (fur_source::get_operand): Ditto.
19475 (fur_source::get_phi_operand): Ditto.
19476 (fur_source::query_relation): Ditto.
19477 (fur_source::register_relation): Ditto.
19478 (class fur_edge): Ditto.
19479 (fur_edge::fur_edge): Ditto.
19480 (fur_edge::get_operand): Ditto.
19481 (fur_edge::get_phi_operand): Ditto.
19482 (fur_stmt::fur_stmt): Ditto.
19483 (fur_stmt::get_operand): Ditto.
19484 (fur_stmt::get_phi_operand): Ditto.
19485 (fur_stmt::query_relation): Ditto.
19486 (class fur_depend): Relocate to gimple-range-fold.h.
19487 (fur_depend::fur_depend): Relocate to gimple-range-fold.cc.
19488 (fur_depend::register_relation): Ditto.
19489 (fur_depend::register_relation): Ditto.
19490 (class fur_list): Ditto.
19491 (fur_list::fur_list): Ditto.
19492 (fur_list::get_operand): Ditto.
19493 (fur_list::get_phi_operand): Ditto.
19494 (fold_range): Ditto.
19495 (adjust_pointer_diff_expr): Ditto.
19496 (gimple_range_adjustment): Ditto.
19497 (gimple_range_base_of_assignment): Ditto.
19498 (gimple_range_operand1): Ditto.
19499 (gimple_range_operand2): Ditto.
19500 (gimple_range_calc_op1): Relocate to gimple-range-gori.cc.
19501 (gimple_range_calc_op2): Ditto.
19502 (fold_using_range::fold_stmt): Relocate to gimple-range-fold.cc.
19503 (fold_using_range::range_of_range_op): Ditto.
19504 (fold_using_range::range_of_address): Ditto.
19505 (fold_using_range::range_of_phi): Ditto.
19506 (fold_using_range::range_of_call): Ditto.
19507 (fold_using_range::range_of_builtin_ubsan_call): Ditto.
19508 (fold_using_range::range_of_builtin_call): Ditto.
19509 (fold_using_range::range_of_cond_expr): Ditto.
19510 (fold_using_range::range_of_ssa_name_with_loop_info): Ditto.
19511 (fold_using_range::relation_fold_and_or): Ditto.
19512 (fold_using_range::postfold_gcond_edges): Ditto.
19513 * gimple-range.h: Add gimple-range-fold.h to include files. Change
19514 GIMPLE_RANGE_STMT_H to GIMPLE_RANGE_H.
19515 (gimple_range_handler): Relocate to gimple-range-fold.h.
19516 (gimple_range_ssa_p): Ditto.
19517 (range_compatible_p): Ditto.
19518 (class fur_source): Ditto.
19519 (class fur_stmt): Ditto.
19520 (class fold_using_range): Ditto.
19521 (gimple_range_calc_op1): Relocate to gimple-range-gori.h
19522 (gimple_range_calc_op2): Ditto.
19524 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
19526 PR tree-optimization/101148
19527 PR tree-optimization/101014
19528 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
19529 (ranger_cache::~ranger_cache): Adjust.
19530 (ranger_cache::block_range): Check if propagation disallowed.
19531 (ranger_cache::propagate_cache): Disallow propagation if new value
19532 can't be stored properly.
19533 * gimple-range-cache.h (ranger_cache::m_propfail): New member.
19535 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
19537 * gimple-range-cache.cc (class ssa_block_ranges): Adjust prototype.
19538 (sbr_vector::set_bb_range): Return true.
19539 (class sbr_sparse_bitmap): Adjust.
19540 (sbr_sparse_bitmap::set_bb_range): Return value.
19541 (block_range_cache::set_bb_range): Return value.
19542 (ranger_cache::propagate_cache): Use return value to print msg.
19543 * gimple-range-cache.h (class block_range_cache): Adjust.
19545 2021-06-23 Andrew MacLeod <amacleod@redhat.com>
19547 * gimple-range.cc (dump_bb): Use range_on_edge from the cache.
19549 2021-06-23 Jeff Law <jeffreyalaw@gmail.com>
19551 * config/h8300/logical.md (<code><mode>3<ccnz>): Use <cczn>
19552 so this pattern can be used for test/compare removal. Pass
19553 current insn to compute_logical_op_length and output_logical_op.
19554 * config/h8300/h8300.c (compute_logical_op_cc): Remove.
19555 (h8300_and_costs): Add argument to compute_logical_op_length.
19556 (output_logical_op): Add new argument. Use it to determine if the
19557 condition codes are used and adjust the output accordingly.
19558 (compute_logical_op_length): Add new argument and update length
19559 computations when condition codes are used.
19560 * config/h8300/h8300-protos.h (compute_logical_op_length): Update
19562 (output_logical_op): Likewise.
19564 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
19567 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
19568 Handle 64bit modes for TARGET_XOP. Use indirect gen_* functions.
19569 * config/i386/mmx.md (mmx_ppermv64): New insn pattern.
19570 * config/i386/i386.md (unspec): Move UNSPEC_XOP_PERMUTE from ...
19571 * config/i386/sse.md (unspec): ... here.
19573 2021-06-23 Martin Liska <mliska@suse.cz>
19576 * optc-save-gen.awk: Put back arm_fp16_format to
19579 2021-06-23 Uroš Bizjak <ubizjak@gmail.com>
19582 * config/i386/i386.md (bsr_rex64): Add zero-flag setting RTX.
19585 (clz<mode>2): Update RTX pattern for additions.
19587 2021-06-23 Jakub Jelinek <jakub@redhat.com>
19589 PR middle-end/101167
19590 * omp-low.c (lower_omp_regimplify_p): Regimplify also PARM_DECLs
19591 and RESULT_DECLs that have DECL_HAS_VALUE_EXPR_P set.
19593 2021-06-22 Sergei Trofimovich <siarheit@google.com>
19595 * doc/rtl.texi: drop unbalanced parenthesis.
19597 2021-06-22 Richard Biener <rguenther@suse.de>
19599 PR middle-end/101156
19600 * gimplify.c (gimplify_expr): Remove premature incorrect
19603 2021-06-22 Jakub Jelinek <jakub@redhat.com>
19605 PR tree-optimization/101159
19606 * tree-vect-patterns.c (vect_recog_popcount_pattern): Fix some
19609 2021-06-22 Jakub Jelinek <jakub@redhat.com>
19611 PR middle-end/101160
19612 * function.c (assign_parms): For decl_result with TYPE_EMPTY_P type
19613 clear crtl->return_rtx instead of keeping it referencing a pseudo.
19615 2021-06-22 Jakub Jelinek <jakub@redhat.com>
19616 Andrew Pinski <apinski@marvell.com>
19618 PR tree-optimization/101162
19619 * fold-const.c (range_check_type): Handle OFFSET_TYPE like pointer
19622 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
19624 * range-op.cc (range_relational_tests): New.
19625 (range_op_tests): Call range_relational_tests.
19627 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
19629 * range-op.cc (operator_cast::lhs_op1_relation): New.
19630 (operator_identity::lhs_op1_relation): Mew.
19632 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
19634 * range-op.cc (operator_minus::op1_op2_relation_effect): New.
19636 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
19638 * range-op.cc (operator_plus::lhs_op1_relation): New.
19639 (operator_plus::lhs_op2_relation): New.
19641 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
19643 * gimple-range-cache.cc (ranger_cache::ranger_cache): Create a
19644 relation_oracle if dominators exist.
19645 (ranger_cache::~ranger_cache): Dispose of oracle.
19646 (ranger_cache::dump_bb): Dump oracle.
19647 * gimple-range.cc (fur_source::fur_source): New.
19648 (fur_source::get_operand): Use mmeber query.
19649 (fur_source::get_phi_operand): Use member_query.
19650 (fur_source::query_relation): New.
19651 (fur_source::register_dependency): Delete.
19652 (fur_source::register_relation): New.
19653 (fur_edge::fur_edge): Adjust.
19654 (fur_edge::get_phi_operand): Fix comment.
19655 (fur_edge::query): Delete.
19656 (fur_stmt::fur_stmt): Adjust.
19657 (fur_stmt::query): Delete.
19658 (fur_depend::fur_depend): Adjust.
19659 (fur_depend::register_relation): New.
19660 (fur_depend::register_relation): New.
19661 (fur_list::fur_list): Adjust.
19662 (fur_list::get_operand): Use member query.
19663 (fold_using_range::range_of_range_op): Process and query relations.
19664 (fold_using_range::range_of_address): Adjust dependency call.
19665 (fold_using_range::range_of_phi): Ditto.
19666 (gimple_ranger::gimple_ranger): New. Use ranger_ache oracle.
19667 (fold_using_range::relation_fold_and_or): New.
19668 (fold_using_range::postfold_gcond_edges): New.
19669 * gimple-range.h (class gimple_ranger): Adjust.
19670 (class fur_source): Adjust members.
19671 (class fur_stmt): Ditto.
19672 (class fold_using_range): Ditto.
19674 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
19676 * range-op.cc (range_operator::wi_fold): Apply relation effect.
19677 (range_operator::fold_range): Adjust and apply relation effect.
19678 (*::fold_range): Add relation parameters.
19679 (*::op1_range): Ditto.
19680 (*::op2_range): Ditto.
19681 (range_operator::lhs_op1_relation): New.
19682 (range_operator::lhs_op2_relation): New.
19683 (range_operator::op1_op2_relation): New.
19684 (range_operator::op1_op2_relation_effect): New.
19685 (relop_early_resolve): New.
19686 (operator_equal::op1_op2_relation): New.
19687 (operator_equal::fold_range): Call relop_early_resolve.
19688 (operator_not_equal::op1_op2_relation): New.
19689 (operator_not_equal::fold_range): Call relop_early_resolve.
19690 (operator_lt::op1_op2_relation): New.
19691 (operator_lt::fold_range): Call relop_early_resolve.
19692 (operator_le::op1_op2_relation): New.
19693 (operator_le::fold_range): Call relop_early_resolve.
19694 (operator_gt::op1_op2_relation): New.
19695 (operator_gt::fold_range): Call relop_early_resolve.
19696 (operator_ge::op1_op2_relation): New.
19697 (operator_ge::fold_range): Call relop_early_resolve.
19698 * range-op.h (class range_operator): Adjust parameters and methods.
19700 2021-06-22 Andrew MacLeod <amacleod@redhat.com>
19702 * Makefile.in (OBJS): Add value-relation.o.
19703 * gimple-range.h: Adjust include files.
19704 * tree-data-ref.c: Adjust include file order.
19705 * value-query.cc (range_query::get_value_range): Default to no oracle.
19706 (range_query::query_relation): New.
19707 (range_query::query_relation): New.
19708 * value-query.h (class range_query): Adjust.
19709 * value-relation.cc: New.
19710 * value-relation.h: New.
19712 2021-06-22 Richard Biener <rguenther@suse.de>
19714 PR tree-optimization/101151
19715 * tree-ssa-sink.c (statement_sink_location): Expand irreducible
19718 2021-06-22 Jojo R <rjiejie@linux.alibaba.com>
19720 * config/riscv/riscv.c (thead_c906_tune_info): New.
19721 (riscv_tune_info_table): Use new tune.
19723 2021-06-22 Richard Biener <rguenther@suse.de>
19725 PR tree-optimization/101158
19726 * tree-vect-slp.c (vect_build_slp_tree_1): Move same operand
19727 checking after checking for matching operation.
19729 2021-06-22 Richard Biener <rguenther@suse.de>
19731 PR tree-optimization/101159
19732 * tree-vect-patterns.c (vect_recog_popcount_pattern): Add
19733 missing NULL vectype check.
19735 2021-06-22 Richard Biener <rguenther@suse.de>
19737 PR tree-optimization/101154
19738 * tree-vect-slp.c (vect_build_slp_tree_2): Fix out-of-bound access.
19740 2021-06-22 Jakub Jelinek <jakub@redhat.com>
19743 * config/i386/i386-protos.h (ix86_last_zero_store_uid): Declare.
19744 * config/i386/i386-expand.c (ix86_last_zero_store_uid): New variable.
19745 * config/i386/i386.c (ix86_expand_prologue): Clear it.
19746 * config/i386/i386.md (peephole2s for 1/2/4 stores of const0_rtx):
19747 Remove "" from match_operand. Emit new insns using emit_move_insn and
19748 set ix86_last_zero_store_uid to INSN_UID of the last store.
19749 Add peephole2s for 1/2/4 stores of const0_rtx following previous
19752 2021-06-22 Martin Liska <mliska@suse.cz>
19754 * auto-profile.c (AUTO_PROFILE_VERSION): Bump as string format
19757 2021-06-22 Martin Liska <mliska@suse.cz>
19759 * gcov-io.h: Remove padding entries.
19761 2021-06-22 liuhongt <hongtao.liu@intel.com>
19763 PR tree-optimization/97770
19764 * tree-vect-patterns.c (vect_recog_popcount_pattern):
19766 (vect_recog_func vect_vect_recog_func_ptrs): Add new pattern.
19768 2021-06-22 liuhongt <hongtao.liu@intel.com>
19771 * config/i386/i386-builtin.def (BDESC): Adjust builtin name.
19772 * config/i386/sse.md (<avx512>_expand<mode>_mask): Rename to ..
19773 (expand<mode>_mask): this ..
19774 (*expand<mode>_mask): New pre_reload splitter to transform
19775 v{,p}expand* to vmov* when mask is zero, all ones, or has all
19776 ones in it's lower part, otherwise still generate
19779 2021-06-22 liuhongt <hongtao.liu@intel.com>
19782 * config/i386/i386-expand.c
19783 (ix86_expand_special_args_builtin): Keep constm1_operand only
19784 if it satisfies insn's operand predicate.
19786 2021-06-21 Jason Merrill <jason@redhat.com>
19789 * df-scan.c (df_ref_record): Check that regno < endregno.
19790 * function.c (assign_parms, expand_function_end): Do nothing with a
19791 TYPE_EMPTY_P result.
19793 2021-06-21 Richard Biener <rguenther@suse.de>
19795 PR tree-optimization/101120
19796 * tree-vect-data-refs.c (bump_vector_ptr): Fold the
19798 * tree-vect-slp.c (vect_transform_slp_perm_load): Add
19799 DR chain DCE capability.
19800 * tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
19801 * tree-vect-stmts.c (vectorizable_load): Remove unused
19802 loads in the DR chain for SLP.
19804 2021-06-21 Jakub Jelinek <jakub@redhat.com>
19806 PR inline-asm/100785
19807 * gimplify.c (gimplify_asm_expr): Don't diagnose errors if
19808 output or input operands were already error_mark_node.
19809 * cfgexpand.c (expand_asm_stmt): If errors are emitted,
19810 remove all inputs, outputs and clobbers from the asm and
19811 set template to "".
19813 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
19815 * config/arm/arm_neon.h (vceq_s8): Replace builtin with __a == __b.
19816 (vceq_s16): Likewise.
19817 (vceq_s32): Likewise.
19818 (vceq_u8): Likewise.
19819 (vceq_u16): Likewise.
19820 (vceq_u32): Likewise.
19821 (vceq_p8): Likewise.
19822 (vceqq_s8): Likewise.
19823 (vceqq_s16): Likewise.
19824 (vceqq_s32): Likewise.
19825 (vceqq_u8): Likewise.
19826 (vceqq_u16): Likewise.
19827 (vceqq_u32): Likewise.
19828 (vceqq_p8): Likewise.
19829 (vceq_f32): Gate __a == __b on __FAST_MATH__.
19830 (vceqq_f32): Likewise.
19831 (vceq_f16): Likewise.
19832 (vceqq_f16): Likewise.
19834 2021-06-21 prathamesh.kulkarni <prathamesh.kulkarni@linaro.org>
19837 * config/arm/iterators.md (NEON_VACMP): Remove.
19838 * config/arm/neon.md (neon_vca<cmp_op><mode>): Use GLTE instead of GTGE
19840 (neon_vca<cmp_op><mode>_insn): Likewise.
19841 (neon_vca<cmp_op_unsp><mode>_insn_unspec): Use NEON_VAGLTE instead of
19844 2021-06-21 Richard Biener <rguenther@suse.de>
19846 PR tree-optimization/101121
19847 * tree-vect-slp.c (vect_build_slp_tree_2): To not fail fatally
19848 when we just lack a stmt with the desired op when doing permutation.
19849 (vect_build_slp_tree): When caching a failed SLP build attempt
19850 assert that at least one lane is marked as not matching.
19852 2021-06-21 liuhongt <hongtao.liu@intel.com>
19855 * config/i386/i386.md: (*anddi_1): Disparage slightly the mask
19856 register alternative.
19857 (*and<mode>_1): Ditto.
19859 (*andn<mode>_1): Ditto.
19860 (*<code><mode>_1): Ditto.
19861 (*<code>qi_1): Ditto.
19862 (*one_cmpl<mode>2_1): Ditto.
19863 (*one_cmplsi2_1_zext): Ditto.
19864 (*one_cmplqi2_1): Ditto.
19865 * config/i386/i386.c (x86_order_regs_for_local_alloc): Change
19866 the order of mask registers to be before general registers.
19868 2021-06-21 Roger Sayle <roger@nextmovesoftware.com>
19871 * config/i386/i386.md: New define_peephole2s to shrink writing
19872 1, 2 or 4 consecutive zeros to memory when optimizing for size.
19874 2021-06-18 Jeff Law <jeffreyalaw@gmail.com>
19876 * config/h8300/h8300.c (h8300_select_cc_mode): Handle SYMBOL_REF.
19877 * config/h8300/logical.md (<code><mode>3 logcial expander): Generate
19878 more efficient code when the source can be trivially simplified.
19880 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
19882 * gimple-range-cache.cc (ranger_cache::range_of_def): Calculate
19883 a range if global is not available.
19884 (ranger_cache::entry_range): Fallback to range_of_def.
19885 * gimple-range-cache.h (range_of_def): Adjust prototype.
19887 2021-06-18 Andrew MacLeod <amacleod@redhat.com>
19889 PR tree-optimization/101014
19890 * gimple-range-cache.cc (ranger_cache::ranger_cache): Remove poor
19892 (ranger_cache::~ranger_cache): Ditto.
19893 (ranger_cache::enable_new_values): Delete.
19894 (ranger_cache::push_poor_value): Delete.
19895 (ranger_cache::range_of_def): Remove poor value processing.
19896 (ranger_cache::entry_range): Ditto.
19897 (ranger_cache::fill_block_cache): Ditto.
19898 * gimple-range-cache.h (class ranger_cache): Remove poor value members.
19899 * gimple-range.cc (gimple_ranger::range_of_expr): Remove call.
19900 * gimple-range.h (class gimple_ranger): Adjust.
19902 2021-06-18 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
19905 * common/config/arm/arm-common.c (arm_canon_arch_option_1): New function
19906 derived from arm_canon_arch.
19907 (arm_canon_arch_option): Call it.
19908 (arm_canon_arch_multilib_option): New function.
19909 * config/arm/arm-cpus.in (IGNORE_FOR_MULTILIB): New fgroup.
19910 * config/arm/arm.h (arm_canon_arch_multilib_option): New prototype.
19911 (CANON_ARCH_MULTILIB_SPEC_FUNCTION): New macro.
19912 (MULTILIB_ARCH_CANONICAL_SPECS): New macro.
19913 (DRIVER_SELF_SPECS): Add MULTILIB_ARCH_CANONICAL_SPECS.
19914 * config/arm/arm.opt (mlibarch): New option.
19915 * config/arm/t-rmprofile (MULTILIB_MATCHES): For armv8*-m, replace use
19916 of march on RHS with mlibarch.
19918 2021-06-18 Marcel Vollweiler <marcel@codesourcery.com>
19920 * config.in: Regenerate.
19921 * config/gcn/gcn.c (print_operand_address): Fix for global_load assembler
19923 * configure: Regenerate.
19924 * configure.ac: Fix for global_load assembler functions.
19926 2021-06-18 Richard Biener <rguenther@suse.de>
19928 PR tree-optimization/101112
19929 * tree-vect-slp.c (vect_slp_linearize_chain): Fix condition
19930 to lookup a pattern stmt def.
19932 2021-06-18 Jakub Jelinek <jakub@redhat.com>
19934 PR middle-end/101062
19935 * stor-layout.c (finish_bitfield_layout): Don't add bitfield
19936 representatives in QUAL_UNION_TYPE.
19938 2021-06-18 Andrew Pinski <apinski@marvell.com>
19940 * tree-ssa-phiopt.c (replace_phi_edge_with_variable):
19941 Add counting of how many times it is done.
19942 (factor_out_conditional_conversion): Likewise.
19943 (match_simplify_replacement): Likewise.
19944 (value_replacement): Likewise.
19945 (spaceship_replacement): Likewise.
19946 (cond_store_replacement): Likewise.
19947 (cond_if_else_store_replacement_1): Likewise.
19948 (hoist_adjacent_loads): Likewise.
19950 2021-06-18 Andrew Pinski <apinski@marvell.com>
19952 * tree-cfg.c (verify_gimple_assign_unary): Reject point and offset
19953 types on NEGATE_EXPR, ABS_EXPR, BIT_NOT_EXPR, PAREN_EXPR and CNONJ_EXPR.
19954 (verify_gimple_assign_binary): Reject point and offset types on
19955 MULT_EXPR, MULT_HIGHPART_EXPR, TRUNC_DIV_EXPR, CEIL_DIV_EXPR,
19956 FLOOR_DIV_EXPR, ROUND_DIV_EXPR, TRUNC_MOD_EXPR, CEIL_MOD_EXPR,
19957 FLOOR_MOD_EXPR, ROUND_MOD_EXPR, RDIV_EXPR, and EXACT_DIV_EXPR.
19959 2021-06-18 Michael Meissner <meissner@linux.ibm.com>
19961 * config/rs6000/rs6000.c (rs6000_emit_minmax): Add support for ISA
19962 3.1 IEEE 128-bit floating point xsmaxcqp/xsmincqp instructions.
19963 * config/rs6000/rs6000.md (s<minmax><mode>3, IEEE128 iterator):
19966 2021-06-17 Aaron Sawdey <acsawdey@linux.ibm.com>
19968 * config/rs6000/genfusion.pl (gen_logical_addsubf): Add
19969 earlyclobber to alts 0/1.
19970 (gen_addadd): Add earlyclobber to alts 0/1.
19971 * config/rs6000/fusion.md: Regenerate file.
19973 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
19975 * cfgloopanal.c (get_loop_hot_path): Make path an auto_vec.
19977 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
19979 * gimple-range-cache.cc: Comment cleanups.
19980 * gimple-range-gori.cc: Comment cleanups.
19981 * gimple-range.cc: Comment/spacing cleanups
19982 * value-range.h: Comment cleanups.
19984 2021-06-17 H.J. Lu <hjl.tools@gmail.com>
19987 * calls.c (expand_call): Replace PUSH_ARGS with
19988 targetm.calls.push_argument (0).
19989 (emit_library_call_value_1): Likewise.
19990 * defaults.h (PUSH_ARGS): Removed.
19991 (PUSH_ARGS_REVERSED): Replace PUSH_ARGS with
19992 targetm.calls.push_argument (0).
19993 * expr.c (block_move_libcall_safe_for_call_parm): Likewise.
19994 (emit_push_insn): Pass the number bytes to push to
19995 targetm.calls.push_argument and pass 0 if ARGS_ADDR is 0.
19996 * hooks.c (hook_bool_uint_true): New.
19997 * hooks.h (hook_bool_uint_true): Likewise.
19998 * rtlanal.c (nonzero_bits1): Replace PUSH_ARGS with
19999 targetm.calls.push_argument (0).
20000 * target.def (push_argument): Add a targetm.calls hook.
20001 * targhooks.c (default_push_argument): New.
20002 * targhooks.h (default_push_argument): Likewise.
20003 * config/bpf/bpf.h (PUSH_ARGS): Removed.
20004 * config/cr16/cr16.c (TARGET_PUSH_ARGUMENT): New.
20005 * config/cr16/cr16.h (PUSH_ARGS): Removed.
20006 * config/i386/i386.c (ix86_push_argument): New.
20007 (TARGET_PUSH_ARGUMENT): Likewise.
20008 * config/i386/i386.h (PUSH_ARGS): Removed.
20009 * config/m32c/m32c.c (TARGET_PUSH_ARGUMENT): New.
20010 * config/m32c/m32c.h (PUSH_ARGS): Removed.
20011 * config/nios2/nios2.h (PUSH_ARGS): Likewise.
20012 * config/pru/pru.h (PUSH_ARGS): Likewise.
20013 * doc/tm.texi.in: Remove PUSH_ARGS documentation. Add
20014 TARGET_PUSH_ARGUMENT hook.
20015 * doc/tm.texi: Regenerated.
20017 2021-06-17 Uroš Bizjak <ubizjak@gmail.com>
20020 * config/i386/i386-expand.c (expand_vector_set_var):
20021 Handle V2FS mode remapping. Pass TARGET_MMX_WITH_SSE to
20022 ix86_expand_vector_init_duplicate.
20023 (ix86_expand_vector_init_duplicate): Emit insv_1 for
20024 QImode for !TARGET_PARTIAL_REG_STALL.
20025 * config/i386/predicates.md (vec_setm_mmx_operand): New predicate.
20026 * config/i386/mmx.md (vec_setv2sf): Use vec_setm_mmx_operand
20027 as operand 2 predicate. Call ix86_expand_vector_set_var
20028 for non-constant index operand.
20029 (vec_setv2si): Ditto.
20030 (vec_setv4hi): Ditto.
20031 (vec_setv8qi): ditto.
20033 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
20035 PR tree-optimization/100790
20036 * gimple-range.cc (range_of_builtin_call): Cleanup clz and ctz
20039 2021-06-17 Martin Liska <mliska@suse.cz>
20041 * doc/invoke.texi: Use consistently -O1 instead of -O.
20043 2021-06-17 Martin Liska <mliska@suse.cz>
20045 * gcov-io.h: Update documentation entry about string format.
20047 2021-06-17 Marius Hillenbrand <mhillen@linux.ibm.com>
20050 * config/s390/vecintrin.h (vec_doublee): Fix to use
20051 __builtin_s390_vflls.
20052 (vec_floate): Fix to use __builtin_s390_vflrd.
20054 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
20056 * dominance.c (get_dominated_to_depth): Return auto_vec<basic_block>.
20057 * dominance.h (get_dominated_to_depth): Likewise.
20058 (get_all_dominated_blocks): Likewise.
20059 * cfgcleanup.c (delete_unreachable_blocks): Adjust.
20060 * gcse.c (hoist_code): Likewise.
20061 * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
20062 * tree-parloops.c (oacc_entry_exit_ok): Likewise.
20063 * tree-ssa-dce.c (eliminate_unnecessary_stmts): Likewise.
20064 * tree-ssa-phiprop.c (pass_phiprop::execute): Likewise.
20066 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
20068 * dominance.c (get_dominated_by_region): Return auto_vec<basic_block>.
20069 * dominance.h (get_dominated_by_region): Likewise.
20070 * tree-cfg.c (gimple_duplicate_sese_region): Adjust.
20071 (gimple_duplicate_sese_tail): Likewise.
20072 (move_sese_region_to_fn): Likewise.
20074 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
20076 * dominance.c (get_dominated_by): Return auto_vec<basic_block>.
20077 * dominance.h (get_dominated_by): Likewise.
20078 * auto-profile.c (afdo_find_equiv_class): Adjust.
20079 * cfgloopmanip.c (duplicate_loop_to_header_edge): Likewise.
20080 * loop-unroll.c (unroll_loop_runtime_iterations): Likewise.
20081 * tree-cfg.c (test_linear_chain): Likewise.
20082 (test_diamond): Likewise.
20084 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
20086 * cfgloop.h (get_loop_hot_path): Return auto_vec<basic_block>.
20087 * cfgloopanal.c (get_loop_hot_path): Likewise.
20088 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
20090 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
20092 * cgraph.c (cgraph_node::collect_callers): Return
20093 auto_vec<cgraph_edge *>.
20094 * cgraph.h (cgraph_node::collect_callers): Likewise.
20095 * ipa-cp.c (create_specialized_node): Adjust.
20096 (decide_about_value): Likewise.
20097 (decide_whether_version_node): Likewise.
20098 * ipa-sra.c (process_isra_node_results): Likewise.
20100 2021-06-17 Trevor Saunders <tbsaunde@tbsaunde.org>
20102 * vec.h (vl_ptr>::using_auto_storage): Handle null m_vec.
20103 (auto_vec<T, 0>::auto_vec): Define move constructor, and delete copy
20105 (auto_vec<T, 0>::operator=): Define move assignment and delete copy
20108 2021-06-17 Aldy Hernandez <aldyh@redhat.com>
20110 * gimple-range.cc (debug_seed_ranger): New.
20111 (dump_ranger): New.
20112 (debug_ranger): New.
20114 2021-06-17 Richard Biener <rguenther@suse.de>
20116 PR tree-optimization/54400
20117 * tree-vectorizer.h (enum slp_instance_kind): Add
20118 slp_inst_kind_bb_reduc.
20119 (reduction_fn_for_scalar_code): Declare.
20120 * tree-vect-data-refs.c (vect_slp_analyze_instance_dependence):
20121 Check SLP_INSTANCE_KIND instead of looking at the
20123 (vect_slp_analyze_instance_alignment): Likewise.
20124 * tree-vect-loop.c (reduction_fn_for_scalar_code): Export.
20125 * tree-vect-slp.c (vect_slp_linearize_chain): Split out
20126 chain linearization from vect_build_slp_tree_2 and generalize
20127 for the use of BB reduction vectorization.
20128 (vect_build_slp_tree_2): Adjust accordingly.
20129 (vect_optimize_slp): Elide permutes at the root of BB reduction
20131 (vectorizable_bb_reduc_epilogue): New function.
20132 (vect_slp_prune_covered_roots): Likewise.
20133 (vect_slp_analyze_operations): Use them.
20134 (vect_slp_check_for_constructors): Recognize associatable
20135 chains for BB reduction vectorization.
20136 (vectorize_slp_instance_root_stmt): Generate code for the
20137 BB reduction epilogue.
20139 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
20141 * gimple-range-gori.cc (gori_compute::has_edge_range_p): Check with
20143 (gori_compute::may_recompute_p): New.
20144 (gori_compute::outgoing_edge_range_p): Perform recomputations.
20145 * gimple-range-gori.h (class gori_compute): Add prototype.
20147 2021-06-17 Andrew MacLeod <amacleod@redhat.com>
20149 * gimple-range-cache.cc (ranger_cache::range_on_edge): Always return
20150 true when a range can be calculated.
20151 * gimple-range.cc (gimple_ranger::dump_bb): Check has_edge_range_p.
20153 2021-06-16 Martin Sebor <msebor@redhat.com>
20155 * doc/invoke.texi (-Wmismatched-dealloc, -Wmismatched-new-delete):
20156 Correct documented defaults.
20158 2021-06-16 Andrew MacLeod <amacleod@redhat.com>
20160 * gimple-range-cache.cc (ranger_cache::ranger_cache): Initialize
20161 m_new_value_p directly.
20163 2021-06-16 Uroš Bizjak <ubizjak@gmail.com>
20166 * config/i386/i386-expand.c (expand_vec_perm_2perm_pblendv):
20167 Handle 64bit modes for TARGET_SSE4_1.
20168 (expand_vec_perm_pshufb2): Handle 64bit modes for TARGET_SSSE3.
20169 (expand_vec_perm_even_odd_pack): Handle V4HI mode.
20170 (expand_vec_perm_even_odd_1) <case E_V4HImode>: Expand via
20171 expand_vec_perm_pshufb2 for TARGET_SSSE3 and via
20172 expand_vec_perm_even_odd_pack for TARGET_SSE4_1.
20173 * config/i386/mmx.md (mmx_packusdw): New insn pattern.
20175 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
20177 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn<mode>):
20178 Change to an expander that emits the correct instruction
20179 depending on endianness.
20180 (aarch64_<sur><addsub>hn<mode>_insn_le): Define.
20181 (aarch64_<sur><addsub>hn<mode>_insn_be): Define.
20183 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
20185 * config/aarch64/aarch64-simd-builtins.def: Split generator
20186 for aarch64_<su>qmovn builtins into scalar and vector
20188 * config/aarch64/aarch64-simd.md (aarch64_<su>qmovn<mode>_insn_le):
20190 (aarch64_<su>qmovn<mode>_insn_be): Define.
20191 (aarch64_<su>qmovn<mode>): Split into scalar and vector
20192 variants. Change vector variant to an expander that emits the
20193 correct instruction depending on endianness.
20195 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
20197 * config/aarch64/aarch64-simd-builtins.def: Split generator
20198 for aarch64_sqmovun builtins into scalar and vector variants.
20199 * config/aarch64/aarch64-simd.md (aarch64_sqmovun<mode>):
20200 Split into scalar and vector variants. Change vector variant
20201 to an expander that emits the correct instruction depending
20203 (aarch64_sqmovun<mode>_insn_le): Define.
20204 (aarch64_sqmovun<mode>_insn_be): Define.
20206 2021-06-16 Jonathan Wright <jonathan.wright@arm.com>
20208 * config/aarch64/aarch64-simd.md (aarch64_xtn<mode>_insn_le):
20209 Define - modeling zero-high-half semantics.
20210 (aarch64_xtn<mode>): Change to an expander that emits the
20211 appropriate instruction depending on endianness.
20212 (aarch64_xtn<mode>_insn_be): Define - modeling zero-high-half
20214 (aarch64_xtn2<mode>_le): Rename to...
20215 (aarch64_xtn2<mode>_insn_le): This.
20216 (aarch64_xtn2<mode>_be): Rename to...
20217 (aarch64_xtn2<mode>_insn_be): This.
20218 (vec_pack_trunc_<mode>): Emit truncation instruction instead
20220 * config/aarch64/iterators.md (Vnarrowd): Add Vnarrowd mode
20221 attribute iterator.
20223 2021-06-16 Martin Jambor <mjambor@suse.cz>
20225 PR tree-optimization/100453
20226 * tree-sra.c (create_access): Disqualify any const candidates
20227 which are written to.
20228 (sra_modify_expr): Do not store sub-replacements back to a const base.
20229 (handle_unscalarized_data_in_subtree): Likewise.
20230 (sra_modify_assign): Likewise. Earlier, use TREE_READONLy test
20231 instead of constant_decl_p.
20233 2021-06-16 Jakub Jelinek <jakub@redhat.com>
20235 PR middle-end/101062
20236 * stor-layout.c (finish_bitfield_representative): For fields in unions
20237 assume nextf is always NULL.
20238 (finish_bitfield_layout): Compute bit field representatives also in
20239 unions, but handle it as if each bitfield was the only field in the
20242 2021-06-16 Richard Biener <rguenther@suse.de>
20244 PR tree-optimization/101088
20245 * tree-ssa-loop-im.c (sm_seq_valid_bb): Only look for
20246 supported refs on edges. Do not assert same ref but
20247 different kind stores are unsuported but mark them so.
20248 (hoist_memory_references): Only look for supported refs
20251 2021-06-16 Roger Sayle <roger@nextmovesoftware.com>
20253 PR rtl-optimization/46235
20254 * config/i386/i386.md: New define_split for bt followed by cmov.
20255 (*bt<mode>_setcqi): New define_insn_and_split for bt followed by setc.
20256 (*bt<mode>_setncqi): New define_insn_and_split for bt then setnc.
20257 (*bt<mode>_setnc<mode>): New define_insn_and_split for bt followed
20258 by setnc with zero extension.
20260 2021-06-16 Richard Biener <rguenther@suse.de>
20262 PR tree-optimization/101083
20263 * tree-vect-slp.c (vect_slp_build_two_operator_nodes): Get
20264 vectype as argument.
20265 (vect_build_slp_tree_2): Adjust.
20267 2021-06-15 Martin Sebor <msebor@redhat.com>
20269 PR middle-end/100876
20270 * builtins.c: (gimple_call_return_array): Account for size_t
20271 mangling as either unsigned int or unsigned long
20273 2021-06-15 Jeff Law <jeffreyalaw@gmail.com>
20275 * compare-elim.c (try_eliminate_compare): Run DCE to clean things
20276 up before eliminating comparisons.
20278 2021-06-15 Aldy Hernandez <aldyh@redhat.com>
20280 * range-op.cc (operator_bitwise_or::wi_fold): Make sure
20281 nonzero|X is nonzero.
20282 (range_op_bitwise_and_tests): Add tests for above.
20284 2021-06-15 Carl Love <cel@us.ibm.com>
20287 * config/rs6000/rs6000-builtin.def (VCMPEQUT): Fix the ICODE for the
20289 (VRLQ, VSLQ, VSRQ, VSRAQ): Remove unused BU_P10_OVERLOAD_2
20292 2021-06-15 Tobias Burnus <tobias@codesourcery.com>
20295 * gimplify.c (enum gimplify_defaultmap_kind): Add GDMK_SCALAR_TARGET.
20296 (struct gimplify_omp_ctx): Extend defaultmap array by one.
20297 (new_omp_context): Init defaultmap[GDMK_SCALAR_TARGET].
20298 (omp_notice_variable): Update type classification for Fortran.
20299 (gimplify_scan_omp_clauses): Update calls for new argument; handle
20300 GDMK_SCALAR_TARGET; for Fortran, GDMK_POINTER avoid GOVD_MAP_0LEN_ARRAY.
20301 * langhooks-def.h (lhd_omp_scalar_p): Add 'ptr_ok' argument.
20302 * langhooks.c (lhd_omp_scalar_p): Likewise.
20303 (LANG_HOOKS_OMP_ALLOCATABLE_P, LANG_HOOKS_OMP_SCALAR_TARGET_P): New.
20304 (LANG_HOOKS_DECLS): Add them.
20305 * langhooks.h (struct lang_hooks_for_decls): Add new hooks, update
20306 omp_scalar_p pointer type to include the new bool argument.
20308 2021-06-15 David Malcolm <dmalcolm@redhat.com>
20310 * doc/analyzer.texi
20311 (Special Functions for Debugging the Analyzer): Add
20312 __analyzer_dump_capacity.
20314 2021-06-15 Jakub Jelinek <jakub@redhat.com>
20317 * expr.c (expand_expr_real_2) <case VEC_PACK_FIX_TRUNC_EXPR,
20318 case VEC_PACK_TRUNC_EXPR>: Clear subtarget when changing mode.
20320 2021-06-15 Richard Biener <rguenther@suse.de>
20322 * cfgloopanal.c (mark_irreducible_loops): Use a dominance
20323 check to identify loop latches.
20324 * cfgloop.c (verify_loop_structure): Likewise.
20325 * loop-init.c (apply_loop_flags): Allow marked irreducible
20326 regions even with multiple latches.
20327 * predict.c (rebuild_frequencies): Simplify.
20329 2021-06-15 Richard Biener <rguenther@suse.de>
20331 * tree-ssa-threadupdate.c
20332 (jump_thread_path_registry::mark_threaded_blocks): Assert we
20333 have marked irreducible regions.
20335 2021-06-14 Martin Sebor <msebor@redhat.com>
20338 * builtins.c (gimple_call_return_array): Check for attribute fn spec.
20339 Handle calls to placement new.
20340 (ndecl_dealloc_argno): Avoid placement delete.
20342 2021-06-14 Peter Bergner <bergner@linux.ibm.com>
20345 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_mma_builtin): Use
20346 create_tmp_reg_or_ssa_name().
20348 2021-06-14 Andrew MacLeod <amacleod@redhat.com>
20350 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust.
20351 (ranger_cache::enable_new_values): Set to specified value and
20352 return the old value.
20353 (ranger_cache::disable_new_values): Delete.
20354 (ranger_cache::fill_block_cache): Disable non 1st order derived
20356 * gimple-range-cache.h (ranger_cache): Adjust prototypes.
20357 * gimple-range.cc (gimple_ranger::range_of_expr): Adjust.
20359 2021-06-14 Uroš Bizjak <ubizjak@gmail.com>
20362 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const):
20363 Return true early when testing with V2HImode.
20364 * config/i386/mmx.md (*punpckwd): Split to sse2_pshuflw_1.
20366 2021-06-14 Christophe Lyon <christophe.lyon@linaro.org>
20368 * config/arm/mve.md (mve_vec_unpack<US>_lo_<mode>): New pattern.
20369 (mve_vec_unpack<US>_hi_<mode>): New pattern.
20370 (@mve_vec_pack_trunc_lo_<mode>): New pattern.
20371 (mve_vmovntq_<supf><mode>): Prefix with '@'.
20372 * config/arm/neon.md (vec_unpack<US>_hi_<mode>): Move to
20374 (vec_unpack<US>_lo_<mode>): Likewise.
20375 (vec_pack_trunc_<mode>): Rename to
20376 neon_quad_vec_pack_trunc_<mode>.
20377 * config/arm/vec-common.md (vec_unpack<US>_hi_<mode>): New
20379 (vec_unpack<US>_lo_<mode>): New.
20380 (vec_pack_trunc_<mode>): New.
20382 2021-06-14 Richard Biener <rguenther@suse.de>
20384 PR tree-optimization/100934
20385 * tree-ssa-dom.c (pass_dominator::execute): Properly
20386 mark irreducible regions.
20388 2021-06-14 Martin Liska <mliska@suse.cz>
20390 * doc/invoke.texi: Put r{...} on the same line as @item.
20392 2021-06-14 Martin Liska <mliska@suse.cz>
20394 * doc/invoke.texi: Add missing newline.
20396 2021-06-14 Martin Liska <mliska@suse.cz>
20398 * doc/invoke.texi: Remove '+' charasters.
20400 2021-06-14 Claudiu Zissulescu <claziss@synopsys.com>
20402 * config.gcc (arc): Add support for with_cpu option.
20403 * config/arc/arc.h (OPTION_DEFAULT_SPECS): Add fpu.
20405 2021-06-14 Richard Biener <rguenther@suse.de>
20407 PR tree-optimization/101031
20408 * tree-ssa-strlen.c (maybe_invalidate): Increment max_size
20409 instead of size when accounting for a possibly string
20412 2021-06-14 Martin Liska <mliska@suse.cz>
20414 * gimple-ssa-evrp.c (pointer_equiv_analyzer::~pointer_equiv_analyzer): Use delete[].
20416 2021-06-14 Aldy Hernandez <aldyh@redhat.com>
20418 * value-query.cc (gimple_range_global): Call get_range_global
20419 if called after inlining.
20421 2021-06-13 Uroš Bizjak <ubizjak@gmail.com>
20424 * config/i386/i386-expand.c (expand_vec_perm_pshufb):
20425 Emit constant permutation insn directly from here.
20427 2021-06-13 Trevor Saunders <tbsaunde@tbsaunde.org>
20429 * attribs.c (find_attribute_namespace): Iterate over vec<> with
20431 * auto-profile.c (afdo_find_equiv_class): Likewise.
20432 * gcc.c (do_specs_vec): Likewise.
20433 (do_spec_1): Likewise.
20434 (driver::set_up_specs): Likewise.
20435 * gimple-loop-jam.c (any_access_function_variant_p): Likewise.
20436 * gimple-ssa-store-merging.c (compatible_load_p): Likewise.
20437 (imm_store_chain_info::try_coalesce_bswap): Likewise.
20438 (imm_store_chain_info::coalesce_immediate_stores): Likewise.
20439 (get_location_for_stmts): Likewise.
20440 * graphite-poly.c (print_iteration_domains): Likewise.
20441 (free_poly_bb): Likewise.
20442 (remove_gbbs_in_scop): Likewise.
20443 (free_scop): Likewise.
20444 (dump_gbb_cases): Likewise.
20445 (dump_gbb_conditions): Likewise.
20446 (print_pdrs): Likewise.
20447 (print_scop): Likewise.
20448 * ifcvt.c (cond_move_process_if_block): Likewise.
20449 * lower-subreg.c (decompose_multiword_subregs): Likewise.
20450 * regcprop.c (pass_cprop_hardreg::execute): Likewise.
20451 * sanopt.c (sanitize_rewrite_addressable_params): Likewise.
20452 * sel-sched-dump.c (dump_insn_vector): Likewise.
20453 * store-motion.c (store_ops_ok): Likewise.
20454 (store_killed_in_insn): Likewise.
20455 * timevar.c (timer::named_items::print): Likewise.
20456 * tree-cfgcleanup.c (cleanup_control_flow_pre): Likewise.
20457 (cleanup_tree_cfg_noloop): Likewise.
20458 * tree-data-ref.c (dump_data_references): Likewise.
20459 (print_dir_vectors): Likewise.
20460 (print_dist_vectors): Likewise.
20461 (dump_data_dependence_relations): Likewise.
20462 (dump_dist_dir_vectors): Likewise.
20463 (dump_ddrs): Likewise.
20464 (create_runtime_alias_checks): Likewise.
20465 (free_subscripts): Likewise.
20466 (save_dist_v): Likewise.
20467 (save_dir_v): Likewise.
20468 (invariant_access_functions): Likewise.
20469 (same_access_functions): Likewise.
20470 (access_functions_are_affine_or_constant_p): Likewise.
20471 (find_data_references_in_stmt): Likewise.
20472 (graphite_find_data_references_in_stmt): Likewise.
20473 (free_dependence_relations): Likewise.
20474 (free_data_refs): Likewise.
20475 * tree-inline.c (copy_debug_stmts): Likewise.
20476 * tree-into-ssa.c (dump_currdefs): Likewise.
20477 (rewrite_update_phi_arguments): Likewise.
20478 * tree-ssa-propagate.c (clean_up_loop_closed_phi): Likewise.
20479 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
20481 (vect_slp_analyze_node_dependences): Likewise.
20482 (vect_slp_analyze_instance_dependence): Likewise.
20483 (vect_record_base_alignments): Likewise.
20484 (vect_get_peeling_costs_all_drs): Likewise.
20485 (vect_peeling_supportable): Likewise.
20486 * tree-vectorizer.c (vec_info::~vec_info): Likewise.
20487 (vec_info::free_stmt_vec_infos): Likewise.
20489 2021-06-13 Jeff Law <jeffreyalaw@gmail.com>
20491 * config/h8300/logical.md (<code>qi3_1<cczn>): New pattern.
20492 (andqi3_1<cczn>): Removed.
20493 (<ors>qi3_1): Do not split for IOR/XOR a single bit.
20494 (H8/SX bit logicals): Split out from other patterns.
20495 * config/h8300/multiply.md (mulqihi3_const<cczn>): Renamed from
20496 mulqihi3_const_clobber_flags.
20497 (mulqihi3<cczn>, mulhisi3_const<cczn>, mulhisi3<cczn>): Similarly
20499 2021-06-13 H.J. Lu <hjl.tools@gmail.com>
20502 * config/i386/i386.c (ix86_expand_prologue): Set red_zone_used
20503 to true if red zone is used.
20504 (ix86_output_indirect_jmp): Replace ix86_red_zone_size with
20505 ix86_red_zone_used.
20506 * config/i386/i386.h (machine_function): Add red_zone_used.
20507 (ix86_red_zone_size): Removed.
20508 (ix86_red_zone_used): New.
20509 * config/i386/i386.md (peephole2 patterns): Replace
20510 ix86_red_zone_size with ix86_red_zone_used.
20512 2021-06-12 Jason Merrill <jason@redhat.com>
20514 * doc/extend.texi (unused variable attribute): Applies to
20515 structure fields as well.
20517 2021-06-12 Eugene Rozenfeld <erozen@microsoft.com>
20519 * auto-profile.c (read_profile): fix a typo in an error string
20521 2021-06-11 Thomas Schwinge <thomas@codesourcery.com>
20523 * tree-pretty-print.h (dump_omp_clauses): Add 'bool = true'
20525 * tree-pretty-print.c (dump_omp_clauses): Update.
20526 (dump_generic_node) <OMP_CLAUSE>: Use it.
20528 2021-06-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
20531 * config/arm/arm_mve.h (__arm_vld1q): Change __ARM_mve_coerce(p0,
20532 int8_t const *) to __ARM_mve_coerce1(p0, int8_t *) in the argument for
20533 the polymorphic variants matching code.
20534 (__arm_vld1q_z): Likewise.
20535 (__arm_vld2q): Likewise.
20536 (__arm_vld4q): Likewise.
20537 (__arm_vldrbq_gather_offset): Likewise.
20538 (__arm_vldrbq_gather_offset_z): Likewise.
20540 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
20542 PR tree-optimization/96392
20543 * fold-const.h (tree_expr_maybe_real_minus_zero_p): Fix prototype.
20545 2021-06-11 Roger Sayle <roger@nextmovesoftware.com>
20547 PR tree-optimization/96392
20548 * fold-const.c (fold_real_zero_addition_p): Take both arguments
20549 of the addition or subtraction, not just the zero. Use this
20550 other argument in tests for signaling NaNs and signed zeros.
20551 (tree_expr_maybe_real_minus_zero_p): New predicate.
20552 * fold-const.h (fold_real_zero_addition_p): Update prototype.
20553 (tree_expr_maybe_real_minus_zero_p): New function prototype.
20554 * match.pd: Update calls to fold_real_zero_addition_p.
20555 Replace HONOR_NANS with tree_expr_maybe_nan_p.
20556 Replace HONOR_SIGNED_ZEROS with tree_expr_maybe_real_minus_zero_p.
20557 Replace HONOR_SNANS with tree_expr_maybe_signaling_nan_p.
20558 * tree-ssa-reassoc.c (eliminate_using_constants): Update
20559 call to fold_real_zero_addition_p.
20561 2021-06-11 Richard Biener <rguenther@suse.de>
20563 PR tree-optimization/101025
20564 * tree-ssa-loop-im.c (sm_seq_valid_bb): Make sure to process
20565 all refs that require dependence checking.
20567 2021-06-11 Richard Biener <rguenther@suse.de>
20569 PR tree-optimization/101028
20570 * tree-vect-slp.c (vect_build_slp_tree_2): When SLP
20571 reassoc discovery fails fatally, mark appropriate lanes
20574 2021-06-11 Richard Biener <rguenther@suse.de>
20576 PR tree-optimization/101026
20577 * tree-vect-slp.c (vect_build_slp_tree_2): Make sure we
20578 have a representative for the associated chain nodes.
20580 2021-06-11 Jakub Jelinek <jakub@redhat.com>
20582 PR rtl-optimization/101008
20583 * simplify-rtx.c (relational_result): New function.
20584 (simplify_logical_relational_operation,
20585 simplify_relational_operation): Use it.
20587 2021-06-11 Jakub Jelinek <jakub@redhat.com>
20590 * config/i386/sse.md (*vec_concat<mode>_0_1): Require TARGET_SSE2.
20592 2021-06-11 Uroš Bizjak <ubizjak@gmail.com>
20595 * config/i386/i386-expand.c (expand_vec_perm_pshufb): Return
20596 false if the permutation can be implemented with constant
20597 permutation instruction in wider mode.
20598 (canonicalize_vector_int_perm): Move above expand_vec_perm_pshufb.
20599 Handle V8QImode and V4HImode.
20601 2021-06-11 Martin Liska <mliska@suse.cz>
20603 PR gcov-profile/100788
20604 * common.opt: Add new option.
20605 * coverage.c (coverage_begin_function): Emit warning instead on
20606 the internal compiler error.
20607 * doc/invoke.texi: Document the option.
20608 * toplev.c (process_options): Enable it by default.
20610 2021-06-11 Richard Biener <rguenther@suse.de>
20612 PR middle-end/101009
20613 * tree-data-ref.c (build_classic_dist_vector_1): Make sure
20614 to set *init_b to true when we encounter a constant equal
20616 (compute_affine_dependence): Also dump the actual DR_REF.
20618 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
20620 PR tree-optimization/100984
20621 * gimple-ssa-evrp.c (ssa_equiv_stack): Use auto_vec for
20622 replacements table.
20623 (ssa_equiv_stack::~ssa_equiv_stack): Remove.
20625 2021-06-11 Kewen Lin <linkw@linux.ibm.com>
20627 * config/rs6000/rs6000.md
20628 (floatsi<SFDF:mode>2_lfiwax_<QHI:mode>_mem_zext): New
20629 define_insn_and_split.
20631 2021-06-11 Richard Biener <rguenther@suse.de>
20633 * tree-vect-slp.c (vect_build_slp_tree_2): Use stablesort
20634 to sort operands of the associative chain.
20636 2021-06-11 Richard Biener <rguenther@suse.de>
20638 * system.h (gcc_stablesort_r): Declare.
20639 * sort.cc (gcc_sort_r): Support stable sort.
20640 (gcc_stablesort_r): Define.
20641 * vec.h (vec<>::stablesort): Add.
20643 2021-06-10 Uroš Bizjak <ubizjak@gmail.com>
20646 * config/i386/i386-expand.c (ix86_split_mmx_punpck):
20647 Handle V2SF mode. Emit SHUFPS to fixup unpack-high for V2SF mode.
20648 (expand_vec_perm_blend): Handle 64bit modes for TARGET_SSE4_1.
20649 (expand_vec_perm_pshufb): Handle 64bit modes for TARGET_SSSE3.
20650 (expand_vec_perm_pblendv): Handle 64bit modes for TARGET_SSE4_1.
20651 (expand_vec_perm_interleave2): Handle 64bit modes.
20652 (expand_vec_perm_even_odd_pack): Handle V8QI mode.
20653 (expand_vec_perm_even_odd_1): Ditto.
20654 (ix86_vectorize_vec_perm_const): Ditto.
20655 * config/i386/i386.md (UNSPEC_PSHUFB): Move from ...
20656 * config/i386/sse.md: ... here.
20657 * config/i386/mmx.md (*vec_interleave_lowv2sf):
20658 New insn_and_split pattern.
20659 (*vec_interleave_highv2sf): Ditto.
20660 (mmx_pshufbv8qi3): New insn pattern.
20661 (*mmx_pblendw): Ditto.
20663 2021-06-10 Peter Bergner <bergner@linux.ibm.com>
20665 * config/rs6000/rs6000-builtin.def (build_pair): New built-in.
20666 (build_acc): Likewise.
20667 * config/rs6000/rs6000-call.c (mma_expand_builtin): Swap assemble
20668 source operands in little-endian mode.
20669 (rs6000_gimple_fold_mma_builtin): Handle VSX_BUILTIN_BUILD_PAIR.
20670 (mma_init_builtins): Likewise.
20671 * config/rs6000/rs6000.c (rs6000_split_multireg_move): Handle endianness
20672 ordering for the MMA assemble and build source operands.
20673 * doc/extend.texi (__builtin_vsx_build_acc, __builtin_mma_build_pair):
20675 (__builtin_mma_assemble_acc, __builtin_mma_assemble_pair): Remove
20678 2021-06-10 Jeff Law <jeffreyalaw@gmail.com>
20680 * config/h8300/h8300.c (select_cc_mode): Handle MEM. Use
20682 * config/h8300/extensions.md: Replace _clobber_flags patterns
20685 2021-06-10 Robin Dapp <rdapp@linux.ibm.com>
20687 * config/s390/vector.md (vcond_mask_<mode><mode>): Change to
20688 (vcond_mask_<mode><tointvec>): this.
20690 2021-06-10 Andrew Stubbs <ams@codesourcery.com>
20691 Thomas Schwinge <thomas@codesourcery.com>
20693 * omp-builtins.def (BUILT_IN_GOACC_ENTER_EXIT_DATA): Split into...
20694 (BUILT_IN_GOACC_ENTER_DATA, BUILT_IN_GOACC_EXIT_DATA): ... these.
20695 * gimple.h (enum gf_mask): Split
20696 'GF_OMP_TARGET_KIND_OACC_ENTER_EXIT_DATA' into
20697 'GF_OMP_TARGET_KIND_OACC_ENTER_DATA' and
20698 'GF_OMP_TARGET_KIND_OACC_EXIT_DATA'.
20699 (is_gimple_omp_oacc): Update.
20700 * gimple-pretty-print.c (dump_gimple_omp_target): Likewise.
20701 * gimplify.c (gimplify_omp_target_update): Likewise.
20702 * omp-expand.c (expand_omp_target, build_omp_regions_1)
20703 (omp_make_gimple_edges): Likewise.
20704 * omp-low.c (check_omp_nesting_restrictions, lower_omp_target):
20707 2021-06-10 Aldy Hernandez <aldyh@redhat.com>
20709 * value-query.cc (value_query::value_on_edge): Rename name to
20711 (range_query::range_on_edge): Same.
20712 (range_query::value_of_expr): Same.
20713 (range_query::value_on_edge): Same.
20714 * value-query.h (class value_query): Same.
20715 (class range_query): Same.
20717 2021-06-10 Richard Biener <rguenther@suse.de>
20719 PR tree-optimization/101003
20720 * tree-vect-slp.c (vect_build_slp_tree_2): Appropriately
20721 use the pattern stmt defs when linearizing a chain.
20723 2021-06-10 Jakub Jelinek <jakub@redhat.com>
20726 * ifcvt.c (noce_get_alt_condition, noce_try_abs): Use
20727 prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
20729 2021-06-10 Clement Chigot <clement.chigot@atos.net>
20731 * config/rs6000/aix71.h (ASM_CPU_SPEC): Add Power10 directive.
20732 * config/rs6000/aix72.h (ASM_CPU_SPEC): Likewise.
20734 2021-06-09 Andrew Pinski <apinski@marvell.com>
20736 PR tree-optimization/100925
20737 * match.pd (a ? CST1 : CST2): Limit transformations
20738 that would produce a negative to integeral types only.
20739 Change !POINTER_TYPE_P to INTEGRAL_TYPE_P also.
20741 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
20744 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
20746 * doc/tm.texi: Correctly update.
20748 2021-06-09 Jeff Law <jeffreyalaw@gmail.com>
20750 * doc/tm.texi: Correctly update.
20752 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
20755 * doc/tm.texi.in (Trampolines): Add a missing blank line.
20757 2021-06-09 Paul Eggert <eggert@cs.ucla.edu>
20760 * doc/invoke.texi (Code Gen Options); Document that -fno-trampolines
20761 and -ftrampolines work only with Ada.
20762 * doc/tm.texi.in (Trampolines): Likewise.
20763 * doc/tm.texi: Regenerated.
20765 2021-06-09 Carl Love <cel@us.ibm.com>
20767 * config/rs6000/altivec.h (vec_signextll, vec_signexti, vec_signextq):
20768 Add define for new builtins.
20769 * config/rs6000/altivec.md(altivec_vreveti2): Add define_expand.
20770 * config/rs6000/rs6000-builtin.def (VSIGNEXTI, VSIGNEXTLL): Add
20771 overloaded builtin definitions.
20772 (VSIGNEXTSB2W, VSIGNEXTSH2W, VSIGNEXTSB2D, VSIGNEXTSH2D,VSIGNEXTSW2D,
20773 VSIGNEXTSD2Q): Add builtin expansions.
20774 (SIGNEXT): Add P10 overload definition.
20775 * config/rs6000/rs6000-call.c (P9V_BUILTIN_VEC_VSIGNEXTI, P9V_BUILTIN_VEC_VSIGNEXTLL,
20776 P10_BUILTIN_VEC_SIGNEXT): Add overloaded argument definitions.
20777 * config/rs6000/vsx.md (vsx_sign_extend_v2di_v1ti): Add define_insn.
20778 (vsignextend_v2di_v1ti, vsignextend_qi_<mode>, vsignextend_hi_<mode>,
20779 vsignextend_si_v2di)[VIlong]: Add define_expand.
20780 Make define_insn vsx_sign_extend_si_v2di visible.
20781 * doc/extend.texi: Add documentation for the vec_signexti,
20782 vec_signextll builtins and vec_signextq.
20784 2021-06-09 Carl Love <cel@us.ibm.com>
20786 * config/rs6000/rs6000.c (__fixkfti, __fixunskfti, __floattikf,
20787 __floatuntikf): Names changed to __fixkfti_sw, __fixunskfti_sw,
20788 __floattikf_sw, __floatuntikf_sw respectively.
20789 * config/rs6000/rs6000.md (floatti<mode>2, floatunsti<mode>2,
20790 fix_trunc<mode>ti2, fixuns_trunc<mode>ti2): Add
20791 define_insn for mode IEEE 128.
20793 2021-06-09 Carl Love <cel@us.ibm.com>
20795 * config/rs6000/altivec.md (altivec_vslq, altivec_vsrq):
20796 Rename to altivec_vslq_<mode>, altivec_vsrq_<mode>, mode VEC_TI.
20797 * config/rs6000/vector.md (VEC_TI): Was named VSX_TI in vsx.md.
20798 (vashlv1ti3): Change to vashl<mode>3, mode VEC_TI.
20799 (vlshrv1ti3): Change to vlshr<mode>3, mode VEC_TI.
20800 * config/rs6000/vsx.md (VSX_TI): Remove define_mode_iterator. Update
20801 uses of VSX_TI to VEC_TI.
20803 2021-06-09 Carl Love <cel@us.ibm.com>
20805 * config/rs6000/dfp.md (floattitd2, fixtdti2): New define_insns.
20807 2021-06-09 Carl Love <cel@us.ibm.com>
20809 * config/rs6000/altivec.h (vec_dive, vec_mod): Add define for new
20811 * config/rs6000/altivec.md (UNSPEC_VMULEUD, UNSPEC_VMULESD,
20812 UNSPEC_VMULOUD, UNSPEC_VMULOSD): New unspecs.
20813 (altivec_eqv1ti, altivec_gtv1ti, altivec_gtuv1ti, altivec_vmuleud,
20814 altivec_vmuloud, altivec_vmulesd, altivec_vmulosd, altivec_vrlq,
20815 altivec_vrlqmi, altivec_vrlqmi_inst, altivec_vrlqnm,
20816 altivec_vrlqnm_inst, altivec_vslq, altivec_vsrq, altivec_vsraq,
20817 altivec_vcmpequt_p, altivec_vcmpgtst_p, altivec_vcmpgtut_p): New
20819 (vec_widen_umult_even_v2di, vec_widen_smult_even_v2di,
20820 vec_widen_umult_odd_v2di, vec_widen_smult_odd_v2di, altivec_vrlqmi,
20821 altivec_vrlqnm): New define_expands.
20822 * config/rs6000/rs6000-builtin.def (VCMPEQUT_P, VCMPGTST_P,
20823 VCMPGTUT_P): Add macro expansions.
20824 (BU_P10V_AV_P): Add builtin predicate definition.
20825 (VCMPGTUT, VCMPGTST, VCMPEQUT, CMPNET, CMPGE_1TI,
20826 CMPGE_U1TI, CMPLE_1TI, CMPLE_U1TI, VNOR_V1TI_UNS, VNOR_V1TI, VCMPNET_P,
20827 VCMPAET_P, VMULEUD, VMULESD, VMULOUD, VMULOSD, VRLQ,
20828 VSLQ, VSRQ, VSRAQ, VRLQNM, DIV_V1TI, UDIV_V1TI, DIVES_V1TI, DIVEU_V1TI,
20829 MODS_V1TI, MODU_V1TI, VRLQMI): New macro expansions.
20830 (VRLQ, VSLQ, VSRQ, VSRAQ, DIVE, MOD): New overload expansions.
20831 * config/rs6000/rs6000-call.c (P10_BUILTIN_VCMPEQUT,
20832 P10V_BUILTIN_CMPGE_1TI, P10V_BUILTIN_CMPGE_U1TI,
20833 P10V_BUILTIN_VCMPGTUT, P10V_BUILTIN_VCMPGTST,
20834 P10V_BUILTIN_CMPLE_1TI, P10V_BUILTIN_VCMPLE_U1TI,
20835 P10V_BUILTIN_DIV_V1TI, P10V_BUILTIN_UDIV_V1TI,
20836 P10V_BUILTIN_VMULESD, P10V_BUILTIN_VMULEUD,
20837 P10V_BUILTIN_VMULOSD, P10V_BUILTIN_VMULOUD,
20838 P10V_BUILTIN_VNOR_V1TI, P10V_BUILTIN_VNOR_V1TI_UNS,
20839 P10V_BUILTIN_VRLQ, P10V_BUILTIN_VRLQMI,
20840 P10V_BUILTIN_VRLQNM, P10V_BUILTIN_VSLQ,
20841 P10V_BUILTIN_VSRQ, P10V_BUILTIN_VSRAQ,
20842 P10V_BUILTIN_VCMPGTUT_P, P10V_BUILTIN_VCMPGTST_P,
20843 P10V_BUILTIN_VCMPEQUT_P, P10V_BUILTIN_VCMPGTUT_P,
20844 P10V_BUILTIN_VCMPGTST_P, P10V_BUILTIN_CMPNET,
20845 P10V_BUILTIN_VCMPNET_P, P10V_BUILTIN_VCMPAET_P,
20846 P10V_BUILTIN_DIVES_V1TI, P10V_BUILTIN_MODS_V1TI,
20847 P10V_BUILTIN_MODU_V1TI):
20848 New overloaded definitions.
20849 (rs6000_gimple_fold_builtin) [P10V_BUILTIN_VCMPEQUT,
20850 P10V_BUILTIN_CMPNET, P10V_BUILTIN_CMPGE_1TI,
20851 P10V_BUILTIN_CMPGE_U1TI, P10V_BUILTIN_VCMPGTUT,
20852 P10V_BUILTIN_VCMPGTST, P10V_BUILTIN_CMPLE_1TI,
20853 P10V_BUILTIN_CMPLE_U1TI]: New case statements.
20854 (rs6000_init_builtins) [bool_V1TI_type_node, int_ftype_int_v1ti_v1ti]:
20856 (altivec_init_builtins): New E_V1TImode case statement.
20857 (builtin_function_type)[P10_BUILTIN_128BIT_VMULEUD,
20858 P10_BUILTIN_128BIT_VMULOUD, P10_BUILTIN_128BIT_DIVEU_V1TI,
20859 P10_BUILTIN_128BIT_MODU_V1TI, P10_BUILTIN_CMPGE_U1TI,
20860 P10_BUILTIN_VCMPGTUT, P10_BUILTIN_VCMPEQUT]: New case statements.
20861 * config/rs6000/rs6000.c (rs6000_handle_altivec_attribute) [E_TImode,
20862 E_V1TImode]: New case statements.
20863 * config/rs6000/rs6000.h (rs6000_builtin_type_index): New enum
20864 value RS6000_BTI_bool_V1TI.
20865 * config/rs6000/vector.md (vector_gtv1ti,vector_nltv1ti,
20866 vector_gtuv1ti, vector_nltuv1ti, vector_ngtv1ti, vector_ngtuv1ti,
20867 vector_eq_v1ti_p, vector_ne_v1ti_p, vector_ae_v1ti_p,
20868 vector_gt_v1ti_p, vector_gtu_v1ti_p, vrotlv1ti3, vashlv1ti3,
20869 vlshrv1ti3, vashrv1ti3): New define_expands.
20870 * config/rs6000/vsx.md (UNSPEC_VSX_DIVSQ, UNSPEC_VSX_DIVUQ,
20871 UNSPEC_VSX_DIVESQ, UNSPEC_VSX_DIVEUQ, UNSPEC_VSX_MODSQ,
20872 UNSPEC_VSX_MODUQ): New unspecs.
20873 (mulv2di3, vsx_div_v1ti, vsx_udiv_v1ti, vsx_dives_v1ti,
20874 vsx_diveu_v1ti, vsx_mods_v1ti, vsx_modu_v1ti, xxswapd_v1ti): New
20876 (vcmpnet): New define_expand.
20877 * doc/extend.texi: Add documentation for the new builtins vec_rl,
20878 vec_rlmi, vec_rlnm, vec_sl, vec_sr, vec_sra, vec_mule, vec_mulo,
20879 vec_div, vec_dive, vec_mod, vec_cmpeq, vec_cmpne, vec_cmpgt, vec_cmplt,
20880 vec_cmpge, vec_cmple, vec_all_eq, vec_all_ne, vec_all_gt, vec_all_lt,
20881 vec_all_ge, vec_all_le, vec_any_eq, vec_any_ne, vec_any_gt, vec_any_lt,
20882 vec_any_ge, vec_any_le.
20884 2021-06-09 Carl Love <cel@us.ibm.com>
20886 * config/rs6000/altivec.md (altivec_vrl<VI_char>mi): Fix
20887 bug in argument generation.
20889 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
20891 * config/arm/iterators.md (<supf>): Remove VCLZQ_U, VCLZQ_S.
20893 * config/arm/mve.md (mve_vclzq_<supf><mode>): Add '@' prefix,
20894 remove <supf> iterator.
20895 (mve_vclzq_u<mode>): New.
20896 * config/arm/neon.md (clz<mode>2): Rename to neon_vclz<mode>.
20897 (neon_vclz<mode): Move to ...
20898 * config/arm/unspecs.md (VCLZQ_U, VCLZQ_S): Remove.
20899 * config/arm/vec-common.md: ... here. Add support for MVE.
20901 2021-06-09 Christophe Lyon <christophe.lyon@linaro.org>
20903 * config/arm/mve.md (mve_vhaddq_<supf><mode>): Prefix with '@'.
20904 (@mve_vrhaddq_<supf><mode): Likewise.
20905 * config/arm/neon.md (neon_v<r>hadd<sup><mode>): Likewise.
20906 * config/arm/vec-common.md (avg<mode>3_floor, uavg<mode>3_floor)
20907 (avg<mode>3_ceil", uavg<mode>3_ceil): New patterns.
20909 2021-06-09 imba-tjd <109224573@qq.com>
20911 * doc/invoke.texi: Fix typo.
20913 2021-06-09 Roger Sayle <roger@nextmovesoftware.com>
20915 PR middle-end/53267
20916 * fold-const-call.c (fold_const_call_sss) [CASE_CFN_FMOD]:
20917 Support evaluation of fmod/fmodf/fmodl at compile-time.
20919 2021-06-09 Richard Biener <rguenther@suse.de>
20921 PR tree-optimization/100981
20922 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
20923 gimple_get_lhs to also handle calls.
20924 * tree-vect-slp-patterns.c (complex_pattern::build): Transfer
20927 2021-06-09 Richard Biener <rguenther@suse.de>
20929 PR tree-optimization/97832
20930 * tree-vectorizer.h (_slp_tree::failed): New.
20931 * tree-vect-slp.c (_slp_tree::_slp_tree): Initialize
20933 (_slp_tree::~_slp_tree): Free failed.
20934 (vect_build_slp_tree): Retain failed nodes and record
20935 matches in them, copying that back out when running
20936 into a cached fail. Dump start and end of discovery.
20937 (dt_sort_cmp): New.
20938 (vect_build_slp_tree_2): Handle associatable chains
20939 together doing more aggressive operand swapping.
20941 2021-06-09 H.J. Lu <hjl.tools@gmail.com>
20944 * config.gcc (gcc_cv_initfini_array): Set to yes for Linux and
20946 * doc/install.texi: Require glibc 2.1 and binutils 2.12 for
20947 Linux and GNU targets.
20949 2021-06-09 Richard Biener <rguenther@suse.de>
20951 * tree-vect-stmts.c (vect_is_simple_use): Always get dt
20954 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
20956 * config/arc/arc.md (loop_end): Change it to
20957 define_insn_and_split.
20959 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
20961 * config/arc/arc.md (maddhisi4): Use VMAC2H instruction.
20962 (machi): New pattern.
20963 (umaddhisi4): Use VMAC2HU instruction.
20964 (umachi): New pattern.
20966 2021-06-09 Claudiu Zissulescu <claziss@synopsys.com>
20968 * config/arc/arc-protos.h (arc_split_move_p): New prototype.
20969 * config/arc/arc.c (arc_split_move_p): New function.
20970 (arc_split_move): Clean up.
20971 * config/arc/arc.md (movdi_insn): Clean up, use arc_split_move_p.
20972 (movdf_insn): Likewise.
20973 * config/arc/simdext.md (mov<VWH>_insn): Likewise.
20975 2021-06-09 Uroš Bizjak <ubizjak@gmail.com>
20978 * config/i386/i386.c (print_operand_address_as): Rename "no_rip"
20979 argument to "raw". Do not emit segment overrides when "raw" is true.
20981 2021-06-09 Martin Liska <mliska@suse.cz>
20983 * doc/gcov.texi: Create a proper JSON files.
20984 * doc/invoke.texi: Remove dots in order to make it a valid
20987 2021-06-09 Xionghu Luo <luoxhu@linux.ibm.com>
20989 * config/rs6000/rs6000-p8swap.c (pattern_is_rotate64): New.
20990 (insn_is_load_p): Use pattern_is_rotate64.
20991 (insn_is_swap_p): Likewise.
20992 (quad_aligned_load_p): Likewise.
20993 (const_load_sequence_p): Likewise.
20994 (replace_swapped_aligned_load): Likewise.
20995 (recombine_lvx_pattern): Likewise.
20996 (recombine_stvx_pattern): Likewise.
20998 2021-06-09 Andrew MacLeod <amacleod@redhat.com>
21000 * gimple-range-gori.cc (gori_compute::outgoing_edge_range_p): Use a
21001 fur_stmt source record.
21002 * gimple-range.cc (fur_source::get_operand): Generic range query.
21003 (fur_source::get_phi_operand): New.
21004 (fur_source::register_dependency): New.
21005 (fur_source::query): New.
21006 (class fur_edge): New. Edge source for operands.
21007 (fur_edge::fur_edge): New.
21008 (fur_edge::get_operand): New.
21009 (fur_edge::get_phi_operand): New.
21010 (fur_edge::query): New.
21011 (fur_stmt::fur_stmt): New.
21012 (fur_stmt::get_operand): New.
21013 (fur_stmt::get_phi_operand): New.
21014 (fur_stmt::query): New.
21015 (class fur_depend): New. Statement source and process dependencies.
21016 (fur_depend::fur_depend): New.
21017 (fur_depend::register_dependency): New.
21018 (class fur_list): New. List source for operands.
21019 (fur_list::fur_list): New.
21020 (fur_list::get_operand): New.
21021 (fur_list::get_phi_operand): New.
21022 (fold_range): New. Instantiate appropriate fur_source class and fold.
21023 (fold_using_range::range_of_range_op): Use new API.
21024 (fold_using_range::range_of_address): Ditto.
21025 (fold_using_range::range_of_phi): Ditto.
21026 (imple_ranger::fold_range_internal): Use fur_depend class.
21027 (fold_using_range::range_of_ssa_name_with_loop_info): Use new API.
21028 * gimple-range.h (class fur_source): Now a base class.
21029 (class fur_stmt): New.
21030 (fold_range): New prototypes.
21031 (fur_source::fur_source): Delete.
21033 2021-06-08 Andrew Pinski <apinski@marvell.com>
21035 PR tree-optimization/25290
21036 * tree-ssa-phiopt.c (xor_replacement): Delete.
21037 (tree_ssa_phiopt_worker): Delete use of xor_replacement.
21038 (match_simplify_replacement): Allow one cheap preparation
21039 statement that can be moved to before the if.
21041 2021-06-08 Pat Haugen <pthaugen@linux.ibm.com>
21043 * config/rs6000/power10.md (power10-fused-load, power10-fused-store,
21044 power10-fused_alu, power10-fused-vec, power10-fused-branch): New.
21046 2021-06-08 Jeff Law <jeffreyalaw@gmail.com>
21048 * config/h8300/logical.md (andqi3_1): Move BCLR case into define_insn_and_split.
21049 Create length attribute on define_insn_and_split. Only split for cases which we
21051 (andqi3_1<cczn>): Renamed from andqi3_1_clobber_flags. Only handle AND here and
21052 fix length computation.
21053 (b<code><mode>msx): Combine QImode and HImode H8/SX patterns using iterator.
21055 2021-06-08 Richard Biener <rguenther@suse.de>
21057 PR tree-optimization/100923
21058 * tree-ssa-sccvn.c (valueize_refs_1): Take a pointer to
21059 the operand vector to be valueized.
21060 (valueize_refs): Likewise.
21061 (valueize_shared_reference_ops_from_ref): Adjust.
21062 (valueize_shared_reference_ops_from_call): Likewise.
21063 (vn_reference_lookup_3): Likewise.
21064 (vn_reference_lookup_pieces): Likewise. Re-valueize
21065 with honoring availability when we are about to create
21066 the ao_ref and valueized before.
21067 (vn_reference_lookup): Likewise.
21068 (vn_reference_insert_pieces): Adjust.
21070 2021-06-08 Richard Biener <rguenther@suse.de>
21072 * tree-vectorizer.h (_slp_instance::root_stmt): Change to...
21073 (_slp_instance::root_stmts): ... a vector.
21074 (SLP_INSTANCE_ROOT_STMT): Rename to ...
21075 (SLP_INSTANCE_ROOT_STMTS): ... this.
21076 (slp_root::root): Change to...
21077 (slp_root::roots): ... a vector.
21078 (slp_root::slp_root): Adjust.
21079 * tree-vect-slp.c (_slp_instance::location): Adjust.
21080 (vect_free_slp_instance): Release the root stmt vector.
21081 (vect_build_slp_instance): Adjust.
21082 (vect_analyze_slp): Likewise.
21083 (_bb_vec_info::~_bb_vec_info): Likewise.
21084 (vect_slp_analyze_operations): Likewise.
21085 (vect_bb_vectorization_profitable_p): Likewise. Adjust
21086 costs for the root stmt.
21087 (vect_slp_check_for_constructors): Gather all BIT_INSERT_EXPRs
21089 (vect_slp_analyze_bb_1): Simplify by marking all root stmts
21091 (vectorize_slp_instance_root_stmt): Adjust.
21092 (vect_schedule_slp): Likewise.
21094 2021-06-08 Aldy Hernandez <aldyh@redhat.com>
21096 * gimple-ssa-evrp.c (class ssa_equiv_stack): New.
21097 (ssa_equiv_stack::ssa_equiv_stack): New.
21098 (ssa_equiv_stack::~ssa_equiv_stack): New.
21099 (ssa_equiv_stack::enter): New.
21100 (ssa_equiv_stack::leave): New.
21101 (ssa_equiv_stack::push_replacement): New.
21102 (ssa_equiv_stack::get_replacement): New.
21103 (is_pointer_ssa): New.
21104 (class pointer_equiv_analyzer): New.
21105 (pointer_equiv_analyzer::pointer_equiv_analyzer): New.
21106 (pointer_equiv_analyzer::~pointer_equiv_analyzer): New.
21107 (pointer_equiv_analyzer::set_global_equiv): New.
21108 (pointer_equiv_analyzer::set_cond_equiv): New.
21109 (pointer_equiv_analyzer::get_equiv): New.
21110 (pointer_equiv_analyzer::enter): New.
21111 (pointer_equiv_analyzer::leave): New.
21112 (pointer_equiv_analyzer::get_equiv_expr): New.
21113 (pta_valueize): New.
21114 (pointer_equiv_analyzer::visit_stmt): New.
21115 (pointer_equiv_analyzer::visit_edge): New.
21116 (hybrid_folder::value_of_expr): Call PTA.
21117 (hybrid_folder::value_on_edge): Same.
21118 (hybrid_folder::pre_fold_bb): New.
21119 (hybrid_folder::post_fold_bb): New.
21120 (hybrid_folder::pre_fold_stmt): New.
21121 (rvrp_folder::pre_fold_bb): New.
21122 (rvrp_folder::post_fold_bb): New.
21123 (rvrp_folder::pre_fold_stmt): New.
21124 (rvrp_folder::value_of_expr): Call PTA.
21125 (rvrp_folder::value_on_edge): Same.
21127 2021-06-08 Jakub Jelinek <jakub@redhat.com>
21130 * tree-inline.c (copy_tree_body_r): For OMP_CLAUSE_DEPEND don't
21131 check TREE_CODE if OMP_CLAUSE_DECL is NULL.
21133 2021-06-08 Richard Biener <rguenther@suse.de>
21135 PR middle-end/100951
21136 * tree-vect-generic.c (expand_vector_piecewise): Build a
21137 VECTOR_CST if all elements are constant.
21138 (expand_vector_condition): Likewise.
21139 (lower_vec_perm): Likewise.
21140 (expand_vector_conversion): Likewise.
21142 2021-06-08 Martin Liska <mliska@suse.cz>
21144 * doc/invoke.texi: Document new param evrp-sparse-threshold.
21146 2021-06-08 Martin Liska <mliska@suse.cz>
21148 * genautomata.c (create_automata): Fix typo.
21150 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
21152 PR tree-optimization/100794
21153 * tree-predcom.c (tree_predictive_commoning_loop): Add parameter
21154 allow_unroll_p and only allow unrolling when it's true.
21155 (tree_predictive_commoning): Add parameter allow_unroll_p and
21157 (run_tree_predictive_commoning): Likewise.
21158 (pass_predcom::gate): Check flag_tree_loop_vectorize and
21159 global_options_set.x_flag_predictive_commoning.
21160 (pass_predcom::execute): Adjust for allow_unroll_p.
21162 2021-06-08 Kewen Lin <linkw@linux.ibm.com>
21164 * tree-predcom.c (execute_pred_commoning): Remove update_ssa call.
21165 (tree_predictive_commoning_loop): Factor some cleanup stuffs into
21166 lambda function cleanup, remove scev_reset call, and adjust return
21168 (tree_predictive_commoning): Adjust for different changed values,
21169 only set flag TODO_update_ssa_only_virtuals if changed.
21170 (pass_data pass_data_predcom): Remove TODO_update_ssa_only_virtuals
21171 from todo_flags_finish.
21173 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
21175 * gimple-range-cache.cc (class sbr_sparse_bitmap): New.
21176 (sbr_sparse_bitmap::sbr_sparse_bitmap): New.
21177 (sbr_sparse_bitmap::bitmap_set_quad): New.
21178 (sbr_sparse_bitmap::bitmap_get_quad): New.
21179 (sbr_sparse_bitmap::set_bb_range): New.
21180 (sbr_sparse_bitmap::get_bb_range): New.
21181 (sbr_sparse_bitmap::bb_range_p): New.
21182 (block_range_cache::block_range_cache): initialize bitmap obstack.
21183 (block_range_cache::~block_range_cache): Destruct obstack.
21184 (block_range_cache::set_bb_range): Decide when to utilze the
21185 sparse on entry cache.
21186 * gimple-range-cache.h (block_range_cache): Add bitmap obstack.
21187 * params.opt (-param=evrp-sparse-threshold): New.
21189 2021-06-07 Andrew MacLeod <amacleod@redhat.com>
21191 * bitmap.c (bitmap_set_aligned_chunk): New.
21192 (bitmap_get_aligned_chunk): New.
21193 (test_aligned_chunk): New.
21194 (bitmap_c_tests): Call test_aligned_chunk.
21195 * bitmap.h (bitmap_set_aligned_chunk, bitmap_get_aligned_chunk): New.
21197 2021-06-07 Uroš Bizjak <ubizjak@gmail.com>
21200 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
21202 (ix86_expand_vector_init_one_nonzero): Ditto.
21203 (ix86_expand_vector_init_one_var): Ditto.
21204 (ix86_expand_vector_init_general): Ditto.
21205 * config/i386/mmx.md (vec_initv4qiqi): New expander.
21207 2021-06-07 Jeff Law <jeffreyalaw@gmail.com>
21209 * config/h8300/movepush.md: Change most _clobber_flags
21210 patterns to instead use <cczn> subst.
21211 (movsi_cczn): New pattern with usable CC cases split out.
21212 (movsi_h8sx_cczn): Likewise.
21214 2021-06-07 Martin Liska <mliska@suse.cz>
21216 * common/common-target.def: Split long lines and replace them
21218 * target.def: Likewise.
21219 * doc/tm.texi: Re-generated.
21221 2021-06-07 Jakub Jelinek <jakub@redhat.com>
21224 * fold-const.c (fold_read_from_vector): Return NULL if trying to
21225 read from a CONSTRUCTOR with vector type elements.
21227 2021-06-07 Jakub Jelinek <jakub@redhat.com>
21229 PR middle-end/100898
21230 * tree-inline.c (copy_bb): Only use gimple_call_arg_ptr if memcpy
21231 should copy any arguments. Don't call gimple_call_num_args
21232 on id->call_stmt or call_stmt more than once.
21234 2021-06-07 liuhongt <hongtao.liu@intel.com>
21237 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): Refine
21239 (<insn>v4siv4di2): Delete constraints for define_expand.
21241 2021-06-07 liuhongt <hongtao.liu@intel.com>
21244 * config/i386/i386-expand.c (ix86_expand_builtin): Remove
21245 assignment of cfun->machine->has_explicit_vzeroupper.
21246 * config/i386/i386-features.c
21247 (ix86_add_reg_usage_to_vzerouppers): Delete.
21248 (ix86_add_reg_usage_to_vzeroupper): Ditto.
21249 (rest_of_handle_insert_vzeroupper): Remove
21250 ix86_add_reg_usage_to_vzerouppers, add df_analyze at the end
21252 (gate): Remove cfun->machine->has_explicit_vzeroupper.
21253 * config/i386/i386-protos.h (ix86_expand_avx_vzeroupper):
21255 * config/i386/i386.c (ix86_insn_callee_abi): New function.
21256 (ix86_initialize_callee_abi): Ditto.
21257 (ix86_expand_avx_vzeroupper): Ditto.
21258 (ix86_hard_regno_call_part_clobbered): Adjust for vzeroupper
21260 (TARGET_INSN_CALLEE_ABI): Define as ix86_insn_callee_abi.
21261 (ix86_emit_mode_set): Call ix86_expand_avx_vzeroupper
21263 * config/i386/i386.h (struct GTY(()) machine_function): Delete
21264 has_explicit_vzeroupper.
21265 * config/i386/i386.md (enum unspec): New member
21267 (ABI_DEFAULT,ABI_VZEROUPPER,ABI_UNKNOWN): New
21268 define_constants for insn callee abi index.
21269 * config/i386/predicates.md (vzeroupper_pattern): Adjust.
21270 * config/i386/sse.md (UNSPECV_VZEROUPPER): Deleted.
21271 (avx_vzeroupper): Call ix86_expand_avx_vzeroupper.
21272 (*avx_vzeroupper): Rename to ..
21273 (avx_vzeroupper_callee_abi): .. this, and adjust pattern as
21274 call_insn which has a special vzeroupper ABI.
21275 (*avx_vzeroupper_1): Deleted.
21277 2021-06-07 liuhongt <hongtao.liu@intel.com>
21280 * df-scan.c (df_get_call_refs): When call_insn is a fake call,
21281 it won't use stack pointer reg.
21282 * final.c (leaf_function_p): When call_insn is a fake call, it
21283 won't affect caller as a leaf function.
21284 * reg-stack.c (callee_clobbers_any_stack_reg): New.
21285 (subst_stack_regs): When call_insn doesn't clobber any stack
21286 reg, don't clear the arguments.
21287 * rtl.c (shallow_copy_rtx): Don't clear flag used when orig is
21289 * shrink-wrap.c (requires_stack_frame_p): No need for stack
21290 frame for a fake call.
21291 * rtl.h (FAKE_CALL_P): New macro.
21293 2021-06-06 Eric Botcazou <ebotcazou@adacore.com>
21295 * config/sparc/sparc-protos.h (order_regs_for_local_alloc): Rename
21297 (sparc_order_regs_for_local_alloc): ...this.
21298 (sparc_leaf_reg_remap): Declare.
21299 * config/sparc/sparc.h (ADJUST_REG_ALLOC_ORDER): Adjust.
21300 (LEAF_REG_REMAP): Reimplement as call to sparc_leaf_reg_remap.
21301 * config/sparc/sparc.c (leaf_reg_remap): Delete.
21302 (order_regs_for_local_alloc): Rename to...
21303 (sparc_order_regs_for_local_alloc): ...this.
21304 (sparc_leaf_reg_remap): New function.
21305 (sparc_conditional_register_usage): Do not modify leaf_reg_remap.
21307 2021-06-06 David Edelsohn <dje.gcc@gmail.com>
21309 * config/rs6000/rs6000.c (rs6000_xcoff_asm_output_aligned_decl_common):
21310 Use assemble_name to output BSS section name.
21312 2021-06-06 Uroš Bizjak <ubizjak@gmail.com>
21314 * config/i386/constraints.md (Bs):
21315 Remove boolean operators from match_test RTX.
21318 (M): Use "mode" variable instead of GET_MODE (op) in match_test RTX.
21321 2021-06-06 Martin Liska <mliska@suse.cz>
21323 * doc/extend.texi: Add missing @headitem.
21324 * doc/invoke.texi: Likewise.
21325 * doc/objc.texi: Likewise.
21327 2021-06-06 Martin Liska <mliska@suse.cz>
21329 * genhooks.c (emit_findices): Remove unused function.
21330 (emit_documentation): Do not call emit_findices
21331 and do not search for @Fcode directives.
21333 2021-06-06 Martin Liska <mliska@suse.cz>
21335 * doc/invoke.texi: Remove extra character.
21337 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
21339 * config/sh/sh.md (doloop_end_split): Fix empty split condition.
21341 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
21343 * config/sparc/sparc.md (*snedi<W:mode>_zero_vis3,
21344 *neg_snedi<W:mode>_zero_subxc, *plus_snedi<W:mode>_zero,
21345 *plus_plus_snedi<W:mode>_zero, *minus_snedi<W:mode>_zero,
21346 *minus_minus_snedi<W:mode>_zero): Fix empty split condition.
21348 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
21350 * config/or1k/or1k.md (*movdi): Fix empty split condition.
21352 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
21354 * config/mips/mips.md (<anonymous>, bswapsi2, bswapdi2): Fix empty
21357 2021-06-05 Kewen Lin <linkw@linux.ibm.com>
21359 * config/m68k/m68k.md (*zero_extend_inc, *zero_extend_dec,
21360 *zero_extendsidi2): Fix empty split condition.
21362 2021-06-05 Jeff Law <jeffreyalaw@gmail.com>
21364 * config/h8300/addsub.md: Fix split condition in define_insn_and_split
21366 * config/h8300/bitfield.md: Likewise.
21367 * config/h8300/combiner.md: Likewise.
21368 * config/h8300/divmod.md: Likewise.
21369 * config/h8300/extensions.md: Likewise.
21370 * config/h8300/jumpcall.md: Likewise.
21371 * config/h8300/movepush.md: Likewise.
21372 * config/h8300/multiply.md: Likewise.
21373 * config/h8300/other.md: Likewise.
21374 * config/h8300/shiftrotate.md: Likewise.
21375 * config/h8300/logical.md: Likewise. Fix split pattern to use
21376 code iterator that somehow slipped through.
21378 2021-06-04 Tobias Burnus <tobias@codesourcery.com>
21380 PR middle-end/100905
21381 * tree-nested.c (convert_nonlocal_omp_clauses,
21382 convert_local_omp_clauses): Handle OMP_CLAUSE_BIND.
21384 2021-06-04 Martin Sebor <msebor@redhat.com>
21386 PR middle-end/100732
21387 * gimple-fold.c (gimple_fold_builtin_sprintf): Avoid folding calls
21388 with either source or destination argument of invalid type.
21389 * tree-ssa-uninit.c (maybe_warn_pass_by_reference): Avoid checking
21390 calls with arguments of invalid type.
21392 2021-06-04 Martin Sebor <msebor@redhat.com>
21394 * attribs.c (init_attr_rdwr_indices): Use VLA bounds in the expected
21396 (attr_access::vla_bounds): Also handle VLA bounds.
21398 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
21400 * config/i386/predicates.md (GOT_memory_operand):
21401 Implement using match_code RTXes.
21402 (GOT32_symbol_operand): Ditto.
21404 2021-06-04 Uroš Bizjak <ubizjak@gmail.com>
21407 * config/i386/i386-expand.c (ix86_expand_vector_init_duplicate):
21409 (ix86_expand_vector_init_general): Ditto.
21410 Use SImode instead of word_mode for logic operations
21411 when GET_MODE_SIZE (mode) < UNITS_PER_WORD.
21412 (expand_vec_perm_even_odd_1): Assert that V2HI mode should be
21413 implemented by expand_vec_perm_1.
21414 (expand_vec_perm_broadcast_1): Assert that V2HI and V4HI modes
21415 should be implemented using standard shuffle patterns.
21416 (ix86_vectorize_vec_perm_const): Handle V2HImode. Add V4HI and
21417 V2HI modes to modes, implementable with shuffle for one operand.
21418 * config/i386/mmx.md (*punpckwd): New insn_and_split pattern.
21419 (*pshufw_1): New insn pattern.
21420 (*vec_dupv2hi): Ditto.
21421 (vec_initv2hihi): New expander.
21423 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
21425 * config/arm/vfp.md (no_literal_pool_df_immediate,
21426 no_literal_pool_sf_immediate): Fix empty split condition.
21428 2021-06-04 Kewen Lin <linkw@linux.ibm.com>
21430 * config/i386/i386.md (*load_tp_x32_zext, *add_tp_x32_zext,
21431 *tls_dynamic_gnu2_combine_32): Fix empty split condition.
21432 * config/i386/sse.md (*<sse2_avx2>_pmovmskb_lt,
21433 *<sse2_avx2>_pmovmskb_zext_lt, *sse2_pmovmskb_ext_lt,
21434 *<sse4_1_avx2>_pblendvb_lt): Likewise.
21436 2021-06-04 Jakub Jelinek <jakub@redhat.com>
21439 * config/i386/i386-expand.c (ix86_expand_vector_init): Handle
21440 concatenation from half-sized modes with TImode elements.
21442 2021-06-04 Claudiu Zissulescu <claziss@synopsys.com>
21444 * config/arc/arc.c (arc_override_options): Disable millicode
21445 thunks when RF16 is on.
21447 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
21449 * config/rs6000/rs6000.h (PROMOTE_MODE): Remove.
21451 2021-06-04 Haochen Gui <guihaoc@gcc.gnu.org>
21453 * config/rs6000/rs6000-call.c (rs6000_promote_function_mode):
21454 Replace PROMOTE_MODE marco with its content.
21456 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
21458 * config/cris/cris.md (*addi_reload): Fix empty split condition.
21460 2021-06-03 Jim Wilson <jimw@sifive.com>
21462 * config.gcc (riscv*-*-*): If --with-riscv-attribute not used,
21463 turn it on for all riscv targets.
21465 2021-06-03 Uroš Bizjak <ubizjak@gmail.com>
21468 * config/i386/i386-expand.c (ix86_expand_vector_set):
21469 Handle V2HI and V4QI modes.
21470 (ix86_expand_vector_extract): Ditto.
21471 * config/i386/mmx.md (*pinsrw): New insn pattern.
21474 (*pextrw_zext): Ditto.
21476 (*pextrb_zext): Ditto.
21477 (vec_setv2hi): New expander.
21478 (vec_extractv2hihi): Ditto.
21479 (vec_setv4qi): Ditto.
21480 (vec_extractv4qiqi): Ditto.
21481 (vec_setv8qi): Enable only for TARGET_SSE4_1.
21482 (vec_extractv8qiqi): Ditto.
21484 2021-06-03 Aaron Sawdey <acsawdey@linux.ibm.com>
21486 * config/rs6000/genfusion.pl (gen_logical_addsubf): Fix input
21487 order to subf instruction.
21488 * config/rs6000/fusion.md: Regenerate.
21490 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
21492 * calls.c (get_size_range): Use range_of_expr instead of
21493 determine_value_range.
21494 * tree-affine.c (expr_to_aff_combination): Same.
21495 * tree-data-ref.c (split_constant_offset): Same.
21496 * tree-vrp.c (determine_value_range_1): Remove.
21497 (determine_value_range): Remove.
21498 * tree-vrp.h (determine_value_range): Remove.
21500 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
21502 * function-tests.c (test_ranges): Call gimple_range_tests.
21503 * gimple-range-cache.cc (ranger_cache::range_of_expr): Pass stmt
21505 * gimple-range.cc (fur_source::get_operand): Do not call
21506 get_tree_range or gimple_range_global.
21508 (get_tree_range): Move to value-query.cc.
21509 Call get_arith_expr_range.
21510 (gimple_ranger::range_of_expr): Add argument to get_tree_range.
21511 Include gimple-range-tests.cc.
21512 * gimple-range.h (fold_range): Add argument.
21513 (get_tree_range): Remove.
21514 * selftest.h (gimple_range_tests): New.
21515 * value-query.cc (global_range_query::range_of_expr): Add
21517 (range_query::get_tree_range): Move from gimple-range.cc.
21518 * value-query.h (class range_query): Add get_tree_range and
21519 get_arith_expr_range. Make fur_source a friend.
21520 * vr-values.c (vr_values::range_of_expr): Pass stmt to
21522 * gimple-range-tests.cc: New file.
21524 2021-06-03 Aldy Hernandez <aldyh@redhat.com>
21526 * gimple-range.cc (gimple_ranger::export_global_ranges): Call
21527 update_global_range.
21528 * value-query.cc (update_global_range): New.
21529 * value-query.h (update_global_range): New.
21531 2021-06-03 David Malcolm <dmalcolm@redhat.com>
21533 * diagnostic-show-locus.c (diagnostic_show_locus): Don't reject
21534 printing the same location twice if there are fix-it hints,
21535 multiple locations, or a label.
21537 2021-06-03 Andre Vieira <andre.simoesdiasvieira@arm.com>
21539 * tree-vect-loop.c (vect_transform_loop): Use main loop's various'
21540 thresholds to narrow the upper bound on epilogue iterations.
21542 2021-06-03 Christophe Lyon <christophe.lyon@linaro.org>
21544 * config/arm/mve.md (mve_vabsq_f<mode>): Use 'abs' instead of unspec.
21545 (mve_vabsq_s<mode>): Likewise.
21546 * config/arm/neon.md (abs<mode>2): Rename to neon_abs<mode>2.
21547 * config/arm/unspecs.md (VABSQ_F, VABSQ_S): Delete.
21548 * config/arm/vec-common.md (neg<mode>2): Rename to
21549 <absneg_str><mode>2.
21551 2021-06-03 Claudiu Zissulescu <claziss@synopsys.com>
21553 * common/config/arc/arc-common.c (arc_option_optimization_table):
21554 Remove malign-call.
21555 * config/arc/arc.c (arc_unalign_branch_p): Remove unused function.
21556 * config/arc/arc.h (TARGET_MIXED_CODE): Remove macro.
21557 (INDEX_REG_CLASS): Only refer to GENERAL_REGS.
21558 * config/arc/arc.md (abssi2_mixed): Remove pattern.
21559 * config/arc/arc.opt (munalign-prob-threshold): Mark it obsolete.
21560 (malign-call): Likewise.
21561 (mmixed-code): Likewise.
21562 * doc/invoke.texi (ARC): Update doc.
21564 2021-06-03 Martin Liska <mliska@suse.cz>
21566 * common.opt: Use proper Enum values.
21567 * opts.c (COVERAGE_SANITIZER_OPT): Remove.
21568 (parse_sanitizer_options): Handle only sanitizer_opts.
21569 (common_handle_option): Just assign value.
21571 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
21574 * tree-inline.c (inline_forbidden_p): Remove test on return type.
21576 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
21578 * dwarf2out.c (loc_list_from_tree_1) <FUNCTION_DECL>: Also generate
21579 DW_OP_GNU_variable_value referencing an existing DIE at file scope.
21580 (type_byte_size): Inline into...
21581 (add_byte_size_attribute): ...this and call add_scalar_info.
21583 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
21585 * dwarf2out.c (mem_loc_descriptor) <UDIV>: Fix typo.
21586 (typed_binop_from_tree): New function.
21587 (loc_list_from_tree_1) <EXACT_DIV_EXPR>: For an unsigned type,
21588 turn a divide by a power of 2 into a shift.
21589 <CEIL_DIV_EXPR>: For an unsigned type, use a signed divide if the
21590 size of the mode is lower than DWARF2_ADDR_SIZE; otherwise, do a
21591 typed divide by calling typed_binop_from_tree.
21593 2021-06-03 Eric Botcazou <ebotcazou@adacore.com>
21595 * dwarf2out.c (scompare_loc_descriptor): Fix head comment.
21596 (is_handled_procedure_type): Likewise.
21597 (struct loc_descr_context): Add strict_signedness field.
21598 (resolve_args_picking_1): Deal with DW_OP_[GNU_]deref_type,
21599 DW_OP_[GNU_]convert and DW_OP_[GNU_]reinterpret.
21600 (resolve_args_picking): Minor tweak.
21601 (function_to_dwarf_procedure): Initialize strict_signedness field.
21602 (type_byte_size): Likewise.
21603 (field_byte_offset): Likewise.
21604 (gen_descr_array_type_die): Likewise.
21605 (gen_variant_part): Likewise.
21606 (loc_list_from_tree_1) <CALL_EXPR>: Tidy up and set strict_signedness
21607 to true when a context is present before evaluating the arguments.
21608 <COND_EXPR>: Do not generate a useless comparison with zero.
21609 When dereferencing an address, if strict_signedness is true and the
21610 type is small and signed, use DW_OP_deref_type to do the dereference
21611 and then DW_OP_convert to convert back to the generic type.
21613 2021-06-03 Jakub Jelinek <jakub@redhat.com>
21616 * tree-inline.c (copy_tree_body_r): Handle iterators on
21617 OMP_CLAUSE_AFFINITY or OMP_CLAUSE_DEPEND.
21619 2021-06-03 Kewen Lin <linkw@linux.ibm.com>
21621 * config/arc/arc.md (*bbit_di): Remove.
21623 2021-06-02 Christoph Muellner <cmuellner@gcc.gnu.org>
21625 PR rtl-optimization/100264
21626 * ree.c (get_sub_rtx): Ignore SET expressions without register
21627 destinations and remove assertion, as it is not valid anymore
21628 with this new behaviour.
21629 (merge_def_and_ext): Eliminate destination check for register
21630 as such SET expressions can't occur anymore.
21631 (combine_reaching_defs): Likewise.
21633 2021-06-02 Jakub Jelinek <jakub@redhat.com>
21636 * config/xtensa/xtensa.h (LEAF_REG_REMAP): Cast REGNO to int to avoid
21637 -Wtype-limits warnings.
21638 (DWARF_FRAME_REGISTER): Rewrite into ternary operator with addition
21639 in operands to avoid -Wsign-compare warnings.
21641 2021-06-02 Pat Haugen <pthaugen@linux.ibm.com>
21643 * config/rs6000/rs6000-logue.c (rs6000_emit_prologue): Use
21646 2021-06-02 Vineet Gupta <vgupta@synopsys.com>
21648 * config/arc/arc.h (TARGET_CPU_DEFAULT): Change to hs38_linux.
21650 2021-06-02 Ilya Leoshkevich <iii@linux.ibm.com>
21652 * config/s390/s390.md(*ashrdi3_31<setcc><cconly>): Use a single
21654 * config/s390/subst.md(cconly_subst): Use a single constraint
21655 in (match_scratch).
21657 2021-06-02 Martin Liska <mliska@suse.cz>
21659 * ipa-icf.h: Use auto_vec for memory_access_types.
21661 2021-06-02 Jeff Law <jeffreyalaw@gmail.com>
21663 * config/h8300/h8300-protos.h (compute_a_shift_length): Drop unused
21664 argument from prototype.
21665 (output_logical_op): Add rtx_code argument.
21666 (compute_logical_op_length): Likewise.
21667 * config/h8300/h8300.c (h8300_and_costs): Pass additional argument
21668 to compute_a_shift_length.
21669 (output_logical_op); New argument with the rtx code rather than
21670 extracting it from an operand. Handle QImode too.
21671 (compute_logical_op_length): Similary.
21672 (compute_a_shift_length): Drop unused argument.
21673 * config/h8300/h8300.md (logicals): New code iterator.
21674 * config/h8300/logical.md (<code><mode>3 expander): Combine
21675 the "and" expander with the "ior"/"xor" expander.
21676 (bclr<mode>msx): Combine the QI/HI mode patterns.
21677 (<logical><mode>3 insns): Use code iterator rather than match_operator.
21678 Handle QImode as well. Update call to output_logical_op and
21679 compute_logical_op_length to pass in rtx_code
21680 Fix split condition on all define_insn_and_split patterns.
21681 (one_cmpl<mode>2<cczn>): Use <cczn> to support both clobbering
21682 the flags and setting ZN via existing define_subst.
21683 * config/h8300/shiftrotate.md: Drop unused argument from
21684 calls to compute_a_shift_length.
21685 Signed-off-by: Jeff Law <jeffreyalaw@gmail.com>
21687 2021-06-01 Andrew Pinski <apinski@marvell.com>
21689 PR tree-optimization/25290
21690 * tree-ssa-phiopt.c (match_simplify_replacement):
21692 (tree_ssa_phiopt_worker): Use match_simplify_replacement.
21693 (two_value_replacement): Change the comment about
21694 conditional_replacement.
21695 (conditional_replacement): Delete.
21697 2021-06-01 Andrew Pinski <apinski@marvell.com>
21699 PR tree-optimization/95481
21700 * tree-tailcall.c (find_tail_calls): Handle empty typed
21703 2021-06-01 Andrew Pinski <apinski@marvell.com>
21705 * gimplify.c (zero_sized_field_decl): Delete
21706 (zero_sized_type): Delete
21707 (gimplify_init_ctor_eval): Use is_empty_type instead
21708 of zero_sized_field_decl.
21709 (gimplify_modify_expr): Use is_empty_type instead of
21712 2021-06-01 Jason Merrill <jason@redhat.com>
21715 * tree.h (CALL_FROM_NEW_OR_DELETE_P): Adjust comment.
21717 2021-06-01 Jason Merrill <jason@redhat.com>
21720 * diagnostic.h (warning_enabled_at): Declare.
21721 * diagnostic.c (diagnostic_enabled): Factor out from...
21722 (diagnostic_report_diagnostic): ...here.
21723 (warning_enabled_at): New.
21725 2021-06-01 Aldy Hernandez <aldyh@redhat.com>
21727 * gimple-ssa-evrp.c: Enable exporting of global ranges.
21729 2021-06-01 Martin Liska <mliska@suse.cz>
21732 * doc/invoke.texi: Mention that -fgcse-after-reload
21733 is enabled with -O3.
21735 2021-06-01 liuhongt <hongtao.liu@intel.com>
21737 PR tree-optimization/98365
21738 * tree-if-conv.c (strip_nop_cond_scalar_reduction): New function.
21739 (is_cond_scalar_reduction): Handle nop_expr in cond scalar reduction.
21740 (convert_scalar_cond_reduction): Ditto.
21741 (predicate_scalar_phi): Ditto.
21743 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
21745 PR tree-optimization/100781
21746 * gimple-range-cache.cc (ranger_cache::ranger_cache): Enable new
21747 value calculation by default.
21748 (ranger_cache::enable_new_values): New.
21749 (ranger_cache::disable_new_values): New.
21750 (ranger_cache::push_poor_value): Check if new values are allowed.
21751 * gimple-range-cache.h (class ranger_cache): New member/methods.
21752 * gimple-range.cc (gimple_ranger::range_of_expr): Check for debug
21753 statement, and disable/renable new value calculation.
21755 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
21757 * gimple-range-cache.cc (ranger_cache::ssa_range_in_bb): Delete.
21758 (ranger_cache::range_of_def): New.
21759 (ranger_cache::entry_range): New.
21760 (ranger_cache::exit_range): New.
21761 (ranger_cache::range_of_expr): Adjust.
21762 (ranger_cache::range_on_edge): Adjust.
21763 (ranger_cache::propagate_cache): Call exit_range directly.
21764 * gimple-range-cache.h (class ranger_cache): Adjust.
21766 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
21768 * gimple-range-cache.cc (ranger_cache::ranger_cache): Adjust for
21769 gori_compute being a member rather than base class.
21770 dervied call to member call.
21771 (ranger_cache::dump): No longer dump gori_map.
21772 (ranger_cache::dump_bb): New.
21773 (ranger_cache::get_non_stale_global_range): Adjust for gori_compute
21774 being a member rather than base class.
21775 (ranger_cache::set_global_range): Ditto.
21776 (ranger_cache::ssa_range_in_bb): Ditto.
21777 (ranger_cache::range_of_expr): New.
21778 (ranger_cache::range_on_edge): New.
21779 (ranger_cache::block_range): Adjust for gori_computes. Debug changes.
21780 (ranger_cache::propagate_cache): Adjust debugging output.
21781 (ranger_cache::fill_block_cache): Adjust for gori_computes. Debug
21783 * gimple-range-cache.h (class ranger_cache): Make gori_compute a
21784 member, and inherit from range_query instead.
21785 (ranger_cache::dump_bb): New. split from dump.
21786 * gimple-range-gori.cc (gori_compute::ssa_range_in_bb): Delete.
21787 (gori_compute::expr_range_at_stmt): Delete.
21788 (gori_compute::compute_name_range_op): Delete.
21789 (gori_compute::compute_operand_range_switch): Add fur_source.
21790 (gori_compute::compute_operand_range): Add fur_source param, inline
21791 old compute_name_range_op and optimize_logical_operands.
21792 (struct tf_range): Delete.
21793 (gori_compute::logical_combine): Adjust
21794 (gori_compute::optimize_logical_operands): Delete.
21795 (gori_compute::compute_logical_operands_in_chain): Delete.
21796 (gori_compute::compute_logical_operands): Adjust.
21797 (gori_compute::compute_operand1_range): Adjust to fur_source.
21798 (gori_compute::compute_operand2_range): Ditto.
21799 (gori_compute::compute_operand1_and_operand2_range): Ditto.
21800 (gori_compute::outgoing_edge_range_p): Add range_query parameter,
21801 and adjust to fur_source.
21802 * gimple-range-gori.h (class gori_compute): Simplify and adjust to
21803 range_query and fur_source.
21804 * gimple-range.cc (gimple_ranger::range_on_edge): Query range_on_edge
21805 from the ranger_cache..
21806 (gimple_ranger::fold_range_internal): Adjust to base class change of
21808 (gimple_ranger::dump_bb): Adjust dump.
21809 * gimple-range.h (gimple_ranger):export gori computes object.
21811 2021-06-01 Andrew MacLeod <amacleod@redhat.com>
21813 PR tree-optimization/100774
21814 * gimple-range-cache.cc (ranger_cache::get_non_stale_global_range):
21815 Constant values are also not stale.
21816 (ranger_cache::set_global_range): Range invariant values should also
21817 have the correct timestamp.
21819 2021-05-31 Martin Liska <mliska@suse.cz>
21821 * tree-streamer-in.c (unpack_ts_function_decl_value_fields):
21822 Unpack FUNCTION_DECL_DECL_TYPE.
21823 * tree-streamer-out.c (pack_ts_function_decl_value_fields):
21824 Stream FUNCTION_DECL_DECL_TYPE instead of
21825 DECL_IS_OPERATOR_NEW_P.
21826 * tree.h (set_function_decl_type): Use FUNCTION_DECL_DECL_TYPE
21828 (DECL_IS_OPERATOR_NEW_P): Likewise.
21829 (DECL_IS_OPERATOR_DELETE_P): Likewise.
21830 (DECL_LAMBDA_FUNCTION_P): Likewise.
21832 2021-05-31 Richard Biener <rguenther@suse.de>
21835 * internal-fn.c (expand_SHUFFLEVECTOR): Define.
21836 * internal-fn.def (SHUFFLEVECTOR): New.
21837 * internal-fn.h (expand_SHUFFLEVECTOR): Declare.
21838 * doc/extend.texi: Document __builtin_shufflevector.
21840 2021-05-31 Peter Bergner <bergner@linux.ibm.com>
21843 * config/rs6000/predicates.md(mma_assemble_input_operand): Allow
21844 indexed form addresses.
21846 2021-05-29 Jeff Law <jlaw@tachyum.com>
21848 * config/h8300/h8300.c (h8300_emit_stack_adjustment): Drop unused
21849 parameter. Call callers fixed.
21851 (output_plussi): Add FALLTHRU markers.
21852 (h8300_shift_needs_scratch_p): Add gcc_unreachable marker.
21854 2021-05-29 Jakub Jelinek <jakub@redhat.com>
21856 PR middle-end/99928
21857 * gimplify.c (gimplify_scan_omp_clauses): For taskloop simd
21858 combined with parallel, make sure to add shared clause to
21859 parallel for explicit linear clause.
21861 2021-05-29 Aldy Hernandez <aldyh@redhat.com>
21863 PR tree-optimization/100787
21864 * gimple-ssa-evrp.c: Disable exporting of global ranges.
21866 2021-05-28 Jason Merrill <jason@redhat.com>
21868 * tree-iterator.h (struct tree_stmt_iterator): Add operator++,
21869 operator--, operator*, operator==, and operator!=.
21870 (class tsi_range): New.
21872 2021-05-28 Richard Biener <rguenther@suse.de>
21874 PR tree-optimization/100778
21875 * tree-vect-slp.c (vect_build_slp_tree_1): Prevent possibly
21876 trapping ops in different BBs.
21878 2021-05-28 Richard Biener <rguenther@suse.de>
21881 * tree-inline.c (copy_bb): When processing __builtin_va_arg_pack
21882 copy fntype from original call.
21884 2021-05-28 Martin Liska <mliska@suse.cz>
21886 PR gcov-profile/100751
21887 * doc/gcov.texi: Revert partially a hunk that was wrong.
21889 2021-05-28 Cooper Qu <cooper.qu@linux.alibaba.com>
21891 * config/csky/csky-linux-elf.h (HAVE_sync_compare_and_swapqi):
21893 (HAVE_sync_compare_and_swaphi): Likewise.
21894 (HAVE_sync_compare_and_swapsi): Likewise.
21896 2021-05-28 Jakub Jelinek <jakub@redhat.com>
21898 PR middle-end/99928
21899 * tree.h (OMP_CLAUSE_MAP_IMPLICIT): Define.
21901 2021-05-28 Tobias Burnus <tobias@codesourcery.com>
21903 * gimplify.c (gimplify_omp_affinity): New.
21904 (gimplify_scan_omp_clauses): Call it; remove affinity clause afterwards.
21905 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_AFFINITY.
21906 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_AFFINITY.
21907 * tree.c (omp_clause_num_ops, omp_clause_code_name): Add clause.
21908 (walk_tree_1): Handle OMP_CLAUSE_AFFINITY.
21910 2021-05-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
21911 Richard Biener <rguenther@suse.de>
21913 * match.pd <popcount & / + pattern matching>:
21914 When generating popcount directly fails, try doing it in two halves.
21916 2021-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
21918 * Makefile.in (generated_files): Add gimple-match.c and
21921 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
21923 * gensupport.c (alter_predicate_for_insn): Handle MATCH_DUP.
21925 2021-05-28 Joern Rennecke <joern.rennecke@embecosm.com>
21927 * gensupport.c (alter_constraints): Add MATCH_SCRATCH case.
21929 2021-05-28 Kewen Lin <linkw@linux.ibm.com>
21931 PR tree-optimization/99398
21932 * tree-ssa-forwprop.c (simplify_permutation): Optimize some cases
21933 where the fed operands are CTOR/CST and propagated through
21934 VIEW_CONVERT_EXPR. Call vec_perm_indices::new_shrunk_vector.
21935 * vec-perm-indices.c (vec_perm_indices::new_shrunk_vector): New
21937 * vec-perm-indices.h (vec_perm_indices::new_shrunk_vector): New
21940 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
21942 * config/i386/mmx.md (addv2sf3): Do not call
21943 ix86_fixup_binary_operands_no_copy.
21946 (<smaxmin:code>v2sf3): Ditto.
21947 (<plusminus:insn><MMXMODEI:mode>3): Ditto.
21948 (<plusminus:insn><VI_32:mode>3): Remove expander.
21949 (<plusminus:insn><VI_32:mode>3): Rename from
21950 "*<plusminus:insn><VI_32:mode>3".
21951 (mulv4hi): Do not call ix86_fixup_binary_operands_no_copy.
21952 (mulv2hi3): Remove expander.
21953 (mulv2hi3): Rename from *mulv2hi3.
21954 (<s>mulv2hi3_highpart): Remove expander.
21955 (<s>mulv2hi3_highpart): Rename from *<s>mulv2hi3_highpart.
21956 (<smaxmin:code><MMXMODE14:mode>3): Rename from
21957 "*mmx_<smaxmin:code><MMXMODE14:mode>3".
21958 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Remove expander.
21959 (SMAXMIN_MMXMODEI): Remove mode iterator.
21960 (<smaxmin:code>v4hi3): New expander.
21961 (<smaxmin:code>v4qi3): Rename from *<smaxmin:code>v4qi3.
21962 (<smaxmin:code>v2hi3): Rename from *<smaxmin:code>v2hi3.
21963 (<smaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
21964 (SMAXMIN_VI_32): Remove mode iterator.
21965 (<umaxmin:code><MMXMODE24:mode>3): Rename from
21966 "*mmx_<umaxmin:code><MMXMODE24:mode>3".
21967 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Remove expander.
21968 (UMAXMIN_MMXMODEI): Remove mode iterator.
21969 (<umaxmin:code>v8qi3): New expander.
21970 (<umaxmin:code>v4qi3): Rename from *<umaxmin:code>v4qi3.
21971 (<umaxmin:code>v2hi3): Rename from *<umaxmin:code>v2hi3.
21972 (<umaxmin:code><SMAXMIN_VI_32:mode>3): Remove expander.
21973 (UMAXMIN_VI_32): Remove mode iterator.
21974 (<any_shift:insn>v2hi3): Remove expander.
21975 (<any_shift:insn>v2hi3): Rename from *<any_shift:insn>v2hi3.
21976 (<any_logic:code><MMXMODEI:mode>3): Do not call
21977 ix86_fixup_binary_operands_no_copy.
21978 (<any_logic:code><VI_32:mode>3): Remove expander.
21979 (<any_logic:code><VI_32:mode>3): Rename from
21980 "*<any_logic:code><VI_32:mode>3".
21981 (uavg<mode>3_ceil): Do not call ix86_fixup_binary_operands_no_copy.
21982 * config/i386/sse.md (div<VF2:mode>3): Do not call
21983 ix86_fixup_binary_operands_no_copy.
21984 (div<VF1:mode>3): Ditto.
21985 (<maxmin:code><VI8_AVX2_AVX512F:mode>3): Ditto.
21986 (smulhrsv4hi3): Ditto.
21987 (smulhrsv2hi3): Ditto.
21989 2021-05-27 Martin Sebor <msebor@redhat.com>
21991 * ggc.h (gt_ggc_mx): Add overloads for all integers.
21993 * hash-map.h (class hash_map): Add pch_nx_helper overloads for all
21995 (hash_map::operator==): New function.
21997 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
22000 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
22001 For TARGET_XOP bypass SSE comparisons for all supported vector modes.
22002 * config/i386/mmx.md (*xop_maskcmp<MMXMODEI:mode>3): New insn pattern.
22003 (*xop_maskcmp<VI_32:mode>3): Ditto.
22004 (*xop_maskcmp_uns<MMXMODEI:mode>3): Ditto.
22005 (*xop_maskcmp_uns<VI_32:mode>3): Ditto.
22007 2021-05-27 Richard Earnshaw <rearnsha@arm.com>
22010 * config/arm/arm.c (arm_configure_build_target): Remove parameter
22011 opts_set, directly check opts parameters for being non-null.
22012 (arm_option_restore): Update call to arm_configure_build_target.
22013 (arm_option_override): Likewise.
22014 (arm_can_inline_p): Likewise.
22015 (arm_valid_target_attribute_tree): Likewise.
22016 * config/arm/arm-c.c (arm_pragma_target_parse): Likewise.
22017 * config/arm/arm-protos.h (arm_configure_build_target): Adjust
22020 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
22022 * vr-values.c (simplify_conversion_using_ranges): Use
22023 get_range_query instead of get_global_range_query.
22025 2021-05-27 Aldy Hernandez <aldyh@redhat.com>
22027 * gimple-range.cc (get_range_global): Move to value-query.cc.
22028 (gimple_range_global): Same.
22029 (get_global_range_query): Same.
22030 (global_range_query::range_of_expr): Same.
22031 * gimple-range.h (class global_range_query): Move to
22033 (gimple_range_global): Same.
22034 * tree-ssanames.c (get_range_info): Move to value-query.cc.
22035 (get_ptr_nonnull): Same.
22036 * tree-ssanames.h (get_range_info): Remove.
22037 (get_ptr_nonnull): Remove.
22038 * value-query.cc (get_ssa_name_range_info): Move from
22040 (get_ssa_name_ptr_info_nonnull): Same.
22041 (get_range_global): Move from gimple-range.cc.
22042 (gimple_range_global): Same.
22043 (get_global_range_query): Same.
22044 (global_range_query::range_of_expr): Same.
22045 * value-query.h (class global_range_query): Move from
22047 (gimple_range_global): Same.
22049 2021-05-27 Uroš Bizjak <ubizjak@gmail.com>
22052 * config/i386/mmx.md (uavgv4qi3_ceil): New insn pattern.
22053 (uavgv2hi3_ceil): Ditto.
22055 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
22058 * doc/extend.texi (scalar_storage_order): Rephrase slightly.
22060 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
22062 * tree-ssanames.c (get_range_info): Merge both copies of
22063 get_range_info into one that works with irange.
22064 * tree-ssanames.h (get_range_info): Remove version that works on
22067 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
22069 * builtins.c (check_nul_terminated_array): Convert to get_range_query.
22070 (expand_builtin_strnlen): Same.
22071 (determine_block_size): Same.
22072 * fold-const.c (expr_not_equal_to): Same.
22073 * gimple-fold.c (size_must_be_zero_p): Same.
22074 * gimple-match-head.c: Include gimple-range.h.
22075 * gimple-pretty-print.c (dump_ssaname_info): Convert to get_range_query.
22076 * gimple-ssa-warn-restrict.c
22077 (builtin_memref::extend_offset_range): Same.
22078 * graphite-sese-to-poly.c (add_param_constraints): Same.
22079 * internal-fn.c (get_min_precision): Same.
22080 * ipa-fnsummary.c (set_switch_stmt_execution_predicate): Same.
22081 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Same.
22083 * tree-data-ref.c (split_constant_offset): Same.
22084 (dr_step_indicator): Same.
22085 * tree-dfa.c (get_ref_base_and_extent): Same.
22086 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
22087 * tree-ssa-loop-niter.c (refine_value_range_using_guard): Same.
22088 (determine_value_range): Same.
22089 (record_nonwrapping_iv): Same.
22090 (infer_loop_bounds_from_signedness): Same.
22091 (scev_var_range_cant_overflow): Same.
22092 * tree-ssa-phiopt.c (two_value_replacement): Same.
22093 * tree-ssa-pre.c (insert_into_preds_of_block): Same.
22094 * tree-ssa-reassoc.c (optimize_range_tests_to_bit_test): Same.
22095 * tree-ssa-strlen.c (handle_builtin_stxncpy_strncat): Same.
22097 (dump_strlen_info): Same.
22098 (set_strlen_range): Same.
22099 (maybe_diag_stxncpy_trunc): Same.
22100 (get_len_or_size): Same.
22101 (handle_integral_assign): Same.
22102 * tree-ssa-structalias.c (find_what_p_points_to): Same.
22103 * tree-ssa-uninit.c (find_var_cmp_const): Same.
22104 * tree-switch-conversion.c (bit_test_cluster::emit): Same.
22105 * tree-vect-patterns.c (vect_get_range_info): Same.
22106 (vect_recog_divmod_pattern): Same.
22107 * tree-vrp.c (intersect_range_with_nonzero_bits): Same.
22108 (register_edge_assert_for_2): Same.
22109 (determine_value_range_1): Same.
22110 * tree.c (get_range_pos_neg): Same.
22111 * vr-values.c (vr_values::get_lattice_entry): Same.
22112 (vr_values::update_value_range): Same.
22113 (simplify_conversion_using_ranges): Same.
22115 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
22117 * gimple-ssa-warn-alloca.c (alloca_call_type): Use
22118 get_range_query instead of query argument.
22119 (pass_walloca::execute): Enable and disable global ranger.
22121 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
22123 * gimple-ssa-evrp.c (rvrp_folder::rvrp_folder): Call
22125 (rvrp_folder::~rvrp_folder): Call disable_ranger.
22126 (hybrid_folder::hybrid_folder): Call enable_ranger.
22127 (hybrid_folder::~hybrid_folder): Call disable_ranger.
22129 2021-05-26 Aldy Hernandez <aldyh@redhat.com>
22131 * function.c (allocate_struct_function): Set cfun->x_range_query.
22132 * function.h (struct function): Declare x_range_query.
22133 (get_range_query): New.
22134 (get_global_range_query): New.
22135 * gimple-range-cache.cc (ssa_global_cache::ssa_global_cache):
22136 Remove call to safe_grow_cleared.
22137 * gimple-range.cc (get_range_global): New.
22138 (gimple_range_global): Move from gimple-range.h.
22139 (get_global_range_query): New.
22140 (global_range_query::range_of_expr): New.
22141 (enable_ranger): New.
22142 (disable_ranger): New.
22143 * gimple-range.h (gimple_range_global): Move to gimple-range.cc.
22144 (class global_range_query): New.
22145 (enable_ranger): New.
22146 (disable_ranger): New.
22147 * gimple-ssa-evrp.c (evrp_folder::~evrp_folder): Rename
22148 dump_all_value_ranges to dump.
22149 * tree-vrp.c (vrp_prop::finalize): Same.
22150 * value-query.cc (range_query::dump): New.
22151 * value-query.h (range_query::dump): New.
22152 * vr-values.c (vr_values::dump_all_value_ranges): Rename to...
22153 (vr_values::dump): ...this.
22154 * vr-values.h (class vr_values): Rename dump_all_value_ranges to
22155 dump and make virtual.
22157 2021-05-26 Uroš Bizjak <ubizjak@gmail.com>
22159 * config/i386/i386.c (ix86_autovectorize_vector_modes):
22160 Add V4QImode and V16QImode for TARGET_SSE2.
22161 * doc/sourcebuild.texi (Vector-specific attributes):
22162 Add vect64 and vect32 description.
22164 2021-05-26 Bernd Edlinger <bernd.edlinger@hotmail.de>
22166 * gimple-range-gori.cc (range_def_chain::register_dependency):
22167 Resize m_def_chain when needed.
22169 2021-05-26 Christophe Lyon <christophe.lyon@linaro.org>
22171 * config/arm/mve.md (mve_vaddvq_<supf><mode>): Prefix with '@'.
22172 * config/arm/neon.md (reduc_plus_scal_<mode>): Move to ..
22173 * config/arm/vec-common.md: .. here. Add support for MVE.
22175 2021-05-26 Jakub Jelinek <jakub@redhat.com>
22177 * config/epiphany/epiphany.c (epiphany_print_operand_address): Remove
22179 * config/microblaze/microblaze.c (microblaze_legitimize_address,
22181 microblaze_option_override, print_operand): Likewise.
22182 * config/microblaze/microblaze.md (call_internal_plt,
22183 call_value_intern_plt, call_value_intern): Likewise.
22184 * config/arm/aout.h (ASM_OUTPUT_ALIGN): Likewise.
22185 * config/iq2000/iq2000.md (call_internal1, call_value_internal1,
22186 call_value_multiple_internal1): Likewise.
22187 * config/bfin/bfin.c (symbolic_reference_mentioned_p): Likewise.
22189 2021-05-26 Jan-Benedict Glaw <jbglaw@lug-owl.de>
22191 * config/arc/arc.c (arc_address_cost, arc_print_operand_address,
22192 arc_ccfsm_advance, symbolic_reference_mentioned_p,
22193 arc_raw_symbolic_reference_mentioned_p): Remove register
22196 2021-05-26 Jakub Jelinek <jakub@redhat.com>
22199 * omp-low.c: Include omp-offload.h.
22200 (create_omp_child_function): If current_function_decl has
22201 "omp declare target" attribute and is_gimple_omp_offloaded,
22202 remove that attribute from the copy of attribute list and
22203 add "omp target entrypoint" attribute instead.
22204 (lower_omp_target): Mark .omp_data_sizes.* and .omp_data_kinds.*
22205 variables for offloading if in omp_maybe_offloaded_ctx.
22206 * omp-offload.c (pass_omp_target_link::execute): Nullify second
22207 argument to GOMP_target_data_ext in offloaded code.
22209 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
22211 * config/csky/csky.c (csky_can_change_mode_class): Delete.
22212 For csky, HF/SF mode use the low bits of VREGS.
22214 2021-05-26 Eric Botcazou <ebotcazou@adacore.com>
22216 * gimplify.c (gimplify_decl_expr): Do not clear TREE_READONLY on a
22217 DECL which is a reference for OMP.
22219 2021-05-26 Martin Liska <mliska@suse.cz>
22221 PR gcov-profile/100751
22222 * doc/gcov.texi: Document that __gcov_dump can be called just
22223 once and that __gcov_reset resets run-time counters.
22225 2021-05-26 Martin Liska <mliska@suse.cz>
22227 * doc/install.texi: Port relevant part from install-old.texi
22228 and re-generate list of CPUs and systems.
22230 2021-05-26 Martin Liska <mliska@suse.cz>
22232 * Makefile.in: Remove it.
22233 * doc/include/fdl.texi: Update next/previous chapters.
22234 * doc/install.texi: Likewise.
22235 * doc/install-old.texi: Removed.
22237 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
22239 * config/csky/csky.c (ck810_legitimate_index_p): Support
22240 "base + index" with DF mode.
22241 * config/csky/constraints.md ("Y"): New constraint for memory operands
22242 without index register.
22243 * config/csky/csky_insn_fpuv2.md (fpuv3_movdf): Use "Y" instead of "m"
22244 when mov between memory and general registers, and lower their priority.
22245 * config/csky/csky_insn_fpuv3.md (fpuv2_movdf): Likewise.
22247 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
22249 * config/csky/csky.c (TARGET_PROMOTE_PROTOTYPES): Delete.
22251 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
22253 * config/csky/csky.md (untyped_call): Emit clobber for return
22254 registers to mark them used.
22256 2021-05-26 Geng Qi <gengqi@linux.alibaba.com>
22258 * config/csky/csky.md (cskyv2_sextend_ldbs): New.
22260 2021-05-26 Andrew Pinski <apinski@marvell.com>
22262 * match.pd (x < 0 ? ~y : y): New patterns.
22264 2021-05-26 Andrew Pinski <apinski@marvell.com>
22266 * match.pd (A?CST1:CST2): Add simplifcations for A?0:+-1, A?+-1:0,
22267 A?POW2:0 and A?0:POW2.
22269 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
22271 * gimple-range-gori.cc (class logical_stmt_cache): Delete
22272 (logical_stmt_cache::logical_stmt_cache ): Delete.
22273 (logical_stmt_cache::~logical_stmt_cache): Delete.
22274 (logical_stmt_cache::cache_entry::dump): Delete.
22275 (logical_stmt_cache::get_range): Delete.
22276 (logical_stmt_cache::cached_name ): Delete.
22277 (logical_stmt_cache::same_cached_name): Delete.
22278 (logical_stmt_cache::cacheable_p): Delete.
22279 (logical_stmt_cache::slot_diagnostics ): Delete.
22280 (logical_stmt_cache::dump): Delete.
22281 (gori_compute_cache::gori_compute_cache): Delete.
22282 (gori_compute_cache::~gori_compute_cache): Delete.
22283 (gori_compute_cache::compute_operand_range): Delete.
22284 (gori_compute_cache::cache_stmt): Delete.
22285 * gimple-range-gori.h (gori_compute::compute_operand_range): Remove
22287 (class gori_compute_cache): Delete.
22289 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
22291 * gimple-range.cc (fold_using_range::range_of_range_op): Use m_gori
22293 (fold_using_range::range_of_address): Adjust.
22294 (fold_using_range::range_of_phi): Adjust.
22295 * gimple-range.h (class fur_source): Adjust.
22296 (fur_source::fur_source): Adjust.
22298 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
22300 * gimple-range-gori.cc (gori_compute::expr_range_at_stmt): Rename
22301 from expr_range_in_bb and adjust.
22302 (gori_compute::compute_name_range_op): Adjust.
22303 (gori_compute::optimize_logical_operands): Adjust.
22304 (gori_compute::compute_logical_operands_in_chain): Adjust.
22305 (gori_compute::compute_operand1_range): Adjust.
22306 (gori_compute::compute_operand2_range): Adjust.
22307 (ori_compute_cache::cache_stmt): Adjust.
22308 * gimple-range-gori.h (gori_compute): Rename prototype.
22310 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
22312 * gimple-range.cc (gimple_ranger::range_of_expr): Non-null should be
22313 checked only after range_of_stmt, not range_on_entry.
22314 (gimple_ranger::range_on_entry): Check for non-null in any
22315 predecessor block, if it is not already non-null.
22316 (gimple_ranger::range_on_exit): DOnt check for non-null after
22317 range on entry call.
22318 (gimple_ranger::dump_bb): New. Split from dump.
22319 (gimple_ranger::dump): Adjust.
22320 * gimple-range.h (class gimple_ranger): Adjust.
22322 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
22324 * gimple-range-cache.cc (struct range_timestamp): Delete.
22325 (class temporal_cache): Adjust.
22326 (temporal_cache::get_timestamp): Delete.
22327 (temporal_cache::set_dependency): Delete.
22328 (temporal_cache::temporal_value): Adjust.
22329 (temporal_cache::current_p): Take dependencies as params.
22330 (temporal_cache::set_timestamp): Adjust.
22331 (temporal_cache::set_always_current): Adjust.
22332 (ranger_cache::get_non_stale_global_range): Adjust.
22333 (ranger_cache::register_dependency): Delete.
22334 * gimple-range-cache.h (class range_cache): Adjust.
22336 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
22338 * gimple-range-gori.cc (range_def_chain::range_def_chain): init
22340 (range_def_chain::~range_def_chain): Dispose of obstack rather than
22341 each individual bitmap.
22342 (range_def_chain::set_import): New.
22343 (range_def_chain::get_imports): New.
22344 (range_def_chain::chain_import_p): New.
22345 (range_def_chain::register_dependency): Rename from build_def_chain
22347 (range_def_chain::def_chain_in_bitmap_p): New.
22348 (range_def_chain::add_def_chain_to_bitmap): New.
22349 (range_def_chain::has_def_chain): Just check first depenedence.
22350 (range_def_chain::get_def_chain): Process imports, use generic
22351 register_dependency routine.
22352 (range_def_chain::dump): New.
22353 (gori_map::gori_map): Allocate import list.
22354 (gori_map::~gori_map): Release imports.
22355 (gori_map::exports): Check for past allocated block size.
22356 (gori_map::imports): New.
22357 (gori_map::def_chain_in_export_p): Delete.
22358 (gori_map::is_import_p): New.
22359 (gori_map::maybe_add_gori): Handle imports.
22360 (gori_map::dump): Adjust output, add imports.
22361 (gori_compute::has_edge_range_p): Remove def_chain_in_export call.
22362 (gori_export_iterator::gori_export_iterator): New.
22363 (gori_export_iterator::next): New.
22364 (gori_export_iterator::get_name): New.
22365 * gimple-range-gori.h (range_def_chain): Add imports and direct
22366 dependecies via struct rdc.
22367 (range_def_chain::depend1): New.
22368 (range_def_chain::depend2): New.
22369 (class gori_map): Adjust.
22370 (FOR_EACH_GORI_IMPORT_NAME): New.
22371 (FOR_EACH_GORI_EXPORT_NAME): New.
22372 (class gori_export_iterator): New.
22374 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
22376 * gimple-range-cache.cc (ranger_cache::ranger_cache): Move initial
22377 export cache filling to here.
22378 * gimple-range-gori.cc (gori_compute::gori_compute) : From Here.
22380 2021-05-25 Andrew MacLeod <amacleod@redhat.com>
22382 * gimple-range-gori.cc (range_def_chain): Move to gimple-range-gori.h.
22383 (gori_map): Move to gimple-range-gori.h.
22384 (gori_compute::gori_compute): Adjust.
22385 (gori_compute::~gori_compute): Delete.
22386 (gori_compute::compute_operand_range_switch): Adjust.
22387 (gori_compute::compute_operand_range): Adjust.
22388 (gori_compute::compute_logical_operands): Adjust.
22389 (gori_compute::has_edge_range_p ): Adjust.
22390 (gori_compute::set_range_invariant): Delete.
22391 (gori_compute::dump): Adjust.
22392 (gori_compute::outgoing_edge_range_p): Adjust.
22393 * gimple-range-gori.h (class range_def_chain): Relocate here.
22394 (class gori_map): Relocate here.
22395 (class gori_compute): Inherit from gori_map, and adjust.
22397 2021-05-25 Aldy Hernandez <aldyh@redhat.com>
22399 * value-range.cc (range_tests_legacy): Use
22400 build_nonstandard_integer_type instead of int and short.
22402 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
22404 * gimplify.c (gimplify_decl_expr): Clear TREE_READONLY on the DECL
22405 when really creating an initialization statement for it.
22407 2021-05-25 Eric Botcazou <ebotcazou@adacore.com>
22409 * tree-inline.c (setup_one_parameter): Fix thinko in new condition.
22411 2021-05-25 Kito Cheng <kito.cheng@sifive.com>
22413 * config/riscv/riscv.h (ASM_SPEC): Pass -mno-relax.
22415 2021-05-25 Martin Liska <mliska@suse.cz>
22417 PR tree-optimization/92860
22419 * optc-save-gen.awk: Remove exceptions.
22421 2021-05-25 Martin Liska <mliska@suse.cz>
22423 * asan.h (sanitize_coverage_p): New function.
22424 * doc/extend.texi: Document it.
22425 * fold-const.c (fold_range_test): Use sanitize_flags_p
22426 instead of flag_sanitize_coverage.
22427 (fold_truth_andor): Likewise.
22428 * sancov.c: Likewise.
22429 * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
22430 * ipa-inline.c (sanitize_attrs_match_for_inline_p): Handle
22431 -fsanitize-coverage when inlining.
22433 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
22435 * config/csky/csky-modes.def : Fix copyright.
22437 2021-05-25 Cooper Qu <cooper.qu@linux.alibaba.com>
22439 * config/csky/csky-modes.def : Amend copyright.
22440 * config/csky/csky_insn_fpuv2.md : Likewise.
22441 * config/csky/csky_insn_fpuv3.md : Likewise.
22443 2021-05-25 Richard Biener <rguenther@suse.de>
22445 PR middle-end/100727
22446 * calls.c (initialize_argument_information): Explicitely test
22447 for WITH_SIZE_EXPR.
22448 * gimple-expr.c (mark_addressable): Skip outer WITH_SIZE_EXPR.
22450 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
22452 * config/csky/csky.h (FRAME_POINTER_REGNUM): Use
22453 HARD_FRAME_POINTER_REGNUM and FRAME_POINTER_REGNUM instead of
22454 the signle definition. The signle definition may not work well
22455 at simplify_subreg_regno().
22456 (HARD_FRAME_POINTER_REGNUM): New.
22457 (ELIMINABLE_REGS): Add for HARD_FRAME_POINTER_REGNUM.
22458 * config/csky/csky.c (get_csky_live_regs, csky_can_eliminate,
22459 csky_initial_elimination_offset, csky_expand_prologue,
22460 csky_expand_epilogue): Add for HARD_FRAME_POINTER_REGNUM.
22462 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
22464 * config/csky/csky.c (csky_option_override):
22465 Init csky_arch_isa_features[] in advance, so TARGET_DSP
22466 and TARGET_DIV can be set well.
22468 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
22470 * config/csky/constraints.md ("l", "h"): Delete.
22471 * config/csky/csky.h (reg_class, REG_CLASS_NAMES,
22472 REG_CLASS_CONTENTS): Delete LO_REGS and HI_REGS.
22473 * config/csky/csky.c (regno_reg_classm,
22474 csky_secondary_reload, csky_register_move_cost):
22475 Use HILO_REGS instead of LO_REGS and HI_REGS.
22477 2021-05-25 Geng Qi <gengqi@linux.alibaba.com>
22479 * config/csky/constraints.md ("W"): New constriant for mem operand
22480 with base reg, index register.
22481 ("Q"): Renamed and modified "csky_valid_fpuv2_mem_operand" to
22482 "csky_valid_mem_constraint_operand" to deal with both "Q" and "W"
22484 ("Dv"): New constraint for const double value that can be used at
22486 * config/csky/csky-modes.def (HFmode): New mode.
22487 * config/csky/csky-protos.h (csky_valid_fpuv2_mem_operand): Rename
22488 to "csky_valid_mem_constraint_operand" and support new constraint
22490 (csky_get_movedouble_length): New.
22491 (fpuv3_output_move): New.
22492 (fpuv3_const_double): New.
22493 * config/csky/csky.c (csky_option_override): New arch CK860 with fpv3.
22494 (decompose_csky_address): Refine.
22495 (csky_print_operand): New "CONST_DOUBLE" operand.
22496 (csky_output_move): Support fpv3 instructions.
22497 (csky_get_movedouble_length): New.
22498 (fpuv3_output_move): New.
22499 (fpuv3_const_double): New.
22500 (csky_emit_compare): Cover float comparsion.
22501 (csky_emit_compare_float): Refine.
22502 (csky_vaild_fpuv2_mem_operand): Rename to
22503 "csky_valid_mem_constraint_operand" and support new constraint "W".
22504 (ck860_rtx_costs): New.
22505 (csky_rtx_costs): Add the cost calculation of CK860.
22506 (regno_reg_class): New vregs for fpuv3.
22507 (csky_dbx_regno): Likewise.
22508 (csky_cpu_cpp_builtins): New builtin macro for fpuv3.
22509 (csky_conditional_register_usage): Suporrot fpuv3.
22510 (csky_dwarf_register_span): Suporrot fpuv3.
22511 (csky_init_builtins, csky_mangle_type): Support "__fp16" type.
22512 (ck810_legitimate_index_p): Support fp16.
22513 * config/csky/csky.h (TARGET_TLS): ADD CK860.
22514 (CSKY_VREG_P, CSKY_VREG_LO_P, CSKY_VREG_HI_P): Support fpuv3.
22515 (TARGET_SINGLE_FPU): Support fpuv3.
22516 (TARGET_SUPPORT_FPV3): New.
22517 (FIRST_PSEUDO_REGISTER): Change to 202 to hold the new fpuv3 registers.
22518 (FIXED_REGISTERS, CALL_REALLY_USED_REGISTERS, REGISTER_NAMES,
22519 REG_CLASS_CONTENTS): Support fpuv3.
22520 * config/csky/csky.md (movsf): Move to cksy_insn_fpu.md and refine.
22521 (csky_movsf_fpv2): Likewise.
22522 (ck801_movsf): Likewise.
22523 (csky_movsf): Likewise.
22525 (csky_movdf_fpv2): Likewise.
22526 (ck801_movdf): Likewise.
22527 (csky_movdf): Likewise.
22528 (movsicc): Refine. Use "comparison_operatior" instead of
22529 "ordered_comparison_operatior".
22530 (addsicc): Likewise.
22531 (CSKY_FIRST_VFP3_REGNUM, CSKY_LAST_VFP3_REGNUM): New constant.
22532 (call_value_internal_vh): New.
22533 * config/csky/csky_cores.def (CK860): New arch and cpu.
22538 * config/csky/csky_insn_fpu.md: Refactor. Separate all float patterns
22539 into emit-patterns and match-patterns, remain the emit-patterns here,
22540 and move the match-patterns to csky_insn_fpuv2.md or
22541 csky_insn_fpuv3.md.
22542 * config/csky/csky_insn_fpuv2.md: New file for fpuv2 instructions.
22543 * config/csky/csky_insn_fpuv3.md: New file and new patterns for fpuv3
22545 * config/csky/csky_isa.def (fcr): New.
22550 (CK860): New definition for ck860.
22551 * config/csky/csky_tables.opt (ck860): New processors ck860,
22552 ck860f. And new arch ck860.
22557 * config/csky/predicates.md (csky_float_comparsion_operator): Delete
22558 "geu", "gtu", "leu", "ltu", which will never appear at float comparison.
22559 * config/csky/t-csky-elf: Support 860.
22560 * config/csky/t-csky-linux: Likewise.
22561 * doc/md.texi: Add "Q" and "W" constraints for C-SKY.
22563 2021-05-24 Aaron Sawdey <acsawdey@linux.ibm.com>
22565 * config/rs6000/genfusion.pl (gen_logical_addsubf): Refactor to
22566 add generation of logical-add and add-logical fusion pairs.
22567 * config/rs6000/rs6000-cpus.def: Add new fusion to ISA 3.1 mask
22569 * config/rs6000/rs6000.c (rs6000_option_override_internal): Turn on
22570 logical-add and add-logical fusion by default.
22571 * config/rs6000/rs6000.opt: Add -mpower10-fusion-logical-add and
22572 -mpower10-fusion-add-logical options.
22573 * config/rs6000/fusion.md: Regenerate file.
22575 2021-05-24 Aldy Hernandez <aldyh@redhat.com>
22577 * value-range.cc (irange::legacy_equal_p): Check type when
22578 comparing VR_VARYING types.
22579 (range_tests_legacy): Test comparing VARYING ranges of different
22582 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
22584 * config/aarch64/aarch64.c (neoversen1_tunings):
22585 Enable AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
22587 2021-05-24 Wilco Dijkstra <wdijkstr@arm.com>
22589 * config/aarch64/aarch64.c (aarch64_classify_symbol): Use GOT for
22590 extern weak symbols. Limit symbol offsets for non-GOT symbols with
22593 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
22595 * config/arm/neon.md (vec_load_lanesxi<mode>)
22596 (vec_store_lanexoi<mode>): Move ...
22597 * config/arm/vec-common.md: here.
22599 2021-05-24 Christophe Lyon <christophe.lyon@linaro.org>
22601 * config/arm/neon.md (vec_load_lanesoi<mode>)
22602 (vec_store_lanesoi<mode>): Move ...
22603 * config/arm/vec-common.md: here.
22605 2021-05-24 liuhongt <hongtao.liu@intel.com>
22608 * config/i386/i386.c (ix86_gimple_fold_builtin): Replacing
22609 stmt with GIMPLE_NOP when lhs doesn't exist.
22611 2021-05-23 Uroš Bizjak <ubizjak@gmail.com>
22614 * config/i386/mmx.md (*push<VI_32:mode>2_rex64):
22615 New instruction pattern.
22616 (*push<VI_32:mode>2): Ditto.
22617 (push splitter for SSE registers): New splitter.
22619 2021-05-23 Andrew Pinski <apinski@marvell.com>
22621 * match.pd ((A & C) != 0 ? D : 0): Limit to non pointer types.
22623 2021-05-22 Aaron Sawdey <acsawdey@linux.ibm.com>
22625 * config/rs6000/genfusion.pl (gen_addadd): Fix incorrect attr types.
22626 * config/rs6000/fusion.md: Regenerate file.
22628 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
22630 * config/rs6000/genfusion.pl (gen_addadd): New function.
22631 * config/rs6000/fusion.md: Regenerate file.
22632 * config/rs6000/rs6000-cpus.def: Add
22633 OPTION_MASK_P10_FUSION_2ADD to masks.
22634 * config/rs6000/rs6000.c (rs6000_option_override_internal):
22635 Handle default value of OPTION_MASK_P10_FUSION_2ADD.
22636 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2add.
22638 2021-05-21 Jakub Jelinek <jakub@redhat.com>
22640 PR middle-end/99928
22641 * tree.h (OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET): Define.
22642 * gimplify.c (enum gimplify_omp_var_data): Fix up
22643 GOVD_MAP_HAS_ATTACHMENTS value, add GOVD_FIRSTPRIVATE_IMPLICIT.
22644 (omp_lastprivate_for_combined_outer_constructs): If combined target
22645 has GOVD_FIRSTPRIVATE_IMPLICIT set for the decl, change it to
22646 GOVD_MAP | GOVD_SEEN.
22647 (gimplify_scan_omp_clauses): Set GOVD_FIRSTPRIVATE_IMPLICIT for
22648 firstprivate clauses with OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT.
22649 (gimplify_adjust_omp_clauses): For firstprivate clauses with
22650 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT either clear that bit and
22651 OMP_CLAUSE_FIRSTPRIVATE_IMPLICIT_TARGET too, or remove it and
22652 let it be replaced by implicit map clause.
22654 2021-05-21 Jakub Jelinek <jakub@redhat.com>
22656 PR middle-end/99928
22657 * gimplify.c (omp_lastprivate_for_combined_outer_constructs): New
22659 (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LASTPRIVATE>: Use it.
22660 (gimplify_omp_for): Likewise.
22662 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
22664 PR middle-end/90115
22665 * omp-low.c (oacc_privatization_candidate_p): Reject 'static',
22666 'external' in blocks.
22668 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
22670 PR middle-end/90115
22671 * flag-types.h (enum openacc_privatization): New.
22672 * params.opt (-param=openacc-privatization): New.
22673 * doc/invoke.texi (openacc-privatization): Document it.
22674 * omp-general.h (get_openacc_privatization_dump_flags): New
22676 * omp-low.c (oacc_privatization_candidate_p): Add diagnostics.
22677 * omp-offload.c (execute_oacc_device_lower)
22678 <IFN_UNIQUE_OACC_PRIVATE>: Re-work diagnostics.
22679 * target.def (goacc.adjust_private_decl): Add 'location_t'
22681 * doc/tm.texi: Regenerate.
22682 * config/gcn/gcn-protos.h (gcn_goacc_adjust_private_decl): Adjust.
22683 * config/gcn/gcn-tree.c (gcn_goacc_adjust_private_decl): Likewise.
22684 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl):
22685 Likewise. Preserve it for...
22686 (nvptx_goacc_expand_var_decl): ... use here.
22688 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
22690 * doc/sourcebuild.texi (Other attributes): Document '__OPTIMIZE__'
22693 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
22695 PR middle-end/90115
22696 * omp-low.c (oacc_privatization_candidate_p): New function.
22697 (oacc_privatization_scan_clause_chain)
22698 (oacc_privatization_scan_decl_chain): Use it. Also
22699 'gcc_checking_assert' that we're not seeing duplicates.
22701 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
22703 PR middle-end/90115
22704 * omp-offload.c (execute_oacc_device_lower): Skip processing if no
22707 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
22709 PR middle-end/90115
22710 * omp-offload.c (execute_oacc_device_lower): Explain.
22712 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
22714 PR middle-end/90115
22715 * omp-offload.c (execute_oacc_device_lower)
22716 <IFN_UNIQUE_OACC_PRIVATE>: Diagnose and handle for 'level == -1'
22718 * internal-fn.c (expand_UNIQUE): Don't expect
22719 'IFN_UNIQUE_OACC_PRIVATE'.
22721 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
22723 PR middle-end/90115
22724 * omp-low.c (lower_omp_for): Don't evaluate OpenMP 'for' clauses.
22726 2021-05-21 Thomas Schwinge <thomas@codesourcery.com>
22728 PR middle-end/90115
22729 * config/nvptx/nvptx.c (nvptx_goacc_adjust_private_decl)
22730 (nvptx_goacc_expand_var_decl): Tighten.
22732 2021-05-21 Julian Brown <julian@codesourcery.com>
22733 Chung-Lin Tang <cltang@codesourcery.com>
22734 Thomas Schwinge <thomas@codesourcery.com>
22736 PR middle-end/90115
22737 * doc/tm.texi.in (TARGET_GOACC_EXPAND_VAR_DECL)
22738 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Add documentation hooks.
22739 * doc/tm.texi: Regenerate.
22740 * expr.c (expand_expr_real_1): Expand decls using the
22741 expand_var_decl OpenACC hook if defined.
22742 * internal-fn.c (expand_UNIQUE): Handle IFN_UNIQUE_OACC_PRIVATE.
22743 * internal-fn.h (IFN_UNIQUE_CODES): Add OACC_PRIVATE.
22744 * omp-low.c (omp_context): Add oacc_privatization_candidates
22746 (lower_oacc_reductions): Add PRIVATE_MARKER parameter. Insert
22748 (lower_oacc_head_tail): Add PRIVATE_MARKER parameter. Modify
22749 private marker's gimple call arguments, and pass it to
22750 lower_oacc_reductions.
22751 (oacc_privatization_scan_clause_chain)
22752 (oacc_privatization_scan_decl_chain, lower_oacc_private_marker):
22754 (lower_omp_for, lower_omp_target, lower_omp_1): Use these.
22755 * omp-offload.c (convert.h): Include.
22756 (oacc_loop_xform_head_tail): Treat private-variable markers like
22757 fork/join when transforming head/tail sequences.
22758 (struct var_decl_rewrite_info): Add struct.
22759 (oacc_rewrite_var_decl, is_sync_builtin_call): New functions.
22760 (execute_oacc_device_lower): Support rewriting gang-private
22761 variables using target hook, and fix up addr_expr and var_decl
22763 * target.def (adjust_private_decl, expand_var_decl): New hooks.
22764 * config/gcn/gcn-protos.h (gcn_goacc_adjust_gangprivate_decl):
22766 (gcn_goacc_adjust_private_decl): ...this.
22767 * config/gcn/gcn-tree.c (gcn_goacc_adjust_gangprivate_decl):
22769 (gcn_goacc_adjust_private_decl): ...this. Add LEVEL parameter.
22770 * config/gcn/gcn.c (TARGET_GOACC_ADJUST_GANGPRIVATE_DECL): Rename
22771 definition using gcn_goacc_adjust_gangprivate_decl...
22772 (TARGET_GOACC_ADJUST_PRIVATE_DECL): ...to this, using
22773 gcn_goacc_adjust_private_decl.
22774 * config/nvptx/nvptx.c (tree-pretty-print.h): Include.
22775 (gang_private_shared_size): New global variable.
22776 (gang_private_shared_align): Likewise.
22777 (gang_private_shared_sym): Likewise.
22778 (gang_private_shared_hmap): Likewise.
22779 (nvptx_option_override): Initialize these.
22780 (nvptx_file_end): Output gang_private_shared_sym.
22781 (nvptx_goacc_adjust_private_decl, nvptx_goacc_expand_var_decl):
22783 (nvptx_set_current_function): Clear gang_private_shared_hmap.
22784 (TARGET_GOACC_ADJUST_PRIVATE_DECL): Define hook.
22785 (TARGET_GOACC_EXPAND_VAR_DECL): Likewise.
22787 2021-05-21 H.J. Lu <hjl.tools@gmail.com>
22789 * config/i386/i386-modes.def (MAX_BITSIZE_MODE_ANY_INT): Removed.
22791 2021-05-21 Richard Biener <rguenther@suse.de>
22792 H.J. Lu <hjl.tools@gmail.com>
22794 PR middle-end/90773
22795 * expr.c (expand_constructor): Elide expand_constructor if
22796 move by pieces is preferred.
22798 2021-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22800 * config/aarch64/aarch64-builtins.c (aarch64_call_properties):
22801 Take a flag and mode value as arguments.
22802 (aarch64_modifies_global_state_p): Likewise.
22803 (aarch64_reads_global_state_p): Likewise.
22804 (aarch64_could_trap_p): Likewise.
22805 (aarch64_get_attributes): Likewise.
22806 (aarch64_init_simd_builtins): Adjust callsite of above.
22807 (aarch64_init_fcmla_laneq_builtins): Use aarch64_get_attributes to get
22808 function attributes to apply to builtins.
22809 (aarch64_init_crc32_builtins): Likewise.
22810 (aarch64_init_builtin_rsqrt): Likewise.
22812 2021-05-21 Aaron Sawdey <acsawdey@linux.ibm.com>
22814 * config/rs6000/rs6000.md (define_attr "type"): Add types for fusion.
22815 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10): Use new fusion types.
22816 (gen_2logical): Use new fusion types.
22817 * config/rs6000/fusion.md: Regenerate.
22819 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
22822 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
22823 Handle V4QI and V2HI modes.
22824 (ix86_expand_sse_movcc): Ditto.
22825 * config/i386/mmx.md (*<sat_plusminus:insn><VI_32:mode>3):
22826 New instruction pattern.
22827 (*eq<VI_32:mode>3): Ditto.
22828 (*gt<VI_32:mode>3): Ditto.
22829 (*xop_pcmov_<VI_32:mode>): Ditto.
22830 (mmx_pblendvb32): Ditto.
22831 (mmx_pblendvb64): Rename from mmx_pblendvb.
22832 (vec_cmp<VI_32:mode><VI_32:mode>): New expander.
22833 (vec_cmpu<VI_32:mode><VI_32:mode>): Ditto.
22834 (vcond<VI_32:mode><VI_32:mode>): Ditto.
22835 (vcondu<VI_32:mode><VI_32:mode>): Ditto.
22836 (vcond_mask_<VI_32:mode><VI_32:mode>): Ditto.
22838 2021-05-21 Jakub Jelinek <jakub@redhat.com>
22840 PR tree-optimization/94589
22841 * tree-ssa-phiopt.c (spaceship_replacement): For integral rhs1 and
22842 rhs2, treat x <= 4 equivalently to x < 5 etc. In cmp1 and cmp2 (if
22843 not the same as cmp3) treat <= the same as < and >= the same as >.
22844 Don't require that cond2_phi_edge is true edge, instead take
22845 false/true edges into account based on cmp1/cmp2 comparison kinds.
22847 2021-05-21 Uroš Bizjak <ubizjak@gmail.com>
22850 * config/i386/mmx.md (SMAXMIN_MMXMODEI): New mode iterator.
22851 (<smaxmin:code><SMAXMIN_MMXMODEI:mode>3): Macroize expander
22852 from <smaxmin:code>v4hi3> and <smaxmin:code><MMXMODE14:mode>3
22853 using SMAXMIN_MMXMODEI mode iterator.
22854 (*<smaxmin:code>v4qi3): New insn pattern.
22855 (*<smaxmin:code>v2hi3): Ditto.
22856 (SMAXMIN_VI_32): New mode iterator.
22857 (<smaxmin:code><SMAXMIN_VI_32>mode3): New expander.
22858 (UMAXMIN_MMXMODEI): New mode iterator.
22859 (<umaxmin:code><UMAXMIN_MMXMODEI:mode>3): Macroize expander
22860 from <umaxmin:code>v8qi3> and <umaxmin:code><MMXMODE24:mode>3
22861 using UMAXMIN_MMXMODEI mode iterator.
22862 (*<umaxmin:code>v4qi3): New insn pattern.
22863 (*<umaxmin:code>v2hi3): Ditto.
22864 (UMAXMIN_VI_32): New mode iterator.
22865 (<umaxmin:code><UMAXMIN_VI_32>mode3): New expander.
22866 (abs<VI_32:mode>2): New insn pattern.
22867 (ssse3_abs<MMXMODEI:mode>2, abs<MMXMODEI:mode>2): Move from ...
22868 * config/i386/sse.md: ... here.
22870 2021-05-20 Clement Chigot <clement.chigot@atos.net>
22871 David Edelsohn <dje.gcc@gmail.com>
22873 * collect2.c (scan_prog_file): Issue non-fatal warning for
22876 2021-05-20 Jonathan Wakely <jwakely@redhat.com>
22878 * doc/invoke.texi (-Wno-c++11-extensions)
22879 (-Wno-c++14-extensions, -Wno-c++17-extensions)
22880 (-Wno-c++20-extensions, -Wno-c++23-extensions): Document
22883 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
22885 * config/c6x/c6x.c (c6x_output_file_unwind): Use dwarf_debuginfo_p.
22886 * config/darwin.c (darwin_override_options): Likewise.
22887 * config/i386/cygming.h (DBX_REGISTER_NUMBER): Likewise.
22888 * config/i386/darwin.h (DBX_REGISTER_NUMBER): Likewise.
22889 (DWARF2_FRAME_REG_OUT): Likewise.
22890 * config/mips/mips.c (mips_output_filename): Likewise.
22891 * config/rs6000/rs6000.c (rs6000_xcoff_declare_function_name):
22893 (rs6000_dbx_register_number): Likewise.
22894 * dbxout.c: Include flags.h.
22895 * dwarf2cfi.c (cfi_label_required_p): Likewise.
22896 (dwarf2out_do_frame): Likewise.
22897 * except.c: Include flags.h.
22898 * final.c (dwarf2_debug_info_emitted_p): Likewise.
22899 (final_scan_insn_1): Likewise.
22900 * flags.h (dwarf_debuginfo_p): New function declaration.
22901 * opts.c (dwarf_debuginfo_p): New function definition.
22902 * targhooks.c (default_debug_unwind_info): Use dwarf_debuginfo_p.
22903 * toplev.c (process_options): Likewise.
22905 2021-05-20 Indu Bhagat <indu.bhagat@oracle.com>
22907 * common.opt: Change type to support bitmasks.
22908 * flag-types.h (enum debug_info_type): Rename enumerator constants.
22909 (NO_DEBUG): New bitmask.
22910 (DBX_DEBUG): Likewise.
22911 (DWARF2_DEBUG): Likewise.
22912 (XCOFF_DEBUG): Likewise.
22913 (VMS_DEBUG): Likewise.
22914 (VMS_AND_DWARF2_DEBUG): Likewise.
22915 * flags.h (debug_set_to_format): New function declaration.
22916 (debug_set_count): Likewise.
22917 (debug_set_names): Likewise.
22918 * opts.c (debug_type_masks): Array of bitmasks for debug formats.
22919 (debug_set_to_format): New function definition.
22920 (debug_set_count): Likewise.
22921 (debug_set_names): Likewise.
22922 (set_debug_level): Update access to debug_type_names.
22923 * toplev.c: Likewise.
22925 2021-05-20 Martin Sebor <msebor@redhat.com>
22927 PR middle-end/100684
22928 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Handle C++ lambda.
22930 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
22933 * config/i386/i386.md (isa): Remove x64_bmi.
22934 (enabled): Remove x64_bmi.
22935 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
22936 Remove general register alternative.
22937 (*andnot<VI_32:mode>3): Ditto.
22938 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
22939 (*<any_logic:code><VI_32:mode>3): Ditto.
22941 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
22943 * config/arm/arm.c: Include head files tree-vectorizer.h and
22946 2021-05-20 Uroš Bizjak <ubizjak@gmail.com>
22949 * config/i386/mmx.md (Yv_Yw): Revert adding V4QI and V2HI modes.
22950 (*<plusminus:insn><VI_32:mode>3): Use Yw instad of <Yv_Yw> constrint.
22951 (<s>mulv4hi3_highpart): New expander.
22952 (*<s>mulv2hi3_highpart): New insn pattern.
22953 (<s>mulv2hi3_higpart): New expander.
22954 (*<any_shift:insn>v2hi3): New insn pattern.
22955 (<any_shift:insn>v2hi3): New expander.
22956 * config/i386/sse.md (smulhrsv2hi3): New expander.
22957 (*smulhrsv2hi3): New insn pattern.
22959 2021-05-20 Kewen Lin <linkw@linux.ibm.com>
22961 * doc/invoke.texi (vect-inner-loop-cost-factor): Document new
22963 * params.opt (vect-inner-loop-cost-factor): New.
22964 * targhooks.c (default_add_stmt_cost): Replace hardcoded factor
22965 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR, include head file
22966 tree-vectorizer.h and its required ones.
22967 * config/aarch64/aarch64.c (aarch64_add_stmt_cost): Replace
22968 hardcoded factor 50 with LOOP_VINFO_INNER_LOOP_COST_FACTOR.
22969 * config/arm/arm.c (arm_add_stmt_cost): Likewise.
22970 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
22971 * config/rs6000/rs6000.c (rs6000_add_stmt_cost): Likewise.
22972 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
22974 (_loop_vec_info::_loop_vec_info): Init inner_loop_cost_factor.
22975 * tree-vectorizer.h (_loop_vec_info): Add inner_loop_cost_factor.
22976 (LOOP_VINFO_INNER_LOOP_COST_FACTOR): New macro.
22978 2021-05-20 Christophe Lyon <christophe.lyon@linaro.org>
22979 Torbjörn Svensson <torbjorn.svensson@st.com>
22982 * doc/cpp.texi (Common Predefined Macros): Document __FILE_NAME__.
22984 2021-05-20 Jakub Jelinek <jakub@redhat.com>
22986 PR middle-end/99928
22987 * gimplify.c (gimplify_scan_omp_clauses) <case OMP_CLAUSE_LINEAR>: For
22988 explicit linear clause when combined with target, make it map(tofrom:)
22989 instead of no clause or firstprivate.
22991 2021-05-20 Jakub Jelinek <jakub@redhat.com>
22993 PR tree-optimization/94589
22994 * match.pd ((X & Y) == X -> (X & ~Y) == 0): Simplify even in presence
22995 of integral conversions.
22997 2021-05-19 Andrew MacLeod <amacleod@redhat.com>
22999 * gimple-range.cc (fur_source::get_operand): New.
23000 (gimple_range_fold): Delete.
23001 (fold_using_range::fold_stmt): Move from gimple_ranger::calc_stmt.
23002 (fold_using_range::range_of_range_op): Move from gimple_ranger.
23003 (fold_using_range::range_of_address): Ditto.
23004 (fold_using_range::range_of_phi): Ditto.
23005 (fold_using_range::range_of_call): Ditto.
23006 (fold_using_range::range_of_builtin_ubsan_call): Move from
23007 range_of_builtin_ubsan_call.
23008 (fold_using_range::range_of_builtin_call): Move from
23009 range_of_builtin_call.
23010 (gimple_ranger::range_of_builtin_call): Delete.
23011 (fold_using_range::range_of_cond_expr): Move from gimple_ranger.
23012 (gimple_ranger::fold_range_internal): New.
23013 (gimple_ranger::range_of_stmt): Use new fold_using_range API.
23014 (fold_using_range::range_of_ssa_name_with_loop_info): Move from
23015 gimple_ranger. Improve ranges of SSA_NAMES when possible.
23016 * gimple-range.h (gimple_ranger): Remove various range_of routines.
23017 (class fur_source): New.
23018 (class fold_using_range): New.
23019 (fur_source::fur_source): New.
23021 * vr-values.c (vr_values::extract_range_basic): Use fold_using_range
23022 instead of range_of_builtin_call.
23024 2021-05-19 Jonathan Wakely <jwakely@redhat.com>
23026 * doc/cpp.texi (Common Predefined Macros): Update documentation
23027 for the __GXX_EXPERIMENTAL_CXX0X__ macro.
23029 2021-05-19 Alex Coplan <alex.coplan@arm.com>
23032 * config/arm/arm.md (nonsecure_call_internal): Always ensure
23033 callee's address is in a register.
23035 2021-05-19 Geng Qi <gengqi@linux.alibaba.com>
23037 * common/config/riscv/riscv-common.c
23038 (riscv_subset_list::parsing_subset_version): Properly parse the letter
23040 (riscv_subset_list::parse_std_ext,
23041 riscv_subset_list::parse_multiletter_ext): To handle errors generated
23042 in riscv_subset_list::parsing_subset_version.
23044 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
23046 * config/aarch64/aarch64-simd.md: Use "neon_move_narrow_q"
23047 type attribute in patterns generating XTN(2).
23049 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
23051 * config/aarch64/aarch64-simd.md (aarch64_simd_vec_pack_trunc_<mode>):
23052 Remove as duplicate of...
23053 (aarch64_xtn<mode>): This.
23054 (aarch64_xtn2<mode>_le): Move position in file.
23055 (aarch64_xtn2<mode>_be): Move position in file.
23056 (aarch64_xtn2<mode>): Move position in file.
23057 (vec_pack_trunc_<mode>): Define as an expander.
23059 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
23061 * config/aarch64/aarch64-simd-builtins.def: Split builtin
23062 generation for aarch64_<sur>q<r>shr<u>n_n<mode> pattern into
23063 separate scalar and vector generators.
23064 * config/aarch64/aarch64-simd.md
23065 (aarch64_<sur>q<r>shr<u>n_n<mode>): Define as an expander and
23067 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_le): This and...
23068 (aarch64_<sur>q<r>shr<u>n_n<mode>_insn_be): This.
23069 * config/aarch64/iterators.md: Define SD_HSDI iterator.
23071 2021-05-19 Jonathn Wright <jonathan.wright@arm.com>
23073 * config/aarch64/aarch64-simd.md: Use UNSPEC_SQXTUN instead
23075 * config/aarch64/iterators.md: Remove UNSPEC_SQXTUN2.
23077 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
23079 * config/aarch64/aarch64-simd.md (aarch64_<sur>q<r>shr<u>n2_n<mode>):
23080 Implement as an expand emitting a big/little endian
23081 instruction pattern.
23082 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_le): Define.
23083 (aarch64_<sur>q<r>shr<u>n2_n<mode>_insn_be): Define.
23085 2021-05-19 Jonathan Wright <jonathan.wright@arm.com>
23087 * config/aarch64/aarch64-simd.md (aarch64_<sur><addsub>hn2<mode>):
23088 Implement as an expand emitting a big/little endian
23089 instruction pattern.
23090 (aarch64_<sur><addsub>hn2<mode>_insn_le): Define.
23091 (aarch64_<sur><addsub>hn2<mode>_insn_be): Define.
23092 * config/aarch64/iterators.md: Remove UNSPEC_[R]ADDHN2 and
23093 UNSPEC_[R]SUBHN2 unspecs and ADDSUBHN2 iterator.
23095 2021-05-19 Richard Biener <rguenther@suse.de>
23097 PR middle-end/100672
23098 * fold-const.c (fold_negate_expr_1): Use element_precision.
23099 (negate_expr_p): Likewise.
23101 2021-05-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
23103 * config/aarch64/iterators.md (SVE_PRED_LOAD): New iterator.
23104 (pred_load): New int attribute.
23105 * config/aarch64/aarch64-sve.md
23106 (aarch64_load_<ANY_EXTEND:optab><SVE_HSDI:mode><SVE_PARTIAL_I:mode>): Use
23107 SVE_PRED_LOAD enum iterator and corresponding pred_load attribute.
23108 * config/aarch64/aarch64-sve-builtins-base.cc (expand): Update call to
23109 code_for_aarch64_load.
23111 2021-05-19 Richard Biener <rguenther@suse.de>
23113 * cfgexpand.c (discover_nonconstant_array_refs_r): Make
23114 sure TARGET_MEM_REF bases are expanded as memory.
23115 * tree-ssa-operands.c (operands_scanner::get_tmr_operands):
23116 Do not mark TARGET_MEM_REF bases addressable.
23117 * tree-ssa.c (non_rewritable_mem_ref_base): Handle
23118 TARGET_MEM_REF bases as never rewritable.
23119 * gimple-walk.c (walk_stmt_load_store_addr_ops): Do not
23120 walk TARGET_MEM_REF bases as address-takens.
23121 * tree-ssa-dce.c (ref_may_be_aliased): Handle TARGET_MEM_REF.
23123 2021-05-19 Richard Biener <rguenther@suse.de>
23125 * builtins.c (get_object_alignment_1): Strip outer
23127 * tree-dfa.c (get_ref_base_and_extent): Handle outer
23128 WITH_SIZE_EXPR for size processing and process the
23130 * tree-ssa-alias.c (ao_ref_base_alias_set): Strip
23131 outer WITH_SIZE_EXPR.
23132 (ao_ref_base_alias_ptr_type): Likewise.
23133 (refs_may_alias_p_2): Allow WITH_SIZE_EXPR in ref->ref
23134 and handle that accordingly, stripping it for the
23135 core alias workers.
23136 * tree.c (get_base_address): Handle WITH_SIZE_EXPR by
23137 looking through it instead of returning NULL.
23139 2021-05-19 Jakub Jelinek <jakub@redhat.com>
23141 PR middle-end/100576
23142 * builtins.c (check_read_access): Convert bound to size_type_node if
23145 2021-05-19 Richard Biener <rguenther@suse.de>
23147 * tree-cfg.c (verify_types_in_gimple_min_lval): Inline...
23148 (verify_types_in_gimple_reference): ... here. Sanitize.
23149 (verify_gimple_call): Verify references in LHS and arguments.
23150 (verify_gimple_assign_single): Reject WITH_SIZE_EXPR.
23152 2021-05-19 Uroš Bizjak <ubizjak@gmail.com>
23154 * config/i386/i386.h (VALID_INT_MODE_P):
23155 Add V8QI, V4HI and V2SI modes for TARGET_64BIT.
23156 * config/i386/i386.md (isa): Add x64_bmi.
23157 (enabled): Handle x64_bmi.
23158 * config/i386/mmx.md (mmx_andnot<MMXMODEI:mode>3):
23159 Add alternative using 64bit general registers.
23160 (*mmx_<any_logic:code><MMXMODEI:mode>3): Ditto.
23162 2021-05-19 Jakub Jelinek <jakub@redhat.com>
23164 PR middle-end/99928
23165 * tree.h (OMP_MASTER_COMBINED): Define.
23166 * gimplify.c (gimplify_scan_omp_clauses): Rewrite lastprivate
23167 handling for outer combined/composite constructs to a loop.
23168 Handle lastprivate on combined target.
23169 (gimplify_expr): Formatting fix.
23171 2021-05-19 Xionghu Luo <luoxhu@linux.ibm.com>
23173 * passes.def: Add sink_code pass before store_merging.
23174 * tree-ssa-sink.c (pass_sink_code:clone): New.
23176 2021-05-18 Bill Schmidt <wschmidt@linux.ibm.com>
23178 * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Remove call to
23179 rs6000_special_adjust_field_align_p.
23180 * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
23181 * config/rs6000/rs6000-call.c (rs6000_function_arg_boundary):
23182 Remove ABI warning.
23183 (rs6000_function_arg): Likewise.
23184 * config/rs6000/rs6000-protos.h
23185 (rs6000_special_adjust_field_align_p): Remove prototype.
23186 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
23188 * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Remove call to
23189 rs6000_special_adjust_field_align_p.
23191 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
23194 * config/i386/i386.h (VALID_SSE2_REG_MODE):
23195 Add V4QI and V2HI modes.
23196 (VALID_INT_MODE_P): Ditto.
23197 * config/i386/mmx.md (VI_32): New mode iterator.
23198 (mmxvecsize): Handle V4QI and V2HI.
23200 (mov<VI_32:mode>): New expander.
23201 (*mov<mode>_internal): New insn pattern.
23202 (movmisalign<VI_32:mode>): New expander.
23203 (neg<VI_32:mode>): New expander.
23204 (<plusminus:insn><VI_32:mode>3): New expander.
23205 (*<plusminus:insn><VI_32:mode>3): New insn pattern.
23206 (mulv2hi3): New expander.
23207 (*mulv2hi3): New insn pattern.
23208 (one_cmpl<VI_32:mode>2): New expander.
23209 (*andnot<VI_32:mode>3): New insn pattern.
23210 (<any_logic:code><VI_32:mode>3): New expander.
23211 (*<any_logic:code><VI_32:mode>3): New insn pattern.
23213 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
23215 * config/i386/sse.md (<any_extend:insn>v4qiv4di2):
23216 Fix a mode mismatch with operand 1.
23218 2021-05-18 Uroš Bizjak <ubizjak@gmail.com>
23221 * config/i386/i386-expand.c (split_double_mode): Return
23222 temporary register when simplify_gen_subreg fails with
23223 the high half od the paradoxical subreg.
23225 2021-05-18 Richard Biener <rguenther@suse.de>
23227 * cfgexpand.c (expand_one_var): Pass in forced_stack_var
23228 and honor it when expanding.
23229 (expand_used_vars_for_block): Pass through forced_stack_var.
23230 (expand_used_vars): Likewise.
23231 (discover_nonconstant_array_refs_r): Set bits in
23232 forced_stack_vars instead of marking vars TREE_ADDRESSABLE.
23233 (avoid_type_punning_on_regs): Likewise.
23234 (discover_nonconstant_array_refs): Likewise.
23235 (pass_expand::execute): Create and pass down forced_stack_var
23236 bitmap. For parameters and returns temporarily set
23237 TREE_ADDRESSABLE when expand_function_start.
23239 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
23241 * doc/sourcebuild.texi: Document 'dg-note'.
23243 2021-05-18 Tobias Burnus <tobias@codesourcery.com>
23246 * configure: Regenerate.
23247 * configure.ac (BUILD_CFLAG, BUILD_CXXFLAGS): Add $(CFLAGS-$@).
23249 2021-05-18 Thomas Schwinge <thomas@codesourcery.com>
23251 * gimple.h (is_gimple_omp_oacc): Tighten.
23252 * omp-low.c (check_omp_nesting_restrictions): Adjust.
23254 2021-05-18 Richard Biener <rguenther@suse.de>
23256 * tree-ssa-operands.c (mark_address_taken): Simplify.
23258 2021-05-18 Martin Liska <mliska@suse.cz>
23260 * config/gcn/mkoffload.c (STR): Redefine.
23261 * config/i386/intelmic-mkoffload.c (STR): Likewise.
23262 * config/nvptx/mkoffload.c (STR): Likewise.
23264 2021-05-18 Martin Liska <mliska@suse.cz>
23266 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
23267 Use startswith function instead of strncmp.
23268 * common/config/bfin/bfin-common.c (bfin_handle_option): Likewise.
23269 * common/config/riscv/riscv-common.c (riscv_subset_list::parse): Likewise.
23270 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_type): Likewise.
23271 * config/aarch64/aarch64.c (aarch64_process_one_target_attr): Likewise.
23272 * config/alpha/alpha.c (alpha_elf_section_type_flags): Likewise.
23273 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
23274 * config/arm/arm.c (arm_file_start): Likewise.
23275 (arm_valid_target_attribute_rec): Likewise.
23276 (thumb1_md_asm_adjust): Likewise.
23277 * config/arm/driver-arm.c (host_detect_local_cpu): Likewise.
23278 * config/avr/avr.c (STR_PREFIX_P): Likewise.
23279 (avr_set_current_function): Likewise.
23280 (avr_handle_addr_attribute): Likewise.
23281 (avr_asm_output_aligned_decl_common): Likewise.
23282 (avr_asm_named_section): Likewise.
23283 (avr_section_type_flags): Likewise.
23284 (avr_asm_select_section): Likewise.
23285 * config/c6x/c6x.c (c6x_in_small_data_p): Likewise.
23286 (c6x_section_type_flags): Likewise.
23287 * config/darwin-c.c (darwin_cfstring_ref_p): Likewise.
23288 (darwin_objc_declare_unresolved_class_reference): Likewise.
23289 (darwin_objc_declare_class_definition): Likewise.
23290 * config/darwin.c (indirect_data): Likewise.
23291 (darwin_encode_section_info): Likewise.
23292 (darwin_objc2_section): Likewise.
23293 (darwin_objc1_section): Likewise.
23294 (machopic_select_section): Likewise.
23295 (darwin_globalize_label): Likewise.
23296 (darwin_label_is_anonymous_local_objc_name): Likewise.
23297 (darwin_asm_named_section): Likewise.
23298 (darwin_asm_output_dwarf_offset): Likewise.
23299 * config/frv/frv.c (frv_string_begins_with): Likewise.
23300 (frv_in_small_data_p): Likewise.
23301 * config/gcn/mkoffload.c (STR): Likewise.
23303 * config/i386/i386-builtins.c (get_builtin_code_for_version): Likewise.
23304 * config/i386/i386-options.c (ix86_option_override_internal): Likewise.
23305 * config/i386/i386.c (x86_64_elf_section_type_flags): Likewise.
23306 (ix86_md_asm_adjust): Likewise.
23307 * config/i386/intelmic-mkoffload.c (STR): Likewise.
23308 * config/i386/winnt.c (i386_pe_asm_named_section): Likewise.
23309 (i386_pe_file_end): Likewise.
23310 * config/ia64/ia64.c (ia64_in_small_data_p): Likewise.
23311 (ia64_section_type_flags): Likewise.
23312 * config/mips/driver-native.c (host_detect_local_cpu): Likewise.
23313 * config/mips/mips.c (mips_handle_interrupt_attr): Likewise.
23314 (mips16_stub_function_p): Likewise.
23315 (mips_function_rodata_section): Likewise.
23316 * config/msp430/msp430.c (msp430_mcu_name): Likewise.
23317 (msp430_function_section): Likewise.
23318 (msp430_section_type_flags): Likewise.
23319 (msp430_expand_helper): Likewise.
23320 * config/nios2/nios2.c (nios2_small_section_name_p): Likewise.
23321 (nios2_valid_target_attribute_rec): Likewise.
23322 * config/nvptx/mkoffload.c (process): Likewise.
23324 * config/pa/som.h: Likewise.
23325 * config/pdp11/pdp11.c (pdp11_output_ident): Likewise.
23326 * config/riscv/riscv.c (riscv_elf_select_rtx_section): Likewise.
23327 * config/rs6000/rs6000.c (VTABLE_NAME_P): Likewise.
23328 (rs6000_inner_target_options): Likewise.
23329 * config/s390/driver-native.c (s390_host_detect_local_cpu): Likewise.
23330 * config/sparc/driver-sparc.c (host_detect_local_cpu): Likewise.
23331 * config/vax/vax.c (vax_output_int_move): Likewise.
23332 * config/vms/vms-ld.c (startswith): Likewise.
23333 (process_args): Likewise.
23335 * config/vms/vms.c: Likewise.
23337 2021-05-18 Jakub Jelinek <jakub@redhat.com>
23339 PR rtl-optimization/100590
23340 * regcprop.c (copyprop_hardreg_forward_1): Only DCE dead sets if
23341 they are NONJUMP_INSN_P.
23343 2021-05-18 Jakub Jelinek <jakub@redhat.com>
23346 * function.c (push_dummy_function): Set DECL_ARTIFICIAL and
23347 DECL_ASSEMBLER_NAME on the fn_decl.
23349 2021-05-18 Jakub Jelinek <jakub@redhat.com>
23351 PR tree-optimization/94589
23352 * tree-ssa-phiopt.c (spaceship_replacement): Pattern match
23353 phi result used in (res & ~1) == 0 comparison as res >= 0 as
23354 res == 2 would be UB with -ffinite-math-only.
23356 2021-05-18 Martin Liska <mliska@suse.cz>
23358 * Makefile.in: genversion.o should depend on DATESTAMP.
23360 2021-05-18 Claudiu Zissulescu <claziss@synopsys.com>
23362 * config/arc/simdext.md (negv2si2): Remove round bracket.
23364 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
23366 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Define
23367 _Bool as macro expanding to _Bool.
23369 2021-05-18 Andreas Krebbel <krebbel@linux.ibm.com>
23372 * tree.c (build_reference_type_for_mode)
23373 (build_pointer_type_for_mode): Pick pointer mode if MODE argument
23375 (build_reference_type, build_pointer_type): Invoke
23376 build_*_type_for_mode with VOIDmode.
23378 2021-05-17 Andrew MacLeod <amacleod@redhat.com>
23380 PR tree-optimization/100512
23381 * gimple-range-cache.cc (ranger_cache::set_global_range): Mark const
23382 and non-zero pointer ranges as invariant.
23383 * gimple-range.cc (gimple_ranger::range_of_stmt): Remove pointer
23384 processing from here.
23386 2021-05-17 Tom de Vries <tdevries@suse.de>
23389 * config/nvptx/nvptx-protos.h (nvptx_output_atomic_insn): Declare
23390 * config/nvptx/nvptx.c (nvptx_output_barrier)
23391 (nvptx_output_atomic_insn): New function.
23392 (nvptx_print_operand): Add support for 'B'.
23393 * config/nvptx/nvptx.md: Use nvptx_output_atomic_insn for atomic
23396 2021-05-17 Aldy Hernandez <aldyh@redhat.com>
23398 PR tree-optimization/100349
23399 * vr-values.c (bounds_of_var_in_loop): Bail if scev returns
23402 2021-05-17 Tamar Christina <tamar.christina@arm.com>
23404 * config/aarch64/driver-aarch64.c (DEFAULT_ARCH): New.
23405 (host_detect_local_cpu): Use it.
23407 2021-05-17 Martin Liska <mliska@suse.cz>
23409 * doc/invoke.texi: Add 2 missing dots.
23411 2021-05-17 Marius Hillenbrand <mhillen@linux.ibm.com>
23413 PR bootstrap/100552
23414 * configure.ac: Replace pattern substitution with call to sed.
23415 * configure: Regenerate.
23417 2021-05-17 Richard Biener <rguenther@suse.de>
23419 PR middle-end/100582
23420 * tree.c (array_at_struct_end_p): Get to the base of the
23421 reference before looking for the underlying decl.
23423 2021-05-17 Joern Rennecke <joern.rennecke@embecosm.com>
23425 * genoutput.c (validate_insn_alternatives) Make "wrong number of
23426 alternatives" message more specific, and remove assumption on where
23429 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
23431 * config/arm/iterators.md (V16): New iterator.
23432 (VH_cvtto): New iterator.
23433 (v_cmp_result): Added V4HF and V8HF support.
23434 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result>): Use VDQWH.
23435 (vcond<mode><mode>): Likewise.
23436 (vcond_mask_<mode><v_cmp_result>): Likewise.
23437 (vcond<VH_cvtto><mode>): New expander.
23439 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
23441 * config/arm/arm-protos.h (arm_expand_vector_compare): Update
23443 * config/arm/arm.c (arm_expand_vector_compare): Add support for
23445 (arm_expand_vcond): Likewise.
23446 * config/arm/iterators.md (supf): Remove VCMPNEQ_S, VCMPEQQ_S,
23447 VCMPEQQ_N_S, VCMPNEQ_N_S.
23448 (VCMPNEQ, VCMPEQQ, VCMPEQQ_N, VCMPNEQ_N): Remove.
23449 * config/arm/mve.md (@mve_vcmp<mve_cmp_op>q_<mode>): Add '@' prefix.
23450 (@mve_vcmp<mve_cmp_op>q_f<mode>): Likewise.
23451 (@mve_vcmp<mve_cmp_op>q_n_f<mode>): Likewise.
23452 (@mve_vpselq_<supf><mode>): Likewise.
23453 (@mve_vpselq_f<mode>"): Likewise.
23454 * config/arm/neon.md (vec_cmp<mode><v_cmp_result): Enable for MVE
23455 and move to vec-common.md.
23456 (vec_cmpu<mode><mode>): Likewise.
23457 (vcond<mode><mode>): Likewise.
23458 (vcond<V_cvtto><mode>): Likewise.
23459 (vcondu<mode><v_cmp_result>): Likewise.
23460 (vcond_mask_<mode><v_cmp_result>): Likewise.
23461 * config/arm/unspecs.md (VCMPNEQ_U, VCMPNEQ_S, VCMPEQQ_S)
23462 (VCMPEQQ_N_S, VCMPNEQ_N_S, VCMPEQQ_U, CMPEQQ_N_U, VCMPNEQ_N_U)
23463 (VCMPGEQ_N_S, VCMPGEQ_S, VCMPGTQ_N_S, VCMPGTQ_S, VCMPLEQ_N_S)
23464 (VCMPLEQ_S, VCMPLTQ_N_S, VCMPLTQ_S, VCMPCSQ_N_U, VCMPCSQ_U)
23465 (VCMPHIQ_N_U, VCMPHIQ_U): Remove.
23466 * config/arm/vec-common.md (vec_cmp<mode><v_cmp_result): Moved
23468 (vec_cmpu<mode><mode>): Likewise.
23469 (vcond<mode><mode>): Likewise.
23470 (vcond<V_cvtto><mode>): Likewise.
23471 (vcondu<mode><v_cmp_result>): Likewise.
23472 (vcond_mask_<mode><v_cmp_result>): Likewise. Added unsafe math
23475 2021-05-17 liuhongt <hongtao.liu@intel.com>
23478 * config/i386/i386.c (ix86_gimple_fold_builtin): Use
23479 gsi_insert_seq_before instead.
23481 2021-05-17 Christophe Lyon <christophe.lyon@linaro.org>
23483 * doc/sourcebuild.texi (arm_qbit_ok): Rename into...
23484 (arm_sat_ok): ...this.
23486 2021-05-17 Martin Liska <mliska@suse.cz>
23488 * lto-wrapper.c (merge_flto_options): Factor out a new function.
23489 (merge_and_complain): Use it.
23490 (run_gcc): Merge also linker command line -flto=foo argument
23493 2021-05-16 Christophe Lyon <christophe.lyon@linaro.org>
23495 * config/arm/arm.h (CPP_SPEC): Remove error message about
23496 -mlittle-endian/-mbig-endian conflict.
23498 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
23500 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
23501 __ROP_PROTECT__ if -mrop-protect is selected.
23503 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
23505 * config/rs6000/rs6000-internal.h (rs6000_stack): Add
23506 rop_hash_save_offset and rop_hash_size.
23507 * config/rs6000/rs6000-logue.c (rs6000_stack_info): Compute
23508 rop_hash_size and rop_hash_save_offset.
23509 (debug_stack_info): Dump rop_hash_save_offset and rop_hash_size.
23510 (rs6000_emit_prologue): Emit hashst[p] in prologue.
23511 (rs6000_emit_epilogue): Emit hashchk[p] in epilogue.
23512 * config/rs6000/rs6000.md (unspec): Add UNSPEC_HASHST and
23514 (hashst): New define_insn.
23515 (hashchk): Likewise.
23517 2021-05-15 Bill Schmidt <wschmidt@linux.ibm.com>
23519 * config/rs6000/rs6000.c (rs6000_option_override_internal):
23520 Disable shrink wrap when inserting ROP-protect instructions.
23521 * config/rs6000/rs6000.opt (mrop-protect): New option.
23522 (mprivileged): Likewise.
23523 * doc/invoke.texi: Document mrop-protect and mprivileged.
23525 2021-05-15 Hans-Peter Nilsson <hp@axis.com>
23527 * reorg.c (fill_slots_from_thread): Reinstate code typoed out in
23530 2021-05-15 Martin Jambor <mjambor@suse.cz>
23533 2021-05-13 Martin Jambor <mjambor@suse.cz>
23535 PR tree-optimization/100453
23536 * tree-sra.c (sra_modify_assign): All const base accesses do not
23537 need refreshing, not just those from decl_pool.
23538 (sra_modify_assign): Do not refresh into a const base decl.
23540 2021-05-15 Jakub Jelinek <jakub@redhat.com>
23542 PR rtl-optimization/100342
23543 * regcprop.c (copy_value): When copying a source reg in a wider
23544 mode than it has recorded for the value, adjust recorded destination
23545 mode too or punt if !REG_CAN_CHANGE_MODE_P.
23547 2021-05-14 Jason Merrill <jason@redhat.com>
23549 * intl.h: Add comments.
23551 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23553 * config/aarch64/aarch64-simd.md
23554 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Split into...
23555 (aarch64_sqdmlsl2_lane<mode>_internal): ... This...
23556 (aarch64_sqdmlal2_lane<mode>_internal): ... And this.
23557 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Split into ...
23558 (aarch64_sqdmlsl2_laneq<mode>_internal): ... This...
23559 (aarch64_sqdmlal2_laneq<mode>_internal): ... And this.
23560 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>_internal): Split into...
23561 (aarch64_sqdmlsl2_n<mode>_internal): ... This...
23562 (aarch64_sqdmlal2_n<mode>_internal): ... And this.
23564 2021-05-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
23567 * config/arm/arm_neon.h (vtst_s8): Replace call to vtst builtin with it's
23568 boolean logic equivalent.
23569 (vtst_s16): Likewise.
23570 (vtst_s32): Likewise.
23571 (vtst_u8): Likewise.
23572 (vtst_u16): Likewise.
23573 (vtst_u32): Likewise.
23574 (vtst_p8): Likewise.
23575 (vtst_p16): Likewise.
23576 (vtstq_s8): Likewise.
23577 (vtstq_s16): Likewise.
23578 (vtstq_s32): Likewise.
23579 (vtstq_u8): Likewise.
23580 (vtstq_u16): Likewise.
23581 (vtstq_u32): Likewise.
23582 (vtstq_p8): Likewise.
23583 (vtstq_p16): Likewise.
23584 * config/arm/arm_neon_builtins.def: Remove entry for vtst.
23585 * config/arm/neon.md (neon_vtst<mode>): Remove pattern.
23587 2021-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23589 * config/aarch64/aarch64-simd.md (aarch64_sqdmlal2<mode>): Merge into...
23590 (aarch64_sqdml<SBINQOPS:as>l2<mode>): ... This.
23591 (aarch64_sqdmlsl2<mode>): Delete.
23592 (aarch64_sqdmlal2_lane<mode>): Merge this...
23593 (aarch64_sqdmlsl2_lane<mode>): ... And this...
23594 (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>): ... Into this.
23595 (aarch64_sqdmlal2_laneq<mode>): Merge this...
23596 (aarch64_sqdmlsl2_laneq<mode>): ... And this...
23597 (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>): ... Into this.
23598 (aarch64_sqdmlal2_n<mode>): Merge this...
23599 (aarch64_sqdmlsl2_n<mode>): ... And this...
23600 (aarch64_sqdml<SBINQOPS:as>l2_n<mode>): ... Into this.
23602 2021-05-13 Martin Sebor <msebor@redhat.com>
23604 PR middle-end/100574
23605 * builtins.c (access_ref::get_ref): Improve detection of PHIs with
23606 all null arguments.
23608 2021-05-13 Martin Sebor <msebor@redhat.com>
23610 PR tree-optimization/93100
23611 PR middle-end/98583
23612 * tree-ssa-uninit.c (check_defs): Exclude intrinsic functions that
23613 don't modify referenced objects.
23615 2021-05-13 Martin Jambor <mjambor@suse.cz>
23617 PR tree-optimization/100453
23618 * tree-sra.c (sra_modify_assign): All const base accesses do not
23619 need refreshing, not just those from decl_pool.
23620 (sra_modify_assign): Do not refresh into a const base decl.
23622 2021-05-13 Martin Liska <mliska@suse.cz>
23624 * tree-ssa-dom.c: Remove m_simplifier.
23626 2021-05-13 Richard Earnshaw <rearnsha@arm.com>
23629 * config/arm/arm.c (arm_canonicalize_comparison): Correctly
23630 canonicalize DImode inequality comparisons against the
23631 maximum integral value.
23633 2021-05-13 Jakub Jelinek <jakub@redhat.com>
23635 PR tree-optimization/98856
23636 * config/i386/i386.c (ix86_shift_rotate_cost): Add CODE argument.
23637 Expect V2DI and V4DI arithmetic right shifts to be emulated.
23638 (ix86_rtx_costs, ix86_add_stmt_cost): Adjust ix86_shift_rotate_cost
23640 * config/i386/i386-expand.c (expand_vec_perm_2perm_interleave,
23641 expand_vec_perm_2perm_pblendv): New functions.
23642 (ix86_expand_vec_perm_const_1): Use them.
23643 * config/i386/sse.md (ashr<mode>3<mask_name>): Rename to ...
23644 (<mask_codefor>ashr<mode>3<mask_name>): ... this.
23645 (ashr<mode>3): New define_expand with VI248_AVX512BW iterator.
23646 (ashrv4di3): New define_expand.
23647 (ashrv2di3): Change condition to TARGET_SSE2, handle !TARGET_XOP
23648 and !TARGET_AVX512VL expansion.
23650 2021-05-13 Uroš Bizjak <ubizjak@gmail.com>
23653 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Force mode
23654 sizes < 16 to a register when constructing vpcmov pattern.
23655 * config/i386/mmx.md (*xop_pcmov_<mode>): Use MMXMODE124 mode.
23657 2021-05-13 Martin Liska <mliska@suse.cz>
23659 * gcov-io.c (gcov_write_block): Remove.
23660 (gcov_write_words): Likewise.
23661 (gcov_read_words): Re-implement using gcov_read_bytes.
23662 (gcov_allocate): Remove.
23663 (GCOV_BLOCK_SIZE): Likewise.
23664 (struct gcov_var): Remove most of the fields.
23665 (gcov_position): Implement with ftell.
23666 (gcov_rewrite): Remove setting of start and offset fields.
23667 (from_file): Re-format.
23668 (gcov_open): Remove setbuf call. It should not be needed.
23669 (gcov_close): Remove internal buffer handling.
23670 (gcov_magic): Use __builtin_bswap32.
23671 (gcov_write_counter): Use directly gcov_write_unsigned.
23672 (gcov_write_string): Use direct fwrite and do not round
23674 (gcov_seek): Use directly fseek.
23675 (gcov_write_tag): Use gcov_write_unsigned directly.
23676 (gcov_write_length): Likewise.
23677 (gcov_write_tag_length): Likewise.
23678 (gcov_read_bytes): Use directly fread.
23679 (gcov_read_unsigned): Use gcov_read_words.
23680 (gcov_read_counter): Likewise.
23681 (gcov_read_string): Use gcov_read_bytes.
23682 * gcov-io.h (GCOV_WORD_SIZE): Adjust to reflect
23683 that size is not in bytes, but words (4B).
23684 (GCOV_TAG_FUNCTION_LENGTH): Likewise.
23685 (GCOV_TAG_ARCS_LENGTH): Likewise.
23686 (GCOV_TAG_ARCS_NUM): Likewise.
23687 (GCOV_TAG_COUNTER_LENGTH): Likewise.
23688 (GCOV_TAG_COUNTER_NUM): Likewise.
23689 (GCOV_TAG_SUMMARY_LENGTH): Likewise.
23691 2021-05-13 liuhongt <hongtao.liu@intel.com>
23694 * config/i386/sse.md (ssedoublevecmode): Add attribute for
23695 V64QI/V32HI/V16SI/V4DI.
23696 (ssehalfvecmode): Add attribute for V2DI/V2DF.
23697 (*vec_concatv4si_0): Extend to VI124_128.
23698 (*vec_concat<mode>_0): New pre-reload splitter.
23699 * config/i386/predicates.md (movq_parallel): New predicate.
23701 2021-05-13 Alexandre Oliva <oliva@adacore.com>
23703 * targhooks.c (default_zero_call_used_regs): Retry using
23704 successfully-zeroed registers as sources.
23706 2021-05-12 Tobias Burnus <tobias@codesourcery.com>
23708 * omp-low.c (finish_taskreg_scan): Use the proper detach decl.
23710 2021-05-12 Aldy Hernandez <aldyh@redhat.com>
23713 * gimple-range.cc (range_of_builtin_call): Skip out on
23714 processing __builtin_clz when varying.
23716 2021-05-12 Tom de Vries <tdevries@suse.de>
23719 * config/nvptx/nvptx-opts.h (enum ptx_version): New enum.
23720 * config/nvptx/nvptx.c (nvptx_file_start): Print .version according
23721 to ptx_version_option.
23722 * config/nvptx/nvptx.h (TARGET_PTX_6_3): Define.
23723 * config/nvptx/nvptx.md (define_insn "nvptx_shuffle<mode>")
23724 (define_insn "nvptx_vote_ballot"): Use sync variant for
23726 * config/nvptx/nvptx.opt (ptx_version): Add enum.
23727 (mptx): Add option.
23728 * doc/invoke.texi (Nvidia PTX Options): Add mptx item.
23730 2021-05-12 Richard Biener <rguenther@suse.de>
23732 PR tree-optimization/100566
23733 * tree-ssa-sccvn.c (dominated_by_p_w_unex): Properly handle
23734 allow_back for all edge queries.
23736 2021-05-12 liuhongt <hongtao.liu@intel.com>
23739 * config/i386/sse.md (<sse4_1_avx2>_pblendvb): Add
23740 splitters for pblendvb of NOT mask register.
23742 2021-05-12 Richard Biener <rguenther@suse.de>
23744 PR tree-optimization/100519
23745 * tree-ssa-reassoc.c (can_associate_p): Split into...
23746 (can_associate_op_p): ... this
23747 (can_associate_type_p): ... and this.
23748 (is_reassociable_op): Call can_associate_op_p.
23749 (break_up_subtract_bb): Call the appropriate predicates.
23750 (reassociate_bb): Likewise.
23752 2021-05-12 Martin Liska <mliska@suse.cz>
23754 * lto-wrapper.c (merge_and_complain): Merge -flto=arg options.
23755 (run_gcc): Use -flto argument detection for merged
23758 2021-05-12 Martin Liska <mliska@suse.cz>
23760 * lto-wrapper.c (print_lto_docs_link): New function.
23761 (run_gcc): Print warning about missing job server detection
23762 after we know NR of partitions. Do the same for -flto{,=1}.
23763 * opts.c (get_option_html_page): Support -flto option.
23765 2021-05-12 Martin Liska <mliska@suse.cz>
23767 * lto-wrapper.c (get_options_from_collect_gcc_options): Change
23769 (append_option): Remove.
23770 (find_option): Rework to use the vector type.
23771 (remove_option): Remove.
23772 (merge_and_complain): Use vectors for cl_decoded_option data
23774 (append_compiler_options): Likewise.
23775 (append_diag_options): Likewise.
23776 (append_linker_options): Likewise.
23777 (append_offload_options): Likewise.
23778 (compile_offload_image): Likewise.
23779 (compile_images_for_offload_targets): Likewise.
23780 (find_and_merge_options): Likewise.
23781 (run_gcc): Likewise.
23783 2021-05-12 Bernd Edlinger <bernd.edlinger@hotmail.de>
23786 * dwarf2out.c (dwarf2out_finish): Set
23787 have_multiple_function_sections with multi-range text_section.
23789 2021-05-12 Martin Liska <mliska@suse.cz>
23791 PR bootstrap/100560
23792 * Makefile.in: Remove version.h from linker command line.
23794 2021-05-12 Richard Biener <rguenther@suse.de>
23796 PR middle-end/100547
23797 * rtl.h (rtvec_alloc): Make argument size_t.
23798 * rtl.c (rtvec_alloc): Verify the count is less than INT_MAX.
23800 2021-05-12 Jakub Jelinek <jakub@redhat.com>
23802 PR middle-end/100508
23803 * cfgexpand.c (expand_debug_expr): For DEBUG_EXPR_DECL with vector
23804 type, don't reuse DECL_RTL if it has different mode, instead force
23805 creation of a new DEBUG_EXPR.
23807 2021-05-12 Jakub Jelinek <jakub@redhat.com>
23808 Marc Glisse <marc.glisse@inria.fr>
23810 PR tree-optimization/94589
23811 * match.pd ((X & Y) == X -> (X & ~Y) == 0,
23812 (X | Y) == Y -> (X & ~Y) == 0): New GIMPLE simplifications.
23814 2021-05-12 Uroš Bizjak <ubizjak@gmail.com>
23817 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Handle V2SF mode.
23818 * config/i386/mmx.md (MMXMODE124): New mode iterator.
23820 (mmxintvecmode): New mode attribute.
23821 (mmxintvecmodelower): Ditto.
23822 (*mmx_maskcmpv2sf3_comm): New insn pattern.
23823 (*mmx_maskcmpv2sf3): Ditto.
23824 (vec_cmpv2sfv2si): New expander.
23825 (vcond<V2FI:mode>v2si): Ditto.
23826 (mmx_vlendvps): New insn pattern.
23827 (vcond<MMXMODE124:mode><MMXMODEI:mode>): Also handle V2SFmode.
23828 (vcondu<MMXMODE124:mode><MMXMODEI:mode>): Ditto.
23829 (vcond_mask_<mode><mmxintvecmodelower>): Ditto.
23831 2021-05-11 Martin Sebor <msebor@redhat.com>
23833 PR middle-end/21433
23834 * expr.c (expand_expr_real_1): Replace unreachable code with an assert.
23836 2021-05-11 Richard Biener <rguenther@suse.de>
23838 * gimple-fold.c (gimple_fold_call): Do not call
23839 maybe_fold_reference on call arguments or the static chain.
23840 (fold_stmt_1): Do not call maybe_fold_reference on GIMPLE_ASM
23843 2021-05-11 Martin Liska <mliska@suse.cz>
23845 * builtins.def (DEF_HSAIL_BUILTIN): Remove.
23846 (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
23847 (DEF_HSAIL_SAT_BUILTIN): Likewise.
23848 (DEF_HSAIL_INTR_BUILTIN): Likewise.
23849 (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
23850 * doc/frontends.texi: Remove BRIG.
23851 * doc/install.texi: Likewise.
23852 * doc/invoke.texi: Likewise.
23853 * doc/standards.texi: Likewise.
23854 * brig-builtins.def: Removed.
23855 * brig/ChangeLog: Removed.
23856 * brig/Make-lang.in: Removed.
23857 * brig/brig-builtins.h: Removed.
23858 * brig/brig-c.h: Removed.
23859 * brig/brig-lang.c: Removed.
23860 * brig/brigfrontend/brig-arg-block-handler.cc: Removed.
23861 * brig/brigfrontend/brig-atomic-inst-handler.cc: Removed.
23862 * brig/brigfrontend/brig-basic-inst-handler.cc: Removed.
23863 * brig/brigfrontend/brig-branch-inst-handler.cc: Removed.
23864 * brig/brigfrontend/brig-cmp-inst-handler.cc: Removed.
23865 * brig/brigfrontend/brig-code-entry-handler.cc: Removed.
23866 * brig/brigfrontend/brig-code-entry-handler.h: Removed.
23867 * brig/brigfrontend/brig-comment-handler.cc: Removed.
23868 * brig/brigfrontend/brig-control-handler.cc: Removed.
23869 * brig/brigfrontend/brig-copy-move-inst-handler.cc: Removed.
23870 * brig/brigfrontend/brig-cvt-inst-handler.cc: Removed.
23871 * brig/brigfrontend/brig-fbarrier-handler.cc: Removed.
23872 * brig/brigfrontend/brig-function-handler.cc: Removed.
23873 * brig/brigfrontend/brig-function.cc: Removed.
23874 * brig/brigfrontend/brig-function.h: Removed.
23875 * brig/brigfrontend/brig-inst-mod-handler.cc: Removed.
23876 * brig/brigfrontend/brig-label-handler.cc: Removed.
23877 * brig/brigfrontend/brig-lane-inst-handler.cc: Removed.
23878 * brig/brigfrontend/brig-machine.c: Removed.
23879 * brig/brigfrontend/brig-machine.h: Removed.
23880 * brig/brigfrontend/brig-mem-inst-handler.cc: Removed.
23881 * brig/brigfrontend/brig-module-handler.cc: Removed.
23882 * brig/brigfrontend/brig-queue-inst-handler.cc: Removed.
23883 * brig/brigfrontend/brig-seg-inst-handler.cc: Removed.
23884 * brig/brigfrontend/brig-signal-inst-handler.cc: Removed.
23885 * brig/brigfrontend/brig-to-generic.cc: Removed.
23886 * brig/brigfrontend/brig-to-generic.h: Removed.
23887 * brig/brigfrontend/brig-util.cc: Removed.
23888 * brig/brigfrontend/brig-util.h: Removed.
23889 * brig/brigfrontend/brig-variable-handler.cc: Removed.
23890 * brig/brigfrontend/hsa-brig-format.h: Removed.
23891 * brig/brigfrontend/phsa.h: Removed.
23892 * brig/brigspec.c: Removed.
23893 * brig/config-lang.in: Removed.
23894 * brig/gccbrig.texi: Removed.
23895 * brig/lang-specs.h: Removed.
23896 * brig/lang.opt: Removed.
23898 2021-05-11 Richard Biener <rguenther@suse.de>
23901 * ipa-param-manipulation.c
23902 (ipa_param_body_adjustments::modify_call_stmt): Avoid
23903 altering SSA_NAME_DEF_STMT by adjusting the calls LHS
23904 via gimple_call_lhs_ptr.
23906 2021-05-11 Alex Coplan <alex.coplan@arm.com>
23909 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear):
23910 Avoid emitting CFA adjusts on the sp if we have the fp.
23912 2021-05-11 Richard Sandiford <richard.sandiford@arm.com>
23914 * config/aarch64/iterators.md (VMUL_CHANGE_NLANES): Delete.
23915 (VMULD): New iterator.
23916 (VCOND): Handle V4HF and V8HF.
23917 (VCONQ): Fix entry for V2SF.
23918 * config/aarch64/aarch64-simd.md (mul_lane<mode>3): Use VMULD
23919 instead of VMUL. Use a 64-bit vector mode for the indexed operand.
23920 (*aarch64_mul3_elt_<vswap_width_name><mode>): Merge with...
23921 (mul_laneq<mode>3): ...this define_insn. Use VMUL instead of VDQSF.
23922 Use a 128-bit vector mode for the indexed operand. Use stype for
23923 the scheduling type.
23925 2021-05-11 Richard Biener <rguenther@suse.de>
23927 * gimple-fold.c (maybe_fold_reference): Only return
23928 is_gimple_min_invariant values.
23930 2021-05-11 Richard Biener <rguenther@suse.de>
23932 PR middle-end/100509
23933 * gimple-fold.c (fold_gimple_assign): Only call
23934 get_symbol_constant_value on register type symbols.
23936 2021-05-11 Srinath Parvathaneni <srinath.parvathaneni@arm.com>
23937 Joe Ramsay <joe.ramsay@arm.com>
23940 * config/arm/arm_mve.h (__arm_vstrwq_scatter_offset): Fix wrong arguments.
23941 (__arm_vcmpneq): Remove duplicate definition.
23942 (__arm_vstrwq_scatter_offset_p): Likewise.
23943 (__arm_vmaxq_x): Likewise.
23944 (__arm_vmlsdavaq): Likewise.
23945 (__arm_vmlsdavaxq): Likewise.
23946 (__arm_vmlsdavq_p): Likewise.
23947 (__arm_vmlsdavxq_p): Likewise.
23948 (__arm_vrmlaldavhaq): Likewise.
23949 (__arm_vstrbq_p): Likewise.
23950 (__arm_vstrbq_scatter_offset): Likewise.
23951 (__arm_vstrbq_scatter_offset_p): Likewise.
23952 (__arm_vstrdq_scatter_offset): Likewise.
23953 (__arm_vstrdq_scatter_offset_p): Likewise.
23954 (__arm_vstrdq_scatter_shifted_offset): Likewise.
23955 (__arm_vstrdq_scatter_shifted_offset_p): Likewise.
23957 2021-05-11 Jakub Jelinek <jakub@redhat.com>
23959 PR middle-end/100471
23960 * omp-low.c (lower_omp_task_reductions): For OMP_TASKLOOP, if data
23961 is 0, bypass the reduction loop including
23962 GOMP_taskgroup_reduction_unregister call.
23964 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
23966 * config/rs6000/rs6000.c (struct rs6000_cost_data): New member
23967 costing_for_scalar.
23968 (rs6000_density_test): Early return if costing_for_scalar is true.
23969 (rs6000_init_cost): Init costing_for_scalar of rs6000_cost_data.
23971 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
23973 * doc/tm.texi: Regenerated.
23974 * target.def (init_cost): Add new parameter costing_for_scalar.
23975 * targhooks.c (default_init_cost): Adjust for new parameter.
23976 * targhooks.h (default_init_cost): Likewise.
23977 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
23978 (vect_compute_single_scalar_iteration_cost): Likewise.
23979 (vect_analyze_loop_2): Likewise.
23980 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
23981 (vect_bb_vectorization_profitable_p): Likewise.
23982 * tree-vectorizer.h (init_cost): Likewise.
23983 * config/aarch64/aarch64.c (aarch64_init_cost): Likewise.
23984 * config/i386/i386.c (ix86_init_cost): Likewise.
23985 * config/rs6000/rs6000.c (rs6000_init_cost): Likewise.
23987 2021-05-11 Kewen Lin <linkw@linux.ibm.com>
23989 * config/rs6000/rs6000.c (rs6000_vect_nonmem): Renamed to
23990 vect_nonmem and moved into...
23991 (struct rs6000_cost_data): ...here.
23992 (rs6000_init_cost): Use vect_nonmem of cost_data instead.
23993 (rs6000_add_stmt_cost): Likewise.
23994 (rs6000_finish_cost): Likewise.
23996 2021-05-10 Eric Botcazou <ebotcazou@adacore.com>
23998 * range-op.cc (get_bool_state): Adjust head comment.
23999 (operator_not_equal::op1_range): Fix comment.
24000 (operator_bitwise_xor::op1_range): Remove call to gcc_unreachable.
24002 2021-05-10 Martin Sebor <msebor@redhat.com>
24004 PR middle-end/100425
24005 PR middle-end/100510
24006 * gimple-ssa-warn-alloca.c (pass_walloca::firast_time_p): Rename...
24007 (pass_walloca::xlimit_certain_p): ...to this.
24008 (pass_walloca::gate): Execute for any kind of handled warning.
24009 (pass_walloca::execute): Avoid issuing "maybe" and "unbounded"
24010 warnings when xlimit_certain_p is set.
24012 2021-05-10 Pat Haugen <pthaugen@linux.ibm.com>
24014 * config/rs6000/rs6000.c (rs6000_ira_change_pseudo_allocno_class):
24015 Return ALTIVEC_REGS if that is best_class.
24016 (rs6000_compute_pressure_classes): Add ALTIVEC_REGS.
24018 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
24020 * config/arm/arm.h (CPP_SPEC): Remove error message about
24023 2021-05-10 Martin Jambor <mjambor@suse.cz>
24025 * ipa-prop.h (IPA_NODE_REF): Removed.
24026 (IPA_NODE_REF_GET_CREATE): Likewise.
24027 (IPA_EDGE_REF): Likewise.
24028 (IPA_EDGE_REF_GET_CREATE): Likewise.
24029 (IS_VALID_JUMP_FUNC_INDEX): Likewise.
24030 * ipa-cp.c (print_all_lattices): Replaced IPA_NODE_REF with a direct
24031 use of ipa_node_params_sum.
24032 (ipcp_versionable_function_p): Likewise.
24033 (push_node_to_stack): Likewise.
24034 (pop_node_from_stack): Likewise.
24035 (set_single_call_flag): Replaced two IPA_NODE_REF with one single
24036 direct use of ipa_node_params_sum.
24037 (initialize_node_lattices): Replaced IPA_NODE_REF with a direct use of
24038 ipa_node_params_sum.
24039 (ipa_context_from_jfunc): Replaced IPA_EDGE_REF with a direct use of
24041 (ipcp_verify_propagated_values): Replaced IPA_NODE_REF with a direct
24042 use of ipa_node_params_sum.
24043 (self_recursively_generated_p): Likewise.
24044 (propagate_scalar_across_jump_function): Likewise.
24045 (propagate_context_across_jump_function): Replaced IPA_EDGE_REF with a
24046 direct use of ipa_edge_args_sum, moved the lookup after the early
24047 exit. Replaced IPA_NODE_REF with a direct use of ipa_node_params_sum.
24048 (propagate_bits_across_jump_function): Replaced IPA_NODE_REF with
24049 direct uses of ipa_node_params_sum.
24050 (propagate_vr_across_jump_function): Likewise.
24051 (propagate_aggregate_lattice): Likewise.
24052 (propagate_aggs_across_jump_function): Likewise.
24053 (propagate_constants_across_call): Likewise, also replaced
24054 IPA_EDGE_REF with a direct use of ipa_edge_args_sum.
24055 (good_cloning_opportunity_p): Replaced IPA_NODE_REF with a direct use
24056 of ipa_node_params_sum.
24057 (estimate_local_effects): Likewise.
24058 (add_all_node_vals_to_toposort): Likewise.
24059 (propagate_constants_topo): Likewise.
24060 (ipcp_propagate_stage): Likewise.
24061 (ipcp_discover_new_direct_edges): Likewise.
24062 (calls_same_node_or_its_all_contexts_clone_p): Likewise.
24063 (cgraph_edge_brings_value_p): Likewise (in both overloaded functions).
24064 (get_info_about_necessary_edges): Likewise.
24065 (want_remove_some_param_p): Likewise.
24066 (create_specialized_node): Likewise.
24067 (self_recursive_pass_through_p): Likewise.
24068 (self_recursive_agg_pass_through_p): Likewise.
24069 (find_more_scalar_values_for_callers_subset): Likewise and also
24070 replaced IPA_EDGE_REF with direct uses of ipa_edge_args_sum, in one
24071 case replacing two of those with a single query.
24072 (find_more_contexts_for_caller_subset): Likewise for the
24073 ipa_polymorphic_call_context overload.
24074 (intersect_aggregates_with_edge): Replaced IPA_EDGE_REF with a direct
24075 use of ipa_edge_args_sum. Replaced IPA_NODE_REF with direct uses of
24076 ipa_node_params_sum.
24077 (find_aggregate_values_for_callers_subset): Likewise, also reusing
24078 results of ipa_edge_args_sum->get.
24079 (cgraph_edge_brings_all_scalars_for_node): Replaced IPA_NODE_REF with
24080 direct uses of ipa_node_params_sum, replaced IPA_EDGE_REF with a
24081 direct use of ipa_edge_args_sum.
24082 (cgraph_edge_brings_all_agg_vals_for_node): Likewise, moved node
24083 summary query after the early exit and reused the result later.
24084 (decide_about_value): Replaced IPA_NODE_REF with a direct use of
24085 ipa_node_params_sum.
24086 (decide_whether_version_node): Likewise. Removed re-querying for
24087 summaries after cloning.
24088 (spread_undeadness): Replaced IPA_NODE_REF with a direct use of
24089 ipa_node_params_sum.
24090 (has_undead_caller_from_outside_scc_p): Likewise, reusing results of
24092 (identify_dead_nodes): Likewise.
24093 (ipcp_store_bits_results): Replaced IPA_NODE_REF with direct uses of
24094 ipa_node_params_sum.
24095 (ipcp_store_vr_results): Likewise.
24096 * ipa-fnsummary.c (evaluate_properties_for_edge): Likewise.
24097 (ipa_fn_summary_t::duplicate): Likewise.
24098 (analyze_function_body): Likewise.
24099 (estimate_calls_size_and_time): Likewise.
24100 (ipa_cached_call_context::duplicate_from): Likewise.
24101 (ipa_call_context::equal_to): Likewise.
24102 (remap_edge_params): Likewise.
24103 (ipa_merge_fn_summary_after_inlining): Likewise.
24104 (inline_read_section): Likewise.
24105 * ipa-icf.c (sem_function::param_used_p): Likewise.
24106 * ipa-modref.c (compute_parm_map): Likewise.
24107 (compute_parm_map): Replaced IPA_EDGE_REF with a direct use of
24109 (get_access_for_fnspec): Replaced IPA_NODE_REF with a direct use of
24110 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
24112 * ipa-profile.c (check_argument_count): Likewise.
24113 * ipa-prop.c (ipa_alloc_node_params): Replaced IPA_NODE_REF_GET_CREATE
24114 with a direct use of ipa_node_params_sum.
24115 (ipa_initialize_node_params): Likewise.
24116 (ipa_print_node_jump_functions_for_edge): Replaced IPA_EDGE_REF with a
24117 direct use of ipa_edge_args_sum and reused the query result.
24118 (ipa_compute_jump_functions_for_edge): Replaced IPA_NODE_REF with a
24119 direct use of ipa_node_params_sum and replaced IPA_EDGE_REF with a
24120 direct use of ipa_edge_args_sum.
24121 (ipa_note_param_call): Replaced IPA_NODE_REF with a direct use of
24122 ipa_node_params_sum and reused the result of the query.
24123 (ipa_analyze_node): Likewise.
24124 (ipa_analyze_controlled_uses): Replaced IPA_NODE_REF with a direct use
24125 of ipa_node_params_sum.
24126 (update_jump_functions_after_inlining): Replaced IPA_EDGE_REF with
24127 direct uses of ipa_edge_args_sum.
24128 (update_indirect_edges_after_inlining): Replaced IPA_NODE_REF with
24129 direct uses of ipa_node_params_sum and replaced IPA_EDGE_REF with a
24130 direct use of ipa_edge_args_sum. Removed superficial re-querying the
24132 (propagate_controlled_uses): Replaced IPA_NODE_REF with direct uses of
24133 ipa_node_params_sum and replaced IPA_EDGE_REF with a direct use of
24135 (ipa_propagate_indirect_call_infos): Replaced IPA_EDGE_REF with a
24136 direct use of ipa_edge_args_sum.
24137 (ipa_edge_args_sum_t::duplicate): Replaced IPA_NODE_REF with a direct
24138 use of ipa_node_params_sum.
24139 (ipa_print_node_params): Likewise.
24140 (ipa_write_node_info): Likewise and also replaced IPA_EDGE_REF with
24141 direct uses of ipa_edge_args_sum.
24142 (ipa_read_edge_info): Replaced IPA_EDGE_REF with a direct use of
24144 (ipa_read_node_info): Replaced IPA_NODE_REF with a direct use of
24145 ipa_node_params_sum.
24146 (ipa_prop_write_jump_functions): Likewise. Move variable node to the
24147 scopes where it is used.
24149 2021-05-10 Uroš Bizjak <ubizjak@gmail.com>
24151 * config/i386/i386-expand.c (ix86_expand_sse_movcc)
24152 <case E_V2SImode>: Force op_true to register.
24154 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
24156 * config/arm/iterators.md (MVE_FP_COMPARISONS): New.
24157 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_f<mode>)
24158 (mve_vcmp<mve_cmp_op>q_n_f<mode>): New, merge all vcmp_*f*
24160 (mve_vcmpeqq_f<mode>, mve_vcmpeqq_n_f<mode>, mve_vcmpgeq_f<mode>)
24161 (mve_vcmpgeq_n_f<mode>, mve_vcmpgtq_f<mode>)
24162 (mve_vcmpgtq_n_f<mode>, mve_vcmpleq_f<mode>)
24163 (mve_vcmpleq_n_f<mode>, mve_vcmpltq_f<mode>)
24164 (mve_vcmpltq_n_f<mode>, mve_vcmpneq_f<mode>)
24165 (mve_vcmpneq_n_f<mode>): Remove.
24166 * config/arm/unspecs.md (VCMPEQQ_F, VCMPEQQ_N_F, VCMPGEQ_F)
24167 (VCMPGEQ_N_F, VCMPGTQ_F, VCMPGTQ_N_F, VCMPLEQ_F, VCMPLEQ_N_F)
24168 (VCMPLTQ_F, VCMPLTQ_N_F, VCMPNEQ_F, VCMPNEQ_N_F): Remove.
24170 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
24172 * config/arm/iterators.md (MVE_COMPARISONS): New.
24174 (mve_cmp_type): New.
24175 * config/arm/mve.md (mve_vcmp<mve_cmp_op>q_<mode>): New, merge all
24177 (mve_vcmpneq_<mode>, mve_vcmpcsq_n_<mode>, mve_vcmpcsq_<mode>)
24178 (mve_vcmpeqq_n_<mode>, mve_vcmpeqq_<mode>, mve_vcmpgeq_n_<mode>)
24179 (mve_vcmpgeq_<mode>, mve_vcmpgtq_n_<mode>, mve_vcmpgtq_<mode>)
24180 (mve_vcmphiq_n_<mode>, mve_vcmphiq_<mode>, mve_vcmpleq_n_<mode>)
24181 (mve_vcmpleq_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
24182 (mve_vcmpneq_n_<mode>, mve_vcmpltq_n_<mode>, mve_vcmpltq_<mode>)
24183 (mve_vcmpneq_n_<mode>): Remove.
24185 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
24187 * config/arm/arm_mve.h (__arm_vcmp*): Remove 's' suffix.
24188 * config/arm/arm_mve_builtins.def (vcmp*): Remove 's' suffix.
24189 * config/arm/mve.md (mve_vcmp*): Remove 's' suffix in pattern
24192 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
24194 * config/arm/arm_mve_builtins.def (vcmpneq_u): Remove.
24195 (vcmpneq_n_u): Likewise.
24196 (vcmpeqq_u,): Likewise.
24197 (vcmpeqq_n_u): Likewise.
24198 * config/arm/iterators.md (supf): Remove VCMPNEQ_U, VCMPEQQ_U,
24199 VCMPEQQ_N_U and VCMPNEQ_N_U.
24200 * config/arm/mve.md (mve_vcmpneq): Remove <supf> iteration.
24201 (mve_vcmpeqq_n): Likewise.
24202 (mve_vcmpeqq): Likewise.
24203 (mve_vcmpneq_n): Likewise.
24205 2021-05-10 Christophe Lyon <christophe.lyon@linaro.org>
24207 * config/arm/arm_mve.h (__arm_vcmpeq*u*, __arm_vcmpne*u*): Call
24208 the 's' version of the builtin.
24210 2021-05-10 Richard Biener <rguenther@suse.de>
24212 PR tree-optimization/100492
24213 * tree-loop-distribution.c (find_seed_stmts_for_distribution):
24214 Find nothing when the loop contains an irreducible region.
24216 2021-05-10 Richard Biener <rguenther@suse.de>
24218 PR middle-end/100464
24220 * gimple-fold.c (canonicalize_constructor_val): Do not set
24223 2021-05-10 Richard Biener <rguenther@suse.de>
24225 PR tree-optimization/100434
24226 * tree-ssa-dse.c (initialize_ao_ref_for_dse): Handle
24228 (dse_optimize_stmt): Handle call LHS by dropping the
24229 LHS or the whole call if it doesn't have other
24231 (pass_dse::execute): Adjust.
24233 2021-05-10 Martin Liska <mliska@suse.cz>
24235 * Makefile.in: Add missing genversion rule.
24237 2021-05-10 Alex Coplan <alex.coplan@arm.com>
24240 * config/arm/mve.md (*mve_mov<mode>): Simplify output code. Use
24241 vldrw.u32 and vstrw.32 for V2D[IF]mode loads and stores.
24243 2021-05-10 Martin Liska <mliska@suse.cz>
24245 * builtins.c (is_builtin_name): Use startswith
24246 function instead of strncmp.
24247 * collect2.c (main): Likewise.
24248 (has_lto_section): Likewise.
24249 (scan_libraries): Likewise.
24250 * coverage.c (coverage_checksum_string): Likewise.
24251 (coverage_init): Likewise.
24252 * dwarf2out.c (is_cxx): Likewise.
24253 (gen_compile_unit_die): Likewise.
24254 * gcc-ar.c (main): Likewise.
24255 * gcc.c (init_spec): Likewise.
24256 (read_specs): Likewise.
24257 (execute): Likewise.
24258 (check_live_switch): Likewise.
24259 * genattrtab.c (write_attr_case): Likewise.
24260 (IS_ATTR_GROUP): Likewise.
24261 * gencfn-macros.c (main): Likewise.
24262 * gengtype.c (type_for_name): Likewise.
24263 (gen_rtx_next): Likewise.
24264 (get_file_langdir): Likewise.
24265 (write_local): Likewise.
24266 * genmatch.c (get_operator): Likewise.
24267 (get_operand_type): Likewise.
24268 (expr::gen_transform): Likewise.
24269 * genoutput.c (validate_optab_operands): Likewise.
24270 * incpath.c (add_sysroot_to_chain): Likewise.
24271 * langhooks.c (lang_GNU_C): Likewise.
24272 (lang_GNU_CXX): Likewise.
24273 (lang_GNU_Fortran): Likewise.
24274 (lang_GNU_OBJC): Likewise.
24275 * lto-wrapper.c (run_gcc): Likewise.
24276 * omp-general.c (omp_max_simt_vf): Likewise.
24277 * omp-low.c (omp_runtime_api_call): Likewise.
24278 * opts-common.c (parse_options_from_collect_gcc_options): Likewise.
24279 * read-rtl-function.c (function_reader::read_rtx_operand_r): Likewise.
24280 * real.c (real_from_string): Likewise.
24281 * selftest.c (assert_str_startswith): Likewise.
24282 * timevar.c (timer::validate_phases): Likewise.
24283 * tree.c (get_file_function_name): Likewise.
24284 * ubsan.c (ubsan_use_new_style_p): Likewise.
24285 * varasm.c (default_function_rodata_section): Likewise.
24286 (incorporeal_function_p): Likewise.
24287 (default_section_type_flags): Likewise.
24288 * system.h (startswith): Define startswith.
24290 2021-05-10 Martin Liska <mliska@suse.cz>
24292 * bitmap.h (class auto_bitmap): Remove
24293 __cplusplus >= 201103.
24294 * config/aarch64/aarch64.c: Likewise.
24295 * gimple-ssa-store-merging.c (store_immediate_info::store_immediate_info):
24297 * sbitmap.h: Likewise.
24299 2021-05-10 Martin Liska <mliska@suse.cz>
24301 * Makefile.in: Rename gcov-iov to genversion and depend
24302 on version.h (instead of gcov-iov.h).
24303 * gcov-io.h: Include version.h instread of gcov-iov.h.
24304 * gengtype-state.c (read_state_version): Likewise.
24305 * gcov-iov.c: Moved to...
24306 * genversion.c: ...here.
24307 * lto-streamer.h (LTO_major_version): Define it with
24309 * version.c: Removed.
24310 * version.h: Removed.
24312 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
24314 * config/arc/arc.md (UNSPEC_ARC_DMPYWH): Define.
24315 * config/arc/simdext.md (VCT): Add predicates for iterator
24318 (voptab): Likewise.
24319 (vec_widen_<V_US>mult_hi_v4hi): Change pattern predicate.
24320 (<voptab>v2si3): New patterns.
24322 (reduc_plus_scal_v4hi): Likewise.
24323 (reduc_plus_scal_v2si): Likewise.
24324 (vec_duplicatev2si): Likewise.
24325 (vec_duplicatev4hi): Likewise.
24327 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
24329 * config/arc/simdext.md: Format and cleanup file.
24331 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
24333 * config/arc/simdext.md (movmisalignv2hi): Allow misaligned access
24334 only when munaligned-access option is on.
24335 (movmisalign<mode>): Likewise.
24337 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
24339 * common/config/arc/arc-common.c (arc_handle_option): Remove dot
24341 * config/arc/arc.c (arc_reorg): Remove underscore from string.
24343 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
24345 * config/arc/arc.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
24346 (CTZ_DEFINED_VALUE_AT_ZERO): Likewise.
24347 * config/arc/arc.md (clrsbsi2): Cleanup pattern.
24348 (norm_f): Likewise.
24351 (clzsi2): Use fls instruction when available.
24352 (arc_clzsi2): Likewise.
24354 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
24356 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add r26 and r27.
24358 2021-05-10 Claudiu Zissulescu <claziss@synopsys.com>
24360 * doc/extend.texi (__builtin_arc_sr): Swap arguments.
24362 2021-05-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
24364 PR middle-end/100467
24365 * toplev.c (compile_file): Call insn_locations_init before
24366 targetm.asm_out.code_end.
24368 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
24371 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
24373 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
24375 2021-05-07 Jakub Jelinek <jakub@redhat.com>
24376 Andrew Stubbs <amd@codesourcery.com>
24379 * builtins.c (try_store_by_multiple_pieces): Use force_operand for
24380 emit_move_insn operands.
24382 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
24384 * cfgexpand.c (expand_gimple_basic_block): Do not inherit a current
24385 location for the outgoing edges of an empty block.
24386 * dwarf2out.c (add_subscript_info): Retrieve the bounds and index
24387 type by means of the get_array_descr_info langhook, if it is set and
24388 returns true. Remove obsolete code dealing with unnamed subtypes.
24390 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
24392 * gimple-range-cache.cc (ssa_block_ranges): Virtualize.
24393 (sbr_vector): Renamed from ssa_block_cache.
24394 (sbr_vector::sbr_vector): Allocate from obstack abd initialize.
24395 (ssa_block_ranges::~ssa_block_ranges): Remove.
24396 (sbr_vector::set_bb_range): Use varying and undefined cached values.
24397 (ssa_block_ranges::set_bb_varying): Remove.
24398 (sbr_vector::get_bb_range): Adjust assert.
24399 (sbr_vector::bb_range_p): Adjust assert.
24400 (~block_range_cache): No freeing loop required.
24401 (block_range_cache::get_block_ranges): Remove.
24402 (block_range_cache::set_bb_range): Inline get_block_ranges.
24403 (block_range_cache::set_bb_varying): Remove.
24404 * gimple-range-cache.h (set_bb_varying): Remove prototype.
24405 * value-range.h (irange_allocator::get_memory): New.
24407 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
24409 * gimple-range-cache.cc (non_null_ref::non_null_deref_p): Search
24410 dominator tree is available and requested.
24411 (ranger_cache::ssa_range_in_bb): Don't search dom tree here.
24412 (ranger_cache::fill_block_cache): Don't search dom tree here either.
24413 * gimple-range-cache.h (non_null_deref_p): Add dom_search param.
24415 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
24417 * gimple-range.cc (gimple_ranger::range_on_exit): Handle block with
24418 only PHI nodes better.
24420 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
24422 * gimple-range-edge.h (gimple_outgoing_range): Rename from
24424 (gcond_edge_range): Export prototype.
24425 * gimple-range-edge.cc (gcond_edge_range): New.
24426 (gimple_outgoing_range::edge_range_p): Use gcond_edge_range.
24427 * gimple-range-gori.h (gori_compute): Use gimple_outgoing_range.
24429 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
24431 * gimple-range-edge.cc (outgoing_range::calc_switch_ranges): Compute
24432 default range into a temp and allocate only what is needed.
24434 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
24436 * range-op.cc (operator_trunc_mod::wi_fold): x % 0 is UNDEFINED.
24438 2021-05-07 Andrew MacLeod <amacleod@redhat.com>
24440 * gimple-range.h (gimple_range_global): Pick up parameter initial
24441 values, and use-before defined locals are UNDEFINED.
24443 2021-05-07 Eric Botcazou <ebotcazou@adacore.com>
24445 * doc/extend.texi (scalar_storage_order): Mention effect on pointer
24447 * tree.h (reverse_storage_order_for_component_p): Return false if
24448 the type is a pointer.
24450 2021-05-07 Andrew Stubbs <ams@codesourcery.com>
24452 * config/gcn/gcn.c (gcn_scalar_mode_supported_p): Disable TImode.
24454 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
24457 * config/i386/i386-expand.c (ix86_expand_sse_movcc):
24458 Handle V8QI, V4HI and V2SI modes.
24459 * config/i386/mmx.md (mmx_pblendvb): New insn pattern.
24460 * config/i386/sse.md (unspec): Move UNSPEC_BLENDV ...
24461 * config/i386/i386.md (unspec): ... here.
24463 2021-05-07 Tobias Burnus <tobias@codesourcery.com>
24464 Tom de Vries <tdevries@suse.de>
24466 * omp-low.c (lower_rec_simd_input_clauses): Set max_vf = 1 if
24467 a truth_value_p reduction variable is nonintegral.
24469 2021-05-07 Uroš Bizjak <ubizjak@gmail.com>
24472 * config/i386/i386-expand.c (ix86_use_mask_cmp_p):
24473 Return false for mode sizes < 16.
24475 2021-05-07 Jakub Jelinek <jakub@redhat.com>
24478 * config/i386/mmx.md (*xop_pcmov_<mode>): New define_insn.
24480 2021-05-06 Martin Jambor <mjambor@suse.cz>
24482 * ipa-sra.c (ipa_sra_dump_all_summaries): Dump edge summaries even
24483 when there is no function summary.
24484 (ipa_sra_summarize_function): produce edge summaries even when
24487 2021-05-06 Tom Tromey <tom@tromey.com>
24489 * godump.c (string_hash_eq): Remove.
24490 (go_finish): Use htab_eq_string.
24492 2021-05-06 Tom Tromey <tom@tromey.com>
24494 * gengtype-state.c (read_state): Use htab_eq_string.
24495 (string_eq): Remove.
24497 2021-05-06 Tom Tromey <tom@tromey.com>
24499 * gensupport.c (htab_eq_string): Remove.
24501 2021-05-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
24504 * debug.h (gcc_debug_hooks): Add set_ignored_loc function pointer.
24505 * dwarf2out.h (dw_fde_node::ignored_debug): New data item.
24506 * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Add dummy
24507 set_ignored_loc callbacks.
24508 * debug.c (do_nothing_debug_hooks): Likewise.
24509 * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
24510 * dwarf2out.c (text_section_used, cold_text_section_used): Remove.
24511 (in_text_section_p, last_text_label, last_cold_label,
24512 switch_text_ranges, switch_cold_ranges): New data items.
24513 (dwarf2out_note_section_used): Remove.
24514 (dwarf2out_begin_prologue): Set fde->ignored_debug and
24516 (mark_ignored_debug_section): New helper function.
24517 (dwarf2out_end_epilogue, dwarf2out_switch_text_section): Call
24518 mark_ignored_debug_section.
24519 (dwarf2_debug_hooks): Use dwarf2out_set_ignored_loc.
24520 (dwarf2_lineno_debug_hooks): Use dummy for set_ignored_loc.
24521 (size_of_aranges): Adjust formula for multi-part text ranges size.
24522 (output_aranges): Output multi-part text ranges.
24523 (dwarf2out_set_ignored_loc): New callback function.
24524 (dwarf2out_finish): Output multi-part text ranges.
24525 (dwarf2out_c_finalize): Clear new data items.
24526 * final.c (final_start_function_1): Call set_ignored_loc callback.
24527 (final_scan_insn_1): Likewise.
24528 * ggc-page.c (gt_ggc_mx): New helper function.
24529 * stringpool.c (gt_pch_nx): Likewise.
24531 2021-05-06 Richard Biener <rguenther@suse.de>
24533 * timevar.def (TV_TREE_INSERT_PHI_NODES): Remove.
24534 (TV_TREE_SSA_REWRITE_BLOCKS): Likewise.
24535 (TV_TREE_INTO_SSA): New.
24536 * tree-into-ssa.c (insert_phi_nodes): Do not account separately.
24537 (rewrite_blocks): Likewise.
24538 (pass_data_build_ssa): Account to TV_TREE_INTO_SSA.
24540 2021-05-06 Jakub Jelinek <jakub@redhat.com>
24542 * tree-ssa-phiopt.c (value_replacement, minmax_replacement,
24543 abs_replacement, xor_replacement,
24544 cond_removal_in_popcount_clz_ctz_pattern,
24545 replace_phi_edge_with_variable): Change type of phi argument from
24546 gimple * to gphi *.
24548 2021-05-06 Richard Biener <rguenther@suse.de>
24550 * tree-ssa-loop-split.c (split_loop): Delay updating SSA form.
24551 Output an opt-info message.
24552 (do_split_loop_on_cond): Likewise.
24553 (tree_ssa_split_loops): Update SSA form here.
24555 2021-05-06 Richard Biener <rguenther@suse.de>
24557 * tree-inline.c (tree_function_versioning): Fix DECL_BY_REFERENCE
24558 return variable removal.
24560 2021-05-06 Marius Hillenbrand <mhillen@linux.ibm.com>
24562 * config/s390/s390-builtins.def (O_M5, O1_M5, ...): Remove unused macros.
24563 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
24564 (s390_vec_permi_dbl, s390_vpdi): Use the O3_U2 type for the immediate
24566 * config/s390/s390.c (s390_const_operand_ok): Remove unused
24569 2021-05-06 Jakub Jelinek <jakub@redhat.com>
24571 PR tree-optimization/94589
24572 * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Call
24573 spaceship_replacement.
24574 (cond_only_block_p, spaceship_replacement): New functions.
24576 2021-05-06 Richard Biener <rguenther@suse.de>
24579 * tree-emutls.c (gen_emutls_addr): Pass in whether we're
24580 dealing with a debug use and only query existing addresses
24582 (lower_emutls_1): Avoid splitting out addresses for debug
24583 stmts, reset the debug stmt when we fail to find existing
24585 (lower_emutls_phi_arg): Set wi.stmt.
24587 2021-05-06 Christoph Muellner <cmuellner@gcc.gnu.org>
24590 * config/riscv/riscv.c (riscv_block_move_loop): Use cbranch helper.
24591 * config/riscv/riscv.md (cbranch<mode>4): Generate helpers.
24592 (stack_protect_test): Use cbranch helper.
24594 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
24597 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
24598 always return the establisher frame for __builtin_frame_address (0).
24600 2021-05-05 Ivan Sorokin <vanyacpp@gmail.com>
24603 * config/i386/i386-builtins.c (ix86_cpu_model_type_node): New.
24604 (ix86_cpu_model_var): Likewise.
24605 (ix86_cpu_features2_type_node): Likewise.
24606 (ix86_cpu_features2_var): Likewise.
24607 (fold_builtin_cpu): Cache __cpu_model and __cpu_features2 with
24610 2021-05-05 Martin Sebor <msebor@redhat.com>
24612 * passes.def (pass_warn_printf): Run after SSA.
24614 2021-05-05 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24616 * config/arm/neon.md (neon_vtst_combine<mode>): New pattern.
24617 * config/arm/predicates.md (minus_one_operand): New predicate.
24619 2021-05-05 Jeff Law <jlaw@tachyum.com>
24621 * config/avr/avr.md: Remove references to CC_STATUS_INIT.
24623 2021-05-05 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
24625 PR rtl-optimization/100263
24626 * postreload.c (move2add_valid_value_p): Ensure register can
24629 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
24631 PR rtl-optimization/100411
24632 * cfgcleanup.c (try_crossjump_to_edge): Also skip end of prologue
24633 and beginning of function markers.
24635 2021-05-05 Jeff Law <jlaw@tachyum.com>
24637 * config/cr16/cr16.h (NOTICE_UPDATE_CC): Remove.
24638 * config/cr16/cr16.c (notice_update_cc): Remove.
24639 * config/cr16/cr16-protos.h (notice_update_cc): Remove.
24641 2021-05-05 Uroš Bizjak <ubizjak@gmail.com>
24644 * config/i386/i386-expand.c (ix86_expand_int_sse_cmp):
24645 Handle V8QI, V4HI and V2SI modes.
24646 * config/i386/i386.c (ix86_build_const_vector): Handle V2SImode.
24647 (ix86_build_signbit_mask): Ditto.
24648 * config/i386/mmx.md (MMXMODE14): New mode iterator.
24649 (<smaxmin:code><MMXMODE14:mode>3): New expander.
24650 (*mmx_<smaxmin:code><MMXMODE14:mode>3): New insn pattern.
24651 (<umaxmin:code><MMXMODE24:mode>3): New expander.
24652 (*mmx_<umaxmin:code><MMXMODE24:mode>3): New insn pattern.
24653 (vec_cmp<MMXMODEI:mode><MMXMODEI:mode>): New expander.
24654 (vec_cmpu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
24655 (vcond<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
24656 (vcondu<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
24657 (vcond_mask_<MMXMODEI:mode><MMXMODEI:mode>): Ditto.
24659 2021-05-05 Eric Botcazou <ebotcazou@adacore.com>
24661 * dwarf2out.c (loc_list_from_tree_1) <DECL>: During early DWARF, do
24662 not expand the VALUE_EXPR of variables put in the non-local frame.
24663 * gimplify.c (gimplify_type_sizes) <RECORD_TYPE>: If the type is not
24664 to be ignored for debug info, ensure its variable offsets are not.
24666 2021-05-05 Richard Biener <rguenther@suse.de>
24668 PR tree-optimization/79333
24669 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt):
24670 Fold stmt following SSA edges.
24672 2021-05-05 Richard Biener <rguenther@suse.de>
24674 PR middle-end/100394
24675 * calls.c (expand_call): Preserve possibly throwing calls.
24676 * cfgexpand.c (expand_call_stmt): When a call can throw signal
24677 RTL expansion there are side-effects.
24678 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Simplify,
24679 mark all possibly throwing stmts necessary unless we can elide
24681 * tree-ssa-dse.c (pass_dse::execute): Preserve exceptions unless
24682 -fdelete-dead-exceptions.
24683 * tree.h (DECL_PURE_P): Add note about exceptions.
24685 2021-05-05 Alexandre Oliva <oliva@adacore.com>
24687 * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Make it
24690 2021-05-04 David Edelsohn <dje.gcc@gmail.com>
24692 * config/rs6000/rs6000-call.c (rs6000_output_mi_thunk): Use
24693 get_fnname_from_decl for name of thunk.
24694 * config/rs6000/rs6000.c (rs6000_declare_alias): Use assemble_name
24695 and ASM_OUTPUT_LABEL.
24696 (rs6000_xcoff_declare_function_name): Use assemble_name and
24698 (rs6000_xcoff_declare_object_name): Use ASM_OUTPUT_LABEL.
24699 (rs6000_xcoff_encode_section_info): Don't add mapping class
24700 for aliases. Always add [DS] mapping class to primary
24702 (rs6000_asm_weaken_decl): Don't explicitly add [DS].
24704 2021-05-04 Martin Sebor <msebor@redhat.com>
24706 PR middle-end/100307
24707 * builtins.c (compute_objsize_r): Clear base0 for pointers.
24709 2021-05-04 Jeff Law <jlaw@tachyum.com>
24711 * config/bfin/bfin.h (NOTICE_UPDATE_CC): Remove.
24713 2021-05-04 Segher Boessenkool <segher@kernel.crashing.org>
24715 * caller-save.c: Remove CC0.
24716 * cfgcleanup.c: Remove CC0.
24717 * cfgrtl.c: Remove CC0.
24718 * combine.c: Remove CC0.
24719 * compare-elim.c: Remove CC0.
24720 * conditions.h: Remove CC0.
24721 * config/h8300/h8300.h: Remove CC0.
24722 * config/h8300/h8300-protos.h: Remove CC0.
24723 * config/h8300/peepholes.md: Remove CC0.
24724 * config/i386/x86-tune-sched.c: Remove CC0.
24725 * config/m68k/m68k.c: Remove CC0.
24726 * config/rl78/rl78.c: Remove CC0.
24727 * config/sparc/sparc.c: Remove CC0.
24728 * config/xtensa/xtensa.c: Remove CC0.
24729 (gen_conditional_move): Use pc_rtx instead of cc0_rtx in a piece of
24730 RTL where that is used as a placeholder only.
24731 * cprop.c: Remove CC0.
24732 * cse.c: Remove CC0.
24733 * cselib.c: Remove CC0.
24734 * df-problems.c: Remove CC0.
24735 * df-scan.c: Remove CC0.
24736 * doc/md.texi: Remove CC0. Adjust an example.
24737 * doc/rtl.texi: Remove CC0. Adjust an example.
24738 * doc/tm.texi: Regenerate.
24739 * doc/tm.texi.in: Remove CC0.
24740 * emit-rtl.c: Remove CC0.
24741 * final.c: Remove CC0.
24742 * fwprop.c: Remove CC0.
24743 * gcse-common.c: Remove CC0.
24744 * gcse.c: Remove CC0.
24745 * genattrtab.c: Remove CC0.
24746 * genconfig.c: Remove CC0.
24747 * genemit.c: Remove CC0.
24748 * genextract.c: Remove CC0.
24749 * gengenrtl.c: Remove CC0.
24750 * genrecog.c: Remove CC0.
24751 * haifa-sched.c: Remove CC0.
24752 * ifcvt.c: Remove CC0.
24753 * ira-costs.c: Remove CC0.
24754 * ira.c: Remove CC0.
24755 * jump.c: Remove CC0.
24756 * loop-invariant.c: Remove CC0.
24757 * lra-constraints.c: Remove CC0.
24758 * lra-eliminations.c: Remove CC0.
24759 * optabs.c: Remove CC0.
24760 * postreload-gcse.c: Remove CC0.
24761 * postreload.c: Remove CC0.
24762 * print-rtl.c: Remove CC0.
24763 * read-rtl-function.c: Remove CC0.
24764 * reg-notes.def: Remove CC0.
24765 * reg-stack.c: Remove CC0.
24766 * reginfo.c: Remove CC0.
24767 * regrename.c: Remove CC0.
24768 * reload.c: Remove CC0.
24769 * reload1.c: Remove CC0.
24770 * reorg.c: Remove CC0.
24771 * resource.c: Remove CC0.
24772 * rtl.c: Remove CC0.
24773 * rtl.def: Remove CC0.
24774 * rtl.h: Remove CC0.
24775 * rtlanal.c: Remove CC0.
24776 * sched-deps.c: Remove CC0.
24777 * sched-rgn.c: Remove CC0.
24778 * shrink-wrap.c: Remove CC0.
24779 * simplify-rtx.c: Remove CC0.
24780 * system.h: Remove CC0. Poison NOTICE_UPDATE_CC, CC_STATUS_MDEP_INIT,
24781 CC_STATUS_MDEP, and CC_STATUS.
24782 * target.def: Remove CC0.
24783 * valtrack.c: Remove CC0.
24784 * var-tracking.c: Remove CC0.
24786 2021-05-04 Richard Biener <rguenther@suse.de>
24788 PR tree-optimization/100414
24789 * tree-ssa-phiopt.c (get_non_trapping): Do not compute dominance
24791 (tree_ssa_phiopt_worker): But unconditionally here.
24793 2021-05-04 Tobias Burnus <tobias@codesourcery.com>
24795 * omp-low.c (lower_rec_input_clauses, lower_reduction_clauses): Handle
24796 && and || with floating-point and complex arguments.
24798 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
24800 * tree-inline.c (insert_debug_decl_map): Delete.
24801 (copy_debug_stmt): Minor tweak.
24802 (setup_one_parameter): Do not use a variable if the value is either
24803 a read-only DECL or a non-addressable local variable in the caller.
24804 In this case, insert the debug-only variable in the map manually.
24805 (expand_call_inline): Do not generate a CLOBBER for these values.
24806 * tree-inline.h (debug_map): Minor tweak.
24808 2021-05-04 Eric Botcazou <ebotcazou@adacore.com>
24810 * builtins.c (builtin_with_linkage_p): Return true for stp[n]cpy.
24811 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Tidy up.
24813 2021-05-04 Richard Biener <rguenther@suse.de>
24815 PR tree-optimization/100329
24816 * tree-ssa-reassoc.c (can_reassociate_p): Do not reassociate
24818 (insert_stmt_after): Assert we're not running into asm goto.
24820 2021-05-04 Richard Biener <rguenther@suse.de>
24822 PR tree-optimization/100398
24823 * tree-ssa-dse.c (pass_dse::execute): Preserve control
24826 2021-05-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24828 * builtins.c (try_store_by_multiple_pieces): Fix constfun's prototype.
24830 2021-05-04 Alexandre Oliva <oliva@adacore.com>
24832 * builtins.c (try_store_by_multiple_pieces): New.
24833 (expand_builtin_memset_args): Use it. If target_char_cast
24834 fails, proceed as for non-constant val. Pass len's ctz to...
24835 * expr.c (clear_storage_hints): ... this. Try store by
24836 multiple pieces after setmem.
24837 (clear_storage): Adjust.
24838 * expr.h (clear_storage_hints): Likewise.
24839 (try_store_by_multiple_pieces): Declare.
24840 * passes.def: Replace the last copy_prop with ccp.
24842 2021-05-03 Tom de Vries <tdevries@suse.de>
24845 * omp-low.c (lower_rec_input_clauses): Disable SIMT for user-defined
24848 2021-05-03 Richard Biener <rguenther@suse.de>
24850 * tree-ssa-dse.c (dse_classify_store): Track two PHI defs.
24852 2021-05-03 Richard Biener <rguenther@suse.de>
24854 * tree-ssa-dse.c: Do not include domwalk.h but cfganal.h.
24855 (dse_dom_walker): Remove.
24856 (dse_dom_walker::dse_optimize_stmt): Rename...
24857 (dse_optimize_stmt): ... to this, pass in live_bytes sbitmap.
24858 (dse_dom_walker::before_dom_children): Inline ...
24859 (pass_dse::execute): ... here. Perform a reverse program
24862 2021-05-03 H.J. Lu <hjl.tools@gmail.com>
24865 * configure: Regenerated.
24867 2021-05-03 Ilya Leoshkevich <iii@linux.ibm.com>
24870 * config/s390/s390.c (s390_hard_fp_reg_p): New function.
24871 (s390_md_asm_adjust): Handle hard registers.
24873 2021-05-03 Jakub Jelinek <jakub@redhat.com>
24875 PR tree-optimization/100382
24876 * tree-ssa-dse.c: Include tree-eh.h.
24877 (dse_dom_walker::before_dom_children): Don't remove stmts if
24878 stmt_unremovable_because_of_non_call_eh_p is true.
24880 2021-05-02 David Edelsohn <dje.gcc@gmail.com>
24882 * varasm.c (compute_reloc_for_var): Split out from...
24883 (get_variable_section): Use it.
24884 * output.h (compute_reloc_for_var): Declare.
24885 * config/rs6000/rs6000-protos.h
24886 (rs6000_xcoff_asm_output_aligned_decl_common): Change alignment to
24888 * config/rs6000/rs6000.c (rs6000_legitimize_tls_address_aix):
24889 Don't append storage mapping class to symbol.
24890 (rs6000_xcoff_asm_named_section): Add BS and UL mapping classes.
24891 Don't convert TLS BSS to common.
24892 (rs6000_xcoff_unique_section): Don't fall back to select_secton.
24893 (rs6000_xcoff_section_type_flags): Add SECTION_BSS if DECL is
24895 (rs6000_xcoff_asm_globalize_decl_name): Don't strip storage
24897 (rs6000_xcoff_asm_output_aligned_decl_common): Align is unsigned int.
24898 If align is 0 from TLS class, use the same rules as varasm.c
24899 If not common, switch to BSS section manually.
24900 If common, emit appropriate comm or lcomm directive.
24901 (rs6000_xcoff_encode_section_info): Add logic to append all
24902 storage mapping classes.
24903 (rs6000_asm_weaken_decl): Adjust for qualname symbols.
24904 * config/rs6000/xcoff.h (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
24905 rs6000_xcoff_asm_output_aligned_decl_common.
24906 (ASM_OUTPUT_ALIGNED_DECL_LOCAL): Use
24907 rs6000_xcoff_asm_output_aligned_decl_common.
24908 (ASM_OUTPUT_TLS_COMMON): Use
24909 rs6000_xcoff_asm_output_aligned_decl_common.
24911 2021-05-02 Jakub Jelinek <jakub@redhat.com>
24914 * config/nvptx/nvptx.c (nvptx_sese_pseudo): Use nullptr instead of 0
24915 as first argument of pseudo_node_t constructors.
24917 2021-05-02 Jakub Jelinek <jakub@redhat.com>
24920 * config/i386/t-i386 (TM_H): Add $(srcdir)/config/i386/i386-isa.def.
24922 2021-05-01 Aldy Hernandez <aldyh@redhat.com>
24924 * value-range.cc (DEFINE_INT_RANGE_GC_STUBS): Remove.
24925 (gt_pch_nx (int_range<1> *&)): New.
24926 (gt_ggc_mx (int_range<1> *&)): New.
24927 * value-range.h (class irange): Add GTY support for
24930 2021-05-01 Geng Qi <gengqi@linux.alibaba.com>
24932 * doc/options.texi (Negative): Change either or to both and.
24934 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
24936 * config/aarch64/aarch64-simd-builtins.def: Add
24937 float_ml[as][q]_laneq builtin generator macros.
24938 * config/aarch64/aarch64-simd.md (mul_laneq<mode>3): Define.
24939 (aarch64_float_mla_laneq<mode>): Define.
24940 (aarch64_float_mls_laneq<mode>): Define.
24941 * config/aarch64/arm_neon.h (vmla_laneq_f32): Use RTL builtin
24942 instead of GCC vector extensions.
24943 (vmlaq_laneq_f32): Likewise.
24944 (vmls_laneq_f32): Likewise.
24945 (vmlsq_laneq_f32): Likewise.
24947 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
24949 * config/aarch64/aarch64-simd-builtins.def: Add
24950 float_ml[as]_lane builtin generator macros.
24951 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt<mode>):
24953 (mul_lane<mode>3): This, and re-order arguments.
24954 (aarch64_float_mla_lane<mode>): Define.
24955 (aarch64_float_mls_lane<mode>): Define.
24956 * config/aarch64/arm_neon.h (vmla_lane_f32): Use RTL builtin
24957 instead of GCC vector extensions.
24958 (vmlaq_lane_f32): Likewise.
24959 (vmls_lane_f32): Likewise.
24960 (vmlsq_lane_f32): Likewise.
24962 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
24964 * config/aarch64/aarch64-simd-builtins.def: Add float_ml[as]
24965 builtin generator macros.
24966 * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>):
24968 (aarch64_float_mls<mode>): Define.
24969 * config/aarch64/arm_neon.h (vmla_f32): Use RTL builtin
24970 instead of relying on GCC vector extensions.
24971 (vmla_f64): Likewise.
24972 (vmlaq_f32): Likewise.
24973 (vmlaq_f64): Likewise.
24974 (vmls_f32): Likewise.
24975 (vmls_f64): Likewise.
24976 (vmlsq_f32): Likewise.
24977 (vmlsq_f64): Likewise.
24978 * config/aarch64/iterators.md: Define VDQF_DF mode iterator.
24980 2021-04-30 Jonathan Wright <jonathan.wright@arm.com>
24982 * config/aarch64/aarch64-simd-builtins.def: Add
24983 float_ml[as]_n_builtin generator macros.
24984 * config/aarch64/aarch64-simd.md (*aarch64_mul3_elt_from_dup<mode>):
24986 (mul_n<mode>3): This, and re-order arguments.
24987 (aarch64_float_mla_n<mode>): Define.
24988 (aarch64_float_mls_n<mode>): Define.
24989 * config/aarch64/arm_neon.h (vmla_n_f32): Use RTL builtin
24990 instead of inline asm.
24991 (vmlaq_n_f32): Likewise.
24992 (vmls_n_f32): Likewise.
24993 (vmlsq_n_f32): Likewise.
24995 2021-04-30 Jonathan Wright <joanthan.wright@arm.com>
24997 * config/aarch64/aarch64-simd-builtins.def: Add pmull[2]
24998 builtin generator macros.
24999 * config/aarch64/aarch64-simd.md (aarch64_pmullv8qi): Define.
25000 (aarch64_pmull_hiv16qi_insn): Define.
25001 (aarch64_pmull_hiv16qi): Define.
25002 * config/aarch64/arm_neon.h (vmull_high_p8): Use RTL builtin
25003 instead of inline asm.
25004 (vmull_p8): Likewise.
25006 2021-04-30 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
25008 * config/avr/avr.md: Adjust peepholes to match and
25009 generate parallels with clobber of REG_CC.
25010 (mov<mode>_insn): Rename to mov<mode>_insn_split.
25011 (*mov<mode>_insn): Rename to mov<mode>_insn.
25013 2021-04-30 David Edelsohn <dje.gcc@gmail.com>
25015 * varasm.c (use_blocks_for_decl_p): Don't use section anchors
25016 for VAR_DECLs if -fdata-sections enabled.
25018 2021-04-30 Michael Meissner <meissner@linux.ibm.com>
25020 PR bootstrap/100327
25021 * config/rs6000/rs6000.c
25022 (TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P): Define.
25023 (rs6000_libgcc_floating_mode_supported_p): New target hook.
25025 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
25027 * tree-ssa-threadbackward.c (class thread_jumps): Split out code
25029 (class back_threader_registry): ...to here...
25030 (class back_threader_profitability): ...and here...
25031 (thread_jumps::thread_through_all_blocks): Remove argument.
25032 (back_threader_registry::back_threader_registry): New.
25033 (back_threader_registry::~back_threader_registry): New.
25034 (back_threader_registry::thread_through_all_blocks): New.
25035 (thread_jumps::profitable_jump_thread_path): Move from here...
25036 (back_threader_profitability::profitable_path_p): ...to here.
25037 (thread_jumps::find_taken_edge): New.
25038 (thread_jumps::convert_and_register_current_path): Move...
25039 (back_threader_registry::register_path): ...to here.
25040 (thread_jumps::register_jump_thread_path_if_profitable): Move...
25041 (thread_jumps::maybe_register_path): ...to here.
25042 (thread_jumps::handle_phi): Call find_taken_edge and
25043 maybe_register_path.
25044 (thread_jumps::handle_assignment): Same.
25045 (thread_jumps::fsm_find_control_statement_thread_paths): Remove
25046 tree argument to handle_phi and handle_assignment.
25047 (thread_jumps::find_jump_threads_backwards): Set m_name. Remove
25048 set of m_speed_p and m_max_threaded_paths.
25049 (pass_thread_jumps::execute): Remove second argument from
25050 find_jump_threads_backwards.
25051 (pass_early_thread_jumps::execute): Same.
25053 2021-04-30 Aldy Hernandez <aldyh@redhat.com>
25055 * tree-ssa-dom.c (class dom_jump_threader_simplifier): New.
25056 (class dom_opt_dom_walker): Initialize some class variables.
25057 (pass_dominator::execute): Pass evrp_range_analyzer and
25058 dom_jump_threader_simplifier to dom_opt_dom_walker.
25059 Adjust for some functions moving into classes.
25060 (simplify_stmt_for_jump_threading): Adjust and move to...
25061 (jump_threader_simplifier::simplify): ...here.
25062 (dom_opt_dom_walker::before_dom_children): Adjust for
25063 m_evrp_range_analyzer.
25064 (dom_opt_dom_walker::after_dom_children): Remove x_vr_values hack.
25065 (test_for_singularity): Place in dom_opt_dom_walker class.
25066 (dom_opt_dom_walker::optimize_stmt): The argument
25067 evrp_range_analyzer is now a class field.
25068 * tree-ssa-threadbackward.c (class thread_jumps): Add m_registry.
25069 (thread_jumps::thread_through_all_blocks): New.
25070 (thread_jumps::convert_and_register_current_path): Use m_registry.
25071 (pass_thread_jumps::execute): Adjust for thread_through_all_blocks
25072 being in the threader class.
25073 (pass_early_thread_jumps::execute): Same.
25074 * tree-ssa-threadedge.c (threadedge_initialize_values): Move...
25075 (jump_threader::jump_threader): ...here.
25076 (threadedge_finalize_values): Move...
25077 (jump_threader::~jump_threader): ...here.
25078 (jump_threader::remove_jump_threads_including): New.
25079 (jump_threader::thread_through_all_blocks): New.
25080 (record_temporary_equivalences_from_phis): Move...
25081 (jump_threader::record_temporary_equivalences_from_phis): ...here.
25082 (record_temporary_equivalences_from_stmts_at_dest): Move...
25083 (jump_threader::record_temporary_equivalences_from_stmts_at_dest):
25085 (simplify_control_stmt_condition_1): Move to jump_threader class.
25086 (simplify_control_stmt_condition): Move...
25087 (jump_threader::simplify_control_stmt_condition): ...here.
25088 (thread_around_empty_blocks): Move...
25089 (jump_threader::thread_around_empty_blocks): ...here.
25090 (thread_through_normal_block): Move...
25091 (jump_threader::thread_through_normal_block): ...here.
25092 (thread_across_edge): Move...
25093 (jump_threader::thread_across_edge): ...here.
25094 (thread_outgoing_edges): Move...
25095 (jump_threader::thread_outgoing_edges): ...here.
25096 * tree-ssa-threadedge.h: Move externally facing functings...
25097 (class jump_threader): ...here...
25098 (class jump_threader_simplifier): ...and here.
25099 * tree-ssa-threadupdate.c (struct redirection_data): Remove comment.
25100 (jump_thread_path_allocator::jump_thread_path_allocator): New.
25101 (jump_thread_path_allocator::~jump_thread_path_allocator): New.
25102 (jump_thread_path_allocator::allocate_thread_edge): New.
25103 (jump_thread_path_allocator::allocate_thread_path): New.
25104 (jump_thread_path_registry::jump_thread_path_registry): New.
25105 (jump_thread_path_registry::~jump_thread_path_registry): New.
25106 (jump_thread_path_registry::allocate_thread_edge): New.
25107 (jump_thread_path_registry::allocate_thread_path): New.
25108 (dump_jump_thread_path): Make extern.
25109 (debug (const vec<jump_thread_edge *> &path)): New.
25110 (struct removed_edges): Move to tree-ssa-threadupdate.h.
25111 (struct thread_stats_d): Remove.
25112 (remove_ctrl_stmt_and_useless_edges): Make static.
25113 (lookup_redirection_data): Move...
25114 (jump_thread_path_registry::lookup_redirection_data): ...here.
25115 (ssa_redirect_edges): Make static.
25116 (thread_block_1): Move...
25117 (jump_thread_path_registry::thread_block_1): ...here.
25118 (thread_block): Move...
25119 (jump_thread_path_registry::thread_block): ...here.
25120 (thread_through_loop_header): Move...
25121 (jump_thread_path_registry::thread_through_loop_header): ...here.
25122 (mark_threaded_blocks): Move...
25123 (jump_thread_path_registry::mark_threaded_blocks): ...here.
25124 (debug_path): Move...
25125 (jump_thread_path_registry::debug_path): ...here.
25126 (debug_all_paths): Move...
25127 (jump_thread_path_registry::dump): ..here.
25128 (rewire_first_differing_edge): Move...
25129 (jump_thread_path_registry::rewire_first_differing_edge): ...here.
25130 (adjust_paths_after_duplication): Move...
25131 (jump_thread_path_registry::adjust_paths_after_duplication): ...here.
25132 (duplicate_thread_path): Move...
25133 (jump_thread_path_registry::duplicate_thread_path): ..here.
25134 (remove_jump_threads_including): Move...
25135 (jump_thread_path_registry::remove_jump_threads_including): ...here.
25136 (thread_through_all_blocks): Move to...
25137 (jump_thread_path_registry::thread_through_all_blocks): ...here.
25138 (delete_jump_thread_path): Remove.
25139 (register_jump_thread): Move...
25140 (jump_thread_path_registry::register_jump_thread): ...here.
25141 * tree-ssa-threadupdate.h: Move externally facing functions...
25142 (class jump_thread_path_allocator): ...here...
25143 (class jump_thread_path_registry): ...and here.
25144 (thread_through_all_blocks): Remove.
25145 (struct removed_edges): New.
25146 (register_jump_thread): Remove.
25147 (remove_jump_threads_including): Remove.
25148 (delete_jump_thread_path): Remove.
25149 (remove_ctrl_stmt_and_useless_edges): Remove.
25150 (free_dom_edge_info): New prototype.
25151 * tree-vrp.c: Remove x_vr_values hack.
25152 (class vrp_jump_threader_simplifier): New.
25153 (vrp_jump_threader_simplifier::simplify): New.
25154 (vrp_jump_threader::vrp_jump_threader): Adjust method signature.
25155 Remove m_dummy_cond.
25156 Instantiate m_simplifier and m_threader.
25157 (vrp_jump_threader::thread_through_all_blocks): New.
25158 (vrp_jump_threader::simplify_stmt): Remove.
25159 (vrp_jump_threader::after_dom_children): Do not set m_dummy_cond.
25160 Remove x_vr_values hack.
25161 (execute_vrp): Adjust for thread_through_all_blocks being in a
25164 2021-04-30 Christophe Lyon <christophe.lyon@linaro.org>
25166 * genflags.c (gen_insn): Print failed expansion string.
25168 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
25170 * expr.c (alignment_for_piecewise_move): Call mode_for_size
25171 without limit to MAX_FIXED_MODE_SIZE.
25173 2021-04-30 H.J. Lu <hjl.tools@gmail.com>
25175 PR middle-end/90773
25176 * builtins.c (builtin_memset_gen_str): Don't use return from
25177 simplify_gen_subreg.
25179 2021-04-30 Uroš Bizjak <ubizjak@gmail.com>
25182 * config/i386/i386.md (*add<mode>3_carry_0r): New insn pattern.
25183 (*addsi3_carry_zext_0r): Ditto.
25184 (*sub<mode>3_carry_0): Ditto.
25185 (*subsi3_carry_zext_0r): Ditto.
25186 * config/i386/predicates.md (ix86_carry_flag_unset_operator):
25188 * config/i386/i386.c (ix86_rtx_costs) <case PLUS, case MINUS>:
25189 Also consider ix86_carry_flag_unset_operator to calculate
25190 the cost of adc/sbb insn.
25192 2021-04-30 Roman Zhuykov <zhroma@ispras.ru>
25194 PR rtl-optimization/100225
25195 PR rtl-optimization/84878
25196 * modulo-sched.c (sms_schedule): Use note_stores to skip loops
25197 where we have an instruction which touches (writes) any hard
25198 register from df->regular_block_artificial_uses set.
25199 Allow not-single-set instruction only right before basic block
25202 2021-04-30 Geng Qi <gengqi@linux.alibaba.com>
25204 * config/riscv/riscv.opt (march=,mabi=): Negative itself.
25206 2021-04-30 LevyHsu <admin@levyhsu.com>
25208 * config/riscv/riscv.c (riscv_min_arithmetic_precision): New.
25209 * config/riscv/riscv.h (TARGET_MIN_ARITHMETIC_PRECISION): New.
25210 * config/riscv/riscv.md (addv<mode>4, uaddv<mode>4): New.
25211 (subv<mode>4, usubv<mode>4, mulv<mode>4, umulv<mode>4): New.
25213 2021-04-29 Alexandre Oliva <oliva@adacore.com>
25215 * config.gcc: Merged x86 and x86_64 cpu_type-setting cases.
25217 2021-04-29 Alexandre Oliva <oliva@adacore.com>
25219 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Rename to...
25220 (ASM_OUTPUT_MAX_SKIP_ALIGN): ... this. Enclose in do/while(0).
25221 * config/i386/i386.c: Adjust.
25222 * config/i386/i386.md: Adjust.
25223 * config/i386/darwin.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Drop.
25224 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
25225 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
25226 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
25227 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
25228 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
25229 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
25230 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
25231 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
25232 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
25233 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
25235 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
25237 * config/i386/i386-expand.c (ix86_expand_int_compare):
25238 Swap operands of GTU and LEU comparison to emit carry flag comparison.
25239 * config/i386/i386.md (*add<mode>3_carry_0): Change insn
25240 predicate to allow more combine opportunities with memory operands.
25241 (*sub<mode>3_carry_0): Ditto.
25243 2021-04-29 Richard Sandiford <richard.sandiford@arm.com>
25245 PR rtl-optimization/100303
25246 * rtl-ssa/accesses.cc (function_info::make_use_available): Take a
25247 boolean that indicates whether the use will only be used in
25248 debug instructions. Treat it in the same way that existing
25249 cross-EBB debug references would be handled if so.
25250 (function_info::make_uses_available): Likewise.
25251 * rtl-ssa/functions.h (function_info::make_uses_available): Update
25252 prototype accordingly.
25253 (function_info::make_uses_available): Likewise.
25254 * fwprop.c (try_fwprop_subst): Update call accordingly.
25256 2021-04-29 Jeff Law <jlaw@tachyum.com>
25258 * config/nios2/nios2-protos.h (nios2_fpu_insn_enabled): Move outside
25261 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
25262 Richard Biener <rguenther@suse.de>
25265 * config/i386/i386-builtin.def (IX86_BUILTIN_MASKLOADPD)
25266 (IX86_BUILTIN_MASKLOADPS, IX86_BUILTIN_MASKLOADPD256)
25267 (IX86_BUILTIN_MASKLOADPS256, IX86_BUILTIN_MASKLOADD)
25268 (IX86_BUILTIN_MASKLOADQ, IX86_BUILTIN_MASKLOADD256)
25269 (IX86_BUILTIN_MASKLOADQ256): Move from SPECIAL_ARGS
25270 to PURE_ARGS category.
25271 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
25272 Handle PURE_ARGS category.
25273 * config/i386/i386-expand.c (ix86_expand_builtin): Ditto.
25275 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
25277 * configure.ac: Check for the presence of sys/locking.h header and
25278 for whether _LK_LOCK is supported by _locking.
25279 * configure: Regenerate.
25280 * config.in: Likewise.
25281 * gcov-io.h: Define GCOV_LOCKED_WITH_LOCKING if HOST_HAS_LK_LOCK.
25282 * gcov-io.c (gcov_open): Add support for GCOV_LOCKED_WITH_LOCKING.
25283 * system.h: Include <sys/locking.h> if HAVE_SYS_LOCKING_H.
25285 2021-04-29 Uroš Bizjak <ubizjak@gmail.com>
25287 * config/i386/predicates.md (fcmov_comparison_operator):
25288 Do not check for trivial FP comparison operator.
25289 <case GEU, case LTU>: Allow CCGZmode.
25290 <case GTU, case LEU>: Do not allow CCCmode.
25291 (ix86_comparison_operator) <case GTU, case LEU>: Allow only CCmode.
25292 (ix86_carry_flag_operator): Match only LTU and UNLT code.
25293 Do not check for trivial FP comparison operator. Allow CCGZmode.
25295 2021-04-29 Tom de Vries <tdevries@suse.de>
25297 * omp-expand.c (expand_omp_simd): Add step_orig, and replace uses of
25298 fd->loop.step by either step or orig_step.
25300 2021-04-29 Eric Botcazou <ebotcazou@adacore.com>
25302 * config/sparc/sparc.c (gen_load_pcrel_sym): Delete.
25303 (load_got_register): Do the PIC dance here.
25304 (sparc_legitimize_tls_address): Simplify.
25305 (sparc_emit_probe_stack_range): Likewise.
25306 (sparc32_initialize_trampoline): Likewise.
25307 (sparc64_initialize_trampoline): Likewise.
25308 * config/sparc/sparc.md (load_pcrel_sym<P:mode>): Add @ marker.
25309 (probe_stack_range<P:mode>): Likewise.
25310 (flush<P:mode>): Likewise.
25311 (tgd_hi22<P:mode>): Likewise.
25312 (tgd_lo10<P:mode>): Likewise.
25313 (tgd_add<P:mode>): Likewise.
25314 (tgd_call<P:mode>): Likewise.
25315 (tldm_hi22<P:mode>): Likewise.
25316 (tldm_lo10<P:mode>): Likewise.
25317 (tldm_add<P:mode>): Likewise.
25318 (tldm_call<P:mode>): Likewise.
25319 (tldo_hix22<P:mode>): Likewise.
25320 (tldo_lox10<P:mode>): Likewise.
25321 (tldo_add<P:mode>): Likewise.
25322 (tie_hi22<P:mode>): Likewise.
25323 (tie_lo10<P:mode>): Likewise.
25324 (tie_add<P:mode>): Likewise.
25325 (tle_hix22<P:mode>): Likewise.
25326 (tle_lox10<P:mode>): Likewise.
25327 (stack_protect_setsi): Rename to...
25328 (stack_protect_set32): ...this.
25329 (stack_protect_setdi): Rename to...
25330 (stack_protect_set64): ...this.
25331 (stack_protect_set): Adjust calls to above.
25332 (stack_protect_testsi): Rename to...
25333 (stack_protect_test32): ...this.
25334 (stack_protect_testdi): Rename to...
25335 (stack_protect_test64): ...this.
25336 (stack_protect_test): Adjust calls to above.
25338 2021-04-29 H.J. Lu <hjl.tools@gmail.com>
25340 PR middle-end/90773
25341 * builtins.c (builtin_memcpy_read_str): Add a dummy argument.
25342 (builtin_strncpy_read_str): Likewise.
25343 (builtin_memset_read_str): Add an argument for the previous RTL
25344 information and generate the new RTL from the previous RTL info.
25345 (builtin_memset_gen_str): Likewise.
25346 * builtins.h (builtin_strncpy_read_str): Update the prototype.
25347 (builtin_memset_read_str): Likewise.
25348 * expr.c (by_pieces_ninsns): If targetm.overlap_op_by_pieces_p()
25349 returns true, round up size and alignment to the widest integer
25350 mode for maximum size.
25351 (pieces_addr::adjust): Add a pointer to by_pieces_prev argument
25352 and pass it to m_constfn.
25353 (op_by_pieces_d): Add m_push and m_overlap_op_by_pieces.
25354 (op_by_pieces_d::op_by_pieces_d): Add a bool argument to
25355 initialize m_push. Initialize m_overlap_op_by_pieces with
25356 targetm.overlap_op_by_pieces_p ().
25357 (op_by_pieces_d::run): Pass the previous RTL information to
25358 pieces_addr::adjust and generate overlapping operations if
25359 m_overlap_op_by_pieces is true.
25361 (move_by_pieces_d::move_by_pieces_d): Updated for op_by_pieces_d
25363 (store_by_pieces_d::store_by_pieces_d): Updated for op_by_pieces_d
25365 (can_store_by_pieces): Use by_pieces_constfn on constfun.
25366 (store_by_pieces): Use by_pieces_constfn on constfun. Updated
25367 for op_by_pieces_d change.
25368 (clear_by_pieces_1): Add a dummy argument.
25369 (clear_by_pieces): Updated for op_by_pieces_d change.
25370 (compare_by_pieces_d::compare_by_pieces_d): Likewise.
25371 (string_cst_read_str): Add a dummy argument.
25372 * expr.h (by_pieces_constfn): Add a dummy argument.
25373 (by_pieces_prev): New.
25374 * target.def (overlap_op_by_pieces_p): New target hook.
25375 * config/i386/i386.c (TARGET_OVERLAP_OP_BY_PIECES_P): New.
25376 * doc/tm.texi.in: Add TARGET_OVERLAP_OP_BY_PIECES_P.
25377 * doc/tm.texi: Regenerated.
25379 2021-04-29 Richard Biener <rguenther@suse.de>
25381 PR tree-optimization/100253
25382 * tree-vect-stmts.c (vectorizable_load): Do not assume
25383 element alignment when DR_MISALIGNMENT is -1.
25384 (vectorizable_store): Likewise.
25386 2021-04-29 Jakub Jelinek <jakub@redhat.com>
25389 * config/aarch64/aarch64.c (aarch64_add_offset_1_temporaries): Use
25390 absu_hwi instead of abs_hwi.
25392 2021-04-29 Richard Biener <rguenther@suse.de>
25394 PR middle-end/38474
25395 * tree-ssa-structalias.c (add_graph_edge): Avoid direct
25396 forwarding when indirect forwarding through ESCAPED
25399 2021-04-29 Tom de Vries <tdevries@suse.de>
25402 * internal-fn.c (expand_GOMP_SIMT_ENTER_ALLOC)
25403 (expand_GOMP_SIMT_LAST_LANE, expand_GOMP_SIMT_ORDERED_PRED)
25404 (expand_GOMP_SIMT_VOTE_ANY, expand_GOMP_SIMT_XCHG_BFLY)
25405 (expand_GOMP_SIMT_XCHG_IDX): Ensure target is assigned to.
25407 2021-04-29 Richard Biener <rguenther@suse.de>
25409 PR tree-optimization/99912
25410 * tree-ssa-dse.c (dse_dom_walker::m_need_cfg_cleanup): New.
25411 (dse_dom_walker::todo): Likewise.
25412 (dse_dom_walker::dse_optimize_stmt): Move VDEF check to the
25414 (dse_dom_walker::before_dom_children): Remove trivially
25415 dead SSA defs and schedule CFG cleanup if we removed all
25417 (pass_dse::execute): Get TODO as computed by the DOM walker
25418 and return it. Wipe dominator info earlier.
25420 2021-04-29 Richard Biener <rguenther@suse.de>
25423 * ipa-prop.c (ipcp_modif_dom_walker::before_dom_children):
25424 Track blocks to cleanup EH in new m_need_eh_cleanup.
25425 (ipcp_modif_dom_walker::cleanup_eh): New.
25426 (ipcp_transform_function): Release dominator info before
25429 2021-04-29 Martin Sebor <msebor@redhat.com>
25431 PR middle-end/100250
25432 * attribs.c (attr_access::array_as_string): Avoid dereferencing
25433 a pointer when it's null.
25435 2021-04-29 Martin Sebor <msebor@redhat.com>
25437 * Makefile.in (OBJS): Add ipa-free-lang-data.o.
25438 * ipa-free-lang-data.cc: New file.
25439 * tree.c: Move pass free_lang_data to file above.
25440 (build_array_type_1): Declare extern.
25441 * tree.h (build_array_type_1): Declare.
25443 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
25445 * config/aarch64/aarch64-simd-builtins.def: Modify comment to
25446 make consistent with updated RTL pattern.
25447 * config/aarch64/aarch64-simd.md (aarch64_<sur>qmovn<mode>):
25448 Implement using ss_truncate and us_truncate rather than
25450 * config/aarch64/iterators.md: Remove redundant unspecs and
25451 iterator: UNSPEC_[SU]QXTN and SUQMOVN respectively.
25453 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
25455 * config/aarch64/arm_acle.h (__attribute__): Make intrinsic
25456 attributes consistent with those defined in arm_neon.h.
25458 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
25460 * config/aarch64/arm_fp16.h (__attribute__): Make intrinsic
25461 attributes consistent with those defined in arm_neon.h.
25463 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
25465 * config/aarch64/aarch64-simd-builtins.def: Add
25466 float_trunc_rodd builtin generator macros.
25467 * config/aarch64/aarch64-simd.md (aarch64_float_trunc_rodd_df):
25469 (aarch64_float_trunc_rodd_lo_v2sf): Define.
25470 (aarch64_float_trunc_rodd_hi_v4sf_le): Define.
25471 (aarch64_float_trunc_rodd_hi_v4sf_be): Define.
25472 (aarch64_float_trunc_rodd_hi_v4sf): Define.
25473 * config/aarch64/arm_neon.h (vcvtx_f32_f64): Use RTL builtin
25474 instead of inline asm.
25475 (vcvtx_high_f32_f64): Likewise.
25476 (vcvtxd_f32_f64): Likewise.
25477 * config/aarch64/iterators.md: Add FCVTXN unspec.
25479 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
25481 * config/aarch64/aarch64-simd-builtins.def: Add tbx1 builtin
25483 * config/aarch64/aarch64-simd.md (aarch64_tbx1<mode>):
25485 * config/aarch64/arm_neon.h (vqtbx1_s8): USE RTL builtin
25486 instead of inline asm.
25487 (vqtbx1_u8): Likewise.
25488 (vqtbx1_p8): Likewise.
25489 (vqtbx1q_s8): Likewise.
25490 (vqtbx1q_u8): Likewise.
25491 (vqtbx1q_p8): Likewise.
25492 (vtbx2_s8): Likewise.
25493 (vtbx2_u8): Likewise.
25494 (vtbx2_p8): Likewise.
25496 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
25498 * config/aarch64/aarch64-simd-builtins.def: Add tbl1 builtin
25500 * config/aarch64/arm_neon.h (vqtbl1_p8): Use RTL builtin
25501 instead of inline asm.
25502 (vqtbl1_s8): Likewise.
25503 (vqtbl1_u8): Likewise.
25504 (vqtbl1q_p8): Likewise.
25505 (vqtbl1q_s8): Likewise.
25506 (vqtbl1q_u8): Likewise.
25507 (vtbl1_s8): Likewise.
25508 (vtbl1_u8): Likewise.
25509 (vtbl1_p8): Likewise.
25510 (vtbl2_s8): Likewise.
25511 (vtbl2_u8): Likewise.
25512 (vtbl2_p8): Likewise.
25514 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
25516 * config/aarch64/aarch64-simd-builtins.def: Add polynomial
25517 ssri_n buitin generator macro.
25518 * config/aarch64/arm_neon.h (vsri_n_p8): Use RTL builtin
25519 instead of inline asm.
25520 (vsri_n_p16): Likewise.
25521 (vsri_n_p64): Likewise.
25522 (vsriq_n_p8): Likewise.
25523 (vsriq_n_p16): Likewise.
25524 (vsriq_n_p64): Likewise.
25526 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
25528 * config/aarch64/aarch64-simd-builtins.def: Use VALLP mode
25529 iterator for polynomial ssli_n builtin generator macro.
25530 * config/aarch64/arm_neon.h (vsli_n_p8): Use RTL builtin
25531 instead of inline asm.
25532 (vsli_n_p16): Likewise.
25533 (vsliq_n_p8): Likewise.
25534 (vsliq_n_p16): Likewise.
25535 * config/aarch64/iterators.md: Define VALLP mode iterator.
25537 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
25539 * config/aarch64/aarch64-simd-builtins.def: Use VDQV_L
25540 iterator to generate [su]adalp RTL builtins.
25541 * config/aarch64/aarch64-simd.md: Use VDQV_L iterator in
25542 [su]adalp RTL pattern.
25543 * config/aarch64/arm_neon.h (vpadal_s32): Use RTL builtin
25544 instead of inline asm.
25545 (vpadal_u32): Likewise.
25547 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
25549 * config/aarch64/aarch64-simd-builtins.def: Add [su]addlp
25550 builtin generator macros.
25551 * config/aarch64/aarch64-simd.md (aarch64_<su>addlp<mode>):
25553 * config/aarch64/arm_neon.h (vpaddl_s8): Use RTL builtin
25554 instead of inline asm.
25555 (vpaddl_s16): Likewise.
25556 (vpaddl_s32): Likewise.
25557 (vpaddl_u8): Likewise.
25558 (vpaddl_u16): Likewise.
25559 (vpaddl_u32): Likewise.
25560 (vpaddlq_s8): Likewise.
25561 (vpaddlq_s16): Likewise.
25562 (vpaddlq_s32): Likewise.
25563 (vpaddlq_u8): Likewise.
25564 (vpaddlq_u16): Likewise.
25565 (vpaddlq_u32): Liwewise.
25566 * config/aarch64/iterators.md: Define [SU]ADDLP unspecs with
25567 appropriate attributes.
25569 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
25571 * config/aarch64/aarch64-simd-builtins.def: Use VDQ_I iterator
25572 for aarch64_addp<mode> builtin macro generator.
25573 * config/aarch64/aarch64-simd.md: Use VDQ_I iterator in
25574 aarch64_addp<mode> RTL pattern.
25575 * config/aarch64/arm_neon.h (vpaddq_s8): Use RTL builtin
25576 instead of inline asm.
25577 (vpaddq_s16): Likewise.
25578 (vpaddq_s32): Likewise.
25579 (vpaddq_s64): Likewise.
25580 (vpaddq_u8): Likewise.
25581 (vpaddq_u16): Likewise.
25582 (vpaddq_u32): Likewise.
25583 (vpaddq_u64): Likewise.
25585 2021-04-28 Jonathan Wright <jonathan.wright@arm.com>
25587 * config/aarch64/aarch64-simd-builtins.def: Add sq[r]dmulh_n
25588 builtin generator macros.
25589 * config/aarch64/aarch64-simd.md (aarch64_sq<r>dmulh_n<mode>):
25591 * config/aarch64/arm_neon.h (vqdmulh_n_s16): Use RTL builtin
25592 instead of inline asm.
25593 (vqdmulh_n_s32): Likewise.
25594 (vqdmulhq_n_s16): Likewise.
25595 (vqdmulhq_n_s32): Likewise.
25596 (vqrdmulh_n_s16): Likewise.
25597 (vqrdmulh_n_s32): Likewise.
25598 (vqrdmulhq_n_s16): Likewise.
25599 (vqrdmulhq_n_s32): Likewise.
25601 2021-04-28 Tobias Burnus <tobias@codesourcery.com>
25603 * doc/install.texi (--enable-offload-defaulted): Document.
25605 2021-04-28 Senthil Kumar Selvaraj <saaadhu@gcc.gnu.org>
25607 * config/avr/avr-dimode.md: Turn existing patterns into
25608 define_insn_and_split style patterns where the splitter
25609 adds a clobber of the condition code register. Drop "cc"
25610 attribute. Add new patterns to match output of
25612 * config/avr/avr-fixed.md: Likewise.
25613 * config/avr/avr.c (cc_reg_rtx): New.
25614 (avr_parallel_insn_from_insns): Adjust insn count
25615 for removal of set of cc0.
25616 (avr_is_casesi_sequence): Likewise.
25617 (avr_casei_sequence_check_operands): Likewise.
25618 (avr_optimize_casesi): Likewise. Also insert
25619 new insns after jump_insn.
25620 (avr_pass_casesi::avr_rest_of_handle_casesi): Adjust
25621 for removal of set of cc0.
25622 (avr_init_expanders): Initialize cc_reg_rtx.
25623 (avr_regno_reg_class): Handle REG_CC.
25624 (cond_string): Remove usage of CC_OVERFLOW_UNUSABLE.
25625 (avr_notice_update_cc): Remove function.
25626 (ret_cond_branch): Remove usage of CC_OVERFLOW_UNUSABLE.
25627 (compare_condition): Adjust for PARALLEL with
25629 (out_shift_with_cnt): Likewise.
25630 (ashlhi3_out): Likewise.
25631 (ashrhi3_out): Likewise.
25632 (lshrhi3_out): Likewise.
25633 (avr_class_max_nregs): Return single reg for REG_CC.
25634 (avr_compare_pattern): Check for REG_CC instead
25636 (avr_reorg_remove_redundant_compare): Likewise.
25637 (avr_reorg):Adjust for PARALLEL with REG_CC clobber.
25638 (avr_hard_regno_nregs): Return single reg for REG_CC.
25639 (avr_hard_regno_mode_ok): Allow only CCmode for REG_CC.
25640 (avr_md_asm_adjust): Clobber REG_CC.
25641 (TARGET_HARD_REGNO_NREGS): Define.
25642 (TARGET_CLASS_MAX_NREGS): Define.
25643 (TARGET_MD_ASM_ADJUST): Define.
25644 * config/avr/avr.h (FIRST_PSEUDO_REGISTER): Adjust
25646 (enum reg_class): Add CC_REG class.
25647 (NOTICE_UPDATE_CC): Remove.
25648 (CC_OVERFLOW_UNUSABLE): Remove.
25649 (CC_NO_CARRY): Remove.
25650 * config/avr/avr.md: Turn existing patterns into
25651 define_insn_and_split style patterns where the splitter
25652 adds a clobber of the condition code register. Drop "cc"
25653 attribute. Add new patterns to match output of
25655 (sez): Remove unused pattern.
25657 2021-04-28 Richard Earnshaw <rearnsha@arm.com>
25660 * config/arm/arm.c (arm_hard_regno_mode_ok): Only allow VPR to be
25663 2021-04-28 Richard Sandiford <richard.sandiford@arm.com>
25666 * config/aarch64/constraints.md (Utq): Require the address to
25667 be valid for both the element mode and for V2DImode.
25669 2021-04-28 Jakub Jelinek <jakub@redhat.com>
25670 Tobias Burnus <tobias@codesourcery.com>
25672 * configure.ac (OFFLOAD_DEFAULTED): AC_DEFINE if offload-defaulted.
25673 * gcc.c (process_command): New variable.
25674 (driver::maybe_putenv_OFFLOAD_TARGETS): If OFFLOAD_DEFAULTED,
25675 set it if -foffload is defaulted.
25676 * lto-wrapper.c (OFFLOAD_TARGET_DEFAULT_ENV): Define.
25677 (compile_offload_image): If OFFLOAD_DEFAULTED and
25678 OFFLOAD_TARGET_DEFAULT is in the environment, don't fail
25679 if corresponding mkoffload can't be found.
25680 (compile_images_for_offload_targets): Likewise. Free and clear
25681 offload_names if no valid offload is found.
25682 * config.in: Regenerate.
25683 * configure: Regenerate.
25685 2021-04-28 Richard Biener <rguenther@suse.de>
25687 PR tree-optimization/100292
25688 * tree-vect-generic.c (expand_vector_condition): Do not fold
25691 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
25693 * config/rs6000/aix.h (SUBTARGET_DRIVER_SELF_SPECS): New.
25694 * config/rs6000/aix64.opt (m64): New.
25697 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
25699 * config/vax/vax.c (print_operand_address, vax_address_cost_1)
25700 (index_term_p): Handle ASHIFT too.
25702 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
25704 * config/vax/builtins.md (jbb<ccss>i<mode>): Remove operand #3.
25705 (sync_lock_test_and_set<mode>): Adjust accordingly.
25706 (sync_lock_release<mode>): Likewise.
25708 2021-04-27 Maciej W. Rozycki <macro@orcam.me.uk>
25710 * config/vax/vax-protos.h (adjacent_operands_p): Remove
25712 * config/vax/vax.c (adjacent_operands_p): Remove.
25714 2021-04-27 Maciej W. Rozycki <macro@linux-mips.org>
25716 * ifcvt.c (dead_or_predicable) [!IFCVT_MODIFY_TESTS]: Fall
25717 through to the non-conditional execution case if getting the
25718 condition for conditional execution has failed.
25720 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
25722 PR middle-end/100284
25723 * gimple.c (gimple_could_trap_p_1): Remove VEC_COND_EXPR test.
25724 * tree-eh.c (operation_could_trap_p): Handle VEC_COND_EXPR rather
25725 than asserting on it.
25727 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
25729 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): Protect
25730 with TARGET_AIX_OS.
25732 2021-04-27 David Edelsohn <dje.gcc@gmail.com>
25735 * calls.c (precompute_register_parameters): Additionally test
25736 targetm.precompute_tls_p to pre-compute argument.
25737 * config/rs6000/aix.h (TARGET_PRECOMPUTE_TLS_P): Define.
25738 * config/rs6000/rs6000.c (rs6000_aix_precompute_tls_p): New.
25739 * target.def (precompute_tls_p): New.
25740 * doc/tm.texi.in (TARGET_PRECOMPUTE_TLS_P): Add hook documentation.
25741 * doc/tm.texi: Regenerated.
25743 2021-04-27 Jakub Jelinek <jakub@redhat.com>
25746 * config/aarch64/aarch64.c (aarch64_print_operand): Cast -UINTVAL
25747 back to HOST_WIDE_INT.
25749 2021-04-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
25752 * simplify-rtx.c (simplify_context::simplify_subreg): Check the
25753 memory alignment for the outer mode.
25755 2021-04-27 H.J. Lu <hjl.tools@gmail.com>
25757 PR middle-end/90773
25758 * expr.c (op_by_pieces_d::get_usable_mode): New member function.
25759 (op_by_pieces_d::run): Cange a while loop to a do-while loop.
25761 2021-04-27 Alex Coplan <alex.coplan@arm.com>
25764 * config/arm/arm.c (arm_split_compare_and_swap): Fix up codegen
25765 with negative immediates: ensure we expand cbranchsi4_scratch
25766 correctly and ensure we satisfy its constraints.
25767 * config/arm/sync.md
25768 (@atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): Don't
25769 attempt to tie two output operands together with constraints;
25770 collapse two alternatives.
25771 (@atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): Likewise.
25772 * config/arm/thumb1.md (cbranchsi4_neg_late): New.
25774 2021-04-27 Jakub Jelinek <jakub@redhat.com>
25777 * config/aarch64/predicates.md (aarch64_sub_immediate,
25778 aarch64_plus_immediate): Use -UINTVAL instead of -INTVAL.
25779 * config/aarch64/aarch64.md (casesi, rotl<mode>3): Likewise.
25780 * config/aarch64/aarch64.c (aarch64_print_operand,
25781 aarch64_split_atomic_op, aarch64_expand_subvti): Likewise.
25783 2021-04-27 Jakub Jelinek <jakub@redhat.com>
25785 PR tree-optimization/100239
25786 * tree-vect-generic.c (lower_vec_perm): Don't accept constant
25787 permutations with all indices from the first zero element as vec_shl.
25789 2021-04-27 Jakub Jelinek <jakub@redhat.com>
25791 PR rtl-optimization/100254
25792 * cfgcleanup.c (outgoing_edges_match): Check REG_EH_REGION on
25793 last1 and last2 insns rather than BB_END (bb1) and BB_END (bb2) insns.
25795 2021-04-27 Richard Biener <rguenther@suse.de>
25797 PR tree-optimization/99912
25798 * passes.def: Add comment about new TODO_remove_unused_locals.
25799 * tree-stdarg.c (pass_data_stdarg): Run TODO_remove_unused_locals
25802 2021-04-27 Richard Biener <rguenther@suse.de>
25804 PR tree-optimization/99912
25805 * passes.def (pass_all_optimizations): Add pass_dse before
25806 the first pass_dce, move the first pass_dse before the
25807 pass_dce following pass_pre.
25809 2021-04-27 Jakub Jelinek <jakub@redhat.com>
25811 PR tree-optimization/95527
25812 * generic-match-head.c: Include tm.h.
25813 * gimple-match-head.c: Include tm.h.
25814 * match.pd (CLZ == INTEGER_CST): Don't use
25815 #ifdef CLZ_DEFINED_VALUE_AT_ZERO, only test CLZ_DEFINED_VALUE_AT_ZERO
25816 if clz == CFN_CLZ. Add missing val declaration.
25817 (CTZ cmp CST): New simplifications.
25819 2021-04-27 Jakub Jelinek <jakub@redhat.com>
25821 PR tree-optimization/96696
25822 * expr.c (expand_expr_divmod): New function.
25823 (expand_expr_real_2) <case TRUNC_DIV_EXPR>: Use it for truncations and
25824 divisions. Formatting fixes.
25825 <case MULT_EXPR>: Optimize x / y * y as x - x % y if the latter is
25828 2021-04-27 Martin Jambor <mjambor@suse.cz>
25831 * ipa-param-manipulation.c (ipa_param_adjustments::modify_call):
25832 If removing a call statement LHS SSA name, release it.
25834 2021-04-27 Richard Earnshaw <rearnsha@arm.com>
25837 * config/arm/arm.c (THUMB2_WORK_REGS): Check PIC_OFFSET_TABLE_REGNUM
25838 is valid before including it in the mask.
25840 2021-04-27 Richard Sandiford <richard.sandiford@arm.com>
25843 * config/aarch64/aarch64.c (aarch64_comp_type_attributes): Handle
25846 2021-04-27 Richard Biener <rguenther@suse.de>
25848 PR tree-optimization/100051
25849 * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Add
25850 disambiguator based on access size vs. decl size.
25852 2021-04-27 Richard Biener <rguenther@suse.de>
25854 PR tree-optimization/100278
25855 * tree-ssa-pre.c (compute_avail): Give up when we cannot
25856 adjust TBAA beacuse of mismatching bases.
25858 2021-04-27 Jakub Jelinek <jakub@redhat.com>
25861 * config/i386/i386.md (*<insn><mode>3_mask, *<insn><mode>3_mask_1):
25862 For any_rotate define_insn_split and following splitters, use
25863 SWI iterator instead of SWI48.
25865 2021-04-27 Richard Biener <rguenther@suse.de>
25867 PR tree-optimization/99776
25868 * match.pd (bit_field_ref (ctor)): Relax element extract
25869 type compatibility checks.
25871 2021-04-27 Cui,Lili <lili.cui@intel.com>
25873 * common/config/i386/i386-common.c (processor_names):
25874 Sync processor_names with processor_type.
25875 * config/i386/i386-options.c (processor_cost_table):
25876 Sync processor_cost_table with processor_type.
25878 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
25880 * value-range.cc (irange::irange_set_1bit_anti_range): Add assert.
25881 (irange::set): Call irange_set_1bit_anti_range for handling all
25882 1-bit ranges. Fall through on ~[MIN,MAX].
25884 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
25886 * value-range.cc (irange::legacy_num_pairs): Remove.
25887 (irange::invert): Change gcc_assert to gcc_checking_assert.
25888 * value-range.h (irange::num_pairs): Adjust for a cached
25889 num_pairs(). Also, rename all gcc_assert's to
25890 gcc_checking_assert's.
25892 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
25894 * value-range.cc (irange::operator=): Set m_kind.
25895 (irange::copy_to_legacy): Handle varying and undefined sources
25896 as a legacy copy since they can be easily copied.
25897 (irange::irange_set): Set m_kind.
25898 (irange::irange_set_anti_range): Same.
25899 (irange::set): Rename normalize_min_max to normalize_kind.
25900 (irange::verify_range): Adjust for multi-ranges having the
25902 (irange::irange_union): Set m_kind.
25903 (irange::irange_intersect): Same.
25904 (irange::invert): Same.
25905 * value-range.h (irange::kind): Always return m_kind.
25906 (irange::varying_p): Rename to...
25907 (irange::varying_comptaible_p): ...this.
25908 (irange::undefined_p): Only look at m_kind.
25909 (irange::irange): Always set VR_UNDEFINED if applicable.
25910 (irange::set_undefined): Always set VR_UNDEFINED.
25911 (irange::set_varying): Always set m_kind to VR_VARYING.
25912 (irange::normalize_min_max): Rename to...
25913 (irange::normalize_kind): ...this.
25915 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
25917 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
25918 Adjust for constant_p including varying_p.
25919 * tree-vrp.c (vrp_prop::finalize): Same.
25920 (determine_value_range): Same.
25921 * vr-values.c (vr_values::range_of_expr): Same.
25922 * value-range.cc (irange::symbolic_p): Do not check varying_p.
25923 (irange::constant_p): Same.
25925 2021-04-26 Aldy Hernandez <aldyh@redhat.com>
25927 * value-range.cc (irange::legacy_lower_bound): Replace
25928 !undefined_p check with num_ranges > 0.
25929 (irange::legacy_upper_bound): Same.
25930 * value-range.h (irange::type): Same.
25931 (irange::lower_bound): Same.
25932 (irange::upper_bound): Same.
25934 2021-04-26 Richard Biener <rguenther@suse.de>
25936 PR tree-optimization/99956
25937 * gimple-loop-interchange.cc (compute_access_stride):
25938 Try instantiating the access in a shallower loop nest
25939 if instantiating failed.
25940 (compute_access_strides): Pass adjustable loop_nest
25941 to compute_access_stride.
25943 2021-04-26 Christophe Lyon <christophe.lyon@linaro.org>
25945 * doc/sourcebuild.texi (arm_cmse_hw): Document.
25947 2021-04-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25949 * config/aarch64/iterators.md (vwcore): Handle V4BF, V8BF.
25951 2021-04-26 Thomas Schwinge <thomas@codesourcery.com>
25952 Nathan Sidwell <nathan@codesourcery.com>
25953 Tom de Vries <vries@codesourcery.com>
25954 Julian Brown <julian@codesourcery.com>
25955 Kwok Cheung Yeung <kcy@codesourcery.com>
25957 * omp-offload.c (oacc_validate_dims): Implement
25958 '-Wopenacc-parallelism'.
25959 * doc/invoke.texi (-Wopenacc-parallelism): Document.
25961 2021-04-26 Richard Biener <rguenther@suse.de>
25963 * tree-cfg.h (gimplify_build1): Remove.
25964 (gimplify_build2): Likewise.
25965 (gimplify_build3): Likewise.
25966 * tree-cfg.c (gimplify_build1): Move to tree-vect-generic.c.
25967 (gimplify_build2): Likewise.
25968 (gimplify_build3): Likewise.
25969 * tree-vect-generic.c (gimplify_build1): Move from tree-cfg.c.
25971 (gimplify_build2): Likewise.
25972 (gimplify_build3): Likewise.
25973 (tree_vec_extract): Use resimplify with following SSA edges.
25974 (expand_vector_parallel): Avoid passing NULL size/bitpos
25975 to tree_vec_extract.
25976 * expr.c (store_constructor): Deal with zero-element CTORs.
25977 * match.pd (bit_field_ref <vector CTOR>): Make sure to
25978 produce vector constants when possible.
25980 2021-04-26 Richard Biener <rguenther@suse.de>
25982 * tree-complex.c: Include gimple-fold.h.
25983 (expand_complex_addition): Use gimple_build.
25984 (expand_complex_multiplication_components): Likewise.
25985 (expand_complex_multiplication): Likewise.
25986 (expand_complex_div_straight): Likewise.
25987 (expand_complex_div_wide): Likewise.
25988 (expand_complex_division): Likewise.
25989 (expand_complex_conjugate): Likewise.
25990 (expand_complex_comparison): Likewise.
25992 2021-04-26 Richard Biener <rguenther@suse.de>
25994 * tree-ssa-phiopt.c (two_value_replacement): Remove use
25995 of legacy gimplify_buildN API.
25997 2021-04-26 Richard Biener <rguenther@suse.de>
25999 PR tree-optimization/99473
26000 * tree-ssa-phiopt.c (cond_store_replacement): Handle all
26003 2021-04-26 Richard Biener <rguenther@suse.de>
26005 * config/rs6000/rs6000-call.c (rs6000_gimple_fold_builtin):
26006 Use replace_call_with_value.
26008 2021-04-26 Richard Biener <rguenther@suse.de>
26010 * tree-ssa-propagate.h (valid_gimple_rhs_p): Remove.
26011 (update_gimple_call): Likewise.
26012 (update_call_from_tree): Likewise.
26013 * tree-ssa-propagate.c (valid_gimple_rhs_p): Remove.
26014 (valid_gimple_call_p): Likewise.
26015 (move_ssa_defining_stmt_for_defs): Likewise.
26016 (finish_update_gimple_call): Likewise.
26017 (update_gimple_call): Likewise.
26018 (update_call_from_tree): Likewise.
26019 (propagate_tree_value_into_stmt): Use replace_call_with_value.
26020 * gimple-fold.h (update_gimple_call): Declare.
26021 * gimple-fold.c (valid_gimple_rhs_p): Move here from
26022 tree-ssa-propagate.c.
26023 (update_gimple_call): Likewise.
26024 (valid_gimple_call_p): Likewise.
26025 (finish_update_gimple_call): Likewise, and simplify.
26026 (gimplify_and_update_call_from_tree): Implement
26027 update_call_from_tree functionality, avoid excessive
26028 push/pop_gimplify_context.
26029 (gimple_fold_builtin): Use only gimplify_and_update_call_from_tree.
26030 (gimple_fold_call): Likewise.
26031 * gimple-ssa-sprintf.c (try_substitute_return_value): Likewise.
26032 * tree-ssa-ccp.c (ccp_folder::fold_stmt): Likewise.
26033 (pass_fold_builtins::execute): Likewise.
26034 (optimize_stack_restore): Use replace_call_with_value.
26035 * tree-cfg.c (fold_loop_internal_call): Likewise.
26036 * tree-ssa-dce.c (maybe_optimize_arith_overflow): Use
26037 only gimplify_and_update_call_from_tree.
26038 * tree-ssa-strlen.c (handle_builtin_strlen): Likewise.
26039 (handle_builtin_strchr): Likewise.
26040 * tsan.c: Include gimple-fold.h instead of tree-ssa-propagate.h.
26042 2021-04-26 Jakub Jelinek <jakub@redhat.com>
26045 * vmsdbgout.c (ASM_OUTPUT_DEBUG_STRING, vmsdbgout_begin_block,
26046 vmsdbgout_end_block, lookup_filename, vmsdbgout_source_line): Remove
26049 2021-04-25 liuhongt <hongtao.liu@intel.com>
26052 * config/i386/i386-builtin.def (BDESC): Change the icode of
26053 the following builtins to CODE_FOR_nothing.
26054 * config/i386/i386.c (ix86_gimple_fold_builtin): Fold
26055 IX86_BUILTIN_PCMPEQB128, IX86_BUILTIN_PCMPEQW128,
26056 IX86_BUILTIN_PCMPEQD128, IX86_BUILTIN_PCMPEQQ,
26057 IX86_BUILTIN_PCMPEQB256, IX86_BUILTIN_PCMPEQW256,
26058 IX86_BUILTIN_PCMPEQD256, IX86_BUILTIN_PCMPEQQ256,
26059 IX86_BUILTIN_PCMPGTB128, IX86_BUILTIN_PCMPGTW128,
26060 IX86_BUILTIN_PCMPGTD128, IX86_BUILTIN_PCMPGTQ,
26061 IX86_BUILTIN_PCMPGTB256, IX86_BUILTIN_PCMPGTW256,
26062 IX86_BUILTIN_PCMPGTD256, IX86_BUILTIN_PCMPGTQ256.
26063 * config/i386/sse.md (avx2_eq<mode>3): Deleted.
26064 (sse2_eq<mode>3): Ditto.
26065 (sse4_1_eqv2di3): Ditto.
26066 (sse2_gt<mode>3): Rename to ..
26067 (*sse2_gt<mode>3): .. this.
26069 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
26072 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
26075 * config/darwin.c (darwin_binds_local_p): Assume that any
26076 public symbol might be interposed for PIC code. Update function
26077 header comment to reflect current Darwin capability.
26079 2021-04-24 Iain Sandoe <iain@sandoe.co.uk>
26082 * config/darwin.c (darwin_binds_local_p): Assume that any
26083 public symbol might be interposed for PIC code. Update function
26084 header comment to reflect current Darwin capability.
26086 2021-04-24 Richard Sandiford <richard.sandiford@arm.com>
26088 * doc/sourcebuild.texi: Document no-opts and any-opts target
26091 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
26093 * config/bpf/bpf.h (ASM_OUTPUT_ALIGNED_BSS): Use .type and .lcomm.
26095 2021-04-23 YiFei Zhu <zhuyifei1999@gmail.com>
26097 * config/bpf/bpf.h (FUNCTION_BOUNDARY): Set to 64.
26099 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
26102 * config/i386/i386-options.c (ix86_option_override_internal):
26103 Error out when -m96bit-long-double is used with 64bit targets.
26104 * config/i386/i386.md (*pushxf_rounded): Remove pattern.
26106 2021-04-23 Martin Liska <mliska@suse.cz>
26108 * lto-wrapper.c: Remove FIXME about usage of
26109 hardware_concurrency. The function is not on par with
26112 2021-04-23 Uroš Bizjak <ubizjak@gmail.com>
26115 * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
26116 Copy operand 3 to operand 4. Use sse_reg_operand
26117 as operand 3 predicate.
26118 (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): Ditto.
26119 (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
26120 (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
26121 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2):
26122 Copy operand 1 to operand 0.
26123 (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): Ditto.
26124 (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
26125 (LDX_ATOMIC/LDX_ATOMIC FP store peephole2 with mem blockage): Ditto.
26127 2021-04-23 Alex Coplan <alex.coplan@arm.com>
26129 PR rtl-optimization/100230
26130 * early-remat.c (early_remat::sort_candidates): Use delete[]
26131 instead of delete for array allocated with new[].
26133 2021-04-23 Richard Biener <rguenther@suse.de>
26135 * genmatch.c (lower_cond): Remove VEC_COND_EXPR special-casing.
26136 (capture_info::capture_info): Likewise.
26137 (capture_info::walk_match): Likewise.
26138 (expr::gen_transform): Likewise.
26139 (dt_simplify::gen_1): Likewise.
26140 * gimple-match-head.c (maybe_resimplify_conditional_op):
26141 Remove VEC_COND_EXPR special-casing.
26142 (gimple_simplify): Likewise.
26143 * gimple.c (gimple_could_trap_p_1): Adjust.
26144 * tree-ssa-pre.c (compute_avail): Allow VEC_COND_EXPR
26145 to participate in PRE.
26147 2021-04-23 Richard Biener <rguenther@suse.de>
26149 * cfganal.c (connect_infinite_loops_to_exit): First call
26150 add_noreturn_fake_exit_edges.
26151 * ipa-sra.c (process_scan_results): Do not call the now redundant
26152 add_noreturn_fake_exit_edges.
26153 * predict.c (tree_estimate_probability): Likewise.
26154 (rebuild_frequencies): Likewise.
26155 * store-motion.c (one_store_motion_pass): Likewise.
26157 2021-04-23 Richard Biener <rguenther@suse.de>
26159 PR tree-optimization/100222
26160 * predict.c (pass_profile::execute): Remove redundant call to
26161 mark_irreducible_loops.
26162 (report_predictor_hitrates): Likewise.
26164 2021-04-23 Richard Biener <rguenther@suse.de>
26166 * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Avoid
26167 valid_gimple_rhs_p by instead gimplifying to one.
26169 2021-04-23 Richard Biener <rguenther@suse.de>
26171 PR tree-optimization/99971
26172 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
26173 Always use TBAA for loads.
26175 2021-04-23 liuhongt <hongtao.liu@intel.com>
26178 * config/i386/i386-options.c (ix86_option_override_internal):
26179 Clear MASK_AVX256_SPLIT_UNALIGNED_LOAD/STORE in x_target_flags
26180 when X86_TUNE_AVX256_UNALIGNED_LOAD/STORE_OPTIMAL is enabled
26181 by target attribute.
26183 2021-04-23 David Edelsohn <dje.gcc@gmail.com>
26185 * config/rs6000/aix71.h (PREFERRED_DEBUGGING_TYPE): Change to
26187 * config/rs6000/aix72.h (PREFERRED_DEBUGGING_TYPE): Same.
26189 2021-04-22 David Edelsohn <dje.gcc@gmail.com>
26191 * config.gcc (powerpc-ibm-aix6.*): Remove.
26192 * config/rs6000/aix61.h: Delete.
26194 2021-04-22 Martin Liska <mliska@suse.cz>
26196 PR testsuite/100159
26197 PR testsuite/100192
26198 * builtins.c (expand_builtin): Fix typos and missing comments.
26199 * dwarf2out.c (gen_subprogram_die): Likewise.
26200 (gen_struct_or_union_type_die): Likewise.
26202 2021-04-22 Uroš Bizjak <ubizjak@gmail.com>
26205 * config/i386/i386-expand.c (ix86_expand_convert_uns_sidf_sse):
26206 Remove the sign with FE_DOWNWARD, where x - x = -0.0.
26208 2021-04-21 Iain Sandoe <iain@sandoe.co.uk>
26210 * config/i386/darwin.h (TARGET_64BIT): Remove definition
26211 based on TARGET_ISA_64BIT.
26212 (TARGET_64BIT_P): Remove definition based on
26213 TARGET_ISA_64BIT_P().
26215 2021-04-21 Martin Liska <mliska@suse.cz>
26218 2021-04-21 Martin Liska <mliska@suse.cz>
26220 * lto-wrapper.c (cpuset_popcount): Remove.
26221 (init_num_threads): Remove and use hardware_concurrency.
26223 2021-04-21 Martin Liska <mliska@suse.cz>
26226 * main.c (main): Call toplev::finalize in CHECKING_P mode.
26227 * ipa-modref.c (ipa_modref_c_finalize): summaries are NULL
26228 when incremental LTO linking happens.
26230 2021-04-21 Martin Liska <mliska@suse.cz>
26232 * lto-wrapper.c (run_gcc): When -flto=jobserver is used, but the
26233 makeserver cannot be detected, then use -flto=N fallback.
26235 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
26237 * acinclude.m4 (gcc_AC_INITFINI_ARRAY): When cross-compiling,
26238 default to yes for aarch64-linux-gnu.
26239 * configure: Regenerate.
26241 2021-04-21 Martin Liska <mliska@suse.cz>
26243 * lto-wrapper.c (cpuset_popcount): Remove.
26244 (init_num_threads): Remove and use hardware_concurrency.
26246 2021-04-21 Martin Liska <mliska@suse.cz>
26248 * config/i386/i386.c: Remove superfluous || TARGET_MACHO
26249 which remains to be '(... || 0)' and clang complains about it.
26250 * dwarf2out.c (AT_vms_delta): Declare conditionally.
26251 (add_AT_vms_delta): Likewise.
26252 * tree.c (fld_simplified_type): Use rather more common pattern
26253 for disabling of something (#if 0).
26254 (get_tree_code_name): Likewise.
26255 (verify_type_variant): Likewise.
26257 2021-04-21 Martin Liska <mliska@suse.cz>
26259 * config/i386/i386-expand.c (decide_alignment): Use newly named
26260 macro TARGET_CPU_P.
26261 * config/i386/i386.c (ix86_decompose_address): Likewise.
26262 (ix86_address_cost): Likewise.
26263 (ix86_lea_outperforms): Likewise.
26264 (ix86_avoid_lea_for_addr): Likewise.
26265 (ix86_add_stmt_cost): Likewise.
26266 * config/i386/i386.h (TARGET_*): Remove.
26267 (TARGET_CPU_P): New macro.
26268 * config/i386/i386.md: Use newly named macro TARGET_CPU_P.
26269 * config/i386/x86-tune-sched-atom.c (do_reorder_for_imul): Likewise.
26270 (swap_top_of_ready_list): Likewise.
26271 (ix86_atom_sched_reorder): Likewise.
26272 * config/i386/x86-tune-sched-bd.c (ix86_bd_has_dispatch): Likewise.
26273 * config/i386/x86-tune-sched.c (ix86_adjust_cost): Likewise.
26275 2021-04-21 Martin Liska <mliska@suse.cz>
26277 * config/i386/i386-options.c (TARGET_EXPLICIT_NO_SAHF_P):
26279 (SET_TARGET_NO_SAHF): Likewise.
26280 (TARGET_EXPLICIT_PREFETCH_SSE_P): Likewise.
26281 (SET_TARGET_PREFETCH_SSE): Likewise.
26282 (TARGET_EXPLICIT_NO_TUNE_P): Likewise.
26283 (SET_TARGET_NO_TUNE): Likewise.
26284 (TARGET_EXPLICIT_NO_80387_P): Likewise.
26285 (SET_TARGET_NO_80387): Likewise.
26287 * config/i386/i386.h (TARGET_*): Remove.
26288 * opth-gen.awk: Generate new used macros.
26290 2021-04-21 Martin Liska <mliska@suse.cz>
26292 * config/i386/i386.h (PTA_*): Remove.
26293 (enum pta_flag): New.
26294 (DEF_PTA): Generate PTA_* values from i386-isa.def.
26295 * config/i386/i386-isa.def: New file.
26297 2021-04-21 Alex Coplan <alex.coplan@arm.com>
26300 * config/aarch64/aarch64-bti-insert.c (aarch64_bti_j_insn_p): New.
26301 (rest_of_insert_bti): Avoid inserting duplicate bti j insns for
26302 jump table targets.
26304 2021-04-21 H.J. Lu <hjl.tools@gmail.com>
26306 * config.gcc: Install mwaitintrin.h for i[34567]86-*-* and
26307 x86_64-*-* targets.
26308 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_MWAIT_SET):
26310 (OPTION_MASK_ISA2_MWAIT_UNSET): Likewise.
26311 (ix86_handle_option): Handle -mmwait.
26312 * config/i386/i386-builtins.c (ix86_init_mmx_sse_builtins):
26313 Replace OPTION_MASK_ISA_SSE3 with OPTION_MASK_ISA2_MWAIT on
26314 __builtin_ia32_monitor and __builtin_ia32_mwait.
26315 * config/i386/i386-options.c (isa2_opts): Add -mmwait.
26316 (ix86_valid_target_attribute_inner_p): Likewise.
26317 (ix86_option_override_internal): Enable mwait/monitor
26318 instructions for -msse3.
26319 * config/i386/i386.h (TARGET_MWAIT): New.
26320 (TARGET_MWAIT_P): Likewise.
26321 * config/i386/i386.opt: Add -mmwait.
26322 * config/i386/mwaitintrin.h: New file.
26323 * config/i386/pmmintrin.h: Include <mwaitintrin.h>.
26324 * config/i386/sse.md (sse3_mwait): Replace TARGET_SSE3 with
26326 (@sse3_monitor_<mode>): Likewise.
26327 * config/i386/x86gprintrin.h: Include <mwaitintrin.h>.
26328 * doc/extend.texi: Document mwait target attribute.
26329 * doc/invoke.texi: Document -mmwait.
26331 2021-04-21 Martin Liska <mliska@suse.cz>
26333 * config/i386/i386-options.c (DEF_ENUM): Remove it.
26334 * config/i386/i386-opts.h (DEF_ENUM): Likewise.
26335 * config/i386/stringop.def (DEF_ENUM): Likewise.
26337 2021-04-21 Martin Liska <mliska@suse.cz>
26339 * tree-cfg.c (gimple_verify_flow_info): Use qD instead
26340 of print_generic_expr.
26342 2021-04-21 Jakub Jelinek <jakub@redhat.com>
26344 PR rtl-optimization/100148
26345 * cprop.c (constprop_register): Use next_nondebug_insn instead of
26348 2021-04-21 Martin Liska <mliska@suse.cz>
26351 * cgraphunit.c (cgraph_node::analyze): Remove duplicate
26352 free_dominance_info calls.
26354 2021-04-21 Richard Biener <rguenther@suse.de>
26356 * gimple-fold.c (maybe_fold_reference): Remove is_lhs
26357 parameter (and assume it to be false).
26358 (fold_gimple_assign): Adjust, remove all callers of
26359 maybe_fold_reference calling it with is_lhs true.
26360 (gimple_fold_call): Likewise.
26361 (fold_stmt_1): Likewise.
26363 2021-04-21 Richard Biener <rguenther@suse.de>
26365 * fold-const.c (pedantic_non_lvalue_loc): Remove.
26366 (fold_binary_loc): Adjust.
26367 (fold_ternary_loc): Likewise.
26369 2021-04-21 Richard Sandiford <richard.sandiford@arm.com>
26371 PR middle-end/100130
26372 * varasm.c (get_block_for_decl): Make sure that any use of the
26373 retain attribute matches the section's retain flag.
26374 (switch_to_section): Check for retain mismatches even when
26375 changing sections, but do not warn if the given decl is the
26376 section's named.decl.
26377 (output_object_block): Pass the first decl in the block (if any)
26378 to switch_to_section.
26380 2021-04-20 H.J. Lu <hjl.tools@gmail.com>
26382 * config/i386/i386-c.c (ix86_target_macros_internal): Define
26383 __CRC32__ for -mcrc32.
26384 * config/i386/i386-options.c (ix86_option_override_internal):
26385 Enable crc32 instruction for -msse4.2.
26386 * config/i386/i386.md (sse4_2_crc32<mode>): Remove TARGET_SSE4_2
26388 (sse4_2_crc32di): Likewise.
26389 * config/i386/ia32intrin.h: Use crc32 target option for CRC32
26392 2021-04-20 Segher Boessenkool <segher@kernel.crashing.org>
26395 * config/rs6000/rs6000.c (rs6000_machine_from_flags): Do not consider
26398 2021-04-20 Martin Liska <mliska@suse.cz>
26400 * doc/invoke.texi: Fix typo.
26401 * params.opt: Likewise.
26403 2021-04-20 Martin Liska <mliska@suse.cz>
26405 * doc/invoke.texi: Document new param.
26407 2021-04-19 Andrew MacLeod <amacleod@redhat.com>
26409 PR tree-optimization/100081
26410 * gimple-range-cache.h (ranger_cache): Inherit from gori_compute
26411 rather than gori_compute_cache.
26412 * gimple-range-gori.cc (is_gimple_logical_p): Move to top of file.
26413 (range_def_chain::m_logical_depth): New member.
26414 (range_def_chain::range_def_chain): Initialize m_logical_depth.
26415 (range_def_chain::get_def_chain): Don't build defchains through more
26416 than LOGICAL_LIMIT logical expressions.
26417 * params.opt (param_ranger_logical_depth): New.
26419 2021-04-19 Richard Earnshaw <rearnsha@arm.com>
26422 * config/arm/arm.c (arm_configure_build_target): Do not strip
26423 extended FPU/SIMD feature bits from the target ISA when -mfpu
26424 is specified (partial revert of r11-8168).
26426 2021-04-19 Thomas Schwinge <thomas@codesourcery.com>
26428 * params.opt (-param=openacc-kernels=): Add.
26429 * omp-oacc-kernels-decompose.cc
26430 (pass_omp_oacc_kernels_decompose::gate): Use it.
26431 * doc/invoke.texi (-fopenacc-kernels=@var{mode}): Move...
26432 (--param): ... here, 'openacc-kernels'.
26434 2021-04-19 Martin Liska <mliska@suse.cz>
26437 * gengtype.c (finish_root_table): Align function arguments
26438 in between declaration and definition.
26440 2021-04-19 Eric Botcazou <ebotcazou@adacore.com>
26442 * config/i386/winnt.c (i386_pe_seh_cold_init): Properly deal with
26443 frames larger than the SEH maximum frame size.
26445 2021-04-18 Segher Boessenkool <segher@kernel.crashing.org>
26447 PR rtl-optimization/99927
26448 * combine.c (distribute_notes) [REG_UNUSED]: If the register already
26449 is dead, just drop it.
26451 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
26454 * config/i386/winnt-d.c (TARGET_D_TEMPLATES_ALWAYS_COMDAT): Define.
26455 * doc/tm.texi: Regenerate.
26456 * doc/tm.texi.in (D language and ABI): Add @hook for
26457 TARGET_D_TEMPLATES_ALWAYS_COMDAT.
26459 2021-04-17 Iain Buclaw <ibuclaw@gdcproject.org>
26461 * config/darwin-d.c (darwin_d_handle_target_object_format): New
26463 (darwin_d_register_target_info): New function.
26464 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
26465 * config/dragonfly-d.c (dragonfly_d_handle_target_object_format): New
26467 (dragonfly_d_register_target_info): New function.
26468 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
26469 * config/freebsd-d.c (freebsd_d_handle_target_object_format): New
26471 (freebsd_d_register_target_info): New function.
26472 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
26473 * config/glibc-d.c (glibc_d_handle_target_object_format): New
26475 (glibc_d_register_target_info): New function.
26476 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
26477 * config/i386/i386-d.c (ix86_d_handle_target_object_format): New
26479 (ix86_d_register_target_info): Add ix86_d_handle_target_object_format
26480 as handler for objectFormat key.
26481 * config/i386/winnt-d.c (winnt_d_handle_target_object_format): New
26483 (winnt_d_register_target_info): New function.
26484 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
26485 * config/netbsd-d.c (netbsd_d_handle_target_object_format): New
26487 (netbsd_d_register_target_info): New function.
26488 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
26489 * config/openbsd-d.c (openbsd_d_handle_target_object_format): New
26491 (openbsd_d_register_target_info): New function.
26492 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
26493 * config/pa/pa-d.c (pa_d_handle_target_object_format): New function.
26494 (pa_d_register_target_info): Add pa_d_handle_target_object_format as
26495 handler for objectFormat key.
26496 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_object_format): New
26498 (rs6000_d_register_target_info): Add
26499 rs6000_d_handle_target_object_format as handler for objectFormat key.
26500 * config/sol2-d.c (solaris_d_handle_target_object_format): New
26502 (solaris_d_register_target_info): New function.
26503 (TARGET_D_REGISTER_OS_TARGET_INFO): Define.
26505 2021-04-16 Jakub Jelinek <jakub@redhat.com>
26508 * config/aarch64/aarch64.c (aarch64_function_arg_alignment): Change
26509 abi_break argument from bool * to unsigned *, store there the pre-GCC 9
26511 (aarch64_layout_arg, aarch64_gimplify_va_arg_expr): Adjust callers.
26512 (aarch64_function_arg_regno_p): Likewise. Only emit -Wpsabi note if
26513 the old and new alignment after applying MIN/MAX to it is different.
26515 2021-04-16 Tamar Christina <tamar.christina@arm.com>
26518 * config/aarch64/aarch64-sve.md (@aarch64_sve_trn1_conv<mode>): New.
26519 * config/aarch64/aarch64.c (aarch64_expand_sve_const_pred_trn): Use new
26521 * config/aarch64/iterators.md (UNSPEC_TRN1_CONV): New.
26523 2021-04-16 Bill Schmidt <wschmidt@linux.ibm.com>
26525 * doc/extend.texi (PowerPC AltiVec/VSX Built-in Functions): Revise
26526 this section and its subsections.
26528 2021-04-16 Jakub Jelinek <jakub@redhat.com>
26531 * config/aarch64/aarch64.md (*neg_asr_si2_extr, *extrsi5_insn_di): New
26532 define_insn patterns.
26534 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
26536 PR rtl-optimization/98689
26537 * reg-notes.def (UNTYPED_CALL): New note.
26538 * combine.c (distribute_notes): Handle it.
26539 * emit-rtl.c (try_split): Likewise.
26540 * rtlanal.c (rtx_properties::try_to_add_insn): Likewise. Assume
26541 that calls with the note implicitly set all return value registers.
26542 * builtins.c (expand_builtin_apply): Add a REG_UNTYPED_CALL
26545 2021-04-16 Richard Sandiford <richard.sandiford@arm.com>
26547 PR rtl-optimization/99596
26548 * rtlanal.c (rtx_properties::try_to_add_insn): Don't add global
26549 register accesses for const calls. Assume that pure functions
26550 can only read from global registers. Ignore cases in which
26551 the stack pointer has been marked global.
26553 2021-04-16 Jakub Jelinek <jakub@redhat.com>
26556 * tree-vect-loop.c (vect_transform_loop): Don't remove just
26557 dead scalar .MASK_LOAD calls, but also dead .COND_* calls - replace
26558 them by their last argument.
26560 2021-04-15 Martin Liska <mliska@suse.cz>
26562 * doc/invoke.texi: Other params don't use it, remove it.
26564 2021-04-15 Richard Biener <rguenther@suse.de>
26566 * gimple-builder.h: Add deprecation note.
26568 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
26571 * attribs.h (restrict_type_identity_attributes_to): Declare.
26572 * attribs.c (restrict_type_identity_attributes_to): New function.
26574 2021-04-15 Richard Sandiford <richard.sandiford@arm.com>
26577 * attribs.h (affects_type_identity_attributes): Declare.
26578 * attribs.c (remove_attributes_matching): New function.
26579 (affects_type_identity_attributes): Likewise.
26581 2021-04-15 Jakub Jelinek <jakub@redhat.com>
26584 * config/aarch64/aarch64.md (*<LOGICAL:optab>_<SHIFT:optab><mode>3):
26585 Add combine splitters for *<LOGICAL:optab>_ashl<mode>3 with
26586 ZERO_EXTEND, SIGN_EXTEND or AND.
26588 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
26590 PR rtl-optimization/99929
26591 * rtl.h (same_vector_encodings_p): New function.
26592 * cse.c (exp_equiv_p): Check that CONST_VECTORs have the same encoding.
26593 * cselib.c (rtx_equal_for_cselib_1): Likewise.
26594 * jump.c (rtx_renumbered_equal_p): Likewise.
26595 * lra-constraints.c (operands_match_p): Likewise.
26596 * reload.c (operands_match_p): Likewise.
26597 * rtl.c (rtx_equal_p_cb, rtx_equal_p): Likewise.
26599 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
26601 * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Print
26602 more information about variable-length CONST_VECTORs.
26604 2021-04-14 Vladimir N. Makarov <vmakarov@redhat.com>
26606 PR rtl-optimization/100066
26607 * lra-constraints.c (split_reg): Check paradoxical_subreg_p for
26608 ordered modes when choosing splitting mode for hard reg.
26610 2021-04-14 Richard Sandiford <richard.sandiford@arm.com>
26613 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector_sel):
26615 (aarch64_expand_sve_const_vector): Use it for nelts_per_pattern==2.
26617 2021-04-14 Andreas Krebbel <krebbel@linux.ibm.com>
26619 * config/s390/s390-builtins.def (O_M5, O_M12, ...): Add new macros
26620 for mask operand types.
26621 (s390_vec_permi_s64, s390_vec_permi_b64, s390_vec_permi_u64)
26622 (s390_vec_permi_dbl, s390_vpdi): Use the M5 type for the immediate
26624 (s390_vec_msum_u128, s390_vmslg): Use the M12 type for the
26626 * config/s390/s390.c (s390_const_operand_ok): Check the new
26627 operand types and generate a list of valid values.
26629 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
26631 * doc/tm.texi: Regenerate.
26632 * doc/tm.texi.in (D language and ABI): Add @hook for
26633 TARGET_D_REGISTER_OS_TARGET_INFO.
26635 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
26637 * config/aarch64/aarch64-d.c (aarch64_d_handle_target_float_abi): New
26639 (aarch64_d_register_target_info): New function.
26640 * config/aarch64/aarch64-protos.h (aarch64_d_register_target_info):
26642 * config/aarch64/aarch64.h (TARGET_D_REGISTER_CPU_TARGET_INFO):
26644 * config/arm/arm-d.c (arm_d_handle_target_float_abi): New function.
26645 (arm_d_register_target_info): New function.
26646 * config/arm/arm-protos.h (arm_d_register_target_info): Declare.
26647 * config/arm/arm.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
26648 * config/i386/i386-d.c (ix86_d_handle_target_float_abi): New function.
26649 (ix86_d_register_target_info): New function.
26650 * config/i386/i386-protos.h (ix86_d_register_target_info): Declare.
26651 * config/i386/i386.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
26652 * config/mips/mips-d.c (mips_d_handle_target_float_abi): New function.
26653 (mips_d_register_target_info): New function.
26654 * config/mips/mips-protos.h (mips_d_register_target_info): Declare.
26655 * config/mips/mips.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
26656 * config/pa/pa-d.c (pa_d_handle_target_float_abi): New function.
26657 (pa_d_register_target_info): New function.
26658 * config/pa/pa-protos.h (pa_d_register_target_info): Declare.
26659 * config/pa/pa.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
26660 * config/riscv/riscv-d.c (riscv_d_handle_target_float_abi): New
26662 (riscv_d_register_target_info): New function.
26663 * config/riscv/riscv-protos.h (riscv_d_register_target_info): Declare.
26664 * config/riscv/riscv.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
26665 * config/rs6000/rs6000-d.c (rs6000_d_handle_target_float_abi): New
26667 (rs6000_d_register_target_info): New function.
26668 * config/rs6000/rs6000-protos.h (rs6000_d_register_target_info):
26670 * config/rs6000/rs6000.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
26671 * config/s390/s390-d.c (s390_d_handle_target_float_abi): New function.
26672 (s390_d_register_target_info): New function.
26673 * config/s390/s390-protos.h (s390_d_register_target_info): Declare.
26674 * config/s390/s390.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
26675 * config/sparc/sparc-d.c (sparc_d_handle_target_float_abi): New
26677 (sparc_d_register_target_info): New function.
26678 * config/sparc/sparc-protos.h (sparc_d_register_target_info): Declare.
26679 * config/sparc/sparc.h (TARGET_D_REGISTER_CPU_TARGET_INFO): Define.
26680 * doc/tm.texi: Regenerate.
26681 * doc/tm.texi.in (D language and ABI): Add @hook for
26682 TARGET_D_REGISTER_CPU_TARGET_INFO.
26684 2021-04-14 Iain Buclaw <ibuclaw@gdcproject.org>
26686 * config/i386/i386-d.c (ix86_d_has_stdcall_convention): New function.
26687 * config/i386/i386-protos.h (ix86_d_has_stdcall_convention): Declare.
26688 * config/i386/i386.h (TARGET_D_HAS_STDCALL_CONVENTION): Define.
26689 * doc/tm.texi: Regenerate.
26690 * doc/tm.texi.in (D language and ABI): Add @hook for
26691 TARGET_D_HAS_STDCALL_CONVENTION.
26693 2021-04-14 Richard Biener <rguenther@suse.de>
26695 * tree-cfg.c (verify_gimple_assign_ternary): Verify that
26696 VEC_COND_EXPRs have a gimple_val condition.
26697 * tree-ssa-propagate.c (valid_gimple_rhs_p): VEC_COND_EXPR
26698 can no longer have a GENERIC condition.
26700 2021-04-14 Richard Earnshaw <rearnsha@arm.com>
26703 * config/arm/arm.c (arm_configure_build_target): Strip isa_all_fpbits
26704 from the isa_delta when -mfpu has been used.
26705 (arm_options_perform_arch_sanity_checks): It's the architecture that
26706 lacks an FPU not the processor.
26708 2021-04-13 Richard Biener <rguenther@suse.de>
26710 PR tree-optimization/100053
26711 * tree-ssa-sccvn.c (vn_nary_op_get_predicated_value): Do
26712 not use optimistic dominance queries for backedges to validate
26714 (dominated_by_p_w_unex): Add parameter to ignore executable
26715 state on backedges.
26716 (rpo_elim::eliminate_avail): Adjust.
26718 2021-04-13 Jakub Jelinek <jakub@redhat.com>
26721 * config/aarch64/aarch64.md (*aarch64_bfxil<mode>_extr,
26722 *aarch64_bfxilsi_extrdi): New define_insn patterns.
26724 2021-04-13 Jakub Jelinek <jakub@redhat.com>
26727 * simplify-rtx.c (simplify_immed_subreg): For MODE_COMPOSITE_P
26728 outermode, return NULL if the result doesn't encode back to the
26729 original byte sequence.
26730 (simplify_gen_subreg): Don't create SUBREGs from constants to
26731 MODE_COMPOSITE_P outermode.
26733 2021-04-12 Jakub Jelinek <jakub@redhat.com>
26735 PR rtl-optimization/99905
26736 * combine.c (expand_compound_operation): If pos + len > modewidth,
26737 perform the right shift by pos in inner_mode and then convert to mode,
26738 instead of trying to simplify a shift of rtx with inner_mode by pos
26739 as if it was a shift in mode.
26741 2021-04-12 Jakub Jelinek <jakub@redhat.com>
26744 * combine.c (simplify_and_const_int_1): Don't optimize varop
26745 away if it has side-effects.
26747 2021-04-12 Martin Liska <mliska@suse.cz>
26749 * doc/extend.texi: Escape @smallexample content.
26751 2021-04-12 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
26753 * config/s390/s390.md ("*movdi_31", "*movdi_64"): Add
26754 alternative in order to load a DFP zero.
26756 2021-04-12 Martin Liska <mliska@suse.cz>
26758 * doc/extend.texi: Be more precise in documentation
26759 of symver attribute.
26761 2021-04-12 Martin Liska <mliska@suse.cz>
26764 * gimplify.c (gimplify_expr): Right now, we unpoison all
26765 variables before a goto <dest>. We should not do it if we are
26768 2021-04-12 Cui,Lili <lili.cui@intel.com>
26770 * common/config/i386/cpuinfo.h (get_intel_cpu): Handle
26772 * common/config/i386/i386-common.c (processor_names): Add
26774 (processor_alias_table): Add rocketlake.
26775 * common/config/i386/i386-cpuinfo.h (processor_subtypes): Add
26776 INTEL_COREI7_ROCKETLAKE.
26777 * config.gcc: Add -march=rocketlake.
26778 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
26780 * config/i386/i386-options.c (m_ROCKETLAKE) : Define.
26781 (processor_cost_table): Add rocketlake cost.
26782 * config/i386/i386.h (ix86_size_cost) : Define
26784 (processor_type) : Add PROCESSOR_ROCKETLAKE.
26785 (PTA_ROCKETLAKE): Ditto.
26786 * doc/extend.texi: Add rocketlake.
26787 * doc/invoke.texi: Add rocketlake.
26789 2021-04-12 Cui,Lili <lili.cui@intel.com>
26791 * config/i386/i386.h (PTA_ALDERLAKE): Change alderlake ISA list.
26792 * config/i386/i386-options.c (m_CORE_AVX2): Add m_ALDERLAKE.
26793 * common/config/i386/cpuinfo.h (get_intel_cpu): Add AlderLake model.
26794 * doc/invoke.texi: Change alderlake ISA list.
26796 2021-04-11 Hafiz Abid Qadeer <abidh@codesourcery.com>
26798 PR middle-end/98088
26799 * omp-expand.c (expand_oacc_collapse_init): Update condition in
26802 2021-04-10 H.J. Lu <hjl.tools@gmail.com>
26805 * config/i386/serializeintrin.h (_serialize): Defined as macro.
26807 2021-04-10 Jakub Jelinek <jakub@redhat.com>
26810 * expr.c (expand_expr_addr_expr_1): Test is_global_var rather than
26811 just TREE_STATIC on COMPOUND_LITERAL_EXPR_DECLs.
26813 2021-04-10 Jakub Jelinek <jakub@redhat.com>
26815 PR middle-end/99989
26816 * gimple-ssa-warn-alloca.c
26817 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit to
26818 0 with integer precision unconditionally.
26820 2021-04-10 Jakub Jelinek <jakub@redhat.com>
26822 PR rtl-optimization/98601
26823 * rtlanal.c (rtx_addr_can_trap_p_1): Allow in assert unknown size
26824 not just for BLKmode, but also for VOIDmode. For STRICT_ALIGNMENT
26825 unaligned_mems handle VOIDmode like BLKmode.
26827 2021-04-10 Jan Hubicka <hubicka@ucw.cz>
26830 * tree.c (free_lang_data_in_decl): Do not release body of
26831 declare_variant_alt.
26833 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
26835 * config/aarch64/aarch64.c (aarch64_option_restore): If the
26836 architecture was specified explicitly and the tuning wasn't,
26837 tune for the architecture rather than the configured default CPU.
26839 2021-04-09 Richard Sandiford <richard.sandiford@arm.com>
26841 * config/aarch64/aarch64.md (tlsdesc_small_sve_<mode>): Use X30
26842 as the temporary register.
26844 2021-04-09 Martin Liska <mliska@suse.cz>
26846 * doc/extend.texi: Move non-target attributes on the top level.
26848 2021-04-09 Martin Liska <mliska@suse.cz>
26850 * doc/invoke.texi: Document minimum and maximum value of the
26851 argument for both supported compression algorithms.
26853 2021-04-08 David Edelsohn <dje.gcc@gmail.com>
26855 * config/rs6000/rs6000.c (rs6000_xcoff_select_section): Select
26856 TLS BSS before TLS data.
26857 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .comm.
26859 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
26861 * doc/sourcebuild.texi (stdint_types_mbig_endian): Document.
26863 2021-04-08 Richard Sandiford <richard.sandiford@arm.com>
26865 * match.pd: Extend vec_cond folds to handle shifts.
26867 2021-04-08 Maciej W. Rozycki <macro@orcam.me.uk>
26869 * config/vax/vax.md: Fix comment for `*bit<mode>' pattern's
26872 2021-04-08 Alex Coplan <alex.coplan@arm.com>
26875 * config/arm/iterators.md (MVE_vecs): New.
26876 (V_elem): Also handle V2DF.
26877 * config/arm/mve.md (*mve_mov<mode>): Rename to ...
26878 (*mve_vdup<mode>): ... this. Remove second alternative since
26879 vec_duplicate of const_int is not canonical RTL, and we don't
26880 want to match symbol_refs.
26881 (*mve_vec_duplicate<mode>): Delete (pattern is redundant).
26883 2021-04-08 Xionghu Luo <luoxhu@linux.ibm.com>
26885 * fold-const.c (fold_single_bit_test): Fix typo.
26886 * print-rtl.c (print_rtx_insn_vec): Call print_rtl_single
26889 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
26891 PR tree-optimization/97513
26892 * tree-vect-slp.c (vect_add_slp_permutation): New function,
26894 (vectorizable_slp_permutation): ...here. Detect cases in which
26895 all VEC_PERM_EXPRs are guaranteed to have the same stepped
26896 permute vector and only generate one permute vector for that case.
26897 Extend that case to handle variable-length vectors.
26899 2021-04-07 Richard Sandiford <richard.sandiford@arm.com>
26901 PR tree-optimization/99873
26902 * tree-vect-slp.c (vect_slp_prefer_store_lanes_p): New function.
26903 (vect_build_slp_instance): Don't split store groups that could
26904 use IFN_STORE_LANES.
26906 2021-04-07 Jakub Jelinek <jakub@redhat.com>
26909 * varasm.c (output_constant_pool_contents): Don't strip name encoding
26910 from XSTR (desc->sym, 0) or from label before passing those to
26913 2021-04-07 Richard Biener <rguenther@suse.de>
26915 PR tree-optimization/99954
26916 * tree-loop-distribution.c: Include tree-affine.h.
26917 (generate_memcpy_builtin): Try using tree-affine to prove
26919 (loop_distribution::classify_builtin_ldst): Always classify
26922 2021-04-07 Richard Biener <rguenther@suse.de>
26924 PR tree-optimization/99947
26925 * tree-vect-loop.c (vectorizable_induction): Pre-allocate
26926 steps vector to avoid pushing elements from the reallocated
26929 2021-04-07 Richard Biener <rguenther@suse.de>
26931 * tree-ssa-sccvn.h (print_vn_reference_ops): Declare.
26932 * tree-ssa-pre.c (print_pre_expr): Factor out VN reference operand
26934 * tree-ssa-sccvn.c (print_vn_reference_ops): ... into this new
26936 (debug_vn_reference_ops): New.
26938 2021-04-07 Bin Cheng <bin.cheng@linux.alibaba.com>
26940 PR tree-optimization/98736
26941 * tree-loop-distribution.c
26942 * (loop_distribution::bb_top_order_init):
26943 Compute RPO with programing order preserved by calling function
26944 rev_post_order_and_mark_dfs_back_seme.
26946 2021-04-06 Vladimir N. Makarov <vmakarov@redhat.com>
26949 * lra-constraints.c (split_reg): Don't check paradoxical_subreg_p.
26950 * lra-lives.c (clear_sparseset_regnos, regnos_in_sparseset_p): New
26952 (process_bb_lives): Don't update biggest mode of hard reg for
26953 implicit in multi-register group. Use the new functions for
26954 updating dead_set and unused_set by register notes.
26956 2021-04-06 Xianmiao Qu <xianmiao_qu@c-sky.com>
26958 * config/csky/csky_pipeline_ck802.md : Use insn reservation name
26961 2021-04-06 H.J. Lu <hjl.tools@gmail.com>
26963 * config/i386/x86-tune-costs.h (skylake_memcpy): Updated.
26964 (skylake_memset): Likewise.
26965 (skylake_cost): Change CLEAR_RATIO to 17.
26966 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
26967 Replace m_CANNONLAKE, m_ICELAKE_CLIENT, m_ICELAKE_SERVER,
26968 m_TIGERLAKE and m_SAPPHIRERAPIDS with m_SKYLAKE and m_CORE_AVX512.
26970 2021-04-06 Richard Biener <rguenther@suse.de>
26972 PR tree-optimization/99880
26973 * tree-vect-loop.c (maybe_set_vectorized_backedge_value): Only
26974 set vectorized defs of relevant PHIs.
26976 2021-04-06 Richard Biener <rguenther@suse.de>
26978 PR tree-optimization/99924
26979 * tree-vect-slp.c (vect_bb_partition_graph_r): Do not mark
26980 nodes w/o scalar stmts as visited.
26982 2021-04-06 Alex Coplan <alex.coplan@arm.com>
26985 * config/arm/arm.c (arm_libcall_uses_aapcs_base): Also use base
26986 PCS for [su]fix_optab.
26988 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
26990 * config/darwin.c (machopic_legitimize_pic_address): Check
26991 that the current pic register is one of the hard reg set
26992 before setting liveness.
26994 2021-04-03 Iain Sandoe <iain@sandoe.co.uk>
26996 * config/darwin.c (machopic_legitimize_pic_address): Fix
26997 whitespace, remove unused code.
26999 2021-04-03 Jakub Jelinek <jakub@redhat.com>
27001 PR tree-optimization/99882
27002 * gimple-ssa-store-merging.c (bswap_view_convert): Handle val with
27005 2021-04-03 Jakub Jelinek <jakub@redhat.com>
27007 PR rtl-optimization/99863
27008 * dse.c (replace_read): Drop regs_live argument. Instead of
27009 regs_live, use store_insn->fixed_regs_live if non-NULL,
27010 otherwise punt if insns sequence clobbers or sets any hard
27013 2021-04-03 Jakub Jelinek <jakub@redhat.com>
27016 * targhooks.h (default_print_patchable_function_entry_1): Declare.
27017 * targhooks.c (default_print_patchable_function_entry_1): New function,
27018 copied from default_print_patchable_function_entry with an added flags
27020 (default_print_patchable_function_entry): Rewritten into a small
27021 wrapper around default_print_patchable_function_entry_1.
27022 * config/rs6000/rs6000.c (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
27024 (rs6000_print_patchable_function_entry): New function.
27026 2021-04-02 Eric Botcazou <ebotcazou@adacore.com>
27028 * doc/invoke.texi (fdelete-dead-exceptions): Minor tweak.
27030 2021-04-01 Jason Merrill <jason@redhat.com>
27033 * common.opt: Document v15 and v16.
27035 2021-04-01 Richard Biener <rguenther@suse.de>
27037 PR tree-optimization/99863
27038 * gimplify.c (gimplify_init_constructor): Recompute vector
27041 2021-04-01 Jakub Jelinek <jakub@redhat.com>
27043 * doc/extend.texi (symver attribute): Fix up syntax errors
27046 2021-04-01 Jakub Jelinek <jakub@redhat.com>
27048 PR tree-optimization/96573
27049 * gimple-ssa-store-merging.c (init_symbolic_number): Handle
27050 also pointer types.
27052 2021-04-01 Richard Biener <rguenther@suse.de>
27054 PR tree-optimization/99856
27055 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Promote
27056 precision to vector element precision.
27058 2021-04-01 Martin Jambor <mjambor@suse.cz>
27060 PR tree-optimization/97009
27061 * tree-sra.c (access_or_its_child_written): New function.
27062 (propagate_subaccesses_from_rhs): Use it instead of a simple grp_write
27065 2021-03-31 Jan Hubicka <hubicka@ucw.cz>
27068 * cif-code.def (USES_COMDAT_LOCAL): Make CIF_FINAL_NORMAL.
27070 2021-03-31 Pat Haugen <pthaugen@linux.ibm.com>
27073 * config/rs6000/altivec.md (xxspltiw_v4si, xxspltiw_v4sf_inst,
27074 xxspltidp_v2df_inst, xxsplti32dx_v4si_inst, xxsplti32dx_v4sf_inst,
27075 xxblend_<mode>, xxpermx_inst, xxeval): Mark prefixed.
27076 * config/rs6000/mma.md (mma_<vvi4i4i8>, mma_<avvi4i4i8>,
27077 mma_<vvi4i4i2>, mma_<avvi4i4i2>, mma_<vvi4i4>, mma_<avvi4i4>,
27078 mma_<pvi4i2>, mma_<apvi4i2>, mma_<vvi4i4i4>, mma_<avvi4i4i4>):
27080 * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Adjust test.
27081 * config/rs6000/rs6000.md (define_attr "maybe_prefixed"): New.
27082 (define_attr "prefixed"): Update initializer.
27084 2021-03-31 Jakub Jelinek <jakub@redhat.com>
27087 * dwarf2out.c (debug_ranges_dwo_section): New variable.
27088 (DW_RANGES_IDX_SKELETON): Define.
27089 (struct dw_ranges): Add begin_entry and end_entry members.
27090 (DEBUG_DWO_RNGLISTS_SECTION): Define.
27091 (add_ranges_num): Adjust r initializer for addition of *_entry
27093 (add_ranges_by_labels): For -gsplit-dwarf and force_direct,
27094 set idx to DW_RANGES_IDX_SKELETON.
27095 (use_distinct_base_address_for_range): New function.
27096 (index_rnglists): Don't set r->idx if it is equal to
27097 DW_RANGES_IDX_SKELETON. Initialize r->begin_entry and
27098 r->end_entry for -gsplit-dwarf if those will be needed by
27100 (output_rnglists): Add DWO argument. If true, switch to
27101 debug_ranges_dwo_section rather than debug_ranges_section.
27102 Adjust l1/l2 label indexes. Only output the offset table when
27103 dwo is true and don't include in there the skeleton range
27104 entry if present. For -gsplit-dwarf, skip ranges that belong
27105 to the other rnglists section. Change return type from void
27106 to bool and return true if there are any range entries for
27107 the other section. For dwarf_split_debug_info use
27108 DW_RLE_startx_endx, DW_RLE_startx_length and DW_RLE_base_addressx
27109 entries instead of DW_RLE_start_end, DW_RLE_start_length and
27110 DW_RLE_base_address. Use use_distinct_base_address_for_range.
27111 (init_sections_and_labels): Initialize debug_ranges_dwo_section
27112 if -gsplit-dwarf and DWARF >= 5. Adjust ranges_section_label
27113 and range_base_label indexes.
27114 (dwarf2out_finish): Call index_rnglists earlier before finalizing
27115 .debug_addr. Never emit DW_AT_rnglists_base attribute. For
27116 -gsplit-dwarf and DWARF >= 5 call output_rnglists up to twice
27117 with different dwo arguments.
27118 (dwarf2out_c_finalize): Clear debug_ranges_dwo_section.
27120 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
27122 PR tree-optimization/98268
27123 * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Call
27124 recompute_tree_invariant_for_addr_expr after successfully
27125 folding a TARGET_MEM_REF that occurs inside an ADDR_EXPR.
27127 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
27129 PR tree-optimization/99726
27130 * tree-data-ref.c (create_intersect_range_checks_index): Bail
27131 out if there is more than one access function SCEV for the loop
27134 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
27136 PR rtl-optimization/97141
27137 PR rtl-optimization/98726
27138 * emit-rtl.c (valid_for_const_vector_p): Return true for
27140 * rtx-vector-builder.h (rtx_vector_builder::step): Return a
27141 poly_wide_int instead of a wide_int.
27142 (rtx_vector_builder::apply_set): Take a poly_wide_int instead
27144 * rtx-vector-builder.c (rtx_vector_builder::apply_set): Likewise.
27145 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Return
27146 false for CONST_VECTORs that cannot be forced to memory.
27147 * config/aarch64/aarch64-simd.md (mov<mode>): If a CONST_VECTOR
27148 is too complex to force to memory, build it up from individual
27151 2021-03-31 Jan Hubicka <jh@suse.cz>
27154 * cgraph.c (cgraph_node::release_body): Fix overactive check.
27156 2021-03-31 Christophe Lyon <christophe.lyon@linaro.org>
27159 * config/arm/vec-common.md (mul<mode>3): Disable on iwMMXT, expect
27162 2021-03-31 H.J. Lu <hjl.tools@gmail.com>
27164 * config/i386/i386-expand.c (expand_set_or_cpymem_via_rep):
27165 For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, don't convert QImode
27167 (decide_alg): For TARGET_PREFER_KNOWN_REP_MOVSB_STOSB, use
27168 "rep movsb/stosb" only for known sizes.
27169 * config/i386/i386-options.c (processor_cost_table): Use Ice
27170 Lake cost for Cannon Lake, Ice Lake, Tiger Lake, Sapphire
27171 Rapids and Alder Lake.
27172 * config/i386/i386.h (TARGET_PREFER_KNOWN_REP_MOVSB_STOSB): New.
27173 * config/i386/x86-tune-costs.h (icelake_memcpy): New.
27174 (icelake_memset): Likewise.
27175 (icelake_cost): Likewise.
27176 * config/i386/x86-tune.def (X86_TUNE_PREFER_KNOWN_REP_MOVSB_STOSB):
27179 2021-03-31 Richard Sandiford <richard.sandiford@arm.com>
27182 * config/aarch64/aarch64.c
27183 (aarch64_vectorize_preferred_vector_alignment): Query the size
27184 of the provided SVE vector; do not assume that all SVE vectors
27185 have the same size.
27187 2021-03-31 Jan Hubicka <jh@suse.cz>
27190 * cgraph.c (cgraph_node::release_body): Remove all callers and
27192 * cgraphclones.c (cgraph_node::materialize_clone): Do not do it here.
27193 * cgraphunit.c (cgraph_node::expand): And here.
27195 2021-03-31 Martin Liska <mliska@suse.cz>
27197 * ipa-modref.c (analyze_ssa_name_flags): Fix coding style
27198 and one negated condition.
27200 2021-03-31 Jakub Jelinek <jakub@redhat.com>
27201 Richard Sandiford <richard.sandiford@arm.com>
27204 * config/aarch64/aarch64.md (*add<mode>3_poly_1): Swap Uai and Uav
27205 constraints on operands[2] and similarly 0 and rk constraints
27206 on operands[1] corresponding to that.
27208 2021-03-31 Jakub Jelinek <jakub@redhat.com>
27211 * configure.ac (HAVE_LD_BROKEN_PE_DWARF5): New AC_DEFINE if PECOFF
27212 linker doesn't support DWARF sections new in DWARF5.
27213 * config/i386/i386-options.c (ix86_option_override_internal): Default
27214 to dwarf_version 4 if HAVE_LD_BROKEN_PE_DWARF5 for TARGET_PECOFF
27216 * config.in: Regenerated.
27217 * configure: Regenerated.
27219 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27222 * config/aarch64/aarch64.c (aarch64_analyze_loop_vinfo): Check for
27223 available issue_info before using it.
27225 2021-03-30 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27228 * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): Do not allow zero
27231 2021-03-30 Xionghu Luo <luoxhu@linux.ibm.com>
27234 * config/rs6000/altivec.md (altivec_lvsl_reg): Change to ...
27235 (altivec_lvsl_reg_<mode>): ... this.
27236 (altivec_lvsr_reg): Change to ...
27237 (altivec_lvsr_reg_<mode>): ... this.
27238 * config/rs6000/predicates.md (vec_set_index_operand): New.
27239 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
27240 Enable 32bit variable vec_insert for all TARGET_VSX.
27241 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
27242 Enable 32bit variable vec_insert for p9 and above.
27243 (rs6000_expand_vector_set_var_p8): Rename to ...
27244 (rs6000_expand_vector_set_var_p7): ... this.
27245 (rs6000_expand_vector_set): Use TARGET_VSX and adjust assert
27247 * config/rs6000/vector.md (vec_set<mode>): Use vec_set_index_operand.
27248 * config/rs6000/vsx.md (xl_len_r): Use gen_altivec_lvsl_reg_di and
27249 gen_altivec_lvsr_reg_di.
27251 2021-03-30 H.J. Lu <hjl.tools@gmail.com>
27254 * config/i386/ia32intrin.h (__rdtsc): Defined as macro.
27255 (__rdtscp): Likewise.
27257 2021-03-30 Tamar Christina <tamar.christina@arm.com>
27259 PR tree-optimization/99825
27260 * tree-vect-slp-patterns.c (vect_check_evenodd_blend):
27261 Reject non-mult 2 lanes.
27263 2021-03-30 Richard Earnshaw <rearnsha@arm.com>
27266 * config/arm/arm.c (arm_file_start): Fix emission of
27267 Tag_ABI_VFP_args attribute.
27269 2021-03-30 Richard Biener <rguenther@suse.de>
27271 PR tree-optimization/99824
27272 * stor-layout.c (set_min_and_max_values_for_integral_type):
27273 Assert the precision is within the bounds of
27274 WIDE_INT_MAX_PRECISION.
27275 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
27276 the outermost component ref only to lower the access size
27277 and initialize that from the access type.
27279 2021-03-30 Richard Sandiford <richard.sandiford@arm.com>
27282 * config/aarch64/aarch64.md (mov<mode>): Pass multi-instruction
27283 CONST_INTs to aarch64_expand_mov_immediate when called after RA.
27285 2021-03-30 Mihailo Stojanovic <mihailo.stojanovic@typhoon-hil.com>
27287 * config/aarch64/aarch64.md
27288 (<optab>_trunc<fcvt_target><GPI:mode>2): Set the "arch"
27289 attribute to disambiguate between SIMD and FP variants of the
27292 2021-03-29 Jan Hubicka <hubicka@ucw.cz>
27294 * ipa-modref.c (merge_call_lhs_flags): Correct handling of deref.
27295 (analyze_ssa_name_flags): Fix typo in comment.
27297 2021-03-29 Alex Coplan <alex.coplan@arm.com>
27300 * config/aarch64/aarch64-sve-builtins.cc
27301 (function_builder::add_function): Add placeholder_p argument, use
27302 placeholder decls if this is set.
27303 (function_builder::add_unique_function): Instead of conditionally adding
27304 direct overloads, unconditionally add either a direct overload or a
27306 (function_builder::add_overloaded_function): Set placeholder_p if we're
27307 using C++ overloads. Use the obstack for string storage instead
27308 of relying on the tree nodes.
27309 (function_builder::add_overloaded_functions): Don't return early for
27310 m_direct_overloads: we need to add placeholders.
27311 * config/aarch64/aarch64-sve-builtins.h
27312 (function_builder::add_function): Add placeholder_p argument.
27314 2021-03-29 Richard Biener <rguenther@suse.de>
27316 PR tree-optimization/99807
27317 * tree-vect-slp.c (vect_slp_analyze_node_operations_1): Move
27318 assert below VEC_PERM handling.
27320 2021-03-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27323 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>): Use
27324 aarch64_simd_or_scalar_imm_zero to match zeroes. Remove pattern
27325 matching const_int 0.
27326 (move_lo_quad_internal_be_<mode>): Likewise.
27327 (move_lo_quad_<mode>): Update for the above.
27328 * config/aarch64/iterators.md (VQ_2E): Delete.
27330 2021-03-29 Jakub Jelinek <jakub@redhat.com>
27332 PR tree-optimization/99777
27333 * fold-const.c (extract_muldiv_1): For conversions, punt on casts from
27334 types other than scalar integral types.
27336 2021-03-28 David Edelsohn <dje.gcc@gmail.com>
27338 * config/rs6000/rs6000.c (rs6000_output_dwarf_dtprel): Do not add
27339 XCOFF TLS reloc decorations.
27341 2021-03-28 Gerald Pfeifer <gerald@pfeifer.com>
27343 * doc/analyzer.texi (Analyzer Internals): Update link to
27344 "A Memory Model for Static Analysis of C Programs".
27346 2021-03-26 David Edelsohn <dje.gcc@gmail.com>
27348 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Call function.
27349 * config/rs6000/rs6000-protos.h (rs6000_special_adjust_field_align):
27351 * config/rs6000/rs6000.c (rs6000_special_adjust_field_align): New.
27352 (rs6000_special_round_type_align): Recursively check innermost first
27355 2021-03-26 Jakub Jelinek <jakub@redhat.com>
27358 * dwarf2out.h (struct dw_fde_node): Add rule18 member.
27359 * dwarf2cfi.c (dwarf2out_frame_debug_expr): When handling (set hfp sp)
27360 assignment with drap_reg active, queue reg save for hfp with offset 0
27361 and flush queued reg saves. When handling a push with rule18,
27362 defer queueing reg save for hfp and just assert the offset is 0.
27363 (scan_trace): Assert that fde->rule18 is false.
27365 2021-03-26 Vladimir Makarov <vmakarov@redhat.com>
27368 * ira-costs.c (record_reg_classes): Put case with
27369 CT_RELAXED_MEMORY adjacent to one with CT_MEMORY.
27370 * ira.c (ira_setup_alts): Ditto.
27371 * lra-constraints.c (process_alt_operands): Ditto.
27372 * recog.c (asm_operand_ok): Ditto.
27373 * reload.c (find_reloads): Ditto.
27375 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
27377 * config/aarch64/aarch64-protos.h
27378 (cpu_addrcost_table::post_modify_ld3_st3): New member variable.
27379 (cpu_addrcost_table::post_modify_ld4_st4): Likewise.
27380 * config/aarch64/aarch64.c (generic_addrcost_table): Update
27381 accordingly, using the same costs as for post_modify.
27382 (exynosm1_addrcost_table, xgene1_addrcost_table): Likewise.
27383 (thunderx2t99_addrcost_table, thunderx3t110_addrcost_table):
27384 (tsv110_addrcost_table, qdf24xx_addrcost_table): Likewise.
27385 (a64fx_addrcost_table): Likewise.
27386 (neoversev1_addrcost_table): New.
27387 (neoversev1_tunings): Use neoversev1_addrcost_table.
27388 (aarch64_address_cost): Use the new post_modify costs for CImode
27391 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
27393 * config/aarch64/aarch64.opt
27394 (-param=aarch64-loop-vect-issue-rate-niters=): New parameter.
27395 * doc/invoke.texi: Document it.
27396 * config/aarch64/aarch64-protos.h (aarch64_base_vec_issue_info)
27397 (aarch64_scalar_vec_issue_info, aarch64_simd_vec_issue_info)
27398 (aarch64_advsimd_vec_issue_info, aarch64_sve_vec_issue_info)
27399 (aarch64_vec_issue_info): New structures.
27400 (cpu_vector_cost): Write comments above the variables rather
27402 (cpu_vector_cost::issue_info): New member variable.
27403 * config/aarch64/aarch64.c: Include gimple-pretty-print.h
27404 and tree-ssa-loop-niter.h.
27405 (generic_vector_cost, a64fx_vector_cost, qdf24xx_vector_cost)
27406 (thunderx_vector_cost, tsv110_vector_cost, cortexa57_vector_cost)
27407 (exynosm1_vector_cost, xgene1_vector_cost, thunderx2t99_vector_cost)
27408 (thunderx3t110_vector_cost): Initialize issue_info to null.
27409 (neoversev1_scalar_issue_info, neoversev1_advsimd_issue_info)
27410 (neoversev1_sve_issue_info, neoversev1_vec_issue_info): New structures.
27411 (neoversev1_vector_cost): Use them.
27412 (aarch64_vec_op_count, aarch64_sve_op_count): New structures.
27413 (aarch64_vector_costs::saw_sve_only_op): New member variable.
27414 (aarch64_vector_costs::num_vector_iterations): Likewise.
27415 (aarch64_vector_costs::scalar_ops): Likewise.
27416 (aarch64_vector_costs::advsimd_ops): Likewise.
27417 (aarch64_vector_costs::sve_ops): Likewise.
27418 (aarch64_vector_costs::seen_loads): Likewise.
27419 (aarch64_simd_vec_costs_for_flags): New function.
27420 (aarch64_analyze_loop_vinfo): Initialize num_vector_iterations.
27421 Count the number of predicate operations required by SVE WHILE
27423 (aarch64_comparison_type, aarch64_multiply_add_p): New functions.
27424 (aarch64_sve_only_stmt_p, aarch64_in_loop_reduction_latency): Likewise.
27425 (aarch64_count_ops): Likewise.
27426 (aarch64_add_stmt_cost): Record whether see an SVE operation
27427 that cannot currently be implementing using Advanced SIMD.
27428 Record issue information about the scalar, Advanced SIMD
27429 and (where relevant) SVE versions of a loop.
27430 (aarch64_vec_op_count::dump): New function.
27431 (aarch64_sve_op_count::dump): Likewise.
27432 (aarch64_estimate_min_cycles_per_iter): Likewise.
27433 (aarch64_adjust_body_cost): If issue information is available,
27434 try to compare the issue rates of the various loop implementations
27435 and increase or decrease the vector body cost accordingly.
27437 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
27439 * config/aarch64/aarch64.c (aarch64_detect_vector_stmt_subtype):
27440 Assume a zero cost for induction phis.
27442 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
27444 * config/aarch64/aarch64.c (aarch64_embedded_comparison_type): New
27446 (aarch64_adjust_stmt_cost): Add the costs of embedded scalar and
27447 vector comparisons.
27449 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
27451 * config/aarch64/aarch64.c (aarch64_detect_scalar_stmt_subtype):
27453 (aarch64_add_stmt_cost): Call it.
27455 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
27457 * config/aarch64/aarch64-tuning-flags.def (matched_vector_throughput):
27458 New tuning parameter.
27459 * config/aarch64/aarch64.c (neoversev1_tunings): Use it.
27460 (aarch64_estimated_sve_vq): New function.
27461 (aarch64_vector_costs::analyzed_vinfo): New member variable.
27462 (aarch64_vector_costs::is_loop): Likewise.
27463 (aarch64_vector_costs::unrolled_advsimd_niters): Likewise.
27464 (aarch64_vector_costs::unrolled_advsimd_stmts): Likewise.
27465 (aarch64_record_potential_advsimd_unrolling): New function.
27466 (aarch64_analyze_loop_vinfo, aarch64_analyze_bb_vinfo): Likewise.
27467 (aarch64_add_stmt_cost): Call aarch64_analyze_loop_vinfo or
27468 aarch64_analyze_bb_vinfo on the first use of a costs structure.
27469 Detect whether we're vectorizing a loop for SVE that might be
27470 completely unrolled if it used Advanced SIMD instead.
27471 (aarch64_adjust_body_cost_for_latency): New function.
27472 (aarch64_finish_cost): Call it.
27474 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
27476 * config/aarch64/aarch64.c (aarch64_vector_costs): New structure.
27477 (aarch64_init_cost): New function.
27478 (aarch64_add_stmt_cost): Use aarch64_vector_costs instead of
27479 the default unsigned[3].
27480 (aarch64_finish_cost, aarch64_destroy_cost_data): New functions.
27481 (TARGET_VECTORIZE_INIT_COST): Override.
27482 (TARGET_VECTORIZE_FINISH_COST): Likewise.
27483 (TARGET_VECTORIZE_DESTROY_COST_DATA): Likewise.
27485 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
27487 * config/aarch64/aarch64.c (neoversev1_advsimd_vector_cost)
27488 (neoversev1_sve_vector_cost): New cost structures.
27489 (neoversev1_vector_cost): Likewise.
27490 (neoversev1_tunings): Use them. Enable use_new_vector_costs.
27492 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
27494 * config/aarch64/aarch64-protos.h
27495 (sve_vec_cost::scatter_store_elt_cost): New member variable.
27496 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
27497 accordingly, taking the cost from the cost of a scalar_store.
27498 (a64fx_sve_vector_cost): Likewise.
27499 (aarch64_detect_vector_stmt_subtype): Detect scatter stores.
27501 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
27503 * config/aarch64/aarch64-protos.h
27504 (simd_vec_cost::store_elt_extra_cost): New member variable.
27505 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
27506 accordingly, using the vec_to_scalar cost for the new field.
27507 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
27508 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
27509 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
27510 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
27511 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
27512 (thunderx3t110_advsimd_vector_cost): Likewise.
27513 (aarch64_detect_vector_stmt_subtype): Detect single-element stores.
27515 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
27517 * config/aarch64/aarch64-protos.h (simd_vec_cost::ld2_st2_permute_cost)
27518 (simd_vec_cost::ld3_st3_permute_cost): New member variables.
27519 (simd_vec_cost::ld4_st4_permute_cost): Likewise.
27520 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
27521 accordingly, using zero for the new costs.
27522 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
27523 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
27524 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
27525 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
27526 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
27527 (thunderx3t110_advsimd_vector_cost): Likewise.
27528 (aarch64_ld234_st234_vectors): New function.
27529 (aarch64_adjust_stmt_cost): Likewise.
27530 (aarch64_add_stmt_cost): Call aarch64_adjust_stmt_cost if using
27531 the new vector costs.
27533 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
27535 * config/aarch64/aarch64-protos.h (sve_vec_cost): Turn into a
27536 derived class of simd_vec_cost. Add information about CLAST[AB]
27537 and FADDA instructions.
27538 * config/aarch64/aarch64.c (generic_sve_vector_cost): Update
27539 accordingly, using the vec_to_scalar costs for the new fields.
27540 (a64fx_sve_vector_cost): Likewise.
27541 (aarch64_reduc_type): New function.
27542 (aarch64_sve_in_loop_reduction_latency): Likewise.
27543 (aarch64_detect_vector_stmt_subtype): Take a vinfo parameter.
27544 Use aarch64_sve_in_loop_reduction_latency to handle SVE reductions
27545 that occur in the loop body.
27546 (aarch64_add_stmt_cost): Update call accordingly.
27548 2021-03-26 Richard Sandiford <richard.sandiford@arm.com>
27550 * config/aarch64/aarch64-tuning-flags.def (use_new_vector_costs):
27552 * config/aarch64/aarch64-protos.h (simd_vec_cost): Put comments
27553 above the fields rather than to the right.
27554 (simd_vec_cost::reduc_i8_cost): New member variable.
27555 (simd_vec_cost::reduc_i16_cost): Likewise.
27556 (simd_vec_cost::reduc_i32_cost): Likewise.
27557 (simd_vec_cost::reduc_i64_cost): Likewise.
27558 (simd_vec_cost::reduc_f16_cost): Likewise.
27559 (simd_vec_cost::reduc_f32_cost): Likewise.
27560 (simd_vec_cost::reduc_f64_cost): Likewise.
27561 * config/aarch64/aarch64.c (generic_advsimd_vector_cost): Update
27562 accordingly, using the vec_to_scalar_cost for the new fields.
27563 (generic_sve_vector_cost, a64fx_advsimd_vector_cost): Likewise.
27564 (a64fx_sve_vector_cost, qdf24xx_advsimd_vector_cost): Likewise.
27565 (thunderx_advsimd_vector_cost, tsv110_advsimd_vector_cost): Likewise.
27566 (cortexa57_advsimd_vector_cost, exynosm1_advsimd_vector_cost)
27567 (xgene1_advsimd_vector_cost, thunderx2t99_advsimd_vector_cost)
27568 (thunderx3t110_advsimd_vector_cost): Likewise.
27569 (aarch64_use_new_vector_costs_p): New function.
27570 (aarch64_simd_vec_costs): New function, split out from...
27571 (aarch64_builtin_vectorization_cost): ...here.
27572 (aarch64_is_reduction): New function.
27573 (aarch64_detect_vector_stmt_subtype): Likewise.
27574 (aarch64_add_stmt_cost): Call aarch64_detect_vector_stmt_subtype if
27575 using the new vector costs.
27577 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
27580 * tree-emutls.c (get_emutls_init_templ_addr): Mark initializer of weak
27581 TLS declarations as public.
27583 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
27585 * config/aarch64/aarch64-d.c (IN_TARGET_CODE): Define.
27586 * config/arm/arm-d.c (IN_TARGET_CODE): Likewise.
27587 * config/i386/i386-d.c (IN_TARGET_CODE): Likewise.
27588 * config/mips/mips-d.c (IN_TARGET_CODE): Likewise.
27589 * config/pa/pa-d.c (IN_TARGET_CODE): Likewise.
27590 * config/riscv/riscv-d.c (IN_TARGET_CODE): Likewise.
27591 * config/rs6000/rs6000-d.c (IN_TARGET_CODE): Likewise.
27592 * config/s390/s390-d.c (IN_TARGET_CODE): Likewise.
27593 * config/sparc/sparc-d.c (IN_TARGET_CODE): Likewise.
27595 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
27598 * config.gcc (*-*-cygwin*): Add winnt-d.o
27599 (*-*-mingw*): Likewise.
27600 * config/i386/cygwin.h (EXTRA_TARGET_D_OS_VERSIONS): New macro.
27601 * config/i386/mingw32.h (EXTRA_TARGET_D_OS_VERSIONS): Likewise.
27602 * config/i386/t-cygming: Add winnt-d.o.
27603 * config/i386/winnt-d.c: New file.
27605 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
27607 * config/freebsd-d.c: Include memmodel.h.
27609 2021-03-26 Iain Buclaw <ibuclaw@gdcproject.org>
27612 * config.gcc (*-*-openbsd*): Add openbsd-d.o.
27613 * config/t-openbsd: Add openbsd-d.o.
27614 * config/openbsd-d.c: New file.
27616 2021-03-25 Stam Markianos-Wright <stam.markianos-wright@arm.com>
27618 PR tree-optimization/96974
27619 * tree-vect-stmts.c (vect_get_vector_types_for_stmt): Replace assert
27620 with graceful exit.
27622 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
27625 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
27629 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
27630 always_inline in system headers.
27632 2021-03-25 Kewen Lin <linkw@linux.ibm.com>
27634 * tree-vect-loop.c (vect_model_reduction_cost): Init inside_cost.
27636 2021-03-25 Jakub Jelinek <jakub@redhat.com>
27639 * tree-core.h (enum operand_equal_flag): Add OEP_ADDRESS_OF_SAME_FIELD.
27640 * fold-const.c (operand_compare::operand_equal_p): Don't compare
27641 field offsets if OEP_ADDRESS_OF_SAME_FIELD.
27643 2021-03-25 H.J. Lu <hjl.tools@gmail.com>
27647 * config/i386/i386.c (ix86_can_inline_p): Don't check ISA for
27648 always_inline in system headers.
27650 2021-03-25 Richard Biener <rguenther@suse.de>
27652 PR tree-optimization/99746
27653 * tree-vect-slp-patterns.c (complex_pattern::build): Do not mark
27654 the scalar stmt as patterned. Instead set up required things
27657 2021-03-25 Xionghu Luo <luoxhu@linux.ibm.com>
27659 * config/rs6000/rs6000.c (power8_costs): Change l2 cache
27662 2021-03-24 Martin Liska <mliska@suse.cz>
27665 * common/config/i386/i386-common.c (ARRAY_SIZE): Fix off-by-one
27667 * config/i386/i386-options.c (ix86_option_override_internal):
27668 Add run-time assert.
27670 2021-03-24 Martin Jambor <mjambor@suse.cz>
27673 * ipa-cp.c (initialize_node_lattices): Mark as bottom all
27674 parameters with unknown type.
27675 (ipacp_value_safe_for_type): New function.
27676 (propagate_vals_across_arith_jfunc): Verify that the constant type
27677 can be used for a type of the formal parameter.
27678 (propagate_vals_across_ancestor): Likewise.
27679 (propagate_scalar_across_jump_function): Likewise. Pass the type
27680 also to propagate_vals_across_ancestor.
27682 2021-03-24 Christophe Lyon <christophe.lyon@linaro.org>
27685 * config/arm/mve.md (movmisalign<mode>_mve_store): Use Ux
27687 (movmisalign<mode>_mve_load): Likewise.
27689 2021-03-24 Jakub Jelinek <jakub@redhat.com>
27692 * config/arm/vec-common.md (one_cmpl<mode>2, neg<mode>2,
27693 movmisalign<mode>): Disable expanders for TARGET_REALLY_IWMMXT.
27695 2021-03-24 Alexandre Oliva <oliva@adacore.com>
27697 * doc/sourcebuild.texi (sysconf): New effective target.
27699 2021-03-24 Alexandre Oliva <oliva@adacore.com>
27701 * config/i386/predicates.md (reg_or_const_vec_operand): New.
27702 * config/i386/sse.md (ssse3_pshufbv8qi3): Add an expander for
27703 the now *-prefixed insn_and_split, turn the splitter const vec
27704 into an input for the insn, making it an ignored immediate for
27705 non-split cases, and loaded into the scratch register
27708 2021-03-23 Vladimir N. Makarov <vmakarov@redhat.com>
27711 * config/aarch64/constraints.md (Utq, UOb, UOh, UOw, UOd, UOty):
27712 Use define_relaxed_memory_constraint for them.
27714 2021-03-23 Iain Sandoe <iain@sandoe.co.uk>
27717 * config/host-darwin.c (darwin_gt_pch_use_address): Add a
27718 colon to the diagnostic message.
27720 2021-03-23 Ilya Leoshkevich <iii@linux.ibm.com>
27722 * fwprop.c (fwprop_propagation::fwprop_propagation): Look at
27724 (try_fwprop_subst_note): Use set_info instead of insn_info.
27725 (try_fwprop_subst_pattern): Likewise.
27726 (try_fwprop_subst_notes): Likewise.
27727 (try_fwprop_subst): Likewise.
27728 (forward_propagate_subreg): Likewise.
27729 (forward_propagate_and_simplify): Likewise.
27730 (forward_propagate_into): Likewise.
27731 * rtl-ssa/accesses.h (set_info::single_nondebug_use) New
27733 (set_info::single_nondebug_insn_use): Likewise.
27734 (set_info::single_phi_use): Likewise.
27735 * rtl-ssa/member-fns.inl (set_info::single_nondebug_use) New
27737 (set_info::single_nondebug_insn_use): Likewise.
27738 (set_info::single_phi_use): Likewise.
27740 2021-03-23 Christophe Lyon <christophe.lyon@linaro.org>
27742 * doc/sourcebuild.texi (arm_dsp_ok, arm_dsp): Document.
27744 2021-03-23 Jakub Jelinek <jakub@redhat.com>
27747 * config/aarch64/aarch64.c (aarch64_add_offset): Tell
27748 expand_mult to perform an unsigned rather than a signed
27751 2021-03-23 H.J. Lu <hjl.tools@gmail.com>
27754 * config/i386/cpuid.h (__cpuid): Add __volatile__.
27755 (__cpuid_count): Likewise.
27757 2021-03-23 Richard Biener <rguenther@suse.de>
27759 PR tree-optimization/99721
27760 * tree-vect-slp.c (vect_slp_analyze_node_operations):
27761 Make sure we can schedule the node.
27763 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
27765 * config/riscv/riscv.c (riscv_subword): Take endianness into
27766 account when calculating the byte offset.
27768 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
27770 * config/riscv/predicates.md (subreg_lowpart_operator): New predicate
27771 * config/riscv/riscv.md (*addsi3_extended2, *subsi3_extended2)
27772 (*negsi2_extended2, *mulsi3_extended2, *<optab>si3_mask)
27773 (*<optab>si3_mask_1, *<optab>di3_mask, *<optab>di3_mask_1)
27774 (*<optab>si3_extend_mask, *<optab>si3_extend_mask_1): Use
27775 new predicate "subreg_lowpart_operator"
27777 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
27779 * config/riscv/riscv.c (riscv_swap_instruction): New function
27780 to byteswap an SImode rtx containing an instruction.
27781 (riscv_trampoline_init): Byteswap the generated instructions
27784 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
27786 * common/config/riscv/riscv-common.c
27787 (TARGET_DEFAULT_TARGET_FLAGS): Set default endianness.
27788 * config.gcc (riscv32be-*, riscv64be-*): Set
27789 TARGET_BIG_ENDIAN_DEFAULT to 1.
27790 * config/riscv/elf.h (LINK_SPEC): Change -melf* value
27791 depending on default endianness.
27792 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
27793 * config/riscv/linux.h (LINK_SPEC): Likewise.
27794 * config/riscv/riscv.c (TARGET_DEFAULT_TARGET_FLAGS): Set
27795 default endianness.
27796 * config/riscv/riscv.h (DEFAULT_ENDIAN_SPEC): New macro.
27798 2021-03-23 Marcus Comstedt <marcus@mc.pp.se>
27800 * config/riscv/elf.h (LINK_SPEC): Pass linker endianness flag.
27801 * config/riscv/freebsd.h (LINK_SPEC): Likewise.
27802 * config/riscv/linux.h (LINK_SPEC): Likewise.
27803 * config/riscv/riscv.h (ASM_SPEC): Pass -mbig-endian and
27805 (BYTES_BIG_ENDIAN): Handle big endian.
27806 (WORDS_BIG_ENDIAN): Define to BYTES_BIG_ENDIAN.
27807 * config/riscv/riscv.opt (-mbig-endian, -mlittle-endian): New
27809 * doc/invoke.texi (-mbig-endian, -mlittle-endian): Document.
27811 2021-03-23 Stefan Schulze Frielinghaus <stefansf@linux.ibm.com>
27813 * regcprop.c (find_oldest_value_reg): Ask target whether
27814 different mode is fine for replacement register.
27816 2021-03-23 Aldy Hernandez <aldyh@redhat.com>
27818 PR tree-optimization/99296
27819 * value-range.cc (irange::irange_set_1bit_anti_range): New.
27820 (irange::irange_set_anti_range): Call irange_set_1bit_anti_range
27821 * value-range.h (irange::irange_set_1bit_anti_range): New.
27823 2021-03-22 Vladimir N. Makarov <vmakarov@redhat.com>
27826 * config/aarch64/constraints.md (UtQ): Use
27827 define_relaxed_memory_constraint for it.
27828 * doc/md.texi (define_relaxed_memory_constraint): Describe it.
27829 * genoutput.c (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
27830 * genpreds.c (constraint_data): Add bitfield is_relaxed_memory.
27831 (have_relaxed_memory_constraints): New static var.
27832 (relaxed_memory_start, relaxed_memory_end): Ditto.
27833 (add_constraint): Add arg is_relaxed_memory. Check name for
27834 relaxed memory. Set up is_relaxed_memory in constraint_data and
27835 have_relaxed_memory_constraints. Adjust calls.
27836 (choose_enum_order): Process relaxed memory.
27837 (write_tm_preds_h): Ditto.
27838 (main): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
27839 * gensupport.c (process_rtx): Process DEFINE_RELAXED_MEMORY_CONSTRAINT.
27840 * ira-costs.c (record_reg_classes): Process CT_RELAXED_MEMORY.
27841 * ira-lives.c (single_reg_class): Use
27842 insn_extra_relaxed_memory_constraint.
27843 * ira.c (ira_setup_alts): CT_RELAXED_MEMORY.
27844 * lra-constraints.c (valid_address_p): Use
27845 insn_extra_relaxed_memory_constraint instead of other memory
27847 (process_alt_operands): Process CT_RELAXED_MEMORY.
27848 (curr_insn_transform): Use insn_extra_relaxed_memory_constraint.
27849 * recog.c (asm_operand_ok, preprocess_constraints): Process
27851 * reload.c (find_reloads): Ditto.
27852 * rtl.def (DEFINE_RELAXED_MEMORY_CONSTRAINT): New.
27853 * stmt.c (parse_input_constraint): Use
27854 insn_extra_relaxed_memory_constraint.
27856 2021-03-22 Segher Boessenkool <segher@kernel.crashing.org>
27859 * ubsan.c (ubsan_instrument_float_cast): Don't test for unordered if
27862 2021-03-22 Alex Coplan <alex.coplan@arm.com>
27865 * config/arm/arm-protos.h (neon_make_constant): Add generate
27866 argument to guard emitting insns, default to true.
27867 * config/arm/arm.c (arm_legitimate_constant_p_1): Reject
27868 CONST_VECTORs which neon_make_constant can't handle.
27869 (neon_vdup_constant): Add generate argument, avoid emitting
27870 insns if it's not set.
27871 (neon_make_constant): Plumb new generate argument through.
27872 * config/arm/constraints.md (Ui): New. Use it...
27873 * config/arm/mve.md (*mve_mov<mode>): ... here.
27874 * config/arm/vec-common.md (movv8hf): Use neon_make_constant to
27875 synthesize constants.
27877 2021-03-22 Richard Biener <rguenther@suse.de>
27879 * debug.h: Add deprecation warning.
27881 2021-03-22 Richard Biener <rguenther@suse.de>
27883 PR tree-optimization/99694
27884 * tree-ssa-sccvn.c (visit_phi): Ignore edges with the
27887 2021-03-22 Kito Cheng <kito.cheng@sifive.com>
27890 * config/riscv/riscv.c (riscv_expand_block_move): Get RTL value
27891 after type checking.
27893 2021-03-22 Jakub Jelinek <jakub@redhat.com>
27897 * dwarf2out.c (get_full_len): Use get_precision rather than
27899 (add_const_value_attribute): Make sure add_AT_wide argument has
27900 precision prec rather than some very wide one.
27902 2021-03-22 Kewen Lin <linkw@linux.ibm.com>
27904 * config/rs6000/rs6000.md (*rotldi3_insert_sf,
27905 *mov<SFDF:mode><SFDF2:mode>cc_p9, floatsi<mode>2_lfiwax,
27906 floatsi<mode>2_lfiwax_mem, floatunssi<mode>2_lfiwzx,
27907 floatunssi<mode>2_lfiwzx_mem, *floatsidf2_internal,
27908 *floatunssidf2_internal, fix_trunc<mode>si2_stfiwx,
27909 fix_trunc<mode>si2_internal, fixuns_trunc<mode>si2_stfiwx,
27910 *round32<mode>2_fprs, *roundu32<mode>2_fprs,
27911 *fix_trunc<mode>si2_internal): Fix empty split condition.
27912 * config/rs6000/vsx.md (*vsx_le_undo_permute_<mode>,
27913 vsx_reduc_<VEC_reduc_name>_v2df, vsx_reduc_<VEC_reduc_name>_v4sf,
27914 *vsx_reduc_<VEC_reduc_name>_v2df_scalar,
27915 *vsx_reduc_<VEC_reduc_name>_v4sf_scalar): Likewise.
27917 2021-03-22 Xionghu Luo <luoxhu@linux.ibm.com>
27920 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var_p9):
27921 Convert idx to DImode.
27922 (rs6000_expand_vector_set_var_p8): Likewise.
27924 2021-03-21 Jakub Jelinek <jakub@redhat.com>
27927 * dwarf2out.c (insert_float): Change return type from void to
27928 unsigned, handle GET_MODE_SIZE (mode) == 2 and return element size.
27929 (mem_loc_descriptor, loc_descriptor, add_const_value_attribute):
27932 2021-03-20 H.J. Lu <hjl.tools@gmail.com>
27935 * config/i386/i386.c (construct_container): Check cfun != NULL
27936 before accessing silent_p.
27938 2021-03-20 Ahamed Husni <ahamedhusni73@gmail.com>
27940 * asan.c: Fix typos in comments.
27942 2021-03-20 Vladimir N. Makarov <vmakarov@redhat.com>
27944 PR rtl-optimization/99680
27945 * lra-constraints.c (skip_contraint_modifiers): Rename to skip_constraint_modifiers.
27946 (process_address_1): Check empty constraint before using
27949 2021-03-19 Pat Haugen <pthaugen@linux.ibm.com>
27951 * config/rs6000/rs6000.c (power10_cost): New.
27952 (rs6000_option_override_internal): Set Power10 costs.
27953 (rs6000_issue_rate): Set Power10 issue rate.
27954 * config/rs6000/power10.md: Rewrite for Power10.
27956 2021-03-19 Vladimir N. Makarov <vmakarov@redhat.com>
27959 * lra-constraints.c (process_address_1): Don't use unknown
27960 constraint for address constraint.
27962 2021-03-19 Iain Sandoe <iain@sandoe.co.uk>
27965 * config.gcc (powerpc-*-darwin8): Delete the reference to
27966 the now removed darwin8.h.
27968 2021-03-19 Olivier Hainque <hainque@adacore.com>
27971 * config/vxworksae.h (VX_CPU_PREFIX): Define.
27973 2021-03-19 John David Anglin <danglin@gcc.gnu.org>
27975 * config/pa/pa.c (import_milli): Use memcpy instead of strncpy.
27977 2021-03-19 Tamar Christina <tamar.christina@arm.com>
27979 PR tree-optimization/99656
27980 * tree-vect-slp-patterns.c (linear_loads_p,
27981 complex_add_pattern::matches, is_eq_or_top,
27982 vect_validate_multiplication, complex_mul_pattern::matches,
27983 complex_fms_pattern::matches): Remove complex_perm_kinds_t.
27984 * tree-vectorizer.h: (complex_load_perm_t): Removed.
27985 (slp_tree_to_load_perm_map_t): Use complex_perm_kinds_t instead of
27986 complex_load_perm_t.
27988 2021-03-19 H.J. Lu <hjl.tools@gmail.com>
27991 * config/i386/i386-options.c (ix86_init_machine_status): Set
27993 * config/i386/i386.c (init_cumulative_args): Set silent_p to
27995 (construct_container): Return early for return and argument
27996 errors if silent_p is true.
27997 * config/i386/i386.h (machine_function): Add silent_p.
27999 2021-03-19 Jakub Jelinek <jakub@redhat.com>
28002 * config/arm/constraints.md (Ds): New constraint.
28003 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): Use w,Ds
28004 constraint instead of w,Dm.
28006 2021-03-19 Andrew Stubbs <ams@codesourcery.com>
28008 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Fix quotes
28011 2021-03-19 Eric Botcazou <ebotcazou@adacore.com>
28013 PR middle-end/99641
28014 * fold-const.c (native_encode_initializer) <CONSTRUCTOR>: For an
28015 array type, do the computation of the current position in sizetype.
28017 2021-03-18 Vladimir N. Makarov <vmakarov@redhat.com>
28020 * lra-constraints.c (process_address_1): Use lookup_constraint
28021 only for a single constraint.
28023 2021-03-18 Martin Sebor <msebor@redhat.com>
28025 PR middle-end/99502
28026 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): Rename...
28027 (inbounds_memaccess_p): ...to this. Check the ending offset of
28028 the accessed member.
28030 2021-03-18 Andrew Stubbs <ams@codesourcery.com>
28032 * config/gcn/gcn.c (gcn_parse_amdgpu_hsa_kernel_attribute): Add %< and
28033 %> quote markers to error messages.
28034 (gcn_goacc_validate_dims): Likewise.
28035 (gcn_conditional_register_usage): Remove exclaimation mark from error
28037 (gcn_vectorize_vec_perm_const): Ensure perm is fully uninitialized.
28039 2021-03-18 Jan Hubicka <hubicka@ucw.cz>
28041 * config/i386/x86-tune-costs.h (struct processor_costs): Fix costs of
28044 2021-03-18 Sinan Lin <sinan@isrc.iscas.ac.cn>
28045 Kito Cheng <kito.cheng@sifive.com>
28047 * config/riscv/riscv.c (riscv_block_move_straight): Change type
28048 to unsigned HOST_WIDE_INT for parameter and local variable with
28049 HOST_WIDE_INT type.
28050 (riscv_adjust_block_mem): Ditto.
28051 (riscv_block_move_loop): Ditto.
28052 (riscv_expand_block_move): Ditto.
28054 2021-03-18 Nick Clifton <nickc@redhat.com>
28056 * config/v850/v850.c (construct_restore_jr): Increase static
28058 (construct_save_jarl): Likewise.
28059 * config/v850/v850.h (DWARF2_DEBUGGING_INFO): Define.
28061 2021-03-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28063 * config/aarch64/aarch64.c (aarch64_adjust_generic_arch_tuning): Define.
28064 (aarch64_override_options_internal): Use it.
28065 (generic_tunings): Add AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS to
28068 2021-03-17 Sandra Loosemore <sandra@codesourcery.com>
28070 * config/nios2/nios2.c (nios2_custom_check_insns): Clean up
28071 error message format issues.
28072 (nios2_option_override): Likewise.
28073 (nios2_expand_fpu_builtin): Likewise.
28074 (nios2_init_custom_builtins): Adjust to avoid bogus strncpy
28075 truncation warning.
28076 (nios2_expand_custom_builtin): More error message format fixes.
28077 (nios2_expand_rdwrctl_builtin): Likewise.
28078 (nios2_expand_rdprs_builtin): Likewise.
28079 (nios2_expand_eni_builtin): Likewise.
28080 (nios2_expand_builtin): Likewise.
28081 (nios2_register_custom_code): Likewise.
28082 (nios2_valid_target_attribute_rec): Likewise.
28083 (nios2_add_insn_asm): Fix uninitialized variable warning.
28085 2021-03-17 Jan Hubicka <jh@suse.cz>
28087 * config/i386/x86-tune-costs.h (struct processor_costs): Update costs
28088 of gather to match reality.
28089 * config/i386/x86-tune.def (X86_TUNE_USE_GATHER): Enable for znver3.
28091 2021-03-17 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28093 * config/aarch64/aarch64-builtins.c (aarch64_expand_rng_builtin): Use EQ
28094 to compare against CC_REG rather than NE.
28096 2021-03-17 H.J. Lu <hjl.tools@gmail.com>
28099 * config/i386/i386.c (ix86_force_load_from_GOT_p): Support
28100 inline assembly statements.
28101 (ix86_print_operand): Update 'P' handling for -fno-plt.
28103 2021-03-17 Tamar Christina <tamar.christina@arm.com>
28106 * config/aarch64/aarch64.c
28107 (aarch64_simd_clone_compute_vecsize_and_simdlen): Remove unused var.
28109 2021-03-16 Segher Boessenkool <segher@kernel.crashing.org>
28112 * config/rs6000/predicates.md (branch_comparison_operator): Allow
28113 ordered and unordered for CCFPmode, if flag_finite_math_only.
28115 2021-03-16 Jakub Jelinek <jakub@redhat.com>
28118 * config/i386/i386-expand.c (ix86_split_lea_for_addr): Emit a MULT
28119 rather than ASHIFT.
28120 * config/i386/i386.md (mult by 1248 into ashift): New splitter.
28122 2021-03-16 Martin Liska <mliska@suse.cz>
28125 * optc-save-gen.awk: Add flag_ipa_ra to exceptions for
28126 cl_optimization_compare function.
28128 2021-03-16 Ilya Leoshkevich <iii@linux.ibm.com>
28130 * config/s390/s390.c (f_constraint_p): Treat "fv" constraints
28133 2021-03-16 Jakub Jelinek <jakub@redhat.com>
28136 * config/i386/i386.h (struct machine_function): Add
28137 has_explicit_vzeroupper bitfield.
28138 * config/i386/i386-expand.c (ix86_expand_builtin): Set
28139 cfun->machine->has_explicit_vzeroupper when expanding
28140 IX86_BUILTIN_VZEROUPPER.
28141 * config/i386/i386-features.c (rest_of_handle_insert_vzeroupper):
28142 Do the mode switching only when TARGET_VZEROUPPER, expensive
28143 optimizations turned on and not optimizing for size.
28144 (pass_insert_vzeroupper::gate): Enable even when
28145 cfun->machine->has_explicit_vzeroupper is set.
28147 2021-03-16 Jakub Jelinek <jakub@redhat.com>
28150 * config/aarch64/aarch64.c
28151 (aarch64_simd_clone_compute_vecsize_and_simdlen): If not a function
28152 definition, walk TYPE_ARG_TYPES list if non-NULL for argument types
28153 instead of DECL_ARGUMENTS. Ignore types for uniform arguments.
28155 2021-03-15 Richard Biener <rguenther@suse.de>
28157 PR tree-optimization/98834
28158 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle missing
28159 subsetting by truncating the access size.
28161 2021-03-15 Jan Hubicka <hubicka@ucw.cz>
28163 * config/i386/i386-options.c (processor_cost_table): Add znver3_cost.
28164 * config/i386/x86-tune-costs.h (znver3_cost): New gobal variable; copy
28167 2021-03-15 Martin Liska <mliska@suse.cz>
28169 * spellcheck.c: Add missing comma in initialization.
28171 2021-03-14 Uroš Bizjak <ubizjak@gmail.com>
28173 * config/i386/sse.md (*vec_extract<mode>): Merge alternative 0 with
28174 alternative 2 and alternative 1 with alternative 3 using
28175 YW register constraint.
28176 (*vec_extract<PEXTR_MODE12:mode>_zext): Merge alternatives
28177 using YW register constraint.
28178 (*vec_extractv16qi_zext): Ditto.
28179 (*vec_extractv4si): Merge alternatives 4 and 5
28180 using Yw register constraint.
28181 (*ssse3_palignr<mode>_perm): Use Yw instead of v for alternative 3.
28183 2021-03-13 Martin Sebor <msebor@redhat.com>
28185 PR tree-optimization/99489
28186 * builtins.c (gimple_call_alloc_size): Fail gracefully when argument
28187 is not a call statement.
28189 2021-03-13 Jakub Jelinek <jakub@redhat.com>
28191 PR tree-optimization/99544
28192 * match.pd (X + (X << C) -> X * (1 + (1 << C))): Don't simplify
28193 if for vector types multiplication can't be done in type's mode.
28195 2021-03-12 Eric Botcazou <ebotcazou@adacore.com>
28198 * config/sparc/constraints.md (w): Rename to...
28199 (W): ... this and ditch previous implementation.
28200 * config/sparc/sparc.md (*movdi_insn_sp64): Replace W with m.
28201 (*movdf_insn_sp64): Likewise.
28202 (*mov<VM64:mode>_insn_sp64): Likewise.
28203 * config/sparc/sync.md (*atomic_compare_and_swap<mode>_1): Replace
28205 (atomic_compare_and_swap_leon3_1): Likewise.
28206 (*atomic_compare_and_swapdi_v8plus): Likewise.
28207 * config/sparc/sparc.c (memory_ok_for_ldd): Remove useless test on
28208 architecture and add missing address validity check during LRA.
28210 2021-03-12 Tobias Burnus <tobias@codesourcery.com>
28213 * gimplify.c (omp_add_variable): Handle NULL_TREE as size
28214 occuring for assumed-size arrays in use_device_{ptr,addr}.
28216 2021-03-12 Jakub Jelinek <jakub@redhat.com>
28219 * config/i386/constraints.md (YW): New internal constraint.
28220 * config/i386/sse.md (v_Yw): Add V4TI, V2TI, V1TI and TI cases.
28221 (*<sse2_avx2>_<insn><mode>3<mask_name>,
28222 *<sse2_avx2>_uavg<mode>3<mask_name>, *abs<mode>2,
28223 *<s>mul<mode>3_highpart<mask_name>): Use <v_Yw> instead of v in
28225 (<sse2_avx2>_psadbw): Use YW instead of v in constraints.
28226 (*avx2_pmaddwd, *sse2_pmaddwd, *<code>v8hi3, *<code>v16qi3,
28227 avx2_pmaddubsw256, ssse3_pmaddubsw128): Merge last two alternatives
28228 into one, use Yw instead of former x,v.
28229 (ashr<mode>3, <insn><mode>3): Use <v_Yw> instead of x in constraints of
28230 the last alternative.
28231 (<sse2_avx2>_packsswb<mask_name>, <sse2_avx2>_packssdw<mask_name>,
28232 <sse2_avx2>_packuswb<mask_name>, <sse4_1_avx2>_packusdw<mask_name>,
28233 *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, <ssse3_avx2>_palignr<mode>,
28234 <ssse3_avx2>_pshufb<mode>3<mask_name>): Merge last two alternatives
28235 into one, use <v_Yw> instead of former x,v.
28236 (avx2_interleave_highv32qi<mask_name>,
28237 vec_interleave_highv16qi<mask_name>): Use Yw instead of v in
28238 constraints. Add && <mask_avx512bw_condition> to condition.
28239 (avx2_interleave_lowv32qi<mask_name>,
28240 vec_interleave_lowv16qi<mask_name>,
28241 avx2_interleave_highv16hi<mask_name>,
28242 vec_interleave_highv8hi<mask_name>,
28243 avx2_interleave_lowv16hi<mask_name>, vec_interleave_lowv8hi<mask_name>,
28244 avx2_pshuflw_1<mask_name>, sse2_pshuflw_1<mask_name>,
28245 avx2_pshufhw_1<mask_name>, sse2_pshufhw_1<mask_name>,
28246 avx2_<code>v16qiv16hi2<mask_name>, sse4_1_<code>v8qiv8hi2<mask_name>,
28247 *sse4_1_<code>v8qiv8hi2<mask_name>_1, <sse2_avx2>_<insn><mode>3): Use
28248 Yw instead of v in constraints.
28249 * config/i386/mmx.md (Yv_Yw): New define_mode_attr.
28250 (*mmx_<insn><mode>3, mmx_ashr<mode>3, mmx_<insn><mode>3): Use <Yv_Yw>
28251 instead of Yv in constraints.
28252 (*mmx_<insn><mode>3, *mmx_mulv4hi3, *mmx_smulv4hi3_highpart,
28253 *mmx_umulv4hi3_highpart, *mmx_pmaddwd, *mmx_<code>v4hi3,
28254 *mmx_<code>v8qi3, mmx_pack<s_trunsuffix>swb, mmx_packssdw,
28255 mmx_punpckhbw, mmx_punpcklbw, mmx_punpckhwd, mmx_punpcklwd,
28256 *mmx_uavgv8qi3, *mmx_uavgv4hi3, mmx_psadbw): Use Yw instead of Yv in
28258 (*mmx_pinsrw, *mmx_pinsrb, *mmx_pextrw, *mmx_pextrw_zext, *mmx_pextrb,
28259 *mmx_pextrb_zext): Use YW instead of Yv in constraints.
28260 (*mmx_eq<mode>3, mmx_gt<mode>3): Use x instead of Yv in constraints.
28261 (mmx_andnot<mode>3, *mmx_<code><mode>3): Split last alternative into
28262 two, one with just x, another isa avx512vl with v.
28264 2021-03-12 Martin Liska <mliska@suse.cz>
28266 * doc/invoke.texi: Add missing param documentation.
28268 2021-03-11 David Malcolm <dmalcolm@redhat.com>
28271 * Makefile.in (ANALYZER_OBJS): Add analyzer/feasible-graph.o and
28272 analyzer/trimmed-graph.o.
28273 * doc/analyzer.texi (Analyzer Paths): Rewrite description of
28274 feasibility checking to reflect new implementation.
28275 * doc/invoke.texi (-fdump-analyzer-feasibility): Document new
28277 * shortest-paths.h (shortest_paths::get_shortest_distance): New.
28279 2021-03-11 David Malcolm <dmalcolm@redhat.com>
28281 * digraph.cc (selftest::test_shortest_paths): Update
28282 shortest_paths init for new param. Add test of
28283 SPS_TO_GIVEN_TARGET.
28284 * shortest-paths.h (enum shortest_path_sense): New.
28285 (shortest_paths::shortest_paths): Add "sense" param.
28286 Update for renamings. Generalize to use "sense" param.
28287 (shortest_paths::get_shortest_path): Rename param.
28288 (shortest_paths::m_sense): New field.
28289 (shortest_paths::m_prev): Rename...
28290 (shortest_paths::m_best_edge): ...to this.
28291 (shortest_paths::get_shortest_path): Update for renamings.
28292 Conditionalize flipping of path on sense of traversal.
28294 2021-03-11 David Malcolm <dmalcolm@redhat.com>
28296 * digraph.cc (selftest::test_shortest_paths): Add test coverage
28297 for paths from B and C.
28298 * shortest-paths.h (shortest_paths::shortest_paths): Handle
28299 unreachable nodes, rather than asserting.
28301 2021-03-11 David Edelsohn <dje.gcc@gmail.com>
28304 * config/rs6000/rs6000.c (rs6000_xcoff_file_start): Don't create
28305 xcoff_tbss_section_name.
28306 * config/rs6000/xcoff.h (ASM_OUTPUT_TLS_COMMON): Use .lcomm.
28307 * xcoffout.c (xcoff_tbss_section_name): Delete.
28308 * xcoffout.h (xcoff_tbss_section_name): Delete.
28310 2021-03-11 Richard Biener <rguenther@suse.de>
28312 PR tree-optimization/99523
28313 * tree-cfg.c (dump_function_to_file): Dump SSA names
28314 w/o identifier to the decls section as well, not only those
28315 without a VAR_DECL.
28317 2021-03-11 Jakub Jelinek <jakub@redhat.com>
28320 * ipa-icf-gimple.c (func_checker::compare_gimple_call): For internal
28321 function calls with lhs fail if the lhs don't have compatible types.
28323 2021-03-11 Hans-Peter Nilsson <hp@axis.com>
28325 * config/cris/cris.h (HARD_FRAME_POINTER_REGNUM): Define.
28326 Change FRAME_POINTER_REGNUM to correspond to a new faked
28327 register faked_fp, part of GENNONACR_REGS like faked_ap.
28328 (CRIS_FAKED_REGS_CONTENTS): New helper macro.
28329 (FIRST_PSEUDO_REGISTER, FIXED_REGISTERS, CALL_USED_REGISTERS):
28330 (REG_ALLOC_ORDER, REG_CLASS_CONTENTS, REGNO_OK_FOR_BASE_P)
28331 (ELIMINABLE_REGS, REGISTER_NAMES): Adjust accordingly.
28332 * config/cris/cris.md (CRIS_FP_REGNUM): Renumber to new faked
28334 (CRIS_REAL_FP_REGNUM): New constant.
28335 * config/cris/cris.c (cris_reg_saved_in_regsave_area): Check
28336 for HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM.
28337 (cris_initial_elimination_offset): Handle elimination changes
28338 to HARD_FRAME_POINTER_REGNUM instead of FRAME_POINTER_REGNUM
28339 and add one from FRAME_POINTER_REGNUM to
28340 HARD_FRAME_POINTER_REGNUM.
28341 (cris_expand_prologue, cris_expand_epilogue): Emit code for
28342 hard_frame_pointer_rtx instead of frame_pointer_rtx.
28344 2021-03-10 David Edelsohn <dje.gcc@gmail.com>
28347 * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Add check for DCmode.
28348 * config/rs6000/rs6000.c (rs6000_special_round_type_align): Same.
28350 2021-03-10 Vladimir N. Makarov <vmakarov@redhat.com>
28353 * lra-constraints.c (process_address_1): Don't check unknown
28354 constraint, use X for empty constraint.
28356 2021-03-10 Alex Coplan <alex.coplan@arm.com>
28358 * config/aarch64/aarch64.c (aarch64_vfp_is_call_or_return_candidate):
28359 Fix typo in comment describing "is_ha" argument.
28361 2021-03-10 John David Anglin <danglin@gcc.gnu.org>
28363 * doc/sourcebuild.texi: Document LRA target selector.
28365 2021-03-10 David Malcolm <dmalcolm@redhat.com>
28367 * doc/ux.texi: Add subsection contrasting interactive versus
28368 batch usage of GCC.
28370 2021-03-10 Joel Hutton <joel.hutton@arm.com>
28373 * tree-vect-stmts.c (vectorizable_store): Fix scatter store mask
28375 (vectorizable_load): Fix gather load mask check condition.
28377 2021-03-10 Richard Biener <rguenther@suse.de>
28379 PR tree-optimization/99510
28380 * tree.c (check_aligned_type): Check that the candidate
28381 has TYPE_USER_ALIGN set instead of matching with the
28384 2021-03-10 Eric Botcazou <ebotcazou@adacore.com>
28386 * config/sparc/sparc.c (sparc_regmode_natural_size): Return 4 for
28387 float and vector integer modes only if the mode is not larger.
28389 2021-03-10 Hans-Peter Nilsson <hp@axis.com>
28391 * config/cris/cris.h (DWARF_FRAME_REGISTERS): Define.
28393 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
28395 * ira.c (ira_setup_alts, ira_get_dup_out_num): Process digital
28397 * ira-lives.c (single_reg_class): Ditto.
28399 2021-03-09 Sebastian Huber <sebastian.huber@embedded-brains.de>
28401 * config.gcc (aarch64-*-rtems*): Include general rtems.h after
28402 the architecture-specific rtems.h.
28403 (aarch64-*-rtems*): Likewise.
28404 (arm*-*-rtems*): Likewise.
28405 (epiphany-*-rtems*): Likewise.
28406 (riscv*-*-rtems*): Likewise.
28408 2021-03-09 Jakub Jelinek <jakub@redhat.com>
28410 PR tree-optimization/99305
28411 * tree-ssa-phiopt.c (conditional_replacement): Test integer_pow2p
28412 before integer_all_onesp instead of vice versa.
28414 2021-03-09 Richard Earnshaw <rearnsha@arm.com>
28416 * common/config/arm/arm-common.c (arm_config_default): Change type
28417 of 'i' to unsigned.
28419 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
28422 * lra-constraints.c (process_address_1): Process constraint 'g'
28423 separately and digital constraints containing more one digit.
28425 2021-03-09 Nick Clifton <nickc@redhat.com>
28427 * config/rx/rx.h (DBX_DEBUGGING_INFO): Define.
28428 (DWARF"_DEBUGGING_INFO): Define.
28430 2021-03-09 Eric Botcazou <ebotcazou@adacore.com>
28433 * calls.c (initialize_argument_information): When the argument
28434 is passed by reference, do not make a copy in a thunk only if
28435 the argument is already in memory. Remove redundant test for
28436 the case of callee copy.
28438 2021-03-09 Vladimir N. Makarov <vmakarov@redhat.com>
28441 * lra-constraints.c (process_address_1): Process 0..9 constraints
28442 in process_address_1.
28444 2021-03-09 Andreas Krebbel <krebbel@linux.ibm.com>
28446 * config/s390/s390.c (struct s390_processor processor_table):
28447 Binutils name string must not be empty.
28449 2021-03-09 Claudiu Zissulescu <claziss@synopsys.com>
28451 * config/arc/arc.c (arc_attr_type): Remove function.
28453 2021-03-09 Martin Liska <mliska@suse.cz>
28456 * config/i386/i386-options.c (ix86_option_override_internal):
28457 Set isa_flags for OPTS argument and not for the global
28460 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
28462 * config/rs6000/predicates.md (ds_form_mem_operand): Check
28465 2021-03-09 Aaron Sawdey <acsawdey@linux.ibm.com>
28468 * config/rs6000/predicates.md (ds_form_mem_operand) New
28470 * config/rs6000/genfusion.pl (gen_ld_cmpi_p10) Use
28471 ds_form_mem_operand in ld/lwa patterns.
28472 * config/rs6000/fusion.md: Regenerate file.
28474 2021-03-08 Martin Sebor <msebor@redhat.com>
28476 PR middle-end/98266
28477 * gimple-array-bounds.cc (inbounds_vbase_memaccess_p): New function.
28478 (array_bounds_checker::check_array_bounds): Call it.
28480 2021-03-08 Martin Sebor <msebor@redhat.com>
28482 PR middle-end/97631
28483 * tree-ssa-strlen.c (maybe_warn_overflow): Test rawmem.
28484 (handle_builtin_stxncpy_strncat): Rename locals. Determine
28485 destination size from allocation calls. Issue a more appropriate
28487 (handle_builtin_memcpy): Pass true as rawmem to maybe_warn_overflow.
28488 (handle_builtin_memset): Same.
28490 2021-03-08 Peter Bergner <bergner@linux.ibm.com>
28493 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_permute): Add an assert
28494 to ensure we do not have an Altivec style address.
28495 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): Disable if passed
28496 an Altivec style address.
28497 (*vsx_le_perm_store_<mode>): Likewise.
28498 (splitters after *vsx_le_perm_store_<mode>): Likewise.
28499 (vsx_load_<mode>): Disable special expander if passed an Altivec
28501 (vsx_store_<mode>): Likewise.
28503 2021-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28506 * config/aarch64/predicates.md (aarch64_simd_shift_imm_vec_qi): Define.
28507 (aarch64_simd_shift_imm_vec_hi): Likewise.
28508 (aarch64_simd_shift_imm_vec_si): Likewise.
28509 (aarch64_simd_shift_imm_vec_di): Likewise.
28510 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le): Use
28511 predicate from above.
28512 (aarch64_shrn<mode>_insn_be): Likewise.
28513 (aarch64_rshrn<mode>_insn_le): Likewise.
28514 (aarch64_rshrn<mode>_insn_be): Likewise.
28515 (aarch64_shrn2<mode>_insn_le): Likewise.
28516 (aarch64_shrn2<mode>_insn_be): Likewise.
28517 (aarch64_rshrn2<mode>_insn_le): Likewise.
28518 (aarch64_rshrn2<mode>_insn_be): Likewise.
28520 2021-03-08 Vladimir N. Makarov <vmakarov@redhat.com>
28523 * lra-constraints.c (skip_contraint_modifiers): New function.
28524 (process_address_1): Use it before lookup_constraint call.
28526 2021-03-08 Martin Liska <mliska@suse.cz>
28529 * config/i386/i386-options.c (ix86_option_override_internal):
28530 Enable UINTR and HRESET for -march that supports it.
28532 2021-03-08 Ilya Leoshkevich <iii@linux.ibm.com>
28534 * config/s390/s390.c (f_constraint_p): New function.
28535 (s390_md_asm_adjust): Implement TARGET_MD_ASM_ADJUST.
28536 (TARGET_MD_ASM_ADJUST): Likewise.
28538 2021-03-08 Tobias Burnus <tobias@codesourcery.com>
28541 * tree-nested.c (convert_local_reference_stmt): Avoid calling
28542 lookup_field_for_decl for Fortran module (= namespace context).
28544 2021-03-08 Andreas Krebbel <krebbel@linux.ibm.com>
28546 * config/s390/s390.c (s390_expand_vec_compare): Implement <0
28547 comparison with arithmetic right shift.
28548 (s390_expand_vcond): No need for a force_reg anymore.
28549 s390_vec_compare will do it.
28550 * config/s390/vector.md ("vec_cmp<mode><tointvec>"): Accept also
28551 immediate operands.
28553 2021-03-07 Jakub Jelinek <jakub@redhat.com>
28556 * config/i386/constraints.md (Yw): Use SSE_REGS if TARGET_SSE
28557 but TARGET_AVX512BW or TARGET_AVX512VL is not set. Adjust description
28559 * config/i386/sse.md (v_Yw): New define_mode_attr.
28560 (*<insn><mode>3, *mul<mode>3<mask_name>, *avx2_<code><mode>3,
28561 *sse4_1_<code><mode>3<mask_name>): Use <v_Yw> instead of v
28563 * config/i386/mmx.md (mmx_pshufw_1, *vec_dupv4hi): Use Yw instead of
28564 xYw in constraints.
28566 2021-03-06 Julian Brown <julian@codesourcery.com>
28568 * tree-pretty-print.c (dump_generic_node): Emit non-generic
28569 address space info for aggregates.
28571 2021-03-06 Hans-Peter Nilsson <hp@axis.com>
28573 * config/cris/cris.h (MAX_FIXED_MODE_SIZE): Don't define.
28575 2021-03-05 Jakub Jelinek <jakub@redhat.com>
28577 PR middle-end/99322
28578 * tree-cfg.c (bb_to_omp_idx): New variable.
28579 (execute_build_cfg): Release the bb_to_omp_idx vector after
28580 cleanup_tree_cfg returns.
28581 (handle_abnormal_edges): Remove bb_to_omp_idx argument, adjust
28582 for bb_to_omp_idx being a vec<int> instead of pointer to array
28584 (make_edges): Remove bb_to_omp_idx local variable, don't pass
28585 it to handle_abnormal_edges, adjust for bb_to_omp_idx being a
28586 vec<int> instead of pointer to array of ints and don't free/release
28588 (remove_bb): When removing a bb and placing forced label somewhere
28589 else, ensure it is put into the same OpenMP region during cfg
28590 pass if possible or to entry successor as fallback. Unregister
28591 bb from bb_to_omp_idx.
28593 2021-03-05 Vladimir N. Makarov <vmakarov@redhat.com>
28596 * lra-constraints.c (process_address_1): Skip decomposing address
28597 for asm insn operand with unknown constraint.
28599 2021-03-05 Martin Jambor <mjambor@suse.cz>
28602 * cgraph.c (cgraph_edge::set_call_stmt): Do not update all
28603 corresponding speculative edges if we are about to resolve
28604 sepculation. Make edge direct (and so resolve speculations) before
28605 removing it from call_site_hash.
28606 (cgraph_edge::make_direct): Relax the initial assert to allow calling
28607 the function on speculative direct edges.
28609 2021-03-05 Eric Botcazou <ebotcazou@adacore.com>
28611 PR rtl-optimization/99376
28612 * rtlanal.c (nonzero_bits1) <arithmetic operators>: If the number
28613 of low-order zero bits is too large, set the result to 0 directly.
28615 2021-03-04 Jakub Jelinek <jakub@redhat.com>
28617 PR middle-end/93235
28618 * expmed.c (store_bit_field_using_insv): Return false of xop0 is a
28619 SUBREG and a SUBREG to op_mode can't be created.
28621 2021-03-04 Alex Coplan <alex.coplan@arm.com>
28624 * config/aarch64/aarch64-sve-builtins.cc
28625 (function_resolver::require_vector_type): Handle error_mark_node.
28627 2021-03-04 Ilya Leoshkevich <iii@linux.ibm.com>
28629 * cfgexpand.c (expand_asm_loc): Pass new parameter.
28630 (expand_asm_stmt): Likewise.
28631 * config/arm/aarch-common-protos.h (arm_md_asm_adjust): Add new
28633 * config/arm/aarch-common.c (arm_md_asm_adjust): Likewise.
28634 * config/arm/arm.c (thumb1_md_asm_adjust): Likewise.
28635 * config/cris/cris.c (cris_md_asm_adjust): Likewise.
28636 * config/i386/i386.c (ix86_md_asm_adjust): Likewise.
28637 * config/mn10300/mn10300.c (mn10300_md_asm_adjust): Likewise.
28638 * config/nds32/nds32.c (nds32_md_asm_adjust): Likewise.
28639 * config/pdp11/pdp11.c (pdp11_md_asm_adjust): Likewise.
28640 * config/rs6000/rs6000.c (rs6000_md_asm_adjust): Likewise.
28641 * config/vax/vax.c (vax_md_asm_adjust): Likewise.
28642 * config/visium/visium.c (visium_md_asm_adjust): Likewise.
28643 * doc/tm.texi (md_asm_adjust): Likewise.
28644 * target.def (md_asm_adjust): Likewise.
28646 2021-03-04 Richard Biener <rguenther@suse.de>
28648 PR middle-end/97855
28649 * tree-pretty-print.c: Poison pp_printf.
28650 (dump_decl_name): Avoid use of pp_printf.
28651 (dump_block_node): Likewise.
28652 (dump_generic_node): Likewise.
28654 2021-03-04 Martin Sebor <msebor@redhat.com>
28656 PR middle-end/96963
28657 PR middle-end/94655
28658 * builtins.c (handle_array_ref): New helper.
28659 (handle_mem_ref): New helper.
28660 (compute_objsize_r): Factor out ARRAY_REF and MEM_REF handling
28661 into new helper functions. Correct a workaround for vectorized
28664 2021-03-03 Pat Haugen <pthaugen@linux.ibm.com>
28666 * config/rs6000/dfp.md (extendddtd2, trunctddd2, *cmp<mode>_internal1,
28667 floatditd2, ftrunc<mode>2, fix<mode>di2, dfp_ddedpd_<mode>,
28668 dfp_denbcd_<mode>, dfp_dxex_<mode>, dfp_diex_<mode>,
28669 *dfp_sgnfcnc_<mode>, dfp_dscli_<mode>, dfp_dscri_<mode>): Update size
28670 attribute for Power10.
28671 * config/rs6000/mma.md (*movoo): Likewise.
28672 * config/rs6000/rs6000.md (define_attr "size"): Add 256.
28673 (define_mode_attr bits): Add DD/TD modes.
28674 * config/rs6000/sync.md (load_quadpti, store_quadpti, load_lockedpti,
28675 store_conditionalpti): Update size attribute for Power10.
28677 2021-03-03 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
28680 * config/sparc/t-sparc (tree-ssanames.o-warn): Don't error for
28681 -Wuninitialized, -Wmaybe-uninitialized.
28682 (wide-int.o-warn): Likewise.
28684 2021-03-03 Richard Earnshaw <rearnsha@arm.com>
28686 * common/config/arm/arm-common.c: Include configargs.h.
28687 (arm_config_default): New function.
28688 (arm_target_mode): Renamed from arm_target_thumb_only. Handle
28689 processors that do not support Thumb. Take into account the
28690 --with-mode configuration setting for selecting the default.
28691 * config/arm/arm.h (OPTION_DEFAULT_SPECS): Remove entry for 'mode'.
28692 (TARGET_MODE_SPEC_FUNCTIONS): Update for function name change.
28694 2021-03-03 Martin Liska <mliska@suse.cz>
28696 PR gcov-profile/97461
28697 * gcov-io.h (GCOV_PREALLOCATED_KVP): Remove.
28699 2021-03-03 Eric Botcazou <ebotcazou@adacore.com>
28702 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
28703 point back the hard frame pointer to its default location when the
28704 frame is larger than SEH_MAX_FRAME_SIZE.
28706 2021-03-03 Jakub Jelinek <jakub@redhat.com>
28709 * config/i386/predicates.md (logic_operator): New define_predicate.
28710 * config/i386/i386.md (mov + mem using comm arith peephole2):
28711 Punt if operands[1] is EXT_REX_SSE_REGNO_P, AVX512BW is not enabled
28712 and the inner mode is [QH]Imode.
28714 2021-03-03 Jakub Jelinek <jakub@redhat.com>
28717 * dwarf2out.c (dw_loc_list_struct): Add end_entry member.
28718 (new_loc_list): Clear end_entry.
28719 (output_loc_list): Only use DW_LLE_startx_length for -gsplit-dwarf
28720 if HAVE_AS_LEB128, otherwise use DW_LLE_startx_endx. Fix comment
28722 (index_location_lists): For dwarf_version >= 5 without HAVE_AS_LEB128,
28723 initialize also end_entry.
28725 2021-03-03 Jakub Jelinek <jakub@redhat.com>
28728 * cfgrtl.c (fixup_partitions): When changing some bbs from hot to cold
28729 partitions, if in non-layout mode after reorder_blocks also move
28730 affected blocks to ensure a single partition transition.
28732 2021-03-03 Jason Merrill <jason@redhat.com>
28735 * cgraphunit.c (process_function_and_variable_attributes): Don't
28736 warn about flatten on an alias if the target also has it.
28737 * cgraph.h (symtab_node::get_alias_target_tree): New.
28739 2021-03-02 David Edelsohn <dje.gcc@gmail.com>
28741 * config/rs6000/rs6000.md (tls_get_tpointer_internal): Prepend
28742 period to symbol name.
28743 (tls_get_addr_internal<mode>): Same.
28745 2021-03-02 David Malcolm <dmalcolm@redhat.com>
28748 * diagnostic-show-locus.c
28749 (selftest::test_one_liner_many_fixits_2): Fix accidental usage of
28752 2021-03-02 Martin Sebor <msebor@redhat.com>
28754 PR middle-end/99276
28755 * builtins.c (warn_for_access): Remove stray warning text.
28757 2021-03-02 Martin Sebor <msebor@redhat.com>
28759 PR middle-end/99295
28760 * doc/extend.texi (attribute malloc): Reword and clarify nonaliasing
28763 2021-03-02 Jakub Jelinek <jakub@redhat.com>
28766 * dwarf2out.c (output_macinfo_op): Use DW_MACRO_*_str* even with
28767 -gdwarf-5 -gstrict-dwarf. For -gsplit-dwarf -gdwarf-5 use
28768 DW_MACRO_*_strx instead of DW_MACRO_*_strp. Handle
28769 DW_MACRO_define_strx and DW_MACRO_undef_strx.
28770 (save_macinfo_strings): Use DW_MACRO_*_str* even with
28771 -gdwarf-5 -gstrict-dwarf. Handle DW_MACRO_define_strx and
28772 DW_MACRO_undef_strx.
28774 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
28776 * config/s390/s390-builtin-types.def (BT_FN_V4SF_V8HI_UINT): New
28778 (BT_FN_V8HI_V8HI_UINT): Likewise.
28779 (BT_FN_V8HI_V4SF_V4SF_UINT): Likewise.
28780 * config/s390/s390-builtins.def (B_NNPA): New macro definition.
28781 (s390_vclfnhs, s390_vclfnls, s390_vcrnfs, s390_vcfn, s390_vcnf):
28782 New builtin definitions.
28783 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Bump
28784 vector extension version.
28785 * config/s390/s390.c (s390_expand_builtin): Check if builtins are
28786 available with current -march level.
28787 * config/s390/s390.md (UNSPEC_NNPA_VCLFNHS_V8HI)
28788 (UNSPEC_NNPA_VCLFNLS_V8HI, UNSPEC_NNPA_VCRNFS_V8HI)
28789 (UNSPEC_NNPA_VCFN_V8HI, UNSPEC_NNPA_VCNF_V8HI): New constants.
28790 * config/s390/vecintrin.h (vec_extend_to_fp32_hi): New macro.
28791 (vec_extend_to_fp32_lo): Likewise.
28792 (vec_round_from_fp32): Likewise.
28793 (vec_convert_to_fp16): Likewise.
28794 (vec_convert_from_fp16): Likewise.
28795 * config/s390/vx-builtins.md (vclfnhs_v8hi): New insn pattern.
28796 (vclfnls_v8hi): Likewise.
28797 (vcrnfs_v8hi): Likewise.
28798 (vcfn_v8hi): Likewise.
28799 (vcnf_v8hi): Likewise.
28801 2021-03-02 Andreas Krebbel <krebbel@linux.ibm.com>
28803 * common/config/s390/s390-common.c (processor_flags_table): New entry.
28804 * config.gcc: Enable arch14 for --with-arch and --with-tune.
28805 * config/s390/driver-native.c (s390_host_detect_local_cpu): Pick
28806 arch14 for unknown CPU models.
28807 * config/s390/s390-opts.h (enum processor_type): Add PROCESSOR_ARCH14.
28808 * config/s390/s390.c (s390_issue_rate): Add case for PROCESSOR_ARCH14.
28809 (s390_get_sched_attrmask): Likewise.
28810 (s390_get_unit_mask): Likewise.
28811 * config/s390/s390.h (enum processor_flags): Add PF_NNPA and PF_ARCH14.
28812 (TARGET_CPU_ARCH14, TARGET_CPU_ARCH14_P, TARGET_CPU_NNPA)
28813 (TARGET_CPU_NNPA_P, TARGET_ARCH14, TARGET_ARCH14_P, TARGET_NNPA)
28814 (TARGET_NNPA_P): New macro definitions.
28815 * config/s390/s390.md ("cpu_facility", "enabled"): Add arch14 and nnpa.
28816 * config/s390/s390.opt: Add PROCESSOR_ARCH14.
28818 2021-03-02 Jakub Jelinek <jakub@redhat.com>
28820 PR middle-end/95757
28821 * tree-vrp.c (register_edge_assert_for): Remove superfluous ()s around
28822 condition. Call register_edge_assert_for_1 for == 0, != 0, == 1 and
28823 != 1 comparisons if name is lhs of a comparison.
28825 2021-03-01 Iain Sandoe <iain@sandoe.co.uk>
28829 * config/darwin-protos.h (darwin_should_restore_cfa_state): New.
28830 * config/darwin.c (darwin_should_restore_cfa_state): New.
28831 * config/darwin.h (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New.
28832 * doc/tm.texi: Regenerated.
28833 * doc/tm.texi.in: Document TARGET_ASM_SHOULD_RESTORE_CFA_STATE.
28834 * dwarf2cfi.c (connect_traces): If the target requests, restore
28835 the CFA expression after a DW_CFA_restore.
28836 * target.def (TARGET_ASM_SHOULD_RESTORE_CFA_STATE): New hook.
28838 2021-03-01 Martin Liska <mliska@suse.cz>
28841 * optc-save-gen.awk: Add 4 more exceptions.
28843 2021-03-01 Nathan Sidwell <nathan@acm.org>
28846 * tree.h (TYPE_ALIGN_RAW): New accessor.
28847 (TYPE_ALIGN): Use it.
28849 2021-03-01 Jan Hubicka <jh@suse.cz>
28852 * ipa-fnsummary.c (compute_fn_summary): Fix sanity check.
28854 2021-03-01 Eric Botcazou <ebotcazou@adacore.com>
28857 * config/i386/i386.c (ix86_compute_frame_layout): For a SEH target,
28858 point the hard frame pointer to the SSE register save area instead
28859 of the general register save area. Perform only minimal adjustment
28860 for small frames if it is initially not correctly aligned.
28861 (ix86_expand_prologue): Remove early saves for a SEH target.
28862 * config/i386/winnt.c (struct seh_frame_state): Document constraint.
28864 2021-02-28 Jakub Jelinek <jakub@redhat.com>
28867 * ipa.c (symbol_table::remove_unreachable_nodes): Fix a comment
28868 typo - referneced -> referenced.
28869 * tree.c (component_ref_size): Fix comment typo -
28870 refernce -> reference.
28871 * tree-ssa-alias.c (access_path_may_continue_p): Fix comment typo -
28872 traling -> trailing.
28873 (aliasing_component_refs_p): Fix comment typos -
28874 refernce -> reference and refernece -> reference and
28875 traling -> trailing.
28876 (nonoverlapping_refs_since_match_p): Fix comment typo -
28877 referneces -> references.
28878 * doc/invoke.texi (--param modref-max-bases): Fix a typo -
28879 referneces -> references.
28881 2021-02-27 Iain Sandoe <iain@sandoe.co.uk>
28883 * config/host-darwin.c (darwin_gt_pch_use_address): Modify
28884 diagnostic message to avoid use of a contraction and format
28887 2021-02-27 Jakub Jelinek <jakub@redhat.com>
28890 * gcse.c (gcse_or_cprop_is_too_expensive): Use %wu instead of
28891 HOST_WIDE_INT_PRINT_UNSIGNED in warning format string.
28892 * ipa-devirt.c (ipa_odr_read_section): Use %wd instead of
28893 HOST_WIDE_INT_PRINT_DEC in inform format string. Fix comment
28896 2021-02-26 Richard Biener <rguenther@suse.de>
28898 PR middle-end/99281
28899 * expr.c (store_field): For calls with return-slot optimization
28900 and addressable return type expand the store directly.
28902 2021-02-26 Richard Biener <rguenther@suse.de>
28905 * builtins.c (warn_string_no_nul): Fix diagnostic formatting.
28907 2021-02-26 Peter Bergner <bergner@linux.ibm.com>
28910 * config/rs6000/rs6000-call.c (rs6000_init_builtins): Replace assert
28913 2021-02-26 Aaron Sawdey <acsawdey@linux.ibm.com>
28915 * config.gcc: Add rs6000-pcrel-opt.o.
28916 * config/rs6000/rs6000-pcrel-opt.c: New file.
28917 * config/rs6000/pcrel-opt.md: New file.
28918 * config/rs6000/predicates.md: Add d_form_memory predicate.
28919 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_PCREL_OPT.
28920 * config/rs6000/rs6000-passes.def: Add pass_pcrel_opt.
28921 * config/rs6000/rs6000-protos.h: Add reg_to_non_prefixed(),
28922 pcrel_opt_valid_mem_p(), output_pcrel_opt_reloc(),
28923 and make_pass_pcrel_opt().
28924 * config/rs6000/rs6000.c (reg_to_non_prefixed): Make global.
28925 (rs6000_option_override_internal): Add pcrel-opt.
28926 (rs6000_delegitimize_address): Support pcrel-opt.
28927 (rs6000_opt_masks): Add pcrel-opt.
28928 (pcrel_opt_valid_mem_p): New function.
28929 (reg_to_non_prefixed): Make global.
28930 (rs6000_asm_output_opcode): Reset prepend_p_to_next_insn.
28931 (output_pcrel_opt_reloc): New function.
28932 * config/rs6000/rs6000.md (loads_extern_addr): New attr.
28933 (pcrel_extern_addr): Set loads_extern_addr.
28934 Add include for pcrel-opt.md.
28935 * config/rs6000/rs6000.opt: Add -mpcrel-opt.
28936 * config/rs6000/t-rs6000: Add rules for pcrel-opt.c and
28939 2021-02-26 YunQiang Su <yunqiang.su@cipunited.com>
28942 * config/mips/mips.c (mips_expand_ext_as_unaligned_load):
28943 If TARGET_64BIT and dest is SUBREG, we check the width, if it
28944 equal to SImode, we use SImode operation, just like what we are
28947 2021-02-26 Marek Polacek <polacek@redhat.com>
28949 * builtins.c (warn_for_access): Fix typos.
28951 2021-02-25 Iain Sandoe <iain@sandoe.co.uk>
28953 * config/aarch64/aarch64.md (<optab>_rol<mode>3): Add a '#'
28954 mark in front of the immediate quantity.
28955 (<optab>_rolsi3_uxtw): Likewise.
28957 2021-02-25 Richard Earnshaw <rearnsha@arm.com>
28960 * config/arm/thumb2.md (nonsecure_call_reg_thumb2_fpcxt): New pattern.
28961 (nonsecure_call_value_reg_thumb2_fpcxt): Likewise.
28962 (nonsecure_call_reg_thumb2): Restrict to using r4 for the callee
28963 address and disable when the FPCXT is not available.
28964 (nonsecure_call_value_reg_thumb2): Likewise.
28966 2021-02-25 Nathan Sidwell <nathan@acm.org>
28969 * doc/invoke.texi (flang-info-module-cmi): Renamed option.
28971 2021-02-25 Tamar Christina <tamar.christina@arm.com>
28973 * tree-vect-slp.c (optimize_load_redistribution_1): Abort on NULL nodes.
28975 2021-02-25 Richard Biener <rguenther@suse.de>
28977 PR tree-optimization/99253
28978 * tree-vect-loop.c (check_reduction_path): First compute
28979 code, then verify out-of-loop uses.
28981 2021-02-25 Jakub Jelinek <jakub@redhat.com>
28984 * match.pd ((T)(A) + CST -> (T)(A + CST)): Add :s to convert.
28986 2021-02-25 Jakub Jelinek <jakub@redhat.com>
28988 PR tree-optimization/80635
28989 * tree-vrp.c (vrp_simplify_cond_using_ranges): Also handle
28990 VIEW_CONVERT_EXPR if modes are the same, innerop is integral and
28991 has mode precision.
28993 2021-02-25 Richard Biener <rguenther@suse.de>
28995 * tree-vect-slp.c (optimize_load_redistribution_1): Delay
28996 load_map population.
28997 (vect_match_slp_patterns_2): Revert part of last change.
28998 (vect_analyze_slp): Do not interleave optimize_load_redistribution
28999 with pattern detection but do it afterwards. Dump the
29000 whole SLP graph after pattern recognition and load
29001 redistribution optimization finished.
29003 2021-02-24 Jakub Jelinek <jakub@redhat.com>
29006 * omp-low.c (struct omp_context): Add teams_nested_p and
29007 nonteams_nested_p members.
29008 (scan_omp_target): Diagnose teams nested inside of target with other
29009 directives strictly nested inside of the same target.
29010 (check_omp_nesting_restrictions): Set ctx->teams_nested_p or
29011 ctx->nonteams_nested_p as needed.
29013 2021-02-24 Vladimir N. Makarov <vmakarov@redhat.com>
29015 PR inline-asm/99123
29016 * lra-constraints.c (uses_hard_regs_p): Don't use decompose_mem_address.
29018 2021-02-24 Hans-Peter Nilsson <hp@axis.com>
29020 * config/cris/cris.c (cris_expand_prologue): Set
29021 current_function_static_stack_size, if flag_stack_usage_info.
29023 2021-02-24 Pat Haugen <pthaugen@linux.ibm.com>
29025 * config/rs6000/rs6000.c (next_insn_prefixed_p): Rename.
29026 (rs6000_final_prescan_insn): Adjust.
29027 (rs6000_asm_output_opcode): Likewise.
29029 2021-02-24 Martin Sebor <msebor@redhat.com>
29031 PR middle-end/97172
29032 * attribs.c (attr_access::free_lang_data): Clear attribute arg spec
29033 from function arguments.
29035 2021-02-24 Tamar Christina <tamar.christina@arm.com>
29037 PR tree-optimization/99220
29038 * tree-vect-slp.c (optimize_load_redistribution_1): Remove
29039 node from cache when it's about to be deleted.
29041 2021-02-24 Jakub Jelinek <jakub@redhat.com>
29043 PR tree-optimization/99225
29044 * fold-const.c (fold_binary_loc) <case NE_EXPR>: In (x & (1 << y)) != 0
29045 to ((x >> y) & 1) != 0 simplifications use build_one_cst instead of
29046 build_int_cst (..., 1). Formatting fixes.
29048 2021-02-24 Tamar Christina <tamar.christina@arm.com>
29050 PR tree-optimization/99149
29051 * tree-vect-slp-patterns.c (vect_detect_pair_op): Don't recreate the
29053 (vect_slp_reset_pattern): Remove.
29054 (complex_fma_pattern::matches): Remove call to vect_slp_reset_pattern.
29055 (complex_mul_pattern::build, complex_fma_pattern::build,
29056 complex_fms_pattern::build): Fix ref counts.
29057 * tree-vect-slp.c (vect_free_slp_tree): Undo SLP only pattern relevancy
29058 when node is being deleted.
29059 (vect_match_slp_patterns_2): Correct result of cache hit on patterns.
29060 (vect_schedule_slp): Invalidate SLP_TREE_REPRESENTATIVE of removed
29062 * tree-vectorizer.c (vec_info::new_stmt_vec_info): Initialize value.
29064 2021-02-24 Matthias Klose <doko@ubuntu.com>
29067 2020-12-07 Matthias Klose <doko@ubuntu.com>
29069 * genextract.c (print_header): Undefine ENABLE_RTL_CHECKING
29070 and ENABLE_RTL_FLAG_CHECKING.
29072 2021-02-24 Richard Biener <rguenther@suse.de>
29075 * builtins.c (fold_builtin_next_arg): Avoid NULL arg.
29077 2021-02-23 Peter Bergner <bergner@linux.ibm.com>
29079 * config/rs6000/mma.md (mma_assemble_pair): Rename from this...
29080 (vsx_assemble_pair): ...to this.
29081 (*mma_assemble_pair): Rename from this...
29082 (*vsx_assemble_pair): ...to this.
29083 (mma_disassemble_pair): Rename from this...
29084 (vsx_disassemble_pair): ...to this.
29085 (*mma_disassemble_pair): Rename from this...
29086 (*vsx_disassemble_pair): ...to this.
29087 * config/rs6000/rs6000-builtin.def (BU_MMA_V2, BU_MMA_V3,
29088 BU_COMPAT): New macros.
29089 (mma_assemble_pair): Rename from this...
29090 (vsx_assemble_pair): ...to this.
29091 (mma_disassemble_pair): Rename from this...
29092 (vsx_disassemble_pair): ...to this.
29093 (mma_assemble_pair): New compatibility built-in.
29094 (mma_disassemble_pair): Likewise.
29095 * config/rs6000/rs6000-call.c (struct builtin_compatibility): New.
29096 (RS6000_BUILTIN_COMPAT): Define.
29097 (bdesc_compat): New.
29098 (mma_expand_builtin): Use VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
29099 (rs6000_gimple_fold_mma_builtin): Use MMA_BUILTIN_DISASSEMBLE_PAIR
29100 and VSX_BUILTIN_ASSEMBLE_PAIR.
29101 (rs6000_init_builtins): Register compatibility built-ins.
29102 (mma_init_builtins): Use VSX_BUILTIN_ASSEMBLE_PAIR,
29103 VSX_BUILTIN_ASSEMBLE_PAIR_INTERNAL, VSX_BUILTIN_DISASSEMBLE_PAIR and
29104 VSX_BUILTIN_DISASSEMBLE_PAIR_INTERNAL.
29105 * doc/extend.texi (__builtin_mma_assemble_pair): Rename from this...
29106 (__builtin_vsx_assemble_pair): ...to this.
29107 (__builtin_mma_disassemble_pair): Rename from this...
29108 (__builtin_vsx_disassemble_pair): ...to this.
29110 2021-02-23 Martin Liska <mliska@suse.cz>
29113 * ipa-icf.c (sem_variable::merge): Do not merge 2 variables
29114 with different alignment. That leads to an invalid red zone
29115 size allocated in runtime.
29117 2021-02-23 Jakub Jelinek <jakub@redhat.com>
29119 PR tree-optimization/99204
29120 * fold-const.c (fold_read_from_constant_string): Check that
29121 tree_fits_uhwi_p (index) rather than just that index is INTEGER_CST.
29123 2021-02-23 Segher Boessenkool <segher@kernel.crashing.org>
29124 Kewen Lin <linkw@gcc.gnu.org>
29126 * config/rs6000/rs6000.md (*rotl<mode>3_insert_3): Renamed to...
29127 (rotl<mode>3_insert_3): ...this.
29128 (plus_ior_xor): New code_iterator.
29129 (define_split for GPR rl*imi): New splitter.
29130 * config/rs6000/vsx.md (vsx_init_v4si): Use gen_rotldi3_insert_3
29131 for integer merging.
29133 2021-02-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29135 * config/aarch64/aarch64-tuning-flags.def (cse_sve_vl_constants):
29137 * config/aarch64/aarch64.md (add<mode>3): Force CONST_POLY_INT immediates
29138 into a register when the above is enabled.
29139 * config/aarch64/aarch64.c (neoversev1_tunings):
29140 AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
29141 (aarch64_rtx_costs): Use AARCH64_EXTRA_TUNE_CSE_SVE_VL_CONSTANTS.
29143 2021-02-22 Hans-Peter Nilsson <hp@axis.com>
29145 * config/cris/cris.c (cris_print_operand) <'T'>: Change
29146 valid operand from is now an addi mult-value to shift-value.
29147 * config/cris/cris.md (*addi): Change expression of scaled
29148 operand from mult to ashift.
29149 * config/cris/cris.md (*addi_reload): New insn_and_split.
29151 2021-02-22 John David Anglin <danglin@gcc.gnu.org>
29154 * config/pa/pa.c (TARGET_ASM_CAN_OUTPUT_MI_THUNK): Define as
29155 hook_bool_const_tree_hwi_hwi_const_tree_true.
29156 (pa_asm_output_mi_thunk): Add support for nonzero vcall_offset.
29158 2021-02-22 Andre Vieira <andre.simoesdiasvieira@arm.com>
29160 PR rtl-optimization/98791
29161 * ira-conflicts.c (process_regs_for_copy): Don't create allocno copies
29162 for unordered modes.
29164 2021-02-22 Martin Liska <mliska@suse.cz>
29166 * tree-inline.c (inline_forbidden_p): Set
29167 inline_forbidden_reason.
29169 2021-02-22 Richard Biener <rguenther@suse.de>
29171 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Dump
29174 2021-02-22 Richard Biener <rguenther@suse.de>
29176 PR tree-optimization/99165
29177 * gimple-ssa-store-merging.c (pass_store_merging::process_store):
29178 Accumulate changed to ret.
29180 2021-02-21 Uros Bizjak <ubizjak@gmail.com>
29183 2020-12-09 Uroš Bizjak <ubizjak@gmail.com>
29185 * config/i386/i386.h (REG_ALLOC_ORDER): Remove
29187 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
29190 * config/s390/vector.md (trunctf<DFP_ALL:mode>2_vr): New
29192 (trunctf<DFP_ALL:mode>2): Likewise.
29193 (trunctdtf2_vr): Likewise.
29194 (trunctdtf2): Likewise.
29195 (extend<DFP_ALL:mode>tf2_vr): Likewise.
29196 (extend<DFP_ALL:mode>tf2): Likewise.
29197 (extendtftd2_vr): Likewise.
29198 (extendtftd2): Likewise.
29200 2021-02-20 Ilya Leoshkevich <iii@linux.ibm.com>
29202 * config/s390/vector.md (*fprx2_to_tf): Rename to fprx2_to_tf,
29203 add memory alternative.
29204 (tf_to_fprx2): New pattern.
29206 2021-02-19 Martin Sebor <msebor@redhat.com>
29209 * attribs.c (init_attr_rdwr_indices): Guard vblist use.
29210 (attr_access::free_lang_data): Remove a spurious test.
29212 2021-02-19 Nathan Sidwell <nathan@acm.org>
29214 * doc/invoke.texi (flang-info-module-read): Document.
29216 2021-02-19 Martin Liska <mliska@suse.cz>
29218 PR translation/99167
29219 * params.opt: Fix typo.
29221 2021-02-19 Richard Biener <rguenther@suse.de>
29223 PR middle-end/99122
29224 * tree-inline.c (inline_forbidden_p): Do not inline functions
29225 with VLA arguments or return value.
29227 2021-02-19 Jakub Jelinek <jakub@redhat.com>
29230 * config/arm/arm.md (*stack_protect_combined_set_insn,
29231 *stack_protect_combined_test_insn): If force_const_mem result
29232 is not valid general operand, force its address into the destination
29235 2021-02-19 Jakub Jelinek <jakub@redhat.com>
29238 * tree-cfg.c (gimple_merge_blocks): If bb a starts with eh landing
29239 pad or non-local label, put FORCED_LABELs from bb b after that label
29240 rather than before it.
29242 2021-02-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
29245 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3): Use
29246 expand_vector_broadcast' to emit the vec_duplicate operand.
29248 2021-02-18 Vladimir N. Makarov <vmakarov@redhat.com>
29250 PR rtl-optimization/96264
29251 * lra-remat.c (reg_overlap_for_remat_p): Check also output insn
29254 2021-02-18 H.J. Lu <hjl.tools@gmail.com>
29257 * varasm.c (get_section): Replace SUPPORTS_SHF_GNU_RETAIN with
29258 looking up the retain attribute.
29259 (resolve_unique_section): Likewise.
29260 (get_variable_section): Likewise.
29261 (switch_to_section): Likewise. Warn when a symbol without the
29262 retain attribute and a symbol with the retain attribute are
29263 placed in the section with the same name, instead of the used
29265 * doc/extend.texi: Document the "retain" attribute.
29267 2021-02-18 Nathan Sidwell <nathan@acm.org>
29270 * doc/invoke.texi (flang-info-include-translate): Document header
29273 2021-02-18 Richard Biener <rguenther@suse.de>
29275 PR middle-end/99122
29276 * ipa-fnsummary.c (analyze_function_body): Set
29277 CIF_FUNCTION_NOT_INLINABLE for VLA parameter calls.
29278 * tree-inline.c (insert_init_debug_bind): Pass NULL for
29279 error_mark_node values.
29280 (force_value_to_type): Do not build V_C_Es for WITH_SIZE_EXPR
29282 (setup_one_parameter): Delay force_value_to_type until when
29285 2021-02-18 Hans-Peter Nilsson <hp@axis.com>
29287 PR tree-optimization/99142
29288 * match.pd (clz cmp 0): Gate replacement on single_use of clz result.
29290 2021-02-18 Jakub Jelinek <jakub@redhat.com>
29292 * wide-int-bitmask.h (wide_int_bitmask::wide_int_bitmask (),
29293 wide_int_bitmask::wide_int_bitmask (uint64_t),
29294 wide_int_bitmask::wide_int_bitmask (uint64_t, uint64_t),
29295 wide_int_bitmask::operator ~ () const,
29296 wide_int_bitmask::operator | (wide_int_bitmask) const,
29297 wide_int_bitmask::operator & (wide_int_bitmask) const): Use constexpr
29299 * config/i386/i386.h (PTA_3DNOW, PTA_3DNOW_A, PTA_64BIT, PTA_ABM,
29300 PTA_AES, PTA_AVX, PTA_BMI, PTA_CX16, PTA_F16C, PTA_FMA, PTA_FMA4,
29301 PTA_FSGSBASE, PTA_LWP, PTA_LZCNT, PTA_MMX, PTA_MOVBE, PTA_NO_SAHF,
29302 PTA_PCLMUL, PTA_POPCNT, PTA_PREFETCH_SSE, PTA_RDRND, PTA_SSE, PTA_SSE2,
29303 PTA_SSE3, PTA_SSE4_1, PTA_SSE4_2, PTA_SSE4A, PTA_SSSE3, PTA_TBM,
29304 PTA_XOP, PTA_AVX2, PTA_BMI2, PTA_RTM, PTA_HLE, PTA_PRFCHW, PTA_RDSEED,
29305 PTA_ADX, PTA_FXSR, PTA_XSAVE, PTA_XSAVEOPT, PTA_AVX512F, PTA_AVX512ER,
29306 PTA_AVX512PF, PTA_AVX512CD, PTA_NO_TUNE, PTA_SHA, PTA_PREFETCHWT1,
29307 PTA_CLFLUSHOPT, PTA_XSAVEC, PTA_XSAVES, PTA_AVX512DQ, PTA_AVX512BW,
29308 PTA_AVX512VL, PTA_AVX512IFMA, PTA_AVX512VBMI, PTA_CLWB, PTA_MWAITX,
29309 PTA_CLZERO, PTA_NO_80387, PTA_PKU, PTA_AVX5124VNNIW, PTA_AVX5124FMAPS,
29310 PTA_AVX512VPOPCNTDQ, PTA_SGX, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES,
29311 PTA_AVX512VBMI2, PTA_VPCLMULQDQ, PTA_AVX512BITALG, PTA_RDPID,
29312 PTA_PCONFIG, PTA_WBNOINVD, PTA_AVX512VP2INTERSECT, PTA_PTWRITE,
29313 PTA_AVX512BF16, PTA_WAITPKG, PTA_MOVDIRI, PTA_MOVDIR64B, PTA_ENQCMD,
29314 PTA_CLDEMOTE, PTA_SERIALIZE, PTA_TSXLDTRK, PTA_AMX_TILE, PTA_AMX_INT8,
29315 PTA_AMX_BF16, PTA_UINTR, PTA_HRESET, PTA_KL, PTA_WIDEKL, PTA_AVXVNNI,
29316 PTA_X86_64_BASELINE, PTA_X86_64_V2, PTA_X86_64_V3, PTA_X86_64_V4,
29317 PTA_CORE2, PTA_NEHALEM, PTA_WESTMERE, PTA_SANDYBRIDGE, PTA_IVYBRIDGE,
29318 PTA_HASWELL, PTA_BROADWELL, PTA_SKYLAKE, PTA_SKYLAKE_AVX512,
29319 PTA_CASCADELAKE, PTA_COOPERLAKE, PTA_CANNONLAKE, PTA_ICELAKE_CLIENT,
29320 PTA_ICELAKE_SERVER, PTA_TIGERLAKE, PTA_SAPPHIRERAPIDS, PTA_ALDERLAKE,
29321 PTA_KNL, PTA_BONNELL, PTA_SILVERMONT, PTA_GOLDMONT, PTA_GOLDMONT_PLUS,
29322 PTA_TREMONT, PTA_KNM): Use constexpr instead of const.
29324 2021-02-18 Jakub Jelinek <jakub@redhat.com>
29326 PR middle-end/99109
29327 * gimple-array-bounds.cc (build_zero_elt_array_type): Rename to ...
29328 (build_printable_array_type): ... this. Add nelts argument. For
29329 overaligned eltype, use TYPE_MAIN_VARIANT (eltype) instead. If
29330 nelts, call build_array_type_nelts.
29331 (array_bounds_checker::check_mem_ref): Use build_printable_array_type
29332 instead of build_zero_elt_array_type and build_array_type_nelts.
29334 2021-02-18 Jakub Jelinek <jakub@redhat.com>
29337 * config/i386/i386.c (distance_non_agu_define): Don't call
29338 extract_insn_cached here.
29339 (ix86_lea_outperforms): Save and restore recog_data around call
29340 to distance_non_agu_define and distance_agu_use.
29341 (ix86_ok_to_clobber_flags): Remove.
29342 (ix86_avoid_lea_for_add): Don't call ix86_ok_to_clobber_flags.
29343 (ix86_avoid_lea_for_addr): Likewise. Adjust function comment.
29344 * config/i386/i386.md (*lea<mode>): Change from define_insn_and_split
29345 into define_insn. Move the splitting to define_peephole2 and
29346 check there using peep2_regno_dead_p if FLAGS_REG is dead.
29348 2021-02-17 Julian Brown <julian@codesourcery.com>
29350 * gimplify.c (gimplify_scan_omp_clauses): Handle ATTACH_DETACH
29353 2021-02-17 Xi Ruoyao <xry111@mengyan1223.wang>
29356 * config/mips/mips.c (mips_symbol_insns): Do not use
29357 MSA_SUPPORTED_MODE_P if mode is MAX_MACHINE_MODE.
29359 2021-02-16 Vladimir N. Makarov <vmakarov@redhat.com>
29361 PR inline-asm/98096
29362 * stmt.c (resolve_operand_name_1): Take inout operands into account
29363 for access to labels by names.
29364 * doc/extend.texi: Describe counting operands for accessing labels.
29366 2021-02-16 Richard Biener <rguenther@suse.de>
29368 PR tree-optimization/38474
29369 * tree-ssa-structalias.c (variable_info::address_taken): New.
29370 (new_var_info): Initialize address_taken.
29371 (process_constraint): Set address_taken.
29372 (solve_constraints): Use the new address_taken flag rather
29373 than is_reg_var for sorting variables.
29374 (dump_constraint): Dump the variable number if the name
29377 2021-02-16 Jakub Jelinek <jakub@redhat.com>
29380 * tree-vect-stmts.c (vectorizable_simd_clone_call): For num_calls != 1
29381 multiply by 4096 and for inbranch by 8192.
29382 * config/i386/i386.c (ix86_simd_clone_usable): For TARGET_AVX512F,
29383 return 3, 2 or 1 for mangle letters 'b', 'c' or 'd'.
29385 2021-02-15 Maya Rashish <coypu@sdf.org>
29387 * config/aarch64/aarch64.c (aarch64_init_builtins):
29388 Call SUBTARGET_INIT_BUILTINS.
29390 2021-02-15 Peter Bergner <bergner@linux.ibm.com>
29392 PR rtl-optimization/98872
29393 * init-regs.c (initialize_uninitialized_regs): Skip initialization
29394 if CONST0_RTX is NULL.
29396 2021-02-15 Richard Sandiford <richard.sandiford@arm.com>
29398 PR rtl-optimization/98863
29399 * rtl-ssa/functions.h (function_info::bb_live_out_info): Delete.
29400 (function_info::build_info): Turn into a declaration, moving the
29401 definition to internals.h.
29402 (function_info::bb_walker): Declare.
29403 (function_info::create_reg_use): Likewise.
29404 (function_info::calculate_potential_phi_regs): Take a build_info
29406 (function_info::place_phis, function_info::create_ebbs): Declare.
29407 (function_info::calculate_ebb_live_in_for_debug): Likewise.
29408 (function_info::populate_backedge_phis): Delete.
29409 (function_info::start_block, function_info::end_block): Declare.
29410 (function_info::populate_phi_inputs): Delete.
29411 (function_info::m_potential_phi_regs): Move information to build_info.
29412 * rtl-ssa/internals.h: New file.
29413 (function_info::bb_phi_info): New class.
29414 (function_info::build_info): Moved from functions.h.
29415 Add a constructor and destructor.
29416 (function_info::build_info::ebb_use): Delete.
29417 (function_info::build_info::ebb_def): Likewise.
29418 (function_info::build_info::bb_live_out): Likewise.
29419 (function_info::build_info::tmp_ebb_live_in_for_debug): New variable.
29420 (function_info::build_info::potential_phi_regs): Likewise.
29421 (function_info::build_info::potential_phi_regs_for_debug): Likewise.
29422 (function_info::build_info::ebb_def_regs): Likewise.
29423 (function_info::build_info::bb_phis): Likewise.
29424 (function_info::build_info::bb_mem_live_out): Likewise.
29425 (function_info::build_info::bb_to_rpo): Likewise.
29426 (function_info::build_info::def_stack): Likewise.
29427 (function_info::build_info::old_def_stack_limit): Likewise.
29428 * rtl-ssa/internals.inl (function_info::build_info::record_reg_def):
29429 Remove the regno argument. Push the previous definition onto the
29430 definition stack where necessary.
29431 * rtl-ssa/accesses.cc: Include internals.h.
29432 * rtl-ssa/changes.cc: Likewise.
29433 * rtl-ssa/blocks.cc: Likewise.
29434 (function_info::build_info::build_info): Define.
29435 (function_info::build_info::~build_info): Likewise.
29436 (function_info::bb_walker): New class.
29437 (function_info::bb_walker::bb_walker): Define.
29438 (function_info::add_live_out_use): Convert a logarithmic-complexity
29439 test into a linear one. Allow the same definition to be passed
29441 (function_info::calculate_potential_phi_regs): Moved from
29442 functions.cc. Take a build_info parameter and store the
29443 information there instead.
29444 (function_info::place_phis): New function.
29445 (function_info::add_entry_block_defs): Update call to record_reg_def.
29446 (function_info::calculate_ebb_live_in_for_debug): New function.
29447 (function_info::add_phi_nodes): Use bb_phis to decide which
29448 registers need phi nodes and initialize ebb_def_regs accordingly.
29449 Do not add degenerate phis here.
29450 (function_info::add_artificial_accesses): Use create_reg_use.
29451 Assert that all definitions are listed in the DF LR sets.
29452 Update call to record_reg_def.
29453 (function_info::record_block_live_out): Record live-out register
29454 values in the phis of successor blocks. Use the live-out set
29455 when processing the last block in an EBB, instead of always
29456 using the live-in sets of successor blocks. AND the live sets
29457 with the set of registers that have been defined in the EBB,
29458 rather than with all potential phi registers. Cope correctly
29459 with branches back to the start of the current EBB.
29460 (function_info::start_block): New function.
29461 (function_info::end_block): Likewise.
29462 (function_info::populate_phi_inputs): Likewise.
29463 (function_info::create_ebbs): Likewise.
29464 (function_info::process_all_blocks): Rewrite into a multi-phase
29466 * rtl-ssa/functions.cc: Include internals.h.
29467 (function_info::calculate_potential_phi_regs): Move to blocks.cc.
29468 (function_info::init_function_data): Remove caller.
29469 * rtl-ssa/insns.cc: Include internals.h
29470 (function_info::create_reg_use): New function. Lazily any
29471 degenerate phis needed by the linear RPO view.
29472 (function_info::record_use): Use create_reg_use. When processing
29473 debug uses, use potential_phi_regs and test it before checking
29474 whether the register is live on entry to the current EBB. Lazily
29475 calculate ebb_live_in_for_debug.
29476 (function_info::record_call_clobbers): Update call to record_reg_def.
29477 (function_info::record_def): Likewise.
29479 2021-02-15 Martin Liska <mliska@suse.cz>
29481 * toplev.c (init_asm_output): Free output of
29482 gen_command_line_string function.
29483 (process_options): Likewise.
29485 2021-02-15 Martin Liska <mliska@suse.cz>
29487 * params.opt: Add 2 missing Param keywords.
29489 2021-02-15 Eric Botcazou <ebotcazou@adacore.com>
29491 * df-core.c (df_worklist_dataflow_doublequeue): Use proper cast.
29493 2021-02-15 Jakub Jelinek <jakub@redhat.com>
29495 PR tree-optimization/99079
29496 * match.pd (A % (pow2pcst << N) -> A & ((pow2pcst << N) - 1)): Remove
29497 useless tree_nop_conversion_p (type, TREE_TYPE (@3)) check. Instead
29498 require both type and TREE_TYPE (@1) to be integral types and either
29499 type having smaller or equal precision, or TREE_TYPE (@1) being
29500 unsigned type, or type being signed type. If TREE_TYPE (@1)
29501 doesn't have wrapping overflow, perform the subtraction of one in
29504 2021-02-14 Jan Hubicka <hubicka@ucw.cz>
29505 Richard Biener <rguether@suse.de>
29508 * ipa-reference.c (ipa_init): Only conditinally initialize
29509 reference_vars_to_consider.
29510 (propagate): Conditionally deninitialize reference_vars_to_consider.
29511 (ipa_reference_write_optimization_summary): Sanity check that
29512 reference_vars_to_consider is not allocated.
29514 2021-02-13 Levy Hsu <admin@levyhsu.com>
29517 * config/riscv/riscv-shorten-memrefs.c (pass_shorten_memrefs): Add
29518 extend parameter to get_si_mem_base_reg declaration.
29519 (get_si_mem_base_reg): Add extend parameter. Set it.
29520 (analyze): Pass extend arg to get_si_mem_base_reg.
29521 (transform): Likewise. Use it when rewriting mems.
29522 * config/riscv/riscv.c (riscv_legitimize_move): Check for subword
29523 loads and emit sign/zero extending load followed by subreg move.
29525 2021-02-13 Jim Wilson <jimw@sifive.com>
29528 * config/riscv/riscv.c (riscv_compressed_lw_address_p): Drop early
29529 exit when !reload_completed. Only perform check for compressed reg
29530 if reload_completed.
29531 (riscv_rtx_costs): In MEM case, when optimizing for size and
29532 shorten memrefs, if not compressible, then increase cost.
29534 2021-02-13 Jakub Jelinek <jakub@redhat.com>
29536 PR rtl-optimization/98439
29537 * recog.c (pass_split_before_regstack::gate): Enable even when
29538 pass_split_before_sched2 is enabled if -fselective-scheduling2 is
29541 2021-02-13 Jakub Jelinek <jakub@redhat.com>
29544 * config/i386/mmx.md (*mmx_pshufd_1): Add a combine splitter for
29545 swap of V2SImode elements in memory into DImode memory rotate by 32.
29547 2021-02-12 Martin Sebor <msebor@redhat.com>
29549 * tree-pretty-print.c (print_generic_expr_to_str): Update comment.
29551 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
29553 * rtl-ssa/accesses.cc (function_info::make_use_available): Use
29554 m_temp_obstack rather than m_obstack to allocate the temporary use.
29556 2021-02-12 Richard Sandiford <richard.sandiford@arm.com>
29558 * df-problems.c (df_lr_bb_local_compute): Treat partial definitions
29559 as read-modify operations.
29561 2021-02-12 Richard Biener <rguenther@suse.de>
29563 PR middle-end/38474
29564 * ipa-fnsummary.c (unmodified_parm_1): Only walk when
29565 fbi->aa_walk_budget is bigger than zero. Update
29566 fbi->aa_walk_budget.
29567 (param_change_prob): Likewise.
29568 * ipa-prop.c (detect_type_change_from_memory_writes):
29569 Properly account walk_aliased_vdefs.
29570 (parm_preserved_before_stmt_p): Canonicalize updates.
29571 (parm_ref_data_preserved_p): Likewise.
29572 (parm_ref_data_pass_through_p): Likewise.
29573 (determine_known_aggregate_parts): Account own alias queries.
29575 2021-02-12 Martin Liska <mliska@suse.cz>
29577 * opts-common.c (decode_cmdline_option): Release werror_arg.
29578 * opts.c (gen_producer_string): Release output of
29579 gen_command_line_string.
29581 2021-02-12 Richard Biener <rguenther@suse.de>
29583 PR tree-optimization/38474
29584 * params.opt (-param=max-store-chains-to-track=): New param.
29585 (-param=max-stores-to-track=): Likewise.
29586 * doc/invoke.texi (max-store-chains-to-track): Document.
29587 (max-stores-to-track): Likewise.
29588 * gimple-ssa-store-merging.c (pass_store_merging::m_n_chains):
29590 (pass_store_merging::m_n_stores): Likewise.
29591 (pass_store_merging::terminate_and_process_chain): Update
29592 m_n_stores and m_n_chains.
29593 (pass_store_merging::process_store): Likewise. Terminate
29594 oldest chains if the number of stores or chains get too large.
29595 (imm_store_chain_info::terminate_and_process_chain): Dump
29598 2021-02-11 Eric Botcazou <ebotcazou@adacore.com>
29600 * config/i386/winnt.c (i386_pe_seh_unwind_emit): When switching to
29601 the cold section, emit a nop before the directive if the previous
29602 active instruction can throw.
29604 2021-02-11 Peter Bergner <bergner@linux.ibm.com>
29607 * config/rs6000/predicates.md (mma_assemble_input_operand): Restrict
29608 memory addresses that are legal for quad word accesses.
29610 2021-02-11 Andrea Corallo <andrea.corallo@arm.com>
29613 * config/arm/thumb2.md (*doloop_end_internal): Generate
29614 alternative sequence to handle long range branches.
29616 2021-02-11 Joel Hutton <joel.hutton@arm.com>
29618 PR tree-optimization/98772
29619 * optabs-tree.c (supportable_half_widening_operation): New function
29620 to check for supportable V8QI->V8HI widening patterns.
29621 * optabs-tree.h (supportable_half_widening_operation): New function.
29622 * tree-vect-stmts.c (vect_create_half_widening_stmts): New function
29623 to create promotion stmts for V8QI->V8HI widening patterns.
29624 (vectorizable_conversion): Add case for V8QI->V8HI.
29626 2021-02-11 Richard Biener <rguenther@suse.de>
29628 * sparseset.h (SPARSESET_ELT_BITS): Remove.
29629 (SPARSESET_ELT_TYPE): Use unsigned int.
29630 * fwprop.c: Do not include sparseset.h.
29632 2021-02-10 Jakub Jelinek <jakub@redhat.com>
29635 * varasm.c (declare_weak): For -fsyntax-only, allow even
29636 TREE_ASM_WRITTEN function decls.
29638 2021-02-10 Jakub Jelinek <jakub@redhat.com>
29641 * config/i386/sse.md (fix<fixunssuffix>_truncv2sfv2di2,
29642 <insn>v8qiv8hi2, <insn>v8qiv8si2, <insn>v4qiv4si2, <insn>v4hiv4si2,
29643 <insn>v8qiv8di2, <insn>v4qiv4di2, <insn>v2qiv2di2, <insn>v4hiv4di2,
29644 <insn>v2hiv2di2, <insn>v2siv2di2): Force operands[1] into REG before
29645 calling simplify_gen_subreg on it.
29647 2021-02-10 Martin Liska <mliska@suse.cz>
29649 * config/nvptx/nvptx.c (nvptx_option_override): Use
29650 flag_patchable_function_entry instead of the removed
29651 function_entry_patch_area_size.
29653 2021-02-10 Martin Liska <mliska@suse.cz>
29655 PR tree-optimization/99002
29656 PR tree-optimization/99026
29657 * gimple-if-to-switch.cc (if_chain::is_beneficial): Fix memory
29658 leak when adjacent cases are merged.
29659 * tree-switch-conversion.c (switch_decision_tree::analyze_switch_statement): Use
29661 (make_pass_lower_switch): Remove trailing whitespace.
29662 * tree-switch-conversion.h (release_clusters): New.
29664 2021-02-10 Richard Biener <rguenther@suse.de>
29666 PR rtl-optimization/99054
29667 * cfgrtl.c (rtl-optimization/99054): Return an auto_vec.
29668 (fixup_partitions): Adjust.
29669 (rtl_verify_edges): Likewise.
29671 2021-02-10 Jakub Jelinek <jakub@redhat.com>
29673 PR middle-end/99007
29674 * gimplify.c (gimplify_scan_omp_clauses): For MEM_REF on reductions,
29675 temporarily disable gimplify_ctxp->into_ssa around gimplify_expr
29678 2021-02-10 Richard Biener <rguenther@suse.de>
29681 * ipa-pure-const.c (propagate_malloc): Use an auto_vec<>
29684 2021-02-10 Richard Biener <rguenther@suse.de>
29686 PR tree-optimization/99024
29687 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Only
29688 clear loop->aux if it is associated with the destroyed loop_vinfo.
29690 2021-02-10 Martin Liska <mliska@suse.cz>
29692 PR tree-optimization/99002
29693 * gimple-if-to-switch.cc (find_conditions): Fix memory leak
29696 2021-02-10 Martin Liska <mliska@suse.cz>
29699 * ipa-icf.c (sem_item::add_reference): Fix memory leak when
29700 a reference exists.
29702 2021-02-10 Jakub Jelinek <jakub@redhat.com>
29705 * dwarf2out.c (prune_unused_types_walk): Mark DW_TAG_variable DIEs
29706 at class scope for DWARF5+.
29708 2021-02-09 Eric Botcazou <ebotcazou@adacore.com>
29710 PR rtl-optimization/96015
29711 * reorg.c (skip_consecutive_labels): Minor comment tweaks.
29712 (relax_delay_slots): When deleting a jump to the next active
29713 instruction over a barrier, first delete the barrier if the
29714 jump is the only way to reach the target label.
29716 2021-02-09 Andre Vieira <andre.simoesdiasvieira@arm.com>
29718 * config/aarch64/aarch64-cost-tables.h: Add entries for vect.mul.
29719 * config/aarch64/aarch64.c (aarch64_rtx_mult_cost): Use vect.mul for
29720 vector multiplies and vect.alu for SSRA.
29721 * config/arm/aarch-common-protos.h (struct vector_cost_table): Define
29722 vect.mul cost field.
29723 * config/arm/aarch-cost-tables.h: Add entries for vect.mul.
29724 * config/arm/arm.c: Likewise.
29726 2021-02-09 Richard Biener <rguenther@suse.de>
29728 PR tree-optimization/98863
29729 * tree-ssa-sccvn.h (vn_avail::next_undo): Add.
29730 * tree-ssa-sccvn.c (last_pushed_avail): New global.
29731 (rpo_elim::eliminate_push_avail): Chain pushed avails.
29732 (unwind_state::avail_top): Add.
29733 (do_unwind): Rewrite unwinding of avail entries.
29734 (do_rpo_vn): Initialize last_pushed_avail and
29735 avail_top of the undo state.
29737 2021-02-09 Jakub Jelinek <jakub@redhat.com>
29739 PR middle-end/99004
29740 * calls.c (maybe_warn_rdwr_sizes): Change s0 and s1 type from
29741 const char * to char * and free those pointers after use.
29743 2021-02-09 Richard Biener <rguenther@suse.de>
29745 PR tree-optimization/99017
29746 * tree-vect-slp.c (vect_bb_vectorization_profitable_p): Allow
29747 zero vector cost entries.
29749 2021-02-08 Andre Vieira <andre.simoesdiasvieira@arm.com>
29751 PR middle-end/98974
29752 * tree-vect-stmts.c (vectorizable_condition): Remove shadow vec_num
29753 parameter in vectorizable_condition.
29755 2021-02-08 Richard Biener <rguenther@suse.de>
29758 * tree.c (walk_tree_1): Walk VECTOR_CST elements.
29760 2021-02-08 Martin Liska <mliska@suse.cz>
29763 * cfgexpand.c (pass_expand::execute): Parse per-function option
29764 flag_patchable_function_entry and use it.
29765 * common.opt: Remove function_entry_patch_area_size and
29766 function_entry_patch_area_start global variables.
29767 * opts.c (parse_and_check_patch_area): New function.
29768 (common_handle_option): Use it.
29769 * opts.h (parse_and_check_patch_area): New function.
29770 * toplev.c (process_options): Parse and use
29771 function_entry_patch_area_size.
29773 2021-02-08 Martin Sebor <msebor@redhat.com>
29775 * doc/extend.texi (attribute malloc): Correct typos.
29777 2021-02-05 Nathan Sidwell <nathan@acm.org>
29780 * gcc.c (driver::maybe_run_linker): Check for input file
29781 accessibility if not linking.
29783 2021-02-05 Richard Biener <rguenther@suse.de>
29785 PR tree-optimization/98855
29786 * tree-vectorizer.h (add_stmt_cost): New overload.
29787 * tree-vect-slp.c (li_cost_vec_cmp): New.
29788 (vect_bb_slp_scalar_cost): Cost individual loop regions
29789 separately. Account for the scalar instance root stmt.
29791 2021-02-05 Tom de Vries <tdevries@suse.de>
29794 * tree-switch-conversion.c (jump_table_cluster::emit): Add loc
29796 (bit_test_cluster::emit): Reuse location_t for newly created
29798 (switch_decision_tree::try_switch_expansion): Preserve
29800 * tree-switch-conversion.h: Change function signatures.
29802 2021-02-05 Jakub Jelinek <jakub@redhat.com>
29805 * config/i386/i386-options.c (m_NONE, m_ALL): Define.
29806 * config/i386/x86-tune.def (X86_TUNE_BRANCH_PREDICTION_HINTS,
29807 X86_TUNE_PROMOTE_QI_REGS): Use m_NONE instead of 0U.
29808 (X86_TUNE_QIMODE_MATH): Use m_ALL instead of ~0U.
29810 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29812 * config/aarch64/aarch64-simd-builtins.def (get_high): Define builtin.
29813 * config/aarch64/aarch64-simd.md (aarch64_get_high<mode>): Define.
29814 * config/aarch64/arm_neon.h (__GET_HIGH): Delete.
29815 (vget_high_f16): Reimplement using new builtin.
29816 (vget_high_f32): Likewise.
29817 (vget_high_f64): Likewise.
29818 (vget_high_p8): Likewise.
29819 (vget_high_p16): Likewise.
29820 (vget_high_p64): Likewise.
29821 (vget_high_s8): Likewise.
29822 (vget_high_s16): Likewise.
29823 (vget_high_s32): Likewise.
29824 (vget_high_s64): Likewise.
29825 (vget_high_u8): Likewise.
29826 (vget_high_u16): Likewise.
29827 (vget_high_u32): Likewise.
29828 (vget_high_u64): Likewise.
29830 2021-02-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29832 * config/aarch64/aarch64-simd-builtins.def (get_low): Define builtin.
29833 * config/aarch64/aarch64-simd.md (aarch64_get_low<mode>): Define.
29834 * config/aarch64/arm_neon.h (__GET_LOW): Delete.
29835 (vget_low_f16): Reimplement using new builtin.
29836 (vget_low_f32): Likewise.
29837 (vget_low_f64): Likewise.
29838 (vget_low_p8): Likewise.
29839 (vget_low_p16): Likewise.
29840 (vget_low_p64): Likewise.
29841 (vget_low_s8): Likewise.
29842 (vget_low_s16): Likewise.
29843 (vget_low_s32): Likewise.
29844 (vget_low_s64): Likewise.
29845 (vget_low_u8): Likewise.
29846 (vget_low_u16): Likewise.
29847 (vget_low_u32): Likewise.
29848 (vget_low_u64): Likewise.
29850 2021-02-05 Kito Cheng <kito.cheng@sifive.com>
29852 * gcc.c (print_multilib_info): Check all required argument is provided
29855 2021-02-05 liuhongt <hongtao.liu@intel.com>
29858 * config/i386/i386-expand.c (ix86_expand_sse_cmp): Don't
29859 generate integer mask comparison for 128/256-bits vector when
29860 op_true/op_false is NULL_RTX or CONSTM1_RTX/CONST0_RTX. Also
29861 delete redundant !maskcmp condition.
29862 (ix86_expand_int_vec_cmp): Ditto but no redundant deletion
29864 (ix86_expand_sse_movcc): Delete definition of maskcmp, add the
29865 condition directly to if (maskcmp), add extra check for
29866 cmpmode, it should be MODE_INT.
29867 (ix86_expand_fp_vec_cmp): Pass NULL to ix86_expand_sse_cmp's
29868 parameters op_true/op_false.
29869 (ix86_use_mask_cmp_p): New.
29871 2021-02-05 liuhongt <hongtao.liu@intel.com>
29874 * config/i386/x86-tune.def (X86_TUNE_AVX256_UNALIGNED_LOAD_OPTIMAL):
29875 Remove m_GENERIC from ~list.
29876 (X86_TUNE_AVX256_UNALIGNED_STORE_OPTIMAL): Ditto.
29878 2021-02-04 David Malcolm <dmalcolm@redhat.com>
29881 * diagnostic-show-locus.c (compatible_locations_p): Require
29882 locations in the same macro map to be either both from the
29883 macro definition, or both from the macro arguments.
29885 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
29887 * config/aarch64/aarch64-simd-builtins.def: Add
29888 [su]mull_hi_lane[q] builtin generator macros.
29889 * config/aarch64/aarch64-simd.md
29890 (aarch64_<su>mull_hi_lane<mode>_insn): Define.
29891 (aarch64_<su>mull_hi_lane<mode>): Define.
29892 (aarch64_<su>mull_hi_laneq<mode>_insn): Define.
29893 (aarch64_<su>mull_hi_laneq<mode>): Define.
29894 * config/aarch64/arm_neon.h (vmull_high_lane_s16): Use RTL
29895 builtin instead of inline asm.
29896 (vmull_high_lane_s32): Likewise.
29897 (vmull_high_lane_u16): Likewise.
29898 (vmull_high_lane_u32): Likewise.
29899 (vmull_high_laneq_s16): Likewise.
29900 (vmull_high_laneq_s32): Likewise.
29901 (vmull_high_laneq_u16): Likewise.
29902 (vmull_high_laneq_u32): Liekwise.
29904 2021-02-04 Jonathan Wright <jonathan.wright@arm.com>
29906 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_hi_n
29907 builtin generator macros.
29908 * config/aarch64/aarch64-simd.md
29909 (aarch64_<su>mull_hi_n<mode>_insn): Define.
29910 (aarch64_<su>mull_hi_n<mode>): Define.
29911 * config/aarch64/arm_neon.h (vmull_high_n_s16): Use RTL builtin
29912 instead of inline asm.
29913 (vmull_high_n_s32): Likewise.
29914 (vmull_high_n_u16): Likewise.
29915 (vmull_high_n_u32): Likewise.
29917 2021-02-04 Richard Biener <rguenther@suse.de>
29919 PR tree-optimization/98855
29920 * tree-vect-loop.c (vectorizable_phi): Do not cost
29921 single-argument PHIs.
29922 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Likewise.
29923 * tree-vect-stmts.c (vectorizable_bswap): Also perform
29924 costing for SLP operation.
29926 2021-02-04 Martin Liska <mliska@suse.cz>
29928 * doc/extend.texi: Mention -mprefer-vector-width in target
29931 2021-02-03 Martin Sebor <msebor@redhat.com>
29933 PR tree-optimization/98937
29934 * tree-ssa-strlen.c (strlen_dom_walker::~strlen_dom_walker): Define.
29935 Flush pointer_query cache.
29937 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
29939 * config/rs6000/genfusion.pl (gen_2logical): Add missing
29940 fixes based on patch review.
29941 * config/rs6000/fusion.md: Regenerate file.
29943 2021-02-03 Aaron Sawdey <acsawdey@linux.ibm.com>
29945 * config/rs6000/t-rs6000: Comment out auto generation of
29948 2021-02-03 Andrew Stubbs <ams@codesourcery.com>
29950 * config/gcn/gcn-opts.h (enum processor_type): Add PROCESSOR_GFX908.
29951 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Add gfx908.
29952 (output_file_start): Add gfx908.
29953 * config/gcn/gcn.opt (gpu_type): Add gfx908.
29954 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Add march=gfx908.
29955 (MULTILIB_DIRNAMES): Add gfx908.
29956 * config/gcn/mkoffload.c (EF_AMDGPU_MACH_AMDGCN_GFX908): New define.
29957 (main): Recognize gfx908.
29958 * config/gcn/t-omp-device: Add gfx908.
29960 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
29962 * config/aarch64/aarch64-simd-builtins.def: Add
29963 [su]mlsl_hi_lane[q] builtin macro generators.
29964 * config/aarch64/aarch64-simd.md
29965 (aarch64_<su>mlsl_hi_lane<mode>_insn): Define.
29966 (aarch64_<su>mlsl_hi_lane<mode>): Define.
29967 (aarch64_<su>mlsl_hi_laneq<mode>_insn): Define.
29968 (aarch64_<su>mlsl_hi_laneq<mode>): Define.
29969 * config/aarch64/arm_neon.h (vmlsl_high_lane_s16): Use RTL
29970 builtin instead of inline asm.
29971 (vmlsl_high_lane_s32): Likewise.
29972 (vmlsl_high_lane_u16): Likewise.
29973 (vmlsl_high_lane_u32): Likewise.
29974 (vmlsl_high_laneq_s16): Likewise.
29975 (vmlsl_high_laneq_s32): Likewise.
29976 (vmlsl_high_laneq_u16): Likewise.
29977 (vmlsl_high_laneq_u32): Likewise.
29978 (vmlal_high_laneq_u32): Likewise.
29980 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
29982 * config/aarch64/aarch64-simd-builtins.def: Add
29983 [su]mlal_hi_lane[q] builtin generator macros.
29984 * config/aarch64/aarch64-simd.md
29985 (aarch64_<su>mlal_hi_lane<mode>_insn): Define.
29986 (aarch64_<su>mlal_hi_lane<mode>): Define.
29987 (aarch64_<su>mlal_hi_laneq<mode>_insn): Define.
29988 (aarch64_<su>mlal_hi_laneq<mode>): Define.
29989 * config/aarch64/arm_neon.h (vmlal_high_lane_s16): Use RTL
29990 builtin instead of inline asm.
29991 (vmlal_high_lane_s32): Likewise.
29992 (vmlal_high_lane_u16): Likewise.
29993 (vmlal_high_lane_u32): Likewise.
29994 (vmlal_high_laneq_s16): Likewise.
29995 (vmlal_high_laneq_s32): Likewise.
29996 (vmlal_high_laneq_u16): Likewise.
29997 (vmlal_high_laneq_u32): Likewise.
29999 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
30001 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_hi_n
30002 builtin generator macros.
30003 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_hi_n<mode>_insn):
30005 (aarch64_<su>mlsl_hi_n<mode>): Define.
30006 * config/aarch64/arm_neon.h (vmlsl_high_n_s16): Use RTL builtin
30007 instead of inline asm.
30008 (vmlsl_high_n_s32): Likewise.
30009 (vmlsl_high_n_u16): Likewise.
30010 (vmlsl_high_n_u32): Likewise.
30012 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
30014 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_hi_n
30015 builtin generator macros.
30016 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_hi_n<mode>_insn):
30018 (aarch64_<su>mlal_hi_n<mode>): Define.
30019 * config/aarch64/arm_neon.h (vmlal_high_n_s16): Use RTL builtin
30020 instead of inline asm.
30021 (vmlal_high_n_s32): Likewise.
30022 (vmlal_high_n_u16): Likewise.
30023 (vmlal_high_n_u32): Likewise.
30025 2021-02-03 Jonathan Wright <jonathan.wright@arm.com>
30027 * config/aarch64/aarch64-simd-builtins.def: Add RTL builtin
30029 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal_hi<mode>):
30031 (aarch64_<su>mlal_hi<mode>_insn): This.
30032 (aarch64_<su>mlal_hi<mode>): Define.
30033 * config/aarch64/arm_neon.h (vmlal_high_s8): Use RTL builtin
30034 instead of inline asm.
30035 (vmlal_high_s16): Likewise.
30036 (vmlal_high_s32): Likewise.
30037 (vmlal_high_u8): Likewise.
30038 (vmlal_high_u16): Likewise.
30039 (vmlal_high_u32): Likewise.
30041 2021-02-03 Ilya Leoshkevich <iii@linux.ibm.com>
30043 * lra-spills.c (remove_pseudos): Call lra_update_insn_recog_data()
30044 after calling alter_subreg() on a (mem).
30046 2021-02-03 Martin Liska <mliska@suse.cz>
30049 * lto-streamer-out.c (produce_lto_section): Fill up missing
30051 * lto-streamer.h (struct lto_section): Add _padding field.
30053 2021-02-03 Richard Biener <rguenther@suse.de>
30055 * lto-streamer.c (lto_get_section_name): Free temporary
30057 * tree-loop-distribution.c
30058 (loop_distribution::merge_dep_scc_partitions): Free edge data.
30060 2021-02-03 Jakub Jelinek <jakub@redhat.com>
30062 PR middle-end/97487
30063 * ifcvt.c (noce_can_force_operand): New function.
30064 (noce_emit_move_insn): Use it.
30065 (noce_try_sign_mask): Likewise. Formatting fix.
30067 2021-02-03 Jakub Jelinek <jakub@redhat.com>
30069 PR middle-end/97971
30070 * lra-constraints.c (process_alt_operands): For inline asm, don't call
30071 fatal_insn, but instead return false.
30073 2021-02-03 Jakub Jelinek <jakub@redhat.com>
30075 PR tree-optimization/98287
30076 * config/i386/mmx.md (<insn><mode>3): For shifts don't enable expander
30079 2021-02-03 Tamar Christina <tamar.christina@arm.com>
30081 PR tree-optimization/98928
30082 * tree-vect-loop.c (vect_analyze_loop_2): Change
30083 STMT_VINFO_SLP_VECT_ONLY to STMT_VINFO_SLP_VECT_ONLY_PATTERN.
30084 * tree-vect-slp-patterns.c (complex_pattern::build): Likewise.
30085 * tree-vectorizer.h (STMT_VINFO_SLP_VECT_ONLY_PATTERN): New.
30086 (class _stmt_vec_info): Add slp_vect_pattern_only_p.
30088 2021-02-02 Richard Biener <rguenther@suse.de>
30090 * gimple-loop-interchange.cc (prepare_data_references):
30092 * gimple-loop-jam.c (tree_loop_unroll_and_jam): Likewise.
30093 * tree-ssa-loop-im.c (hoist_memory_references): Likewise.
30094 * tree-vect-stmts.c (vectorizable_condition): Do not
30096 (vectorizable_comparison): Likewise.
30098 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30100 * config/aarch64/aarch64-simd-builtins.def (ursqrte): Define builtin.
30101 * config/aarch64/aarch64-simd.md (aarch64_ursqrte<mode>): New pattern.
30102 * config/aarch64/arm_neon.h (vrsqrte_u32): Reimplement using builtin.
30103 (vrsqrteq_u32): Likewise.
30105 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30107 * config/aarch64/aarch64-simd-builtins.def (sqxtun2): Define builtin.
30108 * config/aarch64/aarch64-simd.md (aarch64_sqxtun2<mode>_le): Define.
30109 (aarch64_sqxtun2<mode>_be): Likewise.
30110 (aarch64_sqxtun2<mode>): Likewise.
30111 * config/aarch64/arm_neon.h (vqmovun_high_s16): Reimplement using builtin.
30112 (vqmovun_high_s32): Likewise.
30113 (vqmovun_high_s64): Likewise.
30114 * config/aarch64/iterators.md (UNSPEC_SQXTUN2): Define.
30116 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30118 * config/aarch64/aarch64-simd-builtins.def (bfdot_lane, bfdot_laneq): Use
30120 (bfmlalb_lane, bfmlalt_lane, bfmlalb_lane_q, bfmlalt_lane_q): Use FP flags.
30122 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30124 * config/aarch64/aarch64-simd-builtins.def (fcmla_lane0, fcmla_lane90,
30125 fcmla_lane180, fcmla_lane270, fcmlaq_lane0, fcmlaq_lane90, fcmlaq_lane180,
30126 fcmlaq_lane270, scvtf, ucvtf, fcvtzs, fcvtzu, scvtfsi, scvtfdi, ucvtfsi,
30127 ucvtfdi, fcvtzshf, fcvtzuhf, fmlal_lane_low, fmlsl_lane_low,
30128 fmlal_laneq_low, fmlsl_laneq_low, fmlalq_lane_low, fmlslq_lane_low,
30129 fmlalq_laneq_low, fmlslq_laneq_low, fmlal_lane_high, fmlsl_lane_high,
30130 fmlal_laneq_high, fmlsl_laneq_high, fmlalq_lane_high, fmlslq_lane_high,
30131 fmlalq_laneq_high, fmlslq_laneq_high): Use FP flags.
30133 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30135 * config/aarch64/aarch64-builtins.c (FLAG_LOAD): Define.
30136 * config/aarch64/aarch64-simd-builtins.def (ld1x2, ld2, ld3, ld4, ld2r,
30137 ld3r, ld4r, ld1, ld1x3, ld1x4): Use LOAD flags.
30139 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30141 * config/aarch64/aarch64-simd-builtins.def (combine, zip1, zip2,
30142 uzp1, uzp2, trn1, trn2, simd_bsl): Use AUTO_FP flags.
30144 2021-02-02 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30146 * config/aarch64/aarch64-simd-builtins.def (clrsb, clz, ctz, popcount,
30147 vec_smult_lane_, vec_smlal_lane_, vec_smult_laneq_, vec_smlal_laneq_,
30148 vec_umult_lane_, vec_umlal_lane_, vec_umult_laneq_, vec_umlal_laneq_,
30149 ashl, sshl, ushl, srshl, urshl, sdot_lane, udot_lane, sdot_laneq,
30150 udot_laneq, usdot_lane, usdot_laneq, sudot_lane, sudot_laneq, ashr,
30151 ashr_simd, lshr, lshr_simd, srshr_n, urshr_n, ssra_n, usra_n, srsra_n,
30152 ursra_n, sshll_n, ushll_n, sshll2_n, ushll2_n, ssri_n, usri_n, ssli_n,
30153 ssli_n, usli_n, bswap, rbit, simd_bsl, eor3q, rax1q, xarq, bcaxq): Use
30154 NONE builtin flags.
30156 2021-02-02 Jakub Jelinek <jakub@redhat.com>
30158 PR tree-optimization/98848
30159 * tree-vect-patterns.c (vect_recog_over_widening_pattern): Punt if
30160 STMT_VINFO_DEF_TYPE (last_stmt_info) is vect_reduction_def.
30162 2021-02-02 Kito Cheng <kito.cheng@sifive.com>
30165 * expr.c: Check mode before calling store_expr.
30167 2021-02-02 Christophe Lyon <christophe.lyon@linaro.org>
30169 * config/arm/iterators.md (supf): Remove VORNQ_S and VORNQ_U.
30171 * config/arm/mve.md (mve_vornq_s<mode>): New entry for vorn
30172 instruction using expression ior.
30173 (mve_vornq_u<mode>): New expander.
30174 (mve_vornq_f<mode>): Use ior code instead of unspec.
30175 * config/arm/unspecs.md (VORNQ_S, VORNQ_U, VORNQ_F): Remove.
30177 2021-02-02 Alexandre Oliva <oliva@adacore.com>
30179 * tree-nested.c (convert_nonlocal_reference_op): Move
30180 current_function_decl restore after re-gimplification.
30181 (convert_local_reference_op): Likewise.
30183 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30185 * config/aarch64/aarch64-simd-builtins.def (rshrn, rshrn2):
30187 * config/aarch64/aarch64-simd.md (aarch64_rshrn<mode>_insn_le):
30189 (aarch64_rshrn<mode>_insn_be): Likewise.
30190 (aarch64_rshrn<mode>): Likewise.
30191 (aarch64_rshrn2<mode>_insn_le): Likewise.
30192 (aarch64_rshrn2<mode>_insn_be): Likewise.
30193 (aarch64_rshrn2<mode>): Likewise.
30194 * config/aarch64/aarch64.md (unspec): Add UNSPEC_RSHRN.
30195 * config/aarch64/arm_neon.h (vrshrn_high_n_s16): Reimplement
30197 (vrshrn_high_n_s32): Likewise.
30198 (vrshrn_high_n_s64): Likewise.
30199 (vrshrn_high_n_u16): Likewise.
30200 (vrshrn_high_n_u32): Likewise.
30201 (vrshrn_high_n_u64): Likewise.
30202 (vrshrn_n_s16): Likewise.
30203 (vrshrn_n_s32): Likewise.
30204 (vrshrn_n_s64): Likewise.
30205 (vrshrn_n_u16): Likewise.
30206 (vrshrn_n_u32): Likewise.
30207 (vrshrn_n_u64): Likewise.
30209 2021-02-01 Sergei Trofimovich <siarheit@google.com>
30211 PR tree-optimization/98499
30212 * ipa-modref.c (analyze_ssa_name_flags): treat RVO
30213 conservatively and assume all possible side-effects.
30215 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30217 * config/aarch64/aarch64-simd-builtins.def (vec_unpacks_hi,
30218 vec_unpacku_hi_): Define builtins.
30219 * config/aarch64/arm_neon.h (vmovl_high_s8): Reimplement using
30221 (vmovl_high_s16): Likewise.
30222 (vmovl_high_s32): Likewise.
30223 (vmovl_high_u8): Likewise.
30224 (vmovl_high_u16): Likewise.
30225 (vmovl_high_u32): Likewise.
30227 2021-02-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30229 * config/aarch64/aarch64-simd-builtins.def (sabdl, uabdl):
30231 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl<mode>): New
30233 * config/aarch64/aarch64.md (unspec): Define UNSPEC_SABDL,
30235 * config/aarch64/arm_neon.h (vabdl_s8): Reimplemet using
30237 (vabdl_s16): Likewise.
30238 (vabdl_s32): Likewise.
30239 (vabdl_u8): Likewise.
30240 (vabdl_u16): Likewise.
30241 (vabdl_u32): Likewise.
30242 * config/aarch64/iterators.md (ABDL): New int iterator.
30243 (sur): Handle UNSPEC_SABDL, UNSPEC_UABDL.
30245 2021-02-01 Martin Sebor <msebor@redhat.com>
30247 * tree.h (BLOCK_VARS): Add comment.
30248 (BLOCK_SUBBLOCKS): Same.
30249 (BLOCK_SUPERCONTEXT): Same.
30250 (BLOCK_ABSTRACT_ORIGIN): Same.
30251 (inlined_function_outer_scope_p): Same.
30253 2021-02-01 Martin Sebor <msebor@redhat.com>
30255 PR middle-end/97172
30256 * attribs.c (attr_access::free_lang_data): Define new function.
30257 * attribs.h (attr_access::free_lang_data): Declare new function.
30259 2021-02-01 Richard Biener <rguenther@suse.de>
30261 * vec.h (auto_vec::auto_vec): Add memory stat parameters
30263 * bitmap.h (auto_bitmap::auto_bitmap): Likewise.
30265 2021-02-01 Tamar Christina <tamar.christina@arm.com>
30267 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>,
30268 aarch64_<su>mlsl<mode>, aarch64_<su>mlsl_n<mode>): Flip mult operands.
30270 2021-02-01 Richard Biener <rguenther@suse.de>
30272 PR rtl-optimization/98863
30273 * config/i386/i386-features.c (convert_scalars_to_vector):
30274 Set DF_RD_PRUNE_DEAD_DEFS.
30276 2021-01-31 Eric Botcazou <ebotcazou@adacore.com>
30278 * system.h (SIZE_MAX): Define if not already defined.
30280 2021-01-30 Aaron Sawdey <acsawdey@linux.ibm.com>
30282 * config/rs6000/genfusion.pl (gen_2logical): New function to
30283 generate patterns for logical-logical fusion.
30284 * config/rs6000/fusion.md: Regenerated patterns.
30285 * config/rs6000/rs6000-cpus.def: Add
30286 OPTION_MASK_P10_FUSION_2LOGICAL.
30287 * config/rs6000/rs6000.c (rs6000_option_override_internal):
30288 Enable logical-logical fusion for p10.
30289 * config/rs6000/rs6000.opt: Add -mpower10-fusion-2logical.
30291 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
30293 * config/rs6000/rs6000.opt: Add periods to new AIX options.
30295 2021-01-30 David Edelsohn <dje.gcc@gmail.com>
30297 * config/rs6000/rs6000.opt (mabi=vec-extabi): New.
30298 (mabi=vec-default): New.
30299 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Define
30300 __EXTABI__ for AIX Vector extended ABI.
30301 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Print AIX Vector
30303 (conditional_register_usage): If AIX vec_extabi enabled, vs20-vs31
30305 * doc/invoke.texi (PowerPC mabi): Add AIX vec-extabi and vec-default.
30307 2021-01-30 Jakub Jelinek <jakub@redhat.com>
30309 * config/i386/i386-features.c (remove_partial_avx_dependency): Clear
30310 DF_DEFER_INSN_RESCAN after calling df_process_deferred_rescans.
30312 2021-01-29 Vladimir N. Makarov <vmakarov@redhat.com>
30315 * lra-constraints.c (in_class_p): Don't narrow class only for REG
30318 2021-01-29 Will Schmidt <will_schmidt@vnet.ibm.com>
30320 * config/rs6000/rs6000-call.c (rs6000_expand_binup_builtin): Add
30321 clauses for CODE_FOR_vsx_xvcvuxddp_scale and
30322 CODE_FOR_vsx_xvcvsxddp_scale to the parameter checking code.
30324 2021-01-29 Andrew MacLeod <amacleod@redhat.com>
30326 PR tree-optimization/98866
30327 * gimple-range-gori.h (gori_compute:set_range_invariant): New.
30328 * gimple-range-gori.cc (gori_map::set_range_invariant): New.
30329 (gori_map::m_maybe_invariant): Rename from all_outgoing.
30330 (gori_map::gori_map): Rename all_outgoing to m_maybe_invariant.
30331 (gori_map::is_export_p): Ditto.
30332 (gori_map::calculate_gori): Ditto.
30333 (gori_compute::set_range_invariant): New.
30334 * gimple-range.cc (gimple_ranger::range_of_stmt): Set range
30335 invariant for pointers evaluating to [1, +INF].
30337 2021-01-29 Richard Biener <rguenther@suse.de>
30339 PR rtl-optimization/98863
30340 * config/i386/i386-features.c (remove_partial_avx_dependency):
30341 Do not perform DF analysis.
30342 (pass_data_remove_partial_avx_dependency): Remove
30345 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
30347 * config/aarch64/aarch64-simd-builtins.def: Add [su]mull_n
30348 builtin generator macros.
30349 * config/aarch64/aarch64-simd.md (aarch64_<su>mull_n<mode>):
30351 * config/aarch64/arm_neon.h (vmull_n_s16): Use RTL builtin
30352 instead of inline asm.
30353 (vmull_n_s32): Likewise.
30354 (vmull_n_u16): Likewise.
30355 (vmull_n_u32): Likewise.
30357 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30359 * config/aarch64/aarch64-simd-builtins.def (sabdl2, uabdl2):
30361 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
30363 (aarch64_<sur>abdl2<mode>): ... This.
30364 (<sur>sadv16qi): Adjust use of above.
30365 * config/aarch64/arm_neon.h (vabdl_high_s8): Reimplement using
30367 (vabdl_high_s16): Likewise.
30368 (vabdl_high_s32): Likewise.
30369 (vabdl_high_u8): Likewise.
30370 (vabdl_high_u16): Likewise.
30371 (vabdl_high_u32): Likewise.
30373 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30375 * config/aarch64/aarch64-simd-builtins.def (sabal2): Define
30377 (uabal2): Likewise.
30378 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal2<mode>): New
30380 * config/aarch64/aarch64.md (unspec): Add UNSPEC_SABAL2 and
30382 * config/aarch64/arm_neon.h (vabal_high_s8): Reimplement using
30384 (vabal_high_s16): Likewise.
30385 (vabal_high_s32): Likewise.
30386 (vabal_high_u8): Likewise.
30387 (vabal_high_u16): Likewise.
30388 (vabal_high_u32): Likewise.
30389 * config/aarch64/iterators.md (ABAL2): New mode iterator.
30390 (sur): Handle UNSPEC_SABAL2, UNSPEC_UABAL2.
30392 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30394 * config/aarch64/aarch64-simd-builtins.def (sabal): Define
30397 * config/aarch64/aarch64-simd.md (aarch64_<sur>abal<mode>_4):
30399 (aarch64_<sur>abal<mode>): ... This
30400 (<sur>sadv16qi): Adust use of the above.
30401 * config/aarch64/arm_neon.h (vabal_s8): Reimplement using
30403 (vabal_s16): Likewise.
30404 (vabal_s32): Likewise.
30405 (vabal_u8): Likewise.
30406 (vabal_u16): Likewise.
30407 (vabal_u32): Likewise.
30409 2021-01-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30411 * config/aarch64/aarch64-simd-builtins.def (saddlv, uaddlv):
30413 * config/aarch64/aarch64-simd.md (aarch64_<su>addlv<mode>):
30415 * config/aarch64/arm_neon.h (vaddlv_s8): Reimplement using
30417 (vaddlv_s16): Likewise.
30418 (vaddlv_u8): Likewise.
30419 (vaddlv_u16): Likewise.
30420 (vaddlvq_s8): Likewise.
30421 (vaddlvq_s16): Likewise.
30422 (vaddlvq_s32): Likewise.
30423 (vaddlvq_u8): Likewise.
30424 (vaddlvq_u16): Likewise.
30425 (vaddlvq_u32): Likewise.
30426 (vaddlv_s32): Likewise.
30427 (vaddlv_u32): Likewise.
30428 * config/aarch64/iterators.md (VDQV_L): New mode iterator.
30429 (unspec): Add UNSPEC_SADDLV, UNSPEC_UADDLV.
30430 (Vwstype): New mode attribute.
30432 (VWIDE_S): Likewise.
30433 (USADDLV): New int iterator.
30434 (su): Handle UNSPEC_SADDLV, UNSPEC_UADDLV.
30436 2021-01-29 Jonathan Wright <jonathan.wright@arm.com>
30438 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_lane[q]
30439 builtin generator macros.
30440 * config/aarch64/aarch64-simd.md (aarch64_vec_<su>mlsl_lane<Qlane>):
30442 * config/aarch64/arm_neon.h (vmlsl_lane_s16): Use RTL builtin
30443 instead of inline asm.
30444 (vmlsl_lane_s32): Likewise.
30445 (vmlsl_lane_u16): Likewise.
30446 (vmlsl_lane_u32): Likewise.
30447 (vmlsl_laneq_s16): Likewise.
30448 (vmlsl_laneq_s32): Likewise.
30449 (vmlsl_laneq_u16): Likewise.
30450 (vmlsl_laneq_u32): Likewise.
30452 2021-01-29 Richard Biener <rguenther@suse.de>
30454 * doc/invoke.texi (--param max-gcse-memory): Document unit
30456 * gcse.c (gcse_or_cprop_is_too_expensive): Adjust.
30457 * params.opt (--param max-gcse-memory): Adjust default and
30458 document unit of size.
30460 2021-01-29 Richard Biener <rguenther@suse.de>
30462 PR rtl-optimization/98863
30463 * gcse.c (gcse_or_cprop_is_too_expensive): Use unsigned
30464 HOST_WIDE_INT for the memory estimate.
30466 2021-01-29 Bin Cheng <bin.cheng@linux.alibaba.com>
30467 Richard Biener <rguenther@suse.de>
30469 PR tree-optimization/97627
30470 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
30471 Do not analyze fake edges.
30473 2021-01-29 Richard Biener <rguenther@suse.de>
30475 PR rtl-optimization/98144
30476 * df.h (df_mir_bb_info): Add con_visited member.
30477 * df-problems.c (df_mir_alloc): Initialize con_visited,
30478 do not fully populate IN and OUT.
30479 (df_mir_reset): Likewise.
30480 (df_mir_confluence_0): Set con_visited.
30481 (df_mir_confluence_n): Properly handle implicitely
30482 fully populated IN and OUT as designated by con_visited
30483 and update con_visited accordingly.
30485 2021-01-29 Jakub Jelinek <jakub@redhat.com>
30488 * config/arm/vec-common.md (mve_vshlq_<supf><mode>,
30489 vashl<mode>3, vashr<mode>3, vlshr<mode>3): Add
30490 && !TARGET_REALLY_IWMMXT to conditions.
30492 2021-01-29 Jakub Jelinek <jakub@redhat.com>
30495 * cfgbuild.c (find_bb_boundaries): Reset debug_insn when seeing
30498 2021-01-28 Marek Polacek <polacek@redhat.com>
30501 * stor-layout.c (finalize_type_size): If we reset TYPE_USER_ALIGN in
30502 the main variant, maybe reset it in its variants too.
30503 * tree.c (check_base_type): Return true only if TYPE_USER_ALIGN match.
30504 (check_aligned_type): Check if TYPE_USER_ALIGN match.
30506 2021-01-28 Christophe Lyon <christophe.lyon@linaro.org>
30509 * config/arm/arm.c (arm_rtx_costs_internal): Adjust cost of vector
30510 of constant zero for comparisons.
30512 2021-01-28 Michael Meissner <meissner@linux.ibm.com>
30514 * config/rs6000/rs6000.c (rs6000_mangle_decl_assembler_name): Add
30515 support for mapping built-in function names for long double
30516 built-in functions if long double is IEEE 128-bit.
30518 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
30520 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlsl_n
30521 builtin generator macros.
30522 * config/aarch64/aarch64-simd.md (aarch64_<su>mlsl_n<mode>):
30524 * config/aarch64/arm_neon.h (vmlsl_n_s16): Use RTL builtin
30525 instead of inline asm.
30526 (vmlsl_n_s32): Likewise.
30527 (vmlsl_n_u16): Likewise.
30528 (vmlsl_n_u32): Likewise.
30530 2021-01-28 Jonathan Wright <jonathan.wright@arm.com>
30532 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal_n
30533 builtin generator macros.
30534 * config/aarch64/aarch64-simd.md (aarch64_<su>mlal_n<mode>):
30536 * config/aarch64/arm_neon.h (vmlal_n_s16): Use RTL builtin
30537 instead of inline asm.
30538 (vmlal_n_s32): Likewise.
30539 (vmlal_n_u16): Likewise.
30540 (vmlal_n_u32): Likewise.
30542 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30544 * config/aarch64/aarch64-simd-builtins.def (shrn2): Define
30546 * config/aarch64/aarch64-simd.md (aarch64_shrn2<mode>_insn_le):
30548 (aarch64_shrn2<mode>_insn_be): Likewise.
30549 (aarch64_shrn2<mode>): Likewise.
30550 * config/aarch64/arm_neon.h (vshrn_high_n_s16): Reimlplement
30552 (vshrn_high_n_s32): Likewise.
30553 (vshrn_high_n_s64): Likewise.
30554 (vshrn_high_n_u16): Likewise.
30555 (vshrn_high_n_u32): Likewise.
30556 (vshrn_high_n_u64): Likewise.
30558 2021-01-28 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30560 * config/aarch64/aarch64-simd-builtins.def (shrn): Define
30562 * config/aarch64/aarch64-simd.md (aarch64_shrn<mode>_insn_le):
30564 (aarch64_shrn<mode>_insn_be): Likewise.
30565 (aarch64_shrn<mode>): Likewise.
30566 * config/aarch64/arm_neon.h (vshrn_n_s16): Reimplement using
30568 (vshrn_n_s32): Likewise.
30569 (vshrn_n_s64): Likewise.
30570 (vshrn_n_u16): Likewise.
30571 (vshrn_n_u32): Likewise.
30572 (vshrn_n_u64): Likewise.
30573 * config/aarch64/iterators.md (vn_mode): New mode attribute.
30575 2021-01-28 Richard Biener <rguenther@suse.de>
30577 PR rtl-optimization/80960
30578 * dse.c (check_mem_read_rtx): Call get_addr on the
30581 2021-01-28 Xionghu Luo <luoxhu@linux.ibm.com>
30582 David Edelsohn <dje.gcc@gmail.com>
30585 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
30586 Don't generate VIEW_CONVERT_EXPR for fcode ALTIVEC_BUILTIN_VEC_INSERT
30588 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
30590 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Remove the
30591 wrapper call rs6000_expand_vector_set_var for cleanup. Call
30592 rs6000_expand_vector_set_var_p9 and rs6000_expand_vector_set_var_p8
30594 (rs6000_expand_vector_set_var): Delete.
30595 (rs6000_expand_vector_set_var_p9): Make static.
30596 (rs6000_expand_vector_set_var_p8): Make static.
30598 2021-01-28 Xing GUO <higuoxing@gmail.com>
30600 * common/config/riscv/riscv-common.c
30601 (riscv_subset_list::parsing_subset_version): Fix -march option parsing
30602 when `p` extension exists.
30604 2021-01-27 Vladimir N. Makarov <vmakarov@redhat.com>
30606 PR rtl-optimization/97684
30607 * ira.c (ira): Call ira_set_pseudo_classes before
30608 update_equiv_regs when it is necessary.
30610 2021-01-27 Jakub Jelinek <jakub@redhat.com>
30613 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Use
30614 %w0, %w1 and %2 instead of %0, %1 and %2.
30616 2021-01-27 Aaron Sawdey <acsawdey@linux.ibm.com>
30618 * config/rs6000/genfusion.pl: New script to generate
30619 define_insn_and_split patterns so combine can arrange fused
30620 instructions next to each other.
30621 * config/rs6000/fusion.md: New file, generated fused instruction
30622 patterns for combine.
30623 * config/rs6000/predicates.md (const_m1_to_1_operand): New predicate.
30624 (non_update_memory_operand): New predicate.
30625 * config/rs6000/rs6000-cpus.def: Add OPTION_MASK_P10_FUSION and
30626 OPTION_MASK_P10_FUSION_LD_CMPI to ISA_3_1_MASKS_SERVER and
30628 * config/rs6000/rs6000-protos.h (address_is_non_pfx_d_or_x): Add
30630 * config/rs6000/rs6000.c (rs6000_option_override_internal):
30631 Automatically set OPTION_MASK_P10_FUSION and
30632 OPTION_MASK_P10_FUSION_LD_CMPI if target is power10.
30633 (rs600_opt_masks): Allow -mpower10-fusion
30634 in function attributes.
30635 (address_is_non_pfx_d_or_x): New function.
30636 * config/rs6000/rs6000.h: Add MASK_P10_FUSION.
30637 * config/rs6000/rs6000.md: Include fusion.md.
30638 * config/rs6000/rs6000.opt: Add -mpower10-fusion
30639 and -mpower10-fusion-ld-cmpi.
30640 * config/rs6000/t-rs6000: Add dependencies involving fusion.md.
30642 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
30644 * config/aarch64/aarch64-simd-builtins.def: Add [su]mlal
30645 builtin generator macros.
30646 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlal<mode>):
30648 (aarch64_<su>mlal<mode>): This.
30649 * config/aarch64/arm_neon.h (vmlal_s8): Use RTL builtin
30650 instead of inline asm.
30651 (vmlal_s16): Likewise.
30652 (vmlal_s32): Likewise.
30653 (vmlal_u8): Likewise.
30654 (vmlal_u16): Likewise.
30655 (vmlal_u32): Likewise.
30657 2021-01-27 Richard Biener <rguenther@suse.de>
30659 PR tree-optimization/98854
30660 * tree-vect-slp.c (vect_build_slp_tree_2): Also build
30661 PHIs from scalars when the number of CTORs matches the
30662 number of children.
30664 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
30666 * config/aarch64/aarch64-simd-builtins.def: Add mls_n builtin
30668 * config/aarch64/aarch64-simd.md (*aarch64_mls_elt_merge<mode>):
30670 (aarch64_mls_n<mode>): This.
30671 * config/aarch64/arm_neon.h (vmls_n_s16): Use RTL builtin
30673 (vmls_n_s32): Likewise.
30674 (vmls_n_u16): Likewise.
30675 (vmls_n_u32): Likewise.
30676 (vmlsq_n_s16): Likewise.
30677 (vmlsq_n_s32): Likewise.
30678 (vmlsq_n_u16): Likewise.
30679 (vmlsq_n_u32): Likewise.
30681 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
30683 * config/aarch64/aarch64-simd-builtins.def: Add mls builtin
30685 * config/aarch64/arm_neon.h (vmls_s8): Use RTL builtin rather
30687 (vmls_s16): Likewise.
30688 (vmls_s32): Likewise.
30689 (vmls_u8): Likewise.
30690 (vmls_u16): Likewise.
30691 (vmls_u32): Likewise.
30692 (vmlsq_s8): Likewise.
30693 (vmlsq_s16): Likewise.
30694 (vmlsq_s32): Likewise.
30695 (vmlsq_u8): Likewise.
30696 (vmlsq_u16): Likewise.
30697 (vmlsq_u32): Likewise.
30699 2021-01-27 Jonathan Wright <jonathan.wright@arm.com>
30701 * config/aarch64/aarch64-simd-builtins.def: Add mla_n builtin
30703 * config/aarch64/aarch64-simd.md (*aarch64_mla_elt_merge<mode>):
30705 (aarch64_mla_n<mode>): This.
30706 * config/aarch64/arm_neon.h (vmla_n_s16): Use RTL builtin
30708 (vmla_n_s32): Likewise.
30709 (vmla_n_u16): Likewise.
30710 (vmla_n_u32): Likewise.
30711 (vmlaq_n_s16): Likewise.
30712 (vmlaq_n_s32): Likewise.
30713 (vmlaq_n_u16): Likewise.
30714 (vmlaq_n_u32): Likewise.
30716 2021-01-27 liuhongt <hongtao.liu@intel.com>
30719 * config/i386/sse.md (sse2_gt<mode>3): Drop !TARGET_XOP in condition.
30720 (*sse2_eq<mode>3): Ditto.
30722 2021-01-27 Jakub Jelinek <jakub@redhat.com>
30724 * tree-pass.h (PROP_trees): Rename to ...
30725 (PROP_gimple): ... this.
30726 * cfgexpand.c (pass_data_expand): Replace PROP_trees with PROP_gimple.
30727 * passes.c (execute_function_dump, execute_function_todo,
30728 execute_one_ipa_transform_pass, execute_one_pass): Likewise.
30729 * varpool.c (ctor_for_folding): Likewise.
30731 2021-01-27 Jakub Jelinek <jakub@redhat.com>
30733 PR tree-optimization/97260
30734 * varpool.c: Include tree-pass.h.
30735 (ctor_for_folding): In GENERIC return DECL_INITIAL for TREE_READONLY
30736 non-TREE_SIDE_EFFECTS automatic variables.
30738 2021-01-26 Paul Fee <paul.f.fee@gmail.com>
30740 * doc/cpp.texi (__cplusplus): Document value for -std=c++23
30742 * doc/invoke.texi: Document -std=c++23 and -std=gnu++23.
30743 * dwarf2out.c (highest_c_language): Recognise C++20 and C++23.
30744 (gen_compile_unit_die): Recognise C++23.
30746 2021-01-26 Jakub Jelinek <jakub@redhat.com>
30749 * dwarf2asm.c (dw2_assemble_integer): Cast DWARF2_ADDR_SIZE to int
30752 2021-01-26 Jakub Jelinek <jakub@redhat.com>
30755 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
30756 Use UINTVAL (shft_amnt) and UINTVAL (mask) instead of INTVAL (shft_amnt)
30757 and INTVAL (mask). Add && INTVAL (mask) > 0 condition.
30759 2021-01-26 Richard Biener <rguenther@suse.de>
30761 * gimple-pretty-print.c (dump_binary_rhs): Handle
30762 VEC_WIDEN_{PLUS,MINUS}_{LO,HI}_EXPR.
30764 2021-01-26 Richard Biener <rguenther@suse.de>
30766 PR middle-end/98726
30767 * tree.h (vector_cst_int_elt): Remove.
30768 * tree.c (vector_cst_int_elt): Use poly_wide_int for computations,
30771 2021-01-26 Andrew Stubbs <ams@codesourcery.com>
30773 * config/gcn/gcn.c (gcn_expand_reduc_scalar): Use move instructions
30774 for V64DFmode min/max reductions.
30776 2021-01-26 Jakub Jelinek <jakub@redhat.com>
30778 * dwarf2asm.c (dw2_assemble_integer): Handle size twice as large
30779 as DWARF2_ADDR_SIZE if x is not a scalar int by emitting it as
30780 two halves, one with x and the other with const0_rtx, ordered
30781 depending on endianity.
30783 2021-01-26 Alexandre Oliva <oliva@adacore.com>
30785 * gimplify.c (gimplify_decl_expr): Skip asan marking calls for
30786 temporaries not seen in binding block, and not about to be
30787 added as gimple variables.
30789 2021-01-25 Martin Sebor <msebor@redhat.com>
30792 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust warning text.
30794 2021-01-25 Martin Liska <mliska@suse.cz>
30796 * value-prof.c (get_nth_most_common_value): Use %s instead
30799 2021-01-25 Jakub Jelinek <jakub@redhat.com>
30802 * configure.ac (HAVE_AS_GDWARF_5_DEBUG_FLAG): Only define if
30803 readelf -wi is able to read the emitted .debug_info back.
30804 * configure: Regenerated.
30806 2021-01-25 Martin Liska <mliska@suse.cz>
30808 PR gcov-profile/98739
30809 * common.opt: Add missing sign symbol.
30810 * value-prof.c (get_nth_most_common_value): Restore handling
30811 of PROFILE_REPRODUCIBILITY_PARALLEL_RUNS and
30812 PROFILE_REPRODUCIBILITY_MULTITHREADED.
30814 2021-01-25 Richard Biener <rguenther@suse.de>
30816 PR middle-end/98807
30817 * tree.c (vector_element_bits): Always use precision of
30818 the element type for boolean vectors.
30820 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
30822 * config/rtems.h (STARTFILE_SPEC): Remove qnolinkcmds.
30823 (ENDFILE_SPEC): Evaluate qnolinkcmds.
30825 2021-01-25 Sebastian Huber <sebastian.huber@embedded-brains.de>
30827 * config/rtems.h (STARTFILE_SPEC): Remove nostdlib and
30828 nostartfiles handling since this is already done by
30829 LINK_COMMAND_SPEC. Evaluate qnolinkcmds.
30830 (ENDFILE_SPEC): Remove nostdlib and nostartfiles handling since this
30831 is already done by LINK_COMMAND_SPEC.
30832 (LIB_SPECS): Remove nostdlib and nodefaultlibs handling since
30833 this is already done by LINK_COMMAND_SPEC. Remove qnolinkcmds
30836 2021-01-25 Jakub Jelinek <jakub@redhat.com>
30839 * fold-const-call.c (host_size_t_cst_p): Renamed to ...
30840 (size_t_cst_p): ... this. Check and store unsigned HOST_WIDE_INT
30841 value rather than host size_t.
30842 (fold_const_call): Change type of s2 from size_t to
30843 unsigned HOST_WIDE_INT. Use size_t_cst_p instead of
30844 host_size_t_cst_p. For strncmp calls, pass MIN (s2, SIZE_MAX)
30845 instead of s2 as last argument.
30847 2021-01-25 Tamar Christina <tamar.christina@arm.com>
30849 * config/arm/iterators.md (rotsplit1, rotsplit2, conj_op, fcmac1,
30850 VCMLA_OP, VCMUL_OP): New.
30851 * config/arm/mve.md (mve_vcmlaq<mve_rot><mode>): Support vec_dup 0.
30852 * config/arm/neon.md (cmul<conj_op><mode>3): New.
30853 * config/arm/unspecs.md (UNSPEC_VCMLA_CONJ, UNSPEC_VCMLA180_CONJ,
30854 UNSPEC_VCMUL_CONJ): New.
30855 * config/arm/vec-common.md (cmul<conj_op><mode>3, arm_vcmla<rot><mode>,
30856 cml<fcmac1><conj_op><mode>4): New.
30858 2021-01-23 Jakub Jelinek <jakub@redhat.com>
30861 * config/rs6000/mmintrin.h (__m64): Add __may_alias__ attribute.
30863 2021-01-22 Jonathan Wright <jonathan.wright@arm.com>
30865 * config/aarch64/aarch64-simd-builtins.def: Add mla builtin
30867 * config/aarch64/arm_neon.h (vmla_s8): Use RTL builtin rather
30869 (vmla_s16): Likewise.
30870 (vmla_s32): Likewise.
30871 (vmla_u8): Likewise.
30872 (vmla_u16): Likewise.
30873 (vmla_u32): Likewise.
30874 (vmlaq_s8): Likewise.
30875 (vmlaq_s16): Likewise.
30876 (vmlaq_s32): Likewise.
30877 (vmlaq_u8): Likewise.
30878 (vmlaq_u16): Likewise.
30879 (vmlaq_u32): Likewise.
30881 2021-01-22 David Malcolm <dmalcolm@redhat.com>
30883 * doc/invoke.texi (GCC_EXTRA_DIAGNOSTIC_OUTPUT): Add @findex
30886 2021-01-22 Jakub Jelinek <jakub@redhat.com>
30889 * dwarf2out.c (output_file_names): For -gdwarf-5, if there are no
30890 filenames to emit, still emit the required 0 index directory and
30891 filename entries that match DW_AT_comp_dir and DW_AT_name of the
30894 2021-01-22 Marek Polacek <polacek@redhat.com>
30897 * doc/invoke.texi: Update C++ ABI Version 15 description.
30899 2021-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30901 PR tree-optimization/98766
30902 * tree-ssa-math-opts.c (convert_mult_to_fma): Use maybe_le when
30903 comparing against type size with param_avoid_fma_max_bits.
30905 2021-01-22 Richard Biener <rguenther@suse.de>
30907 PR middle-end/98793
30908 * tree.c (vector_element_bits): Key single-bit bool vector on
30909 integer mode rather than not vector mode.
30911 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
30914 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
30915 Generate ARRAY_REF(VIEW_CONVERT_EXPR) for P8 and later
30917 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): Update
30918 to call different path for P8 and P9.
30919 (rs6000_expand_vector_set_var_p9): New function.
30920 (rs6000_expand_vector_set_var_p8): New function.
30922 2021-01-22 Xionghu Luo <luoxhu@linux.ibm.com>
30926 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
30927 Ajdust variable index vec_insert from address dereference to
30928 ARRAY_REF(VIEW_CONVERT_EXPR) tree expression.
30929 * config/rs6000/rs6000-protos.h (rs6000_expand_vector_set_var):
30931 * config/rs6000/rs6000.c (rs6000_expand_vector_set_var): New function.
30933 2021-01-22 Martin Liska <mliska@suse.cz>
30935 PR gcov-profile/98739
30936 * profile.c (compute_value_histograms): Drop time profile for
30937 -fprofile-reproducible=multithreaded.
30939 2021-01-22 Nathan Sidwell <nathan@acm.org>
30941 * gcc.c (process_command): Don't check OPT_SPECIAL_input_file
30944 2021-01-22 Richard Biener <rguenther@suse.de>
30946 PR middle-end/98773
30947 * tree-data-ref.c (initalize_matrix_A): Revert previous
30948 change, retaining failing on HOST_WIDE_INT_MIN CHREC_RIGHT.
30950 2021-01-22 Jakub Jelinek <jakub@redhat.com>
30952 PR tree-optimization/90248
30953 * match.pd (X cmp 0.0 ? 1.0 : -1.0 -> copysign(1, +-X),
30954 X cmp 0.0 ? -1.0 : +1.0 -> copysign(1, -+X)): Remove
30956 (X * (X cmp 0.0 ? 1.0 : -1.0) -> +-abs(X),
30957 X * (X cmp 0.0 ? -1.0 : 1.0) -> +-abs(X)): New simplifications.
30959 2021-01-22 Jakub Jelinek <jakub@redhat.com>
30961 PR tree-optimization/98255
30962 * tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
30963 extend index - low_bound from sizetype's precision rather than index
30965 (get_addr_base_and_unit_offset_1): Likewise.
30966 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
30967 * gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
30969 2021-01-22 Richard Biener <rguenther@suse.de>
30971 PR tree-optimization/98786
30972 * tree-ssa-phiopt.c (factor_out_conditional_conversion): Avoid
30973 adding new uses of abnormals. Verify we deal with a conditional
30976 2021-01-22 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
30979 * optc-save-gen.awk: Add arm_fp16_format to checked_options.
30981 2021-01-22 liuhongt <hongtao.liu@intel.com>
30985 * config/i386/sse.md (VI_128_256): New mode iterator.
30986 (*avx_cmp<mode>3_1, *avx_cmp<mode>3_2, *avx_cmp<mode>3_3,
30987 *avx_cmp<mode>3_4, *avx2_eq<mode>3, *avx2_pcmp<mode>3_1,
30988 *avx2_pcmp<mode>3_2, *avx2_gt<mode>3): New
30989 define_insn_and_split to lower avx512 vector comparison to avx
30990 version when dest is vector.
30991 (*<avx512>_cmp<mode>3,*<avx512>_cmp<mode>3,*<avx512>_ucmp<mode>3):
30992 define_insn_and_split for negating the comparison result.
30993 * config/i386/predicates.md (float_vector_all_ones_operand):
30995 * config/i386/i386-expand.c (ix86_expand_sse_movcc): Use
30996 general NOT operator without UNSPEC_MASKOP.
30998 2021-01-21 Vladimir N. Makarov <vmakarov@redhat.com>
31000 PR rtl-optimization/98777
31001 * lra-int.h (lra_pmode_pseudo): New extern.
31002 * lra.c (lra_pmode_pseudo): New global.
31004 * lra-eliminations.c (eliminate_regs_in_insn): Use it.
31006 2021-01-21 Ilya Leoshkevich <iii@linux.ibm.com>
31008 * fwprop.c (fwprop_propagation::classify_result): Allow
31009 (subreg (mem)) simplifications.
31011 2021-01-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31013 * config/aarch64/aarch64-simd.md (aarch64_sqdml<SBINQOPS:as>l<mode>):
31015 (aarch64_sqdmlal<mode>): ... This...
31016 (aarch64_sqdmlsl<mode>): ... And this.
31017 (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Split into...
31018 (aarch64_sqdmlal_lane<mode>): ... This...
31019 (aarch64_sqdmlsl_lane<mode>): ... And this.
31020 (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Split into...
31021 (aarch64_sqdmlsl_laneq<mode>): ... This...
31022 (aarch64_sqdmlal_laneq<mode>): ... And this.
31023 (aarch64_sqdml<SBINQOPS:as>l_n<mode>): Split into...
31024 (aarch64_sqdmlsl_n<mode>): ... This...
31025 (aarch64_sqdmlal_n<mode>): ... And this.
31026 (aarch64_sqdml<SBINQOPS:as>l2<mode>_internal): Split into...
31027 (aarch64_sqdmlal2<mode>_internal): ... This...
31028 (aarch64_sqdmlsl2<mode>_internal): ... And this.
31030 2021-01-21 Christophe Lyon <christophe.lyon@linaro.org>
31032 * config/arm/arm_mve.h (__arm_vcmpneq_s8): Fix return type.
31034 2021-01-21 Andrea Corallo <andrea.corallo@arm.com>
31037 * doc/sourcebuild.texi (arm_thumb2_no_arm_v8_1_lob): Document.
31039 2021-01-21 liuhongt <hongtao.liu@intel.com>
31041 PR rtl-optimization/98694
31042 * regcprop.c (copy_value): If SRC had been assigned a mode
31043 narrower than the copy, we can't link DEST into the chain even
31044 they have same hard_regno_nregs(i.e. HImode/SImode in i386
31047 2021-01-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31049 * config/aarch64/aarch64-simd.md (aarch64_get_lane<mode>):
31050 Convert to define_insn_and_split. Split into simple move when moving
31053 2021-01-20 Segher Boessenkool <segher@kernel.crashing.org>
31055 * config/rs6000/rs6000.c (rs6000_emit_le_vsx_store): Change assert.
31056 Adjust comment. Simplify code.
31058 2021-01-20 Jakub Jelinek <jakub@redhat.com>
31061 * dwarf2out.c (reset_indirect_string): Also reset indirect strings
31062 with DW_FORM_line_strp form.
31063 (prune_unused_types_update_strings): Don't add into debug_str_hash
31064 indirect strings with DW_FORM_line_strp form.
31065 (adjust_name_comp_dir): New function.
31066 (dwarf2out_finish): Call it on CU DIEs after resetting
31067 debug_line_str_hash.
31069 2021-01-20 Vladimir N. Makarov <vmakarov@redhat.com>
31071 PR rtl-optimization/98722
31072 * lra-eliminations.c (eliminate_regs_in_insn): Check that target
31073 has no 3-op add insn to transform insns containing two pluses.
31075 2021-01-20 Richard Biener <rguenther@suse.de>
31077 * hwint.h (add_hwi): New function.
31078 (mul_hwi): Likewise.
31079 * tree-data-ref.c (initialize_matrix_A): Properly translate
31080 tree constants and avoid HOST_WIDE_INT_MIN.
31081 (lambda_matrix_row_add): Avoid undefined integer overflow
31082 and return true on such overflow.
31083 (lambda_matrix_right_hermite): Handle overflow from
31084 lambda_matrix_row_add gracefully. Simplify previous fix.
31085 (analyze_subscript_affine_affine): Likewise.
31087 2021-01-20 Eugene Rozenfeld <erozen@microsoft.com>
31089 PR tree-optimization/96674
31090 * match.pd: New patterns: x < y || y == XXX_MIN --> x <= y - 1
31091 x >= y && y != XXX_MIN --> x > y - 1
31093 2021-01-20 Richard Sandiford <richard.sandiford@arm.com>
31095 PR tree-optimization/98535
31096 * tree-vect-slp.c (duplicate_and_interleave): Use quick_grow_cleared.
31097 If the high and low permutes are the same, remove the high permutes
31098 from the working set and only continue with the low ones.
31100 2021-01-20 Jakub Jelinek <jakub@redhat.com>
31102 PR tree-optimization/98721
31103 * builtins.c (access_ref::inform_access): Don't assume
31104 SSA_NAME_IDENTIFIER must be non-NULL. Print messages about
31105 object whenever allocfn is NULL, rather than only when DECL_P
31106 is true. Use %qE instead of %qD for that. Formatting fixes.
31108 2021-01-20 Richard Biener <rguenther@suse.de>
31110 PR tree-optimization/98758
31111 * tree-data-ref.c (int_divides_p): Use lambda_int arguments.
31112 (lambda_matrix_right_hermite): Avoid undefinedness with
31113 signed integer abs and multiplication.
31114 (analyze_subscript_affine_affine): Use lambda_int.
31116 2021-01-20 David Malcolm <dmalcolm@redhat.com>
31119 * dwarf2out.c (output_line_info): Rename static variable
31120 "generation", moving it out of the function to...
31121 (output_line_info_generation): New.
31122 (init_sections_and_labels): Likewise, renaming the variable to...
31123 (init_sections_and_labels_generation): New.
31124 (dwarf2out_c_finalize): Reset the new variables.
31126 2021-01-19 Martin Sebor <msebor@redhat.com>
31128 PR middle-end/98664
31129 * tree-ssa-live.c (remove_unused_scope_block_p): Keep scopes for
31130 all functions, even if they're not declared artificial or inline.
31131 * tree.c (tree_inlined_location): Use macro expansion location
31132 only if scope traversal fails to expose one.
31134 2021-01-19 Richard Sandiford <richard.sandiford@arm.com>
31136 PR rtl-optimization/92294
31137 * alias.c (compare_base_symbol_refs): Take an extra parameter
31138 and add the distance between two symbols to it. Enshrine in
31139 comments that -1 means "either 0 or 1, but we can't tell
31140 which at compile time".
31141 (memrefs_conflict_p): Update call accordingly.
31142 (rtx_equal_for_memref_p): Likewise. Take the distance between symbols
31145 2021-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31147 * config/aarch64/aarch64-simd-builtins.def (sqshl, uqshl,
31148 sqrshl, uqrshl, sqadd, uqadd, sqsub, uqsub, suqadd, usqadd, sqmovn,
31149 uqmovn, sqxtn2, uqxtn2, sqabs, sqneg, sqdmlal, sqdmlsl, sqdmlal_lane,
31150 sqdmlsl_lane, sqdmlal_laneq, sqdmlsl_laneq, sqdmlal_n, sqdmlsl_n,
31151 sqdmlal2, sqdmlsl2, sqdmlal2_lane, sqdmlsl2_lane, sqdmlal2_laneq,
31152 sqdmlsl2_laneq, sqdmlal2_n, sqdmlsl2_n, sqdmull, sqdmull_lane,
31153 sqdmull_laneq, sqdmull_n, sqdmull2, sqdmull2_lane, sqdmull2_laneq,
31154 sqdmull2_n, sqdmulh, sqrdmulh, sqdmulh_lane, sqdmulh_laneq,
31155 sqrdmulh_lane, sqrdmulh_laneq, sqshrun_n, sqrshrun_n, sqshrn_n,
31156 uqshrn_n, sqrshrn_n, uqrshrn_n, sqshlu_n, sqshl_n, uqshl_n, sqrdmlah,
31157 sqrdmlsh, sqrdmlah_lane, sqrdmlsh_lane, sqrdmlah_laneq, sqrdmlsh_laneq,
31158 sqmovun): Use NONE flags.
31160 2021-01-19 Richard Biener <rguenther@suse.de>
31163 * ipa-modref.c (analyze_stmt): Only record a summary for a
31166 2021-01-19 Richard Biener <rguenther@suse.de>
31168 PR middle-end/98638
31169 * tree-ssanames.c (fini_ssanames): Zero SSA_NAME_DEF_STMT.
31171 2021-01-19 Daniel Hellstrom <daniel@gaisler.com>
31173 * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
31174 built-in define __FIX_LEON3FT_TN0018.
31176 2021-01-19 Richard Biener <rguenther@suse.de>
31179 * tree-inline.c (tree_function_versioning): Set input_location
31180 to UNKNOWN_LOCATION throughout the function.
31182 2021-01-19 Tobias Burnus <tobias@codesourcery.com>
31185 * omp-low.c (lower_omp_target): Handle nonpointer is_device_ptr.
31187 2021-01-19 Martin Jambor <mjambor@suse.cz>
31190 * ipa-sra.c (ssa_name_only_returned_p): New parameter fun. Check
31191 whether non-call exceptions allow removal of a statement.
31192 (isra_analyze_call): Pass the appropriate function to
31193 ssa_name_only_returned_p.
31195 2021-01-19 Geng Qi <gengqi@linux.alibaba.com>
31197 * config/riscv/arch-canonicalize (longext_sort): New function for
31198 sorting 'multi-letter'.
31199 * config/riscv/multilib-generator: Adjusting the loop of 'alt' in
31200 'alts'. The 'arch' may not be the first of 'alts'.
31201 (_expand_combination): Add underline for the 'ext' without '*'.
31202 This is because, a single-letter extension can always be treated well
31203 with a '_' prefix, but it cannot be separated out if it is appended
31206 2021-01-18 Vladimir N. Makarov <vmakarov@redhat.com>
31209 * ira.c (ira): Skip abnormal critical edge splitting.
31211 2021-01-18 Jakub Jelinek <jakub@redhat.com>
31213 PR tree-optimization/98727
31214 * tree-ssa-math-opts.c (match_arith_overflow): Fix up computation of
31215 second .MUL_OVERFLOW operand for signed multiplication with overflow
31216 checking if the second operand of multiplication is not constant.
31218 2021-01-18 David Edelsohn <dje.gcc@gmail.com>
31220 * doc/invoke.texi (-gdwarf): TPF defaults to version 2 and AIX
31221 defaults to version 4.
31223 2021-01-18 David Malcolm <dmalcolm@redhat.com>
31225 * attribs.h (fndecl_dealloc_argno): New decl.
31226 * builtins.c (call_dealloc_argno): Split out second half of
31228 (fndecl_dealloc_argno): New.
31229 * doc/extend.texi (Common Function Attributes): Document the
31230 interaction between the analyzer and the malloc attribute.
31231 * doc/invoke.texi (Static Analyzer Options): Likewise.
31233 2021-01-17 David Edelsohn <dje.gcc@gmail.com>
31235 * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Override
31236 dwarf_version to 4.
31237 * config/rs6000/aix72.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
31239 2021-01-17 Martin Jambor <mjambor@suse.cz>
31242 * cgraph.c (clone_of_p): Check also former_clone_of as we climb
31245 2021-01-17 Mark Wielaard <mark@klomp.org>
31247 * common.opt (gdwarf-): Init(5).
31248 * doc/invoke.texi (-gdwarf): Document default to 5.
31250 2021-01-16 Kwok Cheung Yeung <kcy@codesourcery.com>
31252 * builtin-types.def
31253 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT): Rename
31255 (BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR):
31256 ...this. Add extra argument.
31257 * gimplify.c (omp_default_clause): Ensure that event handle is
31258 firstprivate in a task region.
31259 (gimplify_scan_omp_clauses): Handle OMP_CLAUSE_DETACH.
31260 (gimplify_adjust_omp_clauses): Likewise.
31261 * omp-builtins.def (BUILT_IN_GOMP_TASK): Change function type to
31262 BT_FN_VOID_OMPFN_PTR_OMPCPYFN_LONG_LONG_BOOL_UINT_PTR_INT_PTR.
31263 * omp-expand.c (expand_task_call): Add GOMP_TASK_FLAG_DETACH to flags
31264 if detach clause specified. Add detach argument when generating
31266 * omp-low.c (scan_sharing_clauses): Setup data environment for detach
31268 (finish_taskreg_scan): Move field for variable containing the event
31269 handle to the front of the struct.
31270 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_DETACH. Fix
31272 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
31273 OMP_CLAUSE_DETACH clause.
31274 (convert_local_omp_clauses): Handle OMP_CLAUSE_DETACH clause.
31275 * tree-pretty-print.c (dump_omp_clause): Handle OMP_CLAUSE_DETACH.
31276 * tree.c (omp_clause_num_ops): Add entry for OMP_CLAUSE_DETACH.
31278 (omp_clause_code_name): Add entry for OMP_CLAUSE_DETACH. Fix
31280 (walk_tree_1): Handle OMP_CLAUSE_DETACH.
31282 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
31284 * config/nios2/t-rtems: Reset all MULTILIB_* variables. Shorten
31285 multilib directory names. Use MULTILIB_REQUIRED instead of
31286 MULTILIB_EXCEPTIONS. Add -mhw-mul -mhw-mulx -mhw-div
31287 -mcustom-fpu-cfg=fph2 multilib.
31289 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
31291 * config/nios2/nios2.c (NIOS2_FPU_CONFIG_NUM): Adjust value.
31292 (nios2_init_fpu_configs): Provide register values for new
31293 -mcustom-fpu-cfg=fph2 option variant.
31294 * doc/invoke.texi (-mcustom-fpu-cfg=fph2): Document new option
31297 2021-01-16 Sebastian Huber <sebastian.huber@embedded-brains.de>
31299 * config/nios2/nios2.c (nios2_custom_check_insns): Remove
31300 custom instruction warnings.
31302 2021-01-16 Jakub Jelinek <jakub@redhat.com>
31304 PR tree-optimization/96669
31305 * match.pd ((CST << x) & 1 -> x == 0): New simplification.
31307 2021-01-16 Jakub Jelinek <jakub@redhat.com>
31309 PR tree-optimization/96271
31310 * passes.def: Pass false argument to first two pass_cd_dce
31311 instances and true to last instance. Add comment that
31312 last instance rewrites no longer addressed locals.
31313 * tree-ssa-dce.c (pass_cd_dce): Add update_address_taken_p member and
31315 (pass_cd_dce::set_pass_param): New method.
31316 (pass_cd_dce::execute): Return TODO_update_address_taken from
31317 last cd_dce instance.
31319 2021-01-15 Carl Love <cel@us.ibm.com>
31321 * config/rs6000/altivec.h (vec_mulh, vec_div, vec_dive, vec_mod):
31323 * config/rs6000/altivec.md (VIlong): Move define to file vsx.md.
31324 * config/rs6000/rs6000-builtin.def (DIVES_V4SI, DIVES_V2DI,
31325 DIVEU_V4SI, DIVEU_V2DI, DIVS_V4SI, DIVS_V2DI, DIVU_V4SI,
31326 DIVU_V2DI, MODS_V2DI, MODS_V4SI, MODU_V2DI, MODU_V4SI,
31327 MULHS_V2DI, MULHS_V4SI, MULHU_V2DI, MULHU_V4SI, MULLD_V2DI):
31328 Add builtin define.
31329 (MULH, DIVE, MOD): Add new BU_P10_OVERLOAD_2 definitions.
31330 * config/rs6000/rs6000-call.c (VSX_BUILTIN_VEC_DIV,
31331 VSX_BUILTIN_VEC_DIVE, P10_BUILTIN_VEC_MOD, P10_BUILTIN_VEC_MULH):
31332 New overloaded definitions.
31333 (builtin_function_type) [P10V_BUILTIN_DIVEU_V4SI,
31334 P10V_BUILTIN_DIVEU_V2DI, P10V_BUILTIN_DIVU_V4SI,
31335 P10V_BUILTIN_DIVU_V2DI, P10V_BUILTIN_MODU_V2DI,
31336 P10V_BUILTIN_MODU_V4SI, P10V_BUILTIN_MULHU_V2DI,
31337 P10V_BUILTIN_MULHU_V4SI]: Add case
31338 statement for builtins.
31339 * config/rs6000/rs6000.md (bits): Add new attribute sizes V4SI, V2DI.
31340 * config/rs6000/vsx.md (VIlong): Moved from config/rs6000/altivec.md.
31341 (UNSPEC_VDIVES, UNSPEC_VDIVEU): New unspec definitions.
31342 (vsx_mul_v2di): Add if TARGET_POWER10 statement.
31343 (vsx_udiv_v2di): Add if TARGET_POWER10 statement.
31344 (dives_<mode>, diveu_<mode>, div<mode>3, uvdiv<mode>3,
31345 mods_<mode>, modu_<mode>, mulhs_<mode>, mulhu_<mode>, mulv2di3):
31346 Add define_insn, mode is VIlong.
31347 * doc/extend.texi (vec_mulh, vec_mul, vec_div, vec_dive, vec_mod):
31348 Add builtin descriptions.
31350 2021-01-15 Eric Botcazou <ebotcazou@adacore.com>
31352 * final.c (final_start_function_1): Reset force_source_line.
31354 2021-01-15 Jakub Jelinek <jakub@redhat.com>
31356 PR tree-optimization/96669
31357 * match.pd (((1 << A) & 1) != 0 -> A == 0,
31358 ((1 << A) & 1) == 0 -> A != 0): Generalize for 1s replaced by
31359 possibly different power of two constants and to right shift too.
31361 2021-01-15 Jakub Jelinek <jakub@redhat.com>
31363 PR tree-optimization/96681
31364 * match.pd ((x < 0) ^ (y < 0) to (x ^ y) < 0): New simplification.
31365 ((x >= 0) ^ (y >= 0) to (x ^ y) < 0): Likewise.
31366 ((x < 0) ^ (y >= 0) to (x ^ y) >= 0): Likewise.
31367 ((x >= 0) ^ (y < 0) to (x ^ y) >= 0): Likewise.
31369 2021-01-15 Alexandre Oliva <oliva@adacore.com>
31371 * opts.c (gen_command_line_string): Exclude -dumpbase-ext.
31373 2021-01-15 Tamar Christina <tamar.christina@arm.com>
31375 * config/aarch64/aarch64-simd.md (cml<fcmac1><conj_op><mode>4,
31376 cmul<conj_op><mode>3): New.
31377 * config/aarch64/iterators.md (UNSPEC_FCMUL,
31378 UNSPEC_FCMUL180, UNSPEC_FCMLA_CONJ, UNSPEC_FCMLA180_CONJ,
31379 UNSPEC_CMLA_CONJ, UNSPEC_CMLA180_CONJ, UNSPEC_CMUL, UNSPEC_CMUL180,
31380 FCMLA_OP, FCMUL_OP, conj_op, rotsplit1, rotsplit2, fcmac1, sve_rot1,
31381 sve_rot2, SVE2_INT_CMLA_OP, SVE2_INT_CMUL_OP, SVE2_INT_CADD_OP): New.
31382 (rot): Add UNSPEC_FCMUL, UNSPEC_FCMUL180.
31383 (rot_op): Renamed to conj_op.
31384 * config/aarch64/aarch64-sve.md (cml<fcmac1><conj_op><mode>4,
31385 cmul<conj_op><mode>3): New.
31386 * config/aarch64/aarch64-sve2.md (cml<fcmac1><conj_op><mode>4,
31387 cmul<conj_op><mode>3): New.
31389 2021-01-15 David Malcolm <dmalcolm@redhat.com>
31393 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
31394 Escape the tempfile name when constructing the expected output.
31396 2021-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31398 * config/aarch64/aarch64-simd.md (*aarch64_<su>mlsl_hi<mode>):
31400 (aarch64_<su>mlsl_hi<mode>): ... This.
31401 (aarch64_<su>mlsl_hi<mode>): Define.
31402 (*aarch64_<su>mlsl<mode): Rename to...
31403 (aarch64_<su>mlsl<mode): ... This.
31404 * config/aarch64/aarch64-simd-builtins.def (smlsl, umlsl,
31405 smlsl_hi, umlsl_hi): Define builtins.
31406 * config/aarch64/arm_neon.h (vmlsl_high_s8, vmlsl_high_s16,
31407 vmlsl_high_s32, vmlsl_high_u8, vmlsl_high_u16, vmlsl_high_u32,
31408 vmlsl_s8, vmlsl_s16, vmlsl_s32, vmlsl_u8,
31409 vmlsl_u16, vmlsl_u32): Reimplement with builtins.
31411 2021-01-15 Uroš Bizjak <ubizjak@gmail.com>
31413 * config/i386/i386-c.c (ix86_target_macros):
31414 Use cpp_define_formatted for __SIZEOF_FLOAT80__ definition.
31416 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
31419 * config.gcc (aarch64*-*-*): Add aarch64-cc-fusion.o to extra_objs.
31420 * Makefile.in (RTL_SSA_H): New variable.
31421 * config/aarch64/t-aarch64 (aarch64-cc-fusion.o): New rule.
31422 * config/aarch64/aarch64-protos.h (make_pass_cc_fusion): Declare.
31423 * config/aarch64/aarch64-passes.def: Add pass_cc_fusion after
31425 * config/aarch64/aarch64-cc-fusion.cc: New file.
31427 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
31429 * recog.h (insn_change_watermark::~insn_change_watermark): Avoid
31430 calling cancel_changes for changes that no longer exist.
31432 2021-01-15 Richard Sandiford <richard.sandiford@arm.com>
31434 * rtl-ssa/functions.h (function_info::ref_defs): Rename to...
31435 (function_info::reg_defs): ...this.
31436 * rtl-ssa/member-fns.inl (function_info::ref_defs): Rename to...
31437 (function_info::reg_defs): ...this.
31439 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
31442 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
31444 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
31447 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
31450 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
31452 2021-01-15 Richard Biener <rguenther@suse.de>
31454 PR tree-optimization/96376
31455 * tree-vect-stmts.c (get_load_store_type): Disregard alignment
31456 for VMAT_INVARIANT.
31458 2021-01-15 Martin Liska <mliska@suse.cz>
31460 * doc/install.texi: Document that some tests need pytest module.
31461 * doc/sourcebuild.texi: Likewise.
31463 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
31466 * config/arm/arm_neon.h (vceqz_p64, vceqq_p64, vceqzq_p64): New.
31468 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
31470 * config/arm/mve.md (mve_vshrq_n_s<mode>_imm): New entry.
31471 (mve_vshrq_n_u<mode>_imm): Likewise.
31472 * config/arm/neon.md (vashr<mode>3, vlshr<mode>3): Move to ...
31473 * config/arm/vec-common.md: ... here.
31475 2021-01-15 Christophe Lyon <christophe.lyon@linaro.org>
31477 * config/arm/mve.md (mve_vshlq_<supf><mode>): Move to
31479 * config/arm/neon.md (vashl<mode>3): Delete.
31480 * config/arm/vec-common.md (mve_vshlq_<supf><mode>): New.
31481 (vasl<mode>3): New expander.
31483 2021-01-15 Richard Biener <rguenther@suse.de>
31485 PR tree-optimization/98685
31486 * tree-vect-slp.c (vect_schedule_slp_node): Refactor handling
31487 of vector extern defs.
31489 2021-01-14 David Malcolm <dmalcolm@redhat.com>
31492 * diagnostic.c (diagnostic_kind_text): Break out this array
31494 (diagnostic_build_prefix): ...here.
31495 (fancy_abort): Detect when diagnostic_initialize has not yet been
31496 called and fall back to a minimal implementation of printing the
31497 ICE, rather than segfaulting in internal_error.
31499 2021-01-14 David Malcolm <dmalcolm@redhat.com>
31501 * diagnostic.c (diagnostic_initialize): Eliminate
31502 parseable_fixits_p in favor of initializing extra_output_kind from
31503 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
31504 (convert_column_unit): New function, split out from...
31505 (diagnostic_converted_column): ...this.
31506 (print_parseable_fixits): Add "column_unit" and "tabstop" params.
31507 Use them to call convert_column_unit on the column values.
31508 (diagnostic_report_diagnostic): Eliminate conditional on
31509 parseable_fixits_p in favor of a switch statement on
31510 extra_output_kind, passing the appropriate values to the new
31511 params of print_parseable_fixits.
31512 (selftest::test_print_parseable_fixits_none): Update for new
31513 params of print_parseable_fixits.
31514 (selftest::test_print_parseable_fixits_insert): Likewise.
31515 (selftest::test_print_parseable_fixits_remove): Likewise.
31516 (selftest::test_print_parseable_fixits_replace): Likewise.
31517 (selftest::test_print_parseable_fixits_bytes_vs_display_columns):
31519 (selftest::diagnostic_c_tests): Call it.
31520 * diagnostic.h (enum diagnostics_extra_output_kind): New.
31521 (diagnostic_context::parseable_fixits_p): Delete field in favor
31523 (diagnostic_context::extra_output_kind): ...this new field.
31524 * doc/invoke.texi (Environment Variables): Add
31525 GCC_EXTRA_DIAGNOSTIC_OUTPUT.
31526 * opts.c (common_handle_option): Update handling of
31527 OPT_fdiagnostics_parseable_fixits for change to diagnostic_context
31530 2021-01-14 Tamar Christina <tamar.christina@arm.com>
31532 * tree-vect-slp-patterns.c (class complex_operations_pattern,
31533 complex_operations_pattern::matches,
31534 complex_operations_pattern::recognize,
31535 complex_operations_pattern::build): New.
31536 (slp_patterns): Use it.
31538 2021-01-14 Tamar Christina <tamar.christina@arm.com>
31540 * internal-fn.def (COMPLEX_FMS, COMPLEX_FMS_CONJ): New.
31541 * optabs.def (cmls_optab, cmls_conj_optab): New.
31542 * doc/md.texi: Document them.
31543 * tree-vect-slp-patterns.c (class complex_fms_pattern,
31544 complex_fms_pattern::matches, complex_fms_pattern::recognize,
31545 complex_fms_pattern::build): New.
31547 2021-01-14 Tamar Christina <tamar.christina@arm.com>
31549 * internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
31550 * optabs.def (cmla_optab, cmla_conj_optab): New.
31551 * doc/md.texi: Document them.
31552 * tree-vect-slp-patterns.c (vect_match_call_p,
31553 class complex_fma_pattern, vect_slp_reset_pattern,
31554 complex_fma_pattern::matches, complex_fma_pattern::recognize,
31555 complex_fma_pattern::build): New.
31557 2021-01-14 Tamar Christina <tamar.christina@arm.com>
31559 * internal-fn.def (COMPLEX_MUL, COMPLEX_MUL_CONJ): New.
31560 * optabs.def (cmul_optab, cmul_conj_optab): New.
31561 * doc/md.texi: Document them.
31562 * tree-vect-slp-patterns.c (vect_match_call_complex_mla,
31563 vect_normalize_conj_loc, is_eq_or_top, vect_validate_multiplication,
31564 vect_build_combine_node, class complex_mul_pattern,
31565 complex_mul_pattern::matches, complex_mul_pattern::recognize,
31566 complex_mul_pattern::build): New.
31568 2021-01-14 Tamar Christina <tamar.christina@arm.com>
31570 * tree-vect-slp.c (optimize_load_redistribution_1): New.
31571 (optimize_load_redistribution, vect_is_slp_load_node): New.
31572 (vect_match_slp_patterns): Use it.
31574 2021-01-14 Tamar Christina <tamar.christina@arm.com>
31576 * tree-vect-slp-patterns.c (complex_add_pattern::build):
31579 2021-01-14 Thomas Schwinge <thomas@codesourcery.com>
31581 * config/gcn/mkoffload.c (main): Create an offload image only in
31582 64-bit configurations.
31584 2021-01-14 H.J. Lu <hjl.tools@gmail.com>
31587 * config/i386/i386-options.c (ix86_option_override_internal):
31588 Issue an error for -fcf-protection with CF_BRANCH when compiling
31589 for 32-bit non-TARGET_CMOV targets.
31591 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
31594 * config/i386/i386-options.c (ix86_valid_target_attribute_inner_p):
31595 Remove declaration and initialization of shadow variable "ret".
31596 (ix86_option_override_internal): Remove delcaration of
31597 shadow variable "i". Redeclare shadowed variable to unsigned.
31598 * common/config/i386/i386-common.c (pta_size): Redeclare to unsigned.
31599 * config/i386/i386-builtins.c (get_builtin_code_for_version):
31600 Update for redeclaration.
31601 * config/i386/i386.h (pta_size): Ditto.
31603 2021-01-14 Richard Biener <rguenther@suse.de>
31605 PR tree-optimization/98674
31606 * tree-data-ref.c (base_supports_access_fn_components_p): New.
31607 (initialize_data_dependence_relation): For two bases without
31608 possible access fns resort to type size equality when determining
31609 shape compatibility.
31611 2021-01-14 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
31614 * config/arm/arm_neon.h: Replace calls to __builtin_vcge* by
31615 <=, >= operators in vcle and vcge intrinsics respectively.
31616 * config/arm/arm_neon_builtins.def: Remove entry for
31619 2021-01-14 Uroš Bizjak <ubizjak@gmail.com>
31622 * config/i386/i386-options.c (ix86_function_specific_save):
31623 Remove redundant assignment to opts->x_ix86_branch_cost.
31624 * config/i386/i386.c (ix86_prefetch_sse):
31625 Rename from x86_prefetch_sse. Update all uses.
31626 * config/i386/i386.h: Update for rename.
31627 * config/i386/i386-options.h: Ditto.
31629 2021-01-14 Jakub Jelinek <jakub@redhat.com>
31632 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3,
31633 *sse4_1_zero_extendv4hiv4si2_3, *sse4_1_zero_extendv2siv2di2_3):
31634 Use Bm instead of m for non-avx. Add isa attribute.
31636 2021-01-14 Jakub Jelinek <jakub@redhat.com>
31638 PR tree-optimization/96688
31639 * match.pd (~(X >> Y) -> ~X >> Y): New simplification if
31640 ~X can be simplified.
31642 2021-01-14 Richard Sandiford <richard.sandiford@arm.com>
31644 * tree-vect-stmts.c (vect_model_load_cost): Account for unused
31645 IFN_LOAD_LANES results.
31647 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31649 * config/aarch64/aarch64-simd.md (aarch64_<su>xtl<mode>):
31651 (aarch64_xtn<mode>): Likewise.
31652 * config/aarch64/aarch64-simd-builtins.def (sxtl, uxtl, xtn):
31655 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
31657 (vmovl_s16): Likewise.
31658 (vmovl_s32): Likewise.
31659 (vmovl_u8): Likewise.
31660 (vmovl_u16): Likewise.
31661 (vmovl_u32): Likewise.
31662 (vmovn_s16): Likewise.
31663 (vmovn_s32): Likewise.
31664 (vmovn_s64): Likewise.
31665 (vmovn_u16): Likewise.
31666 (vmovn_u32): Likewise.
31667 (vmovn_u64): Likewise.
31669 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31671 * config/aarch64/aarch64-simd.md (aarch64_<su>qxtn2<mode>_le):
31673 (aarch64_<su>qxtn2<mode>_be): Likewise.
31674 (aarch64_<su>qxtn2<mode>): Likewise.
31675 * config/aarch64/aarch64-simd-builtins.def (sqxtn2, uqxtn2):
31677 * config/aarch64/iterators.md (SAT_TRUNC): Define code_iterator.
31678 (su): Handle ss_truncate and us_truncate.
31679 * config/aarch64/arm_neon.h (vqmovn_high_s16): Reimplement using
31681 (vqmovn_high_s32): Likewise.
31682 (vqmovn_high_s64): Likewise.
31683 (vqmovn_high_u16): Likewise.
31684 (vqmovn_high_u32): Likewise.
31685 (vqmovn_high_u64): Likewise.
31687 2021-01-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
31689 * config/aarch64/aarch64-simd.md (aarch64_xtn2<mode>_le):
31691 (aarch64_xtn2<mode>_be): Likewise.
31692 (aarch64_xtn2<mode>): Likewise.
31693 * config/aarch64/aarch64-simd-builtins.def (xtn2): Define
31695 * config/aarch64/arm_neon.h (vmovn_high_s16): Reimplement using
31697 (vmovn_high_s32): Likewise.
31698 (vmovn_high_s64): Likewise.
31699 (vmovn_high_u16): Likewise.
31700 (vmovn_high_u32): Likewise.
31701 (vmovn_high_u64): Likewise.
31703 2021-01-13 Stafford Horne <shorne@gmail.com>
31705 * config/or1k/or1k.h (ASM_PREFERRED_EH_DATA_FORMAT): New macro.
31707 2021-01-13 Stafford Horne <shorne@gmail.com>
31709 * config/or1k/linux.h (TARGET_ASM_FILE_END): Define macro.
31711 2021-01-13 Stafford Horne <shorne@gmail.com>
31713 * config/or1k/or1k.h (TARGET_CPU_CPP_BUILTINS): Add builtin
31714 define for __or1k_hard_float__.
31716 2021-01-13 Stafford Horne <shorne@gmail.com>
31718 * config/or1k/or1k.h (NO_PROFILE_COUNTERS): Define as 1.
31719 (PROFILE_HOOK): Define to call _mcount.
31720 (FUNCTION_PROFILER): Change from abort to no-op.
31722 2021-01-13 Jakub Jelinek <jakub@redhat.com>
31724 PR tree-optimization/96691
31725 * match.pd ((~X | C) ^ D -> (X | C) ^ (~D ^ C),
31726 (~X & C) ^ D -> (X & C) ^ (D ^ C)): New simplifications if
31727 (~D ^ C) or (D ^ C) can be simplified.
31729 2021-01-13 Richard Biener <rguenther@suse.de>
31731 PR tree-optimization/92645
31732 * match.pd (BIT_FIELD_REF to conversion): Delay canonicalization
31733 until after vector lowering.
31735 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
31737 * config/aarch64/aarch64-sve.md (fnma<mode>4): Extend from SVE_FULL_I
31739 (@aarch64_pred_fnma<mode>, cond_fnma<mode>, *cond_fnma<mode>_2)
31740 (*cond_fnma<mode>_4, *cond_fnma<mode>_any): Likewise.
31742 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
31744 * config/aarch64/aarch64-sve.md (fma<mode>4): Extend from SVE_FULL_I
31746 (@aarch64_pred_fma<mode>, cond_fma<mode>, *cond_fma<mode>_2)
31747 (*cond_fma<mode>_4, *cond_fma<mode>_any): Likewise.
31749 2021-01-13 Richard Biener <rguenther@suse.de>
31751 PR tree-optimization/92645
31752 * tree-vect-slp.c (vect_build_slp_tree_1): Relax supported
31753 BIT_FIELD_REF argument.
31754 (vect_build_slp_tree_2): Record the desired vector type
31755 on the external vector def.
31756 (vectorizable_slp_permutation): Handle required punning
31757 of existing vector defs.
31759 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
31761 * rtl-ssa/accesses.h (def_lookup): Fix order of comparison results.
31763 2021-01-13 Richard Sandiford <richard.sandiford@arm.com>
31765 * config/sh/sh.md (movsf_ie): Remove operands[2] test.
31767 2021-01-13 Samuel Thibault <samuel.thibault@ens-lyon.org>
31769 * config.gcc [$target == *-*-gnu*]: Enable
31770 'default_gnu_indirect_function'.
31772 2021-01-13 Jakub Jelinek <jakub@redhat.com>
31775 * optabs.c (expand_vec_perm_const): Don't force v0 and v1 into
31776 registers before calling targetm.vectorize.vec_perm_const, only after
31778 * config/i386/i386-expand.c (ix86_vectorize_vec_perm_const): Handle
31779 two argument permutation when one operand is zero vector and only
31780 after that force operands into registers.
31781 * config/i386/sse.md (*avx2_zero_extendv16qiv16hi2_1): New
31782 define_insn_and_split pattern.
31783 (*avx512bw_zero_extendv32qiv32hi2_1): Likewise.
31784 (*avx512f_zero_extendv16hiv16si2_1): Likewise.
31785 (*avx2_zero_extendv8hiv8si2_1): Likewise.
31786 (*avx512f_zero_extendv8siv8di2_1): Likewise.
31787 (*avx2_zero_extendv4siv4di2_1): Likewise.
31788 * config/mips/mips.c (mips_vectorize_vec_perm_const): Force operands
31790 * config/arm/arm.c (arm_vectorize_vec_perm_const): Likewise.
31791 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Likewise.
31792 * config/ia64/ia64.c (ia64_vectorize_vec_perm_const): Likewise.
31793 * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const): Likewise.
31794 * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const): Likewise.
31795 * config/gcn/gcn.c (gcn_vectorize_vec_perm_const): Likewise. Use std::swap.
31797 2021-01-13 Martin Liska <mliska@suse.cz>
31799 PR tree-optimization/98455
31800 * gimple-if-to-switch.cc (condition_info::record_phi_mapping):
31801 Record also virtual PHIs.
31802 (pass_if_to_switch::execute): Return TODO_cleanup_cfg only
31805 2021-01-13 Jonathan Wakely <jwakely@redhat.com>
31807 * doc/invoke.texi (C++ Modules): Fix typos.
31809 2021-01-13 Richard Biener <rguenther@suse.de>
31811 PR tree-optimization/98640
31812 * tree-ssa-sccvn.c (visit_nary_op): Do not try to
31813 handle plus or minus from a truncated operand to be
31816 2021-01-13 Jakub Jelinek <jakub@redhat.com>
31819 * config/i386/i386.md (*btr<mode>_1, *btr<mode>_2): New
31820 define_insn_and_split patterns.
31821 (splitter after *btr<mode>_2): New splitter.
31823 2021-01-13 Martin Liska <mliska@suse.cz>
31826 * cgraphunit.c (analyze_functions): Remove dead code.
31828 2021-01-13 Qian Jianhua <qianjh@cn.fujitsu.com>
31830 * config/aarch64/aarch64-cost-tables.h (a64fx_extra_costs): New.
31831 * config/aarch64/aarch64.c (a64fx_addrcost_table): New.
31832 (a64fx_regmove_cost, a64fx_vector_cost): New.
31833 (a64fx_tunings): Use the new added cost tables.
31835 2021-01-13 Jakub Jelinek <jakub@redhat.com>
31838 * config/i386/predicates.md (pmovzx_parallel): New predicate.
31839 * config/i386/sse.md (*sse4_1_zero_extendv8qiv8hi2_3): New
31840 define_insn_and_split pattern.
31841 (*sse4_1_zero_extendv4hiv4si2_3): Likewise.
31842 (*sse4_1_zero_extendv2siv2di2_3): Likewise.
31844 2021-01-13 Julian Brown <julian@codesourcery.com>
31846 * config/gcn/gcn.c (gcn_conditional_register_usage): Remove dead code
31847 to fix v0 register.
31849 2021-01-13 Julian Brown <julian@codesourcery.com>
31851 * config/gcn/gcn.c (gcn_md_reorg): Fix case where EXEC reg is live
31854 2021-01-13 Julian Brown <julian@codesourcery.com>
31856 * config/gcn/gcn-valu.md (recip<mode>2<exec>, recip<mode>2): Use unspec
31857 for reciprocal-approximation instructions.
31858 (div<mode>3): Use fused multiply-accumulate operations for reciprocal
31859 refinement and division result.
31860 * config/gcn/gcn.md (UNSPEC_RCP): New unspec constant.
31862 2021-01-13 Julian Brown <julian@codesourcery.com>
31864 * config/gcn/gcn-valu.md (subdf): Rename to...
31867 2021-01-12 Martin Liska <mliska@suse.cz>
31869 * gcov.c (source_info::debug): Fix printf format for 32-bit hosts.
31871 2021-01-12 Andrea Corallo <andrea.corallo@arm.com>
31873 * function-abi.h: Fix typo.
31875 2021-01-12 Christophe Lyon <christophe.lyon@linaro.org>
31879 * config/arm/arm.h (ARM_HAVE_NEON_V8QI_LDST): New macro.
31880 (ARM_HAVE_NEON_V16QI_LDST, ARM_HAVE_NEON_V4HI_LDST): Likewise.
31881 (ARM_HAVE_NEON_V8HI_LDST, ARM_HAVE_NEON_V2SI_LDST): Likewise.
31882 (ARM_HAVE_NEON_V4SI_LDST, ARM_HAVE_NEON_V4HF_LDST): Likewise.
31883 (ARM_HAVE_NEON_V8HF_LDST, ARM_HAVE_NEON_V4BF_LDST): Likewise.
31884 (ARM_HAVE_NEON_V8BF_LDST, ARM_HAVE_NEON_V2SF_LDST): Likewise.
31885 (ARM_HAVE_NEON_V4SF_LDST, ARM_HAVE_NEON_DI_LDST): Likewise.
31886 (ARM_HAVE_NEON_V2DI_LDST): Likewise.
31887 (ARM_HAVE_V8QI_LDST, ARM_HAVE_V16QI_LDST): Likewise.
31888 (ARM_HAVE_V4HI_LDST, ARM_HAVE_V8HI_LDST): Likewise.
31889 (ARM_HAVE_V2SI_LDST, ARM_HAVE_V4SI_LDST, ARM_HAVE_V4HF_LDST): Likewise.
31890 (ARM_HAVE_V8HF_LDST, ARM_HAVE_V4BF_LDST, ARM_HAVE_V8BF_LDST): Likewise.
31891 (ARM_HAVE_V2SF_LDST, ARM_HAVE_V4SF_LDST, ARM_HAVE_DI_LDST): Likewise.
31892 (ARM_HAVE_V2DI_LDST): Likewise.
31893 * config/arm/mve.md (*movmisalign<mode>_mve_store): New pattern.
31894 (*movmisalign<mode>_mve_load): New pattern.
31895 * config/arm/neon.md (movmisalign<mode>): Move to ...
31896 * config/arm/vec-common.md: ... here.
31898 2021-01-12 Vladimir N. Makarov <vmakarov@redhat.com>
31901 * lra-eliminations.c (eliminate_regs_in_insn): Add transformation
31902 of pattern 'plus (plus (hard reg, const), pseudo)'.
31904 2021-01-12 Richard Biener <rguenther@suse.de>
31906 PR tree-optimization/98550
31907 * tree-vect-slp.c (vect_record_max_nunits): Check whether
31908 the group size is a multiple of the vector element count.
31909 (vect_build_slp_tree_1): When we need to fail because
31910 the vector type choosen causes unrolling do so lazily
31911 without affecting matches only at the end to guide group splitting.
31913 2021-01-12 Martin Liska <mliska@suse.cz>
31916 * optc-save-gen.awk: Compare also n_target_save vars with
31919 2021-01-12 Martin Liska <mliska@suse.cz>
31921 * gcov.c (source_info::debug): New.
31922 (print_usage): Add --debug (-D) option.
31923 (process_args): Likewise.
31924 (generate_results): Call src->debug after
31925 accumulate_line_counts.
31926 (read_graph_file): Properly assign id for EXIT_BLOCK.
31927 * profile.c (branch_prob): Dump function body before it is
31930 2021-01-12 Jakub Jelinek <jakub@redhat.com>
31932 PR tree-optimization/98629
31933 * tree-ssa-math-opts.c (arith_overflow_check_p): Don't update use_stmt
31934 unless returning non-zero.
31936 2021-01-12 Jakub Jelinek <jakub@redhat.com>
31938 PR tree-optimization/95731
31939 * tree-ssa-reassoc.c (optimize_range_tests_cmp_bitwise): Also optimize
31940 x < 0 && y < 0 && z < 0 into (x | y | z) < 0 for signed x, y, z.
31941 (optimize_range_tests): Call optimize_range_tests_cmp_bitwise
31942 only after optimize_range_tests_var_bound.
31944 2021-01-12 Jakub Jelinek <jakub@redhat.com>
31946 * configure.ac: Ensure c/Make-lang.in comes first in @all_lang_makefrags@.
31947 * configure: Regenerated.
31949 2021-01-12 liuhongt <hongtao.liu@intel.com>
31952 * config/i386/i386-builtins.h (BUILTIN_DESC_SWAP_OPERANDS):
31954 * config/i386/i386-expand.c (ix86_expand_sse_comi): Delete
31957 2021-01-12 Alexandre Oliva <oliva@adacore.com>
31959 * ssa-iterators.h (end_imm_use_stmt_traverse): Forward
31961 (auto_end_imm_use_stmt_traverse): New struct.
31962 (FOR_EACH_IMM_USE_STMT): Use it.
31963 (BREAK_FROM_IMM_USE_STMT, RETURN_FROM_IMM_USE_STMT): Remove,
31965 * gimple-ssa-strength-reduction.c: ... here, ...
31966 * graphite-scop-detection.c: ... here, ...
31967 * ipa-modref.c, ipa-pure-const.c, ipa-sra.c: ... here, ...
31968 * tree-predcom.c, tree-ssa-ccp.c: ... here, ...
31969 * tree-ssa-dce.c, tree-ssa-dse.c: ... here, ...
31970 * tree-ssa-loop-ivopts.c, tree-ssa-math-opts.c: ... here, ...
31971 * tree-ssa-phiprop.c, tree-ssa.c: ... here, ...
31972 * tree-vect-slp.c: ... and here, ...
31973 * doc/tree-ssa.texi: ... and the example here.
31975 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
31977 * config/aarch64/aarch64-sve.md (sdiv_pow2<mode>3): Extend from
31978 SVE_FULL_I to SVE_I. Generate an UNSPEC_PRED_X.
31979 (*sdiv_pow2<mode>3): New pattern.
31980 (@cond_<sve_int_op><mode>): Extend from SVE_FULL_I to SVE_I.
31981 Wrap the ASRD in an UNSPEC_PRED_X.
31982 (*cond_<sve_int_op><mode>_2): Likewise. Replace the UNSPEC_PRED_X
31983 predicate with a constant PTRUE, if it isn't already.
31984 (*cond_<sve_int_op><mode>_z): Replace with...
31985 (*cond_<sve_int_op><mode>_any): ...this new pattern.
31987 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
31989 * config/aarch64/aarch64-sve.md (*cond_bic<mode>_2): Extend from
31990 SVE_FULL_I to SVE_I.
31991 (*cond_bic<mode>_any): Likewise.
31993 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
31995 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart)
31996 (@aarch64_pred_<MUL_HIGHPART:optab><mode>): Extend from SVE_FULL_I
31999 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
32001 * config/aarch64/aarch64-sve.md (<su>abd<mode>_3): Extend from
32002 SVE_FULL_I to SVE_I.
32003 (*aarch64_cond_<su>abd<mode>_2): Likewise.
32004 (*aarch64_cond_<su>abd<mode>_any): Likewise.
32005 (@aarch64_pred_<su>abd<mode>): Likewise. Use UNSPEC_PRED_X
32006 for the max and min but not for the minus.
32007 (*aarch64_cond_<su>abd<mode>_3): New pattern.
32009 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
32011 * config/aarch64/iterators.md (SVE_24I): New iterator.
32012 * config/aarch64/aarch64-sve.md (*aarch64_adr<mode>_shift): Extend from
32013 SVE_FULL_SDI to SVE_24I. Use containers rather than elements.
32015 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
32017 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_BINARY:optab><mode>)
32018 (*cond_<SVE_INT_BINARY:optab><mode>_2): Extend from SVE_FULL_I
32020 (*cond_<SVE_INT_BINARY:optab><mode>_3): Likewise.
32021 (*cond_<SVE_INT_BINARY:optab><mode>_any): Likewise.
32022 (*cond_<SVE_INT_BINARY:optab><mode>_2_const): Likewise.
32023 (*cond_<SVE_INT_BINARY:optab><mode>_any_const): Likewise.
32025 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
32027 * config/aarch64/aarch64-sve.md (<SVE_INT_BINARY_IMM:optab><mode>3)
32028 (@aarch64_pred_<SVE_INT_BINARY_IMM:optab><mode>)
32029 (*post_ra_<SVE_INT_BINARY_IMM:optab><mode>3): Extend from SVE_FULL_I
32032 2021-01-11 Richard Sandiford <richard.sandiford@arm.com>
32034 * config/aarch64/aarch64-sve.md (<ASHIFT:optab><mode>3)
32035 (v<ASHIFT:optab><mode>3, @aarch64_pred_<optab><mode>)
32036 (*post_ra_v<ASHIFT:optab><mode>3): Extend from SVE_FULL_I to SVE_I.
32038 2021-01-11 Martin Liska <mliska@suse.cz>
32041 * symtab-clones.h (clone_info::release): Release
32042 symtab::m_clones with ggc_delete as it's a GGC memory.
32044 2021-01-11 Matthias Klose <doko@ubuntu.com>
32046 * Makefile.in (LINK_PROGRESS): Show the link target.
32048 2021-01-11 Richard Biener <rguenther@suse.de>
32050 PR tree-optimization/91403
32051 * tree-vect-data-refs.c (vect_analyze_group_access_1): Cap
32052 single-element interleaving group size at 4096 elements.
32054 2021-01-11 Richard Biener <rguenther@suse.de>
32056 PR tree-optimization/98526
32057 * tree-vect-loop.c (vect_model_reduction_cost): Remove costing
32058 of the actual reduction op for the regular case.
32059 (vectorizable_reduction): Cost the stmts
32060 vect_transform_reduction produces here.
32062 2021-01-11 Andreas Krebbel <krebbel@linux.ibm.com>
32064 * tree-ssa-forwprop.c (simplify_vector_constructor): For
32065 big-endian, use UNPACK[_FLOAT]_HI.
32067 2021-01-11 Tamar Christina <tamar.christina@arm.com>
32069 * tree-vect-slp-patterns.c (class complex_pattern,
32070 class complex_add_pattern): Add parameters to matches.
32071 (complex_add_pattern::build): Free memory.
32072 (complex_add_pattern::matches): Move validation end of match.
32073 (complex_add_pattern::recognize): Likewise.
32075 2021-01-11 Tamar Christina <tamar.christina@arm.com>
32077 * tree-vect-slp-patterns.c (linear_loads_p): Fix externals.
32079 2021-01-11 Tamar Christina <tamar.christina@arm.com>
32081 * tree-vect-slp-patterns.c (is_linear_load_p): Fix ambiguity.
32083 2021-01-11 Jakub Jelinek <jakub@redhat.com>
32085 PR tree-optimization/95867
32086 * tree-ssa-math-opts.h: New header.
32087 * tree-ssa-math-opts.c: Include tree-ssa-math-opts.h.
32088 (powi_as_mults): No longer static. Use build_one_cst instead of
32089 build_real. Formatting fix.
32090 * tree-ssa-reassoc.c: Include tree-ssa-math-opts.h.
32091 (attempt_builtin_powi): Handle multiplication reassociation without
32092 powi_fndecl using powi_as_mults.
32093 (reassociate_bb): For integral types don't require
32094 -funsafe-math-optimizations to call attempt_builtin_powi.
32096 2021-01-11 Jakub Jelinek <jakub@redhat.com>
32098 PR tree-optimization/95852
32099 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): Change
32100 mul_stmts parameter type to vec<gimple *> &. Before cond_stmt
32101 allow in the bb any of the stmts in that vector, div_stmt and
32102 up to 3 cast stmts.
32103 (arith_cast_equal_p): New function.
32104 (arith_overflow_check_p): Add cast_stmt argument, handle signed
32105 multiply overflow checks.
32106 (match_arith_overflow): Adjust caller. Handle signed multiply
32109 2021-01-11 Jakub Jelinek <jakub@redhat.com>
32111 PR tree-optimization/95852
32112 * tree-ssa-math-opts.c (maybe_optimize_guarding_check): New function.
32113 (uaddsub_overflow_check_p): Renamed to ...
32114 (arith_overflow_check_p): ... this. Handle also multiplication
32115 with overflow check.
32116 (match_uaddsub_overflow): Renamed to ...
32117 (match_arith_overflow): ... this. Add cfg_changed argument. Handle
32118 also multiplication with overflow check. Adjust function comment.
32119 (math_opts_dom_walker::after_dom_children): Adjust callers. Call
32120 match_arith_overflow also for MULT_EXPR.
32122 2021-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32124 * config/aarch64/arm_neon.h (vmovl_s8): Reimplement using
32125 __builtin_convertvector.
32126 (vmovl_s16): Likewise.
32127 (vmovl_s32): Likewise.
32128 (vmovl_u8): Likewise.
32129 (vmovl_u16): Likewise.
32130 (vmovl_u32): Likewise.
32131 (vmovn_s16): Likewise.
32132 (vmovn_s32): Likewise.
32133 (vmovn_s64): Likewise.
32134 (vmovn_u16): Likewise.
32135 (vmovn_u32): Likewise.
32136 (vmovn_u64): Likewise.
32138 2021-01-11 Martin Liska <mliska@suse.cz>
32140 * gimple-if-to-switch.cc (struct condition_info): Use auto_var.
32141 (if_chain::is_beneficial): Delete clusters
32142 (find_conditions): Make second argument of conditions_in_bbs a
32143 pointer so that we control over it's lifetime.
32144 (pass_if_to_switch::execute): Delete them.
32146 2021-01-11 Kewen Lin <linkw@linux.ibm.com>
32148 * ira.c (move_unallocated_pseudos): Check other_reg and skip if
32151 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
32153 * config/vax/vax.md (cc): Remove mode attribute.
32154 (subst_<cc>, subst_f<cc>): Rename to...
32155 (subst_<mode>, subst_f<VAXccnz:mode>): ... these respectively.
32156 (*cbranch<VAXint:mode>4_<VAXcc:mode>): Update for `cc' removal.
32157 (*cbranch<VAXfp:mode>4_<VAXccnz:mode>): Likewise.
32158 (*branch_<mode>, *branch_<mode>_reversed): Likewise.
32160 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
32162 * config/vax/vax.md (subst_f<cc>): Add mode to operands and
32163 `const_double_zero'.
32165 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
32167 * config/pdp11/pdp11.md (PDPfp): New mode iterator.
32168 (fcc_cc, fcc_ccnz): Use it. Add mode to `const_double_zero' and
32171 2021-01-09 Maciej W. Rozycki <macro@linux-mips.org>
32173 * genemit.c (gen_exp) <CONST_DOUBLE>: Handle `const_double_zero'
32175 * read-rtl.c (rtx_reader::read_rtx_code): Handle machine mode
32176 with `const_double_zero'.
32177 * doc/rtl.texi (Constant Expression Types): Document it.
32179 2021-01-09 Jakub Jelinek <jakub@redhat.com>
32182 * tree-cfg.c (verify_gimple_assign_binary): Allow lhs of
32183 POINTER_DIFF_EXPR to be any integral type.
32185 2021-01-09 Jakub Jelinek <jakub@redhat.com>
32187 PR rtl-optimization/98603
32188 * function.c (instantiate_virtual_regs_in_insn): For asm goto
32189 with impossible constraints, drop all SETs, CLOBBERs, drop PARALLEL
32190 if any, set ASM_OPERANDS mode to VOIDmode and change
32191 ASM_OPERANDS_OUTPUT_CONSTRAINT and ASM_OPERANDS_OUTPUT_IDX.
32193 2021-01-09 Alexandre Oliva <oliva@gnu.org>
32196 * final.c (notice_source_line): Narrow down the condition to
32197 skip a line-0 marker.
32199 2021-01-08 Sergei Trofimovich <siarheit@google.com>
32201 * ipa-modref.c (merge_call_side_effects): Fix
32202 linebreak split by reordering two print calls.
32204 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
32206 * config/s390/vector.md (*tf_to_fprx2_0): Rename from
32207 "*mov_tf_to_fprx2_0" for consistency, fix constraint.
32208 (*tf_to_fprx2_1): Rename from "*mov_tf_to_fprx2_1" for
32209 consistency, fix constraint.
32211 2021-01-08 Ilya Leoshkevich <iii@linux.ibm.com>
32213 * config/s390/s390-c.c (s390_def_or_undef_macro): Accept
32214 callables instead of mask values.
32215 (struct target_flag_set_p): New predicate.
32216 (s390_cpu_cpp_builtins_internal): Define or undefine
32217 __LONG_DOUBLE_VX__ macro.
32219 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
32222 * config/i386/i386.c (x86_function_profiler): Use R10 and R11
32223 to call mcount in large model with PIC for NO_PROFILE_COUNTERS
32226 2021-01-08 Richard Biener <rguenther@suse.de>
32228 * tree-ssa-sccvn.c (pass_fre::execute): Reset the SCEV hash table.
32230 2021-01-08 Richard Biener <rguenther@suse.de>
32232 * tree-vect-slp.c (scalar_stmts_to_slp_tree_map_t): Fix.
32233 (vect_build_slp_tree): On cache hit release the matched
32234 scalar stmts vector.
32235 * tree-vect-stmts.c (vectorizable_store): Properly free
32236 vec_oprnds before possibly gathering them again.
32238 2021-01-08 Richard Biener <rguenther@suse.de>
32240 PR tree-optimization/98544
32241 * tree-vect-slp.c (vect_optimize_slp): Always materialize
32242 permutes at a permute node.
32244 2021-01-08 H.J. Lu <hjl.tools@gmail.com>
32247 * config/i386/i386.c (x86_function_profiler): Use R10 to call
32248 mcount in large model. Sorry for large model with PIC.
32250 2021-01-08 Jakub Jelinek <jakub@redhat.com>
32253 * config/i386/i386.opt (ix86_cmodel, ix86_incoming_stack_boundary_arg,
32254 ix86_pmode, ix86_preferred_stack_boundary_arg, ix86_regparm,
32255 ix86_veclibabi_type): Remove x_ prefix, use TargetVariable instead of
32256 TargetSave and initialize for variables with enum types.
32257 (mfentry, mstack-protector-guard-reg=, mstack-protector-guard-offset=,
32258 mstack-protector-guard-symbol=): Add Save.
32259 * config/i386/i386-options.c (ix86_function_specific_save,
32260 ix86_function_specific_restore): Don't save or restore x_ix86_cmodel,
32261 x_ix86_incoming_stack_boundary_arg, x_ix86_pmode,
32262 x_ix86_preferred_stack_boundary_arg, x_ix86_regparm,
32263 x_ix86_veclibabi_type.
32265 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
32267 * config/aarch64/aarch64-sve.md (*cnot<mode>): Extend from
32268 SVE_FULL_I to SVE_I.
32269 (*cond_cnot<mode>_2, *cond_cnot<mode>_any): Likewise.
32271 2021-01-08 Richard Sandiford <richard.sandiford@arm.com>
32273 * config/aarch64/aarch64-sve.md (*cond_uxt<mode>_2): Extend from
32274 SVE_FULL_I to SVE_I.
32275 (*cond_uxt<mode>_any): Likewise.
32277 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32279 * config/aarch64/iterators.md (Vwhalf): New iterator.
32280 * config/aarch64/aarch64-simd.md (aarch64_<sur>adalp<mode>_3):
32282 (aarch64_<sur>adalp<mode>): ... This. Make more
32284 (<sur>sadv16qi): Adjust callsite of the above.
32285 * config/aarch64/aarch64-simd-builtins.def (sadalp, uadalp): New
32287 * config/aarch64/arm_neon.h (vpadal_s8): Reimplement using
32289 (vpadal_s16): Likewise.
32290 (vpadal_u8): Likewise.
32291 (vpadal_u16): Likewise.
32292 (vpadalq_s8): Likewise.
32293 (vpadalq_s16): Likewise.
32294 (vpadalq_s32): Likewise.
32295 (vpadalq_u8): Likewise.
32296 (vpadalq_u16): Likewise.
32297 (vpadalq_u32): Likewise.
32299 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32301 * config/aarch64/aarch64-simd.md (aarch64_<su>abd<mode>_3):
32303 (aarch64_<su>abd<mode>): ... This.
32304 (<sur>sadv16qi): Adjust callsite of the above.
32305 * config/aarch64/aarch64-simd-builtins.def (sabd, uabd): Define
32307 * config/aarch64/arm_neon.h (vabd_s8): Reimplement using
32309 (vabd_s16): Likewise.
32310 (vabd_s32): Likewise.
32311 (vabd_u8): Likewise.
32312 (vabd_u16): Likewise.
32313 (vabd_u32): Likewise.
32314 (vabdq_s8): Likewise.
32315 (vabdq_s16): Likewise.
32316 (vabdq_s32): Likewise.
32317 (vabdq_u8): Likewise.
32318 (vabdq_u16): Likewise.
32319 (vabdq_u32): Likewise.
32321 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32323 * config/aarch64/aarch64-simd-builtins.def (saba, uaba): Define
32325 * config/aarch64/arm_neon.h (vaba_s8): Implement using builtin.
32326 (vaba_s16): Likewise.
32327 (vaba_s32): Likewise.
32328 (vaba_u8): Likewise.
32329 (vaba_u16): Likewise.
32330 (vaba_u32): Likewise.
32331 (vabaq_s8): Likewise.
32332 (vabaq_s16): Likewise.
32333 (vabaq_s32): Likewise.
32334 (vabaq_u8): Likewise.
32335 (vabaq_u16): Likewise.
32336 (vabaq_u32): Likewise.
32338 2021-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
32340 * config/aarch64/aarch64-simd.md (aba<mode>_3): Rename to...
32341 (aarch64_<su>aba<mode>): ... This. Handle uaba as well.
32342 Change RTL pattern to match.
32344 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
32346 * common/config/riscv/riscv-common.c (riscv_current_subset_list): New.
32347 * config/riscv/riscv-c.c (riscv-subset.h): New.
32348 (INCLUDE_STRING): Define.
32349 (riscv_cpu_cpp_builtins): Add new style architecture extension
32351 * config/riscv/riscv-subset.h (riscv_subset_list::begin): New.
32352 (riscv_subset_list::end): New.
32353 (riscv_current_subset_list): New.
32355 2021-01-08 Kito Cheng <kito.cheng@sifive.com>
32357 * common/config/riscv/riscv-common.c (RISCV_DONT_CARE_VERSION):
32358 Move to riscv-subset.h.
32359 (struct riscv_subset_t): Ditto.
32360 (class riscv_subset_list): Ditto.
32361 * config/riscv/riscv-subset.h (RISCV_DONT_CARE_VERSION): Move
32362 from riscv-common.c.
32363 (struct riscv_subset_t): Ditto.
32364 (class riscv_subset_list): Ditto.
32365 * config/riscv/t-riscv ($(common_out_file)): Add file
32368 2021-01-07 Jakub Jelinek <jakub@redhat.com>
32371 * config/i386/i386.md (*bmi_blsi_<mode>_cmp, *bmi_blsi_<mode>_ccno):
32372 New define_insn patterns.
32374 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
32376 * config/aarch64/aarch64-sve.md (@cond_<SVE_INT_UNARY:optab><mode>)
32377 (*cond_<SVE_INT_UNARY:optab><mode>_2): Extend from SVE_FULL_I to SVE_I.
32378 (*cond_<SVE_INT_UNARY:optab><mode>_any): Likewise.
32380 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
32382 PR tree-optimization/98560
32383 * internal-fn.def (IFN_VCONDU, IFN_VCONDEQ): Use type vec_cond.
32384 * internal-fn.c (vec_cond_mask_direct): Get the data mode from
32386 (vec_cond_direct): Likewise argument 2.
32387 (vec_condu_direct, vec_condeq_direct): Delete.
32388 (expand_vect_cond_optab_fn): Rename to...
32389 (expand_vec_cond_optab_fn): ...this, replacing old macro.
32390 (expand_vec_condu_optab_fn, expand_vec_condeq_optab_fn): Delete.
32391 (expand_vect_cond_mask_optab_fn): Rename to...
32392 (expand_vec_cond_mask_optab_fn): ...this, replacing old macro.
32393 (direct_vec_cond_mask_optab_supported_p): Treat the optab as a
32395 (direct_vec_cond_optab_supported_p): Likewise.
32396 (direct_vec_condu_optab_supported_p): Delete.
32397 (direct_vec_condeq_optab_supported_p): Delete.
32398 * gimple-isel.cc: Include internal-fn.h.
32399 (gimple_expand_vec_cond_expr): Check that IFN_VCONDEQ is supported
32402 2021-01-07 Richard Sandiford <richard.sandiford@arm.com>
32404 PR tree-optimization/98560
32405 * gimple-isel.cc (gimple_expand_vec_cond_expr): If we fail to use
32406 IFN_VCOND{,U,EQ}, fall back on IFN_VCOND_MASK.
32408 2021-01-07 Uroš Bizjak <ubizjak@gmail.com>
32410 * config/i386/i386.md (insn): Merge from plusminus_insn, shift_insn,
32411 rotate_insn and optab code attributes.
32412 Update all uses to merged code attribute.
32413 * config/i386/sse.md: Update all uses to merged code attribute.
32414 * config/i386/mmx.md: Update all uses to merged code attribute.
32416 2021-01-07 Jakub Jelinek <jakub@redhat.com>
32418 PR tree-optimization/98568
32419 * gimple-ssa-store-merging.c (bswap_view_convert): New function.
32420 (bswap_replace): Use it.
32422 2021-01-06 Vladimir N. Makarov <vmakarov@redhat.com>
32424 PR rtl-optimization/97978
32425 * lra-int.h (lra_hard_reg_split_p): New external.
32426 * lra.c (lra_hard_reg_split_p): New global.
32427 (lra): Set up lra_hard_reg_split_p after splitting a hard reg.
32428 * lra-assigns.c (lra_assign): Don't check allocation correctness
32429 after hard reg splitting.
32431 2021-01-06 Martin Sebor <msebor@redhat.com>
32434 * builtins.c (new_delete_mismatch_p): New overload.
32435 (new_delete_mismatch_p (tree, tree)): Call it.
32437 2021-01-06 Alexandre Oliva <oliva@adacore.com>
32439 * Makefile.in (T_GLIMITS_H): New.
32440 (stmp-int-hdrs): Depend on it, use it.
32441 * config/t-vxworks (T_GLIMITS_H): Override it.
32442 (vxw-glimits.h): New.
32444 2021-01-06 Richard Biener <rguenther@suse.de>
32446 PR tree-optimization/98513
32447 * value-range.cc (intersect_ranges): Compare the upper bounds
32448 for the expected relation.
32450 2021-01-06 Gerald Pfeifer <gerald@pfeifer.com>
32453 2020-12-28 Gerald Pfeifer <gerald@pfeifer.com>
32455 * doc/standards.texi (HSAIL): Remove section.
32457 2021-01-05 Samuel Thibault <samuel.thibault@ens-lyon.org>
32459 * configure: Re-generate.
32461 2021-01-05 Jakub Jelinek <jakub@redhat.com>
32463 * doc/invoke.texi (-std=c++20): Adjust for the publication of
32464 ISO 14882:2020 standard.
32465 * doc/standards.texi: Likewise.
32467 2021-01-05 Jakub Jelinek <jakub@redhat.com>
32469 PR tree-optimization/94802
32470 * expr.h (maybe_optimize_sub_cmp_0): Declare.
32471 * expr.c: Include tree-pretty-print.h and flags.h.
32472 (maybe_optimize_sub_cmp_0): New function.
32473 (do_store_flag): Use it.
32474 * cfgexpand.c (expand_gimple_cond): Likewise.
32476 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
32478 * mux-utils.h (pointer_mux::m_ptr): Tweak description of contents.
32479 * rtlanal.c (simple_regno_set): Tweak description to clarify the
32482 2021-01-05 Richard Biener <rguenther@suse.de>
32484 PR tree-optimization/98516
32485 * tree-vect-slp.c (vect_optimize_slp): Permute the incoming
32486 lanes when materializing on a VEC_PERM node.
32487 (vectorizable_slp_permutation): Dump the permute properly.
32489 2021-01-05 Richard Biener <rguenther@suse.de>
32491 * tree-vect-slp.c (vect_slp_region): Move debug counter
32492 to cover individual subgraphs.
32494 2021-01-05 Richard Biener <rguenther@suse.de>
32496 PR tree-optimization/98428
32497 * tree-vect-slp.c (vect_build_slp_tree_1): Properly reject
32498 vector lane extracts for loop vectorization.
32500 2021-01-05 Jakub Jelinek <jakub@redhat.com>
32502 PR tree-optimization/98514
32503 * tree-ssa-reassoc.c (bb_rank): Change type from long * to
32505 (operand_rank): Change type from hash_map<tree, long> to
32506 hash_map<tree, int64_t>.
32507 (phi_rank): Change return type from long to int64_t.
32508 (loop_carried_phi): Change block_rank variable type from long to
32510 (propagate_rank): Change return type, rank parameter type and
32511 op_rank variable type from long to int64_t.
32512 (find_operand_rank): Change return type from long to int64_t
32513 and change slot variable type from long * to int64_t *.
32514 (insert_operand_rank): Change rank parameter type from long to
32516 (get_rank): Change return type and rank variable type from long to
32517 int64_t. Use PRId64 instead of ld to print the rank.
32518 (init_reassoc): Change rank variable type from long to int64_t
32519 and adjust correspondingly bb_rank and operand_rank initialization.
32521 2021-01-05 Jakub Jelinek <jakub@redhat.com>
32523 PR tree-optimization/96928
32524 * tree-ssa-phiopt.c (xor_replacement): New function.
32525 (tree_ssa_phiopt_worker): Call it.
32527 2021-01-05 Jakub Jelinek <jakub@redhat.com>
32529 PR tree-optimization/96930
32530 * match.pd ((A / (1 << B)) -> (A >> B)): If A is extended
32531 from narrower value which has the same type as 1 << B, perform
32532 the right shift on the narrower value followed by extension.
32534 2021-01-05 Jakub Jelinek <jakub@redhat.com>
32536 PR tree-optimization/96239
32537 * gimple-ssa-store-merging.c (maybe_optimize_vector_constructor): New
32539 (get_status_for_store_merging): Don't return BB_INVALID for blocks
32540 with potential bswap optimizable CONSTRUCTORs.
32541 (pass_store_merging::execute): Optimize vector CONSTRUCTORs with bswap
32544 2021-01-05 Richard Biener <rguenther@suse.de>
32546 PR tree-optimization/98381
32547 * tree.c (vector_element_bits): Properly compute bool vector
32549 * tree-vect-loop.c (vectorizable_live_operation): Properly
32550 compute the last lane bit offset.
32552 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
32555 * config/i386/sse.md (sse_cvtps2pi): Redefine as define_insn_and_split.
32556 Clear the top 64 bytes of the input XMM register.
32557 (sse_cvttps2pi): Ditto.
32559 2021-01-05 Uroš Bizjak <ubizjak@gmail.com>
32562 * config/i386/xopintrin.h (_mm256_cmov_si256): New.
32564 2021-01-05 H.J. Lu <hjl.tools@gmail.com>
32567 * config/i386/xmmintrin.h (_mm_extract_pi16): Cast to unsigned
32570 2021-01-05 Claudiu Zissulescu <claziss@synopsys.com>
32572 * config/arc/arc.md (maddsidi4_split): Use ACC_REG_FIRST.
32573 (umaddsidi4_split): Likewise.
32575 2021-01-05 liuhongt <hongtao.liu@intel.com>
32578 * config/i386/sse.md (*sse2_pmovskb_zexthisi): New
32579 define_insn_and_split for zero_extend of subreg HI of pmovskb
32581 (*sse2_pmovskb_zexthisi): Add new combine splitters for
32582 zero_extend of not of subreg HI of pmovskb result.
32584 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
32587 * explow.c (convert_memory_address_addr_space_1): Handle UNSPECs
32589 * config/aarch64/aarch64.c (aarch64_expand_mov_immediate): Use
32590 convert_memory_address to convert symbolic immediates to ptr_mode
32591 before forcing them to memory.
32593 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
32595 PR rtl-optimization/97144
32596 * recog.c (constrain_operands): Initialize matching_operand
32597 for each alternative, rather than only doing it once.
32599 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
32601 PR rtl-optimization/98403
32602 * rtl-ssa/changes.cc (function_info::finalize_new_accesses): Explain
32603 why we don't remove call clobbers.
32604 (function_info::apply_changes_to_insn): Don't attempt to add
32605 call clobbers here.
32607 2021-01-05 Richard Sandiford <richard.sandiford@arm.com>
32609 PR tree-optimization/98371
32610 * tree-vect-loop.c (vect_reanalyze_as_main_loop): New function.
32611 (vect_analyze_loop): If an epilogue loop appears to be cheaper
32612 than the main loop, re-analyze it as a main loop before adopting
32615 2021-01-05 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
32618 * configure.ac (NETLIBS): Determine using AX_LIB_SOCKET_NSL.
32619 * aclocal.m4, configure: Regenerate.
32620 * Makefile.in (NETLIBS): Define.
32621 (BACKEND): Remove $(CODYLIB).
32623 2021-01-05 Jakub Jelinek <jakub@redhat.com>
32625 PR rtl-optimization/98334
32626 * simplify-rtx.c (simplify_context::simplify_binary_operation_1):
32627 Optimize (X - 1) * Y + Y to X * Y or (X + 1) * Y - Y to X * Y.
32629 2021-01-05 Bernd Edlinger <bernd.edlinger@hotmail.de>
32631 * tree-inline.c (expand_call_inline): Restore input_location.
32632 Return result from recursive call.
32634 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
32636 PR tree-optimization/95401
32637 * config/aarch64/aarch64-sve-builtins.cc
32638 (gimple_folder::load_store_cookie): Use bits rather than bytes
32639 for the alignment argument to IFN_MASK_LOAD and IFN_MASK_STORE.
32640 * gimple-fold.c (gimple_fold_mask_load_store_mem_ref): Likewise.
32641 * tree-vect-stmts.c (vectorizable_store): Likewise.
32642 (vectorizable_load): Likewise.
32644 2021-01-04 Richard Biener <rguenther@suse.de>
32646 PR tree-optimization/98308
32647 * tree-vect-stmts.c (vectorizable_load): Set invariant mask
32650 2021-01-04 Jakub Jelinek <jakub@redhat.com>
32652 PR tree-optimization/95771
32653 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Handle types
32654 with precision smaller than int's precision and types with precision
32655 twice as large as long long. Formatting fixes.
32657 2021-01-04 Richard Biener <rguenther@suse.de>
32659 PR tree-optimization/98464
32660 * tree-ssa-sccvn.c (vn_valueize_for_srt): Rename from ...
32661 (vn_valueize_wrapper): ... this. Temporarily adjust vn_context_bb.
32662 (process_bb): Adjust.
32664 2021-01-04 Matthew Malcomson <matthew.malcomson@arm.com>
32667 * doc/invoke.texi (-fsanitize=address): Fix wording describing
32668 clash with -fsanitize=hwaddress.
32670 2021-01-04 Richard Biener <rguenther@suse.de>
32672 PR tree-optimization/98282
32673 * tree-ssa-sccvn.c (vn_get_stmt_kind): Classify tcc_reference on
32674 invariants as VN_NARY.
32676 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
32679 * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Accept
32680 aarch64_simd_reg_or_zero for operand 2. Use the combinez patterns
32681 to handle zero operands.
32683 2021-01-04 Richard Sandiford <richard.sandiford@arm.com>
32685 * config/aarch64/aarch64.c (offset_6bit_signed_scaled_p): New function.
32686 (offset_6bit_unsigned_scaled_p): Fix typo in comment.
32687 (aarch64_sve_prefetch_operand_p): Accept MUL VLs in the range
32690 2021-01-04 Richard Biener <rguenther@suse.de>
32692 PR tree-optimization/98393
32693 * tree-vect-slp.c (vect_build_slp_tree): Properly zero matches
32694 when hitting the limit.
32696 2021-01-04 Richard Biener <rguenther@suse.de>
32698 PR tree-optimization/98291
32699 * tree-vect-loop.c (vectorizable_reduction): Bypass
32700 associativity check for SLP reductions with VF 1.
32702 2021-01-04 Jakub Jelinek <jakub@redhat.com>
32704 PR tree-optimization/96782
32705 * match.pd (x == ~x -> false, x != ~x -> true): New simplifications.
32707 2021-01-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
32709 * collect-utils.c (collect_execute): Check dumppfx.
32710 * collect2.c (maybe_run_lto_and_relink, do_link): Pass atsuffix
32711 to collect_execute.
32712 (do_link): Add new parameter atsuffix.
32713 (main): Handle -dumpdir option. Skip one argument for
32714 -o, -isystem and -B options.
32715 * gcc.c (make_at_file): New helper function.
32716 (close_at_file): Use it.
32718 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
32720 * config/darwin.h (MIN_LD64_NO_COAL_SECTS): Adjust.
32721 Amend handling for LD64_VERSION fallback defaults.
32723 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
32725 * config.gcc: Compute default version information
32726 from the configured target. Likewise defaults for
32728 * config/darwin10.h: Removed.
32729 * config/darwin12.h: Removed.
32730 * config/darwin9.h: Removed.
32731 * config/rs6000/darwin8.h: Removed.
32733 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
32735 * config/darwin9.h (ASM_OUTPUT_ALIGNED_COMMON): Delete.
32737 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
32739 * config/darwin9.h (STACK_CHECK_STATIC_BUILTIN): Move from here..
32740 * config/darwin.h (STACK_CHECK_STATIC_BUILTIN): .. to here.
32742 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
32744 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move from
32746 * config/darwin.h (LINK_GCC_C_SEQUENCE_SPEC): ... to here.
32748 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
32750 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Move the spec
32751 for the Darwin10 unwinder stub from here ...
32752 * config/darwin.h (LINK_COMMAND_SPEC_A): ... to here.
32754 2021-01-02 Iain Sandoe <iain@sandoe.co.uk>
32756 * config/darwin.h (DSYMUTIL_SPEC): Default to DWARF
32757 (ASM_DEBUG_SPEC):Only define if the assembler supports
32759 (PREFERRED_DEBUGGING_TYPE): Default to DWARF.
32760 (DARWIN_PREFER_DWARF): Define.
32761 * config/darwin9.h (PREFERRED_DEBUGGING_TYPE): Remove.
32762 (DARWIN_PREFER_DWARF): Likewise
32763 (DSYMUTIL_SPEC): Likewise.
32764 (COLLECT_RUN_DSYMUTIL): Likewise.
32765 (ASM_DEBUG_SPEC): Likewise.
32766 (ASM_DEBUG_OPTION_SPEC): Likewise.
32768 2021-01-02 Jan Hubicka <jh@suse.cz>
32770 * cfg.c (free_block): ggc_free bb.
32772 2021-01-01 Jakub Jelinek <jakub@redhat.com>
32774 * gcc.c (process_command): Update copyright notice dates.
32775 * gcov-dump.c (print_version): Ditto.
32776 * gcov.c (print_version): Ditto.
32777 * gcov-tool.c (print_version): Ditto.
32778 * gengtype.c (create_file): Ditto.
32779 * doc/cpp.texi: Bump @copying's copyright year.
32780 * doc/cppinternals.texi: Ditto.
32781 * doc/gcc.texi: Ditto.
32782 * doc/gccint.texi: Ditto.
32783 * doc/gcov.texi: Ditto.
32784 * doc/install.texi: Ditto.
32785 * doc/invoke.texi: Ditto.
32787 2021-01-01 Jakub Jelinek <jakub@redhat.com>
32789 * ChangeLog-2020: Rotate ChangeLog. New file.
32792 Copyright (C) 2021 Free Software Foundation, Inc.
32794 Copying and distribution of this file, with or without modification,
32795 are permitted in any medium without royalty provided the copyright
32796 notice and this notice are preserved.